2017-01-03 13:48:25 royger: and potential less cache misses :D 2017-01-03 13:48:57 I doubt very much that the strategy of using x32 to save memory, or even a few CPU cycles, is remotely worth the effort. 2017-01-03 13:48:58 ls, cat, etc. are short-lived programs, you're not going to oom a machine with them. What ooms a machine is big, memory-hungry applications, and for those you need 64-bit pointers anyway. 2017-01-03 13:48:58 If you need more memory, it's probably a better investment to go to the store and buy some RAM than spend time making x32 work for applications that don't even take much RAM in the first place. 2017-01-03 13:51:03 skarnet: OK, pick another example then, ffmpeg for example is a long running applications that uses less than 4GiB of memory (unless you are doing crazy encodings maybe), and could really benefit from faster pointer arithmetics, and this also applies to VLC for example 2017-01-03 13:51:15 it's not to save memory, it's to make operations on pointers faster 2017-01-03 13:51:33 please provide benchmarks so I can have a good laugh 2017-01-03 13:51:58 https://indico.cern.ch/event/214319/contributions/432623/attachments/339806/474205/x32ABI.pdf 2017-01-03 13:52:43 "Application uses millions of pointers" yeah, maybe that's the problem here? 2017-01-03 13:53:27 skarnet: http://linuxplumbersconf.org/2011/ocw/sessions/531 2017-01-03 13:56:32 5-11%, eh? that's actually not too bad. I still suspect the benchmark tools (povray? bwaves?...) are specific cases, like the CERN application, thatbenefit from x32 particularly well because they're using a lot of pointers. 2017-01-03 13:57:41 my point is just that the gains have to be balanced against the cost of making the thing work. Those costs are often underestimated. 2017-01-03 13:59:55 I agree that the benefits might not out weight the trouble, but the gain is probably mostly from decreased cache misses than 64 vs 32 pointer operations 2017-01-03 14:00:20 RAM is cheap but cache isn't 2017-01-03 14:01:42 oh, even RAM isn't cheap. But you know what's more expensive than RAM and cache and CPU cycles? Developer time. That's the most expensive resource of all. 2017-01-03 14:12:15 Kruge, YOU AROUND? 2017-01-03 14:12:20 Sorry 2017-01-03 14:12:25 *you around? 2017-01-03 14:12:26 :) 2017-01-03 14:26:54 Kruge, i've opened a feature request for openvas-smb since a while 2017-01-03 14:26:55 https://bugs.alpinelinux.org/issues/5011 2017-01-03 15:02:50 I'm here 2017-01-03 15:04:36 Just read your issue 2017-01-03 15:06:43 I'm really keen to get this moving, but don't have a massive amount of experience in this sort of thing 2017-01-03 15:16:51 I take it ncopa's suggestion of winegcc didn't work? 2017-01-03 15:39:50 it's bb-offtopic, so responding here. ncopa, until ash bug is fixed, maybe we should switch to internal glob for now in AL bb? 2017-01-03 15:40:02 przemoc: im on it 2017-01-03 15:47:21 BTW which repo server remains the most up-to-date nowadays? 2017-01-03 15:48:09 I'm mostly using dl-5, as it's fast, but apparently not the most fresh one 2017-01-03 15:48:56 fr.a.o 2017-01-03 15:49:06 reminds me, we should fix that 2017-01-03 15:50:46 sorry, I mischecked, bb 1.26.1-r1 is not built for x86_64 yet 2017-01-03 15:51:01 so maybe it's not lagging at all 2017-01-03 15:51:07 because the builder is stuck 2017-01-03 15:51:18 its probably syncing every hour 2017-01-03 16:08:49 przemoc: updated busybox should be available now 2017-01-03 16:08:52 sorry about that 2017-01-03 16:10:47 no, I'm sorry that I haven't posted such patch yesterday, but I thought I'll have time to tinker with bb to find the real issue (glob used where it shouldn't be), but it seems unlikely 2017-01-03 16:12:09 at least today 2017-01-03 18:07:20 jirutka - please don't pull nginx-naxsi yet - I've backported http-sysguard from tengine so it works as a dynamic module - will rebase 2017-01-03 22:15:30 um.. how well is gdb on alpine expected to work? 2017-01-03 22:20:24 I'm getting "unknown signal" debugging even the simplest multi-threaded program 2017-01-03 22:23:53 :D 2017-01-03 22:24:09 iirc, its because musl does not have a thread debugging library 2017-01-03 22:24:17 but don't quote me on that :P 2017-01-03 22:26:02 I'll try to resist the temptation 2017-01-03 23:46:40 is it possible that the pthread protections of grsec are interfering? 2017-01-03 23:51:03 possible 2017-01-04 09:27:06 morning 2017-01-04 09:27:58 howdy 2017-01-04 11:09:13 hmm, isci kernel modules isn't being loaded for my intel C600 SAS Controller by init since upgrading in edge 2017-01-04 11:09:34 it's in initramfs-grsec, but not loading 2017-01-04 11:21:15 3.5 boots fine from a stick, isci loads. i might just re-install and reload world 2017-01-04 11:26:06 /dev/sd?? device are renamed when booting 3.5 I wonder if that has anything to do with it. 2017-01-04 11:29:13 (for root, 3.4:/dev/sda1 -> 3.5:/dev/sdd1) 2017-01-04 19:18:30 can nginx-naxsi & nxapi be pulled on github please ? 2017-01-04 19:37:20 BitL0G1c: I’ve reviewed it 2017-01-04 19:39:59 many thx 2017-01-04 20:19:41 hey jirutka ncopa clandmeter, what do you think about https://github.com/sabotage-linux/netbsd-curses ? 2017-01-04 20:20:07 asie wanted to post this first, but he's too shy to talk on this channel, no idea why 2017-01-04 20:20:09 skrzyp: dunno, this is not my field 2017-01-04 20:20:25 so I'm pasting the links in behalf of him 2017-01-04 20:20:37 skrzyp: because i don't feel i'm in a position enough to make far-reaching suggestions like this 2017-01-04 20:20:46 or to ping everyone like that 2017-01-04 20:20:52 skrzyp: hm, we have more shy people here, another one is Alba Pompoe, he reads the channel, but only emails me :P 2017-01-04 20:20:54 asie: linux serious business 2017-01-04 20:20:57 skrzyp: yes 2017-01-04 20:20:58 actually, yes 2017-01-04 20:21:04 operating systems *are* serious business 2017-01-04 20:21:52 well, Linux is no so serious 2017-01-04 20:21:54 not 2017-01-04 20:24:05 right now i'm working on cleaning up more of my local aports 2017-01-04 20:29:48 here's a question - if i'm going to be pushing OpenTTD (which has a few additional packages for the default sound/graphics assets, which are distributed separately) 2017-01-04 20:29:57 should I push the aports as multiple commits in one PR, or as multiple PRs, one commit each? 2017-01-04 20:31:01 asie: multiple commits in one PR 2017-01-04 20:31:26 thanks 2017-01-04 20:31:38 huh 2017-01-04 20:31:47 that's way better than void 2017-01-04 20:31:59 they require to amend everything in single commit 2017-01-04 20:34:01 squashing multiple separate changes into single commit is anti-pattern 2017-01-04 20:35:41 elaborate 2017-01-04 20:36:09 you can't revert easily that way 2017-01-04 20:36:25 that 2017-01-04 20:36:38 if you have a change to X Y and Z, each independent of each other, and need to revert Y, its easier if its a separate commit 2017-01-04 20:37:32 and not just reverting, even reviewing patches or analyzing history 2017-01-04 20:37:43 yep makes bisecting easier too 2017-01-04 20:40:22 okay, there we go 2017-01-04 20:41:42 build failed 2017-01-04 20:41:45 bam 2017-01-04 20:42:06 oj 2017-01-04 20:42:21 that's not nice 2017-01-04 20:42:22 ah jo 2017-01-04 20:44:37 hm... libiculx exists in 3.5 but not edge 2017-01-04 20:46:08 yes, icu58 2017-01-04 20:46:16 this is provided by icu-libs that exists in edge 2017-01-04 20:46:29 not in icu 58 2017-01-04 20:46:36 they split it out in that version, to separate packages 2017-01-04 20:47:09 ..or at least, not installed by default 2017-01-04 20:47:11 researching 2017-01-04 20:48:04 https://pkgs.alpinelinux.org/contents?branch=edge&name=icu-libs&arch=x86_64&repo=main indeed, iculx is now missing! 2017-01-04 20:49:14 --without-icu-layout can fix it, but then there's no right-to-left support 2017-01-04 20:49:29 "The layout engine code has been removed; the ParagraphLayout is not deprecated and remains (and must now be built on top of HarfBuzz)." 2017-01-04 20:49:51 it appears --without-icu-layout is the best course of action until openttd moves to harfbuzz 2017-01-04 20:50:02 from https://github.com/alpinelinux/aports/commits/master/main/icu it doesn’t look like intentional change 2017-01-04 20:50:13 no, it is 2017-01-04 20:50:26 could jirutka help asie fixing icu? 2017-01-04 20:50:31 skrzyp: "fixing" 2017-01-04 20:50:35 http://site.icu-project.org/download/58 2017-01-04 20:50:35 aha, so it’s a change in upstream? 2017-01-04 20:50:38 yes 2017-01-04 20:50:43 now iculx has to be built separately 2017-01-04 20:50:49 the new version is a wrapper on harfbuzz, essentially 2017-01-04 20:50:55 aha 2017-01-04 20:51:12 I’ll look at it later, can’t do it now 2017-01-04 20:51:31 so, right now we just want to disable icu in openttd? 2017-01-04 20:51:36 skrzyp: nah 2017-01-04 20:51:43 icu-le-hb and icu-lx packages have to be created, as per http://userguide.icu-project.org/layoutengine/paragraph 2017-01-04 20:51:48 i'll look into it 2017-01-04 20:56:03 icu-le-hb is done, now icu-lx 2017-01-04 20:56:45 how_fast.ogv 2017-01-04 20:57:39 not sure how to handle icu-lx though 2017-01-04 20:58:57 it's compiled from icu, so requires the same patches/etc.; however, it depends on icu-le-hb 2017-01-04 20:59:06 is there a clean way to handle this in apk? 2017-01-04 20:59:11 (and icu-le-hb depends on icu) 2017-01-04 22:15:52 asie: how's going with that? 2017-01-04 22:15:56 skrzyp: i'm stuck 2017-01-04 22:16:00 i asked a question above 2017-01-04 22:16:06 i don't want to just cp -R icu/ icu-lx/ 2017-01-04 22:16:12 as that's bad maintenance 2017-01-04 22:16:27 but i don't think apk lets you do "build this package, then build a package depending on it for this subpackage" 2017-01-04 22:16:30 jirutka: ^ 2017-01-04 22:16:47 as icu-lx is part of icu (so it needs the same patches as icu), but depends on icu-le-hb now (which depends on icu) 2017-01-04 22:17:10 so circular dependency…? 2017-01-04 22:17:17 not circular 2017-01-04 22:17:23 per se 2017-01-04 22:17:26 icu -> icu-le-hb -> icu-lx 2017-01-04 22:17:31 except that icu-lx is part of the icu sourcetree 2017-01-04 22:17:47 asie: wait wait 2017-01-04 22:17:51 you can create subpackages 2017-01-04 22:17:54 yes 2017-01-04 22:18:05 but they don't work here 2017-01-04 22:18:08 this is a main principle for alpine 2017-01-04 22:18:13 no you don't understand at all 2017-01-04 22:18:24 unless i make icu-le-hb icu's subpackage, which is nonsense 2017-01-04 22:18:33 (because icu-le-hb is *not* part of icu's distribution) 2017-01-04 22:18:43 it might, however, be the best solution for the time being 2017-01-04 22:18:56 i think that i understand the problem… this is very bad situation, it can be solved, but ncopa will not be happy… 2017-01-04 22:19:03 because you cannot make a subpackage which depends on a different package which depends on the main package 2017-01-04 22:19:21 i think the sanest solution for the time being is to make icu-le-hb a part of the icu package, and have the icu package build icu, icu-le-hb and icu-lx 2017-01-04 22:19:32 however, then we get versioning problems 2017-01-04 22:19:37 as icu-le-hb and icu's releases are not synchronized 2017-01-04 22:19:49 so we have icu 58.2 and icu-le-hb 1.0.3 right now 2017-01-04 22:20:35 however, most distros simply deprecated icu-lx altogether i think 2017-01-04 22:20:47 that's the intention - icu developers want people to replace icu's layout engine with harfbuzz 2017-01-04 22:21:17 in FreeBSD, openttd is the only package depending on icu-lx 2017-01-04 22:21:21 and it's an optional dependency in the ./configure 2017-01-04 22:21:35 so it might be best to just drop icu-lx and wait for openttd devs to adopt harfbuzz? 2017-01-04 22:21:39 is it possible to build icu-lx in a separate package? 2017-01-04 22:21:42 yes 2017-01-04 22:21:46 but that's cp -r icu/ icu-lx/ 2017-01-04 22:21:49 which is a bit ugly 2017-01-04 22:22:05 (i assume both packages use the same src/common, which is patched by icu/) 2017-01-04 22:22:19 of course, as i said, i can just tell openttd not to use icu-lx... it's a game, it's not a huge loss 2017-01-04 22:22:23 and nothing else seems to really use icu-lx anymore 2017-01-04 22:22:39 i think i'll do that for now and we'll have this discussion again when something else using icu-lx pops up. ok? 2017-01-04 22:23:06 it is, but since icu-lx is deprecated, moving icu-le-hb to icu is plain wrong and messing with circular dependency is problematic, it may be the best option 2017-01-04 22:23:17 yes 2017-01-04 22:23:24 and only one thing uses it in all of freebsd, which is a game 2017-01-04 22:23:34 in other distros i think they just disabled openttd's usage of icu 2017-01-04 22:23:53 what do you lose in OpenTTD when you remove icu-lx? 2017-01-04 22:24:00 right-to-left language support 2017-01-04 22:24:06 asie: but, eh 2017-01-04 22:24:22 I'm not sure if people from RTL countries will use their locale in OpenTTD 2017-01-04 22:24:31 skrzyp: there's no perfect solution 2017-01-04 22:24:38 and the amount of hacks necessary to get RTL to work in OpenTTD is not worth the effort 2017-01-04 22:24:40 if it's just OpenTTD 2017-01-04 22:24:53 and if they want to, they would just file a bug 2017-01-04 22:25:02 that’s IMO not a problem, we usually don’t build pkgs with nls and alternative locales support to make them smaller 2017-01-04 22:25:03 most Alpine users are from EU, some USA 2017-01-04 22:25:10 anyhow, testing OpenTTD without ICU and updating my PR 2017-01-04 22:25:17 should I remove non-English languages for OpenTTD as well, then? 2017-01-04 22:25:26 asie: ^ as they said, no NLS mostly 2017-01-04 22:25:28 i can do that... i could also remove TTF support as it's non-essential 2017-01-04 22:25:32 dunno in this case 2017-01-04 22:25:37 well, it's just .lang files 2017-01-04 22:25:37 asie: leave TTF 2017-01-04 22:25:41 could be easily split out to a subpackage 2017-01-04 22:25:45 it’s a game, not just simple CLI tool 2017-01-04 22:25:46 but remove nls/langs 2017-01-04 22:25:46 "openttd-lang" 2017-01-04 22:25:54 i think that might be a good option, as it's langfiles 2017-01-04 22:25:59 it's not a compile-time option 2017-01-04 22:26:07 yes, splitting to subpkgs is a good option 2017-01-04 22:26:11 ok, on it 2017-01-04 22:27:12 for the default "openttd", i assume only keep the default language/ 2017-01-04 22:27:14 ? 2017-01-04 22:27:16 (which is English (UK)) 2017-01-04 22:27:32 yes 2017-01-04 22:28:24 asie: btw. 2017-01-04 22:28:35 asie: openttd-open{gfx,sfx} is not required 2017-01-04 22:28:39 skrzyp: i know 2017-01-04 22:28:43 i didn't make them in that way 2017-01-04 22:28:51 if openttd fails to find it, it proposes to download 2017-01-04 22:28:55 they depend on openttd (for reasons of being useless otherwise), but openttd doesn't depend on them 2017-01-04 22:29:42 you can add README in post install scrit to notify user that he would need gfx/sfx but it's not installed by default because game proposes to downlaod them on its own 2017-01-04 22:29:53 jirutka: can we print messages to user when installing pkgs? 2017-01-04 22:31:41 skrzyp: I do this with post-install script, but clandmeter has warned me that it’s not the preferred way; however, the preferred way is incomplete and undocumented… 2017-01-04 22:32:19 jirutka: in void, theres was insititution called "INSTALL.msg" which was being directly printed by xbps 2017-01-04 22:32:25 without scripting 2017-01-04 22:32:37 also UINSTALL.msg for removing of course 2017-01-04 22:32:42 UNINSTALL* 2017-01-04 22:32:52 skrzyp: so, basically, it’s IMO ok-ish for now, until we have proper documented alternative; but keep the msg very short and follow the convention like in https://github.com/alpinelinux/aports/blob/master/community/R/R.post-install 2017-01-04 22:33:11 skrzyp: this is insufficient for upgrade messages 2017-01-04 22:33:27 skrzyp: b/c upgrade messages should be conditional, based on the version you’re upgrading from 2017-01-04 22:34:41 we have something like README.alpine and some pkg that prints these messages, but it’s not installed by default, so basically useless 2017-01-04 22:34:58 and it’s only for post-install messages iirc 2017-01-04 22:35:25 oh, hey, it turns out openttd-lang is bigger than openttd 2017-01-04 22:35:46 well, it’s a game, probably contain A LOT of texts 2017-01-04 22:35:54 yeah 2017-01-04 22:35:57 and a lot of languages 2017-01-04 22:36:11 56, apparently 2017-01-04 22:37:12 yes 2017-01-04 22:37:28 openttd had some bounties to translate it to almost every language 2017-01-04 22:37:58 even some weird Afrikaans or fifth Chinese dialects 2017-01-04 22:38:25 >fifth Chinese dialects 2017-01-04 22:38:30 there's only the regular two: traditional and simplified 2017-01-04 22:38:38 those are both normal in translations 2017-01-04 22:38:52 that's just a joke 2017-01-04 22:39:04 but you got it, they're including some very rarely used languages 2017-01-04 22:39:11 that they are, a few 2017-01-04 22:39:24 also, mind to NOT move town names list to languages 2017-01-04 22:39:30 it might break network lay 2017-01-04 22:39:32 play* 2017-01-04 22:41:44 i'm not doing anything like that 2017-01-04 22:42:04 unless it's actually in the langfiles 2017-01-04 23:27:15 okay, openttd PR updated 2017-01-04 23:27:27 let's hope it compiles 2017-01-04 23:32:46 ok 2017-01-04 23:36:28 2017-01-04 23:36:45 jirutka: who's able to confirm and merge? 2017-01-04 23:37:16 they'll get to it in time, no rush 2017-01-04 23:37:30 i'm going to sleep in a bit anyway, so i'll only reply to change requests tomorrow 2017-01-04 23:37:50 today* 2017-01-04 23:39:38 > checking libxdg-basedir... not found 2017-01-04 23:39:39 > personal home directory... .openttd 2017-01-04 23:40:06 wouldn't be better to put it in $XDG_CONFIG_HOME/openttd? (~/.config/ by default) 2017-01-04 23:40:27 it's by no means a standard 2017-01-04 23:40:35 and redirecting that is usually a worse idea than not doing so 2017-01-04 23:41:00 on my computer right now i have .gimp-2.8, .gnome, .java, .local, .mozilla, .pki, .ssr, .visualvm, .wine, .abuild, etc. 2017-01-04 23:41:17 it's standard declared by freedesktop.org 2017-01-04 23:41:23 but openttd has support for libxdg-basedir 2017-01-04 23:41:30 so it's not hardocded redirection 2017-01-04 23:41:32 which apparently very few people care about, in practice 2017-01-04 23:41:38 I do care 2017-01-04 23:41:38 also i know, but most distros, guides, etc. assume .openttd 2017-01-04 23:41:47 .lcoal is not .config btw. 2017-01-04 23:41:49 i know 2017-01-04 23:41:57 .wine does not contain config files only 2017-01-04 23:42:02 but yes, openttd does support both 2017-01-04 23:42:07 my preference is .openttd as it's not strictly config 2017-01-04 23:42:10 but i'll wait for someone more experienced 2017-01-04 23:42:28 but when used with xdg, it might move data to .local/share/openttd 2017-01-04 23:42:31 and config to .config 2017-01-04 23:42:56 like everyone should do for consistency, especially when I'm moving/syncing XDG dirs between computers 2017-01-04 23:43:15 and I also can switch env vars to store config in ~/config, without dot 2017-01-04 23:43:20 yes, it does that 2017-01-04 23:43:24 but the question is - does alpine adhere to xdg? 2017-01-04 23:43:37 if yes, it might be a good idea to make a general effort to move more packages to xdg standards 2017-01-04 23:43:40 if no, well 2017-01-04 23:44:04 I think I might look deeper into this topic 2017-01-04 23:44:28 awesome and dunst are the two packages in alpine which use libxdg-basedir 2017-01-04 23:44:43 but many more use .config or .local/share 2017-01-04 23:44:45 it's still about 50/50 though 2017-01-04 23:45:28 there's no need to strictly use libxdg's by software, it just might point to $XDG_* cariables 2017-01-04 23:47:31 I think we've got some monthly record with these messages during last 12 hours 2017-01-04 23:47:48 activity is not bad if it's productive 2017-01-04 23:47:56 reviewed 2017-01-04 23:48:31 jirutka: and what can you say about XDG? (see backlog) 2017-01-04 23:49:18 I don’t know desktop stuff on Linux, I use it only on servers etc. 2017-01-04 23:49:52 what do you use on desktop then? 2017-01-04 23:50:02 jirutka: shouldn't -doc also be noarch? 2017-01-04 23:50:11 it's noarch by default afaik 2017-01-04 23:50:14 oh 2017-01-04 23:50:15 asie: no, -doc is noarch by default 2017-01-04 23:50:41 ok, thanks 2017-01-04 23:51:25 testing fixes 2017-01-04 23:52:19 jirutka: I'm not going to hate whatever option you'd picked, but do you know any other contributor/developer who uses alpine on desktops? 2017-01-04 23:52:39 If no, we should take some desktop mesaurements with asie and axoi 2017-01-04 23:53:14 skrzyp: try clandmeter, ^7heo, ncopa… 2017-01-04 23:53:27 I thought ncopa uses macOS 2017-01-04 23:55:27 skrzyp: not sure if he use *only* macOS 2017-01-04 23:56:31 right, I'm using AmigaOS too but it doesn't count as a main OS though :P 2017-01-04 23:56:47 okay, updated PR 2017-01-04 23:57:05 but I've attended some NetBSD convention last year 2017-01-04 23:57:11 and most developers used a Mac 2017-01-04 23:57:34 jirutka: one more question - should openttd-opengfx/opensfx depend on openttd in the first place? 2017-01-04 23:57:35 no one wanted to help us with installing NetBSD on my laptop 2017-01-04 23:57:43 there's no formal dependency - they're only files in a particular directory 2017-01-04 23:58:07 jirutka: and OpenTTD asks for downloading them from their website automatically 2017-01-04 23:58:16 after that, it stores these assets in homedir 2017-01-04 23:58:34 skrzyp: imho openttd-opengfx/opensfx should not depend on openttd 2017-01-04 23:58:49 agreed 2017-01-04 23:58:49 asie: ^ 2017-01-04 23:58:51 but why openttd should depend on them? 2017-01-04 23:58:52 fixing 2017-01-04 23:58:57 skrzyp: shouldn't, i already responded 2017-01-04 23:58:59 okay 2017-01-04 23:59:01 yes 2017-01-04 23:59:05 they're not a dependency nor are they the only option 2017-01-04 23:59:09 sorry, I've got confused for a bit 2017-01-04 23:59:17 though very few people actually use original TTD assets nowadays 2017-01-04 23:59:17 skrzyp: I also use macOS as desktop OS 2017-01-04 23:59:24 it is still a possibility, and so they should still be separate pkgs 2017-01-04 23:59:28 (also, they're distributed separately) 2017-01-04 23:59:51 jirutka: good for you - it's okay but hillarously works better on not-Macs 2017-01-05 00:00:04 also caught a typo 2017-01-05 00:00:11 asie: beep beep, have you checked licensing for gfx/sfx? 2017-01-05 00:00:56 skrzyp: OpenGFX yes 2017-01-05 00:01:29 OpenSFX no, good point 2017-01-05 00:01:52 is there a code for Creative Commons Sampling Plus? 2017-01-05 00:02:18 of course they're freely available, but they shouldn't inherit license variable from base openttd package 2017-01-05 00:02:25 skrzyp: openttd-opengfx IS actually GPL2 2017-01-05 00:02:35 ah okay 2017-01-05 00:02:37 right 2017-01-05 00:02:40 it's mostly code 2017-01-05 00:02:44 GRF code but still 2017-01-05 00:03:28 fixed, thanks 2017-01-05 00:06:50 skrzyp: works better on not-Macs? you mean Hackintosh? 2017-01-05 00:07:05 yes 2017-01-05 00:07:22 asie: that’s surprising for me 2017-01-05 00:07:25 well 2017-01-05 00:07:31 it's not quite as nice 2017-01-05 00:07:38 first of all, you have to pick the right machine 2017-01-05 00:07:41 macOS has a limited set of drivers 2017-01-05 00:07:51 even though there are some community drivers, it still involves doing research 2017-01-05 00:07:55 second, legal issues 2017-01-05 00:07:56 asie: can you recommend some notebook with good support for Hackintosh? 2017-01-05 00:07:56 jirutka: mostly yes 2017-01-05 00:08:02 jirutka: Thinkpad X230 2017-01-05 00:08:03 jirutka: X220 has a special tutorial 2017-01-05 00:08:05 (ThinkPad) 2017-01-05 00:08:05 and Thinkpad X220 2017-01-05 00:08:09 X230 is very similar on the internals, so that too 2017-01-05 00:08:14 not exactly 2017-01-05 00:08:18 X230 is Ivy Bridge 2017-01-05 00:08:34 Sandy and Ivy Bridge are often compatible enough to have swappable CPUs between the two 2017-01-05 00:08:36 but almost evey IvyBridge machine can be customized 2017-01-05 00:08:37 with some BIOS hackery 2017-01-05 00:08:41 to load Hackintosh 2017-01-05 00:08:52 asie: I haven't needed BIOS hackery 2017-01-05 00:08:57 no i mean 2017-01-05 00:08:58 except whitelisting WiFi card 2017-01-05 00:08:59 read what i said 2017-01-05 00:09:06 "3Sandy and Ivy Bridge are often compatible enough to have swappable CPUs between the two" 2017-01-05 00:09:07 which is only one problem on thinkpads 2017-01-05 00:09:07 asie: I have MacBook Pro 2011 Early, it’s still working fine, but I really don’t know what to buy when I need to replace it, b/c the current MacBooks are… well… not devices I’d like to buy 2017-01-05 00:09:26 jirutka: Thinkpads, HP Elitebooks or Dells Latitude 2017-01-05 00:09:38 if I didn't know Apple's philosophy, I would hope they'd revert their decisions in 2017 2017-01-05 00:09:40 these 3 kinds are exceptionally praised on hackintosh communities 2017-01-05 00:09:41 but they can't 2017-01-05 00:09:57 skrzyp: my dad has Dell Latitude, already a second model, and I really don’t like these notebooks 2017-01-05 00:10:03 generally, business laptops are probably the closest thing you can get in terms of quality 2017-01-05 00:10:07 so thinkpads and elitebooks especially 2017-01-05 00:10:14 Thinkpads looks fine, I had R60 many years ago 2017-01-05 00:10:25 jirutka: keep in mind you'll need some tinkering on the first 2 days, coding DSDT tables on the worst case 2017-01-05 00:10:26 you won't get a display this good, but moving on from a 2011 MacBook means it should still be a quality improvement 2017-01-05 00:10:29 but after that it's wine 2017-01-05 00:10:32 >wine 2017-01-05 00:10:34 that doesn't imply fine 2017-01-05 00:10:35 heh 2017-01-05 00:10:39 fine* 2017-01-05 00:10:59 right now you don't actually need (and please avoid) modifying MacOS system partition 2017-01-05 00:11:12 so there are no problems with sleep mode, WiFi card, touchpad etc? 2017-01-05 00:11:16 everything is done via injecting Kexts (kernel extensions, modules) from bootloader 2017-01-05 00:11:21 and DSDT tables 2017-01-05 00:11:35 jirutka: everything above is provided by 3rd party extensions 2017-01-05 00:11:45 but does it work really reliably? 2017-01-05 00:12:01 it is, with good DSDT patching it worked for me even better than on Linus 2017-01-05 00:12:06 in terms of battery saving 2017-01-05 00:12:22 but I had some issues with waking up before I upgraded to newest BIOS version 2017-01-05 00:12:28 and hacked DSDT table routines 2017-01-05 00:12:29 jirutka: generally, once you set it up, yes, aside from the problem of updates 2017-01-05 00:12:35 the yearly bigger update can sometimes break things 2017-01-05 00:12:45 asie: updates are not an issue right now 2017-01-05 00:12:55 I’m used to that MBP is perfectly stable, no crashes or anything like that 2017-01-05 00:12:55 because your /System isn't touched 2017-01-05 00:13:12 but of course it's better to wait some months after release to update Kexts 2017-01-05 00:13:27 I’ve read about Hackintosh few years ago and it didn’t look like perfectly stable solution 2017-01-05 00:13:31 it's not perfectly stable 2017-01-05 00:13:37 it's perfectly stable in the same way linux is, really 2017-01-05 00:13:37 jirutka: the only crash I had was when I plugged Samsung (xD) phone and triggered HoRNDIS kext 2017-01-05 00:13:50 beyond that, it was perfectly stable for me 2017-01-05 00:14:04 unless I was trying to intentionally crash the Mach kernel 2017-01-05 00:14:39 jirutka: but not that macOS won't work on Coreboot 2017-01-05 00:14:46 generally, the less you deviate from hardware apple puts in their devices, the stabler it will be 2017-01-05 00:15:08 I know there's large Coreboot userbase in Czech Republic, they even have own events and confrences 2017-01-05 00:15:49 asie: but sometimes it's not true, for example they found a way to run GTX 1080/1070 cards w/ full acceleration 2017-01-05 00:15:56 which are of course not supported by apple 2017-01-05 00:16:01 but on desktop machines 2017-01-05 00:16:13 on laptops, avoid anything except Intel SoC GPUs 2017-01-05 00:19:16 jirutka: I can somewhat assist you if you decide to buy kind of supported laptop, but the best thing you can do is reading RehabMan's tutorials on tonymacx86 and browse his GitHub 2017-01-05 00:19:29 also, remember to NOT use so called "distributions" 2017-01-05 00:19:47 skrzyp: distributions? 2017-01-05 00:19:55 customized macOS images with added questionable kexts 2017-01-05 00:19:59 usually very outdated 2017-01-05 00:20:06 and you can't trust them 2017-01-05 00:20:45 jirutka: just download macOS Sierra.app from appstore right now, plug your USB stick, do /Applications/macOS Sierra.app/.../createinstallmedia /dev/disk1 and wait 2017-01-05 00:21:17 createinstallmediaa is a tool on their macOS "app" (a folder with installer an image) which "burns" InstallESD.dmg on USB stick 2017-01-05 00:21:30 I’m too paranoid to download any modified macOS image ;) 2017-01-05 00:21:49 after that you need to put clover (do not use installed, especially on Mac) by just copying EFI folder from their distribution 2017-01-05 00:21:57 actually I’m not very comfortable even with loading unknown kexts to kernel 2017-01-05 00:21:58 clover is a bootloader for macOS on non-Macs 2017-01-05 00:22:12 these kexts are open sourced 2017-01-05 00:22:16 and code is signed 2017-01-05 00:22:34 almost every kext you need is on RehabMan's github 2017-01-05 00:22:47 along with MaciASL (DSDT editor) and some tools 2017-01-05 00:23:57 hmm 2017-01-05 00:25:02 good to you for having a mac 2017-01-05 00:25:44 without that you'll be pushed to spawn up an macOS image in QEMU (obtained by some weird way, I won't explain that publicly) or beg someone with mac 2017-01-05 00:42:12 or could use https://github.com/kholia/OSX-KVM 2017-01-05 00:42:25 if one were inclined 2017-01-05 00:43:43 mitchty: but still 2017-01-05 00:43:50 mitchty: you need an image 2017-01-05 00:44:02 yep 2017-01-05 00:44:03 I said that previously, QEMU is a KVM frontend 2017-01-05 00:44:39 thats more if anyone wants to try doing it, bit more useful example :) 2017-01-05 00:44:50 asie: btw. Apple's EULA is overrriden by local law, at least Polish 2017-01-05 00:46:10 mitchty: but in VirtualBox we can even skip the Enoch/Clover just by modifying some EFI variables and adding fake PCI devices using VboxManage 2017-01-05 00:48:05 skrzyp: sure, i try to avoid oracle stuff if i can in general 2017-01-05 00:48:36 you can do the same in qemu though 2017-01-05 00:48:42 but it's not covered by this tutorial 2017-01-05 00:48:52 http://www.contrib.andrew.cmu.edu/%7Esomlo/OSXKVM/ 2017-01-05 00:48:55 check there 2017-01-05 00:52:47 mitchty: me too, I build OpenJDK myself to avoid installing Oracle JDK on macOS and I installed VirtualBox after hours of experimenting with xhyve… 2017-01-05 00:54:14 nice 2017-01-05 00:54:22 but openjdk was available from Homebrew? 2017-01-05 00:54:24 wasn't it? 2017-01-05 00:54:33 xhyve is still experimental though 2017-01-05 00:54:50 I used Parallels on macOS which even integrates with Vagrant better than Virtualbox 2017-01-05 00:54:52 jirutka: same, xhyve is great, just no grub boot support got old fast 2017-01-05 00:55:20 i use nix on osx instead of homebrew, not sure on java stuff as i never really use java, it might be there 2017-01-05 00:56:05 from what I heard, nix on osx is using only "unstable" binaries, and support for it is a bit picky 2017-01-05 00:56:26 the bad thing is that xhyve is not maintained, kinda… Docker comp. forked it, but I really don’t trust docker… 2017-01-05 00:56:29 its not that bad, not that it isn't finicky 2017-01-05 00:56:41 homebrew does not provide openjdk 2017-01-05 00:57:32 but docker uses alpine omg omg omg 2017-01-05 00:58:37 and openjdk is in nix at least https://gist.github.com/mitchty/e3d6daa861a43c74cfb1936a8b80f96b 2017-01-05 00:59:25 thats on a sierra system as a note 2017-01-05 01:04:21 IT'S A SIERRA SYSTEM 2017-01-05 01:06:18 well this is getting ot but yeah, getting a sierra box working has been fun, the linker changes in 10.12 are rather onerous 2017-01-05 01:07:54 wasn't that made to support Swift 3 external modules? 2017-01-05 01:08:23 like, well, linking C/anything binary library to Swift code and calling its functions directly like a class 2017-01-05 01:08:50 that also makes me think seriously about writing xmp-based player for iOS 2017-01-05 01:09:07 asie would be amazed 2017-01-05 01:09:12 if i had something with iOS 2017-01-05 01:09:19 i probably would 2017-01-05 01:12:24 what happened to your eye pad? 2017-01-05 01:14:04 eye sold it 2017-01-05 01:15:48 because u-pad-ł? 2017-01-05 01:15:57 no 2017-01-05 01:16:02 because i no longer needed it 2017-01-05 01:28:37 jirutka: btw. don't you need to verify asie's PR after he fixed the changed, to make it pass by GH? 2017-01-05 01:39:01 skrzyp: yes, but tomorrow, I’m too tired now 2017-01-05 01:39:38 not a problem 2017-01-05 01:39:40 just asking 2017-01-05 06:48:04 ncopa: I think the Alpine docs are almost useless, I had to actually read initrd's /init to figure out which boot options are needed to spawn up working LUKS rootfs 2017-01-05 07:19:56 for i in `cat /etc/apk/world| grep -v "\-doc" ` ; do apk search -q ${i}-doc ; done | xargs sudo apk add 2017-01-05 07:20:02 I'm going to die for this some day 2017-01-05 08:03:23 skrzyp, no need to do that 2017-01-05 08:03:36 skarnet, just do "apk add docs" 2017-01-05 08:03:41 skrzyp, ^ 2017-01-05 08:03:54 and all the -doc packages for installed packages get pulled in 2017-01-05 08:04:00 what about -dev? 2017-01-05 08:04:43 no similar setup for those, developers generally know which -dev packages they want 2017-01-05 08:04:51 okay 2017-01-05 08:04:56 thank you 2017-01-05 08:05:27 clandmeter: I think I might take maintainership on unmaintained/dosbox today's evening 2017-01-05 08:05:53 it was pulled to unmaintained because last version is from 2010, but it works well regardless of that 2017-01-05 08:06:05 I can add snapshot() if needed 2017-01-05 08:19:41 . 2017-01-05 08:19:47 : 2017-01-05 08:19:53 .:. 2017-01-05 08:20:08 .:' 2017-01-05 08:20:13 umh.. 2017-01-05 08:20:27 .:.:. <- Cisco 2017-01-05 08:20:55 that's their motd on newer switches btw 2017-01-05 08:30:57 ncopa: looks like default grsec-rbac config is fucked up: http://sprunge.us/KYiY 2017-01-05 08:53:40 is it acceptable to produce binary with changed PAX flags in APKBUILD? 2017-01-05 08:53:45 not related to previous question 2017-01-05 09:40:39 skrzyp: i havent used rbac. i am open for suggestions how to fix it 2017-01-05 09:41:04 skrzyp: yes, it is acceptable to change pax flags in apkbuild. we do so for nodejs for example 2017-01-05 09:41:11 okay 2017-01-05 09:41:36 so why RBAC is in default distributon with broken state? 2017-01-05 09:41:55 now I'm trying to understand and fix the default policy 2017-01-05 09:42:02 what package ships the default config? 2017-01-05 09:42:15 gradm? 2017-01-05 09:42:15 gradm 2017-01-05 09:42:20 but it's a kernel-side feature 2017-01-05 09:43:56 according git lot it was kaniini that did the original packaging and default config 2017-01-05 09:44:03 i havent tested it or touched it since 2017-01-05 09:44:20 git log -- testing/gradm 2017-01-05 09:44:34 i suppose it worked at that time 2017-01-05 09:45:15 skrzyp: patches are welcome ;) 2017-01-05 09:56:26 ncopa: it will be probably a kernel patch 2017-01-05 09:56:33 at least kernel config 2017-01-05 09:56:41 gimme 20 minutes 2017-01-05 10:37:39 still can't boot :( 2017-01-05 10:39:04 init gets to "loading hardware drivers", then monitor gets probed/reset, and then the monitor loses signal 2017-01-05 10:42:27 ScrumpyJack: what hw? 2017-01-05 10:42:34 i think you need the framebuffer 2017-01-05 10:42:40 ght fbcon 2017-01-05 10:42:44 the fbcon 2017-01-05 10:42:52 i'll check 2017-01-05 10:43:10 <^7heo> moin 2017-01-05 10:43:14 i think it actually boots just switches to framebuffer mode 2017-01-05 10:43:31 might also be that you an add nomodesetting to boot cmdline 2017-01-05 10:44:20 it doesn't boot no - no tcp stack loads (I don't have a serial tty to the box) 2017-01-05 10:48:39 OMG I am sooo blonde! 2017-01-05 11:03:51 ncopa: I've managed to run RBAC after modifying 2 lines in default config 2017-01-05 11:04:12 let's see what happen 2017-01-05 11:11:48 ok, so if i disable KMS, monitor works, but the graphics card is an nvidia, and nouveau requires KMS no? 2017-01-05 11:14:13 well, since 5AM I've: 2017-01-05 11:14:17 - fixed RSBAC 2017-01-05 11:14:36 - fixed booting from cryptoroot 2017-01-05 11:14:41 - fixed dosbox 2017-01-05 11:14:49 - installed and configured alpine 2017-01-05 11:17:52 - applied for a medal 2017-01-05 11:21:41 right now, we have 2 lies on alpine webpage 2017-01-05 11:22:00 1. "rootfs on ZFS" - completely broken 2017-01-05 11:22:08 2. "RSBAC support" - ha ha 2017-01-05 11:22:34 I hope that list won't grow anymore soon 2017-01-05 11:25:01 alpine says it has root on ZFS? 2017-01-05 11:25:20 isn't it legally dubious to even ship ZoL right now because CDDL + GPL in the same kernel blob? 2017-01-05 11:25:47 oh what am I saying, alpine ships firefox with official branding 2017-01-05 11:26:37 > New features and noteworthy changes 2017-01-05 11:26:40 > Support for ZFS as root 2017-01-05 11:26:57 awilfox: it's shipped as module 2017-01-05 11:27:05 not directly compiled in 2017-01-05 11:27:16 ahh 2017-01-05 11:27:16 awilfox: and you can request permission from mozilla to ship firefox 2017-01-05 11:27:17 okay 2017-01-05 11:27:20 but still 2017-01-05 11:27:29 that module is not provided with base image 2017-01-05 11:27:35 installed is not configured to use it 2017-01-05 11:27:45 mkinitfs got fix very recently 2017-01-05 11:28:13 skrzyp: the root on ZFS was a mistake 2017-01-05 11:28:30 i thought i pushed that fix so that it was included in the 3.5.0 release 2017-01-05 11:28:45 i plan to fix that for 3.5.1 2017-01-05 11:29:36 ncopa: but what about installed? 2017-01-05 11:29:54 btw. the installer says /boot on XFS isn't supported, but latest SYSLINUX supports it 2017-01-05 11:30:10 (I use grub due to full disk encryption with /boot, but still) 2017-01-05 11:30:23 I had to manually patch setup-disk 2017-01-05 11:36:37 did you send the patch? 2017-01-05 11:41:36 not yet 2017-01-05 11:41:49 does the versino of syslinux we ship support xfs? 2017-01-05 11:41:52 still setting everything up 2017-01-05 11:42:04 i have on my todolist to set up zfs on root 2017-01-05 11:42:07 to test it 2017-01-05 11:42:13 i want do that on my laptop 2017-01-05 11:42:49 if we're using latest syslinux 6.xxx then yes 2017-01-05 11:43:09 of course it does not support zfs 2017-01-05 11:43:17 so grub is needed 2017-01-05 11:46:23 i suppose you can have /boot on separate partition too 2017-01-05 11:46:45 but then you have unencrypted kernel 2017-01-05 11:47:00 right now only the MBR and bootblock are unencrypted 2017-01-05 11:47:08 and I have a copy of it on encrypted drive 2017-01-05 11:47:12 it's checked every boot 2017-01-05 11:47:24 but that solution is not ready to publish yet 2017-01-05 13:13:02 /etc/security/limiys.conf is not respected 2017-01-05 13:13:12 limits.conf 2017-01-05 13:13:22 even when I added 'shadow' package to replace busybox login by PAM-aware login 2017-01-05 13:34:23 ncopa: https://bugs.busybox.net/show_bug.cgi?id=9556 Denys is against ash avoiding glob when not needed, so busybox internal glob in ash won't be turned off anytime soon. 2017-01-05 13:50:43 to be honest I really don't like it, because in variables sometimes bigger data are stored, like some file downloaded from net for instance, so even mere idea that echo "$VAR" does anything more than print damn value of variable makes me cringe a bit... 2017-01-05 15:08:45 RSBAC is undocummented as hell 2017-01-05 15:37:04 also, having multi-call busybox is not so "secure" because you need to grant huge amount of permissions for it to cover all the applets 2017-01-05 15:39:49 same with RBAC as it resolve symlinks as the original binary 2017-01-05 15:40:34 of course 2017-01-05 15:41:00 but grsec MAC is way simpler than NSA/Linux 2017-01-05 15:43:19 oh indeed 2017-01-05 15:47:41 right now I just made a global autolearning policy 2017-01-05 15:47:50 and set it for day or two 2017-01-05 20:20:09 ncopa: didn't you had a fix for 4.8 grsec kernel with gcc 6.2 ? 2017-01-06 02:03:46 - fixed segfaulting mpv 2017-01-06 02:03:51 - updated ffmpeg and mpv 2017-01-06 04:11:36 does alpine support the concept of alternative providers/packages/commands (similar to https://linux.die.net/man/8/alternatives) in any way? 2017-01-06 04:15:52 - updated dosbox, updated i3, added i3-gaps 2017-01-06 04:15:57 but I'm still not committing 2017-01-06 04:16:16 mikeee_: I don't think so 2017-01-06 04:16:34 but you can do that trough "provides/replaces/conflicts=" statement 2017-01-06 04:18:05 hum… where would I specify that? 2017-01-06 04:18:40 https://wiki.alpinelinux.org/wiki/APKBUILD_Reference 2017-01-06 04:19:07 thx 2017-01-06 06:11:15 mikeee_, "alternatives (8)" like system does not exist currently. we are working on apk-tools update, which would have something equivalent. 2017-01-06 06:11:56 nice 2017-01-06 06:12:00 good to know 2017-01-06 06:15:18 is there any public commits about that? would love to watch the progress 2017-01-06 06:15:20 it's coming with some other major changes, so no ETA currently. my secret hope is that it'd be usable by april for next alpine branching. but let's see how it goes in real life. 2017-01-06 06:15:25 no public commits yet 2017-01-06 06:16:39 awilfox, will publish tree when it becomes somewhat usable 2017-01-06 06:17:01 ETA son 2017-01-06 06:17:04 :) 2017-01-06 06:17:06 currenty i have design on paper, and pretty much all details figured out. some code exists, but not much 2017-01-06 06:21:04 cool! keep us posted 2017-01-06 06:34:34 awilfox, right. yes. in any case apk will contain in future "mkpkg" etc. to create the packages. so it is easier to be tied in to other build system than abuild if necessary. 2017-01-06 06:34:59 i understand you use it in connection with portage 2017-01-06 06:35:58 yeah I do. improvements are welcome in that area, but really, I'm more interested in the user-facing parts of apk; i.e. what it is like for my users and I to use the thing 2017-01-06 06:36:15 having a real alternatives(8)-like system would be excellent :3 2017-01-06 06:36:37 will try to keep the most common things compatible in CLI. though we hope to clean up the 'info' and 'search' things to be more consistent. 2017-01-06 06:37:13 awilfox: just to select if ypu want busybox vi, kosheer vi, and vim? :P 2017-01-06 06:37:16 and especially the default output to be more user-friendly. add some option then to make it machine-friendly and parseable. 2017-01-06 06:37:57 fabled: if I may make a tiny request: it would be nice to have a way to print only the requested metadata with no lines about it... uh example: 2017-01-06 06:38:17 apk info -d zsh says "zsh-5.x.y description: The Z shell is a korn derivative blah blah best shell" 2017-01-06 06:38:25 yes, that's planned 2017-01-06 06:38:31 it would be cool to have a flag to just say "The Z shell..." 2017-01-06 06:38:33 cool :D 2017-01-06 06:38:41 > blah blah best shell 2017-01-06 06:38:43 that's the "machine-friendly" part of the above :) 2017-01-06 06:38:44 :D 2017-01-06 06:40:06 awilfox: just parse APKINDEX 2017-01-06 06:41:10 skrzyp: this was for a bash/zsh-completions thing.. 2017-01-06 06:41:20 skrzyp: would rather not parse APKINDEX manually in a shell completion script... 2017-01-06 06:41:34 and also that would not allow for multiple repositories 2017-01-06 06:41:59 awilfox: but apk does the same :D 2017-01-06 06:48:49 Looking for a help with Git. Let's assume that I cloned aports tree, modified many packages but not commited them yet. How can I commit each package separaterly to new branch? 2017-01-06 07:07:39 okay, sorted that out 2017-01-06 11:34:04 <_ikke_> LebedevRI: Too bad no reaction yet to your mail 2017-01-06 11:34:26 ACTION is being ripped apart 2017-01-06 11:34:59 it was sent on a bad day i suppose, i'll resend some later time 2017-01-06 11:35:52 <_ikke_> No problem 2017-01-06 11:36:28 <_ikke_> I don't expect there to be too many people that run alpine on their desktop, use darktable and are subscribed to the mailing list :P 2017-01-06 11:57:55 _ikke_: Alpine on our desktop ? Aren't we all here ? :D 2017-01-06 11:58:22 <_ikke_> coredumb: It's the combination 2017-01-06 11:59:10 to have Alpine on the desktop, we should start with Linux on the desktop 2017-01-06 11:59:18 is it that year yet? 2017-01-06 11:59:36 <_ikke_> They said 2k17 will be the year 2017-01-06 12:00:33 I don't install anything than alpine anyway 2017-01-06 12:01:26 else* 2017-01-06 14:29:51 i'm prototyping adding chroot building to abuild with bubblewrap 2017-01-06 14:30:06 seems something resembling http://sprunge.us/iIZL might just be enough 2017-01-06 14:33:39 that will be revolutionary 2017-01-06 14:33:57 awesome! 2017-01-06 14:34:55 http://bugs.alpinelinux.org/issues/625 :) 2017-01-06 14:35:38 http://bugs.alpinelinux.org/issues/801 for the abuild issue about it 2017-01-06 14:36:38 probably the oldest open bugs :) 2017-01-06 14:38:25 it's also going to disable networking while building (so 'fetch' really needs to get all sources) 2017-01-06 14:38:43 and read-only mounting most things 2017-01-06 14:38:51 so the build is really isolated 2017-01-06 14:39:38 and build stuff in an edge VM so you don't have to have edge on your development machine? 2017-01-06 14:42:55 unrelated: who should I annoy to get my mail server whitelisted again? 2017-01-06 15:37:19 ncopa: what is wrong with dl-cdn? 2017-01-06 15:37:40 it does not list 3.5 2017-01-06 17:03:14 ncopa: when developers are reviewing PRs? are there some merge windows or something? 2017-01-06 19:39:42 hi guys, hope this is the right channel. I want to use alpine on my NAS. It generally boots already, but I need a custom kernel to fully support the hw. Can anyone point me in the right direction (github etc.) as https://wiki.alpinelinux.org/wiki/Custom_Kernel is not filled with any content. 2017-01-06 19:47:44 haxon: do you have prior experience on building packages for alpine? 2017-01-06 19:50:21 no, I did not look into that. But is the kernel a apk package at all? Isn't it just copied to the USB drive etc.? 2017-01-06 19:51:34 I know kernel compiling for typical linux systems, but the modloop file is new to me 2017-01-06 19:51:55 so you have root on tmpfs? 2017-01-06 19:52:54 yes, as the NAS boots from an internel flash memory connected by USB, I wanted to keep it that way 2017-01-06 19:52:59 $ update-kernel -h 2017-01-06 19:56:57 you likely want to use the -b option 2017-01-06 19:57:34 it will install kernel+initramfs+modloop for you 2017-01-06 19:58:00 ok, great, I'll try that 2017-01-06 19:58:00 ... after building custom kernel 2017-01-07 09:07:24 hmm 2017-01-07 09:07:39 I think rxvt-unicode should have its terminfo in separate package, or ncurses-term 2017-01-07 09:08:08 because when I'm connecting to my server from urxvt, it's going crazy 2017-01-07 09:11:55 yes, many distros have terminfos for things like rxvt-unicode in separate packages, which is generally a good idea 2017-01-07 09:17:01 tl;dr subpackage 2017-01-07 09:17:05 give me 3 minutes 2017-01-07 09:27:26 vault% apk info -L rxvt-unicode-terminfo 2017-01-07 09:27:29 WARNING: Ignoring /home/skrzyp/packages/non-free/x86_64/APKINDEX.tar.gz: No such file or directory 2017-01-07 09:27:31 rxvt-unicode-terminfo-9.22-r1 contains: 2017-01-07 09:27:33 usr/share/terminfo/r/rxvt-unicode 2017-01-07 09:27:35 usr/share/terminfo/r/rxvt-unicode-256color 2017-01-07 09:38:02 ready 2017-01-07 09:38:09 700 get 2017-01-07 09:58:02 <_ikke_> would be nice if the terminfo would be part of something default 2017-01-07 09:58:10 <_ikke_> (for urxvt) 2017-01-07 09:58:17 <_ikke_> anoying to have to install it on every server I go to 2017-01-07 10:00:34 _ikke_: of course but I'm not sure if we want to fsck up our ncurses-terminfo APKBUILD with adding external urxvt tarball, making it just for terminfo files and copying them to terminfo base 2017-01-07 10:00:36 that would require to either redistribute the terminfo file with the terminfo package or download urxvt there, no? hardly an ideal soluton 2017-01-07 10:01:05 keep in mind that using plain terminfo distributed with apkbuild might break compatibility if urxvt changes something 2017-01-07 10:01:21 yeah 2017-01-07 10:01:31 the best solution would be to add rxvt-unciode-* to ncurses-terminfo on upstream 2017-01-07 10:01:35 though it's unlikely to happen as it would also break comms to any outdated version of urxvt 2017-01-07 10:01:37 's terminfo file 2017-01-07 10:01:46 or migrate our ncurses to netbsd-curses 2017-01-07 10:01:47 as such, that's only a theoretical concern 2017-01-07 10:01:53 which has tic-based terminfo database 2017-01-07 10:02:02 i don't think that can be done for every package, but at least most 2017-01-07 10:02:03 see: libedit 2017-01-07 10:02:04 apparently with urxvt support 2017-01-07 10:02:10 which is used by some, but not many 2017-01-07 10:02:18 yeah, isn't libedit using some weird ncurses-specific hacks? 2017-01-07 10:02:35 no 2017-01-07 10:03:20 <_ikke_> perhaps a group the installs the terminfo subpackage of most used terms? 2017-01-07 10:03:22 <_ikke_> that 2017-01-07 10:03:34 hmm 2017-01-07 10:03:36 hmmmm 2017-01-07 10:03:40 good idea 2017-01-07 10:04:00 terminfo as a metapackage depending on ncurses-term, rxvt-unicode-terminfo and other 2017-01-07 10:04:13 or just make ncurses-term depending on rxvt-unicode-terminfo 2017-01-07 10:04:30 that last one is ugly 2017-01-07 10:04:33 a terminfo metapackage is better 2017-01-07 10:04:45 though ncurses (non-terminfo) already bundles the most important terminfos in alpine 2017-01-07 10:04:54 so i'm not sure if most people need the ncurses terminfo package per se 2017-01-07 10:05:00 asie: ncurses-terminfo-base* 2017-01-07 10:05:39 which has removed AT&T and IBM terms becuase the only people using it on alpine might be me and you 2017-01-07 10:05:44 not me 2017-01-07 10:05:51 it also removes terms for many other terminals 2017-01-07 10:05:58 including a few i can't even find references to in google, outside of other terminfos 2017-01-07 10:09:04 asie: so this is why I prefer tic-based terminfos, because they contain description for each terminal in its header 2017-01-07 10:25:41 _ikke_: who can decide about which way we should choose? 2017-01-07 10:25:44 ncopa: ? 2017-01-07 10:25:50 clandmeter: ? 2017-01-07 10:28:36 ehh, don't ping people randomly skrzyp 2017-01-07 10:28:37 they have lives 2017-01-07 10:29:47 asie: but they're described as main developers on wiki 2017-01-07 10:30:07 so if they have lives, they could respond to pings when they're available 2017-01-07 10:30:36 they also probably read the chat 2017-01-07 10:30:45 i feel pings should be reserved for proper emergencies 2017-01-07 10:30:50 like "oh no the latest update broke EVERYTHING" 2017-01-07 10:31:55 asie: I've posted the fix to broke segfaulting on mpv by updating it and rebuilding against newer ffmpeg (also libass needs to be updated, gimme 10 minutes) and no one cares 2017-01-07 10:32:32 "no one cares" 2017-01-07 10:32:36 "@n.copa | libass brok abi in minor version?" 2017-01-07 10:33:02 apparently someone has started looking into it already 2017-01-07 11:00:03 asie: pings are ok, people don't have to respond to them instantly either. And they're good to call attention to specific matters when you have a whole night of noisy backlogs to catch up with. 2017-01-07 11:00:30 ok 2017-01-07 13:09:09 ncopa: have you done some official efforts to bring Alpine onto Scaleway, or did they just made these images on their own? 2017-01-07 14:50:05 <^7heo> ncopa: the mkinitfs fails also in manual testing... 2017-01-07 14:50:29 <^7heo> ncopa: what version is in 3.5? 2017-01-07 14:53:17 3.0.6 iirc 2017-01-07 14:53:20 but I'm not sure 2017-01-07 14:54:16 <^7heo> 3.0.6 isn't tagged in git. 2017-01-07 14:54:22 <^7heo> not sure we can have that. 2017-01-07 14:55:29 <^7heo> oh there WAS a 3.0.6 2017-01-07 14:55:32 <^7heo> but no tag... 2017-01-07 14:55:38 <^7heo> well, we need to tag that asap. 2017-01-07 14:56:04 <^7heo> ncopa: can you please tag mkinitfs a628a8f922a18a2256532b4f99183bb7e9a1fd68 with 3.0.6? 2017-01-07 14:56:29 <^7heo> also 8f8abdd89e59c212d96994fb41b63df23153e4c6 with 3.0.7? 2017-01-07 14:57:15 they are already tagged like that 2017-01-07 14:57:37 <^7heo> not in my repo =/ 2017-01-07 14:57:40 <^7heo> gosh 2017-01-07 14:58:10 <^7heo> but you're right. 2017-01-07 15:00:28 <^7heo> damn I expected -t to be default... 2017-01-07 15:00:32 <^7heo> okay, sorry for the noise. 2017-01-07 15:01:35 `git fetch --all -p` is your friend 2017-01-07 15:01:46 which I aliased as fap ;) 2017-01-07 15:02:02 <^7heo> no, it might be wasting bandwith, I'm on 4G sometimes. 2017-01-07 15:02:46 "always fap before working in your repo" 2017-01-07 15:03:12 <^7heo> that's a good rule tho. 2017-01-07 15:03:45 yet you can be amazed how often people break it... 2017-01-07 15:04:00 <^7heo> the repo, or? 2017-01-07 15:04:28 the rule, of course 2017-01-07 15:05:22 <^7heo> man git-pull says: By default, tags that point at objects that are downloaded from the remote repository are fetched and stored locally. 2017-01-07 15:05:49 <^7heo> przemoc: well, that's because a lot of us like to keep semen and keyboards separated. 2017-01-07 15:06:08 <^7heo> so I really wonder why I didn't have those tags... 2017-01-07 15:06:11 only signed tags 2017-01-07 15:06:56 sorry, not signed 2017-01-07 15:06:58 annotated 2017-01-07 15:07:06 (but they are usually signed) 2017-01-07 15:07:30 <^7heo> wait, those tags aren't annotated? 2017-01-07 15:07:45 I don't know, maybe they are, haven't tested 2017-01-07 15:07:46 <^7heo> oooh right. 2017-01-07 15:07:58 <^7heo> no no they aren't 2017-01-07 15:08:03 <^7heo> git tag -n9 doesn't show anything. 2017-01-07 15:19:47 apparently you haven't fetched them for some reason. `git show -s v3.0.7^{tag}` clearly shows that there is annotated tag and `git tag -n` affirms it 2017-01-07 17:59:08 jirutka: "IT'S REVIEW TIME!" 2017-01-07 18:00:07 skrzyp: now, review time just ended for me, it’s 19.00 here so I’m gonna to dinner and watch TV news ;) 2017-01-07 18:00:25 s/now/no/ 2017-01-07 18:00:36 jirutka: but we're in the same TZ 2017-01-07 18:00:57 aha, I don’t remember what TZ other people are :) 2017-01-07 18:00:59 jirutka: next review window wen? 2017-01-07 18:01:24 oh no i missed him 2017-01-07 18:01:55 [`] 2017-01-07 18:02:16 cities [*] 2017-01-07 18:30:54 What is the order of building miltiple APKBUILDs on single PR? 2017-01-07 18:31:02 from newest to oldest commit? 2017-01-07 18:31:26 tried to swap commit and their dates, git push -f, and it didn't helped 2017-01-07 18:31:31 <_ikke_> skrzyp: It will try to figure out the build order 2017-01-07 18:32:15 _ikke_: right now I have a PR which needs to build 2 libraries first needed by last commited APKBUILD 2017-01-07 18:32:20 https://github.com/alpinelinux/aports/pull/695 2017-01-07 18:32:50 <_ikke_> skrzyp: https://github.com/alpinelinux/aports/blob/master/.travis/build-pkgs#L30 2017-01-07 18:33:15 <_ikke_> skrzyp: That should work if you set up the correct dependencies 2017-01-07 18:33:50 these dependencies are corrent, but they're updates, not new aports 2017-01-07 18:34:40 nice shell hackery btw 2017-01-07 18:34:41 <_ikke_> It looks at changed packages 2017-01-07 18:35:06 <_ikke_> jirutka's work 2017-01-07 18:36:22 but apparently it doesn't work, when I had the libass->ffmpeg->mpv order it started to build it from bottom to top, right now it builds from top to bottom when I changed dates :D 2017-01-07 18:36:48 jirutka: tell us when you finish the TV news 2017-01-07 18:36:50 :P 2017-01-07 18:36:51 <_ikke_> It just failed 2017-01-07 18:37:10 <_ikke_> https://travis-ci.org/alpinelinux/aports/builds/189836127#L6488 2017-01-07 18:37:43 I know 2017-01-07 18:37:56 because it wants to access functions from newsest ffmpeg 2017-01-07 18:38:07 which is also pushed in that PR 2017-01-07 18:38:21 and it needs to be upgraded and rebuilt with latest libass 2017-01-07 18:38:26 which is also there 2017-01-07 18:38:27 <_ikke_> Weird, it doesn't show an error message, just that it failed 2017-01-07 18:38:52 skrzyp: if they’re in the same repository (main, community, or testing), then it sorts them by dependencies order 2017-01-07 18:39:04 https://travis-ci.org/alpinelinux/aports/builds/189836127#L671 2017-01-07 18:39:21 jirutka: mpv is in community/, libass and ffmpeg are in main/ 2017-01-07 18:40:06 skrzyp: the script uses https://github.com/alpinelinux/lua-aports to sort packages 2017-01-07 18:43:08 https://github.com/alpinelinux/lua-aports/blob/master/aports/db.lua#L254 2017-01-07 18:44:13 skrzyp: updated ffmpeg doesn’t provide /usr/bin/ffplay … why? 2017-01-07 18:44:46 s/provide/install/ 2017-01-07 18:46:11 Enabled programs: 2017-01-07 18:46:14 ffmpeg ffprobe ffserver 2017-01-07 18:46:35 libass-0.13.5 – libass.so.5.3.3, libass-0.13.6 – libass.so.9.0.0, please kill the upstream someone… 2017-01-07 18:47:06 dafuq xD 2017-01-07 18:47:34 are those changes automated? http://git.alpinelinux.org/cgit/aports/commit/main/zfs-vanilla/APKBUILD?id=67284a3ab92f22e8bfde3157fbc8fb48cdb3e39e 2017-01-07 18:47:38 I dunno, looks like they’re generating lib version randomly… 2017-01-07 18:48:07 jomatv6: dunno, but probably not 2017-01-07 18:49:36 hmm... zfs and spl seem to be the only two packages differentiating between our two kernel flavours 2017-01-07 18:51:01 I'm trying to package the wireguard kmod for alpine ... my current WIP is https://github.com/jomat/aports/tree/master/testing/wireguard 2017-01-07 18:55:32 well, I'll do it like zfs and send another aports patch :-) 2017-01-07 19:09:26 btw. 2017-01-07 19:09:37 anyone had experience with Ansible on Alpine? 2017-01-07 19:09:54 I'm getting the strange error when using "users" module: 2017-01-07 19:09:56 "msg": "Password: usermod: PAM: Authentication information cannot be recovered\n", 2017-01-07 19:10:06 but regular useradd works fine 2017-01-07 19:11:20 yes 2017-01-07 19:11:57 i think you need some packages like useradd and usermod 2017-01-07 19:12:01 aeh no 2017-01-07 19:12:04 not packages 2017-01-07 19:12:05 shadow* 2017-01-07 19:12:15 shadow@community 2017-01-07 19:12:17 and yes, my roles is adding shadow and cracklib 2017-01-07 19:12:23 before adding users 2017-01-07 19:12:29 but apparently it doesn't work 2017-01-07 19:12:34 and linux-pam is also added 2017-01-07 19:12:40 and then you still need /etc/pam.d/useradd and usermod correct 2017-01-07 19:13:23 the pam.d structure on Alpine is kinda weird 2017-01-07 19:13:37 ls /etc/pam.d/ 2017-01-07 19:13:38 base-account base-session other 2017-01-07 19:13:39 https://0.jmt.gr/?ad2de40f507d7b20#ry2mPxOKl1H4WYtvpPV6ntJ9ErYbQbzqrsxxHWK+UBk= 2017-01-07 19:13:41 base-auth base-session-noninteractive 2017-01-07 19:13:43 base-password login 2017-01-07 19:14:28 wekk 2017-01-07 19:14:33 well* 2017-01-07 19:14:48 so why it's not included in linux-pam package? 2017-01-07 19:15:06 nobody files a bug 2017-01-07 19:15:50 what happen if I do? 2017-01-07 19:15:57 it'll be fixed 2017-01-07 19:16:00 probably 2017-01-07 19:16:23 the apk-tools bug with noexec on cache is still unresolved though 2017-01-07 19:21:20 https://bugs.alpinelinux.org/issues/6642 2017-01-07 19:22:30 fuck 2017-01-07 19:22:33 wrong project 2017-01-07 19:22:50 thanks :-) 2017-01-07 19:25:24 https://bugs.alpinelinux.org/issues/6643 - proper one 2017-01-07 19:25:33 i'm not able to remove the previous one, but left a note 2017-01-07 19:27:23 Ansible is piece of crap, it’s naive to expect that it will work on non-gnu system 2017-01-07 19:28:32 i see it as pssh on steroids... :-) 2017-01-07 19:29:01 jirutka: what can you recommend as an alternative? I think ansible is the best (or at least most UNIX way) from the holy configuration management triad (ansible-chef-puppet) 2017-01-07 19:29:18 have you ever tried salt? 2017-01-07 19:29:29 but I'm getting paid for making Ansible work, so my opinion might be biased 2017-01-07 19:30:04 jomatv6: not yet 2017-01-07 19:30:55 i'm getting paid for making things work... ansible helps me with this task and supplies me with lots of free time :-) 2017-01-07 19:31:26 sort of 2017-01-07 19:31:44 skrzyp: I don’t know about any sane alternative; I’d recommend to write plain shell scripts with some good utils, conventions and discipline and use some tools for transfering them and running to multiple systems 2017-01-07 19:31:44 jirutka - continuous integration does not seem to be working - can you pull ffmpeg please so xpra will build ? 2017-01-07 19:32:13 but my CEO is probably only one person in this country who uses Ansible on production and announces that publicly on various conventions and conferences 2017-01-07 19:32:27 so I'm sometimes working more with ansible than with actual issues 2017-01-07 19:32:51 jirutka: hurr durr indempotence :) 2017-01-07 19:33:20 BitL0G1c: I did ffmpeg update PR too 2017-01-07 19:34:18 aha - must be the cause 2017-01-07 19:34:37 jirutka: "Salt is sort of halfway between Puppet and Ansible – it supports Python, but also forces users to write all CLI commands in either Python, or the custom DSL called PyDSL. It uses a master server and deployed agents called minions to control and communicate with the target servers, but this is implemented using the ZeroMq messaging lib at the transport laye" - WHAAAAAAAAT 2017-01-07 19:35:21 my flatmate loves salt 2017-01-07 19:35:34 skrzyp: I have some ideas how to implement better CFM tool, but currently I don’t have strong enough motivation for it; I manage dozens of (mostly virtual) systems, but don’t have many systems in the same configuration; moreover I’m currently moving services to LXC containers with Alpine and these are very simple 2017-01-07 19:36:14 skrzyp: when I need to deploy some more complicated software, I just write an abuild instead of YAML+Jinja2 mess 2017-01-07 19:36:42 so the only left are just plain configuration files which I can edit by hand 2017-01-07 19:36:46 jomatv6: so he's being salty? 2017-01-07 19:36:49 :P 2017-01-07 19:36:56 yep :-) 2017-01-07 19:37:05 jirutka: abuild for configuration deployment? 2017-01-07 19:37:07 o.0 2017-01-07 19:37:18 Ansible’s indempotence is mostly an illusion 2017-01-07 19:37:26 and you can do the same even with plain shell scripts 2017-01-07 19:37:37 ofc when you know shell programming… 2017-01-07 19:37:52 but you should always make sure if you really need to do that action 2017-01-07 19:37:56 without actually doing it 2017-01-07 19:38:06 for example, this is how I deploy GitLab https://github.com/jirutka/user-aports/tree/v3.4/bundles/gitlab-ce 2017-01-07 19:38:32 skrzyp: abuild for deployment, not configuration deployment 2017-01-07 19:39:10 btw I wrote all these roles myself https://github.com/gentoo-ansible 2017-01-07 19:39:14 my network setup is quite complex... with lots of vpn tunnels bgp peers and ibgp between the machines... thats awesome to have it templated instead of touching dozens of config files by hand :-) 2017-01-07 19:39:35 I’ve been using Ansible for a year, wasted enormous time with it until I finally realized that it’s completely wrong approach 2017-01-07 19:39:45 jirutka: that kind of setup is quite nice if you're working with IT-related company 2017-01-07 19:40:21 jirutka: but we're supporting 3 different companies - marketing/mailing stuff, webdev and some others, they just want to get their shut fixed 2017-01-07 19:40:47 and Ansible was proven to be the best way in that kind of environments 2017-01-07 19:41:18 jomatv6: then I’d write a special purpose software to manage it – if it’s simple enough, then just shell scripts, if it’s more complex, then in real programming lang, Lua, Ruby or maybe Python are good for that 2017-01-07 19:42:14 then again, if i'd use python, i'd probably also use jinja, and then why not ansible from the beginning? :-D 2017-01-07 19:42:32 Jinja2 is part of the problem… 2017-01-07 19:42:36 it’s really not good 2017-01-07 19:42:39 but I think you might want to separate information from code 2017-01-07 19:42:47 jirutka: the thing is, there's an illusion that deploying pre-made software (which includes configuring it etc.) is easier and/or faster than writing stuff yourself, even when said stuff is relatively simple 2017-01-07 19:42:50 like network setup 2017-01-07 19:43:11 so it might also be a part of documentation 2017-01-07 19:43:25 so, decision makers will always prefer it if you can say "we use X or Y" instead of "it's a set of homemade Lua scripts" 2017-01-07 19:43:30 skrzyp: Ansible is “scripting” in YAML (data serialization format) and Jinja2 (templating lang for HTML)… this is completly wrong and source all of the problems with Ansible 2017-01-07 19:43:48 even when rewriting the whole mess would actually be faster than making the existing package work 2017-01-07 19:44:03 it’s very limited, very verbose and it’s extremely hard to reason what it does 2017-01-07 19:44:14 jirutka: I was not the person who picked that for our company 2017-01-07 19:44:20 don't blame me for using it :< 2017-01-07 19:44:31 even after a year of writing ansible roles I cannot be sure what this piece of yaml/jinja will do 2017-01-07 19:44:51 The fallacy is also widespread among developers: look at the number of packages using autoconf, despite it bringing more problems than it solves. 2017-01-07 19:45:18 skarnet: but the ansible code can be understood at least 2017-01-07 19:45:19 there are about 5 places when you can define variables and it’s totally non-intuitive what has precedence when in specific situations 2017-01-07 19:45:26 I mean, ansible's sources 2017-01-07 19:45:44 while autoconf poops to Makefiles 2017-01-07 19:46:08 ansible’s sources? you mean YAML or Ansible code-base? 2017-01-07 19:46:13 jirutka: yaml + jinja2 is the centerpiece for a lot of tools other than ansible that don't suffer from all the problems ansible have ... 2017-01-07 19:46:23 Ansible code-base was really mess before 2.0 (I don’t know 2.0+) 2017-01-07 19:46:24 ansible is just bad ... period 2017-01-07 19:46:36 jirutka: I stared from 2.0 2017-01-07 19:46:45 the fact they use yaml and jinja is irrelevant in this case 2017-01-07 19:46:57 I should write an article about it 2017-01-07 19:47:23 I can talk 1+ hours of why Ansible is bad, but don’t have time for it know 2017-01-07 19:47:43 also it makes me sad, when I remember how much time I wasted on it 2017-01-07 19:47:56 PRs queue is growing :) 2017-01-07 19:48:04 and I have some others uncommited 2017-01-07 19:48:11 jirutka: just in case, will you be at FOSDEM? 2017-01-07 19:48:19 and why I was so stupid… scripting in YAML + Jinja2 should be enough reason why to avoid it, and yet I didn’t 2017-01-07 19:48:44 skarnet: yes, if I manage to arrange all the stuff in time, then I’ll be 2017-01-07 19:48:55 skarnet: http://n-gate.com/fosdem/2016/12/29/0/ 2017-01-07 19:49:12 cool, we'll meet there then :) 2017-01-07 19:49:19 skarnet: but I don’t think that I can talk about it, I don’t speak English very well :/ 2017-01-07 19:50:03 it's ok, I'll speak and you'll listen :D 2017-01-07 19:50:11 agree :) 2017-01-07 19:50:12 i still need to cleanup my package for ungoogled-chromium 2017-01-07 19:50:42 skrzyp: yeah, it's easier to criticize than to actually organize shit 2017-01-07 19:51:24 skarnet: but, you’re French… I’m kinda afraid that you may have even worse pronunciation than me :P 2017-01-07 19:51:57 so you won't be able to pass the information to each other 2017-01-07 19:52:02 that's okay too 2017-01-07 19:52:42 skarnet: any opinion about https://en.wikipedia.org/wiki/Tiny_C_Compiler ? 2017-01-07 19:52:47 jirutka: aï ave no aïdee wat yoo arrre taling aboot 2017-01-07 19:52:54 XD 2017-01-07 19:53:25 XD 2017-01-07 19:54:03 tcc? very cool toy, yet another demonstration of how brilliant Fabrice Bellard is, but don't expect to do serious work with it, because it makes a lot of standard-breaking assumptions :) 2017-01-07 19:54:18 I think asie could win at accent with everyone in this channel 2017-01-07 19:56:09 could please someone look at https://bugs.alpinelinux.org/issues/6635 ? 2017-01-07 19:57:00 smells like a DNS freeze 2017-01-07 19:57:23 just like sendmail 15 years ago 2017-01-07 19:57:30 yes, but why since v3.5? 2017-01-07 19:58:26 I have no idea. Regression that nobody caught because all the sane people are using dropbear anyway. ;) 2017-01-07 19:58:45 btw I’d like to meet the person who designed here-document behaviour in the shell… and smash his face on keyboard… it’s impossible to parse it without hackery 2017-01-07 19:59:11 are you writing a shell? 2017-01-07 19:59:38 this `cat < yes 2017-01-07 20:00:19 no, I’m writing parser for POSIX sh (in Lua using lpeg), so I can programmatically modify APKBUILDs 2017-01-07 20:00:56 ooooh, nice rabbit hole you're exploring 2017-01-07 20:01:02 yeah… 2017-01-07 20:01:31 some parts of POSIX sh are really insane 2017-01-07 20:01:41 tell me something I don't know 2017-01-07 20:01:44 XD 2017-01-07 20:02:12 I suggest forgetting about modifying deep shell structures, and sticking to parsing and modifying key=value assignments :P 2017-01-07 20:02:19 and maybe simple functions if you have to 2017-01-07 20:02:46 that’s not sufficient, read some APKBUILDs… 2017-01-07 20:03:09 why do you need to modify them programmatically? 2017-01-07 20:03:18 APKBUILDs were designed to be a human interface 2017-01-07 20:03:47 if you need automation on them, define some stricter syntax for automatable APKBUILDs 2017-01-07 20:04:06 and have your automation expect that stricter syntax 2017-01-07 20:04:19 so we can semi-automatically update rust/python/lua/whatever abuilds with automatic dependency check 2017-01-07 20:04:22 and blow up noisily when an APKBUILD doesn't meet it when supposed to 2017-01-07 20:04:41 that’s an option, but when I write sh parser, I want to do it properly 2017-01-07 20:04:53 that's my point: a sh parser is too much work 2017-01-07 20:04:53 it’s a challenge ;) 2017-01-07 20:05:22 bonus points: code sh parser in 4096 binary 2017-01-07 20:05:26 4096 bytes* 2017-01-07 20:05:39 I have already working implementation, just few things missing (e.g. this insane here-doc) and need to design proper AST 2017-01-07 20:05:43 if you like a challenge, I have a hundred projects more interesting than a sh parser in lua 2017-01-07 20:06:05 I’m not writing it by hand, I’m using lpeg that makes it quite easy 2017-01-07 20:06:52 and don’t care about error recovering, if the script is invalid, then it will just fail, no message where is the error 2017-01-07 20:07:03 b/c this is not needed for the main use case 2017-01-07 20:07:42 this also simplifies it quite a lot 2017-01-07 20:07:55 just print an angry line and fail when you encounter a heredoc in an APKBUILD 2017-01-07 20:08:06 here-docs are very useful 2017-01-07 20:08:21 yes they are, and they're also a big special case 2017-01-07 20:08:42 but I may ban writing insanity like `< that won't be enough, quoting rules are also weird inside heredocs 2017-01-07 20:09:36 I’ll probably do, b/c I have no idea how to properly deal with that and don’t wanna waste too much time on it 2017-01-07 20:09:53 fortunately I don’t need to interpret 2017-01-07 20:10:26 "you want me to handle your APKBUILD automatically? then don't use heredocs you n00b, or it's all manual abump for you" 2017-01-07 20:10:53 again, here-docs are very useful and I use them very often, so I definitely don’t wanna ban them completely 2017-01-07 20:12:24 parsing is hard 2017-01-07 20:12:26 actually, what I really need is just modify content of top-level variables, at least for now, but it’s more interesting to write complete parser… yes, I’m kinda insane… 2017-01-07 20:12:51 and I really wanted to try lpeg :) 2017-01-07 20:13:19 also it’s a practice before writing parser of AsciiDoc, probably in Rust :) 2017-01-07 20:13:46 libs_private: [('L', '/usr/local/opt/openssl/lib'), ('L', '/usr/local/Cellar/qt/4.8.7_1/lib'), ('l', 'z'), ('l', 'm -framework ApplicationServices -framework CoreFoundation -framework Security')] 2017-01-07 20:13:48 parsing is hard 2017-01-07 20:14:27 this is from some Homebrew recipe? 2017-01-07 20:14:39 no 2017-01-07 20:14:48 that's pkgconf's quality work on OS X 2017-01-07 20:14:55 aha 2017-01-07 20:15:16 what's odd is 2017-01-07 20:15:21 -framework Framework 2017-01-07 20:15:23 that's actually the correct output 2017-01-07 20:15:32 when printed as a list of arguments 2017-01-07 20:15:43 but really, writing parser for CF grammar in classic way (bottom-up) really sucks, but PEG is fun ;) 2017-01-07 20:15:44 but clearly not the right memory representation 2017-01-07 20:15:49 it’s more intuitive 2017-01-07 20:22:18 inb4 EBNF 2017-01-07 20:24:56 inb4? 2017-01-07 20:25:09 "in before". 2017-01-07 20:25:19 aha 2017-01-07 20:26:35 but it's a bit wider meaning 2017-01-07 20:27:02 jp2gmd 2017-01-07 20:27:14 oops, wrong window. 2017-01-07 20:27:46 jpg2md* 2017-01-07 20:28:19 :| 2017-01-07 21:53:44 ACTION uses augeas to modify APKBUILDs programmatically 2017-01-08 08:12:21 what is augeas? 2017-01-08 08:15:41 <_ikke_> http://augeas.net/ 2017-01-08 08:50:15 jirutka: ffmpeg's ffplay requires SDL2 now 2017-01-08 09:02:53 jirutka: also, you're conflicting me now :P 2017-01-08 11:44:36 skrzyp: you must to rebase it… 2017-01-08 11:49:35 what exactly? 2017-01-08 11:51:35 don’t worry, I already did it, but I’ve tried GitHub new supoort for resolving conflicts, so it created merge commit instead of rebasing 2017-01-08 11:51:42 I’ll rebase it before merging 2017-01-08 11:52:08 eww 2017-01-08 11:52:26 it doesn't sound like a good solution 2017-01-08 11:52:28 but okay 2017-01-08 11:52:39 at least it makes "dirty" commits 2017-01-08 11:53:30 yes, i don’t like it either 2017-01-08 11:53:47 but as I said, I’ll rebase it and get rid of this merge commit before merging it into master 2017-01-08 11:53:55 okay 2017-01-08 11:54:03 what about those other new aports? 2017-01-08 11:54:34 what aports? 2017-01-08 11:56:15 kakoune, i3wm-gaps, rxvt-unicode-terminfo, dosbox 2017-01-08 12:23:50 I’ll look at them later 2017-01-08 12:24:18 I’ve just fixed travis script to build repositories in defined order (main > community > testing) 2017-01-08 13:12:32 jirutka: i3wm-gaps fixed 2017-01-08 13:14:14 hmm 2017-01-08 13:14:44 why github does not connect my email with account in commits? 2017-01-08 13:15:41 skrzyp: probably b/c you’ve used another email address that is not registered on GH 2017-01-08 13:15:57 jirutka: now I see 2017-01-08 13:16:05 vault% cat ~/.gitconfig 2017-01-08 13:16:07 [user] 2017-01-08 13:16:09 email = j.skrzypnik@openmailbox.orxg 2017-01-08 13:16:11 > orxg 2017-01-08 13:16:13 ;_; 2017-01-08 13:17:49 could you fix by `git commit --amend --author="Jakub Skrzypnik "` on that 3 commits? 2017-01-08 13:38:48 every Alpine contributor should change their first name to Jakub, for clarity 2017-01-08 13:39:05 ^ this 2017-01-08 13:50:11 kek 2017-01-08 13:57:02 ale kurwa 2017-01-08 13:57:15 wkurwia mnie że ncopa korzysta z pohakowanego newapkbuild 2017-01-08 13:57:23 i nie czyści tych konstrukcji automatycznych 2017-01-08 13:57:33 no bo po chuj w metapakiecie iteracja po niestniejących patchach 2017-01-08 13:57:35 xDD 2017-01-08 14:08:36 behave yourself, skrzyp. it's not your backyard 2017-01-08 14:09:12 przemoc: oh 2017-01-08 14:09:26 przemoc: that was meant to be send to other buffer 2017-01-08 14:09:52 przemoc: but still, the thing is real 2017-01-08 14:10:21 using patch-loops in metapackages actually clutters the APKBUILD 2017-01-08 14:10:38 and slows down the buildservers for a bit 2017-01-08 14:11:04 also, having unused/uninitialized variables also makes some APKBUILDs too large 2017-01-08 14:12:31 przemoc: also, take a look at your status window 2017-01-08 14:15:17 if you have time to do the cleanings, please do and send patches or PRs. whining brings practically nothing. 2017-01-08 14:16:25 it can be done automatically 2017-01-08 14:16:46 as someone was hyping for programatically modyfing aports yesterday 2017-01-08 14:17:40 jirutka wrote he's writing some stuff, but it's still in the works from what I understood quickly going through backlog 2017-01-08 14:41:09 skrzyp: you will burn in the hell. 2017-01-08 14:41:39 I'm already burned out 2017-01-08 14:41:41 so I don't care 2017-01-08 14:53:01 I'll take Polish over Dutch any day 2017-01-08 14:56:52 jirutka - can you pull nginx-naxsi please ? 2017-01-08 15:50:43 BitL0G1c: have you fixed these review complains? 2017-01-08 16:22:58 yes 2017-01-08 16:23:48 skrzyp - if I remove any more from module() it doesn't build 2017-01-08 16:27:20 not disabling the pax protections dropped ram usage by 10% (perhaps it closed an exploit) 2017-01-09 13:09:30 clandmeter: hi dude, you about? 2017-01-09 18:06:33 can nginx-naxsi be pulled please ? - enabling the pax protections closes a security hole 2017-01-09 18:13:09 stop being such a naxsi about it 2017-01-09 19:38:57 Hello... 2017-01-09 19:39:23 I am thinking about packaging the yarn package manager (for nodejs) 2017-01-09 19:40:42 as there are now packages for lts and latest nodejs, should I not make this package depend on nodejs at all, or is there a way to specify alternates? 2017-01-09 20:14:37 ncopa: yo, got your email 2017-01-09 20:14:47 hi! 2017-01-09 20:14:54 thanks for showing up! 2017-01-09 20:14:59 :D 2017-01-09 20:15:11 i got a step forward 2017-01-09 20:15:34 i did zfs set mountpoint=/sysroot ... 2017-01-09 20:15:48 I wrote down a bunch of steps somewhere but I forget where I put the notes ;__; 2017-01-09 20:16:14 now i get error: cannot import '9680....292' 2017-01-09 20:17:09 so it seems that it cannot import the pool with the given uuid 2017-01-09 20:17:46 I think I mostly followed the arch guide 2017-01-09 20:17:48 https://wiki.archlinux.org/index.php/Installing_Arch_Linux_on_ZFS 2017-01-09 20:18:13 buuut, let me look at my current setup and see where the differences are 2017-01-09 20:18:29 Warning: 2017-01-09 20:18:29 Always use id names when working with ZFS, otherwise import errors will occur. 2017-01-09 20:18:39 there is my problem i think 2017-01-09 20:19:47 my grub entry looks like this 2017-01-09 20:20:00 linux /vmlinuz-grsec root=zfsroot/ROOT/default rootfstype=zfs modules=zfs quiet 2017-01-09 20:20:06 I'm not sure if modules=zfs is necessary 2017-01-09 20:20:21 it is 2017-01-09 20:20:27 (for now) 2017-01-09 20:21:22 then my zfs root volume is zfsroot/ROOT/default, obviously 2017-01-09 20:22:03 but the mountpoint for it is set to legacy, since the initrd does mounting 2017-01-09 20:22:37 aha 2017-01-09 20:22:53 so i need to set mountpoint=legacy 2017-01-09 20:23:02 otherwise it'll mount to / on import and that's probably why you're seeing the issue with / not being empty 2017-01-09 20:23:07 ype 2017-01-09 20:23:10 er, yep 2017-01-09 20:23:13 exactly 2017-01-09 20:23:34 that was the thing i was missing 2017-01-09 20:23:43 the arch doc was very useful too 2017-01-09 20:23:47 i followed gentoo doc :) 2017-01-09 20:23:57 ah, excellent 2017-01-09 20:24:05 i didnt use disk-by-id 2017-01-09 20:24:11 i used sda5 2017-01-09 20:24:19 ohhh 2017-01-09 20:24:24 which caused the import error i think 2017-01-09 20:24:35 i suppose i have to recreate the pool then? 2017-01-09 20:24:51 and restore from backup again... 2017-01-09 20:25:00 are you doing raid or anything fancy? 2017-01-09 20:25:08 seems like there's a way to pop the drive out and replace it 2017-01-09 20:25:23 no raid, no nothing 2017-01-09 20:25:26 its a single disk 2017-01-09 20:25:38 zpool export tank; zpool import -d /dev/disk/by-id -aN ? 2017-01-09 20:26:45 did we have a /dev/disk/by-id script for mdev? 2017-01-09 20:26:46 hm, my zpool status calls return sda, sdc, sdb 2017-01-09 20:27:20 but it was created with disk id? 2017-01-09 20:27:37 perhaps not 2017-01-09 20:28:38 this box doesn't actually have a /dev/disk 2017-01-09 20:28:48 does udev make those? 2017-01-09 20:28:59 yes 2017-01-09 20:29:09 so you dont use udev 2017-01-09 20:29:11 good :) 2017-01-09 20:29:43 udev is.. running 2017-01-09 20:29:46 weird 2017-01-09 20:29:53 ok so you are using udev 2017-01-09 20:30:25 what i did notice though, i can run: zpool import tank 2017-01-09 20:30:51 i wonder if we should use the LABEL instead of UUID? 2017-01-09 20:31:34 oh, where are you specifying the uuid? 2017-01-09 20:31:36 but its interesting that it works for you 2017-01-09 20:32:08 well, your patch does so 2017-01-09 20:32:20 http://git.alpinelinux.org/cgit/mkinitfs/commit/?id=be20737c8c6ca8b4ba94302f6b6c9725bc98565d 2017-01-09 20:34:15 oh yes, that 2017-01-09 20:34:24 hm 2017-01-09 20:34:33 that seems to work actually 2017-01-09 20:38:56 like works works? 2017-01-09 20:39:10 like it works from boot usb 2017-01-09 20:40:39 ah yeah, looks like LABEL should work too 2017-01-09 20:40:55 hrm 2017-01-09 20:41:02 uuid is better 2017-01-09 20:41:10 but i still get same error 2017-01-09 20:41:26 atleast i have the keyboard driver in my initramfs now :) 2017-01-09 20:41:39 thank makes things alot easier :) 2017-01-09 20:41:56 blkid doesn't even show my zfs volumes 2017-01-09 20:42:09 apk add blkid 2017-01-09 20:42:25 to get the util-linux variant instead of busybox 2017-01-09 20:42:36 :| 2017-01-09 20:43:09 that works 2017-01-09 20:43:22 yeah, all my drives in zfs raid have the same UUID, but the UUID_SUB differs 2017-01-09 20:43:39 PARTUUID looks like a normal UUID 2017-01-09 20:44:37 so the uuid is probably correct 2017-01-09 20:46:28 ok 2017-01-09 20:46:31 seems like a race 2017-01-09 20:46:52 when i re-run it, i get error that zpool already exist 2017-01-09 20:47:11 and i can mount -t zfs tank/alpine/root /sysroot 2017-01-09 20:47:52 so i kinda have a working desktop again :) 2017-01-09 20:48:32 Klowner: big thanks 2017-01-09 20:48:36 awesome 2017-01-09 20:48:44 i think i will be able to figure out the rest now 2017-01-09 20:48:51 otherwise i will ping you again ;) 2017-01-09 20:48:53 you are most welcome 2017-01-09 20:48:58 kk, I'll be around, just holler 2017-01-09 20:49:05 my internet has been stupid so email me if I disappear 2017-01-09 20:49:06 i think your patch is correct to use uuid 2017-01-09 20:49:16 but i think we will do the symlinke mtab unconditional 2017-01-09 20:49:30 my patch has some issues with my set up since I have multiple drives in a pool 2017-01-09 20:49:39 that way it will work without roofsttype=zfs 2017-01-09 20:49:45 importing one imports them all, but it seems to work fine 2017-01-09 20:49:55 i think that is kind of expected 2017-01-09 20:50:05 and tbh, i think that is what makes it work for you 2017-01-09 20:50:09 i think there is a race 2017-01-09 20:50:30 i will investigate 2017-01-09 20:50:35 and try polish it a bit 2017-01-09 20:50:45 and try have it acutally working for 3.5.1 release 2017-01-09 20:50:58 big thanks! and good night 2017-01-09 20:51:11 i'll continue tomorrow 2017-01-09 20:52:19 very welcome, good luck! 2017-01-09 20:59:38 good news, it works as in works works as soon as i replace the old root=UUID= with root=tank/alpine/root 2017-01-09 20:59:59 now i go for real :) 2017-01-09 21:03:20 I was trying to build performance co-pilot, but man.. down the rabbit hole on package dependencies 2017-01-09 21:12:54 allllmost got elfutils to build 2017-01-09 21:44:07 I also wish there were a simpler pattern for dealing with python2/python3 packages that don't use distutils/setuptools :/ 2017-01-10 09:04:23 guys 2017-01-10 09:04:34 does `build-base` apk contain `libstdc++` ? 2017-01-10 09:04:43 no 2017-01-10 09:05:00 well 2017-01-10 09:05:06 nowadays it might indirectly 2017-01-10 09:05:29 build-base depends on g++ now (since gcc 6) 2017-01-10 09:05:34 and g++ depends on libstdc++ 2017-01-10 09:06:33 ah got them depends="binutils gcc make libc-dev fortify-headers g++" 2017-01-10 09:06:47 ty fabled 2017-01-10 09:07:08 also wanted to ask 2017-01-10 09:07:19 why has py-scipy=0.17.0-r0 been flagged? 2017-01-10 09:08:59 SciPy 0.18.1 was released on 2016-09-19. https://www.scipy.org/ 2017-01-10 09:09:03 https://pkgs.alpinelinux.org/flagged?origin=py-scipy&branch=&repo=&maintainer=# 2017-01-10 09:09:08 see click the message icon 2017-01-10 09:09:41 and I can install it via `apk add py-scipy@testing ? 2017-01-10 09:10:04 tyvm ^^ 2017-01-10 09:10:29 flagging is note for the package maintainer that there's new upstream version and that our binary package is outdated 2017-01-10 09:30:26 ah cool :) 2017-01-10 09:30:37 I thought it had some vuln 2017-01-10 10:07:30 question: is the enabling of CONFIG_GRKERNSEC_SYSFS_RESTRICT going to be kept? it breaks many desktop/laptop applications, even things as simple as battery monitoring 2017-01-10 10:08:37 asie: just configure your rsbac lol 2017-01-10 10:09:00 it overrides this setting and others 2017-01-10 10:09:04 skrzyp: how? 2017-01-10 10:09:09 i know sysctls exist for many settings, but not his 2017-01-10 10:09:11 this* 2017-01-10 10:09:12 gradm 2017-01-10 10:09:30 i'll look into it, thanks 2017-01-10 10:09:38 also 2017-01-10 10:09:40 grsec wikibooks 2017-01-10 10:11:26 asie: but I think Alpine devs don't know too much about grsec too, I suppose they've added it just for respect for "muh secure" distro 2017-01-10 10:12:03 every package should add its own default rsbac policy 2017-01-10 10:12:20 right now RSBAC is disabled by default, and grsec settings are lowered to minimum 2017-01-10 10:13:07 but IT news sites are fapping to "most secure distro" which Alpine might be, but not right now 2017-01-10 10:14:41 much hype 2017-01-10 10:15:00 kind of 2017-01-10 10:15:15 just make some alpine-related searches 2017-01-10 10:15:23 mostly Google 2017-01-10 10:15:26 and you'll see 2017-01-10 10:18:54 i admit that i havent used rsbac 2017-01-10 10:19:01 and i admint i dont know much about it 2017-01-10 10:20:10 but why did you pulled whole grsec without understanding it? PaX also needs to be fixed (by paxctl) for some binaries 2017-01-10 10:20:17 but i think at least a couple of the devs are not totally clueless when it comes to PaX 2017-01-10 10:20:43 i kept the rsbac there in case someone want use it 2017-01-10 10:20:49 even if i dont care much about it 2017-01-10 10:21:57 i do use the other grsec features like protecting /sys and /proc, chroot hardedning etc 2017-01-10 10:22:58 we do paxctl some binaries 2017-01-10 10:23:21 specifically things that uses JIT 2017-01-10 10:24:27 paxctl'ing is not a process, setting flags is :) 2017-01-10 10:24:53 it's like saying "i'm chowning this directory" 2017-01-10 10:27:43 maybe i am clueless with pax too then 2017-01-10 10:28:12 how is PaX needing to be fixed for some binaries? 2017-01-10 10:29:48 not that important, but I had to fix some 3rd party binaries' headers to make them work on PaX-enabled kernel, even when they are not using JIT or anything 2017-01-10 10:30:24 what did you need to fix? 2017-01-10 10:30:34 and how did you fix them 2017-01-10 10:31:09 using paxctl, but it was for unpackaged stuff 2017-01-10 10:31:19 and it's not meant to be packaged 2017-01-10 10:31:30 those we cannot do much about 2017-01-10 10:31:42 what flags did you need to set? 2017-01-10 10:31:57 but I can, just read PaX docs and it worked as intended :) 2017-01-10 10:32:03 (also, some of my new aports' PRs are still waiting) 2017-01-10 10:32:11 one challenge is if you run like docker with grsec 2017-01-10 10:32:29 and you run a docker image which needs disable pax features 2017-01-10 10:32:34 MPROTECT mostly, but I think it's not elegant solution when it can be patched 2017-01-10 10:32:41 but it's not always possible 2017-01-10 10:32:50 because with docker images you dont really have control over the filesystem 2017-01-10 10:32:52 yes 2017-01-10 10:33:24 over the filesystem inside container? 2017-01-10 10:33:33 yes 2017-01-10 10:33:49 weird 2017-01-10 10:33:59 I have no experience with docker though 2017-01-10 10:34:06 i'd like a knob to disable pax/grsec features per container 2017-01-10 10:34:31 so you for example can disable mprotect for an entire container 2017-01-10 10:34:54 i think its possible in theory with rsbac currently 2017-01-10 10:35:03 but its kind of inconvenient 2017-01-10 10:36:37 if you need disable mprotect it means the software tries to have both write and execute access to a memory page 2017-01-10 10:37:02 its normally JIT that needs that 2017-01-10 10:38:14 but I know what W^X/PaX is 2017-01-10 10:41:58 sry for little noise... 2017-01-10 10:54:57 does anybody know if there are leagal problems with shipping zfs kernel modules with alpine iso images? 2017-01-10 10:58:13 ha ha :) 2017-01-10 10:58:31 and you're asking about this NOW? 2017-01-10 10:59:03 they can be distributed as separate packages with modules, but not as being compiled directly into kernel 2017-01-10 10:59:30 not sure about same providing them on the same image as kernel 2017-01-10 10:59:33 we already ship them as separate package 2017-01-10 10:59:57 that is what my question is, what if we provide the modules in the kernel driver tree 2017-01-10 11:00:08 not compiled directly into the kernel 2017-01-10 11:00:11 but as module 2017-01-10 11:00:45 i moved my laptop over to zfs root yesterday 2017-01-10 11:01:13 but its kind of inconvenient to set it up unless you can `modprobe zfs` 2017-01-10 11:02:03 so to be able to actually do root on zfs we need to either ship zfs with the boot media or users needs to build their own boot media 2017-01-10 11:02:52 ask Ubuntu people, they're also supporting ZFS on / from base dist now 2017-01-10 11:03:21 on systemrescuecd, zfs on ISO is distributed via overlay FS on /lib/modules 2017-01-10 11:05:45 hm 2017-01-10 11:05:50 that would be an option 2017-01-10 11:06:24 ship a second squashfs image that you can overlay mount 2017-01-10 11:06:34 but I don't think it's due to license troubles, more like way to build customizable images 2017-01-10 13:54:18 fabled: a question about nlplug-findfs. do we need to mount and scan the device if -a (apkovl) or -b (bootrepo) is not specified? 2017-01-10 13:54:42 probably not 2017-01-10 13:55:23 iirc we didnt do that earlier 2017-01-10 13:55:45 i just discovered that it tried to mount 2017-01-10 13:55:46 if we do know, then it was probably an overlook 2017-01-10 13:57:19 yeah, i think we dont need mount 2017-01-10 13:57:28 append_line will just return doing nothing if its found 2017-01-10 13:57:40 ok 2017-01-10 13:57:45 i have a fix for it 2017-01-10 13:57:53 <^7heo> more commits v_v 2017-01-10 13:58:06 yup 2017-01-10 13:58:20 <^7heo> (on the other hand it's probably worth fixing things before I commit my new stuff) 2017-01-10 13:58:21 ^7heo: i discovered while running your test script 2017-01-10 13:58:25 <^7heo> ncopa: ok :) 2017-01-10 13:58:30 <^7heo> ncopa: I also have a fix afaik 2017-01-10 13:58:32 <^7heo> that's not in yet. 2017-01-10 13:58:39 <^7heo> but nothing concerning the mounting. 2017-01-10 13:58:41 i also think there are are a couple of bugs in your script 2017-01-10 13:58:48 <^7heo> very possibly. 2017-01-10 13:58:56 trying to make the test.sh work 2017-01-10 13:59:16 <^7heo> well, that would be freaking awesome. 2017-01-10 13:59:16 but it has already proved to be useful 2017-01-10 13:59:25 <^7heo> \o/ 2017-01-10 14:01:09 <^7heo> for the test.sh script, I haven't reviewed it much yet. My focus was almost entirely on nlplug-findfs 2017-01-10 14:01:27 <^7heo> but it's very good that you're using it (and potentially reviewing it too :)) 2017-01-10 14:01:33 <^7heo> more eyes, better code. 2017-01-10 14:27:12 ^7heo: does mkinitfs pass the test on github now? 2017-01-10 14:27:15 how can i check? 2017-01-10 14:29:30 <^7heo> ncopa: to check, you only have to send a PR on github or commit and trigger travis 2017-01-10 14:30:48 i suppose i need to apply this: https://github.com/alpinelinux/mkinitfs/pull/3/commits/73a1e878273d0ee8cf6709ed28ca1b9d02c7cb75 2017-01-10 14:34:13 <^7heo> I can also apply the commit myself in my PR 2017-01-10 14:34:24 <^7heo> if you git format-patch it 2017-01-10 14:34:42 <^7heo> and then I'll trigger a new travis build by updating my PR 2017-01-10 14:34:44 <^7heo> if you want 2017-01-10 14:46:56 ^7heo: the test script also revelaed fd leak in nlplug-findfs and in cryptsetup 2017-01-10 14:47:28 <^7heo> ah? 2017-01-10 14:47:31 <^7heo> ncopa: how? 2017-01-10 14:47:35 <^7heo> ncopa: I was hunting that one... 2017-01-10 14:47:41 <^7heo> ncopa: so I'm quite interested how... 2017-01-10 14:47:53 <^7heo> did you turn on the debug output and parsed through all the things? 2017-01-10 14:48:52 there was some messages like: 2017-01-10 14:48:54 File descriptor 10 (/dev/random) leaked on lvm invocation. Parent PID 23962: ./nlplug-findfs 2017-01-10 14:48:54 File descriptor 9 (/dev/urandom) leaked on lvm invocation. Parent PID 23962: ./nlplug-findfs 2017-01-10 14:48:55 File descriptor 10 (/dev/random) leaked on lvm invocation. Parent PID 23962: ./nlplug-findfs 2017-01-10 14:49:02 its actually lvm that reports it 2017-01-10 14:49:04 <^7heo> definitely. 2017-01-10 14:49:12 <^7heo> I have seen that at each boot. 2017-01-10 14:49:14 it also reported on some uevent 2017-01-10 14:49:15 <^7heo> since I can remember. 2017-01-10 14:49:23 <^7heo> I opened a bug in b.a.o for it afaik 2017-01-10 14:49:31 i have fixed the uevent thingy 2017-01-10 14:49:37 <^7heo> ok 2017-01-10 14:49:40 but the other is a bug in cryptsetup 2017-01-10 14:49:44 and i know how to fix it 2017-01-10 14:49:48 <^7heo> well, I'm really interested to have that test working 2017-01-10 14:49:54 test works 2017-01-10 14:49:54 <^7heo> and then commit all the fixes I did. 2017-01-10 14:49:56 i pushed it 2017-01-10 14:49:59 <^7heo> awesome :) 2017-01-10 14:50:03 <^7heo> I'd have to pull then. 2017-01-10 14:50:04 i think.. 2017-01-10 14:50:08 <^7heo> did you use my commit? 2017-01-10 14:50:13 <^7heo> or will have I have to reset my tree? 2017-01-10 14:50:19 i used parts of it 2017-01-10 14:50:33 i suppose you can create a backup branch 2017-01-10 14:50:44 <^7heo> I have that already ;) 2017-01-10 14:50:47 <^7heo> all my PRs are in branches 2017-01-10 14:50:49 <^7heo> it's just... 2017-01-10 14:50:49 git branch my-stuff-that-got-partially-merged 2017-01-10 14:50:55 yeab i know.. 2017-01-10 14:51:06 <^7heo> when I'll rebase... 2017-01-10 14:51:12 <^7heo> it's gonna do sh*t 2017-01-10 14:51:14 <^7heo> :P 2017-01-10 14:51:17 :D 2017-01-10 14:51:20 <^7heo> but let's see. 2017-01-10 14:51:35 <^7heo> I'm *very* happy you got the time to check out my script and fix the stuff you know about. 2017-01-10 14:51:45 <^7heo> now I can continue working much more easily. 2017-01-10 14:51:48 thats the price for doing few bigger commits than many smaller commits 2017-01-10 14:52:01 many small commits makes it easier to rebase 2017-01-10 14:52:04 <^7heo> yep 2017-01-10 14:52:08 <^7heo> definitely. 2017-01-10 14:52:14 <^7heo> I work with the rebase workflow usually 2017-01-10 14:52:19 <^7heo> (not that I don't use git-merge ever, heh) 2017-01-10 14:52:25 <^7heo> so I value small commits. :) 2017-01-10 14:52:32 <^7heo> it's also better for reviewing. 2017-01-10 14:52:38 <^7heo> and easier on newbies. 2017-01-10 14:52:38 yup 2017-01-10 14:52:43 <^7heo> of course it takes more time... 2017-01-10 14:52:53 <^7heo> but I prefer to spend the time to have a good history rather than rush things. 2017-01-10 14:52:56 <^7heo> anyway 2017-01-10 14:52:57 +1 2017-01-10 14:52:58 <^7heo> pulling now. 2017-01-10 14:53:09 i'll bug the cryptsetup devs 2017-01-10 14:53:14 <^7heo> :) 2017-01-10 14:53:20 <^7heo> so it's cryptsetup that's flawed? 2017-01-10 14:55:26 <^7heo> ncopa: #5961 2017-01-10 15:12:49 heh, I remember seeing those leaking file descriptor warnings from lvm on RHEL4 long time ago... 2017-01-10 15:13:20 I wonder what's the point in complaining about them 2017-01-10 15:13:42 <^7heo> that it gets fixed? :) 2017-01-10 15:13:59 <^7heo> I mean, if cryptsetup is leaking file descriptors it might very well be a security issue. 2017-01-10 15:15:00 lvm executes anyway as root? 2017-01-10 15:15:40 <^7heo> we can not make that assumption. 2017-01-10 18:30:19 <^7heo> ncopa: does the test automatically run on travis? 2017-01-10 18:30:24 <^7heo> because I don't think it does. 2017-01-10 18:35:03 <^7heo> I updated my PR so it does. 2017-01-10 19:02:52 <^7heo> ncopa: the PR passes!!!one! 2017-01-10 19:03:05 <^7heo> ncopa: thank you so much. 2017-01-10 19:05:33 ^7heo: I will dance for you 2017-01-10 19:05:36 \o\ 2017-01-10 19:05:41 /o/ 2017-01-10 19:05:45 >->o 2017-01-10 19:05:48 oops 2017-01-10 19:05:48 <^7heo> scadu: thanks! :p 2017-01-10 19:06:15 <^7heo> scadu: looks like you fell 2017-01-10 19:06:41 ^7heo: yeah, I'm not good at this 2017-01-10 19:09:12 <^7heo> :p 2017-01-10 19:56:19 ^7heo: i think i'll cut 3.0.9 release now 2017-01-10 19:56:27 there were a bunch of more fixes 2017-01-10 19:58:01 after that i think im more or less ready to look at the detached header support 2017-01-10 20:13:17 <^7heo> if it passes the tests it's fine IMHO 2017-01-10 20:14:31 <_ikke_> test success -> ship it 2017-01-10 20:16:05 "if it compiles, it's tested. If it links, it's shipped!" 2017-01-10 20:21:08 <^7heo> :p 2017-01-10 21:08:07 out of curiosity could anyone have a look at this PR/change and if they think its ok run the docker commands to generate the bootstrap tar.xz files? its all in the snapshot function so should be easy (it'll take a while to cross compile things though) 2017-01-10 21:08:14 the change: 2017-01-10 21:08:15 https://github.com/alpinelinux/aports/pull/205/commits/4303907ccde80654c59e857c157640532260e9d8 2017-01-10 21:08:47 and all that needs to happen to generate the tar.xz files the apkbuild depends on https://github.com/alpinelinux/aports/pull/205#issuecomment-269838953 2017-01-10 21:09:34 then I can finally get this port upstreamed and into testing 2017-01-10 21:10:03 no rush though 2017-01-10 21:34:33 also didn't want to just have this all fall exclusively on jirutka, cheers 2017-01-10 21:35:02 a bit loaded with other work, sorry :-/ 2017-01-10 21:35:23 no worries ncopa, its no rush honest :) 2017-01-10 21:37:17 as a note it takes about 6.5 hours to compile the entire cross compilation chain for each arch, so its not like it'll be a quick task anyway, I let it run overnight when i test 2017-01-10 21:37:33 thats with an i7 skylake might be faster with a better proc 2017-01-10 21:43:28 fortunately only needs to run ocassionally 2017-01-10 21:44:35 it's machine time, it doesn't count if you can do something else in the meantime. 2017-01-10 21:45:19 true, just means it took a long time to debug the problem with the prior bootstrapped ghc 2017-01-10 22:06:42 mitchty: what? I’m a bit out-of-context now 2017-01-10 22:07:35 mitchty: I’m not reading IRC now, I need to focus on work and implementation of POSIX shell parser in free time for some abuild automation 2017-01-10 22:08:39 and this is going surprisingly well btw, even despite the fact that people who designed shell grammar were probably totally insane 2017-01-10 22:09:09 jirutka: no worries, for my ghc port i just need to regenerate the bootstrap ghc compiler and get them uploaded 2017-01-10 22:09:20 <_ikke_> jirutka: (or just have different concerns than you :P) 2017-01-10 22:09:28 i doubt shell was ever "designed" based on the old source I read :) 2017-01-10 22:10:52 jirutka: its no big rush though, getting this ported has helped teach me more patience about cross compiling 2017-01-10 22:10:55 you mean that it manifested itself? XD 2017-01-10 22:11:24 ah, ghc, I should totally look at it, this was a huge effort 2017-01-10 22:11:30 well it was interesting reading "shell" before stuff like >, <, | and whatnot operators existed 2017-01-10 22:12:28 yeah, i've learnt more about llvm IR, PIC, elf, the GOT, linking, arm assembly, and more 2017-01-10 22:12:40 than i ever imagined i'd need to when i started this task 2017-01-10 22:16:51 wow! 2017-01-10 22:17:23 we should create a Hall of fame on alpinelinux.org 2017-01-10 22:20:52 the ultimate vanity position: the one who builds halls of fame :) 2017-01-10 22:22:48 <^7heo> I don't care, my PR passes the tests! pp 2017-01-10 22:22:56 <^7heo> :p* 2017-01-11 15:55:33 ls 2017-01-11 15:55:37 oops 2017-01-11 18:20:20 I'd like to add a IPFS mirror to the mirror list in rsync.alpinelinux.org/alpine, who should I contact? 2017-01-11 20:26:25 any reason that the headers of libffi are in /usr/lib/libffi-3.2.1/include instead of /usr/include ? 2017-01-12 08:06:24 <^7heo> fabled: also, I have inivited dalias here. I dunno for you, but I sure do not wish to spend most of my time doing PR with docker users... 2017-01-12 08:07:37 github is experiencing issues 2017-01-12 08:08:35 <^7heo> fcolista: https://status.github.com/ might be more informative ;) 2017-01-12 08:09:43 bah 2017-01-12 08:10:03 <^7heo> :p 2017-01-12 08:10:20 should exists a standard way to know the status.. 2017-01-12 08:10:35 i use to check twitter for that 2017-01-12 08:11:34 <^7heo> ah, you mean a standard way accross companies... 2017-01-12 08:11:48 <^7heo> what about using rss/atom? 2017-01-12 08:12:10 <^7heo> or something alike, more specific, w/o xml 2017-01-12 10:30:31 http://pastebin.com/sE6rkmpj does that make any sense for someone ? 2017-01-12 10:48:04 Hi. Want to ask someone who have a little time to review the pr#716 - php7 upgrade. It includes fixes and improvements that have been discovered in the past few months. APKBUILD internals was cleaned-up to reuse it to build php5 with minimal changes, so we can make php7 & php5 live in sync 2017-01-12 10:51:49 <^7heo> ncopa: ping 2017-01-12 10:52:11 <^7heo> could someone please merge https://github.com/alpinelinux/mkinitfs/pull/3? :) 2017-01-12 10:52:25 <^7heo> Klowner: you might be happy to know that there is progress in mkinitfs :) 2017-01-12 10:54:33 <^7heo> ncopa: 5a84d57 nice one :D 2017-01-12 11:41:44 ^7heo: pong 2017-01-12 11:41:52 will looke at that a bit later 2017-01-12 11:42:19 yeah its normally a good idea to wipe passwords after use 2017-01-12 11:43:53 <_ikke_> What is the standard to mention CVE info in commits? 2017-01-12 11:47:24 we dont have any written standard, but i find it useful to at least write the CVE in the commit 2017-01-12 11:47:45 because then you can search the gitlog on a given CVE 2017-01-12 11:48:14 <_ikke_> Right. I thought I read something about automatic cve detection 2017-01-12 11:48:24 if it is for the stable branch(es) then i tend to like to have the CVE in the suject line (first line) 2017-01-12 11:48:43 because then the CVE is visible in the cgit log 2017-01-12 11:48:52 <_ikke_> right, like abuild -s does 2017-01-12 11:48:59 <_ikke_> abump* 2017-01-12 11:49:01 eaxctly 2017-01-12 11:49:25 http://git.alpinelinux.org/cgit/aports/log/?h=3.4-stable 2017-01-12 12:00:22 <^7heo> ncopa: there's not much in that PR anymore since you fixed the script and the nlplug-findfs so it works with it. 2017-01-12 12:00:40 <^7heo> ncopa: it just enables the testing automatically on every PR from now on. 2017-01-12 13:14:58 http://pastebin.com/sE6rkmpj does that make any sense for someone ? < in case you've not seen it 2017-01-12 13:16:32 or were knowingly ignoring me :) 2017-01-12 15:08:52 coredumb: remove update_config_sub? 2017-01-12 15:11:48 clandmeter: oh I'm stupid I should remove the prepare() completely 2017-01-12 15:18:45 i am about to update the rsync.alpinelinux.org in few minutes 2017-01-12 15:19:09 we have set up new instance on a bit stronger hardware 2017-01-12 15:19:57 i will need to stop the builders for a while 2017-01-12 15:20:43 i need to: 1) stop builders, 2) update dns 3) update .ssh/known_hosts on all builders 4) restart the builders 2017-01-12 15:37:31 stopping the builders now 2017-01-12 15:43:34 builders are down for 20mins 2017-01-12 16:26:51 builders are up again 2017-01-12 16:27:04 and we are not over on new rsync.alpinelinux.org host 2017-01-12 16:28:14 those were some pretty long 20 minutes :D 2017-01-12 16:29:03 sorry bout that 2017-01-12 16:29:06 :) 2017-01-12 16:29:22 oh 2017-01-12 16:29:24 i said wrong 2017-01-12 16:29:33 we *are* over to new rsync.alpinelinux.org 2017-01-12 16:29:41 grats 2017-01-12 16:30:31 i wonder if we should bwlimit rsync 2017-01-12 16:30:39 bwlimit per rsync 2017-01-12 16:31:17 why, are the pipes narrow? 2017-01-12 16:31:28 not really 2017-01-12 16:31:32 its 200mbit i think 2017-01-12 16:31:49 i was thinking of adding a 100mbit limit 2017-01-12 16:32:02 because the mirrors will likely rsync at the same times 2017-01-12 16:32:20 how about making sure they don't do that instead 2017-01-12 16:32:55 (your solution will fail if/when you add a third mirror) 2017-01-12 16:32:55 its kinda nice to know when they mirror 2017-01-12 16:33:04 yes i know 2017-01-12 16:33:10 sure. Do them sequentially. 2017-01-12 16:33:21 it was more of "just becuase" 2017-01-12 16:33:37 we dont have control over all mirrors 2017-01-12 16:34:11 hm 2017-01-12 16:34:17 i suppose try distribute the sync time makes more sense 2017-01-12 16:34:19 in that case yes, you could bwlimit 2017-01-12 16:34:39 i don tthink it is a problem atm though 2017-01-12 16:35:03 the mirrors pull the data, you don't push it to them? 2017-01-12 16:35:10 they pull yes 2017-01-12 16:35:16 normally from cron jobs 2017-01-12 16:35:26 yeah, then you can't properly sync the times 2017-01-12 16:35:49 which is why i thought that it could make sense to bwlimit 2017-01-12 16:36:15 it does indeed, but it's suboptimal if there's only one sync at a time 2017-01-12 16:36:22 i think what we really want is tier1 and tier2 mirrors 2017-01-12 16:36:36 that is, tier1 mirrors that we do have control over 2017-01-12 16:36:40 can you just limit the max of concurrent rsync clients to 1 or something? 2017-01-12 16:36:41 one on each continent 2017-01-12 16:36:59 and only the tier1 mirrors can rsync from master mirror 2017-01-12 16:37:13 so if another one connects while a mirror is rsyncing, the second mirror is delayed 2017-01-12 16:37:17 until the first one has finished 2017-01-12 16:37:30 we could probably also do push to tier1 from master 2017-01-12 16:37:31 I don't know how the rsync client reacts to TCP delays though 2017-01-12 16:37:45 might timeout and give up 2017-01-12 16:37:57 it's probably tunable 2017-01-12 16:38:07 then we do some kind of geodns 2017-01-12 16:38:19 so you get a tier1 mirror based on your source ip 2017-01-12 16:39:04 ah, designing big infrastructure... I feel in 2002 again 2017-01-12 16:39:09 or maybe even by dns round-robin to make it stupid and simple 2017-01-12 16:39:41 2002 was a good year 2017-01-12 16:39:48 :) 2017-01-12 17:10:00 <^7heo> ncopa: !!! 2017-01-12 17:10:13 <^7heo> ncopa: review? 2017-01-12 17:11:00 <^7heo> skarnet: 2002 was a palindrome year. That's why. 2017-01-12 21:58:05 Hey, is there the tool/daemon auditd (from RedHat) packaged with Alpine Linux? The only "audit" package I can find with "apk search" is "aaudit" (Alpine Auditor). 2017-01-12 21:58:26 auditd - if for security auditing, i.e. for intercepting/capturing users activity in the system, e.g. log syscalls, file changes, etc. 2017-01-12 21:58:41 aaudit - a package in Alpine - this package is Alpine Auditor - what for purpose is this package? For the same purpose as auditd, i.e. security auditing? 2017-01-12 21:59:19 czart: we are not using selinux 2017-01-12 21:59:35 but grsecurity, with its RSBAC MAC 2017-01-12 21:59:55 auditd is selinux utility 2017-01-12 22:00:06 or at least strongly connected with it 2017-01-12 22:00:27 skrzyp: OK, thanks. So, is there any alternative to auditd on Alpine? 2017-01-12 22:01:00 skrzyp: at least, I would have something to intercept/capture syscalls, in particular exec* family calls. 2017-01-12 22:01:27 sysdig for example? 2017-01-12 22:01:46 I am using Alpine as Docker container, and I wish to have some audit facility in a container. 2017-01-12 22:01:58 skrzyp: Thanks, I will check this one. 2017-01-12 22:05:36 skrzyp: Btw, are selinux and grsecurity two different philosophies or alternatives? I am just curious. I met only SELinux term when pouring over Docker docs so far. 2017-01-12 22:07:00 I'd say they are two different ways of wasting your time and not significantly improve security, but my opinion is pretty unpopular. ;) 2017-01-12 22:07:03 poring over* 2017-01-12 22:09:20 skarnet: OK, so those are security related topics... So, Linux could work without them or particular functionalities would be missing without them? Or, they solely improve security? 2017-01-12 22:10:14 Linux can, and does, work without them. I don't have the numbers, but chances are there are more kernels running without any kind of security modules than there are running SELinux or grsecurity. 2017-01-12 22:11:42 SELinux and grsecurity add "security functionality", i.e. a configurable way of fine-tuning various permissions. 2017-01-12 22:12:16 ITT: skarnet disturbs young minds 2017-01-12 22:12:19 :) 2017-01-12 22:12:32 I'm a pervert like that 2017-01-12 22:12:54 skarnet: Thanks, clear enough :) 2017-01-12 22:15:26 And if you're a dev, my advice is to forget all this and learn secure Unix coding practices. If you're an admin and have to handle software you don't control, it's another story and SELinux/grsecurity *may* be useful to you. 2017-01-12 22:18:41 skarnet: Yeah, I rather perform admin responsibilities to achieve dev goals. Thanks. 2017-01-12 22:28:07 https://bugzilla.redhat.com/show_bug.cgi?id=1409531 btw seems on topic 2017-01-12 22:28:09 :) 2017-01-12 22:36:48 coredumb: do you know what "incoming process" means in this context? Any tangible example? 2017-01-12 22:44:31 czart: as far as I understand it's the processes run with "runc exec" 2017-01-12 22:52:15 coredumb: Hmm, docker utilizes libcontainer/runc, so it may be triggered when "docker exec" is used, e.g. when invoking "docker exec -it /bin/sth" on the host machine. 2017-01-12 22:52:43 Interesting, anyway... 2017-01-12 22:58:27 yep runc has been the backend for a couple versions or so 2017-01-12 23:00:02 But if CVE occurs during "initialization", it might be some "runc exec" internal call, as you said. 2017-01-13 00:09:02 Do I need to install git on my Alpine to clone (check out) a github repository? Or, is there any other way to download source codes? 2017-01-13 00:09:49 I must download: https://github.com/draios/sysdig to install sysdig from source codes. 2017-01-13 00:09:56 skrzyp: ^ 2017-01-13 00:19:32 depending on the software, you generally want a release tarball more than a github commit, because a lot of software uses autoconf and non-release commits generally don't include a configure script 2017-01-13 00:19:44 and you don't want to install autoconf on your machine 2017-01-13 00:20:25 so, get the latest release tarball: wget https://github.com/draios/sysdig/archive/0.13.0.tar.gz 2017-01-13 00:20:30 czart: ^ 2017-01-13 00:22:46 skarnet: good to know, thanks :D 2017-01-13 00:27:00 Indeed, in the official sysdig installation instruction: "Download the sysdig github repository to your local machine", which is imprecise... 2017-01-13 00:49:53 skarnet: to use wget with https, the openssl need to be installed: "apk add openssl" 2017-01-13 00:50:41 (the hint from https://github.com/Yelp/dumb-init/issues/73) 2017-01-13 00:51:35 apk add libressl 2017-01-13 00:51:55 Alpine switched from openssl to libressl a short time ago 2017-01-13 00:52:25 and yes, it's a useful package to have - you won't get much networking done without it. 2017-01-13 00:59:18 skarnet: OK, thanks. 2017-01-13 01:51:07 I have installed cmake (apk add cmake), but when I use it, I get errors: http://pastebin.com/7rLCgDyt Also, I installed g++ and gcc (apk add g++ and apk add gcc). Is the order of package installation vital to avoid the errors or the cause is different? 2017-01-13 01:52:07 Do I have to install other packages? (As described in: https://wiki.alpinelinux.org/wiki/How_to_get_regular_stuff_working) 2017-01-13 02:04:23 That is, I used cmake in an attempt to build sysdig... 2017-01-13 02:16:55 OK. Resolved, make package was not there, so inevitable was to install it (apk add make). 2017-01-13 04:28:55 how do I get a package added? I want jed in the package database 2017-01-13 05:09:10 I stuck building sysdig on Alpine in the middle (48% build progress), while building libsinsp library a compilation error occurs: http://pastebin.com/n5d0hwTS 2017-01-13 05:11:51 My step-by-step installation notes and instructions/commands are listed here: http://pastebin.com/ewx3pthg 2017-01-13 05:12:29 Could anybody examine this issue and determine what is wrong? 2017-01-13 08:26:45 morning. happy friday! 2017-01-13 08:31:34 Morning o/ 2017-01-13 09:49:58 \o 2017-01-13 09:59:43 czart: looks like getaddrinfo_a is a GNU extension 2017-01-13 10:00:07 i think you can report the compile error to sysdig upstream 2017-01-13 10:04:06 ncopa: morenig, what do you think about an ABI x32 alpine? 2017-01-13 10:05:17 xming: i think its more work than it is worh 2017-01-13 10:05:20 worth* 2017-01-13 10:10:44 yeah probably, but I have been trying to do that 2017-01-13 10:10:58 it's progressing 2017-01-13 12:45:11 <^7heo> ncopa: could someone else (fabled or skarnet or jirutka maybe) review the PR 3? :) 2017-01-13 12:45:23 <^7heo> I'd love to rebase my other features before the we 2017-01-13 12:45:26 <^7heo> and submit them asap. 2017-01-13 12:47:52 not me, sorry, very busy atm. 2017-01-13 12:52:09 <^7heo> skarnet: thanks for the swift answer. 2017-01-13 13:09:22 ^7heo, PR3? full link? 2017-01-13 13:09:35 <^7heo> fabled: sure. 2017-01-13 13:09:54 <^7heo> (as soon as my connection can :P) 2017-01-13 13:10:33 <^7heo> fabled: https://github.com/alpinelinux/mkinitfs/pull/3 2017-01-13 13:14:21 <^7heo> jirutka: if you have time to check it out, esp. the travis part, that would be really awesome. 2017-01-13 13:14:26 <^7heo> (and yes the travis part is most of it) 2017-01-13 13:15:09 <^7heo> fabled: do you have write access on that repo? 2017-01-13 13:19:55 anyone, maybe fabled: my Alpine (3.5 freshly updated) doesn't boot for some reason: OpenRC starts, but as soon as the console fonts change, the screen blacks out and nothing more happens. 2017-01-13 13:20:15 I'd like to investigate with init=/bin/sh, but apparently it doesn't work on syslinux. 2017-01-13 13:20:37 So, does anyone know how to give a command line to the kernel with the installed syslinux? 2017-01-13 13:21:14 boot: vanilla init=/bin/sh -> standard boot, not giving me a shell 2017-01-13 13:21:37 and it's kinda blocking me right now :/ 2017-01-13 13:27:12 <^7heo> Damn 2017-01-13 13:27:17 <^7heo> I just realized something 2017-01-13 13:27:38 <^7heo> my vmlinuz-grsec is in the /boot partition... 2017-01-13 13:27:59 so? 2017-01-13 13:28:11 is that bad? 2017-01-13 13:28:40 <^7heo> well 2017-01-13 13:28:48 <^7heo> my whole disk is encrypted... 2017-01-13 13:28:52 <^7heo> to the exception of /boot 2017-01-13 13:28:55 <^7heo> for obvious reasons. 2017-01-13 13:29:02 <^7heo> so... 2017-01-13 13:29:08 <^7heo> yeah it's kinda bad. 2017-01-13 13:29:16 <^7heo> moreover I lost my bowl... 2017-01-13 13:29:42 I still don't understand what the problem is 2017-01-13 13:31:57 and I really would appreciate some hint on how to boot with init=/bin/sh 2017-01-13 13:32:28 else I'll have to boot on another VM and mount the disk and that makes everything an order of magnitude more painful 2017-01-13 13:36:35 init=/bin/busybox sh ? 2017-01-13 13:38:20 <^7heo> xming: why wouldn't /bin/sh work? 2017-01-13 13:38:30 it should 2017-01-13 13:38:55 <^7heo> skarnet: the problem is that since the kernel I run is in a clear partition... 2017-01-13 13:39:01 <^7heo> skarnet: anyone can replace it. 2017-01-13 13:39:13 unless it's a symlink and it's broken 2017-01-13 13:39:22 <^7heo> skarnet: so I could as well stop encrypting. 2017-01-13 13:40:11 xming: it's a symlink and it's not broken, and if it was, I'd get a kernel panic because init not found. That's not what's happening: the init= arg is simply ignored. 2017-01-13 13:40:35 <^7heo> skarnet: I can't find any mention of the init= arg in the syslinux doc 2017-01-13 13:40:39 ^7heo: an attacker still needs your key to decrypt everything. 2017-01-13 13:41:09 <^7heo> skarnet: if the kernel is spoofed, the kernel can do "behind the curtain" operations while I use my computer normally... 2017-01-13 13:41:13 yeah, if syslinux doesn't allow me to give a kernel command line, what the hell is it good for? 2017-01-13 13:41:22 <^7heo> skarnet: so, it's kinda like I provide the key by myself, yes. 2017-01-13 13:41:46 ah, yes. You're worried about social engineering via a mitm kernel. :D 2017-01-13 13:41:59 <^7heo> no need for social engineering. 2017-01-13 13:42:04 <^7heo> Just change the kernel 2017-01-13 13:42:10 <^7heo> wait for me to use my computer 2017-01-13 13:42:21 <^7heo> done, you get me running your softwrae. 2017-01-13 13:42:24 <^7heo> software* 2017-01-13 13:42:33 <^7heo> and all decrypted. 2017-01-13 13:43:10 you still need to be phished for your pw. the evil kernel must pass as your own. 2017-01-13 13:43:19 <^7heo> well, yes. 2017-01-13 13:43:25 <^7heo> and it's pretty easy to do so. 2017-01-13 13:43:45 <^7heo> take my kernel (which source is open and free), modify what you want to include, compile, copy. 2017-01-13 13:43:54 build date. 2017-01-13 13:44:01 <^7heo> yeah okay the build date will change. 2017-01-13 13:44:07 <^7heo> but that's not like it's checked anywhere. 2017-01-13 13:44:10 <^7heo> nor is its hash. 2017-01-13 13:44:20 you know what you have to do, then. 2017-01-13 13:44:31 <^7heo> I thought about that. 2017-01-13 13:44:45 <^7heo> Or even using my kernel hash as part of the cryptsetup key. 2017-01-13 13:45:09 <^7heo> but that would obviously open new problems. 2017-01-13 13:46:16 <^7heo> Anyway... 2017-01-13 13:46:18 ^7heo: so there is no, erm, sane way to workaround the problem of not encrypted /boot? 2017-01-13 13:46:31 <^7heo> well 2017-01-13 13:46:47 <^7heo> I know it's there in /boot for a reason 2017-01-13 13:46:57 <^7heo> but I wish I would have realized it before. 2017-01-13 13:47:40 <^7heo> the problem I have here is: there is no way to have your verification in the encrypted partition (so, secure) while using it BEFORE decrypting the partition. 2017-01-13 13:47:44 <^7heo> because that would actually work. 2017-01-13 13:48:02 <^7heo> unless the binary is signed 2017-01-13 13:48:43 <^7heo> but then the trust falls back on the public key. 2017-01-13 13:48:51 <^7heo> if it's changed, the signed binary can be changed too. 2017-01-13 13:49:11 yep, I fall in the same loop thinking about this some time ago 2017-01-13 13:49:19 <^7heo> I see a solution tho. 2017-01-13 13:49:19 then I just forgot ;f 2017-01-13 13:49:24 ^7heo: hm? 2017-01-13 13:49:31 <^7heo> nah nevermind 2017-01-13 13:50:02 <^7heo> I was thinking about cryptsetup checking the public key (in /boot) against the public key (in encrypted part) 2017-01-13 13:50:09 <^7heo> BEFORE unlocking the partition 2017-01-13 13:50:48 <^7heo> and halting with a giant message "encryption compromised - change your passphrase from a trusted setup" 2017-01-13 13:51:00 <^7heo> that would help a little 2017-01-13 13:51:23 <^7heo> but it still wouldn't prevent the key from leaking to the (attacker's) kernel, obviously. 2017-01-13 13:52:02 <^7heo> what I would really like is to have the kernel prevented from ever being used, so user would see the real issue before it's too late 2017-01-13 13:52:11 <^7heo> (and before having to change the key, the passphrase, etc) 2017-01-13 13:53:24 skarnet: does it help to add fbcon to /etc/modules 2017-01-13 13:53:45 ncopa: does it help to press F1 when the keyboard is missing 2017-01-13 13:53:50 I CANNOT BOOT 2017-01-13 13:54:30 "OpenRC starts, but as soon as the console fonts change, the screen blacks out and nothing more happens." 2017-01-13 13:54:47 <^7heo> yeah that's a funny thing to have "Keyboard missing - Press F1 to continue to boot" 2017-01-13 13:54:53 and the system is unusable: no console, no sshd 2017-01-13 13:54:54 it sounds like it actually boots but switches to framebuffer 2017-01-13 13:55:08 <^7heo> ncopa: what he wants to achieve is fixing the machine WITHOUT having to boot another medium. 2017-01-13 13:55:19 <^7heo> ncopa: so he needs a shell for that. 2017-01-13 13:55:35 if I had a sshd, I could fix everything, and idc for the console anyway 2017-01-13 13:55:44 but it's obviously not reaching that point 2017-01-13 13:55:50 <^7heo> skarnet: can't you use serial? :) 2017-01-13 13:56:07 I could, but would it help? 2017-01-13 13:56:10 you might be able to log in "blind" 2017-01-13 13:56:25 <^7heo> and start ssh 2017-01-13 13:56:28 as if the screen was disconnected 2017-01-13 13:56:41 and do: modprobe fbcon 2017-01-13 13:56:47 <^7heo> or that 2017-01-13 13:56:52 <^7heo> better than SSH :D 2017-01-13 13:57:07 I'll try 2017-01-13 13:57:14 not like I have any better ideas 2017-01-13 13:57:19 <^7heo> but anyway, even without serial 2017-01-13 13:57:28 <^7heo> you can probably login blind granted that you wait long enough 2017-01-13 13:57:52 <^7heo> damn I'm f***ing hungry 2017-01-13 13:58:08 now that you remind me, im actually hungry too 2017-01-13 13:58:11 :) 2017-01-13 13:58:16 <^7heo> Huhu 2017-01-13 13:58:25 <^7heo> I'm a little pissed because I can't find my bowl 2017-01-13 13:58:29 lol, logging in blind and typing modprobe fdcon worked. 2017-01-13 13:58:35 fbcon* 2017-01-13 13:58:35 <^7heo> looks like someone took it, but I can't be sure until I find it. 2017-01-13 13:58:42 <^7heo> skarnet: :) 2017-01-13 13:58:54 <^7heo> skarnet: grats, you saved time against trying with a different medium. 2017-01-13 13:58:56 <^7heo> (and pain) 2017-01-13 13:58:59 skarnet: would be great if you could help me figure out how/what should have autoloaded fbcon 2017-01-13 13:59:09 ncopa: you've earned a coffee on me at fosdem. 2017-01-13 13:59:19 and yes, I'm going to try and figure that out. 2017-01-13 13:59:27 <^7heo> you're all going at FOSDEM? 2017-01-13 13:59:32 yes. 2017-01-13 13:59:36 yes 2017-01-13 13:59:40 and you're not invited. 2017-01-13 13:59:43 ACTION hides 2017-01-13 13:59:45 <^7heo> ok ok 2017-01-13 13:59:48 <^7heo> that settles it then. 2017-01-13 13:59:56 of course you are. 2017-01-13 14:00:03 <^7heo> Yeah but brussel is far 2017-01-13 14:00:07 <^7heo> and I have no place to stay there. 2017-01-13 14:00:26 i kinda thought we had an mdev.conf rule that would make sure fbcon was autoloaded 2017-01-13 14:00:45 skarnet: do you use mdev of eudev or something else? 2017-01-13 14:01:15 what do I look like, a Gentoo ricer? 2017-01-13 14:01:26 <^7heo> ncopa: in case you're not thinking of it: it's an updated install of an older alpine. I had many more issues with updated installs than with fresh installs. 2017-01-13 14:02:06 skarnet: you look like someone that might be crazy enough to use s6 ;) 2017-01-13 14:02:22 (crazy is positive in this ctx) 2017-01-13 14:02:41 <^7heo> why not ctxt? 2017-01-13 14:02:59 well if I had modified my Alpine to use s6, it would have booted up to dropbear, because, you know, s6 actually works 2017-01-13 14:03:06 <^7heo> ahahaha 2017-01-13 14:03:10 lol 2017-01-13 14:21:37 ScrumpyJack: are you still ok to maintain abook? 2017-01-13 14:21:43 im about to move it to community 2017-01-13 14:21:56 http://patchwork.alpinelinux.org/patch/2496/ 2017-01-13 14:24:56 ncopa: well /etc/init.d/hwdrivers looks fine, but it clearly doesn't modprobe fbcon, so I don't know. 2017-01-13 14:25:08 maybe there's a race condition somewhere? 2017-01-13 14:25:12 i think we load it from mdef rule 2017-01-13 14:25:15 mdev 2017-01-13 14:25:30 i think the idea is that if we get any /dev/fb* device, then we should load fbcon 2017-01-13 14:25:46 i think we do so already with mdev (but i havent checked) 2017-01-13 14:25:59 but i dont know if we added corresponding rule to eudev 2017-01-13 14:25:59 fb0 root:video 0660 @modprobe -q -b fbcon 2017-01-13 14:26:25 do you have /dev/fb0? 2017-01-13 14:26:40 I do, but mdev permissions have not been applied 2017-01-13 14:26:50 aha 2017-01-13 14:27:05 /proc/sys/kernel/hotplug is empty and you don't run s6-devd 2017-01-13 14:27:11 lol 2017-01-13 14:27:13 so obviously mdev is never called 2017-01-13 14:27:27 right 2017-01-13 14:27:33 do you have udevd running? 2017-01-13 14:27:53 do I look like I would have udevd running? 2017-01-13 14:28:18 no, but i had to ask you know 2017-01-13 14:28:30 i dont want to judge anyone by the looks... 2017-01-13 14:28:40 I have no hotplug manager apparently, not sure if it's normal or not 2017-01-13 14:28:47 it is not 2017-01-13 14:29:11 I don't think I did anything to change from the default... but I'll fix that all right 2017-01-13 14:29:40 are you using xorg? 2017-01-13 14:29:42 not today though, I'll use a quickfix in /etc/modules for now 2017-01-13 14:29:44 no 2017-01-13 14:29:52 I don't even care about the console 2017-01-13 14:29:58 then mdev might be a good choice for now 2017-01-13 14:30:06 actually 2017-01-13 14:30:06 what do you mean "for now" 2017-01-13 14:30:26 I very much hope it will continue to be a supported choice 2017-01-13 14:30:34 til we get s6-devd... 2017-01-13 14:31:06 which is only a netlink listener for mdev ;) 2017-01-13 14:31:20 at some point I should take over vdev and make it usable 2017-01-13 14:31:40 but my plate is full for the whole next year at least :/ 2017-01-13 14:31:44 rc-update add mdev sysinit 2017-01-13 14:32:02 * rc-update: mdev already installed in runlevel `sysinit'; skipping 2017-01-13 14:32:07 huh 2017-01-13 14:32:11 oh.. 2017-01-13 14:32:32 what you have in /etc/inittab? 2017-01-13 14:32:44 ::sysinit:/sbin/openrc sysinit 2017-01-13 14:32:45 ::sysinit:/sbin/openrc boot 2017-01-13 14:32:45 ::wait:/sbin/openrc default 2017-01-13 14:32:48 nothing different from the default 2017-01-13 14:33:34 do you have the "::sysinit:/sbin/openrc sysinit" there? 2017-01-13 14:33:59 what part of "nothing different from the default" was unclear 2017-01-13 14:34:21 im just doublechecking 2017-01-13 14:34:33 if everything is default, the mdev should been the hotplugger 2017-01-13 14:34:48 an inittab without that line would probably not be able to boot at all :D 2017-01-13 14:35:01 yeah, something is emptying my hotplug helper 2017-01-13 14:35:21 if mdev openrc script actually execute, then the restult should be that mdev is hotplugger 2017-01-13 14:35:27 so i wonder why it was not 2017-01-13 14:35:56 if I'm the only one with this issue, it's no big deal, I probably have something somewhere that empties my hotplug helper 2017-01-13 14:36:07 sounds like that yes 2017-01-13 14:36:18 could be entry in sysctl.conf 2017-01-13 14:36:28 or something else 2017-01-13 14:36:37 nothing there 2017-01-13 14:36:49 could also happen if you have udev in /proc/cmdline apparently 2017-01-13 14:36:57 please 2017-01-13 14:37:23 # check if udev is specified on cmd line 2017-01-13 14:37:24 if get_bootparam "udev"; then 2017-01-13 14:37:24 ewarn "Skipping mdev as udev requested in kernel cmdline" 2017-01-13 14:37:24 return 0 2017-01-13 14:37:24 fi 2017-01-13 14:38:01 im out of ideas 2017-01-13 14:38:05 alpine:/# grep udev /proc/cmdline || echo "seriously, what did you think?" 2017-01-13 14:38:05 seriously, what did you think? 2017-01-13 14:38:23 but don't worry about it 2017-01-13 14:38:32 it's probably something I did at some point and don't remember 2017-01-13 14:38:39 I'll find it, or I won't 2017-01-13 14:38:41 doesn't matter 2017-01-13 14:38:48 ok 2017-01-13 14:38:57 i'll continue with my friday then :) 2017-01-13 14:38:58 as long as it works for others there's no problem 2017-01-13 14:39:05 thanks for the fbcon hint 2017-01-13 14:39:18 np 2017-01-13 14:39:46 ah, question 2017-01-13 14:39:56 modloop systematically fails for me 2017-01-13 14:39:59 is it normal? 2017-01-13 14:40:04 no 2017-01-13 14:40:11 I'm not using root-on-tmpfs 2017-01-13 14:40:20 oh 2017-01-13 14:40:24 then i suppose its normal 2017-01-13 14:40:48 remove the modloop from startup 2017-01-13 14:40:54 you dont need it 2017-01-13 14:40:58 ok 2017-01-13 14:41:14 its a hack for root-on-tmpfs 2017-01-13 14:41:42 would be nice to automatically delete it from sysinit for root-on-disk installs 2017-01-13 14:41:56 because I never used root-on-tmpfs and still always had modloop 2017-01-13 14:42:11 yes it should be deleted 2017-01-13 14:42:25 i might have fixed it already 2017-01-13 15:40:05 ncopa: yes to abook 2017-01-13 15:40:55 anything else that needs a maintainer? 2017-01-13 15:44:48 @ncopa: Thanks. I will report the issue to sysdig, and share a link to the ticket when it's ready. So that the issue can be tracked. See you later. 2017-01-13 15:50:03 ScrumpyJack: check in testing 2017-01-13 18:01:02 kk 2017-01-14 08:00:14 Alpine's missing scripts for handling acpid events 2017-01-14 15:02:27 I think we need to do some maintainers cleanups in upcoming future. many of them are maintainers only on paper, not contributing to aports for years. 2017-01-14 15:05:19 well, it's nothing wrong with people losing interest in things, but they should cede maintaintership before leaving alpine devel boat (mountain?). apparently it's rarely the case. 2017-01-14 15:06:29 void (i think it was void) had this nice idea of having packages include a script, if possible, to notify what the latest version is 2017-01-14 15:06:35 to be able to quickly list unupdated packages 2017-01-14 15:07:36 asie: https://release-monitoring.org/distro/Alpine/ 2017-01-14 15:07:50 asie: as far as I know it is already integrated 2017-01-14 15:07:57 jirutka: could you please confirm? 2017-01-14 15:08:23 asie: maintainers should receive e-mail notifications 2017-01-14 15:08:35 yeah, jirutka integrated aports with anitya some time ago. 2017-01-14 15:08:36 http://lists.alpinelinux.org/alpine-devel/5456.html 2017-01-14 15:08:48 the problem is with maintainers, as I already said 2017-01-14 15:09:27 I have also my own version tracking tool that got stalled a bit 2017-01-14 15:09:49 https://github.com/vracker 2017-01-14 15:11:29 przemoc: you’re right, many maintainers are not active anymore :( 2017-01-14 15:12:37 scadu: yes, I’ve integrated Anitya with Aports few months ago, most of the outdated flags are now automatically created by it; curently it’s registered about 60 % of packages from main and community 2017-01-14 15:14:26 hm, I should finally write something about Anitya on wiki 2017-01-14 15:16:10 actually anitya flooded http://pkgs.alpinelinux.org/flagged ;) there are so many outdated pkgs there that it's hard to clean it up in a timely fashion... 2017-01-14 15:17:18 yes, unfortunately… but at least we know it now 2017-01-14 15:17:27 this is the first step for improvement 2017-01-14 15:17:53 I'm not complaining, at least not complaining about anitya doing its job ;) 2017-01-14 15:19:39 IIRC if something is flagged, it cannot be reflagged with newer version, therefore new version field can become outdated too 2017-01-14 15:19:58 this does not apply for Anitya ;) 2017-01-14 15:20:17 it compares the version in the flag and flag it again when the version is older 2017-01-14 15:20:42 I see, poor maintainers, flooded with mails... 2017-01-14 15:20:55 so it’s quiet efficient in spamming maintainers :P 2017-01-14 15:21:30 we’d like to implement something like daily digests 2017-01-14 15:21:33 I think that sending digest mails combining all updates in the day would be nicer 2017-01-14 15:21:36 exactly 2017-01-14 15:21:43 instead of sending email right once the package is flagged 2017-01-14 15:25:13 https://github.com/alpinelinux/aports-turbo/issues/22 2017-01-14 17:01:56 jirutka: bumped a few packages. https://github.com/alpinelinux/aports/pulls/scadu 2017-01-14 17:07:51 oh, travis failed on i3wm :< 2017-01-14 17:09:01 no, it's just because it lacks libtoolize 2017-01-14 17:09:08 but previous iterations were okay 2017-01-14 17:09:26 so I guess it some kind of build environment bug 2017-01-14 17:09:31 or packaging for autoshit 2017-01-14 17:09:33 jirutka: 2017-01-14 18:26:16 scadu: why did you redo some of the work that was already done? 2017-01-14 18:26:21 http://patchwork.alpinelinux.org/ anyone? 2017-01-14 18:27:54 przemoc: frankly I didn't look at patchwork. sorry then. 2017-01-14 18:29:01 przemoc: it seems that patchwork is abandoned a little nowadays :< 2017-01-14 18:34:28 well, we need more services to manage the project. it certainly will help in preventing conflicts like those. 2017-01-14 18:38:45 it's not abandoned, it was mostly cleaned/applied on Friday by ncopa. fabled also applies stuff from there now and then. 2017-01-14 18:42:37 no big issue, but it's simply tad sad when in project like AL, where manpower is so limited, work gets duplicated 2017-01-14 18:45:17 przemoc: I know. It gets me sad too. I will check patchwork before commiting anything in the future. 2017-01-14 18:46:06 but it is easy to forget when you used to use on service. 2017-01-15 07:22:02 przemoc, pantchwork is preferred by fabled/ncopa, but we've been backlogged recently. clandmeter/jirutka seem to prefer github. not sure about other committers. 2017-01-15 08:01:13 fabled: seems like it would be easier to have only one 2017-01-15 08:04:27 coredumb, we started accept github PRs since some prefer that. however, it's rather unfortunate that it created this disparity. should probably talk on mailing list what to do about it 2017-01-15 08:07:32 fabled: yep 2017-01-15 09:48:03 can we also talk about whitelisting on the aports ML server? 2017-01-15 09:48:32 (because I'd like to be able to actually send patches) 2017-01-15 09:49:15 where is your IP coming from ? 2017-01-15 09:49:22 I send OK to the ML myself 2017-01-15 09:50:51 I'm talking about fucking PBL 2017-01-15 09:51:06 And I'm considered as a spammer pretty much everywhere else >_< 2017-01-15 09:51:34 join the club then 2017-01-15 09:52:08 what do you mean you send OK to the ML? 2017-01-15 09:52:47 I'm not even greylisted, I'm *black*listed because the server checks the PBL, which is an insult to net neutrality, and denies my IP since it's dynamic 2017-01-15 09:53:25 I have thrown a tantrum about this before, it was loud enough that nangel woke up and whitelisted me 2017-01-15 09:53:42 yeah I'm not in the PBL indeed 2017-01-15 09:53:51 Somehow during the past months somebody changed something, and now I'm not whitelisted anymore 2017-01-15 09:54:07 yep I remember you where 2017-01-15 09:54:22 that's the benefit of having a small team, you know: communication between the members, so there's no undoing or duplication of work 2017-01-15 09:54:40 since Alpine is a small team, that's a huge advantage we have over other distros 2017-01-15 09:55:02 you made me gigle with "net neutrality" btw :D 2017-01-15 09:55:36 it should make you angry instead 2017-01-15 09:55:58 I'm angry 2017-01-15 09:56:02 good 2017-01-15 09:56:04 I'm a tiger 2017-01-15 09:56:09 rawr 2017-01-15 09:56:10 let the anger flew trough you 2017-01-15 09:56:29 I'm hungry 2017-01-15 09:56:41 :P 2017-01-15 09:56:51 still breakfasting here 2017-01-15 09:56:59 hey Palpatine, might using the Dark Side to force choke the PBL out of the list.alpinelinux.org MX ? 2017-01-15 09:57:03 mind* 2017-01-15 09:58:01 skarnet: I AM the MX 2017-01-15 09:58:10 :D 2017-01-15 09:58:11 oh, then it's your fault then 2017-01-15 09:59:41 That's the progress. 2017-01-15 10:00:06 skrzyp: Please fix yourself. 2017-01-15 10:08:11 scadu: for i3wm, you can took my i3wm-gaps, it's already updated 2017-01-15 10:14:24 skrzyp: thanks. I'll take a look. 2017-01-15 11:24:52 fabled: przemoc: I admin that I don’t watch Patchwork at all, b/c it’s plain horrible, zero automation, difficult reviewing… I don’t like wasting my time in such way 2017-01-15 11:27:16 you admin? 2017-01-15 11:27:19 :D 2017-01-15 11:27:26 s/admin/admit/ :P 2017-01-15 14:31:26 aports ML is original place to submit patches, and patchwork is kind of its extension to ease managing them (surely not without its flaws). I do use github, but I think that projects like AL should not make external services core of their workflow. practically everyone has e-mail account, but not everyone has github account and I don't think it would be good to require it if someone wants to 2017-01-15 14:31:31 become AL contributor (if, hypothetically, GH would become the only place for AL contributions). if AL steering committee will decide to move to GH for good, then I will abide it. 2017-01-15 14:34:32 przemoc: yeah, I know that, but try too look at it from the core maintainer point of view 2017-01-15 14:34:55 for symmetry, it seems some of my patches yesterday were duplicating what was already sent in some GH PRs (at least by looking at commit's author date). well, lacking parts were APKBUILD improvements, but jirutka filled them in. 2017-01-15 14:37:20 we can mirror patches from ML to GH, this will solve many issues, but someone must do it… 2017-01-15 14:37:29 I do builds myself before sending patches. what features of github make you like it so much beside travis (which isn't actually a github thing)? 2017-01-15 14:38:47 maybe you do, many do not… we can’t rely on it; CI gives more guarantee, I can also examine the log, check if there isn’t something weird, some so name differences etc. 2017-01-15 14:39:29 so AL should have its own CI 2017-01-15 14:39:53 and it’s nothing but waste of time to checkout every patch, build it on my machine and check log; I do it only when I need to 2017-01-15 14:40:16 another benefit of GH is much usable interface for doing proper code review 2017-01-15 14:41:32 I dare to say mail is also usable, but I can understand that not everyone shares this view 2017-01-15 14:42:12 own CI would be good also to test archs other than x86-64 2017-01-15 14:42:16 yes 2017-01-15 14:42:49 actually, we can test other archs even on Travis, using qemu-user, but this will not work for packages that takes a lot of time to build 2017-01-15 14:43:17 https://github.com/jirutka/alpine-chroot-install/ 2017-01-15 14:43:41 I’ll try to enable other archs for aports soon 2017-01-15 14:44:15 AL build infrastructure is already in place, so it's about repurposing its resources when it'ss free for doing test builds from patchwork. 2017-01-15 14:44:39 and set up this integration… 2017-01-15 14:44:53 but Patchwork API is unusable piece of shit 2017-01-15 14:45:22 I mean Patchwork that we run, Freedesktop’s fork has better API 2017-01-15 14:46:41 maybe it should remain at mail level, replies would show in patchwork anyway 2017-01-15 14:46:42 AL build infra is not ready for this 2017-01-15 14:47:11 mails are not very usable for this 2017-01-15 14:47:22 just try it… you’ll see 2017-01-15 14:48:09 jirutka: you mean freedesktop has actually managed to fork something and not make it worse? o.O 2017-01-15 14:48:26 skarnet: i know that it’s surprising, but yes, in this case 2017-01-15 14:50:28 przemoc: on GH you can keep adding commits to opened pull request, so contributor open PR, (1) we’ll write review, (2) contributor fixes the issues and commits changes to the PR, goto (1), merge 2017-01-15 14:52:21 actually I hate when people add commits instead of fixing broken commits (rebasing stuff). if something is not part of public history, it should be as clean as possible, because it eases code review, management and debugging in future if needed. 2017-01-15 14:52:35 przemoc: but on Patchwork, you can’t add patches to single “unit”, Patchwork can hardly identify patches that are part of one series… 2017-01-15 14:52:38 that's a good workflow, but as a developer, can I do that and still use my local git command line, w/o being tied to the GH web interface? 2017-01-15 14:52:59 przemoc: have you ever done code-review? 2017-01-15 14:53:09 yes 2017-01-15 14:53:18 przemoc: it’s much more clean when contributor add another commits instead of rebasing single commit 2017-01-15 14:53:24 przemoc: and ofc I squash it before merging 2017-01-15 14:53:58 przemoc: when contributor just overrides one commit, it’s hard to track changes he did 2017-01-15 14:54:24 it's ok-ish only if you squash. if you not, then I don't want in git log fixes for problems that can be avoided in the first place, because something wasn't merged yet 2017-01-15 14:54:24 skarnet: what do you mean, writing review or just commiting? 2017-01-15 14:54:46 przemoc: ofc, I totally agree with that 2017-01-15 14:54:48 you can easily diff two version of the patch 2017-01-15 14:55:03 just sending patches, as I do via the ML currently. 2017-01-15 14:55:05 if you have mail, you won't lose old version 2017-01-15 14:55:08 przemoc: ah, diffing diffs, so readable! 2017-01-15 14:55:32 no, diffing commits that came from applying diffs 2017-01-15 14:55:35 skarnet: well… this reminds my one item in my TODO list… I’ll try to do it before FOSDEM :) 2017-01-15 14:56:04 przemoc: so you must do it manually…? 2017-01-15 14:56:07 but usually, if you requested some fixes, you can tell whether they well applied or not 2017-01-15 14:56:16 s/well/were/ 2017-01-15 14:56:24 cool. Because if I can have a cli for patch-sending, one of my objections (the practical argument) to using GH falls :) 2017-01-15 14:56:45 the principle argument stands, though: don't let companies control FOSS, so always have an available alternative 2017-01-15 14:57:02 and as is, the alternative we have apparently isn't viable 2017-01-15 14:57:22 przemoc: sometimes I review dozens of PRs in a day and come back to them after several days, I really don’t remember the original patch and what I wrote… 2017-01-15 14:57:39 skarnet: I agree with that 2017-01-15 14:59:30 now please excuse me, I go back to https://github.com/jirutka/sh-parser :) the grammar is basically complete, I must polish AST now 2017-01-15 15:00:21 squashing is not always desirable, having good commit history w/o useless junk like "fix ...", "forgotten ..." helps everyone. actually squashing was only added quite recently to GH (recently considering how old GH is) 2017-01-15 15:00:36 I understand desire for automation 2017-01-15 15:00:50 przemoc: we never merge via GH interface 2017-01-15 15:01:00 przemoc: it’s only a mirror, we must push to git.a.o 2017-01-15 15:01:35 przemoc: I always rebase/squash manually using git 2017-01-15 15:02:13 so what GH automates if you do yourself almost everything manually? beside CI I mean. 2017-01-15 15:02:51 because your arguments were that I need to do something manually, yet apparently you do most manually too 2017-01-15 15:03:23 I'm all for automation, but let's not color GH as some ultimate tool solving almost all problems, because it does not do that 2017-01-15 15:03:28 przemoc: almost anything? 2017-01-15 15:04:01 przemoc: I enter just single command to apply patch from GH and push it to git.a.o 2017-01-15 15:05:38 pardon me, but I’d like to do something useful, we have already discussed this many times on IRC and it’s going to nowhere; 2017-01-15 15:05:40 https://imgs.xkcd.com/comics/standards.png 2017-01-15 15:06:28 ok, but it's not github-related that it's a single command only. 2017-01-15 15:07:38 no, but I’ve already explained why is GH way much more convenient for us and you just don’t want or can’t understand it 2017-01-15 15:08:17 skrzyp: this. 2017-01-15 15:09:17 github related is CI that already works and much more usable interface for doing code-reviews, but for maintainer and contributors 2017-01-15 15:10:02 ofc we can set up CI for Patchwork, somehow… but once again, *someone must do it*… no one did it so far 2017-01-15 15:10:20 and to be honest, we have much bigger issues in Alpine than this… 2017-01-15 15:10:54 GH way works fine and we still have an alternative, so if GH become evil, we’re not doomed 2017-01-15 15:14:26 btw one of my students works on Simple CI as his thesis, it’ll be designed with Alpine in mind; I have also a topic for minimalistic git manager with support for PRs, something that we may use instead of GH with same core benefits as GH provides, but this is for next year 2017-01-15 15:15:17 s/but for maintainer…/both for maintainers…/ 2017-01-15 15:21:12 skarnet: what would you prefer: a) automatically open a pull request on alpinelinux/aports once you push new branch to skarnet/aports, or b) use special command to open a pull request? 2017-01-15 15:21:26 ok, it's not that I don't want to understand, but apparently we greatly differ in how we value stuff GH provides. CI is IMHO the only improvement over what we have now, because everything else we do in CLI anyway. EOT. 2017-01-15 15:23:31 jirutka: in either case, would I have to fork the aports repo on github to start with? 2017-01-15 15:23:46 skarnet: yes, but only once 2017-01-15 15:24:11 travis ci is not dependent on gh 2017-01-15 15:24:21 skrzyp: we all know… 2017-01-15 15:25:19 skarnet: in case (a), it’d be also needed to set up webhook on your “fork” 2017-01-15 15:25:30 ok, so that point doesn't matter because I work on a local copy of the aports git repo, but I can make my gh copy a mirror of my local copy so that's ok. I prefer b) because it's more flexible 2017-01-15 15:26:21 skarnet: do you mind using bloated tool written in GoShit, or do you prefer simple shell script? 2017-01-15 15:26:39 Are you implying Golang is bloated? 2017-01-15 15:26:54 the tool would be run on my Alpine VM, which is minimal, so I'd rather keep it small :) 2017-01-15 15:26:55 asie: that tool is bloated 2017-01-15 15:27:00 jirutka: oh 2017-01-15 15:27:01 skarnet: okay 2017-01-15 15:27:04 when Go started being Shit? 2017-01-15 15:27:10 which tool ? 2017-01-15 15:27:12 skrzyp: when it was born? XD 2017-01-15 15:27:13 hub 2017-01-15 15:27:19 https://hub.github.com/ 2017-01-15 15:27:19 hub? 2017-01-15 15:27:29 hub pull-request "Testing pull-request" -b Org:master -h OrgFork:testing 2017-01-15 15:27:42 but this can be implemented even in few lines of shell with curl 2017-01-15 15:27:49 skrzyp: Go is fine for big applications, but its runtime is "ok, let me start by allocating 700MB and THEN think" 2017-01-15 15:28:08 so you don't want to use it for small stuff :D 2017-01-15 15:28:34 upx(D) 2017-01-15 15:28:36 I’d say that you don’t want to use it for anything… :P 2017-01-15 15:29:10 700MB ? seriously wtf ? 2017-01-15 15:29:36 protip 2017-01-15 15:29:36 skarnet: 700MB? nonsense 2017-01-15 15:29:46 not even Minecraft uses 700MB, and it's rather bloated and in *Java* 2017-01-15 15:29:46 it's not JVM we're talking about come on 2017-01-15 15:30:00 Go is one of the lightest runtimes I've written web tools in 2017-01-15 15:30:08 the same would happen if your system will be provided without libc 2017-01-15 15:30:12 for C 2017-01-15 15:30:18 my image/paste servers take up 4MB of RAM each. Equivalent tools in, let's say, Node.js (V8 is also very efficient), take up 60-80MB each 2017-01-15 15:30:32 4MB for go apps? 2017-01-15 15:30:41 I'll believe it when I see it 2017-01-15 15:30:54 asie: have you ever seen Lua? 2017-01-15 15:30:55 and even then I'll hire a Photoshop expert to do the checks for me 2017-01-15 15:31:39 skarnet: 8MB each, 4MB of which is shared 2017-01-15 15:31:46 i just checked with htop 2017-01-15 15:31:50 jirutka: yes he's also experienced in Lua as well 2017-01-15 15:32:06 jirutka: yes, can we get a world in which Mike Pall and the Lua developers actually agree on things again 2017-01-15 15:32:20 because LuaJIT and Lua 5.3 being two different languages is bad for both 2017-01-15 15:32:26 skrzyp: just that Lua runtime is definitely much smaller than Go… 2017-01-15 15:32:34 yes, but it's interpreted as opposed to natively compiled 2017-01-15 15:32:40 yes, that’s bad, but still better than GoShit :P 2017-01-15 15:32:47 if you use "shit" in relation to Go 2017-01-15 15:32:58 i can't help but wonder what you call Java, Python, Ruby, Perl, Haskell 2017-01-15 15:33:00 and possibly even bash 2017-01-15 15:33:09 xDD 2017-01-15 15:33:13 asie: yeah, and still Lua has smaller runtime than Go XD 2017-01-15 15:33:23 Java = shit 2017-01-15 15:33:35 Python = not very good 2017-01-15 15:33:36 that means you consider Java as bad as Go 2017-01-15 15:33:53 did you know that Forth and Lisp runtimes can be much smaller than Lua? "XD" 2017-01-15 15:34:04 in fact, I'm pretty sure picoc - a C interpreter - is smaller than Lua! 2017-01-15 15:34:10 asie: then every Go program I've seen (including Docker and a few things found on the web) all allocate 700ish MB at start 2017-01-15 15:34:16 bad coding is possible on every language 2017-01-15 15:34:17 Ruby = not very good, but I like it more than Python 2017-01-15 15:34:26 Perl = plain ugly 2017-01-15 15:34:31 Haskell = really cool 2017-01-15 15:34:45 also, Go keeps improving with every version 2017-01-15 15:34:51 Go 1.8 gets rid of most (not all, obviously) STWs 2017-01-15 15:34:53 let's start rating languages by jirutka scale from now on 2017-01-15 15:34:58 :P 2017-01-15 15:35:00 jirutka: "Haskell" "really cool" 2017-01-15 15:35:02 asie: that's hardly an accomplishment :D 2017-01-15 15:35:04 you're talking a language with a 70MB .so 2017-01-15 15:35:14 if your argument is about runtime size, you just lost all credibility to me :) 2017-01-15 15:35:17 in that regard 2017-01-15 15:35:25 aha, then please ask precisely 2017-01-15 15:35:28 ah 2017-01-15 15:35:33 but we kept talking about RAM usage 2017-01-15 15:35:36 are we awarding points again? 2017-01-15 15:35:36 your argument was about RAM usage 2017-01-15 15:35:41 i think 2017-01-15 15:35:44 okay, then Haskell is very bad 2017-01-15 15:35:48 okay then 2017-01-15 15:35:51 that's fair 2017-01-15 15:35:57 i just disagree that Go is bad in terms of runtime usage. 2017-01-15 15:36:01 there are reasons to consider it bad 2017-01-15 15:36:03 the winner is Rust for me 2017-01-15 15:36:11 skarnet: daily shitposting 2017-01-15 15:36:12 show me the programs and their /proc/smaps 2017-01-15 15:36:22 eh, they're ~200 LoC of code each 2017-01-15 15:36:25 as i said 2017-01-15 15:36:26 simple image server, simple paste server 2017-01-15 15:36:34 pardon me, but I really really hate talking about Go and other crimes against humanity, so leaving now 2017-01-15 15:36:40 okay, sorry for being a burden 2017-01-15 15:36:41 coud you be bothered to post a URL? 2017-01-15 15:36:44 and sorry for confusing you with skarnet 2017-01-15 15:36:49 skarnet: http://paste.asie.pl is the pasteserver 2017-01-15 15:36:53 the image server has no public interface 2017-01-15 15:37:18 it's not hard to have that take up a few MB of RAM, especially when it's storing everything on postgres anyway 2017-01-15 15:37:30 ... 2017-01-15 15:37:44 skarnet: this is single purpose script for opening PR, but it’s not written in shell, but Bash… if you want, I can rewrite it for you https://github.com/Idnan/github-pr 2017-01-15 15:38:18 GC languages will always take more RAM than non-GC languages, obviously 2017-01-15 15:38:23 with GC, it's a tradeoff between speed and memory usage 2017-01-15 15:38:30 asie: I meant a URL to a. the source code, b. a paste with the /proc/$pid/smaps of the running Go executable. Not the interface of the program, which I don't care about. 2017-01-15 15:38:35 ah 2017-01-15 15:38:38 skarnet: this one can also list your PRs etc. but it’s not very lightweight, written in Python and has some dependencies https://github.com/dataxu/github-pr 2017-01-15 15:38:48 jirutka: thanks, I'll bookmark it and look at it later. 2017-01-15 15:38:54 the bash thing. 2017-01-15 15:39:00 skarnet: smaps? not maps? 2017-01-15 15:39:08 smaps. 2017-01-15 15:39:16 easier to read by a human. 2017-01-15 15:39:17 i must be blind, i can't find such a file 2017-01-15 15:39:39 is there a tool for this? 2017-01-15 15:39:40 gotta love grsecurity 2017-01-15 15:39:43 oh, yeah 2017-01-15 15:39:57 why are people still using this 2017-01-15 15:40:19 no file on root either 2017-01-15 15:41:11 eh. whatever. arguing about memory usage is not the way i want to spend my day 2017-01-15 15:41:20 i do it enough when working on that one mod for minecraft which fixes bloated memory usage in mods 2017-01-15 15:41:36 jirutka: I think you already linked me to that page. I remember seeing the "Install: curl | sudo sh" and saying NEXT! 2017-01-15 15:42:01 skarnet: just ignore this stupidity, this script doesn’t need root privileges, just author is idiot 2017-01-15 15:42:18 and you want me to run a thing made by an idiot author? 2017-01-15 15:42:41 skarnet: as I said, if you want, I can rewrite it… 2017-01-15 15:43:04 skarnet: the question is if this interface/feature-set is what you want 2017-01-15 15:43:22 ok, I'll take a look, not now though. 2017-01-15 15:53:20 jirutka: yes, that's the kind of tool I'd like to have. And it looks short enough - I could probably rewrite it myself, so unless you want to make it available for everyone shortly, don't spend your time on this. Thanks for pushing me to look inside the actual script, I really didn't expect it to be that simple. 2017-01-15 15:54:41 skarnet: you’re welcome :) 2017-01-15 17:51:37 Hi there 2017-01-15 17:52:48 Just playing with Alpine Linux and I really love it so far! Could somebody give me a hint about the modloop image? Why is it required, if kernel modules are also included into the initramfs? 2017-01-15 17:53:07 Thanks! 2017-01-15 17:56:26 to the best of my knowledge, they aren't 2017-01-15 17:56:36 the initramfs may include some essential modules needed to boot 2017-01-15 17:56:55 modloop likely contains the full set you want for a completely running/working system 2017-01-15 17:57:39 (verifying now, this is indeed the case) 2017-01-15 17:58:05 from the initramfs /init: 2017-01-15 17:58:07 # load available drivers to get access to modloop media 2017-01-15 17:58:10 ebegin "Loading boot drivers" 2017-01-15 17:58:17 those modules are only needed to get access to modloop itself 2017-01-15 17:58:20 r0l1 ^ 2017-01-15 18:00:23 Thanks for the fast reply! Sounds plausible. Got it now. 2017-01-15 18:02:42 I am trying to implement a safe update process similar to CoreOS A/B update procedure. Is there anything similar present so far? 2017-01-15 18:03:39 from a quick Googling, it seems like a dual-rootfs scheme? 2017-01-15 18:04:41 it seems kind of painful to have to reboot for updates... 2017-01-15 18:04:46 Extactly. With an automatic rollback if the reboot to the new system fails. They managed this with a patched grub bootloader. 2017-01-15 18:04:53 (but afaik, I don't think anything in Alpine does something similar right now) 2017-01-15 18:06:27 Most of the time a reboot is required anyway, due to kernel updates... Worked quite a while with CoreOS and it is a really stable setup. But I don't like the systemd dependency... 2017-01-15 18:06:36 i guess you could do something similar with having a default apk prefix and switch_root(8) 2017-01-15 18:06:51 well, the advantage of grsec kernels is that they don't need updating that often :) 2017-01-15 18:08:34 a quick-n-dirty approach I could come up with would mount the inactive partition as /new or something, modify apk to use /new as default prefix, and execute a script that pivot_root(8)s into that root and execute a post-upgrade script that possibly switches the bootloader config to boot into that partition 2017-01-15 18:08:54 and mounts the previously active partition again as /new in the now-active partition 2017-01-15 18:09:32 but it seems coreos and alpine have fundamentally different approaches to upgrading in the first place seeing as coreos just downloads and extracts a new rootfs 2017-01-15 18:09:43 thats not really similar as alpine has like... packages and stuff 2017-01-15 18:09:44 :p 2017-01-15 18:10:04 the dual rootfs approach makes sense for firmware, i.e. controlled read-only rootfs 2017-01-15 18:10:13 it's an industry standard 2017-01-15 18:10:28 but for a user-controlled rootfs, which is read-write most of the time... 2017-01-15 18:10:44 ... the guarantee that you won't b0rk your rootfs is nonexistent anyway 2017-01-15 18:11:09 so it would be a lot of effort for a very dubious benefit 2017-01-15 18:11:15 r0l1: something you may enjoy is apkovls made by lbu(1) 2017-01-15 18:11:45 it's a way to create a .tar containing all user-made changes from the 'default' rootfs as you'd get when installing alpine + 2017-01-15 18:12:07 you can pass them to the initramfs as kernel params to have them applied automatically on boot 2017-01-15 18:13:38 I'm currently using it for machines that bootstrap almost entirely from the internet (modulo kernel/initramfs) 2017-01-15 18:13:51 every boot they setup a stock alpine install in RAM and then apply the apkovl 2017-01-15 18:19:31 Thanks for the hints! I had a look at lbu and apkovls. Thats really smart way of handling overlays. Didn't know you could even pull them from the internet! That's awesome! I am setting up a small firmware with an read-only rootfs and an tmpfs overlay. Coming from CoreOS, I wanted to implement the A/B update process. I had a look at the alpine-iso Makefile. Based on that I managed to create a custom Makefile with a custom init script, moun 2017-01-15 18:19:34 ting a squashfs base system image and calling switch_root. The boot directory looks as following right now: /boot/a/{vmlinuz, initramfs, system.img} and /boot/b/{vmlinuz, initramfs, system.img} 2017-01-15 18:20:48 An update is quite simple, by just deploying a new system image containing all the three files. 2017-01-15 18:21:16 Copy it to /boot/b and try to reboot. Fallback to a on failure 2017-01-15 18:21:59 Do you think that the Alpine community is interested to include this kind of setup? I could prepare a pull request... 2017-01-15 18:26:41 hmm, it might be worth asking the other developers or even the alpine-devel mailing list :) 2017-01-15 18:30:56 Thanks. I'll do that as soon as I am ready... 2017-01-15 18:34:15 Shiz: neat, didn't know about that, that would actually simplify some of my build testing a lot if I could just boot a vm with everything I need setup for building 2017-01-15 22:16:10 Shiz: just made some great progress so far and the boot process into the squashed root filesystem works as expected. However I am not really sure about the apk install scripts. With a normal Alpine Linux system these scripts are executed during the boot. However in my case they have been executed while building the squashfs root image. Does anybody know, if this has a negative effect? Are there any system related settings, which do need to 2017-01-15 22:16:12 be generated during boot? 2017-01-15 22:16:14 Thanks! 2017-01-15 22:31:54 r0l1: what script during boot loads the apk's? 2017-01-16 00:00:57 The init script of the initramfs: http://git.alpinelinux.org/cgit/mkinitfs/tree/initramfs-init.in 2017-01-16 00:57:00 <^7heo> r0l1: what are you trying to achieve? 2017-01-16 01:01:33 I am implementing an A/B update procedure for embedded systems with an automatic fallback. The root system is bundled into a squashfs archive, which is mounted read-only with an additional overlaytmpfs mount. During an update the new bundle, kernel and initramfs are installed to /boot/b. Afterwards the system reboots into the b system, falling back to /boot/a on failure. 2017-01-16 01:02:10 <^7heo> what is an A/B update? 2017-01-16 01:03:45 https://coreos.com/os/docs/latest/sdk-disk-partitions.html 2017-01-16 01:03:52 https://coreos.com/os/docs/latest/sdk-disk-partitions.html 2017-01-16 01:04:44 https://coreos.com/os/docs/latest/manual-rollbacks.html 2017-01-16 01:10:59 <^7heo> if it's that complicated to explain that it needs 2 links with one pasted twice... 2017-01-16 01:11:02 <^7heo> I'll pass. 2017-01-16 01:11:04 <^7heo> I need to sleep. 2017-01-16 01:19:01 Sorry. The current system is present in an A directory. The new system is installed to B. Afterwards the system reboots into the updated B system. The next update is then installed to A... 2017-01-16 01:19:08 Thanks anyway. 2017-01-16 01:21:05 <^7heo> ah 2017-01-16 01:22:22 <^7heo> so a simulated update to validate the real one. 2017-01-16 01:24:01 Yeah. If the boot succeeds, then the system starts all it's services and starts operating. If something fails, the bootloader will notice and fallback to the previous "old" state before the update. 2017-01-16 01:24:32 <^7heo> I do that myself with multiple computers ;) 2017-01-16 01:24:38 <^7heo> but yeah I'm left with an install to fix if it fails. 2017-01-16 01:25:16 System A running -> installs updated system to B -> reboot to B -> reboot to A on failure 2017-01-16 01:25:22 Cool! How did you solve it? 2017-01-16 01:25:47 <^7heo> Solve the install to fix if it fails? 2017-01-16 01:25:58 <^7heo> I have a usb stick formatted with Alpine 2017-01-16 01:26:12 <^7heo> I boot on it 2017-01-16 01:26:20 <^7heo> I fix the system manually 2017-01-16 01:26:44 Ah ok. Got it. In my case it has to be automatic, because it is a firmware for an embedded system ... 2017-01-16 01:27:36 Just have seen, that you did commit to the mkinitfs project. Could you please fix line 186 2017-01-16 01:27:51 i) init=$OPTARG;; 2017-01-16 01:27:54 to 2017-01-16 01:27:59 i) init="$OPTARG";; 2017-01-16 01:28:03 http://git.alpinelinux.org/cgit/mkinitfs/tree/mkinitfs.in 2017-01-16 01:28:05 thanks! 2017-01-16 01:30:09 <^7heo> r0l1: I'm not saying I wouldn't like an automated system ;) 2017-01-16 01:31:09 I added an additional bundle kernel command line argument to support bundle based boots. This is required for this kind of update procedure. If you're interested to include it into Alpine Linux, I could do a pull request with my additions :) 2017-01-16 01:32:49 <^7heo> well, maybe :) 2017-01-16 01:33:08 <^7heo> please show me what you've done when it's ready :) 2017-01-16 01:35:29 Cool. I'll publish a forked github repo soon. Trying to keep the changes as small as possible. The initramfs init script requires only a few additional lines... 2017-01-16 01:37:25 <^7heo> just giving you some heads up 2017-01-16 01:37:40 <^7heo> the repo can change here and there unexpectedly ;P 2017-01-16 01:41:16 That should be fine ^^ I'll keep track with a copy... 2017-01-16 02:42:10 if I've found a line that needs fixing, what do I do? 2017-01-16 02:43:17 http://git.alpinelinux.org/cgit/mkinitfs/tree/initramfs-init.in -- line 58, openssl needs to be switched to libressl. 2017-01-16 02:44:46 otherwise v3.5.0 will fail to boot an encrypted LBU. 2017-01-16 03:24:58 I found two additional small bugs. We should report them at: https://bugs.alpinelinux.org/projects/alpine/issues 2017-01-16 04:21:31 ~ cat nop.c 2017-01-16 04:21:33 void _start(void){ 2017-01-16 04:21:35 __asm__("nop"); 2017-01-16 04:21:37 } 2017-01-16 04:21:39 ~ gcc -nostdlib -static -s -o nop nop.c 2017-01-16 04:21:41 ~ file nop 2017-01-16 04:21:43 nop: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, stripped 2017-01-16 04:21:45 ~ ls -lah nop 2017-01-16 04:21:47 -rwxr-xr-x 1 skrzyp users 4.7K Jan 16 05:20 nop 2017-01-16 04:21:49 *applause.wav* 2017-01-16 07:18:56 skrzyp: ? 2017-01-16 07:21:31 I just created my first issue :) 2017-01-16 07:25:39 I'm not sure how this works. Can I add the matching commit to fix it? 2017-01-16 08:04:01 rawf: http://www.redmine.org/projects/redmine/wiki/RedmineSettings#Referencing-issues-in-commit-messages 2017-01-16 08:17:28 clandmeter: Thanks. But I wasn't sure on how to submit a commit. 2017-01-16 08:17:53 you can send patches via ML or via GH 2017-01-16 08:32:34 Woohoo. Added my first pull request via GH. 2017-01-16 09:43:41 does mdev react on regulatory domain requests from kernel? 2017-01-16 09:43:49 <^7heo> moin 2017-01-16 09:44:12 <^7heo> what is a regulatory domain request? 2017-01-16 09:44:22 xentec: regulatory domain requests are not sent to the hotplug helper. 2017-01-16 09:44:24 <^7heo> sounds like a product from redhat to me 2017-01-16 09:44:31 it's a wifi thing. 2017-01-16 09:44:35 <^7heo> ah 2017-01-16 09:44:55 <^7heo> thanks skarnet 2017-01-16 09:45:18 uh.. well how do I make tell crda to insert the correct regDom? 2017-01-16 09:45:29 it's not `cat /proc/sys/kernel/hotplug` that's invoked, it's /sbin/crda. (there may be a kernel option to change that binary location but I've never bothered to check it) 2017-01-16 09:46:38 I have no idea. The only implementation of crda I know is in Python and I'm not running that, and the only Linux machine I need wifi on is my embedded router, so I embedded the regdb in the kernel and said screw crda. 2017-01-16 09:47:40 <^7heo> ACTION wonders how much crap he is running just because "it works that way" 2017-01-16 09:47:42 but if you have /sbin/crda, then the kernel invokes it directly and from there it's a question of knowing the interface. 2017-01-16 09:48:33 If you want to RE the interface: wrap /sbin/crda with something that prints its command line first. 2017-01-16 09:51:03 for context: I running alpine on rpi3 in diskless mode and am already somehow discouraged by all this nice red dmesg lines from brcmfmac module 2017-01-16 09:51:07 like brcmfmac: brcmf_cfg80211_reg_notifier: not a ISO3166 code 2017-01-16 09:52:05 looks like something's broken in the way requests are sent to that module 2017-01-16 09:52:31 ISO3166 is just a fancy way of saying "two-letter country code" 2017-01-16 09:52:56 yeah. doesn't surprise me with Broadcom 2017-01-16 09:53:51 what does "iw reg get" tell you? 2017-01-16 09:54:41 even after setting it to DE, it's still 00 2017-01-16 09:54:52 <^7heo> wait a minute... is crda ONLY for selecting what channel should be available in what country? 2017-01-16 09:55:13 ^7heo: and the exact frequency ranges for each channel. 2017-01-16 09:55:58 and other minute details of what extension to 802.11b/g/n/ac is authorized. 2017-01-16 09:56:48 <^7heo> wtf, since when is it so complicated to configure hardware? 2017-01-16 09:56:55 running 'COUNTRY=DE crda' returns 'Failed to set regulatory domain: -22' 2017-01-16 09:57:49 ^7heo: wifi is hard. Radio protocols are immensely complex. That's why I like Ethernet, cable or whatever - any time you go wireless, the complexity goes an order of magnitude higher. 2017-01-16 09:58:26 I think no one here has a HAM license 2017-01-16 09:58:28 xentec: is your crda database in the right place, accessible, etc. ? 2017-01-16 09:58:41 <^7heo> skarnet: right but where is the need for negociation when it's pre-computed data anyway? 2017-01-16 09:59:01 I can't help you much more, because I've only dabbled in regdb long enough to make it work for me and stopped touching it as soon as I could. 2017-01-16 09:59:21 ^7heo: regulations change. Data is stable, but humans are fickle. 2017-01-16 09:59:47 <^7heo> xentec: sttupid question: can you, and did you strace it? 2017-01-16 10:00:10 crda is a Python program. stracing it will produce more output than is readable. 2017-01-16 10:00:31 Better wrap it with environment printers than strace it. 2017-01-16 10:03:16 I think I let it derail to much: i specifically asked about mdev in the first place, because with eudev it works 2017-01-16 10:03:34 also /sbin/crda is a binary and https://pkgs.alpinelinux.org/package/edge/testing/armhf/crda doesnt list a python dependency 2017-01-16 10:03:56 weird. Maybe it has changed in 2 years? 2017-01-16 10:04:39 do you have rules for crda/regdb in your eudev? the mechanism may have changed in recent kernels. 2017-01-16 10:05:17 dunno, stock eudev from alpine repos 2017-01-16 10:05:29 ncopa, I've read about you wanting to use nldev. is it stable? 2017-01-16 10:06:06 afaik nldev is just a netlink listener for mdev 2017-01-16 10:08:03 http://lists.busybox.net/pipermail/busybox/2015-April/082773.html 2017-01-16 10:09:11 yes. I already have one: https://skarnet.org/software/s6-linux-utils/s6-devd.html 2017-01-16 10:36:33 I see that the "who" busybox applet on alp-3. 4 is missing. correct ? 2017-01-16 10:37:57 <_ikke_> grharry: I believe that's because the facilities for that are also missing 2017-01-16 10:38:03 <_ikke_> wtmp and such 2017-01-16 10:38:10 indeed 2017-01-16 10:40:51 lsof -ni | grep sshd ... is missing the console logins ... any other way to find out who is on the system ?? 2017-01-16 10:45:11 I think the canonical answer to that is "why would you want to do that" 2017-01-16 10:46:57 <^7heo> ncopa: ping 2017-01-16 10:47:51 <^7heo> skarnet: grharry is searching for who(1) 2017-01-16 10:48:16 <^7heo> also I dunno for you, but my busybox who applet is missing. 2017-01-16 10:49:28 ^7heo: yes, and _ikke_ gave the right answer. There's no utmp/wtmp on musl for security reasons. 2017-01-16 10:49:29 <^7heo> it is because who is: /usr/bin/who -> /bin/busybox but busybox doesn't have who. 2017-01-16 10:49:39 <^7heo> oh right. 2017-01-16 10:49:42 <^7heo> sorry for spamming. 2017-01-16 10:50:04 <^7heo> why do we ship /usr/bin/who tho? 2017-01-16 10:50:30 <^7heo> (since it's a missing feature) 2017-01-16 10:50:33 do we? ncopa can check the busybox packaging I guess. 2017-01-16 10:50:39 <^7heo> I have it. 2017-01-16 10:50:54 <^7heo> maybe it's an artefact I have from an old, previous setup; that was never removed... 2017-01-16 10:50:58 <^7heo> but I have it. 2017-01-16 10:51:15 <_ikke_> It's part of coreutils 2017-01-16 10:51:27 anyway, the important thing an admin has to do is check the running _processes_ on the system, and variations around "ps" work for that. Knowing who is currently logged, on the other hand, is a privacy violation. 2017-01-16 10:51:35 <^7heo> _ikke_: we don't use GNU Coreutils on alpine. 2017-01-16 10:51:58 <^7heo> skarnet: depends what you use it for. 2017-01-16 10:52:05 <_ikke_> ^7heo: the package is there 2017-01-16 10:52:33 <^7heo> _ikke_: there are many more pacakge in alpine than necessary, let alone sane. 2017-01-16 10:53:14 <^7heo> _ikke_: also obviously I was talking about the busybox version. 2017-01-16 10:53:21 <^7heo> (i.e. a symlink) 2017-01-16 10:54:01 ^7heo: it is a privacy violation no matter what you use it for. The only way it's not is if your users have given you consent prior to your checking. That's why sysadmins in companies make everyone sign a charter, which often includes consent for that kind of things. 2017-01-16 10:54:22 <^7heo> hmm 2017-01-16 10:54:59 <^7heo> skarnet: well it's not a violation of privacy if it's to monitor your kid's login time, I'm pretty sure of that. 2017-01-16 10:55:17 <^7heo> and your own personal machine that you have @home. 2017-01-16 10:55:27 <^7heo> (not that it should matter, but just to be precise) 2017-01-16 10:55:51 <_ikke_> what about checking if non-authorized users are logged in? 2017-01-16 10:55:54 "The privacy violation" thing ... seems very vague to me ... 2017-01-16 10:56:11 <^7heo> grharry: it's not vague at all. 2017-01-16 10:56:20 <_ikke_> You have to be german to understand it :P 2017-01-16 10:56:26 <^7heo> also I'm curious why utmp/wtmp is a security problem. 2017-01-16 10:56:40 <^7heo> _ikke_: s/german/intelligent/ 2017-01-16 10:57:02 <^7heo> _ikke_: but yeah that's an issue. With so many stupid people around, it's fucking up our rights big time. 2017-01-16 10:57:17 <^7heo> freedom is the only think that wears off when you do NOT use it. 2017-01-16 10:57:53 ^7heo: if you have to check "who" to monitor your kid's login time, I question your parenting methods. :P 2017-01-16 10:58:14 <^7heo> skarnet: could be a script doing it, for, you know, automation purposes. 2017-01-16 10:58:37 <^7heo> skarnet: just giving an example anyway. I have no kids, just a dog, and he doesn't yet know how to login in sh(1) 2017-01-16 10:59:26 this turns out to be a political/philosophical discussion and it includes quite a bit of dark and agnostic corners ... 2017-01-16 10:59:53 "pretty girls are everywhere" 2017-01-16 10:59:59 <^7heo> the separation between technical and political is only made for the simple minds to understand parts of complex matters. 2017-01-16 11:00:13 <^7heo> in the end, a contract could be written (described) in procedural code. 2017-01-16 11:00:20 my point stands. It's nobody else's business when and how long I'm logged in. If I'm logged in to a system you own, with your permission, it's STILL not your business when I'm logged in. It's your business to know what I run on your machine, and you have "ps" for that, but that's it. 2017-01-16 11:00:29 <^7heo> (contract as in text binding two or more entities, by law) 2017-01-16 11:00:53 no need to search in "Germania" for intelligence or the contrary ... 2017-01-16 11:01:35 <^7heo> skarnet: that's true; but I'm still curious why utmp/wtmp is a security problem. 2017-01-16 11:02:38 <^7heo> skarnet: also the issue with ps is that the best you can do is a poll-based check. 2017-01-16 11:03:23 <^7heo> skarnet: so if the software that monitors what users do uses ps, say, every second, there still is plenty of time for a software to run without being seen. 2017-01-16 11:04:07 On a system that I own ... its my right to know when-what-who ... after all I have the right to format it ... break it ... or piss on it ! 2017-01-16 11:04:18 "ps" is of course an approximation, you have a lot more tools to do that. 2017-01-16 11:04:23 <^7heo> ok :) 2017-01-16 11:04:31 <^7heo> fair enough. Sorry I took it so literally. 2017-01-16 11:04:40 for the security issues of utmp, search the musl ML archives 2017-01-16 11:04:45 <^7heo> ok. 2017-01-16 11:04:50 <^7heo> thanks :) 2017-01-16 11:06:13 grharry: you can do whatever you want to your *system*, that does not give you the right to snoop on *users*. Unless, again, they have given you consent to do so. You can contractually tie the use of your system to acceptation of your right to snoop, which is what "terms of service" usually do. 2017-01-16 11:06:50 That's up to you and the legislation in the place you live in ... 2017-01-16 11:07:08 Are you a lawyer or a law representative ? 2017-01-16 11:07:37 <^7heo> skarnet: I wouldn't feel the troll on a monday ;) 2017-01-16 11:07:44 <^7heo> s/eel/eed/ 2017-01-16 11:07:49 IANAL, but I live in a country that's - for now - sensitive enough about privacy rights, and it's always good to be explicit in your terms of service. 2017-01-16 11:07:55 <^7heo> sorry, still in pre-coffee time here. 2017-01-16 11:08:46 yeah .... EUROPE ... !!!!! Big TROLL TIME !!! 2017-01-16 11:09:00 LOL !!! 2017-01-16 11:09:31 GERMAN-OPE 2017-01-16 11:10:31 <^7heo> I still fail to see what the people who try to argue against freedom-aware and freedom-respectful laws have to win from it. 2017-01-16 11:10:53 <^7heo> I mean, I get that the goverment and the companies who sell surveillance software have to gain from it... 2017-01-16 11:13:09 ^7theo: It's simple ... If you wanna watch/download porn ... eliminating the "who" command prevents your boss from finding that out!!!! LOL 2017-01-16 11:14:59 <^7heo> but I observe this "anti-freedom-you're-a-pathological-paranoiac-nutcase" reaction rather often. 2017-01-16 11:15:14 also ... on next release ... we should remove from alpine ... tcpdump, ngrep, lsof ... 2017-01-16 11:15:27 <^7heo> I don't believe that many people work in surveillance related government positions or firms... 2017-01-16 11:16:24 ^7heo: that's purely politics. Governments and media like to sell the idea that surveillance=security, and lots of people buy it. 2017-01-16 11:16:33 <^7heo> but still 2017-01-16 11:16:51 <^7heo> I know almost nobody who has been mugged in the last year or more. 2017-01-16 11:17:34 I think with the tcpdump, ngrep, and lsof you can snoop "Big-Time" ... what is the "who" command provides that these wont see ?? 2017-01-16 11:18:02 ^7heo, grharry hi 2017-01-16 11:18:14 Hey ncopa ... 2017-01-16 11:18:24 been long time ... !! ;) 2017-01-16 11:18:29 indeed 2017-01-16 11:18:33 U good ?? 2017-01-16 11:18:42 how's life? 2017-01-16 11:18:47 im all good thanks 2017-01-16 11:19:08 ok ... ... kids grow ... me 2 !!! :( 2017-01-16 11:19:35 i suppose thats better than the alternative :) 2017-01-16 11:19:41 re "who" 2017-01-16 11:19:51 the short answer is: doesnt work 2017-01-16 11:20:31 doesnt work with musl libc 2017-01-16 11:20:59 i think the reason is that musl devs dont think it is possible to implement utmpx in a sane way 2017-01-16 11:21:04 or something like that 2017-01-16 11:22:07 <^7heo> ncopa: btw, do you think you'd have time for my PR today? 2017-01-16 11:22:30 probably not :-( 2017-01-16 11:22:33 url? 2017-01-16 11:22:34 <^7heo> :'( 2017-01-16 11:22:45 <^7heo> https://github.com/alpinelinux/mkinitfs/pull/3 2017-01-16 11:22:57 <^7heo> it's not THAT big ;) 2017-01-16 11:24:35 <^7heo> I mean, we could ask jirutka to review... to be sure that there's no unnecessary stuff there. 2017-01-16 11:24:48 <^7heo> But AFAIR he already requested that to be merged months ago. 2017-01-16 11:24:59 <^7heo> So I'd assume he has reviewed it at the time. 2017-01-16 11:25:27 <^7heo> (and the current state of the PR is a subset of what he then reviewed) 2017-01-16 11:25:51 <^7heo> ncopa: e.g. https://github.com/alpinelinux/mkinitfs/pull/3#issuecomment-257150125 2017-01-16 11:25:58 <_errm> hey I opened a PR for a new aport, does anyone have some time to review what I did https://github.com/alpinelinux/aports/pull/714 2017-01-16 11:26:51 <_ikke_> _errm: small remark, you could squash the 2 commits 2017-01-16 11:27:02 <^7heo> I was gonna say 2017-01-16 11:27:23 <^7heo> the -depends="nodejs" in one commit seem a little much :P 2017-01-16 11:27:36 that's a big change :P 2017-01-16 11:27:53 <_errm> yeah ok, I left them for now as I wasn’t sure if dropping the dependency was the correct thing to do .... 2017-01-16 11:28:40 <^7heo> skarnet: not in the code :) 2017-01-16 11:28:44 <^7heo> skarnet: but I get your point. 2017-01-16 11:29:31 <_errm> since nodejs-current has replaces="nodejs nodejs-lts" I would have thought that it should work, but in practice no ... 2017-01-16 11:30:35 <_errm> in debian land its possible to have multiple packages provide a meta package for cases like this… does apk have something similar? 2017-01-16 11:31:20 apk supports "provides" 2017-01-16 11:32:22 <_errm> ok so if nodejs-current had provides="nodejs" it should work? 2017-01-16 11:33:35 yes 2017-01-16 11:35:05 <_errm> Is it ok to change that in this PR, or should I do a seperate one … 2017-01-16 11:36:01 _errm: well, probably it will be better to do it in the same PR 2017-01-16 11:36:52 <_ikke_> separate commit, same PR 2017-01-16 12:18:23 how do I download apk source code? 2017-01-16 12:31:16 http://git.alpinelinux.org/cgit/apk-tools/ 2017-01-16 13:33:28 <_errm> does this look ok now https://github.com/alpinelinux/aports/pull/714 2017-01-16 13:54:29 What is the best way to help with Alpine ? I am a software engineer looking to help 2017-01-16 14:00:23 <^7heo> Borgoaia: reading the existing apkbuild manifests, and submit github PRs/git diffs via ML to maintain them, simplify them and make them more consistent. 2017-01-16 14:01:08 <^7heo> Borgoaia: also we need a few tools, mostly related to the doc' 2017-01-16 14:01:21 What kind of tools ? 2017-01-16 14:03:09 I am looking at https://wiki.alpinelinux.org/wiki/Alpine_Linux:Contribute :) 2017-01-16 14:04:24 <^7heo> well lately it has been discussed to migrate/mirror the doc from the Wiki to some more developer-friendly place. 2017-01-16 14:04:34 <^7heo> i.e. using git or whatnot. 2017-01-16 14:05:04 <^7heo> I personally do not want to contribute to a doc' that is a blob of binary somewhere on a server and that you can only edit/view with a browser. 2017-01-16 14:05:21 <^7heo> it seems the opposite of what alpine stands for. 2017-01-16 14:05:36 <^7heo> but some people here disagree with my POV 2017-01-16 14:05:51 <^7heo> I wish I would find time to do that. 2017-01-16 14:05:58 <^7heo> (mirroring at first) 2017-01-16 14:08:04 okay :) 2017-01-16 14:08:17 <^7heo> also 2017-01-16 14:08:21 Like hosting it on github 2017-01-16 14:08:27 <^7heo> we have something already 2017-01-16 14:08:31 Is there any need for c++ skills ? 2017-01-16 14:08:52 <^7heo> definitely not C++ 2017-01-16 14:08:55 <^7heo> if anything C. 2017-01-16 14:09:14 <^7heo> but coming from a C++ background will require discipline and carefulness. 2017-01-16 14:09:51 <^7heo> Borgoaia: https://github.com/adocs/adocs 2017-01-16 14:10:25 <^7heo> here is the "very rough" attempt at having docs in a dev friendly format. 2017-01-16 14:10:51 okay :) i see! 2017-01-16 14:11:01 What kind of C work is there ? 2017-01-16 14:11:01 <^7heo> it contains almost nothing, but I wanted to develop something to mirror the existing wiki, first. 2017-01-16 14:11:21 <^7heo> well, I dunno what we do in C exclusively but... 2017-01-16 14:11:24 I like the thing you made 2017-01-16 14:11:29 <^7heo> like? 2017-01-16 14:11:40 please be nice and don't discourage people who want to contribute. We can always find use for skills, even C++ skills. We always need more people to work on the power treadmills, the turnover is pretty high there - they collapse fast. 2017-01-16 14:11:49 The github. I know it is a first step 2017-01-16 14:12:01 <^7heo> Borgoaia: ah yeah. It's jomat who did it. 2017-01-16 14:12:09 <^7heo> skarnet: was I sounding too harsh? 2017-01-16 14:12:37 <^7heo> skarnet: I just don't want to endorse people to come over with C++ and start writing OOP everywhere. 2017-01-16 14:12:43 I had C in school it is just a little while ago i used it reall. 2017-01-16 14:12:47 <^7heo> skarnet: and start over-engineering stuff like in GNU 2017-01-16 14:13:07 https://en.wikipedia.org/wiki/Joke 2017-01-16 14:13:08 I would be happy to write in C again 2017-01-16 14:13:29 Is it kernel modules or something that needs to be done in C ? 2017-01-16 14:13:52 basically all the low-level software layers need to be. 2017-01-16 14:13:58 That includes the kernel and also low userspace. 2017-01-16 14:14:06 Ah very nice 2017-01-16 14:15:03 skarnet and ^7heo can i ask what kind of programming you do so i get a better understanding of the things i can help with ? 2017-01-16 14:16:35 if you're looking to contribute to Alpine in particular, I'd wait for an answer from the big bosses (ncopa, fabled, clandmeter etc.) The rest of us are mostly Alpine users and occasionally contributors. 2017-01-16 14:18:09 okay 2017-01-16 14:19:28 <^7heo> Borgoaia: I currently contribute to the boot system 2017-01-16 14:19:51 That sounds awesome :) 2017-01-16 14:19:51 <^7heo> Borgoaia: I implemented a deported-header feature for LUKS in it 2017-01-16 14:20:12 <^7heo> but I need to refactor it with the patches that were introduced while it was waiting 2017-01-16 14:20:36 i think adding some form of boot-level auth/crypto would be nice 2017-01-16 14:20:41 things i'm randomly looking at from the side 2017-01-16 14:20:57 <^7heo> Shiz: definitely. 2017-01-16 14:21:05 <^7heo> Shiz: that's my next goal after mkinitfs 2017-01-16 14:21:25 i'm currently looking at securing a server stack at boot-level 2017-01-16 14:21:31 there's some interesting options :P 2017-01-16 14:21:34 I see. Yeah i would like to do some programming tasks for alpine. Just starting small so i don't get throw into the sea at once. 2017-01-16 14:21:40 <^7heo> Shiz: you coming to fosdem? 2017-01-16 14:21:50 that's in two weeks isn't it 2017-01-16 14:21:57 <^7heo> yep 2017-01-16 14:22:00 <^7heo> brussels 2017-01-16 14:22:19 i think i can make that, but i don't have a staying place there yet :P 2017-01-16 14:22:49 <^7heo> book a hotel before they're all 300€ 2017-01-16 14:22:50 <^7heo> :D 2017-01-16 14:23:10 D: 2017-01-16 14:23:21 <^7heo> got a room with 2 beds last night for 95€ 2017-01-16 14:23:28 i was there the previous two years, gotta check my cash reserves right now 2017-01-16 14:23:46 ACTION just came back from two weeks of Japan 2017-01-16 14:23:54 <^7heo> It shouldn't cost much aside the tickets and the hotel, should it? 2017-01-16 14:24:00 <^7heo> ACTION has never been in Brussels 2017-01-16 14:24:06 yeah, but even then 2017-01-16 14:24:08 <^7heo> ACTION has no idea what the prices are like over there 2017-01-16 14:24:11 cash reserves are very low right now 2017-01-16 14:24:13 :p 2017-01-16 14:24:16 <^7heo> oh ok 2017-01-16 14:24:51 Get a airbnb ? 2017-01-16 14:24:58 <^7heo> depending on the price I can probably beer and/or sandwitch you. 2017-01-16 14:25:13 did ncopa's talk get approvedb tw? 2017-01-16 14:25:31 i'll see if i can arrange something thru $dayjob 2017-01-16 14:25:33 <^7heo> no idea. 2017-01-16 14:25:41 with a bit of luck i'll have them pay for it 2017-01-16 14:25:44 :p 2017-01-16 14:26:05 <^7heo> Shiz: yes it apparently has. 2017-01-16 14:26:16 <^7heo> https://fosdem.org/2017/schedule/speaker/natanael_copa/ 2017-01-16 14:28:01 it was 2017-01-16 14:28:10 (sitting in a mtg now..) 2017-01-16 14:28:44 <^7heo> when meetings are so intense that you have to abbreviate their names. 2017-01-16 14:29:18 <^7heo> (or you can't handle them) 2017-01-16 14:30:26 I have plane tickets to FOSDEM! 2017-01-16 14:30:56 <^7heo> yeah me too :P 2017-01-16 14:31:04 <^7heo> jirutka: or did you just get them? 2017-01-16 14:31:29 ncopa: this will be useful to you: https://goo.gl/weoMT0 2017-01-16 14:32:08 <^7heo> skarnet: 2012 called. It wants its buzzwords back. 2017-01-16 14:32:34 who will be there? ncopa, clandmeter?, skarnet, ^7heo, anyone else? 2017-01-16 14:32:53 <^7heo> I used to write applications to companies while taking care to put ALL those words in. 2017-01-16 14:32:54 we can arrange mini Alpine conf here :P 2017-01-16 14:33:19 <^7heo> yeah definitely. 2017-01-16 14:33:26 that's basically what ncopa's talk is about, isn't it? XD 2017-01-16 14:33:44 <^7heo> an alkonf. 2017-01-16 14:33:45 I’ll arrive at Friday morning and going back at Sunday evening 2017-01-16 14:33:53 <^7heo> (as opposed to alconf) 2017-01-16 14:34:06 <^7heo> jirutka: same same. 2017-01-16 14:35:07 anyone interested in some sightseeing at Friday? this will be my first time in Brussels 2017-01-16 14:35:14 <^7heo> skarnet: just to give you an idea about how old your buzzwords are: there's not even container and microservices in there... 2017-01-16 14:35:40 <^7heo> jirutka: oh wait no I arrive SAT morning. 2017-01-16 14:35:44 <^7heo> not FRI, sorry 2017-01-16 14:36:53 ^7heo: feel free to design a 2017 card! 2017-01-16 14:37:21 <^7heo> it's easy, just parse the systemd git log 2017-01-16 14:37:23 <^7heo> :D 2017-01-16 14:38:58 skarnet: ^7heo: and send it as a Docker image! sending just raw images in SVG or PNG is not cool, you know, some people may not have an image viewer, so you must pack it into Docker image with fullblown Ubuntu, image editor and dozens of useless stuff… 2017-01-16 14:39:41 seems to be some interest from $dayjob in attending 2017-01-16 14:39:48 may be able to get there for free after all 2017-01-16 14:40:21 I also have it as a “business trip” from university :) 2017-01-16 14:40:38 together with one co-worker 2017-01-16 14:41:07 <^7heo> Shiz: good for you mate :) 2017-01-16 14:41:24 can't complain in general 2017-01-16 14:41:26 <^7heo> I paid for it meself. 2017-01-16 14:41:36 not a lot of people get to do working with alpine as their dayjobs... :p 2017-01-16 14:41:37 <^7heo> but it isn't too spensive anyway. 2017-01-16 14:41:45 <^7heo> right. 2017-01-16 14:42:11 i think train tickets would probably be ~40eur for me 2017-01-16 14:42:15 well, I haven’t said to me boss that I’m going here mainly to meet mates from Alpine :P 2017-01-16 14:42:53 37eur 2017-01-16 14:44:39 93 EUR plain tickets :( 2017-01-16 14:44:56 PRG <-> BRU 2017-01-16 14:46:22 cheapest flight would be 105eur for me 2017-01-16 14:46:40 where are you from? 2017-01-16 14:46:52 .nl 2017-01-16 14:50:13 <^7heo> jirutka: plane :D 2017-01-16 14:50:19 <^7heo> jirutka: or plain plane tickets. 2017-01-16 14:50:25 <^7heo> :D 2017-01-16 14:52:00 XD 2017-01-16 15:17:16 <^7heo> ncopa: does Alpine have a booth at the FOSDEM? 2017-01-16 15:22:15 question on an apkbuild file, if i wanted to ensure that the apk never gets installed on a < 3.5 system, is there a canonical way or should I uses a script to test things at install time? 2017-01-16 15:23:11 no canonical way since apkbuilds don't really know the concept of an alpine version 2017-01-16 15:23:12 nor do they care 2017-01-16 15:23:16 <^7heo> yeah 2017-01-16 15:23:23 <^7heo> the alpine version is a file residing in /etc/ 2017-01-16 15:23:25 <^7heo> and that is it. 2017-01-16 15:23:27 <^7heo> (AFAIK) 2017-01-16 15:24:41 <^7heo> I don't see any Alpine booth at the FOSDEM 2017-01-16 15:24:56 <^7heo> do we have any pre-agreed meeting point? 2017-01-16 15:25:08 <^7heo> LFS? :) 2017-01-16 15:25:24 clearly the FSFE booth 2017-01-16 15:25:26 ACTION runs 2017-01-16 15:25:39 <^7heo> :D 2017-01-16 15:25:47 <^7heo> I was gonna make the joke too but I resisted :D 2017-01-16 15:25:51 gotcha, in this case it looks like the gcc that 3.5 includes fixes a few things that makes the configuration totally different (simple), i'm thinking requiring 3.5+ makes more sense in that instance 2017-01-16 15:25:57 <^7heo> Apparently, resistance is futile! 2017-01-16 15:26:18 mitchty: you should probably test by features 2017-01-16 15:26:21 not by alpine version 2017-01-16 15:26:26 testing by identity is generally broken imo 2017-01-16 15:26:29 test by capabilities 2017-01-16 15:26:35 <^7heo> Shiz: http://2.bp.blogspot.com/-07P0Cs7n2gs/UtqGs6OwL-I/AAAAAAAAf9Q/6CqFtXtGD2I/s1600/resistance+is+futile.png 2017-01-16 15:26:55 <^7heo> (an HD version would make a good wallpaper) 2017-01-16 15:27:11 awful 2017-01-16 15:27:15 <^7heo> :D 2017-01-16 15:27:23 Shiz: i'd be replicating autoconf at that point 2017-01-16 15:27:29 <^7heo> Shiz: I think it's pretty good. 2017-01-16 15:27:42 no, because autoconf tends to do identity-based testing 2017-01-16 15:27:50 ("checking if awk is GNU awk...") 2017-01-16 15:28:20 <^7heo> mitchty: you can't replicate autohell, even if you tried to replicate systemd and failed. 2017-01-16 15:28:23 correct, but in this case the apk built on 3.5 won't work on 3.4 due to gcc not supporting the feature 2017-01-16 15:28:41 and hacks i did on 3.4 don't carry over to 3.5 2017-01-16 15:28:58 <^7heo> mitchty: let's try a different approach: why do you want to install your apk on alpine < 3.5? 2017-01-16 15:29:19 ^7heo: i don't, i want to stop people from trying something that won't work 2017-01-16 15:29:26 <^7heo> there must be a reason that's not an arbitrary number that only makes sense to humans, or? 2017-01-16 15:29:40 <^7heo> mitchty: what wouldn't work? 2017-01-16 15:29:45 i think you should just not check it at all 2017-01-16 15:29:50 running the apk on < 3.5 2017-01-16 15:29:51 it's not your responsibility as an apk 2017-01-16 15:29:56 <^7heo> mitchty: why doesn't it work? 2017-01-16 15:29:57 it makes more sense to include a dependency on gcc 2017-01-16 15:30:00 >= a certain version 2017-01-16 15:30:07 if you must use identity testing 2017-01-16 15:30:11 ^7heo: beause of PIC 2017-01-16 15:30:15 just add it to depends= 2017-01-16 15:30:17 :p 2017-01-16 15:30:36 <^7heo> mitchty: PIC in gcc? 2017-01-16 15:30:40 <^7heo> mitchty: or? 2017-01-16 15:30:46 ^7heo: yes 2017-01-16 15:30:57 <^7heo> mitchty: then put a constraint on the APK against the GCC version. 2017-01-16 15:31:17 <^7heo> mitchty: that would work. 2017-01-16 15:31:50 <^7heo> it's not ideal but it's certainly better than trying to put a constraint on an alpine version. 2017-01-16 15:33:17 is there an example apkbuild, the reference doesn't have an example of how to test versions https://wiki.alpinelinux.org/wiki/APKBUILD_Reference#depends 2017-01-16 15:35:17 <^7heo> mitchty: AFAIK you can 'depends' on a specific version 2017-01-16 15:35:34 found one that has it 2017-01-16 15:35:35 py-google-apputils/APKBUILD:11:depends="python2 py-dateutil>=1.4 2017-01-16 15:35:51 i'll do that thanks 2017-01-16 15:36:00 <^7heo> yeah 2017-01-16 15:36:19 <^7heo> we *should* document that but we are too busy fighting over what we should use for doc :D 2017-01-16 15:36:29 text files work 2017-01-16 15:36:34 <^7heo> indeed 2017-01-16 15:36:40 <^7heo> my pointexactly 2017-01-16 15:36:46 <^7heo> s/te/t e/ 2017-01-16 15:37:14 <^7heo> but apparently it's more important to be able to edit with a browser, even if it means that it requires a browser to read the doc' 2017-01-16 15:37:35 i'd just edit the wiki to add here's how you constrain versions so the reference has it 2017-01-16 15:37:35 <^7heo> I don't want to contribute to something I can't use, so I do not contribute to the current docs =/ 2017-01-16 15:37:43 <^7heo> awesome. 2017-01-16 15:37:49 <^7heo> thanks for doing that 2017-01-16 15:38:02 well i have to create an account first :) 2017-01-16 15:38:06 i'm really lazy 2017-01-16 15:38:06 <^7heo> indeed. 2017-01-16 15:38:21 <^7heo> sorry it's so unnecessarily complicated. 2017-01-16 15:40:46 no worries, just never saw any version checks in the reference so was unexpected 2017-01-16 15:41:07 that said, this really makes this ghc port beyond simple 2017-01-16 15:41:23 this is the first version that works without hacks 2017-01-16 15:41:56 but i need to rebuild the bootstrap compilers and test on arm, so i have another 2 days of waiting for an arm box to compile stuff 2017-01-16 16:25:24 re #6685: do you see any issue in backporting the fix? 2017-01-16 17:11:23 fcolista, i think #6685 should have target 3.5.1 and be fixed in 3.5 at least 2017-01-16 17:11:42 i dont see any issue with backporting it to older too if those are broken 2017-01-16 17:13:07 if requested... 2017-01-16 19:24:40 does github account 'alpinelinux' accepts pull request? 2017-01-16 19:30:34 <^7heo> "yes" 2017-01-16 19:31:20 <^7heo> (the quotes mean that the answer is relative to a best effort interpretation of your question.) 2017-01-16 19:36:41 the 'fetch' applet of apk with '--recursive' does not seem to be able to handle virtual dependencies even when they are satisfied on the command line 2017-01-17 00:50:52 21 CVEs for libxml2… wow 2017-01-17 05:39:51 <_ikke_> wut 2017-01-17 10:34:20 rnalrd: sorry for not noticing the ABI bump and not providing suitable pkgrel bumps for depending packages. I'm a bit busy lately, so if you can bump packages yourself that would be great. otherwise I'll send suitable patches, but possibly not earlier than in 10-12 hours. 2017-01-17 10:34:38 np, tnx 2017-01-17 10:35:47 i'll just wait since i'm busy as well 2017-01-17 10:37:19 it seems that redoing patches already in patchwork is ongoing on github. it's complete waste of time. can we mention in GH mirror somehow that there is something like patchwork, so people would look there before? 2017-01-17 10:37:38 jirutka: ^ 2017-01-17 10:38:13 https://github.com/alpinelinux/aports/blob/master/.github/CONTRIBUTING.md 2017-01-17 10:38:41 hurrr durrr github vs patchwork battle, day 2137 2017-01-17 10:39:17 as I said, I’m not watching Patchwork, I don’t have enough time even to review all PRs on GH 2017-01-17 10:39:31 [`] 2017-01-17 10:40:40 hm, I thought that there’s something about patchwork in contributing.md, but it’s not 2017-01-17 10:40:44 we should update it 2017-01-17 10:41:43 that won't fix anything 2017-01-17 10:42:37 maybe won't fix, but there is a chance it will lessen doubled work, so it is some improvement 2017-01-17 10:42:47 I don't think so 2017-01-17 10:43:02 me neither, but I don’t know what else to do with that 2017-01-17 10:43:16 i agree with jirutka 2017-01-17 10:43:21 i dont know what to do with it 2017-01-17 10:43:22 i mean 2017-01-17 10:43:24 inb4 no one reads contributing.md 2017-01-17 10:43:27 unless you receive red pop-up with "U FUG, HAVE YOU CHECKED PATCHWORK? Y STILL WANNA COMMIT?" 2017-01-17 10:43:56 i dont want require everyone to patcwork 2017-01-17 10:44:00 and there may be a problem in both directions… what if someone send patch to patchwork without looking into GH? 2017-01-17 10:44:10 yes 2017-01-17 10:44:30 yeah, I already admitted that I committed this sin once 2017-01-17 10:44:38 we could either drop patchwork all together, or we could drop github PRs 2017-01-17 10:44:55 there’s one possible solution, implement some service that would automatically open PR for patches on Patchwork 2017-01-17 10:45:02 jirutka: this 2017-01-17 10:45:06 maybe przemoc can do it instead of complaining ;) 2017-01-17 10:45:12 XD 2017-01-17 10:45:25 who should be the owner of the PR? 2017-01-17 10:45:54 of the pr mirrored from patchwork? 2017-01-17 10:46:07 yes 2017-01-17 10:46:25 some special account, maybe algitbot :P 2017-01-17 10:46:32 could work 2017-01-17 10:47:19 jirutka: what about code-review? you've mentioned that it is impossible or hard to do a series of commits under the same thing 2017-01-17 10:47:41 jirutka: so if I understand correctly where will be like over9000 PRs 2017-01-17 10:47:59 that’s why it’s not easy to implement such thing… 2017-01-17 10:48:32 I think we are making problems that shouldn't exist ;f 2017-01-17 10:49:22 yes 2017-01-17 10:50:36 i’m sorry, I just woke up, just to fix incident with one service, and I’m still half sleeping 2017-01-17 10:50:42 so afk 2017-01-17 10:50:49 jirutka, thanks! 2017-01-17 10:54:16 ayyyy 2017-01-17 10:54:37 why we couldn't just move issue tracker from redmine to github? 2017-01-17 10:54:44 so everything will be there 2017-01-17 10:54:49 skrzyp: just like Void did? 2017-01-17 10:54:51 and eliminate patchwork then 2017-01-17 10:54:58 yes 2017-01-17 10:55:00 void did the same 2017-01-17 10:55:08 eliminate patchwork? 2017-01-17 10:55:16 11:53 < ncopa (~ncopa@alpine/ncopa) has quit (Quit: Leaving) 2017-01-17 10:55:18 11:54 ayyyy 2017-01-17 10:55:20 11:54 why we couldn't just move issue tracker from redmine to github? 2017-01-17 10:55:22 11:54 so everything will be there 2017-01-17 10:55:24 11:54 skrzyp: just like Void did? 2017-01-17 10:55:26 11:54 > ncopa (~ncopa@alpine/ncopa) has joined #alpine-devel 2017-01-17 10:55:33 because issues in github are limited 2017-01-17 10:55:34 also, you can drive github PRs/issue tracker via email 2017-01-17 10:55:35 yeah i had to reboot sorry 2017-01-17 10:55:48 clandmeter: but they're sufficient for such a project 2017-01-17 10:56:19 let ncopa determine that, he works with it the most. 2017-01-17 10:56:20 skrzyp: we need a way for people to: 1 report issues, 2 contribute with patches, without require them to sign up an apple id 2017-01-17 10:56:25 or google id 2017-01-17 10:56:29 or GitHub account 2017-01-17 10:56:44 but they need email account anyways 2017-01-17 10:56:48 so this is a form of id 2017-01-17 10:57:02 github acc is not intrusive anyhow 2017-01-17 10:57:30 but we dont require the email be google or outlook or any specific vendor 2017-01-17 10:57:43 they don't require that too 2017-01-17 10:57:54 i've registered my github acc via openmailbox 2017-01-17 10:58:14 clandmeter: I'm not sure if we completely need issue tracker / patchwork if ncopa does the most of work and no one ever looks on remine 2017-01-17 10:58:34 which is because these services are too fragmented and poorly advertised on our webpage 2017-01-17 10:58:41 we dont want require people to register to GitHub to be able to contribute 2017-01-17 10:59:03 11:56 but they need email account anyways 2017-01-17 10:59:21 so, let's just make anonymous html commit form 2017-01-17 10:59:41 you can have email account without registering to github 2017-01-17 11:00:00 *facepalm* 2017-01-17 11:00:11 scadu: can you…? 2017-01-17 11:00:28 email account != GitHub account 2017-01-17 11:01:00 but both are form of identification, so if you want people to be anonymous, eliminate any method of identification 2017-01-17 11:01:24 that is not the problem 2017-01-17 11:01:25 and to send email, you need to actually register email account first 2017-01-17 11:01:39 ncopa: how many active commiters does not have account on github? 2017-01-17 11:01:47 you can even have your own mail server. there are people refusing to create accounts elsewhere. 2017-01-17 11:02:13 przemoc: I don't think if rms@ ever wants to commit to Alpine 2017-01-17 11:02:19 przemoc: and there are people like skarnet with his own mail server which is blacklisted every week 2017-01-17 11:02:22 OTOH there are people like skarnet, who cannot send patches, because they are blacklisted... 2017-01-17 11:02:31 XD 2017-01-17 11:02:31 so we need to setup extra infrastructure because people dont want to signup anywwhere. 2017-01-17 11:02:36 it sounds the other way around 2017-01-17 11:02:59 thats not contributing, thats demanding 2017-01-17 11:03:14 let's get rid of github then and stick with patchwork [cool] 2017-01-17 11:03:26 no no no 2017-01-17 11:03:29 so if all those ppl who dont want to sign up anywhere start to create solution we can use, everybody is happy. 2017-01-17 11:03:35 just send the patches via regular snail mail 2017-01-17 11:03:40 clandmeter: I am demanding because I want to send a patch and my server is unjustifiably blacklisted? 2017-01-17 11:03:42 and fidonet 2017-01-17 11:03:55 (inb4 no one knows fido) 2017-01-17 11:04:20 skarnet: im talking about procedure, not broken systems.. 2017-01-17 11:04:40 i dont manage the email infra, so i cannot help with that. 2017-01-17 11:05:11 You can help by agreeing that the infra should accommodate contributors instead of making them jump through hoops 2017-01-17 11:05:35 i help a lot 2017-01-17 11:05:41 i dont see others help that much 2017-01-17 11:06:33 then why contribution infrastructure still exists, while other people aren't helping you? 2017-01-17 11:11:52 i choose github becuase it "just works" (most of the time) and most of us have an account. I really dont care which system we use, but setting up our own cost a lot of time and nobody seems to want to support it. 2017-01-17 11:13:13 i know ncopa and fabled prefer mailing list so we are not tied to a third party aka "eat your own dogg food" solution(s). So they will have to fix this issue imho. 2017-01-17 11:14:07 err i mean the they are the opposite of course :) 2017-01-17 11:14:19 I'm not against github. if it will become the main and only place for sending patches, I'll abide such decision (and others possibly too). sorry for being the bad guy and noticing that people tend to create patches for patchwork or github and this leads occasionally to needlessly duplicated work, which is a waste of effort for project as small as AL, but it's better to start fixing that sooner 2017-01-17 11:14:25 than later. that's why I think that mentioning patchwork in contributing won't make things worse at least and there is chance that some people will read it (but possibly not present GH contributors, only future ones). I'm ok with ditching patchwork, but not sure how other contributors will see it. 2017-01-17 11:15:00 przemoc: right, its good to mention it. 2017-01-17 11:15:21 create a pr ;-) 2017-01-17 11:16:00 maybe tonight then 2017-01-17 11:16:05 I prefer gh since we have it 2017-01-17 11:16:18 I'm interested how many contributors are not gh members anyway 2017-01-17 11:16:22 its not only the issue of having 2 systems. 2017-01-17 11:17:31 for instance currently im busy and dont have time to review patches. Then only jirutka is doing review and things slow down... 2017-01-17 11:20:13 yep, we know that 2017-01-17 11:20:19 and ncopa has less time to spend on alpine too... 2017-01-17 11:20:34 but it won't be quicker with two systems to maintain 2017-01-17 11:27:37 <^7heo> moin 2017-01-17 11:38:22 i think the big dislike for patchwork was that it does not always close the issue after it's been committed 2017-01-17 11:38:32 it might be just a config option to fix that 2017-01-17 11:39:51 it possibly does not close if the patch was amended 2017-01-17 11:41:07 yeah, something like that 2017-01-17 11:41:30 the other thing was commenting on the patches. personally i like to do it in mail client. but apparently many people now use just web mail and it's not so convenient there. 2017-01-17 11:41:58 there's also gerrit etc. for getting commits reviewed 2017-01-17 11:42:00 but personally 2017-01-17 11:42:30 i just want the get job done. after finding the 'hub' tool that can pick github pull requests to my local git tree, it's been okish 2017-01-17 11:43:00 i'm ok to review code via web interface, but i want to merge and do the final checks using my tools on my local tree 2017-01-17 11:47:44 fabled: you can use simple curl to fetch pr's 2017-01-17 11:47:50 as i mentioned before 2017-01-17 11:47:57 I guess no one thinks to make GH repo the main git repo, it would remain the mirror of g.a.o. and you don't been need hub tool to get PRs. you can add fetch to your origin remote like: fetch = +refs/pull/*/head:refs/remotes/origin/pr/* and you'll fetch all pull requests then 2017-01-17 11:49:42 the only downside of this fetch thing is that you'll get all PRs, not only opened ones 2017-01-17 11:50:32 i lost the irc history for the last 50 mins sorry 2017-01-17 11:51:12 ncopa: http://irclogger.com/.alpine-devel/2017-01-17#1484650884 2017-01-17 11:51:26 algitbot: oh funny you 2017-01-17 11:51:42 http://dev.alpinelinux.org/irclogs/%23alpine-devel-2017-01.log 2017-01-17 11:52:03 clandmeter: I was looking for this. thanks. 2017-01-17 11:52:10 :) 2017-01-17 11:55:13 thanks 2017-01-17 12:19:50 skrzyp: ehm, it’s not true that no one ever looks on redmine… I do for example and I hope that I’m not the only one 2017-01-17 12:20:14 i use redmine as checklist when tagging releases 2017-01-17 12:20:19 to make sure i dont forget things 2017-01-17 12:20:38 jirutka: *crickets.wav* 2017-01-17 12:21:04 so yeah, i do use redmine too 2017-01-17 12:21:55 scadu: “let's get rid of github then and stick with patchwork [cool]” … than bye bye from me, I really don’t have so much time to waste it with totally inefficient tools like Patchwork… 2017-01-17 12:22:15 "clandmeter: so if all those ppl who dont want to sign up anywhere start to create solution we can use, everybody is happy." — exactly! 2017-01-17 12:23:21 jirutka: it was sarcasm :> 2017-01-17 12:23:34 <^7heo> heey jirutka :) 2017-01-17 12:24:24 przemoc: do you know what is even more waste of effort than occasional duplicated patches? working with Patchwork, without any automation… 2017-01-17 12:24:25 <^7heo> scadu: yeah the main chan is funny :) 2017-01-17 12:24:30 <^7heo> scadu: it's full of random people 2017-01-17 12:24:46 i can't follow 4 alpine channels at once 2017-01-17 12:24:57 5 even 2017-01-17 12:24:59 <^7heo> 5? 2017-01-17 12:25:01 <^7heo> what 5? 2017-01-17 12:25:17 #alpine-devel #alpine-commits #alpine-linux #alpine-offtopic #alpine-pl 2017-01-17 12:25:48 police? 2017-01-17 12:25:50 fabled: well, I don’t use webmail and still sending patches via mail is like using candles instead of lightbulbs… 2017-01-17 12:25:56 clandmeter: perl 2017-01-17 12:26:04 :D 2017-01-17 12:26:18 <^7heo> :D 2017-01-17 12:26:32 fabled: and even skarnet is fine with GH once I showed him how to open PRs from command line, without touching web interface at all 2017-01-17 12:26:45 incorrect interpreted... :) 2017-01-17 12:27:03 ^7heo: wut XD 2017-01-17 12:27:17 ^7heo: oh nvm 2017-01-17 12:27:42 <^7heo> wat wut? 2017-01-17 12:27:47 <^7heo> (in the wut wut?) 2017-01-17 12:28:14 fabled: "but i want to merge and do the final checks using my tools on my local tree" — that’s what we do even with GH, b/c it’s just a mirror, we push only to git.a.o; and I personally don’t see problem with that, actually I’d like to disable Merge button even on projects that use only GH, b/c it makes people lazy to clean commits before merging 2017-01-17 12:28:38 i would create an alpine-nl but skarnet would not approve it. 2017-01-17 12:29:03 native languages? 2017-01-17 12:29:30 przemoc: "I guess no one thinks to make GH repo the main git repo" … correct, I fully understand arguments about not being dependent on 3rd party service, it’s wise to have alternatives 2017-01-17 12:29:59 jirutka: everyone forgot that git is a distributed vcs 2017-01-17 12:30:06 and even if you're using 3rd party remote 2017-01-17 12:30:09 its distributed, so it doesnt really matter. 2017-01-17 12:30:12 data is still yours 2017-01-17 12:30:25 exactly 2017-01-17 12:30:41 huh, I didn’t know abou #alpine-pl 2017-01-17 12:30:51 poland strong 2017-01-17 12:31:04 borland drunk 2017-01-17 12:31:28 I've crated it but haven't officially announced it yet 2017-01-17 12:31:39 will ncopa sue me on the freenode police? 2017-01-17 12:31:50 created* 2017-01-17 12:32:44 all your channels belong to us 2017-01-17 12:32:59 skrzyp: "everyone forgot that git is a distributed vcs", hm, maybe for a minute… that’s true, it should nto matter if we push to GH or git.a.o 2017-01-17 12:34:45 clandmeter: create it! and host all NL discussions there, freeing the rest of the world! 2017-01-17 12:35:12 allemaal? 2017-01-17 12:35:24 nondejus 2017-01-17 12:35:32 ale o co chodzi? 2017-01-17 12:35:42 skrzyp: but we don’t have sync GH → git.a.o… to be honest, I think that it’s better to not have it, as I said I have not very good experience with people using Merge button on GH 2017-01-17 12:36:29 just configure access control properly 2017-01-17 12:36:40 to limit people only to these using git-merge 2017-01-17 12:36:45 skrzyp: my co-worker often merges crap just because he’s lazy and clicking big fat Merge button is easier than write command in terminal and rebase these fucking commits with nonsense commit messages 2017-01-17 12:37:00 sorry, I’ve just read new commits in our repo today and I’m upset again 2017-01-17 12:37:49 MERG BATTON ==> "Marge branch 'master' of http://foobar.com/foo.git" with no commits sometimes 2017-01-17 12:38:52 btw there are 3 official langs in Belgium: French, Germany and Dutch… so, I’ll be the only one from Alpine doomed here XD 2017-01-17 12:39:11 jirutka: now try to guess about switzerland 2017-01-17 12:39:27 jirutka: to make you feel better: https://twitter.com/clfln 2017-01-17 12:40:02 skrzyp: XD 2017-01-17 12:40:12 pardon, skarnet: XD 2017-01-17 12:40:31 jirutka: the avg fosdem audiance is larger then the german speaking belgium population. 2017-01-17 12:40:44 :DD 2017-01-17 12:41:15 heh, but you and ncopa are Dutch native, skarnet and ^7heo french native 2017-01-17 12:41:35 ncopa and dutch? 2017-01-17 12:41:57 i think you are mistaken 2017-01-17 12:42:02 ncopa is from Netherlands, isn’t he? 2017-01-17 12:42:06 no 2017-01-17 12:42:12 huh, aha 2017-01-17 12:42:14 <^7heo> clandmeter: he never said "ncopa and Dutch" :P 2017-01-17 12:42:27 <^7heo> jirutka: ncopa is from finland, oder? 2017-01-17 12:42:32 aaah 2017-01-17 12:42:34 ncopa, I'd instakick jirutka to high heavens if I were you 2017-01-17 12:42:37 no 2017-01-17 12:42:43 <^7heo> sweden? 2017-01-17 12:42:53 <^7heo> oh wait wait wait 2017-01-17 12:42:54 <^7heo> last one 2017-01-17 12:42:57 <^7heo> noway! 2017-01-17 12:42:59 literally no 2017-01-17 12:43:03 <^7heo> rway! 2017-01-17 12:43:06 I’m lost 2017-01-17 12:43:08 <^7heo> :D 2017-01-17 12:43:18 he's from yesway 2017-01-17 12:43:26 noway 2017-01-17 12:43:28 <^7heo> :D 2017-01-17 12:43:58 how’s that possible that ncopa is not on Wikipedia yet? :P 2017-01-17 12:44:22 <^7heo> because we need him to retain his sanity. 2017-01-17 12:45:05 jirutka: did you see his new twitter pic? 2017-01-17 12:45:12 yes I did 2017-01-17 12:45:49 oh crap, must go to lunch and to work 2017-01-17 12:46:26 https://pbs.twimg.com/profile_images/811963744796508164/-mYFOhr4.jpg 2017-01-17 12:46:49 I can’t guess country from a photo 2017-01-17 13:31:53 scadu: what 2017-01-17 13:32:25 skrzyp: see -offtopic 2017-01-17 13:38:51 <^7heo> skrzyp: you know, where you've been highlighted... 2017-01-17 13:41:41 jirutka, are you going to fosdem? 2017-01-17 13:41:45 <^7heo> yes 2017-01-17 13:41:56 <^7heo> (sorry to answer for him but he said he had his tickets) 2017-01-17 13:42:01 <^7heo> (plain tickets even :D) 2017-01-17 13:42:19 cool 2017-01-17 13:42:50 <^7heo> roughly 23 hours ago. 2017-01-17 13:42:56 <^7heo> (if you wanna check it out) 2017-01-17 13:43:38 thanks 2017-01-17 13:53:53 <^7heo> you're welcome 2017-01-17 14:12:31 I’m back 2017-01-17 14:12:43 I'm front 2017-01-17 14:13:14 ncopa: yes, I can confirm that I’m really going to FOSDEM, mainly to meet you and other people from Alpine :) 2017-01-17 14:13:54 very nice :) 2017-01-17 14:14:34 anyone else going? i know skarnet is 2017-01-17 14:15:02 ncopa: clandmeter: ^7heo: I have an idea, what about creating channel #alpine-fosdem? for organizational things, to not spam main channel with that 2017-01-17 14:15:24 ncopa: someone wrote yesterday that he’s also going, but don’t remember the nick 2017-01-17 14:15:35 ncopa: also ^7heo is going 2017-01-17 14:15:43 im in #alpine-fosdem now 2017-01-17 14:15:54 i dont think im going 2017-01-17 14:16:01 been there, didnt really like it. 2017-01-17 14:16:03 :-( 2017-01-17 14:16:10 weird ppl 2017-01-17 14:16:11 ncopa: ah, Shiz is the one who wrote yeasterdasy that he’s also going 2017-01-17 14:16:11 :p 2017-01-17 14:16:26 clandmeter: what, you’re not going? :( 2017-01-17 14:16:33 clandmeter: I hoped that I’ll meet you too 2017-01-17 14:16:41 im not sure 2017-01-17 14:16:52 what about fabled? 2017-01-17 14:17:17 jirutka, can't make it unfortunately 2017-01-17 14:18:31 fabled, two small suggestions for apk fetch 2017-01-17 14:18:41 jirutka: and ill have to give too many beers to ncopa 2017-01-17 14:18:50 lol 2017-01-17 14:18:59 fabled, one, either ignore v: deps when -R, or see if their satisfiers are specified 2017-01-17 14:19:07 <^7heo> oh awilfox is here too. 2017-01-17 14:19:18 i am afraid i'll have to give way too many to clandmeter too. would end up bad :) 2017-01-17 14:19:22 fabled, two, make a --installed option to autofetch everything installed on computer 2017-01-17 14:19:24 <^7heo> awilfox == TheWilfox ? 2017-01-17 14:19:31 yes ^7heo 2017-01-17 14:19:40 <^7heo> oh, that's easier :P 2017-01-17 14:20:06 awilfox, two is basically "apk fetch $(cat /etc/apk/world)" 2017-01-17 14:20:06 awilfox: apk fetch -R $(cat /etc/apk/world) 2017-01-17 14:20:06 <^7heo> why the transition from 'a' to 'the'? 2017-01-17 14:20:14 <^7heo> since when did you become unique? :D 2017-01-17 14:20:18 <^7heo> since adelie? :D 2017-01-17 14:20:25 awilfox, not really sure what you mean with the first request 2017-01-17 14:21:58 ^7heo, awilfox is a pun on my name since I am foxy and A. Wilcox 2017-01-17 14:22:32 ^7heo, in 2014 on staticbox (old irc network not around any more) one of the opers was saying "but you're not just *a* wilfox, you're *the* wilfox!" 2017-01-17 14:22:34 it stuck 2017-01-17 14:22:48 <^7heo> awilfox: is your name really "Foxy And A. Wilcox"!? 2017-01-17 14:22:52 <^7heo> ACTION hides 2017-01-17 14:22:55 XD 2017-01-17 14:23:20 <^7heo> yeah so I was not the first to make that joke... 2017-01-17 14:23:30 <^7heo> and one more disappointment with my life... 2017-01-17 14:23:32 fabled, hm, well I used apk fetch $(apk info) iirc 2017-01-17 14:24:08 fabled, also in first request apk fetch -R foo says "foo: failed to select package" if foo depends on v:something 2017-01-17 14:24:41 fabled, if you do apk fetch -R ProviderOfSomething foo, that does not help it 2017-01-17 14:25:00 ^7heo, I have been here long time too: http://bugs.alpinelinux.org/issues/478#note-7 2017-01-17 14:25:47 <^7heo> wow, 6 years... 2017-01-17 14:25:55 <^7heo> okay you're beating me on that one. 2017-01-17 14:25:58 me too http://bugs.alpinelinux.org 2017-01-17 14:26:07 <^7heo> I'm here only for 3-4 years. 2017-01-17 14:26:22 http://bugs.alpinelinux.org/users/107 2017-01-17 14:26:25 <^7heo> Registered on: 03/08/2009 2017-01-17 14:26:27 clandmeter: hah :P 2017-01-17 14:26:32 :) 2017-01-17 14:27:17 lol 2017-01-17 14:27:30 <^7heo> awilfox: sorry, I thought for a moment you had the profile named "Mika Havela" 2017-01-17 14:27:42 <^7heo> awilfox: I missed the #note-7 part of the url. quite some confusion there. 2017-01-17 14:27:50 hahaha 2017-01-17 14:27:57 <^7heo> awilfox: cute fox btw... 2017-01-17 14:28:02 I am definitely A. Wilcox, it says so on my ID card 2017-01-17 14:28:50 so Mr. Definitely... 2017-01-17 14:29:13 I wouldn't say that 2017-01-17 14:29:13 <^7heo> http://bugs.alpinelinux.org/users/3 2017-01-17 14:29:30 <^7heo> clandmeter: ^ 2017-01-17 14:29:42 yes, i just checked 2017-01-17 14:29:45 where did #2 go? 2017-01-17 14:29:48 <^7heo> looks like someone screwed up 2 times before creating the first accounts :D 2017-01-17 14:29:56 <^7heo> and #1 too.. 2017-01-17 14:29:59 nah 2017-01-17 14:30:02 #1 is admin acc 2017-01-17 14:30:10 wow, Mika Havela is #4 2017-01-17 14:30:13 <^7heo> I'm http://bugs.alpinelinux.org/users/782 2017-01-17 14:30:18 <^7heo> clandmeter: ah ok 2017-01-17 14:30:32 <^7heo> I feel young :P 2017-01-17 14:30:49 <^7heo> but tbh I waited quite some time to create a redmine account 2017-01-17 14:30:49 fabled is #6 and ncopa is #7... ncopa you need to work on your ladder rankings 2017-01-17 14:31:20 <^7heo> http://bugs.alpinelinux.org/users/42 2017-01-17 14:31:25 <^7heo> isn't that some spam account? 2017-01-17 14:31:32 ncopa: where is the old svn repo? i wonder if im in it. 2017-01-17 14:32:15 SVN? uh, I don’t want to know that Alpine ever used SVN… XD 2017-01-17 14:32:34 clandmeter: the old svn repo is gone i think 2017-01-17 14:32:35 there was a pre git era :) 2017-01-17 14:32:46 I’m totally Linus person with opinion about CVS and SVN 2017-01-17 14:32:47 that was before git was invented 2017-01-17 14:32:51 you didnt keep a copy? 2017-01-17 14:33:01 clandmeter: i probably did 2017-01-17 14:33:04 :) 2017-01-17 14:33:17 what, the world didn’t started with invention of Git? :P 2017-01-17 14:33:22 on a backup tape probably 2017-01-17 14:33:38 or a cdrom or maybe a floppy :D 2017-01-17 14:33:58 <^7heo> jirutka: nah, it's versionned with something called "TBB" 2017-01-17 14:34:07 ^7heo: what’s that? 2017-01-17 14:34:18 <^7heo> jirutka: a versionning system for the universe. 2017-01-17 14:34:30 <^7heo> jirutka: every time you want a new version, you create a new "BB" 2017-01-17 14:34:57 <^7heo> jirutka: but you have to wait for expansion to proceed so the universe collapes 2017-01-17 14:35:04 <^7heo> jirutka: then you enter your commit message 2017-01-17 14:35:15 <^7heo> jirutka: it generates a unique hash 2017-01-17 14:35:25 <^7heo> jirutka: along with unique new rules of physics 2017-01-17 14:35:36 <^7heo> jirutka: and done, you have a new version. 2017-01-17 14:36:00 <^7heo> (also, BB == Big Bang) 2017-01-17 14:39:02 lol 2017-01-17 15:24:31 ncopa: could you please tag new version? https://github.com/ncopa/lua-optarg 2017-01-17 15:24:56 ncopa: probably 0.2 2017-01-17 15:26:02 do i need to modify the luarock file for that? 2017-01-17 15:26:48 seems like not 2017-01-17 15:27:49 ncopa: nope, you don’t need 2017-01-17 15:28:55 ncopa: btw this is how I release luarocks: https://github.com/jirutka/gversion.lua/blob/master/script/deploy-rock 2017-01-17 15:29:44 ncopa: I’ll take care about optarg luarock, just need a git tag 2017-01-17 15:30:34 how do I force abuild to build a package with a different md5sum ? 2017-01-17 15:30:41 I just needed to change/patch my source file 2017-01-17 15:31:13 jirutka, seems like --longopt=value is broke 2017-01-17 15:33:44 this works: `--longopt value` 2017-01-17 15:34:47 hm, I haven’t changed anything that may cause this 2017-01-17 15:35:11 <^7heo> ncopa: who would write --longopt=value? 2017-01-17 15:35:21 <^7heo> ncopa: I thought the GNU way was --longopt value 2017-01-17 15:35:44 <^7heo> ncopa: and moreover, longopts are sucky, their implementation is orders of magnitude too painful for the gain. 2017-01-17 15:35:53 both variants should work 2017-01-17 15:36:38 <^7heo> s/should/&n't/ 2017-01-17 15:37:25 <_ikke_> ACTION kicks ^7heo 2017-01-17 15:37:40 ncopa: yes, this is broken since beginning 2017-01-17 15:37:55 so let’s see if i can fix it 2017-01-17 15:39:11 <^7heo> _ikke_: what now? 2017-01-17 15:39:44 <^7heo> _ikke_: I wrote stuff that is against the DRPK? 2017-01-17 15:39:53 <_ikke_> lol 2017-01-17 15:40:23 DRPK? 2017-01-17 15:41:15 <_ikke_> DPRK (Democratic People Republic of (north) Korea) 2017-01-17 15:43:14 wat? 2017-01-17 15:45:25 <^7heo> yeah 2017-01-17 15:45:26 <^7heo> For some reason my brain wrote DPRK and my fingers wrote DRPK. 2017-01-17 15:45:26 <^7heo> It wluod not be the frsit tmie taht hpepnas 2017-01-17 15:45:26 <^7heo> sutipd biarn 2017-01-17 15:45:26 <^7heo> jirutka: http://awesci.com/cambridge-university-says-spelling-does-not-matter/ 2017-01-17 15:45:26 <^7heo> "Aoccdrnig to rscheearch at Cmabrigde uinervtisy, it deosn’t mttaer waht oredr the ltteers in a wrod are, the olny iprmoetnt tihng is taht the frist and lsat ltteres are at the rghit pclae." 2017-01-17 15:45:50 <^7heo> And tehy cnoitune wtih: "The rset can be a tatol mses and you can sitll raed it wouthit a porbelm." 2017-01-17 15:46:04 <^7heo> And tehn: "Tihs is bcuseae we do not raed ervey lteter by it slef but the wrod as a wlohe." 2017-01-17 15:46:58 <^7heo> jsut syanig... 2017-01-17 15:47:14 <^7heo> (but it does not work with acronyms) 2017-01-17 15:47:22 <^7heo> (for obvious reasons) 2017-01-17 15:52:43 for ovubois raonses? 2017-01-17 15:54:36 omg how the hell this damn thing works 2017-01-17 15:59:15 jirutka: something like this? http://tpaste.us/Gnvg 2017-01-17 16:06:32 ncopa: looks ok, I wanted to make it less messy, but semes to be impossible without rewriting :/ 2017-01-17 16:07:28 this works better: http://tpaste.us/2zRN 2017-01-17 16:07:38 incase you do --foo=foo=bar 2017-01-17 16:08:01 I’ve been just testing this use case :) 2017-01-17 16:08:02 i mean --foo=bar=baz 2017-01-17 16:08:15 ^7heo: my experiance is when you get older it gets more frequent :| 2017-01-17 16:08:31 jirutka, it does not support -o=outfile though 2017-01-17 16:08:33 I’d just use `+` in `[^=]*` 2017-01-17 16:08:41 ncopa: that’s correct imo 2017-01-17 16:08:48 yeah 2017-01-17 16:08:50 probably 2017-01-17 16:08:53 ncopa: who combines short option with = ? 2017-01-17 16:09:12 exactly 2017-01-17 16:09:30 and yes, '+' is better than '*' as you mentioned 2017-01-17 16:09:41 i think i'll just commit that 2017-01-17 16:11:47 yeah, go on :) 2017-01-17 16:13:43 done 2017-01-17 16:13:50 can you please test it abit? 2017-01-17 16:13:54 then i'll tag it 2017-01-17 16:13:56 thanks 2017-01-17 16:16:52 looks okay 2017-01-17 16:17:05 I’ve tried few inputs and it seems that it works 2017-01-17 16:30:27 jirutka, tag pushed 2017-01-17 16:30:28 thanks! 2017-01-17 16:35:17 ncopa: thanks, I’ve uploaded new rock to luarocks https://luarocks.org/modules/jirutka/optarg/0.2-1 2017-01-17 17:24:39 hmmm, someone said my name but my lastlog appears to have scrolled away 2017-01-17 17:44:08 btw, recompiled all the qemu linux-user packages with static link 2017-01-17 17:44:19 which i think it makes more sense 2017-01-17 17:45:12 yeah, that’s more useful! 2017-01-17 17:45:46 ncopa: I don’t remember, have I already showed you https://github.com/jirutka/alpine-chroot-install/#install-alpine-linux-in-chroot ? 2017-01-17 17:46:49 i think i have seen the alpine-linux-in-chroot script yes 2017-01-17 17:46:52 ncopa: I love that comparison with debootstrap, 32 sec vs. 360 sec it quite appealing :) 2017-01-17 17:48:57 wasnt really aware that it was for qemu user 2017-01-17 17:49:09 32 sec is cool 2017-01-17 17:49:17 still too much imho :) 2017-01-17 17:49:26 that’s including time for Travis to bootstrap… 2017-01-17 17:49:45 installing base alpine with apk.static takes few seconds 2017-01-17 17:50:11 thanks to it it’s not a problem when someone does not support Alpine, like CIs, I can just install it into chroot in few seconds… actually it’s installed faster than apt-get can update repositories or emerge resolve dependencies, lol :) 2017-01-17 17:50:16 i created a ticket for native alpine once in travis 2017-01-17 17:50:21 they said no 2017-01-17 17:50:39 lol 2017-01-17 17:50:50 not surprised, they support only ubuntu and not even latest version, the default is still Precise, Trusty as opt-in, nothing newer 2017-01-17 17:51:02 oh, we now ship alpine rootfs images 2017-01-17 17:51:27 I must even install qemu-user directly from a package for newer debian, b/c that outdated version on Travis does not work correctly 2017-01-17 17:51:32 could maybe simplify since you dont need run apk.static 2017-01-17 17:51:46 qemu-user is used to emulate armhf or other arch 2017-01-17 17:51:52 where can I find rootfs image? 2017-01-17 17:52:09 https://alpinelinux.org/downloads/ 2017-01-17 17:52:25 mini root filesystem 2017-01-17 17:52:28 but still, I really like that we can use the same tool as for installing packages even to install base system, this is awesome and actually one of the things I love about Alpine 2017-01-17 17:52:38 yes 2017-01-17 17:52:42 that is very nice 2017-01-17 17:52:47 actually 2017-01-17 17:52:52 I don’t know about any other distro that can do this 2017-01-17 17:52:55 the static alpine qemu user packages 2017-01-17 17:53:06 i think they are useful for others too 2017-01-17 17:53:14 yeah, definitely 2017-01-17 17:53:26 i mean, you can curl | tar -zx it 2017-01-17 17:53:48 and you have a static binary thats smaller than static binary from debian 2017-01-17 17:53:51 (i assume) 2017-01-17 17:54:00 i havent checked the difference though 2017-01-17 17:54:52 but heh, its inda cool that you can set up the system on the same time gentoo spend on calculating the deps :) 2017-01-17 17:55:31 actually, it’s very sad how slow emerge is 2017-01-17 17:55:55 it doesn’t take so much time on every machine, but it really takes dozens of seconds on many of my machines 2017-01-17 17:55:57 i havent used it since we stopped build alpine with it 2017-01-17 17:57:18 I still have some systems with Gentoo, but gradually migrating everything to Alpine 2017-01-17 17:59:40 please fill: “your bird day or birthdate” … wait, what? XD 2017-01-17 17:59:49 <_ikke_> lol 2017-01-17 17:59:53 <_ikke_> WHen did you bird? 2017-01-17 18:00:34 alpine static qemu user is like 300k smaller 2017-01-17 18:01:00 except i386 2017-01-17 18:01:07 hmm, is debian qemu-user-static even completely static? without dependency on glibc? :) 2017-01-17 18:01:24 debian: 2017-01-17 18:01:24 root@0986f804abac:/# du -sh /usr/bin/qemu-i386-static 2017-01-17 18:01:25 3.2M /usr/bin/qemu-i386-static 2017-01-17 18:01:28 alpine: 2017-01-17 18:01:44 2.4M /usr/bin/qemu-i386 2017-01-17 18:01:52 :) 2017-01-17 18:02:19 yeah, i dont think static link with gnu libc makes much sense 2017-01-17 18:02:59 i need to go 2017-01-17 18:03:02 see u 2017-01-17 18:05:00 see ya 2017-01-17 18:48:05 oh crap, I’ve upgraded one VM from 3.4.6 to 3.5.0 and it fails to boot 2017-01-17 18:48:44 mounting LABEL=root on /sysroot failed: No such file or directory 2017-01-17 18:48:51 <_ikke_> auch 2017-01-17 19:06:55 hm, it seems that it’s my fault, the label is incorrect 2017-01-17 19:12:11 static binaries and glibc is the surest route to madness 2017-01-17 19:13:03 my favorite was a static binary that could still dynamic link in libraries 2017-01-17 19:17:46 glibc is madness itself 2017-01-17 19:50:40 i see no real issue with that 2017-01-17 19:50:49 its not unfeasible for a static binary to dlopen() modules 2017-01-17 20:43:41 problem is it didn't work just like a dynamic binary due to glibc's versioning 2017-01-17 22:57:59 Hi. What's the easiest way to get abuild verbose (to see why a dep is not satisfied)? 2017-01-17 22:58:49 what does it print? 2017-01-17 23:00:31 jirutka: I know ppc64le is not present, but that's what I got: https://paste.fedoraproject.org/529188/69399414/raw/ 2017-01-17 23:00:47 (the ppc64le support I mean...) 2017-01-17 23:01:05 I wondering if pass1 had to go firstly... 2017-01-17 23:01:55 it's the command line I've isolated from ./bootstrap.sh script 2017-01-17 23:02:11 hm, it says that binutils-ppc64le is missing… what more information should it provide? 2017-01-17 23:02:42 but I have no experience with bootstrapping Alpine from scratch… 2017-01-17 23:02:44 missing exactly where? in which place? 2017-01-17 23:02:57 places specified in /etc/apk/repositories 2017-01-17 23:03:26 hm, because I do have it in : alpine01:~/packages/main/x86_64$ ls -l 2017-01-17 23:03:26 total 4304 2017-01-17 23:03:27 -rw-r--r-- 1 gromero gromero 829 Jan 17 22:20 APKINDEX.tar.gz 2017-01-17 23:03:27 -rw-r--r-- 1 gromero gromero 4399149 Jan 17 22:20 binutils-ppc64le-2.27-r1.apk 2017-01-17 23:03:53 do you have ~/packages/main in /etc/apk/repositories? 2017-01-17 23:03:57 (with absolute path) 2017-01-17 23:04:45 fabled, are you here? 2017-01-17 23:05:14 nope, just added 2017-01-17 23:05:32 now I'm getting: 2017-01-17 23:05:33 >>> ERROR: gcc-pass2-ppc64le: all failed 2017-01-17 23:05:33 >>> gcc-pass2-ppc64le: Uninstalling dependencies... 2017-01-17 23:05:33 WARNING: Ignoring /home/gromero/packages/main/x86_64/APKINDEX.tar.gz: UNTRUSTED signature 2017-01-17 23:05:35 run apk update and try abuild again 2017-01-17 23:05:37 aha 2017-01-17 23:05:46 add your key to /etc/apk/keys 2017-01-17 23:06:21 https://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package#Setup_your_system_and_account there’s described how to prepare environemnt 2017-01-17 23:06:34 relevant part is abuild-keygen -a -i 2017-01-17 23:06:51 IIRC this generates key AND also adds it to your /etc/apk/keys 2017-01-17 23:07:25 but you probably already have a key, so it should be sufficient to just copy it into /etc/apk/keys (your public key, not private) 2017-01-17 23:07:35 just copied 2017-01-17 23:08:12 `apk update`, now it should not complain about untrusted signature 2017-01-17 23:08:24 then try abuild again 2017-01-17 23:08:33 ok 2017-01-17 23:08:55 still getting WARNING: Ignoring /home/gromero/packages/main/x86_64/APKINDEX.tar.gz: UNTRUSTED signature 2017-01-17 23:08:57 on update... 2017-01-17 23:09:07 hm, weird 2017-01-17 23:09:38 I'll try to generate the key again 2017-01-17 23:09:44 are you absolutely sure that this index is signed by the key you’ve copied to /etc/apk/keys? 2017-01-17 23:10:20 yep, that seems to be the problem... 2017-01-17 23:10:27 maybe you can try to regenerate it, but to be honest I don’t remember how to do it separately, abuild automatically regenerated apkindex when you install pkg into your repo 2017-01-17 23:11:01 yep, that's what I would ask you... I'll check here. thanks 2017-01-17 23:13:50 jirutka: just signed again (probably I changed the key) with: 2017-01-17 23:13:51 abuild-sign -k gromero-587e99ee.rsa /home/gromero/packages/main/x86_64/APKINDEX.tar.gz 2017-01-17 23:14:08 jirutka: I works. thanks a lot. 2017-01-17 23:14:13 *It works 2017-01-17 23:16:55 you’re welcome! :) 2017-01-17 23:17:00 :) 2017-01-18 08:03:12 Where can I find this: http://man7.org/linux/man-pages/man3/getaddrinfo_a.3.html within the scope of Alpine Linux? I have downloaded Linux kernel source and extracted it to /linux-kernel. When I perform: 'cd / && grep -r "gaicb" .', there are no search results! 2017-01-18 08:04:10 Do I need to download any package with apk (packet manager) or any additional source codes? From where? 2017-01-18 08:06:36 Here is my entire gcc configuration: http://pastebin.com/St4JWR1L (this contains the header include paths) 2017-01-18 08:11:23 czart, that's c-library function. it would found in 'musl' package, but i believe it's not implemented in musl as it's a GNU glibc extension. 2017-01-18 08:13:52 <^7heo> moin people 2017-01-18 08:18:34 @fabled: OK, I will check. Thanks. 2017-01-18 08:55:44 @fabled: Is any degree of relationship between "lib-dev" package and GNU glibc? I would only obtained some information with 'apk info lib-dev': "Meta package to pull in correct libc". 2017-01-18 08:57:02 czart, we used to do uclibc before musl. we had also at somepoint some private trees doing glibc. so in those cases libc-dev was different accordingly. currently we do only musl in the public tree. 2017-01-18 08:57:30 in any case, you cannot mix and match glibc and musl easily 2017-01-18 09:02:26 @fabled: The things (i.e. "gaicb") appear to be contained in libanl - "libanl is an asynchronous name lookup library." (http://tldp.org/LDP/lfs/5.0/html/appendixa/glibc.html) - which is contained in glibc. 2017-01-18 09:03:43 @fabled: So, the question is about how to workaround and build "libanl" on Alpine - is it even possible with accordance what you have said about mixing 'musl' and 'glibc'? 2017-01-18 09:10:10 czart, file portability bug against the failing project 'getaddrinfo_a' is not portable; in addition you might request musl libc to implement this function, 2017-01-18 09:10:37 though, not sure if there's substitute libanl available that implements it in portable way 2017-01-18 09:11:00 OK, I will search the web. Thanks. 2017-01-18 10:46:57 hi guys...a programming question that you would solve in one sec: how do i get the latest version of a software when they are formatted as: 2017-01-18 10:47:12 1.1, 1.2, 1.17, 1.8, 1.9 2017-01-18 10:47:32 ordering by the biggest return 1.9 2017-01-18 10:47:41 but the latest is 1.17 2017-01-18 10:47:52 the obiuvs thing would be: 2017-01-18 10:48:05 check the second digit, if not exist, add a 0 2017-01-18 10:48:11 so 1.9 would become 1.09 2017-01-18 10:48:26 depends on the version numbering scheme 2017-01-18 10:48:32 X.XX 2017-01-18 10:48:46 it's purely application defined 2017-01-18 10:49:27 what apk does is in: http://git.alpinelinux.org/cgit/apk-tools/tree/src/version.c 2017-01-18 10:49:35 it's based on the gentoo version string format 2017-01-18 10:49:41 Hi, found two identical packages main/perl-mail-tools & testing/perl-mailtools. just keep in mind :) 2017-01-18 10:50:39 in my case it's not a general purpose numerbing schema 2017-01-18 10:50:49 it's just x.xx 2017-01-18 10:50:57 So i can _ 2017-01-18 10:51:23 1. iterate the integer part, and the the decimal 2017-01-18 10:51:30 find the biggest 2017-01-18 10:51:33 and join together 2017-01-18 10:51:41 or 2017-01-18 10:52:01 2. adding a 0 in the decimal part when the second decimal does not exist 2017-01-18 10:52:24 there's still corner cases 2017-01-18 10:52:29 that is app specific 2017-01-18 10:52:35 how to compare e.g. 1.09 vs. 1.1 ? 2017-01-18 10:53:04 and if you write too much assumptions in your code, it's bound to break 2017-01-18 10:53:14 good point 2017-01-18 10:53:25 if you can live with apk rules, you can use "apk version -t" 2017-01-18 10:53:33 to compare two versions 2017-01-18 10:53:56 i'm parsing a web page containing the list of package versions 2017-01-18 10:54:37 remove the html tag, figure the latest version, then got notified 2017-01-18 10:59:45 fcolista, you can search for semver compare 2017-01-18 11:00:01 what language are you implementing it in? 2017-01-18 11:00:06 python 2017-01-18 11:00:16 i think there is a py-semver already 2017-01-18 11:00:40 https://pypi.python.org/pypi/semver 2017-01-18 11:00:40 https://pypi.python.org/pypi/semver 2017-01-18 11:00:44 gr8 2017-01-18 11:01:13 semver.max_ver() 2017-01-18 11:01:15 ok 2017-01-18 11:01:25 thx ncopa ! 2017-01-18 11:21:51 ncopa: I missed the "Create a patch" wiki page, so I opened issues instead of sending patches. I'm sorry! Should I resend this patches and ask for closing the issues? 2017-01-18 11:22:30 hi Ganwell, which issue is it? 2017-01-18 11:24:50 Ganwell: I suppose its #6698? 2017-01-18 11:25:11 ncopa: and another, just a sec. 2017-01-18 11:25:32 dont bother resend same patch 2017-01-18 11:26:05 duplicates are worse 2017-01-18 11:26:23 re cmus ffmpeg, id preer not use the old ffmpeg if possible 2017-01-18 11:26:33 ncopa: 6698, 6696, 6673 2017-01-18 11:27:19 but i suppose it make sense to use 2.8 til there is a new release of cmus 2017-01-18 11:27:20 ncopa: about cmus. Is it ok to use a git based version, I consider stable? 2017-01-18 11:27:55 i think i'd use the last version + patch for latest ffmpeg 2017-01-18 11:28:05 unless there are many patches to fix new ffmpeg support 2017-01-18 11:28:15 i suppose we can use ffmpeg2.8 for now 2017-01-18 11:28:28 i prefer use releases over git snapshot 2017-01-18 11:28:29 ncopa: ok great. 2017-01-18 11:29:28 <^7heo> ncopa: any chance to get https://github.com/alpinelinux/mkinitfs/pull/3 merged? 2017-01-18 11:29:46 <^7heo> jirutka: could you please review https://github.com/alpinelinux/mkinitfs/pull/3 quickly and mark it as ready if it so is? :) 2017-01-18 11:29:56 ^7heo, ah, yes 2017-01-18 11:30:03 ncopa: Yes. I probably could also do the patch for current ffmpeg support, but I'm glad I we can use ffmpeg2.8. Thanks a lot! 2017-01-18 11:30:49 jirutka: dont bother, i'll merge ^7heo's PR 2017-01-18 11:31:04 ^7heo: ah, sorry, I’ve read it 2017-01-18 11:31:37 ^7heo: and it’s basically just copy of .travis from aports repo, isn’t it? 2017-01-18 11:33:49 ^7heo: some stuff is not needed here and you can simplified it a lot by using https://github.com/jirutka/alpine-chroot-install/, there’s how I call it from script: https://github.com/bigclownlabs/bc-bridge/blob/master/script/ci-install#L28-L34 2017-01-18 11:33:57 :) 2017-01-18 11:34:27 ok, maybe it was a good idea that jirutka reviewd the travis stuff after all 2017-01-18 11:34:30 thanks! 2017-01-18 11:34:42 ^7heo: I’ll refactor aports/.travis later to use this script, to not repeating the same code on many places 2017-01-18 11:35:24 ^7heo: another way is to use minirootfs from https://alpinelinux.org/downloads/, that may simplify it a bit 2017-01-18 11:36:34 sorry for long delays, many problems in work these days, you know :/ 2017-01-18 11:36:51 <^7heo> jirutka: yeah no worries. 2017-01-18 11:37:05 <^7heo> jirutka: so it would be okay to merge it in this state for now, and revamp the travis alter? 2017-01-18 11:37:08 <^7heo> s/al/la/ 2017-01-18 11:39:06 ^7heo: yeah, there’s nothing wrong in it; you can remove setup-alpine , just keep `mkdir -p "${ALPINE_ROOT}${CLONE_DIR}", mount --bind "$CLONE_DIR" "${ALPINE_ROOT}${CLONE_DIR}"` 2017-01-18 11:39:27 ^7heo: and remove .travis/abuild-apk 2017-01-18 11:40:00 fold_start and fold_end (in commons.sh) are not used in the scripts you need, so can be also removed 2017-01-18 11:40:02 <^7heo> jirutka: I'll do that in a later PR 2017-01-18 11:40:09 <^7heo> I really need that one to be merged 2017-01-18 11:40:17 <^7heo> it's half a year old already. 2017-01-18 11:41:20 as well as print(), and replace its usage in `die()` with `printf '\033[1;31mERROR:\033[0m %s\n' "$1" >&2` 2017-01-18 11:41:48 <^7heo> ncopa: can we merge that one and I'll do the requested changes that jirutka asked in a later one? :) 2017-01-18 11:42:10 <^7heo> ncopa: I think it's quite important to automatically test mkinitfs's build asap 2017-01-18 11:42:32 I’d personally prefer to clean it before merging, but you know, I’m perfectionist :P 2017-01-18 11:42:37 <^7heo> yeah 2017-01-18 11:42:55 <^7heo> Let's do that in 2018 you're right. 2017-01-18 11:43:01 :P 2017-01-18 11:43:10 <^7heo> I think we do not need testing for mkinitfs, it's not really critical if it breaks. 2017-01-18 11:43:14 <^7heo> who cares about booting anyway? 2017-01-18 11:43:27 ofc we need it 2017-01-18 11:43:33 <^7heo> ACTION bangs his head on the desk 2017-01-18 11:43:37 <^7heo> I know :D 2017-01-18 11:43:42 <^7heo> I'm just saying 2017-01-18 11:43:45 actually I thought that it’s already merged :P 2017-01-18 11:43:45 <^7heo> this PR works 2017-01-18 11:43:55 <^7heo> I can open a PR RIGHT UP with that one merged 2017-01-18 11:43:57 <^7heo> however 2017-01-18 11:44:01 <^7heo> as long as this isn't merged 2017-01-18 11:44:08 <^7heo> anyone can open a PR that breaks mkinitfs 2017-01-18 11:44:16 <^7heo> and get it in, it it gets reviewed. 2017-01-18 11:44:29 sigh, so just merge it 2017-01-18 11:44:33 <^7heo> I can not :D 2017-01-18 11:44:37 aha XD 2017-01-18 11:44:41 <^7heo> I don't have the ownership rights there... 2017-01-18 11:44:47 <^7heo> I wouldn't be making such a fuss otherwise. 2017-01-18 11:44:58 <^7heo> I'd have merged it for 1 or 2 weeks 2017-01-18 11:45:05 <^7heo> and maybe be facing your wrath 2017-01-18 11:45:13 <^7heo> but at least we would have merging. 2017-01-18 11:45:23 <^7heo> Sorry for all the spam, I didn't get much sleep. 2017-01-18 11:45:53 give mi a minute 2017-01-18 11:46:07 <^7heo> :) 2017-01-18 11:52:16 <^7heo> jirutka: thanks! 2017-01-18 11:53:36 build succeeded 2017-01-18 11:53:57 <^7heo> yeah ofc it did. 2017-01-18 11:54:02 <^7heo> it did already before. 2017-01-18 11:54:16 <^7heo> now I don't know what you changed, but I'll guess I'll see during the rebase... 2017-01-18 11:54:28 <^7heo> I guess having to rebase is a small price to change to get it in. 2017-01-18 11:57:04 <^7heo> so you removed fold_start() and fold_end() in common.sh, the CLONE_DIR mounting (why?) and some stuff from the setup-alpine script 2017-01-18 11:57:12 <^7heo> and the lua-aports apk too 2017-01-18 11:57:48 clone_dir mounting is already in setup-alpine 2017-01-18 11:57:57 you don’t use lua-aports in your script 2017-01-18 11:57:58 <^7heo> I see. 2017-01-18 11:58:05 <^7heo> No I do not. 2017-01-18 11:58:16 <^7heo> again, I just took the stuff from the aports 2017-01-18 11:58:19 actually even abuild-sdk or how it it called, but it also pulls sudo etc. and I didn’t want to risk that I will break it :P 2017-01-18 11:58:25 so I kept that 2017-01-18 11:58:27 I know 2017-01-18 11:58:32 <^7heo> yeah 2017-01-18 11:58:36 <^7heo> thanks a lot for getting it in :) 2017-01-18 11:58:43 <^7heo> now I can rebase more stuff. 2017-01-18 11:59:15 <^7heo> what about the fold starts and ends? 2017-01-18 11:59:24 <^7heo> don't we want to keep that for the travis output? 2017-01-18 11:59:24 so I did just very quick clean up of what I was confident that I can remove without creating very embarassing moment that I’ve messed with that when I should not and break the stuff :P 2017-01-18 11:59:37 <^7heo> thanks :D 2017-01-18 11:59:42 <^7heo> I owe you a beer. :) 2017-01-18 11:59:58 these are used only in buildrepo script that 2017-01-18 12:00:20 <^7heo> ah ok 2017-01-18 12:00:28 <^7heo> so dead code for me. 2017-01-18 12:00:44 yeah, exactly 2017-01-18 12:01:57 tavis automatically folds output by commands executed in travis.yml, we use these fold commands in buildrepo b/c that script may build multiple packages and output is usually quite long 2017-01-18 12:01:59 <^7heo> jirutka: did you also remove the .travis/build-apk file? 2017-01-18 12:02:27 yes, also dead code 2017-01-18 12:02:27 <^7heo> (I think you did, just asking confirmation) 2017-01-18 12:02:28 <^7heo> ok 2017-01-18 12:02:31 <^7heo> yep 2017-01-18 12:04:36 <^7heo> ok, getting a lot of branches at the same point. 2017-01-18 12:07:22 <^7heo> omg it's so annoying to pop a stash done on a version that's nowhere anymore... 2017-01-18 12:07:26 <^7heo> SERIOUSLY. 2017-01-18 12:09:55 <^7heo> wow, done. 2017-01-18 12:10:03 <^7heo> fortunately I hadn't rebased everything yet. 2017-01-18 12:13:13 <^7heo> that'll teach me to use stashes for too long. 2017-01-18 12:20:39 heh, I know that feeling 2017-01-18 12:21:36 ufff, finally finished some urgent stuff of today’s morning/midday and can go to lunch 2017-01-18 12:22:33 <^7heo> enjoy your lunch 2017-01-18 12:29:22 Hi. What's the easiest way to set a package in aports/main to point to a specific tag (like apk-tools-28a9dcda568c575c569ffa4b68775034b655230a.tar.bz2)? 2017-01-18 12:29:54 giving that I'm using a specific release tag in aports 2017-01-18 12:41:35 hi again, found another duplicate packages main/perl-time-date & community/perl-timedate 2017-01-18 12:42:03 from it real names it seems that perl-mail-tools and perl-time-date - are wrong-named packages. 2017-01-18 12:42:26 Want to ask to move community/perl-timedate and testing/perl-mailtools to main/ 2017-01-18 12:42:29 <^7heo> vakartel: feel free to open PRs if you think it's really duplicates. 2017-01-18 12:42:40 <^7heo> (or send ML patches) 2017-01-18 12:44:33 need to move perl-timedate & perl-mailtools to main. Can I do it with PR? 2017-01-18 13:23:24 made a PR but needs to review it. https://github.com/alpinelinux/aports/pull/763 2017-01-18 13:55:03 <^7heo> ncopa: why did you do pipe to bc instead of using $(()) in test.sh? 2017-01-18 13:56:46 <^7heo> I initially had the `| bc` syntax as well, and replaced it by $(()) to reduce dependencies (bc is not a busybox applet - so it can be missing) and allow easier portability. 2017-01-18 13:57:33 <^7heo> oh wait nevermind 2017-01-18 13:57:40 <^7heo> I am reading the patch in reverse.. 2017-01-18 13:57:42 <^7heo> sorry 'bout that 2017-01-18 13:57:45 I can confirm that $(()) is specified in POSIX 2017-01-18 13:58:08 both the syntax and (non)required behaviour 2017-01-18 13:58:17 <^7heo> jirutka: my mistake, I confused ncopa's patch and mine. 2017-01-18 13:58:21 aha 2017-01-18 13:58:21 yes, better not introduce dependency of bc 2017-01-18 13:58:40 <^7heo> yeah 2017-01-18 13:58:44 <^7heo> sorry I misread diff 2017-01-18 13:59:26 ACTION knows POSIX Shell Command Language specification quite well since writing shell parser :P 2017-01-18 13:59:52 ACTION even things he’d prefer to not know… :/ 2017-01-18 14:00:03 <^7heo> ncopa: also the operation "clean" should not be left over now. 2017-01-18 14:00:31 <^7heo> oh wait 2017-01-18 14:00:39 <^7heo> it's independant from the clean flag 2017-01-18 14:00:50 <^7heo> this is confusing 2017-01-18 14:01:47 brb 2017-01-18 14:12:19 how’s our perl expert? fcolista? 2017-01-18 14:12:37 <^7heo> wb ncopa 2017-01-18 14:15:30 thanks :) 2017-01-18 14:17:04 <^7heo> is it okay to write a sh test like `[ ... ] && echo "success" || echo "fail"`? 2017-01-18 14:18:16 ^7heo: ok-ish 2017-01-18 14:18:55 ^7heo: when test fail, then the script should return non-zero status 2017-01-18 14:20:04 <^7heo> right 2017-01-18 14:20:10 better to write some helper func that will switch some global variable when first test fail 2017-01-18 14:20:12 <^7heo> that is being taken care of in the next step 2017-01-18 14:20:22 <^7heo> (exactly that actually) 2017-01-18 14:20:33 <^7heo> so aside the return code, is there anyting not okay with that? 2017-01-18 14:21:13 echo normally never fail, so no, this should work 2017-01-18 14:21:24 <^7heo> ok 2017-01-18 14:21:35 have you looked at https://github.com/sstephenson/bats ? 2017-01-18 14:21:53 <^7heo> nope 2017-01-18 14:22:00 take a look 2017-01-18 14:22:22 may be better than (re)inventing poor’s man testing fw :P 2017-01-18 14:22:30 what's up ? 2017-01-18 14:22:44 ^7heo: you have a `set -e` at the top so things like `[ ... ] && echo "success"` 2017-01-18 14:22:50 without the || 2017-01-18 14:22:58 will make script abort and exit with error 2017-01-18 14:22:58 <^7heo> jirutka: honestly, I don't like to download binaries randomly from internet and executing them. 2017-01-18 14:23:07 <^7heo> jirutka: I'd rather take an Open Source project, even incomplete. 2017-01-18 14:23:07 ncopa: well, not necessarily… 2017-01-18 14:23:47 ncopa: it depends on shell, some shell does handle [ ] && … specially, but I don’t remember how exactly ash 2017-01-18 14:23:48 <^7heo> ncopa: how would `echo` exit with an error? 2017-01-18 14:24:01 <^7heo> ah you mean the TEST would. 2017-01-18 14:24:04 ncopa: generally I’m trying to avoid this 2017-01-18 14:24:28 ^7heo: [ false ] && anything → 1 2017-01-18 14:24:40 ^7heo: that’s how `&&` works ;) 2017-01-18 14:25:19 <^7heo> well 2017-01-18 14:25:19 jirutka, do you know if there are anything like bats but for posix shell? 2017-01-18 14:25:24 or busybox ash 2017-01-18 14:25:30 <^7heo> also, not like bats 2017-01-18 14:25:33 <^7heo> open source. 2017-01-18 14:25:38 ncopa: unfortunately I don’t, I’d also prefer something non-bash 2017-01-18 14:26:11 <^7heo> ah misread 2017-01-18 14:26:25 <^7heo> damn, it's stupid to distribute something with bin/bats 2017-01-18 14:26:43 the sad fact is that most people silly assume that bash is the only one right shell and it’s everywhere 2017-01-18 14:26:56 <^7heo> especially when people call their script with /usr/bin/env bats and then /usr/bin/env bash 2017-01-18 14:27:06 <^7heo> doesn't really work well for readbility 2017-01-18 14:27:56 <^7heo> s/call their script/invoke the shell in their script with/ 2017-01-18 14:28:01 <^7heo> gosh I hate when I don't sleep enough 2017-01-18 14:28:06 fcolista: https://github.com/alpinelinux/aports/pull/763 2017-01-18 14:28:08 <^7heo> I do so many mistakes today 2017-01-18 14:28:23 me too 2017-01-18 14:28:33 <^7heo> and I write SO much 2017-01-18 14:28:39 <^7heo> I'm almost giving myself nausea 2017-01-18 14:28:42 <^7heo> I'm sorry people =/ 2017-01-18 14:29:21 fcolista: for sure he should at least add provides= and replaces= with old names, but I’m not sure if these pkgs should be really renamed etc. I don’t know perl ecosystem and pkgs at all 2017-01-18 14:29:55 jirutka, those are generated by apkbuild-cpan...so the name is how it comes from the CPAN 2017-01-18 14:30:12 looks that one of those two was built by hand 2017-01-18 14:30:37 well, apkbuild-cpan is quite outdated 2017-01-18 14:32:17 i would prefer to keep the one in main 2017-01-18 14:33:26 and remove the one in testing 2017-01-18 14:33:48 perl-mail-tools should have a provide= 2017-01-18 14:34:29 what is supposed to be set in "provides"? Is the package name find? 2017-01-18 14:34:31 *fine 2017-01-18 14:34:33 ? 2017-01-18 14:35:34 Ah ok 2017-01-18 14:36:30 ok 2017-01-18 14:36:57 think not ok. cpan name is Mailtools. so the package name must be the same 2017-01-18 14:37:05 no 2017-01-18 14:37:13 it's ok the patches you sent 2017-01-18 14:37:30 i agree: package name should be perl-mailtools 2017-01-18 14:37:55 I don't think is needed set a provides="" 2017-01-18 14:38:06 but replaces="" yes 2017-01-18 14:38:16 replaces="perl-mail-tools" 2017-01-18 14:39:14 I thought to add provides= for existing compatibility. but made depends fixes instead 2017-01-18 14:41:01 in any case one of them must be deleted to not support two identical 2017-01-18 14:42:05 agreed 2017-01-18 14:54:36 to less handwork it is better to remove wrong-named ones and move good-named to main. 2017-01-18 15:03:44 vakartel: ofc it’s good that you’ve already fixed dependencies, but users may have some custom abuilds, so it’s always good thing to also add provides 2017-01-18 15:08:04 why we need to provide wrong names? I fix it for existing, so people can fix those custom ones. 2017-01-18 15:10:49 for backward compatibility 2017-01-18 15:11:07 you’re renaming packages in the main repository 2017-01-18 15:15:10 as a variant, for notify them, can increase pkgrel and make a dummy contents in wrong-named packs with post-install notification. 2017-01-18 15:18:49 and then remove wrong-named at next release. 2017-01-18 15:25:48 ok, I add provides= and replaces= for perl-mailtools and perl-timedate and and bump their pkgrel 2017-01-18 16:47:07 Hi. I understand that gcc final pass (after pass1 and pass2) would depend on libc (glibc or musl), but looking at the gcc APKBUILD file that dependency is not there. Any clue on what I'm missing here? 2017-01-18 16:51:42 dependency? 2017-01-18 16:52:03 you expected somethign like: depends="musl"? 2017-01-18 16:52:49 abuild will dynamically autodetect the shared objects dependencies 2017-01-18 16:53:38 $ apk info --depends gcc | tpaste 2017-01-18 16:53:39 http://tpaste.us/MMO5 2017-01-18 16:53:59 gcc depends on whatever package that provides so:libc.musl-x86_64.so.1 2017-01-18 16:58:47 ncopa: hm, i see. yes, I was expecting something like depends="musl". Also, I would expect that final pass depends on pass1 and pass2... which else seems not tha case... 2017-01-18 16:59:29 ncopa: btw, I'm firstly trying to compile with glibc (gnu) 2017-01-18 16:59:51 I'm on a ppc64le btw, doing a cross compilation 2017-01-18 17:23:01 gromero: to bring up alpine on a new architecture, you need a musl-cross toolchain to start from 2017-01-18 17:34:01 kaniini: I've built cross gcc-pass1 and gcc-pass2. final pass just broke: http://pastebin.com/raw/HyG9s3bc 2017-01-18 17:34:56 but even so I would try to build using gcc-pass2 the musl package specifying the CROSS_COMPILE, like 2017-01-18 17:34:58 CTARGET=ppc64le CROSS_COMPILE=powerpc64le-alpine-linux-gnu- BOOTSTRAP=nolibc APKBUILD=/home/gromero/git/aports/main/musl/APKBUILD /usr/bin/abuild -r 2017-01-18 17:35:46 (abuild, apk-tools, and gcc APKBUILD is already patched to include ppc64le) 2017-01-18 17:36:01 *are already patched 2017-01-18 17:48:47 i dont really see why gnu libc needs to be involved 2017-01-18 17:49:52 gromero: did you see the aports/scripts/bootstrap.sh script? 2017-01-18 17:50:39 i'm going to port alpine to DEC alpha soon ;) 2017-01-18 17:52:42 better try with DEC VAX 2017-01-18 17:52:51 ncopa: yes, i did see. on facing some issues on that I tried to build gcc isolated... but I think I'll try to use musl as you said 2017-01-18 17:53:14 ncopa: btw, do I have to create any key before using the bootstrap.sh ? 2017-01-18 17:54:50 i think you do 2017-01-18 17:54:53 ncopa: getting that: https://hastebin.com/raw/suwayebupo (musl seems to be present) 2017-01-18 17:55:10 kaniini: cool :) 2017-01-18 17:55:17 brb 2017-01-18 17:55:26 skrzyp: i do not have any vax stuff 2017-01-18 17:56:06 gromero: i think you might need create apk key and copy the .pub to /etc/apk/keys 2017-01-18 17:59:09 ncopa: I've tried to add --allow-untrusted w/o success 2017-01-18 18:00:35 better create keys 2017-01-18 18:00:37 i need to go 2017-01-18 18:00:38 see u 2017-01-18 18:24:16 how could I determine exactly which key is used sign the packages created by bootstrap.sh? 2017-01-18 18:24:22 *to sign 2017-01-18 18:25:15 PACKAGER_PRIVKEY is set 2017-01-18 18:25:41 the pub and priv key is in ~/.abuild 2017-01-18 18:26:03 and the pub key is also in /etc/apk/keys 2017-01-18 18:26:21 but I still get untrusted errors on running bootstrap.sh... 2017-01-18 18:30:51 gromero_: have you read https://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package#Setup_your_system_and_account ? 2017-01-18 18:36:05 jirutka: yes, i did. just found that index is signed with key gromero-587e99ee.rsa, and if I use --keys-dir it works. (i created my key using abuild-keygen -a -i as we discussed yesterday...). Here is my config: http://pastebin.com/raw/8RCWsfdy 2017-01-18 18:36:42 but I'm wondering with --keys-dir is necessary... something is missing 2017-01-18 18:38:03 *why 2017-01-18 18:38:57 s/del/add will give the same result also 2017-01-18 18:39:02 gromero_: you’re using --root, I’m not sure, but IIRC it also changes the path when it tries to find keys by default, to $ROOT/etc/apk/keys 2017-01-18 18:39:11 so --keys-dir may be necessary in this case 2017-01-18 18:39:20 and it seems that your paste confirms that 2017-01-18 18:39:39 you’ve used it in the second command and it haven’t printed the error message about untrusted key 2017-01-18 18:40:42 hm, right. actually bootstrap.sh is using --root by default... 2017-01-18 18:46:36 bootstrap.sh seems to not understand where the key is (even adding --keys-dir to it)... 2017-01-18 18:52:06 have you tried to add the keys to $ROOT/etc/apk/keys? 2017-01-18 18:56:29 has ppc64le been added to apk-tools as an arch? 2017-01-18 18:58:18 jirutka: it worked, thanks a lot 2017-01-18 18:58:32 kaniini: yes, it's added 2017-01-18 18:58:35 your welcome 2017-01-18 18:58:39 I just opened a pull request to track ppc64le inclusion (#765), but, before that, I sent musl patches to alpine-aports. Should I concentrate all these patches into just one pull request? I.e, give up the mailing list add the patch into the pull request? 2017-01-18 18:59:41 well, it has been indeed added to apk-tools, but not backported to the abuild 2017-01-18 18:59:54 so if you use apk from the repository, then it does not have this 2017-01-18 19:00:29 do you need it to be ported to the apk-tools package in edge? 2017-01-18 19:03:41 kaniini: when you said "you need a musl-cross toolchain to start from", did you mean this https://github.com/GregorR/musl-cross ? 2017-01-18 19:03:59 that is one way to do it 2017-01-18 19:04:08 if you have problems with the bootstrap.sh it may be a way to do it 2017-01-18 19:04:28 kaniini: I see. thanks 2017-01-18 19:05:06 kaniini: are you the one who argued against grsecurity in favour of vanilla kernel with some security enhancements? 2017-01-18 19:05:15 don’t remember who was that 2017-01-18 19:05:31 jirutka: yes 2017-01-18 19:06:03 kaniini: great! do you have some time to prepare kernel config? :) 2017-01-18 19:06:25 jirutka: possibly this weekend 2017-01-18 19:07:04 i have a logistical question concerning the ppc64le port 2017-01-18 19:07:23 will IBM be operating builders for it? 2017-01-18 19:07:56 kaniini: I personally don’t have problem with grsec kernels (yet), but when I listen to stories of my friend how badly is bleeding edge kernel sometimes broken, I’m starting to be afraid… and if you say that you can get similar security with stable vanilla kernel, it would be really great to have it 2017-01-18 19:08:12 kaniini: perfect, I’m looking forward to it :) 2017-01-18 19:08:43 jirutka: i do not have problem with offering grsec as option for those who want it, it's the people that having it as default for so long attracted that are the problem 2017-01-18 19:08:50 kaniini, yes. This would be our next questions to you. I can provide some builders for the project, for sure. 2017-01-18 19:09:00 Does a remote machine work? 2017-01-18 19:09:01 kaniini: yes, I understand that point 2017-01-18 19:09:31 if they want to take the risk of running experimental code for unrealistic attack vector mitigation, that's their freedom to choose 2017-01-18 19:10:38 "good enough" security should also be a possibility 2017-01-18 19:11:39 leitao: yes i don't see why it would be a problem, it's just making sure there is a builder available. if the direct vendor (in this case IBM) is providing it, then that's probably the best possible option really 2017-01-18 19:11:52 btw it would be absolutely perfect if you can also write few words about the setup, how it’s more secure than others and how it compares with grsecurity; you know, for PR :P so we can write on the web that it’s not just some stable vanilla kernel, but that it’s also very secure :) 2017-01-18 19:12:54 but it is literally just the same hardening as ubuntu server and chromium os 2017-01-18 19:13:20 kaniini, sure. If a remote machine works, then I can have it/them really soon. How many builders is necessary? 2017-01-18 19:13:30 just one probably 2017-01-18 19:13:47 kaniini, ok. 2017-01-18 19:13:51 we only really use multiple builders for 'slow' architectures 2017-01-18 19:13:57 that’s pity, I hoped that it’d be better than Ubuntu 2017-01-18 19:13:58 like arm 2017-01-18 19:14:22 jirutka: i do have plans to enhance the yama LSM to provide more grsecurity features 2017-01-18 19:14:28 jirutka: like /proc restrictions 2017-01-18 19:14:41 kaniini, right. ppc64le are fast servers. They are the fastest builders in Debian nowadays. 2017-01-18 19:14:44 kaniini: yeah, yama LSM, that’s what you were talking about before, I forgot the name 2017-01-18 19:17:08 of course, the grsecurity crowd who have shown up say unpleasant things when you say that anything less than grsecurity may be good enough 2017-01-18 19:18:16 jirutka: ubuntu security hardening is not that bad, really 2017-01-18 19:18:31 jirutka: they could do with less non-PIE binaries, but other than that 2017-01-18 19:18:48 the main things left to do is restrict access to /proc 2017-01-18 19:19:03 which could be done with permissions, but that makes /proc completely unusable 2017-01-18 19:23:45 kaniini: I personally don’t have any problems with the grsecurity’s approach for hardening /proc; just needed to add e.g. nrpe to procps group etc. 2017-01-18 19:24:00 as an admin, I find grsecurity's restricted /proc completely unusable already 2017-01-18 19:24:32 kaniini: do I understand it correctly, is grsecurity and /proc just about permissions? 2017-01-18 19:24:48 skarnet: b/c you refuses to learn anything new, as you told us… 2017-01-18 19:25:23 skarnet: I use grsecurity for 5+ years on metal and VMs, without any noticeable limitation or problem 2017-01-18 19:25:36 maybe you don't use it the same way I do 2017-01-18 19:25:42 that’s possible 2017-01-18 19:26:40 jirutka: i meant vs. vanilla 2017-01-18 19:27:02 anyway, this can be as an option that you can disable, right? at least with grsec you can disable some features with arguments on kernel line, don’t remember if /proc hardening 2017-01-18 19:30:35 yeah sysctl or something 2017-01-18 19:43:54 kaniini: btw one friend of mine works on https://github.com/crossdistro (he told me about it yesterday) – some tools for generating kernel config; I haven’t look at it yet, just that it may be interesting for you 2017-01-18 19:45:44 eh, wrong link, this https://github.com/crossdistro/kernel-tools 2017-01-18 22:43:45 :/ no fastcgi anymore? 2017-01-18 22:44:39 wherE? 2017-01-18 22:46:07 or why did my cgit break... 2017-01-18 22:55:07 I have no idea 2017-01-18 23:18:20 jirutka: I'm getting the following error on gcc final pass: https://hastebin.com/raw/evujixixis 2017-01-18 23:19:03 basically, Running split function libgcc... is looking for libgcc_s.so.* in .../lib 2017-01-18 23:19:12 but it must be in lib64 instead 2017-01-18 23:19:30 is it familiar in some way to you? 2017-01-18 23:20:16 since: 2017-01-18 23:20:17 alpine01:~/git/aports/main/gcc$ find . -type f -name libgcc_s.so\* 2017-01-18 23:20:17 ./pkg/gcc-ppc64le/usr/powerpc64le-alpine-linux-musl/lib64/libgcc_s.so 2017-01-18 23:20:17 ./pkg/gcc-ppc64le/usr/powerpc64le-alpine-linux-musl/lib64/libgcc_s.so.1 2017-01-18 23:20:54 no, we don’t use lib64 at all 2017-01-18 23:20:59 . /usr/lib is correct 2017-01-18 23:22:53 thanks 2017-01-18 23:40:34 jirutka: what sets lib64 in ./pkg? 2017-01-18 23:41:25 gromero_: this can be usually configured in configure autohells script or cmake 2017-01-18 23:41:39 ah! ok 2017-01-18 23:45:05 what is suddenly all this ppc64le? did i miss something? 2017-01-18 23:46:06 clandmeter: gromero_ is bootstrapping Alpine on ppc64le :) 2017-01-18 23:46:42 right 2017-01-18 23:46:45 what about leitao? 2017-01-18 23:46:52 or am i mixing things? 2017-01-18 23:46:57 leitao? o.O 2017-01-18 23:47:19 he also seems to talk about it 2017-01-18 23:47:24 aha 2017-01-18 23:47:26 don’t know 2017-01-18 23:47:38 invasion of power 2017-01-18 23:47:40 pc 2017-01-18 23:47:42 :) 2017-01-18 23:48:12 gromero_: which hardware do you use? 2017-01-18 23:48:42 clandmeter: leitao is my colleague. we both are bootstrapping it 2017-01-18 23:48:52 a nice 2017-01-18 23:48:53 clandmeter: it's a POWER8 2017-01-18 23:49:00 :) 2017-01-18 23:49:03 which company if i may ask? 2017-01-18 23:49:10 IBM 2017-01-18 23:49:29 ah ok 2017-01-18 23:49:30 just boostrapping LE 2017-01-18 23:49:31 welcome 2017-01-18 23:49:40 clandmeter: thanks! 2017-01-18 23:50:10 when are you going to upgrade? 2017-01-18 23:50:11 :) 2017-01-18 23:50:24 v9 is around the corner right? 2017-01-18 23:51:00 clandmeter: I guess so, but I don't have details on it... :( 2017-01-18 23:51:56 it's mid 2017 2017-01-18 23:52:34 at least, that's what IBM said at Hot Chips 2017-01-18 23:53:07 they're gonna start around 6k$ too, really nice price point to challenge sparc 2017-01-18 23:53:49 i know our company joined the foundation and could start with power9. 2017-01-18 23:54:42 the openpower foundation you mean? is POWER9 available to them already? 2017-01-18 23:54:49 thought that wasn't even coming out until early 2018 2017-01-18 23:54:59 i have no clue about dates 2017-01-18 23:55:02 post-ORNL deployment 2017-01-18 23:55:05 just rumors from within 2017-01-18 23:55:28 ah, I have my eyes on this for many reasons :) 2017-01-18 23:55:46 gromero_: just a curious, can I ask, IBM has some interests with Alpine or you’re doing it just as a hoppy? 2017-01-18 23:56:53 jirutka: is that the "new" hobby? 2017-01-18 23:56:58 pretty sure IBM already helped port musl out, so I think they are going through and trying to add support for ppc64le to major distros 2017-01-18 23:57:33 it helps us end users have more choice, and helps them in marketing :) 2017-01-18 23:57:34 eh, sry, I mean hobby, just a typo 2017-01-18 23:57:39 :) 2017-01-18 23:57:50 hehe, I read that as "as a happy" 2017-01-18 23:58:28 I must be really tired, “b” and “p” are pretty far from each on keyboard 2017-01-18 23:58:36 jirutka: you should press the b's harder so they dont drop down ;-) 2017-01-18 23:58:41 I am personally interested in seeing if IBM still keeps big endian POWER alive for non-AIX systems 2017-01-18 23:58:42 XD 2017-01-18 23:58:43 jirutka: no, *not* a hobby :) 2017-01-18 23:59:16 awilfox: RHEL has BE support 2017-01-18 23:59:20 I have significant investment in big-endian POWER systems and am pretty much maintaining them myself now since major distros all decided to go le 2017-01-18 23:59:21 gromero_: oh, that’s wonderful, I wouldn’t expect this from company like IBM :) 2017-01-18 23:59:27 gromero_: ah, good to know. 2017-01-18 23:59:55 debian dropped BE support, and soon after I saw one other major one drop it... don't remember which 2017-01-19 00:00:01 made me a bit nervous 2017-01-19 00:00:15 yes 2017-01-19 00:00:28 Ubuntu just support LE from start IIRC 2017-01-19 00:00:28 it is a big selling point for me that it is BE, very fast for network processing and also some audio algorithms 2017-01-19 00:01:20 jirutka: :) Happy to contribute to the Alpine community 2017-01-19 00:02:41 gromero_: you are interested in alpine or more in getting it boostrapped for musl? 2017-01-19 00:03:07 clandmeter: focused on alpine boostrap on PPC64 LE 2017-01-19 00:03:57 clandmeter: I understand that getting musl working is a pre-requisite, right? 2017-01-19 00:04:01 awilfox: BE for network processing to avoid all those ntohs() and similar? 2017-01-19 00:04:06 gromero_: what’s the real difference between LE and BE, the reason why IBM supports both? 2017-01-19 00:04:12 if you do need any assistance with power stuff that isn't specific to LE let me know, I have been spending most of my holiday bootstrapping my musl-based distro on ppc64 2017-01-19 00:04:18 pavlix: yes 2017-01-19 00:04:33 pavlix: they are no-ops on BE architectures 2017-01-19 00:04:36 jirutka: ah, that predates my existance, better ask leitao tomorrow :) 2017-01-19 00:04:39 awilfox: obviously 2017-01-19 00:04:49 pavlix: I just wanted to write you to ask yourself and here we go, you’re here :) 2017-01-19 00:05:10 awilfox: pardon my ignorance but does it actually make a change for real applications? 2017-01-19 00:05:15 awilfox: great. thanks for you help. will do 2017-01-19 00:06:27 jirutka: here I am, asking an entirely different question :) 2017-01-19 00:06:47 jirutka: yes, leitao would know for sure all the details on BE/LE :) 2017-01-19 00:07:49 gromero_: most of the alpine boostrapping is/has been done by fabled and ncopa, so if you have questions they are the right persons (not excluding anybody else here) 2017-01-19 00:08:21 im going to bed, gnite! 2017-01-19 00:08:26 clandmeter: gn! 2017-01-19 00:08:32 night! 2017-01-19 00:09:03 pavlix: it can, yes. moreso on ARM where there are many cycles for memory fetch/write (which is ixp425 and similar chips run arm in its little-known big-endian mode) 2017-01-19 00:09:36 clandmeter: yes, I'm talking to ncopa. thanks for the information! :) 2017-01-19 00:09:45 pavlix: but when you are doing stateful filtering on 10 or 40 gig lines.. well.. there is a reason most enterprise routers are power or mips-based 2017-01-19 00:10:50 pavlix: on gigE, it probably doesn't matter as much. 2017-01-19 00:17:08 jirutka: gcc-ppc64le-6.3.0-r1.apk generated :) 2017-01-19 00:17:18 gromero_: congratz! \o/ 2017-01-19 00:17:50 awilfox: I see 2017-01-19 00:18:51 jirutka: well, no really possible without great support from great people in here :) 2017-01-19 00:19:24 :) 2017-01-19 00:20:40 awilfox: and you choose endianess at boot time by some CPU configuration mechanism? 2017-01-19 00:21:18 pavlix: yes, setenv little-endian?{true|false} 2017-01-19 00:21:47 pavlix: well, on Power. on ARM, I believe it is jumper-like interface (not software settable) 2017-01-19 00:22:20 note that I only touch POWER4 and POWER6 personally. not sure if it has changed since 2017-01-19 00:22:44 by the time I could afford off-lease POWER8 systems... POWER9 was announced... and I would rather save up for one of them. 2017-01-19 00:24:25 awilfox: Adélie linux? :) (looking at github) 2017-01-19 00:24:57 that is the distribution I founded, yes. 2017-01-19 00:25:18 we use apk-tools too because it is the best package format I could find 2017-01-19 00:26:09 and the tools are pretty well written; there are some issues with it but Timo is working hard on a new version and it sounds like it will be even better! 2017-01-19 00:33:34 hm, no boostrap breaks when trying to install g++-ppc64le and gcc-gnat-ppc64le: https://hastebin.com/raw/ayiwimovah 2017-01-19 00:33:39 *now bootstrap 2017-01-19 00:34:12 since libstdc++-6.2.1-r1 is installed? 2017-01-19 00:35:49 and boostrap wishes to install g++-ppc64le-6.3.0-r1.apk and gcc-gnat-ppc64le-6.3.0-r1.apk... 2017-01-19 00:36:38 abuild-apk add --repository /home/gromero/packages//main gcc-gnat gcc-ppc64le g++-ppc64le gcc-gnat-ppc64le 2017-01-19 00:36:54 gcc-gnat should not be here imho 2017-01-19 00:38:45 it's from cross_compile_base(): ${SUDO_APK} add --repository "$REPODEST/main" gcc-gnat gcc-$TARGET_ARCH g++-$TARGET_ARCH gcc-gnat-$TARGET_ARCH \ 2017-01-19 00:38:53 in boostrap.sh 2017-01-19 00:39:05 wierd 2017-01-19 00:39:06 that’s weird, why is this in the bootstrap script 2017-01-19 00:39:14 i see 2017-01-19 00:39:32 is there an Ada package lurking in alpine-base? 2017-01-19 00:40:28 no, alpine-base does not depend on ada 2017-01-19 00:40:35 alpine01:~/git/aports/scripts$ sudo apk add alpine-base 2017-01-19 00:40:35 OK: 692 MiB in 145 packages 2017-01-19 00:40:35 alpine01:~/git/aports/scripts$ apk info | fgrep ada 2017-01-19 00:40:54 i believe there is no ada in alpine-base 2017-01-19 00:41:10 that is weird then :) gnat should not be a prerequesite for bootstrap then 2017-01-19 00:41:15 maybe it just does all gcc subpackages 2017-01-19 00:41:49 what if you remove all occurrences of gcc-gnat (not gcc-gnat-$TARGET_ARCH) in the script? 2017-01-19 00:42:44 I must ask fabled about this, I’m really curious why is gcc Ada support in the bootstrap script at all 2017-01-19 00:43:20 pavlix: any idea what may be the reason? 2017-01-19 00:43:39 awilfox: create_cross_compiler() installs gcc-gnat 2017-01-19 00:44:04 so even if I remove it from cross_compile_base() it gets installed... 2017-01-19 00:44:16 (and remove manually...) 2017-01-19 00:44:32 jirutka: it's been a long time since I used a cross compiler 2017-01-19 00:44:33 gromero_: I mean remove it directly from the bootstrap script 2017-01-19 00:44:35 gromero_: what if as jirutka says, you remove all 'gcc-gnat' 2017-01-19 00:44:43 from the script 2017-01-19 00:44:50 ok 2017-01-19 00:45:23 maybe something like: sed '/gcc-gnat/d' bootstrap-nognat.sh 2017-01-19 00:45:33 or however you like 2017-01-19 00:45:44 oh wait.. you can't do that 2017-01-19 00:45:49 that would remove the entire line 2017-01-19 00:45:50 nope, there’s also `gcc-gnat-$TARGET_ARCH` 2017-01-19 00:46:00 ah don't listen to me, I am starting to become tired :P 2017-01-19 00:46:01 yeah, that’s problem too :) 2017-01-19 00:46:05 awilfox: same errors... 2017-01-19 00:46:08 jirutka: and I have no idea why you would need Ada for bootstrap 2017-01-19 00:46:17 pavlix: me neither 2017-01-19 00:46:53 jirutka: you should only need to be able to build C AFAIK 2017-01-19 00:46:53 gromero_: you have ppc64le 6.3.0, but your host seems to have 6.2.1 2017-01-19 00:46:58 gromero_: are you running edge? 2017-01-19 00:47:12 jirutka: (unless you're boostrapping haskell and such stuff) 2017-01-19 00:47:28 awilfox: yes, my host is 6.2.1, but bootstrap create 6.3.0 cross files 2017-01-19 00:47:43 awilfox: edge? 2017-01-19 00:47:44 gromero_: they must match exactly for alpine bootstrap 2017-01-19 00:48:05 awilfox: no, it's standard (if I got your question correctly...( 2017-01-19 00:48:09 gromero_: http://wiki.alpinelinux.org/wiki/Edge 2017-01-19 00:48:10 to be more clear, when you do this change http://tpaste.us/EL6O, it still tries to install gcc-gnat? 2017-01-19 00:48:48 eh,sry for messed patch 2017-01-19 00:49:10 jirutka: nope, it does not try to install gcc-gnat, however the error is the same 2017-01-19 00:49:23 I guess it's not edge as awilfox says 2017-01-19 00:49:27 here http://tpaste.us/0N9E 2017-01-19 00:49:40 yes, awilfox is right about edge 2017-01-19 00:50:14 hmm 2017-01-19 00:50:38 edge means the latest Alpine version, unstable 2017-01-19 00:51:11 for example http://dl-cdn.alpinelinux.org/alpine/edge/main instead of http://dl-cdn.alpinelinux.org/alpine/v3.5/main 2017-01-19 00:51:57 right. that's the problem so. I'll upgrade. ta jirutka awilfox 2017-01-19 01:00:27 awilfox: worked as you said :) 2017-01-19 01:01:30 building now ">>> linux-headers" 2017-01-19 01:01:59 :D 2017-01-19 01:02:02 that is great! 2017-01-19 01:02:41 :D 2017-01-19 07:30:02 clandmeter any objection to http://patchwork.alpinelinux.org/patch/2808/ ? 2017-01-19 08:15:53 rnalrd: nope 2017-01-19 08:16:06 k, tnx 2017-01-19 12:33:38 So what about php7.1 upgrade? Today php 7.1.1 is out 2017-01-19 12:42:44 jirutka, hi there. I just updated PR#765. Should I toggle the changes-request label? 2017-01-19 12:58:53 as variant, I can upgrade 7.0 branch to 7.0.15 and create a new package testing/php71. But all changes I made for 7.1 would be in updated 7.0 because there are too many fixes for it. And those fixes also need to be done in php5 2017-01-19 12:59:52 vakartel: merge it into community/php7? 2017-01-19 13:00:41 not sure how others, I’m strongly against introducing pkg php71, this is a way to hell 2017-01-19 13:01:41 so what about to upgrade 7.0 -> 7.1 ? 2017-01-19 13:02:15 that’s needed 2017-01-19 13:02:36 I just push a comit to upgrade 7.1.0 -> 7.1.1 (PR716) 2017-01-19 13:03:21 but it still not accepted 2017-01-19 13:04:03 please discuss this with Andy Postnikov, he’s the current main maintainer of php stuff 2017-01-19 13:04:44 I see that he already approved the PR 2017-01-19 13:06:11 He review it 8 days ago and his verdict 'This looks totally great!' 2017-01-19 13:06:51 So I just don't know what to do else 2017-01-19 13:27:42 btw we may change default php version to php7 instead of php5, but it must be done with caution 2017-01-19 13:27:46 vakartel: ^ 2017-01-19 13:35:31 In one of my previous commits I made php5 & php7 (cli, cgi, dev-tools) live on same host without conflicts. Then I drop it. For now the only php-fpm[57] can. I can make the same when php7 and php5 would be synced via apkbuild. 2017-01-19 13:36:10 vakartel: that’s good that you’ve doped it, b/c that trigger was horrible hack with terrible consequences 2017-01-19 13:37:09 this reminds me that we must discuss the update-alternatives topic really soon 2017-01-19 13:37:30 the absence of such mechanism is still bigger and bigger problem 2017-01-19 13:38:03 The trigger was in history. :) for now the only need is: replaces= 2017-01-19 13:38:19 to replace symlinks 2017-01-19 13:39:06 it's better update-alternatives variant 2017-01-19 13:40:51 replaces is not a solution, just another half-working workaround 2017-01-19 13:43:29 <^7heo> let's do half assed "solutions"! 2017-01-19 13:43:34 <^7heo> I want my 15 minutes of fame too! 2017-01-19 13:43:36 <^7heo> ACTION hides 2017-01-19 13:47:13 .... I think to try to push all php versions to apkbuild: 7.0 (as stable), 7.1 (as current) and 5.6.29 (as historical stable) 2017-01-19 13:51:55 vakartel: do you mean v3.5 as stable and edge as current? not two separate abuilds in edge? 2017-01-19 13:55:25 I mean all php versions (5 7.0 7.1) can be built via one apkbuild. I successfuly test building php5 with my current php7 apkbuild 2017-01-19 13:56:23 I’m not sure if it’s a good idea to have single abuild for php5 and php7 2017-01-19 13:56:37 but we’ll see 2017-01-19 13:58:04 it's nearly the same, php5 has 2 old extensions (mysql, mssql) and a different patch for configs 2017-01-19 13:59:05 and some things like pear, phar are the same in all versions 2017-01-19 14:00:37 aha, I thought that there are more differences 2017-01-19 14:04:27 Hi there. What is happening exactly when I see a error like this: 2017-01-19 14:04:30 libltdl-2.4.6-r1 (missing): 2017-01-19 14:04:30 required by: world[libltdl-2.4.6-r1] 2017-01-19 14:04:49 when running: # apk add libltdl-2.4.6-r1 2017-01-19 14:05:25 try apk add libltdl 2017-01-19 14:06:30 asie, it works. :-) Thanks 2017-01-19 14:17:40 clandmeter, i'm testing syslinux uefi, and seems it boots on x86_64 for me 2017-01-19 14:17:55 (that is syslinux prompt boots, need to test next alpine) 2017-01-19 14:18:12 fabled: that’s great! :) 2017-01-19 14:33:38 leitao: it means it doesnt exist 2017-01-19 14:33:48 in that specific case, probably the specific version requrested doesn't exist 2017-01-19 14:36:52 abump does not support bump of pkgrel right? 2017-01-19 14:36:54 http://git.alpinelinux.org/cgit/abuild/tree/abump.in 2017-01-19 14:37:00 it seems pkgver only 2017-01-19 14:37:40 fcolista: yes, b/c it quite doesn’t make sense to use abump for bumping pkgrel 2017-01-19 14:37:58 jirutka, that's not true 2017-01-19 14:38:16 fcolista: pkgrel is usually bumped when you change something in apkbuild, so when you’re in it, you can also bump pkgrel yourself in the editor, right? :) 2017-01-19 14:38:31 it make sense when ABI change, and you need to bump pkgrel for all packages dependent 2017-01-19 14:38:37 aha, right 2017-01-19 14:38:48 would make the rebuild easier and faster 2017-01-19 14:38:53 and scripted 2017-01-19 14:39:02 you’re right 2017-01-19 14:39:19 but then it’d be even more help to have script that will automatically bump depending pkgs 2017-01-19 14:40:00 this is yet another thing blocked by apk rewrite :( 2017-01-19 14:40:02 yes, ofc, but having one tool that does the small jobs atomically is more unix-like 2017-01-19 14:40:03 ;) 2017-01-19 14:40:23 b/c new apk will support resolving dependencies, of all pkgs, not just installed 2017-01-19 14:40:25 but yes, such script would help as well 2017-01-19 14:40:31 yes, I absolutely argree 2017-01-19 14:41:14 so I’m taking it back, you’re right that it’d be indeed useful to improve abump for bumping pkgrel 2017-01-19 14:42:56 question for you guys on my ghc port, i found out that 8.0.2 has some fun bug that causes explosive memory use when compiling one of the ported packages (stack) and on arm if you don't have 8GiB, it won't compile 2017-01-19 14:43:23 uh, that’s nasty 2017-01-19 14:43:25 should i just keep the submitted port at 8.0.1 for now and backport the changes from 8.0.2 that make it a lot easier to have ported? 2017-01-19 14:43:42 yeah, it was a bugfix that caused it but looks like it had other side effects 2017-01-19 14:43:56 it works fine as a note 2017-01-19 14:44:08 I’m not sure if we already build armhf on new machine with a lot of RAM 2017-01-19 14:44:20 clandmeter: ^ 2017-01-19 14:45:19 I think that keeping 8.0.1 and backporting few patches would be better for now, but don’t know the problem in detail, so hard to decide 2017-01-19 14:45:23 I can keep it at 8.0.1 for now, 8.0.2 isn't that big of a change 2017-01-19 14:45:40 its nicer cause less patches but not that huge of a difference 2017-01-19 14:45:51 the important is if we even can build it, i.e. if armhf is already build on new infra; if it is, then it’s basically up to you 2017-01-19 14:45:54 ok, i'll need to update the bootstrap thing again and ghc and retest 2017-01-19 14:45:59 if it’s not, then we wouldn’t be able to build it 2017-01-19 14:46:13 technically I can't build it with 4GiB of ram :/ 2017-01-19 14:46:28 sorry, I need to go now 2017-01-19 14:46:29 stack not ghc but i use stack for other things 2017-01-19 14:46:33 no worries 2017-01-19 14:46:50 I hope that I’ll have more time now and be able to finally review ghc 2017-01-19 14:46:58 if no one else do it meanwhile 2017-01-19 14:47:16 well i'd hold off if the idea is 8.0.1 with a few more patches 2017-01-19 14:47:32 i need to patch the bootstrap compiler with some stuff again and validate that 8.0.1 with patches works on 3.5 2017-01-19 14:47:46 i imagine it'll be fine, but like to be thorough 2017-01-19 14:48:33 i'll quick get a branch to test locally and let it build over the workday, it takes a year to build the cross built cross compiler 2017-01-19 14:52:24 the bug in question if anyone cares https://ghc.haskell.org/trac/ghc/ticket/13059 2017-01-19 17:30:07 gromero: which bootloader will the ppc64le port use? 2017-01-19 17:35:13 leitao: which bootloader will the ppc64le port use? 2017-01-19 17:35:39 kaniini, does alpine supports grub? 2017-01-19 17:36:52 not presently, but i think it may be packaged to support the arm64 port 2017-01-19 17:36:55 due to uefi 2017-01-19 17:38:21 kaniini, right, that would be our ideal bootloader. We support others, but, grub seems to be the best supported in Power at this moment. 2017-01-19 18:38:23 hey, I am getting this: #error "unknown long double size, cannot define BFP_FMT" 2017-01-19 18:38:24 during the build-cross-native when bootstrapping. 2017-01-19 18:39:39 I saw that perhaps setting libgcc_cv_dfp=no would fix this, but setting it in the APKBUILD configure did not work (I guess it just set for the gcc, not libgcc). 2017-01-19 18:40:14 Where is the best place to set it? 2017-01-19 18:41:31 btw, it's on our ppc64le bootstrap ^ :) 2017-01-19 18:42:23 i think that goes in the autoconf stuff 2017-01-19 18:42:28 so uh 2017-01-19 18:43:01 ferseiti: something like 2017-01-19 18:43:14 echo libgcc_cv_dfp=no > config.site 2017-01-19 18:43:31 CONFIG_SITE=config.site ./configure ... 2017-01-19 18:43:39 in compile() 2017-01-19 18:43:46 instead of the current ./configure ... stanza 2017-01-19 18:50:46 Shiz, I will try that, thanks 2017-01-19 20:17:29 hey must be blind ... were's java web start ? 2017-01-19 20:18:39 wat start? 2017-01-19 20:18:52 icedtea-web < ohhh 2017-01-19 21:19:33 hey folks 2017-01-19 21:19:50 I know there's this install script for raid/lvm/crypt disks 2017-01-19 21:19:57 from someone from here 2017-01-19 21:20:06 (sorry I forgot your name) 2017-01-19 21:20:12 can someone remind me the URL ? 2017-01-19 21:20:18 might be mine 2017-01-19 21:20:28 but i tried it today and it looks like it needs to be updated for 3.5... 2017-01-19 21:20:31 if you're using raid, at least 2017-01-19 21:21:17 mmmmh in my brain memory it's not from you 2017-01-19 21:21:19 :D 2017-01-19 21:22:24 gosh I'm too lazy to do it by hand 2017-01-19 21:27:29 leitao: is the linux config in PR765 the same as for other arches, just with arch-depend flags changed to power8? 2017-01-19 21:31:05 jirutka, no. 2017-01-19 21:31:14 The configs were generated by make def_powernv 2017-01-19 21:31:18 it is the default by the kernel 2017-01-19 21:32:11 aha, I don’t think that configs in linux-vanilla are defaults 2017-01-19 21:32:37 I’ll tell ncopa to look at it, I don’t know how he created these configs 2017-01-20 00:08:29 ... 2017-01-20 00:08:38 upgraded 3.4.6 to 3.5.0 2017-01-20 00:08:52 Error relocating /sbin/lvm: dm_report_group_push: symbol not found 2017-01-20 00:09:09 mount: mounting /dev/mapper/pryderi-root on /sysroot failed: No such file or directory 2017-01-20 00:09:12 Mounting root failed. 2017-01-20 00:10:44 :| 2017-01-20 00:10:54 regenerate initramfs? 2017-01-20 00:12:20 http://i.imgur.com/mvEM4IM.png 2017-01-20 00:12:34 well that is *two* machines now... 2017-01-20 00:13:07 it already did that, I saw it scroll by as a post-install trigger 2017-01-20 00:13:16 I guess I can try again... time to get the iso 2017-01-20 06:12:07 one thing that sorely hurts me in alpine linux is the lack of multilib - specifically, for wine 2017-01-20 06:12:49 are there any plans to look into that? 2017-01-20 06:28:09 apk right now isnt really multilib aware 2017-01-20 06:28:25 asie, we have no immediate plans for multilib. it's usually more pain than worth. you can always install 32-bit chroot. 2017-01-20 06:29:01 i think apk does not need to be aware even. but abuild would need to generate the so:* virtual names differently... 2017-01-20 06:29:10 in any case, the toolchain is not multilib enabled either 2017-01-20 07:13:56 mhm 2017-01-20 09:28:39 hi, yesterday I made a fix for perl-unicode-collate -doc subpackage that was in conflict with perl-doc. Now I tried to install all perl-*-doc subpackages and found 18 more with the same problem. Do I need make patches for each one or make one big patch to fix them all? 2017-01-20 11:01:27 vakartel, I think it would be acceptable one single patch 2017-01-20 11:05:27 <^7heo> dicsovering new ops on alpine-* ever day. 2017-01-20 11:05:38 <^7heo> I'd wage bets on who I'm gonna discover next... 2017-01-20 11:14:34 anyone here have issues with eudev not wanting to start from runlevel? 2017-01-20 11:15:32 <^7heo> I didn't update for quite some time, as I was afraid of the behavior of nlplug-findfs 2017-01-20 11:15:36 <^7heo> so I can't know on that one 2017-01-20 12:17:57 ^7heo: fixed it by rc-update -u 2017-01-20 12:19:59 <^7heo> clandmeter: :) 2017-01-20 12:20:06 <^7heo> clandmeter: makes sense. 2017-01-20 12:45:58 to speed-up adding php7.1, I want to propose to not touch existing community/php7 but add new testing/php71. What do you think about it? 2017-01-20 12:49:16 propose better to build znc with iconv support to deal with different irc encodings on-the-fly 2017-01-20 13:59:54 vakartel: i would prefer to only have one php7 2017-01-20 14:00:08 i think both php5 and php7 is much already 2017-01-20 14:00:37 i suppose we can just upgrade php7 to 7.1 in edge 2017-01-20 14:01:35 I prefer it too, but 7.0 is current stable and 7.0 and 7.1 have some incompatibility 2017-01-20 14:01:57 we keep the 7.0 in stable 2017-01-20 14:02:07 and only upgrade to 7.1 in edge 2017-01-20 14:02:31 then users can chose to stay on stable (7.0) or live on the edge with the latest 2017-01-20 14:05:41 what if two php7 would live in different repositories 7.0 in community and 7.1 in testing ? 2017-01-20 14:06:08 or 7.0 in main and 7.1 in community )) 2017-01-20 14:08:46 wooh, finally i managed to compile wmic 2017-01-20 14:44:01 vakartel: i am afraid of increased maintenance work 2017-01-20 14:44:21 we already need patch and rebuild 2 different phps when there are security issues 2017-01-20 14:45:31 I made and now testing common apkbuild that builds php7.0 as php7 and php7.1 as php71 2017-01-20 14:46:35 in future I want to use it to made all three 5 7 71 in one apkbuild 2017-01-20 14:46:55 in future i dont want php5 or php7.0 2017-01-20 14:47:53 i think we should just keep two versions; latest and one considered LTS 2017-01-20 14:48:02 and/or the previous one if there's no LTS 2017-01-20 14:48:13 for postgresql, keeping two versions is necessary for DB upgrades 2017-01-20 14:48:31 if there are 3 different php versions, (5.x, 7.0, 7.1), 4 stable branches + edge, then we need to patch and rebuild php 15 times 2017-01-20 14:48:44 so for this need to change the line: _versions="5 7 71" to _versions=7 2017-01-20 14:49:07 9 times :) 2017-01-20 14:49:42 it wouldnt be any problem if php was the only application we support 2017-01-20 14:49:48 we have same issue with kernels 2017-01-20 14:50:48 problem is that support work increases exponential when you add supported versions 2017-01-20 14:51:29 not to mention, what do you want to do with php7-* modules? 2017-01-20 14:52:00 adding supported version increases the workload significantly 2017-01-20 14:54:13 I try to not distract you, but for last 1.5months near every day I receive emails like 'please upgrade php7 to 7.1' :) 2017-01-20 14:55:09 that kind of emails will not reduce if add more supported version 2017-01-20 14:55:13 you will get more of similar emails 2017-01-20 14:55:51 "please upgrade php7.1 to 7.1.x+1" 2017-01-20 14:56:00 )) 2017-01-20 14:56:01 "please upgrade php7.0 to 7.0.x+1" 2017-01-20 14:56:15 "please upgrade php5 ...." 2017-01-20 14:56:26 "please add php5.4" ... 2017-01-20 14:57:12 what do you do when php 7.2 comes out? 2017-01-20 14:57:15 I really want to reduce handwork with this apkbuild. 2017-01-20 14:57:42 ncopa: you can't just remove the version instantly, there must be two versions 2017-01-20 14:57:49 otherwise you get a catch-22 2017-01-20 14:57:54 such as 7.0 and 7.1, 7.1 and 7.2, etc. 2017-01-20 14:58:06 5.x is a bit of a special case, i'm not sure how much code is actually /stuck/ on 5.x 2017-01-20 14:58:11 (like python 2 code doesn't work at all on python 3 in general) 2017-01-20 14:58:26 up distro version, up version of php 2017-01-20 14:58:40 only 1 php required 2017-01-20 14:58:45 eleksir: thats what i want 2017-01-20 14:58:47 basicly - change _versions line and add one more archive to sources. )) 2017-01-20 14:58:49 what about postgresql? 2017-01-20 14:58:55 and php5 is exception 2017-01-20 14:58:56 which literally requires both the old and new version to upgrade a db 2017-01-20 14:59:02 (it checks the binaries) 2017-01-20 14:59:12 (and runs them) 2017-01-20 14:59:15 asie: the idea was to have a special postgresql-previous package 2017-01-20 14:59:28 i would just use postgresql[number], and alias postgresql to the latest 2017-01-20 14:59:42 just dump db and restore it on new psql 2017-01-20 14:59:54 psql never been easy 2017-01-20 14:59:57 eleksir: that's not how the official upgrade tool works, and what if you want to dump everything, not just the db? 2017-01-20 15:00:02 like, all the data in one swoop 2017-01-20 15:00:02 *easy to upgrade 2017-01-20 15:00:05 the upgrade tool makes it easy 2017-01-20 15:00:12 ncopa: the thing is 2017-01-20 15:00:20 if someone is dependent on someone else to update their PHP code 2017-01-20 15:00:27 they need to delay a distro version upgrade until everything is done 2017-01-20 15:00:40 right? 2017-01-20 15:00:50 this is why you keep two versions - to provide a smooth transition experience 2017-01-20 15:00:59 and not "everything is incompatible now" 2017-01-20 15:01:03 "i need to panic" 2017-01-20 15:01:07 you cannot devel for nonexistant version of php, right? 2017-01-20 15:01:21 do you really think everyone writes all their php code themselves 2017-01-20 15:01:35 so if new version pops up you adopt ypur code and delay will appear 2017-01-20 15:01:40 "your code" 2017-01-20 15:01:48 most php scripts in the wild are not "your" 2017-01-20 15:01:57 this is anti-user 2017-01-20 15:02:29 there's a reason php developers themselves maintain all these versions 2017-01-20 15:02:37 when you run some app - it already yours, ruby|php especially 2017-01-20 15:02:53 eh, i can only imagine some newbie trying to update a forum script by hand 2017-01-20 15:03:29 dev stend nobody cancels - keep you edge there, and when it will be ready push it to prod 2017-01-20 15:04:24 we did have postgresql-old9.x packages 2017-01-20 15:04:29 novice should be mature enough to fuckup prod app, righ? 2017-01-20 15:04:35 which had the upgrade tools only 2017-01-20 15:05:09 seems they are in unmaintained now 2017-01-20 15:05:18 so the only way to upgrade is dump/restore 2017-01-20 15:05:20 eh, we only really need N and N+1 2017-01-20 15:05:33 N-1 2017-01-20 15:05:40 depends on your frame of reference, but yes 2017-01-20 15:05:58 current and previous 2017-01-20 15:06:15 so we could have postgresql-previous 2017-01-20 15:06:24 the problem is that both needs to be installed at the same time 2017-01-20 15:06:42 that kind of problem, right 2017-01-20 15:06:59 well, postgresql-previous makes it easier 2017-01-20 15:07:15 http://git.alpinelinux.org/cgit/aports/tree/unmaintained/postgresql-old9.3/APKBUILD 2017-01-20 15:07:24 i just dump db in that case, btw time to upgrade, though :( another downtime 2017-01-20 15:07:41 dump/restore is the safest way 2017-01-20 15:07:50 recommended way 2017-01-20 15:08:18 but if you have alot data you may need take down the service for significant time 2017-01-20 15:08:26 yup 2017-01-20 15:08:55 our billing system has 60G+ psql database and every upgrade is pain 2017-01-20 15:09:04 i suppose we could do postgresql-previous 2017-01-20 15:09:44 which installs it in /usr/lib/postgersql-$version/ or similar 2017-01-20 15:09:52 so they both can be installed in parallel 2017-01-20 15:12:55 it should use default location of previous server database data folder, then 2017-01-20 15:14:17 ah another pain :( 2017-01-20 15:16:50 kernel 4.9 will be the next longterm 2017-01-20 15:16:59 https://plus.google.com/+gregkroahhartman/posts/dBWUjQipE72 2017-01-20 15:17:56 and what about grsec? they are very closed about co-operation :( 2017-01-20 15:18:28 we'll update the kernel as soon there is a free 4.9 patch available 2017-01-20 15:18:41 what do the grsec devs think about alpine anyway 2017-01-20 15:18:49 grsecurity patch is still 4.8.17 2017-01-20 15:18:58 asie: i dont think they care much 2017-01-20 15:19:36 they asked me change "grsecurity" to "unofficial grsecurity port" or similar on our about page 2017-01-20 15:19:58 they dont recommend using unofficial ports 2017-01-20 15:20:14 i have asked if we could get a discount 2017-01-20 15:20:27 asked how much it would cost if we bought stable patch 2017-01-20 15:20:31 they are a bit on asshole - they playing gpl bad way 2017-01-20 15:20:37 *of 2017-01-20 15:20:53 but the response was that will not be possible because the price is volume based 2017-01-20 15:21:12 yes, its a bit in the grey area with gpl license 2017-01-20 15:21:32 FSF, where are you? 2017-01-20 15:21:36 "the price is volume based" 2017-01-20 15:22:41 he also thought it woudl not be good for alpine if we had a sponsor paying for it since we'd end up depending on that sponsor 2017-01-20 15:22:45 basically 2017-01-20 15:22:52 that is correct 2017-01-20 15:22:53 he didnt want alpine to have the stable patch 2017-01-20 15:22:54 yup, they have price - pay it and you'll be serviced :) 2017-01-20 15:23:09 yes, i'm not surprised he doesn't want you to - everyone would just grab the patch from alpine 2017-01-20 15:23:28 well they could 2017-01-20 15:23:43 yeah, that is probably what they are afraid of 2017-01-20 15:23:49 yes, and the whole point of grsec's current system is to prevent circulation of stable patches outside paying customers 2017-01-20 15:23:57 they can't do it with a license, so they do it in other ways 2017-01-20 15:24:04 yes 2017-01-20 15:24:30 even then, assuming alpine linux would pay the volume cost 2017-01-20 15:24:35 everyone else could still take the patches 2017-01-20 15:24:40 yes 2017-01-20 17:32:21 Hello gentlemen. I am having some difficulties when cross-building some packages. They need things like .h files and libs, but do not find them. I am not sure they are being installed in the sysroot. Any ideas? 2017-01-20 17:35:12 I have already built the packages that contain those files btw. It's packages like libressl and ncurses 2017-01-20 17:36:50 .h files are usually in -dev subpackages 2017-01-20 17:37:23 jirutka, yes, but the abuild command is supposed to install dependencies, right? 2017-01-20 17:37:36 if they are specified, yes 2017-01-20 17:38:24 jirutka, is there a way for me to install those packages into the sysroot just so I can see if it is being done correctly? 2017-01-20 17:40:11 or I am just not making sense here? :) 2017-01-20 17:40:20 yes, as any other package 2017-01-20 17:40:56 for example, there’s ncurses package and ncruses-dev, that contains .h files etc. 2017-01-20 17:41:11 yes 2017-01-20 17:41:26 also you can use https://pkgs.alpinelinux.org/contents to find which package contains specific files 2017-01-20 17:41:33 it supports wildcards in search 2017-01-20 17:43:57 build dependencies are specified in `makedepends` variable in APKBUILD, it’s basically a superset of `depends`; when you run `apkbuild -r`, it installs both runtime and build time dependencies, so it’s probably some bug when something needed is missing 2017-01-20 17:45:21 jirutka, I don't think it's the package's problem. It's git. I think it might be something wrong in my environment 2017-01-20 17:46:55 jirutka, how do I `apk add` into the sysroot? 2017-01-20 17:47:12 option --root 2017-01-20 17:52:34 ok, thanks 2017-01-20 20:18:15 out of curiosity is there a way to provide abuild the gzip program to use for compression? 2017-01-20 20:21:38 <_ikke_> mitchty: looks like it's hardcoded: http://git.alpinelinux.org/cgit/abuild/tree/abuild.in#n1347 2017-01-20 20:22:26 _ikke_: ah, boo, well i could sed -i it :) 2017-01-20 21:15:29 mitchty: you’d like to use that alternative gzip with parallelization? 2017-01-20 21:16:05 jirutka: yep, would be easy to make it a variable defaulted to gzip 2017-01-20 21:16:22 <_ikke_> pigz 2017-01-20 21:16:27 <_ikke_> ? 2017-01-20 21:16:57 _ikke_: indeed 2017-01-20 21:42:38 ncopa: are you around? 2017-01-20 22:10:22 kaniini: linux 4.9 became LTS version https://lkml.org/lkml/2017/1/19/339 2017-01-20 23:25:22 good evening. topics about the wiki operate on this channel? 2017-01-20 23:29:27 sure 2017-01-20 23:29:34 but you can also ask in #alpine-linux 2017-01-20 23:39:20 How i can help? 2017-01-20 23:39:41 unnamed: you mean help with our wiki? 2017-01-20 23:40:32 jirutka: hi! yes, help with wiki. 2017-01-20 23:41:25 unnamed: well, it depends on your knowledge of Alpine; there are a lot of outdated articles and missing information on wiki 2017-01-20 23:42:27 so improvement of this is highly appreciate 2017-01-20 23:52:41 My knowledge is low, you can consider me a newbie. But i can help with updates, litle updates, and with translations. 2017-01-20 23:53:32 I am avaliable. 2017-01-20 23:54:17 or help the writer with the little things 2017-01-20 23:54:45 this is the best way to learn 2017-01-20 23:57:17 great! 2017-01-21 00:00:25 I think that the best may be to start with tutorials and howtos on wiki, try to follow them and when something doesn’t work, i.e. something has been changed since the article was written, try to fix it and/or ask us here why it doesn’t work as written in article 2017-01-21 00:01:03 or if something is not clear enough 2017-01-21 00:01:05 Thanks :D i will do my best. 2017-01-21 00:02:06 we are sometimes blind to some details that may be stoppers for newbies, so the best to fix these missing pieces is when some newbie try it and let us know when something is unclear 2017-01-21 00:03:36 understood. 2017-01-21 00:04:01 i will finish my new fresh install and create my account. 2017-01-21 04:06:49 ACTION $ find #alpine-devel -user "running alpine+zfs+luks"  2017-01-22 22:00:38 ok guess some of the base packages changed versions in a minor release between 8.0.1 and 8.0.2, i'm kinda sick of debugging ghc builds, so i updated my ghc pr to be 8.0.2, if stack doesn't build on armhf, no big deal, cabal should be ok for now 2017-01-22 22:02:07 so whenever anyone gets a chance to run the snapshot function with docker to regenerate the boostrap compilers for armhf and x86_64 then I can update ghc-bootstrap and (hopefully) get ghc into testing at least 2017-01-23 05:54:32 ncopa: you around? 2017-01-23 09:00:25 fabled, i'm upgrading v4l-utils at latest version 2017-01-23 09:00:52 i had issue in compiling with TEMP_FAILURE_RETRY and strndupa() function that is not defined 2017-01-23 09:00:59 http://tpaste.us/7BXO 2017-01-23 09:01:06 would this work? ^ 2017-01-23 09:01:12 Im' thinking that is not clean 2017-01-23 09:01:45 since strndupa function should be avoided, apparently: 2017-01-23 09:01:46 https://gitweb.gentoo.org/proj/musl.git/commit/?id=b880a39368148d7f7e2906a6d07d73fe606f7c8a 2017-01-23 09:02:04 but i don't know how to properly fix in C 2017-01-23 09:02:33 i saw that those issues are frequent with musl, and found patches that i adapted to v4l-utils 2017-01-23 09:02:44 but i want to be sure that is not garbage, and that does not create sec issues 2017-01-23 11:20:22 kaniini: im here 2017-01-23 11:21:40 ncopa: could you please look at https://github.com/alpinelinux/aports/pull/765#issuecomment-274050831 ? 2017-01-23 11:57:04 jirutka: i'm kind of ok with that PR 2017-01-23 11:57:16 i dont think it will break anything existing 2017-01-23 15:25:37 ncopa: well, but how are existing configs created? are these just defaults or customized? 2017-01-23 15:27:12 configs for what? 2017-01-23 15:27:15 for packages? 2017-01-23 15:34:45 no, kernel config 2017-01-23 15:36:07 well, they were originally made by hand 2017-01-23 15:36:17 things that i thought could be useful for general purpose 2017-01-23 15:36:34 not all things makes sense to enable on all archs though 2017-01-23 15:37:13 i suppose defaultconfig is a good start 2017-01-23 15:37:18 hm, so it’d be better to run make menuconfig with eixisting config and update it for power8? 2017-01-23 15:37:29 possibly 2017-01-23 15:37:49 you might need cross compiler or power8 system for that though 2017-01-23 15:38:07 we could tak the vanilla x86_64 config and do make oldconfig from that 2017-01-23 15:38:29 ncopa: maybe this can help us with kernel configs… https://github.com/crossdistro/kernel-tools#kconfig 2017-01-23 15:42:36 ncopa, when i run checkapk to check the abi change and i got: 2017-01-23 15:42:36 -usr/lib/libopus.so.0.5.3 2017-01-23 15:42:37 +usr/lib/libopus.so.0.5.4 2017-01-23 15:42:46 does it mean that the ABI changed? 2017-01-23 15:43:14 Or there are other kind of check i should do? 2017-01-23 15:45:54 fcolista: likely not 2017-01-23 15:46:07 ah.a 2017-01-23 15:46:12 check what the provides is 2017-01-23 15:46:24 provides=so:libopus.so.0 i guess 2017-01-23 15:47:38 readelf -d libopus.so.0.5.4| grep SONAME 2017-01-23 15:47:38 0x000000000000000e (SONAME) Library soname: [libopus.so.0] 2017-01-23 15:47:53 ^^^^ this kind of check would be enough? 2017-01-23 15:47:58 yes 2017-01-23 15:48:03 k 2017-01-23 15:48:19 libopus.so.0 -> libopbus.so.1 means abi breakage 2017-01-23 15:48:20 I thought that checkapk did that 2017-01-23 15:48:33 if its the same as old then evereything is ok 2017-01-23 15:48:37 apk will only diff the packages 2017-01-23 15:49:25 so, what i should do before upgrading a package is 2017-01-23 15:49:42 1. readelf the actual so 2017-01-23 15:49:46 2. upgrade the package 2017-01-23 15:49:59 3. re-readelf the new version and compare it 2017-01-23 15:51:08 apk info -a should show the provides 2017-01-23 15:51:16 wit the same info 2017-01-23 15:51:26 i though checkapk would show the soname 2017-01-23 15:51:29 maybe it doesnt 2017-01-23 15:52:31 ok good 2017-01-23 15:52:31 so 2017-01-23 15:52:32 apk info -a opus | grep -A1 provid 2017-01-23 15:52:32 opus-1.1.3-r0 provides: 2017-01-23 15:52:32 so:libopus.so.0=0.5.3 2017-01-23 15:52:44 upgrade opus 2017-01-23 15:52:54 readelf the provides 2017-01-23 15:53:07 those are basically the steps 2017-01-23 16:01:20 another question: how do i know what's the repository the package comes from? 2017-01-23 16:01:32 if it's main/community or testing? 2017-01-23 16:09:10 i suppose i should use pkgs.alpinelinux.org for that purpose 2017-01-23 16:13:05 clandmeter, you around? 2017-01-23 16:20:49 Hello. Why is hibernation disabled in the vanilla kernel? I can understand that hibernation is disabled in the grsec kernel considering it isn't compatible with GRKERNSEC_KMEM and PAX_MEMORY_SANITIZE etc. See http://git.alpinelinux.org/cgit/aports/tree/main/linux-vanilla/config-vanilla.x86_64?h=3.5-stable#n551 -> # CONFIG_HIBERNATION is not set 2017-01-23 16:28:18 triple: we try keep them close 2017-01-23 16:28:44 feel free to post an issue on bugs.a.o 2017-01-23 16:28:53 might make sense to enable it on vanilla kernel 2017-01-23 16:35:30 where can i get the json response from pkgs.alpinelinux.org ? 2017-01-23 16:38:31 http://i.imgur.com/afttFoM.jpg 2017-01-23 16:38:51 :D 2017-01-23 17:22:22 umh 2017-01-23 17:22:23 curl https://api.alpinelinux.org/packages | jq . 2017-01-23 17:22:23 % Total % Received % Xferd Average Speed Time Time Time Current 2017-01-23 17:22:23 Dload Upload Total Spent Left Speed 2017-01-23 17:22:23 curl: (51) SSL: no alternative certificate subject name matches target host name 'api.alpinelinux.org' 2017-01-23 17:22:23 0 0 0 0 0 0 0 0 --:--:-- --:--:-- --:--:-- 0 2017-01-23 17:22:36 api.alpinelinux.org it's an alias for pkgs.alpinelinux.org 2017-01-23 17:22:53 api.alpinelinux.org. 2854 IN CNAME pkgs.alpinelinux.org. 2017-01-23 17:22:53 pkgs.alpinelinux.org. 2854 IN A 88.159.20.184 2017-01-23 17:23:12 certificate should be updated...and in turn, the api fixed 2017-01-23 17:57:34 ncopa: OK, fair enough. Thanks. 2017-01-23 18:05:16 fcolista: where did you get api.alpinelinux.org, because i don't think that is something that actually works 2017-01-23 18:05:29 » curl -k https://api.alpinelinux.org/packages 2017-01-23 18:05:32 404 Not Found 2017-01-23 18:07:31 Shiz, without -k 2017-01-23 18:07:45 -k just skips tls cert validation 2017-01-23 18:08:49 yes, i saw. Anyway it's not working, that's my point 2017-01-23 18:09:06 https://github.com/insteps/aport-api/blob/master/docs/api-json.md 2017-01-23 18:09:22 it wouldn't be working even if it had a valid cert 2017-01-23 18:09:26 which is /my/ point 2017-01-23 18:09:35 I don't care about the cert 2017-01-23 19:08:48 I just happened to stumble into this same area. the aport-api project is referring to api.alpinelinux.org and the various endpoints under it. is that no longer supported/working? 2017-01-23 21:46:57 jirutka: what was that software that checked versions of things you were using for the alpine linux repos? 2017-01-23 21:48:25 mitchty: anitya 2017-01-23 21:48:35 mitchty: https://release-monitoring.org/distro/Alpine/ 2017-01-23 21:48:40 but 2017-01-23 21:48:51 how it works actually? 2017-01-23 21:49:02 does it have XPath for each homepage? 2017-01-23 21:49:05 jirutka: 2017-01-23 21:49:51 mitchty: why the past tense? it’s currently used in Aports ;) 2017-01-23 21:50:01 skrzyp: worse, regexps… 2017-01-23 21:50:23 *facepalm* 2017-01-23 21:50:24 skrzyp: take a look https://github.com/fedora-infra/anitya 2017-01-23 21:50:51 jirutka: I see… 2017-01-23 21:50:59 woohoo 2017-01-23 21:51:01 kill me please 2017-01-23 21:51:03 it’s insane, but it quite makes sense, b/c it’s the most universal… and some projects have really idiotic pages 2017-01-23 21:51:12 so XPath would not be usable for all 2017-01-23 21:51:18 mhm 2017-01-23 21:51:24 right 2017-01-23 21:51:30 however, it works quite good 2017-01-23 21:51:40 scadu: jirutka danke, that was what i was looking for, might have a need for it at work :) 2017-01-23 21:51:43 currently we have about 60 % of pkgs from main and community mapped to Anitya 2017-01-23 21:52:14 Anitya provides two APIs – fedmsg (Fedora message bus, based on ZeroMQ) and RESTful API 2017-01-23 21:52:46 we use fedmsg for push checks, https://github.com/alpinelinux/aports-turbo/blob/master/tools/anitya-watch.lua 2017-01-23 21:53:15 and RESTful API for one-time check of all pkgs https://github.com/alpinelinux/aports-turbo/blob/master/tools/anitya-check-all.lua 2017-01-23 21:53:54 ah cool, that will help immensely, idling in this channel pays off in more ways than one :) 2017-01-23 21:54:38 Alpine is currently 2nd in number of mapped packages :) https://release-monitoring.org/distro/Alpine/ 2017-01-23 21:55:08 restlol api 2017-01-23 21:55:17 restless api* 2017-01-23 22:59:24 jirutka: wasn't release-monitoring.org originally for Fedora? 2017-01-23 22:59:42 pavlix: dunno, but currently it’s open for every distro 2017-01-23 23:00:04 pavlix: that’s nice from Fedora guys btw; Gentoo has some own solution that is not open for everyone IIRC 2017-01-23 23:01:08 jirutka: yep, Gentoo is probably a bit more focused in-house than Fedora 2017-01-23 23:01:31 jirutka: when we're talking about infrastructure stuff 2017-01-24 03:17:26 ACTION twitch 2017-01-24 03:17:54 well when the ibm peeps come back, the real answer is --enable-decimal-float=no 2017-01-24 03:18:11 not whatever that -mno-long-double-128 stuff is :p 2017-01-24 06:15:39 <^7heo> moin here 2017-01-24 06:17:54 hi 2017-01-24 06:18:03 fabled: can i pick your brain on something 2017-01-24 06:22:32 kaniini, yes? 2017-01-24 06:29:11 <^7heo> moin 2017-01-24 06:31:38 fabled: i'm having a strange experience crossbuilding the base system for powerpc 2017-01-24 06:32:10 fabled: gcc is passing -pie to the linker, but using the non-PIE crtbegin.o and so on files 2017-01-24 06:34:02 fabled: tracing it deeper, i note that the -dumpspecs output shows a correct output, which means the specfile expansion is not seeing that it is -pie 2017-01-24 06:34:58 hmm 2017-01-24 06:35:01 raccoon:~$ powerpc-alpine-linux-musl-gcc -pie -o test3 test.c 2017-01-24 06:35:03 raccoon:~$ scanelf -t ./test3 2017-01-24 06:35:05 TYPE TEXTREL FILE 2017-01-24 06:35:07 ET_DYN - ./test3 2017-01-24 06:35:13 if i put -pie on commandline, textrel is gone 2017-01-24 06:36:25 perhaps --enable-default-pie is incomplete in someway 2017-01-24 06:40:20 it might be yes 2017-01-24 06:40:24 i think this 207-static-pie.patch is wrong 2017-01-24 06:40:37 it's very recent addition, and maybe they messed it up on some arch 2017-01-24 06:40:48 i am going to remove it and see if my toolchain is correct 2017-01-24 06:40:53 (locally) 2017-01-24 06:41:49 kaniini, 207's idea is to allow pie + static defined together, and to then imply -Bsymbolic 2017-01-24 06:42:05 yes, but it touches other aspects of the spec string 2017-01-24 06:42:25 the modification to the spec string may be wrong :) 2017-01-24 06:42:30 it also requires rcrt1 as startup file 2017-01-24 06:42:34 ACTION nods 2017-01-24 06:42:46 it works on x86_64 :) 2017-01-24 06:43:00 but if there's a bug in it, patches for the patch are welcome 2017-01-24 06:43:11 i am thinking it may be broken for cross 2017-01-24 06:43:22 in some way 2017-01-24 06:43:33 it's possible 2017-01-24 06:43:43 it's also possible enable-default-pie is broken for cross 2017-01-24 06:43:54 the thing is that gcc tries to build itself non-pie iirc 2017-01-24 06:44:32 most of the enable-default-pie commits were adding no-pie flags to self build 2017-01-24 06:45:14 the problem is that the spec string expansion is wrong 2017-01-24 06:45:35 so as an initial thesis, disable the thing which changes the expansion ;) 2017-01-24 06:50:52 yeah 2017-01-24 06:51:09 can you see from dumpspecs on what goes wrong? 2017-01-24 06:52:09 the -dumpspecs output produces a specstring that looks reasonable (but with the rcrt1.o stuff added) 2017-01-24 06:52:22 unfortunately right now, gcc is rebuilding 2017-01-24 07:00:09 kaniini, looks like the initial commit for enable-default-pie does not modify specs of powerpc config 2017-01-24 07:02:56 indeed. textrels remain after removing that patch 2017-01-24 07:03:03 fabled: can you link me to the commit? 2017-01-24 07:03:18 hum. what's the ppc config name in gcc? 2017-01-24 07:03:31 commit e9abca4f4a48fa8b1fd9778f6a3cd748e099e3bb 2017-01-24 07:03:37 rs6000 2017-01-24 07:03:56 e.g in https://github.com/gcc-mirror/gcc/commit/e9abca4f4a48fa8b1fd9778f6a3cd748e099e3bb 2017-01-24 07:04:00 oh 2017-01-24 07:04:03 rs6000 is there 2017-01-24 07:06:52 rs/6000 is the pre-ppc 2017-01-24 07:07:36 skrzyp: yes, well, that is what gcc calls all ppc (even ppc64) 2017-01-24 07:07:46 take it up with them :D 2017-01-24 07:07:52 fabled: http://sprunge.us/CjMR 2017-01-24 07:08:16 http://sprunge.us/hcOK 2017-01-24 07:08:18 :( 2017-01-24 07:08:19 kaniini: hurr durr arm == arm64 2017-01-24 07:08:31 :D 2017-01-24 07:10:17 ACTION is sad 2017-01-24 07:13:28 like 2017-01-24 07:13:33 i don't even understand 2017-01-24 07:14:11 -dumpspecs http://sprunge.us/RDej 2017-01-24 07:20:45 its very strange 2017-01-24 07:20:49 if i do -no-pie 2017-01-24 07:20:52 then it works 2017-01-24 07:20:54 if i do -pie 2017-01-24 07:20:56 then it works 2017-01-24 07:21:02 if i do nothing, i get half-pie 2017-01-24 07:21:24 i wonder how this --enable-default-pie works 2017-01-24 07:24:40 i also noticed that -pie is different from -fpie 2017-01-24 07:27:08 only difference seems to be the crt filenames 2017-01-24 07:29:05 indeed, -pie needs to link against Scrt1.o and so on 2017-01-24 07:29:22 without -pie we get linker flag -pie + -no-pie .o's 2017-01-24 07:29:38 i don't understand how the spec expansion works 2017-01-24 07:30:39 kaniini, -dumpspecs ? 2017-01-24 07:34:57 http://sprunge.us/RDej 2017-01-24 07:35:34 %{!shared: %{pg|p|profile:gcrt1.o%s;pie:Scrt1.o%s;:crt1.o%s}} %{mnewlib:ecrti.o%s;:crti.o%s} %{static:crtbeginT.o%s;shared|pie:crtbeginS.o%s;:crtbegin.o%s} 2017-01-24 07:35:57 which expands to 2017-01-24 07:36:05 crt1.o crtbegin.o 2017-01-24 07:42:59 kaniini, ah 2017-01-24 07:43:01 it's gcc bug 2017-01-24 07:43:15 gcc uses 'endfile' spec and similar to construct those names normally 2017-01-24 07:43:20 however rs6000 specs override it 2017-01-24 07:43:30 and the way it's overridden is not default-pie compatible 2017-01-24 07:43:50 i think 2017-01-24 07:44:46 hmm 2017-01-24 07:45:27 or are the .o files also non-pie? 2017-01-24 07:46:15 Scrt1.o is pie; crt1.o is not; generated .o files are pie 2017-01-24 07:46:48 raccoon:~$ powerpc-alpine-linux-musl-gcc -o test.o -c test.c 2017-01-24 07:46:50 raccoon:~$ scanelf -t ./test.o 2017-01-24 07:46:52 TYPE TEXTREL FILE 2017-01-24 07:46:54 ET_REL - ./test.o 2017-01-24 07:49:11 it's using the rs6000/sysv4.h instead of gnu-user.h definitions 2017-01-24 07:50:39 aha 2017-01-24 07:50:42 i wonder how to fix this 2017-01-24 07:53:08 possibly something like http://sprunge.us/eVfO but i have no idea if it works or not 2017-01-24 07:53:16 you are probably better off filing a bug against gcc 2017-01-24 07:53:53 the gnu spec works differently though 2017-01-24 07:54:06 it has: 2017-01-24 07:54:07 %{shared:crtendS.o%s;: %{" PIE_SPEC ":crtendS.o%s} \ 2017-01-24 07:54:07 %{" NO_PIE_SPEC ":crtend.o%s}} crtn.o%s \ 2017-01-24 07:54:45 see gcc/config/gnu-user.h search for GNU_USER_TARGET_STARTFILE_SPEC and GNU_USER_TARGET_ENDFILE_SPEC 2017-01-24 07:55:21 indeed 2017-01-24 07:55:26 i will file a gcc bug 2017-01-24 07:55:36 and also try that spec patch 2017-01-24 07:57:03 i have no idea if it even works, depends on what PIE_SPEC contains 2017-01-24 07:57:50 from what i can see, PIE_SPEC should be defined to something 2017-01-24 07:57:59 hopefully the right something 2017-01-24 07:59:00 btw, if you are doing powerpc commits you might include "refer #3596" in them... see http://bugs.alpinelinux.org/issues/3596 :) 2017-01-24 08:01:27 pass2 built 2017-01-24 08:01:38 bootstrap musl built 2017-01-24 08:09:06 xgcc: fatal error: braced spec 'static:crtbeginT.o%s;shared|no-pie|r|shared:;:crtbeginS.o%s;:crtbegin.o%s} ' is invalid at ':' 2017-01-24 08:09:09 damn 2017-01-24 08:11:38 however. 2017-01-24 08:15:39 i see what to do 2017-01-24 08:17:03 ok 2017-01-24 08:17:13 rinse, repeat 2017-01-24 08:17:26 i think it is possible to use the gnu-user ones with some modification 2017-01-24 08:17:36 or perhaps stock 2017-01-24 08:20:13 yes 2017-01-24 08:20:21 we can literally just use the specs in gnu-user.h 2017-01-24 11:04:22 <_errm_> I am working on a APKBUILD, and I am getting a bit confused about the correct paths for stuff is there a way to drop into a shell part way through a function? 2017-01-24 11:05:19 <^7heo> errm... 2017-01-24 11:05:33 <^7heo> you mean that you would like interactive shell debugging in the middle of your APKBUILD? 2017-01-24 11:06:54 if i do nothing, i get half-pie 2017-01-24 11:07:13 that's awesome, I would love to get half pie by doing nothing too 2017-01-24 11:08:46 _errm_: just spawn /bin/sh -i in your function at the point you want, but you won't get the parent shell's variables 2017-01-24 11:09:03 <^7heo> skarnet: ◔ 2017-01-24 11:09:10 <^7heo> skarnet: here, half pi. 2017-01-24 11:09:17 <^7heo> skarnet: you're welcome. 2017-01-24 11:12:07 you're the cos of all that. Zero. 2017-01-24 11:17:45 <^7heo> that's mean 2017-01-24 11:17:53 <^7heo> and was not deserved... 2017-01-24 11:17:55 <^7heo> :'( 2017-01-24 11:24:05 triggered by trig 2017-01-24 11:36:26 <_errm_> thanks skanet ... 2017-01-24 11:37:35 <_errm_> I guess any variables I wan’t to know about can be exported first ... 2017-01-24 11:39:54 <^7heo> skarnet: trigonometry? 2017-01-24 11:40:20 yes 2017-01-24 11:40:24 <^7heo> meh 2017-01-24 11:41:11 sorry, I couldn't find the right tone with cos/cause and triggered/trigonometry, I need to work on my pun skills 2017-01-24 11:42:32 <^7heo> not too bad. 2017-01-24 12:01:56 I am trying to config openvpn with ldap plugin "/usr/lib/openvpn-auth-ldap.so". However openvpn breaks down on its face with seg-fault. Logs say denied resource overstep by requesting 4096 for RLIMIT_CORE against limit 0 for /usr/sbin/openvpn .... bla bla bla ... 2017-01-24 12:02:22 ;( 2017-01-24 12:02:49 Any solution to this ? 2017-01-24 12:03:32 grharry: first step is report it on bugs.a.o 2017-01-24 12:03:36 which package is it? 2017-01-24 12:03:41 might be it only needs rebuild 2017-01-24 12:06:40 grharry: i just pushed a rebuild of openvpn-auth-ldap 2017-01-24 12:07:04 can you please try fetch it from rsync.a.o and test if it solves the issue? 2017-01-24 12:07:12 if it does i'll cherry-pick it for 3.5-stable too 2017-01-24 12:08:25 This message is printed when a program is crashed and try to write a coredump. grsecurity prevent this. 2017-01-24 12:09:16 <^7heo> vakartel: stop using GNU programs. 2017-01-24 12:10:22 <^7heo> vakartel: also, man ulimit. 2017-01-24 12:11:51 ^7heo: he was just commenting grharrys problem 2017-01-24 12:11:57 <^7heo> ah 2017-01-24 12:12:12 <^7heo> I ignore that user, so I didn't see anything. 2017-01-24 12:12:22 <^7heo> sorry about that. 2017-01-24 12:12:59 <^7heo> vakartel: ^ 2017-01-24 12:13:10 <^7heo> ncopa: thanks for the heads up :) 2017-01-24 12:25:47 ahhh!!! dsl problems here again ! 2017-01-24 12:25:58 grharry: please test the http://rsync.alpinelinux.org/alpine/v3.5/main/x86_64/openvpn-auth-ldap-2.0.3-r5.apk 2017-01-24 12:26:06 i rebuilt it against openvpn 2.4 2017-01-24 12:26:34 sorry 2017-01-24 12:26:37 thats wrong in thinkg 2017-01-24 12:26:40 should be edge 2017-01-24 12:27:57 yes! I am on openvpn-2.3.11-r0 and alpine 3.4.6 2017-01-24 12:29:45 hum 2017-01-24 12:29:46 ok 2017-01-24 12:34:18 rnalrd: hi, I'm the guy with the DOS-lineendings patch. Can you get the patch somehow past git? Or shall I search for a different solution? 2017-01-24 12:34:50 jirutka, hi There. It seems that most of the changes requested by https://github.com/alpinelinux/aports/pull/765 was already done by William. 2017-01-24 12:37:58 should I open another pull request with the missing parts, or, just refresh this PR? It seems that ncopa's review is not necessary anymore, since the kernel .config was already integrated. 2017-01-24 12:38:14 anyway, let me know how you want to handle it. 2017-01-24 13:10:24 Ganwell, I tried your suggestion but it didn't work 2017-01-24 13:11:08 I also tested with core.autocrlf and core.eol and I was able to build locally, but not on buildservers which have different settings 2017-01-24 13:11:51 rnalrd: I'd prefer to convert the header with DOS-lineendings to unix. Is it ok to do that in prepare()? 2017-01-24 13:12:04 sounds good to me 2017-01-24 13:15:02 Hi. When cross compiling linux-vanilla to ppc64le on a x86_64 machine I'm getting a "gcc: error: unrecognized command line option '-mlittle-endian'" when building some modules in arch/powerpc/boot. It seems the compiler used - x86_64 gcc - is not set correct. Any known issue when porting linux-vanilla to new archs? 2017-01-24 13:16:12 *correctly 2017-01-24 13:22:15 i used it just fine for arm and aarch64; though, i have no experience when endianess is different 2017-01-24 13:38:25 fabled: ah, ok. I see that CROSS_COMPILE is never used in APKBUILD in prepare(), menuconfig(), and build()... 2017-01-24 14:03:39 want to ask somebody to review a new aport testing/proftpd :: PR753 2017-01-24 14:04:52 vakartel, fwiw, we have proftpd in unmaintained 2017-01-24 14:06:04 oops 2017-01-24 14:07:23 <^7heo> well, maybe you can just git rm unmaintained/proftpd :P 2017-01-24 14:10:33 Do I need to make myself as a new maintainer or I need to copy old-one from unmaintained package? 2017-01-24 14:11:53 <^7heo> well, I guess we'll check the PR then. 2017-01-24 14:24:35 oops I have also missed packages in unmaintained...... 2017-01-24 16:37:23 lets see if the textrels are gone now 2017-01-24 16:37:48 yes, textrels gone 2017-01-24 16:46:14 so:libc.musl-ppc.so.1 2017-01-24 16:46:19 hmmm, what has happened here 2017-01-24 16:46:31 iirc we need it to be libc.musl-powerpc.so.1 2017-01-24 16:50:11 but i do have a working compiler now :) 2017-01-24 16:50:41 raccoon:~$ powerpc-alpine-linux-musl-gcc -o test test.c 2017-01-24 16:50:43 raccoon:~$ scanelf -t ./test 2017-01-24 16:50:45 TYPE TEXTREL FILE 2017-01-24 16:50:47 ET_DYN - ./test 2017-01-24 16:50:49 so thats good 2017-01-24 17:00:59 hmm, bootstrap failed on building ppc gcc binary 2017-01-24 18:00:58 seems something is broken in final cross in gcc APKBUILD these days. i am tracing through it 2017-01-24 18:11:54 >>> gcc*: python gdb extension 2017-01-24 18:11:56 >>> gcc*: paxmark cc1 2017-01-24 18:11:58 >>> ERROR: gcc*: package failed 2017-01-24 18:12:00 hmm 2017-01-24 18:12:04 that makes sense 2017-01-24 18:17:41 indeed. paxctl reports that the cross-built binary is not a valid ELF 2017-01-24 18:17:49 which makes sense, given it's for a different architecture :) 2017-01-24 18:29:42 hmm, fails later. i wonder if my fix is wrong 2017-01-24 18:29:52 lets test the logic 2017-01-24 18:43:28 it was 2017-01-24 18:55:19 hmm 2017-01-24 18:55:32 >>> ERROR: g++*: gpp failed 2017-01-24 18:55:36 getting there :) 2017-01-24 18:56:28 ah another paxmark ;) 2017-01-24 18:59:31 that should get it done 2017-01-24 19:14:56 hmm 2017-01-24 19:28:46 kaniini, paxmark should work on cross-building 2017-01-24 19:29:28 kaniini, it uses paxctl which should be working on any arch ELF binaries. and attr which works on xattrs 2017-01-24 19:35:19 and it worked for me when i cross-built aarch64... 2017-01-24 19:35:32 i wonder if there's something different in the powerpc elf binaries 2017-01-24 19:43:14 paxctl it does not like 2017-01-24 19:43:30 fabled: maybe we disable paxctl on power 2017-01-24 19:44:58 hum. either fix it, or make it ignore power ELFs 2017-01-24 19:46:23 i'll look into it 2017-01-24 19:48:36 fabled: a thought is that perhaps it does not know about PIE with power elfs 2017-01-24 19:49:06 if you look paxctl code, it's pretty generic. it operates only on one generic ELF part. 2017-01-24 19:49:17 so either that header is not present, or there's something else funky in the ELF binary 2017-01-24 19:49:27 if it's not ELF (e.g. a.out) it'll fail 2017-01-24 19:50:25 what does scanelf or readelf -a say about that binary? 2017-01-24 19:50:39 ah, scanelf likes it 2017-01-24 19:50:41 readelf -a ? 2017-01-24 19:51:36 no errors 2017-01-24 19:52:12 http://sprunge.us/DeIM 2017-01-24 19:55:57 if (state->size < sizeof(Elf64_Ehdr) || (!is_elf32(state) && !is_elf64(state))) { 2017-01-24 19:56:02 this check may be wrong for power 2017-01-24 19:57:55 is_elf32/is_elf64 branch is wrong 2017-01-24 19:59:29 but where is is_elf32 from 2017-01-24 20:02:51 if (ehdr->e_ehsize != sizeof(ElfW(Elf, _Ehdr))) { 2017-01-24 20:02:52 bingo 2017-01-24 20:03:36 ehdr->e_ehsize (13312) != (52) 2017-01-24 20:03:37 endian 2017-01-24 20:03:40 :) 2017-01-24 20:04:00 fabled: paxctl assumes little-endian 2017-01-24 20:04:18 fabled: ppc is big-endian (except ppc64le) 2017-01-24 20:07:13 fabled: or, rather: paxctl does not work with cross built binaries when the endian does not match 2017-01-24 20:08:02 fabled: i think that we should not invoke paxmark when cross building because of this endianness issue -- paxmarking a binary cross-endian may damage it 2017-01-24 20:12:17 hum 2017-01-24 20:12:39 but it's built to cross-handle 32bit vs. 64bit 2017-01-24 20:12:59 yes, it can handle that. it seems to be byte ordering that it can't handle 2017-01-24 20:13:35 btw. have you seen https://github.com/alpinelinux/aports/pull/765#issuecomment-274921108 ? 2017-01-24 20:13:46 or breno works for you? 2017-01-24 20:14:31 i have seen it, but those guys are mainly interested in getting alpine going on power8 only (which is little-endian) 2017-01-24 20:14:37 ah, ok 2017-01-24 20:14:47 i'd still rather have paxctl work cross-endian 2017-01-24 20:14:56 we may want to cross-compile alpine to some archs 2017-01-24 20:15:07 i'm not sure how safe that is 2017-01-24 20:15:16 or, more accurately 2017-01-24 20:15:23 gcc can do it :) 2017-01-24 20:15:28 i am not sure how much work is inolved in making that safe 2017-01-24 20:15:30 yeah 2017-01-24 20:15:34 understand 2017-01-24 20:15:40 file a bug as minimum 2017-01-24 20:15:52 where does one file bugs for paxctl 2017-01-24 20:16:33 grsec forums ? 2017-01-24 20:16:44 then again 2017-01-24 20:16:50 i think we might want to drop it 2017-01-24 20:16:53 i will just send pageexec@freemail.hu 2017-01-24 20:16:55 a mail 2017-01-24 20:16:59 we are using xattrs now anyways 2017-01-24 20:17:06 hmm 2017-01-24 20:17:11 yes, that is also my thinking 2017-01-24 20:17:17 maybe on crossbuild, just do xattr only 2017-01-24 20:18:08 or just drop paxctl stuff in paxmark 2017-01-24 20:18:19 hmm 2017-01-24 20:18:21 it has: # Create PT_PAX marking (remove after migration) 2017-01-24 20:18:24 yes 2017-01-24 20:18:28 lets just do that 2017-01-24 20:18:40 i remember need that for a while 2017-01-24 20:18:48 lemme check when the kernels got updated 2017-01-24 20:18:48 and then i will back out the paxmark change to gcc 2017-01-24 20:19:34 april 2015 2017-01-24 20:19:43 yeah, i think our migration period is over... 2017-01-24 20:20:05 we even have now: +# CONFIG_PAX_PT_PAX_FLAGS is not set 2017-01-24 20:20:24 yep 2017-01-24 20:20:26 cool 2017-01-24 20:20:28 going to nuke it then 2017-01-24 20:20:31 thx! 2017-01-24 20:20:44 remove the dependency on paxctl too 2017-01-24 20:20:50 ppc64 non little-endian is broken on musl :( 2017-01-24 20:20:57 :/ 2017-01-24 20:21:08 but ppc32 is good 2017-01-24 20:21:38 it's broken for elfv1 2017-01-24 20:21:50 yes, elfv2 requires power8 2017-01-24 20:21:51 but --with-abi=elfv2 works 2017-01-24 20:22:01 which costs $$$$$$$$$$ 2017-01-24 20:22:10 ACTION wonders if ibm sponsored that port ;) 2017-01-24 20:23:03 fabled: is there any reason to keep paxctl around? 2017-01-24 20:23:13 fabled: all it seems to do is edit ELF files 2017-01-24 20:26:23 raccoon:~/aports/main/paxmark$ apk info --rdepends paxctl 2017-01-24 20:26:24 paxctl-0.9-r0 is required by: 2017-01-24 20:26:26 paxmark-0.8-r2 2017-01-24 20:26:30 i think it can go 2017-01-24 20:33:56 yeah, i think it can go; as long as no apkbuild depends on it 2017-01-24 20:33:59 even in makedepends 2017-01-24 20:34:39 seems it's not 2017-01-24 20:34:51 except paxmark, and one thing in unmaintained. so yeah, it can go 2017-01-24 20:35:26 yep i already looked :) 2017-01-24 20:36:32 thanks 2017-01-24 20:42:27 welp 2017-01-24 20:42:34 i have a base image 2017-01-24 20:42:45 decent start :) 2017-01-24 20:43:23 ppc/ppc64 (non power8) is still a useful thing for alpine 2017-01-24 20:43:34 because you can get an xserve off ebay for $100 2017-01-24 20:43:51 and it's still competitive to recent xeon servers 2017-01-24 22:13:44 <^7heo> gosh this is hard to read. 2017-01-25 09:28:38 i wonder if we should reject #6684 2017-01-25 09:29:11 the problem is that if we support install to disk without netowrk then we need to include the kernel package as well 2017-01-25 09:29:16 and the iso becomes bigger 2017-01-25 09:30:27 hm 2017-01-25 09:30:34 13MB 2017-01-25 09:42:03 oh this one is interesting too 2017-01-25 09:42:07 #6610 2017-01-25 09:42:18 we dont ship c99 wrapper 2017-01-25 09:42:30 in fact we dont ship any c compiler at all in the default setup 2017-01-25 09:43:43 i wonder if while at it, i should just remove the different issue priority alternatives, Low, Normal, Urgent, Immediate 2017-01-25 09:43:55 then every new issue will get priority "high" and eveneryone is happy 2017-01-25 09:46:06 lol 2017-01-25 09:46:49 why don't set only one "immediate" for the real immediate issues (like 0-days) and nothing else then? 2017-01-25 09:47:47 <^7heo> ncopa: isn't the iso being 100+MB already due to inclusion of all modules/firmwares that we might need? 2017-01-25 09:48:16 ^7heo: the virt kernel is much smaller 2017-01-25 09:48:20 <^7heo> ah yeah 2017-01-25 09:48:25 the alpine-virt image is only 25MB 2017-01-25 09:48:27 <^7heo> ncopa: how about hiding priority from people? 2017-01-25 09:48:41 <^7heo> ncopa: so nobody sees what prio it has, everybody happy. 2017-01-25 09:48:46 i tested, adding all needed to do install without netowrk will make the iso 41MB 2017-01-25 09:48:49 <^7heo> ncopa: and internally we see it when logged. 2017-01-25 09:48:59 that could work 2017-01-25 09:49:00 but 2017-01-25 09:49:13 i was kinda hopong to not spend my time on hacking redmine... 2017-01-25 09:49:34 <^7heo> oh, I wouldn't. 2017-01-25 09:49:50 <^7heo> I would just put a very ugly hack 2017-01-25 09:50:00 <^7heo> i.e. reverse proxy that filters the HTML on the fly. 2017-01-25 09:50:09 <^7heo> it's web afterall. 2017-01-25 09:50:17 <^7heo> you can't make it good, so why spend time on it anyway? 2017-01-25 09:50:31 <^7heo> I mean, sed can do the job 2017-01-25 09:50:38 :) 2017-01-25 09:52:41 <^7heo> if echo "$HTML" | grep "" | grep -vq "ncopa\|clandmeter\|..."; then 2017-01-25 09:52:59 <^7heo> echo "$HTML" | sed '//d' 2017-01-25 09:53:01 <^7heo> else 2017-01-25 09:53:03 <^7heo> echo "$HTML" 2017-01-25 09:53:04 <^7heo> fi 2017-01-25 09:56:47 <^7heo> I mean, all that has to be matched is '
Priority:
[^<]*
' 2017-01-25 09:57:40 <^7heo> just sed 's/
Priority:
[^<]*
//' and you're good :P 2017-01-25 10:03:44 ncopa, this spam: https://bugs.alpinelinux.org/issues/6727 can be deleted? 2017-01-25 10:03:58 yeah 2017-01-25 10:04:15 i deleted it 2017-01-25 10:04:18 k 2017-01-25 10:04:40 6724 too 2017-01-25 10:37:00 lol 2017-01-25 10:37:01 https://bugs.alpinelinux.org/issues/6722 2017-01-25 10:37:11 there are ppl who's still using floppy.. 2017-01-25 10:41:37 <_ikke_> There is still a company / site selling them 2017-01-25 10:42:03 <_ikke_> Try to get whatever reserve they can get their hands on 2017-01-25 11:48:37 Hi. Is there any issue on Alpine regarding static linkage?. It seems not possible to generate a static executable as usual , by passing -static to gcc. For instance: https://hastebin.com/raw/fuxawagajo 2017-01-25 11:49:35 did you try gcc then ld separately? 2017-01-25 11:56:58 gromero, we do static-pie by default 2017-01-25 11:57:11 file will say it's dynamically linked, even though it's really static 2017-01-25 11:58:57 use 'ldd' tool or 'readelf -d' to see what dynamic dependencies the binary has 2017-01-25 12:00:40 the benefit is that in alpine you get full ASLR working for static binaries :) 2017-01-25 12:06:23 fabled: thanks 2017-01-25 12:07:15 <^7heo> _ikke_: you mean REAL floppies? 2017-01-25 12:07:17 gromero, if it's more important to get 'file' show the right result, you could use '-no-pie -static' to generate regular executable 2017-01-25 12:07:28 but then ASLR is disable 2017-01-25 12:10:47 fabled: I see. At the moment ASLR is not important since I'm just performing some tests on our ppc64le port regarding the static libs, so I was looking a way to be sure if the static executable generated was really a static one. ta :) 2017-01-25 12:43:25 im splitting pu the openssh package a bit 2017-01-25 12:43:29 up* 2017-01-25 12:43:41 will introduce an openssh-server and openssh-keygen package 2017-01-25 12:43:53 so you can insstall the openssh server without pulling in the client 2017-01-25 12:57:38 ncopa, fabled, is it ok with you? http://sprunge.us/XTLj 2017-01-25 12:57:53 v4l-utils fails the build due to undefined strndupa() 2017-01-25 12:58:04 as well as TEMP_FAILURE_RETRY 2017-01-25 12:58:34 i wonder if strndupa() defined as such it's risky 2017-01-25 12:59:28 ah, it's for v4l-utils package... 2017-01-25 13:00:14 <_ikke_> ^7heo: I mean the smaller 3 1/2 " ones 2017-01-25 13:02:00 <^7heo> ah yeah 2017-01-25 15:04:11 ncopa: thanks for the lua-socket fix for #6603 2017-01-25 15:04:19 ACTION confirmed it works on 3.5 2017-01-25 18:11:16 ppc64le strace patch broke armhf :( 2017-01-25 18:17:41 architecture jenga 2017-01-25 18:18:15 m68k FPU libc optimization code for Atari Falcon broke x86_64's firefox 2017-01-25 18:18:47 no, the patch broke strace on armhf 2017-01-25 18:19:06 but it was for ppc 2017-01-25 18:19:07 but maybe not the patch itself but the rebuild 2017-01-25 18:19:13 i'm not sure :/ 2017-01-25 18:34:48 On strace patch for ppc64le, I intend to fix it soon. It's just a case of making the guard a PPC64-only change 2017-01-25 18:35:44 thanks 2017-01-25 18:35:50 np 2017-01-25 18:38:06 kaniini: you haven't caught thee joke ;< 2017-01-25 22:34:45 Hi. I'm getting a timeout on nlplug-findfs while on a usb device (on qemu). Is there a way to tell nlplug-findfs to ignore that device? 2017-01-25 22:37:21 this timeout simply breaks the init script after our kernel boots... btw, our current kernel is: 2017-01-25 22:37:23 Linux (none) 4.4.44 #2-Alpine SMP Wed Jan 25 18:59:36 GMT 2017 ppc64le Linux 2017-01-25 22:37:24 :) 2017-01-25 22:39:12 a paste https://hastebin.com/raw/relavusati , if it is of any help 2017-01-26 00:36:42 hey there 2017-01-26 00:37:11 what's the way to reinstall repo kernel ondisk from install media ? 2017-01-26 00:37:57 apk -p /mnt add linux-grsec is spitting in my face that it doesn't find files in /tmp and the likes ... 2017-01-26 00:39:43 yeah works better with /dev and /proc bindmounted in /mnt 2017-01-26 00:42:58 is there some flag to set from the boot menu to get some debug ? got a menu bootloop after a kernel install :( 2017-01-26 07:49:03 any idea what could make syslinux boot loop on menu when upgrading kernel on a soft raid1 setup ? 2017-01-26 07:50:28 even reinstalling the kernel from install didn't help, I've had to rerun setup-disk -m sys completely :( 2017-01-26 07:56:06 hi guys...any hint on that? http://tpaste.us/8Qo8 2017-01-26 07:56:16 i'm upgrading nghttp2 to 1.19.0 2017-01-26 07:56:21 got this error 2017-01-26 07:56:42 in the changelog it's stated: nghttpx: Enable SCT with OpenSSL 1.1.0 2017-01-26 07:56:59 dunno if it is related 2017-01-26 07:58:53 oh..actually that function is not declared :) 2017-01-26 08:59:05 Damn I have a second server that doesn't reboot after the installation on a raid1 2017-01-26 08:59:25 same procedure as the first one that gets boot loop when upgrading kernel :O 2017-01-26 09:00:06 does it matter if mbr.bin is dd'ed to the disk after the setup-disk step ? 2017-01-26 09:02:56 setup-disk doesnt take care of it? 2017-01-26 09:03:18 i never use raid installation on servers, always use usb to boot. 2017-01-26 09:11:48 clandmeter: apparently not ... 2017-01-26 09:13:11 on first server just doing the setup-disk part didn't make it boot 2017-01-26 09:13:46 I actually tested several variations of GPT/msdos boot partitions setup and the only one I could make boot was msdos 2017-01-26 09:14:08 apparently second server which is exactly the same config has decided to make my balls hurt 2017-01-26 09:15:09 while first one can't take a kernel upgrade though... 2017-01-26 09:15:23 I don't remember having it that difficult on previous installers :/ 2017-01-26 09:16:07 i couldnt use gpt at all with the installer last time i tried on my desktop 2017-01-26 09:16:31 clandmeter: well seems like a thing then 2017-01-26 09:18:48 clandmeter: though this syslinux menu boot loop on kernel update worries me more as that means I can't trust a kernel upgrade to not make the server unbootable:( 2017-01-26 09:21:45 i would ping ncopa cause he is that one who created those scripts. 2017-01-26 09:23:45 ncopa: ping ^ 2017-01-26 09:45:05 hi 2017-01-26 09:45:29 coredumb: what happened? 2017-01-26 09:45:31 morning ncopa 2017-01-26 09:45:39 ncopa: 2 issues 2017-01-26 09:45:57 apart not being able to make GPT disks on raid1 boot at all 2017-01-26 09:46:12 Got 3 identical servers dell r220 2017-01-26 09:46:26 two 2TB satas in there no HW raid 2017-01-26 09:46:42 do you use GPT on them? 2017-01-26 09:46:48 cannot 2017-01-26 09:46:53 ok 2017-01-26 09:46:55 msdos only is able to boot 2017-01-26 09:46:58 thats a known issue 2017-01-26 09:47:06 in legacy bios mode yes 2017-01-26 09:47:21 yeah it's in bios mode not uefi 2017-01-26 09:47:26 so first one booted 2017-01-26 09:47:51 kernel upgrade gave me an endless syslinux menu boot loop 2017-01-26 09:48:05 hum 2017-01-26 09:48:13 how was it installed? 2017-01-26 09:48:19 what version of alpine was installed 2017-01-26 09:48:28 I tried repairing redoing apk del/add in chroot from install CD 2017-01-26 09:48:35 didn't work 2017-01-26 09:48:40 which alpine version? 2017-01-26 09:48:43 3.4? 3.5? 2017-01-26 09:48:45 edge? 2017-01-26 09:48:55 3.5 official image 2017-01-26 09:49:05 ncopa, re: EFI boot 2017-01-26 09:49:10 with 2 disks in raid1 2017-01-26 09:49:12 fixed only by reissuing setup-disk -m sys 2017-01-26 09:49:20 ncopa: yes 2017-01-26 09:49:25 that the first one 2017-01-26 09:49:25 syslinux seems to do efi boot just fine on x86_64. though secureboot might be problematic 2017-01-26 09:49:43 i wonder if we should use syslinux for also uefi boot on arches where supported 2017-01-26 09:49:46 second one, I reproduced the exact same msdos label/parts and everything 2017-01-26 09:49:57 on reboot "boot disk not found" or the likes 2017-01-26 09:50:08 not even tried yet on 3rd one 2017-01-26 09:50:36 coredumb: sounds weird 2017-01-26 09:50:47 i will have to try reproduce it in qemu or similar 2017-01-26 09:50:59 to figure out what happens 2017-01-26 09:51:24 did you install the syslinux mbr? 2017-01-26 09:51:30 ncopa: about the kernel upgrade, I upgraded to my own 4.8 kernel, thought it could come from this but upon reinstalling the 4.4 official version it did the same result until I rerun setup-disk 2017-01-26 09:51:55 ncopa: yes I did both after the setup-disk step 2017-01-26 09:52:19 dd if=/usr/share/syslinux/mbr.bin ... or similar 2017-01-26 09:52:20 I haven't dug up the script to see what's actually happening 2017-01-26 09:52:24 yep sure 2017-01-26 09:52:36 did you mark the boot partition bootable in fdisk? 2017-01-26 09:52:46 in parted 2017-01-26 09:52:50 but yes 2017-01-26 09:53:01 do you have the /boot on separate partition or on the same partition? 2017-01-26 09:53:06 same 2017-01-26 09:53:23 so no separate /root 2017-01-26 09:53:38 sorry 2017-01-26 09:53:42 so no separate /boot 2017-01-26 09:53:57 how did you create the raid1? 2017-01-26 09:54:04 I basically have sd{a,b}1 a 10GB for OS part and sd{a,b}2 another raid1 for lvm usage and datas 2017-01-26 09:54:19 mdadm --create --level=1 ...... 2017-01-26 09:54:28 for that setup to work you need use --metadata=0.90 or what it was 2017-01-26 09:54:39 yep been set indeed 2017-01-26 09:54:43 at least for the partition with /boot 2017-01-26 09:54:50 what filesystem? 2017-01-26 09:54:54 ext4? 2017-01-26 09:55:06 sure it warns about it so I've set 0.90 on md0 and let it default on md1 2017-01-26 09:55:09 yes ext4 2017-01-26 09:55:58 did you dd if=/usr/share/syslinux/mbr.bin of=/dev/sda 2017-01-26 09:56:00 and sdb too? 2017-01-26 09:56:02 on both 2017-01-26 09:56:19 and are both sda1 and sdb1 marked as bootable with parted? 2017-01-26 09:56:24 both AND md0 device 2017-01-26 09:56:25 :D 2017-01-26 09:56:42 yes I've replicated table with sfdisk 2017-01-26 09:57:14 did you try reboot it with old kernel to make sure that it booted the first setup? 2017-01-26 09:57:16 the old kernel? 2017-01-26 09:57:27 i wonder if it was the kernel upgrade that broke things 2017-01-26 09:57:30 btw my kernel package is here http://git.mauras.ch/repos.mauras.ch/aports/src/master/linux-grsec you'll see there's nothing fancy, it's basically the same configuration as default kernel but for 4.8 grsec test with all grsec features enabled 2017-01-26 09:57:46 shouldnt break syslinux 2017-01-26 09:57:53 no shouldn't 2017-01-26 09:58:03 and I couldn't get any debug output from syslinux 2017-01-26 09:58:11 it was just looping on the menu 2017-01-26 09:58:36 couln't make it output any error 2017-01-26 09:59:27 as I said it got fixed only by redoing setup-disk step 2017-01-26 09:59:43 reinstalling the "old" 4.4.44 kernel didn't fix anything 2017-01-26 09:59:57 basically apk del and apk add in chroot 2017-01-26 10:00:52 so second server doesn't even boot 2017-01-26 10:01:02 and it's still pristine after installation 2017-01-26 10:01:10 I'm a bit puzzled by all that :( 2017-01-26 10:01:23 yeah 2017-01-26 10:01:25 hum 2017-01-26 10:01:48 got stuck on that until 3am last night 2017-01-26 10:01:54 then I got to bed 2017-01-26 10:01:56 :D 2017-01-26 10:02:27 all these reboots are quite long :( 2017-01-26 10:02:37 ok 2017-01-26 10:02:42 i will try reproduce it 2017-01-26 10:02:44 in qemu 2017-01-26 10:03:00 I run this kernel on a bunch of 3.5 installs already btw 2017-01-26 10:03:10 in VMs I got no issue whatsoever 2017-01-26 10:03:21 its not kernel problem 2017-01-26 10:03:27 I don't think so 2017-01-26 10:03:29 since it loops in bootloader 2017-01-26 10:03:30 just saying 2017-01-26 10:03:33 :) 2017-01-26 10:03:59 and as it rebooted fine couple of times before I upgrade the kernel .... 2017-01-26 10:04:30 ok. that piece of info is important 2017-01-26 10:04:31 thanks 2017-01-26 10:04:38 I have an equivalent setup with 1TB disks I installed under 3.4 last year and I didn't have any issue with it 2017-01-26 10:05:20 same partition setup with raid1 and 10GB first boot partition 2017-01-26 10:05:38 you set up the disks with parted? 2017-01-26 10:05:53 yes 2017-01-26 10:06:28 mklabel msdos 2017-01-26 10:07:58 did you use sfdisk to clone the partition table? 2017-01-26 10:08:21 how did you set up the disks in first place? 2017-01-26 10:09:21 ncopa: mklabel msdos + mkpart + set xxxxx on sda 2017-01-26 10:09:34 then sfdisk -d /dev/sda | sfdisk /dev/sdb 2017-01-26 10:09:49 first server I had to reboot to have all devices ready 2017-01-26 10:09:56 second worked first time .... 2017-01-26 10:10:08 maybe that's linked to the "not boot issue" 2017-01-26 10:11:14 gtg will get back in couple of hours 2017-01-26 10:11:22 can you verity that sdb1 has the boot flag is enabled? 2017-01-26 10:11:41 sfdisk -d /dev/sda | sfdisk /dev/sdb does not seem to set it 2017-01-26 10:18:46 weird 2017-01-26 10:19:02 looks like it didnt set bootable flag on the partition 2017-01-26 10:19:34 says missing operating system 2017-01-26 10:32:29 ok, i have no idea whats going on 2017-01-26 10:32:35 doesnt work as epxected 2017-01-26 10:33:02 No, if no bootable flag, it would say "No bootable device". In this case there is no mbr on boot partition. #cat /usr/share/syslinux/mbr.bin > /dev/sda1 2017-01-26 10:33:19 thats wrong 2017-01-26 10:33:32 you should copy the mbr.bin to /dev/sda 2017-01-26 10:33:33 not sda1 2017-01-26 10:33:53 and with fdisk or parted toggle the "boot" flag of the first partition 2017-01-26 10:33:55 oh, yes :) 2017-01-26 10:34:11 once i done that it does start syslinux 2017-01-26 10:34:15 or extlinux 2017-01-26 10:34:28 but extlinux says it cannot find any config file 2017-01-26 10:34:42 i suspect that mdadm or extlinux is broke 2017-01-26 10:35:45 extlinux -i /dev/sda 2017-01-26 10:35:53 no 2017-01-26 10:36:05 or for raid1: extlinux -i -r /dev/sda; extlinux -i -r /dev/sdb 2017-01-26 10:36:09 extlinux --install --raid /mnt/boot 2017-01-26 10:36:52 it seems that extlinux fails to read the ext4 partition for some reason 2017-01-26 10:40:11 just install 3.5 to bare-metal 2x1Tb hdds raid1 with lvm and ext4 - works fine 2017-01-26 10:40:43 does it work after you do apk upgrade and get new kernel? 2017-01-26 10:41:00 one sec 2017-01-26 10:41:13 or apk fix linux-grsec to reinstall the kernel 2017-01-26 10:53:13 my inet is too slow (( kernel's progress 50% 2017-01-26 11:10:26 reboot it remotely after upgrade, host pings ok, and seems works fine: #uname -a 2017-01-26 11:10:26 Linux nano 4.4.44-1-grsec #2-Alpine SMP Mon Jan 23 09:35:57 GMT 2017 x86_64 Linux 2017-01-26 11:35:47 sigh, i think i have found a segfault in sed 2017-01-26 11:36:39 gnu sed 2017-01-26 11:36:44 nspr build fails 2017-01-26 11:36:51 busybox sed seems to work 2017-01-26 12:19:47 https://lists.gnu.org/archive/html/screen-devel/2017-01/msg00025.html 2017-01-26 12:20:45 this doesn't seem to affect AL, as screen isn't suid 2017-01-26 12:21:22 as you were 2017-01-26 12:23:58 ncopa: I can confirm you that sdb has the flag 2017-01-26 12:24:06 sfdisk works "as expected" 2017-01-26 12:29:43 ScrumpyJack: we dont set suid, i checked 2017-01-26 12:30:31 how do I know which source-package a .apk package is built from? 2017-01-26 12:32:22 Or, how do I know which package provides a specific subpackage? 2017-01-26 12:35:08 $ apk search --origin -x libressl-dev 2017-01-26 12:35:08 libressl-2.4.4-r0 2017-01-26 12:37:00 fabled, cool, thanks! 2017-01-26 14:01:16 ncopa: think I'll try this evening 2017-01-26 14:22:00 Question: How do I "ldconfig" ?? I can't locate any "ld.so.conf.d" or something like that! 2017-01-26 14:24:16 http://www.openwall.com/lists/musl/2015/04/22/4 2017-01-26 14:25:25 thanx! 2017-01-26 14:50:52 oh no, a sedfault 2017-01-26 14:52:30 ha 2017-01-26 14:55:25 <^7heo> guys 2017-01-26 14:55:35 <^7heo> I can't post on the ML anymore it seems. 2017-01-26 14:55:44 <^7heo> ncopa: ^ 2017-01-26 14:56:20 <^7heo> Klowner: hi! :) 2017-01-26 14:56:36 ^7heo: howdy howdy 2017-01-26 14:57:15 <^7heo> :) 2017-01-27 01:31:22 When running 'apk add' with '--root ' flag I encountered the errors: http://pastebin.com/2YVB9QYm - why it that? Anybody has clue? 2017-01-27 01:39:49 you must --initdb; --root is for specifying where you want / to be 2017-01-27 01:39:55 it is mostly only useful for chroot 2017-01-27 01:47:08 awilfox: Thanks, I will follow instructions from: https://wiki.alpinelinux.org/wiki/Installing_Alpine_Linux_in_a_chroot 2017-01-27 01:47:44 good idea :) 2017-01-27 01:50:00 :) 2017-01-27 01:50:53 czart: wiki jest często gęsto nieaktualna, także ostrożnie 2017-01-27 01:52:23 skrzyp: OK, thanks :) 2017-01-27 02:00:02 grr, this chinese laptop is UEFI only 2017-01-27 02:00:04 because of course it is 2017-01-27 02:02:04 kaniini: is there some 3rd party BIOS calls emulation as uefi application? 2017-01-27 02:02:17 like Clover for regular BIOSes 2017-01-27 02:03:48 not that i know of :( 2017-01-27 02:03:55 but there is gummiboot in alpine. 2017-01-27 02:05:52 "wiki jest często gęsto nieaktualna, także ostrożnie" -> loosely PL to ENG -> wiki content remains outdated, so take precautions 2017-01-27 02:07:43 no one cares, they know about it 2017-01-27 02:08:06 at some point I was actually browsing the C code of mkinitfs 2017-01-27 02:08:16 and initrd scripts 2017-01-27 02:08:32 because it's not docummented anyhow 2017-01-27 02:19:55 seems that the kernel does not really support UEFI boot though. hmm ;) 2017-01-27 02:22:39 by efistub? 2017-01-27 02:25:27 gummiboot 2017-01-27 02:25:34 it boots, but it's not usable 2017-01-27 02:25:40 because the console does not come up 2017-01-27 02:25:53 i will play with it later 2017-01-27 02:25:55 lol 2017-01-27 02:29:51 efistub is good 2017-01-27 02:41:35 skrzyp: I haven't got opportunity to touch the low-level aspects. mkinitfs seems based mainly on Linux kernel modules, so maybe individual modules are documented somewhere else. Do you think anybody needs to know about the internals? If you are a developer you have to read the code anyway and "clean" code is something what tends to provide "self-documenting" code. 2017-01-27 02:42:16 mkinitfs does need better documentation really 2017-01-27 02:42:56 kaniini: but for developers or casual users? 2017-01-27 02:43:00 or both? 2017-01-27 02:45:19 czart: people who wish to become developers in that area of the distribution (which is not uncommon i would say, as alpine is beginning to "grow up") 2017-01-27 02:52:34 kaniini: It sounds exiting that its growing up! I may have a try to pursue documenting... But, where to start? Is there any "Alpine developers guide" or something like that? 2017-01-27 02:52:46 it's growing up* 2017-01-27 05:46:10 czart: there's opportunity imo to start formally organizing these types of projects 2017-01-27 06:39:47 fabled, ncopa: what is your opinion on https://github.com/vakartel/abuild/commit/cd032b0fd02a0c2ed1c4fd5d493508eecbbbd141 ? 2017-01-27 06:40:02 something i noticed while snooping around github 2017-01-27 06:40:29 it seems okay 2017-01-27 06:52:11 morning 2017-01-27 06:59:18 >>> WARNING: pkgconf: APKBUILD does not run any tests - define a check() function! 2017-01-27 06:59:20 woooo 2017-01-27 07:01:30 nice! 2017-01-27 07:17:59 fabled, ncopa: https://github.com/alpinelinux/abuild/compare/master...kaniini:check 2017-01-27 07:20:02 kaniini, looks good but question: "checkdepends" should they be installed only for check, or also for build? 2017-01-27 07:20:22 i wonder if there's any potential concern that checkdepends pull in stuff we don't want during build 2017-01-27 07:20:25 https://github.com/kaniini/aports/commit/f1316341b981158f77c8535ef1cc16a080d61104 2017-01-27 07:20:53 fabled: yes, i'm not sure how to do that part optimally. but the build system may need the test runner to be installed in order to generate testsuite files 2017-01-27 07:21:17 in theory, $checkdepends should only be installed in non-cross situation where check would be run 2017-01-27 07:21:26 but i may misunderstand that code ;) 2017-01-27 07:21:44 it has two implications 2017-01-27 07:22:04 1) checkdepends (when check will be done) will be present in build stage too, and it may cause the configure to pick up extra library dependencies 2017-01-27 07:22:19 e.g. libfoo is mandatory check dependency, and optional build dependency 2017-01-27 07:22:47 i wonder if it should be split 2017-01-27 07:23:12 2) it may affect the build graph and generate cycles if there's core packages requiring something compicated for 'check' 2017-01-27 07:23:23 but i think we just then need a flag to omit check 2017-01-27 07:23:47 lua-aports should probably be included to take note of the new 'checkdepends' so it can order the build correctly 2017-01-27 07:23:48 but for example, some configure require the test runner to be present in order to enable tests 2017-01-27 07:24:01 so i am not sure if maybe we should have 2017-01-27 07:24:11 checkdepends_build, checkdepends_run 2017-01-27 07:24:13 maybe? 2017-01-27 07:24:16 oh, that's true too 2017-01-27 07:24:28 configure probably checks for test deps 2017-01-27 07:25:30 anyway this is still obviously proof of concept stuff, and it's not ready to go into abuild yet 2017-01-27 07:26:06 maybe we need 2017-01-27 07:26:18 i think the changes are correct 2017-01-27 07:26:18 run_checks() {} function 2017-01-27 07:26:29 to determine if $checkdepends is used 2017-01-27 07:26:37 also skip checks if BOOTSTRAP=yes in env 2017-01-27 07:26:49 yep 2017-01-27 07:26:52 or just defined 2017-01-27 07:26:56 for example 2017-01-27 07:27:08 pkgconf depends on itself (indirectly) to run tests 2017-01-27 07:27:15 yes 2017-01-27 07:27:18 because the testsuite is lua-based 2017-01-27 07:30:16 one option is to make the 'checkdepends' optional, and if they fail to install we just skip tests 2017-01-27 07:31:58 i'm not sure, this is getting deeper into abuild itself than i've gone in a very long time haha 2017-01-27 07:33:13 i'll look at it more after i sleep 2017-01-27 07:51:05 kaniini: could you be more specific about what you have said recently? 2017-01-27 07:59:44 czart: right now there's not much way in terms of a formal docs project, but the wiki is a starting point 2017-01-27 08:01:36 kaniini: you mean e.g. https://wiki.alpinelinux.org/wiki/Developer_Documentation ? 2017-01-27 08:02:56 yes 2017-01-27 08:06:37 kaniini: OK, thanks, I also found: https://wiki.alpinelinux.org/wiki/Alpine_Linux:Contribute 2017-01-27 08:07:47 ncopa: http://pastebin.com/QqveHpW6 < This should be fine for reboot right ? 2017-01-27 08:13:12 anyone here using mpd on 3.5? 2017-01-27 08:24:43 clandmeter: I do but I'm on edge 2017-01-27 08:26:57 ncopa: same result ... "No boot device available" did I forget something before rebooting ? 2017-01-27 08:43:26 ncopa: 3.4.6 seems OK with the same procedure! 2017-01-27 08:46:14 coredumb: thats weird 2017-01-27 08:46:56 ncopa: now I upgraded to 3.5 and reboot on its way 2017-01-27 08:47:19 still boots alright 2017-01-27 08:47:38 how is sda and sdb partitioned? 2017-01-27 08:48:11 exactly the same 2017-01-27 08:48:21 gimme a sec I made a screenshot 2017-01-27 08:49:39 ncopa: http://i.imgur.com/ggt2hIV.png 2017-01-27 08:49:51 I didn't redo anything on 3.4.6 installer 2017-01-27 08:49:58 just remounted the raid 2017-01-27 08:50:14 cleaned up old install, rerun setup-disk 2017-01-27 08:50:27 and readded mbr.bin from 3.4.6 syslinux 2017-01-27 08:52:38 upgraded to my 4.8.17 kernel 2017-01-27 08:52:42 still boots fine 2017-01-27 08:52:45 (as expected) 2017-01-27 08:53:01 ncopa: FWIW I used 3.5 alpine-standard iso 2017-01-27 08:53:07 if it makes any difference... 2017-01-27 09:04:52 coredumb: i tried to set up similar in qemu yesterday 2017-01-27 09:05:01 but i had some issues with getting it working 2017-01-27 09:05:11 smells like there is a bug in syslinux or something 2017-01-27 09:05:45 i would appreciate if you could help me find out exactly what is going on 2017-01-27 09:06:38 ncopa: I still have a server to install 2017-01-27 09:06:56 would you like access to its idrac to do some testing ? 2017-01-27 09:07:24 im a bit busy today 2017-01-27 09:07:27 sorry 2017-01-27 09:07:52 i need upgrade the build server hosts 2017-01-27 09:07:53 and actually seeing the bug and how it breaks my kernel updates on the first one, I'm considering reinstalling the first one on 3.4.6 ... 2017-01-27 09:08:06 ncopa: well then let me know what I can do 2017-01-27 09:08:23 I'm not in a hurry to get this third one installed btw 2017-01-27 09:17:03 also dl-cdn seems to not be up2date with nl .... 2017-01-27 09:17:52 ah no my bad 2017-01-27 09:17:57 edge community * 2017-01-27 09:22:47 jirutka: is that normal that borgbackup 1.0.9 didn't make it in 3.5/community being a security update ? Still only 1.0.8 2017-01-27 09:23:28 now I see that it's not been said that the update was related to security ... 2017-01-27 09:26:40 coredumb: do you have a CVE for it? 2017-01-27 09:33:23 ncopa: I don't think there's CVEs for borg 2017-01-27 09:33:37 I check http://borgbackup.readthedocs.io/en/stable/changes.html 2017-01-27 09:34:23 if it is a security issue, then they should probably request a CVE 2017-01-27 09:34:25 http://borgbackup.readthedocs.io/en/stable/changes.html#changelog 2017-01-27 09:34:52 ncopa: yeah they should 2017-01-27 09:39:31 ncopa: I opened an issue about that we'll see 2017-01-27 10:23:15 ncopa: oh apparently they did lol :D 2017-01-27 10:24:21 i just heard yes 2017-01-27 10:25:39 http://www.cvedetails.com/vulnerability-list/vendor_id-16008/product_id-35461/year-2017/Borg-Borg.html 2017-01-27 10:27:46 i am taking down the build servers now 2017-01-27 10:29:45 ok 2017-01-27 10:37:32 phew 2017-01-27 10:37:32 bld servers are back up 2017-01-27 10:37:36 \o/ 2017-01-27 10:54:57 ncopa: how difficult would that be to add the borg CVEs to the 1.0.9 commit ? 2017-01-27 10:55:42 extremely difficult 2017-01-27 10:55:44 :) 2017-01-27 10:55:59 we will make sure the fix gets to the v3.5 2017-01-27 10:56:02 dont worry 2017-01-27 10:56:10 \o/ 2017-01-27 11:01:41 coredumb: pushed 2017-01-27 11:01:50 cool 2017-01-27 11:02:40 ncopa: let me how I could help with this syslinux issue 2017-01-27 11:03:24 it would be good i you could debug it :-/ 2017-01-27 11:03:26 alternatively 2017-01-27 11:03:34 if you can reproduce it in a qemu vm 2017-01-27 11:03:49 and write some step by step instructions 2017-01-27 11:04:17 I couldn't even get a single error from the syslinux menu :O 2017-01-27 11:04:45 Let me finish my new hypervisor installation and I'll try to reproduce it in a VM 2017-01-27 11:04:59 preferible in qemu 2017-01-27 11:05:22 i have access to virtualbox and vmware fusion on mac too 2017-01-27 11:05:30 yeah libvirt KVM guest it will 2017-01-27 11:05:37 I don't do any other 2017-01-27 11:05:48 cool 2017-01-27 11:05:51 thanks 2017-01-27 11:06:01 libvirt kvm is pretty nice 2017-01-27 11:06:07 i use it on my work desktop 2017-01-27 11:06:13 :) 2017-01-27 11:06:21 I use it everywhere :D 2017-01-27 11:06:39 it just consumes a bit memory 2017-01-27 11:09:31 you mean on your desktop ? 2017-01-27 11:09:37 well indeed 2017-01-27 11:09:58 ok let's migrate this IRC machine first :D 2017-01-27 11:32:27 <^7heo> moin alles 2017-01-27 11:38:34 moin 2017-01-27 11:39:21 mmmh can one still use libvirt without dbus and polkit on 3.5 ? 2017-01-27 11:45:32 yeah seems so 2017-01-27 14:03:59 <^7heo> ncopa: just for you to know, my new target for Alpine is the usb armory 2017-01-27 14:04:27 <^7heo> ncopa: it has a module in which you can FUSE (as in physically fuse) the public key of your software 2017-01-27 14:04:43 <^7heo> ncopa: so your boot software has to be signed with your key 2017-01-27 14:04:51 <^7heo> (ofc it is permanent) 2017-01-27 14:04:55 <^7heo> so yeah, NICE. 2017-01-27 14:05:26 <^7heo> now I have to secure my key 2017-01-27 14:20:51 ^7heo, cool. i looked at usb armory earlier in t2.fi few years ago 2017-01-27 14:21:19 <^7heo> t2.fi looks familiar 2017-01-27 14:21:22 <^7heo> but I can't tell what it means 2017-01-27 14:21:38 infosec event in finland 2017-01-27 14:21:46 it's url for it 2017-01-27 14:22:27 <^7heo> hmmm 2017-01-27 14:22:34 <^7heo> right. 2017-01-27 14:23:13 <^7heo> Looks interesting 2017-01-27 14:23:18 <^7heo> I'll think about '17 2017-01-27 14:23:25 <^7heo> btw fabled, no FOSDEM for you? 2017-01-27 14:23:37 sorry, no can do. 2017-01-27 14:23:49 <^7heo> ok 2017-01-27 14:23:58 <^7heo> fabled: are you involved in the t2.fi? 2017-01-27 14:24:21 i've been creating the challenge on few years. last year i had a talk on linux hardening. 2017-01-27 14:24:38 so i often go there. not necessarily every year, though 2017-01-27 14:25:02 <^7heo> ok 2017-01-27 14:25:09 <^7heo> I'm asking because if I attend it 2017-01-27 14:25:20 <^7heo> and I pay for the food in my ticket's price 2017-01-27 14:25:48 <^7heo> and I can't eat it 2017-01-27 14:25:52 <^7heo> I would be pretty annoyed. 2017-01-27 14:27:17 depends on how picky you are on food... :) 2017-01-27 14:27:37 the lunch has been the hotel restaurants buffet on the venue 2017-01-27 14:29:01 i'm sure you can always ask about it for your dietary requirements. the conference caps at 99 attendees (including staff) so you get a bit more personalized discussion if needed... 2017-01-27 14:29:58 <^7heo> I'm not picky on food, I really like all foods 2017-01-27 14:30:10 <^7heo> I just do not eat animal products. 2017-01-27 14:30:27 <^7heo> due to their extremely poor quality when made industrially 2017-01-27 14:30:33 <^7heo> (and they are always made that way) 2017-01-27 14:30:54 <^7heo> I see, 99 people max 2017-01-27 14:31:00 <^7heo> it's not exactly like the FOSDEM :P 2017-01-27 14:32:08 nope 2017-01-27 14:32:34 <^7heo> Ok 2017-01-27 14:32:43 <^7heo> is it happening in Finnish or English? 2017-01-27 14:33:13 english 2017-01-27 14:33:33 <^7heo> Ok so at least I can attend. 2017-01-27 14:33:54 http://t2.fi/conference/ and http://t2.fi/about/eula/ gives an idea 2017-01-27 14:34:24 <^7heo> Did you do the '16 challenge? 2017-01-27 14:34:57 i was involved in the early discussions, but dropped out because i did a talk 2017-01-27 14:35:36 <^7heo> ok 2017-01-27 14:35:42 <^7heo> I should try it 2017-01-27 14:35:49 i was doing the '13, '14 and '15 challenges 2017-01-27 14:36:10 <^7heo> I didn't yet do any infosec challenge ever 2017-01-27 14:36:32 the t2 challenges are not the easiest ones... 2017-01-27 14:36:35 <^7heo> yet I went at a handful of security conferences 2017-01-27 14:37:04 <^7heo> fabled: spending time on something easy isn't necessarily the best way to learn. 2017-01-27 14:37:59 <^7heo> So basically the challenge is the ticket... 2017-01-27 14:38:00 <^7heo> nice. 2017-01-27 14:38:12 true. and yes, if you have infosec knowledge/background, they'll be interesting to look at 2017-01-27 14:38:36 yes, i got involved in t2 by solving the challenge on years before '13 2017-01-27 14:38:51 then they banned me from winning a free ticket again 2017-01-27 14:39:10 <^7heo> every year you solved the challenge? 2017-01-27 14:39:46 yes. but never as 1st one. i got the ticket based on 'best writeup' on two years 2017-01-27 14:40:07 "if you can hack our database to register yourself as invited, then you're de facto invited" 2017-01-27 14:40:16 :) 2017-01-27 14:40:51 <^7heo> ahaha 2017-01-27 14:50:01 <^7heo> I guess I'd better visit finland first 2017-01-27 14:58:49 <^7heo> fabled: is the ë in Finnish the same sound as in German? 2017-01-27 15:00:30 we don't havae *. only *, * and * 2017-01-27 15:00:40 <^7heo> wat? 2017-01-27 15:02:59 he means å, ä, ö 2017-01-27 15:03:21 I don't think Germans use ë either 2017-01-27 15:03:37 they have ü though 2017-01-27 15:11:26 <^7heo> kunkku: you're right, I'm too tired. 2017-01-27 15:11:30 <^7heo> kaniini: I meant ö 2017-01-27 16:27:29 whoever here made this commit: http://git.alpinelinux.org/cgit/aports/commit/?h=3.5-stable&id=7d48b85c0e4d91087ef45c331059a9a606ae9588 2017-01-27 16:27:35 many thanks :D 2017-01-27 17:13:50 Anybody familiar with Docker to answer: "Is the 'scratch' image (i.e. when it appears in a Docker file: FROM scratch) bounded up with boot2docker anyhow? I am confused, since I use Alpine Linux, which uses "FROM scratch" to create the "alpine" official image. On the other hand, when I run 'uname -r' within the respective container, I get: '4.4.39-boot2docker'.". I asked at #docker, but no respond received... 2017-01-27 17:21:25 _ikke_: hey. https://github.com/darktable-org/darktable/releases/tag/release-2.2.2 2017-01-27 17:24:58 LebedevRI: reading through your release notes 2017-01-27 17:25:08 I can supply raw shots from an EOS 750D if they're needed? 2017-01-27 17:25:51 they are needed, yes 2017-01-27 17:27:23 asking because it seems the sample repo alraedy contains a 750D image 2017-01-27 17:27:39 it's not under cc0 is it :) 2017-01-27 17:27:57 right. 2017-01-27 17:28:15 also, canon names these XXXD cameras with three different names in different areas of the world, and i think we want all three samples 2017-01-27 17:28:57 they're the same model though :p 2017-01-27 17:29:07 even the EXIF data says Rebel T6i / EOS 750D / Kiss X 2017-01-27 17:29:09 in model ID 2017-01-27 17:29:31 it's a mess :) 2017-01-27 17:34:40 LebedevRI: uploaded :) 2017-01-27 17:34:47 not sure if it's subject to approval or anything, but there you go 2017-01-27 17:36:34 it is, validated, thank you 2017-01-27 18:09:55 <_ikke_> Shiz: If you could test darktable on awesome, that would be nice 2017-01-27 18:10:21 <_ikke_> on alpine* 2017-01-27 18:10:39 <_ikke_> LebedevRI: Thanks, I'll update it 2017-01-27 18:53:07 czart: "scratch" is empty 2017-01-27 18:53:25 so "from scratch" will get you a completely empty image 2017-01-27 18:54:19 czart: `uname -r` will tell you the kernel information 2017-01-27 18:54:39 and has nothing to do with the alpine image or anything to do with "scratch" 2017-01-27 18:54:59 if you run centos, debian, ubuntu whatever linux image, uname will still return the same 2017-01-27 18:55:23 because your containers share kernel with hose 2017-01-27 18:55:25 host* 2017-01-27 19:01:19 clandmeter: since you and jirutka showed some happiness when I succeed to build the gcc on a POWER8 machine, I would like to share this paste with you and the Alpine community: https://hastebin.com/raw/yoyoweluxu 2017-01-27 19:01:33 thanks for all the help ;-) 2017-01-27 19:02:50 gromero: that is awesome! 2017-01-27 19:03:12 ncopa: :) 2017-01-27 19:03:20 can i tweet that? 2017-01-27 19:03:34 that hastebin url 2017-01-27 19:03:57 gromero: or even better, you tweet and i retweet 2017-01-27 19:04:08 <^7heo> hastebin is funny in German 2017-01-27 19:04:14 ncopa: well, I guess so. we have to contribute back out patches yet... but I guess so :) 2017-01-27 19:04:24 its progress 2017-01-27 19:04:39 <^7heo> like systemd 2017-01-27 19:04:41 ncopa: I'll paste that in a less volatile pastebin? 2017-01-27 19:04:42 <^7heo> progress 2017-01-27 19:04:53 and proof that its not completely impossible 2017-01-27 19:05:51 damn ^7heo you're such a troll :D 2017-01-27 19:05:52 ^7heo: one day i will kick you for cursing if you continue use the s* word 2017-01-27 19:05:54 <^7heo> gromero: why not put it in the doc? 2017-01-27 19:06:02 <^7heo> ncopa: :p 2017-01-27 19:06:27 <^7heo> coredumb1: it's,trollday 2017-01-27 19:06:37 <^7heo> s-,- - 2017-01-27 19:06:50 ncopa: http://pastebin.com/raw/cKRkWRMi 2017-01-27 19:07:09 ncopa: ta! :-) 2017-01-27 19:07:35 <^7heo> gromero: but seriously, in the doc it would be cool too 2017-01-27 19:08:25 ^7heo: do mind to explain that better? 2017-01-27 19:09:00 <^7heo> well if you are okay with a wiki used only over a browser, you can paste stuff in our wiki 2017-01-27 19:14:45 gromero: do you have twitter? 2017-01-27 19:15:03 ncopa: just a note, I don't have an active twitter account, so feel free to tweet it ;-) 2017-01-27 19:17:01 https://twitter.com/n_copa/status/825059837901942785 2017-01-27 19:19:30 ;-) 2017-01-27 19:19:37 cool 2017-01-27 19:28:42 hi, i have the following problem building a new package for alpine. i want to package gnuradio but fetching vie wget from gnuradio.org shows the following error: "wget: error getting response: Address family not supported by protocol". i know it's not an alpine issue but downloading via abuild fails too... any idea how i can handle this? 2017-01-27 19:29:13 it sounds like it is either v4 or v6 only, and your computer only supports the other one 2017-01-27 19:29:22 you can look for a mirror maybe 2017-01-27 19:30:17 awilfox: yes, it has to do with ipv6 and ipv4 but that's bad... 2017-01-27 19:31:02 but actually a have an ipv4 and ipv6 address... 2017-01-27 19:32:11 since alpine abuild uses curl for downloading it starts downloading but aborts after some seconds 2017-01-27 19:41:17 ok, i was wrong... i am working not locally and the remote host where i am developing has only ipv4... :( 2017-01-27 19:44:28 if curl has a -4 option, use it 2017-01-27 19:44:54 busybox wget doesn't, because "prefer ipv4 addresses" is a global busybox option 2017-01-27 19:46:04 skarnet: but i need that feature in an APKBUILD... 2017-01-27 19:46:25 it fails in abuild too 2017-01-27 19:47:09 curl might have a global configuration file, or a configuration environment variable 2017-01-27 19:48:23 skarnet: but this will fail not only for me when the package possibly becomes upstream in alpine... 2017-01-27 19:48:53 i will enable ipv6 now as a workaround but i believe this sad anyway... 2017-01-27 19:49:08 that's a problem that Alpine needs to fix on a more general scale, not only for your package 2017-01-27 19:49:22 debian had that problem at some point 2017-01-27 19:49:26 too 2017-01-27 19:50:17 but the happy news is, gnuradio compiles in alpine when doing it without abuild... \o/ 2017-01-27 19:51:44 want to do some SDR stuff because of RFID research and since i am moving over to alpine from arch i am really missing some packages. 2017-01-27 19:53:46 argh, have an ipv6 address now and it fails... the archive download exits after about a second and extracting fails... :( 2017-01-27 19:56:00 "wget: error getting response: Network unreachable"... 2017-01-27 19:56:05 i am confused 2017-01-27 20:18:21 okay, im agoinf crazy now and just wynt to go on packaging gnuradio. where can i put the archive so abuild checksum and abuild -r does not need to download everytime? 2017-01-27 20:21:06 sorry, forgot... got it... 2017-01-27 20:29:31 can the libreadline package re-enable static compilation, or was that done to work around a bug? 2017-01-27 20:46:59 ncopa: Thanks for pointing this out! In fact, I was not aware of that, and this ("your containers share kernel with host") clarifies many things that I was struggling to understand. For example, now it's clear why containers can be so very lightweight, as they share the kernel with host. I will continue investigation tomorrow - need to have some sleep. Thanks! :) 2017-01-27 20:50:37 czart: be careful on that, one thing that can happen is if your kernel supports syscalls that don't exist where you're going to run, aka a 4.9 kernel with, i dunno, a madvise call that 4.0 didn't have, you can get fun behavior, it also means your containers aren't really running exactly in the environment they expect, aka ubuntu 12.04 with a 4.5 kernel is probably not going to behave as you'd think 2017-01-27 20:50:45 that kinda stuff has bitten me a lot in the past year 2017-01-27 20:51:04 <_ikke_> Doesn't abump add the built package to the local repository? 2017-01-27 20:52:09 <_ikke_> I see it in the directory, but the index seems to have the old version 2017-01-27 21:13:45 http://git.alpinelinux.org/cgit/aports/tree/main/readline/APKBUILD#n50 2017-01-27 21:13:49 line 50 2017-01-27 21:14:00 requesting that static library versions of ncurses be re-enabled 2017-01-27 22:22:12 djw: please file an issue on bugs.alpinelinux.org its friday here and on monday i have forgotten it 2017-01-27 22:22:38 djw: i think i'll add a -static subpackage. we normally want keep the statci libs out 2017-01-27 22:23:43 _ikke_: abump should git add the package after upgrade 2017-01-27 22:23:49 and git commit 2017-01-27 22:27:49 ncopa: thank you. 2017-01-27 22:28:40 ncopa: should I make multiple bug reports? There is also the bug that libreadline doesn't support pkg-config 2017-01-27 22:30:18 um, might be enough with one since i'll likely fix them both in one shot 2017-01-27 22:30:50 readline not shipping pkg-config is technically an upstream bug 2017-01-27 22:31:22 hm. how to report a bug. the bug issue tracker doesn't really explain it 2017-01-27 22:32:03 you need create an account i think 2017-01-27 22:32:14 i think you can send email 2017-01-27 22:32:14 :-( ok. 2017-01-27 22:32:23 and it will create anonymous issue 2017-01-27 22:32:28 sending an email would be ideal. I'm used to the debian bug tracker 2017-01-27 22:32:40 try send an email to bugs@alpinelinux.org 2017-01-27 22:36:05 i havent tested if the email gate works nowdays 2017-01-27 22:37:41 ok, atleast alpine@bugs.alpinelinux.org works 2017-01-27 22:39:30 ok, sent 2017-01-27 22:41:14 http://bugs.alpinelinux.org/issues/6768 2017-01-27 22:41:53 doesnt collect your email, so you wont get notification 2017-01-28 00:18:59 is travis-ci.org really free for building packages for open source projects? i currently build a lot of 'bigger' packages for alpine and it takes hours on my machines... 2017-01-28 00:56:00 ncopa: seems like executing setup-disk step twice make it boot >_< 2017-01-28 00:56:28 and kernel upgrade doesn't break syslinux on reboot 2017-01-28 01:29:57 hanez: yes 2017-01-28 01:34:49 seems to timeout after an hour 2017-01-28 01:35:09 least for the ghc port it barely gets past stage 1 of the 3 stages of compiling 2017-01-28 01:35:31 Shiz: so when i register i can use it for building all alpine packages i publish? 2017-01-28 01:36:12 my goal is to get them upstream anyway but it really takes some time testing all that stuff 2017-01-28 05:50:42 http://imgur.com/a/Km7gQ 2017-01-28 05:50:47 don't buy these drives whatever you do 2017-01-28 05:50:53 yet another one bites the dust in my build server 2017-01-28 05:54:12 <_ikke_> ncopa: sorry, I meant the alpine repository 2017-01-28 05:54:17 <_ikke_> not the git repository 2017-01-28 05:55:43 [ 187.908815] blk_update_request: I/O error, dev sdc, sector 109304064 is generally a bad sign that your raid has failed 2017-01-28 05:55:53 thanks seagate 2017-01-28 09:10:56 kaniini: I've stopped buying seagates years ago 2017-01-28 09:11:05 they always had issues :( 2017-01-28 15:35:33 https://github.com/alpinelinux/alpine-conf/blob/master/setup-disk.in#L290 < funny this 2017-01-28 15:36:52 I'm using UDEREF without any problem on VMware 2017-01-28 15:38:40 and AFAIK default linux-grsec doesn't enable UDEREF right ? 2017-01-28 16:31:45 is there a plan to switch /usr/bin/python to python3 as default? 2017-01-28 16:38:04 unlikely 2017-01-28 16:38:27 https://www.python.org/dev/peps/pep-0394/ 2017-01-28 16:40:30 Shiz: okay, nice. gnuradio would need a lot of patching else. didn't knew that pep. you made my day... ;) thx 2017-01-28 16:40:46 i would still recommend gnuradio patch their stuff, since it does say it can change in the future 2017-01-28 16:40:48 <_ikke_> Shiz: for arch python links to python3 2017-01-28 16:40:49 and software should be prepared 2017-01-28 16:40:58 _ikke_: it even says so in the PEP... 2017-01-28 16:41:05 however, the python devs recommendation is to link python to python2 2017-01-28 16:41:33 Shiz: yes you're right! arch is sometimes horrible patched... 2017-01-28 16:42:07 distros should do what the software developers recommend in such a case... 2017-01-28 16:42:21 <_ikke_> hanez: Well, arch switched before this PEP was there 2017-01-28 16:42:26 ah, ok 2017-01-28 16:42:50 <_ikke_> "however, end users should be aware that python refers to python3 on at least Arch Linux (that change is what prompted the creation of this PEP)," 2017-01-28 16:42:57 _ikke_: in gentoo you can select what it makes more defficult to know what the default is 2017-01-28 16:43:09 oh, ok 2017-01-28 16:43:52 <_ikke_> Shouldn't be a problem, you should be specific if it matters 2017-01-28 16:43:53 but like Shiz says... then gnuradio should use /usr/bin/python2 in their code then let us patching everything... 2017-01-28 16:44:06 yes, they should 2017-01-28 16:44:24 _ikke_: gnuradio relies on that /usr/bin/python is python2 2017-01-28 16:45:57 <_ikke_> Shiz: That unlikely is also not true. THe PEP expects that the default will change 2017-01-28 16:46:40 _ikke_: that will be fine but then gnuradio devs should fix it. not all distris schould patch... 2017-01-28 16:46:53 <_ikke_> right 2017-01-28 16:47:04 <_ikke_> arch has choosen to switch, so they are taken the burden on them 2017-01-28 16:56:28 _ikke_: well, the plan isn't currently there 2017-01-28 16:56:31 it will be in the future 2017-01-28 16:56:33 :p 2017-01-28 16:57:52 Shiz: i will take care of that issue in the future then... ;) \o/ 2017-01-28 19:03:22 mitchty: Thanks for pointing this out. I summarized likely cause of such an incompatibility in points: 1) kernels, 2) ABI (Application Binary Interface) 3) C standar library implementation. Here is full text: http://pastebin.com/hLAxPqNf . Is that what you meant? Wish you comment on that further? 2017-01-28 19:06:40 ncopa: with respect to host/container kernel sharing, I need to read more about Linux containers and LXC (the technology referenced on Alpine wiki: https://en.wikipedia.org/wiki/Alpine_Linux) 2017-01-29 03:23:57 i'm designing a service that, given a git repo with APKBUILDs in it, will build that repo and publish it 2017-01-29 15:06:28 czart_: mostly, the essence at the kernel level is that containers are leaky abstractions, and you used backward compatibility when forward is what you're looking for, aka stuff compiled on older will work in future 2017-01-29 15:07:06 if you bring glibc into the mix you get to deal with glibc versioning too, which is even more annoying, c++ then you can likely add libstdc++ into the fun 2017-01-29 15:14:15 just a ping for the ghc port https://github.com/alpinelinux/aports/pull/205 if anyone could rerun the snapshot function and upload the resultant binaries that would help, or if you want I can provide the armhf/x86_64 binaries I used to test with 2017-01-29 19:35:23 <_ikke_> I'm trying to create a package for ranger (curses file explorer), but it sefaults. Apparently curses.ungetmouse(0, 0, 0, 0, 0) is causing it (found by stepping through with pdb until it segfaulted( 2017-01-29 19:38:41 <_ikke_> from gdb: 0x00007ffff7d8fa07 in __stack_chk_fail () from /lib/ld-musl-x86_64.so.1 2017-01-29 19:39:38 sounds like a buffer overflow 2017-01-29 19:40:17 <_ikke_> Anoying thing with debugging curses programs is that they mess with the terminal 2017-01-29 19:51:50 <_ikke_> http://sprunge.us/XMKU 2017-01-29 21:46:24 sigh, i finally got around to packaging upx, and it works, but everything it compresses segfaults 2017-01-29 22:31:02 upx does not work well with musl binaries 2017-01-29 22:31:10 and with recent glibc ones 2017-01-29 22:31:27 on dynamically linked binaries, it throws notcompressibleexception 2017-01-29 22:31:49 yep, i don't know enough about its compression scheme to debug it atm either, i think i hit all of the above 2017-01-29 22:32:23 the backtraces in gdb are... reminding me of debugging a virus that is trying to hide itself 2017-01-29 22:33:02 i'm using it for atari/tos, dos/* and ps1/exe binaries 2017-01-29 22:33:34 would it pay to submit this to testing as it is? i haven't tried it on anything but a static musl binary and dynamic 2017-01-29 22:35:13 hmm 2017-01-29 22:38:40 its been a while since i used r2 but i could probably learn enough about how this works eventually, just not how I want to spend my time if its busted in general 2017-01-29 22:52:12 r2 is actually good 2017-01-29 22:52:14 hmm 2017-01-29 22:52:34 what upx does if you feed him with binary not using standard library? 2017-01-29 22:53:39 ok so upx that i built seems to work about as well with binaries that are old glibc and some other static ones i've compressed via glibc upx 2017-01-29 22:54:11 i had a static version of jq i built with alpine that i normally compress outside of apine linux 2017-01-29 22:54:19 that works just like upx on the outside 2017-01-29 22:54:51 and upx on the same hello world dynamic musl binary makes any dynamic binary break too so its no worse than glibc compiled upx 2017-01-29 23:26:09 https://github.com/alpinelinux/aports/compare/master...mitchty:upx?expand=1 if anyone wants give it a go, i'll sit on it for a bit and test 2017-01-30 06:00:45 fabled: what is your opinion on http://lists.alpinelinux.org/alpine-devel/5542.html 2017-01-30 06:01:08 morning 2017-01-30 06:01:43 kaniini, no strong opinion. i personally wouldn't like to use it. but then again seems there's use cases for it 2017-01-30 07:09:17 https://github.com/MartinBriza/MediaWriter 2017-01-30 07:09:18 this is neat 2017-01-30 07:50:59 morning 2017-01-30 08:54:47 _ikke_: i t does sound like a buffer overflow. try build it with debugging enabled (either `DEBUG=1 abuild -r` or add $pkgname-dbg as fisr subpackage) 2017-01-30 08:55:08 then you run `gdb --core ./core /usr/bin/` 2017-01-30 08:55:17 abd `bt` to get a backtrace 2017-01-30 09:56:27 <_ikke_> ncopa: thanks, will do 2017-01-30 09:56:50 <_ikke_> ncopa: note that it's a python package, will duilding with debug do anything? 2017-01-30 09:57:07 <_ikke_> building* 2017-01-30 09:57:13 <_ikke_> or should I rebuild curses? 2017-01-30 09:57:20 <_ikke_> / python 2017-01-30 09:57:58 probably anything in the chain build/install with dbg 2017-01-30 10:59:52 Anybody with openvpn + openvpn-auth-ldap installation and any alpine release, please run " openvpn --dev null --plugin /usr/lib/openvpn-auth-ldap.so" to check if it breaks or not. I suspect that libobjc is to blame. 2017-01-30 11:10:56 hi 2017-01-30 11:19:48 <_ikke_> LebedevRI: Hey 2017-01-30 11:20:07 whops, wrong channel: D 2017-01-30 11:22:56 <^7heo> nice one :D 2017-01-30 13:50:10 clandmeter: http://git.alpinelinux.org/cgit/aports/tree/testing/gitea/gitea.confd#n1 typo, you have written GETEA :) 2017-01-30 13:50:57 :) 2017-01-30 13:51:24 and as a tiny hint, http://git.alpinelinux.org/cgit/aports/tree/testing/gitea/gitea.ini#n14 this option is used only for pgsql 2017-01-30 13:53:05 too late, i just pushed a fix 2017-01-30 13:54:09 hum... need to try the package but i'm pretty sure http://git.alpinelinux.org/cgit/aports/tree/testing/gitea/APKBUILD#n65 is missing directories. 2017-01-30 13:54:34 it does 2017-01-30 13:54:37 at least it should also include the "options" directory 2017-01-30 13:54:42 correct 2017-01-30 13:54:52 but good thing is, i fixed that bin helper 2017-01-30 13:54:56 cant remember its name 2017-01-30 13:55:05 so we could pack everything in a single file 2017-01-30 13:55:40 ah you moved the options into the custom path 2017-01-30 13:56:01 i think i didnt touch it yet? 2017-01-30 13:56:03 or i did? 2017-01-30 13:56:09 http://git.alpinelinux.org/cgit/aports/tree/testing/gitea/APKBUILD#n63 2017-01-30 13:56:40 yes i did but i remember i still had an issue 2017-01-30 13:57:11 mosez: i didnt spend time on it as it does not work on alpine's repo 2017-01-30 13:57:21 its just too slow on main directory listing 2017-01-30 13:57:24 and don't add the bindata build flag <1.1.0, currently the custom templates and custom public files are broken. 2017-01-30 13:57:32 and the releases page is still broken. 2017-01-30 13:57:54 yeah, that's something we try to resolve before 1.1.0 release 2017-01-30 13:58:17 especially the release page 2017-01-30 13:58:20 i thought that releases fix was already commited? 2017-01-30 13:58:40 a fix yes... but it can still be slow, or i'm just not up2date for that 2017-01-30 13:58:44 but it seems all releases/tags are still listed on the first page (including the pager) 2017-01-30 13:59:11 so the pager has no use 2017-01-30 13:59:18 but the aports repo is indeed a good example repo to test everything :D 2017-01-30 14:00:28 aports has a horizontal design, so it will cause issues the way you handle it like you do now. 2017-01-30 14:01:14 yeah gitea will get lots of refactorings because many things are simply totally poor performancewise 2017-01-30 14:01:33 you will get it right, doesnt look that hard to do it. 2017-01-30 14:01:55 if its caching or just disable some lookups 2017-01-30 14:01:55 first release have been nearly at all bug fixing, same for 1.0.1 2017-01-30 14:02:18 overall its pretty fast 2017-01-30 14:02:30 we have to find a balance between proper code and intelligent caching 2017-01-30 14:09:45 mosez: atleast you guy's are looking at the performance issues, seems upstream didnt really care about it at all. 2017-01-30 14:15:11 clandmeter: one of the many reasons for the fork :P 2017-01-30 14:16:01 unknwon clearly said that gogs is his pet project and he is doing on it what he wants and what he prefers. 2017-01-30 17:13:03 ncopa: is there any objection to doing an abuild 2.30 release some time this week? 2017-01-30 17:22:34 kaniini, i'd like to fix apkbuild-cpan too for check and any perl warnings it has 2017-01-30 17:22:40 should not take too much time 2017-01-30 17:23:04 ACTION wants to start doing the testsuite conversions is why ;) 2017-01-30 17:23:36 :) 2017-01-30 17:23:45 i'll try to fix it tm 2017-01-30 17:33:42 hmm 2017-01-30 17:34:13 i think initramfs should use framebuffer console unconditionally when booting from UEFI. 2017-01-30 17:34:49 i suspect alpine boots fine on my laptop, but cannot be useable because of the lack of a framebuffer console 2017-01-30 17:35:44 what if I have a server without a console and boot from UEFI? 2017-01-30 17:35:55 (this is not just hypothetical, I may have this soon) 2017-01-30 17:36:13 okay, let me rephrase 2017-01-30 17:36:20 if EFI framebuffer is present, it should use it 2017-01-30 17:36:28 if EFI framebuffer is not present, it shouldn't 2017-01-30 17:36:31 (obviously) 2017-01-30 17:36:31 makes sense :) 2017-01-30 17:37:03 i think it boots in text mode which winds up doing nothing useful since there is no text mode in pure EFI 2017-01-30 17:37:28 but i'm not sure 2017-01-30 17:38:05 kaniini, it's tricky, there are setups where fbcon is unwanted 2017-01-30 17:38:30 but yes, in most setups it makes sense. we have a bug about it and we did make eudev load fbcon now be default when fb0 exists 2017-01-30 17:38:44 odd 2017-01-30 17:38:54 then i'm not sure why i don't get a console on my machine 2017-01-30 17:39:11 fedora boots just fine, for example 2017-01-30 17:40:02 is it dual-gpu? 2017-01-30 17:40:35 no, just intel 2017-01-30 17:40:52 it's a piece of junk atom laptop 2017-01-30 17:40:53 kaniini, see also http://bugs.alpinelinux.org/issues/6490 2017-01-30 17:41:26 see also http://bugs.alpinelinux.org/issues/6723 2017-01-30 19:53:39 Hi 2017-01-30 19:53:56 I'm trying to install alpine linux on VMWare Fusion 2017-01-30 19:54:53 The installation fails every time while complaining about inability to mount a boot media (3.5.1 ISO downloaded from the web site) 2017-01-30 19:55:07 I want to try and build some package for alpine 2017-01-30 19:55:24 themik1024: sounds bad, i'll test to boot it in vmware fusion 2017-01-30 19:55:53 themik1024: what host machine is it? 2017-01-30 19:56:03 i have osx here 2017-01-30 19:56:11 Fusion: 8.5.3 2017-01-30 19:56:21 macOS 10.12.3 2017-01-30 19:56:57 themik1024: have you tried docker for mac? 2017-01-30 19:57:05 probably the easiest way 2017-01-30 19:57:10 Yes. That works 2017-01-30 19:57:19 Hmmm 2017-01-30 19:57:22 Right 2017-01-30 19:57:24 which iso did you download? 2017-01-30 19:57:36 I can just create an image with all the packages 2017-01-30 19:57:46 both vanilla and regulae 2017-01-30 19:57:48 both vanilla and regular 2017-01-30 19:57:51 None worked 2017-01-30 19:58:13 both vanilla and standard, sorry 2017-01-30 19:58:26 <_ikke_> ncopa: Should I rebuild python with debugging to get the trace for curses? 2017-01-30 19:58:47 _ikke_: possibly 2017-01-30 19:59:02 <_ikke_> ok 2017-01-30 19:59:22 _ikke_: if you can generate a core, then you do: gdb --core /path/to/core /usr/bin/python 2017-01-30 19:59:26 and run bt 2017-01-30 19:59:43 if problem happens in your module, then you should have a stacktrace 2017-01-30 19:59:56 at least give an indication where things goes wrong 2017-01-30 20:03:15 <_ikke_> Where can I find the coredump? (do I need to enable them)? 2017-01-30 20:03:39 ulimit -c unlimited 2017-01-30 20:03:53 <_ikke_> thanks 2017-01-30 20:07:43 @ncopa: Thank you for the idea. The Docker way works 2017-01-30 20:59:35 https://github.com/OpenRC/openrc/issues/116 2017-01-30 21:00:51 as I told awilfox, I think straying from simple Makefiles for a low-level piece of infrastructure such as a service manager (which OpenRC is supposed to be) leads to bootstrapping nightmares 2017-01-30 21:01:26 if you build from scratch and want OpenRC, now you need to build Python first, oh happy days 2017-01-30 21:02:04 and Ninja, and Meson 2017-01-30 21:02:15 yeah, not even mentioning them 2017-01-30 21:02:49 so my instinct is to go to there and nuke the proposal with extreme prejudice, but I'm being told it wouldn't be well received 2017-01-30 21:12:23 in terms of alpine, it would just be a $makedepends_host dependency. in general, i do not like the idea, but it would not be fatal. 2017-01-30 21:15:05 as long as there's no circle, no amount of spaghetti dependencies is fatal. Doesn't mean it's a good thing. 2017-01-30 21:15:31 and circles can happen really fast without anyone noticing. 2017-01-30 21:15:42 I mean cycles. 2017-01-31 00:07:03 skarnet: you should like that though -- it gives reason to look more intensely at s6 2017-01-31 00:25:25 I would love that if I was ready to do the service manager conversion work 2017-01-31 00:25:47 well 2017-01-31 00:26:06 williamh is kind of stupid because alpine used to run an ancient version of openrc with a ton of patches on top 2017-01-31 00:26:16 and it was *me* who did most of the work of cleaning that up 2017-01-31 00:27:03 if meson breaks our ability to bootstrap the distribution (due to it not supporting crossbuilds of openrc), then we would be forced to keep openrc back 2017-01-31 00:27:21 he can certainly ask others opinion, but that's the policy 2017-01-31 00:28:07 oh, I +100 the opinion that this move is a very bad idea 2017-01-31 00:28:21 we have held openrc back in the past because they broke us 2017-01-31 00:28:26 why would we not do so in the future 2017-01-31 05:56:39 <_ikke_> ncopa: I'm getting a coredump, but the bt does not have any debug symbols 2017-01-31 05:57:53 <_ikke_> ncopa: ah, should've specified py3 2017-01-31 06:02:16 <_ikke_> https://hastebin.com/raw/unerivamec 2017-01-31 08:32:46 ncopa: the downloads page seems not up to date 2017-01-31 09:43:47 clandmeter: looks up2date to me? it shows 3.5.1 release 2017-01-31 09:43:58 looks closer :) 2017-01-31 09:44:03 last of the releases 2017-01-31 09:44:07 arm related i think 2017-01-31 09:44:26 you're right 2017-01-31 09:44:28 sigh 2017-01-31 09:44:54 ok i'll fix that 2017-01-31 09:56:06 we'll have to redo the css when adding new arch :| 2017-01-31 09:56:26 yeah, i dont like that 2017-01-31 09:56:41 there was alot that broke when i did 3.5.0 release 2017-01-31 09:57:20 i'd also like to have the description of the iso flavor in aports/scripts/mkimg.* 2017-01-31 09:57:52 we should also make it possible to specify the order of them 2017-01-31 09:59:52 _ikke_: it looks like the problem is in python3 itself 2017-01-31 10:00:32 <_ikke_> Hmm, ok. Any way to get more information from python/? 2017-01-31 10:07:55 could possibley be a bug in ungetmouse too 2017-01-31 10:08:08 seems like the stack canary is overwritten 2017-01-31 10:08:17 the only variable there is MEVENT 2017-01-31 10:08:36 <_ikke_> The sentinel? 2017-01-31 10:08:44 yes 2017-01-31 10:09:10 something in PyCurses_UngetMouse is corrupting memory 2017-01-31 10:09:12 well 2017-01-31 10:09:28 memory gets corrupted in that function 2017-01-31 10:09:37 can you run "info threads" in gdb? 2017-01-31 10:09:51 it could be a race as well 2017-01-31 10:18:14 <_ikke_> Ok, I will soon 2017-01-31 11:33:48 <_ikke_> ncopa: info threads only returns 1 thread 2017-01-31 11:40:53 ok so its not a race 2017-01-31 11:40:57 good 2017-01-31 11:40:59 makes it simpler 2017-01-31 11:57:48 Hello - I'd like to create a patch/fix for the exim package in testing. I've worked out the change (http://bugs.alpinelinux.org/issues/6775#note-2) I'm just not sure where to take it next 2017-01-31 12:01:15 Do I just mail alpine-aports? 2017-01-31 12:02:23 ashb, yes, send "git format-patch" type patch to alpine-aports; github pull requests are also accepted 2017-01-31 12:02:39 Is one easier or preferred? 2017-01-31 12:03:05 if you have preference use that 2017-01-31 12:03:40 fair. Thanks 2017-01-31 12:19:54 Should I bump the pkgrel in the same commit? Or is that handled by some build process? 2017-01-31 12:20:10 ashb: you need to bump the pkgrel 2017-01-31 12:20:18 in your commit 2017-01-31 12:42:17 Sent (hopefully?) I used my gmail and haven't subscribed so I guess it'll need moderating 2017-01-31 13:04:30 <^7heo> gogs updated to latest. 2017-01-31 13:04:35 <^7heo> (and locally tested) 2017-01-31 13:05:26 <^7heo> actually, not. 2017-01-31 13:05:31 <^7heo> for some reason it does not work. 2017-01-31 13:20:09 _ikke_: it would be nice if we could try narrow down the pythin curses issue. i wonder if we could make a minimal python script that triggers the error 2017-01-31 13:34:45 <^7heo> ok found out. 2017-01-31 13:48:39 <^7heo> ncopa: what's the policy when a package needs a hardcoded home? 2017-01-31 13:49:01 <_ikke_> ncopa: I can try to work on that 2017-01-31 14:25:28 kaniini, i think i got the apkbuild-cpan updated for 'abuild check', you may want to check my recent abuild commits 2017-01-31 16:29:08 hmm, the cpan apkbuilds don't return error? 2017-01-31 17:02:06 <_errm_> I just submitted a patch to gsl to make the gsl bindings for ruby work … https://github.com/alpinelinux/aports/pull/817 2017-01-31 17:02:41 <_errm_> I would appreciate any comments about whether this is an appropriate way to fix this issue ... 2017-01-31 17:20:39 Hello, I’m trying to use bootstrap.sh to create an arm cross-compiler, is it possible? What option should I give to the script? 2017-01-31 17:21:09 (and is it the best way to do this) 2017-01-31 17:34:48 ugh, chromium update is totally broke 2017-01-31 17:35:17 <^7heo> ncopa: same for gogs and I'm on it. 2017-01-31 17:36:13 What are the available target arch for bootstrap.sh? 2017-01-31 17:36:13 i wonder if paxmark works 2017-01-31 17:37:35 paxmark should be working, we just removed the old paxctl calls which the kernel does not support anyway 2017-01-31 17:38:09 kaniini: i know. i had to upgrade the build server kernels (which was pretty old anyway) 2017-01-31 17:38:27 but there was some issue with fakeroot 2017-01-31 17:38:35 humm 2017-01-31 17:38:45 ncopa-desktop:~$ attr -l /usr/lib/chromium/chrome 2017-01-31 17:38:45 ncopa-desktop:~$ 2017-01-31 17:39:00 odd 2017-01-31 17:39:12 i wonder if it got installed without attrs 2017-01-31 17:39:46 yes, that what it looks like 2017-01-31 17:40:15 ncopa-desktop:~$ sudo attr -s pax.flags -V m /usr/lib/chromium/chrome 2017-01-31 17:40:15 Attribute "pax.flags" set to a 1 byte value for /usr/lib/chromium/chrome: 2017-01-31 17:40:15 m 2017-01-31 17:40:19 and now it works 2017-01-31 17:40:23 so something is broken 2017-01-31 17:40:40 someone complained about the same with firefox-esr on #musl 2017-01-31 17:41:20 so, we do paxmark in package() function (in fakeroot) 2017-01-31 17:41:31 for some reason the attr does not make it to the install 2017-01-31 17:41:45 i wonder if fakeroot is not supporting attrs correctly 2017-01-31 17:41:53 since it proxies some syscalls to suid 2017-01-31 17:41:54 we should check if the xattr is included in the .apk 2017-01-31 17:42:04 i think it *should* work 2017-01-31 17:42:17 ACTION checks bugs.debian.org 2017-01-31 17:42:24 well, it will not set the mark on the underlying filesystem 2017-01-31 17:42:39 but it should pick it up 2017-01-31 17:44:14 http://dpaste.com/0PCW1DV 2017-01-31 17:44:34 seems like it works as it should in fakeroot 2017-01-31 17:44:38 maybe its a bug in apk? 2017-01-31 17:44:53 could be 2017-01-31 17:45:54 is there some way to list xattrs in tar? 2017-01-31 17:46:46 When listing the archive, if both `--xattrs' and `--verbose' options are given, files that have extended attributes are marked with an asterisk following their permission mask. For example: 2017-01-31 17:46:48 2017-01-31 17:46:50 -rw-r--r--* smith/users 110 2016-03-16 16:07 file 2017-01-31 17:46:52 When two or more `--verbose' options are given, a detailed listing of extended attributes is printed after each file entry. Each attribute is listed on a separate line, which begins with two spaces and the letter `x' indicating extended attribute. It is followed by a colon, length of the attribute and its name, e.g.: 2017-01-31 17:46:54 2017-01-31 17:46:56 -rw-r--r--* smith/users 110 2016-03-16 16:07 file 2017-01-31 17:46:58 x: 7 user.mime_type 2017-01-31 17:47:00 x: 32 trusted.md5sum 2017-01-31 17:47:00 ncopa-desktop:~$ gunzip -c chromium-56.0.2924.76-r0.apk | strings | grep pax.flags 2017-01-31 17:47:00 ncopa-desktop:~$ 2017-01-31 17:47:05 tar -l --xattrs --verbose 2017-01-31 17:47:36 not there 2017-01-31 17:47:48 so it happens on .apk creation 2017-01-31 17:47:52 will have to look at that tm 2017-01-31 17:47:54 need to go 2017-01-31 17:47:57 see you 2017-01-31 17:48:15 <_ikke_> bye 2017-01-31 17:48:41 if someone asks, the workaround for now is: attr -s pax.flags -V $flag /usr/bin/$program 2017-01-31 17:48:57 where đlag is for exampl 'm' 2017-01-31 17:49:04 in the chromium case 2017-01-31 17:49:12 check the APKBUILD what paxmark says it should be 2017-01-31 17:49:18 cu 2017-01-31 18:59:16 What are the available target arch for bootstrap.sh? 2017-01-31 18:59:19 in aports 2017-01-31 18:59:29 I’m trying to use bootstrap.sh to create an arm cross-compiler, is it possible? What option should I give to the script? 2017-01-31 19:00:18 armhf 2017-01-31 19:37:27 ncopa: i am sketching out a replacement to openrc (sometime down the line) anyway, that builds an alpine-like interface ontop of s6/s6-rc 2017-01-31 19:46:14 kaniini: I tried that one, but the script fails 2017-01-31 19:46:29 what happens? it should work 2017-01-31 19:47:04 https://dpaste.de/jjM6/raw 2017-01-31 19:47:07 kaniini: ^ 2017-01-31 19:47:10 This is the output 2017-01-31 19:47:47 binutils APKBUILD may need fixing. hmm 2017-01-31 19:47:57 this is a good question for fabled i guess 2017-01-31 19:48:01 Oh :/ 2017-01-31 19:49:22 I think it’s a bit too complicated as a first time for me to fix the APKBUILD /o\ 2017-01-31 19:51:19 armhf was bootstrapped before that 2017-01-31 19:53:59 afaik bootstrap.sh requires whole target-triplet 2017-01-31 19:54:59 it takes either 2017-01-31 19:55:14 you could try supplying a triplet, it may help 2017-01-31 19:55:23 for armhf it would be armv6-alpine-linux-muslgnueabihf 2017-01-31 19:55:54 that is certainly a mouthful ;) 2017-01-31 19:56:10 xentec: the other archs have aliases for this. i'll look into adding an alias for armhf 2017-01-31 19:57:14 please do :) 2017-01-31 20:18:17 ashb: Seems like exim's Makefile template (src/EDITME) changed style, so it needs to remade it to a new line-by-line 2017-01-31 20:19:23 vakartel: oh? That file still works 2017-01-31 20:19:36 (thanks for taking a look btw!) 2017-01-31 20:20:24 well, it compiled anyway 2017-01-31 20:20:50 yes, but as I see, all libs described in LDFLAGS=, not in LOOKUP_PGSQL_LIBS= and so on 2017-01-31 20:21:38 https://github.com/Exim/exim/blob/master/src/src/lookups/Makefile#L24-L25 is where it uses LOOKUP_PGSQL_LIBS 2017-01-31 20:21:50 libs in LDFLAGS etc is for if you aren't building them as modules from what I could tell 2017-01-31 20:22:45 sorry... described in LOOKUP_LIBS= (LOOKUP_LIBS=-lldap -llber -lmysqlclient -lpq -lgds -lsqlite3) 2017-01-31 20:23:08 Ah yes, I asked about that on in #exim - let me find what they said 2017-01-31 20:23:20 *one 2017-01-31 20:24:39 it's used in the final link stage. see build*/Makefile (once you've tried to do a build) 2017-01-31 20:24:55 So it's extra libs needed for lookup plugins only used if you aren't using dynamic 2017-01-31 20:26:10 yeah -- they are only used in the link step of the main `exim` binary 2017-01-31 20:26:12 aha... I just made a test build and LOOKUP_LIBS has zero effect ... (( 2017-01-31 20:27:01 They'd only help if we set LOOKUP_PGSQL=yes (instead of `2` to build a module. Why did they pick 2 rather than "module" as the value I wonder) 2017-01-31 20:30:16 Youbi: if all you want is a cross-compiler, not something alpine-specific you should check out musl-cross-make: https://github.com/richfelker/musl-cross-make 2017-01-31 20:58:35 neat, that looks a bit simpler compared to the musl-cross stuff 2017-01-31 21:14:24 ashb: think, "yes","no" are the names for 1,0 in makefiles. so for module build they use a different unnamed value 2 2017-01-31 21:14:33 ahh 2017-01-31 21:16:39 ashb: just dig in lookups/makefile, and seems your variant is the only working one. 2017-01-31 21:18:34 ashb: 2 things to do :: remade LOOKUP_SQLITE_PC to LIBS/INCLUDES and add linux-headers to makedepends in APKBUILD (it shows warning linux/ip.h not found without it) 2017-01-31 21:19:24 ah okay, yes I didn't look at sqlite 2017-01-31 22:33:12 vakartel: I'd still need to use _LIBS form rather than _PC (which is handled, but I need that -Wl option) 2017-01-31 22:33:33 I should do say: LOOKUP_SQLITE_LIBS=-Wl,--no-as-neeed $(shell pkg-config --libs sqlite3) 2017-01-31 22:33:34 Thoughts? 2017-01-31 22:34:06 yep 2017-01-31 22:37:51 with _PC form you can't add additional linker flags, the only from pkg-config are used 2017-01-31 22:40:31 I can do "LOOKUP_SQLITE_LIBS=-Wl,--no-as-needed $(shell pkg-config --libs sqlite3)" though 2017-01-31 22:41:28 Do you think that is clearer, or should I just leave it as hard-coded values? 2017-01-31 23:15:51 think -lsqlite3 would be enough 2017-01-31 23:25:32 ok