2015-08-01 00:14:37 It looks like execline and imagemagick both contain /usr/bin/import. I'm not sure what the best fix is. Any solution I can think of involves breaking compatibility, adding complexity, or both. 2015-08-01 00:33:28 ^^^ Filed it as Bug #4504 2015-08-02 07:02:28 Hi guys, it seems the LXC (linux-containers) template is broken for some reason. https://github.com/lxc/lxc/issues/609 2015-08-02 07:03:32 More clear: the alpine template which is packaged with upstream LXC 2015-08-02 10:52:18 i just had to upgrade a debian system 2015-08-02 10:52:23 apt-get is so slow and i have no idea why 2015-08-02 10:52:31 apk is propobly 40 times faster 2015-08-02 10:52:36 good job alpine :) 2015-08-02 11:24:42 leo-unglaub: well I'm sure 40x is even understated 2015-08-02 11:24:45 ^^ 2015-08-02 12:01:21 hehe 2015-08-03 06:21:13 good morning 2015-08-03 06:24:13 morning 2015-08-03 06:28:10 today i am going to look into why /etc/crypttab does not work with more than one entry 2015-08-03 07:18:36 ncopa: cgit has issues again? 2015-08-03 07:19:31 apparently it has 2015-08-03 07:20:14 it segfaulted 2015-08-03 07:20:37 and cgit segfaulted too 2015-08-03 07:20:51 smells like a security problem 2015-08-03 07:22:08 i thought my pkgs is broke 2015-08-03 07:22:14 broken 2015-08-03 07:22:23 but it seems we dont have a single commit this weekend? 2015-08-03 07:23:04 ncopa: did you tell them to take the weekend off? 2015-08-03 07:23:38 the blackhats? 2015-08-03 07:24:05 i did. i asked them to please wait with attacking our cgit server til monday 2015-08-03 07:24:08 they never listen.... 2015-08-03 07:24:09 :) 2015-08-03 07:24:33 :) 2015-08-03 07:27:46 actually 2015-08-03 07:27:53 lighttpd seems to be up 2015-08-03 07:27:59 cgit segfaulted though 2015-08-03 07:29:05 oh we do have commits this weekend. 2015-08-03 07:29:39 ncopa: more of our infra is down? 2015-08-03 07:29:50 not that i know 2015-08-03 07:29:59 i'm looknig at the logs for cgit 2015-08-03 07:30:02 algitbot didnt announce anytning 2015-08-03 07:30:15 my git log shows changes 2015-08-03 07:31:15 oh shit... im looking in the wrong channel! 2015-08-03 07:43:35 ncopa: anyway i can ask algitbot to try to rebuild? 2015-08-03 07:44:04 you can ask him 2015-08-03 07:44:20 but i cannot guarantee he will obey ;) 2015-08-03 07:46:36 i thought that cmd didnt work anymore? 2015-08-03 07:48:32 seems like you can do: algitbot retry 2015-08-03 08:13:12 ncopa, this can be done from everyone? Or just for who is irc op? 2015-08-03 08:13:31 *from who 2015-08-03 09:00:08 fcolista: everyone can do it 2015-08-03 09:09:56 ncopa: maybe a "algitbot help" would be usefull. 2015-08-03 11:48:51 Hey :) 2015-08-03 11:49:02 ncopa: :) 2015-08-03 11:50:41 hi 2015-08-03 11:52:19 hey 2015-08-03 11:52:29 i was told you'd be interested in speaking to me? 2015-08-03 11:53:45 ok? i dont remember why though :) 2015-08-03 12:12:26 about documentation 2015-08-03 12:13:29 alacerda: hi 2015-08-03 12:14:01 ok? 2015-08-03 12:14:09 hi nootropics 2015-08-03 12:14:32 we have wiki only for now, but i think it would be nice to have some more official docs generated from asciidoc 2015-08-03 12:14:45 jey 2015-08-03 12:14:52 ncopa: I was thinking of totally rewriting it 2015-08-03 12:14:59 *hey 2015-08-03 12:15:05 and writing a better installer 2015-08-03 12:15:10 i have always liked the freebsd handbook 2015-08-03 12:15:17 define "better" 2015-08-03 12:15:54 ncopa: alpine linux currently has no installer that supports, UEFI, LKM/LUKS 2015-08-03 12:17:40 or USB install 2015-08-03 12:17:43 and I'd like to change that 2015-08-03 12:21:25 true 2015-08-03 12:21:27 agree 2015-08-03 12:22:10 i would like to be able to backwards in the setup "wizard" too 2015-08-03 12:22:31 so i've got ideas, i'd like to rewrite the docs from scratch, write an updated installer and write a usb mode where you can install alpine persistantly on a usb and choose to either boot from scratch or to your persistant partition 2015-08-03 12:22:36 much like tails 2015-08-03 12:22:50 isnt that what Ctrl+c is for? 2015-08-03 12:23:03 :) 2015-08-03 12:23:25 i would also like to have autocompletion for things 2015-08-03 12:23:42 Agreed 2015-08-03 12:23:56 clandmeter: lol yes exiting the entire install jus to change one thing 2015-08-03 12:24:01 also, currently, if you use arrow keys to edit the input line, things go pretty bad... 2015-08-03 12:24:28 i have been thinking of using linenoise for readline functionality 2015-08-03 12:24:37 i was just thinking of that 2015-08-03 12:24:45 couldnt remember the name though 2015-08-03 12:24:50 :) 2015-08-03 12:25:00 with you guys permission i'd like to turn alpine into something much more user friendly 2015-08-03 12:25:08 been thinking of making a lua module for linenoise 2015-08-03 12:25:09 well 2015-08-03 12:25:10 where anyone can plug in a usb and instantly have a grsec + hardened toolchain install 2015-08-03 12:25:19 I absolutely LOVE this project. 2015-08-03 12:25:20 I really do 2015-08-03 12:25:52 i dont mind a more user friendly setup 2015-08-03 12:25:57 but i wonder what the price is 2015-08-03 12:26:03 :) 2015-08-03 12:26:10 you know the price... 2015-08-03 12:26:16 there will be no price. 2015-08-03 12:26:23 i mean "price" not price 2015-08-03 12:26:30 no I know 2015-08-03 12:26:47 you plug it in, a small menu pops up "Want to boot from your persistant partition or fresh" 2015-08-03 12:26:54 enter your pass if you choose the first 2015-08-03 12:27:02 does it mean that we will need gazillion of deps? 2015-08-03 12:27:06 Nope 2015-08-03 12:27:12 I hate hundreds of deps 2015-08-03 12:27:13 will our mini iso turn into a 3GB dvd :) 2015-08-03 12:27:16 nope. 2015-08-03 12:27:23 I'll write the menu in bash 2015-08-03 12:27:31 your mini iso will at most add 5MB 2015-08-03 12:27:52 I'm hgappy to fork the project if you'd rather but there's some REALLY intelligent people here 2015-08-03 12:28:01 and I'm not good enough to continue development of this project 2015-08-03 12:28:03 whats out, we are bash bashers :) 2015-08-03 12:28:10 ay :3 2015-08-03 12:28:13 bash is <3 2015-08-03 12:28:17 whiptail is <3 2015-08-03 12:28:26 we do bash bashers 2015-08-03 12:28:35 bash is big and slow 2015-08-03 12:28:51 dpeends on how you write it 2015-08-03 12:29:01 besides, the more deps we install for it, the more attack surface 2015-08-03 12:29:10 i'd really like to get involved with this project in some offical capacity 2015-08-03 12:29:15 but its fine if you dont want me to 2015-08-03 12:29:23 we do like posix shell though... 2015-08-03 12:29:41 and Lua 2015-08-03 12:29:41 did we mention lua already? 2015-08-03 12:30:06 Lua is about half size of bash 2015-08-03 12:30:09 I actually haven't given lua a fair chance 2015-08-03 12:30:12 i'd like to try it 2015-08-03 12:30:23 but you guys seriously underestimate how much i love this project 2015-08-03 12:30:31 i'd like to add modes for total transparent tor proxying etc 2015-08-03 12:30:37 :) 2015-08-03 12:31:00 alacerda has this setup-box project 2015-08-03 12:31:17 which as a set of prilfes for different setups 2015-08-03 12:31:21 ^ 2015-08-03 12:31:25 profiles* 2015-08-03 12:31:27 I've wrote my own one 2015-08-03 12:31:28 called init 2015-08-03 12:31:34 just for personal boxes 2015-08-03 12:31:58 but i guess what im asking is; can i join the team? its fine if you dont want me to i'm still hapyp to help, just would be super nice 2015-08-03 12:32:25 i dont mind a bit help with this 2015-08-03 12:32:35 :) 2015-08-03 12:32:48 and a new installer is something i have been thinking of too 2015-08-03 12:32:49 i'm passionate about this stuff, security, freedom, privacy 2015-08-03 12:33:00 ncopa: is there a chance you'd let me to join teh team? 2015-08-03 12:33:15 normally it works like this 2015-08-03 12:33:18 you send patches 2015-08-03 12:33:22 contribut minor stuff 2015-08-03 12:33:27 then you contribute major stuff 2015-08-03 12:33:28 I'm happy to do that. 2015-08-03 12:33:41 we wait and see what you are good for 2015-08-03 12:33:49 I've considered donating but I'm a really really poor student :( 2015-08-03 12:33:52 Alright. 2015-08-03 12:33:55 and we wait to see if you are someone who will stay 2015-08-03 12:33:59 Makes sense, I didn't mean right away I mean if it pans out well 2015-08-03 12:34:11 or if you are gone after 3 months 2015-08-03 12:34:12 i suck at C though 2015-08-03 12:34:14 we dont accepts poor ppl's money, so keep it. 2015-08-03 12:34:19 :) 2015-08-03 12:34:25 like i can do a bit, but I can't do much kernel hacking 2015-08-03 12:35:00 yeah, we normally prefer contributions in form of patches and wiki articles :) 2015-08-03 12:35:12 Alright awesome 2015-08-03 12:35:19 wiki articles is something I can 100% do 2015-08-03 12:35:39 what we do need 2015-08-03 12:35:44 but yeah thank you for hearing me out and this sounds cheesy as fuck but for real amazing job on the project so far 2015-08-03 12:35:56 and write a man page for algitbot 2015-08-03 12:36:01 i have no idea what that is 2015-08-03 12:36:17 we do need someone that can take a bit responsability for documentation 2015-08-03 12:36:34 right now the documentation is on wiki 2015-08-03 12:36:34 I can absolutely do that, I've got time that I waste all the time and I want to lern things 2015-08-03 12:36:37 yep. 2015-08-03 12:36:51 only issue is medical wise I'm a bit up and down so some days I might not be available 2015-08-03 12:37:26 wiki does not guarantee that someone has the bigger overview plan with documentation 2015-08-03 12:37:38 people normally only see their tiny corner 2015-08-03 12:37:43 no like 2015-08-03 12:37:45 i wan tto rewrite 2015-08-03 12:37:47 EVERYTHING 2015-08-03 12:38:01 i want to make it as good as the arch linux wiki 2015-08-03 12:38:04 but with more of a focus on security 2015-08-03 12:38:05 and privacy 2015-08-03 12:38:15 sounds good to me 2015-08-03 12:38:17 :) 2015-08-03 12:38:31 I'd also like to add smart card support to alpine 2015-08-03 12:38:34 if thats okay with you guys 2015-08-03 12:38:34 we coudl also do official documentation as asciidoc 2015-08-03 12:38:40 i dont mind that 2015-08-03 12:38:49 I've written in my spoare time at school 2015-08-03 12:38:50 a small 2015-08-03 12:38:54 30-40 page guide 2015-08-03 12:38:56 called Paranoia.md 2015-08-03 12:39:01 I plan on releasing soon/when it's done 2015-08-03 12:43:14 sorry I was in a meeting 2015-08-03 12:43:17 reading the backlog 2015-08-03 12:43:33 i wonder if it would be possible to move documentation to where we want it in smaller steps 2015-08-03 12:43:40 instead of rip out and replace 2015-08-03 12:44:11 incremental changes tend to work better in practice than revolutions 2015-08-03 12:46:34 No of course np, yeah that works fine with me 2015-08-03 12:46:44 in a bit of an arguement on twitter sorry if im afk 2015-08-03 12:53:42 alacerda: any htoughts? 2015-08-03 12:54:09 nothing to add. Just a +1 to thi: 2015-08-03 12:54:11 "incremental changes tend to work better in practice than revolutions" 2015-08-03 12:54:15 agreed 2015-08-03 12:54:26 i was thinking of hosting a private gitlab repo with all my changes 2015-08-03 12:54:29 so youg uys can look over them 2015-08-03 12:56:03 2015-08-03 12:56:41 sure 2015-08-03 12:57:12 sounds good to me 2015-08-03 12:57:32 but i'd recommend to start with a minor project first 2015-08-03 12:57:37 and see how that works out first 2015-08-03 12:57:45 I'd like to start with the installer script 2015-08-03 12:57:48 before starting soemthing big 2015-08-03 12:57:56 to me that's pretty small 2015-08-03 12:58:30 do you think we can do incremental changes to installer too? 2015-08-03 12:58:45 my plan was to write a real installer to put alpine onto a usb stick 2015-08-03 12:58:47 or cd 2015-08-03 12:58:50 ok 2015-08-03 12:58:55 then work on the actual installer 2015-08-03 12:59:00 ncopa: did you mean https://github.com/hoelzro/lua-linenoise ? 2015-08-03 12:59:29 clandmeter: ha.. i didnt know one existed 2015-08-03 12:59:30 coold 2015-08-03 12:59:35 cool* 2015-08-03 12:59:50 and it even has releases 2015-08-03 12:59:56 and recent ones too 2015-08-03 13:00:08 is lua worth learning? 2015-08-03 13:00:21 if you like alpine it is 2015-08-03 13:00:24 WOW 2015-08-03 13:00:27 that's rude 2015-08-03 13:00:54 well we use lua pretty much 2015-08-03 13:01:03 for many things 2015-08-03 13:01:23 got any examples? 2015-08-03 13:01:28 where others use perl python or ruby we use lua 2015-08-03 13:01:33 im a python/perl/c guy myself 2015-08-03 13:01:36 oh lol yeah 2015-08-03 13:01:52 build server scripts 2015-08-03 13:02:07 pkgs.alpinelinux.org is lua 2015-08-03 13:02:16 the website generator scripts are lua 2015-08-03 13:02:45 algitbot: hi 2015-08-03 13:02:55 the irc bot uses lua... 2015-08-03 13:07:22 so if you contribute with 5 lines python code, then be prepared for a question like: "can this be rewritten in lua?" ;) 2015-08-03 13:11:22 oh lol 2015-08-03 13:11:22 ok 2015-08-03 13:35:19 2015-08-03 13:35:26 hey 2015-08-03 13:43:41 ncopa: no love for ACF 2015-08-03 13:43:50 you didn't mention that it's written in Lua :) 2015-08-03 13:48:02 oh 2015-08-03 13:48:05 thats true 2015-08-03 13:48:07 ACF 2015-08-03 13:48:08 and awall 2015-08-03 14:46:00 alacerda: ncopa helping my mate set up arch then i'll start writing the thing 2015-08-03 14:50:19 no stress 2015-08-03 14:51:46 good nootropics, take your time ; ) 2015-08-03 16:54:54 also 2015-08-03 16:54:58 i had a qestion for alacerda and ncopa 2015-08-03 16:55:06 when you install arch, the command you run is 2015-08-03 16:55:08 pacstrap 2015-08-03 16:55:12 whats alpines equivlent 2015-08-03 17:00:13 I think that questions belongs to #alpine-linux but you are probably looking for setup-disk 2015-08-03 17:00:14 nootropics, i've never used arch linux. But if i am not mistaken setup-alpine and setup-box is what you are looking for. 2015-08-03 17:00:33 thanks nmeum :) 2015-08-03 17:01:51 Yeah setup-alpine does everything, nmeum thats for seting up disk 2015-08-03 17:01:56 i just want the packages 2015-08-03 17:01:59 copied to a paritition 2015-08-03 17:40:10 ncopa: Thanks for looking into LXC template issue hopefully the patch (pull request) of mine will be accepted soon 2015-08-03 17:43:04 hey :) 2015-08-03 20:21:17 nootropics: just note we want to replace setup-* with setup-box one day 2015-08-03 20:21:25 oh nice 2015-08-03 20:21:30 writing a small bash librarry atm 2015-08-03 20:21:34 for the project and because lol easy 2015-08-03 20:22:03 so if you really want to rewrite anything, help alacerda with setup-box instead 2015-08-03 20:22:13 im going to lol 2015-08-03 20:23:35 there is no pacstrap equivalent, because there is no need to have it 2015-08-03 20:23:40 ... 2015-08-03 20:23:44 so how do i get packages 2015-08-03 20:23:47 onto the partition.. 2015-08-03 20:23:47 everyone can mount procfs, sysfs and run apk 2015-08-03 20:23:53 apk --help 2015-08-03 20:24:02 did you really check what pacstrap does? 2015-08-03 20:24:05 Yes... 2015-08-03 20:24:12 I know what pacstrap does. 2015-08-03 20:24:17 good 2015-08-03 20:24:18 I'm just saying it'd be nice to have something similar. 2015-08-03 20:24:27 we're not arch 2015-08-03 20:24:37 lol. 2015-08-03 20:26:16 setup-alpine is expected to be semi-friendly, if you need something more, jump into apk-tools help messages 2015-08-03 20:26:53 It doesn't support encryption, I didn't see anywhere on any wiki anywhere about mounting proc/sysfs gentoo style 2015-08-03 20:27:01 even my friend said it's an issue with the documentation 2015-08-03 20:27:08 you're right you're not arch but do you really want to deny the extra users? 2015-08-03 20:28:04 gentoo style? 2015-08-03 20:28:13 Mounting the proc and sysfs. 2015-08-03 20:28:13 yes 2015-08-03 20:28:20 what is gentooish about it? 2015-08-03 20:28:34 it's common stuff for setting any distribution if you don't use the installer 2015-08-03 20:28:57 Why would i not use the installer when the installer works fine 2015-08-03 20:29:10 and before kaniini jumps at you, let me say that for him – we don't want every people in the world to use alpine 2015-08-03 20:29:21 because installers are always fishy 2015-08-03 20:29:43 okay, why dont you want more people to use it barthalion 2015-08-03 20:30:03 and I'm certain I'm not alone here by saying that I don't remember last time I used an installer of any distro 2015-08-03 20:30:17 pacstrap → pacman -r 2015-08-03 20:30:18 Gentoo/*BSD 2015-08-03 20:30:25 debian installer → debootstrap 2015-08-03 20:30:30 are the only ones I've had to manually mount the *fs 2015-08-03 20:32:20 barthalion: you're giving me information I already know. 2015-08-03 20:32:32 and yet you're asking questions 2015-08-03 20:33:15 lol. I'm fully aware arch can be installed manually same with debian, the issue is the ease of use for installing it. 2015-08-03 20:33:48 if you want ease of use, download ubuntu 2015-08-03 20:33:53 ... are you joking? 2015-08-03 20:34:07 KISS is rarely about direct user experience 2015-08-03 20:34:29 lol. yeah ubuntu has a hardened toolchain and eveyrthing compiled against musl libc 2015-08-03 20:34:33 of course 2015-08-03 20:35:34 barthalion: if you dont want more people to use it then you're kind of missing the point 2015-08-03 20:36:10 it's not me who missed the point of not wanting literally everyone to use alpine 2015-08-03 20:36:22 Did I say everyone? 2015-08-03 20:36:24 No I didn't. 2015-08-03 20:36:27 I said MORE people. 2015-08-03 20:36:41 yeah, pacstrap clone is way to go 2015-08-03 20:37:00 lol what do you suggest then 2015-08-03 20:37:02 EOT from my side 2015-08-03 20:37:30 could someone of you please bump claws-mail to 3.12? 2015-08-03 20:37:53 i am on my laptop without dev tools on it an i cannot create a patch here 2015-08-03 20:40:53 leo-unglaub: pushed 2015-08-03 20:41:51 and built 2015-08-03 20:42:03 barthalion: thanks! 2015-08-03 20:42:24 np 2015-08-03 20:42:38 barthalion: " Linux distribution designed for power users who appreciate security, simplicity and resource efficiency." 2015-08-03 20:44:25 nootropics: don't get me wrong, my first patch got accepted 3 years ago 2015-08-03 20:44:36 why does that have anything to do with any of this? 2015-08-03 20:44:42 I'm pretty sure I know what users we want 2015-08-03 20:44:59 LOL 2015-08-03 20:45:02 you decide what users you get? 2015-08-03 20:45:13 did I say get or want? 2015-08-03 20:45:47 wow you got me there! 2015-08-03 20:45:54 going back to my corner of EOT 2015-08-03 20:46:01 ok how's this, no matter what your stance is I'm going to write tools to help make alpine easier to install 2015-08-03 20:46:56 great, don't forget about setup-box :) 2015-08-03 20:47:20 Meh I'm going to focus on first making a bootable image. 2015-08-03 20:50:25 barthalion: what you've basically said is, dont do $x that'll make us get users from !$desiredPool 2015-08-03 20:53:32 barthalion: did you get a change look into the postgres-old-upgrade idea? :) 2015-08-03 20:57:50 chris|: I returned yesterday from Spain, so not really, I'll try to do it tomorrow or day after 2015-08-03 20:59:58 barthalion: if I wrote one bash script that installed alpine with a menu, either encrypted or non encrypted with full package installation, repo setups, internet conectivity checker etc, would that make you happy 2015-08-03 21:02:53 1. setup-box 2015-08-03 21:03:03 2. I don't care, I wouldn't use it anyway 2015-08-03 21:03:05 link to current copy? 2015-08-03 21:03:10 ncopa probably will 2015-08-03 21:03:20 barthalion: you seem to have an issue with people who use installers 2015-08-03 21:03:42 barthalion: sure, no hury, just wanted to make sure the idea was still viable 2015-08-03 21:09:03 I noticed this installer a little while ago - distribution independent https://calamares.io/ 2015-08-03 21:09:16 lol what 2015-08-03 21:09:36 oh nice it's like, uhh whats that one in python 2015-08-03 21:28:41 hey YoursTruly 2015-08-03 21:40:04 nootropics: http://git.alpinelinux.org/cgit/user/alacerda/setup-box/ 2015-08-03 23:15:26 alright so for my ghc port, the latest release of ghc seems to be working well, i have a script to generate a bootstrap apk that can be used to build ghc itself etc.. i'd like to get this into alpine linux proper if at all possible, i was working with kaniini before but not sure whom I should work with to get this in place due to the weirdness of the cross compilation but would anyone be willing to help me out this week 2015-08-03 23:15:27 getting things upstreamed? 2015-08-04 05:34:39 I'm busy all week but will try to get it done for you asap 2015-08-04 06:08:26 kaniini cool no worries, also if there is anything that is a bit off in how i setup the apkbuilds or anything let me know, right now its fairly flexible though 2015-08-04 06:09:07 nah its all good we just need to get it integrated 2015-08-04 06:10:00 sure, and i'm guessing i'll get to be maintainer for it which is no big deal really, i tend to watch releases like a hawk 2015-08-04 06:11:46 and once x86_64 is in place i can work on arm 2015-08-04 06:11:58 i386 i kinda don't care about though tbh 2015-08-04 06:12:23 not sure what the overall policy there is though but i could add it, just don't see much gain 2015-08-04 06:17:11 and i have a wip apkbuild for the newish build tool stack i'm working on getting setup as well 2015-08-04 06:17:41 my thoughts are mostly, get build tools and ghc working then let people sort out dependencies as they see fit 2015-08-04 06:25:36 also curious on one other thing, i'd like to get the port working with llvm as well but ghc 7.10 only supports generating ir compatible with 3.5 so somewhat curious how older versions of things work out, aka if an llvm-3.5 with binaries suffixed say llc-3.5 would be ok to add as well, but thats not a huge priority 2015-08-04 06:28:17 it'd be up to you as maintainer but someone might want it on ia-32 2015-08-04 06:29:18 morning 2015-08-04 06:29:39 i have been thinking of an albootstrap script 2015-08-04 06:29:44 moins ncopa I got your notification on GH 2015-08-04 06:29:53 Waf3lijz3r: mornig 2015-08-04 06:30:00 I was not aware they also use a mailing list for LXC :-) 2015-08-04 06:30:03 sorry ididnt tell you I sent patch to lxc-devel 2015-08-04 06:30:09 no problem 2015-08-04 06:30:24 a lot of people will benefit from this fix 2015-08-04 06:30:39 i have sent stuff there earlier, i asked in #lxcontainers if they prefer git send-email or github pull requests 2015-08-04 06:31:01 distro packagers can always pick that patch 2015-08-04 06:31:20 or build a package for apk-tools 2015-08-04 06:31:36 yeah because it also should be fixed in stable packages 2015-08-04 06:33:38 then it makes sense to push it also to the ML 2015-08-04 07:31:41 kaniini: i wonder what the uwsgi locking problem was about? 2015-08-04 07:31:42 http://git.alpinelinux.org/cgit/aports/commit/main/uwsgi?id=38778fca7461a8afc34109a1e5efc150113ef716 2015-08-04 07:32:04 and http://git.alpinelinux.org/cgit/aports/commit/main/uwsgi?id=90b28189ad9ed879e1bb2d65ace59272e281d05c 2015-08-04 07:32:30 musl didn't provide the function it uses 2015-08-04 07:32:56 how can i test it? 2015-08-04 07:33:32 without it dlopening the python module fails iirc 2015-08-04 07:34:33 ok 2015-08-04 07:35:21 thanks 2015-08-04 08:34:55 hey friends :) 2015-08-04 08:38:56 hi 2015-08-04 09:01:04 this is my /etc/crypttab https://paste.debian.net/plainh/5550c860 2015-08-04 09:01:18 and the swap entry gets decrtypted during boot, but not the data entry 2015-08-04 09:01:20 any ideas why? 2015-08-04 09:01:26 i dont see an error in my syntax 2015-08-04 09:01:31 also the uuid is correct 2015-08-04 09:01:35 i tripple checked it 2015-08-04 09:17:26 does mdadm service start before cryptsetup? 2015-08-04 09:17:55 which service is it that parses crypttab? 2015-08-04 09:18:55 you may need add rc_before=cryptsetup in /etc/conf.d/mdadm 2015-08-04 09:19:14 or rc_after=mdadm in /etc/conf.d/cryptsetup 2015-08-04 09:40:47 in /etc/conf.d/mdadm? 2015-08-04 09:40:50 are you sure about that? 2015-08-04 09:41:01 i think you mean /etc/init.d/mdadm 2015-08-04 09:50:08 no, i mean conf.d 2015-08-04 09:50:31 mdadm-raid 2015-08-04 09:58:11 i dont even have /etc/conf.d/mdadm-raid 2015-08-04 09:58:15 hmm, very strange 2015-08-04 11:19:39 leo-unglaub: you can create it then 2015-08-04 11:20:03 openrc has support to change order as a part of the service config 2015-08-04 11:20:25 echo "rc_before=cryptsetup" > /etc/conf.d/mdadm-raid 2015-08-04 12:06:13 hi 2015-08-04 12:06:26 alpine eudev still broken if you run diskless :( 2015-08-04 12:07:15 says something like: "/dev needs be mounted as devtmpfs" 2015-08-04 12:30:32 AmatCoder: it on bugs already? 2015-08-04 12:31:29 clandmeter: no 2015-08-04 12:31:47 better report it so it doesnt get forgotten 2015-08-04 12:32:09 ok, I will do 2015-08-04 12:37:18 ncopa: stgraber closed the issue but not merged the pull request for master branch ... weird :-) 2015-08-04 13:05:17 Waf3lijz3r: because he picked my identical commit 2015-08-04 13:05:39 https://github.com/lxc/lxc/commit/e00f9e4e7ab5a76cfd4068e09579f9f471666a90 2015-08-04 13:06:18 my guess is he liked my commit message better ;) 2015-08-04 13:08:02 wow, i almost thought that Waf3lijz3r is my current random nick... 2015-08-04 13:10:04 Waf3lijz3r: thanks anyway. i think upstream fixes issues faster when they see that there are many people interested in fixing it 2015-08-04 13:26:42 ncopa: probably :-) 2015-08-04 13:27:03 it is merged in master now whohoo :_) 2015-08-04 13:44:38 hmm, I am thinking about eudev diskless issue... 2015-08-04 13:45:16 'lbu ci' does not save modifications into /etc/init.d folder 2015-08-04 13:46:23 so 'udev', 'udev-postmount' and friends do not survive across reboots 2015-08-04 13:47:22 I believe that is the issue 2015-08-04 13:56:44 nah ,it is not that >( 2015-08-04 13:59:21 I have not clues, I have filled a bug report: Bug #4510 2015-08-04 14:10:57 AmatCoder: unfortunally, eudev requires /dev to be tmpfs 2015-08-04 14:11:10 so you need a fix in initramfs 2015-08-04 14:12:05 basically, eudev does not work with v3.2 2015-08-04 14:12:11 only with edge and v3.3 2015-08-04 14:13:02 I am trying on edge 2015-08-04 14:52:08 interesting, busybox date -R and coreutils date -R give different results too 2015-08-04 19:06:42 is there something like optdepends in apk? 2015-08-04 19:19:28 no 2015-08-04 19:26:47 yeah, stubled upon the subpackages already, thanks though :) 2015-08-04 19:59:34 it's not the same thing 2015-08-04 20:00:16 subpackages exist also in Arch, we just name it differently there 2015-08-04 20:01:21 but yeah, usually you want a split package in case of optional dependency 2015-08-04 20:13:56 that is what I meant by that comment 2015-08-05 01:02:30 ... huh. runit's patched for /run/service but the readme refers to /etc/service 2015-08-05 01:04:02 oh 2015-08-05 01:04:06 that's 2015-08-05 01:04:09 a roundabout way of 2015-08-05 01:06:27 I'm so confused 2015-08-05 01:11:32 i think i put some notes in the docs 2015-08-05 01:13:08 http://hastebin.com/ifuzubiyot.bash 2015-08-05 01:14:24 if you install apk-post-packages I tend to put my notes in /usr/share/doc/$pkg/README.alpine - & when you install the $pkg-doc it prints the README.alpine 2015-08-05 01:15:31 what I'm ?? at is more the fact that afaict /run/service doesn't do anything except be a link for sv to look at 2015-08-05 01:15:58 yes i thought that first of all & did not include the symlink 2015-08-05 01:16:03 because sv is patched to look at /run/service by default specifically 2015-08-05 01:16:16 but without it you cannot start stop restart services with sv 2015-08-05 01:16:22 so then the initscript has to go and set up the link 2015-08-05 01:16:45 read the notes 2015-08-05 01:16:53 in fact the initscript tries to clobber anything that happens to be at /run/service 2015-08-05 01:17:04 what's wrong with just having sv look at /etc/service directly? 2015-08-05 01:17:13 it does 2015-08-05 01:17:57 if you remove the /run/service symlink then sv cannot stop / start services 2015-08-05 01:18:03 http://git.alpinelinux.org/cgit/aports/tree/testing/runit/run-service-dir.patch 2015-08-05 01:18:39 yes 2015-08-05 01:18:55 I just built it to default to /etc/service and it doesn't seem to be overtly broken 2015-08-05 01:20:13 try sv t service-name 2015-08-05 01:22:14 it's fine 2015-08-05 01:31:45 ok - I'll change the patch & the init script 2015-08-05 01:33:16 I can send a patch if you'd like, although it's a p simple change 2015-08-05 01:36:16 http://patchwork.alpinelinux.org/patch/340/ 2015-08-05 01:36:51 thanks for spotting it - it never occurred to me 2015-08-05 01:39:00 https://gist.github.com/muhmuhten/58379109c96d73f24e10 yeah this is what I came up with 2015-08-05 01:39:24 looks, uh, identical 2015-08-05 01:45:03 i've been using runit in lxc - it's very good 2015-08-05 01:46:34 it is 2015-08-05 01:57:19 socklog is quite good too - it has built in log rotation & runs inside runit 2015-08-05 02:02:09 is socklog supposed to be fitted with svlogd on both ends 2015-08-05 02:04:00 it takes over logging for everything 2015-08-05 02:05:06 & runs as user nobody 2015-08-05 02:07:50 oh it's a syslogd 2015-08-05 02:08:45 svlogd runs as user log 2015-08-05 02:09:09 socklog runs as nobody 2015-08-05 02:09:49 the post install script sets up the socklog service 2015-08-05 02:10:55 your logs will then be in http://hastebin.com/oluloxosus.mel 2015-08-05 02:14:58 huh 2015-08-05 02:15:04 I should talk a look at it sometime, then 2015-08-05 02:15:53 it's nice in lxc - one less service running as root 2015-08-05 02:20:10 atm I just have a lot of svlogd processes running as nobody 2015-08-05 02:21:15 i have one svlogd running as log 2015-08-05 02:22:05 one socklog running as nobody 2015-08-05 02:22:33 & runsv running socklog-unix 2015-08-05 03:48:50 oh ugh right I've been running nginx-naxsi instead of stock nginx because it wasn't compiled with some modules I was using 2015-08-05 03:49:01 I could probably just rebuild that 2015-08-05 04:06:37 ammunta - which modules were you missing ? 2015-08-05 04:07:13 the mail proxy ? 2015-08-05 04:15:53 not that one 2015-08-05 04:15:56 real_ip though 2015-08-05 04:18:05 eh I'll just go through the diff 2015-08-05 04:20:54 actually hm it's just that, I can work around that p easily 2015-08-05 07:17:05 #docker 2015-08-05 07:58:16 good morning! 2015-08-05 08:42:54 morning 2015-08-05 13:29:02 i have update problem on stable 3.2.2 linux-vanilla http://sprunge.us/IHAa 2015-08-05 13:47:36 crow: tmpfs root? 2015-08-05 13:48:03 sounds like modloop is mounted 2015-08-05 14:13:12 ncopa no this is disk install, /dev/vg0/apu4c.rootfs on / type ext4 (rw,relatime,data=ordered) 2015-08-05 14:13:34 ok 2015-08-05 14:13:42 what do you have in /lib/modules? 2015-08-05 14:14:06 for some reason it is not able to install files there 2015-08-05 14:14:21 can you check if modloop got mounted? 2015-08-05 14:14:26 mount | grep modloop 2015-08-05 14:55:50 ncopa is not loaded modloop 2015-08-05 14:59:09 http://sprunge.us/ROTj 2015-08-05 15:06:23 ncopa need to go now, lets continue tomorrow please .) 2015-08-05 16:16:08 why is unrar in the main repository? The unrar license isn't a free software license, shouldn't it be in non-free? Or does 'free' only mean 'not propritary'? 2015-08-05 16:20:08 the latter 2015-08-05 16:20:30 we could probably deal with this better if apk-tools get more sophisticated repo management 2015-08-05 16:20:38 but it's up to ncopa and fabled 2015-08-05 20:23:27 does anybody know why the mandoc.db 2015-08-05 20:24:04 …file is removed in the mdocml.trigger file before creating a new one using makewhatis? makewhatis should be able to build the mandoc.db database incremental 2015-08-05 20:26:51 it just slows down package installation a lot if you have a bunch of man pages installed on your system 2015-08-05 21:00:33 no idea 2015-08-05 21:01:44 I can remove it if you say so, I never used mdocml 2015-08-05 21:02:36 I send a patch, I am not 100% sure though but it seems to work for me, maybe somebody else could test my patch? 2015-08-05 21:02:54 s/send/sent/ 2015-08-05 21:07:39 well, looks legit to me 2015-08-05 21:09:22 I'll apply it, I hope ncopa won't shout too much at me 2015-08-05 21:09:31 did you test it? Did it work for you as well? 2015-08-05 21:11:17 first things first, I'm doing it right now :P 2015-08-05 21:16:25 it doesn't bring any significant improvement here 2015-08-05 21:16:35 but I'm spoiled brat with ssd 2015-08-05 21:16:52 it doesn't make things worse either 2015-08-05 21:19:21 maybe it is even better to invoke makewhatis(8) as a cron? 2015-08-05 21:19:29 then it wouldn't block package installation at all 2015-08-05 21:20:17 huh, I just got an error form the mdocml trigger during package installation 2015-08-05 21:20:38 yelling ncopa ahead 2015-08-05 21:20:42 what error? 2015-08-05 21:21:08 triggers are already deferred at the end of packages install, so I think it's fine as it is 2015-08-05 21:21:35 oops I actually fucked it up. /usr/share/man has to be the last argument to makewhatis. Don't know why I didn't notice this before 2015-08-05 21:21:58 you didn't merge this yet, did you? ;_; 2015-08-05 21:22:03 I did, but it's edge 2015-08-05 21:22:09 welcome to world of broken systems :P 2015-08-05 21:22:10 fixing it now 2015-08-05 21:22:26 I am sorry 2015-08-05 21:22:39 nothing to be sorry about 2015-08-05 21:25:17 alright, I'm off to bed 2015-08-05 21:25:19 goodnight 2015-08-05 21:25:24 night 2015-08-05 21:28:04 hm, it doesn't even build anymore (wtf?). I will take a look at this tomorrow and I still think that using a cron to generate the mandoc.db is a good idea. I believe other distributions also do it that way to speedup package installation 2015-08-05 21:30:05 it's never patch author's fault, it's commiter fault because he/she did unreliable code review 2015-08-05 21:30:38 maybe, I guess whatis isn't very crucial to have it immediately 2015-08-05 21:30:54 but cron job has disadvantage of being run even when it's not needed 2015-08-05 21:38:54 while a trigger runs once, exactly when it is required 2015-08-06 06:39:21 morning 2015-08-06 06:39:57 the reason i remove the mandoc.db is that it was only adding stuff, never removing 2015-08-06 06:40:23 so it would leave entries of removed packages 2015-08-06 06:40:46 i was planning to write a feature request for mdocml for a cleanup function 2015-08-06 06:41:09 go over the db and verify that all entries has a file on file system, and remove those who has not 2015-08-06 07:35:18 hum 2015-08-06 07:35:21 it seems to work 2015-08-06 07:36:52 wut, I broke checksums and build didn't fail? 2015-08-06 07:39:22 because pkgrel was not bumpted to builder didnt try to build the fix 2015-08-06 07:46:24 ah 2015-08-06 08:14:07 please git pull git://git.alpinelinux.org/user/fab/aports 2015-08-06 08:14:15 there are a couple of updates 2015-08-06 08:14:40 thanks 2015-08-06 09:17:19 why we need upgrade libsoup? 2015-08-06 09:17:38 i normally upgrade gnome stuf to stable only 2015-08-06 09:17:47 eg 2.50 -> 2.52 2015-08-06 09:18:12 also 2015-08-06 09:18:13 >>> libsoup: Checking sanity of /home/ncopa/aports/main/libsoup/APKBUILD... 2015-08-06 09:18:13 >>> ERROR: libsoup: libsoup-2.51.3.tar.xz is missing in checksums 2015-08-06 09:23:52 -url="http://www.hexchat.org" 2015-08-06 09:23:52 +url="https://hexchat.github.iol" 2015-08-06 09:26:32 faffolter: i rebased them against current master, fixed 2 checksums (libsoup and youtube-dl) and fixed the typo in hexchat url 2015-08-06 09:26:36 faffolter: Thanks! 2015-08-06 09:41:34 morning 2015-08-06 10:15:51 ncopa: it seems that i'm out of practice doing the packaging thing. so it seems to work with my personal move to gitolite 2015-08-06 10:23:44 a coworker asked me to update nuttcp. the problem is that upstream is non-responsive and don't provide a tarball for the latest version, only the source 2015-08-06 10:25:21 nuttcp-6.1.2 is in testing and he is looking for 7.3.3. would it be an option to replace the source in the tarball? 2015-08-06 10:44:36 ncopa do you have any more idee for my upgrade problem? 2015-08-06 11:07:02 faffolter: http://nuttcp.net/nuttcp/nuttcp-7.3.3/ 2015-08-06 11:07:31 it is probably a single file 2015-08-06 11:09:32 crow: i have no idea. check why it can not write to /lib/modules 2015-08-06 11:14:11 faffolter: i pushed nuttcp 7.3.3 2015-08-06 12:50:00 hey friends :) 2015-08-06 12:51:56 sup 2015-08-06 13:02:18 ncopa i dont have that much information now when doing update && upgrade http://sprunge.us/SJdM 2015-08-06 13:11:18 crow: try: apk fix 2015-08-06 13:15:50 ncopa it worked, why i dont understand: http://sprunge.us/bFOA 2015-08-06 13:16:17 i did not jet reboot the box. 2015-08-06 14:21:54 ncopa: thanks for nuttcp 2015-08-06 14:26:20 faffolter: np 2015-08-06 18:03:01 quick question, is there any documentation or otherwise simple workflow on how to work on snapshot APKBUILDs if you are not a developer, aka. do not have access to dev.alpinelinux.org? 2015-08-06 18:12:44 chris|: not 100% sure what you mean with 'snapshot APKBUILDs' but incase you want to submit patches for APKBUILDs see http://wiki.alpinelinux.org/wiki/Creating_patches 2015-08-06 18:14:05 nmeum: submitting patches is not the problem, the problem is testing a patch that relies on a snapshot 2015-08-06 18:22:40 <_mjones> git pull the aports tree. 2015-08-06 18:23:08 <_mjones> You can select what revision to use from git. Or am I misunderstanding your need? 2015-08-06 18:25:41 I think chris| is trying to use an unreleased upstream version in source= and thus needs to upload it to dev.alpinelinux.org/archive, but I am not sure either… 2015-08-06 18:26:08 nmeum: yes, indeed 2015-08-06 18:26:19 <_mjones> What package? 2015-08-06 18:26:34 I am under the assumption that that is what abuild snapshot is used for 2015-08-06 18:26:41 <_mjones> You mean you need it uploaded in order to submit an updated APKBUILD? 2015-08-06 18:27:40 you don't need to upload it, the snapshot function does that but you need to test it somehow 2015-08-06 18:27:57 _mjones: I was hoping for a way to work of a local cache or similar first, actually 2015-08-06 18:27:59 the easiet way to test it is to simply use your own server instead of dev.alpinelinux.org and replace the urls accordingly 2015-08-06 18:28:39 <_mjones> If I understand the need properly, then what nmeum said. 2015-08-06 18:28:40 nmeum: yeah, I am already preparing that since there seems to be no way around it 2015-08-06 18:28:40 or just generate the snapshot archive but don't upload it and instead specify the tarball in source 2015-08-06 18:29:11 you could also just specify the file in source= and not upload it at all (for testing purposes of cause) 2015-08-06 18:29:32 after all snapshot() only generates an archive using git-archive(1) 2015-08-07 06:02:01 <_mjones> Never heard of this distro: http://antergos.com/ Slick website. 2015-08-07 06:18:25 what's it do 2015-08-07 06:19:13 oh it's another arch deriv 2015-08-07 06:20:06 <_mjones> I don't know what it is except a rolling-release and I've never heard of it. 2015-08-07 06:23:09 it's Arch with gui installer and one additional repo 2015-08-07 06:23:15 nothing interesting 2015-08-07 06:23:24 quite regular arch 2015-08-07 06:25:16 <_mjones> I suppose the gui installer is a value-add since the usual arch install is....retro. 2015-08-07 06:25:41 <_mjones> Retro in a good way. Really takes me back. 2015-08-07 06:27:01 no installer is the best installer ;) 2015-08-07 06:39:32 hello 2015-08-07 06:39:46 hi algitbot 2015-08-07 06:40:16 algitbot: how are you today? 2015-08-07 06:40:22 lol 2015-08-07 07:23:38 fcolista: found a new friend? 2015-08-07 07:23:52 looks like...but he's sad 2015-08-07 07:24:17 its a he? 2015-08-07 07:24:38 umh. We should ask to the father, ncopa.. 2015-08-07 07:25:02 i could ask the same question again ;-) 2015-08-07 07:25:36 how many ppl actually met each other irl here? 2015-08-07 07:34:34 I met Roger 2015-08-07 07:35:31 pretty much it 2015-08-07 07:35:54 thats one more then me 2015-08-07 07:46:17 clandmeter, re: installflash 2015-08-07 07:46:31 there are versioning now 2015-08-07 07:46:34 *there is 2015-08-07 07:53:23 fcolista: ? 2015-08-07 07:53:36 do we actually still need it? 2015-08-07 07:53:37 https://github.com/WnP/InstallFlash/releases/tag/v1.2.0 2015-08-07 07:53:43 dunno 2015-08-07 07:54:04 i was going to add patches sent against installflash and i've realized that 2015-08-07 07:54:12 manually copying it will work jsut fine 2015-08-07 07:54:29 there is a howto on wiki 2015-08-07 07:54:49 do you think we can simply remove the package? 2015-08-07 07:55:01 im not against this approach, just wonder if its worth the hassle to maintain it. 2015-08-07 07:55:15 there's no maintaner actually :) 2015-08-07 07:55:26 ok move it and add a link in git to the wiki article 2015-08-07 07:55:42 move it to unmaintained i mean :) 2015-08-07 07:56:10 if the dev wants it maintained, he can send us patches. 2015-08-07 07:56:28 "ok move it and add a link in git to the wiki article" 2015-08-07 07:56:41 which link ? 2015-08-07 07:57:01 http://wiki.alpinelinux.org/wiki/Installing_Adobe_flash_player_for_Firefox 2015-08-07 07:57:25 there is an apkbuild in non-free 2015-08-07 07:57:33 ah ok 2015-08-07 07:57:35 so you can make an apk and install the plugin with apk 2015-08-07 07:58:12 http://git.alpinelinux.org/cgit/aports/tree/non-free/adobe-flashplayer/APKBUILD 2015-08-07 07:58:16 yes 2015-08-07 07:58:45 so, non need of InstallFlash on testing actually 2015-08-07 07:59:06 i wonder if we have any users that really need an apk to install this :) 2015-08-07 07:59:29 well 2015-08-07 07:59:36 we don't have by default non-free repo 2015-08-07 07:59:46 so the firs approach would be "apk search flash" 2015-08-07 07:59:51 *first 2015-08-07 08:00:30 since it does not return adobe flash, someone would add testing repo and re-do the same 2015-08-07 08:00:37 and finds installflash 2015-08-07 08:00:37 well, installflash is surely nice 2015-08-07 08:00:47 but I don't see any reason to have it written in go 2015-08-07 08:00:56 non-free repo is not advertised 2015-08-07 08:01:18 its non free, so we cannot provide the apk 2015-08-07 08:01:23 users need to build it themselves 2015-08-07 08:01:23 +1 barthalion 2015-08-07 08:01:53 but setting up a dev env is much more work then just follow the wiki steps. 2015-08-07 08:01:54 this is what i really _hate_ : go get github.com/WnP/InstallFlash || return 1 2015-08-07 08:02:13 it should be written in posix shell, go is overkill 2015-08-07 08:02:27 fcolista: the apkbuild is probably broken 2015-08-07 08:02:32 it is clandmeter 2015-08-07 08:03:09 its go causing headaches 2015-08-07 08:03:29 thinking im using the src, but in the end it just pulls git... 2015-08-07 08:03:56 you really need to use a .tar.gz for that purpose 2015-08-07 08:04:03 its up to the developer to decide what language he want to use. for sure go is overkill, but its his choice. 2015-08-07 08:04:16 and its my choice not to support go in aports :) 2015-08-07 08:07:19 ok..this go is bloating and breaks apkbuild. 2015-08-07 08:07:32 I'm not going to apply the patch, and i'll leave it as-is 2015-08-07 08:10:53 ncopa, do you agree with %339 ? 2015-08-07 08:10:54 339 New [alpine-aports] main/busybox: enable support for syslog.conf 2015-08-07 08:12:54 ok with me 2015-08-07 08:18:02 ok. 2015-08-07 08:29:38 patchwork donw 2015-08-07 08:29:41 *done 2015-08-07 08:30:44 re: glock and dockerize (358 and 359), snapshot made by the Chris does not upload the file correctly, and i don't have time/willingness to debug it. Just rejected asking to use the builtin snapshot function of abuild. 2015-08-07 08:33:46 I've not applied %353 since build fails (but is not due to the patch....the apkbuild is broken) 2015-08-07 09:22:19 stateless 2015-08-07 09:22:46 stateless: i have an issue with firefox and fortify 2015-08-07 09:23:52 stateless: http://tpaste.us/3lMw 2015-08-07 09:24:11 if i remove the /usr/include/fortify dir it will continue to build 2015-08-07 09:33:18 ok i have a workaround 2015-08-07 10:04:40 fcolista: well, I cannot debug the upload issues myself for obvious reasons ;) 2015-08-07 10:04:59 I tested it locally and it worked fine so I submitted the patches 2015-08-07 10:08:37 ncopa, what's the workaround? 2015-08-07 10:09:39 http://tpaste.us/GdjJ 2015-08-07 10:09:54 ah 2015-08-07 10:10:03 i add an #ifndef _SYSTEM_WRAPPER_... 2015-08-07 10:10:16 always fun with generated code... 2015-08-07 10:14:06 something else breaks... 2015-08-07 10:14:16 will have to check after lunch 2015-08-07 10:48:09 could somebody please close/reject patch 324 for me or can I do that on my own somehow? 2015-08-07 11:10:31 stateless: do you know if the pragma issue is fixable in fortify headers? 2015-08-07 11:12:17 my workaround only works if the wrapped headers is explicitly pulled in by firefox headers 2015-08-07 11:12:37 not sure, I will have to investigate 2015-08-07 11:12:43 I am not sure it is 2015-08-07 11:12:54 and things still break if system headers pulls in wrapped headers 2015-08-07 11:16:33 ok, i might have a workaround for fortify-headers 2015-08-07 11:28:33 hum 2015-08-07 11:28:34 nope 2015-08-07 11:28:38 i dont know how to fix it 2015-08-07 11:28:56 this is why they wrap all system headers: https://bugzilla.mozilla.org/show_bug.cgi?id=273336 2015-08-07 11:50:52 ok, so maybe just disable fortify -headers on firefox for now :-/ 2015-08-07 11:53:30 what's the difference between the git-manpages and git-doc package. If I am not mistaken they ship the same manpages except that the man pages shipped by git-manpages are not compressed with gzip. Is that intentional? 2015-08-07 11:55:37 ncopa, re : %361 and all the snapshot() function (see %359 and %359), looks that there are permissions problem when rsync is going to copy the file. 2015-08-07 11:55:56 ACTION is going to paste the error 2015-08-07 11:57:13 ACTION vomits because of go 2015-08-07 11:57:20 Enter passphrase for key '/home/fcolista/.ssh/id_dsa': 2015-08-07 11:57:20 sending incremental file list 2015-08-07 11:57:20 consul-0.5.2.tar.gz 2015-08-07 11:57:20 32,310,893 100% 6.95MB/s 0:00:04 (xfr#1, to-chk=0/1) 2015-08-07 11:57:20 rsync: mkstemp "/archive/consul/.consul-0.5.2.tar.gz.AaTyhr" failed: Permission denied (13) 2015-08-07 11:57:45 what's the user rsync/abuild uses for upload the file? 2015-08-07 11:58:56 fcolista: what's your problem with go? 2015-08-07 11:59:16 it's a nightmare packaging the software 2015-08-07 11:59:42 yeah, it's kind of gitolite 84 logstatsplain 2015-08-07 11:59:42 d--------- glade 36 logstatsplain 2015-08-07 11:59:42 d--------- glade3 36 logstatsplain 2015-08-07 11:59:42 d--------- glamor-egl 36 logstatsplain 2015-08-07 11:59:45 oops 2015-08-07 12:00:07 I wanted to say that it is kind of difficult but possible 2015-08-07 12:01:48 it is pretty much straightforward if you include the dep libs in the source tarball, but otherwise it is very much a hell I hoped had been reserved for java 2015-08-07 12:02:52 chris|, are you the consul patcher ? 2015-08-07 12:03:00 yes 2015-08-07 12:03:05 ah ok 2015-08-07 12:03:08 hi :) 2015-08-07 12:03:58 well…you could also just create a package for each dependency and build each go package using the same GOPATH e.g /usr/lib/gopath 2015-08-07 12:04:09 hi :) 2015-08-07 12:05:48 could being the operative word here because if you look at go software, it is pretty much a given not to depend on a version of something but rather use a git hash as reference 2015-08-07 12:06:41 packaging a popular lib that way is pretty much a guarantee for disaster if you must make if conform to the requirements of all packages you want to build against it 2015-08-07 12:06:46 imho 2015-08-07 12:07:12 indeed, I totally forgot about the versioning issue 2015-08-07 12:12:53 hey friends 2015-08-07 12:13:01 have you seen the firefox emergency update? 2015-08-07 12:17:59 leo-unglaub: i saw firefox emergency 2015-08-07 12:18:01 yes 2015-08-07 12:18:08 very scary indeed 2015-08-07 12:18:15 i am trying to build firefox as we speak 2015-08-07 12:18:24 but bumped into issues related fortify-headers 2015-08-07 12:21:21 which one? 2015-08-07 12:21:22 interresting, i bumped into issues compiling aurora 2015-08-07 12:21:27 I mean, what emergency 2015-08-07 12:21:53 ah, okay 2015-08-07 12:21:55 found it 2015-08-07 12:23:27 barthalion: some "explore and upload files from the filesystem" 2015-08-07 12:23:38 do to an sandbox error 2015-08-07 12:23:56 what annoys me much is 2015-08-07 12:24:05 it happens in the embeded pdf reader 2015-08-07 12:24:14 and you cannot disable the pdf reader 2015-08-07 12:24:32 you cannot exclude the code for pdf reader 2015-08-07 12:24:39 i actually tried 2015-08-07 12:24:53 we used to have a firefox-pdfjs subpackage 2015-08-07 12:25:06 but if that was not installed, werid things happened with firefox 2015-08-07 12:27:44 ncopa: yeah, ich agree with you 1000000% 2015-08-07 12:28:04 i hate those big not correctly modularized programs so fucking much 2015-08-07 12:28:16 its horrible what is in a browser now adays 2015-08-07 12:28:35 getting firefox with just html and js rendering without audio+video codecs is impossible 2015-08-07 12:28:48 but the really sad thing: Chrome is still worse! 2015-08-07 12:28:56 and without pdf reader embedded 2015-08-07 12:29:12 yeah, or without sync or developer tools, .. 2015-08-07 12:29:57 it would have been great if firefox had support for some kind of plugins/addos so you could get such additional functionallity if you wanted 2015-08-07 12:30:00 :) 2015-08-07 12:30:40 yeah, but that is not goiung to happen because if they did all those "news sites" would bash firefox for not having all the features 2015-08-07 12:35:07 chrome has sync, developer tools and embedded pdf reader 2015-08-07 12:35:14 as well as chromium 2015-08-07 12:36:14 but jesus, this one is insane 2015-08-07 12:36:19 I'm back to links 2015-08-07 12:36:38 barthalion: yeah, it is horrible 2015-08-07 12:36:55 when they introduced the sandboxing for that i told people that this is going to crash big time 2015-08-07 12:37:05 and they just made fun of me and called me paranoid 2015-08-07 12:37:18 what could possible go wrong?? 2015-08-07 12:39:29 I'm actually not sure if Firefox does any sandboxing on Linux 2015-08-07 12:39:44 probably not, it's Chromium that does a lot about it 2015-08-07 12:42:27 click on a link to run an application from network on your local computer. what could possibly go wrong? 2015-08-07 12:43:38 i have a ff update soonish 2015-08-07 12:43:50 i wonder if i should rename firefox to firefox-esr while at it 2015-08-07 12:44:02 sounds fine 2015-08-07 12:44:25 actually no 2015-08-07 12:44:31 people who just do apk upgrade would not get it 2015-08-07 12:44:48 and they'd run vulnerable til the apk add firefox-esr && apk del firefox 2015-08-07 12:44:49 ncopa: no, dont remane it 2015-08-07 12:45:08 we sayed that firefox stays esr and aurora becomes firefox-next or something like that 2015-08-07 12:45:41 i think I'll update testing/firefox to 39.0.3 though 2015-08-07 12:46:19 ah, okay 2015-08-07 12:46:24 it should be stable enought 2015-08-07 12:48:19 ncopa: this is also something awesome! https://github.com/xoreaxeaxeax/sinkhole 2015-08-07 12:48:41 x86 design flaw makes basically all virtual maschines on x86 attackable 2015-08-07 12:48:45 and it is not patchable *g* 2015-08-07 12:54:32 its a pdf... why am i scared to open that... 2015-08-07 12:55:50 haha, it is not relaed to the firefox issue. but i also opened it only on my phone *g* 2015-08-07 12:56:45 does that mem sinkhole affect 32 bit x86 only or x86_64 too? 2015-08-07 12:57:42 i am still trying to fully understand it. this blackhat stuff is so fucking complicated 2015-08-07 12:58:46 the pdf is worth reading 2015-08-07 12:58:53 the examples are awesome 2015-08-07 13:01:08 when i have upgraded firefox 2015-08-07 13:49:42 phew, my laptop shouldn't be vulnerabile since it's based on sandy bridge 2015-08-07 14:33:53 could somebody please close/reject patch 324 for me or can I do that on my own somehow? 2015-08-07 14:42:46 nmeum, done. 2015-08-07 14:42:51 thank you 2015-08-07 15:29:24 what's the difference between the git-manpages and git-doc package. If I am not mistaken they ship the same manpages except that the man pages shipped by git-manpages are not compressed with gzip. Is that intentional? 2015-08-07 15:29:36 s/./?/ 2015-08-07 16:16:16 <_mjones> Core 2 is quite old at this point. 2015-08-07 16:18:53 <_mjones> Of course the upgrades have had less and less raw performance improvement each time. Down here at 14nm and ~4GHz, things get difficult. 2015-08-07 16:21:19 <_mjones> On an unrelated note, I really could have used a uefi installer for alpine last week. Current best practice for other distros seems to be to ship a dual image that will boot either way. Is there a plan for uefi boot, and what work needs to be done? 2015-08-07 19:53:59 which software is being used for pkgs.alpinelinux.org? 2015-08-07 20:38:34 nmeum: lua 2015-08-07 20:38:51 turbo lua to be more precise 2015-08-07 20:39:06 that's not a software, that's a programming language :p 2015-08-07 20:39:36 its not a piece of software 2015-08-07 20:39:58 what is it then? 2015-08-07 20:40:26 I was just interessted in seeing the code or is it closed source? 2015-08-07 20:40:29 trubo lua is :) 2015-08-07 20:40:39 http://www.turbolua.org/ 2015-08-07 20:41:13 never heard about turbolua but it seems to be a framework 2015-08-07 20:41:20 it is 2015-08-07 20:41:32 https://github.com/clandmeter/aports-turbo 2015-08-07 20:41:52 that's what I was looking for, thanks :) 2015-08-07 20:42:04 we use lua for lots of things in alpine 2015-08-07 20:42:12 yeah, I noticed that 2015-08-07 20:42:14 of which pkgs is one 2015-08-07 20:43:01 apt-get is so slow.... 2015-08-07 20:43:25 im using crouton to get ubuntu installed on my new chromebook 2015-08-07 20:43:53 ones i find out how thats down, ill be happy to replace it with an alpine chroot :) 2015-08-08 17:41:33 Should shells (like mksh, dash, …) be installed to /usr/bin or /bin? Because some shells (oksh, zsh, bash, …) are installed to /bin while others (mksh, dash, …) are installed to /usr/bin 2015-08-08 17:42:45 <_mjones> Depends on whether they are compiled static. 2015-08-08 17:43:04 <_mjones> Although there are some historic reasons for this, and thus room for argument, 2015-08-08 17:43:33 <_mjones> the static programs go in /bin and /sbin, and the dynamics in /usr/bin and /usr/sbin/ 2015-08-08 17:44:16 <_mjones> On alpine, base shell (shells?) are busybox anyway, and thus static, and should be in /bin/. 2015-08-08 18:09:55 is anybody able to build the testing/mksh aport? Fails for me… 2015-08-08 18:10:00 s/mksh/oksh/ 2015-08-08 18:44:09 <_mjones> Seems to me I looked at that one the other day. 2015-08-08 18:44:12 <_mjones> Let's see. 2015-08-08 18:46:24 <_mjones> No, it fails. Looks like fortify headers are responsible. 2015-08-09 04:37:00 _mjones: alpine busybox isn't actually even static 2015-08-09 15:14:12 hi 2015-08-09 15:14:21 why is /var/empty in armhf not created? 2015-08-09 15:14:32 I am using openntpd and it won't start 2015-08-09 15:18:44 openssh creates it at least 2015-08-09 15:50:31 <_mjones> I'm not familiar with that behavious of openssh. 2015-08-09 15:50:42 <_mjones> What platform are you using with the armhd load? 2015-08-09 15:50:49 <_mjones> armhf even? 2015-08-09 15:51:40 <_mjones> ACTION is eyeing those odroids. 2015-08-09 16:04:01 rpi 1 modle b 2015-08-09 16:04:03 model* 2015-08-09 16:04:38 I added in conf.d/ntpd NTPD_OPTS="-s", but it is not used 2015-08-09 16:07:37 <_mjones> I don't understand about /var/empty. It's not on any of my other platforms. 2015-08-09 16:36:21 _mjones: I found the error 2015-08-09 16:37:12 openssh checks for /var/empty, see: http://git.alpinelinux.org/cgit/aports/tree/main/openssh/sshd.initd 2015-08-09 16:37:38 openntpd does it not => when I use openntpd without openssh, the error reveals :) 2015-08-09 16:42:55 <_mjones> You mean when you install openntpd but haven't got openssh installed. 2015-08-09 16:43:04 <_mjones> I see, I prefer dropbear for many things. 2015-08-10 04:44:35 hello 2015-08-10 04:44:50 make segfaults and i don't know the reason 2015-08-10 04:45:02 i tried apk fix make but no help 2015-08-10 04:50:08 ? 2015-08-10 06:10:25 morning 2015-08-10 06:10:42 tpanmajia: wht version 2015-08-10 06:10:47 what arch 2015-08-10 06:10:51 x86_64? 2015-08-10 06:10:53 x86? 2015-08-10 06:11:03 alpine edge? v3.2? 2015-08-10 06:11:11 how can i reproduce? 2015-08-10 06:12:26 alpine 3.2.2 on i486 2015-08-10 06:13:30 the partition where alpine is installed is crypted 2015-08-10 06:14:52 type: LUKS1 cipher: aes-xts-plain64 keysize: 256 bits 2015-08-10 06:15:09 can you run make --help? 2015-08-10 06:17:48 what does cmake say? 2015-08-10 06:17:51 dmesg 2015-08-10 06:17:56 i mean what does dmesg say? 2015-08-10 06:18:07 32bit make works for me in a container 2015-08-10 06:18:40 there is new ABI incompat ncurses 2015-08-10 06:18:42 interesting 2015-08-10 06:42:35 oh 2015-08-10 06:54:23 ncopa, what do you think about %364? Can I apply? 2015-08-10 06:57:33 %364 2015-08-10 06:58:43 #4462 2015-08-10 06:59:54 the idea sounds good to me 2015-08-10 07:00:08 but we need to create the chrony user and group at install time too 2015-08-10 07:00:19 the pkgusers/pkggroups only creates those at build time 2015-08-10 07:00:23 so we need pre-install script 2015-08-10 07:01:13 +1. I'm going to add it 2015-08-10 07:01:20 thanks! 2015-08-10 07:08:34 ncopa, what do you think? http://sprunge.us/MHSj 2015-08-10 07:08:48 Since is chrony and is in main, i'd like your feedback before pushing 2015-08-10 07:10:50 are the last hunk needed? 2015-08-10 07:11:00 chown ? 2015-08-10 07:11:08 yes 2015-08-10 07:11:21 have you tested that its needed? 2015-08-10 07:11:39 i think so. If you are going to upgrade, /var/log/chrony belongs to root 2015-08-10 07:11:50 sometimes a process opens the logfile before dropping privs 2015-08-10 07:11:59 what about chrony.drifts ? 2015-08-10 07:12:05 which is in etc ? 2015-08-10 07:12:29 in theory /etc is wrong in first place 2015-08-10 07:12:29 How do I know that privs are dropped, how can i check? 2015-08-10 07:12:39 either read the code, or test 2015-08-10 07:13:05 maybe .drift should be in /var/lib/chrony 2015-08-10 07:13:25 i suppose chown /var/log/chrony is ok 2015-08-10 07:13:26 yes 2015-08-10 07:13:36 drift belongs in /var 2015-08-10 07:13:49 +1 2015-08-10 07:13:55 i think it is in /etc for convenience reasons for tmpfs installs 2015-08-10 07:14:09 umh, yes, right 2015-08-10 07:14:58 i suppose it is not critical to include it in apkovl 2015-08-10 07:15:29 i think it was more usefule back in the days you had dialup connection 2015-08-10 07:16:03 so, what's the cleanest thing to do? a post-install who add .drift to lbu, or leave .drift in /etc, fixing permissions? 2015-08-10 07:16:32 and modify .conf as well 2015-08-10 07:16:45 cleanest is probably to have drift in /var/lib and simply not add it to lbu at all 2015-08-10 07:16:59 ok. 2015-08-10 07:17:06 check where other distros put drift 2015-08-10 07:17:18 /var/lib 2015-08-10 07:17:20 ok 2015-08-10 07:17:21 already checked 2015-08-10 07:17:24 sounds good then 2015-08-10 07:17:31 gotta modofy default conf then 2015-08-10 07:17:39 yeah 2015-08-10 07:17:41 hum 2015-08-10 07:17:47 that might be a problem when upgrading 2015-08-10 07:17:51 exactly 2015-08-10 07:18:10 maybe create a symlink ? 2015-08-10 07:18:50 umh. Is not a definitive solution, though 2015-08-10 07:19:46 one option is to have a pre/post install script that checks if /etc/chrony/chrony.drift exists 2015-08-10 07:19:58 and if it does, it move it to /var/lib/ and create symlink 2015-08-10 07:20:57 but i dont like that either 2015-08-10 07:21:25 because then there is no way to actually keep the drift in /etc/chrony if that is what you really want 2015-08-10 07:21:37 alpine shoulnd not try get in the way 2015-08-10 07:21:54 umh. Is not a definitive solution, though 2015-08-10 07:22:04 sorry :) 2015-08-10 07:22:09 ok 2015-08-10 07:22:12 how about this 2015-08-10 07:22:45 if pre-install/pre-upgrade detects an existing /etc/chrony/chrony.drift, then it will chown that file only 2015-08-10 07:23:02 oh 2015-08-10 07:23:09 i think have the solution 2015-08-10 07:23:27 we let chrony own /etc/chrony 2015-08-10 07:23:44 that's what the patch i sent you does :) 2015-08-10 07:23:49 but we chown root:root /etc/chrony/* 2015-08-10 07:23:58 oh 2015-08-10 07:24:05 excetl .drift 2015-08-10 07:24:10 except* 2015-08-10 07:24:42 so, why not simply chown chrony:chrony /etc/chrony/chrony.drift ? 2015-08-10 07:25:00 because it might not exist at buidl time 2015-08-10 07:25:11 it is 2015-08-10 07:25:15 ok 2015-08-10 07:25:19 touch "$pkgdir"/etc/chrony/chrony.drift 2015-08-10 07:25:24 that might work 2015-08-10 07:25:25 this is part of APKBUILD 2015-08-10 07:25:39 chown chrony:chront chrony.drift might work then 2015-08-10 07:26:11 thats better 2015-08-10 07:26:15 touch "$pkgdir"/etc/chrony/chrony.drift && chown chrony:chront chrony.drift ? 2015-08-10 07:26:20 yes 2015-08-10 07:26:25 i am not convinced that it will work 2015-08-10 07:26:31 but if it does it will be better 2015-08-10 07:26:50 in case chrony daemon unlinks it and recreate it, then it will not work 2015-08-10 07:27:26 right 2015-08-10 07:27:31 i suppose reading the code is the only way 2015-08-10 07:27:56 if it only does it at startup then we might get away with touch && chown in init.d script 2015-08-10 07:28:41 if it "flushes" drift once in a while by unlink and recreate then we have a problem with that 2015-08-10 07:31:45 what it does, according with reference.c, is creating a new tmp drift copy and then copy the file to the correct drift 2015-08-10 07:31:58 for what i can understand (I'm not a C coder) 2015-08-10 07:32:31 reference.c, line 375, funcion "update_drift_file" 2015-08-10 07:34:09 ok 2015-08-10 07:34:21 then will it not be enough to simply chown the file 2015-08-10 07:34:37 it sets permission, moreover 2015-08-10 07:34:37 the chrony process needs write acess to the dir 2015-08-10 07:34:42 yes 2015-08-10 07:34:59 so let's go for chown chrony the etc dir 2015-08-10 07:35:16 and chown root the file !chrony.drift 2015-08-10 07:35:25 all the files !chrony.drift 2015-08-10 07:35:27 sounds good 2015-08-10 07:35:49 write a comment in APKBUILD why we do it that way 2015-08-10 07:35:50 also 2015-08-10 07:35:56 make default config use /var/lib 2015-08-10 07:36:15 this is only for backwards compat 2015-08-10 07:36:40 ok. Gotta set /var/lib/chrony owned by chrony as well 2015-08-10 07:40:55 that's it: http://tpaste.us/240w 2015-08-10 08:43:12 yes 2015-08-10 08:43:31 do you need change default config to make it set path to drift file to /var/lib? 2015-08-10 08:46:29 yes 2015-08-10 08:46:41 default you mean the one we ship? 2015-08-10 08:47:18 Gotta check actually if using the .conf file shipped with the package is better 2015-08-10 08:47:49 btw: + chown $pkgusers:$pkggroups "$pkgdir"/etc/chrony/chrony.drift 2015-08-10 08:47:52 this is wrong 2015-08-10 08:48:07 because we "+ touch "$pkgdir"/var/lib/chrony/chrony.drift" 2015-08-10 08:48:21 and since "$pkgdir"/etc/chrony/chrony.drift does not exists, build fails. 2015-08-10 08:48:33 and we don't want to touch "$pkgdir"/etc/chrony/chrony.drift 2015-08-10 08:49:03 then i think a post-install script like this would work also for upgrade: 2015-08-10 08:49:21 if [ -f /etc/chrony/chrony.drift ], then chown chrony:chrony /etc/chrony/chrony.drift; fi 2015-08-10 09:08:56 why do we need that? 2015-08-10 09:09:37 upgrade? 2015-08-10 09:09:56 yes, why is it needed when upgrading? 2015-08-10 09:10:30 ii understand correctly, at the moment chrnoy.drift belongs to root 2015-08-10 09:10:50 when upgrade, conf file is not overwritten with the new path 2015-08-10 09:11:15 so chrony is not able to read chrony.drift since run as chrony user 2015-08-10 09:11:20 am i wrong? 2015-08-10 09:33:18 is that a config option? 2015-08-10 09:33:33 isnt run-as-non-root a config option? 2015-08-10 09:34:59 --with-user=USER Specify default chronyd user [root] 2015-08-10 09:35:06 this is the configure option 2015-08-10 09:36:13 hum 2015-08-10 09:36:14 ok 2015-08-10 09:37:25 i am not sure how to deal with 2015-08-10 09:37:28 it 2015-08-10 09:37:34 yeah 2015-08-10 09:37:41 i suppose if [ -f /etc/chrony/chrony.drift ], then chown chrony:chrony /etc/chrony/chrony.drift; fi should work 2015-08-10 09:38:12 since we have /etc/chrony dir set as root:root 2015-08-10 09:38:35 ok 2015-08-10 09:38:45 but the udpate will change permissions of /etc/chrony? 2015-08-10 09:39:11 no 2015-08-10 09:39:26 is already set as root:root 2015-08-10 09:40:05 only drift file is going to be touched with chrony permissions 2015-08-10 09:40:08 but didnt we conclude earlier that it needs to be owned by chrony:chrony? 2015-08-10 09:40:30 and we chown root /etc/chrony/* except drift? 2015-08-10 09:40:46 yes 2015-08-10 09:40:48 for backwards compat, for thsoe who upgrade? 2015-08-10 09:40:57 you're right 2015-08-10 09:41:09 chown $pkgusers:$pkggroups "$pkgdir"/etc/chrony 2015-08-10 09:41:14 chown root:root "$pkgdir"/etc/chrony/* 2015-08-10 09:41:24 except .drift if its there 2015-08-10 09:41:39 and the post-script chown chrony:chrony the drift file 2015-08-10 09:41:45 *post-install 2015-08-10 09:41:51 ah yes 2015-08-10 09:42:05 because there is no /etc/chrony/chrony.drift 2015-08-10 09:42:09 shipped with pkg 2015-08-10 09:42:11 right 2015-08-10 09:42:23 makes sense 2015-08-10 09:42:28 so the post-install make the things consistent 2015-08-10 09:42:37 makes upgrade smooth 2015-08-10 09:43:02 i am slightly in doubt the post-install/upgrade is needed 2015-08-10 09:43:22 it is needed only for backward compatibility 2015-08-10 09:43:28 since /etc/chrony is owned by chrony, and chronyd will create new and rename 2015-08-10 09:43:50 i think it should work without chown chrony:chrony /etc/chrony/chrony.drift 2015-08-10 09:44:05 ah..yes...might be 2015-08-10 09:44:11 i can test 2015-08-10 09:49:48 there is really no couchdb package on alpine? oO 2015-08-10 09:51:21 http://git.alpinelinux.org/cgit/aports/plain/unmaintained/apache-couchdb only unmaintained :((( 2015-08-10 09:51:22 fcolista: it will work without chown in post-install 2015-08-10 09:51:34 that's gr8 2015-08-10 09:51:40 so i really have to do it on debian or something else -.- 2015-08-10 09:52:09 fcolista: the new file will have new ownership and there will be a warning logged 2015-08-10 09:52:25 https://github.com/mlichvar/chrony/blob/master/reference.c#L405 2015-08-10 09:52:34 LOG(LOGS_WARN, LOGF_Reference, "Could not replace old driftfile %s with new one %s.tmp", 2015-08-10 09:52:43 no 2015-08-10 09:52:45 sorry 2015-08-10 09:52:55 https://github.com/mlichvar/chrony/blob/master/reference.c#L395 2015-08-10 09:53:02 "Could not change ownership or permissions of temporary driftfile %s.tmp", 2015-08-10 09:53:07 that is what will happen 2015-08-10 09:53:20 it change permission with 777 ? 2015-08-10 09:53:45 *does it 2015-08-10 09:53:47 no, it masks it with 0777 2015-08-10 09:53:56 uses old perms & 0777 2015-08-10 09:54:00 oh 2015-08-10 09:54:02 k 2015-08-10 09:54:06 basically strip out suid bit 2015-08-10 09:54:13 yeah 2015-08-10 09:54:25 i think it will just work 2015-08-10 09:54:29 very nice 2015-08-10 09:54:48 thsi is what i'm gonna commit then: http://sprunge.us/AKHP 2015-08-10 09:55:07 that's sad.. no up2date mariadb, no couchdb... no ceph on alpine :( 2015-08-10 09:55:45 we have mariadb 2015-08-10 09:55:51 fcolista: looks good 2015-08-10 09:56:14 couchdb was really a pain to build iirc 2015-08-10 09:56:21 yeah 2015-08-10 09:56:30 which was probably we we moved it to unmaintained 2015-08-10 09:56:32 re mariadb 2015-08-10 09:56:36 a bloaty monster 2015-08-10 09:56:41 i think we can upgrade mariadb now? 2015-08-10 09:56:54 we only needed the old to migrate from mysql 2015-08-10 09:57:26 i just know that the current mariadb version is pretty outdated :) 2015-08-10 09:57:32 ACTION is happy for chrony package 2015-08-10 09:58:08 hum 2015-08-10 09:58:14 i have ceph in my wip dir 2015-08-10 09:58:21 apparently i started on it once 2015-08-10 09:58:35 cpeh doesn't build with musl iirc 2015-08-10 09:58:55 i think i started to make patches 2015-08-10 09:59:01 ah, the stateless resolver thingy 2015-08-10 09:59:18 that was one of the issues 2015-08-10 10:00:08 barthalion: i think we can upgrade mariadb now, to latest? 2015-08-10 10:01:18 common/buffer.cc:354:68: error: cannot convert 'off_t* {aka long int*}' to 'int*' for argument '2' to 'ssize_t safe_splice(int, int*, int, int*, size_t, unsigned int)' 2015-08-10 10:01:18 ssize_t r = safe_splice(fd, off, pipefds[1], NULL, len, flags); 2015-08-10 10:01:18 ^ 2015-08-10 10:01:24 thats the current ceph error 2015-08-10 10:01:50 ./common/safe_io.h:48:8: error: 'loff_t' has not been declared 2015-08-10 10:01:51 loff_t *off_out, size_t len, unsigned int flags) 2015-08-10 10:01:51 ^ 2015-08-10 10:01:57 loff_t 2015-08-10 10:02:16 i believe ceph is fixable but it will still require some work 2015-08-10 10:07:49 ncopa: sure 2015-08-10 10:08:10 mosez: just because we don't follow latest branch, it doesn't mean we are not up to date… 2015-08-10 10:08:54 i'll update mariadb after lunch 2015-08-10 10:09:04 probably after i'm don with the 3.18.20 kernel 2015-08-10 10:13:59 someone use btrfs here with alpinelinux? 2015-08-10 10:18:17 yeah, me 2015-08-10 10:18:31 I am a bit suprised that it didn't break so far 2015-08-10 10:26:01 mmeum did you test if snapshot restore works? 2015-08-10 10:27:41 no, I haven't used snapshots so far 2015-08-10 10:28:03 but subvolumes work so I assume that snapshots work as well 2015-08-10 10:49:54 hey :) 2015-08-10 11:17:36 barthalion: i never said that :) 2015-08-10 11:18:48 does anybody have ideas how to handle python ffi? #4512 2015-08-10 11:19:48 problem is that they do find_library(name) 2015-08-10 11:19:56 where name is the link time namem 2015-08-10 11:19:58 name* 2015-08-10 11:20:31 i think the only technical correct way is to have the build time deps available at runtime 2015-08-10 11:20:51 which basically means that you need the -dev package 2015-08-10 12:09:35 is it hard to get couchdb maintained again? 2015-08-10 12:10:05 i don't want to build this docker container with debian as they are still offering only 1.2 or 1.4 (and 1.4 only on sid) 2015-08-10 12:10:22 the only other option is to compile it on my own but i would like to avoid that :( 2015-08-10 12:11:26 i suppose the way to go is to try build couchdb with musl, then report each failure upstream to couch db 2015-08-10 12:12:00 nmeum subvolumes are new to me, they are used for mountpoints? so you have / and /home /var on diff subvolumes 2015-08-10 12:13:41 ncopa: was using fossil in project, would it be ok to enable json features, http://tpaste.us/GELV ? 2015-08-10 12:17:25 vkris: applied. thanks 2015-08-10 12:17:48 there is also support for markdown, but not using it 2015-08-10 12:17:58 thanks 2015-08-10 12:18:06 does someone have docker file for irssi hier? 2015-08-10 12:18:47 i found this https://github.com/docker-library/docs/tree/master/irssi but i dont understand anything from this docker things 2015-08-10 12:36:03 fcolista: chrony fails to start here 2015-08-10 12:36:08 getpwnam(chrony) failed 2015-08-10 12:36:23 new install ncopa? 2015-08-10 12:36:29 no, upgrade 2015-08-10 12:36:33 on my desktop 2015-08-10 12:36:41 in my vm starts 2015-08-10 12:36:51 new install? 2015-08-10 12:36:52 no 2015-08-10 12:36:54 upgrade 2015-08-10 12:37:35 ok 2015-08-10 12:37:37 i know why 2015-08-10 12:37:43 i dont have the system user chrony 2015-08-10 12:37:57 which is only created on apk add 2015-08-10 12:37:58 should be created 2015-08-10 12:38:00 not apk upgrade 2015-08-10 12:38:05 aaah 2015-08-10 12:38:21 we need an upgrade script then 2015-08-10 12:38:39 i just symlink it 2015-08-10 12:38:52 +1 2015-08-10 12:58:41 ncopa: what's your opinion 2015-08-10 13:01:18 nmeum: aoub twhat? 2015-08-10 13:01:21 about* 2015-08-10 13:01:22 ncopa: what's your opinion on the patches adding shells to /etc/shells like 379 and 380 2015-08-10 13:01:26 sorry hit return too fast 2015-08-10 13:01:30 :) 2015-08-10 13:01:51 in my opinion i should fix the time on my desktop computer 2015-08-10 13:02:18 now time is correct 2015-08-10 13:02:19 cool 2015-08-10 13:02:40 screen saver thought i hadnt touched the keyboard in 2 hours 2015-08-10 13:02:56 %379 2015-08-10 13:04:49 nmeum: i wonder if it would be an idea to have the shells in /bin 2015-08-10 13:04:58 but i suppose that does not matter 2015-08-10 13:05:15 i mean we have /bin/bash and not /usr/bin/bash 2015-08-10 13:05:18 I wondered as well 2015-08-10 13:05:47 I wasn't sure were to store shell binaries, personally I would also store them in /bin 2015-08-10 13:06:09 but many packages already add store shells in /usr/bin so I didn't change that 2015-08-10 13:06:15 s/add/// 2015-08-10 13:06:47 maybe we should clean that up while at it? 2015-08-10 13:06:55 yeah, sure 2015-08-10 13:07:13 do you want to do that or should I update the patches? 2015-08-10 13:07:44 please update the patches 2015-08-10 13:07:53 i wonder if it is possible to implement this as a trigger 2015-08-10 13:07:57 but i suppose it is not 2015-08-10 13:08:02 btw: the only shell I didn't add to /etc/shells is oksh because the oksh package doesn't build currently (seems to be a fortify headers issuee) 2015-08-10 13:08:19 ok 2015-08-10 13:08:45 I don't think that it is possible to implement this as a trigger 2015-08-10 13:08:46 notoksh 2015-08-10 13:08:52 crow: hm? 2015-08-10 13:08:52 :) 2015-08-10 13:08:58 ah, that's a joke 2015-08-10 13:09:33 i just got back from surgery, so im allowed to make stupid jokes 2015-08-10 13:10:14 clandmeter, what happened? 2015-08-10 13:10:28 nothing serious i hope 2015-08-10 13:10:37 some small issues in the bottom :) 2015-08-10 13:10:54 ok :) 2015-08-10 13:11:31 oh good. i was afraid you got your humor parts of the brain transplanted.... :) 2015-08-10 13:11:53 ncopa: I will update the patches in a few hours, you could already accept %385 (if you want to) because it already stores the shell in /bin 2015-08-10 13:11:59 lol, well some ppl could defenitly need it :) 2015-08-10 13:18:53 ncopa: should the fish-shell also be installed to /bin? I think it makes more sense here to create a symlink from /bin/fish to /usr/bin/fish… 2015-08-10 13:22:46 hm, maybe it doesn't… 2015-08-10 13:23:04 i dont know really 2015-08-10 13:23:44 I doesn't really matter in the end but I will just change the bindir to /bin 2015-08-10 13:32:17 I updated the fish patch but you applied the old one, didn't you? 2015-08-10 13:32:29 %390 2015-08-10 13:37:50 ncopa: I added a patch to install mksh to /bin, other distribution also install it to /bin so I think it is a good idea to do so %391 2015-08-10 13:46:35 nmeum ? 2015-08-10 13:48:16 nmeum i need to read about subvolumes 2015-08-10 13:55:02 snapshot behave similar to subvolumes in some regards 2015-08-10 13:56:18 BTW: commit f9e3822c20601e77293aed701c7ccbca09beb413 should be 3.2, 3.1 etc see https://secure.php.net/ChangeLog-5.php#5.6.12 2015-08-10 13:57:21 *backported 2015-08-10 14:06:44 hacking a bit python 2015-08-10 14:06:50 python is pretty nice 2015-08-10 14:08:11 implementing an is_elf(filepath) func is just 3 lines 2015-08-10 14:10:45 ncopa: do you prefer python over lua? 2015-08-10 14:11:01 i hate python, slow as hell and a crappy standard library 2015-08-10 14:12:11 no, i prefer lua due to its smaller and faster 2015-08-10 14:12:19 even if that means i have to write a few more lines code 2015-08-10 14:12:33 i am trying to fix pythons find_library implementation 2015-08-10 14:12:50 #4512 2015-08-10 14:14:04 just saying that python language is pretty nice 2015-08-10 14:14:11 i understand why its popular 2015-08-10 14:14:37 my findlib implementation: http://tpaste.us/G0N8 2015-08-10 14:51:44 ncopa: nice 2015-08-10 14:51:59 but i still love to write small parts in c and glue them together with lua 2015-08-10 14:57:38 findlib is broken 2015-08-10 14:57:40 my findlib 2015-08-10 14:57:44 have better fix 2015-08-10 14:57:57 which uses libc' logic to find libs 2015-08-10 17:02:20 is there no ddclient package? 2015-08-10 17:02:48 http://sourceforge.net/projects/ddclient/files/ddclient/ddclient-3.8.3/ says "* added Alpine Linux init scritp - patch send by @Tal on github." 2015-08-10 17:02:53 but I cannot find the package 2015-08-10 17:30:48 fcolista: please ignore and close %392 accidentally send you the wrong format-patch 2015-08-10 18:43:52 fcolista: did you have any luck with the go snapshots? 2015-08-10 20:23:20 BTW: What's the status of this entire udev replacement disscussion? Has a decision been made yet? 2015-08-10 20:29:31 mdev? 2015-08-10 20:58:09 shafire: well…using mdev with xorg is a bit of a pain 2015-08-10 20:58:46 there has been a discussin on the devel mailinglist regarding better alternatives 2015-08-10 21:29:26 hi 2015-08-10 21:29:46 if I want to send in a corrected patch in patchwork 2015-08-10 21:29:49 how do I do that? 2015-08-10 22:00:26 hm, does setup-apkrepos trying to find fastest mirror blocks on, uh, finding the fastest mirror 2015-08-10 22:01:32 if one of the mirrors is taking longer than the others, it's probably reasonable to conclude that it's not the fastest? 2015-08-10 22:04:11 hm ... I'll think about it 2015-08-11 06:08:39 ammunta: interesting idea 2015-08-11 06:09:30 in theory, you could set timeout to the number of current fastest mirror 2015-08-11 06:15:44 I just put default timeout of 10s on it so it didn't take forever because my connection to one of the mirrors was doing an odd thing 2015-08-11 06:16:46 but yeah 2015-08-11 06:19:18 personally I'd let it run a bit longer and show the figure because it's useful to know e.g. this mirror is 0.16 and that mirror is 0.18 but you know that the times are dominated by connection latency 2015-08-11 06:27:46 lol 2015-08-11 06:27:56 Changed in version 2.7.3: Since lots of code check for sys.platform == 'linux2', and there is no essential change between Linux 2.x and 3.x, sys.platform is always set to 'linux2', even on Linux 3.x. In Python 3.3 and later, the value will always be set to 'linux', so it is recommended to always use the startswith idiom presented above. 2015-08-11 06:29:27 so python's sys.platform is a lie because many people use it wrong 2015-08-11 06:29:40 they should have just dropped the version number 2015-08-11 06:45:01 pffh 2015-08-11 06:45:18 well they *did* just drop the version number 2015-08-11 06:45:30 eventually 2015-08-11 08:15:55 good morning 2015-08-11 08:22:22 hi 2015-08-11 08:22:31 how do I send in a corrected patch in patchwork? 2015-08-11 08:30:35 http://wiki.alpinelinux.org/wiki/Creating_patches#Resend_an_updated_patch 2015-08-11 08:31:43 I see 2015-08-11 08:33:46 next time :) 2015-08-11 11:02:45 could someone please close %390 ? 2015-08-11 11:21:32 nmeum: done. thanks 2015-08-11 11:22:32 %386 might also be rejected since php was already updated (take a look at my comment) 2015-08-11 20:11:15 which arm boards do you use for the builders again? 2015-08-12 12:05:57 ncopa, any strong opinion on 388? 2015-08-12 12:11:12 %388 2015-08-12 12:12:09 it is needed so krb5 and heimdal package can share the config 2015-08-12 12:12:39 agreed. Just need your feedback for what is in main. 2015-08-12 12:33:56 hum, are we really going to use cron for mkwhatis? 2015-08-12 12:34:16 fcolista: I don't think we have finished discussing this 2015-08-12 12:34:19 ncopa: ↑ 2015-08-12 12:34:32 did you read my comment? 2015-08-12 12:36:06 ncopa haven't replied you yet and he's one of ~4 people around who decides in such cases 2015-08-12 12:37:00 fcolista: so please wait a little more before you push something that's under discussion 2015-08-12 12:38:11 sure…I was a bit suprised myself that it was merged already since ncopa was a bit skeptical 2015-08-12 12:45:11 reverted it for now 2015-08-12 12:45:38 i am against using cron for makewhatis 2015-08-12 12:45:55 if it's not something addressable in mdocml, I'd also prefer to have a subpackage for trigger and maybe another for cron 2015-08-12 12:46:12 but cron just feels wrong, this is perfect case for using triggers 2015-08-12 12:47:25 nmeum: it was you who sent the patch to ml, right? 2015-08-12 12:47:40 i think it would be possible to make it a subpkg 2015-08-12 13:06:37 nmeum: you can read man pages without the mandoc.db 2015-08-12 13:06:45 but i tdont think you can use whatis or apropos 2015-08-12 13:08:20 yeah, I send the patch 2015-08-12 13:08:39 and you are right, with makewhatis you can read manpages but you cannot use whatis or apropos 2015-08-12 13:08:46 s/with/&out/ 2015-08-12 13:09:10 but the question is how would you like to invoke makewhatis in a subpackage? 2015-08-12 13:09:18 as a trigger or as a cron? 2015-08-12 13:09:21 trigger 2015-08-12 13:09:44 we could fork it to backgrouund 2015-08-12 13:09:54 if it becomes a problem 2015-08-12 13:10:17 why not a cron? 2015-08-12 13:10:31 because apk add apropos man-pages 2015-08-12 13:10:35 apropos read 2015-08-12 13:10:38 does not work 2015-08-12 13:10:51 you'll have to wait til tomorrow to actually use it 2015-08-12 13:10:56 sure on newly added packages it doesn't work automatically 2015-08-12 13:11:07 but how often do you search for a man page you just added? 2015-08-12 13:11:19 always :) 2015-08-12 13:11:27 thats the only reason i install the man page 2015-08-12 13:11:36 and pretty much every other major distribution uses a cron 2015-08-12 13:12:06 pretty much every other major distro does not ship manpages in separate package 2015-08-12 13:12:45 I don't see why this has an effect on this? 2015-08-12 13:13:15 also: it is always possible to invoke makewhatis manually 2015-08-12 13:13:20 point is, users normally dont install -doc. if they do, then do it becuase they want a specific pake 2015-08-12 13:13:22 it is 2015-08-12 13:13:27 we could also ship two subpackages one which uses a cron and one which uses a trigger 2015-08-12 13:14:04 thats also an option 2015-08-12 13:14:46 also, i dont expect all alpine users to use cron at all, in eg. docker containers 2015-08-12 13:15:03 why would you need man pages in a docker container? 2015-08-12 13:15:43 also: the mandoc.db isn't that important, not having it only prevents you from using apropos and whatis 2015-08-12 13:16:13 and you can still generate that database manaully 2015-08-12 13:16:18 but just invoke makewhatis 2015-08-12 13:16:20 *b 2015-08-12 13:16:22 *by 2015-08-12 13:17:05 i dont like the idea of you spend X hours on setting up a box 2015-08-12 13:17:11 while doing so you mithg want use manpages 2015-08-12 13:17:19 when you are done, the box just runs 2015-08-12 13:17:21 for years 2015-08-12 13:17:28 nothing added/nothing removed 2015-08-12 13:17:43 why should it check if something happened *every* night for years? 2015-08-12 13:17:50 i just dont like that 2015-08-12 13:18:13 that scenario is unrealistic since you need to install package updates for security reasons and installing package updates might result in the addition or removal of man pages 2015-08-12 13:18:36 I could also increase the interval to only invoke it every week or so 2015-08-12 13:19:13 i think, if you dont need the db within a week, then you dont need it at all 2015-08-12 13:19:48 well…you can uset the database in the meantime it simply isn't uptodate 2015-08-12 13:21:06 and btw: The amount of time the trigger needs grows linear with the amount of man pages you have on your system, so at some point the installation of -doc package might take 2 minutes or so 2015-08-12 13:21:31 however, I can also understand your point and if you prefer to ship a trigger by default than I will simply compile that package myself and use a cron instead :) 2015-08-12 13:22:14 i can agree that 38sec+ for every apk add/del is unacceptable 2015-08-12 13:22:20 so i am interested in solving it 2015-08-12 13:23:38 and tbh, i dont want force people work in doing things one specific way 2015-08-12 13:23:56 so i am open for ideas how we can let user chose how he wants it 2015-08-12 13:24:36 as I said: creating two subpackages (one with a trigger, one with a cron) would grant the user more choices in this regard 2015-08-12 13:24:58 how about this: we split out apropos/whatis/makewhatis into a subpackage with a trigger 2015-08-12 13:25:14 the trigger checks if /etc/periodic/daily/makewhatis exists 2015-08-12 13:25:31 if it does, then it does not run makewhatis 2015-08-12 13:25:38 otherwise it runs it 2015-08-12 13:26:20 yeah, that's sound ok, but it only makes sense if you ship a second package which installs /etc/periodic/daily/makewhatis 2015-08-12 13:26:46 which is a oneliner :) 2015-08-12 13:26:52 well, 2lines 2015-08-12 13:27:14 yeah 2015-08-12 13:27:40 I would infact create 3 subpackages: One which ships apropos and whatis, one which ships makewhatis with a trigger and one which ships makewhatis with a cron 2015-08-12 13:28:08 and then you could simply let the last two packages conflict with each other so you can't install both 2015-08-12 13:29:15 also: If you come up with a solution to speed up the trigger let me know 2015-08-12 13:31:42 this might work: http://tpaste.us/AoQg 2015-08-12 13:32:29 this has one issue: what do you do if that trigger still runs in the background but a new -doc package is installed an the trigger is invoked again? 2015-08-12 13:32:36 yes 2015-08-12 13:32:43 its sql database 2015-08-12 13:33:00 i figured sqlite handles locking 2015-08-12 13:33:56 but then the second trigger fails and the newly added man page might not be added 2015-08-12 13:34:08 or how is that handling implemented? does it simply wait until the lock is removed again? 2015-08-12 13:34:22 *lock handling 2015-08-12 13:35:02 it handles it the same way other sql databases handles concurrent reads/writes i suppose 2015-08-12 13:36:12 I get an I/O error when I try to run the process twice 2015-08-12 13:36:18 ok 2015-08-12 13:36:22 > /usr/share/man//mandoc.db: disk I/O error 2015-08-12 13:37:04 that can probably be fixed in mdocml code 2015-08-12 13:37:24 might be it locks the entire db 2015-08-12 13:37:25 do you really want to patch the mdocml code for this? 2015-08-12 13:38:03 if it is simple, why not? 2015-08-12 13:39:31 but currently, i am mostly interested in fixing the musl timezone offset bug 2015-08-12 13:40:11 yeah, go ahead 2015-08-12 13:40:32 re mdocml, i think first step is split out apropos and makewhatis + trigger 2015-08-12 13:40:40 should solve 80% of the problem 2015-08-12 13:41:18 after that we can think of adding a cron alternative for makewhatis 2015-08-12 13:43:03 barthalion, ncopa, sorry. I miss that this was still under discussion. I'll pay more attention next time. 2015-08-12 13:43:14 ncopa: I will submit a patch for this split 2015-08-12 13:45:21 fcolista: np. thanks alot for taking care of the patches 2015-08-12 14:08:15 nmeum: did you send the patch for git src change? 2015-08-12 14:08:30 clandmeter: yes, is there a problem with it 2015-08-12 14:08:31 ? 2015-08-12 14:08:41 i have the same question 2015-08-12 14:08:43 :) 2015-08-12 14:08:47 what was the problem? 2015-08-12 14:08:53 is the src different? 2015-08-12 14:09:52 it previously used the github repo to fetch the source but the github repo is only a mirror 2015-08-12 14:10:41 im not against changing the src, but if there is no diff, you odnt need to bump pkgrel 2015-08-12 14:11:02 now everybody has to update git, because of cosmetics 2015-08-12 14:11:24 also, the rename :: is still in the uri, whihc is now useless. 2015-08-12 14:15:46 well you have push access to the repository, just remove the rename :: if it annoys you 2015-08-12 15:01:48 just back from hols so apologies if this has already been discussed: 2015-08-12 15:02:12 just apk upgraded and i get Failed to get bus connection: Failed to execute child process "dbus-launch" (No such file or directory) 2015-08-12 15:02:25 when running a number of apps 2015-08-12 15:07:33 it was fixed by adding the dbus-x11 package 2015-08-12 15:08:07 loooool, Kdbus was rejected by the kernel developers because it sucked to hard 2015-08-12 15:08:21 yay, again? 2015-08-12 15:08:26 how bad must your software be that it is worse than dbus?? 2015-08-12 15:08:29 do you have a link to lkml? 2015-08-12 15:08:43 ScrumpyJack: sounds like a bug that it isn't pulled in by something 2015-08-12 15:16:32 this looks related http://git.alpinelinux.org/cgit/aports/commit/?id=14ac979cf1fa72d638e104e5daee9b44ed5269c5 2015-08-12 15:17:33 leo-unglaub: do you have a source on this? 2015-08-12 15:18:27 can't find anything recent on lkml 2015-08-12 15:18:54 nmeum: only on a german news site 2015-08-12 15:19:18 leo-unglaub: could you post the link? 2015-08-12 15:21:33 http://www.golem.de/news/linux-kernel-integration-von-kdbus-wird-wieder-verschoben-1506-114928.html 2015-08-12 15:24:01 i kind of agree with Andy Lutomirski 2015-08-12 15:24:13 ah, there are links in the article 2015-08-12 15:24:30 just putting something in the kernel because the userland version is horribly written is no solution 2015-08-12 15:25:21 basically the article seems to be based on this thread https://lkml.org/lkml/2015/6/23/22 2015-08-12 15:45:40 that is from june 2015 :p 2015-08-12 15:46:15 and there is reasons to put a low latency message bus into the kernel 2015-08-12 15:46:27 modeling it after d-bus, however, probably not so great 2015-08-12 15:46:47 the main benefit is that the scheduler can be aware of the message bus contents 2015-08-12 15:47:02 so for example if a process has pending messages, it can be scheduled early 2015-08-12 15:47:14 which improves interactivity 2015-08-12 15:47:51 this could of course be done with unix sockets instead, but that would be more complex to handle and in many cases would break edge-triggering 2015-08-12 18:39:41 meh, june, I wanted some fresh drama 2015-08-12 18:51:25 hm, the load_modules() function in /etc/init.d/alsa is totally broken 2015-08-12 18:51:47 there is no -l flag for modprobe 2015-08-12 18:51:55 I don't see how this ever worked at all 2015-08-12 18:59:02 the function was written by ncopa in 2009 and never touched since, is it even needed anymore? 2015-08-12 19:49:17 fcolista: did you have any more luck with the snapshot upload? 2015-08-12 20:07:03 god, I fucking hate computers 2015-08-12 20:08:33 with good reason 2015-08-12 20:33:33 nmeum: so much this 2015-08-12 20:34:18 the more time you spend on them, the more often you think you should have become a doctor of medicine :P 2015-08-12 21:06:24 but then you realize that people can have really disgusting diseases and you are glad that you just have to stare into a screen all day ;) 2015-08-12 21:39:09 hoi 2015-08-13 07:11:23 what was the solution when application wants to write to /dev/null 2015-08-13 07:11:29 and WARNING: Device "/dev/null" is not available for reading and writing. 2015-08-13 07:12:31 what even happened 2015-08-13 07:13:30 testing/fossil gives that error 2015-08-13 07:16:44 it also warns "WARNING: Device "/dev/urandom" is not available for reading. This means that the pseudo-random number generator used by SQLite will be poorly seeded." 2015-08-13 07:17:19 If its fossil related than I can add it to bugs.a.o 2015-08-13 07:19:02 ncopa: could you also push testing/fossil to main/, did basic tests+json test, seems to work 2015-08-13 08:03:48 good morning 2015-08-13 08:10:10 hello 2015-08-13 09:16:34 morning 2015-08-13 09:17:49 hey ncopa := 2015-08-13 09:17:51 :) 2015-08-13 09:18:10 vkris: moved fossil 2015-08-13 09:29:47 ncopa, i'm going tp apply 412 2015-08-13 09:29:49 %412 2015-08-13 09:30:44 fcolista: ack 2015-08-13 09:30:57 thanks 2015-08-13 09:31:09 np 2015-08-13 10:05:54 ncopa: do you want to wait with openssh7? 2015-08-13 10:06:05 or do you plan to jump to the next version very soon? 2015-08-13 10:08:16 is there any reason to wait with openssh7? 2015-08-13 10:08:25 i was planning to do it soonish 2015-08-13 10:08:31 might need rebase the HPN patches though 2015-08-13 10:09:15 2015-08-13 10:09:39 i dont know of any reason 2015-08-13 10:09:47 maybe kaniini has some more insight there 2015-08-13 10:39:22 running taskd gives me http://sprunge.us/IHFZ 2015-08-13 10:40:20 what sort of error is that? 2015-08-13 10:44:51 is it caused by missing libs? 2015-08-13 10:47:56 I'm on 3.2, but i grabbed it from edge/testing 2015-08-13 10:49:54 looks like C++ symbol mangling 2015-08-13 11:12:29 works fine on a box that is edge only 2015-08-13 11:12:52 can we move taskd into main? 2015-08-13 11:26:39 ncopa, 381 and 382 does not build 'cause of missing options="suid". 2015-08-13 11:26:47 In this case, should i add it and push? 2015-08-13 11:27:08 Or try to make it work without suid ? 2015-08-13 11:27:14 (if possible9 2015-08-13 11:29:44 %381 2015-08-13 11:30:34 %382 2015-08-13 11:31:01 fcolista: i suppose you can set options="suid" for now 2015-08-13 11:34:29 k 2015-08-13 11:39:35 %365 2015-08-13 11:39:46 do you see problem with that? 2015-08-13 11:40:10 i think is safe to apply 2015-08-13 11:40:14 since git-doc does that 2015-08-13 12:35:51 fcolista: make sense 2015-08-13 12:36:05 ack on %365 2015-08-13 12:36:16 pushed :) 2015-08-13 12:41:02 fcolista: thanks 2015-08-13 12:41:11 i am pretty happy 2015-08-13 12:41:26 we have fixed various nasty bugs 2015-08-13 12:51:02 could someone take a look at %420 and %421? 2015-08-13 14:01:31 FYI: I'll be on vacation for 1.5 weeks now. I might drop by here at times, and will read mails occasionally. 2015-08-13 21:58:30 https://github.com/mitchty/alpine-linux-ghc-bootstrap huh, neat 2015-08-14 06:31:49 fcolista: Why did you change the checksum of testing/slock? Those don't seem to match here 2015-08-14 06:43:26 nmeum, look at commit 19ee0e50d4173016c303e85483f2b09f675e380c 2015-08-14 06:43:43 the checksum was wrong 2015-08-14 06:43:57 and is calculated each time when you abuild 2015-08-14 06:44:14 if doesn't match, package does not build 2015-08-14 06:44:28 do abuild checksum and then abuild -r 2015-08-14 06:57:16 I know that but the checksum you specified doesn't seem to match either 2015-08-14 06:58:34 nmeum, can you please describe the steps you are doing? 2015-08-14 07:00:09 fcolista: the file listed in source has a new checksum each time you download it, it seems to be generate on the fly by cgit, thus the checksum is different each time you download the file 2015-08-14 07:00:17 I changed the url in %432 2015-08-14 07:01:31 ah. This does the same that github does. Adds the timestamp, and this causes the file to be different each time, and checksum fails. 2015-08-14 07:01:45 ok 2015-08-14 07:02:22 nmeum, do you send the patches with gmail? 2015-08-14 07:03:04 no, why? 2015-08-14 07:05:23 fcolista: why would you assume that I send patches with gmail? 2015-08-14 07:05:39 I'm not assuming, i'm asking. 2015-08-14 07:05:47 Got whitespaces error 2015-08-14 07:05:59 slock miss checksum 2015-08-14 07:06:09 remeber to to abuild checksum before sending the patch 2015-08-14 07:06:24 I did 2015-08-14 07:06:40 65eee712e2f07d9d5b6b241a703a9340d2456579ed253a73bf354ba37d2e900a safer-oom-killer-disablement.patch" 2015-08-14 07:06:40 -sha512sums="0e66c5451c72a5df391a61eefa6b1a1493705fc9c0b04d9db934a56c11852c284ecca8706f47ceb1d1c8cfc5241c818c6026afb30fc9cf3938e44ea282338bee slock-1.2.tar.gz 2015-08-14 07:06:40 +sha512sums="8b403da254cf9802de760fcd39a3c848c4d6b0022f0200efd6210ed14b6d65a2978a4a9c89c661f4a386a17e5d871c531278e4240a3af0a2e42c0ffc95d0ad10 slock-1.2.tar.gz 2015-08-14 07:06:56 ? 2015-08-14 07:07:02 - is the old one 2015-08-14 07:07:04 I changed the source url thuse the checksum is different 2015-08-14 07:07:06 + is the new one 2015-08-14 07:07:10 yes 2015-08-14 07:07:15 after you changed source url 2015-08-14 07:07:20 did you abuild checksum ? 2015-08-14 07:07:24 Seems not 2015-08-14 07:07:28 yeah, I did 2015-08-14 07:07:39 I built the entire package after that 2015-08-14 07:09:02 umh 2015-08-14 07:09:05 the checksum is valid here 2015-08-14 07:09:38 ok, got it 2015-08-14 07:09:42 great :) 2015-08-14 07:10:12 the old file was in cache 2015-08-14 07:10:20 and caused to not be re-downloaded 2015-08-14 07:10:28 ah, that explains it 2015-08-14 07:30:00 why does protobuf have a py subpackage which isn't enabled? http://git.alpinelinux.org/cgit/aports/tree/main/protobuf/APKBUILD#n47 2015-08-14 08:15:10 good morning my friends!! 2015-08-14 08:18:54 firefox 40 is here :) 2015-08-14 08:50:14 nmeum: have you seen the install_if feature? 2015-08-14 08:50:58 nmeum: see how kmod has an install_if for the bash completion 2015-08-14 08:51:33 if you "apk add kmod bash" then will it also pull in the bash completion for kmod 2015-08-14 09:17:09 ncopa: oh, didn't know about that feature, that is indeed pretty usefull for the vim subpackages 2015-08-14 09:17:45 make sure that one of the deps are versioned 2015-08-14 09:18:00 eg. install_if="$pkgname=$pkgver-r$pkgrel ..." 2015-08-14 09:18:30 or it will cause problems 2015-08-14 09:32:45 hm, maybe it would be a good idea to add a default function for vim subpackages and {bash,zsh,fish}-completions subpackages to abuild (similar to the default doc subpackage handler) 2015-08-14 09:32:55 because the procedure to create these subpackages is always the same 2015-08-14 09:55:38 could be 2015-08-14 09:55:57 but there are not as many completions subpackages as -doc or -dev 2015-08-14 09:56:09 i'll think of it 2015-08-14 09:56:43 i wonder if we could implement abuild extensions 2015-08-14 09:56:57 so you could in APKBUILD do: extensions="completions" 2015-08-14 10:01:25 sounds like a good idea but you shouldn't make those extension too complex or have too many of them otherwise you end of like gentoos ebuild format which also has >9000 extension and many of them are poorly documented 2015-08-14 10:01:37 s/end/end up/ 2015-08-14 10:02:25 lunch time :) see you later 2015-08-14 10:02:41 nmeum: that is exactly what i want avoid... 2015-08-14 10:02:45 lunch here too 2015-08-14 10:05:26 ncopa: with general purpose extension this is a pretty high risk maybe it's possible to reduce that risk by explicitly limiting their functionality 2015-08-14 10:05:57 but in the end adding more functionality to abuilds will definitly make them more complex 2015-08-14 12:36:20 [257540.826512] Out of memory: Kill process 13815 (m4) score 519 or sacrifice child 2015-08-14 12:36:26 thank you autoconf 2015-08-14 12:36:34 for DOSing my desktop 2015-08-14 12:37:01 made my kernel kill my running vms 2015-08-14 12:41:30 ops 2015-08-14 13:50:27 uff...this go makes me crazy.. 2015-08-14 13:50:35 %425 2015-08-14 13:53:01 look, no go get 2015-08-14 13:53:15 eh, godep restore instead 2015-08-14 13:53:38 yes 2015-08-14 13:53:42 and a lot of them 2015-08-14 13:53:48 Godeps.json in aports? are we going to maintain it? 2015-08-14 13:54:41 I don't think so 2015-08-14 14:21:21 how do other distributions package go software? There must be a someone decent way to do it 2015-08-14 14:21:42 archlinux does it wrong, they just create a GOROOT and run go get which sucks balls 2015-08-14 14:22:48 ncopa: I noticed there are sooo many packages not shipping manpages in a subpackage what's your opinion making abuild output a warning if you install files to /usr/share/man in a package that doesn't end with '-doc' 2015-08-14 14:23:52 %441 2015-08-14 14:24:31 nmeum: sounds like a good idea to me 2015-08-14 14:24:34 with warning 2015-08-14 14:25:03 yeah, warning not error otherwise it would break packages like man-pages 2015-08-14 14:25:40 ncopa: do you want to do it or would you prefer it if I send you a patch? 2015-08-14 14:26:30 feel free to send patch 2015-08-14 14:28:33 where should I send the patch when I finished working on it? 2015-08-14 14:28:40 i am currently watching a talk about reproducable builds 2015-08-14 14:28:50 very interresting, but very hard to get right 2015-08-14 14:28:53 nmeum: alpine-devel@lists.alpinelinux.org 2015-08-14 14:29:11 alright 2015-08-14 14:29:12 leo-unglaub: reproducable buidls sounds difficult indeed 2015-08-14 14:29:41 https://www.youtube.com/watch?v=bb3sCmr2xOk 2015-08-14 14:29:45 from the CCCCamp 2015-08-14 14:29:54 i see the benefit 2015-08-14 14:30:08 but its so hard to get going because so much makefiles contain crap 2015-08-14 14:33:02 makefiles don't really *have* to contain that much crap 2015-08-14 14:33:36 ammunta: in theory ... 2015-08-14 14:33:49 but in reallity they contain stuff like timestamps, or directory listings 2015-08-14 14:34:02 and for example on every fs directory listings can have different orders 2015-08-14 14:34:07 so the build becomes different 2015-08-14 14:34:16 ... makefiles? 2015-08-14 14:34:35 yes 2015-08-14 14:34:48 are we talking about makefiles or the crap automake spits out 2015-08-14 14:34:52 i have seen makefiles that use the current timestamp to create a tmp directory and build in there 2015-08-14 14:35:40 that sounds hideous and more like someone misdesigned their buildsystem than 2015-08-14 14:35:49 hehehe 2015-08-14 14:35:53 yes, it is stupid 2015-08-14 14:36:04 but look at some of the 22 000 packages in debuan 2015-08-14 14:36:15 ncopa: while at it: what about a warning or error (I prefer an error) when a uncompressed man page is installed to /usr/share/man? 2015-08-14 14:36:33 packages like that should just be skipped over with angry mail to upstream 2015-08-14 14:36:59 nmeum: man pages should be auto compressed 2015-08-14 14:37:05 ammunta: if you drop every package that has a sucky buildsystem you have 50 packages in aports 2015-08-14 14:37:12 those who does not have it can just be rebuilt 2015-08-14 14:37:13 no firefox, no chrome, no nothing 2015-08-14 14:37:28 when we do new release, v3.3 we'll rebuild from scratch 2015-08-14 14:37:37 so those will all have properly compressed man pages 2015-08-14 14:37:55 for edge i didnt think it was worth a complete world rebuild 2015-08-14 14:37:59 leo-unglaub: okay yes 2015-08-14 14:38:04 ncopa: if you don't use the default doc funciton than they are not autocompressed, am I wrong? 2015-08-14 14:38:15 correct 2015-08-14 14:38:22 the angry mail to upstream should still happen though 2015-08-14 14:38:33 nmeum: you are correct, not wrong 2015-08-14 14:38:55 ncopa: in that case a checking for uncompressed manpages would be a bit useful, wouldn't it? 2015-08-14 14:39:06 nmeum: you are right 2015-08-14 14:39:16 :) 2015-08-14 14:39:17 add a warning for it 2015-08-14 14:39:28 hm 2015-08-14 14:39:32 maybe error 2015-08-14 14:39:34 do error 2015-08-14 14:39:41 if it becomes a problem we slack it 2015-08-14 14:39:59 +1 for error 2015-08-14 14:40:31 the question is: what's the best way to detect uncompressed man pages? Looking for man-pages with .[0-9] prefix? 2015-08-14 14:40:39 *with a 2015-08-14 14:41:59 "we can assume that, either, everyone is owned," 2015-08-14 15:04:12 ok its weekend 2015-08-14 15:04:23 nmeum: a *big* thanks for all patches 2015-08-14 15:04:45 and a big thanks for all good ideas ;) 2015-08-14 15:04:52 have a nice weekend 2015-08-14 15:06:02 ncopa: no problem, I like to help out :) 2015-08-14 15:06:18 also: I have the patch ready now, sending it now 2015-08-14 15:50:36 hi 2015-08-14 15:51:00 how do you decide if the service is ending with d or not? 2015-08-14 15:51:18 for example cron but sshd? 2015-08-14 15:57:28 This question belongs in #alpine-linux but the easiest way to do so is to simply look in /etc/init.d/ 2015-08-14 15:57:50 or a bit more advanced: apk info -L | grep init.d 2015-08-14 16:20:31 no 2015-08-14 16:20:34 when I pack a package 2015-08-14 16:21:00 why is ssh ending with sshd but cron not, but both are running as daemon? 2015-08-14 16:21:16 ending with "d" 2015-08-14 16:28:20 shafire: I don't think that it matters a lot, just use the name users would expect 2015-08-14 16:29:38 thats not consistent :p 2015-08-14 16:30:52 well…the naming scheme is consistent currently either…sshd is called sshd because the binary is called sshd, so cron should also be called crond cause the binary is called crond but it isn't 2015-08-14 16:31:28 this needs to be changed :) 2015-08-14 16:32:56 shafire_: yeah, I would be interessted in ncopa opinion on this, but he is enjoying his weekend…so why don't you submit a patch for this? 2015-08-14 16:33:16 I will ask monday again;) 2015-08-14 16:33:28 this also applies to the syslog service 2015-08-14 16:40:09 shafire_: I created a patch for you %452 2015-08-14 16:40:44 :) 2015-08-14 17:14:52 oha, firefox just died do to grsec 2015-08-14 17:15:48 http://dpaste.com/2M8PNP7.txt 2015-08-14 18:45:32 mh 2015-08-14 18:45:41 is there any styleguide for APKBUILD? 2015-08-14 18:45:57 every APKBUILD looks very different - many people, many styles 2015-08-14 18:46:04 some uses || return 1, some not 2015-08-14 18:46:15 some all in one line, some everything in a new line 2015-08-14 19:05:28 shafire_: not sure about a style guide, but abuild does not run with errexit, so ||return 1 is important for proper error handling 2015-08-14 19:07:05 I dont remember any guide either 2015-08-14 19:11:03 I would not object to a guide that advocates the usage of either spaces or tabs in an APKBUILD for indention, but not both ;) 2015-08-14 19:12:29 I would like to see init scripts have start_pre and stop_pre instead of overriding start/stop functions 2015-08-14 19:31:51 well, I for one must confess that I do not put too much effort into init scripts because I personaly don't use them 2015-08-14 19:33:13 but if there is something like a comprehensive guide or similar on openrc best practices, I'd be willing to look it over 2015-08-14 19:51:19 what do you use? 2015-08-14 19:51:40 s6 2015-08-14 20:40:23 I did go through some init scripts 2015-08-14 20:40:34 from many, the start/stop functions can be removed :S 2015-08-14 20:46:34 feel free to supplie patches :) 2015-08-14 20:53:57 %463 2015-08-14 20:54:44 Even though packing go packages sucks go is awesome in many other regards, for instance it's easy as fuck to cross compile go code :3 2015-08-14 20:55:39 Theoretically I could even cross compile a binary for plan9 or android now 2015-08-14 21:03:45 can I remove the gentoo header when I simplified it radically? 2015-08-14 21:05:38 nmeum: how did you find alpine? 2015-08-14 21:06:05 shafire_: I wouldn't do that since gentoo licenses its OpenRC services under GPL that could get pretty messy… 2015-08-14 21:06:22 unless you really modified it in a way that no significant part of the original service is left over 2015-08-14 21:06:51 but if you want to be on the safe side just keep that header, it doesn't hurt 2015-08-14 21:07:47 shafire_: don't know…just stumbled on it 2015-08-14 21:08:07 I am using alpine since 2.6 :) 2015-08-14 21:08:10 or 2.5 don't know 2015-08-14 21:08:20 only because of lbu 2015-08-14 21:14:46 someone working on openjdk 8? 2015-08-14 21:23:48 not that I know of, but that doesnt say much 2015-08-14 21:46:48 chris|: What are you running on s6? I've been taking a good look at it, but haven't figured out how hotplug ties into everything. 2015-08-14 21:58:30 s6 seems p neat but damn all the stuff feels so intrusive 2015-08-14 21:58:47 maybe if I installed it into its own tree 2015-08-14 22:11:42 ammunta: A full s6 setup would make OpenRC useless, but it they can coexist. It's really to pick pieces out of s6. 2015-08-14 22:12:49 I'm p okay with making openrc useless 2015-08-14 22:13:35 for alpine 4? :P 2015-08-14 22:20:41 I mean, I'm doing runit thing atm 2015-08-14 22:21:28 s6 probably wouldn't be a huge issue and possibly actually working deps might be nice 2015-08-14 22:40:27 I will leave this here https://gitweb.gentoo.org/proj/openrc.git/tree/s6-guide.md 2015-08-15 07:55:18 ammunta: it's not that bad, you are not required to have the whole of the s6*utils installed in order to use the supervisor 2015-08-15 07:56:01 if you want to make extensive use of execline scripts, I'd recommend those though 2015-08-15 12:25:12 hi 2015-08-15 12:47:32 someone has an idea for #4538 2015-08-15 12:50:16 shafire: what happens if you install lvm2-dev and then try? 2015-08-15 12:50:59 oh 2015-08-15 12:51:00 this works 2015-08-15 12:51:47 something missing in the main package? 2015-08-15 12:52:26 looks like its checking /lib/ while it should be /lib/device-mapper/ 2015-08-15 12:53:33 the linker should read the symlink and use the real file instead of the symlink 2015-08-15 12:56:11 clandmeter: same for #4539 2015-08-15 12:58:34 does anyone know why nginx is not build with file-aio enabled? otherwise, I'd send in a patch for it 2015-08-15 13:00:14 chris|: if there is nothing in git log, you can asume it should be ok. 2015-08-15 13:05:33 clandmeter: okay, done 2015-08-15 13:05:38 thanks 2015-08-15 13:20:09 hi all 2015-08-15 13:20:24 is there reasons to not use cgmanager in lxc ? 2015-08-15 13:34:48 can I update the arm release also with setup-bootable 2015-08-15 13:34:48 ? 2015-08-15 13:35:29 http://forum.alpinelinux.org/forum/general-discussion/upgrading-kernel-raspberry-pi 2015-08-15 13:35:31 like him 2015-08-15 13:54:12 k0r10n - I think cgmanager is useful if you want to run unprivileged containers as a user (it's not required to run unprivileged as root) - but I don't think anyone has got unprivileged containers working yet with grsecurity. This is probably why it's not been packaged 2015-08-15 13:55:31 BitL0G1c: thnx 2015-08-15 15:21:19 Can I link in the commit message to an issue in redmine? 2015-08-15 15:22:17 yes #ISSUENUMBER 2015-08-15 15:22:50 ok 2015-08-15 16:39:43 found the solution for the lvm shared library error 2015-08-15 16:39:51 need to set library_dir in lvm.conf :S 2015-08-16 22:48:41 [28332.278046] grsec: Invalid alignment/Bus error occurred at 00007c0e64e5c000 in /sbin/apk[apk:27300] uid/euid:0/0 gid/egid:0/0, parent /bin/busybox[ash:24368] uid/euid:0/0 gid/egid:0/0 2015-08-16 22:48:53 does that raise anything ? 2015-08-16 22:49:03 happens during an apk upgrade 2015-08-16 22:52:17 mmmmh bad memory ... 2015-08-17 09:03:18 ncopa, there's a patch for busybox: %452. Can i apply? Do you agree? 2015-08-17 09:04:11 that is very scary change 2015-08-17 09:04:32 it will break lots of stuff 2015-08-17 09:04:40 setup scripts 2015-08-17 09:04:51 yeah 2015-08-17 09:05:04 i would prefer not just rename for cosmetic reasons 2015-08-17 09:06:01 according with the patch's comment, i supposed that this was discussed. 2015-08-17 09:08:24 hey :) 2015-08-17 09:08:29 it's not cosmetic, it's for the consistency :p 2015-08-17 09:08:57 I asked that day for a packaging guideline because every APKBUILD looks sometimes different to the others 2015-08-17 09:12:02 those was probably named like this log before anybody thought of consitency 2015-08-17 09:12:16 maybe those was the only services running at that time :) 2015-08-17 09:19:52 ncopa, remember me how to check if an upgrade breaks abi 2015-08-17 09:20:16 it's due to %460 2015-08-17 09:20:18 libdrm upgrade 2015-08-17 09:20:38 I am in favor for a strict guideline :) 2015-08-17 09:24:18 fcolista: checkapk 2015-08-17 09:25:02 ncopa, http://tpaste.us/A10Z 2015-08-17 09:25:05 what does it mean? 2015-08-17 09:25:14 (besides it adds new libraries) 2015-08-17 09:25:21 does it breaks abi? 2015-08-17 09:28:20 Can someone help with http://lists.alpinelinux.org/alpine-devel/4975.html 2015-08-17 09:28:30 I have no idea too and this is not the first time someone asked this 2015-08-17 09:28:36 no idea either* 2015-08-17 09:33:40 fcolista: there are only + and no - 2015-08-17 09:33:44 so it only adds stuff 2015-08-17 09:33:55 so it is safe 2015-08-17 09:34:03 ok 2015-08-17 09:34:13 thx 2015-08-17 09:35:28 shafire: i dont think setup-bootable should be used at all for arm 2015-08-17 09:35:56 we probably need fix the upgrade documentation and tell how to upgrade arm 2015-08-17 09:36:03 yes 2015-08-17 09:36:10 i think it is done differently 2015-08-17 09:40:10 fcolista: you merged the wrong version of the redshift upgrade patch, it contains a typo in license 2015-08-17 09:40:28 nmeum, please check. I changed GLP to GPL 2015-08-17 09:40:35 this before committing 2015-08-17 09:41:15 it's called GPLv3 or not? 2015-08-17 09:41:37 fcolista: nope http://git.alpinelinux.org/cgit/aports/tree/testing/redshift/APKBUILD#n8 2015-08-17 09:42:07 so git commit --amend -v didn't worked? 2015-08-17 09:42:37 seems so 2015-08-17 09:42:43 well…it's still license="GLP3" so probably not 2015-08-17 09:42:56 you could have merged %473 instead 2015-08-17 09:43:13 i'm applying patch one for one, still not reached 473 2015-08-17 09:43:18 nayway i'll fix it asap 2015-08-17 09:43:21 *anyway 2015-08-17 09:43:24 it's not a big deal 2015-08-17 09:43:44 yeah, I know I was wondering… 2015-08-17 09:44:03 473 was the fixed version of the patch you merged 2015-08-17 09:44:08 but anyway, it isn't a big deal 2015-08-17 09:50:58 nmeum: I'm thinking of adding this change to the -doc check: http://tpaste.us/ApZE 2015-08-17 09:51:38 it will check for compressed always, and not only when usr/share/man exist but is not a -doc 2015-08-17 09:52:10 hm 2015-08-17 09:52:43 also, it will instead of using find only check the first level 2015-08-17 09:53:05 i think it will speed up things if there many files which are all ok 2015-08-17 09:55:41 ncopa: the think is after using my patch I found out that some packages install manpage to non-standard locations like /usr/share/man/fr 2015-08-17 09:55:58 your change would check if man pages is non-standard locations are compressed 2015-08-17 09:56:22 fakeroot is an example of this, it installed uncompressed french and german man pages 2015-08-17 09:56:28 *installs 2015-08-17 09:57:05 ok 2015-08-17 09:57:22 probably safer to do find then 2015-08-17 09:57:34 or we coudl check 2 levels too 2015-08-17 09:58:47 also: do we really want to install non-english man pages? 2015-08-17 09:58:55 http://tpaste.us/AWPK 2015-08-17 09:58:57 i dunno 2015-08-17 09:59:13 since we disable nls I would also suggest that we disable the installation of those man pages 2015-08-17 10:04:30 you could just add an additional check for this using 'find /usr/share/man -maxdepth 1 \! -name man[0-8]' while you are at it 2015-08-17 10:04:55 this is my current version: http://tpaste.us/3ga5 2015-08-17 10:05:43 @alpine-staff: what is the alpine-build-setup for arm? i know the wandboard is used but with which setup in detail? 2015-08-17 10:06:04 using micro-sd or sata-hd? 2015-08-17 10:07:42 ncopa: looks find to me but could you change `-name '*.[0-9]'` to `-name '*.[0-8]'` there should be no section 8 2015-08-17 10:07:46 *section 9 2015-08-17 10:16:18 morning 2015-08-17 10:19:54 clandmeter: could you put tpaste.us through https? 2015-08-17 10:21:43 ScrumpyJack: care to add an issue? https://github.com/clandmeter/turbo-paste/issues 2015-08-17 10:22:06 ill need to generate a cert, which needs an email address, which i didnt setup yet. 2015-08-17 10:22:49 sure 2015-08-17 10:22:57 thx 2015-08-17 10:23:16 ill see if i can add it today. first want to finish somehting else. 2015-08-17 10:23:16 looks like it's sitting on the same host/vm as b.a.o 2015-08-17 10:23:34 it is 2015-08-17 10:23:43 btw firefox has crashed twice this morning 2015-08-17 10:24:01 FF 40.0 2015-08-17 10:25:12 I also see ocassional crashes with FF 40.0 but I can't reproduce them reliably at the moment 2015-08-17 11:03:26 ScrumpyJack: yeah, it does 2015-08-17 11:03:33 something with grsecurity 2015-08-17 11:11:30 if you are refering to the kern.alert message generated when ff crashes, i think that's just grsec preventing the core dump 2015-08-17 11:11:58 no, i mean some memory allocation errors 2015-08-17 11:31:23 can lvm2 be backported to 3.2? 2015-08-17 11:47:09 is anybody able to use mpv with --vo=opengl-hq on a nvidia system with nouveau I get the following erorr: '[vo/opengl-hq] GLX not found' 2015-08-17 11:49:26 shafire, knot's apkbuild is broken. 2015-08-17 11:49:51 i just pusehd some fixes 2015-08-17 11:49:54 care to test? 2015-08-17 11:50:07 ok 2015-08-17 11:50:16 why the building work at my test machine? :S 2015-08-17 11:50:20 worked* 2015-08-17 11:50:36 you have some packages installed 2015-08-17 11:50:40 that you should not 2015-08-17 11:50:57 and needs to be explicitly set on makedepends 2015-08-17 11:51:05 mh 2015-08-17 11:52:31 I see 2015-08-17 11:54:55 looks %443 is already applied 2015-08-17 12:34:58 ncopa, there's something wrong with pango? Kodi 15 fails at pango.trigger step. 2015-08-17 12:37:18 what is the error? 2015-08-17 12:38:14 no error. It returns "1 error." 2015-08-17 12:38:26 i mean: no message 2015-08-17 12:39:09 but might be related to this: 2015-08-17 12:39:10 http://git.alpinelinux.org/cgit/aports/commit/?id=f36103ae1c1b38f71eeafcdf6207047b24d40966 2015-08-17 12:39:17 Executing gtk-update-icon-cache-2.24.28-r0.trigger 2015-08-17 12:39:17 Executing pango-1.36.8-r2.trigger 2015-08-17 12:39:17 1 errors; 1485 MiB in 506 packages 2015-08-17 12:39:17 >>> ERROR: kodi: all failed 2015-08-17 12:39:17 >>> kodi: Uninstalling dependencies... 2015-08-17 12:39:28 sudo apk fix 2015-08-17 12:39:52 pango is installed at runtime 2015-08-17 12:40:00 is not installed in the buildenv 2015-08-17 12:41:25 umh 2015-08-17 12:41:38 when is the trigger executed? After package install? 2015-08-17 12:50:44 can lvm2 be backported to 3.2? 2015-08-17 12:55:28 in case the question was missed, let me ask once again: 2015-08-17 12:55:32 @alpine-staff: what is the alpine-build-setup for arm? i know the wandboard is used but with which setup in detail? 2015-08-17 12:55:32 using micro-sd or sata-hd? 2015-08-17 12:55:45 boot from sata? 2015-08-17 12:56:44 StarWarsFan, run-from-ram booting SDcard; LXC on SATA disk 2015-08-17 12:57:04 i think the other builder was with USB disks 2015-08-17 12:57:46 which wandboard is used? 2015-08-17 12:58:19 the quad one? 2015-08-17 12:58:32 StarWarsFan: http://www.wandboard.org/images/blog/2015-wandboard-holiday.png <- :p no help 2015-08-17 12:58:32 i'd like to seup a similar build-machine for our stuff too... 2015-08-17 13:09:27 i c, thx a lot 2015-08-17 13:10:29 shafire, quad 2015-08-17 13:14:52 fabled: did you look at %475 yet? What do you think of it? 2015-08-17 13:15:18 s/fabled/fcolista/ (sorry) 2015-08-17 13:16:38 nmeum, honestly, i don't like it 2015-08-17 13:16:55 but i'd prefer ncopa to express 2015-08-17 13:17:43 that alright, but why don't you like it? 2015-08-17 13:17:47 *that's 2015-08-17 13:17:57 +source="http://dev.alpinelinux.org/archive/$pkgname/core-$_realver.tar.gz 2015-08-17 13:17:57 + http://dev.alpinelinux.org/archive/$pkgname/extras-$_realver.tar.gz 2015-08-17 13:17:57 + http://dev.alpinelinux.org/archive/$pkgname/libselinux-$_realver.tar.gz 2015-08-17 13:17:57 + http://dev.alpinelinux.org/archive/$pkgname/f2fs-tools-$_realver.tar.gz 2015-08-17 13:18:08 those should be splitted in other packages 2015-08-17 13:18:21 imho 2015-08-17 13:18:39 but they are not needed by anything else and can't really be used as a standalone 2015-08-17 13:26:26 f2fs-tools can't? 2015-08-17 13:26:34 we have it already in repositories IIRC 2015-08-17 13:26:43 oh 2015-08-17 13:27:12 that's the linux kernel version, it's different (iirc) 2015-08-17 13:28:58 nmeum: is that so we can do android development from alpine? 2015-08-17 13:29:38 yeah, you can use adb, fastboot etc with this 2015-08-17 13:30:51 external/libselinux external/f2fs-tools 2015-08-17 13:31:07 can we build those as separate? 2015-08-17 13:31:09 hum 2015-08-17 13:31:16 we dont support selinux at all 2015-08-17 13:31:45 is it needed for android or for the mkbootimg/astboot/adb? 2015-08-17 13:32:14 fastboot 2015-08-17 13:35:22 fastboot is needed for android and libselinux is needed because android uses selinux I guess 2015-08-17 13:35:57 have you been able to build anything for andoid with this? 2015-08-17 13:36:15 i was able to get the sdk manager running with java 2015-08-17 13:36:23 this isn't a build tool chain, it just include tools 2015-08-17 13:36:31 like adb for debugging and fastboot for flashing things 2015-08-17 13:36:35 ok 2015-08-17 13:36:46 but if you do android development that this is kind of handy 2015-08-17 13:36:55 I just use it for flashing and debugging purposes 2015-08-17 13:37:11 what do you intend to use it for? 2015-08-17 13:37:27 (just curious) 2015-08-17 13:37:51 i think i built some of it manually to get the android sdk running 2015-08-17 13:38:01 i got stuck some place 2015-08-17 13:38:29 I run a custom rom and recovery on my android phone fastboot is needed to flash new recovery version and I use adb to copy file and to get a root shell 2015-08-17 13:38:38 ok 2015-08-17 13:43:07 nmeum: i suppose we can use system f2fs-tools instead of andoid tools fork? 2015-08-17 13:44:37 the packages from edge will come in 3.2.4 or when? 2015-08-17 13:44:40 I haven't tried but I don't think that it will work 2015-08-17 13:46:31 i will try 2015-08-17 13:46:42 seems like f2fs-tools has a library 2015-08-17 13:47:01 usr/lib/libf2fs.so.0.0.0 2015-08-17 13:47:20 hum 2015-08-17 13:47:22 no headers 2015-08-17 13:48:01 shafire: edge/main will come in v3.3 2015-08-17 13:48:26 in november we will branch 3.3-stable when v3.3.0 is released 2015-08-17 13:48:40 possible to get lvm2 backported to 3.2? 2015-08-17 13:48:59 we already have lvm2 in v3.2 2015-08-17 13:49:07 i need to look at the bug 2015-08-17 13:49:12 but very possible 2015-08-17 13:49:39 if: 1) it fixes a specific issue/bug, 2) does not introduce any new bugs/issues 2015-08-17 13:50:41 ok i think system f2fs-tools iw very possible 2015-08-17 13:50:45 is* 2015-08-17 13:51:45 ok, you might as well try to use the upstream libselinux version instead of androids fork then 2015-08-17 13:52:18 do you know what android changed in their f2fs-tools fork? 2015-08-17 13:52:24 no 2015-08-17 13:52:41 i dont know if then modified anything at all 2015-08-17 13:53:02 probably just lazy 2015-08-17 13:53:47 maybe we should find that out, I don't wanna end up with a bricked phone :) 2015-08-17 13:53:51 :) 2015-08-17 13:53:54 I guess I will just run diff(1) on the two repos 2015-08-17 13:54:14 check what version they forked 2015-08-17 13:56:52 according to the CHANGELOG file they probably forked f2fs-tools 1.0.0 2015-08-17 13:57:38 here is the diff if you compare both master branches http://sprunge.us/ENaW 2015-08-17 13:57:42 (looks pretty different too me) 2015-08-17 13:58:01 they even have different commandline flags 2015-08-17 13:59:24 but we could infact create a seperate package for that and call it android-f2fs-tools or something and let android-tools depend on it as fcolista suggested 2015-08-17 13:59:48 nmeum, ncopa, imho this is a cleaner way 2015-08-17 14:04:03 nmeum: yes. i also think we need prefix the soruce tarballs with android-tools- 2015-08-17 14:04:17 i'm on it 2015-08-17 14:04:32 great 2015-08-17 14:06:22 pushed 2015-08-17 14:07:18 this is relevant too: https://wiki.debian.org/AndroidTools 2015-08-17 14:15:23 ncopa: thanks 2015-08-17 14:15:45 i am interested in android development from alpine 2015-08-17 14:17:41 http://www.linbit.com/en/n/news - DRBD9 Release 2015-08-17 14:19:48 ncopa: yeah, it would be cool if android development would be possible on alpine 2015-08-17 14:20:09 vkris: does anyboy actually use drbd? 2015-08-17 14:20:37 maybe a first step would be to port openjdk8 since android uses java8 (iirc) 2015-08-17 14:21:02 I tried the docs, seems a bit lot to setup, but I may be wrong 2015-08-17 14:21:25 but really don't know about who uses them 2015-08-17 14:21:39 openjdk7 worked for the manager iirc 2015-08-17 14:22:10 though the customer list seems long 2015-08-17 14:28:45 i remember using vmware on top of drbd long time ago 2015-08-17 14:28:54 was too slow to be useful 2015-08-17 14:28:58 things might have changed 2015-08-17 14:29:08 and it was on ubuntu... 2015-08-17 14:31:48 vkris: can you verify if it will work with the current .initd script? 2015-08-17 14:32:39 current v8.4.3 ? 2015-08-17 14:34:00 cannot promise, did not finnish the docs/howtos, would give a try though 2015-08-17 14:36:14 is it needed, http://git.linbit.com/gitweb.cgi?p=drbd-utils.git;a=summary ? 2015-08-17 14:38:22 yes 2015-08-17 14:38:37 With DRBD module version 8.4.5, we split out the management tools 2015-08-17 14:38:37 into their own repository at http://git.linbit.com/drbd-utils.git 2015-08-17 14:38:37 (tarball at http://oss.linbit.com/drbd) 2015-08-17 14:38:37 That started out as "drbd-utils version 8.9.0", 2015-08-17 14:38:37 and provides compatible drbdadm, drbdsetup and drbdmeta tools 2015-08-17 14:38:38 for DRBD module versions 8.3, 8.4 and 9 (prereleases as of now). 2015-08-17 14:38:51 i wonder what version of drbd we ship with the 3.18 kernel 2015-08-17 14:39:08 version: 8.4.5 2015-08-17 14:39:16 test 2015-08-17 14:39:39 ok we probably just want rename the drbd package to drbd-utils 2015-08-17 14:39:54 and use the kernel module shipped with mainline kernel 2015-08-17 14:43:41 for v9 they have http://git.linbit.com/gitweb.cgi?p=drbdmanage.git;a=summary 2015-08-17 14:55:09 vkris: i suppose it needs kernel module too 2015-08-17 14:55:20 so we probably waith with drbd 9 til our kernel has support for it 2015-08-17 14:55:49 ok 2015-08-17 15:26:20 ncopa: fwiw, we use DRBD (but not in AL yet) 2015-08-17 15:31:16 ncopa: did you take a look at lvm2 bugs? 2015-08-17 18:23:47 Hi. Are there any news for the opensmtpd package in alpine linux? http://bugs.alpinelinux.org/issues/4012 2015-08-18 07:12:04 good morning 2015-08-18 07:13:34 morning 2015-08-18 07:20:59 openssh 7 released 2015-08-18 07:21:31 yeah, i think somewhere last week 2015-08-18 07:22:32 i noticed today when my pub key were no longer working 2015-08-18 07:22:36 *was 2015-08-18 07:23:34 oh, why was it no longer working?? 2015-08-18 07:23:38 how old was that key? 2015-08-18 07:23:49 4 years is standard or not? :D 2015-08-18 07:23:52 i'm using it since several years 2015-08-18 07:24:09 my but keys don't have a lifetime? :S 2015-08-18 07:24:11 anyway, you need to add the directive: PubkeyAcceptedKeyTypes $TYPEKEY 2015-08-18 07:24:20 or was the key type removed? 2015-08-18 07:24:21 in ssh_config 2015-08-18 07:26:14 I get this often: 2015-08-18 07:26:15 Aug 15 08:39:13 dpr-rpi12 auth.crit sshd[1570]: fatal: ssh_dispatch_run_fatal: no matching key exchange method found [preauth] Aug 15 08:39:13 dpr-rpi12 auth.err sshd[1570]: error: mm_request_receive: socket closed 2015-08-18 07:29:36 morning 2015-08-18 07:39:21 ncopa: don't forget lvm2 2015-08-18 07:41:47 hey ncopa 2015-08-18 07:42:05 shafire: will look at it soonish 2015-08-18 07:42:13 leo-unglaub: hi 2015-08-18 09:15:04 shafire: if i understand lvm2 problem correctly, then it should be enought to cherry-pick http://dup.pw/aports/27a1bb5 and http://dup.pw/aports/0a8f934cc to 3.2-stable ? 2015-08-18 09:15:25 yes 2015-08-18 09:15:48 will you keep the current pkgver? 2015-08-18 09:16:56 i think so, unless there are a good reason to not do that? 2015-08-18 09:18:10 should be ok 2015-08-18 09:18:25 wow I see openjdk8 :O 2015-08-18 09:32:08 ncopa: and openvswitch please too back to 3.2 http://lists.alpinelinux.org/alpine-devel/4977.html 2015-08-18 09:36:15 yeah, openjdk8 :) 2015-08-18 09:37:41 I hope openjdk9 will come faster 2015-08-18 09:45:19 i hope it will be possible to bootstrap openjdk with gcc-java 2015-08-18 09:52:37 I'm getting a apk file ownership conflict between man-pages and shadow-doc 2015-08-18 09:52:40 http://st.ilet.to/err 2015-08-18 09:57:03 thanks 2015-08-18 10:08:47 how do i resolve the conflict? 2015-08-18 11:10:19 ScrumpyJack - I'm working on shadow today & will look at it - am trying trying to find the segfault in newuidmap for unprivileged containers 2015-08-18 11:24:48 ScrumpyJack - rebuild with these added to package() http://hastebin.com/yuragocexo.bash 2015-08-18 11:30:13 nmeum: who do you suggest shoudl own passwd.5 and getspnam.3? 2015-08-18 11:30:19 ERROR: man-pages-4.02-r4: trying to overwrite usr/share/man/man5/passwd.5.gz owned by shadow-doc-4.2.1-r2. 2015-08-18 11:30:19 ERROR: man-pages-4.02-r4: trying to overwrite usr/share/man/man3/getspnam.3.gz owned by shadow-doc-4.2.1-r2. 2015-08-18 11:33:48 i suppose man-pages? 2015-08-18 11:33:51 ncopa: good question, I would remove the man page from shadow-doc because passwd5) is also useful on a busybox system without shadow 2015-08-18 11:33:59 *passwd(5) 2015-08-18 11:34:03 yeah 2015-08-18 11:34:11 getspnam too probably 2015-08-18 11:34:15 yeah 2015-08-18 11:34:16 probably 2015-08-18 11:34:50 gotta go to lunch 2015-08-18 11:36:42 bon apetit 2015-08-18 11:38:25 ScrumpyJack: i pushed fix for man-pages/shadow-doc conflict. thanks 2015-08-18 11:39:14 thanks 2015-08-18 11:39:53 thanks both of you :0 2015-08-18 11:44:31 i just sent a patch for sc but i think it's messed up 2015-08-18 13:13:49 im packaging, and the build uses libtool. linking is failing. gcc can't find the .o files that were previously compiled 2015-08-18 13:14:32 they were put in .libs, but gcc is looking for then in pwd 2015-08-18 13:18:38 libtool uses a libtoola rchive script *.la 2015-08-18 13:18:55 which points to .libs/ 2015-08-18 13:19:05 libtool should fix that at make install 2015-08-18 13:19:25 each object file seems to be made twice. should that be the case? http://st.ilet.to/libt 2015-08-18 13:20:14 i think first is makefile invoking libtool, and second is libtool invoking gcc 2015-08-18 13:20:36 that makes sense 2015-08-18 13:21:01 but why the different output paths? 2015-08-18 13:22:29 i think it is so you can run the compiled program from build dir 2015-08-18 13:23:14 otherwise you'd have to set LD_LIBRARY_PATH or install the shared lib in /usr/lib before you could run your program 2015-08-18 13:23:17 how can i bet libtool to look in .libs for the objects? 2015-08-18 13:23:19 st.ilet.to/liberr 2015-08-18 13:24:18 i think you shouldnt 2015-08-18 13:24:28 isnt there .lo files? 2015-08-18 13:24:36 i think the .lo should be used instead of .o 2015-08-18 13:25:45 there didn't seem to be a way to path the objects https://www.gnu.org/software/libtool/manual/html_node/Link-mode.html 2015-08-18 13:32:30 here is the Makefile line i think: 2015-08-18 13:32:35 %: %.o 2015-08-18 13:32:42 $(LIBTOOL) --tag=CC --mode=link $(CC) -o $@ -static $^ $(OBJS) $(LDFLAGS) -lz ${LIBS} 2015-08-18 13:34:57 is that target wrong? 2015-08-18 17:59:21 ncopa: what's your opinion on removing testing/installflash? It doesn't build anymore and I don't think that it is needed at all 2015-08-18 17:59:31 see also: %353 2015-08-19 07:14:41 good morning 2015-08-19 07:48:40 morning 2015-08-19 07:53:48 nmeum: i removed installflash 2015-08-19 07:54:10 nmeum: i think we shoudl wait with rebuilding packagse for getting man pages compressed 2015-08-19 07:54:20 til we have solved the man page compression problems 2015-08-19 07:54:43 we should probably compress all variants 2015-08-19 07:54:54 perl *.3pm pages 2015-08-19 07:55:01 posix pages, *.3p 2015-08-19 07:55:16 seems to exst som *.[0-9]am too but i dont know what that is 2015-08-19 07:55:41 we also need fix the compressor to deal with symlinks and hardlinks 2015-08-19 08:04:55 hi 2015-08-19 08:05:24 where do I put my contributor name? as last one? as first one? alphabetical? 2015-08-19 08:05:40 shafire: does not matter imho 2015-08-19 08:05:47 first, last or aplha 2015-08-19 08:06:07 i'd suggest last 2015-08-19 08:06:18 I don't like unspecified processes :-( 2015-08-19 08:06:34 :) 2015-08-19 08:06:45 add it last then 2015-08-19 08:07:38 ok 2015-08-19 08:07:58 In patchwork there is a openvswitch patch, please backport it again 2015-08-19 08:14:37 shafire: i think the patch is wrong 2015-08-19 08:21:02 I don't know :S 2015-08-19 08:22:19 does /etc/init.d/ovs-vswitchd have osvsd-server as depends? 2015-08-19 08:22:28 and ovs-modules? 2015-08-19 08:32:26 ncopa: thanks for your response (the .iso question) 2015-08-19 08:33:35 ScrumpyJack: sorry for delay 2015-08-19 08:34:26 so alpine-mini would be bare alpine linux and alpine.iso would become a livecd? 2015-08-19 08:34:51 something like alpine-base and alpine-livecd? 2015-08-19 08:37:19 i was thinking alpine-bootonly and alpine-livecd 2015-08-19 08:37:24 or similar 2015-08-19 08:38:21 alpine-base works for me :) bootonly sounds restrictive with the word "only" :) 2015-08-19 08:40:38 we have an alpine-base.apk too 2015-08-19 08:46:27 are there any example pacakges with symlink man pages? 2015-08-19 08:48:34 dosfstools-doc 2015-08-19 08:51:14 tcl has hardlinks 2015-08-19 08:51:15 ok 2015-08-19 10:45:14 ncopa: what exactly is the problem with symlinked and hardlinked man pages? Does abuild try to compress them currently? 2015-08-19 11:07:04 nmeum: problem with symlink is that it does not touch symlink 2015-08-19 11:07:20 so if there is a symlink pointing to a man page 2015-08-19 11:07:37 the symlink breaks when the man page it points to is renamed to .gz 2015-08-19 11:08:41 problem with hardlink is that it ends up making multiple copies of same file instead of using same file with different names 2015-08-19 11:09:20 so if there are 16 hardlinks pointing to samefile, we end up with 16 copies instead of 1 2015-08-19 11:10:59 example of symlink problem: 2015-08-19 11:11:06 $ ls -l /usr/share/man/man8/fsck.vfat.8 2015-08-19 11:11:06 lrwxrwxrwx 1 root root 10 Aug 19 13:10 /usr/share/man/man8/fsck.vfat.8 -> fsck.fat.8 2015-08-19 11:11:48 fsck.vfat.8 is a symlink pointing to the now non-existing fsck.fat.8 because fsck.fat.8 was compressed to fsck.fat.8.gz 2015-08-19 11:20:35 hm, yeah that sucks 2015-08-19 21:50:00 go 1.5 might make packaging easier since it has (experimental) support for vendoring package dependencies \o/ 2015-08-19 21:50:13 go projects just need to start using this 2015-08-20 07:08:30 morning 2015-08-20 07:08:44 i will reboot bld1.a.o and bld2.a.o now 2015-08-20 07:10:28 bld1 is going down. 2015-08-20 07:10:29 3 2015-08-20 07:10:29 2 2015-08-20 07:10:30 1 2015-08-20 07:10:31 0 2015-08-20 08:42:08 should be all back now 2015-08-20 10:02:25 nfs seems to work good with v3.2 2015-08-20 10:02:36 i'm gonna use nfs for distfiles and build logs 2015-08-20 11:51:31 seems algitbot died 2015-08-20 11:52:10 argh... 2015-08-20 11:52:38 anybody using nfs on an alpine instance running as a xen domu? 2015-08-20 11:53:03 upgraded a machine from 3.1 to 3.2 and now nfs is not working anymore (again) 2015-08-20 11:53:19 dmesg tells me this: 2015-08-20 11:53:46 http://sprunge.us/KSTH 2015-08-20 11:54:04 possible that https://lkml.org/lkml/2013/4/16/168 is required? 2015-08-20 11:58:22 this is a full show stopper for us right now 2015-08-20 11:58:39 alpine 2.7 could not be used as buildnode, because no java7 available 2015-08-20 11:58:56 alpine 3.2 could not be used as buildnode, because nfs is broken 2015-08-20 11:59:06 so the whole build system is dead :-( :-( :-( 2015-08-20 11:59:34 i need a big coffee now... :-/ 2015-08-20 12:19:36 ok, filed #4543 2015-08-20 12:37:12 StarWarsFan: has mouting nfs in side a container ever been supported? 2015-08-20 12:37:37 https://lists.debian.org/debian-kernel/2014/09/msg00477.html 2015-08-20 12:40:50 why not? i'm using this some years now... 2015-08-20 12:41:51 so it worked with 3.14 kernels? 2015-08-20 12:42:12 oh, not server! i'm using nfs-client 2015-08-20 12:42:25 ok 2015-08-20 12:42:50 right, updated the machine from 3.1 to 3.2 yesterday and since then it's not working 2015-08-20 12:43:06 alpine 3.1 to alpine 3.2 2015-08-20 12:44:02 added a note regarding this to the issue 2015-08-20 12:46:25 is mounting nfs inside a container supposed to work? 2015-08-20 12:46:26 definitely not for unprivileged containers 2015-08-20 12:46:26 definetely not for any containers that are supposed to be secure 2015-08-20 12:47:38 ok. and tells me what? 2015-08-20 12:47:55 that i am still not sure if it is supposed to work 2015-08-20 12:48:52 strange. can't understand where the problem in general is... 2015-08-20 12:49:19 but if it's not working anymore, i've got a real problem 2015-08-20 12:51:26 do you have different nfs mount for each container? 2015-08-20 12:51:34 or is it a single shared nfs? 2015-08-20 12:51:50 i am about to set up nfs on build servers 2015-08-20 12:52:08 so servers can have a shared /var/cache/distfiles 2015-08-20 12:52:24 the way in intend to do it is mounting it on the host 2015-08-20 12:52:40 and then do bind mount in container config 2015-08-20 12:53:00 i c 2015-08-20 12:53:15 at the moment each container has it's own mount 2015-08-20 12:53:33 but there are alle the same target 2015-08-20 12:54:39 so the builders put their output into these folders on the nfs mount 2015-08-20 12:54:52 and one central jenkins job is updating the apkindex 2015-08-20 12:55:40 the whole build system is based on jenkins (which is why i need java7 on alpine 2.7, but this is another story) 2015-08-20 12:56:53 ok 2015-08-20 12:57:00 i need the nfs mount for the build logs too 2015-08-20 12:57:25 the nfs server is a another physical box? 2015-08-20 12:57:43 right 2015-08-20 12:58:08 running on ubuntu i think, not quite sure... 2015-08-20 13:00:30 ok 2015-08-20 13:00:35 where is what i did 2015-08-20 13:00:52 on the lxc host, mount -t nfs .... /mnt 2015-08-20 13:01:53 in container i added: 2015-08-20 13:01:55 lxc.mount.entry = /mnt var/cache/distfiles none bind,create=dir 0 0 2015-08-20 13:02:30 on the host i have: 2015-08-20 13:02:37 192.168.8.2:/var/cache/distfiles /mnt nfs rw,relatime,vers=3,rsize=32768,wsize=32768,namlen=255,hard,proto=tcp,timeo=600,retrans=2,sec=sys,mountaddr=192.168.8.2,mountvers=3,mountport=48466,mountproto=udp,local_lock=none,addr=192.168.8.2 0 0 2015-08-20 13:03:04 that seem to work 2015-08-20 13:03:15 i c 2015-08-20 13:03:32 StarWarsFan: can you try mount the nfs on the lxc host some place? 2015-08-20 13:03:56 i think this setup is safer too, then container does not need mount permissions 2015-08-20 13:04:29 sure, but will be later today 2015-08-20 13:05:03 to be safe not screwing up something when working on such thing in between business tasks... 2015-08-20 13:05:14 ;-) 2015-08-20 13:15:39 ncopa: just to let you know: found the performed update on the log: 2015-08-20 13:15:42 (55/337) Upgrading linux-grsec (3.14.48-r0 -> 3.18.20-r1) 2015-08-20 13:15:56 so up to 3.14.48-r0 it was working 2015-08-20 13:29:45 the dmesg you posted showed 3.18 2015-08-20 13:30:07 right, that's where the problem happens 2015-08-20 13:30:33 don't have a dmesg for 3.14 as it was working there... ;-) 2015-08-20 13:30:56 i believe it is an upstream problem 2015-08-20 13:31:07 mainline kernel problem 2015-08-20 13:31:11 maybe 2015-08-20 13:31:41 and i belive it was only "luck" that it worked on 3.14 kernel 2015-08-20 13:32:24 interesting... 2015-08-20 13:32:45 i'm using this setup since alpine ~2.5 2015-08-20 13:34:04 remember the other issues regarding this 2015-08-20 13:34:12 #2842, #2972 2015-08-20 14:15:23 it is probably a regression in mainline kernel then 2015-08-20 14:18:57 StarWarsFan: btw, we have openjdk8 now 2015-08-20 14:26:57 ncopa, we should probably take some earlier openjdk update version that is GA. 2015-08-20 14:27:00 and move to main 2015-08-20 14:27:04 it seems to be working 2015-08-20 14:27:09 at least basics 2015-08-20 14:28:32 seems jdk8u60-b24 is latest GA 2015-08-20 14:28:38 oh 2015-08-20 14:28:45 no, it's due to be GA this month 2015-08-20 14:29:03 jdk8u40-b32 is latest GA release 2015-08-20 14:31:08 GA? 2015-08-20 14:31:31 there is an unresolved issue with openjdk (including openjdk7) with the cacerts 2015-08-20 14:31:38 apparently there are 2 different ways to solve it 2015-08-20 14:31:51 import ca certs with keytool 2015-08-20 14:32:14 or use p11-kit to generate the java ca cert bundle 2015-08-20 14:32:26 ncopa: for 2.7? 2015-08-20 14:32:33 StarWarsFan: no 2015-08-20 14:32:40 openjdk8 is for v3.3 2015-08-20 14:32:41 see, that's the point 2015-08-20 14:32:52 i need java7 or higher for alpine 2.7 2015-08-20 14:33:11 why can you not use alpine v3.1? 2015-08-20 14:33:23 because it need to build packages for alpine 2.7 2015-08-20 14:33:45 v2.7 is EOL in nov 2015-08-20 14:34:08 real cross compiling would be great... 2015-08-20 14:34:57 hm... 2015-08-20 14:39:02 ncopa, the bundle should be generated from a hook from update-ca-certificates 2015-08-20 14:39:39 ncopa, GA = General Availability 2015-08-20 14:39:50 ncopa, that's what oracle uses as the 'now it's released' 2015-08-20 14:51:03 algitbot: you are not supposed to post upload messages in here... 2015-08-20 14:58:13 if no one is going to say it, i am going to ... that buildbot is crazy *g* 2015-08-20 19:58:47 ncopa: hm, can't find any option to bind a folder on the host to the domu 2015-08-20 19:59:58 to say it again: it's a xen-host, not lxc! 2015-08-20 20:00:04 and to make it more complex: 2015-08-20 20:00:26 it's a xen host with an alpine-domu 2015-08-20 20:00:58 on this container an alpine lxc-container is running 2015-08-20 20:01:05 so i have two level ;-) 2015-08-20 20:01:46 ubuntu-xen-host <- alpine-domu <- alpine-lxc-container 2015-08-20 20:02:13 both alpine containers use nfs-client until updating to alpine 3.2 2015-08-20 20:35:26 wut 2015-08-20 20:35:35 what is algitbot doing here? 2015-08-21 07:25:45 good morning! 2015-08-21 07:55:29 morning 2015-08-21 07:55:37 i'm gonna mess with the build server 2015-08-21 07:55:45 trying to enable jumbo frames, mtu 9000 2015-08-21 07:55:48 for the nfs stuff 2015-08-21 08:16:05 mtu 9000? 2015-08-21 08:52:53 yes, jumbo frames 2015-08-21 08:53:05 i'm slightly annoeyd by lxc 2015-08-21 08:53:22 the bond supports mtu 9000 2015-08-21 08:53:35 the bridge interface refuses due to veth set to 1500 2015-08-21 08:53:48 apparently you can manually set mtu in lxc configuration 2015-08-21 08:54:07 but i think it would be nice to clone the existing bridge interface mtu 2015-08-21 08:54:12 and it should be rather trivial 2015-08-21 09:28:21 hmm, i see 2015-08-21 09:28:38 to be honest, build environments in general are not optimal at all 2015-08-21 09:28:45 there has to be a better solution 2015-08-21 09:36:29 ncopa: your revert yesturday worked on my desktop 2015-08-21 09:36:34 but my laptop is still broken 2015-08-21 09:36:36 any ideas? 2015-08-21 09:36:48 try apk fix kmod 2015-08-21 09:39:03 yeah, that did it 2015-08-21 09:39:04 ! 2015-08-21 09:39:06 thanks! 2015-08-21 09:53:07 i am off to the red cross 2015-08-21 09:53:10 see you guys later 2015-08-21 09:53:14 if not, have a nice weekend :) 2015-08-21 09:59:44 i think i ixed algitbot to not post build status messages here again 2015-08-21 10:45:16 algitbot is punished 2015-08-21 12:02:53 ncopa: regarding the nfs-problem from yesterday 2015-08-21 12:03:06 unfortunately its not that easy to solve 2015-08-21 12:03:11 ok? 2015-08-21 12:03:34 because their is no option to access a folder on the host out of the container 2015-08-21 12:03:41 remember: its a xen-domu 2015-08-21 12:03:48 as i understand you have dom0, and domu 2015-08-21 12:03:58 right, but xen, not lxc 2015-08-21 12:04:04 ok 2015-08-21 12:04:11 another point: 2015-08-21 12:04:21 a collegue of mine checked the dmesg output 2015-08-21 12:04:22 so you want mount the dom0 to the container? 2015-08-21 12:04:34 and stated, that there will be the nfs-server startet 2015-08-21 12:04:36 but why? 2015-08-21 12:04:43 there should be only the nfs-client 2015-08-21 12:05:00 oh, got a phone call... 2015-08-21 12:05:00 brb 2015-08-21 12:05:58 is there a util in the repo that handles the lzma lossless data compression format used by lzip? 2015-08-21 12:14:45 ScrumpyJack - lz4 ? 2015-08-21 12:17:23 http://catchchallenger.first-world.info/wiki/Quick_Benchmark:_Gzip_vs_Bzip2_vs_LZMA_vs_XZ_vs_LZ4_vs_LZO 2015-08-21 12:18:16 BitL0G1c: nope 2015-08-21 12:18:49 if we don't have anything, there is a busybox patch http://download.savannah.gnu.org/releases/lzip/busybox/ 2015-08-21 12:22:22 gnu tar can, apparently 2015-08-21 12:23:15 no it can't :) 2015-08-21 12:32:48 lzo-2.09-r0 webpage: 2015-08-21 12:32:48 http://www.oberhumer.com/opensource/lzo 2015-08-21 12:39:04 ScrumpyJack - have a look @ xz - in Debian it also supports lzma https://packages.debian.org/jessie/xz-utils 2015-08-21 12:50:05 nope, doesn't work either 2015-08-21 12:50:25 patch sent for lzip 2015-08-22 15:08:23 fcolista: could you take a look at %496? 2015-08-22 18:20:23 can we include --with-http_auth_request_module in nginx apk configure options? 2015-08-23 08:22:15 hmm, currently mysql segfaults 2015-08-23 08:22:19 can anyone confirm that? 2015-08-23 08:23:07 https://paste.debian.net/plain/303715 2015-08-23 08:40:05 leo-unglaub: no, not on "latest" edge with 3.14 kernel 2015-08-23 08:41:46 can you get more meaningful backtrace? 2015-08-23 15:21:10 Is there currently anyone working to package gnuradio? 2015-08-23 15:36:18 Ok, so then it's me right now :-) 2015-08-23 17:00:49 <3 jomat! 2015-08-23 18:00:37 qdeyXHmyEe5v: I think I have the first binaries :-) 2015-08-23 18:03:56 qdeyXHmyEe5v: it still needs py-cheetah and qwt packaged, but that also worked quite straight forward 2015-08-23 18:08:12 yep, gnuradio-companion starts :-D 2015-08-23 18:08:54 nice! 2015-08-23 18:09:03 you also do osmo-sdr? 2015-08-23 18:09:14 and libhackrf? 2015-08-23 18:09:31 my goal is to operate the rad1o :-) 2015-08-23 18:09:40 ack 2015-08-23 18:10:15 thx! i had this on my todolist for way too long. 2015-08-23 18:10:29 (gnuradio, not rad1o) 2015-08-23 21:16:13 qdeyXHmyEe5v: Gnuradio disabled components: […] * HackRF Jawbreaker 2015-08-23 21:17:00 qdeyXHmyEe5v: dunno how long I'll continue today, I'd share the work done by now if you're interested… 2015-08-23 21:17:13 but its still very messy 2015-08-23 21:18:11 i know building gr is hard :/ 2015-08-23 21:18:30 unfortunately my alpine setup cannot handle gr atm. :/ 2015-08-23 21:18:36 (not building, running) 2015-08-23 21:38:16 jomat: yeah, can we see it? 2015-08-23 21:56:15 nmeum: https://github.com/jomat/alpine-gnuradio 2015-08-23 22:00:53 doesn't look too messy on a first look 2015-08-23 22:01:02 but yeah it isn't finished obviously 2015-08-23 22:01:43 a lot of dependencies are missing 2015-08-23 22:01:50 there's still no hackrf-support 2015-08-23 22:02:17 i haven't checked if the content of the packages makes sense on a second sight 2015-08-23 22:02:52 but i already hooked up a sine generator to a audiosink and run it, and that works without exploding :-) 2015-08-23 22:03:14 and i don't know if the mode_t.patch 2015-08-23 22:03:25 … is good… i doubt it 2015-08-23 22:04:37 and i don't know yet how to handle python2 and 3 with py-packages 2015-08-24 07:14:45 what's up with build-edge-x86 ? 2015-08-24 07:50:55 oh 2015-08-24 07:51:12 i think i know whats going on 2015-08-24 08:32:49 i have set up nfs for buildlogs 2015-08-24 08:33:01 so all build logs shouldbe available from same location 2015-08-24 08:33:12 http://build.alpinelinux.org/buildlogs/ 2015-08-24 08:35:10 good morning 2015-08-24 08:36:16 i have now narrowed it down, my mariadb segfault 2015-08-24 08:36:29 if you try to login and the login fails, the program segfaults 2015-08-24 08:36:36 ncopa: really great, thanks for that 2015-08-24 08:37:22 https://paste.debian.net/plain/304391 2015-08-24 08:37:45 leo-unglaub: its the mariadb server that segfaults? 2015-08-24 08:38:02 ncopa: mariadb-client 2015-08-24 08:38:14 aka mysql and not mysqld 2015-08-24 08:39:19 ncopa: try the following 2015-08-24 08:39:27 mysql -u foobar -p 188.40.92.4 2015-08-24 08:39:38 and then try 12345678 as a password 2015-08-24 08:41:28 mysql wihtout any opts segfaults too 2015-08-24 08:42:12 yeah, because without any options it tryes to connect to 127.0.0.1 with root as a username and an empty password 2015-08-24 08:45:12 if the login suceeds, there is no segfault 2015-08-24 08:45:29 does it happen with v3.2 too? 2015-08-24 08:45:31 or only edge 2015-08-24 08:45:36 i only have edge 2015-08-24 08:45:39 ok 2015-08-24 08:45:46 we need fix it there anyway 2015-08-24 08:49:14 i just checked on my remaining debian servers 2015-08-24 08:49:20 there mariadb does not have that problem 2015-08-24 08:49:26 so its not an upstream problem 2015-08-24 08:49:35 propobly something in muscle 2015-08-24 08:50:57 i have some wild guesses 2015-08-24 08:51:13 I am building it locally with debug symbols 2015-08-24 09:21:24 ncopa: how can you have a wild guess about a segfault? do you have jedi powers? 2015-08-24 09:21:43 how can someone see a segfault and have a guess when the programm is as bg as mariadb *g* 2015-08-24 09:24:15 my guess is that they make some GNU assumptions 2015-08-24 09:24:27 there are a couple of functions that are defined in posix 2015-08-24 09:24:38 but have special implementation in gnu 2015-08-24 09:24:58 you can enable gnu behavioud by setting _GNU_SOURCE with musl 2015-08-24 09:25:06 and include the proper headers 2015-08-24 09:25:17 otherwise program goes boom 2015-08-24 09:25:58 i think basename(3) and strerror_r(3) 2015-08-24 09:27:42 hmmm 2015-08-24 09:27:44 i see 2015-08-24 09:27:54 i hate it if implementations dont follow a standard 2015-08-24 09:28:02 (gdb) bt 2015-08-24 09:28:02 #0 0x00000ca2e3bab284 in strmake () 2015-08-24 09:28:02 #1 0x00000ca2e3bb3a87 in my_strerror () 2015-08-24 09:28:17 ha 2015-08-24 09:28:46 seems like i actually do have jedi powers... 2015-08-24 09:28:48 :) 2015-08-24 09:29:03 :) may the source be with you 2015-08-24 09:30:13 https://github.com/MariaDB/server/blob/10.1/strings/my_vsnprintf.c#L835 2015-08-24 09:30:25 my get is 2015-08-24 09:30:30 _GNU_SOURCE is defined 2015-08-24 09:30:42 so mysql expects GNU behaviour 2015-08-24 09:30:51 however, musl does not implement both 2015-08-24 09:31:00 it only implement the standard compliant version 2015-08-24 09:31:04 (the only sane thing to do) 2015-08-24 09:33:31 hm, can we just undefine _GNU_SOURCE? 2015-08-24 09:33:58 because it's kinda correct assumption that strmake is there if _GNU_SOURCE is defined 2015-08-24 09:34:22 no 2015-08-24 09:34:32 strmake is mysql thingy 2015-08-24 09:34:37 problem is strerror_r 2015-08-24 09:34:39 I see 2015-08-24 09:34:45 both glibc and musl implement it 2015-08-24 09:34:50 glibc has 2 variants of it 2015-08-24 09:34:55 one standard compliant 2015-08-24 09:35:01 and one gnu - non compliant 2015-08-24 09:35:07 meh 2015-08-24 09:35:15 musl only have the standard compliant 2015-08-24 09:35:31 yeah, I figured it out 2015-08-24 09:35:35 with glibc you get the gnu one if you do _GNU_SOURCE 2015-08-24 09:35:46 so proper is to check for __GLIBC__ i think 2015-08-24 09:35:54 http://tpaste.us/ARjW 2015-08-24 09:36:01 and fallback to standard complaint otherwise 2015-08-24 09:36:58 hm, what ends up used with your patch? 2015-08-24 09:37:27 POSIX strerror_r? 2015-08-24 09:37:29 yeah 2015-08-24 09:42:46 ncopa: do you want me to send the patch upstream? i have an account in there jira 2015-08-24 09:43:12 i think it can be simplified 2015-08-24 09:43:27 if gnu libc is the only broken implelemtnation 2015-08-24 09:43:48 then we coudl check for #defined(__GLIBC__) && defined(_GNU_SOURCE) 2015-08-24 09:45:02 leo-unglaub: but it woudl be nice if you report the bug 2015-08-24 09:45:05 and explain the issue 2015-08-24 09:46:02 ncopa: sure thing, i will take care of it 2015-08-24 09:46:12 thanks for knowing the problem 2015-08-24 09:46:42 fir first patch i posted is broken too btw 2015-08-24 09:46:45 has some typos 2015-08-24 09:51:35 i will explain them the problem 2015-08-24 09:52:03 but i am still pissed at gnu for doing that 2015-08-24 09:52:20 on the other hand, i can understand that if a standard sucks that you try to fix it in some way 2015-08-24 09:53:26 this is a better fix i think: http://tpaste.us/A9K4 2015-08-24 10:23:36 brb lunch 2015-08-24 10:40:59 ncopa: I considered upgrading main/go from 1.4 to 1.5 the thing is it needs to be bootstraped now either using gcc-go (which doesn't seem to work on musl) or go1.4. So what's the best way to make the go package depend on an older version of itself? 2015-08-24 10:49:56 nmeum: can we have separate go1.4 package? 2015-08-24 10:50:21 that would be possible the question is if it is the best solution 2015-08-24 10:50:29 I think so 2015-08-24 10:50:42 it solves chicken and egg problem if we ever add another arch 2015-08-24 10:51:40 right 2015-08-24 10:52:29 I would personally prefer to bootstrap it with gcc-go but I don't know how much work it woudl be to patch gcc in order to make gcc-go work under musl 2015-08-24 10:55:14 I guess I will just rename the current go1.4 package to go-bootstrap and add a new go package for 1.5 2015-08-24 10:55:28 ncopa: ^ does that sound good to you? 2015-08-24 11:04:45 how is go 1.4 built? 2015-08-24 11:05:31 I'm ok with a go-bootstrap or go1.4 package 2015-08-24 11:08:01 go1.4 sounds better to me, but it's up to you 2015-08-24 11:08:19 i tend to agree 2015-08-24 11:08:23 go1.4 sounds better 2015-08-24 11:09:08 damn... i need go 1.5 :( 2015-08-24 11:10:27 ncopa: go 1.4 doesn't need to be bootstraped because the compiler toolchain is mainly written in C with go 1.5 it is almost completly written in Go itself 2015-08-24 11:11:48 nmeum: makes sense 2015-08-24 11:12:00 i wanted to build dron ci within my alpine container, but the newest version relies on golang 1.5 -.- 2015-08-24 11:12:09 s/dron/drone/ 2015-08-24 11:12:28 as I said: I am working on a go 1.5 package but I don't have much time this week 2015-08-24 11:14:45 ncopa: BTW could you accept %496? 2015-08-24 11:15:46 yeah, ic... so i've got to wait for that 2015-08-24 11:16:22 or i should use https://github.com/docker-library/golang/blob/bcc3800fbed7f5116ff27ae82e0d881efd50aa69/1.5/Dockerfile#L9-L16 for the meantime... 2015-08-24 11:19:12 btw, does the binary golang work with musl libc + libc6-compat? 2015-08-24 11:31:48 of course it doesnt work -.- 2015-08-24 11:32:27 mosez: have you tested it? 2015-08-24 11:32:54 musl provides partial glibc compat 2015-08-24 11:33:16 i know that adobe flashplayer works 2015-08-24 11:33:18 i have downloaded the binary, added libc6-compat as well 2015-08-24 11:33:25 ok 2015-08-24 11:33:32 # ./go 2015-08-24 11:33:32 bash: ./go: No such file or directory 2015-08-24 11:33:40 # ldd go /lib64/ld-linux-x86-64.so.2 (0x7f35eafc7000) libpthread.so.0 => /lib64/ld-linux-x86-64.so.2 (0x7f35eafc7000) libc.so.6 => /lib64/ld-linux-x86-64.so.2 (0x7f35eafc7000) 2015-08-24 11:35:12 ah, if i symlink /lib to /lib64 it seems to work :D 2015-08-24 11:35:37 it works? 2015-08-24 11:35:46 thats awesome 2015-08-24 11:36:12 ./go outputs the help for sure, let me check if it can compile as well. 2015-08-24 11:36:30 might be it can be used for bootstrapping go itself 2015-08-24 11:37:35 yeah, looks like that :) 2015-08-24 11:37:50 libc6-compat and the binary download just as a build dependency :) 2015-08-24 11:41:45 stf: just compiled hackrf, looks good, next step is gr-osmosdr now :-) 2015-08-24 11:47:56 leo-unglaub: i just pushed mariadb fix 2015-08-24 11:48:49 nice, i will give it a try as soon as the build is done+ 2015-08-24 11:49:03 i already tested it 2015-08-24 11:49:07 it seems to work 2015-08-24 11:49:16 did you file a bug to mariadb? 2015-08-24 11:49:28 i am about to do a pull request 2015-08-24 11:49:55 ncopa: not yet, i am still in a meeting and have to pretend to listen to the other guy who is sealing all my donougts .... that fat bastard 2015-08-24 11:51:12 https://github.com/MariaDB/server/pull/93 2015-08-24 11:51:20 ncopa: so i can confirm that https://gist.github.com/tboerger/f87236c71c4d8cb7f49f works on edge :) 2015-08-24 11:51:48 thats pretty cool IMHO :) 2015-08-24 11:52:59 nmeum: https://gist.github.com/tboerger/f87236c71c4d8cb7f49f works :) 2015-08-24 11:53:22 ah, well using the compiled version of go1.5 actually causes some problems 2015-08-24 11:53:29 for instance cgo shouldn't work 2015-08-24 11:53:53 but if it works for you that's fine 2015-08-24 11:57:12 nmeum: i think this can get used to build the go package natively :) 2015-08-24 11:58:35 I wouldn't use the go 1.5 binary version to for bootstraping 2015-08-24 11:58:44 why not? 2015-08-24 11:58:59 for me it sounds like a nice idea :) 2015-08-24 11:59:49 ncopa: patch works! 2015-08-24 12:10:58 j\o/mat :) 2015-08-24 12:12:23 ncopa: so, meeting is done 2015-08-24 12:12:34 should i report the bug now or do you want to do it? 2015-08-24 12:12:55 awesome... the tool built with the prebuilt binary core dumps now -.- 2015-08-24 12:14:17 :) 2015-08-24 12:14:39 i don't wanna switch to another distro :( 2015-08-24 12:15:06 mosez: i think the way forward is to build go properly with go1.4 2015-08-24 12:15:44 ncopa: unfortunately no time to dig into the nfs problem from last week 2015-08-24 12:15:57 hope to be able to do this in the next days... :-/ 2015-08-24 12:16:26 so you could not mount nfs on the lxc host? (the domu) and do bind mounts in the containers? 2015-08-24 12:16:56 in my setup i do so 2015-08-24 12:17:37 btw, i sent a patch to lxc upstream for proper jumbo frame support in containers 2015-08-24 12:17:47 as i said, it's xen, not lxc 2015-08-24 12:18:15 oh, you dont use lxc at all? 2015-08-24 12:18:30 does https://build.opensuse.org/package/show/devel:languages:go/go help how to handle the go 1.5 build? :) 2015-08-24 12:19:05 there is no problem with the go 1.5 build 2015-08-24 12:19:18 I simply didn't have time to continue working on it so far 2015-08-24 12:21:19 good and sad at the same timwe 2015-08-24 12:21:27 you are free to submit a patch for the upgrade if you have more time then I do 2015-08-24 12:23:13 just put alpine on a usb stick 2015-08-24 12:23:28 cache is /media/usb/cache 2015-08-24 12:23:41 setup-alpine and lbu committed 2015-08-24 12:24:16 when i reboot, packages in /etc/apk/world don't load 2015-08-24 12:24:32 the load if i run apk fix 2015-08-24 12:24:44 ncopa: that's complicated. it's a xen-host with an alpine-domu. in this alpine-domu i use an alpine lxc container. 2015-08-24 12:25:06 a little bit strange but working for some years until the upgrade to 3.2 2015-08-24 12:25:20 should the packages in world load at boot? what's the prefered method of getting them to load? 2015-08-24 12:25:51 so on the alpine-xen-domU i need to use nfs-client and on the lxc container, which is running on the xen-domU too 2015-08-24 12:26:09 again: this worked for some years like a charm 2015-08-24 12:26:42 and again: nfs-client! the whole security-diskussion regarding nfs on an container is around nfs-server i think... 2015-08-24 12:27:08 even my dmesg output on the regarding issue shows the startup of the nfs-server 2015-08-24 12:27:21 and i have no idea why the server is starting up, as there should be only the nfs client 2015-08-24 12:28:19 StarWarsFan: what u propose is: on your alpline-domu (the lxc "host") you mount nfs 2015-08-24 12:28:51 on your containers you do do bind mount of the mounted dir 2015-08-24 12:29:24 exactly this is what i did for some years and what is no longer working since the 3.2 upgrade 2015-08-24 12:29:29 :-( 2015-08-24 12:29:46 so you do not mount nfs from within each container? 2015-08-24 12:29:55 but only from the alpine-domu? 2015-08-24 12:30:24 [ 28.251209] NFSD: attempt to initialize umh client tracking in a container! 2015-08-24 12:30:59 that message typically happens when you try mount nfs from the container 2015-08-24 12:32:34 stf: \o/ I see a live fft waterfall! So glad it seems to work :-) 2015-08-24 12:33:06 \o/ 2015-08-24 12:35:28 i mount an nfs-share on an alpine-xen-domu, so this is an nfs-client 2015-08-24 12:35:49 on this alpine-xen-comu and alpine-lxc-container is running, 2015-08-24 12:35:55 which also mounts an nfs-share 2015-08-24 12:36:03 so this is anohter nfs-client 2015-08-24 12:36:12 and both are not working anymore 2015-08-24 12:38:45 the dmesg output on the issue is from the alpine-xen-domu. so is it possible that the output there is from the startup of the lxc-container? 2015-08-24 13:17:27 StarWarsFan - in the hosts /etc/sysctl.conf set: kernel.dmesg_restrict=1 (to help keep dmesg separate in LXC) 2015-08-24 13:17:52 ah ok, thx 2015-08-24 13:29:55 In Debian Jessie "# verify the static apk binary signature" fails for Alpine LXC templates http://hastebin.com/ekirijoger.sm - commenting out "openssl dgst......." fixes it. The 3 files are available to be verified - so perhaps it is a problem with the signature 2015-08-24 14:00:38 BitL0G1c: it's been reported and fixed already in lxc git 2015-08-24 14:00:57 gotta go, so I can't give you exact issue. take a look at their github 2015-08-24 14:03:44 ok - np 2015-08-24 14:11:14 was https://github.com/lxc/lxc/issues/609 2015-08-24 14:50:03 ncopa: do really think that it is a good idea to put all android tools into testing/android-tools? 2015-08-24 14:50:32 maybe we should differentiate between platform and build tools? 2015-08-24 14:50:40 archlinux does it this way, it might be a good idea 2015-08-24 14:50:57 ok 2015-08-24 14:50:59 might be 2015-08-24 14:51:06 just a suggestion :p 2015-08-24 15:41:19 https://www.archlinux.org/packages/community/x86_64/android-tools/ I see just three executables here 2015-08-24 16:45:13 almost finished with the go 1.5 upgrade 2015-08-24 16:45:45 barthalion: https://wiki.archlinux.org/index.php/Android#Android_SDK_core_components 2015-08-24 17:21:48 ncopa: I updated #4543 2015-08-24 17:25:26 mosez: I finished working on the go 1.5 upgrade, if you want to test it (which would be great) than take a look at %504 and %503 2015-08-24 18:03:52 nmeum: well, it's sdk, different thing than tools 2015-08-24 20:15:33 ncopa: gotcha, solved the nfs issue :-) 2015-08-24 20:15:50 unfortunately some kind of PEBKAC 2015-08-24 20:15:58 anyway, the issue could be closed 2015-08-24 20:16:20 oh, and please remove comment #5, I forgot to add some content there... 2015-08-24 21:46:26 is the "autoreconf" tool available on alpine? 2015-08-24 21:48:22 oops found it 2015-08-25 07:07:44 good morning! 2015-08-25 07:18:40 mornin 2015-08-25 07:39:58 awww, there is no xinetd package on alpine? 2015-08-25 07:39:58 ncopa, there are few pending patches that I don't know what to do. E.g. %425 has issue with go. %452, %479, %500 and %501 i'd like your feedback 2015-08-25 07:40:44 tbh, i have no idea how to deal with go packages... 2015-08-25 07:40:55 %452 2015-08-25 07:41:35 keeping track of dependencies and mantain them is a huge work 2015-08-25 07:41:48 each dependency should be a different apkbuild 2015-08-25 07:42:03 rather than pick them via github 2015-08-25 07:42:12 yeah 2015-08-25 07:42:13 but a lot of them are without versioning 2015-08-25 07:42:23 so...it's a big mess 2015-08-25 07:44:52 opensuse relies on 0.0.0-GITSHASUM in that case :) 2015-08-25 07:47:15 mosez, we do a custom versioning in such case, called 0_git$YYMMDD 2015-08-25 07:47:25 sorry, 0_git$YYYYMMDD 2015-08-25 07:47:39 but still, mantain them is time-consuming 2015-08-25 07:47:57 since you need a snaphost and upload it on dev.alpinelinux.org 2015-08-25 07:48:09 so, would be better to keep them at minimum 2015-08-25 07:48:47 i suppose its time for us to add support for proper git in abuild 2015-08-25 07:49:01 to the fetch function 2015-08-25 07:49:07 ncopa, you mean to do it automatically? 2015-08-25 07:49:14 so you can do something like: source="git://...." 2015-08-25 07:49:15 yes 2015-08-25 07:49:27 iirc clandmeter started to implement something like that 2015-08-25 07:49:32 but we need a way to specify git repo, and a git tag 2015-08-25 07:50:01 but we need be caseful 2015-08-25 07:50:09 careful 2015-08-25 07:50:27 so we dont end up with a random git snap everywhere 2015-08-25 07:50:38 which will be a nightmare to maintain 2015-08-25 07:50:49 yeah 2015-08-25 07:50:53 the tendency i see is that upstream projects are lazy to maintain their stuff 2015-08-25 07:51:00 and push maintenance downstream 2015-08-25 07:51:08 for packagers 2015-08-25 07:51:27 or use tools that free's them from maintenance (eg ruby gems etc) 2015-08-25 07:51:55 so they can do the fun jobs only (new features development) 2015-08-25 07:52:22 then they tell end users: "use latest git you moron" 2015-08-25 07:52:40 i don't know i want encourage that 2015-08-25 07:53:17 in the end we might end up doing the release engineering for them 2015-08-25 07:53:25 and i defeinitively dont want that 2015-08-25 07:53:36 +1 2015-08-25 07:53:49 %479 is consider rejected right? 2015-08-25 07:54:11 Or "Change requested" ? 2015-08-25 07:54:23 According with your last email on ml, ncopa 2015-08-25 07:54:43 reject 2015-08-25 07:54:53 k 2015-08-25 07:54:54 but maybe wait for feedback frist 2015-08-25 07:55:17 give him a day or two in case he has wellfounded reasons to do it that way 2015-08-25 07:55:26 very good 2015-08-25 07:55:54 %503 is pushed directly to main 2015-08-25 07:56:00 i don't like that 2015-08-25 07:57:19 it ir a rename or copy of the existing go aport? 2015-08-25 07:58:00 source="git://...." how would that work with a http repo? 2015-08-25 07:58:25 it's new 2015-08-25 07:58:29 no copy or rename 2015-08-25 08:02:06 clandmeter: i dont know, we'd need some way to specify a git repo and tag 2015-08-25 08:02:13 or svn 2015-08-25 08:02:17 or mercurial 2015-08-25 08:02:31 or 2015-08-25 08:02:37 morning@all 2015-08-25 08:02:46 hi StarWarsFan 2015-08-25 08:02:52 ncopa: right, great that nfs problem is solved :-) 2015-08-25 08:03:09 StarWarsFan: I'm glad you solved it :) 2015-08-25 08:03:21 :) 2015-08-25 08:03:30 xonotic is huge 2015-08-25 08:03:55 and %505 does not apply 2015-08-25 08:04:09 another thing: after an upgrade from 2.7 to 3.2 i cannot do "su -" anymore 2015-08-25 08:04:18 ~ » su - 2015-08-25 08:04:23 su: must be suid to work properly 2015-08-25 08:06:52 seems that there's something wrong with the permissions of /usr/bin/su 2015-08-25 08:07:08 unfortunately i need to sit on the box to fix this... 2015-08-25 08:07:56 %505 does not apply because pkgrel does not match. Gotta wait fpr %479, and then ask to Stuart to rebase the patch 2015-08-25 08:08:43 hm, is it possible that there is something mismatching regarding the su-command? 2015-08-25 08:08:49 because of: 2015-08-25 08:09:38 ~ » ls -l /bin/su 2015-08-25 08:09:38 lrwxrwxrwx 1 root root 12 Dec 31 2013 /bin/su -> /bin/busybox 2015-08-25 08:09:38 ~ » ls -l /usr/bin/su 2015-08-25 08:09:38 lrwxrwxrwx 1 root root 11 Aug 23 19:16 /usr/bin/su -> /bin/bbsuid 2015-08-25 08:10:02 if i use the 2nd "su" directly, it works 2015-08-25 08:12:12 you dont have password for root? 2015-08-25 08:12:40 i think i sent a patch for it to upstream busybox 2015-08-25 08:13:00 the bbsuid is a wrapper so the entire busybox applet does not need to be suid root 2015-08-25 08:13:30 ncopa: would it effect performance if sqlite has SQLITE_ENABLE_UPDATE_DELETE_LIMIT enabled ? 2015-08-25 08:13:45 I asked upstream but not got reply yet 2015-08-25 08:14:54 its enabled in wheezy 2015-08-25 08:15:10 http://www.sqlite.org/lang_update.html 2015-08-25 08:15:19 dunno 2015-08-25 08:16:27 morning 2015-08-25 08:17:50 anyone have any thoughts on why packages in world don't get installed when booting a usb stick? 2015-08-25 08:18:08 ScrumpyJack, cache broken? 2015-08-25 08:18:13 Or not existing? 2015-08-25 08:22:39 fcolista: %503 is basically a rename of main/go 2015-08-25 08:22:47 that's why I pushed it to main directly 2015-08-25 08:22:58 and main/go needs that otherwise 1.5 doesn't build 2015-08-25 08:23:03 so you need to push that to main 2015-08-25 08:24:59 the alternative would be to have both go-bootstrap and go 1.5 in testing/ instead of main/ 2015-08-25 08:25:05 and keep go 1.4 in main/ for now 2015-08-25 08:25:26 I don't see any reason to do that but if you want to do it that way go ahead 2015-08-25 08:25:36 nmeum, did you git mv for the patch? 2015-08-25 08:25:52 no, because that would have removed the main/go package 2015-08-25 08:26:16 that's why i couldn't see a rename 2015-08-25 08:26:17 ok 2015-08-25 08:26:45 main/go-bootstrap is just main/go version 1.4 with some removals to make maintance simpler 2015-08-25 08:27:40 Ok nmeum, got it. From your comment on %504, other tests are needed. Do you have reasons to push it to main if you are not sure it work? 2015-08-25 08:28:56 yeah, I would keep it in edge for a while to test edge cases but it works…as I said: I only pushed it to main/ because it's a simple package upgrade…you can also move it to testing…i don't know what's better 2015-08-25 08:29:42 ok. Maybe ncopa would say something about that. 2015-08-25 08:29:48 yeah 2015-08-25 08:40:03 ncopa: of course there is a password for root 2015-08-25 08:47:13 shouldn't we be moving all go to different aports, before it starts being issue to main/ development ? 2015-08-25 09:05:01 ? 2015-08-25 10:44:07 fcolista: apk fix installs all the packages so i assume the cache exists. 2015-08-25 10:44:45 there is a /cache folder on the usb stick 2015-08-25 10:44:54 with all the packages 2015-08-25 10:45:16 perhaps it doesn't get mounted quickly enough on /media/usb/cache? 2015-08-25 10:52:33 hmm, what's the difference between /media/usb/cache and /media/usb/apks? 2015-08-25 10:56:04 what happened to man-doc ? 2015-08-25 10:58:49 ScrumpyJack: what if you add "apk fix" to /etc/local.d/apkfix.start file, chmod +x it 2015-08-25 11:01:50 error: unsatifiable contraints: man-doc missing 2015-08-25 11:27:22 hello masters 2015-08-25 11:28:02 i am searching for some small relatively maintained lua posix package 2015-08-25 11:28:13 i remberd alpine has posix module 2015-08-25 11:28:50 i would like to ask whether this module is normal lua posix or your custom / fork reimplementation? 2015-08-25 11:32:14 this one? https://github.com/luaposix/luaposix 2015-08-25 11:35:03 ScrumpyJack, please try apk cache sync 2015-08-25 11:36:09 ScrumpyJack: i see 2015-08-25 11:37:02 if it's that one, then yes, there is a maintained package, with a small patch http://git.alpinelinux.org/cgit/aports/tree/main/lua-posix/fix-sched-header.patch 2015-08-25 11:37:43 fcolista: tried that, i get ERROR $pkgname: IO ERROR after each fetch 2015-08-25 11:37:53 disk full? 2015-08-25 11:37:59 disk ro 2015-08-25 11:38:00 50% 2015-08-25 11:38:03 your usb has problem probably 2015-08-25 11:38:04 i can touch 2015-08-25 11:38:12 what does dmesg say? 2015-08-25 11:38:18 i'll check 2015-08-25 11:39:08 dmesg advises me to fun fsck on the usb stick :) 2015-08-25 11:39:22 so my stick is at fault it seems 2015-08-25 11:39:41 mine also says that but works just fine 2015-08-25 11:40:32 i'll run it anyway, can't hurt 2015-08-25 11:40:43 (I feel i'm going to regret saying that :) 2015-08-25 11:40:55 this message might appear both for usb physical issue that for a FS issue 2015-08-25 11:41:22 FS issue might be caused from a physical problem 2015-08-25 11:41:24 copy off what is important 2015-08-25 11:41:26 if that still works 2015-08-25 11:41:27 +1 2015-08-25 11:42:19 ScrumpyJack: well i am searching for standalone lua posix package i would use on freebsd and lua posix has quite deps so i wanted to make my life easier 2015-08-25 11:44:05 nmeum: awesome, looks great! 2015-08-25 12:04:52 dirty bit was set. no more errors in dmesg, but same symptoms when running apk cache clean or apk update/upgrade 2015-08-25 12:10:54 ok, it's getting better. i had a routing issue *blink* 2015-08-25 12:11:35 ScrumpyJack: are you sure you set the cache symlink correctly? 2015-08-25 12:13:14 apk cache sync is working 2015-08-25 12:13:20 where is the symlink? 2015-08-25 12:17:43 after apk cache sync, i get http://st.ilet.to/contr 2015-08-25 12:30:05 i think itshould be in /etc/apk 2015-08-25 12:30:12 its complaining about missing deps 2015-08-25 12:30:20 do you have proper repo's added? 2015-08-25 12:30:31 apk update could help too 2015-08-25 12:54:47 is there already a way to simply install the golang 1.4 stuff? 2015-08-25 12:55:36 where can i find instructions how to build an run a custom package index? :) 2015-08-25 12:56:34 on the wiki 2015-08-25 12:56:57 this might get you started http://wiki.alpinelinux.org/wiki/Abuild_tree 2015-08-25 12:57:23 yes, i have a symlink in /etc/apk/cache to /media/usb/cache 2015-08-25 13:04:49 nmeum: thanls 2015-08-25 13:04:53 thanks :) 2015-08-25 13:06:21 maybe i can use that for packaging applications as well... we will see 2015-08-25 13:46:13 mosez: yup, that's the idea :) 2015-08-25 13:46:48 mosez: in case you missed it http://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package 2015-08-25 13:58:48 scrumpyjack: i already read that... just the own package index thing was missing :) 2015-08-25 14:51:02 ah i see, so what did you do? 2015-08-25 14:55:20 postfix-ldap-3.0.1-r2 is missing /usr/lib/postfix/dict_ldap.so file 2015-08-25 14:59:09 vkris: won't that get run at each login? 2015-08-25 15:53:46 anyone know the argument types for /lib/rc/bin/rc-depend - with "-t" switch ? 2015-08-25 16:01:03 ncopa 2015-08-25 16:01:32 postfix-ldap-3.0.1-r2 is missing /usr/lib/postfix/dict_ldap.so file 2015-08-25 16:17:29 ScrumpyJack: yes, but should not be of any issue 2015-08-25 16:18:15 would be best to ask fabled if it would be safe 2015-08-25 16:18:49 I did same on my test machine here and it solved similar issue 2015-08-25 16:26:11 also see /etc/local.d/README as when and how it runs 2015-08-25 19:03:45 ncopa: what's the status on the man page compression symlink issue? 2015-08-26 03:04:26 firefox seems to be missing some symbols http://hastebin.com/talonoziba.md 2015-08-26 07:06:50 good morning alpine! 2015-08-26 07:40:53 morning :) 2015-08-26 08:27:04 i think /lib/ld-musl-x86_64.so.1 is leaking memory 2015-08-26 08:28:16 https://paste.debian.net/plain/305019 2015-08-26 08:29:20 leo-unglaub, no leaks in that 2015-08-26 08:29:20 ==2432== LEAK SUMMARY: 2015-08-26 08:29:20 ==2432== definitely lost: 0 bytes in 0 blocks 2015-08-26 08:29:20 ==2432== indirectly lost: 0 bytes in 0 blocks 2015-08-26 08:29:29 what's the issue? 2015-08-26 08:30:00 ==2432== ERROR SUMMARY: 21 errors from 5 contexts (suppressed: 0 from 0) 2015-08-26 08:30:06 those are not leaks 2015-08-26 08:30:11 but 2015-08-26 08:30:13 4,791 are still reachable 2015-08-26 08:30:25 the first error about invalid free is safe to ignore 2015-08-26 08:31:01 why can the first error be ignored? 2015-08-26 08:31:03 not sure about the other errors, they are likely application errors 2015-08-26 08:31:16 leo-unglaub, the first error is related to how musl allocator works internally 2015-08-26 08:31:26 but valgrind overrides musl free, and sees it as error 2015-08-26 08:31:33 ah, i see 2015-08-26 08:31:42 so maybe the other errors are just followups? 2015-08-26 08:31:48 because every program i try has the same errors 2015-08-26 08:31:55 literally everyone 2015-08-26 08:32:01 from geany to firefox to xfce, .. 2015-08-26 08:32:13 leo-unglaub, still-reachable memory is when application is not free():ing all memory 2015-08-26 08:32:37 yes, the 1st free() comes from musl ld.so implementation and happens with all dynamic linked executables 2015-08-26 08:34:25 leo-unglaub, epoll_pwait call comes from libisc (bind). but the error is valgrind error; sigmask is explicitly allowed to benull 2015-08-26 08:35:18 socketcall.bind error sounds like it's application error, since the memory was originally calloc()ed to zero. so the application must've overwritten the field with unknown bytes. 2015-08-26 08:49:16 hmm, mabe its an application error 2015-08-26 08:49:25 but then all tested applications must have the same error 2015-08-26 09:13:58 ncopa: you are working on buildlogs? 2015-08-26 09:15:00 no 2015-08-26 09:15:14 i have got 3 other emergencies to deal with first 2015-08-26 09:15:30 there is an empty buildlogs dir on root of mirror 2015-08-26 09:15:56 i decided to put the buildlogs on bld2.a.o for now 2015-08-26 09:15:59 and share it via nfs 2015-08-26 09:16:14 i think the buildlogs dir on mirror can be deleted 2015-08-26 09:17:36 ok, you will have to do that, cuase if i do it will return :) 2015-08-26 09:39:35 ncopa: sqlite upstream said to use main sqlite-src.zip if SQLITE_ENABLE_UPDATE_DELETE_LIMIT is needed 2015-08-26 09:40:25 I tried and it creates binary 100+bytes more, would there be interest to add it 2015-08-26 09:41:02 clandmeter: ^^ 2015-08-26 09:41:11 i dont have any specific interest in it 2015-08-26 09:41:15 will it break abi compat? 2015-08-26 09:41:22 once we enable it we can not disable it 2015-08-26 09:41:38 what apps needs it? 2015-08-26 09:42:06 any application that needs LIMIT in UPDATE statement 2015-08-26 09:42:53 update tb1 set value=3 where abc="abc" limit 1' 2015-08-26 09:43:45 though after seeing current defaults, I removed the limit from my applications 2015-08-26 09:44:05 fearing other distros might not enable it 2015-08-26 09:45:15 usually if you have your tables with unique ids and use that in where clause, one does not need it 2015-08-26 09:45:55 vkris: limit is still very common on it to make sure the where is working properly and the table does not get dammaged 2015-08-26 09:46:01 i have seen this use a lot 2015-08-26 09:46:34 yes, I kinda caught that habbit 2015-08-26 09:46:54 but now think if its is really necessary 2015-08-26 09:47:13 if you write good code its not necesarry 2015-08-26 09:47:33 but you know ;) mistakes happen *G* 2015-08-26 09:47:59 sqlite still would not allow that in triggers 2015-08-26 09:48:59 but issues are if usually does add limit, they might not work across distros 2015-08-26 09:49:10 if usually one * 2015-08-26 09:51:54 and on top sqlite show "Error: near "limit": syntax error 2015-08-26 09:52:26 only, which may lead to developers thinking something is wrong in there statement 2015-08-26 09:52:38 their* 2015-08-26 09:55:01 does other distros enable it? 2015-08-26 09:55:50 was enabled on knoppix 2015-08-26 09:56:05 would gradually check.. as and when I test others 2015-08-26 09:56:14 then debian and ubuntu probably do 2015-08-26 09:56:46 if debian/ubuntu/fedora do then i think we should too 2015-08-26 09:56:59 unless if it causes major pain 2015-08-26 09:57:24 due to need use other build system or other tarball or whatever 2015-08-26 09:58:51 here what I dit 2015-08-26 09:59:06 I used the export statments in main/sqlite/APKBUILD but added DSQLITE_ENABLE_UPDATE_DELETE_LIMIT 2015-08-26 09:59:09 1. 2015-08-26 09:59:11 export LTLINK_EXTRAS="-ldl" 2015-08-26 09:59:12 https://build.opensuse.org/package/view_file/server:database/sqlite3/sqlite3.spec?expand=1 :) 2015-08-26 09:59:16 export CFLAGS="$CFLAGS -DSQLITE_ENABLE_UPDATE_DELETE_LIMIT -DSQLITE_ENABLE_FTS3=1 -DSQLITE_ENABLE_COLUMN_METADATA=1 -DSQLITE_ENABLE_FTS3_PARENTHESIS -DSQLITE_SECURE_DELETE -DSQLITE_ENABLE_UNLOCK_NOTIFY -DSQLITE_ENABLE_RTREE=1 -DSQLITE_USE_URI -Iext/fts3" 2015-08-26 09:59:21 2. 2015-08-26 09:59:22 ../sqlite-src-3081101/configure 2015-08-26 09:59:28 3. 2015-08-26 09:59:29 make amalgamation-tarball 2015-08-26 10:00:07 mkdir bld && ../sqlite-src-3081101/configure 2015-08-26 10:00:14 so suse does not add it 2015-08-26 10:00:57 I regen amalgamation-tarball and use it to build sqlite3 2015-08-26 10:01:42 ncopa: yeah, looks like that 2015-08-26 10:02:40 i can try updates on patchwork only with building the package manually, right? 2015-08-26 10:03:09 http://pkgs.fedoraproject.org/cgit/sqlite.git/tree/sqlite.spec#n126 2015-08-26 10:03:22 mosez: yes 2015-08-26 10:03:30 seems like fedora does not enable it either 2015-08-26 10:03:38 vkris: i think we dont enable it 2015-08-26 10:04:09 you app wil not work on suse or fedora 2015-08-26 10:04:21 ok, I would meanwhile check when was SQLITE_ENABLE_UPDATE_DELETE_LIMIT add to src tree 2015-08-26 10:04:24 added* 2015-08-26 10:05:20 knoppix has, 3.7.13 2015-08-26 10:08:14 ncopa: sad :( 2015-08-26 10:08:26 currently i have no build environment 2015-08-26 10:09:36 vcs has an log entry - 2008-10-10 2015-08-26 10:14:25 ahem, http://pkgs.fedoraproject.org has git for every ports and that too without src tree in it 2015-08-26 10:15:04 every pkg* 2015-08-26 10:21:44 ncopa: any plans to merge go 1.5 to edge? 8) 2015-08-26 10:31:46 mosez: if you need it right now you can compile it yourself otherwise you just have to wait a few days until ncopa has time to look at it :) 2015-08-26 10:44:17 awww -.- 2015-08-26 10:45:07 i would be happy if that gets done soon... me and the author of drone ci wants to build docker containers for drone on alpine. and we need 1.5 for that as it relies on 1.5 -.- 2015-08-26 10:45:32 but anyway, than i have to get back to it next week or something like that 2015-08-26 10:45:57 as I said: compile it yourself it isn't that hard 2015-08-26 10:47:13 It boils down to installing abuild, cloning the aports git repo, applying the patch(es) and building the thing 2015-08-26 10:47:16 that's it 2015-08-26 10:53:16 nmeum: sure, for local testing this is fine... but for automated build process it sucks pretty hard :) 2015-08-26 10:54:34 well…no…you can create your own repo and add it to /etc/apk/repositories 2015-08-26 10:54:39 that's what I do 2015-08-26 11:00:56 true 2015-08-26 11:05:56 i will likley not have time to look at go1.5 today :-( 2015-08-26 11:06:27 haha, pidgin is leaking as well: https://paste.debian.net/plain/305173 2015-08-26 13:32:29 hmmm, my firefox keeps crashing :( 2015-08-26 13:32:38 something is fighting there with grsec 2015-08-26 14:50:52 i'm not sure it is grsec that triggers the firefox crashing 2015-08-26 15:29:35 ncopa: i think it is 2015-08-26 15:29:47 because every time it crashes i get a grsec message 2015-08-26 15:30:30 https://paste.debian.net/plain/305495 2015-08-26 17:35:30 eh, you get this message because it wants to coredump 2015-08-26 22:23:36 https://grsecurity.net/announce.php for those not following 2015-08-26 22:46:46 hm, that sucks balls 2015-08-26 23:08:18 ncopa: out of curiosity: do you backport the grsec patches from stable to 3.14 or from testing? 2015-08-27 00:20:39 i don't really think grsecurity is a big component of alpine's security strategy 2015-08-27 00:21:01 most features are disabled 2015-08-27 00:21:09 we mainly only use it to get PaX 2015-08-27 00:21:37 i think, generally other components are more important 2015-08-27 00:21:46 such as robustness in the core libraries 2015-08-27 00:23:03 and PIE 2015-08-27 00:23:36 the bottom line is that we've basically forked grsec anyway 2015-08-27 00:24:00 so it really changes nothing 2015-08-27 00:29:11 also, i think in the real world, an alpine box running linux-vanilla is pretty much as secure as an alpine box running linux-grsec except against an advanced persistant threat 2015-08-27 00:29:31 in which case all the cloak-and-dagger type shit is disabled in our grsec builds anyway because we want people to be able to use the OS 2015-08-27 00:57:17 hi everyone, I was wondering if someone could clarify something for me regarding http://git.alpinelinux.org/cgit/aports/tree/main/musl/ldconfig 2015-08-27 00:57:43 I'm packaging musl as system libc for the distribution I help develop and I was curious as to what the license for that file is 2015-08-27 00:57:51 is it the same as musl itself? 2015-08-27 00:57:54 yes 2015-08-27 00:58:23 okay then, thank you :) 2015-08-27 03:59:03 https://grsecurity.net/announce.php sigh 2015-08-27 03:59:49 I see coredumb already posted it here 2015-08-27 06:09:45 nmeum: we track changes in 3.14 kernels yes and port to our 3.18 kernel 2015-08-27 06:11:33 well gonna have to become a sponspor for that now :( 2015-08-27 07:18:05 ncopa: so are you planning to become a sponsor now? 2015-08-27 07:44:57 good morning from the alps to alpine :) 2015-08-27 07:48:18 morning leo-unglaub 2015-08-27 07:48:35 nmeum: i wish, but alpine users does not pay me that well 2015-08-27 07:49:09 yeah, I thought so 2015-08-27 07:49:54 do you want to use test branch instead than or drop support for it altogether? 2015-08-27 07:50:12 ncopa: oh, users dont donate to an open source project? 2015-08-27 07:50:17 thats shocling ;) 2015-08-27 07:54:01 ncopa: do you have donation informations on the website? 2015-08-27 07:54:07 i could not find them 2015-08-27 07:58:41 we dont 2015-08-27 07:59:30 a friend of mine tryed it to live of his open source work ... he got 50€ in one year in donations ... 2015-08-27 10:39:01 leo-unglaub: what was the "open source work" that your friend tried to live off? 2015-08-27 10:41:56 ScrumpyJack: he created the newsletter tool "avisota" 2015-08-27 10:58:31 seemns german only 2015-08-27 13:35:02 grsecurity "Sponsorship begins at 200 USD/mo" 2015-08-27 13:35:29 ACTION doubts we would receive that much if we were to open a way for users to donate to alpinelinux :( 2015-08-27 13:39:28 alternativly we could also stop using linux-lts and switch to stable, then we could use gentoo-hardened patches or grsecurity testing 2015-08-27 13:47:54 hi 2015-08-27 13:48:07 does alpine also have a real-time kernel? 2015-08-27 13:56:48 I think we used to have one 2015-08-27 13:56:50 check testing 2015-08-27 14:01:46 we only have the grsec, vanilla and rpi kernels 2015-08-27 14:02:05 and one vritgrsec for VMs 2015-08-27 14:02:09 virtualgrsec 2015-08-27 14:02:13 virtgrsec 2015-08-27 14:02:20 i looking at the firefox issue 2015-08-27 14:02:25 there are 2 workarounds 2015-08-27 14:02:42 set layers.offmainthreadcomposition.enabled to False 2015-08-27 14:02:52 or build with bundled cairo 2015-08-27 14:13:31 ncopa: what did they change in their bundeled cairo version? 2015-08-27 14:13:45 maybe we could port that change to our system cairo version? 2015-08-27 14:13:54 thats what i am looking at 2015-08-27 14:14:01 they seem to have alot of fixes 2015-08-27 14:14:22 $ ls *.patch | wc -l 2015-08-27 14:14:22 116 2015-08-27 14:14:26 116 to be exact 2015-08-27 14:14:27 oh god 2015-08-27 14:14:33 also for pixman 2015-08-27 14:14:57 many are prefixed with win32- and quartz- 2015-08-27 14:15:05 if they have 116 patches then we might be better of by using the bundeled version 2015-08-27 14:15:51 I don't understand why mozilla doesn't try to get those patches included in the upstream cairo version 2015-08-27 14:17:20 they do 2015-08-27 14:17:24 they try 2015-08-27 14:17:31 like this https://bugs.freedesktop.org/show_bug.cgi?id=21706 2015-08-27 14:17:40 i found that reference in one of the patches 2015-08-27 14:19:49 oh looks like freebsd also have the issue 2015-08-27 14:19:57 http://lists.freebsd.org/pipermail/freebsd-gecko/2015-August/005616.html 2015-08-27 14:36:31 tdtrask: that wouldn't solve the problem 2015-08-27 14:37:06 cause being a sponsor doesn't allow you to distribute the patch yourself 2015-08-27 14:37:20 else this would defeat the point 2015-08-27 14:37:33 then it wouldn't solve the problem for Alpine 2015-08-27 14:37:38 or arch 2015-08-27 14:37:40 or gentoo 2015-08-27 14:38:09 or they'd have to provide us users with a binary kernel only 2015-08-27 14:38:23 which would indeed be another problem 2015-08-27 14:38:51 sorry I'm getting on my irc log I wasn't up the whole morning :) 2015-08-27 14:39:07 > cause being a sponsor doesn't allow you to distribute the patch yourself 2015-08-27 14:39:19 I don't think they can prevented that since it's licensed under GPL2 2015-08-27 14:39:30 and the GPL2 allows you to distribute copies 2015-08-27 14:39:55 yeah but they can remove you as a sponsor 2015-08-27 14:40:49 then, what's the point of being open source / using GPL2? 2015-08-27 14:41:31 well…but If the sponsors use the stable patches on an embedded system (e.g a router) and I buy that system than I can request the source code for the kernel from the sponsor 2015-08-27 14:41:44 the point of requiring sponsership is to force companies to contribute if they want the latest fixes 2015-08-27 14:41:49 including the grsec stable patches 2015-08-27 14:42:19 ACTION doubts they will drop you as sponsor for modifying the code 2015-08-27 14:42:25 if you're still paying 2015-08-27 14:42:58 tdtrask: well they stop accepting your money 2015-08-27 14:43:30 nmeum: well you're allowed to request the sources 2015-08-27 14:43:37 but we know how this works 2015-08-27 14:43:51 well…they are required to give me access to the source 2015-08-27 14:44:00 most of the time this goes nowhere as they don't respect the GPL anyway 2015-08-27 14:44:03 of cause I would have to sue them and it would take ages until I see it but theoretically… 2015-08-27 14:44:11 indeed 2015-08-27 14:45:44 Harald Welte (the guy who wrote iptables) actually founded a site called GPL-violations.org which basically sued everyone who was using a modified version of the linux kernel but didn't publishe the sources 2015-08-27 14:45:53 they had some success but the site dosen't seem to exist anymore 2015-08-27 14:46:12 https://en.wikipedia.org/wiki/Gpl-violations 2015-08-27 14:47:40 yep know it 2015-08-27 14:47:56 but still companies keep violating the GPL 2015-08-27 14:49:46 yeah, but the point I am trying to make is that it is pretty useless to only give sponsors access to grsecurity stable patches since the would be required to publish those patches when distributing copies of their kernel to end users 2015-08-27 14:49:59 *they 2015-08-27 16:37:07 nmeum: it's especially totally useless because it's not like people who care about gpl violations and people who care about being totally updated necessarily overlap much 2015-08-27 19:47:42 i think it's really better to just openly ponder what purpose grsec really has in alpine :) 2015-08-27 19:48:06 the only really compelling feature there is PaX 2015-08-27 20:21:46 at this point anyway 2015-08-27 20:53:34 will PaX patches continue to be available? 2015-08-27 21:06:07 not sure 2015-08-27 21:06:15 it is probably forkable at this point 2015-08-27 21:09:44 in any case, I fail to see how restricting access to stable patches would address the problem they are facing 2015-08-27 23:27:11 kunkku: indeed, it really doesn't. i think they just want more money. 2015-08-27 23:27:35 on top of that, i don't think they really like us evry much anyway since we do not ship "official" builds of their patches 2015-08-28 06:22:03 ncopa, seems there's again cyclic dependency. this time between gtk+2.0 and librsvg 2015-08-28 06:22:13 i'm trying to get arm edge builder into better shape 2015-08-28 06:22:23 seems there's huge bunch of packages failing to build 2015-08-28 06:23:18 guile fails due to gcc bug it seems. gcc has similar https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=796855 2015-08-28 06:23:40 and llvm fails too 2015-08-28 06:23:45 fabled: i'll look at gtk/librsvg 2015-08-28 06:23:48 thanks 2015-08-28 06:24:35 fabled: i have a workaround for firefox craches 2015-08-28 06:24:40 cool 2015-08-28 06:24:45 what's the issue? 2015-08-28 06:25:53 system cairo + OMTC 2015-08-28 06:26:04 either use bundled cairo or disable OMTC 2015-08-28 06:26:23 this freebsd patch explains how: http://lists.freebsd.org/pipermail/freebsd-gecko/2015-August/005616.html 2015-08-28 06:26:39 so you can go to about:config 2015-08-28 06:26:44 and set layers.offmainthreadcomposition.enabled to False 2015-08-28 06:28:15 ok 2015-08-28 06:28:17 i'll test that 2015-08-28 06:28:18 thanks 2015-08-28 07:03:28 ncopa, there's upstream commit that fixes the issue our fix-memory-map-for-PIE-applications.patch fixed 2015-08-28 07:03:33 but due to different reasons 2015-08-28 07:03:59 upstream commit "mm: split ET_DYN ASLR from mmap ASLR" 2015-08-28 07:04:27 http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=d1fd836dcf00d2028c700c7e44d2c23404062c90 2015-08-28 07:14:53 i'm rebooting arm-edge builder 2015-08-28 07:14:55 to new kernel 2015-08-28 07:26:56 ncopa, mmc is not included in initramfs features by default? 2015-08-28 07:29:57 seems not 2015-08-28 07:32:15 that broke builder 2015-08-28 07:32:15 this is probably better default: http://tpaste.us/24y1 2015-08-28 07:32:36 yeah 2015-08-28 07:32:56 remove bootchart and floppy 2015-08-28 07:32:57 add mmc 2015-08-28 07:33:01 looks good 2015-08-28 07:33:16 i dontk onw if we should remove ext2 and ext3 2015-08-28 07:33:21 i suppose they are okish to keep 2015-08-28 07:35:08 i saw an interesting bootoption for netbsd: resize_root 2015-08-28 07:35:44 if you boot it with resize_root=YES then will it try resizefs before remounting root as rw 2015-08-28 07:36:24 it means that you can dd an image and boot it with resize_root 2015-08-28 07:36:31 too bad vfat does not support resizing 2015-08-28 07:38:16 do we add mmc to 3.2-stable too? 2015-08-28 07:38:29 i dont want remove floppy though 2015-08-28 07:45:51 ncopa, please add 2015-08-28 07:46:06 i'd like to run 3.2 unchanged on builders 2015-08-28 07:46:15 mmc is practically needed on all arm boards 2015-08-28 07:52:26 fabled: i fixed the abuild fetch to nfs race the other day 2015-08-28 07:52:35 i ended up with writing an abuild-fetch.c 2015-08-28 07:52:44 which is the only way to make it work 2015-08-28 07:53:09 ok. nice. 2015-08-28 07:53:30 the annoying thing is that if we had a dedicated nfs server (which did not run any builder) it whoudl have just worked 2015-08-28 07:53:55 but since we run builders on the nfs server, it didnt work 2015-08-28 07:53:58 annoying 2015-08-28 07:54:04 where the .c code is? 2015-08-28 07:54:40 oh, it's only as patch against abuild in aports 2015-08-28 07:54:43 not in abuild.git 2015-08-28 07:55:13 i just had not pushed it 2015-08-28 07:55:14 i just did 2015-08-28 07:55:32 http://git.alpinelinux.org/cgit/abuild/tree/abuild-fetch.c 2015-08-28 07:55:59 i also moved the wget fallback and saveas- and $archive:: logic there 2015-08-28 08:01:38 i think i have fix for guile (upstream commits from debian) 2015-08-28 08:01:54 nice 2015-08-28 08:55:35 ncopa: what about that go 1.5 upgrade will you have to time to look at it before the weekend? 2015-08-28 08:58:32 :D 2015-08-28 08:59:39 i wonder what's the llvm/armhf issue 2015-08-28 09:00:01 it passes initial bootstrap build, but the resulting compiler crashes when doing self-compile 2015-08-28 09:00:14 lunch 2015-08-28 09:00:44 hi 2015-08-28 09:00:52 does this affect us? 2015-08-28 09:00:52 https://grsecurity.net/announce.php 2015-08-28 09:03:02 shafire: yes. it affects us 2015-08-28 09:03:27 nmeum: i will try do go 1.5 before weekend 2015-08-28 09:03:33 was this already in discussion? 2015-08-28 09:03:41 not really 2015-08-28 09:03:51 well 2015-08-28 09:03:58 there has been some blabla 2015-08-28 09:04:05 but no conclusions 2015-08-28 09:05:34 maybe one of the sponsors will make it available 2015-08-28 09:06:23 there are basically three options 1) Backport patches from grsecurity testing to 3.14 2) Switch to grsecurity testing (arch and gentoo use it) 3) drop grsecurity and use vanilla linux instead 2015-08-28 09:07:58 approx that yes 2015-08-28 09:08:24 we can 1) backport testing patches to the stable kernel we use (kind of what we do already) 2015-08-28 09:09:02 2) backport pax patch only instead of full grsecurity stuff 2015-08-28 09:09:13 3) use plain vanilla 2015-08-28 09:09:37 4) ask spender for stable patches as a non-profit org. 2015-08-28 09:14:36 5) be a sponsor (200$/month) 2015-08-28 09:15:04 ncopa: why don't you consider using the testing patches? 2015-08-28 09:17:30 because we need maintain stable branches 2015-08-28 09:17:45 we cannot run latest kernel in stable branches 2015-08-28 09:18:29 or more precise, we dont do feature updates in stable branches, only bugfix/secfixes 2015-08-28 09:18:51 so we will not push the 4.x kernel to 2.7-stable 2015-08-28 09:18:57 nor to 3.2-stable 2015-08-28 09:19:49 nmeum: i got checksum error on go 1.5 2015-08-28 09:19:55 tools.tar.gz: FAILED 2015-08-28 09:20:41 maybe some other package has tools.tar.gz? 2015-08-28 09:21:14 shafire: sponsor patch redistribution seems to be an issue right now 2015-08-28 09:22:07 sponsors who redistribute the patched binary need to also redistribute the patch 2015-08-28 09:22:33 and sponsors can not be forbidden to not redistribute the patch 2015-08-28 09:23:08 you cannot say: here you have this GPL'ed patch, but you are not allowed to redistribute it 2015-08-28 09:24:05 well, first they need to ask for non-binary patch 2015-08-28 09:24:10 doesn't mean they do 2015-08-28 09:27:42 and how many sponsor are actually distributing a binary 2015-08-28 09:27:55 most of the sponsor are web hosting companies 2015-08-28 09:40:09 https://forums.grsecurity.net/viewtopic.php?f=3&t=3938&p=13940#p13940 2015-08-28 09:45:08 nmeum: go-bootstrap failed on x86: http://bld1.alpinelinux.org/buildlogs/build-edge-x86/main/go-bootstrap/go-bootstrap-1.4.2-r0.log 2015-08-28 09:54:20 oh 2015-08-28 09:54:26 let me take a look at that 2015-08-28 10:01:38 ncopa: I don't have an x86 computer…could you test if the old go 1.4 package still builds under x86? Because the build function from go 1.4 and go-bootstrap is actually almost identical 2015-08-28 10:03:47 container? 2015-08-28 10:04:09 lxc-create -n alpine32 -t alpine -- --arch x86 2015-08-28 10:04:14 yeah, I will do that 2015-08-28 10:18:27 ncopa: there seems to be something wrong with nginx logging. looks like it just randomly stops logging and continiu after a restart. 2015-08-28 10:19:17 ncopa: somebody complained our www was down (via twitter) but i cant verify it via logs. 2015-08-28 10:20:38 having to pay $200/month for source code access kind of defeats the purpose of "open source" 2015-08-28 10:22:22 actually it doesn't 2015-08-28 10:22:30 there is a difference between zero cost and free software 2015-08-28 10:30:17 well yes 2015-08-28 10:33:03 ncopa: I just tested it and our old main/go 1.4.2 aport doesn't built under x86 either anymore 2015-08-28 10:33:42 since it built sucessfully at some point in time I am assuming that this might be caused by an upgrade of a related package? 2015-08-28 10:36:24 grsecurity as an open source project is somewhat exceptional 2015-08-28 10:36:30 https://grsecurity.net/contribute.php 2015-08-28 10:36:44 usually the "Contribute" page explains how to send patches 2015-08-28 10:36:53 they tell how to send $$$ 2015-08-28 10:38:18 https://grsecurity.net/spender.mp3 <- lol 2015-08-28 12:18:27 any ideas why "abuild -r" is not working for a non-root? 2015-08-28 12:18:47 i've setup the account as described on http://wiki.alpinelinux.org/wiki/Include:Setup_your_system_and_account_for_building_packages 2015-08-28 12:20:10 additional information: the machine was upgraded from 2.7 to edge 2015-08-28 12:27:31 StarWarsFan, what is the error that is beign prompted ? 2015-08-28 12:27:42 nothing 2015-08-28 12:28:17 :| 2015-08-28 12:28:19 thats the problem 2015-08-28 12:28:49 StarWarsFan: i think you need either sudo or be in 'abuild' group 2015-08-28 12:29:04 sudo is not possible 2015-08-28 12:29:10 sudo abuild -r 2015-08-28 12:29:10 >>> ERROR: : Do not run abuild as root 2015-08-28 12:29:27 and the user is in the abuild group 2015-08-28 12:30:19 StarWarsFan, did you log in again after add your user to the abuild group? 2015-08-28 12:30:35 sure, rebootet the whole machine for other reasons... 2015-08-28 12:31:59 abuild -r 2015-08-28 12:31:59 >>> cui-samba: Checking sanity of /home/jenkins/ci/workspace/eisfair-ng/edge/testing/x86_64/cui-samba/cui-samba/APKBUILD... 2015-08-28 12:31:59 >>> cui-samba: Analyzing dependencies... 2015-08-28 12:31:59 >>> ERROR: cui-samba: all failed 2015-08-28 12:32:19 without the -r option, it works 2015-08-28 12:32:40 but that's not usable on an CI system... 2015-08-28 12:32:42 apk update ? 2015-08-28 12:33:03 no change 2015-08-28 12:50:30 StarWarsFan: can you do: abuild-apk add 2015-08-28 12:51:30 trying... 2015-08-28 12:55:15 ncopa: yes, that works 2015-08-28 12:56:57 can you do: ( . ./APKBUILD; abuild-apk add $source 2015-08-28 12:56:58 ) 2015-08-28 12:57:15 sorry 2015-08-28 12:57:27 can you do: ( . ./APKBUILD; abuild-apk add $makedepends ) 2015-08-28 12:58:22 http://sprunge.us/BbYj 2015-08-28 12:59:15 all the jobs here: 2015-08-28 12:59:16 https://ssl.nettworks.org/ci/job/eisfair-ng/job/edge/job/testing/job/x86_64/ 2015-08-28 13:33:22 it tries to install cui-samba 2015-08-28 13:33:28 when building it 2015-08-28 13:33:35 is it a circular dep? 2015-08-28 13:33:47 does something in makedepends depend on cui-samba? 2015-08-28 13:34:13 alternatively: apk del cui-samba 2015-08-28 13:34:14 that's only an example 2015-08-28 13:34:19 every job fails that way 2015-08-28 13:34:33 but i verify this right now... 2015-08-28 13:35:42 no, there is no circular dep 2015-08-28 13:36:16 <@ncopa> can you do: ( . ./APKBUILD; abuild-apk add $makedepends ) 2015-08-28 13:36:16 http://sprunge.us/BbYj 2015-08-28 13:36:34 yes i know 2015-08-28 13:36:38 but no idea why :-/ 2015-08-28 13:36:39 was that the cui-samba package you . ./APKBUILD ? 2015-08-28 13:36:50 right 2015-08-28 13:36:56 what happens if you do: apk fix 2015-08-28 13:37:09 as the user or as root? 2015-08-28 13:37:18 as user: abuild-apk fix 2015-08-28 13:37:44 abuild-apk fix 2015-08-28 13:37:44 (1/1) Reinstalling musl-dev (1.1.10-r4) 2015-08-28 13:37:44 ERROR: musl-dev-1.1.10-r4: trying to overwrite usr/include/iconv.h owned by libiconv-dev-1.12-r8. 2015-08-28 13:37:44 1 errors; 1644 MiB in 512 packages 2015-08-28 13:38:13 is this v2.7? 2015-08-28 13:38:31 it was 2.7. upgraded the box to edge last week 2015-08-28 13:38:38 try apk upgrade -U -a then 2015-08-28 13:38:50 as root? 2015-08-28 13:38:54 yes 2015-08-28 13:38:55 actually 2015-08-28 13:39:05 check if you have uclibc leftovers 2015-08-28 13:39:14 ls /lib/libc.0.9.32* something 2015-08-28 13:39:59 http://sprunge.us/CgKD 2015-08-28 13:40:19 somethign is holding back uclibc 2015-08-28 13:40:37 then i'd recommend: apk add apk-tools-static 2015-08-28 13:41:01 apk.static upgrade --no-selfupgrade -U -a 2015-08-28 13:41:19 then try remove the uclibc leftovers: apk del uclibc 2015-08-28 13:41:25 root@builder07:~ apk.static upgrade --no-selfupgrade -U -a 2015-08-28 13:41:25 apk.static: unrecognized option: no-selfupgrade 2015-08-28 13:42:45 --no-self-upgrade 2015-08-28 13:43:20 you probably have packages installed that depends on uclibc, but has no musl built variant 2015-08-28 13:43:31 you have to identify those 2015-08-28 13:43:38 i c 2015-08-28 13:43:39 and remove them and remove the uclibc 2015-08-28 13:43:44 ah ok 2015-08-28 13:43:53 apk version -l '?' 2015-08-28 13:43:58 might give indication 2015-08-28 13:44:17 http://sprunge.us/UVhj 2015-08-28 13:45:04 mysql could be removed for sure 2015-08-28 13:46:27 mysql is replaced with mariadb 2015-08-28 13:46:44 right 2015-08-28 13:46:58 libiconv-dev-1.12-r8 2015-08-28 13:47:04 remove that 2015-08-28 13:47:17 what is greyfix? 2015-08-28 13:47:53 it needs to be rebuilt probably 2015-08-28 13:48:21 i c 2015-08-28 13:48:26 or you might be able to replace it with gross 2015-08-28 13:48:35 which also does greylisting 2015-08-28 13:48:53 can't say that, another developers package... 2015-08-28 13:54:48 looks better now! 2015-08-28 14:00:37 no checking what's missing... 2015-08-28 14:01:06 by the way, sometimes at the end of an apk step, 2015-08-28 14:01:12 i see this: 2015-08-28 14:01:13 Executing mdocml-1.13.3-r4.trigger 2015-08-28 14:01:19 which takes some minutes 2015-08-28 14:01:27 what's happening in that time? 2015-08-28 14:16:27 it is building index in sqlite db 2015-08-28 14:16:29 for apropos 2015-08-28 14:16:35 and whatis 2015-08-28 14:17:10 ah ok, thx 2015-08-28 14:17:42 does anyone have spare time to help me figure out why our grsecurity patch is so much bigger than the official grsecurity patches? 2015-08-28 14:17:54 ncopa: I still don't understand why the 3.18.20 patch is 6.1 MB 2015-08-28 14:17:54 when 3.18.9 was 4.2MB 2015-08-28 14:32:48 sry, can't help you with that :-/ 2015-08-28 14:55:44 whats this /usr/lib/libparted.so: undefined reference to `dm_task_get_info@Base' 2015-08-28 14:56:16 did lvm2 break abi? 2015-08-28 17:00:17 is there a way to do malloc_trim(0) in musl ? 2015-08-29 20:53:00 is anybody working on an openssh update? 2015-08-29 22:41:18 wait where do I actually subscribe to the mailing list 2015-08-29 22:41:29 ... or, uh, otherwise reply to things 2015-08-29 22:43:12 http://git.alpinelinux.org/cgit/aports/tree/testing/the_server_searcher anyway this got ... misnamed 2015-08-29 22:43:16 the directory did 2015-08-29 22:55:46 fixed 2015-08-29 23:08:33 05:25 having to pay $200/month for source code access kind of defeats the purpose of "open source" 2015-08-29 23:08:35 05:26 actually it doesn't 2015-08-29 23:08:37 05:27 there is a difference between zero cost and free software 2015-08-29 23:08:39 true only on a technicality 2015-08-29 23:09:16 in practice, a GPL software that is behind a paywall except for a "testing" version is still non-free 2015-08-29 23:09:46 likewise, depending on a non-free software for the default kernel environment, is also probably a poor idea 2015-08-29 23:10:09 irregardless of perceived security advantages (of which, grsec only makes a small part of the overall security position in alpine) 2015-08-29 23:12:28 nmeum: would you like to research isolating the grsecurity components we actually make use of in a form that can be rebased ontop of linux-vanilla allowing us to not have to worry any further about spender's decision? 2015-08-29 23:13:03 nmeum: at this point, it is essentially PaX 2015-08-29 23:15:37 perhaps PaX can also be cleaned up 2015-08-29 23:15:41 and upstreamed 2015-08-29 23:16:01 either way, alpine is not part of the brad spengler ego/koolaid club 2015-08-29 23:16:08 kaniini: thanks 2015-08-29 23:16:37 and if grsecurity is the reason you are here, you've totally missed the point 2015-08-29 23:16:39 :) 2015-08-29 23:20:43 speaking of which isn't the correct solution to their problem further legal action 2015-08-29 23:21:28 like sure they can't afford it by themselves but surely there are groups with bones to pick that would be happy to take the excuse 2015-08-29 23:21:56 it's a non-problem 2015-08-29 23:22:03 the reality is spender wants more money 2015-08-29 23:22:14 and is using this as an excuse to paywall the patches 2015-08-29 23:22:20 the solution, is to fork grsecurity 2015-08-29 23:22:36 a good way to do that would be to isolate each individual component and convert it into a patch series 2015-08-29 23:22:50 well yeah 2015-08-29 23:23:06 their excuse doesn't really hold much water is my complaint 2015-08-29 23:23:10 instead of schilling spender as some people here are doing, 2015-08-29 23:23:20 they should instead do this work, or do nothing 2015-08-29 23:23:52 well I'm actually not so sure tthat spender just wants more money 2015-08-29 23:23:59 grsecurity isn't even very good 2015-08-29 23:24:16 there's a lot of patches he's released on -test, that were simply wrong 2015-08-29 23:25:00 this isn't to say that there isn't stuff in grsec we don't want to lose 2015-08-29 23:25:09 i am just saying, working with -test is a non-starter 2015-08-29 23:25:20 if we have to do that, we are better off dropping the patch 2015-08-29 23:25:27 his/their support and kernel release maintenance prices are just not high enough to make me think they are just wanting more money with that 2015-08-29 23:26:02 his motivation is questionable at best, sorry 2015-08-29 23:26:10 clearly 2015-08-29 23:26:17 either way, we cannot provide an acceptable QA quality with -test patches 2015-08-29 23:26:21 so we need to solve this problem 2015-08-29 23:26:25 agreed 2015-08-29 23:26:26 or we need to drop the patch 2015-08-29 23:26:37 snrk 2015-08-29 23:26:51 shipping an alpine with a -test patch is worse than shipping without because people will think it's secure when it is possibly really broken 2015-08-29 23:27:00 would have liked for him to give more details about the redistrubutability of the patch by sponsors 2015-08-29 23:27:16 kaniini: agreed 2015-08-29 23:27:23 well, either way i doubt he really likes what we are doing now 2015-08-29 23:27:23 -test is a no go imho 2015-08-29 23:27:42 does anyone else use grsec for similar things 2015-08-29 23:27:56 we only really use grsec for patch, and some procfs hardening 2015-08-29 23:28:06 most of the grsec features we used to use, have been upstreamed as part of Yama 2015-08-29 23:28:33 enhancing Yama to add the missing features and then having a PaX build and a non-PaX build, would arguably, be a better situation than present 2015-08-29 23:29:02 I love TPE :( 2015-08-29 23:29:40 i'm not sure why, TPE is basically worthless 2015-08-29 23:29:53 why ? 2015-08-29 23:30:11 all it does is prevent someone running gcc to build some 0day they downloaded 2015-08-29 23:30:22 exactly 2015-08-29 23:30:24 more than likely, they will exploit a 0day in a suid application instead to get root 2015-08-29 23:30:28 so, basically worthless 2015-08-29 23:30:45 also, grsec most likely doesn't save you from that unless the system is compiled as PIE 2015-08-29 23:30:58 which we do 2015-08-29 23:31:03 and -vanilla will save you there too 2015-08-29 23:31:35 you can't even imagine the number of people I have here that like to try out the new local exploits they find 2015-08-29 23:31:53 do you run a shell box lol 2015-08-29 23:32:21 I run servers with a lot of developers 2015-08-29 23:32:23 :) 2015-08-29 23:32:42 you are probably not running a stock alpine-grsec kernel image then 2015-08-29 23:32:52 nope 2015-08-29 23:32:53 ... can tpe actually save you from anything on a system that isn't hilariously broken anyway 2015-08-29 23:33:01 ammunta: probably not 2015-08-29 23:33:36 coredumb: i think TPE could be implemented using an LSM though 2015-08-29 23:33:53 coredumb: there is, for example, https://github.com/cormander/tpe-lkm 2015-08-29 23:33:59 it's been implemented as a module yes 2015-08-29 23:34:01 this one 2015-08-29 23:34:11 been using it on some old EL6 kernels 2015-08-29 23:34:26 i've contributed patches to it :) 2015-08-29 23:34:58 running late here 2015-08-29 23:35:04 gnite 2015-08-29 23:36:06 anyway i think that the way to go is to take the current -stable, and split out each logical component 2015-08-29 23:36:20 and release it as an open-source patchset 2015-08-29 23:36:33 grsecurity attitude is same as openssl, i wouldn't trust it 2015-08-29 23:36:34 https://www.grsecurity.net/~paxguy1/ < btw will these patches still be available ? 2015-08-29 23:36:56 who knows 2015-08-29 23:37:05 presumably 2015-08-29 23:38:26 i think starting from pax-linux patch, and then porting over the interesting GRSEC features by hand is not that difficult 2015-08-29 23:38:40 think so too 2015-08-29 23:39:06 but if pax gets non-public as well ... 2015-08-29 23:39:31 if pax gets non-public, oh well 2015-08-29 23:39:43 pax at this point is very mature 2015-08-29 23:39:51 announcement is signed by both of them 2015-08-29 23:40:36 sure 2015-08-29 23:41:01 but other people will step forward if there is a cleanly documented replacement 2015-08-29 23:41:13 "The PaX Team" is hardly the only person who can work on a VMM 2015-08-29 23:41:30 :) 2015-08-29 23:46:39 that said, if the venue for such a patch series is something like github, i would suggest preemptively blocking spender 2015-08-29 23:46:41 :) 2015-08-29 23:47:35 preemptively blocking? 2015-08-29 23:47:48 yes, you can block people on github now. 2015-08-29 23:47:58 whoever does it, i would suggest saving yourself a lot of abuse and just blocking him early 2015-08-29 23:48:00 ;) 2015-08-29 23:49:15 ah, I was imagining something like blocking contributions from instead of 2015-08-29 23:49:18 :&& 2015-08-29 23:51:08 anyway, spengler works for https://www.optiv.com/ 2015-08-29 23:51:21 so really a lot of grsec "features" are cloak and dagger stuff and disabled in alpine 2015-08-29 23:51:31 because they don't hold relevance in the real world 2015-08-29 23:51:34 cloak and dagger? 2015-08-29 23:52:03 theoretical attacks 2015-08-29 23:52:53 ah 2015-08-29 23:54:07 wasn't too clear whether that was more along the lines of "well if the nsa is doing an active attack on you" or security theatre or 2015-08-30 08:42:51 hey :) 2015-08-30 08:43:10 can someone explain to me what the entire grsecurity licence thing is about? 2015-08-30 08:43:17 are we still able to get the current patches? 2015-08-30 08:49:28 no, we don't want -test patches 2015-08-30 08:49:53 i think most likely what will happen is we will separate PaX from grsecurity, and then reimplement the relevant grsecurity features as a patch ontop of the Yama LSM 2015-08-30 09:13:01 hmmm 2015-08-30 09:13:17 i still dont get why grsecurity is reacting this way just because of one stupid vendor 2015-08-30 09:17:22 $$$ 2015-08-30 09:18:02 hmmm, yeah 2015-08-30 09:18:06 fucking money 2015-08-30 09:18:31 damn, i have to go 2015-08-30 09:18:34 see you later 2015-08-30 09:19:25 bye 2015-08-30 09:23:07 kaniini: is that yama lsm really interesting? 2015-08-30 09:23:19 i fail to see it from the kernel doc 2015-08-30 09:23:23 :( 2015-08-30 09:28:18 specially the part where it disable all other LSM 2015-08-30 10:04:41 re 2015-08-30 10:04:49 kaniini: so what exactly is your plan? 2015-08-30 10:05:09 reimplementing stuff or asking grsecurity for access to the patches because we are 100% open source 2015-08-30 10:08:25 kaniini: I don't that PaX is the only thing we need/use under the hood there are many other improvements, for instance chroot security improvements which we also use currently 2015-08-30 10:09:17 kaniini: and GPL software that is behind a paywall isn't non-free it's free for everyone having access to it 2015-08-30 10:09:50 The GPL doesn't prevent you from selling your software 2015-08-30 10:10:19 nmeum: correct, but its still a shitty move from grsecurity 2015-08-30 10:10:34 yeah, it's a shitty move but it doesn't make the thing non-free 2015-08-30 10:11:24 i would not be surprised if some government forced the trouble on grsecurity on purpose, because grsecurity is currently the thing that makes linux exploids the hardest 2015-08-30 10:11:33 but thats just a theory ;) 2015-08-30 10:12:13 very unlikely 2015-08-30 10:12:52 just think about how often grsecurity has protected alpine from some zero days 2015-08-30 10:13:03 but yes, its just a theory ;) 2015-08-30 10:35:03 nmeum: as i said, feel free to solve the problem 2015-08-30 10:35:29 nmeum: if you want to just repeat brad spengler's talking points, then feel free to invite him here so he can explain himself. 2015-08-30 10:37:03 leo-unglaub: *my* plan is to just fork the damn stable patchset for now, and investigate options for improving the maintainability situation later. 2015-08-30 10:37:57 leo-unglaub: we clearly cannot depend on spengler, unless he creates something in writing saying we will have access explicitly to the -stable patches to redistribute in alpine 2015-08-30 10:38:20 kaniini: my point simply is that software isn't non-free (as in freedom) just because you have to pay for it, none the less it's a shitty move and personally I will probably start building my own kernel from the gentoo-hardened sources 2015-08-30 10:38:49 nmeum: except that if we pay him, and then redistribute the patches, 2015-08-30 10:39:04 nmeum: there is a likelyhood he will stop accepting our money and deny us future access 2015-08-30 10:39:18 nmeum: so, it does fail the alpine guidelines for "freedom" 2015-08-30 10:39:41 so, the options come back to: 2015-08-30 10:40:02 1. fork grsecurity to ensure that we can provide patches that pass the alpine definition of "freedom" 2015-08-30 10:40:11 instead of paying him we could simply asks existing sponsors for the source code and since it's GPL2 they have to give it to us if they redistribute it in binary versions 2015-08-30 10:40:24 that is impractical 2015-08-30 10:40:31 and we can "afford" to pay him 2015-08-30 10:40:55 given that, his sponsors probably do not wish to lose access, they are unlikely to redistribute to us anyway 2015-08-30 10:41:07 well…they have to it's GPL2 2015-08-30 10:41:09 and there is no requirement for them to give us grsecurity patches, just the naked kernel sources they use 2015-08-30 10:41:13 so that is useless 2015-08-30 10:41:56 no, the grsecurity patches are licensed under GPL2 they are legally required to give us the source if they redistribute a kernel with those patches applied in binary form 2015-08-30 10:42:04 no, they aren't 2015-08-30 10:42:09 you clearly do not understand GPL2 2015-08-30 10:42:12 what make you think that? 2015-08-30 10:42:15 they can give us the post-patched source 2015-08-30 10:42:17 and it counts 2015-08-30 10:42:31 you mean the pre-patched source? 2015-08-30 10:42:34 and no, it wouldn't count 2015-08-30 10:42:39 it does count 2015-08-30 10:42:41 seriously 2015-08-30 10:42:53 how many BSPs have you requested source for 2015-08-30 10:43:11 they usually give you the source tree, no patches 2015-08-30 10:43:37 I am not a lawyer and I don't assume that you are one but if you modify the source of a GPL2 programm and then redistrubte binary version of it than you have to give everyone who uses the binary version access to the modifications 2015-08-30 10:43:41 further, it would be more expensive for us, to hire a lawyer, to sue some grsecurity sponsor into giving us the patches 2015-08-30 10:43:47 *then' 2015-08-30 10:43:52 then it would be to just pay him for the patches 2015-08-30 10:44:06 again, the problem isn't acquisition, the problem is spender's bipolar disorder 2015-08-30 10:44:24 yeah, I agree on that 2015-08-30 10:44:49 and, something makes me think he is not likely to cooperate anyway 2015-08-30 10:45:02 there is far more alpine seats than there is gentoo-hardened or arch seats running grsec 2015-08-30 10:45:13 yet he mentions them explicitly :) 2015-08-30 10:45:57 so, that leaves the options as: 2015-08-30 10:46:07 1. try to buy the patches from spender, redistribute it and see if a shitshow happens 2015-08-30 10:46:59 2. try to ship using the -test patches and hope that we get enough QA on it, and hope a shitshow doesn't develop with the -test patches (untenable at best) 2015-08-30 10:47:35 3. drop -grsec and tell people to get -grsec patches from upstream by buying them, offer the APKBUILD so they can build it themselves (maybe) 2015-08-30 10:47:51 4. likewise, but with -test patches 2015-08-30 10:48:16 5. just drop -grsec and hope the other exploit mitigations we have in alpine are sufficient (likely true for 90% of the time) 2015-08-30 10:48:34 before doing 1. we might just ask him if it would be ok to redistribute them and then decide to become a sponsor based on his reply even though I highly doubt that we can afford it. 2015-08-30 10:48:46 regarding 5: what other exploit mitigations do we have on a vanilla kernel? 2015-08-30 10:48:54 6. fork -grsec, major win for distributions, distributions come together and replace spender 2015-08-30 10:49:58 regarding 2.: we can't use -test patches on the stable releases, might work for edge though 2015-08-30 10:50:49 nmeum: you have to think holistically, but, since the entire distribution is -fPIE and -fstack-protector-all and -relro, in combination with a well-audited base system (especially musl comes into play here), i'd say we're pretty good 2015-08-30 10:51:08 nmeum: keep in mind, we can get PaX without grsecurity, and that fills in 9/10% of the gap 2015-08-30 10:51:28 an LSM could be used to provide the chroot etc protections 2015-08-30 10:51:46 but using chroots is so 1990s, namespaces provide stronger guarantees already 2015-08-30 10:55:52 i strongly suggest (6) if you care about grsec, as this action will probably result in grsec fading away anyway 2015-08-30 11:16:11 6 will be a lot of work unless other developers start contributing to it 2015-08-30 11:26:24 this grsecurity thing is pidding me off ... 2015-08-30 11:26:37 a huge billion dollar company is fucking them over and we all have to pay for it ... 2015-08-30 11:40:21 i don't think it was just them 2015-08-30 11:40:40 i think that was just the latest example 2015-08-30 11:41:34 ACTION waits patiently for a hadoop job to complete 2015-08-30 11:51:44 yeah, probobly more than one case 2015-08-30 11:51:46 but still 2015-08-30 12:45:46 kaniini: problem is that PaX patch seems to become non-public as well 2015-08-30 12:46:25 actually where has been the stable PaX patch available lately ? 2015-08-30 12:47:00 feels like they shared Git with spender but have never seen public releases of the stable PaX patch ... 2015-08-30 12:47:42 there's also RBAC to replace by another MAC system... 2015-08-30 12:52:24 by the way what's the exact difference between the -stable and -test kernel patches? -stable is only used for linux-lts version and -test for mainline/stable linux-kernels? 2015-08-30 12:52:43 or does -test contain experimental grsec functionality? 2015-08-30 14:12:32 I'm unable to build gcc on stable-3.2. the output of abuild -r is here: http://qt.is/~pluesch/gcc_build_output.log 2015-08-30 14:12:49 can someone point me to the right direction what could be wrong with it? 2015-08-30 14:19:05 nmeum: the only difference is that one is for lts and the other for stable iirc 2015-08-30 14:45:06 wow. it seems that gcc is not building because of zsh 2015-08-30 14:46:33 have the same issues with other packages. for example: zsh 2015-08-30 14:46:43 zsh doesn't build if you use zsh XD 2015-08-30 15:08:58 oh is that weird thing zsh thing still happening 2015-08-30 15:09:42 yeah it seems 2015-08-30 15:10:04 ammunta: have you noticed it too? 2015-08-30 15:15:13 I tried tracking it down but then I realised that autoconf output is abominations 2015-08-30 15:18:05 imho it's something with array handling 2015-08-30 15:18:22 pluesch: wait…how's that possible? You made zsh your /bin/sh? 2015-08-30 15:18:47 nmeum: what? no 2015-08-30 15:19:06 so you just run abuild(1) inside a zsh session? 2015-08-30 15:19:12 yes 2015-08-30 15:19:19 bc users login shell points to zsh 2015-08-30 15:19:49 and what error do you get becasue abuild is using /bin/sh as its shebang 2015-08-30 15:20:24 >I'm unable to build gcc on stable-3.2. the output of abuild -r is here: http://qt.is/~pluesch/gcc_build_output.log 2015-08-30 15:20:45 same happens with the package zsh 2015-08-30 15:21:02 but the error is different 2015-08-30 15:22:02 let me get a log real quick 2015-08-30 15:26:20 http://qt.is/~pluesch/zsh_build_output.log 2015-08-30 15:26:40 just to make it clear... this only happens when you start the ssh session to the server with zsh as the shell 2015-08-30 15:27:00 it doesn't happen when you start the ssh with ash as the shell and switch to zsh by executing /bin/zsh 2015-08-30 15:36:36 looks like abuild is something weird 2015-08-30 15:36:41 doing* 2015-08-30 15:37:10 pluesch: you said "it seems that gcc is not building because of zsh" 2015-08-30 15:37:37 nmeum: yes 2015-08-30 15:37:57 so you cannot build gcc, because you are using zsh? 2015-08-30 15:38:04 yes 2015-08-30 15:38:30 but you can build it when you invoke abuild from /bin/sh? 2015-08-30 15:38:45 fom /bin/ash to be more specific 2015-08-30 15:40:24 and, the important part is, I can also build it when I login to my buildserver with login shell set to /bin/ash then change to /bin/zsh and invoce abuild -r 2015-08-30 15:40:39 at least that's true for the zsh package, I'm trying it out on the gcc package now too 2015-08-30 15:41:06 abuild is using /bin/sh as a shebang, I don't how the fact that you are using zsh could cause an issue 2015-08-30 15:41:14 *know 2015-08-30 15:41:37 it seems that gcc works with this trick too 2015-08-30 15:41:54 so the ssh login part with login shell set to /bin/zsh is the point were something is going wrong 2015-08-30 15:42:09 but at the moment I have no idea what 2015-08-30 15:42:27 it sounds pretty strange… 2015-08-30 15:43:12 does /bin/sh -c /usr/bin/abuild work? 2015-08-30 15:45:43 let me try 2015-08-30 15:47:19 same error 2015-08-30 15:54:30 some difference in environment when zsh started as login shell? 2015-08-30 16:01:33 fabled: zsh as login shell: http://qt.is/~pluesch/env_zsh_loginshell.txt (were abuild -r is not working for some packages) 2015-08-30 16:01:35 zsh not as login shell but started later with /bin/zsh: http://qt.is/~pluesch/env_zsh.txt 2015-08-30 16:03:01 pluesch, $SHELL is different 2015-08-30 16:03:07 i suspect autoconf uses that 2015-08-30 16:03:47 it's gcc's autconf launch that runs stuff with $SHELL and breaks when $SHELL=zsh 2015-08-30 16:04:21 I can confirm that by export SHELL=/bin/ash and try again right? 2015-08-30 16:04:27 yes 2015-08-30 16:07:24 nmeum: you don't have to make zsh /bin/sh 2015-08-30 16:07:49 ammunta: I never suggested that 2015-08-30 16:07:50 if it's in $SHELL, configure is likely to try to reexec itself with that 2015-08-30 16:08:09 fabled: okay. you are right. just tried it out. setting SHELL before invoking abuild builds without an error 2015-08-30 16:08:19 because configure does a lot of ridic things in the name of Absolute Portability 2015-08-30 16:08:35 wow, that configure script is pretty shitty then 2015-08-30 16:08:49 it is 2015-08-30 16:08:56 fucking gnu autoconf 2015-08-30 16:09:08 so is there still something wrong with the zsh package on alpine linux or is this just really gnu autoconfs problem? 2015-08-30 16:09:30 because I remember to be able to compile zsh on for example archlinux with shell set to zsh just fine 2015-08-30 16:09:37 I can test it again just to make sure 2015-08-30 16:09:39 "Remember, /bin/sh is not Posix compliant on many systems, such as FreeBSD 4, NetBSD 3, AIX 3, Solaris 10, or Tru64." 2015-08-30 16:10:04 pluesch: there's *probably* a problem with the zsh build 2015-08-30 16:10:33 autoconf may be a piece of shit but it'd be really weird if this were something that would just up and break 2015-08-30 16:10:47 shell scripts usually aren't supposed to segfault the shell 2015-08-30 16:11:12 it's not segfaulting. it failed to create some path 2015-08-30 16:11:32 config.status: executing default-1 commands 2015-08-30 16:11:33 mkdir build-x86_64-alpine-linux-musl/libiberty 2015-08-30 16:11:33 Configuring in build-x86_64-alpine-linux-musl/libiberty 2015-08-30 16:11:33 mkdir: can't create directory 'build-x86_64-alpine-linux-musl/libiberty': No such file or directory 2015-08-30 16:11:33 zsh:cd:7: no such file or directory: build-x86_64-alpine-linux-musl/libiberty 2015-08-30 16:11:33 Makefile:2494: recipe for target 'configure-build-libiberty' failed 2015-08-30 16:11:43 ah, never mind 2015-08-30 16:11:55 unrelated, then 2015-08-30 16:12:07 dunno why though 2015-08-30 16:14:10 "On ancient BSD systems, touch or any command that results in an empty file does not update the timestamps, so use a command like echo as a workaround. Also, GNU touch 3.16r (and presumably all before that) fails to work on SunOS 4.1.3 when the empty file is on an NFS-mounted 4.2 volume. However, these problems are no longer of practical concern." 2015-08-30 16:15:02 an acquaintance, on reading this, decided to test agains BSD 2.11 2015-08-30 16:15:07 it. was not broken. 2015-08-30 16:15:27 that is truly not of practical concern 2015-08-30 16:16:45 probably like 100% of software that builds with autoconf wouldn't even run on half the systems it supports anyway 2015-08-30 17:23:01 I didn't actually test that ocaml much more than "well it can bootstrap itself" ... but, testing :&& 2015-08-30 17:24:32 maybe I'll package up opam too? 2015-08-30 17:24:46 it seems like a useful thing to 2015-08-30 17:54:34 package website down? 2015-08-30 18:03:27 clandmeter: ↑ 2015-08-30 18:27:40 dammit parallel build nondeterminism 2015-08-31 06:50:46 whats this: /usr/lib/libparted.so: undefined reference to `dm_task_get_info@Base' 2015-08-31 06:57:07 hum 2015-08-31 06:57:16 on v3.2: 2015-08-31 06:57:30 $ nm -D /lib/libdevmapper.so.1.02 | grep dm_task_g 2015-08-31 06:57:30 et_info 2015-08-31 06:57:30 000000000002d436 T dm_task_get_info 2015-08-31 06:57:30 0000000000029d59 T dm_task_get_info 2015-08-31 06:57:30 000000000002d436 T dm_task_get_info_base 2015-08-31 06:57:31 0000000000029d59 T dm_task_get_info_v1_02_97 2015-08-31 06:57:33 000000000002d4af T dm_task_get_info_with_deferred_remove 2015-08-31 06:58:08 on edge: 2015-08-31 06:58:10 $ nm -D /lib/libdevmapper.so.1.02 | grep dm_task_get_ 2015-08-31 06:58:11 info 2015-08-31 06:58:11 0000000000035308 T dm_task_get_info_with_deferred_remove 2015-08-31 06:58:11 0000000000031a2f T dm_task_get_info 2015-08-31 06:58:40 I dont know if this is due to changes in gcc or changes in lvm2? 2015-08-31 07:03:44 I don't think we have rebuilt parted @ Arch since February 2015-08-31 07:05:53 also we still have dm_task_get_info_base there 2015-08-31 07:06:50 i think they do symbol versioning or something 2015-08-31 07:07:23 https://lists.fedorahosted.org/pipermail/lvm2-commits/2015-May/004024.html 2015-08-31 08:07:31 morning@all 2015-08-31 08:07:47 ncopa: found another point regarding the dependency problems we discussed last week 2015-08-31 08:08:06 question: is /etc/alpine-release existing on edge-installations? 2015-08-31 08:08:18 if yes, which value is in there? 2015-08-31 08:09:06 StarWarsFan, seems yes, and it seems to contain the latest x.y.0 release number; it seems to come from alpine-base package before branching 2015-08-31 08:44:31 fabled: ok 2015-08-31 08:44:40 because i've upgraded a box from 2.7 to edge 2015-08-31 08:44:52 and there is no file /etc/alpine-release anymore 2015-08-31 08:45:06 this was an x86-64 box 2015-08-31 08:45:29 but another machine, running x86 and also updated from 2.7 to edge got the file 2015-08-31 08:45:37 so just a little bit confusing... 2015-08-31 08:45:47 might be a problem during the dist-upgrade... 2015-08-31 09:52:14 ok drats 2015-08-31 09:52:33 i should have disable that firefox config option that makes it crash while at it 2015-08-31 14:01:50 isn't iconv packaged yet? 2015-08-31 14:03:34 (ignore this) 2015-08-31 15:51:10 does anybody else get a utf8 conversion error from gpg-connect-agent? 2015-08-31 16:30:56 the musl 1.1.11 upgrade seems to have caused an iconv issue…can I downgrade packages somehow to confirm this? 2015-08-31 16:48:23 so, yeah I can confirm it, musl 1.1.11 breaks gpg-connect-agents utf8 handling and possibly other software 2015-08-31 16:48:26 ncopa: ^ 2015-08-31 16:53:40 (it's an iconv issue) 2015-08-31 16:56:13 also: f10434640de3727d363237e28408c4c9dc1f1114 didn't bump pkgrel and thus didn't cause a rebuild 2015-08-31 16:57:19 nmeum, f104... did not need pkgrel as the patch was mentioned in earlier commit and it failed to build 2015-08-31 16:57:30 ah 2015-08-31 16:57:33 didn't see that 2015-08-31 16:57:42 doesn't fix the iconv issue though ;) 2015-08-31 16:57:46 nope 2015-08-31 16:57:52 what's the issue? 2015-08-31 16:58:51 oh 2015-08-31 16:59:55 after upgrading musl gpg-connect-agent printed the following warning: "gpg-connect-agent: conversion from 'utf-8' to 'UTF-8-CODE-UNITS'". the code printing that warning is using iconv.h, iconv.h is provided by musl and the upgrade somehow broke the iconv code in musl, downgrading musl fixed it 2015-08-31 17:00:03 so it's definitly an issue with the new musl version 2015-08-31 17:00:17 asking about that on #musl 2015-08-31 17:00:20 I assume that it also breaks other software but I only saw that with musl 2015-08-31 17:01:00 If I see it correctly we used to have some iconv patch in aports, I am looking into it right now 2015-08-31 17:01:15 http://git.alpinelinux.org/cgit/aports/tree/main/musl/iconv.c 2015-08-31 17:01:16 we still do 2015-08-31 17:01:25 I assume that this code needs to be fixed for the new musl version 2015-08-31 17:01:40 thus is not really upstreams fault but our fault 2015-08-31 17:11:17 fabled: ^ 2015-08-31 17:11:32 am I wrong? 2015-08-31 17:11:43 iconv.c is different 2015-08-31 17:11:46 it's the command line tool 2015-08-31 17:12:04 the iconv error from other code is directly from the iconv() function call provided by musl 2015-08-31 17:12:04 ah 2015-08-31 17:12:29 yeah, lol it provides a main function, I didn't read the thing m( 2015-08-31 19:15:59 https://www.mozilla.org/en-US/firefox/40.0.3/releasenotes/ 2015-08-31 21:22:18 hi! I still haven't found time to convert my server to alpine yet (damn, 8 months passed already), but just read this few days old sad news: http://grsecurity.net/announce.php 2015-08-31 21:23:18 I was never closely following grsecurity patches. does test and stable differ in quality that substantially? 2015-08-31 21:33:20 good evening 2015-08-31 21:33:32 has someone of you an aport for lua gtk?