2017-02-01 02:08:53 <^7heo> damn 2017-02-01 02:09:19 <^7heo> why did fabled implement 3/4 of the patch from ncopa in his patches without implementing all the rest? 2017-02-01 02:09:25 <^7heo> this is just making my job horrible... 2017-02-01 07:20:24 fabled: Hi, do you have access to an s390x VM ? 2017-02-01 07:20:35 tmh1999, no 2017-02-01 07:20:47 well 2017-02-01 07:20:50 we have qemu for it 2017-02-01 07:21:11 I am not really confident with my gdb skills 2017-02-01 07:21:18 and it will prolong the porting process 2017-02-01 07:21:24 can I send you the sysroot ? 2017-02-01 07:21:41 how does one acquire an s390x system without paying an arm and a leg 2017-02-01 07:22:06 last time I checked, qemu s390x can only run on a s390x machine @@ 2017-02-01 07:22:28 kaniini : maybe more 2017-02-01 07:22:41 tmh1999, how do you test it currently? on real-hw? or VM? 2017-02-01 07:22:55 it's VM 2017-02-01 07:23:42 Linux on z or Linux/s390x is actually all VM. there is no concrete machine that run Linux out of the box,according to my understanding. 2017-02-01 07:24:56 actually people only interested in getting the packages up, does not really care much about the booting process 2017-02-01 07:25:19 I mean, getting the packages up on the repo 2017-02-01 07:25:35 right 2017-02-01 07:25:42 then fire Docker away 2017-02-01 07:26:07 the question is how to get the hardware so we can have a builder tho ;/ 2017-02-01 07:26:34 I heard some IBM fellows contacted ncopa for the hw ? 2017-02-01 07:26:44 that's for powerpc iirc 2017-02-01 07:27:04 I guess s390x will coming soon because my progress so slow ... 2017-02-01 07:27:28 tmh1999: yes, some ibm ppl contacted me 2017-02-01 07:27:44 we've been working on the ppc64le enablement for some time (as well as enablement for classic ppc which is more likely to be in the wild right now) 2017-02-01 07:28:28 power8 is out of my budget atm, those pSeries cost almost as much as my down payment did ;) 2017-02-01 07:28:41 they contacted me about both power8 and s390x 2017-02-01 07:29:04 as i understand, it is possible to get access to s390x hw for free for 120 days 2017-02-01 07:29:13 ncopa : correct ! 2017-02-01 07:29:17 I fogot that ... 2017-02-01 07:29:26 for ppc64le i dont know 2017-02-01 07:29:47 120 days might be enough to bootstrap. but we would need to have more permanent access to s390x hardware to keep the port maintained (and therefore qualifiable in a release) 2017-02-01 07:29:55 that's my concern ;) 2017-02-01 07:29:55 yes 2017-02-01 07:29:57 fabled : would you mind ? https://developer.ibm.com/linuxone/request-a-trial/ 2017-02-01 07:30:13 no point in shipping something we can't update after all ;) 2017-02-01 07:30:25 kaniini : 120 is for public. with Alpine team, I guess it would be more. 2017-02-01 07:30:30 i think if we get something up and run within 120 days, then we can likely get some more permanent solution 2017-02-01 07:30:49 ncopa : agree 2017-02-01 07:33:30 maybe all the ibm dudes that are coming and going on this channel can hook us up ;) 2017-02-01 07:34:41 i have talked with them via dockercommunity.slack.com 2017-02-01 07:35:08 and they have offered longer term than 120 days 2017-02-01 07:35:28 atleast for s390x 2017-02-01 07:35:40 i wonder what the future of sparc is now that oracle is killing solaris 2017-02-01 07:36:08 i have some halfway disassembled sun gear sitting in my living room that i could probably get going as builders for a sparc port 2017-02-01 07:39:13 who has s390x machine running at home? :D 2017-02-01 07:40:19 some do 2017-02-01 07:41:37 https://lists.debian.org/debian-devel-announce/2015/07/msg00006.html 2017-02-01 07:41:43 hmm, i think i will leave sparc alone ;) 2017-02-01 07:41:53 yeah :) 2017-02-01 07:42:07 speaking of removing architectures 2017-02-01 07:42:13 i was thinking of dropping 32bit x86... 2017-02-01 07:47:21 if we do that, can we kill grsecurity too? 2017-02-01 07:48:06 i mean, i think we should check and see if anyone still wants x86 2017-02-01 07:48:22 but if nobody wants it, then that reduces the need for PaX considerably 2017-02-01 07:49:54 kaniini: i kinda like PaX 2017-02-01 07:50:32 I'd like to support it 2017-02-01 07:50:46 maybe someday someone will fork it 2017-02-01 07:50:47 on aarc64 its not supported though 2017-02-01 07:51:00 we kind of have forked it 2017-02-01 07:51:07 but on x86-64, pax is not really super needed 2017-02-01 07:51:14 kASLR provides pretty good protections 2017-02-01 07:51:24 but yeah, grsecurity situation is a bit non-ideal 2017-02-01 07:51:35 upstream is working on improving things 2017-02-01 07:51:42 which upstream 2017-02-01 07:51:58 ASLR in general is not super effective 2017-02-01 07:52:35 kaniini, vanilla linux 2017-02-01 07:52:36 it was implemented in pax due to it was cheap and provide some benefit 2017-02-01 07:52:50 ncopa, aslr != kaslr 2017-02-01 07:52:54 i know 2017-02-01 07:53:08 I have a 486 that I couldn't have lived with last week 2017-02-01 07:53:09 when it comes to kaslr, its no longer cheap 2017-02-01 07:53:17 userland aslr is relatively effective, but kernel aslr is very little gain for very high price 2017-02-01 07:53:18 dropping x86 is premature 2017-02-01 07:53:29 that I couldn't have lived without* 2017-02-01 07:53:41 so re kASLR, its not very super effective and it is fairly costly when it comes to effort 2017-02-01 07:53:44 but yes, aslr is not fixing anything; it's more of complication 2017-02-01 07:54:07 so i dont think kASLR is a great idea 2017-02-01 07:54:13 sure, but the same guy who did kASLR is doing a lot of the other PaX protections too 2017-02-01 07:54:25 i should have clarified 2017-02-01 07:54:30 yes, things are getting getter 2017-02-01 07:54:35 copperhead os ppl 2017-02-01 07:54:47 yes things are improving 2017-02-01 07:55:16 i think the main arch where grsecurity has most impact is 32-bit x86 though 2017-02-01 07:55:18 kaniini, yes, that's what i was meaning with 'upstream is getting better' now 2017-02-01 07:55:22 but i like the idea of being one step ahead of mainline 2017-02-01 07:55:41 yes, the biggest impact of PaX is on x86 2017-02-01 07:56:05 so yes, if we drop x86, we should consider grsecurity patch in general too 2017-02-01 07:56:21 my main concern with grsecurity is the maintenance situation of it. if we drop x86, it may be worth also dropping the grsecurity technical debt 2017-02-01 07:56:28 i don't know, just throwing it out there 2017-02-01 07:57:46 maybe we can assist the chromium os guys with porting the rest of the relevant grsecurity features to vanilla (i enabled yama which provides grsecurity ptrace restrictions in vanilla today after talking with jirutka about it) 2017-02-01 08:00:28 ncopa: i have no real opinion either way about dropping x86 (other than if we do drop x86, we should also look at how much value we get from keeping grsec around) 2017-02-01 08:01:22 i lean towards it may be too soon to drop x86 2017-02-01 08:01:41 it definitely is 2017-02-01 08:01:57 kaniini, yes, that's true. would be good to start discussion on getting the good bits to vanilla linux. grsec has not really been co-operative with us... 2017-02-01 08:02:19 that's the understatement of the century... 2017-02-01 08:03:28 i think from pax we mostly want the improved aslr, kernsec, and mprotect 2017-02-01 08:03:28 there seems to be no 4.9 patch out yet 2017-02-01 08:03:45 and 4.9 is longterm 2017-02-01 08:04:49 i think some of the grsecurity features like brute force prevention is kind of nice 2017-02-01 08:04:58 yama has some of that 2017-02-01 08:05:06 bruteforce prevention makes aslr more effective 2017-02-01 08:05:08 the ptrace stuff 2017-02-01 08:06:51 okay 2017-02-01 08:06:53 so 2017-02-01 08:07:22 the patch for 4.9 should been released yesterday, so its likely just around the corner 2017-02-01 08:07:30 lets say for 3.7 release series, that we want to get as much of the grsecurity features into vanilla as possible 2017-02-01 08:07:36 as kind of a starting point 2017-02-01 08:07:55 then at beginning of 3.8, we can see if we want to kill x86 2017-02-01 08:07:59 yes 2017-02-01 08:08:00 i don't think we should do it quite yet 2017-02-01 08:08:07 agreed 2017-02-01 08:08:18 but the time to do it is coming 2017-02-01 08:08:29 we could make the last x86 release longer support 2017-02-01 08:08:47 oh 2017-02-01 08:08:51 one important problem 2017-02-01 08:08:52 wine 2017-02-01 08:09:00 to run 32-bit wine stuff 2017-02-01 08:09:05 people need x86 built things 2017-02-01 08:09:15 why the fuck is killing x86 even an idea 2017-02-01 08:09:16 so either we need mutlilib, or x86 chroots 2017-02-01 08:09:56 we should get the buildbot + chroot based build system done 2017-02-01 08:10:10 so we don't need lxc for each alpine-branch + arch combination 2017-02-01 08:10:46 ncopa, would be nice to get the rootpkg stuff done, should figure out how to configure the repositories in it... :) 2017-02-01 08:11:37 ok, feel free to ignore me. Let it just be mutual. 2017-02-01 08:14:13 hmm, guess that means s6 won't be a future openrc replacement :p 2017-02-01 08:24:09 seems somebody didnt sleep enough last night... 2017-02-01 08:30:16 i think it seems to be more about his inability to post to alpine lists, actually 2017-02-01 08:31:28 he feels ignored about that issue, and that he will be ignored when x86 is killed (even though realistically we have kicked the can down the road to 2019 at the earliest by even looking at it in 3.8) 2017-02-01 08:32:36 is he really ignored, or did he not talk to the correct people? 2017-02-01 08:32:59 he wants something impossible: a whitelist entry for his dynamic ip 2017-02-01 08:33:12 i'm not even going to ask for that 2017-02-01 08:34:03 well, i dont like to loose good ppl here so im open for ideas. 2017-02-01 08:34:13 let me ask if he has ipv6 2017-02-01 08:34:21 maybe we can whitelist his /64 or such 2017-02-01 08:34:34 ...and that's a bad miss 2017-02-01 08:35:15 let me see how frequent his IPv4 changes 2017-02-01 08:35:32 i have no idea about alpine's smtp infra, its not managed/maintained by me. 2017-02-01 08:35:56 iirc it is managed by nangel 2017-02-01 08:36:04 who is very infrequently around 2017-02-01 08:36:10 so maybe we should change that 2017-02-01 08:36:14 i don't know 2017-02-01 08:36:22 i think ncopa also has access 2017-02-01 08:42:46 clandmeter: http://turtle.dereferenced.org/~kaniini/skarnet.txt 2017-02-01 08:46:57 sigh, running aways doesnt fix anything, and i cannot fix anything related to mail. 2017-02-01 08:48:00 i will just write to alpine-devel about it ;) 2017-02-01 08:48:06 how meta 2017-02-01 08:48:56 <_ikke_> haha 2017-02-01 09:15:12 @kaniini │ kASLR provides pretty good protections 2017-02-01 09:15:14 lol 2017-02-01 09:15:37 they are great protections 2017-02-01 09:15:40 the best protections 2017-02-01 09:15:58 give me a break it was 1:30am 2017-02-01 09:16:03 :p 2017-02-01 09:16:19 i was talking about the other stuff kees cook is doing 2017-02-01 09:16:26 kaslr is mostly a waste of time haha 2017-02-01 09:18:03 i still like grsec, even though it's in a sticky situation with the testing shit 2017-02-01 09:18:10 still a lot of stuff not in mainline and unlikely to be... 2017-02-01 09:18:12 (lol MPROTECT) 2017-02-01 09:19:11 while true, it may be possible to isolate the remaining features we want and maintain it ourselves 2017-02-01 09:19:27 which is what the proposal is really about -- try to upstream what we can, keep the rest locally 2017-02-01 09:19:44 so 'linux-grsec' becomes 'linux-hardened' or such 2017-02-01 09:20:25 my other thing there is that it seems rather hard to main ourselves 2017-02-01 09:21:25 right, the idea is more to just isolate the components and send them off to the kernel-hardening@lists.openwall.com guys and let them polish it up 2017-02-01 09:21:45 since they seem to be making decent progress with upstreaming stuff 2017-02-01 09:22:06 but it is a hard sales pitch to make sometimes to kernel people about proactive mitigations 2017-02-01 09:22:46 on the other hand, the current situation with grsecurity means we can't really say for sure that we are shipping reliable grsecurity kernels because they are experimental patches 2017-02-01 09:23:07 one could be good, one could be unstable 2017-02-01 09:23:12 yeah.. 2017-02-01 09:23:18 and spender doesn't seem very interested in fixing that 2017-02-01 09:23:29 why would he be? it is a revenue generator for him 2017-02-01 09:23:50 i would honestly not be surprised if he pulled something like 2017-02-01 09:24:08 "drop that non-genuine alpine grsec kernel and get genuine grsec kernels for $50/mo" 2017-02-01 09:24:44 it's a simple hustle 2017-02-01 09:25:39 I don't think he'd find that $50/mo Alpine's user base worth the work though 2017-02-01 09:25:44 :) 2017-02-01 09:25:53 well 2017-02-01 09:25:56 $5/mo 2017-02-01 09:25:59 $whatever/mo 2017-02-01 09:26:25 I don't think he'd care less about Alpine as a whole anyway 2017-02-01 09:26:28 he could probably get a few g's out of it 2017-02-01 09:27:20 that's nothing compared to what they get from selling support to stable subscribers 2017-02-01 09:27:32 maybe people can go buy grsecurity patches 2017-02-01 09:27:40 and then put them on non-free repo 2017-02-01 09:27:50 there's an idea 2017-02-01 09:27:52 sue spender 2017-02-01 09:27:58 saying grsec = GPL violation 2017-02-01 09:28:09 that they integrate Alpine's package builder in their builder service, that I could see ... still there's no user base for that 2017-02-01 09:28:21 kaniini: sure you want me to do it ? 2017-02-01 09:28:30 where's the GPL violation again ? 2017-02-01 09:28:43 coredumb: you are not allowed to redistribute the patch 2017-02-01 09:29:06 it is a textbook GPL violation 2017-02-01 09:29:22 that's false 2017-02-01 09:29:37 it is not false: if you redistribute it, your contract is terminated 2017-02-01 09:29:57 don't you think if it was as simple as buying the patches we would just do it 2017-02-01 09:30:23 <_ikke_> As I understand it, you are only required to distribute it to your customers 2017-02-01 09:30:27 <_ikke_> with GPL 2017-02-01 09:30:28 it's been said from the start that if you need to comply to GPL you do it by sharing it 2017-02-01 09:30:32 _ikke_: exactly 2017-02-01 09:30:44 and not forced to do so as a patch format 2017-02-01 09:30:53 that's exactly what RH does 2017-02-01 09:31:03 there's no violation of GPL there 2017-02-01 09:31:13 only ppl getting butthurt 2017-02-01 09:31:18 you're really twisting things around i see 2017-02-01 09:31:21 (I can understand why though) 2017-02-01 09:31:32 kaniini: sadly it's how the licensing works 2017-02-01 09:31:45 if you redistribute the grsecurity source that you acquired through your contract 2017-02-01 09:31:52 spender terminates the contract 2017-02-01 09:31:55 yes 2017-02-01 09:32:05 or I pay more to do so 2017-02-01 09:32:09 thusly, we are compelled not to share it 2017-02-01 09:32:21 how's that different from RH ? 2017-02-01 09:32:22 thusly, it puts us in position of violating GPL 2017-02-01 09:32:39 what are you talking about 2017-02-01 09:32:52 redhat kernel source SRPMs are publicly available 2017-02-01 09:33:25 yeah you're right it's the trademark of the binaries 2017-02-01 09:33:27 <_ikke_> RH sells access to their binary repos 2017-02-01 09:33:47 yes -- this is a different issue entirely 2017-02-01 09:34:11 if we (Alpine Linux) buy from spender his grsecurity source tree (or patch or whatever) 2017-02-01 09:34:19 and we redistribute it, we get banned 2017-02-01 09:34:33 not if you negociate a certain price to do so 2017-02-01 09:34:43 so this means we (Alpine Linux) cannot also distribute binaries compiled from said source tree (or patch or whatever) 2017-02-01 09:34:51 because to do so would be a GPL violation 2017-02-01 09:34:59 that indeed 2017-02-01 09:35:15 this behaviour is predatory and also a GPL violation 2017-02-01 09:35:19 but you should be able to comply with the GPL by giving the source tree 2017-02-01 09:35:25 it's not forbidden 2017-02-01 09:35:28 but you can't give the source tree 2017-02-01 09:35:31 he'll terminate your contract 2017-02-01 09:35:34 ^^^^^ 2017-02-01 09:35:44 or you negociate to be able to do so 2017-02-01 09:36:12 that's now what the GPL means 2017-02-01 09:36:13 at all 2017-02-01 09:36:17 not* 2017-02-01 09:36:31 I've never seen it anywhere said that he'd ban a subscriber for complying to the GPL 2017-02-01 09:36:55 he doesn't have to say it because he will not let anyone who would be in that position subscribe 2017-02-01 09:37:03 that's been speculations from the start 2017-02-01 09:37:10 except people have outright said so 2017-02-01 09:37:23 he himself has said so! 2017-02-01 09:37:54 i'm not drunk enough for this 2017-02-01 09:38:10 you got links ? cause he himself told me that if my customers ask me to comply to GPL I can do it 2017-02-01 09:38:46 spender's twitter is protected these days 2017-02-01 09:38:49 after if a customer shares the sources somewhere else ... who knows what happens there 2017-02-01 09:38:57 i'll fix skarnets mail problem as soon i understand what the exact problem is 2017-02-01 09:39:08 he should in theory be whitelisted 2017-02-01 09:39:08 coredumb: it's the "who knows what happens there" that is the problem 2017-02-01 09:39:40 we're not going to do "suck it and see" distribution 2017-02-01 09:39:59 kaniini: and I agree with that 2017-02-01 09:40:13 "who knows what happens there" means if someone makes an alpine derivative, it's not known how spender will react 2017-02-01 09:40:46 and since we're legitimately concerned about that, clearly it's not in the spirit of the GPL now is it 2017-02-01 09:40:46 this can surely be discussed 2017-02-01 09:41:14 maybe not in the spirit but doesn't violate it either 2017-02-01 09:41:21 which was my point all along 2017-02-01 09:41:22 yeah, it sort of does 2017-02-01 09:41:41 because the point of the GPL is that you can fork 2017-02-01 09:41:52 and that compromises the ability to fork 2017-02-01 09:41:54 sort of but wouldn't make it a term violation against the law 2017-02-01 09:42:28 yes it would 2017-02-01 09:42:44 i dont really understand why you're defending this tbh 2017-02-01 09:42:48 it is absolutely inane 2017-02-01 09:43:03 as shiz says, it would be bad. also, if he goes bonkers and sues somebody or some company for making a derivative of alpine including grsec, it is bad for alpine 2017-02-01 09:43:35 I do agree with that 2017-02-01 09:43:46 and yet you say it is GPL compliant 2017-02-01 09:43:58 you have fascinating interpretations of the GPL ;) 2017-02-01 09:44:59 and this is a problem with forking grsecurity based on the 'test' patches 2017-02-01 09:45:21 as long as I can comply with the license to the customers I give/sell binaries to .... well the interpretation stands 2017-02-01 09:45:23 there would be pressure on that fork to prove they have not looked at the embargoed version 2017-02-01 09:46:32 well if you'd use only the test patch - like gentoo does - there wouldn't be any problem 2017-02-01 09:46:52 if you don't have the "embargoed" version that is 2017-02-01 09:46:57 it gets muddy considering alpine has a custom patch 2017-02-01 09:47:13 the problem is that a fork would have new modifications added to it over time 2017-02-01 09:47:15 clearly this is not nice for a distribution 2017-02-01 09:47:21 did those modifications originate from embargoed grsecurity? 2017-02-01 09:47:24 who knows? 2017-02-01 09:47:30 the whole thing is a nightmare 2017-02-01 09:47:42 also, the gpl is not just about you being able to sublicense/redistribute coredumb 2017-02-01 09:47:46 your customers should be able to as well 2017-02-01 09:48:03 and if spender goes bonkers and sues your customers ... 2017-02-01 09:48:54 Shiz: I didn't say the opposite, but if you're allowed to share to your customers by Grsecurity (tm) I don't see how they can prevent your customers to share to whoever they want as they have no contract with them 2017-02-01 09:49:18 they can sue claiming that it is proprietary information 2017-02-01 09:49:27 that is covered by [the contract] and not actually GPL 2017-02-01 09:49:34 this is what we are saying 2017-02-01 09:49:49 kaniini: GPL applies he can't sue the customer as he has no contract with 2017-02-01 09:50:01 yes he can 2017-02-01 09:50:20 he doesn't have to be in the right to sue 2017-02-01 09:50:26 the legal system would be a lot easier otherwise 2017-02-01 09:50:28 :p 2017-02-01 09:50:28 that, and he can argue a few ways 2017-02-01 09:50:39 right but doesn't mean he'd win either ^^ 2017-02-01 09:50:54 for example: he can say that the code was illegally integrated into the GPL source without permission [the contract is what gives that permission] 2017-02-01 09:52:03 and anyway, while he may, in theory, be compliant with some twisted interpretation of the GPL, it does not change that embargoed grsecurity does not qualify as 'free software' per our guidelines either way... 2017-02-01 09:52:20 so we are stuck with modified test patches 2017-02-01 09:52:22 how can he proves that if the contracts of the initiator stipulates that he can comply to GPL by giving the sources ? 2017-02-01 09:52:48 giving does not mean it grants reuse permissions 2017-02-01 09:53:12 kaniini: clearly I agree with you that the stable patch is no "free software" anymore 2017-02-01 09:53:18 GPL grants reuse 2017-02-01 09:53:31 coredumb: i mean free as in freedom 2017-02-01 09:53:36 coredumb: not free as in $0 2017-02-01 09:54:23 just to be certain we are on the same page :) 2017-02-01 10:03:39 privately coredumb pointed out that saying we "ship grsecurity" is likely also a problem with trademarks 2017-02-01 10:03:47 perhaps we should change the wording to "ship grsec" 2017-02-01 10:05:18 was looking at the website at the same time see if it were noted somewhere 2017-02-01 10:05:54 after all, we do not ship official grsecurity nor have any plans to because it's embargoed 2017-02-01 11:08:00 <^7heo> fabled: it's not easy to go through your patches for rebasing the LUKS deported header... 2017-02-01 11:08:10 <^7heo> fabled: if you have time I'd appreciate some help/reviews. :) 2017-02-01 11:08:45 ^7heo, you mean nlplug-findfs? the async changes were needed for correctness and to fix other issues. i'm ok to do a review. 2017-02-01 11:10:13 <^7heo> fabled: I know the changes were needed and mine too are needed. 2017-02-01 11:10:21 <^7heo> fabled: that's not the issue here :) 2017-02-01 11:10:52 <^7heo> fabled: the point is more that it's hard to figure out what you did with the code and how and when via a merging history. 2017-02-01 11:10:57 <^7heo> (as opposed to a normal diff) 2017-02-01 11:11:40 <^7heo> fabled: for example, with the threading. 2017-02-01 11:12:13 <^7heo> fabled: there is no more threading init for example, from what I can see atm. 2017-02-01 11:18:16 wiw.. 2017-02-01 11:18:18 wow 2017-02-01 11:18:18 https://docs.google.com/document/d/1GCK53YDcBWQveod9kfzW-VCxIABGiryG7_z_6jHdVik/pub 2017-02-01 11:25:22 <^7heo> is that the gitlab incident report? 2017-02-01 11:25:27 <^7heo> (I don't open google links) 2017-02-01 11:25:44 yes 2017-02-01 11:25:56 ;v 2017-02-01 11:26:32 just wanted to see theis grafana, but it seems to be broken too ;3 2017-02-01 11:28:32 <^7heo> :P 2017-02-01 11:28:44 <^7heo> maybe it's being ddosed. 2017-02-01 11:31:12 s/theis/their/ 2017-02-01 11:31:22 ^7heo: http://monitor.gitlab.net/dashboard/db/postgres-stats?panelId=10&fullscreen&from=1485871271986&to=now 2017-02-01 11:32:02 <^7heo> scadu: do you want me to participate to the DDoS? 2017-02-01 11:32:03 <^7heo> :D 2017-02-01 11:32:09 anyway, -> -offtopic 2017-02-01 11:32:20 <^7heo> true 2017-02-01 18:45:31 If a package is showing up in pkgs.alpinelinux.org should it be available or will I have to wait some time to it be available to `apk add -U` ? 2017-02-01 18:45:40 http://pkgs.alpinelinux.org/package/edge/testing/x86_64/exim-postgresql is the package 2017-02-01 18:45:50 Build time was 45 mins ago 2017-02-01 18:46:49 <^7heo> ashb: on edge? 2017-02-01 18:46:53 testing 2017-02-01 18:47:00 oh, edge testing I guess 2017-02-01 18:47:13 (still kinda new to Alpine :D) 2017-02-01 18:47:58 <^7heo> afaik there's a little time between the build and the mirror sync 2017-02-01 18:49:10 <^7heo> try on http://rsync.alpinelinux.org/alpine/edge/testing/x86_64/ 2017-02-01 18:49:16 <^7heo> there seem to be the package there. 2017-02-01 18:49:31 i don't think giving out the master mirror url is always a good idea ;) 2017-02-01 18:49:48 <^7heo> kaniini: meh, it's the devel chan. 2017-02-01 18:49:48 rsync++ 2017-02-01 18:50:15 <^7heo> kaniini: I don't think security by obscurity only is a good idea either. 2017-02-01 18:50:21 it's not about security 2017-02-01 18:50:27 <^7heo> so, what's the issue? 2017-02-01 18:50:30 <^7heo> load? 2017-02-01 18:50:33 it's about ensuring people use the actual mirrors 2017-02-01 18:50:46 <^7heo> yeah well this is about determining if the software has been built. 2017-02-01 18:50:47 right 2017-02-01 18:51:03 <^7heo> I assume that ashb is the person who sent the aformentioned package 2017-02-01 18:51:08 <^7heo> and wants to ensure it's now available. 2017-02-01 18:51:15 makes sense 2017-02-01 18:51:18 Yes, that's me 2017-02-01 18:51:28 <^7heo> maybe I'm assuming wrong, but that's why I gave the rsync host. 2017-02-01 18:51:55 sure 2017-02-01 18:51:57 <^7heo> so now that it's on the master mirror, it's just a matter of time before it's available everywhere. 2017-02-01 18:52:00 <^7heo> ashb: ^ 2017-02-01 18:52:08 just don't use the master mirror for any sort of distributed images please ;) 2017-02-01 18:52:21 Of course 2017-02-01 18:52:21 <^7heo> the hell not :D 2017-02-01 18:52:22 Thanks 2017-02-01 18:52:25 <^7heo> you crazy :P 2017-02-01 18:52:36 <^7heo> "please DDoS the infra, people, we are rich!" 2017-02-01 18:52:56 is the CDN (still?) Fastly? 2017-02-01 18:53:51 <^7heo> if I recall correctly at some point it was some french company 2017-02-01 18:53:56 <^7heo> but I think things moved on. 2017-02-01 18:54:10 dns says flasty.net. 2017-02-01 18:54:19 (I could have checked before I asked I know realise) 2017-02-01 18:54:32 Oh and, thanks for merging the patch whoever did that 2017-02-01 18:54:48 <^7heo> probably jirutka :D 2017-02-01 18:55:16 what? I’m still writing review for these two fucking theses, so probably not me 2017-02-01 18:55:21 <^7heo> ah right 2017-02-01 18:55:23 <^7heo> I forgot. 2017-02-01 18:55:50 I think that committee will be quite surprised by length of it XD 2017-02-01 18:56:12 <^7heo> ashb: I dunno, it doesn't seem to have happened recently. 2017-02-01 21:35:47 hello, i'm trying to creating my first packe, but stuck at libtool: http://pastebin.com/BEyuws3L - APKBUILD is: http://pastebin.com/2mHJmJQm 2017-02-01 21:36:32 seems like you didn't set --prefix=/usr 2017-02-01 21:36:40 yep, you didn't 2017-02-01 21:36:49 the default prefix for a lot of ./configure is /usr/local 2017-02-01 21:36:55 but obviously .apk are distro packages 2017-02-01 21:36:57 so they should go in /usr 2017-02-01 21:37:10 so change ./configure to ./configure --prefix=/usr 2017-02-01 21:38:21 thanks! 2017-02-01 21:38:24 install -c .libs/rgxg /home/sdk/aports/main/rgxg/pkg/rgxg/usr/local/bin/rgxg 2017-02-01 21:38:24 new error :-) 2017-02-01 21:38:26 this is bad btw 2017-02-01 21:38:28 why do this? 2017-02-01 21:38:38 and the rm -rf below that 2017-02-01 21:39:29 removed "install -c..." 2017-02-01 21:39:54 the rm is from: https://wiki.alpinelinux.org/wiki/APKBUILD_examples:Simple 2017-02-01 21:40:20 right 2017-02-01 21:44:30 its done? 2017-02-01 21:44:32 >>> rgxg: Updating the main/x86_64 repository index... >>> rgxg: Signing the index... 2017-02-01 21:45:28 there you go 2017-02-01 21:45:30 :p 2017-02-01 21:45:46 now fill in the dependencies if it has any 2017-02-01 21:46:01 there are no 2017-02-01 21:46:35 then you can just omit the empty variables 2017-02-01 21:49:20 now i have to read: https://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package#Commit_your_work and figure out where the working APKBUILD have to go 2017-02-01 21:52:13 so just https://github.com/alpinelinux/aports/tree/master/testing/rgxg ? add the APKBULD file and PR ? 2017-02-01 21:53:53 yes, if you verified it works 2017-02-01 21:54:01 how? :) 2017-02-01 21:54:36 add your local repo directory to /etc/apk/repositories and apk update && apk add rgxg 2017-02-01 21:54:39 and see if it works 2017-02-01 21:55:21 what is my local repo directory? 2017-02-01 21:56:38 the wiki https://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package has no real helptext for that, right? 2017-02-01 21:57:40 it should tell you 2017-02-01 21:57:53 i think the default is /home//packages ? 2017-02-01 21:59:10 WARNING: Ignoring /home/sdk/packages/x86_64/APKINDEX.tar.gz: No such file or directory 2017-02-01 21:59:47 how to build the index? package is there 2017-02-01 22:00:34 apk add /home/sdk/packages/main/x86_64/rgxg-0.1.1-r0.apk 2017-02-01 22:00:35 is working 2017-02-01 22:01:26 sdk:/home/sdk/aports/main/rgxg# rgxg cidr 192.168.0.1/24 2017-02-01 22:01:27 192\.168\.0\.(25[0-5]|2[0-4][0-9]|1[0-9]{2}|[1-9]?[0-9]) 2017-02-01 22:01:30 :-) 2017-02-01 22:04:42 https://github.com/alpinelinux/aports/pull/819 2017-02-01 22:05:01 Shiz, like this? 2017-02-01 22:05:25 lgtm, but 2017-02-01 22:05:29 you should remove the following lines 2017-02-01 22:05:34 depends="" 2017-02-01 22:05:37 makedepends="" 2017-02-01 22:05:42 install="$pkgname.pre-install $pkgname.post-install" 2017-02-01 22:06:17 and i think the license should be normalized 2017-02-01 22:06:26 to license="zlib" 2017-02-01 22:06:45 since that's what other packages under the zlib license use 2017-02-01 22:07:39 https://github.com/alpinelinux/aports/pull/819/commits/85ec2e1d83cd4543728d247286364cf89db9c962 2017-02-01 22:07:49 Thank you Shiz! 2017-02-01 22:08:23 https://travis-ci.org/alpinelinux/aports/builds/197459959 2017-02-01 22:08:26 and now travis is testing it 2017-02-01 22:08:44 exciting 2017-02-01 22:09:26 and it succeeded 2017-02-01 22:09:30 party? 2017-02-01 22:09:51 >>> rgxg*: Tracing dependencies... 2017-02-01 22:09:53 pkgconfig 2017-02-01 22:10:05 i'm not sure if we declare pkgconfig as a dependency explicitly since it's so common 2017-02-01 22:10:08 should ask one of the devs 2017-02-01 22:10:30 looking at the package contents, you may want to add a separate dev package 2017-02-01 22:10:32 for the headers and such 2017-02-01 22:10:42 https://travis-ci.org/alpinelinux/aports/builds/197459959#L513 2017-02-01 22:11:05 if you add $pkgname-dev to subpackages= it should do a lot of that automatically 2017-02-01 22:11:19 i try :) 2017-02-01 22:15:14 https://travis-ci.org/alpinelinux/aports/builds/197462626 2017-02-01 22:23:57 and now, just wait? :) 2017-02-01 22:24:58 https://www.youtube.com/watch?v=W434iQM7Lzw 2017-02-01 22:27:10 no: "Please squash all commits into a single one per PR." 2017-02-01 22:27:56 its ok, we can do that too 2017-02-01 22:28:01 but, https://github.com/alpinelinux/abuild/blob/master/abuild.in#L638 2017-02-01 22:28:40 hmm? 2017-02-01 22:29:40 so https://wiki.alpinelinux.org/wiki/APKBUILD_examples:Simple is bad? 2017-02-01 22:30:16 wow 2017-02-01 22:30:19 we still have that? 2017-02-01 22:30:25 i used that... 2017-02-01 22:30:28 FIXIT! 2017-02-01 22:30:31 :) 2017-02-01 22:30:34 happy to help :) 2017-02-01 22:30:47 you can remove it and from wiki :) 2017-02-01 22:31:02 ill do that wiki 2017-02-01 22:31:14 This action has been automatically identified as harmful, and you have been prevented from executing it. In addition, to protect Alpine Linux, your user account and all associated IP addresses have been blocked from editing. If this has occurred in error, please contact an administrator. A brief description of the abuse rule which your action matched is: 002 2017-02-01 22:31:20 this wiki is killing me 2017-02-01 22:31:26 no github link on userpage 2017-02-01 22:31:31 and no removeing of this line 2017-02-01 22:31:32 wtf? 2017-02-01 22:32:03 strange 2017-02-01 22:32:29 wiki updated 2017-02-01 22:32:45 thanks 2017-02-01 22:32:56 yes, our wiki needs love 2017-02-01 22:33:10 specially regrading github 2017-02-01 22:33:25 but then again, we all need love. 2017-02-01 22:33:25 i wont tell you that i created: https://en.wikipedia.org/wiki/Stadtwiki_Karlsruhe 2017-02-01 22:33:41 so lets fix the wiki, tomorrow... 2017-02-01 22:34:41 https://github.com/alpinelinux/aports/pull/819/commits/a3048192c2095cc652ff0af7df5b2a1a680d634c 2017-02-01 22:35:07 btw, i dont got my validation mail from the wiki 2017-02-01 22:35:14 i tried twice 2017-02-01 22:35:44 oh no, not another email complaint... 2017-02-01 22:36:13 which is your acc? 2017-02-01 22:36:13 sorry 2017-02-01 22:36:21 nah dont worry 2017-02-01 22:36:26 not your fault 2017-02-01 22:36:30 https://wiki.alpinelinux.org/wiki/User:Hauke_L%C3%B6ffler 2017-02-01 22:36:46 i know thats not my fault, i hate mediawiki for over 10 years 2017-02-01 22:37:33 its shite 2017-02-01 22:37:48 so you have not been activated right? 2017-02-01 22:37:52 yes 2017-02-01 22:37:53 i never touch this thing 2017-02-01 22:38:29 wiki is important 2017-02-01 22:38:54 documentation is important 2017-02-01 22:39:12 and vendor lock-in into github no real alternative fact 2017-02-01 22:39:33 done 2017-02-01 22:39:49 so swicht to static html generated with something like hugo? 2017-02-01 22:40:05 i like musl's new wiki 2017-02-01 22:40:11 it might be useful for alpine too 2017-02-01 22:40:29 https://wiki.somasis.com/Home 2017-02-01 22:40:40 done what? 2017-02-01 22:40:40 its backend is git 2017-02-01 22:42:10 i like 2017-02-01 22:42:15 @mediawiki: Email address confirmation canceled 2017-02-01 22:42:29 wtf, lets try again 2017-02-01 22:43:09 Shiz: why its not linked to musl-libc.org? 2017-02-01 22:43:28 clandmeter: not yet :p 2017-02-01 22:43:31 it's under evaluation 2017-02-01 22:44:05 +1 2017-02-01 22:44:06 Your email address has now been confirmed. 2017-02-01 22:44:37 Shiz: i think we are looking for something similar 2017-02-01 22:45:44 i'll buy you a beer for killing the mediawiki 2017-02-01 22:46:24 ill buy you a sixpac if you convert our wiki to something managed in git ;p 2017-02-01 22:47:03 okay 2017-02-01 22:47:07 we have a deal 2017-02-01 22:47:11 with nachos if you need more ;-) 2017-02-01 22:47:21 you should ask Somasis clandmeter 2017-02-01 22:47:26 musl's current wiki is also mediawiki 2017-02-01 22:47:29 and she converted the articles for that 2017-02-01 22:47:35 :p 2017-02-01 22:47:39 pssst, i need the beer 2017-02-01 22:47:48 Shiz: i see 2017-02-01 22:47:49 the only good thing about mediawiki is export 2017-02-01 22:47:57 its using gollum 2017-02-01 22:48:05 do we want to keep the history? 2017-02-01 22:48:30 its not so big: https://wiki.alpinelinux.org/wiki/Special:Statistics 2017-02-01 22:49:19 i exported https://ka.stadtwiki.net/Spezial:Statistik once 2017-02-01 22:50:11 I think what we want is to first export the dev docs 2017-02-01 22:50:41 not sure if it makes sense to export it all. 2017-02-01 22:50:56 jirutka: ^ 2017-02-01 22:52:19 what’s going on? 2017-02-01 22:52:22 can you run https://phabricator.wikimedia.org/source/mediawiki/browse/master/maintenance/dumpBackup.php for me? 2017-02-01 22:53:51 but i have to leave now, can you drop me a mail, i'm serious if beer is involved, and i'm an wiki / knowledge management export 2017-02-01 22:54:06 hm, Gollum… it’s very limited, like too much limited 2017-02-01 22:54:15 but it may be sufficient for us 2017-02-01 22:54:33 however, I would strongly prefer to use AsciiDoc instead of Markdown, Gollum supports both 2017-02-01 22:55:11 and also not rush with migration… 2017-02-01 22:55:41 we should *think* about the structure first, prepare graphics etc. 2017-02-01 22:56:14 pandoc's variant of Markdown 2017-02-01 22:56:41 also keep in mind that converting between document formats is never lossless 2017-02-01 22:56:56 sure, i'm happy to help with (media)wiki 2017-02-01 22:56:58 still Markdown, still very limited 2017-02-01 22:57:12 minimal not limited 2017-02-01 22:57:17 just like alpine :P 2017-02-01 22:57:20 no, limited 2017-02-01 22:57:38 better than mediawiki, easier to convert 2017-02-01 22:57:43 ill leave you two at it, its bed time for me. 2017-02-01 22:58:03 i'm out too 2017-02-01 22:58:07 pro me a line at https://wiki.alpinelinux.org/wiki/User_talk:Hauke_L%C3%B6ffler 2017-02-01 22:58:12 jirutka: go to bed on time for ones! 2017-02-01 22:58:17 i'm happy to help 2017-02-01 22:58:25 well, mediawiki syntax is pretty bad, but it’s actually more expressive than Markdown… well, everything is more expressive XD 2017-02-01 22:58:26 devlaa: ok np thx. 2017-02-01 22:59:07 clandmeter is very right, I must to go sleep now, b/c I must wake up at about 5 AM at Friday to catch up plane to Brussels 2017-02-01 22:59:09 hi 2017-02-02 03:20:04 that GNU sed 4.3 update was nasty 2017-02-02 03:20:17 i had to hack the APKBUILD to delete sed on a builder 2017-02-02 03:20:26 because it crashed while running autoconf 2017-02-02 03:20:27 hahahahah 2017-02-02 03:20:55 but gnu sed 4.3 is now patched with the upstream fix for it, hopefully they ship 4.3.1 soon 2017-02-02 08:00:39 fabled: i think rootbld is getting more necessary -- build-edge-x86 was in an inconsistent state and broke when i updated gnu sed to fix using it with autoconf scripts 2017-02-02 08:01:21 kaniini, yes, that's step #1 2017-02-02 08:01:53 step #2 is to write buildbot rules/logic that builds all branches on all arches, so we don't need LXC per arch+branch combination 2017-02-02 08:02:13 and so that it also can be used to do CI 2017-02-02 08:02:20 i had to do some very strange 2017-02-02 08:02:25 very strange things 2017-02-02 08:02:32 to unbreak build-edge-x86 2017-02-02 08:02:49 involving invoking abuild-apk from prepare() 2017-02-02 08:02:51 :D 2017-02-02 08:03:15 we usually ssh in and fix it by hand in that kind of situations 2017-02-02 08:03:29 yeah 2017-02-02 08:03:33 nobody was around at the time :p 2017-02-02 08:03:48 so i realized i could just temporary add a prepare() to do it 2017-02-02 08:03:51 :D 2017-02-02 08:04:00 :) 2017-02-02 08:04:25 heh 2017-02-02 08:05:02 but the more worrisome news there is 2017-02-02 08:05:18 all of the xulrunner deps were installed 2017-02-02 08:05:29 huh 2017-02-02 08:05:38 so i think some packages were built with those deps available on x86 edge 2017-02-02 08:06:44 so at some point, we will probably get bug reports from some edge/x86 user about some feature that disappeared that was only on edge 2017-02-02 08:08:35 yeah, that's possible 2017-02-02 08:08:54 well, if someone wants them let them file bug so we can add the explicit dependencies 2017-02-02 08:09:02 ACTION nods 2017-02-02 09:41:18 and that’s why I always prefer to explicitly specify with/use flags and not rely on autodetection 2017-02-02 09:48:53 kaniini: rootbld work in progress: http://tpaste.us/vMM5 2017-02-02 09:48:58 most work done by fabled 2017-02-02 09:49:17 we need fix the --repository lines 2017-02-02 09:49:50 the idea with the last hunk is to be able to specify the repository dependencies 2017-02-02 09:50:02 so in testing we have repo_deps="main community" 2017-02-02 09:50:12 in community we have: repo_deps="main" 2017-02-02 09:51:10 then we need to split the --repository line in 3 parts: $mirror/$relbranch/$repo 2017-02-02 09:52:30 i think $mirror should be possible to set system wide, eg in /etc/abuild.conf and be overriden by user config, eg ~/.abuild/abuild.conf 2017-02-02 09:52:52 relbranch should be gitrepo specific 2017-02-02 09:53:26 so i think we should have an aports.git/.abuild.conf or similar 2017-02-02 09:54:20 if reblranch in unset then we can set default based on current git branch 2017-02-02 09:54:32 ncopa, abuild already sources "aports.git/.abuild" 2017-02-02 09:54:49 see functions.sh:90 2017-02-02 09:55:25 ah, nice 2017-02-02 09:55:45 so we could set relbranch there 2017-02-02 09:56:31 we also need to fix the cleanup bits, which requires root permissions 2017-02-02 09:57:00 i wonder if we should turn the root permissions around 2017-02-02 09:57:06 currently we dont allow run abuild as root 2017-02-02 09:57:11 but we need to elevate the privs 2017-02-02 09:58:17 i wonder if we should allow abuild run as root, and if is root, drop privs whenever not needed 2017-02-02 09:58:31 and maybe require you run as root for rootbld 2017-02-02 10:08:00 be careful when upgrading docker pkg… https://twitter.com/jakubjirutka/status/827095696973107200 2017-02-02 10:09:14 <_ikke_> Oh, that's nasty 2017-02-02 10:09:44 “nasty” is a very weak word for this :) 2017-02-02 10:12:49 <_ikke_> https://github.com/docker/docker/issues/14041 for context 2017-02-02 10:14:20 changing your forward policy regardless of the previous state is NOT a correct solution 2017-02-02 10:26:36 i still have :FORWARD ACCEPT [0:0] 2017-02-02 10:27:02 hm, reading the diff more closely, it changes the forward policy only if ip_forwarding has not been enabled yet 2017-02-02 10:27:38 if I understand the code correctly 2017-02-02 10:27:50 <_ikke_> That is in line with the bug report 2017-02-02 10:27:59 i think that makes sense 2017-02-02 10:28:13 <_ikke_> ie, docker creating an insecure situation by enabling forwarding, but not blocking by default traffic 2017-02-02 10:28:16 so it seems to be ok-ish, but still it’s a breaking change that affected users (see thread on twitter) 2017-02-02 10:28:56 <_ikke_> jirutka: Just curious, how do you think docker should handle this situation? 2017-02-02 10:30:11 _ikke_: well, at the first place, when they mess with your network settings and enables ip forward, then they should took care about forward policy from the beginning 2017-02-02 10:30:41 <_ikke_> right, but that hasn't been the case 2017-02-02 10:31:34 thats what they are trying to fix 2017-02-02 10:33:03 _ikke_: yes; so meanwhile there’s a great chance that some users rely on this broken behaviour (and there really are such users), so it’d be imho better to add warning message at start that this will be changed in next release 2017-02-02 10:33:28 and after change write it very explicitly, not bury it in awfully long changelog https://github.com/docker/docker/blob/master/CHANGELOG.md#1130-2017-01-18 2017-02-02 10:33:31 if you have a service that depends on a vulnerable behavior, then that service will break when you fix the vulnerability 2017-02-02 10:35:32 they have changed it in minor release, so no wonder that many users didn’t notice it 2017-02-02 10:35:56 <_ikke_> well, 'bug' fixes often happen in minor releases 2017-02-02 10:36:46 but, yes, messing with system networking it tricky 2017-02-02 10:37:15 but I admit that first I thought that it changes iptables regardless of previous state, that’s not true 2017-02-02 10:38:33 yeah, that’s why I don’t like even the fact that it is messing with system networking 2017-02-02 10:39:00 I would understand adding rules for port forwarding, but not enabling ip_forward in your behalf 2017-02-02 10:39:34 you can config it to not do so i think 2017-02-02 10:39:53 but yeah 2017-02-02 10:40:04 when they do that, it must be done with great caution and that was not the case… that’s why I consider it yet another indication why it’s not trustful/reliable 2017-02-02 10:40:20 yes, there’s some config option to disable messing with iptables 2017-02-02 10:40:36 <_ikke_> --iptables=false 2017-02-02 10:40:54 i suppose you have same issue with libvirt 2017-02-02 10:41:01 which also messes with iptables 2017-02-02 10:41:49 well, that’s typical problem with software that do too much, with the excuse to user experience (= even monkey without any knowledge can use it) 2017-02-02 10:42:00 yes 2017-02-02 10:42:04 exactly 2017-02-02 10:42:14 they try to make it simple for the user 2017-02-02 10:42:17 you probably know that I wrote https://github.com/jirutka/qemu-openrc, so I don’t use libvirt :) 2017-02-02 10:42:17 just install docker and run 2017-02-02 10:42:21 and everything works 2017-02-02 10:42:33 well, actually, I still use it on some legacy servers and really hate it 2017-02-02 10:42:39 :) 2017-02-02 10:42:57 i think docker is kinda cool 2017-02-02 10:43:21 it makes it very very simple for me to bring up a debian toolchain to testcompile things 2017-02-02 10:43:25 I really don’t, but also I admit that this change is not as bad as I thought 2017-02-02 10:43:49 you don’t need docker to make it equally easy with Alpine 2017-02-02 10:44:01 so it means that Docker is just a workaround for broken things… 2017-02-02 10:47:10 yeah, you need to use docker with care 2017-02-02 10:49:12 I should note this: Docker started as a tool for developers and in this field it’s ok-ish; but now it claims that it’s good even for production and that’s what I see as a problem 2017-02-02 10:49:36 jirutka: we all see the problem :D 2017-02-02 10:58:44 i woundt mind using docker in prod 2017-02-02 10:58:49 just need to be a bit careful 2017-02-02 10:59:41 a bit careful that it may break and fall apart anytime? :) 2017-02-02 11:00:25 you have the similar issue when you give a powerful linux server in the hands to a newbie ubuntu user 2017-02-02 11:02:29 hm, actually yes, b/c ubuntu is also quite broken… 2017-02-02 15:48:39 installing cargo seems to be broken on edge? 2017-02-02 15:48:57 the package requires libgit 0.24, but libgit 0.25 is latest 2017-02-02 15:53:47 I’ll bump it to rebuild 2017-02-02 15:59:04 <^7heo> jirutka: how do we handle packages that need environment cleaning for an upgrade? 2017-02-02 15:59:33 <^7heo> let's say, a package was using its home folder as /foo/bar and now it is /var/lib/foobar 2017-02-02 15:59:55 <^7heo> so you need to clean up the environment BEFORE you create it for the package again. 2017-02-02 16:00:28 <^7heo> do we have some post-uninstall scripts? 2017-02-02 16:00:40 <^7heo> and if yes, are they applied to upgrade? 2017-02-02 16:04:10 what do you mean by clean-up environment? 2017-02-02 16:04:17 move the directory on behalf of user? definitely not 2017-02-02 16:04:28 just add post-install message that warns about this change 2017-02-02 16:04:48 oh, that reminds me, we should discuss at FOSDEM how to properly handle post-inst messages, finally 2017-02-02 16:05:06 <^7heo> yeeeo. 2017-02-02 16:05:08 <^7heo> s/o/p/ 2017-02-02 16:05:10 <^7heo> and doc' 2017-02-02 16:17:52 hi 2017-02-02 16:18:58 is there a metadata of all aports (e.g. info contained in APKBUILD, but without need to parse it manually) available somewhere online? 2017-02-02 16:26:09 AMDmi3: like pkg.alpinelinux.org ? 2017-02-02 16:32:10 coredumb: yes, but it machine-readable form (e.g. something like json) and for source APKBUILDS, not binary packages (e.g. I don't need to see -dev/-doc/-progs/-libs subpackages) 2017-02-02 16:32:41 AMDmi3: no idea then 2017-02-02 17:06:58 <^7heo> AMDmi3: find -exec? 2017-02-02 17:08:09 ^7heo: I'd prefer not to due to security reasons 2017-02-02 17:10:20 <^7heo> wat? 2017-02-02 17:11:11 <^7heo> Now if someone could point out how that is relevant, I'd be very happy. 2017-02-02 17:13:19 it's running third party shell scripts in bulk, this is by definition not safe 2017-02-02 17:13:37 also I still see no way to get variables from shell 2017-02-02 17:33:10 what are you trying to do? 2017-02-02 17:33:21 How is `find` running anything other than the command you give it? 2017-02-02 17:33:51 I'm trying to collect metadata of all aports for repology.org 2017-02-02 17:34:00 and by metadata you mean? 2017-02-02 17:34:18 name, version, homepage, download urls, licenses, maintainer etc. 2017-02-02 17:35:20 AMDmi3: there’s https://github.com/alpinelinux/lua-aports 2017-02-02 17:36:21 AMDmi3: and also RESTful API https://github.com/insteps/aport-api 2017-02-02 17:36:36 it’s deployed on alpinelinux.org, but don’t remember address 2017-02-02 17:36:56 Or if you want to do it the hard way, download the APKINDEX.tar.gz, unpack it, check the sig then look at the APKINDEX structured text file 2017-02-02 17:37:36 ashb: that would be an easy way, but that's data on binary packages, while I need info on source APKBUILDS 2017-02-02 17:37:36 there's an `api.` domain but it returns a 502 on / 2017-02-02 17:38:02 AMDmi3: another approach is to use https://github.com/alpinelinux/aports-turbo (that’s what’s running on pkgs.alpinelinux.org) and build your own SQLite DB 2017-02-02 17:38:35 from a quick look all the info on (for example http://pkgs.alpinelinux.org/package/edge/testing/aarch64/py2-opencolorio) is in the APKINDEX 2017-02-02 17:38:47 yes 2017-02-02 17:38:48 https://www.irccloud.com/pastebin/9c0MIngp/ 2017-02-02 17:39:02 but APKINDEX currently does not contain pkg content (files list) 2017-02-02 17:39:25 it depends on what you need 2017-02-02 17:39:30 yes, I've seen that; I could also just parse a website. but again, this is info on binary packages, that's not what I need 2017-02-02 17:39:35 AMDmi3: what do you mean but "info on source"? Specifics? 2017-02-02 17:39:41 I don't quite understand that phrase 2017-02-02 17:40:22 binary packages are duplicated for different architectures and split into sub-packages. I don't need that 2017-02-02 17:41:07 AMDmi3: btw it’s quite easy to just source APKBUILD files and echo variables that you need; for example https://gist.github.com/jirutka/6717f68c7f76c9425b21d0cbe2eaa007#file-abuild-origin-index-L5-L14 2017-02-02 17:41:47 AMDmi3: these are some really ugly hackish scripts I wrote for particular problem when I need to find all reverse dependencies, incl. build dependencies 2017-02-02 17:42:01 AMDmi3: but maybe it may be useful for inspiration 2017-02-02 17:42:04 jirutka: like I've said, it's insecure, and I'd prefer to not call any external commands 2017-02-02 17:42:27 I dispute your claim of it being insecure 2017-02-02 17:42:31 AMDmi3: maybe I missed it, but you have not said what do you need to do with it… 2017-02-02 17:43:18 ashb: it's shell script, which can do all sorts of things when sourced 2017-02-02 17:43:27 yes but you control that script 2017-02-02 17:43:34 AMDmi3: also it’s not very true that it’s insecure, ofc there’s no rm -rf / in any APKBUILD, I hope that no one would merge such thing to aports tree… but I understand that you don’t want to use this approach, but hard to give better advice when I don’t know exact use case ;) 2017-02-02 17:43:42 That's like saying "python code is insecure, it can `rm -rf`" 2017-02-02 17:44:00 ashb: I don't control it 2017-02-02 17:44:01 Also: you don't have to use the upstream script, you can bring it in to your tree 2017-02-02 17:44:06 Then done, you do control it 2017-02-02 17:44:20 or look at what the script does and re-implement it 2017-02-02 17:44:48 AMDmi3: you may also try to use https://github.com/jirutka/sh-parser, it’s still work in progress, but parser itself is basically complete and working, I just need to improve AST… you can use it to parse APKBUILDs without evaluating them and extract variables 2017-02-02 17:45:22 AMDmi3: but again, depends on what you need to extract… some variables may be dynamically generated, so you can’t simply extract the content without doing evaluation 2017-02-02 17:45:59 And do you actually need to look at the APKBUILD - or can you just look at the Project field and group sub-packages yourself? 2017-02-02 17:46:17 ashb: I need a fresh info from all PKGBUILDS. to get it I need to bulk - run all APKBUILDS. which are code I don't control which can do anything including rm -rf. no, I'm not doing it 2017-02-02 17:46:56 why do you even care about APKBUILDs. Surely what is actually in the repo is the interesting thing 2017-02-02 17:47:04 and the source of truth for that is apkindex 2017-02-02 17:47:24 I see now, http://repology.org, this looks interesting… where does it take information about upstream versions? we currently use https://release-monitoring.org/ for that 2017-02-02 17:47:38 ashb: no, this information is derived from APKBUILDS, and this information is mangled. I need the source 2017-02-02 17:48:00 Can you give me a specific example? I don't see it myself 2017-02-02 17:48:00 jirutka: it doesn't, for now in only compares versions in a lot of repositories 2017-02-02 17:48:22 we have already around 60 % packages in main/community mapped in release-monitoring, so I can use this information to fill data into http://repology.org/ 2017-02-02 17:49:02 AMDmi3: how does it map pkgs between repos? this is quite problematic… 2017-02-02 17:49:09 AMDmi3: b/c just a pkg name is not sufficient 2017-02-02 17:49:47 AMDmi3: that’s why release-monitoring allow to specify distro pkg name for each package and when I filled it for Alpine, it was a semi-automatic process, it can’t be reliable done fully automatically :/ 2017-02-02 17:49:53 jirutka: it uses pkg name by default and a ruleset for exceptions. for example, it maps all perl packages with a single rule 2017-02-02 17:50:18 in release-monitoring there are many pkgs with the same name that are actually different projects… 2017-02-02 17:50:22 how do you handle this? 2017-02-02 17:50:41 Or `nc` on Centos and `netcat` on debian 2017-02-02 17:52:47 repology can rename a packages in any way. e.g. place 'nc' in 'centos' under 'netcat' name, so it matches netcats in all other repos 2017-02-02 17:53:01 or vice versa, place all netcats under 'nc' 2017-02-02 17:53:11 nice 2017-02-02 17:53:15 how? 2017-02-02 17:53:31 manually or some heuristic guessing? 2017-02-02 17:53:37 jirutka: manually 2017-02-02 17:53:44 jirutka: this should give an idea: https://github.com/AMDmi3/repology/blob/master/rules.yaml 2017-02-02 17:54:05 cool! 2017-02-02 17:54:17 bulk ones are particularly interesting: https://github.com/AMDmi3/repology/blob/master/rules.yaml#L624 2017-02-02 17:54:39 indeed, this is really interesting approach 2017-02-02 17:54:58 basically I build similar, but much simpler and ad-hoc, rules to fill our pkgs into release-monitoring 2017-02-02 17:55:26 handling this on the side of the application is better approach 2017-02-02 17:55:50 I’ll take a look at repology ASAP, it looks very interesting 2017-02-02 17:55:56 sorry, afk for ~20 mins 2017-02-02 17:56:23 it would be very useful to integrate it with release-monitoring 2017-02-02 17:56:44 to get informaton about upstream versions 2017-02-02 17:57:24 with this database and set of rules it may not be so hard to find matching projects and provide this info 2017-02-02 17:58:04 afk too 2017-02-02 18:08:52 so 2017-02-02 18:09:28 well, I don't think release-monitoring `opt-in' approach is viable 2017-02-02 18:10:26 it requires a lot of work for set-up and maintanance, while packages with different names are actually exceptions and should be treated like such 2017-02-02 18:11:32 as a result, release-monitoring tracks 12k projects and repology tracks 108k, picking up new projects automatically 2017-02-02 18:12:26 also, a mismatch in repology may be fixed by a single person for all repositories at once, while r-m needs input from all distros 2017-02-02 18:12:55 coalescing by project URL/homepage might find some more matches too 2017-02-02 18:13:23 anyway, back to aports. I too tend to think that a custom parser is needed for them. it could also be useful for gentoo ebuilds and arch pkgbuilds, but these seem to be more complex 2017-02-02 18:13:25 though might be too broad in others 2017-02-02 18:15:03 ashb: yes, but only as a way to suggest potential mismatches; these are not really reliable 2017-02-02 18:15:10 AMDmi3: so the info you want is "name, version, homepage, download urls, licenses, maintainer" per origin package, right? 2017-02-02 18:15:30 (YEah, I'd imagine a lot of project urls are just freedesktop.org or similar) 2017-02-02 18:16:43 ashb: yes; currently used fields in addition to name + version are listed here: https://github.com/AMDmi3/repology#repository-support 2017-02-02 18:18:32 So I think everything that is in the APKBUILD file for a package in that list is also in the APKINDEX file 2017-02-02 18:19:27 ashb: yes, but again it's not 1:1 match to APKBUILDs, because there are subpackages 2017-02-02 18:19:39 Look at/group by the `o:` field 2017-02-02 18:19:56 yes, but summary is mangled 2017-02-02 18:20:05 There is no summary in the APKBUILD that I see 2017-02-02 18:20:21 http://git.alpinelinux.org/cgit/aports/tree/main/dovecot/APKBUILD is the specific example I'm looking at 2017-02-02 18:20:33 ashb: pkgdesc 2017-02-02 18:20:48 Ah oh yes. 2017-02-02 18:21:17 if there's a way to remove subpackage addendum from it I'm probably cool with APKINDEX 2017-02-02 18:21:34 oh, there's no downloads also 2017-02-02 18:21:37 I think where `P:` == `O:` 2017-02-02 18:21:44 The download url can be calculated 2017-02-02 18:22:20 ${package}-${ver}.apk in the same folder as the APKINDEX.tar.gz download 2017-02-02 18:22:35 nono, upstream source download url 2017-02-02 18:22:37 Ah 2017-02-02 18:23:05 P: = o: might work, is it guaranteed for such package to exist? 2017-02-02 18:23:34 That I couldn't say I'm afraid 2017-02-02 18:23:46 I don't know enough to say yes or no 2017-02-02 18:24:24 actually that can be checked quite easily 2017-02-02 18:25:11 hmm, I suspect no 2017-02-02 18:25:16 P:lua5.1-microlight o:lua-microlight 2017-02-02 18:27:29 Ah yeah, and there'll be similar things for php5 and php7 2017-02-02 18:27:47 uh huh 2017-02-02 18:28:14 Can you just group by origin and only process it the first time you see a new origin value? 2017-02-02 18:28:58 Oh wait there is a lua-microlight package too 2017-02-02 18:29:31 I don't think that will work; for example, P:boost-serialization, o:boost 2017-02-02 18:30:20 that's a sub project of boost 2017-02-02 18:30:43 http://git.alpinelinux.org/cgit/aports/tree/main/boost/APKBUILD?h=3.5-stable#n18 2017-02-02 18:31:07 yes, in this case where's P:boost, but what if it was just split into sublibraries - picking one of them would be incorrect 2017-02-02 18:33:10 They all share a version, URL, License don't they? 2017-02-02 18:33:52 It's not perfect, but the other eway involves parsing/executing bash in some way (I see what you mean now about insecure, sorry. I didn't get you were talking about the APKBUILD files for some reason.) 2017-02-02 18:34:04 probably, but not desc 2017-02-02 18:34:23 hmm, quick test shows there's no such o: which doesn't have P: for it 2017-02-02 18:34:29 so P: = o: should work 2017-02-02 18:34:48 but, I've just discovered a funny thing, a sec 2017-02-02 18:34:57 the apkindex format is not stable (for example, apk-tools 3.x is planned to switch to btrees) 2017-02-02 18:35:28 kaniini: good to know. It will still be machine parsable without having to run bash though 2017-02-02 18:35:44 haha, yes: o:Libraries for libcdda_paranoia (Paranoia III) 2017-02-02 18:35:58 http://git.alpinelinux.org/cgit/aports/tree/main/cdparanoia/APKBUILD 2017-02-02 18:36:21 should probably be s/pkgname/pkgdesc/ in libs() 2017-02-02 18:36:34 :D 2017-02-02 18:36:49 Whoops :D 2017-02-02 18:37:45 ok, I'll go P:=o: way for now; downloads is not a big loss, there's no way to parse them for most repos 2017-02-02 18:37:58 desc is in the `T:` field 2017-02-02 18:39:10 I mean picking data from APKINDEX entries with P: = o: 2017-02-02 18:40:16 AMDmi3: http://git.alpinelinux.org/cgit/aports/commit/main/cdparanoia/APKBUILD?id=b805835f4e89f39fdd1eed35bb737669fd0df1f0 2017-02-02 18:40:54 kaniini: :+1: :) 2017-02-02 18:42:39 what's url for APKINDEX again? 2017-02-02 18:43:26 edge is rolling, right? 2017-02-02 18:43:48 edge is basically rolling yeah 2017-02-02 18:43:58 http://cdn-dl.alpinelinux.org/alpine/$branch/$repo/$arch/APKINDEX.tar.gz 2017-02-02 18:45:14 hmm, is fetching x86_64 only arch viable? there could be e.g. x86-only packages... 2017-02-02 18:46:26 AMDmi3: is this for repology? 2017-02-02 18:52:49 kaniini: yes 2017-02-02 18:53:11 another question, versions have -rN suffix; is this the only thing used in aports? 2017-02-02 18:53:31 e.g. FreeBSD uses portrevisions and portepochs 2017-02-02 18:54:50 to my knowledge, yes 2017-02-02 18:55:48 thx; also could there be multiple maintainers? 2017-02-02 18:56:24 seems like no 2017-02-02 19:02:46 hmm, there are 4612 APKBUILDs in the repo, but I've only parsed 3828 from indexes 2017-02-02 19:03:21 probably the missing ones are from unmaintained/ 2017-02-02 19:10:38 or only in testing? 2017-02-02 19:10:46 exim for instance is only in testing 2017-02-02 19:14:35 well I've added alpine support; I'll deploy it tomorrow morning 2017-02-02 19:28:30 AMDmi3: in situation where a package is formally maintained by a group we would point it to the mailinglist for that group :) 2017-02-02 19:28:58 kaniini: that's ok 2017-02-02 19:47:38 AMDmi3: I’m back 2017-02-02 19:48:34 AMDmi3: okay, you’re basically solving the same problem as I did few months ago, so I can help with it 2017-02-02 19:48:44 AMDmi3: Alpine uses Gentoo-style versioning format, this is described here https://devmanual.gentoo.org/ebuild-writing/file-format/#file-naming-rules 2017-02-02 19:49:26 AMDmi3: we have just few extra pre-release suffixes: 'cvs', 'svn', 'git', 'hg', 'p' 2017-02-02 19:50:00 AMDmi3: there’s a library I wrote for converting various versioning formats into valid Gentoo-style versioning formats: https://github.com/jirutka/gversion.lua 2017-02-02 19:50:33 AMDmi3: it works for, I would say, ≥96 % of projects; so every project that use some *sane* versioning format 2017-02-02 19:51:26 AMDmi3: as you already found, you should focus just on “origins”, i.e. you can ignore subpackages; subpackages has always the same source a version number 2017-02-02 19:51:38 that will probably work out of box with repology 2017-02-02 19:53:18 AMDmi3: I would really recommend to do integration with Anitya (release-monitoring.org); the side effect is that you will get database of more than half Alpine packages with useful metadata that will help you to match the projects 2017-02-02 19:53:40 AMDmi3: ten you can parse APKINDEX just to sync versions 2017-02-02 19:54:35 AMDmi3: if you want clean long-term solution how to get these data from Aports, then I’m afraid that we don’t have it (yet) 2017-02-02 19:56:14 AMDmi3: however, I think that we can build some solution suitable for repology in short-term 2017-02-02 19:56:18 I don't think release-monitoring may be useful; repology has a lot more repos and will have to match them at once, so even if some data is taken from r-m, everything will need to be re-reviewed for other repos 2017-02-02 19:56:51 regarding alpine, I just did support for all its packages. will test & deploy tomorrow 2017-02-02 19:57:11 there will be some mismatches, but still it's already better than r-m and needs less effort 2017-02-02 19:57:13 AMDmi3: release-monitoring provides mapping between upstream project name, homepage and version to name of packages in various distribution 2017-02-02 19:57:27 I understand that 2017-02-02 19:57:46 AMDmi3: and the provide RESTful API and fedmsg (message bus on ZeroMQ) that send push notifications about updates 2017-02-02 19:58:19 repology has an api as well 2017-02-02 19:58:47 AMDmi3: well, the database may be smaller, but they’re solving IMHO quite different and harder problem – to monitor upstreams… this means that in many cases they must parse version number from text on some random HTML pages of projects 2017-02-02 19:59:55 I've had that in plans too 2017-02-02 20:00:12 AMDmi3: what are use cases and target audience of repology anyway? 2017-02-02 20:00:28 jirutka: it's described in about 2017-02-02 20:00:47 package maintainers and upstream authors primarily 2017-02-02 20:01:42 AMDmi3: I’m looking at it from the position of distribution – we needs to know when some upstream project, we have package for, is updated, and to which version, so we know that we must updated the package; we currently automatically send mail to maintainer that (s)he should bump the pkg; so it’s very important to have accurate data 2017-02-02 20:02:13 AMDmi3: it’s better to have smaller number of monitored packages than have all of them, but with 10 % errors 2017-02-02 20:03:00 I'm of opposite opinion - I see no point in monitoring only a part of packages 2017-02-02 20:03:13 I don't plan to bother with emails, but RSS are planned 2017-02-02 20:03:31 AMDmi3: the reason why there are maybe not so many pkgs on Anitya is that there’s no way how to automatically determine where and how to check for new version 2017-02-02 20:03:49 AMDmi3: not all pkgs are on GitHub, Pypi or other places where you can easily find pkg by name and get version number 2017-02-02 20:03:57 there actually is 2017-02-02 20:03:59 AMDmi3: so this is a manual process in many cases 2017-02-02 20:04:03 no, there is not, from principle 2017-02-02 20:04:10 you can handle many cases, but definitely not all 2017-02-02 20:04:24 you can never handle all cases 2017-02-02 20:04:41 ask Anitya guys… 2017-02-02 20:05:12 well anything which starts with manual won't work 2017-02-02 20:05:21 that’s not exactly true 2017-02-02 20:05:55 also RSS is nice, but more suitable for automation is some messaging system, MQTT, ZeroMQ or something like that 2017-02-02 20:06:09 it is; manually entered data will rot away in months if not in weeks 2017-02-02 20:06:49 this is how we use fedmsg using plain zeromq lib to get “push notifications” of updated projects https://github.com/alpinelinux/aports-turbo/blob/master/tools/anitya-watch.lua 2017-02-02 20:07:17 that can be implemented with repology as well, I just don't see a use for it now 2017-02-02 20:07:30 AMDmi3: yes, but there’s no registry of all software in the world; there are just plain HTML packages without any standardized structe, FTPs etc. 2017-02-02 20:07:44 AMDmi3: projects moves from one site to another, changes URLs etc. there no way how to automatically detect it 2017-02-02 20:08:02 you’re arguing about something that is simply impossible 2017-02-02 20:08:27 and I know it quite well since Ijve implemented integration for Anitya in Aports 2017-02-02 20:08:33 there is and repology does that by aggregating data from a lot of repos. as soon as package is changed somewhere, everyone else know new location 2017-02-02 20:09:32 I’m not saying that there’s no better way how to handle at least common cases than what Anitya do… their solution is actually quite bad… your solution with file of rules how to match pkgs is better, but this solves just part of mapping names of projects and pkgs in various distros 2017-02-02 20:09:58 so how exactly does it do? 2017-02-02 20:10:19 sorry, I gotta go now 2017-02-02 20:10:35 some machine learning, trying to interpret human messages like “This project has been moved to new location: https://example.org”? 2017-02-02 20:10:43 let's continue tomorrow, when I'll also deploy repology aports support 2017-02-02 20:10:48 okay 2017-02-02 20:11:08 you don't need machine learning, you just need to monitor how packages are updated in the repos 2017-02-02 20:15:58 hm, I should write some notes what’s not properly solved on Anitya, for inspiration 2017-02-02 20:16:13 Anitya is quite stagnating now 2017-02-02 20:17:14 yay, there are badges!! http://repology.org/metapackage/openssl/badges 2017-02-02 20:25:24 and the code-base looks really good 2017-02-02 20:30:56 but don’t see any parallelization of fetching… it must took ages to update so many sources 2017-02-02 20:32:42 jirutka: http://repology.org/statistics 2017-02-02 20:32:55 scadu: I’ve seen that 2017-02-02 20:33:06 jirutka: but dunno, maybe there is some schedule 2017-02-02 20:33:30 didn't know that. Looks really cool. 2017-02-02 20:34:10 he did some things better than Anitya 2017-02-02 20:36:51 but it do a different thing, it’s not a first source, i.e. he knows about new version after someone else updated it in some distro repo; same for metadata 2017-02-02 20:39:09 this is much easier domain; but his approach is definitely better than Anitya’s, so I’m interested how he will deal with greater challenge 2017-02-02 20:41:58 jirutka: what's the greater challenge? sorry if I missed something. 2017-02-02 20:42:51 scadu: to monitor the source (upstream), not just intermediate (distros) 2017-02-02 20:42:53 the scope of repology is just to show what distros have what 2017-02-02 20:43:03 exactly 2017-02-02 20:43:19 jirutka: oh, right. 2017-02-02 20:43:39 someone must do the work first 2017-02-02 20:43:53 update pkg in some repo, then repology can help to all others 2017-02-02 20:43:59 but it cannot help the first guy 2017-02-02 20:44:03 (yet) 2017-02-02 20:44:43 but still it’s useful 2017-02-02 20:45:00 jirutka: noted. just came home from a workout waiting for dopamine boost ;v 2017-02-02 20:45:14 XD 2017-02-02 20:48:06 he’s right that when something like this rely on manual work, it gets outdated sooner than later; it’s impossible to automate it completely, but there’s still big space how to improve it in comparison with Anitya; they even don’t check links of project homepages; repology do check it and also it checks redirects 2017-02-03 06:29:31 ACTION wonders if there's anything interesting in void linux 2017-02-03 06:29:44 so far what i see is not really 2017-02-03 06:34:27 i guess the key thing is alternatives support 2017-02-03 06:34:40 which is probably happening in apk-tools 3 2017-02-03 06:34:51 ACTION hmms 2017-02-03 06:35:23 yeah, it's on the list 2017-02-03 06:45:54 apparently apk fetch does it's own dependency resolution 2017-02-03 06:46:50 it uses the same solver code, but applies special rules for -R. the idea is to be able to create bootable media apk repositories with it 2017-02-03 06:47:17 there is some problem with virtuals 2017-02-03 06:48:32 kaniini, https://wiki.alpinelinux.org/wiki/APKBUILD_Reference#provides 2017-02-03 06:49:12 right, i am tyring to find out more info from awilfox ;) 2017-02-03 06:49:44 i think the issue is that if there's more than 1 package which provides a capability, apk does not know how to handle it 2017-02-03 06:50:08 so if you have eudev and udev which provide 'virtual-udev' or such, it gets confused 2017-02-03 06:50:21 or maybe nano/vim/emacs would be a better example 2017-02-03 06:50:27 provides=editor 2017-02-03 06:50:39 apk add --> error 2017-02-03 06:50:40 you need to list explicitly the virtual package providers you want to get 2017-02-03 06:51:18 for fetch it might make sense to have someoption like --depend-on-all-virtual-providers or similar 2017-02-03 06:51:34 ciall adelie-new-image # apk --arch x86_64 --root squashroot-x86_64 fetch -o squashroot-x86_64/packages -R adelie-base 2017-02-03 06:51:35 adelie-base: unable to select package (or it's dependencies) 2017-02-03 06:51:37 ciall adelie-new-image # apk --arch x86_64 --root squashroot-x86_64 fetch -o squashroot-x86_64/packages -R postfix adelie-base 2017-02-03 06:51:39 postfix: unable to select package (or it's dependencies) 2017-02-03 06:51:41 adelie-base: unable to select package (or it's dependencies) 2017-02-03 06:51:43 adelie-base depends on v:mta 2017-02-03 06:51:54 putting postfix, which satisfies v:mta, on the command line 2017-02-03 06:51:56 just makes it worse 2017-02-03 06:52:11 hum 2017-02-03 06:52:36 it might be a bug, iirc, the fetch code handles the dependencies for each name individually 2017-02-03 06:52:49 it does, and it gets tripped up on v: 2017-02-03 06:53:01 may be worth fixing in 2.x 2017-02-03 06:53:03 it immediately marks it as unsolvable 2017-02-03 06:53:06 yes 2017-02-03 06:53:23 i might also add code to mount /proc when --root is given 2017-02-03 06:53:26 hmm 2017-02-03 06:53:37 i had an idea on how to specify defaults for virtuals 2017-02-03 06:53:39 since most of the scripts will not work properly without /proc due to musl realpath() requiring it 2017-02-03 06:53:48 i think install_if rules could be used somehow 2017-02-03 06:54:08 my original thought was to pick the package with highest replaces_priority 2017-02-03 06:54:17 fabled: yeah, the CD creation script I wrote already takes care of that (found that bug real quick hah), it would be nice to have apk do it itself though 2017-02-03 06:54:56 kinda wish apk index could generate signed indexes given a key file too, if you are looking for things to do :P 2017-02-03 06:55:02 yes 2017-02-03 06:55:03 well 2017-02-03 06:55:18 the plan is that index on physical media does not need to be signed 2017-02-03 06:55:28 so anyone can create boot media from official packages in simple way 2017-02-03 06:55:46 but the online indexes need to be signed, and apk will do that for you 2017-02-03 06:55:56 plan is to bundle key management, signing and package creation to apk 2017-02-03 06:56:29 leaving abuild to be strictly about building? 2017-02-03 06:56:31 :) 2017-02-03 06:56:32 https://code.foxkit.us/adelie/image/blob/master/adelie-build-cd#L255-266 <- this is just an unholy mess :( 2017-02-03 06:56:54 so you have something like 'apk mkpkg' that gets the .PKGINFO file and directory where files are + signing key, and produces signed .apk 2017-02-03 06:57:08 fabled: oh wow that would be killer 2017-02-03 06:57:09 yep 2017-02-03 06:57:21 yes, abuild would be left do the building and meta-data generation 2017-02-03 06:58:03 due to the way signing will be done; it will not support old packages, but i'll do package migration tool that can convert old .apk to new format given a signing key 2017-02-03 06:58:14 p.s. if you like to, I can change this to be 'RS2' or such for compatibility for old packages https://code.foxkit.us/adelie/packages/blob/master/sys-apps/apk-tools/files/apk-tools-2.6.6-use-sha256-signature.patch 2017-02-03 06:58:22 because sha256 really needs to be used 2017-02-03 06:58:25 sha1 is old and broken 2017-02-03 06:58:37 and dss1 is worse 2017-02-03 06:59:21 yes, will default to better ciphersuite 2017-02-03 06:59:30 probably sha256+ec 2017-02-03 06:59:38 that'd be very cool! 2017-02-03 06:59:40 probably djb curves 2017-02-03 07:00:00 i have my reservation on nist curves... 2017-02-03 07:00:09 understandable 2017-02-03 07:00:16 djb curves are better, they are unencumbered 2017-02-03 07:00:27 nist curves are patented 2017-02-03 07:01:38 past 1am, night all 2017-02-03 07:01:57 how about we use BDB for everything so people have to frequently run 2017-02-03 07:02:01 apk --rebuilddb 2017-02-03 07:02:02 i made an 'executive summary' on the list of apk things 2017-02-03 07:02:03 ;) 2017-02-03 07:02:11 i am kidding 2017-02-03 07:02:33 add some sleep() calls too 2017-02-03 07:02:39 to make the rpm users feel at home 2017-02-03 07:02:59 something like 2017-02-03 07:03:10 executive summary type of things i have on the list: http://sprunge.us/eHag 2017-02-03 07:03:17 /* timed `yum system-upgrade` on my pentium 3 */ sleep(33); 2017-02-03 07:03:57 what would be nice 2017-02-03 07:04:04 is some way to have plugins or extra commands 2017-02-03 07:04:06 like git 2017-02-03 07:05:31 for example 2017-02-03 07:05:38 fedora has 2017-02-03 07:05:44 something like 2017-02-03 07:05:50 dnf copr-add foo/bar 2017-02-03 07:05:59 and it imports that copr repo (like ubuntu PPA) 2017-02-03 07:06:04 and it's a plugin 2017-02-03 07:06:42 well, in the current code base, all the commands are 'plugins' that are statically linked. adding them could be done in simple way 2017-02-03 07:06:50 though the issue it keep stable abi 2017-02-03 07:07:19 given how the plan is to use the binary file formats, it would likely be doable in that model 2017-02-03 07:09:26 right 2017-02-03 08:41:41 fabled: i think the main other thing we need is something like PPA/Copr/whatever for alpine so people can distribute their own packages easily -- i am working on designing and implementing such a service though :) 2017-02-03 08:42:27 kaniini, yes, keep me posted on that too. it involves both the build infra + the package management handling 2017-02-03 08:45:21 my plan is to leverage docker for it 2017-02-03 08:53:57 right 2017-02-03 08:54:20 we also had idea to potentially make alternate rootbld implementation with docker, but let's see 2017-02-03 08:54:41 i think it'd be cool to be able to provide build service for ppas 2017-02-03 08:54:57 it kinda depends on chroot building 2017-02-03 08:55:18 and i really hope to get buildbot plugin that does abuild dependency handling + distribution to different arch builders 2017-02-03 08:55:34 that would scale -- the current lxc per arch+branch is problematic there 2017-02-03 08:56:20 hi, are there plans to make apk support other compression methods, like xz? 2017-02-03 08:58:28 yes 2017-02-03 11:27:14 alpine support added to repology.org 2017-02-03 11:27:23 outdated packages, for instance: http://repology.org/metapackages/outdated-in-repo/alpine_edge/ 2017-02-03 11:29:01 <_ikke_> Nice 2017-02-03 11:30:48 <_ikke_> Looks interested 2017-02-03 11:30:51 <_ikke_> interesting* 2017-02-03 11:33:41 what did you end up using AMDmi3 2017-02-03 11:47:47 <^7heo> Shiz: some very secure alternative to find :P 2017-02-03 11:48:51 <^7heo> what does "spread" mean? 2017-02-03 11:50:54 <_ikke_> ^7heo: I guess how many repositories have this package 2017-02-03 11:51:39 <^7heo> _ikke_: but how many repositories in what context? 2017-02-03 11:51:50 <^7heo> _ikke_: in alpine? in the world? in unix? in linux? in the website? 2017-02-03 11:52:43 <_ikke_> In the repositories it tracks 2017-02-03 11:53:55 it's number of repository families actually; e.g. all debian and ubuntu versions are a single family 2017-02-03 11:54:22 <_ikke_> ok, makes sense 2017-02-03 11:57:27 <^7heo> AMDmi3: if you're in charge of the UI of that website, I'd say that it could profit from a tooltip saying what it is 2017-02-03 11:58:47 AMDmi3: suggestine: Alpine Edge -> Alpine Linux Edge 2017-02-03 11:58:51 the distro is called alpine linux after all 2017-02-03 11:58:52 :p 2017-02-03 11:59:01 not just alpine 2017-02-03 12:00:19 ^7heo: that was planned 2017-02-03 12:01:10 Shiz: it's even more likely Alpine GNU/Linux, but no, I'm not adding that for any repos 2017-02-03 12:01:16 no, it's just Alpine Linux 2017-02-03 12:01:25 there's (almost?) no gnu in the base install 2017-02-03 12:05:30 <^7heo> Shiz: I dunno, Linux is an OS including GNU, now; so I'm not sure we should use that name. 2017-02-03 12:05:45 <^7heo> Shiz: it used to be a kernel, but you know, ego, lies, wikipedia... 2017-02-03 12:06:32 <_ikke_> AMDmi3: There is also a mail client called Alpine, hence the distinction 2017-02-03 12:07:06 debian actually calls itself debian on the website, alpine always calls itself alpine linux iirc 2017-02-03 12:07:08 :p 2017-02-03 12:07:37 <^7heo> but what _ikke_ says is true, alpine is actually a MUA. 2017-02-03 12:09:29 ok, let it be alpine linux 2017-02-03 12:11:27 done 2017-02-03 12:12:04 <^7heo> thanks for the correctness. 2017-02-03 12:12:25 <^7heo> AMDmi3: if I may, what "family" is alpine linux in? 2017-02-03 12:12:36 in its own 2017-02-03 12:12:39 <^7heo> ok 2017-02-03 12:12:56 if specific versions are added, they'll be in a single family 2017-02-03 12:13:14 <^7heo> and what other family is the one that has gogs too? 2017-02-03 12:16:14 <^7heo> a link for that would be awesome 2017-02-03 12:16:34 <^7heo> (like when you click on the 'spread', you get a list of all families containing the package) 2017-02-03 12:56:32 is this on the radar? mozilla firefox 53+ builds depend on rust: https://groups.google.com/forum/?_escaped_fragment_=msg/mozilla.dev.platform/Gzwh1IbxvHE/qs49_VAdAwAJ#!msg/mozilla.dev.platform/Gzwh1IbxvHE/qs49_VAdAwAJ 2017-02-03 12:56:51 lol what 2017-02-03 12:57:19 <^7heo> that was only a matter of time. 2017-02-03 12:57:37 <^7heo> fortunately jirutka has made an APKBUILD for rust. 2017-02-03 12:58:41 skrzyp: mozilla is going go rewrite firefox core to rust iirc 2017-02-03 12:58:52 <^7heo> at least it's not gonna be in go. 2017-02-03 12:59:12 kek 2017-02-03 12:59:20 but it's another project 2017-02-03 12:59:29 <^7heo> well, I'm waiting for chrome to be written in go :D 2017-02-03 13:02:08 https://www.youtube.com/watch?v=YJFvkUCXPLQ 2017-02-03 13:05:47 <^7heo> really? 2017-02-03 13:05:50 <^7heo> whyyyy v_v 2017-02-03 13:14:38 rust is not complete on alpine afaik 2017-02-03 13:15:04 atleast those were jirutka words recently 2017-02-03 13:18:21 <^7heo> let's see at the fosdem :) 2017-02-03 13:18:30 <^7heo> we have a few matters to discuss. 2017-02-03 13:19:18 foss-damn :P 2017-02-03 13:19:43 <^7heo> yeah, indeed. 2017-02-03 13:20:35 is there a plan to do some alpine-related talks/presentations on that event? 2017-02-03 13:21:35 "Alpine for Dummies: the comprehensive tutorial to make your workstation great again. In leass than 24 hours, including committing your job-related software!" 2017-02-03 13:26:17 ^7heo: do you still use gogs? 2017-02-03 13:26:26 <^7heo> yep 2017-02-03 13:26:37 <^7heo> last version I updated gave me some problems when I tested the package locally 2017-02-03 13:26:45 i just fixed gitea i believe. 2017-02-03 13:26:55 you could give it a try, its in testing. 2017-02-03 13:27:09 <^7heo> gitea? 2017-02-03 13:27:13 <^7heo> ok 2017-02-03 13:27:18 community fork 2017-02-03 13:27:24 but gogs is in go 2017-02-03 13:27:28 :) 2017-02-03 13:28:10 there is no go in gitea ;-) 2017-02-03 13:28:39 <^7heo> what is it in? 2017-02-03 13:28:56 <^7heo> WOW 2017-02-03 13:28:58 its still go 2017-02-03 13:28:59 so gitea is no-go? 2017-02-03 13:29:00 <^7heo> IT LOOKS AWESOME. 2017-02-03 13:29:16 <^7heo> https://try.gitea.io/ 2017-02-03 13:29:21 <^7heo> "404 page not found" 2017-02-03 13:29:26 <^7heo> Plaintext and all!!!! 2017-02-03 13:29:29 <_ikke_> 7haha 2017-02-03 13:29:29 nice try 2017-02-03 13:29:46 <_ikke_> do or do not, there is no try 2017-02-03 13:29:51 <^7heo> :D 2017-02-03 13:31:59 <^7heo> clandmeter: all jokes aside, tho, where can I see it? :) 2017-02-03 13:32:21 ask mosez :) 2017-02-03 13:32:31 i have it running here locally 2017-02-03 13:32:40 <^7heo> care to screenshot? :) 2017-02-03 13:33:38 looks funnty that gittea is hosted on github 2017-02-03 13:34:10 ^7heo: its the same as gogs 2017-02-03 13:34:27 but they do care about users input and performance improvments 2017-02-03 13:34:35 <^7heo> ah 2017-02-03 13:34:37 <^7heo> yeah I see. 2017-02-03 13:34:38 they didnt fork that long ago 2017-02-03 13:34:42 <^7heo> so gogs is going to die slowly. 2017-02-03 13:34:47 <^7heo> good. 2017-02-03 13:34:58 the dev will maintain it, but as his pet project he says. 2017-02-03 13:35:19 <^7heo> yeah 2017-02-03 13:35:28 he was on away for a few months an nobody had commit access... 2017-02-03 13:35:32 <^7heo> well, mosez had problems with getting user feedback in, AFAIR 2017-02-03 13:37:00 correct 2017-02-03 13:37:15 even in go, its still a nice project. 2017-02-03 13:37:40 hi, I’m already at Brussels and w/o sleep, so not enough energy to read all backlog; ad rust, read https://github.com/rust-lang/rust/issues/31322#issuecomment-277082391 2017-02-03 13:38:48 I really like to have finally proper pkg for Rust, but it requires some effort from upstream 2017-02-03 13:38:54 and things moves slowly :/ 2017-02-03 13:39:33 awesome... try down again... really need to setup monitoring :( 2017-02-03 13:42:03 ^7heo, clandmeter: thanks for the hint, try.gitea.io is up and running again. hopefully i can launch some monitoring within the next days to get autmatically notified about that -.- 2017-02-03 13:43:13 mosez: you should fix the program instead of adding a monitor ;-) 2017-02-03 13:43:43 ^ +1 2017-02-03 13:43:58 jirutka: im comming to brussels 2017-02-03 13:44:01 so buy that beer 2017-02-03 13:44:10 clandmeter: GREAT! 2017-02-03 13:44:50 :) 2017-02-03 13:44:55 i'll be riding on his tails 2017-02-03 13:44:57 coattails 2017-02-03 13:45:40 <^7heo> mosez: are you behind it? :) 2017-02-03 13:45:41 did i read cocktails? 2017-02-03 13:45:42 clandmeter: you probably have an idea how many ppl have this broken mindset… just add monitoring… and when it crashes to frequently, then just set up auto-restarting… and hey, why should not init system handle this and just restart every broken process… 2017-02-03 13:45:53 clandmeter: but the 404 error page comes from traefik... i'm just restarting traefik and it's serving fine again. 2017-02-03 13:46:03 maybe i should just replace traefik with haproxy :P 2017-02-03 13:46:19 clandmeter: please add yourself to #alpine-fosdem ;) 2017-02-03 13:46:21 both gogs and gitea seem active to me 2017-02-03 13:46:23 ^7heo: behind what? :) 2017-02-03 13:46:25 so it's hard to differentiate 2017-02-03 13:46:28 <^7heo> mosez: gitea 2017-02-03 13:46:32 i was going to switch to gogs, but if gitea is better... 2017-02-03 13:46:38 mosez: nginx? 2017-02-03 13:46:40 <^7heo> mosez: I use nginx usually. 2017-02-03 13:46:44 <^7heo> yeah what clandmeter said. 2017-02-03 13:46:48 ^7heo: yeah, i have been elected as an owner :) 2017-02-03 13:46:48 it never let me down 2017-02-03 13:46:54 <^7heo> mosez: really?! Great! :) 2017-02-03 13:46:59 except the times i misconfigure it of course. 2017-02-03 13:47:00 gitea, also written in Go… is it really better? 2017-02-03 13:47:11 i'm fine with /using/ software written in go 2017-02-03 13:47:16 at least it doesn't need a special runtime 2017-02-03 13:47:19 <^7heo> jirutka: it's better supported. 2017-02-03 13:47:22 jirutka: it's simply a gogs fork, so of course it's also written in go 2017-02-03 13:47:26 clandmeter: when will arrive? 2017-02-03 13:47:38 between 2-3 2017-02-03 13:47:40 <^7heo> I'd love a C port of gitea 2017-02-03 13:47:45 <^7heo> but yeah, not gonna happen. 2017-02-03 13:47:48 haha... or lua? :D 2017-02-03 13:48:08 turbolua git hosting :P 2017-02-03 13:48:09 <^7heo> yeah that would work. 2017-02-03 13:48:55 probably the same amount of work as in C... 2017-02-03 13:49:02 Rust! 2017-02-03 13:49:30 jirutka: did you raed the firefox link? 2017-02-03 13:49:31 or Elixir would be also cool 2017-02-03 13:49:40 clandmeter: what link? 2017-02-03 13:49:46 but anyway... gitea got 3 owners and 15 reviewers. 8 people out of them got write access to the repo. the primary goal is to being more community driven and more open for others. and to avoid the single person bottleneck that gogs got. 2017-02-03 13:49:58 jirutka: https://groups.google.com/forum/?_escaped_fragment_=msg/mozilla.dev.platform/Gzwh1IbxvHE/qs49_VAdAwAJ#!msg/mozilla.dev.platform/Gzwh1IbxvHE/qs49_VAdAwAJ 2017-02-03 13:50:00 do you merge from gogs? 2017-02-03 13:50:11 clandmeter: it requires me to log-in… 2017-02-03 13:50:29 jirutka: dont need to log in here 2017-02-03 13:50:36 http://tpaste.us/jee5 2017-02-03 13:50:41 http://txt.shiz.me/ZDkzMWQxND 2017-02-03 13:50:43 oh 2017-02-03 13:50:47 :p 2017-02-03 13:51:08 aha, firefox 2017-02-03 13:51:59 well, after FOSDEM I’ll try to talk with rust devs to FINALLY start providing precompiled rustc for musl or statically linked rustc 2017-02-03 13:52:01 shiz: we have already taken lots of pull requests from gogs. 2017-02-03 13:52:15 ah... and nearly all gitea maintainers are former gogs contributors :) 2017-02-03 13:52:24 back to work now :) 2017-02-03 13:52:36 I can try to cross-compile it, again, but it’s not reliable solution, when they don’t test it 2017-02-03 13:52:51 and unfortunately rustc is complex as hell to build :( 2017-02-03 13:53:10 mosez: one last question, can it do ldap 2017-02-03 13:53:12 :p 2017-02-03 13:53:24 it took me quite long time to get it build in this hackish way and they helped me with it a lot 2017-02-03 13:53:53 dont they already provide that? 2017-02-03 13:54:01 why these all people just can't use makefiles? 2017-02-03 13:54:05 at least they provides prebuild cargo for musl now 2017-02-03 13:54:42 i recently build some lib which was based on rust. is it always that slow to build? 2017-02-03 13:54:47 this would not help much, makefiles are awful… but yeah, creating their own build system is not a very good approach 2017-02-03 13:55:00 especially when you need rustc AND cargo to build rustc… 2017-02-03 13:55:11 cargo is written in Rust… and it have dependencies 2017-02-03 13:55:11 doesn't rust itself use makefiles 2017-02-03 13:55:16 it used to... 2017-02-03 13:55:18 so you have double chicken-egg problem 2017-02-03 13:55:20 insane 2017-02-03 13:55:57 used to, but it was even harder to get work 2017-02-03 13:56:02 and it’s already deprecated 2017-02-03 13:56:18 but the probelm is not in build system 2017-02-03 13:56:37 it’s quite complicated 2017-02-03 13:56:48 don’t have capacity to explain now :) 2017-02-03 13:57:36 it’s sad that project like this doesn’t have any dev with experience with packiging for some linux distribution 2017-02-03 13:58:10 so they created something, then asked distributions what they need and after that started thingking how to improve the situation 2017-02-03 13:58:43 it may sound horrible as I talk about it, but it’s still inifinitely better situation than in Go… 2017-02-03 13:59:01 but, I have to admit that I was a little disapointed 2017-02-03 14:02:15 there are many great ideas in Rust, it’s really the most innovative language since long time, but it scares me how big complexity they’re adding into it… I’d love to see some new minimalistic lang based on great ideas from Rust 2017-02-03 14:04:10 configure: It is recommended you avoid passing --disable-rustbuild to get your 2017-02-03 14:04:12 configure: build working as the makefiles will be deleted on 2017-02-02. 2017-02-03 14:04:21 # date 2017-02-03 14:04:22 Fri Feb 3 14:04:04 GMT 2017 2017-02-03 14:04:25 hmm_emoji.gif 2017-02-03 14:04:30 also LLVM is part of the problem… but I think that with recent changes like introduction of MIR they’re slowly preparing ground to be able to change LLVM in future 2017-02-03 14:10:51 shiz: yes, ldap is supported 2017-02-03 14:50:25 Brussels is a modern city, they use containers! https://twitter.com/jakubjirutka/status/827527150651768832 2017-02-03 14:55:05 <^7heo> niiice :D 2017-02-03 14:55:11 <^7heo> I wonder if I should take my camera with me at the fosdem. 2017-02-03 14:57:54 <_ikke_> Shiz: lol 2017-02-03 14:59:08 <^7heo> "the makefiles will be deleted on 2017-02-02" 2017-02-03 15:00:13 <^7heo> is that another of trump's bills? 2017-02-03 15:38:26 <^7heo> s/bill/executive order/ 2017-02-03 15:38:29 <^7heo> (I know I lag) 2017-02-04 20:38:25 entered wrong channel, wondering why #alpine-dev was empty for hours 2017-02-04 20:43:34 <_ikke_> haha 2017-02-04 21:49:57 so @clandmeter, i still want beer, when to convert the wiki? ;) 2017-02-05 11:09:23 Is it me of lame is unable to set tags ? 2017-02-05 11:18:50 <^7heo> are you from the kingdom of lame? 2017-02-05 11:33:03 ^7heo: if you talk about France, you're right 2017-02-05 11:33:05 :D 2017-02-05 11:35:39 <^7heo> coredumb: nah it's just because the way you phrased it. 2017-02-05 11:41:00 :) 2017-02-05 12:43:43 ^7heo: :D 2017-02-06 01:53:58 just a ping on the ghc port https://github.com/alpinelinux/aports/pull/205 would it be better if i resubmit this to the mailing list? 2017-02-06 05:01:03 mitchty: i'll look at it in the morning 2017-02-06 05:01:22 mitchty: i want to reproduce the build before committing it 2017-02-06 11:53:31 <^7heo> that's a flash visit. 2017-02-06 12:20:53 Hi, I was compiling gettext, libintl.so was not compiled because /usr/include/libintl.h exists from musl. Remove that header, libintl.so is compiled. Found this post : http://www.openwall.com/lists/musl/2014/10/01/3. Is there anyway to force building gettext's libintl ? Cant find any option in configure. Thanks. 2017-02-06 12:21:44 strage thing is that, I did compile gettext few days ago. Maybe now vs then, I have installed different packages that may affect the build. 2017-02-06 12:23:16 tmh1999, the cross-build might add the libintl.h accidentally 2017-02-06 12:23:43 yes, install_sysroot_headers just calls make install-headers 2017-02-06 12:23:49 hum 2017-02-06 12:24:21 i should probably fix musl headers-only build to create a real .apk 2017-02-06 12:27:00 so, which package do you think accidentally add libint.h ? 2017-02-06 12:27:02 linux-headers ? 2017-02-06 12:27:16 I did remove it and libintl.h still there 2017-02-06 12:27:26 hope I can safely remove libintl.h to build gettext 2017-02-06 12:27:42 strange thing I dont have this few days ago lol 2017-02-06 12:29:56 I should create an ordered list of package when natively build (like in bootstrap.sh) to avoid problem like these 2017-02-06 12:40:29 tmh1999, i think it comes from bootstrap.sh when it calls musl install_sysroot_headers 2017-02-06 12:43:06 ah yeah right. I meant musl, not linux-headers. 2017-02-06 12:43:29 Thanks 2017-02-06 12:44:18 should fix musl apkbuild to delete it also in that case 2017-02-06 12:45:01 or even better, make the bootstrap.sh + musl apkbuild work together to create headers-only .apk during that early stage 2017-02-06 12:48:42 I will try it 2017-02-06 13:06:43 chromium crashes :-( 2017-02-06 13:10:13 <^7heo> and firefox requires rust. 2017-02-06 13:10:21 <^7heo> ncopa: moin, safe trip home? 2017-02-06 13:13:09 ^7heo: hi! yes! and you? 2017-02-06 13:13:19 was really nice to meet you :) 2017-02-06 13:13:39 grsecurity patch for 4.9 is out http://www.prweb.com/releases/2017/02/prweb14044396.htm 2017-02-06 13:14:03 <^7heo> ncopa: EZjet really rushed the trip 2017-02-06 13:14:09 <^7heo> ncopa: my ears still hurt... 2017-02-06 13:14:14 <^7heo> ncopa: 2h late AND painful ears. 2017-02-06 13:14:33 <^7heo> ncopa: "easy"jet definitely doesn't mean "easy on you". 2017-02-06 13:14:43 <^7heo> ncopa: yeah I liked meeting everyone. 2017-02-06 13:16:20 flight was on time for us. but we were home late 2017-02-06 13:16:29 <^7heo> also easyjet? 2017-02-06 13:16:37 brussels airlines 2017-02-06 13:16:41 <^7heo> ah that's why. 2017-02-06 13:16:42 :) 2017-02-06 13:16:43 they were ok 2017-02-06 13:16:47 i gotta update my kernel 2017-02-06 13:16:50 <^7heo> easyjet was freaking cheap tho. 2017-02-06 13:16:57 <^7heo> 19€ on the way from Berlin to there 2017-02-06 13:17:06 <^7heo> 40€ on the way back 2017-02-06 13:17:21 thats not bad 2017-02-06 13:17:25 <^7heo> But the "fun" part is that the 19€ one was like a breeze 2017-02-06 13:17:28 <^7heo> pleasant and all 2017-02-06 13:17:35 <^7heo> the way back, however, was hellish. 2017-02-06 13:17:56 :) 2017-02-06 13:18:03 sorry bout that 2017-02-06 13:18:27 ^7heo: i looked at decrypted header on the flight back 2017-02-06 13:18:35 and started working on it 2017-02-06 13:18:50 should not be very difficult to implement 2017-02-06 13:18:53 <^7heo> you mean deported header? 2017-02-06 13:18:56 yes 2017-02-06 13:18:59 <^7heo> well, it was already implemented 2017-02-06 13:19:00 <^7heo> and working 2017-02-06 13:19:03 i looked at your original patches 2017-02-06 13:19:06 yes 2017-02-06 13:19:12 <^7heo> but now I have to rebase all the things 2017-02-06 13:19:15 <^7heo> and I have no time 2017-02-06 13:19:19 thats what i was looking at 2017-02-06 13:19:24 <^7heo> ah ok 2017-02-06 13:19:37 <^7heo> I mean, the difficulty I hit was that I didn't follow any of the changes from fabled 2017-02-06 13:19:44 i wonder if we should do it the other way around 2017-02-06 13:19:48 <^7heo> so in the end, it was like "guessing what was happening when" 2017-02-06 13:20:01 instead of adding -H 2017-02-06 13:20:12 we add -D 2017-02-06 13:20:18 <^7heo> and I ended up patching all the things in ONE diff (the "refactor the structs" one) 2017-02-06 13:20:31 <^7heo> and that wasn't really a good move, retrospectively. 2017-02-06 13:20:44 <^7heo> ncopa: why? 2017-02-06 13:20:56 <^7heo> I'd rather have the device with header put as an argument. 2017-02-06 13:21:05 <^7heo> even tho cryptsetup does it differenyly. 2017-02-06 13:21:06 because thats how cryptsetup works internally 2017-02-06 13:21:11 <^7heo> s/yly/tly/ 2017-02-06 13:21:14 <^7heo> yes 2017-02-06 13:21:26 so code gets a bit simpler 2017-02-06 13:21:36 <^7heo> but it's IMHO more important to give the user a consistent behavior 2017-02-06 13:21:45 yes 2017-02-06 13:21:48 <^7heo> rather than mirroring the internal functionality 2017-02-06 13:21:59 how does other distros? 2017-02-06 13:22:01 <^7heo> and without the -H option it's on device /dev/X 2017-02-06 13:22:14 <^7heo> with the -H option, it's still on device /dev/X, but header is deported. 2017-02-06 13:22:23 <^7heo> (hence the way I did it) 2017-02-06 13:22:34 <^7heo> (even if in reality, it's the data payload that is deported) 2017-02-06 13:22:46 <^7heo> I don't really know how other distros do it. 2017-02-06 13:22:52 <^7heo> I don't use other distros. 2017-02-06 13:23:03 might be worth look 2017-02-06 13:23:15 <^7heo> possibly. 2017-02-06 13:23:30 <^7heo> but only if they go my way ;) 2017-02-06 13:23:34 <^7heo> ACTION hides 2017-02-06 13:32:26 I wasn't able to catch the lot of you on Sunday... 2017-02-06 13:32:47 I saw jirutka on Sunday but nobody else 2017-02-06 13:32:53 <^7heo> pavlix: afaik nobody was. 2017-02-06 13:33:01 ncopa ran away faster than you can say "wife" 2017-02-06 13:33:22 <^7heo> yeah but I found him hiding at the airport :D 2017-02-06 13:33:39 <^7heo> (sorry about that, I didn't do it on purpose, it's easyjet's fault) 2017-02-06 13:33:59 the talks I wanted to attend kept me at the opposite end of the campus from ^7heo 2017-02-06 13:34:06 (still debating whether it was a good or a bad thing) 2017-02-06 13:34:40 :D 2017-02-06 13:34:49 <^7heo> (I stayed in K, talking a lot; so presumably a good thing ;)) 2017-02-06 13:35:05 hi skarnet, was really nice to meet you in person 2017-02-06 13:35:14 the pleasure was all mine! 2017-02-06 13:35:27 i only met jirutka and ^7heo on sunday 2017-02-06 13:35:36 i bumbed into jirutka everywhere 2017-02-06 13:35:39 on campus 2017-02-06 13:35:39 <^7heo> :D 2017-02-06 13:35:43 and on airport 2017-02-06 13:35:46 :) 2017-02-06 13:35:50 <^7heo> he probably put a tracker on you. 2017-02-06 13:35:53 <^7heo> I'd get scanned. 2017-02-06 13:35:55 <^7heo> if I were you. 2017-02-06 13:35:58 <^7heo> ACTION hides 2017-02-06 13:36:25 nah, it was only nice 2017-02-06 13:36:40 the picture I took is available at http://skarnet.org/pics/alpine-team.jpg but I generally prefer ncopa's 2017-02-06 13:36:52 (and if anyone on it wants it removed, it will be) 2017-02-06 13:37:16 <^7heo> damn, we said not to put it on a logged chanel =[ 2017-02-06 13:37:28 did we? well I'll just change the URL 2017-02-06 13:37:35 <^7heo> I think we did 2017-02-06 13:37:42 <^7heo> the other pic is on the fosdem chan 2017-02-06 13:38:01 tbh that's why I hosted it on my server and nowhere else - to be able to take it down or move it if necessary 2017-02-06 13:38:12 <^7heo> ;) 2017-02-06 13:38:36 <^7heo> that's a cool photo tho 2017-02-06 13:39:10 that thumb up photo is dangerous: http://www.techradar.com/news/doing-the-peace-sign-in-pictures-could-put-your-fingerprint-at-risk-researchers-find 2017-02-06 13:39:40 apparently they are able to get the fingerprint from photos nowdays 2017-02-06 13:40:03 but you probably need better resolution than that photo :) 2017-02-06 13:40:08 at the resolution my photo is at, yeah, gl to them 2017-02-06 13:40:29 :) 2017-02-06 13:41:17 <^7heo> yeah exactly, I'm not really afraid of that resolution. 2017-02-06 13:41:56 <^7heo> the one you took, tho, ncopa, has a better resolution 2017-02-06 13:42:06 <^7heo> fortunately, it's a little blurry 2017-02-06 13:42:39 <^7heo> and Shiz's arm looks like it's jirutka's. So it is pretty confusing for a machine I guess ;) 2017-02-06 13:43:21 I'm pretty sure Shiz was just trying to prevent clandmeter from being on the picture 2017-02-06 13:43:30 he succeeded on mine and failed on ncopa's 2017-02-06 13:44:12 <^7heo> :D 2017-02-06 13:44:18 i think clandmeter tried to hide 2017-02-06 13:44:37 "I'll drive you to fosdem, but you must protect me from cameras" 2017-02-06 13:44:50 <^7heo> :D 2017-02-06 13:44:57 <^7heo> sounds like a plausible deal. 2017-02-06 14:03:29 skarnet: 404 Not Found 2017-02-06 14:03:52 pavlix: changed it on purpose 2017-02-06 14:04:50 you have to sign nda to view it. 2017-02-06 14:05:05 XD 2017-02-06 14:05:28 I'm in the pictures! 2017-02-06 14:05:37 I'm an Alpine developer! 2017-02-06 14:05:38 :) 2017-02-06 14:05:54 Now I don't have to develop anything to become one any more. :D 2017-02-06 14:05:58 "contributor" 2017-02-06 14:06:08 that includes users who give feedback and are annoying af 2017-02-06 14:06:12 like me 2017-02-06 14:06:13 yeah thats all that it requires to become an Alpine developer. make sure you are included in the right group picture! 2017-02-06 14:08:48 pavlix: take it as a commitment, now you have to contribute :P ;) 2017-02-06 14:09:43 <^7heo> ncopa: :D 2017-02-06 14:11:37 ncopa: hmm, do we have photos from some previous events? 2017-02-06 14:13:43 nope, thats the first i think 2017-02-06 14:14:28 <^7heo> what previous events happened? 2017-02-06 14:14:32 <^7heo> (in the last 10 years) 2017-02-06 14:17:59 ^7heo: 0 2017-02-06 14:18:33 <^7heo> wait, so that was the FIRST ever event for alpine as a distro? 2017-02-06 14:18:50 I was at fosdem 2014 but nobody had the courage to meet me. 2017-02-06 14:19:25 <^7heo> I never came to FOSDEM before 2017-02-06 14:21:55 clandmeter: it would probably be more accurate to say that you were too discreet and nobody noticed you. 2017-02-06 14:26:44 I think that I’ll print that photo and put it on my table :) 2017-02-06 14:26:46 skarnet: yes i did gain a few kilo's since... :) 2017-02-06 14:27:27 <^7heo> clandmeter: so you can't properly hide behind Shiz anymore? :D 2017-02-06 14:27:53 actually i had to watch out not to bump him out of the pic :p 2017-02-06 14:29:20 <^7heo> ahaha 2017-02-06 14:30:45 leo-unglaub: wb! we missed you. 2017-02-06 14:33:20 Hi! 2017-02-06 14:33:27 What’s the purpose of the generic arm image? 2017-02-06 14:34:02 Can it be used as a rootfs or to install alpine on arm boards? 2017-02-06 14:35:12 <^7heo> for stuff like the usb armory I'd say 2017-02-06 14:38:57 Youbi, it's the u-boot bootable image 2017-02-06 14:42:35 jirutka: I heard you are trying to convert my fellow Gentoo users and sysadmins to Alpine, why? 2017-02-06 14:43:55 clandmeter: I was at many FOSDEMs but usually nobody cared. :) 2017-02-06 14:46:02 pavlix: wat? I’m not converting anyone else now 2017-02-06 14:47:05 <^7heo> s/else/*&*/ 2017-02-06 14:49:20 pavlix: and also I would not say “convert”, I still like Gentoo, just not using it for new installations anymore 2017-02-06 14:56:29 <^7heo> I used to say (and I know we're on record here), that gentoo is like a very young girlfriend. 2017-02-06 14:56:56 <^7heo> It's great to spend time with her, and you enjoy very much all the interactions 2017-02-06 14:57:25 <^7heo> but in exchange, you have to take care of her on all your free time, can't see your friends anymore, etc. 2017-02-06 14:57:45 <^7heo> TL;DR: Gentoo is very attractive but a time sink. 2017-02-06 14:58:24 yeah, unfortunately this is true (based on 5y experience with Gentoo) 2017-02-06 14:58:36 "I'm sorry bro, can't go out have a beer with you tonight, I'm spending the evening with Gentoo" 2017-02-06 14:58:46 "but it's been four weeks" 2017-02-06 14:58:50 "well... yeah" 2017-02-06 14:59:18 jirutka: :D 2017-02-06 14:59:53 and also that young girl is very slow, you have to wait dozens of seconds until she say yes, or no; and not even speaking about waiting several hours until she’s ready to go out 2017-02-06 15:00:07 ^7heo: I had a young girlfriend and she took care of herself all time time she had to. :D 2017-02-06 15:00:55 pavlix: I hope you kept her for a long time, because those ones are precious. :) 2017-02-06 15:00:56 pavlix: exceptions exists everywhere… and I guess that’s why you eventually married her ;) 2017-02-06 15:00:57 <^7heo> pavlix: time time? 2017-02-06 15:01:06 skarnet: she's my wife now 2017-02-06 15:01:11 <^7heo> you mean "every time she had to"? 2017-02-06 15:01:14 good job ;) 2017-02-06 15:01:17 <^7heo> yeah, gg. 2017-02-06 15:01:19 skarnet: but it didn't work the same 2017-02-06 15:01:24 skarnet: :D 2017-02-06 15:01:33 <^7heo> also, I met a really cool girl, and she's much more of a relief than a weight 2017-02-06 15:01:39 <^7heo> (still with her) 2017-02-06 15:01:48 <^7heo> but obviously those aren't the majority. 2017-02-06 15:01:53 <^7heo> also, -> offtopic? 2017-02-06 15:02:01 i have many gf, they they are all in #alpine-offtopic ;-) 2017-02-06 15:02:18 ACTION crying in corner, currently don’t know such girl :/ 2017-02-06 15:02:28 clandmeter: +1 2017-02-06 15:02:37 ^7heo: and you sometimes have to help it a bit 2017-02-06 15:02:55 so now no one is going to #alpine-offtopic, b/c don’t want to become clandmeter’s girlfriend :P 2017-02-06 15:03:12 jirutka: The most important thing is not to cross the line with girls that don't fit your workflow. :) 2017-02-06 15:03:19 but clandmeter is a real gentleman 2017-02-06 15:03:47 ^7heo: Or at least I do. My wife is great but sometimes needs some help with that. :) 2017-02-06 15:03:48 well, but still, he doesn’t have boobs and long hair, this is serious dealbreaker :P 2017-02-06 15:04:13 tsss. So shallow. 2017-02-06 15:20:35 kaniini: thats fine, you'll have to upload the ghc-8.0.1-{armhf,x86_64}.txz files up and then do an abuild checksum, but past that everything should work (famous last words), with one possible exception of stack on armhf hitting a compiler bug, but if you've enough ram that too should work 2017-02-06 15:20:51 for ghc-bootstrap specifically that is 2017-02-06 15:21:09 i know it looks like a nuts pr with 1k loc additions but its mostly patches/ancillary files 2017-02-06 16:00:17 I heard alpine isn't capable of building packages without side effects, i.e. that you need to create your own chroot/container environment in order to get it built. 2017-02-06 16:02:32 pavlix: abuild will install necessary packages and uninstall them, but from what i know it does *not* chroot package building 2017-02-06 16:02:44 <^7heo> pavlix: that is false. 2017-02-06 16:02:47 pavlix: i think fabled is currently working on abuild with chroot support 2017-02-06 16:02:57 abuild with chroot support? that's great news! 2017-02-06 16:03:23 but if you are on alpine you dont need chroot of course, but you do need to use the same branch. 2017-02-06 16:03:28 asie: installing and uninstalling packages IMO counts as a side effect 2017-02-06 16:04:11 pavlix: explain what you mean exactly as side effects. 2017-02-06 16:04:21 btw build systems like those for Fedora and openSUSE actually try to make build environment sort of reproducible 2017-02-06 16:06:05 clandmeter: modifications of the host system where I run the command to build the package 2017-02-06 16:06:46 clandmeter: apk does not revert effects of {pre|post}-install scripts, so when you install and then uninstall e.g. nginx, nginx user is kept in passwd (that’s correct when installing pkgs, but it counts as a side effect in the context of this discussion) 2017-02-06 16:07:26 jirutka: thx 2017-02-06 16:07:57 Debian's got this too https://wiki.debian.org/ReproducibleBuilds 2017-02-06 16:08:20 asie: reproducible builds is a step further over reproducible environment 2017-02-06 16:08:38 this is going to be solved by chroot feature in abuild that is currently WIP; but as I said before, I personally don’t consider the current state as a problem, creating chroot by default adds additional complexity that may complicate troubleshooting of packages 2017-02-06 16:08:40 asie: or I don't even know if reproducible is the right word here 2017-02-06 16:09:18 jirutka: creating chroot by default doesn't complicate it for you if you can still turn it off 2017-02-06 16:09:33 the current state may be a problem for users who don’t know how it works and that they should create ephemeral chroot/container for building pkgs 2017-02-06 16:09:38 pavlix: that’s true 2017-02-06 16:10:39 so I’m not against adding this feature into abuild, just don’t see it as some necessary feature that must be here 2017-02-06 16:11:36 I do or don't depending on the specification of the tool. If it's an actual tool to build a package for myself, then I'd consider it a mandatory feature. If it's just a part of the build pipeline, I don't. 2017-02-06 16:11:52 Anyway, how exactly are *officia* binary packages built? 2017-02-06 16:12:48 pavlix: they are built on our building infra that runs on LXC containers 2017-02-06 16:13:09 pavlix: you can see the current state of build servers at http://build.alpinelinux.org/ and follow messages on #alpine-commits 2017-02-06 16:13:28 jirutka: back to networkmanager 2017-02-06 16:13:36 jirutka: what's the easiest way to try building it 2017-02-06 16:13:42 pavlix: it’s very basic, not anything like Open Build Service 2017-02-06 16:14:05 jirutka: if that works? :) 2017-02-06 16:14:07 pavlix: I’ve already sent you link to wiki page how to setup dev environment; then just enter pkg’s directory and run abuild -r 2017-02-06 16:14:16 jirutka: thx 2017-02-06 16:15:07 jirutka: On the other hand it looks like it's too complicated to just do it now... 2017-02-06 16:15:20 no, it’s super easy 2017-02-06 16:16:08 jirutka: will have to download the repo the second time... 2017-02-06 16:16:15 hm, once I wrote some script to easily create LXC container and prepare it for building pkgs, but can’t find it now 2017-02-06 16:16:26 jirutka: and it will be gone once I kill the docker image 2017-02-06 16:16:56 pavlix: I haven’t told you to use docker… ;) 2017-02-06 16:17:01 jirutka: That would be great, especially if it runs outside Alpine. 2017-02-06 16:17:17 pavlix: if you do, then you should clone repo to some directory outside of the container and mount it 2017-02-06 16:17:19 jirutka: Was quicker than setting up chroot manually. :D 2017-02-06 16:17:29 jirutka: I'm not such a docker guru, yet. :D 2017-02-06 16:17:36 i think snappy 1.1.4 is brokenb 2017-02-06 16:17:42 pavlix: I really don’t think so…https://github.com/jirutka/alpine-chroot-install/ 2017-02-06 16:17:48 jirutka: But I definitely should... I'd need to create a couple of scripts, too... 2017-02-06 16:18:33 I'd be really happy if I could use abuild on Gentoo directly :) 2017-02-06 16:18:49 pavlix: to be fair, you should also consider time and knowledge needed to install and use docker; compare it with fetching https://github.com/jirutka/alpine-chroot-install/blob/master/alpine-chroot-install and running it, nothing special is needed 2017-02-06 16:20:28 jirutka: so what's the default behavior? 2017-02-06 16:21:27 pavlix: default behaviour of what? 2017-02-06 16:21:28 jirutka: Trying to run abuild in the docker thing and it requires a non-root account, why? 2017-02-06 16:21:47 jirutka: Default behavior of alpine-chroot-install that I'm supposed to *just run*. 2017-02-06 16:22:11 pavlix: yes, abuild requires non-root account, that’s a good thing 2017-02-06 16:22:25 pavlix: can you read? ;) https://github.com/jirutka/alpine-chroot-install/blob/master/alpine-chroot-install#L23 2017-02-06 16:22:26 jirutka: I don't see why that's a good thing. 2017-02-06 16:22:55 pavlix: but for building pkgs you should install edge, so add `-b edge` 2017-02-06 16:23:20 jirutka: You said *just run it* as opposed to *learn docker*. That is unfair because you just need to learn one or two docker command lines which is not much different from the long command line from your example... 2017-02-06 16:23:35 no 2017-02-06 16:23:38 What's edge? 2017-02-06 16:23:43 edge ~ unstable 2017-02-06 16:23:56 I told you it's potentially too complex for me to do now... :D 2017-02-06 16:23:58 you must run docker-engine as a daemon 2017-02-06 16:23:59 etc. 2017-02-06 16:24:13 and compare complexity of docker and this simple shell script 2017-02-06 16:24:30 And your claim that it's super easy doesn't prove right... especially if abuild tries to be too clever and enforces non-root user and stuff like that... I wonder how many other glitches are there. 2017-02-06 16:24:43 too clever XD 2017-02-06 16:25:07 you want abuild to create isolated environment for you and you consider requring non-root user as “too clever”? lol 2017-02-06 16:25:23 Yep, it prevents me from just running it. 2017-02-06 16:25:36 it prevents you from messing your system by accident 2017-02-06 16:25:41 That's too clever to me, as it steps in my way. Creating a chroot would not be stepping in my way. 2017-02-06 16:25:54 jirutka: No, it's not... the system is ephemeral. 2017-02-06 16:26:06 how can abuild know it? 2017-02-06 16:26:18 it’s not *required*, it’s just recommended… to run it in ephemeral system 2017-02-06 16:26:31 jirutka: That's called *too clever*. It works with some assumptions and steps in my way. 2017-02-06 16:26:36 sry, don’t have time for this right now, need to finish something 2017-02-06 16:27:19 btw have you read the manual? I sent you link to wiki page that explains what you should do… creating non-root account and setting persmissions is one of the steps 2017-02-06 16:27:24 I'm afraid I will need to patch out stuff that steps in my way or otherwise I end up feeling like in Windows... 2017-02-06 16:27:36 omg… 2017-02-06 16:27:57 I don’t remember all reasons why abuild requires non-root account, you must ask fabled 2017-02-06 16:28:28 but your approach is quite wrong; instead of learning how it works you just demands to work it in a way that you’re used to from another system 2017-02-06 16:28:40 there are reasons why it works as it works 2017-02-06 16:29:02 jirutka: There are reasons for just anything out there... that's not an answer for me. 2017-02-06 16:29:24 as I said, need to go now; maybe it’ll be easier if we met this week, I can show you how to do it and write notes for myself to improve documentation 2017-02-06 16:29:44 jirutka: You cannot simply assume my approach is wrong, it can just as well be that the other side is wrong... and it very often is. 2017-02-06 16:30:22 that’s possible, but you’re basically saying that it’s wrong, b/c it does not work as you expects from experience with other systems 2017-02-06 16:30:50 jirutka: I will certainly not be using Alpine just to adapt myself to some weird set of assumptions, I can just as well stay with Gentoo where I can get away with all assumptions and just get things done. :) 2017-02-06 16:31:05 jirutka: Nope. I'm not saying that. That's the difference between us. 2017-02-06 16:31:29 why do you consider running it as non-root user as weird assumption? Gentoo assumes that you run ebuild as root user, isn’t that weird assumption? 2017-02-06 16:32:06 jirutka: Gentoo doesn't assume that, ebuild simply works with stuff that are not available to the user, that's it... 2017-02-06 16:32:28 jirutka: There's a difference between a concusion and an arbitrary limitation. 2017-02-06 16:32:44 fwiw, there are perfectly good technical reasons to do package buildings in a chroot 2017-02-06 16:32:45 omfg 2017-02-06 16:32:52 skarnet: +1 2017-02-06 16:33:07 this is not arbitrary limitation 2017-02-06 16:33:18 if you don't, the host system's environment leaks into the build environment 2017-02-06 16:33:22 it’s just your ignorance… 2017-02-06 16:33:27 skarnet: +1 2017-02-06 16:33:46 currently abuild modifies the host system to install dependencies for the package being built 2017-02-06 16:33:50 jirutka: -> PM 2017-02-06 16:34:07 jirutka: I'm not going to continue discussion in that direction. 2017-02-06 16:34:15 abuild uses sudo to run only specific actions as root user, not everything 2017-02-06 16:34:27 it requires the host system to be edge in order to package stuff properly 2017-02-06 16:34:30 jirutka: Oh does it? 2017-02-06 16:34:32 it’s generally very bad practice to do everything as root 2017-02-06 16:34:44 skarnet: that’s not true 2017-02-06 16:35:08 jirutka: Please don't come up with “general rules of thumb” in a tech discussion. 2017-02-06 16:35:11 oh? ISTR you yourself told me otherwise 2017-02-06 16:35:22 skarnet: we require to build pkgs on edge just when you want to send abuild to the default aports tree branch, that is edge 2017-02-06 16:35:30 well yes 2017-02-06 16:35:38 skarnet: if you want to build package for a stable branch, then you of course run abuild in the stable branch 2017-02-06 16:35:42 that's what I meant, sorry if it wasn't clear 2017-02-06 16:36:01 you must only know what you’re doing, that’s all 2017-02-06 16:36:11 Alpine is for power users, not for “ubuntu users”… 2017-02-06 16:36:26 still it requires the host system to be on the same branch as what you're building for, and that is not good when you're building for edge 2017-02-06 16:36:26 and it’s nothing complicated, you just should run build in environment that you targets 2017-02-06 16:36:45 fwiw our handling of package building is basically the same as debian 2017-02-06 16:36:46 it's not complicated, but it's a constraint I don't like at all 2017-02-06 16:36:46 Can I disable arch checking in abuild command line? 2017-02-06 16:36:46 yes, then you should create another chroot/container 2017-02-06 16:37:10 you can build in a chroot (and there are helpers to do that), or not 2017-02-06 16:37:12 pavlix: why the hell you want to disable arch checking? 2017-02-06 16:37:22 jirutka: I want to attempt to build a package. 2017-02-06 16:37:31 pavlix: and you’re on different arch? 2017-02-06 16:37:40 pavlix: if you want gentoo with apk, #adelie is likely a better fit 2017-02-06 16:37:44 jirutka: And I don't think it's practical to edit it just to do that. 2017-02-06 16:37:49 pavlix: that is literally what those guys do 2017-02-06 16:38:06 pavlix: aha, b/c there’s arch="" 2017-02-06 16:38:22 jirutka: +1 2017-02-06 16:38:40 jirutka: Still looking at networkmanager apk as an example of something to play with. 2017-02-06 16:38:47 pavlix: if you want to fix that apkbuild, you must edit it anyway… so just replace with arch=all" and stop looking for stupid workarounds 2017-02-06 16:38:51 kaniini: Why? 2017-02-06 16:39:03 kaniini: b/c he’s author of networkmanager… 2017-02-06 16:39:04 pavlix: because that's literally what adelie is 2017-02-06 16:39:33 jirutka: You are asking me to stop looking for stupid workarounds while asking me to use a stupid workaround? 2017-02-06 16:39:58 jirutka: Or you consider arch=all anything else than a stupid workaround for missing a tool that will just try to build it? 2017-02-06 16:40:13 i kind of agree 2017-02-06 16:40:23 it would be nice to have an override 2017-02-06 16:40:27 pavlix: do you know what arch= actually is?! 2017-02-06 16:41:20 pavlix: you must set it to something to make it build on build infra anyway 2017-02-06 16:41:20 i am pretty sure he has figured out it's a list of archs to try building on, or 'all', or 'noarch' (which is like 'all', except that you also say there is no architecture-specific binaries) 2017-02-06 16:41:33 jirutka: I'm not building on build infra, so that's irrelevant. 2017-02-06 16:41:35 what he is saying is, he would like to try anyway 2017-02-06 16:41:50 it's a legitimate feature request and could have been implemented already in the time spent arguing 2017-02-06 16:42:16 jirutka: “i am pretty sure he has figured out it's a list of archs to try building on” – I didn't even think of any other expanation... so why are you asking at all? 2017-02-06 16:42:22 to add yet another option to CLI? 2017-02-06 16:42:40 so we will eventually end up with 100+ options just because some users don’t want to touch text editor or what?! 2017-02-06 16:42:54 this is debian approach… instead of directly editing one file, provide a CLI tool to do that… 2017-02-06 16:43:20 it’s total nonsense 2017-02-06 16:43:22 systemd :( 2017-02-06 16:43:26 <^7heo> shhht 2017-02-06 16:43:31 <^7heo> stop swearing please. 2017-02-06 16:43:42 Sorry 2017-02-06 16:44:12 oh no, he said the s*d word! 2017-02-06 16:44:16 :) 2017-02-06 16:44:33 <^7heo> yeah... 2017-02-06 16:44:35 perhaps the correct answer is to assume that the APKBUILD is correct in all cases and add no parameters to ignore part of the APKBUILD 2017-02-06 16:44:47 <^7heo> gosh I hate software that has toomuch abastraction and shallow documentation 2017-02-06 16:44:48 because the goal is to create a correct APKBUILD, and such flags could be forgotten about quickly 2017-02-06 16:44:49 pavlix: for what it's worth, if you inoke abuild steps manually, 2017-02-06 16:44:57 pavlix: it should ignore $arch 2017-02-06 16:45:10 kaniini: thx 2017-02-06 16:45:17 pavlix: which if you are an upstream trying to figure out why there is a problem, this is probably more useful to you anyway 2017-02-06 16:45:34 pavlix: abuild clean unpack prepare build 2017-02-06 16:45:41 pavlix: should probably get you what you need 2017-02-06 16:49:11 kaniini: +1 2017-02-06 16:49:21 kaniini: That would be a relatively good description of my situation. 2017-02-06 16:50:07 And apparently we don't need any new CLI options as we already have all of them right in their place. :) 2017-02-06 16:51:59 kaniini: Is there a CLI to just install dependencies then? 2017-02-06 16:52:22 yes, actually 2017-02-06 16:52:48 abuild deps 2017-02-06 16:52:58 yes 2017-02-06 16:53:00 abuild --help 2017-02-06 16:53:06 everything is here… 2017-02-06 17:09:48 skarnet: well, I understand why you don’t like this constraint, but the only solution is to run builds in chroot; do you really think that it’s responsibility of a build tool to set up chroot? and what about arch, shouldn’t it also integrate qemu-user for emulation of another arch? 2017-02-06 17:11:06 depends on what a "build tool" is, but ideally, a system that creates a binary package for a specific distribution and a specific architecture, should do all that, yes 2017-02-06 17:11:09 skarnet: isn’t the better approach to have a separate tool for this? and then maybe splitting abuild to more scripts, one of them high-level script that would be as a glue for underlaying low-level scripts? 2017-02-06 17:11:21 absolutely 2017-02-06 17:11:35 and kaniini started that approach years ago with build-lab already 2017-02-06 17:11:43 I know 2017-02-06 17:12:13 IMHO abuild currently do a way more than it should and yet we’re currently adding yet another feature directly into it :/ 2017-02-06 17:12:24 but then we need to agree on terminology, and if we say that "abuild" is only the simple builder itself, then we need additional environment (buildlab?) around abuild 2017-02-06 17:12:33 yes 2017-02-06 17:13:05 the reason why debian has pbuilder and alpine has buildlab is so that the env matches always what is in the build file 2017-02-06 17:13:25 otherwise builders can have inconsistent output 2017-02-06 18:57:11 Btw `abuild -F deps` fails for me due to missing group membership... 2017-02-06 18:57:59 skarnet: +1 2017-02-06 19:01:06 abuild group 2017-02-06 19:03:02 yep 2017-02-06 19:03:15 kaniini: but the check is apparently wrong as root shouldn't need any groups 2017-02-06 19:03:34 ACTION thinks he'll have to patch the tools a bit 2017-02-06 19:04:40 Btw the default vim configuration in Alpine is highly nonstandard. 2017-02-06 19:05:25 (But the config is pretty short.) 2017-02-06 19:07:06 pavlix: what is the standard vim config? 2017-02-06 19:07:32 is it the upstream provided config? 2017-02-06 19:08:02 ncopa: I suspect I'm just used to compatible mode and the first line is nocompatible... I don't know what's upstream, I'm just assuming from what's typically in distributions... one is always learning. :) 2017-02-06 19:09:05 for those interested in the ppc64le alpine port, we made the packages we built so far available. http://ftp.unicamp.br/pub/ppc64el/alpine/ 2017-02-06 19:09:51 leitao: thats awesome 2017-02-06 19:10:21 ncopa, yes. I will be working to make a public builder available 2017-02-06 19:10:38 leitao: that’s great! if I only have some HW to test it :) 2017-02-06 19:11:26 jirutka, I can provide a hardware for you, not with alpine at the moment. 2017-02-06 19:11:36 jirutka, http://openpower.ic.unicamp.br/minicloud/ 2017-02-06 19:12:05 leitao: we need 1 builder that can be official ppc64 builder and at least one that developers could access 2017-02-06 19:12:17 ncopa, right. I will be working on that 2017-02-06 19:12:20 cool, I must look at minicloud platform 2017-02-06 19:12:25 we need a builder and a porting machine yes 2017-02-06 19:12:36 we use lxc for the other archs 2017-02-06 19:12:53 yeah, it can be one physical machine. just two ens 2017-02-06 19:12:55 envs* 2017-02-06 19:13:37 ok 2017-02-06 19:14:51 the host system could in theory be ubuntu or fedora, with lxc 2017-02-06 19:15:08 ncopa, if that is the case, we can have it right now 2017-02-06 19:15:13 with the minicloud 2017-02-06 19:15:26 not a big VM, but, we can have it real soon 2017-02-06 19:16:14 i suppose its a start 2017-02-06 19:16:35 There's probably one more reason to use an automatic chroot and that's dependency list verification. 2017-02-06 19:16:37 i wonder if we can run qemu/kvm there to, so we can test kernel 2017-02-06 19:17:07 pavlix: we have build in a "chroot" on the list 2017-02-06 19:17:40 ncopa, You can, using kvm-pr instead of kvm-hv. 2017-02-06 19:18:05 what is kvm-pr? 2017-02-06 19:18:11 paravirt? 2017-02-06 19:18:38 KVM-PR: 2017-02-06 19:18:39 Uses the so-called "PRoblem state" of the ppc CPUs to run the guests, i.e. the VM is run in user mode and all privileged instructions trap and have to be emulated by the host. 2017-02-06 19:18:44 ncopa: Yep, we discussed it here, just adding my ideas to the list of reasons. 2017-02-06 19:18:46 KVM-HV: 2017-02-06 19:18:47 Requires hardware support of modern POWER server CPUs (like the recent POWER7/POWER8 CPUs). 2017-02-06 19:18:47 Can only be used to run sPAPR (pseries) guests. 2017-02-06 19:18:47 Guests that use a lot of privileged instructions are running much faster than with KVM-PR. 2017-02-06 19:19:01 you can only use kvm-hv on a bare metal machine 2017-02-06 19:19:20 for more info, you can check at http://wiki.qemu-project.org/Documentation/Platforms/POWER 2017-02-06 19:20:03 how big is the performance impact on qemu-pr? 2017-02-06 19:20:44 as long as it does not take like 5mins to boot a kernel it should be ok i suppose 2017-02-06 19:20:45 maybe something around 30% 2017-02-06 19:21:52 So it looks like the worst problem with networkmanager is failed dependency on libgudev 2017-02-06 19:23:10 I wouldn't say it's the worst problem of networkmanager, but it's one of the problems, I guess 2017-02-06 19:24:01 leitao: i dont know if i will have time this week, but i am interested in getting ppc64 up and run 2017-02-06 19:24:26 ncopa, cool! Thanks! 2017-02-06 19:25:13 leitao: i also think you have done the hardest parts already so it should be relatively easy now 2017-02-06 19:25:40 ncopa, that is what I think also. Maybe grub2 might be the hardest part. 2017-02-06 19:25:48 how are you handling aarch64 in terms of bootloader? 2017-02-06 19:25:50 skarnet: It looks like that was just my problem... not being used to the tools, yet. 2017-02-06 19:26:17 skarnet: You can built NetworkManager for pretty much everything Linux as long as you have udev and some other crazy tools. 2017-02-06 19:26:24 pavlix: oh, to package it, maybe. Using it is a whole other enchilada. 2017-02-06 19:26:44 skarnet: It's sometimes awkward and it has bugs... 2017-02-06 19:27:00 skarnet: But you can say that about all sorts of tools, right? 2017-02-06 19:27:14 that's one of the things I had started to rewrite last year, but my customer cut me the funds 2017-02-06 19:27:29 he said "we don't need that right now" 2017-02-06 19:27:42 skarnet: Started rewriting NetworkManager? 2017-02-06 19:28:14 a program that does more or less the same, but in a sane way, yeah. But it never got past the concept phase because I was told to drop it. :/ 2017-02-06 19:28:43 leitao: i think we boot the thunder-x with grub in uefi mode. clandmeter is that correct? 2017-02-06 19:29:03 Voilà! NetworkManager builds for me. 2017-02-06 19:29:10 we still dont have solved it with official packages 2017-02-06 19:29:20 xgene supports uboot too 2017-02-06 19:29:25 skarnet: Everybody writes their own NetworkManager. 2017-02-06 19:29:37 maybe we're not talking about the same one. 2017-02-06 19:29:50 skarnet: I've been making talks about various NM rewrites and replacements for ages... until I decided to write my own tools for testing. 2017-02-06 19:30:15 pavlix: i think we ship uboot files that should work, but clandmeter knows probably more 2017-02-06 19:30:16 skarnet: Recently. 2017-02-06 19:31:14 leitao: i don't really like grub TBH, but there dont seem to be many alternatives that is portable 2017-02-06 19:31:25 tbfh the last time I tested NM was in 2010, so it may have significantly improved since 2017-02-06 19:31:32 but in 2010 it was unusable. 2017-02-06 19:31:34 ncopa: Missed the context, uboot files for what? 2017-02-06 19:31:42 skarnet: I made some significant improvements on 2012. 2017-02-06 19:31:46 aarch64, x-gene 2017-02-06 19:31:50 oh, good :) 2017-02-06 19:31:53 skarnet: And others made many improvements since then. 2017-02-06 19:32:14 ncopa, correct. Grub2 is very bloated. yaboot also works with ppc64le 2017-02-06 19:32:19 skarnet: The problem is... back in 2012 I started rewriting it to allow creating a testsuite... but the testsuite was never created. 2017-02-06 19:32:36 skarnet: I left the team because of a new boss I wouldn't work with. 2017-02-06 19:32:47 yeah, well, you know how I usually consider "improvements" from RedHat, right? with 2 or 3 trucks of salt 2017-02-06 19:33:25 not blaming you personally, obviously (I even have no idea what you did), but the RedHat development process as a whole in general 2017-02-06 19:33:56 I'd have to try and use it again to have a valid opinion. 2017-02-06 19:34:09 leitao: ok. similar as the situation on all archs. there are nicer alternatives for the given arch but none that works on all archs 2017-02-06 19:34:47 pavlix: are you a networkmanager dev? 2017-02-06 19:35:35 what am i supposed to do of: undefined reference to `scandirat' ? 2017-02-06 19:36:09 send hate mail to the authors of the software 2017-02-06 19:37:02 coredumb: i'd ask in #musl 2017-02-06 19:37:24 you can do both! 2017-02-06 19:37:48 sounds like a lot of pain 2017-02-06 19:37:54 :) 2017-02-06 19:40:19 ncopa: Yep. I am an occasional NetworkManager contributor and I used to be a core developer at that time paid fulltime by Red Hat. 2017-02-06 19:40:58 ncopa: And I'm overall into C, Python and networking. 2017-02-06 19:42:41 pavlix: cool! nice that you found your way to here :) 2017-02-06 19:43:07 that's serious improvement :D 2017-02-06 19:47:02 pavlix: i admit that i havent really used NM with alpine 2017-02-06 19:47:11 i have modified wpa_supplicant.conf manually 2017-02-06 19:47:15 which is a bit of a pain 2017-02-06 19:54:05 i use connman peronally 2017-02-06 19:54:08 its nice 2017-02-06 19:56:44 coredumb: sounds like the relevant source file needs to define _GNU_SOURCE befor any includes 2017-02-06 19:57:21 if it still doesn't work then, it's not implemented in musl :p 2017-02-06 19:57:26 shouldn't be too hard to add though 2017-02-06 19:57:43 Shiz: does connman work on alpine? 2017-02-06 19:57:52 i think i made it build but never tested if it actually work 2017-02-06 19:58:18 i haven't checked on alpine itself, admittedly 2017-02-06 19:58:31 when i did a cursory check through the apkbuild i think it had a missing dep (dbus) 2017-02-06 20:13:23 Shiz: yeah don't know if it's worth to add though 2017-02-06 20:25:16 ncopa: from experience in the Adélie PowerPC port: yaboot is old, broken, unmaintained, and does not work with static e2fsprogs built against musl. it would take much effort to make it build against musl, and even more effort to make it actually work once it built 2017-02-06 20:26:22 :-/ 2017-02-06 20:26:45 ncopa: a full fancy grub2 with luks+lvm+every fs easy-kernel supports (I guess that is like alpine linux-vanilla) = 2.1M installed 2017-02-06 20:26:53 it is not so bloated 2017-02-06 20:27:02 yaboot is about 500K 2017-02-06 20:27:13 so yes, it is 4x bigger, but yaboot does not support luks, lvm, or any fs except HFS and ext2/3 2017-02-06 20:27:21 grub2 is just not simple and nice 2017-02-06 20:27:25 it does not support extents so ext4 is dangerous 2017-02-06 20:27:28 its a bit grubby 2017-02-06 20:27:42 yaboot is sadly not suitable for a modern linux 2017-02-06 20:27:48 probably not 2017-02-06 20:28:54 https://bpaste.net/show/42385a386a1e 2017-02-06 20:29:23 yaboot does about 10% of what grub2 can do, and yet is only 70% smaller 2017-02-06 20:31:59 oh, also, if you actually care enough to configure it (that is using default which is basically "enable everything" 2017-02-06 20:32:01 ) 2017-02-06 20:32:18 -rw-r--r-- 1 root root 676112 Jan 31 22:30 grubcore.img 2017-02-06 20:32:42 this is what we ship on the CD, which is enough to support multiple FSes and serial/framebuffer (but no graphical stuff or gcry/luks) 2017-02-06 20:32:59 which is almost equivalent to yaboot :) 2017-02-06 21:27:27 How can I use the generic arm image to install alpine on my arm board? 2017-02-06 21:27:39 Is it its exact purpose? 2017-02-06 21:28:30 ncopa: Sorry for my late answers, I'm still at a conference and I'm basically moving every now and then between rooms, bars, places to eat and a hostel. :) 2017-02-06 21:30:05 ncopa: I was going to write a set of wrappers over wpa_supplicant, iproute and such stuff, probably not too different from what skarnet said he wanted to do as well, but mostly for testing purposes. NetworkManager tends to be a bit tricky and fixing it means digging into quite some C code, so a simple Python wrapper might fit my needs better. 2017-02-06 21:32:46 Shiz: I packaged connman for Fedora some time ago. 2017-02-06 21:33:22 Shiz: But I see huge technology debt with connman so I'm not looking into that so much. 2017-02-06 21:34:48 technology debt? 2017-02-06 21:37:53 I think they mean "technical debt" 2017-02-06 21:47:09 same question applies 2017-02-06 22:01:26 Shiz: They're using their own implementations of so many things including DNS and DHCP and they are not on par with the current development of those protocols. DNSSEC being just an example. 2017-02-06 22:01:35 Shiz: I don't know their exact plans on that, though. 2017-02-06 22:03:25 right 2017-02-06 23:01:19 kaniini: You seem to be also involved in Fedora, right? 2017-02-06 23:03:35 not really 2017-02-06 23:03:57 they use some software that originated in alpine (pkgconf) 2017-02-06 23:04:47 fedora's package management is way too slow for me to want to be involved directly :P 2017-02-06 23:11:22 pavlix: right, that whole internal dhcp/dns thing does look quite ugly 2017-02-06 23:11:44 i wish there was a relatively simple thing like connman except without dbus and just fork out to e.g. dhcpcd for dhcp 2017-02-06 23:11:46 :p 2017-02-06 23:14:33 Shiz: I can do that for you. 2017-02-06 23:14:46 pavlix: +1 2017-02-06 23:14:53 Shiz: Provided that you can live with Python as a dependency. 2017-02-06 23:15:03 Shiz: Plus a couple of python libs. 2017-02-06 23:15:23 Shiz: And provided that it won't be enterprise quality. 2017-02-06 23:15:48 lol 2017-02-06 23:15:57 Shiz: dhcpcd and wpa_supplicant would be used 2017-02-06 23:15:57 pavlix: can you write it in Lua instead of Python? :P 2017-02-06 23:16:08 i'd like something in C... 2017-02-06 23:16:11 but/w 49 2017-02-06 23:16:12 jirutka: I would consider it. 2017-02-06 23:16:14 whoops 2017-02-06 23:16:14 or in C 2017-02-06 23:16:21 it's a project i might undertake at one point 2017-02-06 23:16:46 Shiz: C takes too much time to maintain, Python or Lua would be acceptable, C wouldn't for my case. 2017-02-06 23:16:59 i understand 2017-02-06 23:17:19 that's why i might undertake it myself sometime 2017-02-06 23:17:33 Shiz: It might change in the future but then there's connman and NM in C. :) 2017-02-06 23:17:47 yeah, but NM is a no-go for me in either case 2017-02-06 23:17:50 Shiz: Yet another network config daemon... let's soon have 20. :) 2017-02-06 23:17:50 for various reasons 2017-02-06 23:17:56 Shiz: I see. 2017-02-06 23:18:12 Shiz: Why would you insist on C? 2017-02-06 23:18:55 I guess bootstrappability 2017-02-06 23:19:04 that's my usual reason for wanting C 2017-02-06 23:19:34 Lua isn’t easy to bootstrap on new platform? ;) 2017-02-06 23:19:38 i guess i don't want something that's pretty core to an OS to have heavy dependencies 2017-02-06 23:19:45 (and I don't like Lua :p) 2017-02-06 23:21:34 Shiz: It is so hard to get a networking tool do the job right. So my actual goal is to make a test suite. The tool would just be a reference implementation for the test suite. But then it might actually be more correct in many situation than untested tools like NetworkManager. 2017-02-06 23:22:15 that’s very reasonable approach 2017-02-06 23:22:45 Shiz: So if we could work together to define possible configurations and create some “standard” for configuration, I can actually quickly do it in Python and you can have your alternative C implementation support the same stuff. 2017-02-06 23:23:01 pavlix: I would actually love a test suite for that kind of functionality - that could be a sound basis for actual tool implementations 2017-02-06 23:23:11 that sounds like an interesting idea :) 2017-02-06 23:23:24 and maybe I can then implement it in Lua or Rust :) 2017-02-06 23:23:26 skarnet: It's one of my plans after improving the client-server test suite I'm working on. 2017-02-06 23:23:41 i wouldn't be opposed to that 2017-02-06 23:23:43 it’s much easier when you have a good test suite under it 2017-02-06 23:24:06 I'm 100% behind that 2017-02-06 23:24:12 jirutka: Exactly. We just need to create a common standard and actually settle on the terminology and stuff like that. 2017-02-06 23:25:18 jirutka: It sounds easy but it's not easy at all. Everyone has his own nitpicking points. I for example hate when NetworkManager uses “mac address cloning” instead of “mac address setting/spoofing” just because some router vendors misunderstood the original cloning feature. 2017-02-06 23:25:42 skarnet: I would then create some convertors to get also NetworkManager tested. 2017-02-06 23:26:59 I'd be happy if you write down your use cases and give it to me for the beginning. :) 2017-02-06 23:29:31 i guess i've got one aside from the normal standard ones 2017-02-06 23:29:46 well, a few random thoughts more like 2017-02-06 23:30:10 preferring ethernet over wifi, switching to stronger wifi networks depending on signal (but possibly only for certain networks, e.g. still prefer own network over public networks) 2017-02-06 23:30:47 2.4/5 band switching/preferences (you don't always want to connect to a 5GHz network if its signal is bad) 2017-02-06 23:32:53 I wanted my tool to have all that user-configurable (every network had a priority, the tool only switched to a network with higher priority, and at some point there would have been a function to adjust dynamic priority with signal strength) 2017-02-06 23:33:02 I'm not sure what has a place in a test suite here 2017-02-06 23:33:39 Shiz: I would actually have to have detailed descriptions of test cases and we would have to figure out how to test wifi properly. NetworkManager folks talked about it for years but they didn't actually start doing it. 2017-02-06 23:33:46 sure 2017-02-06 23:34:01 also surely there's some kind of tool to simulate a wifi adapter... 2017-02-06 23:34:09 i would be surprised if there wasn't 2017-02-06 23:34:40 Shiz: There is. But I don't have any working example at hand. I'd like to model exact situations and exact expected behaviors. 2017-02-06 23:34:45 is a signal quality really a good metrics? 2017-02-06 23:35:11 jirutka: It's not a good metric but sometimes it's maybe better than no metric. 2017-02-06 23:35:38 i think ti's mainly important because it's one of the few metrics that you can even encode programatically/automatically 2017-02-06 23:35:47 your wifi adapter itself has no idea what a "home network" or "public network' is 2017-02-06 23:35:52 (except maybe encryption, but even then) 2017-02-06 23:36:14 stuff can be filled in with user config of course, but signal quality is one of the few automated metrics 2017-02-06 23:36:16 there is 2017-02-06 23:36:44 maybe the right question is: passive or active checks? 2017-02-06 23:37:03 jirutka: By all means you need to start with passive, I guess. :) 2017-02-06 23:37:15 yes 2017-02-06 23:37:43 hm, yeah, you can’t connect to multiple networks simultaneously :/ 2017-02-06 23:37:55 Anyway, I'm not going to do any of that for the beginning, but I'd need to be able to at least somehow test the wifi... so that would be a good starting point. 2017-02-06 23:37:56 ayup 2017-02-06 23:38:29 For example NetworkManager doesn't always cope with switching between same-SSID different-configuration networks. 2017-02-06 23:38:46 I'd really like to test for stuff like that. 2017-02-06 23:39:06 that seems good 2017-02-06 23:39:09 Anyway, need to get some sleep. 2017-02-06 23:39:21 sleep well :P 2017-02-06 23:39:39 Don't let your ideas get lost! 2017-02-06 23:39:50 sleep well! and they won't 2017-02-06 23:39:54 <^7heo> ghaaaa 2017-02-06 23:40:10 <^7heo> SO MUCH LOG... 2017-02-06 23:41:01 <^7heo> I can not go to the gym and come back to an easy to read log :P 2017-02-06 23:57:02 ncopa : Hi, when you build gcc-java, do you use any options, or just $ abuild -R ? 2017-02-07 00:03:07 <^7heo> pavlix: "won't be entreprise quality" < I sure hope so, it would be disasterous 2017-02-07 00:40:44 tmh1999: i don't think the build servers use any specific options that differ per-package 2017-02-07 01:01:18 Shiz : thanks ! 2017-02-07 01:02:05 I wonder because I have an ancient error which was probably patched in 2009 ... 2017-02-07 09:51:30 what's the best way to debug non booting - like freezing cursor - kernel from extlinux ? 2017-02-07 09:52:17 you can add boot cmdline: noquiet debug 2017-02-07 09:52:29 when does the freeze happen? 2017-02-07 09:52:32 in the initramfs? 2017-02-07 09:54:08 ncopa: no idea it's right after the menu 2017-02-07 09:54:15 just nothing happens 2017-02-07 09:54:25 let me try noquiet debug 2017-02-07 09:55:41 it could be a fbcon issue 2017-02-07 09:55:58 try add fbcon to the modules=... 2017-02-07 09:56:11 it could ? 2017-02-07 09:56:28 should it show up with debug ? 2017-02-07 09:57:14 if it is fbcon issuem then the system actually boots, but the screen is just black 2017-02-07 09:57:34 mmmh 2017-02-07 09:57:44 seems a bit different from what I have 2017-02-07 09:57:53 as I only get the cursor 2017-02-07 09:59:19 <^7heo> coredumb: drop the space between the last word and '?' or write in French :D 2017-02-07 09:59:47 :) 2017-02-07 10:04:02 ^7heo: you know it's actually quite funny that ppl get over this 2017-02-07 10:04:32 I've never cared for spaces or not before punctuation 2017-02-07 10:08:50 sometimes I wish like natural languages to have strict error handling… 2017-02-07 10:33:49 hi! would you accept a PR for git 2.11.1 for both edge AND 3.5-stable? 2017-02-07 10:50:13 ncopa: are you ignoring me again? 2017-02-07 10:51:10 no only skarnet 2017-02-07 10:51:18 :D 2017-02-07 10:51:35 ncopa isn't ignoring me 2017-02-07 10:51:43 you are! 2017-02-07 10:51:56 who? me? 2017-02-07 10:51:58 clandmeter: yes, i try :) 2017-02-07 10:52:15 ACTION consoles clandmeter 2017-02-07 10:52:52 clandmeter: no, coredumb is. (this is confusing.) 2017-02-07 10:52:52 ACTION crawls back into his windmill 2017-02-07 10:53:52 :D 2017-02-07 11:09:22 <^7heo> jirutka: yeah but then "ppl cnt wrt ths" 2017-02-07 11:09:34 ^7heo: ? 2017-02-07 11:09:47 hum firefox does not build with latest libevent 2017-02-07 11:09:55 <^7heo> if they had strict error handling 2017-02-07 11:10:17 and that would be a good thing imo :) 2017-02-07 11:10:27 <^7heo> jirutka: yeah but then "ppl cnt wrt ths" 2017-02-07 11:10:39 <^7heo> (yeah I know) 2017-02-07 11:11:01 i should go to lunch 2017-02-07 11:11:09 <^7heo> enjoy 2017-02-07 11:19:29 same for me, croque monsieur time. 2017-02-07 11:21:12 <^7heo> remind me 2017-02-07 11:21:52 <^7heo> this is like toast with bechamel and ham in the middle right? 2017-02-07 11:22:38 <^7heo> and if I remember right, a croque madame is the same with eggs on top 2017-02-07 11:31:02 ^7heo: its dutch cheese with ham (atleast thats what im having) 2017-02-07 11:37:21 <^7heo> ah yeah 2017-02-07 11:37:26 <^7heo> nah it's not 2017-02-07 11:38:11 <^7heo> wrong name 2017-02-07 11:38:39 <^7heo> but hey, google sells privacy hm? 2017-02-07 11:41:31 we call it Tosti, not sure thats very international. 2017-02-07 11:41:44 it is very offtopic though :) 2017-02-07 11:45:39 <^7heo> yeah, better. 2017-02-07 12:25:02 clandmeter, ncopa: any comments on http://patchwork.alpinelinux.org/patch/2916/ ? 2017-02-07 12:27:23 lgtm 2017-02-07 12:29:50 heh, interesting 2017-02-07 12:29:54 +1 2017-02-07 12:30:51 how many pkgs are we talking about? 2017-02-07 12:36:34 clandmeter, i have no idea how many golang packages we have. probably some tens? 2017-02-07 12:42:15 fabled: if you’re going to merge https://github.com/alpinelinux/aports/pull/716, please review it and test very carefully; I’m afraid that it will break many things 2017-02-07 12:42:45 jirutka, that's why i suggested adding it to testing first :) 2017-02-07 12:43:10 jirutka, seems PL/Python support breaks postgres on x86 2017-02-07 12:43:25 probably some dependency is built with gettext-dev and adds the -lintl 2017-02-07 12:43:37 checking how to link an embedded Python application... -L/usr/lib/python3.5/config-3.5m -lpython3.5m -lintl -ldl -lm 2017-02-07 12:43:38 fabled: yes, but I have no clue what dep 2017-02-07 12:44:09 fabled: I’m quite sure that postgresql abuild is correct, probably some dep is broken for x86 2017-02-07 12:44:36 this is weird 2017-02-07 12:44:38 because: 2017-02-07 12:44:38 $ python3.5-config --libs 2017-02-07 12:44:39 -lpython3.5m -ldl -lm 2017-02-07 12:44:54 oh, for x86_64 2017-02-07 12:44:57 letme test that on x86 2017-02-07 12:45:17 it seems that I must really set up some x86 system 2017-02-07 12:45:31 I don’t have any, I haven’t seen x86 for more than 10 years… 2017-02-07 12:45:50 ttdev-edge-x86:~$ python3.5-config --libs 2017-02-07 12:45:50 -lpython3.5m -lintl -ldl -lm 2017-02-07 12:46:02 either python on x86 was built with extra deps present (and gettext-deV) 2017-02-07 12:46:06 or it's broke somehow 2017-02-07 12:46:19 i think kaniini said we had extra stuff on the official builder earlier 2017-02-07 12:46:25 yes 2017-02-07 12:46:35 pkgbump just python3 ? 2017-02-07 12:46:49 I wanted to suggest that 2017-02-07 12:47:08 assuming it's not any of the python dependencies 2017-02-07 12:47:15 do I understand correctly that build servers does not properly clean environment between builds? 2017-02-07 12:47:25 they should 2017-02-07 12:47:29 it must've been a bug 2017-02-07 12:47:44 or someone fiddling it manually for a while and forgetting to clean up things 2017-02-07 12:48:30 ok 2017-02-07 12:48:37 let's start with python3 pkgrel bumb 2017-02-07 12:49:19 fabled: quick grepping shows http://tpaste.us/5vv0 2017-02-07 12:57:21 fabled: it seems that rebuilding python3 helped :) 2017-02-07 12:57:28 yeah 2017-02-07 13:01:28 +1 for bumping python3 pkgrel 2017-02-07 13:06:22 ncopa, can you clarify https://github.com/alpinelinux/aports/pull/716#issuecomment-277974673 why you mean "stable" branches as php7.1 should go into testing? 2017-02-07 13:16:51 andypost: how do you mean? 2017-02-07 13:16:57 i odnt follow 2017-02-07 13:17:01 dont* 2017-02-07 13:20:17 ncopa, why not create testing/php7.1 and review what will be broken 2017-02-07 13:21:26 but you said about lack of resources to maintain 3 php versions in *stable* 2017-02-07 13:24:10 andypost: im thinking of it 2017-02-07 13:24:17 and i think that might be a good idea 2017-02-07 13:25:16 but please do not create pkgs depending on php71 2017-02-07 13:25:26 +1 2017-02-07 13:25:32 we have already dozens of php packages with separate apkbuild for php5 and php7 2017-02-07 13:25:45 it’s already total mess that is hard to maintain 2017-02-07 13:25:52 +1 2017-02-07 13:28:06 jirutka, I totally agree with you, that's why Valery trying unify process of building php 2017-02-07 13:28:32 andypost: the problem is not with php abuilds, but dependencies 2017-02-07 13:28:58 same for me 2017-02-07 13:29:41 I'm going to review all php(5|7) dependant pkgs this weekend 2017-02-07 13:30:08 ACTION commented on PR 2017-02-07 13:35:14 jirutka, the main problem here is packages like zabbix, postfixadmin that require to be checked about support of php 7.0 & php 7.1 2017-02-07 13:36:06 btw why is 7.1 not backward compatible with 7.0? what are the breaking changes? 2017-02-07 13:38:10 https://secure.php.net/manual/en/migration71.incompatible.php 2017-02-07 13:38:43 jirutka, basically because their release cycle supposes incompatibilities in minors 2017-02-07 13:39:11 only patch-releases have BC 2017-02-07 13:39:58 so they don’t give a fuck about semver, distributions and their users 2017-02-07 13:40:55 wait, you expect everyone in the world to give a fuck about semver? 2017-02-07 13:41:01 you're in for a surprise 2017-02-07 13:41:35 semver is a good idea, doesn't mean everyone "gives a fuck" about it 2017-02-07 13:41:55 we’re talking about language interpreter 2017-02-07 13:42:13 as for distributions, most of them seem not to complain 2017-02-07 13:42:19 debian at least chose to only distribute 7.0 and 7.1 as of current sid 2017-02-07 13:42:46 I don’t know any other language doing backward incompatible changes in minor versions 2017-02-07 13:42:54 ruby 2017-02-07 13:42:58 not true 2017-02-07 13:43:06 you can still run 1.8.7 code under 2.3.0 2017-02-07 13:43:26 you can also run php 5.3 code on php 7.0 - most of my code works 2017-02-07 13:43:29 well, "my" 2017-02-07 13:43:34 code i inherited written in 2006 2017-02-07 13:43:44 so why we can’t just upgrade php7 to 7.1? 2017-02-07 13:43:49 aha, because it’s not backward compatible 2017-02-07 13:43:51 because *someone* might be hit by those changes 2017-02-07 13:44:55 do we have python31, python32, python33, python34, python35, python36, …? no. does anyone need it? no. there ale only very few exceptions, like mailman-core does not work correctly with Python 3.5 (that should be already fixed in master) 2017-02-07 13:45:15 indeed 2017-02-07 13:45:19 php also has "very few exceptions" 2017-02-07 13:45:53 lua? 2017-02-07 13:45:54 sorry, I don’t wanna discuss this; I should not be so negative, so I must not talk about php, beucase it’s totally impossible to be positive when talking about php 2017-02-07 13:45:58 kunkku: yes, lua 2017-02-07 13:46:01 that's a good example 2017-02-07 13:46:13 a lot of distros still ship lua 5.1, 5.2 and 5.3 2017-02-07 13:46:18 because not only does the language break, the c bindings also break 2017-02-07 13:46:22 do you know about any lua library that does not work under 5.3? 2017-02-07 13:46:40 the reason why we still have lua51 is luajit 2017-02-07 13:47:08 lua53 removes bit32 and adds the differentiation between integer and float 2017-02-07 13:47:14 also luajit is not even lua51 2017-02-07 13:47:17 it's a mix of lua 5.1 and 5.2 2017-02-07 13:47:19 it's its own language now 2017-02-07 13:47:28 yes, but this is very different situation 2017-02-07 13:47:49 i mean lua vs. luajit 2017-02-07 13:47:59 yes, those are two different things pretty much 2017-02-07 13:48:11 but lua51, lua52 and lua53 all are incompatible in some key ways with each other 2017-02-07 13:48:22 and, since they are very often used in games 2017-02-07 13:48:30 a game based around lua52 is unlikely to port to lua53 because the effort is not worth it 2017-02-07 13:48:52 i’m quite sure that we can drop lua52 without any problem; we need to keep lua51 just because of luajit 2017-02-07 13:49:45 wireshark still depends on 5.2 in arch 2017-02-07 13:49:47 so does freeciv 2017-02-07 13:49:59 mpv lets you choose 2017-02-07 13:50:26 ah, freeciv updated to lua 5.3 2017-02-07 13:50:37 but wireshark still depends on lua 5.2 2017-02-07 13:51:08 wait, that is also apparently not true. why does nobody document these changes? 2017-02-07 13:51:16 oh, it didn't. http://www.linuxfromscratch.org/patches/blfs/svn/wireshark-2.2.1-lua_5_3_1-1.patch 2017-02-07 13:51:44 but yes, language interpreters are a mess in particular because there's layers and layers of dependencies beneath 2017-02-07 13:52:11 you have lua, but then you have something that uses lua and then scripts for that which often need porting etc 2017-02-07 13:57:36 actually php is the same - php7 is mostly different language as of php5 2017-02-07 13:59:08 and BC break in minor makes language more strict but adoption takes years, mostly because of badly written code 2017-02-07 14:12:44 i kinda like andypost's suggestion 2017-02-07 14:12:54 we add php7-7.1 to testing 2017-02-07 14:13:00 but we name it php7 2017-02-07 14:13:13 so it will be incompatible with the community/php7 2017-02-07 14:13:38 I really don’t like this idea – naming it php7 2017-02-07 14:13:48 but at least it gives you a way to test php 7.1 by selecting testing repo 2017-02-07 14:13:49 it’s confusing 2017-02-07 14:13:56 jirutka: wat do you propose? 2017-02-07 14:14:02 to name it php71 2017-02-07 14:14:20 is there a way to point that packages incompatible? 2017-02-07 14:14:24 <^7heo> ncopa: nah it's really confusing. 2017-02-07 14:14:35 <^7heo> ncopa: we don't have such a good tool in apk to allow us to do that atm. 2017-02-07 14:14:53 you don’t usually specify repository when installing pkgs using apk, so if you have both community and testing repos (not pinned), then what you will get? 2017-02-07 14:14:54 <^7heo> ncopa: we would need apk to report thoroughly what happens when and from where, with every source. 2017-02-07 14:15:00 <^7heo> ncopa: kinda like apt-cache madison 2017-02-07 14:15:47 jirutka: if you have both community and testing, then you'd get the latest, eg 7.1 2017-02-07 14:17:28 jirutka, I think everyone who use testing using pinned packanges otherwise they doing it for reason 2017-02-07 14:17:39 but that said, i named libressl2.4-libs so we could upgrade to libressl2.5 2017-02-07 14:17:47 <^7heo> ncopa: so you'll get testing while you'd expect not. 2017-02-07 14:18:04 <^7heo> ncopa: I personally would expect community to have a higher priority than testing unless explicitely pinned. 2017-02-07 14:18:11 <^7heo> ncopa: no matter what version. 2017-02-07 14:18:30 why would getting packages that are there for testing be surprising when you add the "testing" repo? 2017-02-07 14:18:46 <^7heo> because you might add the testing repo just for pinning with it. 2017-02-07 14:19:16 also Anitya (release-monitoring) knows only pkgname, not repository (for good reason); but it should not be a problem for testing, b/c there are not checked anyway 2017-02-07 14:19:56 <^7heo> ncopa: or would it only happen when you set the testing repo without an "@whatever" in front? 2017-02-07 14:20:02 b/c many users don’t understand what these repos actually are… 2017-02-07 14:20:21 <^7heo> also, yeah, I agree, it's a two dimentional repository system 2017-02-07 14:20:23 ^7heo: if you add pinning repo you'll not get anything from testing unless you explicitly say you want from that repo 2017-02-07 14:20:28 <^7heo> ok yeah 2017-02-07 14:20:43 <^7heo> with versions vertically, and different "types" horizontally 2017-02-07 14:20:49 so 2017-02-07 14:20:51 <^7heo> (or vice versa) 2017-02-07 14:20:51 when they write a bug report or qustion on IRC, they usually don’t paste content of their /etc/apk/repositories, just commands that the run 2017-02-07 14:20:52 to conclude 2017-02-07 14:21:00 what do we agree on so far 2017-02-07 14:21:08 <^7heo> to conclude, it's gonna be confusing. 2017-02-07 14:21:10 that php 7.1 should be in testing repo 2017-02-07 14:21:22 still undecided what we name it 2017-02-07 14:21:27 so they write you `apk add php7`, but you can’t know what it’s gonna install when php7 is in multiple repos 2017-02-07 14:21:30 <^7heo> php-testing? 2017-02-07 14:21:41 <^7heo> I mean, ok, apk add php-testing@testing... 2017-02-07 14:21:53 apk add php7@testing 2017-02-07 14:22:06 would give you from @testing repo 2017-02-07 14:22:06 <^7heo> yeah for those who have it pinned. 2017-02-07 14:22:19 I really prefer to have different from php7 name in testing 2017-02-07 14:22:25 <^7heo> myeah I guess it'd be fine. 2017-02-07 14:22:25 I know, but how many people (especially docker users) actually know and use pinning? 2017-02-07 14:22:34 <^7heo> andypost: but what name? 2017-02-07 14:22:42 and if you dont add it as @testing it will pick the latest version 2017-02-07 14:22:42 ok 2017-02-07 14:22:50 if we dont use the same name 2017-02-07 14:22:56 then i think we should call it php7.1 2017-02-07 14:22:57 ^7heo, php71 or php7.1 as debian doing 2017-02-07 14:23:04 <^7heo> jirutka: docker users don't even know about apk... 2017-02-07 14:23:06 I can’t even count how many times I see bug report that `apk add something` does not work, just because user is too dumb to realize that we have more than single repository and that his package is not in main 2017-02-07 14:23:32 <^7heo> andypost: myeah but that's a bit problematic to start putting the version in the name while we already have a versionning system. 2017-02-07 14:23:46 and you expects from users to not add testing repo without pinning? 2017-02-07 14:24:00 <^7heo> jirutka: well, it's better than users reporting that their package isn't in the right version 2017-02-07 14:24:06 the problem with php7.1 vs php7 is that we get the php7-* + php7.1-* packages 2017-02-07 14:24:07 <^7heo> jirutka: because they install alpine:latest from docker 2017-02-07 14:24:09 ^7heo: exactly 2017-02-07 14:24:17 <^7heo> jirutka: and do not ever use apk. ever. 2017-02-07 14:24:30 ncopa: now, we both agreed that there should not be any pkgs depending on php71 2017-02-07 14:24:31 <^7heo> ncopa: exactly. 2017-02-07 14:24:38 ncopa: it should be just for testing, nothing else 2017-02-07 14:24:44 ncopa: s/now/no/ 2017-02-07 14:24:45 <^7heo> ncopa: and it's gonna be php7.1-7.1.x-rx 2017-02-07 14:24:55 <^7heo> ncopa: which is really ugly imho 2017-02-07 14:25:21 what happens when we drop support for php 7.0? 2017-02-07 14:25:30 php71 should be only temporary solution, right? I hope that no one wants to maintain php70, php71, php72, … in stable branches 2017-02-07 14:25:33 <^7heo> ncopa: we move the testing to community? 2017-02-07 14:25:40 just upgrade php7 to 7.1 2017-02-07 14:25:44 <^7heo> that ^ 2017-02-07 14:25:55 and remove the php7.1 package 2017-02-07 14:25:58 yes 2017-02-07 14:26:01 <^7heo> wait I have a question 2017-02-07 14:26:07 <^7heo> why do we need 7.1 in testing again? 2017-02-07 14:26:19 jirutka, when 72 will be released 7.0 will gone in support 2017-02-07 14:26:21 <^7heo> isn't that why we have releases and edge? 2017-02-07 14:26:22 ^ 2017-02-07 14:26:42 so people can test that their php apps work with php7.1 2017-02-07 14:26:48 I already wrote that in PR, why we should add 7.1 anyway? 2017-02-07 14:26:49 so people can do php app development 2017-02-07 14:27:01 well, ok 2017-02-07 14:27:12 <^7heo> ncopa: in that case, don't we need something more than edge? 2017-02-07 14:27:27 btw just for a comparison, do you know how Fedora handles this? 2017-02-07 14:27:35 <^7heo> ncopa: like, with debian, next release, and edge. 2017-02-07 14:27:35 no? 2017-02-07 14:27:38 they have just single php package, currently 7.0 2017-02-07 14:28:14 I’ve asked pavlix about update-alternatives few weeks ago and for me surprise, he told me that Fedora use it very rarely, they just don’t need it, b/c they’re not giving much freedom to users 2017-02-07 14:28:43 right 2017-02-07 14:28:45 <^7heo> again, we have 3.X and edge. 2017-02-07 14:28:53 thats kind of what i'd like 2017-02-07 14:28:54 I disagree with that approach, but the point is that it may not be very wise to support gazillion parallel versions, just because some users wants it 2017-02-07 14:28:58 <^7heo> to me it looks that we need 3.X, RC, and edge. 2017-02-07 14:29:29 <^7heo> and all we need is just move tags on the tree 2017-02-07 14:29:36 so 2017-02-07 14:29:39 <^7heo> (the RC tag) 2017-02-07 14:29:44 but that’s what we currently have… 2017-02-07 14:29:49 <^7heo> we don't have RC, do we? 2017-02-07 14:29:51 what i want is that we only have one php7 2017-02-07 14:29:53 we have 2017-02-07 14:29:58 just look at tags in repo 2017-02-07 14:29:58 like we have now 2017-02-07 14:30:06 ^ me too 2017-02-07 14:30:30 when we upgrade to 7.1, then you get php 7.1 only 2017-02-07 14:30:33 if the name will remain php7, then how to point to which branch a PR should be landed 2017-02-07 14:30:37 at that point we drop php 7.0 2017-02-07 14:30:40 <^7heo> nah that's not what I mean. 2017-02-07 14:30:42 <^7heo> I mean: 2017-02-07 14:30:44 that’s why I don’t like the idea about separate 7.1 from the beginning; but ok-ish as a temporary solution just in testing repo 2017-02-07 14:30:48 <^7heo> have RC the SAME way we have edge 2017-02-07 14:30:54 <^7heo> not RC_X 2017-02-07 14:31:00 <^7heo> just a moving RC tag. 2017-02-07 14:31:11 <^7heo> (in addition to the RC_X when needed) 2017-02-07 14:31:20 how should that work? 2017-02-07 14:31:21 <^7heo> RC can be named differently if it's not clear. 2017-02-07 14:31:23 <^7heo> well 2017-02-07 14:31:32 <^7heo> like a release, but with moving tags. 2017-02-07 14:31:38 if you really need php 7.0 then you could use package from 3.5-stable 2017-02-07 14:31:46 exactly 2017-02-07 14:32:11 or one can build the abuild yourself 2017-02-07 14:32:13 <^7heo> jirutka: wasn't the problem that 7.1 would be really experimental and potentially crashing all the time? 2017-02-07 14:32:40 <^7heo> jirutka: we don't expect people to build their own packages, that's the point of packages. 2017-02-07 14:32:53 ^7heo, it will not crash but some dependent apps may crash 2017-02-07 14:32:55 that’s what I usually do… for example, we have still bacula 5, this version is ancient and I doubt that anyone else is using it, so instead of adding yet another ancient pinned package to the official repo, I’ve just backported it into my repo 2017-02-07 14:33:00 <^7heo> andypost: ah. 2017-02-07 14:33:07 <^7heo> andypost: well... that's a problem with PHP then, right? 2017-02-07 14:33:42 ^7heo, fyi drupal spent 3 month to adopt 7.1 for core and create testing 2017-02-07 14:33:52 <^7heo> jirutka: you're a dev, and that's fine that you're doing that; but for most users, packages are the only thing they get. 2017-02-07 14:34:08 <^7heo> andypost: because of the changes in the PHP language? 2017-02-07 14:35:01 ^7heo, mostly because of bugs in code, that caused 7.1 to show warnings/errors 2017-02-07 14:35:36 ^7heo, for example when string variable used as array, 7.1 more strict 2017-02-07 14:35:37 <^7heo> andypost: so yes. 2017-02-07 14:35:58 <^7heo> andypost: because PHP coders are poor coders, and made several mistakes, and were writing code that crashed with the new engine. 2017-02-07 14:36:01 <^7heo> makes sense. 2017-02-07 14:36:16 another example http://stackoverflow.com/questions/41596331/phpmyadmin-with-php-7-1-0-tons-of-deprecation-warnings 2017-02-07 14:36:25 <^7heo> yeah, people being dumb. 2017-02-07 14:36:35 <^7heo> that's enough examples, I knew it before that discussion ;) 2017-02-07 14:36:41 <^7heo> thanks for reminding me of the actual context tho. 2017-02-07 14:42:40 does anyone know how many php packages really does not work with php 7.1? 2017-02-07 14:42:59 is it more than 30%? 2017-02-07 14:43:35 i don't think anyone even has 30% of all php packages 2017-02-07 14:44:13 it may be acceptable for edge to keep some pkgs broken b/c of incompatibility with latest php version, if we’re able to somehow solve it until branching 3.6 2017-02-07 14:44:22 https://seld.be/notes/php-versions-stats-2016-2-edition 2017-02-07 14:58:15 <^7heo> jirutka: I would express wordpress not to be working, if drupal spent 3 months... 2017-02-07 15:05:52 and if its not solved til we do v3.6 then we keep it in testing til v3.7 2017-02-07 15:06:31 jirutka: i think we can enforce not creating any php71-foobar packages if we keep the php7 name 2017-02-07 15:24:47 ^7heo, WP still have issues on 7.1 but most pita is contrib themes 2017-02-07 15:25:24 btw How to point for package that it depends on php5|php7? 2017-02-07 15:26:49 andypost: you can’t, that’s the problem 2017-02-07 15:27:24 andypost: you can specify provides="php" in both php pkgs and declare dependency on php instead of php5 or php7 2017-02-07 15:27:58 andypost: but this does not work as you expect when no php package is installed 2017-02-07 15:28:28 andypost: this brings us back to the problem with handling “alternatives”… 2017-02-07 15:29:04 andypost: too bad that we didn’t have enough time to discuss it on FOSDEM :/ 2017-02-07 15:30:08 oh, yep, now I got the whole trouble 2017-02-07 15:37:27 ncopa: can I repoint my repo-mirror to your repo temporarily while I do maintenance on my mirror? 2017-02-07 15:38:06 or anyone else who has a mirror 2017-02-07 15:38:19 Which mirror(s)? 2017-02-07 15:38:36 ScrumpyJack: which is your repo mirror? 2017-02-07 15:38:45 liskamm 2017-02-07 15:43:25 yeah, we had the build depends for xulrunner on there 2017-02-07 15:48:05 ncopa: Your talk at FOSDEM was very inspiring. Espically that it is possible to clean-up the #ifdef chaos. 2017-02-07 15:48:43 <^7heo> Ganwell: it's either despicable or especially ;) 2017-02-07 15:49:04 <^7heo> ACTION hides 2017-02-07 15:49:43 ^7heo: lol, sorry my fingers sometimes type especially strange things. 2017-02-07 15:50:14 <^7heo> yeah I fugired. 2017-02-07 15:50:29 <^7heo> (but at least I type the right amount of each needed letter ;) 2017-02-07 15:51:11 ^7heo: this reminds me to fix aspell in weechat. it broke when I moved to alpine linux. 2017-02-07 15:51:17 ScrumpyJack: im ok if you point it somewhere else while you do maintenance 2017-02-07 15:51:21 <^7heo> (which has been proven by New York University) 2017-02-07 15:51:26 just check that it works first 2017-02-07 15:51:35 Ganwell: thanks :) 2017-02-07 15:52:31 Ganwell: please file an issue re aspell and weechat on bugs.a.o 2017-02-07 16:15:04 fr.alpinelinux.org is removed 2017-02-07 16:16:14 <^7heo> \o/ 2017-02-07 16:16:18 <^7heo> (sorry) 2017-02-07 16:17:21 the 80* most annoying users have been thrown out in one single operation 2017-02-07 16:17:22 good job 2017-02-07 16:17:26 80%* 2017-02-07 16:34:08 <^7heo> huhuh 2017-02-07 16:41:10 skarnet: sorry, i pushed the "are you sure you want to delete this instance" button 2017-02-07 16:41:35 so dont blame ncopa :) 2017-02-07 16:41:52 "are you sure you want to delete France" 2017-02-07 16:41:55 [X] YES 2017-02-07 16:42:00 :D 2017-02-07 16:43:13 "Are you sure you want to format France completely and remove all inhabitants" 2017-02-07 16:43:13 fr was not France. it was "future removal" 2017-02-07 16:43:28 at least that would delete Le Pen too? 2017-02-07 16:43:53 and it was not really removed it was moved to nl2 2017-02-07 16:44:05 (nl = Nuke Later) 2017-02-07 16:44:15 haha 2017-02-07 16:45:01 future removal, nuke later, delete eventually, ultimately kill 2017-02-07 16:46:49 we have this master plan to implement multi tier mirrors. Its planned after the s6 integration ;-) 2017-02-07 16:47:54 time to go back to my windmill. bon appetit fellow climbers. 2017-02-07 17:23:59 How can I use the generic arm image to install alpine on my arm board? 2017-02-07 17:25:13 Youbi, depends on your board model. we have limited arm boards, and the generic image is tested regularly only on wandboard 2017-02-07 17:25:19 cubieboard might work 2017-02-07 17:27:24 <^7heo> I'll test it asap on usb armory 2017-02-07 17:28:45 the problem is that almost all arm boards boot differently 2017-02-07 17:28:58 and they need u-boot support + kernel support + way to install u-boot 2017-02-07 17:29:37 in most cases those are simple to add, but so far we've added them on request basis when we get developer with access to a board 2017-02-07 17:31:23 I have an olimex lime and lime2, I built u-boot and linux kernel, but I miss a chroot of alpine. 2017-02-07 17:32:16 But there are only u-boot + kernel + packages in the generi arm image, how could I install alpine with that? 2017-02-07 17:33:33 afaik, the first cubie has an allwiner a20, like my boards 2017-02-07 17:33:34 Youbi, use the 'rootfs' tar.gz for getting chroot 2017-02-07 17:34:04 https://nl.alpinelinux.org/alpine/v3.5/releases/armhf/alpine-minirootfs-3.5.0-armhf.tar.gz this? 2017-02-07 17:34:13 yes 2017-02-07 17:34:22 the generic image should have u-boot install script for the few boards it currently supports 2017-02-07 17:35:04 <^7heo> fabled: well, we have developer access to the usb armory. 2017-02-07 17:35:19 Oh yeah, I tried this one! The problem is, I get an error because openrc is missing 2017-02-07 17:35:21 <^7heo> and it is really great, security wise. 2017-02-07 17:35:56 <^7heo> because it has a physically fusable moddule where you can fuse your public key in to check the booted firmwares signatures. 2017-02-07 17:36:01 <^7heo> module* 2017-02-07 17:36:11 <^7heo> (ofc can only be fused once) 2017-02-07 17:38:23 How is the rootfs supposed to boot without /sbin/openrc? 2017-02-07 17:40:45 wasn't me, I swear 2017-02-07 17:44:09 <^7heo> :D 2017-02-07 17:44:29 <^7heo> skarnet: really? Not to save 2MB of cheap disk space? 2017-02-07 17:44:41 <^7heo> ACTION hides behind the biggest rock he can find 2017-02-07 17:48:12 Aha, dirty fix: untar openrc package into chroot, and boot :D 2017-02-07 17:49:55 It works :3 2017-02-07 17:55:16 Well, not so much 2017-02-07 18:00:38 hum. i think ncopa did the rootfs for pure chroot installs, e.g. docker, where init is not started 2017-02-07 18:01:42 <^7heo> Youbi: you can also boot from another media and fix from there. 2017-02-07 18:02:11 <^7heo> Youbi: but CAREFUL, booting from external media is a SECURITY FLAW according to CVE-2016-4484 2017-02-07 18:10:16 the minirootfs is for containers and chroots 2017-02-07 18:10:20 very very minimal 2017-02-07 18:28:29 ncopa: So I shouldn’t use it to run on my arm board? 2017-02-07 18:33:38 you can but you need more than that 2017-02-07 18:33:44 it does not include the kernel 2017-02-07 18:34:10 yeah, dont use that 2017-02-07 18:34:16 use the "Generic ARM" 2017-02-07 18:34:26 https://nl.alpinelinux.org/alpine/v3.5/releases/armhf/alpine-uboot-3.5.0-armhf.tar.gz 2017-02-07 18:34:37 i think 3.5.1 is there too 2017-02-07 18:34:41 https://nl.alpinelinux.org/alpine/v3.5/releases/armhf/alpine-uboot-3.5.1-armhf.tar.gz 2017-02-07 19:50:03 is there a prepared way for going without an alpine_dev? i would like to put everything in initramfs cpio(s). 2017-02-07 20:01:33 it seems normal bootup will extract all apk from boot_repository. so i would either need my boot-repository on initramfs, or extract them before creating initramfs 2017-02-07 20:05:37 th: what do you mean? 2017-02-07 20:06:00 i would like to boot alpine just by handing vmlinuz+initramfs over to bootloader 2017-02-07 20:06:11 (no blockdevice, no nfs, ...) 2017-02-07 20:06:33 kind of a rescue scenario in just the two files... 2017-02-07 20:10:15 hmm 2017-02-07 20:10:18 do you have internet on boot? 2017-02-07 20:10:25 if so, you can just specify alpine_repo as an url :) 2017-02-07 20:10:32 ncopa: but in the “Generic ARM” image, there is no rootfs, only the dtb tree and apks, plus uboot. How can I install alpine with that? 2017-02-07 20:10:58 Shiz: nope - i really would like to pass that as cpio to the kernel (initramfs) 2017-02-07 20:12:05 i guess a somewhat hacky way is to copy all .apks to a directory in /media 2017-02-07 20:12:09 e.g. /media/integrated 2017-02-07 20:12:17 and touch /media/integrated/.boot_repository 2017-02-07 20:12:21 in the initramfs 2017-02-07 20:12:22 with /media on initramfs you mean? 2017-02-07 20:12:25 yeah 2017-02-07 20:12:41 you can do that with some additions in /etc/mkinitfs/features.d 2017-02-07 20:12:46 i had skimmed through nlplug-findfs, which is what currently fails 2017-02-07 20:13:09 right, i forgot that 2017-02-07 20:13:17 i think you can just ^D out of the rescue shell and it should continue 2017-02-07 20:13:20 since that step would be free to skip 2017-02-07 20:13:44 i guess /etc/apk/repositories would need to point to /media/integrated then? or is that handled somehere else? 2017-02-07 20:14:05 yea - i can skip that, then i fail for missing /sbin/init- because i had no alpine-base installed, i guess 2017-02-07 20:14:07 the initramfs handles that 2017-02-07 20:14:14 actually 2017-02-07 20:14:20 just passing alpine_repo=/media/integrated 2017-02-07 20:14:23 to the kernel 2017-02-07 20:14:29 would fix most things, if it's already on the initramfs 2017-02-07 20:14:35 okay - so trying to push my apk-repo into the cpio - and then touch the .boot_repository 2017-02-07 20:15:08 you don't need to touch the .boot_repository file in that case either 2017-02-07 20:15:22 just make sure /media/integrated is a directory containing the apks and an APKINDEX.tar.gz file 2017-02-07 20:15:23 by specifying it explicitly. nice. 2017-02-07 20:15:27 (as any normal repository would) 2017-02-07 20:17:19 i had managed to create my custom iso already (checking out alpine-iso, using a conf.mk...) now i probably need to change mkinitfs as a package? 2017-02-07 20:17:28 (from aports repo?) 2017-02-07 20:18:29 nah 2017-02-07 20:18:42 just add something in /etc/mkinitfs/features.d 2017-02-07 20:18:45 something like 2017-02-07 20:19:11 find /media/integrated > /etc/mkinitfs/features.d/repository.files 2017-02-07 20:19:27 and add "repository" to features= in /etc/mkinitfs/mkinitfs.conf 2017-02-07 20:19:37 and then just re-run mkinitfs 2017-02-07 20:20:02 ahh so my docker build-container's /etc/mkinitfs/features.d is respected when i run the `make PROFILE=alpine-foo iso` there? 2017-02-07 20:20:26 uhm i'm not sure, i've honestly never used alpine-iso stuff 2017-02-07 20:21:28 so you would just run a alpine docker, install sdk packages, adapt /etc/mkinitfs/features.d as above, and fire mkinitfs to get my new cpio? 2017-02-07 20:22:03 i actually do not need the iso, but i had followed the rescue-cd example from https://wiki.alpinelinux.org/wiki/How_to_make_a_custom_ISO_image#A_rescue_CD_example 2017-02-07 20:22:10 i'd also install the kernel inside but yeah 2017-02-07 20:22:22 ok - the easier the better... 2017-02-07 20:22:22 and pass mkinitfs the kernel version of the installed kernel 2017-02-07 20:22:29 else it'll use the host kernel version which is wrong 2017-02-07 20:22:32 since dokcer ;p 2017-02-07 20:22:35 yea 2017-02-07 20:23:05 but yeah, that's what i do for my clean initramfs 2017-02-07 20:23:22 has to be said that i'm not sure how that will interact with modloop however, since i don't use kernel modules (own kernel) 2017-02-07 20:23:31 but it should be a good starting point 2017-02-07 20:23:55 and how i would collect the required apk files for integration 2017-02-07 20:24:09 i guess that is where some alpine-iso magic does $things 2017-02-07 20:24:20 yeah prolly 2017-02-07 20:24:33 http://git.alpinelinux.org/cgit/alpine-iso/tree/alpine-extended.packages 2017-02-07 20:24:41 it has a package list, which i presume it then uses apk fetch to download the .apk files 2017-02-07 20:24:51 yea 2017-02-07 21:17:34 andypost: ncopa: ^7heo: from today’s StackOverflow digest, let me introduce you a docker user: http://stackoverflow.com/questions/42087734/install-php7-gd-in-alpine 2017-02-07 21:19:32 and check out the Dockerfile he’s using https://github.com/docker-library/php/blob/9abc1efe542b56aa93835e4987d5d4a88171b232/7.1/fpm/alpine/Dockerfile 2017-02-07 21:19:46 jirutka, it's interesting thing to compile gd for php 2017-02-07 21:20:24 note that the Dockerfile is based on Alpine v3.4… and he’s adding community repo for edge 2017-02-07 21:20:40 XD 2017-02-07 21:20:43 I hope that no other comments are needed… 2017-02-07 21:21:01 and trying to add extension that using different abi 2017-02-07 21:23:26 why these people even use distributions? they can just compile everything from sources inside one huge “perfectly readable” Dockerfile… oh wait… 2017-02-07 21:35:54 jirutka: distributions are out, with Dockerfiles you can make much better completely stupid mistakes. 2017-02-07 21:44:32 jirutka, I answered 2017-02-07 21:47:44 andypost: oh, I was answering too :) 2017-02-08 02:30:37 someone give fabled a big hug for me when he is next online. my VM crashed hard during an apk operation, and it was able to pick up the pieces and just finish the transaction when it rebooted. no other package manager I have ever used has been so easy to recover after a crash during package maintenance (dpkg, pacman, rpm, yast) 2017-02-08 02:31:30 :) 2017-02-08 02:31:42 in what state did it crash? 2017-02-08 02:36:55 Shiz: adding packages (71/90) 2017-02-08 02:37:10 Shiz: think it was qemu bug or error 2017-02-08 06:56:40 ah, there you are, before I go to bed - four hours ago: 2017-02-08 06:56:49 20:30 someone give fabled a big hug for me when he is next online. my VM crashed hard during an apk operation, and it was able to pick up the pieces and just finish the transaction when it rebooted. no other package manager I have ever used has been so easy to recover after a crash during package maintenance (dpkg, pacman, rpm, yast) 2017-02-08 06:57:54 awilfox, thanks :) 2017-02-08 07:57:02 how do we package Go stuff properly? 2017-02-08 07:57:06 see e.g. https://github.com/alpinelinux/aports/pull/725 2017-02-08 07:59:10 fabled: it has glide support 2017-02-08 07:59:25 it's glide doing the downloads? 2017-02-08 07:59:28 glide can pull in the deps 2017-02-08 07:59:41 it *can* if you use it 2017-02-08 08:03:31 fabled: we have a few packages that use it. 2017-02-08 08:04:09 also a few that build the lock file if a go project does not support versioned deps. 2017-02-08 08:25:54 fabled: seems he is also the developer of that project 2017-02-08 08:26:05 most of the go makefiles are broken. 2017-02-08 08:26:28 they depends on git to fetch tag/version 2017-02-08 08:34:36 <^7heo> yes 2017-02-08 08:35:32 i asked him to fix that makefile if possible. 2017-02-08 08:35:36 <^7heo> and worse, if you get past the vendoring problems 2017-02-08 08:36:03 <^7heo> then you have other potential issues 2017-02-08 08:36:18 <^7heo> like go generate used wrong 2017-02-08 08:37:01 <^7heo> (like in drone ci for instance) 2017-02-08 08:37:24 i tried to get that build one time, but i gave up. 2017-02-08 08:39:24 <^7heo> and finally, if nobody can build the package but the developer ("what's wrong with you guys? it works on my machine!!"), you might have no choice but to get the BINARY out of a docke image; since that will then be the preferred method of distribution 2017-02-08 08:39:39 <^7heo> (drone ci again) 2017-02-08 08:39:55 <^7heo> clandmeter: same here 2017-02-08 08:40:41 <^7heo> clandmeter: at work I took the binary out of the drone image, to stop the hemorragy 2017-02-08 08:41:35 <^7heo> s/docke/&r/ 2017-02-08 08:41:58 <^7heo> s/drone image/docker image/ 2017-02-08 08:42:24 I just wanted to try it out, but if it cant build it then good luck with it. 2017-02-08 08:42:40 <^7heo> well 2017-02-08 08:42:57 <^7heo> one you get it, it "works" indeed 2017-02-08 08:44:07 <^7heo> (I still had to figure out a couple of issues for a couple of days with broken doc to make it work) 2017-02-08 08:44:15 i use alpine, so its a noGo for me :) 2017-02-08 08:44:40 <^7heo> (and patch a file to make it work as I wished) 2017-02-08 08:45:22 <^7heo> but yeah, generally speaking, it's broken in multiple ways, at multiple points 2017-02-08 08:46:46 <^7heo> anyway, it's very far from being ideal, or an optimal software 2017-02-08 08:46:52 liskamm 2017-02-08 08:46:55 oop 2017-02-08 08:47:08 <^7heo> like many many go software 2017-02-08 08:47:56 <^7heo> which, in addition to being used permaturely while immature, are also poorly desinged 2017-02-08 08:49:09 <^7heo> to such extent that now, go is for many of us synonym with "broken hype useless pile of crap" 2017-02-08 08:52:43 <^7heo> but to close the topic; for gogs at least, it seems that the community (led by mosez on that one AFAIK) is taking one of the less bad go software around, and making it what we can hope to be a good, usable software. 2017-02-08 08:54:26 <^7heo> (my take, after my experience with terraform, is that it will be incrementally difficult to do so, as software grows; but let's wait and see) 2017-02-08 08:56:33 ^7heo: you proably mean gitea :) 2017-02-08 08:59:23 do they plan adressing the performances for very high numbers of commits/files ? 2017-02-08 08:59:36 that's the first thing they should focus on imo 2017-02-08 08:59:48 coredumb: they do 2017-02-08 08:59:57 i think you use it right? 2017-02-08 09:00:21 the alpine repo doesnt run that well on it. 2017-02-08 09:03:06 clandmeter: I use gogs yes 2017-02-08 09:03:21 haven't yet switched to gitea 2017-02-08 09:03:43 clandmeter: alpine aports repo is nothing on gogs compared to a kernel clone 2017-02-08 09:04:16 i think they already tried to fix the releases page, but from what i saw it was still not really fixed. 2017-02-08 09:04:31 coredumb: if i open the main directory it will never finish loading 2017-02-08 09:06:40 <^7heo> clandmeter: gogs, gitea - same difference 2017-02-08 09:06:47 ^7heo: btw, i added that aport with you as maintainer :) 2017-02-08 09:06:52 was a copy and pasta 2017-02-08 09:06:56 clandmeter: yea that's a pain 2017-02-08 09:06:58 <^7heo> huhu 2017-02-08 09:07:04 <^7heo> ok :) 2017-02-08 09:07:11 but i know you'll be good to it. ;-) 2017-02-08 09:07:17 <^7heo> I like pasta 2017-02-08 09:07:27 <^7heo> tss ;) 2017-02-08 09:07:28 me too, as you have seen. 2017-02-08 09:07:49 <^7heo> complimenting your way out of trouble 2017-02-08 09:08:02 I must admit that my kernel clone with several branches is also slow on disk 2017-02-08 09:08:14 <^7heo> at least that's better than the french way 2017-02-08 10:02:32 ^7heo: gitea is the stuff i'm working on. gogs is the one man show :P 2017-02-08 10:03:24 and hopefully i can continue on that when i'm not exhausted in the evening anymore :) 2017-02-08 10:28:24 mosez: ah you're part of gitea team ? 2017-02-08 10:28:27 nice to meet you 2017-02-08 10:42:50 coredumb: yes, i'm tboerger :) 2017-02-08 10:43:08 nice to meet you too 2017-02-08 10:47:01 o/ 2017-02-08 10:47:15 if you've backloged my question/comment 2017-02-08 10:47:28 and seen* 2017-02-08 10:47:50 There really needs to have performance improvements 2017-02-08 10:48:51 i thought gitea was reasonably fast? 2017-02-08 10:49:03 can't say have no tested 2017-02-08 10:49:24 but as said with clandmeter, huge repos are not usable in Gogs 2017-02-08 10:49:43 coredumb: yeah we are aware of some perfomance issues, especially with big repos. 2017-02-08 10:49:50 yep 2017-02-08 10:49:58 but we have already improved a lot 2017-02-08 10:50:19 but big repos are still performing bad. we are working on that 2017-02-08 10:50:41 if you're working on this I'm more than happy 2017-02-08 10:50:49 that's simply an architecture issue. so we need caching or we need to change the current handling. 2017-02-08 10:51:39 how big is big? 2017-02-08 10:52:03 i've got a ~350M repo that im thinking of moving to gitea 2017-02-08 10:52:05 :P 2017-02-08 10:52:14 and another ~700M one 2017-02-08 10:52:40 for example the aports repo, or kernel repo. 2017-02-08 10:52:48 repos with a lot of files and folders. 2017-02-08 10:52:59 oh okay 2017-02-08 10:53:06 that's not a problem for me then 2017-02-08 10:53:15 my repos are mostly just big on file sizes 2017-02-08 10:53:22 +history 2017-02-08 10:53:27 that shouldn't be an issue 2017-02-08 10:53:37 at least as far as i know 2017-02-08 10:53:52 Shiz: talking about a kernel clone 2017-02-08 10:53:58 with branches 2017-02-08 10:54:18 talking about ~2.5GB in size 2017-02-08 10:55:00 ~45k commits 2017-02-08 10:55:17 my current clone is at +56k commits 2017-02-08 10:55:54 my other small repos are super fine though 2017-02-08 11:01:03 i'm pretty sure we got tracking issues for that, so feel free to watch these issues. 2017-02-08 11:09:56 can i set another "source" in subpackage function? 2017-02-08 11:11:22 nvermind 2017-02-08 11:22:28 mosez: I will :) 2017-02-08 11:49:58 <^7heo> mosez: yeah sure, but technically there are very little differences between both as of now, right? 2017-02-08 11:51:29 <^7heo> Shiz: yeah by "big", gitea means "big flat collection of repos" 2017-02-08 11:51:43 <^7heo> or big flat collection of folders in a repo. 2017-02-08 11:51:55 <^7heo> or stuff alike 2017-02-08 11:52:20 <^7heo> so long story short, if you use different repos or hierarchy in your repo, things go smooth. 2017-02-08 11:52:44 <^7heo> it's made to be like github; and github has circumvented the issue by "hiding the folders 2017-02-08 11:52:47 <^7heo> " 2017-02-08 11:53:28 <^7heo> e.g with https://github.com/alpinelinux/aports/tree/master/main 2017-02-08 11:53:45 <^7heo> "Sorry, we had to truncate this directory to 1,000 files. 1,046 entries were omitted from the list. Sorry, we had to truncate this directory to 1,000 files. 1,046 entries were omitted from the list." 2017-02-08 11:53:54 <^7heo> for some reason it pasted it twice... 2017-02-08 11:54:03 <^7heo> I guess I need better software. 2017-02-08 12:02:39 ^7heo: there are differences for the upcoming release... like git lfs, 2fa, protected branches, oauth/openid consumer, just to mention some features :) 2017-02-08 12:04:04 <^7heo> 2FA \o/ 2017-02-08 12:11:45 \o/ 2017-02-08 12:13:04 and the 2fa part gets even more improved :) 2017-02-08 12:31:28 more improved, is that even possible? :) 2017-02-08 12:45:07 <^7heo> Gramatically or technically? :D 2017-02-08 12:47:16 <^7heo> https://fosdem.org/2017/schedule/event/dinosaurs/ 2017-02-08 12:47:21 <^7heo> worth watching guys 2017-02-08 12:47:24 <^7heo> ncopa: ^ 2017-02-08 12:59:55 there are different kinds of 2fa 2017-02-08 13:00:39 the goal is to support them. i'm not directly working on it and i'm not familar with all the different kinds so maybe i'm the wrong person to discuss the details. that's where it gets improved :) 2017-02-08 13:15:30 ^7heo: yeah, this was quite interesting talk; however, I’m not very sure about some statements that he made about portability of containers 2017-02-08 13:16:52 done by richard, the opensuse chairman... ok. 2017-02-08 13:18:15 btw Twitter statistics from FOSDEM https://gist.github.com/jirutka/5391d609bf129ffac742012e6305b6a4 2017-02-08 13:24:23 ^7heo: who is that guy asking questions at 36min? 2017-02-08 13:36:00 <^7heo> clandmeter: in which talk? 2017-02-08 13:36:23 the dino one 2017-02-08 13:36:23 openQA is written in Perl?! omg why? 2017-02-08 13:36:37 <^7heo> jirutka: because Suse. Because Germany. 2017-02-08 13:36:47 <^7heo> jirutka: Germans tend to like perl a lot for some reason. 2017-02-08 13:36:52 uh… 2017-02-08 13:36:56 <^7heo> jirutka: maybe it reminds them of their impossible to grasp langauge :D 2017-02-08 13:37:03 <^7heo> clandmeter: I wouldn't know, I wasn't there. 2017-02-08 13:38:22 i was refering to his voice 2017-02-08 13:38:40 he sounds a bit like stephen hawking 2017-02-08 13:38:48 heh 2017-02-08 13:39:05 hard to understand, they should provide subtitles :) 2017-02-08 13:40:42 <^7heo> this one if fantastic: https://twitter.com/pythondj/status/828176181581271044 2017-02-08 13:41:24 yeah XD 2017-02-08 13:54:18 ^7heo, jirutka: some suse dudes love perl... 2017-02-08 13:55:19 <^7heo> mosez: it's not only from Suse. 2017-02-08 13:55:35 <^7heo> mosez: I really noticed an increased amount of people liking perl in Germany, compared to everywhere else I've been. 2017-02-08 13:56:06 <^7heo> mosez: also you probably know it already, but Leider Geil by Deichkind is pretty cool :) 2017-02-08 14:03:21 ^7heo: i also know some perl lovers from sweden :P 2017-02-08 14:03:35 <^7heo> yeah some. 2017-02-08 14:03:37 they tried to convince me about the awesome perl 6 2017-02-08 14:03:39 <^7heo> there are crazy people everywhere. 2017-02-08 14:04:45 how would you abbreviate the word “assignment” (like variable assignment)? 2017-02-08 14:04:57 and no, “ass” is not an option :P 2017-02-08 14:06:28 <^7heo> assgnmnt 2017-02-08 14:06:40 <^7heo> (easy, remove the voyels - in German, it doesn't compress much) 2017-02-08 14:07:25 that’s not much shorter then the original word 2017-02-08 14:07:35 <^7heo> nope. 2017-02-08 14:07:43 <^7heo> assmnt 2017-02-08 14:07:47 <^7heo> ass-mount. 2017-02-08 14:08:48 <^7heo> can "let" work? 2017-02-08 14:08:51 <^7heo> like 2017-02-08 14:09:02 <^7heo> the assignment of X 2017-02-08 14:09:08 <^7heo> "let X be ..." 2017-02-08 14:11:09 nope, I need it as an attribute name in AST 2017-02-08 14:11:15 <^7heo> ah 2017-02-08 14:11:27 <^7heo> val? 2017-02-08 14:12:25 <^7heo> what about "assign"? 2017-02-08 14:12:39 <^7heo> or assgn 2017-02-08 14:12:43 <^7heo> if you really need shorter. 2017-02-08 14:13:12 <^7heo> (it's 50% shorter) 2017-02-08 14:41:36 http://stackoverflow.com/questions/42087734/install-php7-gd-in-alpine <- Will he be able to use pecl? I (sadly) know it usually works well. 2017-02-08 14:42:53 interesting, https://lwn.net/Articles/712528/ 2017-02-08 14:43:18 regarding to our recent x86 discussion; seems some distros are dropping i686 support sooner than i thought 2017-02-08 14:43:31 jirutka: ASG? 2017-02-08 15:21:00 <^7heo> yo fabled. 2017-02-08 15:32:19 Ganwell: php:7.1-fpm-alpine is built fom source. I believe they have some helper script to properly install the gd module 2017-02-08 15:35:35 ncopa: well, it seems better to use clean alpine linux container (alpine:3.5) and add missing modules using pecl. 2017-02-08 15:36:41 we dont have php7.1 yet though 2017-02-08 15:39:06 Ganwell: FROM php:7.1-fpm-alpine 2017-02-08 15:39:22 RUN apk add --no-cache libpng-dev && docker-php-ext-install gd 2017-02-08 15:40:50 ncopa: ok. I will edit my answer. 2017-02-08 15:42:07 i think the more correct line is like: 2017-02-08 15:42:32 RUN apk add --no-cache libpng libpng-dev && docker-php-ext-install gd && apk del libpng-dev 2017-02-08 15:43:19 he should probably also add libjpeg-turbo-dev libwebp-dev zlib-dev libxpm-dev 2017-02-08 15:43:38 and freetype-dev 2017-02-08 15:49:08 ^7heo: that talk is good 2017-02-08 15:49:28 i have thought a bit on the same problem, and i tend to agree with him 2017-02-08 15:49:39 not sure i agree that new LSB is the solution 2017-02-08 15:50:00 but he definitively have good points 2017-02-08 15:51:15 i think what we do, have a rolling release edge + stable maintained branches is a good compromise 2017-02-08 15:51:37 and when that is not good enough for you, go ahead and use your containers 2017-02-08 15:51:56 alpine works much better in containers than suse 2017-02-08 16:00:10 ncopa: https://tr.opensuse.org/MicroSUSE 2017-02-08 16:05:24 ncopa: http://stackoverflow.com/a/42100719/3729659 <- I kept using latest official Alpine Linux on top, since I think it is not a good idea to by-pass the distribution. Since it leads exactly to the mistakes this guy did. 2017-02-08 16:16:32 ScrumpyJack: I heard some rumors at fosdem that someone in redhat got the job to make a slimmed down version of fedora (or was it redhat) to be the size of alpine :) 2017-02-08 16:18:21 <^7heo> yeah 2017-02-08 16:18:24 <^7heo> Good luck with that :D 2017-02-08 16:18:44 exactly 2017-02-08 16:18:59 the glibc locale files are bigger than alpine :) 2017-02-08 16:20:08 <^7heo> yep. 2017-02-08 16:20:20 <^7heo> btw, was there some answer to that locale question on the ML? 2017-02-08 16:20:23 <^7heo> I didn't see any. 2017-02-08 16:22:57 i must have lost that email with the other 4426 unred emails... 2017-02-08 16:23:15 s/unread/unopened/ 2017-02-08 16:24:27 <^7heo> ncopa: first s/unre/unrea/; 2017-02-08 16:24:32 <^7heo> ncopa: THEN you can s/unread/unopened/ 2017-02-08 16:24:35 <^7heo> :P 2017-02-08 16:24:56 oh sorry 2017-02-08 16:24:58 yeah 2017-02-08 16:24:59 i mean 2017-02-08 16:25:24 s/unred/unread; s/unread/unopened/ 2017-02-08 16:26:21 uh; does this include emails from Aports about oudated pkgs, or do you automatically delete those? ;) 2017-02-08 16:26:33 <^7heo> ncopa: can I bitch? :D 2017-02-08 16:26:53 <^7heo> ncopa: you forgot to end the first sed command; and you could have skipped the 'unread' part ;) 2017-02-08 16:27:03 <^7heo> ACTION hides 2017-02-08 16:27:21 <^7heo> code review of IRC sed commands. Yay productivity. 2017-02-08 16:40:11 how do i get my ssh key url from github? 2017-02-08 16:42:30 <_ikke_> ncopa: https://github.com/ncopa.keys 2017-02-08 16:42:45 _ikke_: thank you! 2017-02-08 17:03:44 ncopa: Suse were here at the office last week touting microsuse. they want to ship some sort of official microsuse release Q1/2018. If redhat are rumoured to be doing the same thing, I guess they must think it has a commercial value of sorts. 2017-02-08 17:04:23 mind you, suse have a rpi build. can't see that having a commercial value. just bragging rights 2017-02-08 17:06:20 btw I’ve heard that Fedora is unusable on RPi, b/c their pkg management tool can’t fit into memory… 2017-02-08 17:08:48 lol 2017-02-08 17:29:04 isn't the fedora tool named dnf now? 2017-02-08 17:29:15 or am i misremembering 2017-02-08 17:33:18 yes, something like that 2017-02-08 17:36:02 <_ikke_> Does Not Function? :P 2017-02-08 18:06:12 <^7heo> _ikke_: that's very handy! 2017-02-08 19:34:12 dnf always meant did not finish from what I know of the acronym 2017-02-08 19:36:40 hah 2017-02-08 19:38:11 ScrumpyJack: will microsuse have a decent package manager? ;) 2017-02-08 20:06:24 kaniini: zypper is awesome. Every operation gives you an opportunity to think about your life 2017-02-08 20:07:25 <^7heo> scadu: :D 2017-02-08 20:07:35 <^7heo> scadu: plus, sometimes, it gives you things to fix! 2017-02-08 20:11:21 i can literally 2017-02-08 20:11:26 upgrade an entire alpine system 2017-02-08 20:11:34 in the time it takes dnf to start 2017-02-08 20:12:50 XD 2017-02-08 20:13:29 <^7heo> kaniini: typical. 2017-02-08 20:13:39 I tried Tumbleweed on VM. It took an 15 minutes to sync the main repo. 2017-02-08 20:13:58 Tested under three networks 2017-02-08 20:14:01 <_ikke_> auch 2017-02-08 20:14:05 ZOMG reliable 2017-02-08 20:14:30 Let's switch to -offtopic 2017-02-08 21:42:20 interesting, so Portage is not the only sloooow package manager out there 2017-02-09 01:21:39 ncopa: Just a heads up about the BIND vulnerability (just published) - sent you an email with the details. 2017-02-09 09:53:44 hmm asked in the wrong channel... 2017-02-09 09:53:48 are processes not aware of timezone changes untill they have been restarted? 2017-02-09 09:53:56 in this case crond 2017-02-09 10:09:49 clandmeter: I think they need a restart yes 2017-02-09 10:09:55 may depend of the service 2017-02-09 10:23:07 https://en.wikibooks.org/wiki/Grsecurity/Appendix/Grsecurity_and_PaX_Configuration_Options#Generate_some_entropy_during_boot_and_runtime 2017-02-09 10:23:12 do we want enable that? 2017-02-09 10:23:25 it says its not cryptographically safe 2017-02-09 10:24:06 i wouldn't do it 2017-02-09 10:27:02 good, thanks 2017-02-09 10:27:19 that got into mainline kernel 2017-02-09 10:27:24 which is intersting 2017-02-09 10:27:31 various pax features made it to mainline 2017-02-09 11:18:39 ncopa: Greg KH said they are working on adding grsec/pax feature by feature. Its even founded by the linux foundation. It seems they like it, but don't want it as a whole. 2017-02-09 11:27:09 ncopa, latent entropy might solve: http://bugs.alpinelinux.org/issues/6635 2017-02-09 11:48:28 yes 2017-02-09 11:48:37 but im not sure it is the proper solution 2017-02-09 11:48:57 there are also some wlan drivers that can provide entropy 2017-02-09 11:49:01 i disabled that too 2017-02-09 11:49:09 CONFIG_ATH9K_HWRNG 2017-02-09 11:49:30 Ganwell: yes, i think that makes sense 2017-02-09 11:49:44 WLAN_VENDOR_ATME 2017-02-09 11:49:46 WLAN_VENDOR_ATMEL 2017-02-09 11:49:54 fedora says "no" on that 2017-02-09 11:49:57 i wonder why 2017-02-09 12:00:05 another topic: should I proceed with updating python3 to 3.6.0? 2017-02-09 12:01:24 test_db with lto enabled probably will fail, but currently it is not enabled anyway 2017-02-09 12:01:41 clandmeter: indeed, the timezone is process-specific, basically if you change timezones you have to reboot :P 2017-02-09 12:01:53 scadu: i dont know what will break? 2017-02-09 12:02:01 what are the difference from 3.5 -> 3.6? 2017-02-09 12:02:20 if there are big chanec that many apps breaks, then i think we wait 2017-02-09 12:02:43 if there are only a small chance and only few apps or none might break, then i think we go ahead 2017-02-09 12:04:18 ncopa: the list of changes is long, but it seems that 3.5 -> 3.6 update went smooth in Arch 2017-02-09 12:04:27 then i think we go ahead 2017-02-09 12:04:46 ncopa: https://git.archlinux.org/svntogit/packages.git/commit/trunk?h=packages/python&id=e710cf73fbe4558cf6008534146b2e5b1b99fb9a 2017-02-09 12:05:24 scadu: i think we can upgrade 2017-02-09 12:05:37 will need rebuild the py3-* modules 2017-02-09 12:05:46 actually 2017-02-09 12:06:04 jirutka: do you have anything against upgrading python3 from 3.5 to 3.6? 2017-02-09 12:06:12 ncopa: yeah. I'll proceed once python3 build is finished 2017-02-09 12:06:18 if jirutka is ok then im ok with it 2017-02-09 12:06:56 jirutka: re python3, please ping me 2017-02-09 12:06:59 ncopa: noted. thanks. 2017-02-09 12:07:29 i think we want do that sooner or later. question is if we should do it now or wait a bit more 2017-02-09 12:07:44 my thinking is that if arch did it in december, then we can probably just go ahead 2017-02-09 12:08:10 but i want jirutkas opinion 2017-02-09 12:11:17 <^7heo> I personally would be for going from py 3.5 to 3.6. 2017-02-09 12:31:16 skarnet: thx, ill just restart the container :) 2017-02-09 13:07:28 clandmeter, a question 2017-02-09 13:07:47 what's our policy for custom-made-script shipped with a package? 2017-02-09 13:07:54 maybe for autoconfiguration 2017-02-09 13:08:06 scripts that are not on upstream 2017-02-09 13:12:44 fcolista: I dont think we have any policies. 2017-02-09 13:15:25 ncopa: scadu: don’t know about any breaking changes in Python 3.6, so it should be okay to upgrade it; just note that all packages depending on python 3 must be rebuilt 2017-02-09 13:15:38 yes 2017-02-09 13:15:56 thanks jirutka 2017-02-09 13:16:04 python 3.6 is already released? 2017-02-09 13:16:08 apparently 2017-02-09 13:16:19 huh, they released it before Christmass 2017-02-09 13:16:25 I’ve somehow missed it 2017-02-09 13:16:32 arch linux already upgraded 2017-02-09 13:16:36 so i think we can just push it 2017-02-09 13:16:48 scadu: please go ahead 2017-02-09 13:17:04 ncopa: build is in progress 2017-02-09 13:17:11 nautilus i ~/dev/alpine/aports grep -r "makedepends=.*python3" | wc -l 2017-02-09 13:17:14 239 2017-02-09 13:17:16 woohoo 2017-02-09 13:17:29 fcolista: you could try to write it in shell so you dont need to pull in extra deps. 2017-02-09 13:17:32 ha 2017-02-09 13:17:50 scadu: just rebuild pkgs that depends (in runtime) on python3 2017-02-09 13:18:07 nautilus i ~/dev/alpine/aports grep -r "^depends=.*python3" | wc -l 2017-02-09 13:18:10 23 2017-02-09 13:18:11 jirutka: that's better :P 2017-02-09 13:18:15 this is wrong 2017-02-09 13:18:18 :) 2017-02-09 13:18:26 check that module path didnt change 2017-02-09 13:18:48 depends= does not contain dynamically detected dependency on python3 2017-02-09 13:19:04 and libpython if such thing exists 2017-02-09 13:19:09 libpython.so 2017-02-09 13:19:23 yeah 2017-02-09 13:20:47 jirutka: you're right, sorry. 2017-02-09 13:21:36 fabled: how it looks with new apk? we really need proper support for querying dependency tree :( 2017-02-09 13:22:08 scadu: give me a sec 2017-02-09 13:22:59 jirutka: I don't push anything right now. I'll wait for finishing tests and in the meanwhile I'm looking at pkgs.a.o ;v 2017-02-09 13:23:43 clandmeter, yes, it's shell 2017-02-09 13:24:10 jirutka, plan done, code started, but still quite a bit journey to go 2017-02-09 13:24:31 isnt that already possible? compare so: provides with so: depends? 2017-02-09 13:24:43 fabled: do you have any documentation/notes about planned changes, new features etc? 2017-02-09 13:24:51 i have a rough overview 2017-02-09 13:24:59 i posted it here earlier at some point 2017-02-09 13:25:18 jirutka, http://sprunge.us/biJV 2017-02-09 13:27:41 clandmeter: hmm, yeah, this is already supported, apk search -o -r python3… IIRC some search commands takes into account only installed pkgs 2017-02-09 13:32:42 scadu: this list should be hopefully complete http://paste.fedoraproject.org/551668/66471481/raw/ 2017-02-09 13:36:23 scadu: I’ve generated it with: apk search -x -o -r python3 python3-tkinter so:libpython3.so | sed -En 's/^(.*)-\d+[^-]*-r\d+.*/\1/p' | sort | uniq 2017-02-09 13:37:13 jirutka: thanks. will send a pr today evening. hope it's okay 2017-02-09 13:39:15 scadu: correction, I forgot to apk update: http://paste.fedoraproject.org/551675/47539148/raw/ 2017-02-09 13:41:44 jirutka: thanks 2017-02-09 13:41:55 aha, so `apk info -r python3` takes into account only installed pkgs 2017-02-09 13:42:05 :s 2017-02-09 13:43:09 jirutka: what does that sed do? 2017-02-09 13:43:35 clandmeter: removes version number 2017-02-09 13:44:10 what does -q do? 2017-02-09 13:44:20 q? 2017-02-09 13:44:42 try add it :) 2017-02-09 13:45:02 add where? 2017-02-09 13:45:09 apk search 2017-02-09 13:45:28 ahaa 2017-02-09 13:45:43 that’s totally not what I would expect from general --quiet option :) 2017-02-09 13:46:28 glad to help ;-) 2017-02-09 13:48:17 :) 2017-02-09 13:50:45 well, so apk already allows to query direct and reverse dependencies, just the latter is in search command and not info :) I was confused that `apk info -R PKGNAME` lists all the pkg dependencies, but `apk info -r PKGNAME` lists only installed reverse dependencies, not all 2017-02-09 13:53:27 yeah, it's slightly confusing 2017-02-09 13:53:35 and the flags are inconsistent 2017-02-09 13:53:39 due to historical reasons 2017-02-09 13:53:42 yeah 2017-02-09 13:53:56 info was supposed to interrogate installed-db only, search works on the indexes 2017-02-09 13:54:00 we should clean that up in apk3 2017-02-09 13:54:03 yes 2017-02-09 13:54:03 so now I miss only querying build dependencies and printing complete dependency tree of a pkg 2017-02-09 13:54:16 build dependencies are not included anywhere 2017-02-09 13:54:26 we probably should add that 2017-02-09 13:54:28 yes 2017-02-09 13:54:44 i wonder if we should introduce `apk query ....` or similar 2017-02-09 13:54:55 nuke info and search 2017-02-09 13:54:58 and replace with query 2017-02-09 13:55:00 in apk3 2017-02-09 13:55:06 something like that 2017-02-09 13:55:39 i'm working on quagga and libreoffice java support 2017-02-09 13:55:44 will push soon 2017-02-09 13:55:51 java in libreoffice? 2017-02-09 13:55:59 I hate open-space… I work in small open-space at Thursday and I really can’t focus there 2017-02-09 13:56:16 there’s some PR for java support into libreoffice 2017-02-09 13:56:20 yes 2017-02-09 13:56:21 ok 2017-02-09 13:56:35 jirutka, i'm using the good bits from it, but reworking it a bit 2017-02-09 13:56:37 jirutka: you need noise cancellation headphones 2017-02-09 13:56:42 https://github.com/alpinelinux/aports/pull/552/ 2017-02-09 13:58:07 fabled: btw when you have some free time, it’d be perfect if you can look at https://github.com/alpinelinux/aports/pull/205; mitchty did excellent work with pkgs for Haskell, but the patch is quite huge and I still didn’t have enough time to properly review and test it 2017-02-09 13:58:24 i glanced at it, but the bootstrap procedure is yucky 2017-02-09 13:58:30 ncopa: what he needs is https://www.alpine.nl/ 2017-02-09 13:58:31 i'm not sure how to do that 2017-02-09 13:58:50 you can fix everything with Alpine! 2017-02-09 13:58:59 lol 2017-02-09 13:59:02 i should probably talk to mitchty on it 2017-02-09 13:59:02 XD 2017-02-09 13:59:37 okay, I’ll try to fix it with a coffee and then maybe reallocate myself into Silent room; no big monitor here, but at least no noise 2017-02-09 14:02:10 lol https://www.alpine.nl/gehoorbescherming/mutt-muffs/ 2017-02-09 14:06:09 lawl 2017-02-09 14:18:46 http://www.alpine.cz is already taken :( 2017-02-09 14:19:08 but alpinelinux.cz is available :) 2017-02-09 15:22:48 I’ve just installed mariadb on v3.5.1 and /etc/mysql/my.cnf is owned by user with UID 60 that is not in /etc/passwd, instead of user mysql 2017-02-09 15:23:36 abuild looks okay 2017-02-09 15:23:59 kaniini, evolution is at version 3.22.4. evolution-data-server has a suid. Would options="suid" be enoguh? 2017-02-09 15:24:04 there’s pkgusers, pkggroups and pre-install script that creates mysql user/group 2017-02-09 15:24:38 I mean: should we remove the suid, or we should allow it? 2017-02-09 15:26:56 http://sprunge.us/gLhf 2017-02-09 16:05:42 what's the mail repository? The one containing the packages just built ? 2017-02-09 16:05:44 *main 2017-02-09 16:33:51 lawl 2017-02-09 16:34:07 ffu, wrong buffer ;s 2017-02-09 16:41:50 I thought that it was an accident, but it seems that’s not; after apk upgrade, /etc/shadow changed mode from 0640 to 0644 and group from 42 to 0 2017-02-09 16:43:51 ouch. 2017-02-09 17:56:59 maybe i should snag alpinelinux.nl 2017-02-09 18:30:22 https://github.com/alpinelinux/aports/blob/master/main/python3/APKBUILD#L62-L63 anyone knows why it's been done? 2017-02-09 18:50:04 fabled: do we have JavaFX support in openjdk pkg? 2017-02-09 18:53:23 jirutka, dunno, we do icedtea default build 2017-02-09 18:54:34 fabled: I know, but not sure if javafx is enabled by default or no 2017-02-09 18:54:36 t 2017-02-09 18:55:11 to me it looks it's a separate add-on package 2017-02-09 18:55:57 https://wiki.openjdk.java.net/display/OpenJFX/Building+OpenJFX 2017-02-09 21:03:02 jirutka: no worries on the review, that whole port has almost 2 years of effort behind it, I'd just like to get it upstreamed so I can move onto using it more than beating on it :) 2017-02-09 21:08:25 i get that this is all volunteer etc... i'm just good at being a weekly squeaky wheel to make sure i'm not forgotten >.< 2017-02-09 21:26:49 fcolista: i don't have much interest in the gnome packaging anymore :) 2017-02-09 21:26:55 fcolista: so do what you think is best 2017-02-10 07:42:41 chromium crashed @edge 2017-02-10 07:42:46 *crashes 2017-02-10 08:34:32 mitchty, i'm looking at the ghc/cabal/stack thing 2017-02-10 08:34:51 seems ghc is now self-dependant, but supports cross-compiling 2017-02-10 08:35:59 i would like to have ghc from the 'bootstrap' build for x86_64, and use it to build the first native ghc on x86_64; the rest i would like to bootstrap using cross-compilation 2017-02-10 08:36:24 so i'd like the ghc APKBUILD to support cross-compiling / scripts/bootstrap.sh 2017-02-10 09:14:56 fcolista: you need to manually paxmark it 2017-02-10 09:15:36 oh 2017-02-10 09:15:53 how should i figured that? 2017-02-10 09:15:59 its a bug 2017-02-10 09:16:16 ok.. 2017-02-10 09:16:20 we dropped the paxctl -c in paxmark wrapper script 2017-02-10 09:16:27 and rely on xattrs 2017-02-10 09:16:44 but it didnt work as expected under fakeroot 2017-02-10 09:16:53 i think we have fixed it though 2017-02-10 09:17:01 so it only needs rebuild 2017-02-10 09:17:03 meanwhile 2017-02-10 09:17:37 setfattr -n pax.flags -v m /usr/lib/chromium/chrome 2017-02-10 09:18:56 or: attr -q -s pax.flags -V "${xval}" "${f}" 2017-02-10 09:19:21 attr -s pax.flags -V "m" /usr/lib/chromium/chrome 2017-02-10 09:19:53 or: apk add paxmark && paxmark m /usr/lib/chromium/chrome 2017-02-10 09:21:28 ncopa, i've ran the latter and worked 2017-02-10 09:21:30 thx a lot 2017-02-10 09:21:56 it will be fixed next rebuild i think 2017-02-10 09:22:05 same issue with firefox 2017-02-10 11:33:21 finally i have configured the kernel 2017-02-10 11:34:41 how many hours did that cost? 2017-02-10 11:34:54 x86_64 grsec virtgrsec, x86 grsec virtgrsec and armhf grsec 2017-02-10 11:35:18 not sure 2017-02-10 11:35:21 a day or so 2017-02-10 11:36:08 now i need to fix the 3rd party modules :-/ 2017-02-10 11:43:10 <^7heo> moin 2017-02-10 12:48:41 hello, opensmtpd crashes on v3.5. this bug is known https://archives.gentoo.org/gentoo-commits/message/fadd1831df77fad6f72ddda3bfb77a32 2017-02-10 12:50:52 xentec: can you please check if its reported on bugs.alpinelinux.org, and if it is, please add that link 2017-02-10 12:51:02 or create a bug for it 2017-02-10 12:51:12 i dont have time to look at it now, but i want fix it 2017-02-10 12:51:22 set target to 3.5.1 2017-02-10 12:51:25 3.5.2 2017-02-10 12:51:31 oh, forgot there is a bug tracker for it, sorry 2017-02-10 12:51:58 pinging us here in IRC about bugs it totally ok 2017-02-10 12:52:10 but we will forget it unless we get it into the bugtracker 2017-02-10 12:54:11 <^7heo> +1 2017-02-10 12:55:29 ncopa, http://bugs.alpinelinux.org/issues/6578 2017-02-10 12:56:16 xentec: thank you! 2017-02-10 12:56:43 thank you for that 2017-02-10 12:56:47 should be easy to fix now 2017-02-10 12:58:20 i think the linux-grsec-dev package is broken with 4.9.x 2017-02-10 12:58:21 actually i should thank _you_. first for alpine, second for this issue about ranger I just found on top of the tracker: http://bugs.alpinelinux.org/issues/6839 2017-02-10 12:58:51 (because you reminded me of the tracker) ;) 2017-02-10 12:58:59 finally ranger on alpine 2017-02-10 12:59:00 :) 2017-02-10 12:59:26 that ranger thingy looks like a real upstream bug 2017-02-10 12:59:43 either in ranger, python or ncurses 2017-02-10 13:01:42 yeah, before _ikke_, I took a look at it too and for me it seemed that the bug is in python. not the ungetmouse() only but maybe the whole stack handling in this case. so I baked of as this was too scary 2017-02-10 14:22:32 xentec: is this related? https://github.com/OpenSMTPD/OpenSMTPD/issues/765 2017-02-10 14:22:43 yes 2017-02-10 14:22:57 i guess same behaviour on edge? 2017-02-10 14:23:29 not sure if i should just patch it or upgrade it on edge. 2017-02-10 14:23:49 uuh well I take it back. can't say for sure. crash is happening because of stack overflow (recursion out of hand) 2017-02-10 14:24:07 but cant say that both issues have the same cause 2017-02-10 14:25:31 did you verify if that patch actually fixes your issue? 2017-02-10 14:26:18 didn't have time yet, but it patches the exact functions that blow up the stack. 2017-02-10 14:28:12 also running smtpd as shown in github issue doesn't produce any debug messages, but segfaults instantly 2017-02-10 14:29:44 ok ill just patch both edge and stable and we will see if it fixes things. 2017-02-10 14:32:04 also you should mind to update it to 6.0 on edge some day ;) 2017-02-10 14:54:08 xentec: can you try stable? 2017-02-10 14:54:41 tried it with patch, still fails. I'm already debugging it 2017-02-10 14:55:22 over 100k stack frames in gdb not not easy to pass by 2017-02-10 14:55:27 ill leave it to the maintainer to update it to 6x 2017-02-10 14:55:40 but i prefer to have a working copy in stable. 2017-02-10 15:00:37 is there a fast way to tell abuild to create an debug build? 2017-02-10 15:04:08 ah its DEBUG env var. nevermind 2017-02-10 19:54:18 would it be alright for a package in stable to lose a dependency through a bug fix? 2017-02-11 10:11:15 so, according to the list provided by jirutka there is 192 packages that needs rebuild after python upgrade to 3.6 2017-02-11 10:11:29 148 of them are python modules. 2017-02-11 10:13:37 seems like much fun ;f 2017-02-11 10:14:10 there was some discussion about pip handler in apk, but it was refused, right? 2017-02-11 10:15:48 anyway, I'm not sure what should I do. I currently test 3.6.0 and going to bump pkgrel of all those packages, but… 2017-02-11 10:17:09 should I send this as two pull requests (python3 upgrade to 3.6.0 and rebuild of pkgs that needs it) in 193 commits? :f 2017-02-11 10:18:52 jirutka: ^ 2017-02-11 12:45:59 hi folks 2017-02-11 12:48:18 Thinking about Alpine linux as a home server instead of Gentoo that I have been using for a while. What's the security approach with Alpine Linux, is there anyone who follows CVEs and takes care of security updates? 2017-02-11 12:48:48 I know Gentoo has some security folks, Fedora is mostly covered by Red Hat folks, Debian has some mechanisms... 2017-02-11 13:04:34 Wiki doesn't seem to be very verbose in that respect. 2017-02-11 13:31:00 pavlix: my experience is that generally CVEs get fixed rather promptly 2017-02-11 13:31:42 in addition, the grsec kernel prevents many of them being exploited in the first place 2017-02-11 13:39:58 kunkku: Just curious, why is than RPi image without grsec (according to the download page)? 2017-02-11 14:27:45 Just sucessfully used SysRq to recover from some bad X condition. 2017-02-11 14:50:34 Would it make sense to add Fedora to the comparison at https://wiki.alpinelinux.org/wiki/Comparison_with_other_distros ? 2017-02-11 15:00:47 Btw `apk upgrade` from 3.3 to 3.5 didn't work out, apk just hang and as do many busybox-based tools, kill -9 doesn't take the processes down, all is broken in a way. 2017-02-11 15:01:17 I'm not going to debug it any further, though, it's a test installation that I will simply replace with a fresh install. 2017-02-11 15:10:45 pavlix: which steps did you do to upgrade? 2017-02-11 15:21:23 clandmeter: modification of /etc/apt/repositories; apk update; apk upgrade -> didn't complain, did some weird things, sort of worked (while not upgraded) after reboot 2017-02-11 15:21:45 clandmeter: now I hit the surprise that the ISO doesn't include wpa_supplicant... :/ 2017-02-11 15:21:56 pavlix: don't forget -a on upgrade 2017-02-11 15:22:22 some ppl forget that add -a :) 2017-02-11 15:22:31 oh yeah I did 2017-02-11 15:22:33 s/that/to 2017-02-11 15:22:51 still apk upgrade shouldn't freeze 2017-02-11 15:22:52 pavlix: hope you mean /etc/apk/repositories 2017-02-11 15:22:54 ;p 2017-02-11 15:23:13 clandmeter: shouldn't it try to check and complain, by the way? 2017-02-11 15:23:18 Shiz: yep :D 2017-02-11 15:23:25 Shiz: mechanical memory does its job 2017-02-11 15:29:59 ncopa, clandmeter: http://bugs.alpinelinux.org/issues/6578#note-5 2017-02-11 15:33:46 local mail delivery tested on stable and edge 2017-02-11 16:00:28 https://wiki.alpinelinux.org/wiki/Connecting_to_a_wireless_access_point looks wrong 2017-02-11 16:00:55 You don't need to set SSID manually when using wpa_supplicant. 2017-02-11 16:33:53 wiki.alpinelinux.org: This action has been automatically identified as harmful, and therefore disallowed. If you believe your action was constructive, please inform an administrator of what you were trying to do. A brief description of the abuse rule which your action matched is: 001 2017-02-11 16:34:00 That doesn't sound very constructive. 2017-02-11 16:52:36 wiki considered harmful 2017-02-11 17:02:10 Shiz: Is anyone actually editing the wiki? 2017-02-11 17:40:49 jirutka: can you bump pkgrel for cargo? it still depends on libgit2.so.24 2017-02-11 17:42:58 jirutka: thanks! ultra fast as always :) 2017-02-12 12:08:24 Any solutions to the wiki problem? 2017-02-12 15:40:42 pavlix: I guess you'll have more answers tomorrow 2017-02-12 16:14:53 who decided that it would be a good idea to move pod2man and other perl utils to perl-dev? This breaks the man pages for quite a few packages gdb for instance 2017-02-12 16:15:41 and why was this change even made? I thought that -dev subpackages should not contain any binaries... 2017-02-12 16:17:11 here is the commit which introduced that change http://git.alpinelinux.org/cgit/aports/commit/main/perl/APKBUILD?id=d7bc7e1589be482ad883af78832d6845cc889d06 2017-02-12 16:23:01 if you make such a change you at least need to update the makedepends of all packages using pod2man for man pages 2017-02-12 16:23:24 ncopa: why was it neccessary to move pod2man to the -dev package? 2017-02-12 16:29:26 coredumb: more answers, less time :) 2017-02-12 16:41:33 nmeum: -dev subpackages can contain binaries just fine 2017-02-12 16:41:41 see: mysql-dev containing mysql-config 2017-02-12 16:41:47 as long as they're development-related binaries 2017-02-12 16:42:44 I would argue that pod2man isn't a development-related binary and even if it is you definitly need to adjust the makedepends of existing packages if you make such a change 2017-02-12 16:43:37 i disagree with the former and agree with the latter 2017-02-12 16:43:39 :p 2017-02-12 16:45:20 if we decided to stick with that change we need to update all packages using pod2man in order to do that you first of all need to find all packages that use pod2man and doing so isn't easy 2017-02-12 16:58:11 grep is hard 2017-02-12 17:00:37 grep doesn't work in that case 2017-02-12 17:00:56 how so? 2017-02-12 17:01:29 what pattern should grep search for? 2017-02-12 17:01:55 pod2man isn't a single binary/script? 2017-02-12 17:02:04 it is but it is included in the perl package 2017-02-12 17:02:10 and aports using it simply depend on perl 2017-02-12 17:02:14 and now need to depend on perl-dev 2017-02-12 17:03:47 however, there are packages which depend on perl but don't need pod2man 2017-02-12 17:04:09 for each package depending on perl: if grep pod2man, then add a dep to perl-dev; if not grep perl, then remove the dep to perl 2017-02-12 17:05:26 and you want to grep the entire package source tarball for pod2man or on which files would you perform 'grep pod2man'? 2017-02-12 17:06:29 grep the entire source just to be sure. It's machine time, who cares if you get the result in a week. 2017-02-12 17:07:00 it's not subtle, but it's not _hard_. :P 2017-02-12 17:07:50 well I don't have that amount of machine time :p 2017-02-12 17:52:20 skarnet: By the way found one case of maybe wrong dependency+ordering configuration in Alpine/OpenRC where it didn't want to start SSH for me because it couldn't start dhcpcd because it was already started manually outside the service manager. 2017-02-12 17:53:05 skarnet: That means it wanted to pull in dhcpcd, which is, according to my requirements, wrong, as I didn't want to start that tool as I have already solved the network configuration in a different way. 2017-02-12 17:53:51 skarnet: In this case I would basically want to set up ordering of those tools at boot time but not a hard dependency (Requires=) and maybe not even a soft dependency (Wants=). 2017-02-12 17:54:22 skarnet: Just ordering of the tools that I myself configured to run. 2017-02-12 17:54:49 I guess that's what you meant by *not* needing soft deps and non-dep ordering. 2017-02-12 17:59:22 Not really. In that case that's your own fault, and you'll have the same problem with any service manager: you're doing things behind the service manager's back, so obviously it doesn't understand what's going on and fails at some point. 2017-02-12 18:00:01 it is true, however, that there's no reason why ssh should depend on dhcpcd. 2017-02-12 18:00:55 skarnet: “In that case that's your own fault” – That's a very lennartine argument, even though this case is handled by systemd gracefully. 2017-02-12 18:00:57 so the problem was already there but hidden because dhcpcd usually succeeds, and your tinkering exposed it. 2017-02-12 18:01:20 sometimes even Lennart says the truth ;) 2017-02-12 18:01:40 I mean, you are obviously free to start whatever stuff you want at boot time 2017-02-12 18:02:02 On one hand, SSH should not depend on dhcpcd, on the other hand, at boot time, network servers are generally meant to be started *after* network is configured. 2017-02-12 18:02:05 but if what you're doing conflicts with what the service manager thinks it should be doing, it's normal that something breaks 2017-02-12 18:03:41 skarnet: Therefore in systemd I would use After=network-online.target Wants=network-online.target... and the network-online waiting mechanism gets pulled in and it waits for all services that order themselves before it, like NetworkManager-wait-online.servce that waits for NetworkManager configuration *if* NetworkManager service is run. 2017-02-12 18:04:19 what happens with systemd if you're doing the same, i.e. doing your own thing without systemd knowing? 2017-02-12 18:04:26 skarnet: And that's, in my opinion, a brilliant solution of the situation. And I can hardly imagine OpenRC or s6 doing just as well, even though OpenRC has some mechanisms. 2017-02-12 18:05:30 skarnet: That's not the question, I *can* actually configure systemd in a way that SSH doesn't pull in any particular network configuration service but waits for any if it's already configured to run. 2017-02-12 18:06:25 still, the service you're actually using to bring the network up is managed by systemd, isn't it? 2017-02-12 18:06:30 skarnet: Therefore even if I did something behind the scenes, NetworkManager would not be pulled in when I just want to start SSH. 2017-02-12 18:07:36 skarnet: If I want correct order at boot time, yes, I need a network configuration service scheduled for boot time. If we're talking about explicit start at runtime, then SSH doesn't pull in any dependencies and I can do whatever I want. 2017-02-12 18:08:31 I don't understand 1. what you want to accomplish, 2. what prevents you in OpenRC or s6-rc to accomplish it, 3. what point you're trying to make. 2017-02-12 18:08:34 skarnet: Boot time is a non-issue, though, as the startup is not interactive and therefore you don't have to ask whether I did something manually at boot time; I didn't. 2017-02-12 18:08:52 from accomplishing* 2017-02-12 18:10:03 1) I'm starting SSH at generic runtime and I believe it shouldn't pull in any dependencies. On the other hand, at boot time, it should *wait* for network configuration when it's already scheduled to start, especially at boot time. 2017-02-12 18:12:33 2) You implied last time, that non-dependency ordering is not needed. I'm making use of it with systemd to do exactly what I described in #1. I'm not aware of adequate tools with OpenRC and s6, although OpenRC does have some notion as started+inactive service and Gentoo uses it for NetworkManager with OpenRC. 2017-02-12 18:13:29 3) I'm questioning the claim that there's no good use case for non-dependency ordering and similar stuff and looking for more information to update my opinion. :) 2017-02-12 18:17:37 My pov is that "network config" should be a (virtual) service that can be implemented in several ways, and sshd should depend on it. 2017-02-12 18:17:50 Once your network config is done, you can start sshd at any time. 2017-02-12 18:18:31 Since the "network config" service is already up (it's a oneshot, once it's been run once it's considered up until you explicitly bring it down), sshd will start without pulling anything else. 2017-02-12 18:20:27 dinner time, bbl. 2017-02-12 18:25:29 Redmine SPAM <- http://bugs.alpinelinux.org/issues/6847 2017-02-12 18:26:08 skarnet: My POV is that it shouldn't becuase running SSH on localhost is a perfectly valid use case. 2017-02-12 18:27:09 skarnet: Also for experimenters with network configuration and power users doing network related jobs, it's much better to allow the user to handle networking in any way (or no way) and still run network services as they wish. 2017-02-12 20:08:34 pavlix: when I say "network config", that includes localhost, there's no reason to separate it from the other network interfaces - and if it's an issue because you have a dhcp client for instance, you can have "localhost" as a separate service and have sshd depend on it, that's not an issue. (The normal way of running sshd is on 0.0.0.0 anyway, so it doesn't matter if you configure interfaces later 2017-02-12 20:09:51 and experimenters/power users can do whatever they want, I'm only talking about things that are handled by the service manager. Either you integrate your stuff with it or you don't - it's a choice, but mixing the two will always cause problems no matter which service manager you use. 2017-02-12 20:28:08 skarnet: Sure. I'm not saying that limited operating system features are always a problem. Just showing some use cases that are handled more gracefully with tools that don't enforce specific ways to do things. 2017-02-12 20:29:22 I refute your claim that systemd is more flexible than s6-rc. Maybe it is, but the example you have chosen doesn't show it. 2017-02-12 20:30:54 >And I can hardly imagine OpenRC or s6 doing just as well 2017-02-12 20:31:01 depend() { after network } ? :p 2017-02-12 20:32:10 skarnet: I do not see the refutal, though, as everything I said about the case still holds. 2017-02-12 20:32:48 Shiz: That's probably not enough information to show that. 2017-02-12 20:35:10 I thought I was clear. sshd binds to 0.0.0.0. Depends on nothing. Problem solved. 2017-02-12 20:35:24 nmeum: Shiz: to that change with pod2man… it seems that this change is harmful, so it’d be better to revert that commit, WDYT? 2017-02-12 20:35:50 personally disagree, i just think the makedepends should be fixed 2017-02-12 20:36:04 the point is fixing the makedepends is not that easy 2017-02-12 20:36:22 i believe skarnet outlined that it is in fact pretty easy 2017-02-12 20:36:22 jirutka: I usually don't revert commits without speaking to the guy who comitted it first 2017-02-12 20:36:23 :P 2017-02-12 20:36:30 Shiz: and how do you propose to do that…? I mean how to find out what pkgs really needs pod2man? 2017-02-12 20:36:40 see above 2017-02-12 20:36:47 skarnet suggested grepping through the source of all packages 2017-02-12 20:36:55 find packages that depend on perl 2017-02-12 20:37:01 grep through the source for pod2man 2017-02-12 20:37:01 that’s nothing but insane 2017-02-12 20:37:35 you have powerful computers with fast disks to do the work for you, but sure, using them is insane 2017-02-12 20:37:53 it only needs to be done once... 2017-02-12 20:38:25 no, you’re talking about grepping whole source code of hundreds perl packages 2017-02-12 20:38:31 why use computers when you can spend human time pretending there's no solution 2017-02-12 20:38:59 it’s not that easy, pod2man mean appear just in comments or docs 2017-02-12 20:39:33 so what ? it's crude, flag them for inspection then look at them with human eyes 2017-02-12 20:39:37 besides your rockhammer-method might generate quite a few false-positives so a human would need to filter the results manually 2017-02-12 20:39:47 of course it might 2017-02-12 20:39:55 I personally wouldn’t like to do that… so let’s ask Valery Kartel, who did this injudicious change, to fix it, or revert 2017-02-12 20:40:18 it's still better than anything else that has been suggested, i.e. exactly nothing 2017-02-12 20:40:24 jirutka: that seems to be the best solution 2017-02-12 20:40:39 skarnet: fine, if you consider this as a good solution, then please, do it… 2017-02-12 20:40:57 I would if I gave a shit 2017-02-12 20:41:27 I have more important things to do than grepping perl sources… just b/c someone pushed injudicious change that breaks unknown number of packages 2017-02-12 20:42:35 nmeum: could you please write him an email? valery.kartel@gmail.com, and give ncopa@alpinelinux.org to Cc (he approved this patch), and maybe even me (i’m just curious) 2017-02-12 22:00:26 an alternative solution for now would be to have pod2man install_if on perl 2017-02-12 22:00:34 thusly fixing $makedepends for the short term 2017-02-12 22:05:59 jirutka: sure, will do so 2017-02-12 22:07:10 BTW what does it take for a package to make it from testing to a main repo? 2017-02-12 22:11:20 ashb: someone must acknowledge that it’s working and it must have a maintainer 2017-02-12 22:11:52 Right okay. Once I've tested and used the exim package for a couple of weeks I'l step up for that 2017-02-12 22:12:19 ashb: that’s requirement for community, most pkgs should end up in community 2017-02-12 22:12:37 I assume maintainer involves keeping up to date with releases, and maybe looking at any proposed changes? 2017-02-12 22:12:42 ashb: yes 2017-02-12 22:13:56 ashb: pkgs in community should be supported at least for 6 months since released in Alpine stable (i.e. security fixes etc.) 2017-02-12 22:15:20 ashb: pkgs in main should be supported for at least 2 years; I don’t know what are current rules for moving pkgs into main 2017-02-12 22:16:10 when you say supported: there's no backporting of fixes needed right? It's just releasing appropriate point-release or similar version from upstream? 2017-02-12 22:17:21 ashb: that depends… 2017-02-12 22:19:04 ashb: when pkg is included in stable release (e.g. v3.5), then in that branch it should not be bumped to a version with breaking changes 2017-02-12 22:20:27 Ah okay, so occasionally carrying a patch if upstream only releases a fix for a newer version that breaks back-compat with the shipped version. But if it's just introduced new features it's okay to just update to latest point release? 2017-02-12 22:20:37 (give or take some "it depends") 2017-02-12 22:21:02 yes 2017-02-12 23:13:06 why is the address sanitizer in gcc disabled? also why is ssp disabled if i do 'gcc -v' ? 2017-02-12 23:17:56 oh, musl doesn't support asan. i see 2017-02-12 23:18:29 ah, reading the gcc APK file, all is clear 2017-02-13 07:57:51 Morning. Happy Monday! 2017-02-13 10:48:55 seems like config_stack_validation requires elfutils 2017-02-13 12:12:16 fyi, i'm currently working on the haskell PR 2017-02-13 12:12:25 the ghc-bootstrap is done, and i'm doing native build now 2017-02-13 12:13:12 mitchty has done good work with it 2017-02-13 12:13:50 if all builds ok, i'll push x86_64 ghc today 2017-02-13 12:14:13 strangely, i'm getting lot of "ghc-stage1: unable to decommit memory: Invalid argument" during first native self-build 2017-02-13 12:18:34 madvise(0x738ec8900000, 3145728, MADV_FREE) = -1 EINVAL (Invalid argument) 2017-02-13 12:18:45 ncopa, http://bugs.alpinelinux.org/issues/6578#note-5 2017-02-13 12:19:21 oh 2017-02-13 12:19:34 MADV_FREE was added linux 4.5, and the build is on linux 4.4.y 2017-02-13 12:21:08 https://github.com/NixOS/nixpkgs/pull/18196/commits/54125b4af4e0788f1f72580aa4dbda7f57db0ff7 ? 2017-02-13 12:30:07 ah, upstream commit https://git.haskell.org/ghc.git/commitdiff/6576bf83cdf4eac05eb88a24aa934a736c91e3da 2017-02-13 12:30:35 By the way did anyone look at the wiki editing issue? (ncopa, ...) 2017-02-13 12:33:57 pavlix: which issue? 2017-02-13 12:35:09 clandmeter: I can't edit wiki pages. Some antispam feature that goes to far, apparently. 2017-02-13 12:35:30 I was told that on Monday there's a bigger chance to get answers. :) 2017-02-13 12:43:23 pavlix: which is the exact error? 2017-02-13 13:06:40 pavlix: which username? 2017-02-13 13:06:55 ncopa: i think its the same issue as last week 2017-02-13 13:07:06 i had to manually activate a user to let him edit. 2017-02-13 13:44:00 ncopa, i think i'm ready to push ghc, any objections for that? 2017-02-13 13:44:11 i'll pre-install the bootstrap ghc manually first, and then push the ghc aport 2017-02-13 13:45:38 fabled: im ok to push it 2017-02-13 13:53:52 clandmeter: Username is pavlix, like here. 2017-02-13 13:53:56 ncopa, are you fine to accept https://github.com/alpinelinux/aports/pull/839 2017-02-13 13:54:57 clandmeter: And if the user isn't allowed to *save*, he shouldn't be allowed to *edit* so that he doesn't start creating content that can't be saved. Sure, in my case, the new content is in a text file on my disk, now. Just surprised by the misbehavior of mediawiki. 2017-02-13 14:05:26 pavlix: which error did you get? 2017-02-13 14:09:23 clandmeter: I don't remember, but I can try making a change if it's needed to get that fixed. 2017-02-13 14:09:38 (Or maybe I find it in IRC history.) 2017-02-13 14:09:59 wiki.alpinelinux.org: This action has been automatically identified as harmful, and therefore disallowed. If you believe your action was constructive, please inform an administrator of what you were trying to do. A brief description of the abuse rule which your action matched is: 001 2017-02-13 14:10:10 clandmeter: ^ it was in the channel backlog 2017-02-13 14:21:39 andypost: i would prefer get rid of the 71 prefix if possible 2017-02-13 14:22:58 ncopa, but this prevents collisions for the first time 2017-02-13 14:24:03 ncopa: who is maintaining wiki? nangel? 2017-02-13 14:24:23 andypost: you mean for upgrade? 2017-02-13 14:24:34 clandmeter: yes its nangel 2017-02-13 14:24:43 send him email 2017-02-13 14:25:11 ncopa, yep 2017-02-13 14:25:57 andypost: if you rename the package to testing/php7 then apk will think its the same thing, juste a newer version 2017-02-13 14:26:02 which is exactly what it is 2017-02-13 14:26:21 and apk will say :"upgrading from 7.0 to 7.1.x" 2017-02-13 14:28:45 ncopa, but jirutka said there's no way to store same named packages 2017-02-13 14:29:11 it works just fine 2017-02-13 14:29:21 you can have same package name with different version in different repositories 2017-02-13 14:29:27 we just have them in different repo 2017-02-13 14:29:29 andypost: I didn’t said that, I just said that it’s confusing 2017-02-13 14:30:06 and it leads just to problems and inconsistences 2017-02-13 14:30:40 $ abuild-apk policy ghc 2017-02-13 14:30:40 ghc policy: 2017-02-13 14:30:40 8.0.2-r0: 2017-02-13 14:30:40 @testing http://rsync.alpinelinux.org/alpine/edge/testing 2017-02-13 14:30:41 ah, sorry misunderstood 2017-02-13 14:31:17 i dont want keep the php 7.0 package forever 2017-02-13 14:31:22 i want replace it 2017-02-13 14:31:29 ^ +1 2017-02-13 14:32:29 i think that adding testing/php7 instead of php71 will make that easier 2017-02-13 14:33:19 hm, that’s true 2017-02-13 14:33:41 and it will prevent anyone who is tempted to add php71-foobar packages 2017-02-13 14:34:21 my thinking is, if you add 'testing' repo, then you get the latest version of a package 2017-02-13 14:34:27 the stuff that we are testing 2017-02-13 14:34:35 the stuff that is not ready for stable release 2017-02-13 14:36:11 do we have a list of things that prevents us from replace 7.0 with 7.1? 2017-02-13 14:40:29 not yet, I stuck with compiling some of them this weekend 2017-02-13 14:45:41 FYI I’m preparing upgrade of Ruby to 2.4.0 2017-02-13 14:45:54 <^7heo> moin alle 2017-02-13 14:46:21 moin moin ping pong 2017-02-13 14:46:23 <^7heo> how are things? 2017-02-13 14:46:28 <^7heo> :P 2017-02-13 14:46:37 <^7heo> jirutka: slow down on the coffee mate ;) 2017-02-13 14:46:48 I’ve just ate my lunch 2017-02-13 14:46:51 going to office now 2017-02-13 14:46:56 <^7heo> Oh ok ;) 2017-02-13 14:47:24 I must stay home longer b/c they delivered me a carpet 2017-02-13 14:47:48 5 m long, quite funny to get it into my apartment in 3rd floor… 2017-02-13 14:48:47 grr, fcolista hard-coded ruby ABI version in xapian-bindings >_< 2017-02-13 14:51:04 and scadu is preparing upgrade of Python to 3.6 2017-02-13 15:10:16 jirutka: actually 3.6 seems to be working, however I need to build all packages depending on it. should I commit almost 200 commits (~200 pkgs) in 1 pr? 2017-02-13 15:22:55 ^7heo: Looks like when I use rc_parallel="YES", it breaks if I don't rc-update del hwclock boot 2017-02-13 15:23:15 I mean I tested it out with v3.5.0 but now I am using edge. 2017-02-13 15:23:34 <^7heo> pickfire: with openrc?\ 2017-02-13 15:23:40 Yeah 2017-02-13 15:23:44 Now working 2017-02-13 15:23:49 <^7heo> what made it work? 2017-02-13 15:24:01 rc-update del hwclock boot 2017-02-13 15:24:04 <^7heo> ah ok 2017-02-13 15:24:15 <^7heo> so it's not a fix to the situation, it's a workaround. 2017-02-13 15:24:20 <^7heo> also, what "breaks"? 2017-02-13 15:26:10 ^7heo: boot 2017-02-13 15:27:00 bootchart breaks as well (might be my mistake) 2017-02-13 15:27:15 Well, nothing else breaks 2017-02-13 15:27:26 <^7heo> pickfire: what, the machine doesn't boot at tall? 2017-02-13 15:27:31 <^7heo> s/tall/all/ 2017-02-13 15:27:35 Wait 2017-02-13 15:28:19 scadu: yes 2017-02-13 15:28:54 ACTION is uploading the photo 2017-02-13 15:29:01 Network is kinda slow now 2017-02-13 15:29:10 jirutka: will do then. but maybe not today since I have 100 other things to do 2017-02-13 15:29:30 jirutka: I'll try to push pr tomorrow 2017-02-13 15:29:49 scadu: np 2017-02-13 15:31:06 ^7heo: https://transfer.sh/MBpW1/img-20170213-111107-hdr.jpg 2017-02-13 15:32:30 <^7heo> pickfire: wait, hwclock impeaches the mounting of sda2?! 2017-02-13 15:32:52 No 2017-02-13 15:33:14 ^7heo: Isn't about the setup I did, luks 1/2 devices + btrfs raid 0 2/2 devices 2017-02-13 15:33:45 <^7heo> I'm not sure I follow. 2017-02-13 15:34:19 ^7heo: which means if I would like to boot, I need to type out the stuff to be able to boot. 2017-02-13 15:35:03 <^7heo> pickfire: if you can type it, you can get it done automatically ;) 2017-02-13 15:35:07 I did modify the /usr/share/mkinitfs/initramfs-init to add --allow-discards to luks mount, after I did that, it doesn't work as expected. 2017-02-13 15:35:13 <^7heo> pickfire: so it's not a problem in your setup 2017-02-13 15:35:32 But I try typing out the stuff that I type in initramfs-init, does work either. 2017-02-13 15:36:13 For now, I am just figuring out how to start dbus in order to start firefox. 2017-02-13 15:37:26 <^7heo> apk add dbus && /etc/init.d/dbus start 2017-02-13 15:38:16 <^7heo> but my dbus starts because "needed" 2017-02-13 15:38:54 <^7heo> and that's because of wpa_supplicant. 2017-02-13 15:58:35 Oh 2017-02-13 15:58:39 ^7heo: I didn't start dbus 2017-02-13 15:58:49 I just dbus-uuidgen > /etc/machine-id and now firefox works 2017-02-13 15:59:11 firefox needs dbus? 2017-02-13 15:59:15 jirutka: Yes 2017-02-13 15:59:20 that’s nasty 2017-02-13 15:59:20 No, it needs /etc/machine-id 2017-02-13 15:59:28 And dbus generates /etc/machine-id 2017-02-13 15:59:43 jirutka: But anyway, firefox is way smaller than chromium 2017-02-13 16:00:08 ACTION is still setting up wireless connectivity 2017-02-13 16:00:08 ofc, I didn’t want to advocate for chromium! 2017-02-13 16:00:11 Firefox FTW! 2017-02-13 16:00:35 for the win of what ? 2017-02-13 16:01:22 for the win of users and world wide web 2017-02-13 16:10:25 <^7heo> how can users win anything from the world wide web? 2017-02-13 16:10:28 <^7heo> it's bloated and horrible 2017-02-13 16:10:47 well… that’s unfortunately true :/ 2017-02-13 16:28:37 ncopa: do you have some handy script for bumping pkgrel of depending abuilds when doing minor upgrade of a package? except apkgrel 2017-02-13 16:28:59 I know how to automate it, just if there’s already some ready-made script 2017-02-13 16:30:54 jirutka: he should bundle them and put them in git. 2017-02-13 16:31:14 clandmeter: bundle them into single commit? 2017-02-13 16:31:31 in a seperate repo or whatever 2017-02-13 16:31:41 clandmeter: aha, you mean he as ncopa and them as scripts? 2017-02-13 16:32:00 i think he has more scripts he uses 2017-02-13 16:32:06 for instnace to bump kernels/modules 2017-02-13 16:32:13 clandmeter: yeah 2017-02-13 16:32:21 clandmeter: IMO https://github.com/alpinelinux/abuild is the right repo 2017-02-13 16:32:31 maybe 2017-02-13 16:32:46 or create directory “contrib” or something like that in that repo 2017-02-13 16:33:00 yes thats also fine. 2017-02-13 16:33:26 but maybe its easier in a seperate one (make it less official and easier to just commit). just thinking out loud... 2017-02-13 16:33:54 i know he has them, now we need to find a way to make him show it. 2017-02-13 16:34:13 ACTION grabs has digital gun 2017-02-13 16:36:05 XD 2017-02-13 16:36:17 SCRIPTS, OR LIFE! 2017-02-13 16:36:38 or we haxor your twitter! 2017-02-13 16:37:23 put even more cats on it! 2017-02-13 16:38:13 i think he's hiding behind his firewall... 2017-02-13 16:38:25 i think its time to go home :) 2017-02-13 16:39:09 well, everyone has his/her own set of quick&dirty scripts that (s)he don’t wanna show anyone :) 2017-02-13 16:39:27 with dirty ascii art? 2017-02-13 16:39:34 even those XD 2017-02-13 16:39:44 SHARE THEM! 2017-02-13 16:39:50 NO WAY! 2017-02-13 16:40:35 we are still on topic right? 2017-02-13 16:40:52 yes, ofc 2017-02-13 16:58:53 jirutka: 2017-02-13 16:58:56 ncdev-edge-x86_64:~/aports$ tpaste < rebuild-since 2017-02-13 16:58:56 http://tpaste.us/djVW 2017-02-13 16:59:08 thats what i use 2017-02-13 16:59:33 ^ scadu 2017-02-13 17:00:44 so i do: cd ~/aports/main; apk search --exact --origin -r -q 'so:libfoo.so.1' | sh ../rebuild-since 'rebuild against new version of libfoo' ^ 2017-02-13 17:01:10 or make the list of packages to rebuild in some texteditor which i paste 2017-02-13 17:18:03 xentec: as i understand, the only patch we need is: https://github.com/OpenSMTPD/OpenSMTPD/commit/f948b923873a93472dea9b786cf60a3472b0ddc8 2017-02-13 17:18:10 is that correct? 2017-02-13 17:26:45 ncopa: do you have any reason for icu in open-vm-tools pkg? 2017-02-13 17:27:04 ncopa: i.e. may i remove it to strip the pkg size a bit? 2017-02-13 17:27:13 i think it was needed once, for something 2017-02-13 17:27:20 i dont remember what it was needed for 2017-02-13 17:27:49 oka 2017-02-13 17:28:00 im ok to remove it if it does not break anything 2017-02-13 17:28:10 I’m not sure 2017-02-13 17:28:21 to be honest I do not have a clue what does this pkg actually do 2017-02-13 17:28:34 <^7heo> docker - the new package manager. 2017-02-13 17:28:41 <^7heo> (sorry, offtopic atm) 2017-02-13 17:28:45 <^7heo> (just randomly ranting) 2017-02-13 17:28:56 but infra admin wanted to install it on my guest 2017-02-13 17:29:44 ncopa: what if I disable vgauth? will it break shrinking of guest memory etc? 2017-02-13 17:29:57 (i mean disable for myself, not in pkg) 2017-02-13 17:30:05 i have no idea :) 2017-02-13 17:31:01 it looks like a giant backdoor into the system for me… 2017-02-13 17:32:24 ^7heo: yeah, kids who don’t know how to make (distribution) packages now use docker to make “packages”… 2017-02-13 17:33:22 <^7heo> jirutka: yep 2017-02-13 17:39:17 \o/ i was finally able to build 3rdparty kernel mod 2017-02-13 17:41:00 using docker as package manager might make sense in some situations 2017-02-13 17:41:10 but not for system libraries :) 2017-02-13 17:41:30 # uname -a 2017-02-13 17:41:30 Linux localhost 4.9.9-0-grsec #1-Alpine SMP Mon Feb 13 17:10:31 GMT 2017 x86_64 Linux 2017-02-13 17:41:48 i am making progress! 2017-02-13 17:42:04 <^7heo> ncopa: I wouldn't want any package ever to be made with docker. 2017-02-13 17:42:41 <^7heo> ncopa: however, I am happy with docker powering my testing environment 2017-02-13 17:43:04 so, "packages" for your testing env 2017-02-13 17:43:21 <^7heo> that's like calling a text file a package. 2017-02-13 17:43:34 <^7heo> or a video stream. 2017-02-13 17:43:41 like a 'package' with debian gcc toolchain 2017-02-13 17:43:50 thats how i use it 2017-02-13 17:44:09 i think its very conveniant to test how things behave, work in gcc with glibc 2017-02-13 17:44:12 <^7heo> yeah that's not the meaning people give to package, afaik. 2017-02-13 17:44:17 without needing spend 2 hours installing a vm 2017-02-13 17:44:31 <^7heo> oh man, I surely am happy I do not need 2h to install a VM ;) 2017-02-13 17:44:39 wow, docker packages embedded in funny videos of cats! that would be a thing! I’m gonna start a company in Silicon Valley right now :P 2017-02-13 17:44:45 lol 2017-02-13 17:44:48 <^7heo> but yeah for disposable stuff, it's kinda useful. 2017-02-13 17:45:08 <^7heo> that's why docker is useful: disposable stuff. 2017-02-13 17:45:15 im also thinking for things like running redmine 2017-02-13 17:45:26 where you want gem install a bunch of stuff 2017-02-13 17:45:31 ncopa: bundles 2017-02-13 17:45:34 <^7heo> if you need docker to run redmine; ditch redmine. 2017-02-13 17:45:49 <^7heo> that's my approach on things. 2017-02-13 17:45:59 <^7heo> docker neeeded? ditch: keep; 2017-02-13 17:46:11 we are not using it for redmine (yet) but i think its a valid usecase 2017-02-13 17:46:21 ncopa: you donjt need docker or any other containerization at all, just bundle dependencies into single pkg and use normal package manager, just as I package gitlab-ce, mailman, sentry, … https://github.com/jirutka/user-aports/tree/v3.5/bundles 2017-02-13 17:48:26 the phobia from bundling dependencies (in some cases) in distros is one of the reasons why ppl use docker; but this has nothing to do with technical reasons; using docker image instead of normal package is also bundling, but instead of bundling python/ruby modules, you’re bundling entire system… 2017-02-13 17:48:48 it’s like when you want a banana, but get gorilla holding the banana and entire forest with it 2017-02-13 17:48:55 lol 2017-02-13 17:49:21 yeah 2017-02-13 17:49:40 this is common explanation of OOP, but it’s perfect even for containers ;) 2017-02-13 17:49:44 oh crap, Rust meetup! 2017-02-13 17:49:46 i saw the dinosaur talk from fosdem btw 2017-02-13 17:49:46 <^7heo> jirutka: that's a valid analogy. 2017-02-13 17:49:47 I totally forgot to it 2017-02-13 17:49:52 shit 2017-02-13 17:49:54 <^7heo> jirutka: plus, people can relate to gorillas. 2017-02-13 17:50:09 10 minutes… can’t get in time 2017-02-13 17:50:16 by the suse guy 2017-02-13 17:50:23 <^7heo> jirutka: that's what they would be if they would be going to the gym. 2017-02-13 17:50:32 i think he had valid points 2017-02-13 17:50:33 <^7heo> ncopa: didn't I link it? 2017-02-13 17:50:41 that one yes 2017-02-13 17:50:44 <^7heo> :) 2017-02-13 17:51:00 and i agree basically all he said 2017-02-13 17:51:18 <^7heo> say no more ;) 2017-02-13 17:51:32 he also said at the quesions section "i think that is a valid use case where you do want containers" 2017-02-13 17:51:46 "but i dont think its a general solution for all problems" 2017-02-13 17:51:53 or something like that 2017-02-13 17:51:57 and 2017-02-13 17:52:07 "with great power comes great responsability" 2017-02-13 17:52:38 so, yeah, i think distro packaging is normally a better way to package things 2017-02-13 17:52:43 <^7heo> yeah, that's what people DESINING docker have to remind themselves. 2017-02-13 17:52:46 <^7heo> not the users ;) 2017-02-13 17:52:51 <^7heo> it's not like docker gives a lot of power... 2017-02-13 17:52:56 <^7heo> excepted when you want to write an exploit. 2017-02-13 17:53:18 well, docker is a powerful tool 2017-02-13 17:53:26 <^7heo> I dunno. 2017-02-13 17:53:29 you *can* package entire universe if i want 2017-02-13 17:53:33 <^7heo> it's kinda like cgroups + make 2017-02-13 17:53:36 <^7heo> with a shiny name. 2017-02-13 17:53:36 but its not necessary a good idea 2017-02-13 17:53:48 i mean, you *can* use it as a package manager 2017-02-13 17:54:01 <^7heo> yeah, like you can use cgroups + make as a package manager ;) 2017-02-13 17:54:03 ^7heo: no, it’s linux namespaces (this is the most important), plus cgroups and some other isolation on top 2017-02-13 17:54:08 but just because you can do it, it does not mean its a good idea 2017-02-13 17:54:31 <^7heo> jirutka: I thought that namespaces and cgroups were basically one thing 2017-02-13 17:54:39 i'm sorry for leaving this discussion, but must go to rust meatup now 2017-02-13 17:54:47 <^7heo> take care 2017-02-13 17:54:49 <^7heo> run fast 2017-02-13 17:54:51 <^7heo> ttyl. 2017-02-13 17:55:03 ^7heo: now, these are two different things; actually I don’t use cgroups with LXC at all in my setups 2017-02-13 17:55:59 <^7heo> jirutka: looks like I got it backwards yes. 2017-02-13 17:56:09 <^7heo> jirutka: cgroups is a namespace; not the contrary. 2017-02-13 17:57:09 <^7heo> anyway 2017-02-13 17:57:21 <^7heo> docker is kinda like systemd. It started from a good idea. 2017-02-13 17:57:43 <^7heo> But the implementation went amok somewhere down the road. 2017-02-13 17:57:52 <^7heo> and now it implements way too many features. 2017-02-13 17:58:20 <^7heo> fortunately it stopped MUCH before reaching the black-hole-singularity point that systemd is going for. 2017-02-13 19:01:25 jirutka: thanks for highlighting. I'll try to get it done tomorrow. Uh, maybe I'll setup SSH tunnel to my home laptop, so I could do something when being in the office 2017-02-13 23:30:24 today at Rust meetup someone showed me emacs… I was just wow… now I’m thinking about giving it a try 2017-02-13 23:31:04 and yes, it’s really an operating systems that just happens to be also a text editor; that’s not a joke, but reality :) 2017-02-14 00:03:53 jirutka: welcome to the lightside ;) 2017-02-14 00:04:10 zwiPZuf9W2Lt: how are you? :) 2017-02-14 00:04:46 sick with the flu, but i remember you being sceptical about me packaging emacs ;) 2017-02-14 00:05:01 aha :) 2017-02-14 00:05:58 well, I have plenty of hard-believer vim users… 2017-02-14 00:28:16 open-vm-tools is horrible, typical bloated, undocumented, non-transparent crap from company making proprietary craps 2017-02-14 03:27:12 ncopa: looks like other way around. deleting arc4random from libasr made opensmtpd link to the correct symbols in crypto so everything works now 2017-02-14 05:33:51 did anyone here know, that the bootstrap.sh in aports is really broken? I made it working https://gist.github.com/xentec/dbbf3cfdc3342a14000db4bedce193bd 2017-02-14 05:34:05 s/working/work 2017-02-14 05:35:41 xentec: need diff 2017-02-14 05:35:47 xentec: also, what is 'broken' about it 2017-02-14 05:36:57 https://gist.github.com/xentec/dbbf3cfdc3342a14000db4bedce193bd/revisions 2017-02-14 05:37:37 fast way to find out is to try to use it to build something 2017-02-14 05:38:02 it works for me 2017-02-14 05:38:11 what are you trying to do with it 2017-02-14 05:38:51 I've tried to create an optimized rpi3 build with "./bootstrap armv7" 2017-02-14 05:39:32 abuild has to be aware of the architecture ... 2017-02-14 05:39:38 as does apk-tools ... 2017-02-14 05:42:50 try it with armhf then 2017-02-14 05:46:43 works fine 2017-02-14 05:46:48 it's building gcc and so on 2017-02-14 05:46:59 http://pastebin.com/u9KyZAF6 2017-02-14 05:50:43 ah, i see the problem there. 2017-02-14 05:50:45 one second 2017-02-14 05:50:58 well I see now where I got the idea it way broken: it accepts any input but procecces it in a strage way. after short forms of arch broke, I've tried it a full triple which worked at first but broke even more and this is where my fixing began 2017-02-14 05:51:58 armv7) echo "armv7-alpine-linux-musleabihf" ;; 2017-02-14 05:52:00 this seems wrong 2017-02-14 05:52:29 shouldnt it match armv6 i.e. -muslgnueabihf 2017-02-14 05:52:55 my version puts input in the hands of abuilds functions.sh 2017-02-14 05:54:01 you mean armhf) echo "armhf-alpine-linux-muslgnueabihf" ;;? 2017-02-14 05:54:18 yes 2017-02-14 05:54:18 because armv7 is fine 2017-02-14 05:54:25 i was just making sure 2017-02-14 05:55:15 should be "armv6-alpine-linux-muslgnueabihf" 2017-02-14 05:56:36 well.. my bootstrap.sh seems useless now. so much work only because of two wrong characters :( 2017-02-14 05:57:54 :D 2017-02-14 05:58:00 2.29-r7 has the fix 2017-02-14 05:58:07 just got uploaded to master 2017-02-14 06:44:53 kaniini, if you're still there: kmod and util-linux fail to build because of bashcomp() 2017-02-14 06:46:06 i was about to go to bed, do you have a fix for it? 2017-02-14 06:49:57 sadly not as a real patch: http://pastebin.com/kEd89LwJ 2017-02-14 08:28:31 http://tpaste.us/aQMq om nom nom 2017-02-14 08:28:37 clandmeter: ^ any hints? 2017-02-14 08:35:16 clandmeter: python3 APKBUILD currently looks like this: http://tpaste.us/qMow 2017-02-14 08:35:40 I added --enable-optimizations and --enable-lto 2017-02-14 08:35:50 maybe some tests failed because of lto :s 2017-02-14 08:50:04 scadu: i have no idea, but compare the build result of python, i dont know what you changed (or what changed in python itself) 2017-02-14 09:15:53 clandmeter: only those two options. will dig into this. 2017-02-14 09:39:51 i was just about to upgrade asterisk.... but https://issues.asterisk.org/jira/browse/ASTERISK-26791 2017-02-14 12:08:17 meh, dunno. enum.py looks properly and has IntFlag as expected. 2017-02-14 16:20:17 xentec: both of those packages build for me -- how are you building them ? 2017-02-14 16:33:50 kaniini: huh.. now the error is gone here too. as noted in pastebin, it was about bash-compltions missing in $pkgdir. maybe there is a hidden dependency but I cant reproduce it anymore 2017-02-14 16:52:02 fabled: any objection to backporting the tests support? 2017-02-14 17:24:46 kaniini: you mean for edge? I'm ok with that 2017-02-14 17:25:01 we should probably tag a release soonish 2017-02-14 17:56:41 ncopa, kaniini : yeah, that wouuld be good for edge. would be nice to get 'set -e' enabled too 2017-02-14 18:08:15 \q 2017-02-14 18:12:19 psql? :) 2017-02-14 18:13:27 yeah... in the wrong window 2017-02-14 18:20:37 :q 2017-02-14 18:22:03 drop table 2017-02-14 18:24:09 ACTION chooses red pill. 2017-02-14 18:42:58 okay 2017-02-14 18:44:42 actually, reason I'm here is clandmeter pinged me about 2 users who have had problems adding content to the wiki. They get hung up on abuse filter rule 001 2017-02-14 18:45:08 the rule prevents new users from adding links to the wiki 2017-02-14 18:45:30 Right now we don't tell anyone why they are blocked, just that the content is blocked 2017-02-14 18:46:07 I'm curious if anyone has opinions on the risk/benefit of changing the rule text to something more informative, like "New users are not allowed to add links to the wiki" 2017-02-14 18:46:59 The rule's been hit >200 times since we enabled the abuse filter, and only 3 false positives so far 2017-02-14 18:51:38 sounds useful to inform user about it 2017-02-14 18:51:59 spambots are too impatient to wait to become a non-new user 2017-02-14 18:53:01 i suppose another alternative is manual review 2017-02-14 18:53:07 eg git pull requests 2017-02-14 18:54:13 i think i'll push 4.9.9 kernel now 2017-02-14 18:54:21 hold on to your hats 2017-02-14 18:54:47 ncopa: not a big fan of manual review - that still means an admin has to review 2017-02-14 18:55:09 yes, thats the big downside 2017-02-14 18:55:11 its a wetware DDOS 2017-02-14 19:47:27 ncopa: the one with the grsec CFI patch? ;) 2017-02-14 19:50:24 yup 2017-02-14 19:50:40 seems like its broken on 32bit x86 though 2017-02-14 20:13:10 ok, first problem with 4.9.9. zfs does not seem to work 2017-02-14 20:13:28 znvpair.ok invalid module format 2017-02-14 20:19:35 thats an interesting one 2017-02-14 20:19:37 .ok? 2017-02-14 20:21:49 from dmesg 2017-02-14 20:21:52 znvpair: module is not compatible with the KERNEXEC 'or' method and RAP 2017-02-14 20:22:16 ah 2017-02-14 20:22:22 you want KERNEXEC bts, ncopa 2017-02-14 20:23:33 ncopa: PAX_KERNEXEC_PLUGIN_METHOD_BTS 2017-02-14 20:23:36 instead of PAX_KERNEXEC_PLUGIN_METHOD_OR 2017-02-14 20:23:38 in your kconfig 2017-02-14 20:23:42 https://github.com/zfsonlinux/zfs/issues/4649 2017-02-14 20:24:09 ugh, it emits that on either? 2017-02-14 20:24:11 lessee 2017-02-14 20:24:39 CONFIG_PAX_KERNEXEC_PLUGIN=y 2017-02-14 20:24:39 # CONFIG_PAX_KERNEXEC_PLUGIN_METHOD_NONE is not set 2017-02-14 20:24:39 CONFIG_PAX_KERNEXEC_PLUGIN_METHOD_BTS=y 2017-02-14 20:24:49 yeah, so the error is a bit misleading 2017-02-14 20:24:52 looks like it already is method BTS 2017-02-14 20:25:25 yeah, gonna see what stops RAP from working 2017-02-14 20:25:34 it may just need a rebuilt against the new kernel headers 2017-02-14 20:25:42 /build sys 2017-02-14 20:26:34 hum 2017-02-14 20:26:48 seems like out-of-tree modules does not work with RAP at all? 2017-02-14 20:27:32 uhhhh 2017-02-14 20:27:34 #if defined(CONFIG_PAX_KERNEXEC_PLUGIN_METHOD_OR) || defined(CONFIG_PAX_RAP) 2017-02-14 20:27:36 if (!license || !license_is_gpl_compatible(license)) { 2017-02-14 20:27:41 this is stupid 2017-02-14 20:27:54 so the check that errors out is just that license check ^ 2017-02-14 20:28:35 so out of tree modules that aren't GPL-compatible 2017-02-14 20:28:59 i wonder what the reasoning is... 2017-02-14 20:30:27 https://bugs.gentoo.org/show_bug.cgi?id=581726#c2 2017-02-14 20:31:17 i was just gonna link that 2017-02-14 20:31:19 :p 2017-02-14 20:31:47 so apparently, also 2017-02-14 20:31:53 https://bugs.gentoo.org/show_bug.cgi?id=581726#c7 2017-02-14 20:32:03 as long as it's compiled with that kernel out-of-tree it should work fine? 2017-02-14 20:32:14 it's just their way to block binary modules, which is somewhat inaccurate it seems 2017-02-14 20:35:03 a simple fix is adding a comparison against "CDDL" to that block, but it's a hack 2017-02-14 20:36:17 yes 2017-02-14 20:36:32 i think they want filter out precompiled binaries 2017-02-14 20:37:02 yeah, they say as much in the linked issue tracker 2017-02-14 20:37:11 which makes sense, since they aren't protected by RAP so Bad Things happen 2017-02-14 20:37:20 the method however, doesn't make sense :p 2017-02-14 20:37:53 hm 2017-02-14 20:38:08 i wonder why they dont use vermagic for that 2017-02-14 20:58:09 i think its for the error message 2017-02-14 20:58:20 ? 2017-02-14 21:07:11 Shiz: i think they need to test if the module is a precompiled binary or not 2017-02-14 21:07:21 yes, you've said that before :P 2017-02-14 21:07:38 and to give a proper error message 2017-02-14 21:08:00 how to test it properly though 2017-02-14 21:08:05 vermagic? 2017-02-14 21:08:11 doesn't sound right 2017-02-14 21:08:22 i'm sure nvidia's precompiled stuff vermagics up with the e.g. official ubuntu kernels too 2017-02-14 21:08:22 yes but i think that can be overridden with force 2017-02-14 21:08:27 that too 2017-02-14 21:08:52 but vermagic also includes RAP KERNEXEC_BTS 2017-02-14 21:08:57 REFCOUNT GRSEC 2017-02-14 21:09:15 so kernel modules built without those should get error too 2017-02-14 21:09:19 right 2017-02-14 21:09:27 have you asked them why not to check vermagic? :P 2017-02-14 21:09:37 no 2017-02-14 21:09:43 but they are not stupid 2017-02-14 21:10:46 oh 2017-02-14 21:10:58 its probably the plugin license 2017-02-14 21:11:27 hm 2017-02-14 21:11:41 no that does not make sense either 2017-02-14 21:11:43 im not sure on that 2017-02-14 21:11:48 its not mentioned anywhere and it's a gcc plugin right 2017-02-14 21:12:01 that would have no effect on the generated code 2017-02-14 21:12:12 the gcc plugin is gpl2 in the free liicense 2017-02-14 21:12:16 and if so, they would stop compilation at module compile time 2017-02-14 21:12:39 in the commercial license it is gpl3 i think 2017-02-14 21:12:52 i mean if you pay for the grsecurity patch you get a gpl3 license 2017-02-14 21:13:00 doesn't really matter since gcc license != generated code license, so it shouldnt matter 2017-02-14 21:13:06 or gcc plugin license 2017-02-14 21:13:07 :p 2017-02-14 21:13:34 gcc has some plugins exception so that you cannot ship closed source plugins 2017-02-14 21:13:58 but i think it only affects libgcc 2017-02-14 21:14:05 whcih you dont use for the kernel 2017-02-14 21:14:28 i think they exploit the gcc gpl3 exception 2017-02-14 21:14:43 so you cannot use RAP to build userspace apps with the plugin 2017-02-14 21:14:48 unless you pay 2017-02-14 21:15:33 possibly 2017-02-14 21:15:39 has no bearing on zfs though :p 2017-02-14 21:15:59 should not, yes 2017-02-14 21:37:38 clandmeter: re: python3 rebuild and py-enum34 problem. that's the diff of python's APKBUILD: http://tpaste.us/xWeJ as you see not much was changed. issue-27955.patch have been removed since it seems to be already adopted by upstream 2017-02-15 05:52:14 Hey there I upgraded one of my physical server 2017-02-15 05:52:47 after kernel upgrade I've hit a nice extlinux boot loop :( 2017-02-15 05:53:20 I've tried putting back the previous kernel that had no issue from a live cd but gives me the same extlinux boot loop... 2017-02-15 05:53:30 it's on a raid1 array .... any idea ? 2017-02-15 05:54:22 during post install in chroot extlinux complained it couldn't open /dev/md0 though ... 2017-02-15 06:09:14 ok it's indeed better with grsec chroot protections to 0 2017-02-15 06:09:27 at least extlinux doesn't complain 2017-02-15 06:09:36 let's see how a reboot fares ... 2017-02-15 06:11:50 still not working :( 2017-02-15 06:12:08 any idea how to repair extlinux there ? :( 2017-02-15 06:28:56 extlinux -i/-U /boot complete without issue indeed 2017-02-15 13:46:08 im taking down build-edge-armhf and build-3-5-armhf for a couple of days 2017-02-15 15:23:40 <^7heo> ncopa: don't you think it'd be good to add a symlink fsck.vfat -> fsck.fat in the dosfstools package? 2017-02-15 15:36:25 why? 2017-02-15 15:36:34 does any other distro do so? 2017-02-15 16:16:43 <^7heo> ncopa: because people use mkfs.vfat 2017-02-15 16:16:53 <^7heo> ncopa: and it's probably a good idea to be consistent. 2017-02-15 16:23:00 ^7heo: huh? 2017-02-15 16:23:49 <^7heo> clandmeter: huh! :) 2017-02-15 16:24:12 < ^7heo> ncopa: because people use mkfs.vfat ? huh? 2017-02-15 16:24:29 <^7heo> clandmeter: what's not clear with that? 2017-02-15 16:24:53 the question was what other distro's do 2017-02-15 16:25:38 and it looks like they do use --enable-compat-symlinks 2017-02-15 16:27:24 <^7heo> clandmeter: how the hell should I know what --enable-compat-symlinks flag is for, and what binary it is used with? =/ 2017-02-15 16:28:04 <^7heo> clandmeter: and no, the question was not "what do other distros do?". It was "does ANY other distro do so?". 2017-02-15 16:28:27 <^7heo> clandmeter: the first question cannot be answered with yes/no, while the second can. 2017-02-15 16:29:06 <^7heo> clandmeter: and so my answer to "why?" hinted that "I don't know and it isn't relevant" was implied as an answer to "does any other distro do so?". 2017-02-15 16:33:03 they do, so while you were trying to explain yourself i already added the configure option. 2017-02-15 16:37:23 the diff is http://tpaste.us/oZQY 2017-02-15 16:56:07 scadu: better to use this channel. 2017-02-15 16:56:31 i know some users are not in alpine-linux 2017-02-15 16:57:11 clandmeter: fffu, I didn't notice we're on -linux ;x 2017-02-15 16:57:37 clandmeter: I thought you've answered on -devel :P 2017-02-15 16:58:00 i was also using it incorrectly, but i found out after the first line :) 2017-02-15 17:03:08 I'm tired and probably flu is comming, so I wasn't aware where I write 2017-02-15 17:03:27 it's 6PM now and I feel like it was 2AM 2017-02-15 17:32:57 jirutka: So I think I finally know what you meant by eclass hell. :) 2017-02-15 17:35:27 jirutka: On the other hand, it seems to be in the actual usage of the eclass concept and not in the concept itself. So one wouldn't write eclasses for pieces of software (like kernel or myspell). On the other hand I can't deny that the Gentoo way has some advantages like saving a couple of lines for repetitive packages. It looks like a trade of between complexity and repetitiveness, which is quite 2017-02-15 17:35:33 common in IT. :) 2017-02-15 17:54:00 anybody here already used supervise-daemon? 2017-02-15 17:54:34 sounds shitty 2017-02-15 17:54:38 just sayin' 2017-02-15 17:55:32 :) 2017-02-15 17:55:52 clandmeter: your change to dosfstools is incomplete 2017-02-15 17:56:08 clandmeter: after installing 4.1-r1 2017-02-15 17:56:10 raccoon:~# apk info -W /sbin/mkfs.vfat 2017-02-15 17:56:12 /sbin/mkfs.vfat symlink target is owned by busybox-1.26.1-r3 2017-02-15 17:57:13 kaniini: works for me 2017-02-15 17:57:35 ohh, i see 2017-02-15 17:57:41 there is /usr/sbin/mkfs.vfat 2017-02-15 17:57:45 /usr/sbin/mkfs.vfat 2017-02-15 17:57:48 but that means busybox mkfs.vfat 2017-02-15 17:57:51 is still preferred 2017-02-15 17:58:03 or maybe not 2017-02-15 17:58:22 indeed not 2017-02-15 17:58:37 okay nevermind :)P 2017-02-15 18:11:22 clandmeter - runit is good for service supervision - it's small & I've not had a single service fail in the year or so I've been using it 2017-02-15 18:28:35 BitL0G1c: how is it better then just using supervise-daemon? 2017-02-15 18:35:27 I thought supervisord needed python ? 2017-02-15 18:38:51 supervise-daemon is part of openrc 2017-02-15 18:40:00 BitL0G1c: https://github.com/OpenRC/openrc/blob/master/supervise-daemon-guide.md 2017-02-15 18:40:14 yes just reading now https://github.com/OpenRC/openrc/blob/master/supervise-daemon-guide.md 2017-02-15 18:41:29 documentation is limited 2017-02-15 18:44:52 going to try it 2017-02-15 18:48:32 we can now freely speak about it, skarnet is gone :p 2017-02-15 18:48:43 <^7heo> clandmeter: this chan is logged ;) 2017-02-15 18:48:51 it is? 2017-02-15 18:48:56 shit 2017-02-15 18:49:11 rm -rf /dev/backlog :p 2017-02-15 18:49:19 <^7heo> you wish. 2017-02-15 18:49:26 <^7heo> doesn't work like that... 2017-02-15 18:49:26 <^7heo> :D 2017-02-15 18:49:46 what about a hidden chan :p 2017-02-15 18:49:53 <^7heo> shhhhht! 2017-02-15 21:08:51 alp:~# zpool create tank /dev/loop1 2017-02-15 21:08:51 alp:~# 2017-02-15 21:08:53 \o/ 2017-02-15 21:09:13 i think this RAP thing from grsecurity is really really cool 2017-02-15 21:23:00 $ uname -a 2017-02-15 21:23:00 Linux ncopa-macbook 4.9.9-1-grsec #2-Alpine SMP Wed Feb 15 10:52:23 GMT 2017 x86_64 Linux 2017-02-15 21:23:04 it boots 2017-02-15 21:23:06 with zfs 2017-02-15 21:34:10 ncopa: niceeeee 2017-02-15 21:34:23 zfs is pretty slick 2017-02-15 21:35:28 but it does not shutdown... 2017-02-15 21:35:52 eeeeh? 2017-02-15 21:36:12 yup 2017-02-15 21:36:18 will have to investigate tomorrow 2017-02-15 21:37:44 like it hangs? 2017-02-15 21:41:34 yes 2017-02-15 21:41:46 darn 2017-02-15 21:41:48 i wonder if it is related new kernel or the recent udev changes 2017-02-15 21:42:37 <^7heo> there's nothing like working while listening to the Super mario OST. 2017-02-15 21:43:00 <^7heo> or that might have to go on offtopic, sorry. 2017-02-15 21:43:23 <^7heo> I'd be happy to try zfs 2017-02-15 21:43:25 <^7heo> (bwt) 2017-02-15 21:43:28 <^7heo> s/wt/tw/ 2017-02-15 21:52:01 Hey friends :) 2017-02-15 21:53:32 <^7heo> Hey batman. 2017-02-15 22:01:36 Leo! We missed you at fosdem! 2017-02-15 22:01:58 yeah i know ... i never have money to go to those events :( 2017-02-15 22:02:35 I dunno, maybe if you were getting paid for your work, you would... :P 2017-02-15 22:02:49 hahaha 2017-02-15 22:02:52 no comment 2017-02-15 22:08:46 <^7heo> leo-unglaub: you should have told us. 2017-02-15 22:08:55 <^7heo> leo-unglaub: I'd have been happy to crowdfund some plane tickets. 2017-02-15 22:09:13 <^7heo> (easyjet is cheap) 2017-02-15 22:09:22 <^7heo> (not easy on you, but easy on your wallet) 2017-02-15 22:11:29 pretty sure leo-unglaub has already had to perform first aid on people getting out of an easyjet flight 2017-02-15 22:11:53 hahaha 2017-02-15 22:11:56 thats a nice offer 2017-02-16 08:20:26 fcolista: looks like we have the same issue in 3.5 branch 2017-02-16 08:21:42 ok 2017-02-16 08:21:55 this should be fixed 2017-02-16 08:22:35 in edge there are 2 pkgs that dep on it. 2017-02-16 08:24:06 clandmeter: fcolista so I'll proceed with rebuild of pkgs. I've done with rebuilding pkgs from main 2017-02-16 08:24:38 scadu, I'm going to remove the dependecies from py3-enum34 2017-02-16 08:24:42 started with community, but it tries to install python3 from repo, not the one from ~/packages 2017-02-16 08:24:49 fcolista: cool. 2017-02-16 08:24:52 py3-zeroconf 2017-02-16 08:24:53 py3-sqlalchemy-utils 2017-02-16 08:25:22 those 2 also need update 2017-02-16 08:25:30 not sure they are in your list scadu 2017-02-16 08:25:43 clandmeter: do you know how to tell abuild to use main/python3-* package from local repo when rebuilding packages in community? 2017-02-16 08:25:53 clandmeter: probably yes, but I have no access to my desktop computer right now 2017-02-16 08:26:15 do you have local repo added to repository file? 2017-02-16 08:26:29 if the version is higher it should always pull that in. 2017-02-16 08:26:35 clandmeter, , beets/certbot/obnam are pointing to py-enum34. I'm going to substitute it with py2-enum34 2017-02-16 08:27:20 lemme check if they use py2 or ppy3 2017-02-16 08:27:31 if they use py3, it's enough removing the dependency 2017-02-16 08:28:21 clandmeter: oh, I might forgot this one. 2017-02-16 08:29:13 fcolista: beets uses python2? 2017-02-16 08:29:21 Yes clandmeter 2017-02-16 08:29:46 any reason why? 2017-02-16 08:30:17 ...? 2017-02-16 08:30:21 not sure what the policy is regarding prefered python versions. 2017-02-16 08:30:25 what you mean? 2017-02-16 08:30:36 it's built against py2 2017-02-16 08:30:48 we might update it so it can use py3 2017-02-16 08:30:50 iirc we use python3 for packages that works with python3 2017-02-16 08:30:58 otherwise we use python2 2017-02-16 08:31:02 if it works.. 2017-02-16 08:31:04 then my question makes sense :) 2017-02-16 08:32:06 pypi says python 3.5 2017-02-16 08:32:37 pkgver=1.3.18 2017-02-16 08:32:44 the lateset version is 1.4.3 2017-02-16 08:33:10 woohoo 2017-02-16 08:33:43 we can upgrade the package and make it work with py3.5... 2017-02-16 08:34:04 that would be the prefered way. 2017-02-16 08:34:32 first let me fix the py2/3 enum34 2017-02-16 08:34:39 then i'll go for that 2017-02-16 08:35:01 soudns good? 2017-02-16 08:36:02 sure, though i think you dont need to add py2. 2017-02-16 08:36:26 i guess that only makes sense when you have py2 and py3 2017-02-16 08:36:50 Why? 2017-02-16 08:36:54 Ah 2017-02-16 08:36:59 because they are pulled automatically 2017-02-16 08:37:24 ? 2017-02-16 08:37:29 I mean 2017-02-16 08:37:37 fcolista: I've changed the APKBUILD for enum34 by removing subpackages, but I have no diff here ;x 2017-02-16 08:37:38 py2-enum34 is pulled automatically 2017-02-16 08:37:51 doh, I have to finally set up a SSH tunnel to my pc ;s 2017-02-16 08:38:50 fcolista: you did http://git.alpinelinux.org/cgit/aports/commit/main/py-enum34/APKBUILD?id=9167783064cd7989c7fa5301e9edd2703d15fcd8 2017-02-16 08:39:05 which introduced py2 and py3. 2017-02-16 08:39:09 which was not needed at all. 2017-02-16 08:39:16 yes 2017-02-16 08:39:28 so? 2017-02-16 08:39:39 go back to the orginal :) 2017-02-16 08:39:59 ok...my point is: 2017-02-16 08:40:16 the other packages referring to py-enum34 should be rebuilt? 2017-02-16 08:41:10 if you make that change yes. 2017-02-16 08:41:24 that's what i was saying.. 2017-02-16 08:41:25 im not sure what they depend on? 2017-02-16 08:41:34 py2- or py- 2017-02-16 08:42:02 seems that clandmeter didn't read what I wrote :) 2017-02-16 08:42:08 if you would use py- with the install_if, then it would not need to. 2017-02-16 08:42:23 which part? 2017-02-16 08:42:49 clandmeter, , beets/certbot/obnam are pointing to py-enum34. I'm going to substitute it with py2-enum34 2017-02-16 08:42:49 lemme check if they use py2 or ppy3 2017-02-16 08:42:49 if they use py3, it's enough removing the dependency 2017-02-16 08:43:59 but then, what I was thinking is 2017-02-16 08:44:14 py-enum34 pulled automatically py2-enum34 2017-02-16 08:44:43 in py2/py3 APkBUILD 2017-02-16 08:45:33 if those pkgs all depend on py-enum34, then just revert to old pkg format without py2/3. 2017-02-16 08:47:07 if any of them depend on py2-enum34 you will have to update them. 2017-02-16 09:59:47 <^7heo> ncopa: seen that https://www.helpnetsecurity.com/2017/02/15/bypass-aslr-protection-javascript/ ? 2017-02-16 10:14:53 yup 2017-02-16 11:49:06 ncopa, re set -e for abuild 2017-02-16 11:49:13 it turns out set -e is ignored inside functions 2017-02-16 12:10:04 really? 2017-02-16 12:10:10 thats ugly 2017-02-16 12:10:24 and inside subshell? ( set -e; ... ) 2017-02-16 12:24:20 http://stackoverflow.com/questions/19789102/why-is-bash-errexit-not-behaving-as-expected-in-function-calls 2017-02-16 13:18:47 kaniini, what's the armhf->armv6 change? 2017-02-16 14:00:09 jirutka, abuild commit 6b2fd4a891ba5eb7c32eeb37f5d18986d61d9090 breaks things 2017-02-16 14:00:18 "abuild: make default_prepare() always end up in "$builddir"" 2017-02-16 14:00:40 ncopa, ^ 2017-02-16 14:00:46 should we revert? 2017-02-16 14:00:53 or just make the cd conditional 2017-02-16 14:00:59 builddir is not always set 2017-02-16 14:01:13 it is always set… 2017-02-16 14:01:20 even if it is 2017-02-16 14:01:23 https://github.com/alpinelinux/abuild/blob/6b2fd4a891ba5eb7c32eeb37f5d18986d61d9090/abuild.in#L2243 2017-02-16 14:01:25 we have packages not having source ball 2017-02-16 14:01:30 that extracts stuff to that builddir 2017-02-16 14:01:37 e.g. all meta-packages 2017-02-16 14:01:46 hm, right 2017-02-16 14:02:29 probably should cd there only if the dir exists 2017-02-16 14:02:36 reverting this does not solve the real problem here 2017-02-16 14:02:43 and before applying patches check that it exists or die 2017-02-16 14:03:18 could this wait until evening? I’m very busy now 2017-02-16 14:03:32 and I feel that this has more consequencies 2017-02-16 14:06:37 sure, no panic 2017-02-16 14:06:44 just wanted to check what to do 2017-02-16 14:06:56 i will soon push changes to bootstrap.sh 2017-02-16 14:07:08 cleaning things up even more, and preparing for ghc bootstrap 2017-02-16 14:13:25 <^7heo> Did we already talk about integrating language/framework specific package managers into apk? 2017-02-16 14:14:29 <^7heo> (i.e. pip, gems, ppm, npm, etc) 2017-02-16 14:16:10 ^7heo: nope 2017-02-16 14:16:19 <^7heo> jirutka: what is your opinion on this? 2017-02-16 14:16:54 ^7heo: about what exactly? 2017-02-16 14:17:11 <^7heo> integrating language/framework specific package managers into apk 2017-02-16 14:17:21 but what kind of integration do you imagine? 2017-02-16 14:17:41 <^7heo> that apk would be able to perform operations by invoking those package managers. 2017-02-16 14:18:28 <^7heo> therefore seamlessly processing packages from those package managers. 2017-02-16 14:18:49 <^7heo> that would solve the issue "should we have py- packages or use pip?" 2017-02-16 14:19:20 that’s quite difficult problem 2017-02-16 14:19:27 <^7heo> not technically it is not. 2017-02-16 14:19:33 technically 2017-02-16 14:19:34 <^7heo> but it might be time consuming. 2017-02-16 14:19:40 <^7heo> why? 2017-02-16 14:19:42 b/c of fucking C 2017-02-16 14:19:46 <^7heo> ah no but that's fine. 2017-02-16 14:19:53 native extensions 2017-02-16 14:19:54 <^7heo> C can spawn other processes wasily. 2017-02-16 14:19:56 <^7heo> easily* 2017-02-16 14:20:05 no pkg lang manager provides information about native libs that it requires 2017-02-16 14:20:28 <^7heo> so you mean, accessing the dependency graph is gonna be hellish? 2017-02-16 14:20:38 you don’t understand 2017-02-16 14:20:49 you can build dependency graph when you know the dependencies 2017-02-16 14:21:18 <^7heo> is there a follow up to that? :) 2017-02-16 14:21:30 but you usually don’t know what native libs python package A needs from setup.py (to use python as an example) 2017-02-16 14:21:45 for the n+1th time, dependencies should be documented by the package author, not by a language-specific package shitter 2017-02-16 14:22:08 <^7heo> jirutka: yeah, so you mean, we cannot extract the dependency graph from pip (for example) prior to doing the operation, right? 2017-02-16 14:22:18 no, you can, but only for python dependencies 2017-02-16 14:22:20 <^7heo> (that's what I meant earlier) 2017-02-16 14:22:35 <^7heo> yeah so you'll have a partial one. 2017-02-16 14:22:36 but some python packages have native extensions written in C that requires some libs on the system 2017-02-16 14:22:40 <^7heo> I know. 2017-02-16 14:22:50 <^7heo> but I don't see what's different with using pip. 2017-02-16 14:22:52 yes, you can easily solve this for any packages without native deps 2017-02-16 14:23:10 but these can be installed directly using lang pkg manager already 2017-02-16 14:23:12 <^7heo> when you use pip,it already crashes if you don't have those dependencies installed; and it does not do anything to install them. 2017-02-16 14:23:13 and you don’t need apk for it 2017-02-16 14:23:29 sry, very busy now 2017-02-16 14:23:34 and this is for longer discussion 2017-02-16 14:23:40 <^7heo> ok :) 2017-02-16 14:23:45 <^7heo> nah, if I `pip install XXX_C_dep` and I don't have the C dep, pip will just crahs. 2017-02-16 14:23:48 <^7heo> crash* 2017-02-16 14:24:46 <^7heo> but I get what you mean with "It can be technically difficult". You meant "It can be technically difficult to do it right because then we could solve problems that pip isn't able to solve itself." 2017-02-16 15:02:36 i will probably push my bootstrap.sh / gcc/musl changes tomorrow 2017-02-16 15:02:41 simplifying bootstrap a bit 2017-02-16 15:02:55 after that i hope to finish the ghc bootstrap stuff 2017-02-16 15:03:00 mitchty, ^ 2017-02-16 15:03:25 <^7heo> fabled: any take on my question above? 2017-02-16 15:03:39 <^7heo> fabled: since you have a lot of experience with writing apk ;) 2017-02-16 15:03:47 i'm sorry, i need to run now 2017-02-16 15:03:52 please ask again tomorrow 2017-02-16 15:03:57 <^7heo> ok 2017-02-16 16:44:52 fabled gone, hmm 2017-02-16 16:44:57 it was a typo fix :P 2017-02-16 18:12:21 yep, we seem to be in disparate time zones, i was going to ask about some ghc pkg changes 2017-02-16 18:13:37 i found out how debian is splitting out the profiled libraries 2017-02-16 18:13:59 so that should help reduce the overall install size to not be as huge for most people that don't need the profiled debugging libraries 2017-02-16 18:14:39 and then i can make sure the rest get stripped appropriately 2017-02-16 18:15:23 also, for packages from haskell, would that mean I should add support to apk to get the information from cabal and generate an apkbuild? 2017-02-16 18:15:40 i think that's not part of apk 2017-02-16 18:15:48 it can be useful to add to newapkbuild 2017-02-16 18:15:51 i'm curious if so, what happens if/when one package wants version N and another N+M and its incompatible 2017-02-16 18:15:59 which is part of abuild 2017-02-16 18:16:03 ah wrong tool then, whatever does the stuff for perl etc... i'll have a look at that then 2017-02-16 18:16:16 http://git.alpinelinux.org/cgit/abuild/tree/newapkbuild.in 2017-02-16 18:18:11 cool, pretty simple then thanks Shiz 2017-02-16 20:25:41 clandmeter, ncopa is that possible to accept https://github.com/alpinelinux/aports/pull/880 2017-02-16 20:50:23 mitchty: I’m working on a tool for programmatic modification of abuilds, for generating and updating rust, python, …, packages, including proper support for resolving dependencies 2017-02-16 20:57:48 jirutka: ah cool, yeah it was the dependency stuff that I wasn't looking forward to dealing with 2017-02-16 20:57:58 haskell has a LOT of transitive dependencies 2017-02-16 21:01:16 and a note for the upcoming 8.2 release, it'll be able to deal with multiple versions of the same libraries easily, not sure how apk handles that 2017-02-16 21:01:40 sry, must go now; remind yourself later 2017-02-16 21:03:38 no worries, have a good night jirutka 2017-02-16 22:11:31 I am being curious that, since Alpine uses gcc-gcj to bootstrap OpenJDK, what we are going to do when gcc drops gcj ? Actually they did. 2017-02-16 23:04:47 jirutka: I'm working on a php-extension-build helper to avoid mistakes while create and build php extensions. Have you any template I can use for my helper. Or I can send my variant to review. 2017-02-16 23:05:59 vakartel: don’t have any for php, just for python and lua; send me to review 2017-02-16 23:10:25 oke, tomorrow 2017-02-17 02:16:37 jirutka, ncopa Here's a all php extensions (tested everything that can) https://goo.gl/w2qgEz 2017-02-17 02:16:58 andypost: it requires me to login… 2017-02-17 02:17:20 https://goo.gl/w2qgEz 2017-02-17 02:17:26 shared 2017-02-17 02:17:45 better 2017-02-17 02:32:46 jirutka, so mostly all php7* are compatible with 7.1 2017-02-17 04:59:01 anyone knows if Austin Page is here? 2017-02-17 05:02:00 py-cassandra-driver build fails on subpackaging. http://tpaste.us/jeDK 2017-02-17 05:04:18 arch has been defined in _py section and I wonder how it should be handled nowadays. 2017-02-17 05:13:24 still not sure why arch has been defined in _py function. sorry, I just woke up seing that rebuild failed on this :P 2017-02-17 05:14:01 seeing even 2017-02-17 07:14:17 i wonder what is llvm state currently and if it's soon time to look at it as main toolchain or not 2017-02-17 09:13:28 there was a new llvm release 2017-02-17 09:13:42 andypost: nice list! thank you 2017-02-17 10:12:27 boost-python3 is broken 2017-02-17 10:12:40 for some reason it links against python2 2017-02-17 10:12:44 which is weird 2017-02-17 10:14:20 oO 2017-02-17 10:14:38 and I'm still doing rebuild locally 2017-02-17 10:14:53 seems they have same issue on archlinux 2017-02-17 10:14:56 https://git.archlinux.org/svntogit/packages.git/tree/trunk/PKGBUILD?h=packages/boost#n64 2017-02-17 10:46:41 fabled : in scripts/bootstrap.sh, after msg "Cross building base system", we should add gcc, g++, gnat for $TARGET_ARCH, right ? 2017-02-17 10:55:27 chromium crashes again 2017-02-17 10:58:05 fabled: i think i have a possible apk bug 2017-02-17 10:58:30 to reproduce: apk search --exact -r so:libvncclient.so.0 2017-02-17 10:58:58 it should have shown remmina 2017-02-17 10:59:02 but it doesnt 2017-02-17 10:59:39 $ apk info -R remmina | grep libvncclient 2017-02-17 10:59:39 so:libvncclient.so.0 2017-02-17 11:01:07 $ apk dot --errors | grep libvncclient 2017-02-17 11:01:40 should also shown aqemu 2017-02-17 11:36:18 jirutka, http://sprunge.us/YCRS looks ok? 2017-02-17 11:44:53 bummer, ghc was bootstrapped with wrong triplpet 2017-02-17 11:45:26 how to generate a graph of edge main+community repo errors without messing with local repositories: 2017-02-17 11:45:43 docker run --rm -it alpine:edge sh -c "apk update -q --no-progress -U && apk dot --error" | dot -Tsvg > output.svg 2017-02-17 11:49:10 looks md5&sh256 checksums are deprecated, new abuild? 2017-02-17 11:52:14 yeah, we dont generate them anymore 2017-02-17 11:52:25 to avoid cluttering the diffs 2017-02-17 11:52:58 if they are in APKBUILD they will be verified 2017-02-17 11:53:04 we just stop generate them 2017-02-17 11:53:29 <_ikke_> What if you update a package? 2017-02-17 11:53:44 <_ikke_> would apk checksum still update the existing sums? 2017-02-17 11:53:54 will remove md5 and sha256 2017-02-17 11:54:00 <_ikke_> ah ok 2017-02-17 11:54:00 and only add sha512 2017-02-17 11:54:26 md5 is kinda meaningless nowdays 2017-02-17 11:55:15 what we should do though, is make abuild verify gpg signatures if provided 2017-02-17 11:55:47 i wonder what broke chromium 2017-02-17 11:57:21 <^7heo> rust? 2017-02-17 11:57:27 <^7heo> ACTION hides 2017-02-17 11:59:28 could be nss 2017-02-17 12:00:54 <^7heo> (I know it's not ##physics, but really nobody got my joke?) 2017-02-17 12:01:31 apparently not :) 2017-02-17 12:03:26 <^7heo> chromium is a metal, and rust is another name for corrosion; a reaction that usually degrades (or break metals) 2017-02-17 12:04:56 <^7heo> now the joke is: firefox, another browser, has recently got a dependency on rust (lang) while chromium (the browser) has the same name as the element, which is rust resistant (haha) 2017-02-17 12:06:44 <^7heo> so basically it goes down to parallel between chemistry/physics and software ecosystems 2017-02-17 12:07:00 <^7heo> and the politics in browsers 2017-02-17 12:07:11 <^7heo> and derivated languages 2017-02-17 12:10:52 <^7heo> and as a correlated joke: go rust! 2017-02-17 12:25:01 <_ikke_> ^7heo: I did get the joke but just did not respond 2017-02-17 12:25:28 <^7heo> _ikke_: fair enough :P 2017-02-17 13:12:32 hum 2017-02-17 13:12:44 doing ghc cross-compilation is slightly trickier than anticipated 2017-02-17 13:13:08 seems they by default just cross-compile a native compiler in one go 2017-02-17 13:13:17 the stage1 cross-compiler can be done separately 2017-02-17 13:13:22 but it's slightly tricky 2017-02-17 13:13:41 it expects the environment be cross-compiler for target; instead of for the host 2017-02-17 13:14:18 i wonder if i should just not support creating cross-compiler and do the full cross-build as native compiler 2017-02-17 13:19:21 ncopa: apk search --exact -r so:libvncclient.so.0 shows remmina for me 2017-02-17 13:19:52 apk 2.6.8 on armhf 2017-02-17 13:23:36 also works on x86 but show nothing on x86_64 2017-02-17 13:27:55 ScrumpyJack: because i pushed a rebuild for remmina 2017-02-17 13:28:20 problem in repo is solved 2017-02-17 14:51:12 mitchty, where does bootstrap.patch come from? 2017-02-17 15:05:16 mitchty, i'm at http://sprunge.us/cgQa but it still does not work because the cross-compiled native compiler is also prefixed with the target triplet (and paxmarking fails); need to figure out what to do there 2017-02-17 15:07:29 i also note that the previously when no --target was passed the ghc we now have things we are on x86_64-unknown-linux 2017-02-17 15:07:34 instead of -alpine- 2017-02-17 15:09:07 need to run, will continue later. please email me if you have ideas for removing the triplet from cross-compiled ghc 2017-02-17 15:09:22 i note that there was hackery to do it in the docker files of ghc-bootstrap 2017-02-17 15:09:27 bye 2017-02-17 16:01:08 jirutka: pr just sent. 2017-02-17 16:03:25 scadu, stupid question...but 3.6 is retro-compatible with 3.5? 2017-02-17 16:08:34 ffu, I see there are two conflicts. forgot to resolve them :s 2017-02-17 16:10:34 fabled: i wrote it 2017-02-17 16:10:52 you're hitting the same issue the rebootstrap project hits 2017-02-17 16:10:58 and i hacked around in the dockerfile 2017-02-17 16:11:07 fcolista: afaik yes. bigger changes will be introduced in 3.7. sorry, I have shitty mobile connection right now which constantly disconnects. 2017-02-17 16:11:38 ACTION is scared that py3 packages stops to work after upgrade to 3.6 due to possible incompatibility 2017-02-17 16:12:11 it should be in the ghc-bootstrap package 2017-02-17 16:12:42 https://github.com/alpinelinux/aports/blob/master/testing/ghc-bootstrap/bootstrap.patch 2017-02-17 16:12:54 we'd need to split it apart though 2017-02-17 16:13:03 fcolista: we should be fine :p 2017-02-17 16:13:09 it does things that aren't quite right in the name of getting a stage 2 compiler that can build ghc 2017-02-17 16:13:18 scadu, ok 2017-02-17 16:13:39 the way the debian folks are doing things is they generate the stage 1 compiler, then ship that over to the target platforms for compiling on the target 2017-02-17 16:13:49 i was looking to changelog and I don't see changes in the code syntax 2017-02-17 16:14:29 ref https://groups.google.com/forum/#!topic/linux.debian.bugs.dist/q8f-Q59EQq0 2017-02-17 16:15:20 and the hacks i did to get a native musl libc ghc to work were these https://github.com/alpinelinux/aports/blob/master/testing/ghc-bootstrap/Dockerfile.x86_64#L132 2017-02-17 16:15:25 post build rather 2017-02-17 16:15:58 i was going to tackle the makefile rules in the future to figure out where the triples were coming from 2017-02-17 16:16:56 ah he's gone, i should have kept reading 2017-02-17 16:19:50 ACTION & 2017-02-17 17:22:36 fabled: armhf triplet was wrong 2017-02-17 18:16:01 mitchty, right. stage 1 basically the cross-compiler. i was thinking whether or not to package that too. 2017-02-17 18:16:55 mitchty, but can the stage2 by built without the triplet prefix if doing it in separate step? 2017-02-17 18:17:47 fabled: i'm not 100% sure, every time i've tried it it always throws the triple in, and having gone through those makefiles its not been a high priority to fix but it should be doable 2017-02-17 18:18:15 do you think it's easier to fixup the $DESTDIR tree in package() or try to patch the makefile? 2017-02-17 18:18:25 imho, ghc got the cross-compiler stuff wrong 2017-02-17 18:18:41 for now just linking stuff works like i did in the dockerfile 2017-02-17 18:18:43 well, it's just different from what gcc does 2017-02-17 18:18:59 to be honest cross compilation isn't an oft used part of ghc 2017-02-17 18:19:23 i'm being honest when i am saying I got told i was the first to use stage 2 binaries the way I did 2017-02-17 18:19:38 it needs work to be sure 2017-02-17 18:20:01 and for the bootstrap patch, we probably want to split that apart, i rip out a few things to make the bootstrap from glibc easier 2017-02-17 18:20:13 the problem is they tried to simplify the cross-compilation to single step, and broke the other use cases 2017-02-17 18:20:13 aka the terminfo stuff 2017-02-17 18:20:29 i took the one hunk fixing the bootstrap 2017-02-17 18:20:43 i think terminfo works now that i made the makedepends_host correct 2017-02-17 18:21:44 it should, note that only really affects colored terminal output from say the built in cabal stuff 2017-02-17 18:22:04 but its normally there in ghc installs so likely good to keep it in 2017-02-17 18:23:31 i can have a look at the makefiles probably tonight or tomorrow, or see if the debian guys have that sorted out yet 2017-02-17 18:23:41 they hit the same deal with the rebootstrap stuff 2017-02-17 18:24:31 ok 2017-02-17 18:24:36 i'll probably continue on monday 2017-02-17 18:24:40 would love to get ghc done with 2017-02-17 18:24:51 also as a note, i disagree with that madvise patch, i have a better way to fix the EINVAL https://gist.github.com/mitchty/916952e00f0d6f5cc7252ffe136bd27c 2017-02-17 18:25:46 i've been running with it on some older kernels and an rts compiled on MADVISE_FREE kernel/libc, its much less annoying 2017-02-17 18:25:58 i would just try MADV_FREE if it fails with EINVAL; they fallback to using MADV_DONTNEED 2017-02-17 18:26:04 i'll keep poking at it and work on upstreaming it 2017-02-17 18:26:06 permanently fallback that is 2017-02-17 18:26:13 the patch i took was committed upstream already 2017-02-17 18:26:18 but yeah, i agree it's not nice 2017-02-17 18:26:22 yeah i've talked with them about it 2017-02-17 18:26:23 you get 2 syscalls there 2017-02-17 18:28:33 yep, its not a huge problem as its not a hot path, but its an extra 100 or so instructions between user/kernel 2017-02-17 18:30:04 also working on splitting out the profiled libraries into ghc-prof like in debian 2017-02-17 18:30:35 that should help out most peoples install size and let the debugging stuff get installed if/when needed (most don't) 2017-02-17 18:30:42 right 2017-02-17 18:32:09 so the changes so far are simple as http://sprunge.us/URFD 2017-02-17 18:33:01 but needs the normalization of cross-built native compiler binary/path names 2017-02-17 18:44:42 yep looks pretty straightforward 2017-02-17 18:50:43 question for a split function as a sub package, it always seems to run before the primary package 2017-02-17 19:05:17 mitchty, package() should be run first, and the all the split function in the order listed in $subpackages 2017-02-17 19:12:42 ok i'm doing something wrong then, i'll figure it out, i'm probably doing it wrong 2017-02-17 20:59:07 and it was me doing it wrong, as always 2017-02-17 21:08:26 the feeling of pebkac never gets old 2017-02-17 21:16:06 fabled: also as a note, there is work on a new build system for ghc which will replace the makefile stuff, which means I have to probably help out with getting stage 2 builds working since I'm one of the few users 2017-02-17 21:16:09 https://github.com/snowleopard/hadrian 2017-02-18 15:31:11 ncopa: could you please take a look at boost pkg? they has modified the already released tarball, or maybe sourceforge injected some malware into it… 2017-02-18 16:00:28 jirutka: oh, sorry for this one. If it comes to the python3 rebuild -- ncopa's script generated it and I missed it :x 2017-02-18 16:17:47 jirutka: not sure if it was intended. 2017-02-19 05:55:21 looks like pod2man was removed from perl in edge ? in 3.5 repo still has it 2017-02-19 06:05:16 no, it's my bad. it's in perl-dev :) 2017-02-19 14:16:13 what a fail at my side... ncopa, the upstream patch I sent for curl in 3.4-stable is already applied in master (262629094f23) and 3.5-stable (2f1a1212ac05), I just forgot to check that... 2017-02-19 14:17:47 fabled: please be more consistent with cherry-picking stuff to older, yet still supported, AL releases ;) 2017-02-19 14:20:57 I know that 3.4 is security only by now, but these security upgrades introduced serious regression, so I hope there won't be any opposition in fixing that 2017-02-19 14:26:41 I guess I'll have to upgrade AL to 3.5 after all, possibly with v3.5.2 release 2017-02-19 14:51:44 why isn't /sbin/rc a symlink to /sbin/openrc? 2017-02-19 15:10:13 perhaps it's how openrc install it itself? 2017-02-19 15:10:31 (I mean upstream) 2017-02-19 20:54:35 przemoc, ? 2017-02-19 20:57:45 przemoc, oh the curl issue? sorry, that was not me. when i fixed the issue, stables had older curl version; and i did not cherry-pick the upgrades to stable. 2017-02-19 20:58:23 the issue is fixed on master 2017-02-19 20:59:10 see commit 262629094f233d25f4c19ed7a07f556f5763b58d 2017-02-19 21:00:22 also on 3.5-stable 2017-02-19 21:12:15 so i was looking at how the arch guys are doing cabal packages and i'm probably going to update their tool to just dump out apkbuilds as well as pkgbuilds as both build systems are similar enough to not reinvent the wheel 2017-02-20 07:17:37 morning all 2017-02-20 07:33:22 ncopa, fabled: i hadd to apply this patch to build v4l-utils: http://sprunge.us/UIjD 2017-02-20 07:33:26 *had 2017-02-20 09:04:56 fcolista: you should report it upstream 2017-02-20 09:05:28 and you could do: # !defined(TEMP_FAILURE_RETRY) instead of __GLIBC__ 2017-02-20 09:05:34 ncopa, thx. I was scared that this might introduce bugs or security flaws 2017-02-20 09:06:14 no, it looks valid 2017-02-20 09:06:30 are there many places TEMP_FAILURE_RETRY is used? 2017-02-20 09:06:36 only here 2017-02-20 09:07:00 utils/ir-ctl/ir-ctl.c 2017-02-20 09:07:34 you could probably do the loop manually instead of using the macro 2017-02-20 09:07:46 in any case, would porbably be good to report it upstream 2017-02-20 09:09:12 i'm not a C coder, so actually i'm unsure about how to do that. I'll report it to upstream, though. 2017-02-20 09:09:29 how to do it effectively i mean :) 2017-02-20 09:28:16 hm, the most recent ffmpeg upgrade included an soname bump 2017-02-20 09:28:28 but packages depending on ffmpeg weren't rebuild 2017-02-20 10:05:10 ppl who bump libraries should be really carefull. 2017-02-20 10:05:23 nmeum: who bumped it? 2017-02-20 10:27:17 fabled: yeah, as I already wrote in mesg before the one mentioning you explicitly, I realized (too late) it's already fixed in master and 3.5-stable (and even mentioned the sha1s myself). but according to commit dates, curl in 3.4-stable was upgraded to 7.52.1 on 2017-01-09 and you applied the fix in master and 3.5-stable on 2017-01-24, so I dare to say that stables did not have older version when 2017-02-20 10:27:23 you fixed the issue (I presume that your local 3.4-stable branch was simply not up-to-date with remote one). :) 2017-02-20 10:27:26 anyway, thanks for applying 2017-02-20 10:31:18 clandmeter: http://git.alpinelinux.org/cgit/aports/commit/?id=388c5ed7bee76f6d3cf1d9e43e872d2abe7356cb 2017-02-20 10:31:39 there is also already a bug report #6885 2017-02-20 10:32:11 I would have rebuild all effected packages but apk search -r so:... is currently broken isn't it? 2017-02-20 11:12:40 nmeum: why you wrote that `apk search -r ...` is broken? is it a fact? 2017-02-20 11:14:25 > 11:58 <@ncopa> fabled: i think i have a possible apk bug 2017-02-20 11:14:26 > 11:58 <@ncopa> to reproduce: apk search --exact -r so:libvncclient.so.0 2017-02-20 11:14:33 > 11:58 <@ncopa> it should have shown remmina 2017-02-20 11:14:37 > 11:59 <@ncopa> but it doesnt 2017-02-20 11:14:39 that's all I know 2017-02-20 11:14:44 ah, ok 2017-02-20 12:44:02 how can i properly pass the $@ args to a su command? 2017-02-20 12:44:26 seems like su - alpine -c "foo $@" makes su pick up the $2 2017-02-20 12:45:07 su - alpine -c "cd $startdir && abuild $@" 2017-02-20 12:51:04 you cannot, because -c takes only one argument, so you have to quote "$@" expansion yourself within arg passed to -c 2017-02-20 12:51:26 I meant "you cannot easily" 2017-02-20 12:51:57 ncopa: "$@" is only special in that specific form 2017-02-20 12:52:02 what i'd do is 2017-02-20 12:52:16 set -- cd $startdir '&&' abuild "$@" 2017-02-20 12:52:20 s6-setuidgid 2017-02-20 12:52:25 ACTION whistles innocently. 2017-02-20 12:52:49 wait, as przemoc states, -c only takes a single arg 2017-02-20 12:53:00 if it took multiple args, su - alpine -c "$@" with the above thing would work 2017-02-20 13:06:21 nmeum: wasnt that an error with aports intead of apk-tools? 2017-02-20 13:08:36 nmeum: are you sure its the ffmpeg upgrade? 2017-02-20 13:08:40 hm? 2017-02-20 13:09:10 x265 got a bigger upgrade recently 2017-02-20 13:10:18 the mpv error in #6885 claims that it is 2017-02-20 13:10:27 but I haven't look at it closer yet so: no, I am not sure 2017-02-20 13:10:29 *looked 2017-02-20 13:12:43 that bug is not really verbose, better to find out what the error msg is. 2017-02-20 13:13:48 yeah 2017-02-20 13:13:54 does ffmpeg bump soname on patch level changes? 2017-02-20 13:27:19 clandmeter: "sometimes" 2017-02-20 13:29:07 well according to mpv the soname for libavformat and libavutil changed 2017-02-20 13:34:30 clandmeter: 2017-02-20 13:34:32 JEEB │ FFmpeg bumps it randomly 2017-02-20 13:34:34 JEEB │ although usually the issue is that they don't bump :D 2017-02-20 13:34:36 :p 2017-02-20 14:14:31 they bump abi on minor releases? 2017-02-20 14:14:33 wow 2017-02-20 14:14:36 thats painful 2017-02-20 14:15:44 ncopa: stop ignoring me! 2017-02-20 14:16:38 clandmeter: it was a reaction to your comments 2017-02-20 14:58:26 ncopa: and they dont bump when they should, sometimes 2017-02-20 14:58:30 the situation is flaky 2017-02-20 14:59:02 thats why imo a good more safe approach is to rebuild stuff that links against ffmpeg whenever ffmpeg is updated 2017-02-20 15:19:41 Shiz: why not add logic to ffmpeg apkbuild that will check for changes? 2017-02-20 15:20:02 how? 2017-02-20 15:21:21 we are talking about so changes right? 2017-02-20 15:23:00 abi changes 2017-02-20 15:23:08 that doesnt necessarily mean a new api being added or removed 2017-02-20 15:23:17 it can also involve changing the semantics of an existings ymbol 2017-02-20 15:23:47 sure, but i thought the discussion was about soname 2017-02-20 15:24:04 i mean in this particular case. 2017-02-20 15:24:13 i mean, as i see it there's two issues 2017-02-20 15:24:23 1) they bump SONAME on patch versions 2017-02-20 15:24:32 2) they sometimes dont bump SONAME even when abi changes 2017-02-20 15:24:40 right 2017-02-20 15:25:05 i guess you can fix 1) by adding a _soname= to the abuild and checking for it in install() 2017-02-20 15:25:07 1. is easy to solve by keeping a record of it. 2017-02-20 15:25:11 and have it error out if it mismatches so you know when to rebuild 2017-02-20 15:25:53 for abi changes you will need to compare it with prev apk 2017-02-20 15:26:01 realistically, you can't 2017-02-20 15:26:09 since its not like you can trivially check semantic changes 2017-02-20 15:31:20 right, well atleast we can check for soname changes just like we do with checkapk 2017-02-20 16:37:09 Out of curiosity, is there any company offering Alpine support? 2017-02-20 16:44:03 leitao: depends how much you're offering :) 2017-02-20 16:44:48 tdtrask, In fact, I just want to understand how is the support model behind Alpine 2017-02-20 16:45:25 nothing official 2017-02-20 16:52:24 leitao: something, something, free market solutions 2017-02-20 17:14:19 emergency, what can I do when after upgrade the system reboots right after syslinux boot menu? 2017-02-20 17:15:32 verify that kernel is ok 2017-02-20 17:15:50 or just run away before you boss notice 2017-02-20 17:16:16 jirutka: how did you upgrade? 2017-02-20 17:16:22 from alpine v3.4 to v3.5? 2017-02-20 17:16:33 nope, 3.4.? to 3.4.6 2017-02-20 17:16:38 oh 2017-02-20 17:16:44 thats bad :-/ 2017-02-20 17:16:50 what? 2017-02-20 17:17:01 that it fails within stable branch 2017-02-20 17:17:03 should not happen 2017-02-20 17:17:13 i know 2017-02-20 17:17:18 boot from emergency usb 2017-02-20 17:17:31 it’s quite possible that it’s a HW problem… b/c fucking OpenNebula 2017-02-20 17:17:33 and run filesystem check 2017-02-20 17:17:35 this platform is unreliable as hell 2017-02-20 17:18:08 mount the root files system and /boot 2017-02-20 17:18:37 eg mount $rootdev /mnt && mount $bootdev /mnt/boot 2017-02-20 17:18:50 and bind mount /proc: mount --bind /proc /mnt/proc 2017-02-20 17:19:03 then try reinstlal kernel: apk fix --root /mnt linux-grsec 2017-02-20 17:19:44 I’m trying to revert to latest snapshot now 2017-02-20 17:20:22 uff, it boots now 2017-02-20 17:20:39 finally I was able to revert it, b/c it was stuck for 10+ minutes 2017-02-20 17:28:22 I think that it’d be useful to preserve previous kernel and config during upgrade 2017-02-20 17:28:36 so if something went wrong, you can boot the old kernel 2017-02-20 17:29:49 A/B upgrade would be awesome ;3 2017-02-20 17:30:59 btw if anyone tell you to use OpenNebula in production, run away fast 2017-02-20 17:37:39 yes, it would be nice with rollback posibility 2017-02-20 17:37:52 i also think it would be nice with automatic updates 2017-02-20 17:39:25 aanndd lliivveeppaattcchh ;v;v 2017-02-20 17:39:44 I’m afraid that we cannot afford automatic updates now, without proper QA a automatic testing 2017-02-20 17:47:02 yes 2017-02-20 17:47:11 proper QA needs be done first 2017-02-20 17:47:17 and automatic testing 2017-02-20 17:51:06 nmeum: do you use tmux? 2017-02-20 17:51:22 yep, why do you ask? 2017-02-20 17:51:31 have you noticed that first line does not always display output correctly? 2017-02-20 17:51:55 nope 2017-02-20 17:52:07 if i do tmux and try a command with completion 2017-02-20 17:52:15 the first line does not print it correctly 2017-02-20 17:53:16 by first line you mean the first line written by your shell to stdout with the application that match your completion 'pattern'? 2017-02-20 17:53:42 the line at the top of the window 2017-02-20 17:53:47 tmux window 2017-02-20 17:54:14 when the first command i type is at the top of the tmux window, the it is not displayed correctly 2017-02-20 17:54:39 hm, that's strange what terminal emulator are you using? 2017-02-20 17:55:26 ACTION has to go now but will look at the backlog later 2017-02-20 17:57:34 xfce4-terminal 2017-02-20 17:57:53 but it happens from osx terminal too 2017-02-20 17:58:11 is there a way to specify a git repo for an apkbuild? this has zero releases https://github.com/google/bloaty or should i just clone it at build time? 2017-02-20 17:59:40 mitchty: point to specific commit 2017-02-20 17:59:50 Has anything changed in the php7 packages recently? 2017-02-20 18:00:13 mitchty: and made up some version number, I usually use date in format YYYYMMDD in these cases 2017-02-20 18:00:13 I'm suddenly unable to get random_bytes() on a 3.13 kernel, but able to on a 4.4 2017-02-20 18:00:45 mitchty: and also do not forget to write upstream that they should consider making proper releases… 2017-02-20 18:00:59 sounds good, the arch package picked 0.0.0.... https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=bloaty-git#n15 2017-02-20 18:01:30 and theres already an issue thankfully https://github.com/google/bloaty/issues/22 2017-02-20 18:01:55 mitchty: another option is something like 0.0_gitYYYYMMDD 2017-02-20 18:03:17 mitchty: heh, give them some time, it’s Google, they don’t know much about software development… and making releases is very hard, almost rocket science… ah wait… 2017-02-20 18:03:36 jirutka: i expect no traction on that commit tbh :) 2017-02-20 18:03:43 err issue rather 2017-02-20 18:04:00 the tool is nice, if a bit buggy at times 2017-02-20 18:04:06 for the record, I’m not making fun from Google just from this particular case, I see this in Google projects very often… 2017-02-20 18:05:20 jirutka: b-b-b-but you have less problems. you don't have to worry about deploying new release! 2017-02-20 18:09:11 scadu: and even better is no to have any issue tracker, then you don’t have to solve all the bugs that people make in your software! 2017-02-20 18:10:36 \o/ 2017-02-20 18:11:57 jirutka: Google knows a lot about software development. What they're challenged with is interaction with the outside world, tools that aren't internal to Google. When you're working at Google, it's very easy to forget that there actually are other developers outside of the company, and smart ones even, worth interacting with - and forgetfulness happens very fast. 2017-02-20 18:12:16 lol 2017-02-20 18:12:24 this explains a lot 2017-02-20 18:17:40 In case anyone is interested, I tested a docker container with php7 7.0.16 from edge/testing in it on a 3.13 kernel and random_bytes() fails to get sufficient data. The same container works fine on a 4.14 kernel. 2017-02-20 18:19:48 Kruge: not enough entropy? 2017-02-20 18:19:55 That's what it says 2017-02-20 18:20:08 7.0.15 on 3.13 worked ok 2017-02-20 18:21:43 this is probably the same problem as described in http://bugs.alpinelinux.org/issues/6635#note-3 2017-02-20 18:23:20 it’s cased by libressl, fabled reported it in https://github.com/libressl-portable/portable/issues/274 2017-02-20 18:25:29 Well. At least it's a *known* bug. That'll learn me for using testing. 2017-02-20 18:26:59 the workaround/solution(?) for VMs is to add e.g. virtio-rng 2017-02-20 18:27:57 This is docker on top of a physical box, should have sufficient entropy 2017-02-20 18:28:08 don’t know how to deal with that on embedded devices without HW random generator; on physical servers there should be enough entropy 2017-02-20 18:28:13 there should be something early in the boot sequence that blocks until the entropy pool has been initialized, always 2017-02-20 18:28:31 what OS do you have on this physical box? 2017-02-20 18:28:34 skarnet: Agreed. But this box has been up for 12 hours. 2017-02-20 18:28:34 on containers, there should be an option to share the rng with the host's, I think 2017-02-20 18:28:49 jirutka: Ubuntu 14.04, with a 3.13 kernel 2017-02-20 18:28:54 if virtio-rng is the name then that's it 2017-02-20 18:29:16 skarnet: nope, containers share kernel 2017-02-20 18:29:30 skarnet: so I’d think that there should not be any problem with entropy 2017-02-20 18:29:44 if the host has been initialized, then yes indeed 2017-02-20 18:29:47 dinner time, bbl 2017-02-20 18:29:54 virtio-rng cannot be used in containers (obviously) 2017-02-20 18:30:05 php 7.0.15 doesn't have any such problems, just the upgrade to 7.0.16 2017-02-20 18:30:55 have you installed both php versions from edge or v3.5? 2017-02-20 18:31:02 and 7.1.1, too, it seems 2017-02-20 18:31:04 edge 2017-02-20 18:31:22 so they both should be compiled against libressl 2017-02-20 18:31:29 hm, weird, maybe there’s also some bug in php 2017-02-20 18:33:14 7.1.1 fails on a 4.4 kernel 2017-02-20 18:33:34 Just throwing these random (!) testing results out there 2017-02-20 19:16:11 question on provides, is this correct, or do I have to specify the release for those as well? https://github.com/alpinelinux/aports/compare/master...mitchty:ghc-cleanup?expand=1#diff-a8cf59ca3e0622ff4b66c4e7b5e9fce7R36 2017-02-20 19:49:40 also do the provides need to show up in replaces as well like for pkgbuild? 2017-02-20 20:08:39 what the hell is wrong with OpenNTPd?! it has only one job, keep clock in sync… it knows that it’s out of sync and do nothing about it 2017-02-20 20:12:13 jirutka: maybe difference is too big? dunno if there is some magic switch, but if the service is running is should prevent such issue from occurring anyway. 2017-02-20 20:12:44 scadu: exactly; it happened to me on several servers already 2017-02-20 20:13:04 scadu: it’s not that system is started with clock too out-of-sync 2017-02-20 20:13:19 scadu: I’ve fixed clock manually and after a wekk, it was out-of-sync again 2017-02-20 20:13:29 jirutka: lawl 2017-02-20 20:13:31 scadu: so OpenNTPd clearly doesn’t do its job right, but don’t know why 2017-02-20 20:18:33 Feb 19 01:04:45 localhost daemon.info ntpd[2107]: adjusting local clock by -151.431094s 2017-02-20 20:18:33 Feb 19 01:04:45 localhost daemon.info ntpd[2108]: clock is now synced 2017-02-20 20:18:34 Feb 19 08:25:59 localhost daemon.info ntpd[2107]: adjusting local clock by -154.110484s 2017-02-20 20:18:34 Feb 19 08:25:59 localhost daemon.info ntpd[2108]: clock is now synced 2017-02-20 20:18:34 Feb 19 08:30:16 localhost daemon.info ntpd[2107]: adjusting local clock by -154.095786s 2017-02-20 20:18:34 Feb 19 08:30:16 localhost daemon.info ntpd[2108]: clock is now unsynced 2017-02-20 20:18:35 Feb 19 08:34:41 localhost daemon.info ntpd[2107]: adjusting local clock by -154.086257s 2017-02-20 20:18:35 Feb 19 08:34:41 localhost daemon.info ntpd[2108]: clock is now synced 2017-02-20 20:18:36 Feb 19 08:36:18 localhost daemon.info ntpd[2108]: clock is now unsynced 2017-02-20 20:18:36 Feb 20 19:38:43 localhost daemon.info ntpd[2107]: adjusting local clock by -165.072344s 2017-02-20 20:18:37 Feb 20 19:38:43 localhost daemon.info ntpd[2108]: clock is now synced 2017-02-20 20:18:37 Feb 20 20:07:55 localhost daemon.info ntpd[2107]: adjusting local clock by -166.266714s 2017-02-20 20:18:38 Feb 20 20:07:55 localhost daemon.info ntpd[2108]: clock is now unsynced 2017-02-20 20:18:38 what the heck? 2017-02-20 20:19:35 you have another time source auto-updating the clock and conflicting with ntpd? 2017-02-20 20:20:33 don’t think so 2017-02-20 20:21:49 what other time source can I have? 2017-02-20 20:22:12 it’s a VM, but I don’t have any special guest utils 2017-02-20 20:24:47 I don’t understand what’s it doing, ntpctl -s all shows that clock is unsynced and offset is increasing (!), not decreasing 2017-02-20 20:26:08 aha, now it jumped down 2017-02-20 20:28:25 is the vm hosts' time changing? 2017-02-20 20:28:45 its probably saying unsynced due to the jumps, and declares the host clock as unreliable 2017-02-20 20:29:00 I dunno, I don’t have access to hypervisor 2017-02-20 20:29:21 but VM should not see host’s time…? 2017-02-20 20:29:52 might depend, i know vmware with the tools can do that, (and it sucks) 2017-02-20 20:29:52 hm, actually… it uses it to sync when booting/shutting down 2017-02-20 20:30:04 this is KVM 2017-02-20 20:32:13 might want to look at what clock source you have cat /sys/devices/system/clocksource/clocksource0/current_clocksource 2017-02-20 20:32:27 and make sure that jives with what kvm needs 2017-02-20 20:32:47 i had to deal with ntp at the last job, vm's and ntp/time is a rathole 2017-02-20 20:33:21 the actual protocol and algorithms are pretty simple though 2017-02-20 20:33:50 but god help you if your traffic RTT isn't symmetric 2017-02-20 20:34:24 that, was not fun to learn, outbound was like 10ms, inbound like 15ms, that'll ruin everything on its own 2017-02-20 20:34:36 cat /sys/devices/system/clocksource/clocksource0/current_clocksource kvm-clock 2017-02-20 20:34:38 ahaa 2017-02-20 20:35:13 that looks right from what i've used of kvm 2017-02-20 20:35:36 but this means that it uses host’s clock, doesn’t it? 2017-02-20 20:35:55 i think so, been a bit let me look 2017-02-20 20:38:08 yep 2017-02-20 20:39:03 so i'd venture a guess that your hypervisor host is setting the clock via something stupid like ntpdate 2017-02-20 20:39:19 likely due to drift, if its that bad that fast it looks like bad hardware 2017-02-20 20:39:30 available clock sources: kvm-clock hpet acpi_pm 2017-02-20 20:40:13 for slewing as an example you can do 200 jiffies per million to update, if your clock drifts beyond that eventually ntp will give up 2017-02-20 20:40:26 unfortunately I don’t manage these hypervisors and I wouldn’t be surprised if there’s something wrong… 2017-02-20 20:40:45 could try hpet 2017-02-20 20:42:17 wouldn't be my first choice but it looks like lvm-clock syncs with the hypervisor 2017-02-20 20:42:53 is there some proper place where it should be configured on Alpine? 2017-02-20 20:43:24 I’ve looked into hwclock runscript, but it seems that this syncs clock with HW just during boot and shutdown 2017-02-20 20:43:33 i always dumped it in the kernel boot params, clock={tsc,hpet,etc..} 2017-02-20 20:43:51 okay, I’ll try that 2017-02-20 20:44:11 I’ve never needed to change this 2017-02-20 20:44:39 generally you shouldn't need to 2017-02-20 20:44:54 it always just worked, so I don’t know any details about time sync mechanisms 2017-02-20 20:45:14 but when ntp is not syncing time, it indicates either a misconfigured system, or hardware issues 2017-02-20 20:46:34 whoops sorry clocksource= rather, I cannot brain today 2017-02-20 20:47:43 actually i bet it this, the hypervisors clock isn't being synced by ntp 2017-02-20 20:48:04 kvm-clock syncs it periodically to the hypervisor 2017-02-20 20:48:32 hwclock -r shows HW clock time… and yes, it’s out of sync 2017-02-20 20:48:34 and openntp notices and thinks the system clock is jumping by 160seconds each time it wakes up, which for ntp at the extreme end would be 64seconds 2017-02-20 20:49:06 which would make openntp declare the clock a bad ticker and figure its not possible to keep synced after a while 2017-02-20 20:53:27 sounds about right then 2017-02-20 21:03:47 mitchty: okay, thanks a lot for help with troubleshooting! 2017-02-20 21:05:33 aka problems that never happen when you have complete infrastructure under your control… I’ve never had this problem before, but after migrating VMs to faculty’s shitty infra, I’m dealing with one obscure problem after another :/ 2017-02-20 21:06:43 the real question is, why have a ntpd running in a container if the host has synchronized time anyway :P 2017-02-20 21:08:47 it is not a container, but VM 2017-02-20 21:09:35 uh, Travis has some issues… this day is weird 2017-02-20 21:10:50 jirutka: https://www.youtube.com/watch?v=8CPlF-IEkXQ 2017-02-20 21:14:58 if it's a VM, doesn't the vm manager have a way to sync time from the host? 2017-02-20 21:15:20 jirutka: np, good luck on getting it working, sounds like a non fun situation 2017-02-20 21:52:57 skarnet: I dunno, but why? 2017-02-20 21:53:44 because syncing time from an external machine is expensive and inaccurate, whereas transmitting the time from a host to a guest should be fast and accurate? 2017-02-20 21:54:12 especially when host time is wrong… 2017-02-20 21:54:30 well obviously the host should be correctly synced >.> 2017-02-20 21:54:49 I’m using ntp in all VMs for years and never had problem when I managed also hypervisor behind them 2017-02-20 21:55:05 s/behind/under/ 2017-02-20 21:55:06 but imagine a host with 5 guests, I'd rather have the host run 1 NTP client and export its system clock as, say, virtual CMOS to the guests 2017-02-20 21:55:20 than have the host + all guests perform NTP exchanges with a server down the road 2017-02-20 21:55:56 and how exactly would you synchronize time from host system, when using KVM? 2017-02-20 21:56:26 vmWare has some crappy utility for this and as scadu confirmed, it doesn’t work well 2017-02-20 21:58:07 kvm could export the host's hwclock as the guest's hwclock? 2017-02-20 21:58:17 yes 2017-02-21 05:38:33 Hi, looks like even after http://git.alpinelinux.org/cgit/mkinitfs/commit/?id=6ad3bada9a460f52cb208f4855d805a79e3062d2 but I don't quite understand why everytime when I boot, it shows mount /dev/dm-0 invalid argument? 2017-02-21 05:39:13 I tried looking into mkinitfs and check out arch's mkinitramfs and looks like it doesn't work without udev specified. 2017-02-21 05:39:53 One workaround is to mount /dev/sda2 or mount /dev/mapper/root instead of mounting $KOPT_root directly. 2017-02-21 05:40:02 manually* 2017-02-21 05:40:50 But after the root filesystem is mounted in /sysroot, if I unmount it and mount it with mount -t btrfs /dev/dm-0, it works. 2017-02-21 05:41:21 The dmesg log have an additional line when I successfully mounted the /sysroot: 2017-02-21 05:42:07 > BTRFS device fsid ad3b40laso-(UUID-of-block-device) devid 1 transid 7924 /dev/mapper/root 2017-02-21 05:42:57 My setup is btrfs raid 0 on top of luks one of the two raid device. 2017-02-21 05:43:46 I modified the test.sh in alpine-mkinitfs to accomodate to this https://transfer.sh/ApUj6/test.sh (but it shows not enough free space to allocate chunk) 2017-02-21 05:44:28 Is this issue related to nlplug-findfs or mdev (mdev -s does not create /dev/mapper nodes - from gentoo wiki)? 2017-02-21 05:45:53 And if I instead mount -t btrfs /dev/dm-0, it will show something like mount btrfs Invalid Argument. 2017-02-21 05:46:24 My cmdline: BOOT_IMAGE=vmlinuz-grsec root=UUID=ad3b4049-2408-4858-ac0a-c7650827e0f6 modules=btrfs quiet cryptroot=/dev/sdb cryptdm=root rootfstype=btrfs rootflags=compress=lzo,ssd,space_cache,subvolid=258,subvol=/alpine elevator=noop initrd=initramfs-grsec 2017-02-21 08:05:43 pickfire: so you get an emergency shell in initramfs? 2017-02-21 08:05:59 ncopa: Yeah, I get that all the time. 2017-02-21 08:06:05 does the UUID show up with blkid? 2017-02-21 08:06:16 Yes, I can show it here. 2017-02-21 08:06:32 ncopa: https://transfer.sh/fVsjJ/blkid 2017-02-21 08:10:54 is it correct that /dev/sda2 and /dev/mapper/root has the exat same UUID? 2017-02-21 08:11:22 ncopa: I think no. 2017-02-21 08:11:56 ncopa: https://transfer.sh/YpQ63/lsblk 2017-02-21 08:13:51 Looks like /dev/mapper/root is the same as /dev/sda2 but different from /dev/sdb, FYI /dev/sda2 and /dev/mapper/root is the root 2017-02-21 08:14:25 im not familiar with the internals of btrfs, but might make sense if its mirror 2017-02-21 08:14:43 ncopa: No, it's not mirror, it's raid 0 (striping) 2017-02-21 08:14:52 ok 2017-02-21 08:15:23 in memergency shell, does /dev/dm-0 exist? 2017-02-21 08:15:29 Yeah 2017-02-21 08:15:44 is it s ymlink or is /dev/mapper/root a symlink? 2017-02-21 08:15:49 ncopa: Whenever I mount, I just can't mount ${KOPT_root} 2017-02-21 08:15:56 I don't think it's a symlink. 2017-02-21 08:16:25 I mean I can't mount ${KOPT_root} as in block device (eg. /dev/sda2), I can mount the other way around, like /dev/dm-0 or /dev/mapper/root 2017-02-21 08:16:50 so you cannot mount -t btrfs UUID=ad3b4049-2408-4858-ac0a-c7650827e0f6 /sysroot 2017-02-21 08:18:12 ncopa: I think I can't do that as well but I have never tried. 2017-02-21 08:18:41 It's being done in the mounting stage in initramfs-init so no point to do it again (and it's long) 2017-02-21 08:18:43 and if you use root=/dev/mapper/root as boot option? 2017-02-21 08:19:04 i am trying to understand what goes wrong in the initramfs-init 2017-02-21 08:19:07 ncopa: If I do that, I can't mount -t btrfs /dev/mapper/root /sysroot but I can still do the other way around. 2017-02-21 08:19:17 Like mount -t btrfs /dev/sda2 /sysroot 2017-02-21 08:19:29 Or probably mount -t btrfs /dev/dm-0 /sysroot 2017-02-21 08:19:48 how is /dev/mapper/root set up? 2017-02-21 08:19:54 is it a crypt device? 2017-02-21 08:20:17 ncopa: cryptroot=/dev/sdb cryptdm=root 2017-02-21 08:20:43 and that is a part of the btrfs stripe? 2017-02-21 08:20:55 together with sda2? 2017-02-21 08:21:03 and sda2 is not encrypted? 2017-02-21 08:21:30 Yeah, sda2 not encrypted 2017-02-21 08:22:10 sounds like an untraditional setup :) 2017-02-21 08:22:15 Yeah 2017-02-21 08:22:30 /dev/sda2 + /dev/mapper/root (luks /dev/sdb) = /sysroot 2017-02-21 08:23:11 hum ok 2017-02-21 08:23:38 i think that the nlplug-findfs might not handle that correctly 2017-02-21 08:23:43 Why? 2017-02-21 08:24:01 ncopa: nlplug-findfs mounts partition? 2017-02-21 08:24:17 brb 2017-02-21 08:24:34 Okay 2017-02-21 08:24:37 Just ping me 2017-02-21 08:37:33 pickfire: back 2017-02-21 08:37:43 so, the way nlplug-findfs works 2017-02-21 08:37:52 Is 2017-02-21 08:38:27 the only thing it does, is trigger coldplug events and wait for the block device that has the root 2017-02-21 08:38:34 KOPT_root 2017-02-21 08:39:02 i think i've encountered something similar to pickfire's issue on 3.5 2017-02-21 08:39:03 as soon as it finds the specifed filesystem, either by /dev/blah or UUID=... or LABEL=... 2017-02-21 08:39:24 then nlplug-findfs thinks that the job is done 2017-02-21 08:39:25 so 2017-02-21 08:39:37 Shiz: Where? 2017-02-21 08:39:56 Oh 2017-02-21 08:40:02 i'm not sure if it's entirely the same, but my cryptsetup would bork as it couldn't find either the root device or the mapper device to mount root at 2017-02-21 08:40:04 i forgot which of the two 2017-02-21 08:40:10 mounting manually and resuming boot worked fine 2017-02-21 08:40:17 if it finds the uuid of sda2, before it set up the cryptsetup then it will just exit with "i found the filesystem you looked for" 2017-02-21 08:40:22 ncopa: So nlplug-findfs need to coldplug both /dev/sda2 and /dev/mapper/root 2017-02-21 08:40:46 i think the tricky part here is that there are 2 block devices with same UUID 2017-02-21 08:40:56 ncopa: I tried doing it without nlplug-findfs but failed as well, looks like mdev doesn't scan /dev/mapper/* 2017-02-21 08:41:18 Shiz: Yeah, I am manually mounting all the time. 2017-02-21 08:41:25 might be cryptsetup does that /dev/mapper 2017-02-21 08:41:41 Ah, so basically nlplug-findfs is something similar to udev 2017-02-21 08:41:46 yes 2017-02-21 08:42:00 it will trigger the uevents by scanning /sys/*/uevent 2017-02-21 08:42:15 then read the kernel events via netlink socket 2017-02-21 08:42:18 just like udev 2017-02-21 08:42:27 But why don't just use udev? 2017-02-21 08:42:31 it's a temporary device manager + coldplugger 2017-02-21 08:42:44 meant to run in an initramfs 2017-02-21 08:42:44 yes, its only needed to coldplug 2017-02-21 08:42:45 Weird, at first I thought nlplug-findfs is something like mdev + wait. 2017-02-21 08:42:59 after it found the device it will let mdev take over 2017-02-21 08:43:01 udev comes with a whole complex ecosystem that we really don't want to include in an initramfs 2017-02-21 08:43:10 Oh 2017-02-21 08:43:18 yeah udev in an initramfs is no fun 2017-02-21 08:43:46 i sadly cant repo the issue anymore since i replaced the kernel+initramfs with something custom in the meantime 2017-02-21 08:43:48 repro* 2017-02-21 08:44:17 Oh 2017-02-21 08:44:17 but what is a bit unclear to me is if you can do root=/dev/mapper/root 2017-02-21 08:44:32 ncopa: I can do root=/dev/sda2 as well 2017-02-21 08:44:35 i never did that fwiw 2017-02-21 08:44:43 But I don't wanna do that since I need to type longer. 2017-02-21 08:44:52 i did 2017-02-21 08:45:09 you need to type longer? 2017-02-21 08:45:19 actually, i might have done that 2017-02-21 08:45:27 ncopa: Well, I need to manually mount 2017-02-21 08:45:31 hmm 2017-02-21 08:45:40 ok 2017-02-21 08:45:49 And one more thing, /sbin/bootchartd breaks as well. 2017-02-21 08:45:56 ncopa: so root=/dev/mapper/root cryptroot=/dev/sda2 cryptdm=root *should* be the correct setup 2017-02-21 08:45:58 imo 2017-02-21 08:46:07 No 2017-02-21 08:46:11 /dev/sda2 is not encrypted 2017-02-21 08:46:12 in this case no 2017-02-21 08:46:15 uh, right 2017-02-21 08:46:39 pickfire: i think the problem here is that nlplug-findfs needs to wait for 2 different block devices 2017-02-21 08:47:04 ncopa: So now, we need to make nlplug-findfs to accept multiple partition? 2017-02-21 08:47:23 Like nlplug-findfs -flags /dev/sda2 /dev/mapper/root? 2017-02-21 08:47:41 i think so 2017-02-21 08:48:26 i think we should be able to reproduce with 2 or more partitions in stripe setup with either zfs or btrfs 2017-02-21 08:48:36 at least with btrfs 2017-02-21 08:49:10 I am also thinking of adding a new feature, like cryptkey= 2017-02-21 08:49:23 we are working on cryptheader= 2017-02-21 08:49:31 so you can have detached header 2017-02-21 08:49:34 or cryptdata= 2017-02-21 08:50:12 but hum 2017-02-21 08:50:23 Nice, I haven't even seen distribution with those kind of setup. 2017-02-21 08:50:25 it should continu handle events til there are no more events to read 2017-02-21 08:50:31 if you're making a cryptkey= option or something, make sure the private key can't easily be read, like printed at boot time :P 2017-02-21 08:50:46 (stranger things have happened) 2017-02-21 08:51:02 pickfire: do you think you could file a bug report on bugs.a.o? 2017-02-21 08:51:06 skarnet: The private key might be a huge file so printing won't help. 2017-02-21 08:51:08 i will have to go soon 2017-02-21 08:51:13 bugs.a.o? 2017-02-21 08:51:18 bugs.alpinelinux.org 2017-02-21 08:51:22 Okay 2017-02-21 08:51:26 I will do that later. 2017-02-21 08:51:27 #1 2017-02-21 08:51:39 algitbot, do your job! 2017-02-21 08:51:58 explain your btrfs setup, and that you have cryptsetup + stripe 2017-02-21 08:52:05 algitbot is in its stable state: broken 2017-02-21 08:52:19 Okay 2017-02-21 08:52:32 skarnet: i think it will care about #number when number <10 or so 2017-02-21 08:52:35 #100 2017-02-21 08:52:40 #10 2017-02-21 08:52:44 #9 2017-02-21 08:52:49 #42 2017-02-21 08:52:53 algitbot: thanks! 2017-02-21 08:52:58 \o/ 2017-02-21 08:53:05 :) 2017-02-21 08:53:45 fabled: i wanted to ask you about #6329 and #6635 2017-02-21 08:54:24 sigh... no more #1 announcements, the world has gone haywire! 2017-02-21 08:54:26 i wonder if we should compile in JITTERENTROPY into the kernel 2017-02-21 08:54:46 looks like fedora does so 2017-02-21 08:54:55 instead of kernel module 2017-02-21 09:26:21 ncopa, jitterentropy sounds like a good idea 2017-02-21 09:26:46 ncopa, https://lwn.net/Articles/642166/ 2017-02-21 09:30:18 mailx -a chan.tar.xz thamvmk@yahoo.com 2017-02-21 09:30:18 Subject: Chan's images 2017-02-21 09:30:18 Looks like uncle chanaka forgot to download his images. 2017-02-21 09:30:18 . 2017-02-21 10:20:05 i packaged up a few things, mostly related to gnuradio and other sdr things, but they are too ugly to be merged into normal aports, so i made a special repo: aports-ugly: https://github.com/stef/aports-ugly 2017-02-21 10:24:47 i'm happily using rtlsdr sticks, hackrfs and usrps with these under alpine linux 2017-02-21 10:25:19 i hope it will be useful to others even if ugly 2017-02-21 10:49:46 why not try to clean them up? 2017-02-21 10:57:26 looking at the gnuradio apkbuild it doesnt look that bad 2017-02-21 10:59:45 i don't want to argue about what is clean and what is ugly 2017-02-21 11:00:14 for me this works. i'm happy with it. 2017-02-21 11:00:24 and i don't care if it's ugly 2017-02-21 11:01:15 that's what I say to myself every morning in front of the mirror 2017-02-21 11:01:24 :) 2017-02-21 11:01:57 I'm more demanding of my code than of my appearance tho 2017-02-21 11:02:13 tbh i get agitated when confronted with details like spaces vs tabs 2017-02-21 11:04:52 i'd like to avoid all confrontation in this regard. 2017-02-21 11:06:52 people using tabs should be burned at the stake. That's a nice life you have there, it would be a shame if inquisitors were to visit your home 2017-02-21 11:12:31 i disagree, everyone is free to prefer their own vices. live and let live. thus my ugly repo. 2017-02-21 11:13:12 similarly i react to pep8 in python e.g. 2017-02-21 11:14:22 how dare you have personal preferences 2017-02-21 11:30:30 ohoh, you got me... /o\ 2017-02-21 11:36:15 \o/ 2017-02-21 11:48:56 w 25 2017-02-22 00:23:54 Good afternoon - is fabled on by chance (or another mkinitfs dev)? 2017-02-22 05:22:32 *ping* Any devs with repo admin abilities on? It seems busybox is broken in the latest index (on edge). 2017-02-22 05:48:01 <_ikke_> TemptorSent: Should be someone around in a couple of hours (European work time) 2017-02-22 05:49:20 Thanks _ikke_! It looks like something broke with the busybox 1.26.2-r0 build as none of the apks are in the repo but they are in the index. 2017-02-22 05:53:09 <_ikke_> TemptorSent: perhaps an rsync problem 2017-02-22 05:54:17 <_ikke_> TemptorSent: http://dl-cdn.alpinelinux.org/alpine/edge/main/x86_64/busybox-1.26.2-r0.apk 2017-02-22 05:54:19 <_ikke_> This one? 2017-02-22 05:54:38 Perhaps. Hopefully transient, as it's currently bjorking my work on revising alpine-iso 2017-02-22 05:55:05 <_ikke_> That mirror seems to have it 2017-02-22 05:55:11 It appears so. 2017-02-22 05:55:26 Let's see which mirror is broken then :) 2017-02-22 05:55:29 <_ikke_> what mirror are you using? 2017-02-22 05:55:55 looks like it's leaseweb that's fubar. 2017-02-22 05:57:58 <_ikke_> https://mirror.leaseweb.com/alpine/edge/main/x86_64/busybox-1.26.2-r0.apk 2017-02-22 05:58:19 Forcing the main dl-cdn.alpinelinux.org worked. 2017-02-22 05:59:03 And with multiple mirrors listed, it wasn't trying to fail over. 2017-02-22 05:59:22 <_ikke_> Do you know where exactly it is missing? 2017-02-22 06:02:36 No, I'm not sure what it was puking on -- apk upgrade was failing to find the package when mirror was set to leaseweb, but is happy as a clam on the primary cdn. 2017-02-22 06:03:48 It gave a missing error, not a checksum error, but it's possible the file was corrupted. 2017-02-22 06:04:08 Hello fabled. 2017-02-22 06:08:21 @fabled - I'm working on doing a major rework of the alpine-iso build system, including some issues touching mkinitfs, and have found a few issues that I'm hoping you can help address. 2017-02-22 06:14:42 TemptorSent, morning, alpine-iso is deprecated 2017-02-22 06:15:04 TemptorSent, 3.5 and edge use aports.git/scripts/mkimage.sh 2017-02-22 06:15:14 Oh! Um, what has replaced it? 2017-02-22 06:15:37 That handles building the overalys, initfs, and isos? 2017-02-22 06:15:53 it builds isos, overlays etc. it uses mkinitfs to build initfs 2017-02-22 06:16:13 Bloody hell, I wasted a lot of time making alpine-iso usable :) 2017-02-22 06:18:39 Okay, let's see if my modified build environment will still be of any use... 2017-02-22 06:19:12 Waiting for the bits to crawl across the string between the tin-cans in the rain. 2017-02-22 06:21:55 My basic goal is to add the ability to add multiple 'features' to a profile, allowing the building of customized or even one-off isos/discs, usb sticks, and VM images. 2017-02-22 06:22:11 yes, mkimage.sh is much more well suited for that 2017-02-22 06:22:36 In the old alpine-iso system I have it handling zfs, iscsi, nfs, samba, kvm, and xen. 2017-02-22 06:22:37 it's not perfect yet, and there's few pecularities in it. but it's much better than the alpine-iso makefiles 2017-02-22 06:23:09 Yeah, I'm not bad with make, but it got pretty convoluted in places to do very simple things. 2017-02-22 06:24:17 Oh, btw - bug in mkinitfs -- type: bashism - there is a spurious "local" outside a function in the code for "-L" which the shell pukes on 2017-02-22 06:25:06 I speak bash fluently, so it took be a bit of looking to figure out what was going on there. 2017-02-22 06:25:27 i think some of the "local" things have been fixed recently, but if there's any left patches are welcome 2017-02-22 06:28:02 Better yet would probably be just stick bit of code into a function and call it, "list_features()" or some such. 2017-02-22 06:31:44 Hmm, the new build sytem looks more workable... 2017-02-22 06:32:29 Is there a means of stacking profiles easily that I haven't found yet? 2017-02-22 06:32:52 TemptorSent, they are stacked even now 2017-02-22 06:33:04 the profile function just calls the profile function it wants to stack on 2017-02-22 06:33:46 e.g. mkimg.arm.sh: profile_uboot() calls profile_base from mkimg.base.sh 2017-02-22 06:33:56 in fact everything is pretty much on top of profile_base 2017-02-22 06:36:17 Okay, that side of it is easy :) 2017-02-22 06:36:38 The overlays don't appear to stack though? 2017-02-22 06:37:12 oh, no. it only calls a script to create the overlay. we generally ship with minimal overlay if any 2017-02-22 06:38:03 Ahh, got it. 2017-02-22 06:38:42 That's part of what I was implementing, merging overlays to have a ready-to-run system. 2017-02-22 06:40:09 cool 2017-02-22 06:40:22 i would have use for it in few appliances i run too 2017-02-22 06:40:23 I basically dumped each faature into a directory, then provided configuration for which packages to include for modloop, initfs, world, and pkg repo, as well as an overlay base directory and files. 2017-02-22 06:41:41 So my alpine-kvm-host+zfs profile had the kvm-host, zfs, iscsi, nfs, and samba features enabled, with an overlay automatically generated that would start all services by default. 2017-02-22 06:42:03 nice 2017-02-22 06:42:44 And the only packages I had to list explicitly in my profile were the ones for my particular application. 2017-02-22 06:43:40 Took care of fixing the mkinitfs.conf file as well as world file and options to the kernel. 2017-02-22 06:44:36 BTW - is there a way to tell mkinitfs what packages it needs? 2017-02-22 07:01:49 I'm seeing 'kernel_addons', but it's used inconsistently it seems. 2017-02-22 07:02:29 It looks like in some places it's adding modules, while elsewhere (xtables-addons) it's adding a package without a flavor. 2017-02-22 07:06:01 Would it make more sense to simply split out the kernel-specifc (-$kernelflavor) module packages from the support packages and handle them invididually so we don't have the ugliness found in the extended profile? 2017-02-22 07:08:58 TemptorSent, iirc, mkinitfs does not extract any packages itself. it's done in update-kernel via -p, and update-kernel calls mkinitfs 2017-02-22 07:09:40 Okay, that's what I was seeing. 2017-02-22 07:10:52 With the initramfs 'features' being part of mkinitfs, would it perhaps be appropriate to include the package dependencies for each feature so they can be autodiscovered if not autoinstalled? 2017-02-22 07:11:04 yeah, that would be nice 2017-02-22 07:11:39 just add a file $feature.packages if follwing the current scheme. 2017-02-22 07:12:36 yes. ncopa might have additional thoughts... 2017-02-22 07:13:37 I'd be happy help develop the build/provisioning system. 2017-02-22 07:14:26 we would appreciate that. you might want to email alpine-devel or ask here on the design first, so no surprises come like with alpine-iso 2017-02-22 07:14:35 My 'other' distro is funtoo, having started with gentoo back in its infancy. 2017-02-22 07:15:10 Will do -- just a note, the wiki is rather out of date :) 2017-02-22 07:15:41 If nothing else, it knocked a fair bit of rust of my makefile skills :) 2017-02-22 07:16:59 Alright, what I have in mind right off is just setting up a 'features' subdirectory (currently under scripts) that will contain one sub-directory per feature 2017-02-22 07:19:02 each feature gets a script (call it 'feature-$feature.sh' for now), and optionally an overlay directory file-structure (call it apkovl). 2017-02-22 07:20:44 It may also be logical to set up a directory structure for profiles in a similar fashion to make it easier to mix-and-match as appropriate. 2017-02-22 07:23:37 Making things such as xtables-addons, dahdi, snmp, vpn into features and simplifying the actual profiles to a couple of packages and a list of features. 2017-02-22 07:25:22 Allowing the easy creation of meta-features like all-fs or voip-server 2017-02-22 07:27:58 It might be good to consider how to best handle packages that should be installed and auto-started vs those that should be installed, but not configured, or available on media but not installed. 2017-02-22 07:31:49 Would you have ncopa give me a ping please? 2017-02-22 07:33:06 i'll mention this to him. he usually wakes up around this time, or a bit later. he might be here soonish. 2017-02-22 07:33:47 Ahh, okay - I'm in California (PST8PDT) 2017-02-22 07:34:37 Wish I would have known that before, I was burning the midnight oil and up anyway the past couple nights. 2017-02-22 07:41:24 So if I need to have the packages spl-$kernelflavor, zfs-$kernelflavor, and zfs installed before pivot, where do I specify them? All in kernel_addons? 2017-02-22 07:47:31 Hmm, no.. kernel_addons in base seems to handle the $kernelflavor 2017-02-22 07:52:18 Hmm, it doesnt' currenly appear possible to pass additional packages through section_kernels, which seems like a significant problem. 2017-02-22 07:54:08 Am I missing something there? Wouldn't that break anything that requires packages other than the module package to function? 2017-02-22 07:54:22 zfs for instance? 2017-02-22 08:37:56 hi 2017-02-22 08:50:36 howdy 2017-02-22 08:58:55 Good morning ncopa 2017-02-22 09:02:38 I was talking with fabled earlier about some modifications to the iso/img build system. 2017-02-22 09:03:58 ok? 2017-02-22 09:04:06 I spent a couple days hacking the old alpine-iso package into usable shape and added a "features" feature before finding out from fabled that alpine-iso has been depreciated in favor of the new scrips. 2017-02-22 09:05:54 What I'm looking to do is set the profiles to use sets of featurs for the majority of configuration, with application-specific configuration done at the profile level. 2017-02-22 09:06:54 In the old alpine-iso system, I currently have features for zfs, kvm-host, iscsi, nfs, samba, and xen setup. 2017-02-22 09:08:21 It handles both the packages (modloop, initfs, world, repo) as well as a feature-overlay to auto-start services, install base configs, etc. on a feature-by-feature basis. 2017-02-22 09:09:15 Looking at the new abuild/scripts system, it should be very easy to adapt, but for best effect, will require some modifications to the existing scripts and possibly refactoring the profiles. 2017-02-22 09:09:44 And some related changes for mkinitfs. 2017-02-22 09:13:03 Essentially, each feature lives in its own subdirectory of a 'features' directory and contains a configuration script (feature-$feature.sh) and an optional apkovl directory structure (or tar/cpio archive). 2017-02-22 09:16:39 Each feature configures a list of packages to include in the image (modules/initfs/world/local-repo), and optionally initfs features, boot modules, kernel-command-line options, and so forth using hooks. 2017-02-22 09:17:41 ok 2017-02-22 09:17:56 sounds good to me 2017-02-22 09:18:17 Profiles could optionally be moved into their own directory structure as well, and the majority of the packages would be included through features rather than individually. 2017-02-22 09:19:19 The first question I ran into is it currently possible to add a package to the initrd available before boot other than a kernel module? 2017-02-22 09:20:11 It appeared the existing pathway through the sections_kernel is a dead-end for including packages to be used by update-kernel. 2017-02-22 09:20:43 for instance zfs needs zfs-$kf, spl-$kf, and zfs. 2017-02-22 09:21:18 but passing them using kerenel_addons appends the kernel flavor. 2017-02-22 09:22:22 Would it make sense to have separate configuration directivces for modules and initfs packages? 2017-02-22 09:22:54 And how attached are you to the existing naming? 2017-02-22 09:24:24 The change to mkinitfs would be to allow a file specifying packages needed by each initfs feature, allowing discovery and/or autoinstallation 2017-02-22 09:25:04 Also, the option to include a hook script may be desirable in mkinitfs as well. 2017-02-22 09:25:53 If this generally makes sense and you don't see any red flags, I'll start poking at it tomorrow. 2017-02-22 09:26:49 Time for me to get some sleep soon - my eyes are loosing focus. 2017-02-22 09:36:23 sounds good to me 2017-02-22 09:36:31 i think hooks would be nice in initramfs 2017-02-22 09:36:46 you cannot add a "package" to initramfs, but you can add files 2017-02-22 09:37:12 there is a script that will pull in the shared objects for binaries as well 2017-02-22 09:37:40 Alright great, I'll take a deeper look and start tweaking with it in the morning. 2017-02-22 09:37:58 A first cut should be pretty easy. 2017-02-22 09:38:26 ldd each dep from files, right? 2017-02-22 09:58:45 i think hooks would be nice in initramfs 2017-02-22 09:58:53 PLEASE GOD NO 2017-02-22 09:59:11 whenever you add a line of code to initramfs, a kitten dies 2017-02-22 09:59:20 think of the kittens 2017-02-22 10:00:58 I think we're talking mkinitfs for running the hooks, not the ram-disc 2017-02-22 10:03:50 oh, so compile-time hooks? yeah, that's more reasonable 2017-02-22 10:04:06 but that's still complexity that can very quickly get out of hand 2017-02-22 10:05:13 skarnet: I think it would actually help reduce overall complexity and improve maintainabiliy because everything needed for each feature can be encapsulated in one place. 2017-02-22 10:05:39 the devil, as usual, is in the details 2017-02-22 10:05:51 Of course :) 2017-02-22 10:06:19 abstractions are good... as long as they don't leak 2017-02-22 10:06:50 But for instance setting up ZFS very easy with a simple hook. 2017-02-22 10:07:06 "ZFS very easy" -> syntax error 2017-02-22 10:07:15 *lol* 2017-02-22 10:08:37 ZFS is much easier than trying to constantly juggle and reallocate space. 2017-02-22 10:13:00 Essentially, the only thing ending up in the initramfs are the actual files specified by the feature, their deps, and any direct additions or alterations by the hooks at build time. 2017-02-22 10:14:18 TemptorSent: i have actually thinking of a rewrite of initramfs where most things are implemented as plugins 2017-02-22 10:14:37 For instance, modifying the kernel command line, adding a config file, or creating directories/links/devices. 2017-02-22 10:14:56 ncopa: That would fit in nicely with the rest. 2017-02-22 10:15:36 Doing so should be straight-forward and make the code much cleaner. 2017-02-22 10:15:45 i also thought about using something like initramfs from debian, which as support for plugins 2017-02-22 10:16:01 Too heavy IMHO 2017-02-22 10:16:07 ok 2017-02-22 10:16:47 The ability to keep it simple and easily adapted by anyone who needs somethign different is important. 2017-02-22 10:17:43 yup 2017-02-22 10:17:45 ncopa: What about dracut? 2017-02-22 10:18:23 pickfire: its implemented in bash, which means you need bash to be able to upgrade a kernel 2017-02-22 10:18:33 and it relies on udev 2017-02-22 10:18:39 Oh, no. That happens to arch as well. 2017-02-22 10:19:01 dracut is not an option due to that 2017-02-22 10:19:11 Agreed, udev is fine when you need it, but a royal pain when you don't 2017-02-22 10:19:18 ncopa: So now you are not using udev? 2017-02-22 10:19:41 you can opt-in udev if you want 2017-02-22 10:19:48 but its not a requirement 2017-02-22 10:19:54 and, no we dont use it in initramfs 2017-02-22 10:19:59 udev at the pre-pivot stage is rather heavy if you're just running a VM :) 2017-02-22 10:20:09 I tried removing udev from rc-update and use mdev instead, xorg hangs 2017-02-22 10:20:25 xorg needs udev for hotplug 2017-02-22 10:20:27 No wonder setup-xorg-base require udev. 2017-02-22 10:21:08 in theory you can run xorg without udev, but then you cannot hotplug keyboards and mouse 2017-02-22 10:21:51 I'm having fun right now trying to setup someethign to build a single disc with a VM/Storage host, several service VMs, and a workstation VM over the top. 2017-02-22 10:22:43 So in two of those, udev makes sense, but in the service VMs, they'll have fixed resources anyway and even mdev is almost overkill :) 2017-02-22 10:23:17 ncopa: There is https://github.com/slashbeast/mdev-like-a-boss 2017-02-22 10:24:11 Theoretically, nearly all of the commonly-used udev functionality can live in mdev. 2017-02-22 10:24:22 the main issue is libudev 2017-02-22 10:24:43 and the clients that use it, i.e. everything fdo 2017-02-22 10:24:45 exactly. 2017-02-22 10:25:14 in other words, fdo is a royal pita. Who would have thunk? 2017-02-22 10:26:26 TemptorSent: How long does it normally takes for a aport in mailing lists to get merged into upstream? 2017-02-22 10:28:17 pickfire: No idea whatsoever -- I'm new to Alpine. My main distro is funtoo, gentoo before drobins split, debian before that, and all the way back to Slackware on CD from Walnut Creek :) 2017-02-22 10:29:12 TemptorSent: Ah, I am new to Alpine as well. Back then I was using Arch (for quite some time), I have tried Void before switching to Alpine. 2017-02-22 10:30:32 Alpine hits a much needed niche for fast-deployment (binaries), small size, and essentially universal functionality. 2017-02-22 10:31:37 Oh, and actually reasonably up-to-date or bleeding-edge, not stale. 2017-02-22 10:32:47 There's nothing worse than booting into a live environment and having the downloads required to update just as big as the original image file! 2017-02-22 10:37:30 Being able to give someone a file that's possibly less than 100 megs and having that boot and run everything needed to support the application running on remote storage is huge, even better, I can quickly turn the minimal system into a full-fledged environment without things breaking. 2017-02-22 10:40:04 Anyway, I need to kick off and get some sleep. 2017-02-22 10:40:29 Goodnight (er, morning) all, and thanks! 2017-02-22 15:11:30 ncopa: FYI https://github.com/docker/docker/issues/22853#issuecomment-281696430 2017-02-22 15:38:14 FWIW, +1 to what pavlix said. rewriting resolv.conf is a terrible idea, and it should be vm-specific so no bind mount. 2017-02-22 17:37:52 Hi, trying to build a crosscompiler tool chain targeting RPi gen1. Found aports and bootstrap.sh script, but what target triple to use as arg for bootstrap ? 2017-02-22 18:40:26 test 2017-02-22 18:44:07 <^7heo> ahah bahnhof.se 2017-02-22 18:44:20 <^7heo> nice 2017-02-22 18:46:47 ;) 2017-02-22 18:46:56 was thinking the webclient wasnt working 2017-02-22 18:47:10 but perhaps this is the level of activity to expect 2017-02-22 18:48:28 anyone succesfully managed to use the bootstrap script to build a cross compiler ? 2017-02-22 21:16:10 FYI: http://mail.openjdk.java.net/pipermail/announce/2017-February/000221.html 2017-02-22 21:19:49 i'm curious now - OpenJDK already runs on musl/Alpine, no? 2017-02-22 21:21:57 asie: Alpine uses IcedTea 2017-02-22 21:22:01 i think it's not the same 2017-02-22 21:22:13 IcedTea is based on OpenJDK 2017-02-22 21:22:32 but the "upstream" OpenJDK does not have all the patches in place 2017-02-22 21:22:39 ah 2017-02-22 21:22:49 the goal of this project is to have OpenJDK "proper" support musl 2017-02-22 21:22:52 :) 2017-02-22 21:22:53 (and Alpine) 2017-02-22 21:23:01 may be redundant to say, but http://git.alpinelinux.org/cgit/aports/tree/community/openjdk8 2017-02-22 21:23:04 is a good base to look at for patches 2017-02-22 21:23:06 ;p 2017-02-22 21:23:23 thx :) 2017-02-22 21:23:26 like the discussion we had at fosdem 2017-02-22 21:24:27 mikeee_ are you part of that project? 2017-02-22 21:24:53 (i had assumed so by their vaguely resembling name and @*.oracle.com host) 2017-02-22 21:24:57 I'm the one who proposed it (so yes) 2017-02-22 21:25:07 nice 2017-02-22 21:26:57 random question by someone who is mostly unfamiliar with the JDK infrastructure: 2017-02-22 21:27:06 how does OpenJDK relate to the JDK oracle releases on their site? 2017-02-22 21:27:35 (mostly asking because of the prospect of Oracle releasing static musl builds on their site if they were heavily intertwined :p) 2017-02-22 21:27:45 the OracleJDK is to 99.(some amount of 9's)% based on OpenJDK 2017-02-22 21:27:53 any plans to make an oracle musl release? 2017-02-22 21:28:18 that's cool 2017-02-22 21:28:33 glad that project is being undertaken then, good luck with it :) 2017-02-22 21:28:50 this project is for getting the necessary source code changes in place in OpenJDK 2017-02-22 21:29:00 i know there have been questions from the docker community for Oracle support. 2017-02-22 21:29:08 one word of advice: the 'proper' way to fix stuff for musl is by testing for functionality, not identity 2017-02-22 21:29:16 that's explicitly why there is no __musl__ or equivalent macro 2017-02-22 21:29:23 yup, I've noticed that :) 2017-02-22 21:29:51 i also happen to agree with that mindset, but it sometimes confuses peole :p 2017-02-22 21:30:15 adding a check for 'musl <1.1.2' is easier than 'check if function X is broken', i guess 2017-02-22 21:30:27 or 'api Y missing' 2017-02-22 21:30:41 I've only found a single case (related to library loading) where it would be convenient to have a __MUSL__ define or whatever 2017-02-22 21:30:56 most of the changes are just cleaning up the code 2017-02-22 21:30:58 :) 2017-02-22 21:30:59 making more portable 2017-02-22 21:31:17 if it is related to RTLD_LAZY, musl 1.2 should have improvements that are sufficient for JNI 2017-02-22 21:31:22 add an 'it' to the above sentence where it makes sense 2017-02-22 21:31:22 i'm sure the folks at #musl or here would be happy to help you figure out a proper approach if you're having trouble with that one 2017-02-22 21:32:13 I've talked to them a bit about it, and IIUC it's one of those cases where posix doesn't specify the behavior so it's up for discussion how things should be handled 2017-02-22 21:32:24 they (the #musl guys) have been very accommodating though 2017-02-22 21:32:30 and helpful 2017-02-22 21:33:13 right, we are pushing for improvements that should be helpful for java 2017-02-22 21:33:35 right 2017-02-22 21:33:38 in short, the issue in question is how the dynamic loader decides whether an already loaded library is a candidate for fulfilling an (implicity) library dependency 2017-02-22 21:33:59 and specifically how much of the path is used for the string comparison 2017-02-22 21:34:07 the file name only vs. the whole path 2017-02-22 21:34:27 right 2017-02-22 21:34:43 (the improvements are mostly related to making X work when people ignore the config we install, but would fix the java issue too) 2017-02-22 21:35:00 oh, that one. humm 2017-02-22 21:35:36 I could go either way on it personally. we (the java launcher) seems to be relying on undefined behavior 2017-02-22 21:38:35 apart from a bug in musl I already reported, the library loading behavioral difference is the only "hairy" issues I've run into so far, so all in all I'm pretty happy about it all :) 2017-02-22 21:39:54 :) 2017-02-22 21:40:08 is that bug on the ML? 2017-02-22 21:41:27 I submitted a patch and dalias already merged it - w8 2017-02-22 21:41:49 http://git.musl-libc.org/cgit/musl/commit/?id=27b3fd68f67b674440d21ea7ca5cf918d2e1559f 2017-02-22 21:42:00 ah gotcha 2017-02-22 21:42:02 nice 2017-02-22 21:42:04 it's a corner case of library loading failing 2017-02-22 21:51:17 ACTION notes https://github.com/php/php-src/commit/513582814b0ca82d81eb6b98897d745e0f0eebf5 breaks lastest php7 on 3.17 kernels 2017-02-23 02:41:21 ncopa: How do you as a maintainer normally maintain those apacks? 2017-02-23 02:48:17 Is there something like https://github.com/voidlinux/void-packages/blob/1c79acc7969ab2d581c64526b99b45b40424dc05/srcpkgs/vim/update 2017-02-23 02:48:30 https://github.com/voidlinux/void-packages/blob/master/srcpkgs/vim/update 2017-02-23 02:48:43 Or do you normally use RSS feeds and manually check? 2017-02-23 02:52:20 pickfire: there are two services now that check if packages are outdated 2017-02-23 02:52:22 and send emails 2017-02-23 02:52:33 Shiz: Which 2 services? 2017-02-23 02:52:38 pickfire: http://pkgs.alpinelinux.org/flagged 2017-02-23 02:52:40 this shows one of them 2017-02-23 02:52:47 Anitya 2017-02-23 02:52:51 Shiz: But why doesn't it automatically tries to build? 2017-02-23 02:52:54 Anitya? 2017-02-23 02:52:56 i forgot what the other one was called 2017-02-23 02:53:01 https://release-monitoring.org 2017-02-23 02:53:45 Shiz: Ah, that looks interesting. 2017-02-23 02:53:55 So it's not something internal like what void did. 2017-02-23 02:54:02 nope 2017-02-23 02:55:42 Shiz: Where is documentation for APKBUILD? Is it https://wiki.alpinelinux.org/wiki/APKBUILD_Reference#subpackages 2017-02-23 02:56:03 It might be outdated as I saw ${pkgname}diff::noarch 2017-02-23 02:56:10 In vim, not sure what does that do. 2017-02-23 02:56:28 it is outdated, yes 2017-02-23 02:56:36 it just means that that specific subpackage is arch-independent 2017-02-23 02:56:44 it is equivalent to setting arch= within the subpackage function 2017-02-23 02:56:56 Oh 2017-02-23 02:57:08 Shiz: So there is no manual pages for alpine stuff? 2017-02-23 02:57:30 Weird, xxd is included in vim. Shiz should I seperate it? 2017-02-23 02:57:50 well, xxd is part of vim 2017-02-23 02:58:21 i don't think abuild has a man page no, sadly 2017-02-23 02:58:39 i wouldn't be against splitting xxd, personally 2017-02-23 02:58:56 Shiz: What about neovim? 2017-02-23 02:59:07 what about it? it's a separate package, no 2017-02-23 02:59:15 Yeah, but neovim doesn't have xxd. 2017-02-23 02:59:36 then you don't need to split it...? 2017-02-23 02:59:48 :p 2017-02-23 03:00:32 Shiz: Do I need to do abump -R apk? Or just abump apk? 2017-02-23 03:01:39 i set the version manually, myself 2017-02-23 03:01:46 new version -> reset revision to -r0 2017-02-23 03:01:52 Oh 2017-02-23 03:01:54 same version, apkbuild changes -> increment revision 2017-02-23 03:02:20 Shiz: But if some packages changes some needs to be apkgrel right? 2017-02-23 03:02:26 Because they need to be in r1 2017-02-23 03:02:41 sorry, could you rephrase that? 2017-02-23 03:05:54 Shiz: I mean some packages generally needs to be rebuild because their libraries have upgraded. 2017-02-23 03:06:34 right 2017-02-23 03:06:47 i think we only do that if the library soname is bumped though 2017-02-23 03:06:55 or other abi conflicts 2017-02-23 03:10:08 Shiz: I think we are already doing that with abump -R, need to double check. 2017-02-23 03:15:26 Looks like bumping cryptsetup-1.7.3 have the exact same issues that I have when building fcitx. 2017-02-23 03:15:39 +more issues 2017-02-23 03:35:54 Is it just me, or the sha512 sum of boost-1.62.0.tar.bz2 differs from one in boost's APKBUILD ? 2017-02-23 03:47:43 Wow, abump kodi takes like 15 minutes to just install the deps 2017-02-23 03:47:55 tmh1999: Try abuild checksum 2017-02-23 03:53:07 pickfire : Wow thanks :D 2017-02-23 03:53:38 tmh1999: Actually, I noticed that there is a lot of stuff that you can do with abuild and abump but it just isn't documentated in wiki. 2017-02-23 03:53:45 As well as the APKBUILD, totally outdated. 2017-02-23 03:54:19 ncopa: If I get the chance, I would like to help out to improve the speed of the alpine sites, it takes a long time to load. 2017-02-23 03:57:12 I'd like to point out, that all armhf versions on the download page still point to 3.5.0 instead of 3.5.1 2017-02-23 06:31:10 fabled : I was concerned about community/go patch as it fetches binary from my personal repo. 2017-02-23 06:31:28 tmh1999, yes, i did not apply it yet. need to look at that and the gcc one in more detail 2017-02-23 06:31:29 not sure it violates any policy 2017-02-23 06:31:41 yeah 2017-02-23 06:31:43 thanks 2017-02-23 06:32:35 this is how I build the bootstrap go https://github.com/tmh1999/alpine-s390x-repo/blob/master/README.md 2017-02-23 06:33:33 i think we should fix the 'go' aport to support cross compilation 2017-02-23 06:33:48 btw, I build most of packages in main repo, is there any possibilities to host these packages in alpine main repo ? AFAIK, it requires putting my signing keys in alpine-keys 2017-02-23 06:34:46 tmh1999, can you provide a build server for us? 2017-02-23 06:35:02 tmh1999, could email the list? i think several might want to comment that 2017-02-23 06:36:49 Okay I will email the list 2017-02-23 06:37:31 how do you think about support cross compilation for go ? 2017-02-23 06:38:15 go1.4 is pretty old and does not support many arch 2017-02-23 06:38:27 still need go on other arch to bootstrap 2017-02-23 06:38:47 tmh1999: Nice, looks like you had sent a lot of updates to aports. 2017-02-23 06:38:52 Born in 1999? 2017-02-23 06:39:01 pickfire : no :D 2017-02-23 06:39:06 :( 2017-02-23 06:39:08 1999 sounds cool lol 2017-02-23 06:39:25 tmh1999: So you are older or younger than 1999? 2017-02-23 06:39:37 no I'm 23 now :) 2017-02-23 06:39:46 Haha, I am 18/19 :) 2017-02-23 06:39:49 I don't know how to count. 2017-02-23 06:42:59 well who does? 2017-02-23 06:43:48 fabled : If I am not wrong some IBM fellows already contaced ncopa and gave him a build server 2017-02-23 06:45:09 tmh1999, possibly, i remember we spoke few words on it, but can remember how it went 2017-02-23 06:45:23 tmh1999, re: go bootstrap; the apkbuild would need to support cross-compiling the go compiler 2017-02-23 06:45:32 working on 'ghc' to do the same now 2017-02-23 06:45:42 currently i think only gcc package supports that 2017-02-23 06:46:10 Oh you want to use gcc-go ? just like gcc-java. okay got it 2017-02-23 06:46:18 not necessarily 2017-02-23 06:46:44 we do that for java only; gcc-java is cross-built and then used to build openjdk 2017-02-23 06:47:01 that's also problematic since gcc-java is obsoleted and scheduled to be removed 2017-02-23 06:47:20 yeah last time I tried to build gcc-gcj I also see that 2017-02-23 06:48:04 Building stuff on laptop sure is slow. 2017-02-23 06:48:14 Anyway to do like distcc buildfarm? 2017-02-23 06:48:40 especially when you have an intel U series ... 2017-02-23 06:48:56 ghc is nasty, because it requires itself to be built; for it we needed to build initial alpine package on debian docker container with musl compatible toolchain. it was tedious job, but mitchty did good job on it. 2017-02-23 06:49:23 cross-building openjdk7 took me 2 hours, native build took 5 hours, openjdk8 took 4 hour ... 2017-02-23 06:49:36 it's slightly annoying that all the new language try to achieve self-hosting, and write their compilers on the same language. 2017-02-23 06:49:45 correct 2017-02-23 06:49:53 Well, I can use the arm to build for arm. 2017-02-23 06:50:10 pickfire : You are real MVP 2017-02-23 06:50:17 MVP? 2017-02-23 06:50:20 How? 2017-02-23 06:50:53 You da real MVP the meme 2017-02-23 06:56:57 fabled : The linux-vanilla patch not good ? 2017-02-23 06:57:27 tmh1999, it looks good, i just wanted to look at the kernel config in a bit more detail 2017-02-23 06:57:37 i'm doing the "simple" patches first 2017-02-23 06:57:54 Oh right 2017-02-23 06:58:04 thanks 2017-02-23 06:58:16 asciidoc typo fix probably needs pkgrel bump 2017-02-23 06:58:30 pcsc-lite possibly too 2017-02-23 06:58:42 i think pod2man missing affects doc generation 2017-02-23 06:58:57 i can apply them with pkgrel incremented 2017-02-23 06:59:24 yeah better if you just silently do it rather than I pollute the list 2017-02-23 06:59:29 with other patches 2017-02-23 07:00:42 btw what is your timezone ? Just want to adjust my time to adapt to main Alpine dev 2017-02-23 07:01:37 i am on EET, majority of the main devs are on CET 2017-02-23 07:02:03 EET? 2017-02-23 07:02:05 cool thanks 2017-02-23 07:02:14 What EET and what CET? 2017-02-23 07:02:22 google it ... 2017-02-23 07:03:28 Mine is SGT or MYT 2017-02-23 07:11:13 fabled: Not fair, how come my patches not under review? 2017-02-23 07:12:04 pickfire, which ones? i just marked tmh1999's patched that i looked at, but decided i need to look at them in more detail. i'm still going through the backlog 2017-02-23 07:12:26 fabled: By the way, please help me to delete #3208 2017-02-23 07:12:30 No, not that 2017-02-23 07:12:34 %3208 2017-02-23 07:12:50 fabled: Please help me to delete %3028 2017-02-23 07:13:42 done 2017-02-23 07:13:47 youtube-dl is annoying 2017-02-23 07:13:52 they tag every second day 2017-02-23 07:14:21 Ah 2017-02-23 07:14:27 Which one? 2017-02-23 07:14:52 fabled: What is my password for the patchwork thing? 2017-02-23 07:15:34 i don't remember how it works. probably need to create it yourself 2017-02-23 07:15:38 %3030 is something similar to youtube-dl but they uses py3 and is a lot smaller. 2017-02-23 07:15:55 I like you-get more that youtube-dl, it's like 10x smaller. 2017-02-23 07:18:16 tmh1999: https://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package#Commit_your_work 2017-02-23 07:18:33 I have edited it to put a git hook to let others automatically generate the commit message. 2017-02-23 07:20:40 pickfire : Thanks 2017-02-23 07:21:00 Looks like git auto gc everytime I merge, I think time to do a shallow clone. 2017-02-23 07:26:03 fabled: How do I prepare to compile a program to split into like say lxterminal-gtk2 and lxterminal-gtk3? I am sure I can't use something like _py2() and _py3() 2017-02-23 07:27:35 pickfire, not sure, maybe ncopa has comments. we had discussion to add build 'flavors' to abuild so one abuild could would be built multiple times with different settings. but that has not been implemented yet. 2017-02-23 07:28:14 Ah 2017-02-23 07:28:21 pickfire, depends also on how the package supports it 2017-02-23 07:28:30 fabled: With just --enable-gtk3 2017-02-23 07:28:33 On ./configure 2017-02-23 07:28:33 e.g. main/gtk-xfce-engine supports directly building gtk2 + gtk3 from one tree 2017-02-23 07:28:41 Ah 2017-02-23 07:28:55 That's something similar to $python 2017-02-23 07:29:07 It's like python2 install and python3 install 2017-02-23 07:29:42 fabled: I am curious to know why doesn't the packages are split into pkg-doc and pkg-man 2017-02-23 07:30:02 I think this the reason why lots of man pages disappeared in alpine. 2017-02-23 07:33:16 pickfire, we don't have any -man packages; but most was put to -doc package as on embedded installs we don't want man pages or other docs 2017-02-23 07:33:41 pickfire, you can still do 'apk add docs' to get all documentation for installed packages 2017-02-23 07:35:41 Oh, I didn't know about that. 2017-02-23 07:36:13 fabled: I think I wouldn't wanna do that. 2017-02-23 07:36:17 I don't want those /usr/share/doc 2017-02-23 07:36:23 I just want /usr/share/man 2017-02-23 07:37:24 fabled: Actually, that's what I do when cleaning back then in Arch, I just sudo rm -rf /usr/share/doc 2017-02-23 07:37:41 ah 2017-02-23 07:42:29 pickfire : you've run rabbitmq-server on Alpine docker before ? 2017-02-23 07:42:59 No, what's that? 2017-02-23 07:43:25 it's testing/rabbitmq-server 2017-02-23 07:43:33 it's a message queue system 2017-02-23 07:43:42 ok nvm 2017-02-23 07:43:45 Oh shit, now I can't git stash pop 2017-02-23 07:43:52 tmh1999, re %3051 sounds like the right fix is to move the #include inside #if ENABLE_NLS 2017-02-23 07:44:12 okay I will resubmit 2017-02-23 07:44:15 1 min 2017-02-23 07:44:28 technically the the result is same, but after that you should be able to submit it upstream 2017-02-23 07:44:29 strange thing s390x dev didn't do that 2017-02-23 07:44:35 i think they build enable-nls only 2017-02-23 07:44:45 or use glibc only where libintl.h exists always 2017-02-23 07:46:00 Oh no, now I still fail package fcitx, ledger and partclone T_T 2017-02-23 07:46:12 fcitx and cryptsetup (upgrade) have the same issue 2017-02-23 07:46:14 libintl 2017-02-23 07:46:37 Blame gettext (Maintainer: Carlo Landmeter ) 2017-02-23 07:47:08 tmh1999: Oh, so you have the same issue. 2017-02-23 07:48:12 it's complicated than that I think. I remember I read a discussion on musl's list about decision on moving libintl.h in musl rather than using gettext's 2017-02-23 07:49:51 is libintl POSIX? 2017-02-23 07:50:05 tmh1999, musl aport has that comment 2017-02-23 07:50:26 http://git.alpinelinux.org/cgit/aports/tree/main/musl/APKBUILD#n103 2017-02-23 07:50:52 tmh1999, the gcc boehm-gc patch modifies s390 specific stuff only? 2017-02-23 07:51:12 Looks like there is some issues with procps-3.3.12-r1 2017-02-23 07:51:23 How come their packages is in /bin 2017-02-23 07:51:35 Now I have to type /bin/top and /bin/free 2017-02-23 07:51:54 And it's weird that busybox took /usr/bin/topo 2017-02-23 07:52:09 tmh1999, yeah, gcc patch looks safe as it's for s390 specific #ifdeffed area 2017-02-23 07:52:21 /bin/df as well 2017-02-23 07:52:23 fabled : Yeah, it is inside a #ifdef s390 2017-02-23 07:55:49 rabbitmq is the typical example of the difference between "service started" and "service ready" 2017-02-23 07:56:07 because after you start it, it can take ages before getting ready 2017-02-23 07:56:19 Oh 2017-02-23 07:56:26 Something like openrc? 2017-02-23 07:56:50 something openrc doesn't handle :P 2017-02-23 07:57:02 (openrc is a service manager, rabbitmq is a service) 2017-02-23 07:57:03 I think I wanna use it, it takes a lot of time for me to wait for networking. 2017-02-23 07:57:28 I see no reason to wait for networking at boot, it can continue to load after I login. 2017-02-23 07:57:59 that's a problem of serial service managers, yes 2017-02-23 07:58:03 I am not very familiar with openrc but I really hope that those services can run on background after I login or boot is slow. 2017-02-23 07:58:14 not with openrc 2017-02-23 07:58:26 ("not from a Jedi") 2017-02-23 07:59:08 skarnet: What is it related to? 2017-02-23 07:59:15 but my point was that in a service manager, "OK, I just started a service" does not mean "Now I can start other services that depend on the one I just started" 2017-02-23 07:59:39 Or should I manually start those service that I need after I login? 2017-02-23 07:59:51 because there can be an arbitrary amount of time before a service that has just been started is ready to serve queries, typically performed by stuff that depends on it 2017-02-23 08:00:06 and rabbitmq is a typical example of a service that takes a loooong time to get ready 2017-02-23 08:00:13 skarnet: But how come systemd on arch boot so fast? 2017-02-23 08:00:25 systemd boots fast because it cheats 2017-02-23 08:00:39 Haha, I want to know how to cheat as well. 2017-02-23 08:00:42 it starts everything in parallel and assumes everything will get sorted out in the end 2017-02-23 08:00:46 skarnet : not that long, at least last time I wrote some code in it 2 years ago 2017-02-23 08:01:04 tmh1999: afaik it depends on the backlog there is in the queue 2017-02-23 08:01:12 systemd takes like 2s to boot and in alpine it takes like 10s 2017-02-23 08:01:12 but my info is second-hand 2017-02-23 08:01:24 Not including networking in alpine, including networking in arch 2017-02-23 08:01:44 systemd starts everything in parallel 2017-02-23 08:01:46 openrc is serial 2017-02-23 08:01:57 I thought it's parallel as well? 2017-02-23 08:02:01 nope 2017-02-23 08:02:05 skarnet: rc_parallel=YES in my openrc 2017-02-23 08:02:23 it's a hack that works in some cases 2017-02-23 08:02:33 you definitely don't want to use it 2017-02-23 08:02:40 It always work, just the output is a lot ugly. 2017-02-23 08:02:48 it works because you're lucky 2017-02-23 08:02:52 skarnet: I use it, nothing breaks. 2017-02-23 08:03:01 Hehe, I am lucky. ^^ 2017-02-23 08:03:10 exactly 2017-02-23 08:03:27 if it works 99% of the time, it can be a long time before you see it break 2017-02-23 08:03:34 but that doesn't mean it's correct 2017-02-23 08:03:34 skarnet: How do you normally manage wlan0? 2017-02-23 08:04:15 For me, everything after I boot, I ifup wlan0 and then edit /etc/resolv.conf, pretty troublesome 2017-02-23 08:04:31 depends on your definition of "normally", my normalcy is pretty different from other people's 2017-02-23 08:05:01 Ah 2017-02-23 08:05:14 ideally you'd start the dhcp client in parallel after all the local networks have been brought up 2017-02-23 08:05:37 skarnet : actually I don't care much about how fast it is. I just never run rabbitmq-server in a Docker, especially in a openrc one. And it is kind of a pain for me now... 2017-02-23 08:10:14 I think I will re-write some part of rabbitmq-server port 2017-02-23 08:11:17 skarnet, pickfire : btw, what would you do when you start a service inside an openrc Docker container, that openrc is not pid 1 ? 2017-02-23 08:11:41 openrc can never be pid 1 2017-02-23 08:11:51 hum 2017-02-23 08:11:57 I seldom use openrc in docker 2017-02-23 08:12:15 But why not lxc instead of docker? I found docker fat. 2017-02-23 08:12:20 pickfire : then how do you start a service ? 2017-02-23 08:12:30 tmh1999: I don't start services. 2017-02-23 08:12:32 I don't have a choice. People like Docker ... 2017-02-23 08:12:50 pickfire : you da real MVP 2017-02-23 08:13:15 ACTION tried playing around with lxc but can't get networking to work 2017-02-23 08:13:23 tmh1999: google s6-overlay ;) 2017-02-23 08:13:24 ACTION lol 2017-02-23 08:13:34 Well, for now, I just setup my development (arch) on alpine 2017-02-23 08:13:42 Just qemu with kvm 2017-02-23 08:13:57 I tried xen, xen have it's benefit (like better console) 2017-02-23 08:14:23 skarnet : That looks cool ! Will give it a try :) 2017-02-23 08:19:47 pickfire: lxc is actually pretty easy to setup. 2017-02-23 08:20:06 clandmeter: What about lxc with wlan0? 2017-02-23 08:20:10 I need to setup bridge 2017-02-23 08:20:12 T_T 2017-02-23 08:20:32 fabled: did you see that openjdk project? 2017-02-23 08:20:37 Setting up bridge with eth0 is quite easy but not with wlan0 2017-02-23 08:20:56 wlan... i never tried that sorry. 2017-02-23 08:23:21 I like to pronounce the "w" in "wlan" as "woe" 2017-02-23 08:26:13 ACTION woolan is actually a chinese girl name 2017-02-23 08:27:41 tmh1999: You are a chinese? 2017-02-23 08:27:56 pickfire : No 2017-02-23 08:27:59 T_T 2017-02-23 08:28:16 I am still packaging fcitx to allow me to input chinese characters. 2017-02-23 08:28:19 Now I can't type chinese. 2017-02-23 08:28:46 like this 讨 2017-02-23 08:29:09 ACTION typing random chinese character ... 2017-02-23 08:29:38 tmh1999: How did you insert that? Unicode? 2017-02-23 08:30:40 copy from internet :) 2017-02-23 08:34:23 Oh 2017-02-23 08:40:26 Good evening all. 2017-02-23 08:43:31 evening? the sun is just up (although its behind the clouds) 2017-02-23 08:45:46 *lol* It's quarter of one here - technically morning I suppose :) 2017-02-23 09:13:00 clandmeter, how was http://dev.alpinelinux.org/bootstrap/go/go-linux-musl-arm64-bootstrap.tar.bz2 built? 2017-02-23 09:13:34 i think we need to fix go bootstrapping so that it can be cross-compiled from apkbuild 2017-02-23 09:14:01 right, i dont remember as its some time ago. 2017-02-23 09:18:12 clandmeter : you bootstrap it from go/amd64 ? 2017-02-23 09:18:26 i think so 2017-02-23 09:18:44 if I am not wrong, go1.4 (which is the laster version of go support C compiler) does not support arm64 2017-02-23 09:19:33 this is how I bootstrap go/s390x : https://github.com/tmh1999/alpine-s390x-repo/blob/master/README.md 2017-02-23 09:19:39 yes there was something with go1.4, but my memory about it is faint 2017-02-23 09:19:44 go1.7 is the version start supporrt s390x 2017-02-23 09:20:14 that's what I guess from the package, because if you can bootstrap from go1.4, you don't have to pre-build it and put it there 2017-02-23 09:20:34 tmh1999, that looks similar like what i remember i did. 2017-02-23 09:20:48 clandmeter :) 2017-02-23 09:25:02 the problem was our go-boostrap is using 1.4 which does not support aarch64 2017-02-23 09:29:02 yeah that's what I meant 2017-02-23 09:29:06 is go 1.4 the last that can be built without go? 2017-02-23 09:29:15 fabled : correct 2017-02-23 09:29:19 yes 2017-02-23 09:29:39 but go supports cross-compiling, so we just need to fix go package to be cross-compilable for new architectures 2017-02-23 09:29:51 then we can bootstrap them form x86_64 in a scripted way 2017-02-23 09:29:57 fabled : https://golang.org/doc/install/source#go14 2017-02-23 09:30:22 fabled : correct again 2017-02-23 09:30:39 that's what I prefer over gcc-go thing 2017-02-23 09:31:28 one thing I need to ask, the go/amd64 you are talking about is go/glibc or go/musl ? 2017-02-23 09:31:46 oh no 2017-02-23 09:31:50 it's go/musl 2017-02-23 09:31:51 tmh1999, re: https://github.com/tmh1999/alpine-s390x-repo/blob/master/README.md 2017-02-23 09:31:51 sorry 2017-02-23 09:31:58 you have two script sets 2017-02-23 09:32:13 the first probably creates the cross-compiler 2017-02-23 09:32:28 and the second cross-compiles the new native compiler? 2017-02-23 09:32:52 this one is funny. I am not sure the fist is "enough" so I make a native one 2017-02-23 09:33:51 yeah, what you say is right 2017-02-23 09:34:22 ok 2017-02-23 09:34:29 the first one run on amd64 2017-02-23 09:34:34 the second one run on s390x 2017-02-23 09:34:37 seems it's easy to tweak the go aport to support that then 2017-02-23 09:34:49 :) 2017-02-23 09:35:10 *is run 2017-02-23 09:36:09 what are the GOGCCFLAGS for? 2017-02-23 09:36:26 1 minute 2017-02-23 09:37:18 I got it from IBM's go/s390x port a while ago before it was merged into go1.7 2017-02-23 09:37:21 let me check 2017-02-23 09:37:29 but I didn't use it this time 2017-02-23 09:38:01 I *guess* it only makes sense when CGO_ENABLED=1 2017-02-23 09:40:15 Okay, so I build without those *optional* and they are still there 2017-02-23 09:40:30 so, no need to worry about it 2017-02-23 10:37:08 TemptorSent: Good evening 2017-02-23 11:13:55 tmh1999, i'm trying to bootstrap s390x, and native binutils fails to build 2017-02-23 11:14:04 checking for s390x-alpine-linux-musl-gcc... s390x-alpine-linux-musl-gcc 2017-02-23 11:14:04 checking for C compiler default output file name... 2017-02-23 11:14:04 configure: error: in `/home/tteras/aports/main/binutils/src/binutils-2.27': 2017-02-23 11:14:04 configure: error: C compiler cannot create executables 2017-02-23 11:14:04 See `config.log' for more details. 2017-02-23 11:14:05 >>> ERROR: binutils: all failed 2017-02-23 11:15:00 oh 2017-02-23 11:15:03 might be my mistake 2017-02-23 12:14:25 looking at fixing go bootstrap 2017-02-23 12:14:25 i'll probably make go provide go-bootstrap 2017-02-23 12:14:30 so it makedepends on itself 2017-02-23 12:15:20 since i think go needs latest go to do bootstrap for new arch 2017-02-23 12:52:07 hrmpf. Go has same issue as Haskell 2017-02-23 12:52:19 it expects native host build environment, but does cross-compile 2017-02-23 12:53:11 i wonder how to fix that 2017-02-23 12:53:40 when abuild knows it's cross-compiling, it sets up cross-compile environment 2017-02-23 12:53:47 but that's not ok for ghc / go 2017-02-23 12:54:11 because they use gcc for host, and do magic to cross-build ghc/go using itself 2017-02-23 12:54:22 sigh 2017-02-23 13:33:49 farosas: What about rust 2017-02-23 13:33:53 fabled: ^ 2017-02-23 13:34:00 Sorry but bumping the wrong person. 2017-02-23 13:39:20 did not look rust, but i suspect it's the same thing there 2017-02-23 14:00:06 ncopa, seems i need an options flag to request abuild to not setup cross-compiler even though we are cross-compiling... to fix go/ghc... 2017-02-23 14:00:06 any suggestions for the name? 2017-02-23 14:06:48 clandmeter, is the cross-compiled Go a full compiler, or limited one? 2017-02-23 14:10:48 fabled, if i recall correctly its limited. 2017-02-23 14:11:06 the package content looks pretty same 2017-02-23 14:11:50 but its been some time ago, and i didnt make notes :| 2017-02-23 14:13:26 possibly the -tools is missing, i see if it can be cross-compiled also 2017-02-23 14:13:51 seems no 2017-02-23 14:23:35 clandmeter, ncopa, any objection for http://sprunge.us/AHea ? 2017-02-23 14:23:50 my plan is to make go provide go-bootstrap so it can be used to build itself 2017-02-23 14:24:18 <^7heo> fabled: what is that addition for? 2017-02-23 14:24:23 <^7heo> ah 2017-02-23 14:24:27 i have after that follow-up coming 2017-02-23 14:24:30 <^7heo> what kaniini talked about the other day 2017-02-23 14:24:36 <^7heo> go building itself. 2017-02-23 14:24:42 <^7heo> I see. 2017-02-23 14:25:33 <^7heo> go build yourself. ™ 2017-02-23 14:26:26 eventually the go aport will look something like http://sprunge.us/BTDf 2017-02-23 14:26:54 but that needs still a bit of work 2017-02-23 14:26:54 i was able to cross-build go, though 2017-02-23 14:28:48 <^7heo> damn I expected a flat file, not a diff ;) 2017-02-23 14:29:12 <^7heo> but I guess I can obtain that by applying that diff ;) 2017-02-23 14:29:40 just a sec 2017-02-23 14:29:54 http://sprunge.us/HeGM 2017-02-23 14:30:01 <^7heo> ah that's very nice of you. 2017-02-23 14:30:11 <^7heo> I was gonna but I'm currently busy on something else, so that helps a lot :) 2017-02-23 14:30:59 <^7heo> that's quite a big APK tho. 2017-02-23 14:31:32 <^7heo> and half of it is the build section. 2017-02-23 14:31:45 <^7heo> but it looks neta. 2017-02-23 14:31:50 <^7heo> neat even. 2017-02-23 14:32:21 well, the diff tells how little really changed 2017-02-23 14:32:25 most of it was there before... 2017-02-23 14:32:51 need to add some options to abuild to keep native-compiler even when cross-compiling 2017-02-23 14:35:36 <^7heo> yeah 2017-02-23 14:36:09 i'll just push the provides=go-bootstrap thing 2017-02-23 14:37:32 oh 2017-02-23 14:37:47 https://github.com/golang/go/issues/2775 is fixed 2017-02-23 14:37:51 i suppose we can just drop the go src directory ? 2017-02-23 14:41:11 <^7heo> wait we don't have go in main? 2017-02-23 14:41:47 <^7heo> I could have sworn we did. 2017-02-23 14:43:20 <^7heo> fabled: what do you mean by dropping the go src directory? 2017-02-23 14:43:57 <^7heo> from the package? 2017-02-23 14:44:15 check the package function, its mentioned. 2017-02-23 14:45:05 <^7heo> ah, I see. 2017-02-23 14:45:17 <^7heo> yeah I suppose we can drop it. 2017-02-23 14:50:32 pending commit: http://sprunge.us/XFAG 2017-02-23 14:50:44 though i need fix abuild before that 2017-02-23 14:50:52 well, improve abuild 2017-02-23 14:52:04 <^7heo> fabled: btw, about abuild 2017-02-23 14:52:28 <^7heo> fabled: couldn't we invoke operations from other binaries, in some cases? 2017-02-23 14:52:38 <^7heo> like, pip, gems, etc... 2017-02-23 14:53:20 that would be ideal, not sure how easy it is 2017-02-23 14:54:41 need to run 2017-02-23 14:54:41 will continue tm. 2017-02-23 15:00:50 <^7heo> will continue ™ 2017-02-23 16:08:31 fabled: as a note, i'm still looking at getting rid of the triple from the stage 2 compiler 2017-02-23 16:09:12 but, its not as simple as i'd hoped, just annoying that cross compiling means a triple in stage 2, but if you build natively off that its not there, still haven't figured out what expansion is doing it yet 2017-02-23 16:09:48 and the debian guys haven't figured it out yet either so i can't yoink that yet 2017-02-23 16:12:06 mikeee_: nice to see that you are working aginast official support for musl for openjdk 2017-02-23 17:31:33 btw i think we should include adelie's sha256 checksums patch for apk-tools + abuild 2017-02-23 17:31:44 since practical sha1 collision was demonstrated today 2017-02-23 23:06:49 hey friends 2017-02-23 23:06:56 how is everyone? 2017-02-23 23:07:00 Good afternoon. 2017-02-23 23:07:31 Blurry eyed staring at shell scripts :) 2017-02-23 23:07:35 Yourself? 2017-02-23 23:26:57 i am fighting with xcode ... 2017-02-23 23:31:29 That sounds entertaining, at best.. I'm currently working on rewriting most of the mkimage system to support modular profiles, features, and overlays. 2017-02-23 23:38:45 is that the work you wrote about on the mailinglist? 2017-02-24 00:00:30 Nope, I haven't written anything on the mailing list, but I did talk to ncopa and fabled about it. 2017-02-24 00:05:43 ah, sry. i think i mixed you up with someone else on another mailinglist 2017-02-24 06:43:49 Good evening fabled (er, morning for you I suppose?), how are you? 2017-02-24 06:44:09 morning, waking up still and trying to figure go and ghc cross-compilation :) 2017-02-24 06:44:51 Oh joy, you'd better tank up on coffee for that :) 2017-02-24 06:45:43 Anyway, I have much of the build-system rework done. 2017-02-24 06:46:17 Now I need to go back through and figure out how to make it cache under the abuild environment. 2017-02-24 06:46:43 I can't download the kernel every bloody time I want to build an image! 2017-02-24 06:49:12 But other than that it seems to be working properly. 2017-02-24 06:50:16 What are you working on in both go and haskel? 2017-02-24 06:52:11 to cross-build them to new architectures 2017-02-24 06:52:25 TemptorSent, apk caches stuff on /etc/apk/cache by default, just create that directory 2017-02-24 06:52:48 should probably add environment variable to override that 2017-02-24 06:53:14 i have few pending things for apk in the current tree 2017-02-24 06:54:41 fabled: Yeah, it's fine for normal apk usage, but the image builder uses a different root directory and has privlege issues it seems. 2017-02-24 06:55:20 running 'abuild-apk' with a new initdb every time is not exactly efficient. 2017-02-24 06:55:42 It would be very nice to have an environment variable or command line option to select the cache directory. 2017-02-24 06:56:41 Also, it would be VERY nice if the cache directory could also serve as a repository directly -- I think it would just take parsing all the APKINDEX files in the cache directory? 2017-02-24 06:57:18 Are you cross-compiling from x86_64 -> aarch64? 2017-02-24 06:57:53 Or are you having the fun of trying to cross compile for a 32-bit target? 2017-02-24 07:00:09 Seems happy building both xen and virt thus far, have to fix caching before I try anything else! 2017-02-24 07:03:40 TemptorSent, yes, we are working on improved apk also, the repository/cache thing is going to be done better there. but it's a bit longer term project. 2017-02-24 07:04:02 TemptorSent, we have people bootstrapping s390x and ppc64le 2017-02-24 07:04:19 and ghc needs bootstrap to non-x86_64 platforms 2017-02-24 07:05:19 right, i should probably just add an environment variable, or --cache-dir parameter to override the default $apkroot based cache location 2017-02-24 07:16:55 fabled: Yes, that would be immensely helpful, as there is no good reason to download the same package repeatedly if it checksums! 2017-02-24 07:17:01 yes 2017-02-24 07:17:15 i wrote it to my todo when we were doing the scripts 2017-02-24 07:17:18 I'm on a bloody slow connection by modern standards :) 2017-02-24 07:17:22 but somehow never got to it yet 2017-02-24 07:17:33 but i need to make apk release soon, as in this or next week 2017-02-24 07:17:38 will sneak that in too 2017-02-24 07:18:01 Yeah, I know how that goes -- grep -e TODO -e FIXME is my friend :) 2017-02-24 07:18:17 Excellent news! 2017-02-24 07:18:53 Would it be difficult to allow the cache to be valid as a repo directly without having to generate a new index? 2017-02-24 07:22:21 TemptorSent, it's not possible currently. the index is fundamental on how things work 2017-02-24 07:22:37 and to cache we can't just generate new index without signing it 2017-02-24 07:22:46 the plan is to change package and index formats 2017-02-24 07:22:57 so that index on local file system may be unsigned in certain cases 2017-02-24 07:23:15 in which case apk can generate index for the cache easily 2017-02-24 07:24:33 Right, but the APKINDEX files downloaded from the upstream are alreay signed and the files checksummed, correct? 2017-02-24 07:25:37 So I have APKINDEX.xxxxxxxx.tar.gz in my /etc/apk/cache directory, as downloaded from the upstream. 2017-02-24 07:26:03 yes 2017-02-24 07:26:33 cache is still treated specially currently, and it does not work as a repository 2017-02-24 07:26:36 So, at least for those file which already exist, it should be more or less a valid repository? 2017-02-24 07:27:25 yes and no. it can be used without fetching anything online, but it's still not equivalent to repository. 2017-02-24 07:27:32 the difference is subtle, but there 2017-02-24 07:27:47 i want to fix that somehow for next major apk release 2017-02-24 07:27:59 In other words, if I were to grab the APKINDEX.xxxxxxxx.tar.gz file, parse the list of apks and checksums, I should be able to set up a valid repository for THOSE packages at least. 2017-02-24 07:28:39 no 2017-02-24 07:28:45 not without major scripting 2017-02-24 07:29:03 you'd need to rename all files, and rearrange them in the file system 2017-02-24 07:29:10 Right, not trivial to extract, but the data is there. 2017-02-24 07:29:18 yes 2017-02-24 07:30:01 the .apk files are exact copies, and the index file is exact upstream copy. they are just named differently and different path 2017-02-24 07:30:20 So it seems APK should be able to figure out how to treat a cache as a repository, even if it isn't in the same layout. 2017-02-24 07:31:11 you can re-use cache, and pass-in same --repository; and it'll just work 2017-02-24 07:31:18 Mostly I'm looking at provisioning multiple VMs using a single r/w nfs mount or the like. 2017-02-24 07:31:20 the cache discovery is based on the --repository url 2017-02-24 07:31:46 there's few corner cases where sharing the cache can fail (mostly when purging files from it) 2017-02-24 07:32:00 that's another todo item to make shared cache be failsafe 2017-02-24 07:32:12 which is great, except with all the various mirros, each time a new mirror is selected it's invalidated. 2017-02-24 07:32:17 but yes, we have people using shared cach 2017-02-24 07:32:19 yes 2017-02-24 07:32:34 well 2017-02-24 07:32:38 the index gets invalidated 2017-02-24 07:32:42 but packages are valid 2017-02-24 07:33:25 Just a simple fail-over mechanism would be rather useful, so if a package isn't found at the first uri, it will attempt the next mirror 2017-02-24 07:34:31 I got bit by a bad sync the other night and had a broken apk until I changed mirrors 2017-02-24 07:35:19 yeah, we are hoping to fix that also by keeping the old packages a while before deleting them from the repository 2017-02-24 07:35:39 but yes 2017-02-24 07:36:06 mirror handling could be better 2017-02-24 07:36:15 Would it perhaps be worth while to add an abstraction for mirrors, so you could dump a list of mirrors into a file in say /etc/apk/repositories.d? 2017-02-24 07:37:27 Then handle the mirror-set by name rather than url (such as --repository mirror://alpine/edge/main 2017-02-24 07:39:07 kaniini, btw. thanks for catch that armhf- -> armv6- tripplet thingy. i suspect that might fix few packages that were miscompiled for armv7+ 2017-02-24 07:39:14 And iterate the list when checking the cache, leaving it valid if the previous fetched url is listed in the same mirror file as the new one. 2017-02-24 07:40:23 Or at least making it eligable for checksum/sig check. 2017-02-24 07:42:10 morning 2017-02-24 07:48:45 Hmm... no xtables-modules-virtgrsec package... 2017-02-24 07:48:57 Fixed a bug and found a missing package :) 2017-02-24 07:59:38 'morning fcolista 2017-02-24 07:59:51 _0/ 2017-02-24 08:00:09 anyone remembers how to fix chromium-browser crash ? 2017-02-24 08:00:35 fabled: Care to take a look at a patch of what I have so far? 2017-02-24 08:01:02 sorry fcolista, haven't been playing with chromium lately. 2017-02-24 08:01:13 TemptorSent, sure 2017-02-24 08:05:51 fabled: Check your DCC requests :) 2017-02-24 08:13:05 TemptorSent, please send to mailing list, or pastebin link here 2017-02-24 08:13:08 so others can see too 2017-02-24 08:14:01 *lol* That might take a bit more effort :) 2017-02-24 08:15:23 Working on a machine with no links to any of my hosts that have browsers. 2017-02-24 08:17:50 Ahh, how about termbin.com? 2017-02-24 08:19:01 http://termbin.com/895w 2017-02-24 08:20:10 Please excuse the messy git log, I fubarred my checkin sequence so the diffs aren't as clear as they should be. 2017-02-24 08:21:19 Much of the mess is from splitting files apart. I should have done a checkin after copying, before trimming out the non-relevant portions. 2017-02-24 08:22:21 Huh, termbin is pretty cool -- just " | nc termbin.com 9999" 2017-02-24 08:23:47 sounds secure 2017-02-24 08:23:52 I haven't totally torn out the features from the existing profiles yet. 2017-02-24 08:24:15 *lol* about as good as pastebin anyway I suppose :) 2017-02-24 08:24:45 yeah, a few of those services use https at least :) 2017-02-24 08:25:17 for instance, that was the result of "git diff -p master... | nc termbin.com 9999" 2017-02-24 08:25:18 hastebin does for instance... pastebin and sprunge don't, apparently 2017-02-24 08:25:52 Considering the application, I couldn't really care less who sees it. 2017-02-24 08:26:00 it's not about confidentiality 2017-02-24 08:26:17 it's about non-hijacking of the data you sent 2017-02-24 08:26:23 The worst I do is expose my IP 2017-02-24 08:26:49 I don't want an attacker to introduce bugs into code I paste :P 2017-02-24 08:27:18 Right, not something I'd touch for anything that's not meant for eyeballs. 2017-02-24 08:32:50 fwiw the sha256 sum starts 725133 and ends fe1bb2 2017-02-24 08:34:13 Hmm, any idea why there's no gpm package these days? Is it hiding elsewhere? 2017-02-24 08:39:17 Naming conventions are in flux, suggestions welcome. 2017-02-24 08:44:22 wow..^ 2017-02-24 08:44:47 What a mess! 2017-02-24 08:45:30 affects tons of sites :-/ 2017-02-24 08:47:05 because mitming ssl connections was such a good idea in the first place 2017-02-24 08:47:23 indeed 2017-02-24 08:47:42 I think someone missed the point of ssl certs :) 2017-02-24 08:48:05 nobody uses client certs 2017-02-24 08:48:57 No, but the mitm is via root-signed certs intentionally, with no way to validate the originator of the content. 2017-02-24 08:49:13 of course 2017-02-24 08:50:52 The fix would be for CDN/proxies to store an UPSTREAM SIGNED copy of the content then send that over the SSL link, including their own signature so you can maintain cachability. 2017-02-24 08:52:30 Oh, what I wouldn't give for the simple += operator in ash! 2017-02-24 08:52:31 that, or for content creators to make it lightweight enough so they don't need a friggin CDN 2017-02-24 08:52:49 Agreed for the most part. 2017-02-24 08:53:11 CDNs should be used for large static data and not much else IMHO. 2017-02-24 08:53:45 Distributed caching would work better, but there isn't much client-side support that I'm aware of yet. 2017-02-24 08:56:56 ncopa: I'm trying to come up with clear, short naming for the various config variables in the profiles/feature to allow the kernel-flavored packages to be parsed properly as well as handling initfs-required packages vs main system packages. 2017-02-24 08:57:14 ok? 2017-02-24 08:57:30 currently apks/initfs_apks with _flavored appended if they need the kernel flavor. 2017-02-24 08:58:43 It just gets a bit verbose, especially lacking the concat assignment operator in ash. 2017-02-24 09:08:51 ncopa: Can you think of any circumstance when we need to load apks for the initfs that we DON'T want included in the main system apks? 2017-02-24 09:09:35 you want install apks in the initramfs? 2017-02-24 09:10:24 ncopa: They are installed in a temp root to allow mkinitfs to pull the necessary modules and features. 2017-02-24 09:10:58 xtables-addons-$kernelflavor for instance. 2017-02-24 09:11:28 so you are thinking of installing those on the system 2017-02-24 09:11:34 and copy files from running system? 2017-02-24 09:12:15 Currently, all the profiles appear to include the initramfs-required apks in the main system. 2017-02-24 09:12:44 No, I'm working entirely within the scope of the mkimage toolset. 2017-02-24 09:13:07 what is listed as "kernel_addons" is where the ambiguity lies 2017-02-24 09:13:38 oh, yes 2017-02-24 09:13:42 kernel modules 2017-02-24 09:14:05 we dont install them on running system nor do we install the kernel at all when we run from tmpfs 2017-02-24 09:14:13 They actually aren't all kernel modules though, as mkinitfs needs several of the other files for the features. 2017-02-24 09:14:42 you could in theory have the apkovl or modloop on some raid or lvm device 2017-02-24 09:14:50 which after you have booted is no longer used 2017-02-24 09:14:59 take a look at /etc/mkinitfs/features.d/zfs.files for instance. 2017-02-24 09:15:46 yes, you will need zfs on the running system if your root is zfs 2017-02-24 09:15:46 Okay, it seems as though that's an outlier case then. 2017-02-24 09:16:11 there might be strange setups where you dont need install it on the running system 2017-02-24 09:16:14 Right, as well as /usr/sbin/zfs and /usr/sbin/zpool in the initramfs before pivot. 2017-02-24 09:16:36 mdadm also need executable 2017-02-24 09:16:37 and lvm 2017-02-24 09:17:05 Exactly -- which doesn't make a whole lot of sense as part of "kernel_addons" 2017-02-24 09:17:45 Hense the reason I was renaming the variable to initfs_apks, since both the modules and execs are needed 2017-02-24 09:18:29 with initfs_apks_flavored handling the flavor-specific kernel modules (or flavor-specific execs I suppose) 2017-02-24 09:19:29 I just really don't like the "_flavored" part, but I'm not sure what would be both short and still clear. 2017-02-24 09:22:23 xtables_addons is a good example of one that needs both a kerenel-flavored package and a non-flavored package of the same name. 2017-02-24 09:23:07 And adding the flavor at the time of defining the variable breaks stacking (tried that :) ) 2017-02-24 09:25:56 ncopa: would "_kf" be sufficiently clear in place of "_flavored" do you think? 2017-02-24 09:32:26 im not followin sorry 2017-02-24 09:32:33 where do i find kernel_addons 2017-02-24 09:32:45 initfs_apks 2017-02-24 09:33:02 cannot find it in mkinitfs.in? 2017-02-24 09:33:22 neither in initramfs-init.in 2017-02-24 09:33:32 so i dont really understand what you are asking about 2017-02-24 09:34:06 check the base profile 2017-02-24 09:34:12 an in aports/scripts 2017-02-24 09:35:46 the kernel_addons are a way to specify 3rd party kernel modules to be inclued in the boot media 2017-02-24 09:36:00 they may or may nit be needed in the initramfs 2017-02-24 09:36:14 Correct -- in the unmodified tree, it's in mkimg.base.sh 2017-02-24 09:36:19 xtables-addons will not be needed in initramfs 2017-02-24 09:36:26 but you may need it runtime 2017-02-24 09:36:50 those kernel modules may or may not need userspace tools 2017-02-24 09:36:59 in zfs example they need userspace tool 2017-02-24 09:37:23 Right, but you do need zfs and zfs-$kf (and spl-$kv) in the initramfs 2017-02-24 09:37:49 depends on which initramfs 2017-02-24 09:37:58 so specifying it on kernel_addons was impossible. 2017-02-24 09:38:06 you dont need it when you boot from usb or cdrom 2017-02-24 09:38:18 because you get your live system from usb and fat 2017-02-24 09:38:26 because everything kernel_addons saw had the kernel flavor appended. 2017-02-24 09:38:29 so in that case you dont need zfs in the initramfs 2017-02-24 09:38:43 but when you install to disk 2017-02-24 09:38:48 and root is on zfs 2017-02-24 09:39:06 then you will on the disk install includle the zfs userspace tools in the initramfs 2017-02-24 09:39:39 Right, I'm looking at booting off zfs pools or net-booting and mounting the zfs pool, then pivoting. 2017-02-24 09:40:14 yes we have different kernel falvors and i think the idea was that you should be able to build a boot media with multiple kernels 2017-02-24 09:40:23 and that you from boot menu could pick which kernel you wanted 2017-02-24 09:40:36 so you could have a boot usb with both grsec and vanilla kernel 2017-02-24 09:40:42 Actually, booting of CD or USB I still need it if I want to have my actual root on a zfs pool. 2017-02-24 09:41:16 I wholeheartedly agree with that plan :) 2017-02-24 09:41:35 and if you add the xtables-addons you should have the kernel modules for all kernel flavors 2017-02-24 09:41:50 I'm trying to get everything to build with the proper modules and userspace. 2017-02-24 09:42:13 if you boot of usb and have your real rootfs on zfs, then you shoud probably mount /boot on your usb 2017-02-24 09:42:15 Right, I have that theoretically supported now just fine. 2017-02-24 09:42:43 my point is that i think "kernel_addons" makes perfect sense 2017-02-24 09:43:03 I agree, but it doesn't allow for both flavored and unflavored apks to be included 2017-02-24 09:43:28 so zfs is impossible to support directly from what I could figure out. 2017-02-24 09:43:39 where do you include the apk? in the boot repository on the boot media? 2017-02-24 09:44:31 It doesn't end up doing anything but supplying the filesystem in the tmp directory for mkinitfs to use I believe. 2017-02-24 09:44:32 i am not really understand what you are trying to do? 2017-02-24 09:44:55 exactly what problem are you trying to solve? 2017-02-24 09:45:43 you want boot from live-usb that will switch you over to another root that is on zfs? 2017-02-24 09:45:44 For zfs to work, I need to include the packages "zfs zfs-$kf spl-$kf" in the directory supplied to mkinitfs for it's features to pick up and add to the initramfs 2017-02-24 09:46:13 can you please define what you mean with "zfs to work" 2017-02-24 09:46:16 Actually, I'm booting from dvd or a SD card. 2017-02-24 09:47:00 Boot media, start initfs, import zpool, mount zfs fs, pivot root 2017-02-24 09:47:55 is your boot media there mounted as /boot? 2017-02-24 09:48:01 I had it working quite nicely using my higly-altered version of the depreciated alpine-iso. 2017-02-24 09:48:15 Nope 2017-02-24 09:48:34 Initramfs environment only, running from ram. 2017-02-24 09:48:50 it is a custom dvd or sdcard 2017-02-24 09:48:51 Could be PXE boot 2017-02-24 09:49:30 so every time you updated kernel you had to regenerate your boot media? 2017-02-24 09:49:34 Yep, I'm sending the client an iso, telling them to burn it, stick it in th drive, stick a usb stick in the back, and reboot. 2017-02-24 09:50:33 the problem is that /lib/modules/ mus correspond to the kernel that is on your boot media 2017-02-24 09:50:39 must* 2017-02-24 09:50:45 The USB stick can be re-imaged with fresh images as needed once the system is running, and might store overlays. 2017-02-24 09:51:31 Right, I'm building essentially static environments, then running file services and vms on top. 2017-02-24 09:51:47 the problem is that the only task initramfs is supposed to do is finr the rootfs 2017-02-24 09:52:03 so it will only need include the kernel modules and tools needed to set up the rootfs 2017-02-24 09:52:12 once that is done it will exit 2017-02-24 09:52:15 Right, that's all it's doing, bringing up the root pool and pivoting to that. 2017-02-24 09:52:22 then pivot yes 2017-02-24 09:52:34 in this new rootfs 2017-02-24 09:52:41 it needs the rest of the kernel modules 2017-02-24 09:53:10 so the kernel modules found on the /lib/modules must correspond to the kernel it booted from 2017-02-24 09:53:32 Right, that's why generally the packages included to build the initramfs are ALSO included in the live system I believe. 2017-02-24 09:53:54 Right, I'm utilizing the modloop setup to handle that. 2017-02-24 09:54:34 so you have modloop on top of other root filesystem on disk? 2017-02-24 09:55:22 Thus far, I've been leaving the modloop directory mounted with all needed kerenl modules for that kernel version. 2017-02-24 09:55:36 ok that makes sense 2017-02-24 09:56:50 A kernel upgrade looks exactly like building the original image, only with it written to the USB stick (well, unless the box has a good DVD+RW drive that actually does multisession) 2017-02-24 09:57:05 i thikn i understand now 2017-02-24 09:57:29 you replace the boot media sdcard and reboot 2017-02-24 09:57:34 For extra security, I can have burn a fresh iso (sitting on the file server) to disc and swap it in. 2017-02-24 09:57:44 this means that you have a rootfs on disk 2017-02-24 09:57:56 but you dont install the kernel on disk 2017-02-24 09:58:29 and apk upgrade does not upgrade the kernel 2017-02-24 09:58:33 Right, the ZFS contains all my data and optionally my rootfs. 2017-02-24 09:59:24 so you want build a "livecd" which has root="yourdisk" 2017-02-24 09:59:30 interesting setup 2017-02-24 09:59:34 But realistically, I may simply run from the ramdisk and use lbu to manage the configs. 2017-02-24 10:00:05 I need to allow for either option, depending on what my memory and hardware situation is on a given box. 2017-02-24 10:00:25 we dont support that kind of setup (yet) 2017-02-24 10:00:29 When I need to do an update, just drop an entirely new image. 2017-02-24 10:00:34 yup 2017-02-24 10:00:44 but your rootfs is untouched when you replace kernel 2017-02-24 10:00:54 Yeah, I noticed -- that's why I'm making it work. 2017-02-24 10:01:23 if I pivot to the zfs, yes. 2017-02-24 10:02:29 If I run in a ramdisk and simply mount the zfs and load overlays from/to it for configuration, I have absolutely nothing they can fubar :) 2017-02-24 10:03:42 The box I'm working on prototying it for at home is a reasonably powered xeon that's about to get kicked from 8->32gb. 2017-02-24 10:04:51 so in your case 2017-02-24 10:05:01 At that point, I can easily spare a few hundred megs for a fairly well configured server-appliance running in ram, give ZFS 8GB, and have plenty of working room for VMs 2017-02-24 10:05:06 i would think that it should be enough to add 2017-02-24 10:05:51 kernel_addons="zfs spl" 2017-02-24 10:06:04 initrd_features=".... zfs" 2017-02-24 10:06:32 no its initfs_features 2017-02-24 10:06:39 and i think i found a bug 2017-02-24 10:07:32 and kernel_cmdline="rootfstype=zfs root=tank/alpine/root" 2017-02-24 10:08:04 then you would get a boot iso that would boot your specified zfs rootfs 2017-02-24 10:08:55 if you want to run the VM hosting, with the host runs from tmpfs and use the disks for data 2017-02-24 10:09:11 where your rootfs is tmpfs 2017-02-24 10:09:31 and lets say your /var is mounted on zfs 2017-02-24 10:10:21 then you need kernel_addons="zfs spl" and apks=".... zfs" 2017-02-24 10:10:51 once you have set up the live system you can apk add zfs 2017-02-24 10:11:04 and mount the zfs some place 2017-02-24 10:11:07 lbu commit 2017-02-24 10:11:22 add it to /etc/fstab first 2017-02-24 10:11:29 lbu commit 2017-02-24 10:11:34 then i shoudl be mounted on reboot 2017-02-24 10:11:36 Here's what I have currently (as a git diff) http://termbin.com/8i1y 2017-02-24 10:11:47 you dont need the /usr/bin/zfs in the initramfs in that case 2017-02-24 10:11:48 No fstab with zfs 2017-02-24 10:13:32 That sounds about right, except the fact that I'm storing my apkovls on the zfs, so there's a potential catch-22 :) 2017-02-24 10:13:45 ah 2017-02-24 10:13:56 if you put your apkovl on zfs 2017-02-24 10:14:05 then you need it in initramfs 2017-02-24 10:14:38 Basically, all mutable data will be on the ZFS, with boot media image and backup configs on a USB stick. 2017-02-24 10:15:51 Need to be able to recover from a crashed pool (even replacement of all drives) from backups from remote. 2017-02-24 10:16:43 Actually, I basically need to be able to recreate the entire configureation with nothing but the cd and zfs send/receive 2017-02-24 10:17:56 patch that onto a branch and take a look at what I have so far. 2017-02-24 10:18:51 ehy do you split out the profile-*.sh? 2017-02-24 10:18:53 The really fun part is I plan to actually build the vm images and include them in the iso :) 2017-02-24 10:20:03 Mostly to allow for easy modification / derived files at a glance. 2017-02-24 10:20:52 Once everythign is done getting split, the individual files will be easily tracked in git. 2017-02-24 10:21:58 As currently setup, it allows for them to be anywhere in the directory struture, and they can be coalesced if desired. 2017-02-24 10:22:28 whats the point with apks_flavored? 2017-02-24 10:22:59 and initfs_apks and initfs_apks_flavored 2017-02-24 10:23:05 they should not be needed in theory 2017-02-24 10:23:39 the _flavored adds all kernel-flavor version of the package 2017-02-24 10:25:03 cant we just add those manually to apks? 2017-02-24 10:25:10 and the initfs_apks[_flavored] essentially replaced kernel_addons and properly allows for the inclusion of "zfs" in the mkinitfs tmpdir so it can pick up the execs, 2017-02-24 10:25:47 oh 2017-02-24 10:26:04 Not really, at least not if we want to get "zfs zfs-grsec zfs-vanilla zfs-virtgrsec" all in the tmpdir. 2017-02-24 10:26:22 understand 2017-02-24 10:26:29 i think i finally understand the problem 2017-02-24 10:26:55 I suppose we could test for a kernel-flavor tagged version of a package with apk, but that seems like it's going to be fragile at best. 2017-02-24 10:27:44 This allows for completely modular building of the initramfs basically. 2017-02-24 10:27:46 how about this 2017-02-24 10:28:06 we do refactoring (splitting of the profiles-*.sh) in separate diff and commmit 2017-02-24 10:28:08 Each feature specifies what it needs. 2017-02-24 10:28:20 it becomes very difficult re review the new features otherwise 2017-02-24 10:28:42 its very difficult to know what is "moved" code 2017-02-24 10:28:49 unmodified but moved code 2017-02-24 10:28:54 and what is modified code 2017-02-24 10:28:56 Yeah, unfortunately I didn't take a branch after copying the files before editing them. 2017-02-24 10:29:12 this is kind of hopeless to review :-/ 2017-02-24 10:29:38 I might be able to recreate the process? 2017-02-24 10:29:49 fabled: what do you think about splitting the profiles: http://termbin.com/8i1y 2017-02-24 10:31:18 TemptorSent: maybe you should clean up the commit and split it properly before we waste fabled's time to review it 2017-02-24 10:31:30 I also split the actual imagetypes out to their own directory for sanity. 2017-02-24 10:31:42 should be done in different commit 2017-02-24 10:31:57 Any hints on getting git to reorder the commit-log? 2017-02-24 10:32:12 git rebase -i 2017-02-24 10:32:16 I totally bjorked it. 2017-02-24 10:32:36 you can also do git format-patch -o ... 2017-02-24 10:32:46 and then apply them one by one 2017-02-24 10:33:00 do you have multiple commits already? 2017-02-24 10:33:24 if so, then you can do git format-patch --stdout ... | tpaste 2017-02-24 10:33:42 then we can look over the commits one by one 2017-02-24 10:33:49 instead of a huge diff 2017-02-24 10:33:51 Yes, I have it split to multiple commits. 2017-02-24 10:33:59 good 2017-02-24 10:34:02 Okay, lemme try that. 2017-02-24 10:34:45 so i think there is atleast one "bug" here 2017-02-24 10:34:52 in our current scripts 2017-02-24 10:35:19 lets say we have an apkovl on zfs or lvm 2017-02-24 10:35:31 then we need the userspace tool to be included in the initramfs 2017-02-24 10:35:41 and currently it is not possible to do so 2017-02-24 10:35:59 just initfs_features="zfs" will not include the userspace tool in the initramfs 2017-02-24 10:36:07 is that correct? 2017-02-24 10:38:53 Correct as far as I can tell (need to analyze update-kernel further) 2017-02-24 10:39:05 if that i 2017-02-24 10:39:12 The initfs_feature is the part that requires it. 2017-02-24 10:39:20 yes 2017-02-24 10:39:32 if that is the case, can you please report that on bugs.alpinelinux.org 2017-02-24 10:39:42 and how the HELL do I retroactively remove the ISOs I accidentilly commited? 2017-02-24 10:40:03 in git? 2017-02-24 10:40:17 git rebase -i? 2017-02-24 10:40:41 in the list you select "e" or "edit" on the commit with the iso 2017-02-24 10:40:56 then you git rm file.iso and git commit --amend 2017-02-24 10:41:03 and git rebase --continue 2017-02-24 10:42:02 i'd recommend reading a git book or something ;) 2017-02-24 10:44:52 I was trying to filter-branch it, but since I'm in a subdirectory, it balks 2017-02-24 10:49:03 Okay, got filter-branch working... it's going to take a while! 2017-02-24 10:49:22 3 isos worth of blobs to cleanup. 2017-02-24 10:49:32 (the rebase wouldn't have fixed that I don't think) 2017-02-24 10:50:22 Crap, it's going to take HOURS to clean up the mess I made in 2 minutes. 2017-02-24 10:50:32 lol 2017-02-24 10:50:54 i mean 2017-02-24 10:50:59 sorry :-( 2017-02-24 10:51:01 :) 2017-02-24 10:51:02 Oh well, I need to sleep anyway -- it should finish by the morning 2017-02-24 10:51:19 thank you so far 2017-02-24 10:51:22 *lol* Yeah, that'll teach me to get sloppy with git commit -a -m "" 2017-02-24 10:51:23 and good night 2017-02-24 10:51:35 Thank you as well. 2017-02-24 10:51:52 And have a great day. 2017-02-24 10:54:23 1200/47k -- yeah, it's goint to be a while for sure! 2017-02-24 10:55:56 Off to crash. 2017-02-24 14:36:46 mitchty, i think i got ghc cross-compilation done 2017-02-24 14:36:55 clandmeter, and go cross-compilation too 2017-02-24 14:37:06 will likely push sometime today 2017-02-24 14:37:08 nice 2017-02-24 14:37:24 but it needs a bit more testing still, and ghc needs to be updated manually due to changed triplet 2017-02-24 14:37:31 and each compile takes 30mins 2017-02-24 14:48:14 fabled: cool, yeah ghc is not the speediest of compilers 2017-02-24 14:48:32 there are a few flags we could pass in to speed it up a bit but it won't change a lot 2017-02-24 14:49:39 -fmax-simplifier-iterations=not100000 as an example 2017-02-24 14:50:44 also is the cabal package ok for now? I'm working on getting apkbuilds generated for ghc packages using the arch linux tool, I'll replace whats there with what that outputs later 2017-02-24 14:52:30 mitchty, is this ok for ghc: http://sprunge.us/MKXe 2017-02-24 14:53:15 also, re: provides=haskell-* 2017-02-24 14:53:34 are those going to be also real packages, or just ghc provided libraries with a version? 2017-02-24 14:53:56 they're ghc pkgs within ghc itself 2017-02-24 14:54:18 aka base is a pkg that provides the prelude (think type definitions like String) 2017-02-24 14:54:50 ok. then it's good 2017-02-24 14:55:09 i was afraid those exist also as external package that might be overridden 2017-02-24 14:55:29 because provides=foo=$ver will automatically make that package conflict with package named foo 2017-02-24 14:56:10 the only thing that might trip up some ports/triples is the llvm stuff + BuildFlavour = perf-llvm 2017-02-24 14:56:24 we might have to build unregisterised for those 2017-02-24 14:56:43 aka its an old ghc core to c translation 2017-02-24 14:56:52 i think docs said, it does the unregistered build by default for unrecognized triplets 2017-02-24 14:56:55 or similar 2017-02-24 14:57:19 i'm currently re-bootstrapping ghc due to now using the --target 2017-02-24 14:57:22 it caused ghc build id to change, and requires rebootstrap 2017-02-24 14:57:27 but after that it should be good 2017-02-24 14:57:32 ah, ok then it should work on a quick glance 2017-02-24 14:57:45 i did earlier a cross-compile to armhf and it worked 2017-02-24 14:58:00 i've a meeting in 3 minutes though so will have a longer look later 2017-02-24 14:58:16 yep armhf isn't too big of a deal, that one needs llvm for full support as an example 2017-02-24 14:58:18 i'll just push that if it works on my tests 2017-02-24 20:18:16 minor typo https://alpinelinux.org/downloads/ link to the Raspberry Pi is wrong, https://nl.alpinelinux.org/alpine/v3.5/releases/armhf/alpine-rpi-3.5.0-armhf.tar.gz <- 3.5.0 , should be 3.5.1 2017-02-24 21:43:50 https://bugs.alpinelinux.org/issues/6928 on alpine 3.5.1 and 6927 for alpine 3.4.6. 2017-02-25 05:50:12 Anyone aware of whether the linux-virtgrsec works with xtables-addons-grsec? 2017-02-25 05:55:59 it will not 2017-02-25 05:58:18 Okay, thank you -- it appears some logic in the build-system was previously broken then. 2017-02-25 05:58:41 Fixed now, thus throwing a valid error on unavailable package xtables-addons-virtgrsec 2017-02-25 05:59:29 also, is the firmware package really necessary to be downloaded in ALL profiles? 2017-02-25 06:04:28 for -virt* kernels, most likely not 2017-02-25 06:04:45 but i am not certain :) 2017-02-25 06:05:17 Alright, for the moment I've just disabled the inclusion of xtables-addons 2017-02-25 06:06:10 Once I look into fixing mkinitfs itself to handle apk dependencies for its features the problem should be easier to deal with. 2017-02-25 06:07:46 update-kernel should be able to discover which apks it needs based on the initfs features requested. 2017-02-25 06:08:31 yes, would be nice 2017-02-25 06:08:55 we can add capabilities via $provides to select on, actually. 2017-02-25 06:09:02 provides="initfs:foo" 2017-02-25 06:09:09 That's next... for now I have the refactoring of mkimage nearly done for the first cut. 2017-02-25 06:09:42 kernel situation is a mess right now, we need to have a single APKBUILD that builds all the kernels imo 2017-02-25 06:10:08 That may not be possible based on the multiple architechures supported. 2017-02-25 06:10:29 I haven't looked at that at all yet. 2017-02-25 06:10:36 it is possible 2017-02-25 06:10:49 Right now, I'm just trying to get a modular image creation system working 2017-02-25 06:12:03 More to the point, some non-mainline kernels may not be updated regularly and may require additional packages to build. 2017-02-25 06:12:40 So I suppose one very complex APKBUILD could always cover all possiblities, but would probably be a maintainence nightmare. 2017-02-25 06:13:54 Split the common functionality out to a kernel meta-package and allow specific flavors to have their own build including the meta script. 2017-02-25 06:15:46 It's been a while since I had to do any seriously convoluted kernel hacking, but there were enough times I was adding manual patches where I needed to pause the build that it drove me nuts on some other systems. 2017-02-25 06:16:45 (because modifying intermediate-build files in-flight is just SO much fun) 2017-02-25 06:23:56 like i said, i am not sure how to solve it 100% :) 2017-02-25 06:24:51 the kernel situation has other challenges; like "oops, we basically forked grsec" 2017-02-25 06:28:04 TemptorSent: most likely what is needed is a single alpine kernel source tree, with various features enabled/disabled depending on the chosen flavor 2017-02-25 06:28:43 TemptorSent: so for example, common/linux/APKBUILD (and then the other APKBUILDs source from that) 2017-02-25 06:30:23 For mainline and cleanly patched derivitave versions that could work, but for actual custom kernels it would break. 2017-02-25 06:30:58 I'm not sure what the current status of the rpi kernel is for instance. 2017-02-25 06:31:54 Not to mention trying to support a site-specific kernel. 2017-02-25 06:32:27 (as in initramfs built into the kernel itself and a single file with the whole ball of wax for netboot. 2017-02-25 06:40:05 indeed 2017-02-25 06:40:11 i'm not sure what a good solution is 2017-02-25 06:40:24 sounds like you are interested in cleaning up the mess though :) 2017-02-25 06:44:29 awilfox: sooo about that sha256 patch :) 2017-02-25 06:46:02 If it gets in my way, I'll probably fix it :) 2017-02-25 06:56:04 Ouch, wow -- I guess I have more work on mkinitfs than I thought! No NFS support for the initrd could pose a problem of 10. 2017-02-25 06:56:53 And including the scsi module ends up pulling EVERY scsi driver into the initramfs, including random hardware nobody's seen in 10 years I bet! 2017-02-25 06:57:11 That's next, to trim the fat. 2017-02-25 07:01:28 the initrd used to support nfs :) 2017-02-25 07:01:42 if you have questions about that, hl can probably explain it 2017-02-25 07:18:22 nfs is missing from /etc/mkinitfs/features.d at least, which means its modules don't get included on the initramfs (unless they get caught up in a * glob somewhere?) 2017-02-25 07:19:36 Hmm, looks broken at the moment. 2017-02-25 07:19:42 not surprising 2017-02-25 07:21:35 That's going to end up high on my fix-list it looks like. 2017-02-25 07:21:49 It's not currently killing me, but it will be soon! 2017-02-25 08:14:43 Hi, I have some issues installing gnupg after I switched to gnupg1, looks I can't revert back to gnupg2, I need to delete gnupg and then reinstall it to get gnupg2. 2017-02-25 08:15:19 Can I solve this by adding provides="gnupg2" to gnupg aport? 2017-02-25 08:17:14 Not a clue! Hopefully someone that knows the guts better pops on soon. 2017-02-25 08:27:01 I've only been playing with alpine for a week or so total thus far, so about all I know better than I'd like is the initfs and mkimage system. 2017-02-25 08:43:15 Okay, my eyes are about done for -- hopefully fabled and ncopa pop in soon... 2017-02-25 08:45:53 Other than the overlays, I pretty much have the first-cut refactoring done and clean commit logs to review. 2017-02-25 11:06:19 pickfire, we also have replaces= 2017-02-25 23:57:08 seems like chromium is broken on edge ? 2017-02-26 00:28:36 apparently not only on edge but on 3.5 as well... 2017-02-26 04:30:01 jirutka: thanks for catching my fuckup 2017-02-26 04:35:23 Good evening. 2017-02-26 05:02:35 Anyone up for breaking my modifications to the image build system? :) 2017-02-26 07:08:48 coredumb, i remember fcolista mention something on Twitter about PAX. 2017-02-26 07:11:45 chromium package probably needs a rebuild 2017-02-26 09:47:38 kaniini: can someone trigger the rebuild ? 2017-02-26 10:36:50 coredumb, which error do you get? 2017-02-26 10:49:31 clandmeter: an tab crashes 2017-02-26 10:49:32 any* 2017-02-26 10:49:41 even help/settings/anything 2017-02-26 10:49:44 nothing works 2017-02-26 10:49:54 even with a pristine profile 2017-02-26 10:50:04 are you sure its not pax? 2017-02-26 10:50:26 it happeneds after upgrading my edge laptop 2017-02-26 10:50:33 I guess 55.x to 56.x 2017-02-26 10:50:45 I downgraded to 3.5 afterward 2017-02-26 10:50:48 same result 2017-02-26 10:50:58 https://twitter.com/fcolista/status/835088242185801728 2017-02-26 10:51:44 let me try 2017-02-26 10:52:04 pwarf 2017-02-26 10:52:06 just that 2017-02-26 10:52:08 >_< 2017-02-26 10:52:52 it has bee ndisccused here before 2017-02-26 10:53:19 yeah I haven't read here that much lately :D 2017-02-26 10:54:28 thx btw 2017-02-26 10:55:18 we changed the way we handle pax, so it has some outstading issues iirc 2017-02-26 11:01:07 ok 2017-02-26 13:43:09 I'm following https://wiki.alpinelinux.org/wiki/Edge to change my repository from 3.5.1 to edge, but why the alpine-base downgrade from 3.5.1-r0 -> 3.5.0-r0? 2017-02-26 14:50:42 when fabled gets on he might want a look at https://github.com/alpinelinux/aports/pull/935 2017-02-26 14:50:56 i removed a few hacks that are relevant to cross compilation 2017-02-26 19:46:28 Uh... just realized abuild keeps empty folders in packages, is that "new" behaviour? 2017-02-26 21:10:08 humm 2017-02-26 21:10:09 not sure 2017-02-26 21:25:30 . 2017-02-26 21:25:30 It doesn't end up doing anything but supplying the filesystem in the tmp directory for mkinitfs to use I believe[[E7/. 2017-02-26 21:25:52 Woops, dropped kbd :) 2017-02-26 21:27:15 <_ikke_> lol 2017-02-26 21:28:04 Weekends seem to be quiet in alpine-land :) 2017-02-26 21:28:19 <_ikke_> Yup 2017-02-26 21:29:46 I guess I'll just keep plunking away and see what ncopa and fabled think when they turn up. 2017-02-26 21:33:38 weekends are when i break everything 2017-02-26 21:33:51 by committing tons of people's packages 2017-02-26 21:33:53 ;) 2017-02-26 21:33:57 <_ikke_> heh 2017-02-26 21:34:52 kaniini: Nice :) 2017-02-26 21:36:06 Can you help me break my work so I can fix the bugs before the rest of the crowd gets their ands on it then? ;) 2017-02-26 21:36:18 not that courageous :( 2017-02-26 21:36:22 http://termbin.com/t2ed 2017-02-26 21:37:46 I still have a bit of work to do on it obviously, but I think I should have the original functionality working plus the beginnings of the modular system in place. 2017-02-26 21:40:08 does it boot? :D 2017-02-26 21:42:03 pickfire: i applied your patches except the gnupg one, it looks odd to me :) 2017-02-26 21:42:08 So far the images I've tested look good, but I haven't started burn-testing to check functionalities. 2017-02-26 21:43:20 We probably need some sort of regression-testing kit for the build system at some pont. 2017-02-26 21:44:22 i dont have anything readily available that can boot an ISO :( 2017-02-26 21:44:35 <_ikke_> qemu? 2017-02-26 21:45:59 That should work -- in fact, I should probably add a target for it at some point to start the generated image in qemu 2017-02-26 21:46:29 well, other than VMs yeah 2017-02-26 21:47:39 VMs should work too, just need to make sure they're loaded correctly. 2017-02-26 21:48:12 A qemu .img would be a good added output imagetype. 2017-02-26 21:49:10 As would docker, xen-guest, and virtualbox. 2017-02-26 21:56:52 If you take a look at the modified structure, you can see where the imagetypes have been pulled out from the profiles, which should make profiles building multipele image types much easier. 2017-02-26 21:58:41 The two major things I DON'T like about the system: 1. The lack of the append operator in ash leads to a need to restate the previous variable value every time it's appended to. 2. _flavored is excessively long, but I haven't come up with a shorter, clear tag. 2017-02-26 23:23:23 ACTION is trying to create a private repository using 'apk index' and 'abuild-sign', but is getting the error: package mentioned in index not found (try 'apk update') 2017-02-26 23:23:29 anyone seen this before? 2017-02-26 23:24:26 the index only has one package, but the package depends on other packages which are not in the same repo 2017-02-26 23:24:55 but, I would imagine that would work, because we do the same with main/community/testing 2017-02-27 00:43:44 Woah... where's "setvar" in ash? It's SUPPOSED to work according to ash docs... 2017-02-27 00:48:44 ACTION figured out the 'apk index' problem by reading the abuild source 2017-02-27 00:48:54 turns out it was 'noarch' packages that were breaking 2017-02-27 00:49:08 needed to add --rewrite-arch to the 'apk index' command 2017-02-27 00:49:38 turns out the package was in the index, but it was trying to find the apk in the wrong location 2017-02-27 00:49:45 ACTION lives and learns 2017-02-27 00:54:39 Yeah, I hit that once as well. Seems like it should go in it's own repo for sanity sake, especially when trying to build for multiple archs. 2017-02-27 00:56:06 ACTION thinks that's the default setting for apk, but not used in practice 2017-02-27 00:56:42 we end up duplicating the package in the official repos 2017-02-27 00:58:43 Department of Redundancy Department requirements? ;) 2017-02-27 01:08:22 it's a long story :) 2017-02-27 01:10:53 So, any idea where "setvar" went? Does this look like the correct implementaion? setvar() { eval "$1=\"\$2\"" } 2017-02-27 01:11:08 export 2017-02-27 01:11:30 export? 2017-02-27 01:14:00 I *really* don't want to pollute the global namespace, as this is getting called with local variables in places. 2017-02-27 01:49:34 Does anyone have a realy slick way to implement '+=' in ash? 2017-02-27 02:43:16 Okay, full set of list manipuations tools done... 2017-02-27 04:41:02 General variable and list variable handling helpers: http://termbin.com/34qw 2017-02-27 04:41:35 Built for use in mkimage, but probably useful system-wide. 2017-02-27 04:42:31 Code can be shorened significantly with a couple of case statements perhaps. 2017-02-27 04:46:04 Helper function creates aliases for specified variable to make syntax clean in code. 2017-02-27 04:46:45 anyone could help me test this https://gist.github.com/dlintw/304426e3e134ef99772cc7b7639216e0 2017-02-27 04:47:03 APKBUILD failed, but I can not found any error 2017-02-27 04:47:44 For instance: var_list_alias apks; add_apks vim zfs 2017-02-27 04:49:05 Taking a glance... 2017-02-27 04:51:19 Does qmake/cmake spit out anything? 2017-02-27 10:37:18 why the default alpine-sdk doesn't including automake & autoconf? 2017-02-27 10:37:39 Is there any tool like 'namcap' in Arch? 2017-02-27 11:14:42 Has anyone else run into issues trying to pipe through grep in shell scripts? 2017-02-27 11:15:41 TemptorSent: what kind of issues? 2017-02-27 11:17:13 Failing to output when called with variable for grep arguments... 2017-02-27 11:22:20 When I hard-code the args, it will return valid data, otherwise it gives me nothing. 2017-02-27 11:23:03 I'm about going insane trying to figure out why it's puking on me. 2017-02-27 11:24:32 var=$(cat "$_file" | grep -e "^section.*\(\)" -e "^feature.*\(\)") works 2017-02-27 11:24:59 but if I assign the grep arguments to a variable, it pukes. 2017-02-27 11:30:16 ncopa, i've noticed that busybox 'top' sometimes just crashes on our builders, any ideas? 2017-02-27 11:31:26 any thoughts? It's the only thing holding me up right now, and I can't understand what's happening (even with sh -x) 2017-02-27 11:35:47 fabled, ncopa: http://termbin.com/pn8m is the changeset that works, please review. 2017-02-27 11:35:52 ncopa, actually, i think it's fortify catching something 2017-02-27 11:36:44 fabled: do you have a stable way to reproduce it? 2017-02-27 11:36:56 ncopa, not yet. trying to figure it out 2017-02-27 11:37:57 TemptorSent: do you have a testcase that does not wok? 2017-02-27 11:38:19 like: _section="^section.*\(\)" 2017-02-27 11:38:30 ... grep -e "$_section" ... 2017-02-27 11:40:17 ncopa: http://termbin.com/mpry is the diff that breaks. I tried a few work arounds and no luck. 2017-02-27 11:42:23 i wonder if it is the \) that gets expanded 2017-02-27 11:43:18 ncopa, got backtrace: http://sprunge.us/HbOM 2017-02-27 11:43:49 fabled: nice! 2017-02-27 11:45:39 char scrbuf[100]; /* [80] was a bit too low on 8Gb ram box */ 2017-02-27 11:46:01 i wonder how much you need for 48G 2017-02-27 11:47:10 sh -x isn't showing anything wrong with the expansion, just no value assignment happening. 2017-02-27 11:47:32 I've tweaked with the regex, even putting simple patterns with no extended regex fails 2017-02-27 11:48:21 ncopa, that's weird, it's snprintf, and does capping 2017-02-27 11:50:02 fabled: how can it be reproduced? 2017-02-27 11:50:25 it happens for me on official edge build x86-64 2017-02-27 11:51:16 what is the exact screen width you have? 2017-02-27 11:53:58 Okay, this is weird: blah="-e 'a' -e 'b'"; echo "a b c" | grep -e 'a' -e 'b' ; echo "a b c" | grep $blah 2017-02-27 11:54:05 ncopa, i think the backtrace might be incorrect 2017-02-27 11:54:11 Something seems broken. 2017-02-27 11:54:15 i retry with -O0 2017-02-27 11:56:18 Can anyone confirm whether that works for them or not/ 2017-02-27 11:57:10 TemptorSent: i know what happens 2017-02-27 11:57:18 its he ' that are passed ont 2017-02-27 11:57:47 blah="-e a -e b"; echo "a b c" | grep -e 'a' -e 'b' ; echo "a b c" | grep $blah 2017-02-27 11:57:58 in blah, remove the ' 2017-02-27 11:58:48 blah="-e 'a' -e 'b'"; echo "a b c" | grep -e 'a' -e 'b' ; echo "'a' 'b' c" | grep $blah 2017-02-27 11:59:12 you are grepping for "'a'" 2017-02-27 11:59:25 huh. again ghc failed due to wrong cpio (using bb, even though the real cpio was installed) 2017-02-27 11:59:55 Yeah, without any quotes it's trying, but it fubars spaces 2017-02-27 12:00:04 oh 2017-02-27 12:00:16 Let's see if it will try to work with all quotes removed. 2017-02-27 12:00:33 yes, you cannot really keep the spaces 2017-02-27 12:00:35 well 2017-02-27 12:00:38 you can sort of 2017-02-27 12:00:50 but you will have to use set -- ... for that 2017-02-27 12:00:54 and "$@" 2017-02-27 12:01:08 ncopa, the cpio is a mess, the old symlink needs to be removed manually 2017-02-27 12:01:19 ok 2017-02-27 12:01:54 fabled: did you do that i shoudl i do it? 2017-02-27 12:02:00 i did it 2017-02-27 12:02:05 ok 2017-02-27 12:03:02 ncopa, http://sprunge.us/hUjB 2017-02-27 12:03:21 seems stack smashed 2017-02-27 12:03:21 it happens when building stuff 2017-02-27 12:04:04 interesting 2017-02-27 12:05:04 char vsz_str_buf[8]; 2017-02-27 12:06:35 it could be also any function called 2017-02-27 12:08:02 fix appears to be replace grep $args with eval "grep $args" 2017-02-27 12:09:02 How to let APKBUILD can install /usr/src/test/*, it shows 'omitting direcotry' 2017-02-27 12:09:05 ncopa, ok that was right guess, increasing that buffer fixes it 2017-02-27 12:09:13 22270 22268 buildoze R 1048740m2048% 9 4% /usr/lib/ghc-8.0.2/bin/ghc -B/usr/lib/ghc-8.0.2 -hisuf hi -osuf o -hcsuf hc -static -H32m 2017-02-27 12:10:13 looks like it's the huge VSZ 2017-02-27 12:11:36 yeah, there's only two places it writes to vsz_str_buf, both with sprintf(); so that's the culprit 2017-02-27 12:16:55 if (s->vsz >= 100000) 2017-02-27 12:16:55 sprintf(vsz_str_buf, "%6ldm", s->vsz/1024); 2017-02-27 12:17:45 1048740m 2017-02-27 12:18:05 thats 8 chars with no space for trailing \0 2017-02-27 12:18:09 yeah, that format string does not limit it 2017-02-27 12:18:26 that's the stack overrun caught by SSP 2017-02-27 12:19:12 but that is insane big VSZ 2017-02-27 12:19:17 over a TB? 2017-02-27 12:20:21 yeah, sounds like haskell operates by doing huge virtualmemory alloc but not using it really 2017-02-27 12:22:28 huh 2017-02-27 12:22:29 + abuild-fetch -d /var/cache/distfiles http://downloads.haskell.org/~ghc/8.0.2/ghc-8.0.2-src.tar.xz 2017-02-27 12:22:29 curl: (48) An unknown option was passed in to libcurl 2017-02-27 12:23:19 oh 2017-02-27 12:23:35 old libcurl 2017-02-27 12:23:54 $ git diff procps/ | tpaste 2017-02-27 12:23:54 http://tpaste.us/NRKk 2017-02-27 12:23:58 i was thinking something like that 2017-02-27 12:25:06 ncopa, forgot else 2017-02-27 12:25:36 yes, and i also think we maybe should compare against something *1014 2017-02-27 12:25:40 1024 2017-02-27 12:25:48 since we / 1024 2017-02-27 12:27:23 maybe something like this: http://tpaste.us/xWpJ 2017-02-27 12:28:16 Alright, I give up on making plugins load nicely right now. 2017-02-27 12:28:56 Too tired, need this working by tomorrow for a project and I'm brickwalling. 2017-02-27 12:30:09 Can I get a review on the working code please? http://termbin.com/pm8nm 2017-02-27 12:31:27 I'll rewrite the plugin loading later. 2017-02-27 12:36:04 is that diff-log sufficiently clear? 2017-02-27 12:41:53 i havent looked too close to but at least it was in multiple commits 2017-02-27 12:41:58 so i assume its ok 2017-02-27 12:43:51 ncopa: I recreated the commit-sequence from scratch to clarify how files got copied/edited. 2017-02-27 12:47:17 Also, please take a look specifically at the new utils files, which may be useful to include more generally: http://termbin.com/34qw 2017-02-27 12:50:34 Specifically, would it be preferable to use a big case statement and additional option to implement the many similar functions than duplicating "boilerplate" 2017-02-27 12:58:13 kaniini, awilfox : hope the apk fetch --recursive issue you had is fixed with http://git.alpinelinux.org/cgit/apk-tools/commit/?id=0fe3f3964b8b5accc1e79f5f8a7d848e8fc2cba8 2017-02-27 13:16:47 mitchty, i'm trying to do first native compile of ghc on armhf 2017-02-27 13:16:53 ghc-cabal: Cannot find the program 'ld'. User-specified path 2017-02-27 13:16:53 '/usr/bin/armv6-alpine-linux-muslgnueabihf-ld.gold' does not refer to an 2017-02-27 13:16:53 executable and the program is not on the system path. 2017-02-27 13:17:18 it prefixes ld.gold with the triplet even though, i explicitly give --with-ld.gold 2017-02-27 13:18:35 (there's btw. bug in autoconfigure script when using --with-ld.gold, it gives: ./configure: ./configure.lineno: line 6507: With_ld.gold=ld.gold: not found ) 2017-02-27 13:20:12 make --no-print-directory -f ghc.mk phase=0 phase_0_builds 2017-02-27 13:20:12 "inplace/bin/ghc-cabal" configure libraries/binary dist-boot "" --with-ghc="/usr/bin/ghc" --with-ghc-pkg="/usr/bin/ghc-pkg" --package-db=/home/ncopa/aports/testing/ghc/src/ghc-8.0.2/libraries/bootstrapping.conf -v0 --configure-option=--quiet --configure-option=--disable-option-checking --disable-library-for-ghci --enable-library-vanilla --disable-library-profiling --disable-shared --configure-option=CFLAGS="-Wall -marm -fno-s 2017-02-27 13:20:12 tack-protector -Werror=unused-but-set-variable -Wno-error=inline" --configure-option=LDFLAGS=" " --configure-option=CPPFLAGS=" " --gcc-options="-Wall -marm -fno-stack-protector -Werror=unused-but-set-variable -Wno-error=inline " --constraint "binary == 0.8.3.0" --constraint "Cabal == 1.24.2.0" --constraint "hpc == 0.6.0.3" --constraint "ghc-boot-th == 8.0.2" --constraint "ghc-boot == 8.0.2" --constraint 2017-02-27 13:20:12 "hoopl == 3.10.2.1" --constraint "transformers == 0.5.2.0" --constraint "template-haskell == 2.11.1.0" --constraint "terminfo == 0.4.0.2" --with-gcc="gcc" --with-ld="/usr/bin/armv6-alpine-linux-muslgnueabihf-ld.gold" --with-ar="/usr/bin/ar" 2017-02-27 13:20:19 i wonder where it gets the --with-ld from 2017-02-27 13:24:19 Does it work with --with-ld=/usr/bin/ld.gold 2017-02-27 13:26:46 no, it seems to pick up the stage0 ld from previous build, and it's different due to being cross-compiled 2017-02-27 13:27:35 I love those... compile-time patching of intermediary files is fun. 2017-02-27 13:28:09 or potentially it just guesses the linker based on the triplet 2017-02-27 13:28:31 Put a patch in the makefile to patch the intermediary file before proceeding with the build. 2017-02-27 13:29:05 it picks it up from the cross-compiled ghc 2017-02-27 13:29:06 Ugly as sin, but it works for for most of the stubborn ones like that... or patch configure :) 2017-02-27 13:30:58 Any chance the buggy autoconf script is actually breaking proper detection? 2017-02-27 13:30:59 ncdev-edge-armhf:~/aports/testing/ghc/src/ghc-8.0.2$ ghc --info|grep /usr/bin 2017-02-27 13:30:59 ,("ld command","/usr/bin/armv6-alpine-linux-muslgnueabihf-ld.gold") 2017-02-27 13:31:28 i wonder how i can make the cross-compiled compiler to compile/link with the native compiler/linker 2017-02-27 13:32:04 Gotcha -- looks like you need to patch the environment itself perhaps. 2017-02-27 13:32:51 Drop a hook where it uses "ld command". 2017-02-27 13:42:35 Alright, I'd better try to get to sleep before the sun comes up or I won't be able to at all. 2017-02-27 13:42:44 Have a great day. 2017-02-27 13:44:23 seems like it's tweakable in $ghclib/settings text file 2017-02-27 13:44:27 mitchty, care to fix that to be configurable ? 2017-02-27 13:44:35 i'll probably add a sed hack to build it right 2017-02-27 13:46:45 Can you trick it into using your own file? 2017-02-27 13:50:09 yeah, i'll just sed it in package() to fix up the damage 2017-02-27 13:53:11 hum 2017-02-27 13:53:18 but it does not get the -I paths right now :/ 2017-02-27 13:56:17 ncopa, i'll probably tag and push new apk-tools 2017-02-27 13:57:44 mitchty, huh, ghc appends -fno-stack-protector unconditionally? and -no-pie ? 2017-02-27 13:58:04 that kind sucks 2017-02-27 14:00:35 and seems --with-system-ffi is broken too, it does not pass the proper -I flags to gcc when ghc is compiling 2017-02-27 14:01:41 ncopa, i wonder if we could place ffi.h in /usr/include rather then /usr/include/ffi-* 2017-02-27 14:04:39 make --no-print-directory -f ghc.mk phase=1 phase_1_builds 2017-02-27 14:04:39 HC [stage 0] libraries/terminfo/dist-boot/build/System/Console/Terminfo/Base.o 2017-02-27 14:04:39 HC [stage 0] libraries/Cabal/Cabal/dist-boot/build/Distribution/Compat/ReadP.o 2017-02-27 14:04:39 HC [stage 0] libraries/binary/dist-boot/build/Data/Binary/Put.o 2017-02-27 14:04:39 HC [stage 0] libraries/binary/dist-boot/build/Data/Binary/Get/Internal.o 2017-02-27 14:04:40 /tmp/ghc4753_0/ghc_4.c:4:17: error: 2017-02-27 14:04:42 fatal error: ffi.h: No such file or directory 2017-02-27 14:04:44 #include "ffi.h" 2017-02-27 14:04:46 ^ 2017-02-27 14:04:48 compilation terminated. 2017-02-27 14:04:50 `gcc' failed in phase `C Compiler'. (Exit code: 1) 2017-02-27 14:06:48 whoa. it puts headers in usr/lib/libffi-3.2.1/include/ffi.h 2017-02-27 14:12:47 Hi, I'm a newbie to alpine, I want to contribute 'synergy' package, but it require to also modify 'gtest' and add 'gmock' package. anyone can help me? 2017-02-27 14:13:14 ncopa, see: http://lists.busybox.net/pipermail/buildroot/2016-February/151780.html 2017-02-27 14:13:57 oh, sounds like cross-compiled libffi is different from native built 2017-02-27 14:17:17 fabled: tweak what in the settings file? 2017-02-27 14:18:24 oh do what i did in the ghc-bootstrap stuff 2017-02-27 14:19:13 https://github.com/alpinelinux/aports/blob/master/testing/ghc-bootstrap/APKBUILD#L41 2017-02-27 14:19:55 i left the triples in the settings file that the make install left, then strip all that junk out in the apkbuild via sed magique 2017-02-27 14:20:48 and yeah i'm pretty sure --with-system-ffi is broken, most of the ghc build system has warts that not too many people use 2017-02-27 14:21:54 and the reason for no-pie at the moment is the code generator can't yet properly handle pie in all cases iirc 2017-02-27 14:22:32 i think it was something to do with the rts being used but can't recall, there is a lot of history in ghc, its an old compiler 2017-02-27 14:37:10 i'm a bit closer to getting the triple stuff out, but its pretty pervasive in the makefile setup 2017-02-27 14:37:20 mitchty, testing now with http://sprunge.us/iSVF 2017-02-27 14:37:24 lots of places to look 2017-02-27 14:37:56 yep that'd do it 2017-02-27 14:38:15 make sure that C compiler supports -no-pie is YES 2017-02-27 14:39:37 yeah, it's detected right 2017-02-27 14:40:07 good work on the hp2ps and unlit patches 2017-02-27 14:42:10 libffi issue bug https://github.com/libffi/libffi/issues/24 2017-02-27 14:42:34 that was me trawling through the issue tracker finding uncommitted patches, and just copy/pasting the hp2ps fixes 2017-02-27 14:43:45 i've got a bunch of fixes that i'll be trying to get upstreamed 2017-02-27 14:57:42 cool, yeah the ghc build system is a bit of a mess tbh 2017-02-27 15:40:48 understatement of the century ? 2017-02-27 15:47:02 Oh you don't need to give more than sha512sums in apkbuild? Is that a new change? 2017-02-27 16:00:03 https://wiki.alpinelinux.org/wiki/Package_Maintainers, how to contriubte package? 2017-02-27 16:13:08 ashb: yes 2017-02-27 16:13:21 Yes to which? 2017-02-27 16:13:22 New change? 2017-02-27 16:13:24 ashb: yes 2017-02-27 16:13:31 Much nicer :) 2017-02-27 16:13:46 fabled: what is your opinion on adding sha256 hashes to apk index as a bandaid for sha1 collision attacks 2017-02-27 16:14:04 fabled: i think we can just adapt awilfox patch here 2017-02-27 16:48:36 kaniini, i was hoping to be further in apkv3 and postpone the hash changing there 2017-02-27 16:48:52 kaniini, not sure if it's worth fixing it in the current code base or not 2017-02-27 16:49:40 kaniini, at least the full extent of it. just updating the package signatures, would probably be simple enough 2017-02-27 16:50:49 some variation of https://code.foxkit.us/adelie/packages/blob/master/sys-apps/apk-tools/files/apk-tools-2.6.6-use-sha256-signature.patch 2017-02-27 16:51:34 i guess the question is 2017-02-27 16:51:40 how long before apkv3 is ready 2017-02-27 16:52:00 if it's 6-12 months out, then it is likely worth changing the signatures in apk-tools 2.x 2017-02-27 16:55:45 can the sha256 support be added with backwards compatibility? 2017-02-27 16:56:02 i am thinking for the 3.x-stable branches 2017-02-27 16:56:32 that part i don't know about 2017-02-27 17:00:51 i am thinking 2017-02-27 17:01:02 will apk3 be done for v3.6? 2017-02-27 17:01:08 most likely not 2017-02-27 17:01:29 and we will have to support v3.6 for 2 years 2017-02-27 17:10:41 kaniini, that patch does not fix all of it 2017-02-27 17:11:18 but yeah, it is an improvement and the price is not too high 2017-02-27 17:11:37 would need to do it in backwards compatible way though 2017-02-27 17:11:48 sync that with abuild 2017-02-27 17:18:40 ncopa: that's the other concern yes 2017-02-27 17:19:25 ncopa: what if it becomes trivial to forge APKs with the same hash in the next 2 years 2017-02-27 17:19:51 given that there's a very rapidly growing install base, wanting to produce forged APKs becomes more desirable every day 2017-02-27 17:22:36 so, what i think would be good, is add support for index with sha256 already now 2017-02-27 17:22:46 even if we still generate sha1 index 2017-02-27 17:23:21 the patch as such breaks compatibility, but something similar could be done 2017-02-27 17:24:23 though, changing apk_checksum will increase memory usage quite a bit 2017-02-27 17:25:02 and i don't think it would even be needed 2017-02-27 17:25:18 the mechanism how apk does signing and hashing was designed for speed 2017-02-27 17:25:58 i'll think about what can be done 2017-02-27 17:53:33 lld looks neat https://reviews.llvm.org/rL296072 (line 67 is perf figures) 2017-02-27 19:06:27 realistically, given that apk-tools 3 will require special upgrade instructions, the apkv3 distro will likely be alpine 4.x 2017-02-27 19:11:27 yes 2017-02-27 19:12:51 that means, that like alpine 2.7, alpine 3.6 may get LTS status 2017-02-27 19:13:07 in which case the support period may ultimately be longer than 2 years 2017-02-27 19:14:34 yeah, i'll think about it 2017-02-27 19:14:42 fixing the signatures is simple(ish) 2017-02-27 19:15:22 was thinking that one could do something like 'RSA.' for SHA-1 and 'RSA256.' for SHA-256, so then the if statement would be 'RSA' instead of 'RSA', then select sha1 or sha256 EVP based on whether fourth char is '.' or '2' 2017-02-27 19:15:24 but when index is used to install packages, it uses the index' hash, not the package signature 2017-02-27 19:16:16 awilfox, yes, that's how i'll probably do it. add the hash size/type to the signature filename 2017-02-27 19:16:42 yeah :) 2017-02-27 19:17:03 the patch linked above does not really significantly impact memory usage that I see, but if you can make it better, by all means please do 2017-02-27 19:17:15 I will test the fetch -r patch later today probably 2017-02-27 19:17:18 i think you don't need to modify apk_checksum 2017-02-27 19:17:49 sha256 is longer than sha1, so it was necessary in my testing. unless there is a better way to do it 2017-02-27 19:18:39 jirutka: can i get access to the github organization? 2017-02-27 19:18:48 or maybe clandmeter 2017-02-27 19:18:56 awilfox, the package signature is never stored in struct apk_checksum 2017-02-27 19:19:29 noh 2017-02-27 19:19:31 oh 2017-02-27 19:20:02 the identity hash gets changed too 2017-02-27 19:20:34 it seemed to be stored in apk_fileinfo_hash_xattr_array 2017-02-27 19:20:42 and a few other places 2017-02-27 19:21:49 the biggest thing is that it's in apk_db_file, so there's one of those per each installed file 2017-02-27 19:22:21 also in apk_package, so one per known package (from index or installed repo) 2017-02-27 19:22:25 correct 2017-02-27 19:22:38 I made the entire thing SHA-256 for additional security guarantees 2017-02-27 19:23:49 so, the memory usage increase on my system (150k files, 10k packages) would be about 3 megabytes or so 2017-02-27 19:24:29 it also slows down parsing of the indexes a bit, since more data is parsed 2017-02-27 19:24:42 it's not a problem on modern desktop or server 2017-02-27 19:24:57 but probably is noticeable on embedded things 2017-02-27 19:24:59 would you have 10K packages and 150K files on a machine where 3 MB is an issue? 2017-02-27 19:25:22 that's true, not thatmany files 2017-02-27 19:25:30 the 10k packages would be reality 2017-02-27 19:25:50 on Adélie, where this patch is used, I managed to get a full desktop installed with about 1200 packages and 70K files on a machine with just 40 MB RAM (and no swap) 2017-02-27 19:25:55 so it'd be more like 500k more memoryusage 2017-02-27 19:25:55 apk-tools did not OOM 2017-02-27 19:26:20 ok 2017-02-27 19:26:29 it did however OOM at 32 MB RAM on-board. that is mostly because the kernel and shell and silly eudev need so much though :P 2017-02-27 19:26:43 think that is somewhere near 20 MB RAM consumed just to boot the thing to a # prompt 2017-02-27 19:27:20 probably would have been much lighter if I went with mdev and ash 2017-02-27 19:31:31 hum 2017-02-27 19:31:40 seems the identity is still always sha1 2017-02-27 19:33:43 i'll look at it tm 2017-02-27 19:34:12 okay :) 2017-02-28 01:02:52 Please see http://termbin.com/hhz4 for revised mkimage diff, including working fully-modular plugin loader. 2017-02-28 03:19:17 Good evening kaniini 2017-02-28 03:33:45 https://travis-ci.org/alpinelinux/aports/builds/206047165 2017-02-28 03:34:59 I want to contribute synergy, but it is required build and install gtest & gmock at first, how could I do 2017-02-28 03:41:33 dlintw: your problem is that /home/travis/build/alpinelinux/aports/testing/gmock/pkg/gmock/usr/src/gmock isn't being created. most likely you need GNU coreutils as a build dependency 2017-02-28 03:46:46 Any idea what would be causing fakeroot to complain about permission denied when trying to build the minirootfs image? 2017-02-28 03:47:33 @kaniini, I'm not familiar with git, how could I continue modify it? Just check in my new-synergy will let it pull&request automatically? 2017-02-28 03:47:49 dlintw: most likely 2017-02-28 03:48:08 TemptorSent: i don't :/, it should be working 2017-02-28 03:48:58 Great... wonder what I did that may have broke it... 2017-02-28 03:52:44 kaniini: Can you give this a test run and let me know if I jsut munged my local system or if I broke the build? http://termbin.com/wmbs 2017-02-28 03:53:18 isos are building fine, but minirootfs pukes when it hits fakeroot for some reason. 2017-02-28 03:53:32 buildlog ? 2017-02-28 03:57:11 Hang on.. I think it may be a case of horrible error messages. 2017-02-28 03:57:17 Let me check something. 2017-02-28 04:00:55 Yep, found it -- I hadn't updated the script location, so while it should have been complaining 'File not found', it gave me a rather confusing 'permission denied'. 2017-02-28 04:03:01 Okay, minirootfs builds happily now too -- anyone have an arm setup they could run it against and let me know what I broke there? 2017-02-28 04:05:16 Fixed here: http://termbin.com/h99w 2017-02-28 04:06:45 If I haven't broken everything horribly, I'll proceed to fixing the horrendus append problem and add some convenient syntactic sugar for defining plugins and their variables. 2017-02-28 04:11:49 @kaniini, I update the testing/synergy/PKGBUILD on my origin/new-synergy branch, but after I use 'git ci ', it just to strange branch and show detached HEAD, so I can not check into the branch, what's wrong? 2017-02-28 04:18:42 dlintw: it's not synergy that is problem, it is gtest APKBUILD 2017-02-28 04:33:34 @kaniini, after add the coreutils, it finally passed the CI. So, will it go back into the aports? 2017-02-28 04:34:15 let me review it 2017-02-28 06:08:17 Good evening fabled, how are you? 2017-02-28 06:09:11 TemptorSent, morning. no coffee. i'll get over it... eventually. 2017-02-28 06:09:13 :) 2017-02-28 06:09:49 seems go and ghc cross-compiled bootstrap worked 2017-02-28 06:09:51 *lol* No coffee? Not a good morning! 2017-02-28 06:12:16 Anyway, I rewrote the plugin loading entirely and have it seemingly happy as a clam now. Builds standard, xen, minirootfs, and my fileserver-iscsi-zfs profiles cleanly and the images look complete. 2017-02-28 06:12:31 cool 2017-02-28 06:12:51 http://termbin.com/h99w 2017-02-28 06:13:31 The refactoring of all the bloody lists is the last thing I need to do to have things in what I would consider a reasonable state. 2017-02-28 06:14:50 Take a look and let me know what you think, especially regarding any variable names that need to be changed before cleaning it up. 2017-02-28 06:15:58 wow, lots of stuff 2017-02-28 06:16:04 Supporting mkinitfs integration should be pretty straightforward now, we just need to install a hook. 2017-02-28 06:16:35 in this review https://github.com/alpinelinux/aports/pull/939#discussion_r103375482 2017-02-28 06:16:53 How to do "Please make the requested changes and flatten down the commits into a single commit." 2017-02-28 06:17:01 I'm not familiar with git. 2017-02-28 06:17:28 Much of the diff is the splitting of the existing profiles into distinct files by copying then removing all but the desired code. 2017-02-28 06:18:48 The list-handling utilities are the othe lengthy chunk. Those are of general utility and may be worth including somewhere general. 2017-02-28 06:20:13 dlintw: Sorry, my git-fu is mostly maintaining local repositories. It sounds like you need to do a branch merge perhaps? 2017-02-28 06:22:19 The plugin loader can discover everything it needs given a directory to start in, or it can swallow a single-file as needed. 2017-02-28 06:22:46 dlintw: git rebase --interactive 2017-02-28 06:23:23 dlintw, TemptorSent : it's asking to combine several individual commits as one. http://stackoverflow.com/questions/11724666/flatten-commits-on-a-branch second answer 2017-02-28 06:24:11 Thanks fabled, it'll take a few to get used to the workflow. 2017-02-28 06:28:00 I've always just branched and merged, keeping the commit history intact incase I needed to go and back something out later. 2017-02-28 06:35:42 fabled: What were your plans regarding changes to mkinitfs? (and update-kernel?) 2017-02-28 06:36:06 TemptorSent, i have no immediate changes coming up, i would hope mkinitfs to be more modular, though 2017-02-28 06:36:50 Does the structure I build for mkimage look like it might be appropriate to support mkinitfs as well? 2017-02-28 06:39:18 TemptorSent, i'm still looking at the first few patches, my initial thought is if the profile directory needs to be that deep, as in why not drop the 2nd component of profiles/*/*.sh 2017-02-28 06:39:31 kaniini, check https://gist.github.com/dlintw/9bbff2342203b23d67c95f4f17f8af8e I've merged into one commit, but can not push to my branch again. 2017-02-28 06:39:35 or it's need becomes obvious later? 2017-02-28 06:40:06 TemptorSent, also, uboot is not arm specific. yes, currently alpine uses it for arm only, but uboot in general runs on multiple platforms 2017-02-28 06:40:08 fabled: Actually, it will find them at any depth in the hierarchy, so the directory structure is just for organization. 2017-02-28 06:40:10 dlintw: git push --force 2017-02-28 06:40:12 dlintw, push --force 2017-02-28 06:40:51 TemptorSent, i would not make subfolders unless we have >20 profiles or more 2017-02-28 06:41:11 fabled: Okay, good to know. The various bootloaders should probably live in their own world rather than the profiles I suppose. 2017-02-28 06:41:48 Well, it's building in CI. 2017-02-28 06:42:07 fabled: I could quickly see exceeding 20 profiles, especially since they can be rather application specific. 2017-02-28 06:43:31 So at least broad groupings may make sense, such as mobile, desktop, server, embedded, development, vm. 2017-02-28 06:44:42 TemptorSent, btw. I added the --cache-dir option to apk, hoping to tag a new release this week 2017-02-28 06:45:03 oh that reminds me I need to test fetch -r 2017-02-28 06:45:20 tagging a new release this week would make the adelie 1.0 version freeze, which happens sunday :D 2017-02-28 06:45:21 fabled: I haven't started creating many profiles yet because I wanted to make sure the underpinnings were solid and I didn't horribly break too much first :) 2017-02-28 06:46:03 fabled: Excellent! That will make life much easier when building in multiple roots with the same packages. 2017-02-28 06:47:08 fabled: I also am having a hard time tracking down all the places that xtables-addons and dahdi is hardcoded throughout multiple tools! 2017-02-28 06:49:24 TemptorSent, i wonder _what_ should go into the 'feature' layer. e.g. for samba you have feature per subpackage and that does not look like something i'd want to keep 2017-02-28 06:49:52 it's about .apks going to the media 2017-02-28 06:50:18 eventually, my vision is that apk would manage the 'apks' folder on bootable usb stick 2017-02-28 06:50:25 Why not? At least regarding samba, the functionality and packages required is very different. 2017-02-28 06:50:33 automatically based on 'world' 2017-02-28 06:50:36 my point is 2017-02-28 06:50:46 why make them separate feature, when it's matter of just one package 2017-02-28 06:51:00 kaniini, please to review the 'synergy' again. https://github.com/alpinelinux/aports/pull/939 2017-02-28 06:51:23 i would hope the profile to contain the package lists as literally as possible without indirection to features 2017-02-28 06:51:27 Because the feature will provide the configuration and overlay generation as well, enabling services, etc. 2017-02-28 06:52:13 do we always want to start off enabling the services though? 2017-02-28 06:52:18 TemptorSent, oh, sounds like i should finish reading it all to the end 2017-02-28 06:52:26 for xen ISO it makes sense, but does it make sense to enable samba from the start? 2017-02-28 06:52:33 If I need a profile that can act as a client on a windows network, I'd just include the samba_client feature, but if I'm running a domain controller, I'd use samba_server and samba_dc 2017-02-28 06:53:32 My need is a bootable media that I can dump to a client, have them burn and boot, and have the system functional, with upgrades being the swap of a cd or rewrite of a USB stick. 2017-02-28 06:54:00 TemptorSent, so you are aiming basically a image build system that generates the system configuration too? 2017-02-28 06:54:08 Preconfigured, including ssh keys, custome hostname, etc. 2017-02-28 06:54:47 Yep, I haven't done more than put in a basic skeleton for the overlays based on the existing scripts. 2017-02-28 06:55:29 So my intention is to allow for very easy custom profiles out of the main tree. 2017-02-28 06:55:41 ok, yeah the interesting stuff seems to be comming after commit #10 2017-02-28 06:55:43 reading still 2017-02-28 06:56:21 Think docker appliances, vm images, and rpi images that are ready to run with no user intervention. 2017-02-28 06:57:06 Yeah, sorry for the fits and starts a couple places there - I tried to keep everything as clear as possible. 2017-02-28 06:58:42 utils-list.sh could be refactored to use a couple case statements and an additional parameter to avoid the boilerplate, but I'm not sure if that is more desirable in this case or not -- your call. 2017-02-28 06:59:05 i think we need to talk about this a bit more. but yes, i do have similar need -- to create appliance boot images either fully or partially pre-configured 2017-02-28 06:59:47 I also haven't implemented a disjoint operation in the list utils yet, but that's easy enough, allowing all boolean list operations. 2017-02-28 07:01:05 fabled: That's exactly my goal, build a profile for a particular box, make an image, and send it to the client or host. 2017-02-28 07:02:34 The other big win is reducing the number of profiles you need to keep track of individual apks for, making keeping things up to date that much easier. 2017-02-28 07:03:42 And the features can provide the configureation interface themselves, reducing headaches with writing multiple external scripts. 2017-02-28 07:05:13 how do you plan to customize the ssh-keys etc to be host specific? profile file per host? or some other way to have generic profile that puts bits and pieces together based on db or flat file? 2017-02-28 07:05:24 I haven't implemented it yet, but for instance, it would be easy enough to do something like "feature_zfs --rootpool mypoool" if we wanted. 2017-02-28 07:06:13 but this does change the mkimage script from simple 'create boot image' to 'create instance specific configuration and media' type of thing 2017-02-28 07:06:42 I haven't worked that part out in detail yet, initially probably adding a command-line option like "--ssh-keys". 2017-02-28 07:07:09 Yes, it rather significantly increases it's potential scope of use. 2017-02-28 07:08:05 But none of that needs to necessarily be in the main tree, we can just add search directories as needed on the command line to use local configurations. 2017-02-28 07:09:06 I'm just hoping that the basic modular architecture might fly, making life easie for those of use building custom profiles while keeping the release-building functionality intact. 2017-02-28 07:09:42 yes, i definitely think we want custom profile base image generation to be simple 2017-02-28 07:11:19 and i think your work is definitely step to the right direction. i think we'll need to talk about the big picture first, and then on the implementation details a bit 2017-02-28 07:11:22 New feature could simply be distributed with the apks they support along with meta-packages, making it almost trivially easy to integrate and build a custom dist. 2017-02-28 07:12:11 hopefully ncopa, clandmeter, jirutka and kaniini has time to review and comment on it 2017-02-28 07:12:21 Agreed - I'm rather new to all-things Alpine and haven't gotten a good feel for the flavor yet beyond my initial hacking the build system. 2017-02-28 07:13:16 i've done some looking alreayd but it's a lot to absorb 2017-02-28 07:13:17 Coming from funtoo, things are MUCH simpler, but occasionaly frustrating due to the posixly correct shell instead of bash. 2017-02-28 07:14:18 It may make more sense to just apply it to a branch and take a look at the code and directory structure in-situ. 2017-02-28 07:14:27 fabled: i am considering proposing that we assign maintainership of core alpine packages to a mailing list or perhaps directly to the bugtracker itself 2017-02-28 07:14:46 It's not actually nearly as complex as the patches make it seem :) 2017-02-28 07:15:58 TemptorSent, yeah, there's few extra commits. it's good to split it to some, but it's probably few too many... it got me confused because the beginning does things which purpose becomes obvious only later 2017-02-28 07:16:47 kaniini, hum, maybe just Cc all notifications to mailing list? I'd still prefer a person assigned to each package if possible 2017-02-28 07:17:16 well, the reality is, the core packages are basically team-maintained 2017-02-28 07:17:21 though, i can see that for past few months it's been more and more driven by various doer people than the maintainer 2017-02-28 07:17:28 yes 2017-02-28 07:17:29 fabled: Yeah, sorry about that -- I managed to destroy my origial tree trying to purge a half-dozen accidentally commited .iso, so when I rebuilt the patch set, it was a little forced. 2017-02-28 07:18:13 (note: a top-level .gitignore with **/*.iso would be VERY nice :) 2017-02-28 07:20:12 I'm happy to work on cleaing it up further and meeting any styling you prefer, I just want to make sure it's at least close in terms of functionality and basic design first. 2017-02-28 07:23:13 Also, is it intentional to grab the whole mess of modules for the modloop? I see a huge number that are totally unnecessary when I build the images. 2017-02-28 07:30:00 scsi modules especially! Only rarely do we need the scsi adapter modules, but nearly every system needs the basic scsi modules. 2017-02-28 07:35:30 4761 modules in modloop for profile standard! 2017-02-28 07:36:26 Some of the firmware gets trimmed I think... 2017-02-28 07:46:50 75MB worth of modules included in modloop-grsec using the standard profile. 2017-02-28 07:48:10 3973 files in modules directory, 786 in firmware. 2017-02-28 07:51:25 About 7 times the size of the initfs itself, about 1/2 of the apks (160MB), and over a quarter of the total (256MB)! 2017-02-28 07:53:10 yeah, i have thinned down scsi for my custom images 2017-02-28 07:53:24 i think it's legacy cruft from the time when scsi was popular 2017-02-28 07:53:35 and most of the scsi drivers could be moved to scsi-legacy or similar 2017-02-28 07:56:25 how does one get access to the github organization btw 2017-02-28 07:56:48 I was thinking scsi-base iscsi scsi-hardware scsi-legacy 2017-02-28 07:57:01 i need to close a PR manually as i rebased it and jirutka's bot wasn't clever enough to catch it 2017-02-28 07:58:10 fabled: that is another thing that modular features could help with immensely. 2017-02-28 07:59:17 It allows individual features to be fairly fine-grained while more granular packages can encompas the most common configurations. 2017-02-28 08:00:26 So iscsi, for instance, would only pull in modules it actually needs, not all the hardware and and protocols it can't use. 2017-02-28 08:02:30 rnalrd: use options="!check" instead of defining a stub check() {} function :/ 2017-02-28 08:02:39 (speaking of which, I REALLY wish they would clean up the kerenl module directory structure a bit, we don't need protocol drivers sitting in the same directory as the adapters and various proprietary transports.) 2017-02-28 08:03:00 trying to build 'anthy', but get Invalid configuration `i586-alpine-linux-musl`, any tips? 2017-02-28 08:03:04 kaniini, k didn't know 2017-02-28 08:03:27 rnalrd: i should mention it in the warning, it was only in the original proposal 2017-02-28 08:03:30 yeah doesn't really make sense to add it where it's not supported 2017-02-28 08:03:49 although mentioning it may cause people to just opt their packages out of testsuites 2017-02-28 08:03:54 which is bad ! 2017-02-28 08:03:56 :p 2017-02-28 08:04:01 right :) 2017-02-28 08:05:09 TemptorSent, yes, sounds about right, it's legacy reasons why it's like that. 2017-02-28 08:05:42 kaniini, yeah, that "!check" vs. stub check() was my question, there was some incoming patches with stub functions also 2017-02-28 08:06:13 yes, we need to address this somehow 2017-02-28 08:06:16 i'll think on it 2017-02-28 08:06:17 fabled: Oh, how well I know and remember back when the Adaptech hardware support was the only thing solid! 2017-02-28 08:06:32 1540 I believe it was 2017-02-28 08:07:17 "no check function defined -- you should declare a check function, but you could also have !check in options but if you do this to skip an actual testsuite you're a terrible person" seems a little verbose 2017-02-28 08:07:42 *lol* Sounds perfect to me :) 2017-02-28 08:08:17 Don't forget the ANSI RED for TERRIBLE :) 2017-02-28 08:08:49 I'm twitching not having a test-suite for the mkimage system :) 2017-02-28 08:09:12 I'm tempted to at least writ up something for the utility functions. 2017-02-28 08:09:43 the testsuites here are useful because they help us smoketest for bugs caused by whateer 2017-02-28 08:11:08 Yeah, between ash and musl, there are a lot of packages that break in strange ways. 2017-02-28 08:12:11 Bashisms being one of the most pervasive I've seen, as just about every script out there in build systems expeted bash to be available. 2017-02-28 08:12:27 musl has gotten better, when we first started using it, it was less stable than our local uclibc tree 2017-02-28 08:12:40 Ouch. 2017-02-28 08:13:14 i think the effect of having a serious, real-world distro, actually using musl (stuff like sabotage is cute but it's not the same as alpine which is an actual distro used by tons of people) 2017-02-28 08:13:27 really had a positie effect on stabilizing musl 2017-02-28 08:13:49 Yeah, I remember trying to shoehorn it into some embedded hardware a while back and gave up on it, going with a custom application entirely instead. 2017-02-28 08:13:51 i.e. the "oh shit, people are actually using my code now" moment 2017-02-28 08:14:33 i would say most of the bugs were gone by 3.2 2017-02-28 08:14:38 *lol* Amazing how that happens.. at least it's not 10 years later, like some of the projects I've necroed in the past. 2017-02-28 08:14:40 and 2.7 was an LTS, so it wasn't too bad 2017-02-28 08:14:42 :) 2017-02-28 08:16:03 I haven't run into anything truly painful yet since I started playing with it on alpine. 2017-02-28 08:18:22 ash's posixly correctnes is a bit irritating in a few instances, most notibly the lack of the += append oppertor and somewhat irritating inability to use here-commands without resorting to ugly here-document hacks. 2017-02-28 08:19:46 raccoon:~/aports/main/audacious$ abuild check 2017-02-28 08:19:48 >>> WARNING: audacious: APKBUILD does not run any tests! 2017-02-28 08:19:50 Alpine policy will soon require that packages have any relevant testsuites run during the build process. 2017-02-28 08:19:52 To fix, either define a check() function, or declare !check in $options to indicate the package does not have a testsuite. 2017-02-28 08:20:40 Oh, cool! I just noticed ash DOES have the ${var/} construct available. I wish I would have known THAT before. 2017-02-28 08:22:43 I like it.. I think you should add a line about being a horrible person if you cheat and use an available testsuite :) 2017-02-28 08:23:05 er DONT use the available testsuite 2017-02-28 08:26:08 i doubt the suits at docker, inc. would find that funny 2017-02-28 08:26:48 *lol* Guess they can't handle the truth :) 2017-02-28 08:28:14 $ abuild check 2017-02-28 08:28:16 >>> WARNING: abuild: APKBUILD does not run any tests! 2017-02-28 08:28:18 Alpine policy will soon require that packages have any relevant testsuites run during the build process. 2017-02-28 08:28:20 To fix, either define a check() function, or declare !check in $options to indicate the package does not have a testsuite. 2017-02-28 08:28:22 THE IRONY 2017-02-28 08:29:38 *lol* Nice. 2017-02-28 08:30:02 It beggs for an easter-egg :) 2017-02-28 08:31:00 okay, sometime today i need to chat with dalias about funding the dynamic linking improvements 2017-02-28 08:31:17 hope my internet of things connected piggybank didnt get hacked 2017-02-28 08:31:21 that would suuuuuuuck 2017-02-28 08:31:57 "Check, check, one, two, one - two.. sssibilience, is this thing even on? *drops mike*" 2017-02-28 08:32:37 *lol* iot piggy-bank, yeah - *that* sounds secure! 2017-02-28 08:35:04 kaniini: check mailbox 2017-02-28 08:36:38 thx sir 2017-02-28 08:37:13 kaniini, awilfox : something like http://sprunge.us/XfZQ for the better signatures... 2017-02-28 08:37:40 yes, that looks good 2017-02-28 08:37:56 need to think what to do with the identity hashes still 2017-02-28 08:38:18 at least that's a start 2017-02-28 08:38:18 adelie reuses abuild-tar 2017-02-28 08:38:20 haha 2017-02-28 08:38:56 yes, it was too complex to reimplement it in Python 2017-02-28 08:39:32 would also be nice if github contributors would first read the CONTRIBUTING.md before doing contributions (it is highlighted when creating a PR). 2017-02-28 08:39:34 fabled: that looks good 2017-02-28 08:39:35 +fd = openat(ctx->keys_fd, &fi->name[10], O_RDONLY|O_CLOEXEC); 2017-02-28 08:39:51 this seems wrong 2017-02-28 08:40:04 with the longer SHA256. shouldnt it be [13] 2017-02-28 08:40:29 or maybe i am confused :) 2017-02-28 08:40:57 it might be better to have a temp var for it 2017-02-28 08:41:53 const char *filename = fi->name + 6 + sizeof(signature_type[i].type); 2017-02-28 08:41:55 or such 2017-02-28 08:42:29 clandmeter: yeaaaaaaaah, that aint happenin 2017-02-28 08:42:43 clandmeter: the github generation doesn't have time for such things like reading 2017-02-28 08:42:55 :) 2017-02-28 08:43:06 anyway 2017-02-28 08:43:07 they should add a checkbox 2017-02-28 08:43:08 ACTION nap 2017-02-28 08:43:34 yes I RTFM 2017-02-28 08:44:56 kaniini, awilfox : yeah, http://sprunge.us/hSFJ 2017-02-28 08:45:10 that one is tested to work 2017-02-28 08:45:57 fabled: nice! :D 2017-02-28 08:48:56 pushed 2017-02-28 09:03:45 I think tonight's going to be an early night... 2017-02-28 09:06:17 No watching the sun rise a second day in a row for me any more... 20 years ago it was a way of life, now I can barely see the screen after too long. 2017-02-28 09:47:15 same here, eyes start to hurt... 2017-02-28 09:47:18 getting too old. 2017-02-28 09:48:10 Yeah, never thought I'd live long enough to worry about it when I was younger, now, I'm half-wishing I was right :) 2017-02-28 09:49:16 I blame the PC for my short sighted glasses 2017-02-28 09:50:23 Heh, yeah -- I earned my approaching blindness the old-fashioned way, injuries and looking at the sun. 2017-02-28 09:50:48 The monitor just makes it obvious how far down I've come. 2017-02-28 09:51:12 despite them growing every year :) 2017-02-28 09:51:46 Then again, my old Viewsonic was actually higher resolution than my current flat pannel, so I'm not sure how far we've really come. 2017-02-28 09:52:49 i do notice the quality difference between my screens at work vs home. at home i have dell screens which are far superior. 2017-02-28 09:53:18 these BenQ's are so so... 2017-02-28 09:53:41 I've got ASUS's top of the line displays, thery're not bad, but still no where near actually black in the field. 2017-02-28 09:55:11 when you get older you find out how important a good chair, screen and keyboard are 2017-02-28 09:55:49 Yeah, I'm struggling horribly with the chair part... even the supposedly good ones are made like crap these days. 2017-02-28 09:57:00 I just got one of those "purple" seat cushions, which helps immensely (I hadn't realized I had been sitting for 4 hours straigh when the sun came up ;) 2017-02-28 09:57:31 But these days my coding sessions are usually a couple of hours, not a couple of days. 2017-02-28 09:57:47 what time are you in now? 2017-02-28 09:58:08 NIN101_:57 Pacific 2017-02-28 09:58:25 wtf? 01:57 Pacific 2017-02-28 09:59:57 Two or so I can handle fine, but watching the sun come up after being up at 10 isn't cutting it for me these days. 2017-02-28 10:00:39 Tomorrow I need to hurry up and write a document for public comment on some legislation and finish up coding the mkimage changes so I can get an iso to my client. 2017-02-28 10:02:15 Most of the work is done for making the framework usable, now it's just a matter of finishing up the overlay support and buildng out the profiles I need. 2017-02-28 10:06:02 Most pressing configuration is a zfs backed storage and virtualization image for workstation class hardware, followed by a postgres/postgis VM, some web-services VM appliances, and a utility environment for ETL work. 2017-02-28 10:07:17 I might have to use funtoo for the gis stuff, depending on how libc-agnostic those libraries are. 2017-02-28 10:12:11 ye youngsters can't complain about screens 2017-02-28 10:12:24 back in my day we had CRTs 2017-02-28 10:12:44 Yeah, that's the old monitor I'm remembering fondly :) 2017-02-28 10:12:52 yeah, i used to repair them :) 2017-02-28 10:13:05 any third-hand Chinese-sweatshop-made LCD will be more comfortable than a CRT without trying 2017-02-28 10:13:10 Even better were the old sun monitors -- one of those things damn near killed me! 2017-02-28 10:13:24 Sony Trinitron FTW! 2017-02-28 10:13:41 well those were 60 Hz 2017-02-28 10:13:45 Yup, invar shadow mask. 2017-02-28 10:13:50 every try old Mac screens? 2017-02-28 10:14:05 56Hz. How to get blind fast. 2017-02-28 10:14:42 the 800x600@56Hz VESA mode was an invention from hell. 2017-02-28 10:14:45 I cut my teath on old amber wyse terminals and the classic green-screens when I wasn't using a TV modulator :) 2017-02-28 10:14:56 vt100s were fine. 2017-02-28 10:15:01 120Hz :) 2017-02-28 10:15:40 but the sony's were really the best tubes we ever used. 2017-02-28 10:15:50 I turned my nose up at PCs in those days -- I was living in Atari and Amiga land with much better graphic handling. 2017-02-28 10:15:59 after that mitsubishi 2017-02-28 10:16:18 Yeah, Sony made nice tubes :) 2017-02-28 10:16:39 and there was still Lucky Goldstar.... 2017-02-28 10:16:42 as for chairs, since I'm spending about 3/4 of my time sitting at my desk, I splurged on an ergonomic chair, and it's one of the best investments I ever made 2017-02-28 10:16:44 Flat too compared to a lot of the beach-balls out there. 2017-02-28 10:17:10 What manufacturer? 2017-02-28 10:17:12 http://www.hermanmiller.com/products/seating/performance-work-chairs/embody-chairs.html 2017-02-28 10:18:07 it's not cheap, far from it - but I can sit all day and all night, and I have no back pains. 2017-02-28 10:18:58 Huh, that one looks novel -- I'll have to look into it! 2017-02-28 10:19:00 ah yes, herman miller 2017-02-28 10:19:06 the chair you buy used to survive 2017-02-28 10:20:01 Yeah, I had great chairs in my office back in the mid 90s, draftsman style foot ring, big stable star, good padding, no bloody bolts coming up through the bottom poking you in the ass... 2017-02-28 10:20:39 TemptorSent: I tested a whole panel of ergo chairs, the Aeron was pretty good but the Embody took the cake for me 2017-02-28 10:20:56 it may change depending on your build though 2017-02-28 10:21:05 Yeah, I tried an earlier Aeron and wasn't that impressed. 2017-02-28 10:21:30 I liked the Aeron tbh. But the Embody was better for me. 2017-02-28 10:22:36 other viable alternatives are the newly made gamer chairs 2017-02-28 10:23:00 I'm on the larger side of the spectrum, 6'4 before I shrank (getting old sucks). 2017-02-28 10:23:02 haven't tested them, but gamers/streamers - who also sit all day - seem to be comfortable with them 2017-02-28 10:23:24 (and it's not because they're padded with bank notes) 2017-02-28 10:24:04 Yeah, I've been eyeballing those, but I actually need to work in my chair, not just sit at the keyboard, but working at the desk and bench. 2017-02-28 10:25:15 work that requires you to sit and that doesn't mean typing at a keyboard? 2017-02-28 10:25:31 *lol* Yeah, lots of it. 2017-02-28 10:26:01 wow, what kind of job do you have XD 2017-02-28 10:26:39 Going over old documents, maps, etc. on one hand doing research, and sitting in front of a soldering iron and foredom on the other side. 2017-02-28 10:27:29 I've mostly gotten out of the IT world actually, just picking up the project out of needing the results 2017-02-28 10:28:45 can't say I particularly like the IT world, but for me it beats handling data on dead trees and hardware. :) 2017-02-28 10:29:12 I've mostly keep up some basic skills on funtoo, but I haven't been doing any major development work in almost 20 years. 2017-02-28 10:29:43 hey. _I_'m supposed to be the old one here! 2017-02-28 10:31:13 Back in the day I was on the bleeding edge, running a "internet presence provider" before web-hosting was a thing. I still rembember when I first got my hands on NCSA Mosaic and thought it was pretty cool for browsing some of the gopher sites that actually had images :) 2017-02-28 10:32:48 nice. I also used NCSA Mosaic, but as an early Web browser - I arrived on the scene when gopher and wais were dying. 2017-02-28 10:33:39 Which was a hell of a step up from connecting to the local nearest netcom number, using a slow text client, downloading the image to netcom, dropping out of gopher, and manually runnign zmodem on both ends. 2017-02-28 10:34:08 Yeah, we won't talk about UUCP and FidoNet gateways on BBSes then ;) 2017-02-28 10:34:34 I used UUCP! but it was going out of fashion when I arrived. 2017-02-28 10:35:06 I actually still had a fairly regular session into the late 90s. 2017-02-28 10:36:14 those were not good days. I don't miss them. 2017-02-28 10:37:02 I do miss them actually... We had a much closer tech community then, and everybody pretty much either knew you or knew someone who knew you. 2017-02-28 10:38:22 sure, and people wanted me to run BIND 4 and handwrite a sendmail.cf 2017-02-28 10:38:32 fuck that shit, and I'm being polite 2017-02-28 10:38:33 We literally controlled the internet in those early days... it only took a dozen of us to publish a blackhole route to essentially partion a spammer and his entire provieder off the internet. 2017-02-28 10:39:06 sendmail.cf was an art form and torture all in one. 2017-02-28 10:39:26 I never found beauty in pain 2017-02-28 10:39:28 That's what happens when you try to use something akin to make for configuration :) 2017-02-28 10:39:39 make is child's play 2017-02-28 10:40:25 I was learning Unix at this time and I couldn't help wondering "why does all this need to be so hard when the underlying designs are simple" 2017-02-28 10:40:57 "the concepts are good, why do our tools suck so much" 2017-02-28 10:41:25 My partner and I had actually written one of the first autoconfiguration suites for provisioning virtual hosting including dns, sendmail, pop3/imap, ftp, web, database!, and radius. 2017-02-28 10:42:32 Now, it's part and parcel, but then, each of those services required individual configuration, usually in some twisted fashion. 2017-02-28 10:47:25 I think that must have been about 1995, as Postgres95 was just coming out replacing the earlier ingres/postgres. Personal Home Pages / Forms Interface (PHP/FI) came out about then too, so we were offering database backed dynamic webpages and complete virtual server hosting in the mid 1990s, and our top-of-the-line server was a dual-processor pentium-pro 200 with 128 MEGABytes of memory and a 9GIGAByte 5.25" full height HDD that cost upwards of $1200 2017-02-28 10:48:44 The couple small Quantum Fireball and Seagate Barracuda drives were much smaller and cheaper :) 2017-02-28 10:48:53 Yeah, those were the days. 2017-02-28 10:49:09 When your entire server could fit on a single DDS2 tape. 2017-02-28 10:51:15 Does anyone else remember actually using tar to write to old quarter inch tape? 2017-02-28 10:54:10 again, I have about zero nostalgia of these days 2017-02-28 10:54:19 I kinda laugh when I look at a "lightweight" linux distribution these days and compare it to what were were running then. I was running usefull systems on 4MB of RAM still, including a full set of daemons, print server, etc, and still had a solid gui in the late 1980s 2017-02-28 10:55:33 So why do we need 75mb of kernel modules these days again? 2017-02-28 10:55:33 now I agree, and one of my goals is to bring the lightweight back. :) 2017-02-28 10:56:23 I was able to do a lot of useful stuff with VERY minimal hardware. 2017-02-28 10:57:47 I have the print out of the first linux kernel I ran somewhere around here.. it's less than a cm thick. 2017-02-28 10:58:17 Linus laughed at me when I asked him to sign it for posterity at a usenix conference ;) 2017-02-28 10:58:46 of course he did. He never was one for romanticization of those things. 2017-02-28 10:58:48 It was only a class project gone wrong. 2017-02-28 10:59:28 He didn't think it was going to be a big deal at the time. 2017-02-28 11:00:05 team meeting, bbl. 2017-02-28 11:00:09 Sun was still dominating the server world, with SGI running away with the early web production. 2017-02-28 11:00:17 I'm off to bed. 2017-02-28 11:00:40 It's been great talking, have a good day. 2017-02-28 11:14:42 clandmeter, is buildbot-slave now part of buildbot? could the package in testing be removed? 2017-02-28 11:15:15 fabled: idk, i havent touched it for some time. 2017-02-28 11:15:21 i noticed 2017-02-28 11:15:24 that's why i ask 2017-02-28 11:16:05 im not sure about the status, but if you think its ready to be purged be my guest. 2017-02-28 11:16:59 http://docs.buildbot.net/latest/manual/worker-transition.html 2017-02-28 11:17:03 seems it's changed to buildbot-worker 2017-02-28 11:17:07 looks like fcolista moved it before 2017-02-28 11:19:09 fedora seems to still have a slave pkg 2017-02-28 11:19:49 but they are still on 0.8.x 2017-02-28 11:20:36 yeah, it's 'slave' on 0.8 branch; 'worker' on 0.9+ 2017-02-28 11:22:45 are you going to work on buildbot integration? 2017-02-28 11:27:36 kaniini: That gnupg patch is to allow gnupg1 to be removed and install gnupg2 instead when apk add gnupg (haven't tested it yet). 2017-02-28 11:28:33 clandmeter, that's long term plan yes. not sure if i work on it parallel with apk3, or do apk3 first. 2017-02-28 11:28:47 would be kinda simpler to fix both at same time 2017-02-28 11:28:58 but not sure how much it's work 2017-02-28 11:29:01 need to look at it a bit 2017-02-28 11:40:02 pickfire: can gpg1 and gpg2 still coexist on the same system? 2017-02-28 13:14:05 1144 unread messages in #alpine-devel… I need a vacation to read full backlog :( 2017-02-28 13:16:19 kaniini: "i need to close a PR manually as i rebased it and jirutka's bot wasn't clever enough to catch it" … this sometimes happens when renaming files, it’s already reported https://github.com/jirutka/github-pr-closer/issues/1, but no one fixed it yet :P 2017-02-28 13:53:20 jirutka: sounds like you need to log off once in a while :) 2017-02-28 13:56:05 logging off? you don't say! I want to feed the IRC log right into my brain when I'm sleeping, to avoid missing anything! 2017-02-28 15:08:48 hey friends, i am just reading up on the entire Google Chrome TLS 1.3 thing. If you dont know: https://bugs.chromium.org/p/chromium/issues/detail?id=694593 2017-02-28 15:08:48 maybe i am currently drunk as hell and completly stupid, but isn't one of the usecase of TLS that your connection does not get intercepted by any proxy? If i see that correctly, i think the chromium version in alpine should force enable it again when google activly enabled MITM here 2017-02-28 15:08:52 what are your thoughts? 2017-02-28 16:15:59 leo-unglaub: you are just naive. TLS MITM is unfortunately very common. It should be burned down to the ground with napalm, but unfortunately it's not the case. 2017-02-28 16:16:42 nah, i know that it is very common for tls 1.2 .. speciall with bluecoat proxys 2017-02-28 16:17:05 but thats why they improved the handshake in 1.3 so that most of those things would not work anymore 2017-02-28 16:17:21 ah, I need to read up on the 1.3 mechanisms 2017-02-28 16:19:31 Just found qt's include file put on wrong place 2017-02-28 16:19:54 should I provide patch to correct them? 2017-02-28 16:21:17 skarnet: its kind of interresting to read up on the 1.3 handshake .. its more secure and also faster beceause the client directly sends the key 2017-02-28 16:21:34 so you dont have to wait for the server for that one 2017-02-28 16:21:43 its one less reply response for a handshake 2017-02-28 16:29:39 leo-unglaub: I agree with reverting this in chromium 2017-02-28 16:34:37 dlintw: which qt? 2017-02-28 16:34:40 qt5? 2017-02-28 16:35:56 yes qt5 2017-02-28 16:36:49 In Arch the include path is /usr/include/qt/QtPlatformCompositorSupport 2017-02-28 16:37:01 i wonder if that was done on purpose to not clash with qt4 2017-02-28 16:37:06 but in Alpine, it reduced to /usr/include/QtPlatformCompositorSupport 2017-02-28 16:39:40 how do you know that it is not Arch that is wrong here? 2017-02-28 16:41:20 Well, if we want to support multiple qtX version, I guess there should put a base qt/* then old version qt4/* 2017-02-28 16:41:47 If one day, the qt6 release, then we'll create qt5/* 2017-02-28 16:42:48 Sorry, I must go to sleep. Good night. 2017-02-28 16:51:23 ncopa: https://bugs.alpinelinux.org/issues/6928 vsyscall=emulate for 3.4.6 and 3.5.1 ? 2017-02-28 16:51:45 tru_tru: for any kernel that has this issue 2017-02-28 16:52:09 i would guess its needed for any alpine 2017-02-28 16:52:25 ok thx 2017-02-28 17:15:27 i cannot wait for my new Ryzen CPU ... the first thing i am going to do is to build the entire pkgs tree and see how long it takes :) 2017-02-28 17:16:33 leo-unglaub: so AMD's still alive? I recall this all hype around Bulldozer which actually wasn't that powerful as might expected 2017-02-28 17:17:33 hehe, me to ... but this time amd got it right ... 2017-02-28 17:18:22 i am interrested in how fast all packages are going to build .. when you dont count the download time i asume around 3 hours? 2017-02-28 17:18:54 make sure you build them on a tmpfs :-) 2017-02-28 17:19:05 yes, all in memory 2017-02-28 17:19:12 64GB DDR4 memory 2017-02-28 17:19:37 does anyone know how long the official build server needs for an complete release in x64 2017-02-28 17:21:00 leo-unglaub: meh, it took my 4 or 5 hours to build just libreoffice on my machine ;x 2017-02-28 17:22:21 hmm, can you build this java stuff with make -j x? or is it single threaded` 2017-02-28 19:15:46 is there a reason for alpine's mpv not supporting DVD files by default? (libdvdread/libdvdnav) 2017-02-28 19:39:05 also, speaking of alpine's new "tests for every package" policy i heard - is there a "checkdepends" equivalent to "makedepends", or should i just use the latter? 2017-02-28 19:40:18 asie: yes, there is 2017-02-28 19:40:48 is it called "checkdepends"? 2017-02-28 19:40:53 yes ;) 2017-02-28 19:40:56 heh 2017-02-28 19:40:58 thanks 2017-02-28 19:41:13 packaging Anki; thought I was done, but it turns out I need to package Anki's test framework first 2017-02-28 19:42:18 OpenTTD seems to only come with internal AI regression tests and not general code tests 2017-02-28 19:42:59 AI as artificial intelligence? how they test it? :) 2017-02-28 19:43:57 oh, it just tests things such as data structures etc 2017-02-28 19:44:05 but they clearly seem internal 2017-02-28 20:29:52 another thing i spotted: does ffmpeg-libs need to pull in sdl2-dev? isn't it only used by the main ffmpeg package? 2017-02-28 22:41:17 How to solve 'QtDBus', required by 'virtual:world', not found problem? 2017-02-28 22:56:54 How to continue build by abuild command when I just apply a quick patch? 2017-02-28 23:03:15 dlintw: not sure if I understand your question, but you can invoke the build phases directly, e.g. abuild prepare, abuild build, … 2017-02-28 23:04:53 so, abuild -k build will just let the 'make' continue? 2017-02-28 23:06:07 no 2017-02-28 23:06:21 it would just invoke the build phase 2017-02-28 23:06:57 I don’t know about any option how to “continue” the aborted build 2017-02-28 23:08:08 but you can invoke the phases manually and get what you want, e.g. abuild build rootpkg cleanup 2017-02-28 23:09:01 How to solve this missing gcin-dev https://gist.github.com/dlintw/709de5b6851be79cf7a44ab8d8f69dd1 2017-02-28 23:10:10 the phases are: sanitycheck builddeps clean fetch unpack prepare mkusers build check rootpkg cleanup 2017-02-28 23:10:11 Does that mean I should remove the subpackages of $pkgname-dev? 2017-02-28 23:10:24 yes 2017-02-28 23:11:18 this means that the subpkg -dev is empty, i.e. if you haven’t define custom dev() function, then it means that default_dev didn’t find any “dev files” (header files etc.) 2017-02-28 23:12:12 As a Chinese, I can not accept a linux system without Chinese input method, so, I'm trying to build the 'gcin' package. 2017-02-28 23:16:38 So, the default 'abuild' equal to 'abuild unpack prepare build cleanup ...'? 2017-02-28 23:17:23 more or less… 2017-02-28 23:17:45 it’s a little bit more complicated because of support for cross-compilation 2017-02-28 23:18:27 So, what's the correct sequence, could you describe in https://wiki.alpinelinux.org/wiki/Abuild_and_Helpers? 2017-02-28 23:21:20 Is there any package for lxpanel replacement already in Alpine packages?