2017-03-01 01:57:37 yGweSm1OzVHe: Yes, they can. But here on alpine they can't. 2017-03-01 07:01:42 fabled : I just sent another kernel config for s390x, if you have any complain I am here :) 2017-03-01 07:01:51 v1 was pretty bloat ... 2017-03-01 07:05:44 tmh1999, ok... i was planning to do diff against x86_64 config and try to make the generic options the same; there was some differences mostly 'y' vs 'm' 2017-03-01 07:08:20 thanks 2017-03-01 07:08:30 take care when it comes to s390x-specific 2017-03-01 07:10:51 yeah, many of the differences are valid. it was not too many differences, so i might just go ahead and apply it 2017-03-01 07:12:20 I have cross-checked from Alpine, SUSE Fedora and Ubuntu 2017-03-01 07:12:24 guess that's my best 2017-03-01 07:18:24 fabled : is there any command/options in abuild/apk-tools to remove old packages in local repo, after pulling new aports & building ? 2017-03-01 07:18:39 ok.thanks a lot! i'll try to get it checked today. if there's only minor differences, do you prefer me to push with changes, or check with you the differences first? 2017-03-01 07:18:58 tmh1999, abuild cleanoldpkg ? 2017-03-01 07:22:49 fabled : awesome ! 2017-03-01 07:23:02 never came across this option ../ 2017-03-01 07:23:54 fabled : it's no hurry on the kernel config. 2017-03-01 07:24:21 you do whichever way fit your way and schedule 2017-03-01 08:07:49 heh, leo-unglaub assumes he can rebuild world in 3 hours with his new AMD processor. I wonder after how many days he will just give up :) 2017-03-01 09:10:03 fabled: Good evening! Take a look at the latest couple entries at http://termbin.com/pqrb for an example of how the features allow flexible configuration and simplify the profiles. 2017-03-01 12:18:32 you people are cordially invited to our hackercamp in hungary: https://camp.hsbp.org 2017-03-01 17:54:31 out of curiosity, is there anything akin to nixos hydra to let you self host packages you create? 2017-03-01 17:54:57 or rather do a build then host out the build apks 2017-03-01 17:55:35 or should i just hack some CI stuff around doing that on my own? 2017-03-01 18:00:21 mitchty: no, but it’s not so hard to make your own https://github.com/jirutka/user-aports#how-to-use 2017-03-01 18:00:27 eh, this https://github.com/jirutka/user-aports#how-to-setup-your-own-repository 2017-03-01 18:00:58 cool i'll have a look 2017-03-01 18:02:15 that reminds me that i should ask fabled if/how it can be simplified, I mean mainly to avoid using SSHFS 2017-03-01 18:03:40 i'm asking mostly as i'm testing getting generated apkbuilds for haskell stuff working, and well, there is enough crazy in that to warrant the effort 2017-03-01 18:06:20 and on that note, for haskell packages would it be OK to require that every haskell compiled port has to ensure that it builds/doesn't interfere version wise with any other shared package? 2017-03-01 18:06:53 because needless to say, you can end up with package a requiring version b of some package, and package c requiring version b' 2017-03-01 18:07:19 note, thats only for building really 2017-03-01 18:09:30 Haskell doesn’t use shared libs, right? 2017-03-01 18:09:40 it can, but not by default 2017-03-01 18:09:56 the default is to statically link in haskell packages, dynamic link stuff like c libraries 2017-03-01 18:10:13 so when you install A that depends on B, C, …, Z, then B, C, …, Z are build into the output binary, there are not standalone at fs? 2017-03-01 18:10:21 correct 2017-03-01 18:10:22 so the same as in Rust 2017-03-01 18:10:56 well its actually "worse", you can technically have both versions in the same binary, but thats getting into some stuff that isn't relevant to packaging :) 2017-03-01 18:10:59 then I’d not create pkg for every dependency and just download them in the abuild A 2017-03-01 18:11:37 otherwise it’d be total insanity with no benefits at all 2017-03-01 18:11:51 so i was starting down the road of using the arch tool that generates pkgbuilds for dependencies 2017-03-01 18:12:19 that way instead of rebuilding dependencies in each package you'd only need to build it once in general 2017-03-01 18:12:36 but suppose I should ask how we'd want to do it 2017-03-01 18:12:58 have to map how cabal does dependency resolution to what apk has, or avoid it entirely 2017-03-01 18:13:36 I’m thinking about adding support for third-party deps into abuild; e.g. cargo:rotor-1.2.3, this would be basically just a shortcut for defining full source URL of rotor in sources 2017-03-01 18:14:39 b/c it doesn’t make any sense to create pkg for every dependency and replicate all the work that cargo/cabal/… do, dealing with conflicting versions etc. 2017-03-01 18:14:41 i'd need to think about that a bit, but it would probably be similar to rust in that regard, as long as it bridges it should be ok 2017-03-01 18:15:00 when such dependencies do not end up on filesystem separately 2017-03-01 18:15:08 these are just build dependencies 2017-03-01 18:15:17 well, in haskell its going to get even more ML..y in the next release where modules are generic 2017-03-01 18:16:07 and you can have module A be honored by package B, but only for Integers, and then module C be honored by package D for ByteStrings 2017-03-01 18:16:18 heh 2017-03-01 18:16:29 https://ghc.haskell.org/trac/ghc/wiki/Backpack 2017-03-01 18:16:52 http://blog.ezyang.com/2016/10/try-backpack-ghc-backpack/ is a better intro though 2017-03-01 18:17:50 anyway that works if we don't want to clutter up abuild with a ton of haskell-* packages 2017-03-01 18:18:12 i was just thinking of improving the build time of some things with some shared packages 2017-03-01 18:18:28 its really noticeable on arm as an example 2017-03-01 18:19:06 yeah, I imagine that you can combine both approaches – make abuild for haskell dependency when it makes sense 2017-03-01 18:20:34 hm, maybe it’d be better to add this shortcut to source, not to makedepends 2017-03-01 18:21:10 that’d be closer to what it really do and also less controversial 2017-03-01 18:21:19 well would it be possible to punt to say cabal/cargo and then cache the build? 2017-03-01 18:21:54 what do you mean? 2017-03-01 18:22:12 well say some package requires A B C as dependencies, but only for build 2017-03-01 18:22:33 then package foo requires A B and say D for building 2017-03-01 18:22:54 you can build A B C as if they were virtual packages, and have a way to cache their inputs as an apk you can makedepends on 2017-03-01 18:23:19 that way package foo could reuse the build of A and B if possible 2017-03-01 18:23:34 hopefully i explained that well enough 2017-03-01 18:23:42 something like ccache…? ;) 2017-03-01 18:23:59 pretty much, only that works with not C >.< 2017-03-01 18:24:31 we could cheat by storing some of this in a common location for cabal/ghc 2017-03-01 18:25:09 not sure if it’s a good idea… 2017-03-01 18:25:41 well for cabal its not a big deal, we could use the new-build stuff which has nix style dependency tracking 2017-03-01 18:26:05 aka if you build a package with ffi for example, and need it without for another, both can coexist at once 2017-03-01 18:26:16 if ffi's version changes also no big deal 2017-03-01 18:27:01 anyway, cabal/ghc is a bit flexible 2017-03-01 18:27:14 we’ve already discussed this, leaving FHS, there are few ppl including me who wants to try it 2017-03-01 18:27:38 it’d solve many problems 2017-03-01 18:27:51 nix is really fun though, but in this regard its more build inputs 2017-03-01 18:28:03 who do we have to kidnap and waterboard until they say yes to ditching FHS? 2017-03-01 18:29:02 I’d prefer just to talk… 2017-03-01 18:29:22 I can provide a bathtub. (My budget goes to computer stuff, my torture tools are amateur.) 2017-03-01 18:29:53 1th we must agree on it, 2nd someone must modify/develop tooling 2017-03-01 18:30:13 you're always ruining all the fun 2017-03-01 18:30:56 /usr/bin/env and /bin/sh is the only FHS stuff in nixos iirc, its not a whole lot 2017-03-01 18:31:05 those make sense 2017-03-01 18:31:46 in either case for build inputs let me try doing a few tests with compiling all build inputs into say /var/tmp/ghc-cache and see what happens 2017-03-01 21:59:01 Hello. I'm trying to make packages for https://wireguard.io VPN. It consist of userland tools and kernel module. So, I wonder if there is a tutorial on packaging kernel modules for Alpine somewhere? 2017-03-01 22:01:10 IIUC, Alpine doesn't use dkms to build modules. So I guess, I should package binary module for current kernel in edge. Is it correct? 2017-03-01 22:06:03 fabled : when "$CBUILD" != "$CTARGET", we have pkgname="go-bootstrap" instead of go, so why do we provide go-bootstrap=$pkgver-r$pkgrel at the same time ? 2017-03-01 22:06:29 that means, we will have go-bootstrap-$pkgver-r$pkgrel, and go-bootstrap=$pkgver-r$pkgrel 2017-03-01 22:06:34 care the - and = 2017-03-01 22:16:57 Right, I remove the first provides="...", then it works 2017-03-01 22:17:01 I will send a patch 2017-03-01 23:22:02 linux-vanilla in edge is behind 3.5 branch https://pkgs.alpinelinux.org/packages?name=linux-vanilla&branch=&repo=&arch=&maintainer= 2017-03-01 23:32:39 fabled: there’s something wrong, abuild does not detect dynamically linked libs for lua pkgs anymore 2017-03-01 23:47:46 hm, this didn’t work quite a while… lua5.1-lunix in v3.5 does not depend on lua5.1 2017-03-01 23:48:28 scanelf does not find dependency on lua lib, but ldd does; probably I just don’t understand how it works 2017-03-02 00:50:34 laskin: there is already an APKBUILD submitted pending someone sponsoring it for wireguard, but it has problems 2017-03-02 01:23:03 laskin: you can see it here, 2017-03-02 01:23:06 laskin: https://github.com/alpinelinux/aports/pull/708 2017-03-02 01:44:21 fabled: When you get a chance, please take a look at http://termbin.com/ey7v -- that is the last major restructuring to modularize and cleanup the profile system. The individual profiles could probably use review and tweaking. 2017-03-02 01:49:08 supporting various bootloaders is now very straightforward 2017-03-02 02:40:13 ls 2017-03-02 02:40:20 Oops, wrong term ;) 2017-03-02 04:06:24 Who's the local guru on the bootloader situation for various architectures? 2017-03-02 04:32:18 not me :D 2017-03-02 04:46:15 kaniini - finishing off wireguard now 2017-03-02 04:47:33 sick 2017-03-02 04:47:44 let me know, i will send it to the builders 2017-03-02 04:52:26 BitL0G1c: why tcp-wrappers by the way 2017-03-02 04:54:54 kaniini - stunnel can be built with tcp wrapper support - I also noticed there is an nginx module for tcpwrappers 2017-03-02 04:56:27 if it works ok with just ip addresses - eventually I could replace the dns functionality that is missing - I saw a tiny dns library implemented in one c file 2017-03-02 04:57:44 but do we want to do that 2017-03-02 04:57:56 that is my question really 2017-03-02 04:58:14 red hat still uses tcp wrappers 2017-03-02 05:06:20 kaniini - wireguard rebased into 3 x APKBUILD 2017-03-02 05:09:11 ok 2017-03-02 05:09:16 well 2017-03-02 05:09:28 regarding tcp wrappers; redhat also uses systemd -- should we use systemd? 2017-03-02 05:09:37 hahahahaha no 2017-03-02 05:09:57 systemd is why i stopped using debian 2017-03-02 05:10:09 but redhat is doing it 2017-03-02 05:10:23 (see how this isn't a supporting argument?) 2017-03-02 05:10:55 my understanding is that tcp wrappers has very little security benefit and has had security CVEs in the past 2017-03-02 05:11:21 ok 2017-03-02 05:11:33 personally, that would lean me to not want to be the rabbit who put this in alpine 2017-03-02 05:12:51 looking at one other package first and then i will put wireguard in 2017-03-02 05:13:02 ok 2017-03-02 05:17:50 dear god this package is fucked (not wireguard) 2017-03-02 05:17:55 completely wrong 2017-03-02 05:20:48 last cve on tcpwrappers was in 2007 - ubuntu use it too 2017-03-02 05:27:45 but why should *alpine* use it 2017-03-02 05:34:11 BitL0G1c: i think you forgot to update the grsec version, as it is still building tools :/ 2017-03-02 05:49:28 oops 2017-03-02 05:58:49 I think I need to make a new pr for wireguard - local copies are ok - remote not 2017-03-02 05:58:57 okay 2017-03-02 05:59:03 let me know the PR number and i'll review 2017-03-02 06:07:41 tmh1999, re go pkgname, the intention is to ship go-bootstrap when compiled because it is limited package, so only native compiling provides the full package. however, to be able to recompile itself the native compiled version needs to provide go-bootstrap too. this makes the real go compiler to build itself on further rebuilds 2017-03-02 06:10:07 Does anyone know off hand how to tell get fakeroot to run one instance of faked for multiple instantiations of fakeroot? 2017-03-02 06:10:26 Or do I get to hack the script to make it work? 2017-03-02 06:11:32 :D 2017-03-02 06:14:19 Or, to be even more twisted, can I LD_PRELOAD fakeroot for just the subshell I need it in, keeping the same FAKEROOTKEY? 2017-03-02 06:22:08 kaniini - see https://github.com/alpinelinux/aports/pull/948 2017-03-02 06:26:01 looks good 2017-03-02 06:27:17 How do I use the subpackages? 2017-03-02 06:27:38 I don't understand the "$pkgname-scrips-py::noarch" 2017-03-02 06:27:45 I mean the part, what does that do? 2017-03-02 06:28:00 function to run 2017-03-02 06:28:09 kaniini: What if it is not there? 2017-03-02 06:28:18 then it may be a builtin :) 2017-03-02 06:28:32 Oh 2017-03-02 06:28:48 kaniini: %3089 that's what I send, weird. It have no name. 2017-03-02 06:30:21 fabled : I understand the steps of recompiling natively. It's just that I can't find other/better solution for the conflict error I got from $ apk install go-bootstrap when trying to natively compile the full go compiler 2017-03-02 06:32:34 fabled : do you have the same conflict problem on aarch64 ? 2017-03-02 06:32:38 Looks like rust doesn't build yet on musl T_T 2017-03-02 06:32:40 But go does 2017-03-02 06:33:52 algitbot: Hello, where is your source. 2017-03-02 06:35:32 kaniini: If a package has scrips (python, perl...), should it be included with the main packages as well? 2017-03-02 06:35:35 Like john 2017-03-02 06:35:49 it depends 2017-03-02 06:35:57 if they are optional enhancements, may make sense to split them 2017-03-02 06:36:25 I think I should split them 2017-03-02 06:36:46 kaniini: Then for a all in one install do I use separate package like alpine-sdk or just set the version? 2017-03-02 06:37:09 Like john-vjumbo1 vs john-vr1 2017-03-02 06:37:50 im not sure what youre trying to do 2017-03-02 06:38:20 kaniini: Like I want a package to include all the stuff john currently in. 2017-03-02 06:38:22 have* 2017-03-02 06:38:30 I want to keep john minimal 2017-03-02 06:38:51 And then like apk install john-jumbopack to install all those stuff 2017-03-02 06:38:51 you could make a subpackage that pulls everything else in i guess 2017-03-02 06:38:55 A 2017-03-02 06:38:59 Oka 2017-03-02 07:13:44 fabled : https://wiki.alpinelinux.org/wiki/APKBUILD_Reference#provides. APKINDEX shows : P:go, V:1.7.4-r2, A:s390x, ..., D:binutils gcc so:libc.musl-s390x.so.1, p:go-bootstrap=1.7.4-r2. This APKINDEX is the result of natively building go, using cross-built go-bootstrap with the patch I sent. 2017-03-02 07:14:08 I think it matches just fine the description in the APKBUILD references 2017-03-02 07:28:12 $ apk search bootstrap also show go package 2017-03-02 08:38:43 Okay, I think this will work for now for the fakeroot utility: http://termbin.com/cfin 2017-03-02 08:39:27 I'd love it if someone who is a bit more familiar with how fakeroot works could take a look at it and let me know what needs to be done to make it solid for general use. 2017-03-02 08:40:54 The idea being that you can wrap a section in a script that needs to be in the fakeroot with fkrt_enable/fkrt_disable at will. 2017-03-02 08:43:24 It shouldn't take much more work to make it handle multiple distinct fakeroot environments that can be selected among at will if needed. 2017-03-02 09:52:32 BitL0G1c: thank you for wireguard. Using it right now, works perfectly. 2017-03-02 09:58:50 ncopa: what are the implications of vsyscall=emulate ? and why it is required for running centos6 docker images? 2017-03-02 09:59:45 maybe it is required for other docker images, ... but not needed for centos7 ones, nor debian{7|8} 2017-03-02 10:01:25 tru_tru: http://stackoverflow.com/questions/19938324/what-are-vdso-and-vsyscall 2017-03-02 10:02:32 vsyscall has some security implication, which why i think most distros disables it by default 2017-03-02 10:02:43 centos6 uses ancient glibc that uses vsyscalls 2017-03-02 10:03:21 i dont know what version of glibc centos7 or if it needs vsyscalls 2017-03-02 10:03:36 you should probably ask in some centos channel 2017-03-02 10:05:48 some binaries that links to static dietlibc has also the same problem apparently 2017-03-02 10:06:57 glibc-2.17-157.el7_3.1.x86_64 for c7, glibc-2.12-1.192.el6.x86_64 for c6 2017-03-02 10:07:53 I don't need any additionnal vsyscall=... on c6/c7 to run the same docker images, hence my question. 2017-03-02 10:08:45 that depends on the kernel config 2017-03-02 10:10:02 i think centos7 works 2017-03-02 10:10:17 its only centos6 that dependson vsyscalls in kernel 2017-03-02 10:11:40 you have the same issue if you use arch linux as host 2017-03-02 10:11:43 or recent debian 2017-03-02 10:12:41 thx for the info 2017-03-02 10:13:08 centos6 needs kernel with vsyscalls support 2017-03-02 10:13:34 http://pastebin.centos.org/68626/ <- grep -i syscall on c6/c7 kernel config 2017-03-02 10:14:34 old kernels which supports it 2017-03-02 10:15:15 https://github.com/CentOS/sig-cloud-instance-images/issues/62 2017-03-02 10:15:21 http://tpaste.us/QD1E 2017-03-02 10:15:23 same issue as you mentionned 2017-03-02 10:16:13 yes, newer kernels on newer systems disables vsyscalls by default 2017-03-02 10:16:23 but let you emulate it with vsyscall=emulate 2017-03-02 10:18:54 -> FAQ or heads-up warning for containers hosting :P 2017-03-02 10:51:52 Okay, util-fkrt is mostly done: http://termbin.com/ipnf 2017-03-02 11:01:11 Revision of mkimage, refactoring nearly complete: http://termbin.com/4pca 2017-03-02 12:43:11 <^7heo> ncopa: did you see that yet? https://www.mirbsd.org/permalinks/wlog-10_e20170301-tg.htm#e20170301-tg_wlog-10 2017-03-02 12:43:55 <^7heo> "certain content may no longer legally be pushed to GitHub" 2017-03-02 12:44:59 <^7heo> I don't think we have problems with firefox being called firefox in alpine... yet. 2017-03-02 12:45:04 <^7heo> But that might be a bigger problem right there. 2017-03-02 13:03:48 ok? 2017-03-02 13:03:55 so we stop use github? 2017-03-02 13:07:57 ^7heo: did i read that right? its no longer allowed to put sources under GPL on github? 2017-03-02 13:12:33 ncopa: it definitely requires clarification on their side 2017-03-02 13:17:21 hackernews comments has two laywers in the top comments that refute that blogpost 2017-03-02 13:17:56 afaics it affects mostly sw that is uploaded by someone who has no authority over sublicensing (aka is no the author) 2017-03-02 13:24:05 yGweSm1OzVHe: refute it in what regard? 2017-03-02 13:24:14 the blogpost has multiple points 2017-03-02 13:25:05 the lawyers comments cast some doubts on the interpretation in the blogpost 2017-03-02 13:25:20 so definitely something to clarify 2017-03-02 13:26:36 mhm 2017-03-02 13:28:00 (i'm sorry to quote hn, usually reality is quite far from that crowd, but maybe this time the lawyers have a point) 2017-03-02 14:43:15 what's abuild up2date for? 2017-03-02 14:53:10 Is there pkgbase support in APKBUILD for multiple package? 2017-03-02 14:54:12 <^7heo> ncopa: you read that right. 2017-03-02 14:54:39 <^7heo> yGweSm1OzVHe: "it affects mostly sw that is uploaded by someone who has no authority over sublicensing" 2017-03-02 14:54:45 <^7heo> yGweSm1OzVHe: like, say, package managers? 2017-03-02 14:54:57 <^7heo> s/manage/develop/ 2017-03-02 15:36:29 apk Unable to lock database, how ? ps -ef can not found apk running 2017-03-02 15:38:39 why firefox's UI seems haven't correct 'font'? Are it lack dependency? 2017-03-02 15:39:10 There is no option to choose firefox-gtk2 and firefox-gtk3 2017-03-02 15:39:18 Firefox in alpine is weird. 2017-03-02 15:45:41 So, what's your browser in alpine? Are you using alpine for your desktop? 2017-03-02 15:50:25 dlintw: Yeah, alpine for desktop, arch back then. I uses firefox. 2017-03-02 15:51:02 I uses alpine for both desktop and server (raspberry pi) 2017-03-02 15:51:06 dlintw: What about you? 2017-03-02 15:53:08 My desktop for x86_64 is ArchLInux. I'm trying to use alpine for desktop for i686, but it seems difficult for me to use buggy firefox. 2017-03-02 15:53:35 I don't know why the firefox@testing can not show menu font correctly. 2017-03-02 15:55:54 pickfire, do your desktop use the 'ram','disk' or 'sys' mode? I found the 'sys' mode save more memory for my old note book. 2017-03-02 15:56:27 dlintw: No, firefox not buggy but most stuff are buggy like pandas that I need to use vm. 2017-03-02 15:56:52 dlintw: I use sys mode for both pi and laptop, I don't know how the others are useful. 2017-03-02 15:57:43 So far, alpine is nice to replace arch. Just broken boot and network configuration isn't that easy. 2017-03-02 15:57:50 so, what's your firefox version? are you using firefox@testing? is there require install other package before firefox? 2017-03-02 15:58:16 One more thing, no chinese input which I really hate. (tried building fcitx but libintl error which is the same like cryptsetup) 2017-03-02 15:58:36 I just install openbox, lxpanel, ttf-droid ttf-droid-nonlatin ttf-freefont 2017-03-02 15:58:42 firefox 51 2017-03-02 15:59:10 Ah, lxde stuff. I am planning to update those old lxde stuff and add gtk3 subpackages as well. 2017-03-02 15:59:50 I just use ttf-dejavu, ttf-mononoki and wqy-zenhei 2017-03-02 15:59:54 github is just another channel for getting packages in the distro, i don't think it is a major loss if we lose it 2017-03-02 15:59:56 You can try my main/lxterminal testing/gcin https://github.com/dlintw/aports 2017-03-02 15:59:58 And manually build dwm. 2017-03-02 16:00:08 Nice 2017-03-02 16:00:12 gcin? 2017-03-02 16:00:37 gcin is input method which for Taiwanese, there are also some common input methods. 2017-03-02 16:01:56 Oh, I didn't know about that. 2017-03-02 16:03:34 pickfire do you have your chinese input module package in your github? 2017-03-02 16:03:37 dlintw: Where is gcin? I can't find it. I only tried building fcitx, not ibus. 2017-03-02 16:03:48 dlintw: No, I didn't upload to github. 2017-03-02 16:03:56 I mean my repo not on github. 2017-03-02 16:04:33 Why put it on github? 2017-03-02 16:05:13 check myrepo in new-gcin2 & lxterminal-gtk branch 2017-03-02 16:05:58 I put source on github, it is easier for other's review (pull&request). Instead of e-mail interface. 2017-03-02 16:06:26 Ah, I didn't know you are from taiwan. 2017-03-02 16:06:40 pickfire 你看的到中文嗎? 2017-03-02 16:06:45 dlintw: Yes 2017-03-02 16:06:52 But can't reply in chinese as well. 2017-03-02 16:07:05 Why? what's your interface? 2017-03-02 16:07:11 lxterminal? 2017-03-02 16:07:17 No 2017-03-02 16:07:18 st 2017-03-02 16:07:37 what's your window-manager? 2017-03-02 16:07:50 https://transfer.sh/bi2LV/2017-03-03-000720-1366x768-scrot.png 2017-03-02 16:07:54 https://transfer.sh/bi2LV/2017-03-03-000720-1366x768-scrot.png 2017-03-02 16:07:55 dwm 2017-03-02 16:09:24 dlintw: 猜我从那来？ 2017-03-02 16:09:43 難 2017-03-02 16:10:12 Haha 2017-03-02 16:10:14 China,America, now, many place are Chinese. 2017-03-02 16:10:18 No 2017-03-02 16:10:21 Lol 2017-03-02 16:10:26 Taiwan? 2017-03-02 16:10:28 No 2017-03-02 16:10:33 I bet you can't guess it correctly. 2017-03-02 16:10:51 dlintw: Tip, I learn simplified chinese. 2017-03-02 16:12:19 Oh, are you Ivan Tham? 2017-03-02 16:12:38 Yeah 2017-03-02 16:13:00 I bet you saw where I am from. 2017-03-02 16:13:08 Malaysia 2017-03-02 16:13:20 Haha 2017-03-02 16:13:28 Nice to meet you, I must go to sleep. 2017-03-02 16:13:37 Okay, bye. 2017-03-02 16:16:16 ^7theo as a pkg manager you upload your buildscripts, not the sources of hthe packages. 2017-03-02 16:16:46 (laggy net, sorry for typos) 2017-03-02 16:53:14 laskin - no problem - thx for testing 2017-03-02 20:16:47 im tagging 3.5.2 now 2017-03-02 20:21:27 what about abuild :P 2017-03-02 20:28:50 should tag a release of that too 2017-03-02 20:29:01 but will not be able to look at that this week 2017-03-02 20:29:20 is the current git master of abuild good for release? 2017-03-02 20:29:27 no big breakages? 2017-03-02 20:29:38 should probably wait til monday though 2017-03-02 20:36:31 <_ikke_> What could go wrong :P 2017-03-02 20:47:23 <_ikke_> \o/ 2017-03-02 21:18:29 hey, this may be a farfetched place to ask, but since you all seem so nice: does anybody know why autoconf doesn't produce new releases anymore? 2017-03-02 21:22:50 maybe they finally realized how horrible mistake they did and stopped? okay, just kidding, I’m not so naive… 2017-03-02 21:26:14 :D 2017-03-02 21:26:41 mikeee_: I suspect it's considered "stable" and "mature" by the upstream, thus won't see a release unless a major bug turns up. 2017-03-02 21:27:44 and apparently they don’t consider horrible design as a major bug… 2017-03-02 21:28:18 *lol* Nope, that's a FEATURE, not a bug :) 2017-03-02 21:31:58 Hey guys, can a get a few eyeballs and testing on this fkrt (fakeroot) utility script: http://termbin.com/p8o8 2017-03-02 21:33:03 It lets you setup multiple fakeroot environments and switch between them from a shell or in a script. 2017-03-02 21:33:32 ncopa: basically we have backported everything anyway :P 2017-03-02 21:33:34 fkrt_enable 2017-03-02 21:38:27 TemptorSent: why do you prefix even local variables with fkrt_fakeroot…? 2017-03-02 21:39:40 jirutka: Refactoring artifacts :) 2017-03-02 21:40:04 and some variables inside funcs are not declared as local and it’s not clear if it’s an intention or not, they are not differentiated from locals (e.g. by using capitals) 2017-03-02 21:40:09 jirutka: Obviously, the local vars can get shortened. 2017-03-02 21:40:38 mixed tabs and spaces… 2017-03-02 21:41:37 fkrt_use_abs_lib_path is basically a constant (it’s not changed anywhere), so `[ $fkrt_use_abs_lib_path -ne 0 ]` doesn’t make much sense 2017-03-02 21:41:40 jirutka: Thanks, it started as /usr/bin/fakeroot :) 2017-03-02 21:42:06 `[ "$LD_PRELOAD" ]` … this is not very safe… use -n 2017-03-02 21:43:46 AFAIK, [ "$str" ] is equal to [ -n "$str" ] in this case? 2017-03-02 21:44:03 "Implement setvar function (Why is this missing from our version of ASH?)" … well, b/c it’s not in POSIX standard… actually I don’t know about any shell that implements this func 2017-03-02 21:45:02 what are all these `[ "$1" ] && _inst="_$1"`? 2017-03-02 21:45:03 jirutka: The ash man page states setvar is a builtin. 2017-03-02 21:45:22 "AFAIK, [ "$str" ] is equal to [ -n "$str" ] in this case?" … no, it’s not 2017-03-02 21:45:56 ah, you’re right about setvar 2017-03-02 21:46:11 jirutka: Can you point me at a document for that? The man page for test states that equivilence explicitly. 2017-03-02 21:48:26 pardon, it seems that you’re right about [ "$foo" ] 2017-03-02 21:48:31 jirutka: the [ "$1" ] && _inst="_$1" statments check if an argument has been passed and set the instance suffix after adding a "_" 2017-03-02 21:48:40 but it must be really quoted string 2017-03-02 21:49:36 jirutka: Right, for safety, I always quote my vars unless I REALLY intend to pass the raw value. 2017-03-02 21:50:21 well, you can write it as `local _inst="${1:+"_$1"}"` sufficient? 2017-03-02 21:50:35 s/sufficient?// 2017-03-02 21:54:29 sometimes you define local variables that are not used anyhwere 2017-03-02 21:54:42 e.g. mypid in fkrt_faked_stop 2017-03-02 21:59:26 jirutka: Good point on the '+' operation, I'll swap those out. 2017-03-02 22:35:19 jirutka: Revised per your comments, please take a look: http://termbin.com/wm7uc 2017-03-02 22:36:19 so `fkrt_ld_preload` is meant to be global? 2017-03-02 22:37:11 jirutka: Yes, that's what gets set into LD_PRELOAD when the a fakeroot is active. 2017-03-02 22:37:51 then i’d name it in CAMEL_CASE, to be clear that it’s a global variable, not local 2017-03-02 22:39:52 hm, actually… 2017-03-02 22:40:44 all such variables are prefixed with fkrt_, so it’s probably clear enough 2017-03-02 22:41:47 `[ -p "$fkrt_database_file" ] $$ fkrt_wait_in_trap=1` … shouldn’t be there && or || instead of $$? 2017-03-02 22:42:09 jirutka: Yeah, I cleaned up the prefixing to clarify that. All variables and functions start with fkrt_ 2017-03-02 22:43:24 Woah, missed that one! 2017-03-02 22:43:32 why you’re doing this: `fkrt_inst_list="$fkrt_inst_list ${_inst#_}"; fkrt_inst_list="${fkrt_inst_list## }"`? 2017-03-02 22:43:46 Currently, the db load/save are dead code. 2017-03-02 22:44:11 Trimming leading spaces :) 2017-03-02 22:44:30 do you really need to trim them…? 2017-03-02 22:44:48 I'm tempted to just drag in my list util to handle it, nicely. 2017-03-02 22:45:33 probably not absolutely necessary, but it keeps everything clean for inspection in the shell anyway :) 2017-03-02 22:45:53 I don’t think that a single leading space can hurt… 2017-03-02 22:46:43 Not until it gets used by something else later like awk :) 2017-03-02 22:46:54 and yes, util function for appending to list would be better then this 2017-03-02 22:47:46 jirutka: Easy enough to change it, since I'm using the util for mkimage anyway. 2017-03-02 22:49:10 btw where’s this script actually used? 2017-03-02 22:50:07 jirutka: I'm building it for use in mkimage for building overlays, but the intent is to have a general-purpose set of utilities for any scriping needs. 2017-03-02 22:50:38 It also works right at the shell by sourcing and running the functions. 2017-03-02 22:51:31 and what is it actually doing? it seems that it just sets LD_PRELOAD… 2017-03-02 22:55:12 jirutka: Pretty much yes, along with starting faked and keeping track of keys and pids. 2017-03-02 22:56:38 what does it do when fkrt_lib is not found? it does abort, right? 2017-03-02 22:59:08 jirutka: Here's the revision and including the list utils: http://termbin.com/plcu 2017-03-02 23:00:14 uff, why is there so many redundant code?! 2017-03-02 23:00:36 Yes, it bails out with a return value of 1 and prints a warning on failure (actually, warning function defined by abuild utils) 2017-03-02 23:01:27 then why it’s so unnecessary verbose? it can be shorten as https://hastebin.com/kolukoxoti.sh 2017-03-02 23:01:32 jirutka: Agreed, I don't like the redundancy, but it ends up being simpler per function than than putting it all in a big case. 2017-03-02 23:01:58 and I doubt if global variable fkrt_lib_found is needed at all 2017-03-02 23:02:30 no, I don’t mean splitting code in multiple functions, but a lot of dead code or unnecessary verbose code 2017-03-02 23:03:51 jirutka: the fkrt_lib_found signifies that init has run successfully. I suppose I could use fkrt_ld_preload for that purpose instead actually. 2017-03-02 23:05:55 probably, there are so many vars that I can’t keep it in head 2017-03-02 23:06:03 actually, I can probably eliminate the variable entirely and merge the if statement. 2017-03-02 23:06:34 the actual lib is written in C, isn’t it? wouldn’t it be better to rewrite this script into C? 2017-03-02 23:07:15 it doesn’t look like a task for which shell is a good tool 2017-03-02 23:08:49 The issue is that I need to be able to toggle the fakeroot environment on and off in the shell scripts for building images. 2017-03-02 23:09:47 And I might have several fakeroot environments going simultaneously during the build. 2017-03-02 23:10:29 This lets me keep them all straight from within the script, using passing the section I'm building as the instance name. 2017-03-02 23:11:57 I probably could have written the script from scratch instead of starting with the existing fakeroot script, but I'm not that familiar with fakeroot/faked, so I figured I'd start with somethign that worked and extend it. 2017-03-02 23:13:13 I dont' currently have a need for the db save/restore abilities, but I figured I'd leave them in and make them work in case a need pops up later. 2017-03-02 23:14:21 the only thing that these 273 (486) lines do is setting three variables: LD_PRELOAD, FAKED_MODE, FAKEROOTKEY, right? 2017-03-02 23:14:40 ah and killing processes 2017-03-02 23:15:44 and keeping track of pids 2017-03-02 23:17:05 init sets up the environment, then faked_start runs a new faked for each instance, tracking pid, key, options, etc. for each. 2017-03-02 23:17:22 maybe it may be simpler to store state in file(s) then in variables 2017-03-02 23:18:21 enable runs init and starts the instance if not already running, then sets LD_PRELOAD FAKEROOTKEY and FAKED_MODE for that instance. 2017-03-02 23:18:36 disable unsets the environment variables. 2017-03-02 23:19:29 faked_stop shuts down an instance of faked and cleans up the environment. 2017-03-02 23:19:39 cleanup stops all instances. 2017-03-02 23:20:21 faked_kill does the actual killing of processes. 2017-03-02 23:21:01 and cleanup is registered as the signal handler for INT/EXIT so we don't leave a mess behind when it exits. 2017-03-02 23:22:21 part of the problem with storing state in files is that fakeroot is by design screwing with the apparent permissions and we don't necessarily have a designated storage location that isn't faked. 2017-03-02 23:23:49 Needles to say, a couple arrays in bash would have made this all very simple indeed, but we're POSIXly correct (or mostly so) on Alpine, so it takes a bit of trickery to support. 2017-03-03 08:40:27 Morning. Happy Friday! 2017-03-03 12:48:51 Actually, why is /etc/alpine-release useful? 2017-03-03 12:49:04 I see no reason to duplicated the content of /etc/os-release 2017-03-03 12:57:43 <^7heo> pickfire: the reason is to have an easy to parse version for use by alpine scripts 2017-03-03 12:57:50 <^7heo> and os-release is for interoperability 2017-03-03 12:57:57 Lol 2017-03-03 12:58:04 <^7heo> ?! 2017-03-03 12:58:07 ^7heo: Alpine script can always . /etc/os-release 2017-03-03 12:59:11 <^7heo> yeah 2017-03-03 12:59:19 <^7heo> and if it's not in shell 2017-03-03 12:59:26 <^7heo> it can always implement a shell parser, too, right? 2017-03-03 12:59:35 <^7heo> like, I dunno, from C. 2017-03-03 12:59:39 <^7heo> such a good idea. 2017-03-03 13:34:03 vakartel: the hpn patches for openssh was disabled by mistake. we should not remove them 2017-03-03 13:34:33 have a nice weekend everyone! 2017-03-03 13:34:53 lel 2017-03-03 13:51:23 <^7heo> damn I really am failing to see what's so funny, those days. 2017-03-03 13:51:26 <^7heo> "Hello" 2017-03-03 13:51:29 <^7heo> "lol" 2017-03-03 13:51:46 <^7heo> Is everyone high or? 2017-03-03 14:00:56 Yeah, from C then complicated. 2017-03-03 17:09:02 hi 2017-03-03 17:10:01 let's say I would like to be able to build my own installation image... I would like to preconfigure wifi and SSH including authorized_keys... all that in a reliable and reproducible way 2017-03-03 17:10:24 so that I can easily rebuild it for another network and another SSH user 2017-03-03 17:10:36 what's the best starting point? 2017-03-03 17:11:03 (I do expect some work to be done before it's ready.) 2017-03-03 17:40:02 pavlix: Good morning - I'm currently working on building out the image building system. 2017-03-03 17:41:48 pavlix: I have mkimage modularized and I'm currently working on advancing the overlay building significantly 2017-03-03 17:44:39 TemptorSent: So what is the status, can one try it already and tweak it to their liking or not? 2017-03-03 17:46:07 pavlix: Not quite finished yet, but if you want to lend some eyeballs and some shell script, I could use the help. 2017-03-03 17:46:44 TemptorSent: If you don't mind that I can only lend little time of mine, then I could... 2017-03-03 17:49:12 pavlix: No problem, I have much the same needs. 2017-03-03 17:50:17 pavlix: Hang on a sec, let me get my current work ready to commit and I'll post it. 2017-03-03 17:59:40 TemptorSent: Sure. 2017-03-03 18:03:27 pavlix: I'm almost done with the bulk of the coding for the overlays, but haven't tested any of that yet, so it's currently not going to work as is (barring a random miracle, of course :) ) 2017-03-03 18:07:44 pavlix: There is still a fair bit of refactoring to be done and cleanup of naming before it's ready to stick a fork in. 2017-03-03 18:09:07 TemptorSent: you don’t need anything special to build an Alpine image; just apk and chroot 2017-03-03 18:11:27 jirutka: Sure, it's possible to manually build an image, but certainly not fast, easily reproducable, or particularly convenient. 2017-03-03 18:11:39 not manually 2017-03-03 18:12:40 jirutak: We're talking not just the packages, but the overlay, kernel config, initfs config, modules, etc. 2017-03-03 18:13:18 have you looked into https://github.com/alpinelinux/aports/tree/master/.travis/ ? 2017-03-03 18:13:46 jirutak: If you have only one configuration to deal with, it's fine, but if you need to build multiple images, it's painful. 2017-03-03 18:14:14 jirtutka: I'm working on aports/mkimage. 2017-03-03 18:14:32 hm, maybe I’ll understand once I try your tool 2017-03-03 18:14:42 currently I don’t see how fakeroot can help with this 2017-03-03 18:15:43 jirutka: Fakeroot allows the overlays to be built with the desired permissiions and ownership propigated to the overlay tarball. 2017-03-03 18:17:44 have you discussed it with fabled? IIRC he said that fake-root is just a huge hack and he plans to move into another solution 2017-03-03 18:19:35 jirutka: Long time no see. 2017-03-03 18:20:38 pavlix: yeah… 2017-03-03 18:21:04 pavlix: I assume that we’ll see each other tomorrow at InstallFest…? 2017-03-03 18:21:49 jirutka: I'm definitely going to stop by and probably both of the days. 2017-03-03 18:22:19 jirutka: Yes, I have been talking with fabled on the mkimage work. 2017-03-03 18:22:22 pavlix: me too, at least to Petr’s talk about Alpine and Pavel’s talk about Rust 2017-03-03 18:22:29 TemptorSent: good good 2017-03-03 18:22:52 jirutka: I haven't looked at the schedule, yet, I'm mostly going because of the people. 2017-03-03 18:23:49 pavlix: me too, there are not many interesting talks for me 2017-03-03 18:24:31 jirutka: I'm looking at your .travis scripts, by the way. but I don't know what to make of it, yet. 2017-03-03 18:24:49 unfortunately it’s like a typical Czech conference, most of the talks are just about hyped technologies from their users or PR talks by companies :/ 2017-03-03 18:25:27 pavlix: these in aports repo are a bit outdated, I’ll rewrite it to reuse https://github.com/jirutka/alpine-chroot-install/ 2017-03-03 18:25:38 jirutka: You can't expect another FOSDEM in CZ, right? :D 2017-03-03 18:25:54 pavlix: why not? :P 2017-03-03 18:26:35 jirutka: Because it's a unique conference with its history and stuff like that... many people only come to Europe for FOSDEM. 2017-03-03 18:26:43 pavlix: maybe LinuxDays will become a smaller FOSDEM someday, this year most of the talks will be in English and many foreign speakers 2017-03-03 18:27:43 jirutka: Whatever. I'm not going to search for my crystal ball today. :) 2017-03-03 18:27:55 pavlix: do you know what I’d like to see? a conference with talks about ideas, not about products 2017-03-03 18:28:22 jirutka: Then you're not a typical conference visitor, I'm afraid. 2017-03-03 18:28:44 jirutka: But you can still create such a conference, or create such a track at LinuxDays. 2017-03-03 18:29:32 most of the talks are like “hey, I’ve found Ansible few months ago and I use it for my servers, it’s super cool and I must tell you how cool it is… well, no, I’ve never used another similar tool, I don’t have an idea how it actually works and how it can be solved in another way…” 2017-03-03 18:32:15 pavlix: Okay, here's a patchset against the aports/scripts/mkimage: http://termbin.com/zp0n 2017-03-03 18:35:39 jirutka: Take a look at the above, it probalby makes more sense in context. 2017-03-03 18:36:34 TemptorSent: It looks pretty big for a person who wasn't in the project before. 2017-03-03 18:37:39 pavlix: Yeh, I had a need and I started coding to meet it... then I found out that the old alpine-iso system was depreciated, so I started over with mkimage :) 2017-03-03 18:38:39 :) 2017-03-03 18:38:49 TemptorSent: That sounds like a good idea. 2017-03-03 18:39:08 ufff, that’s too big, can’t review it 2017-03-03 18:39:09 TemptorSent: Are the changes intended to be accepted upstream in any form near the current status? 2017-03-03 18:40:19 pavlix: Hopefully after cleanups most of it can be upstreamed. 2017-03-03 18:42:01 pavlix: It needs to be at least on parity for building release images obviously, as well as settle down on naming before it's ready for the tree. Fabled has similar needs to ours, so he's interested in seeing a solution move forward. 2017-03-03 18:42:27 TemptorSent: Are you going to use it yourself before getting it merged? 2017-03-03 18:42:30 pavlix: There is also a scope of work to be attacked in modularizing mkinitfs. 2017-03-03 18:43:09 TemptorSent: I'm interested in the classic RAID+LVM+initramfs setup as well, have you considered that? 2017-03-03 18:43:12 pavlix: Yes, I have an immediate need for the ability to burn custom images for a client, so it's going to be in use. 2017-03-03 18:43:45 TemptorSent: If you're going to actually use it, it would be good to keep a occasionally rebased version at some public Git service. 2017-03-03 18:43:53 pavlix: Already started on the work needed to handle that. 2017-03-03 18:44:29 pavlix: Agreed, I want to get it usable before dumping on github or whatnot. 2017-03-03 18:44:31 TemptorSent: Now that sounds great. The scripts don't do RAID6 which I thing is a trivial change... 2017-03-03 18:44:50 TemptorSent: Plus it doesn't do single-partition software raid with LVM and rootfs on it 2017-03-03 18:45:25 TemptorSent: I'm talking about setup-disk. 2017-03-03 18:45:30 pavlix: Take a look at the features and overlays systems. 2017-03-03 18:45:45 TemptorSent: Where/why? 2017-03-03 18:46:12 pavlix: I haven't attacked the setup-disk portion yet, as that's beyond the image building for the moment. 2017-03-03 18:47:01 TemptorSent: Yep. But if you're going to do RAID/LVM/initramfs/rootfs, you'll probably be interested in disk installation as well, right? 2017-03-03 18:47:04 pavlix: Apply the patch to a branch on a convenient aports tree. 2017-03-03 18:47:44 TemptorSent: In my experience a published branch works better than a patch or patchset, because it stays the same until it gets rebased and updated. 2017-03-03 18:47:48 pavlix: Actually, one of my big needs is to be able to build a fully configured media-based boot system. 2017-03-03 18:48:03 TemptorSent: I've done this with https://pagure.io/fork/pavlix/koji/branch/python3 2017-03-03 18:48:20 pavlix: Agreed, I just mean for taking a look at how the system works. 2017-03-03 18:48:21 TemptorSent: What do you mean by media-based boot system? 2017-03-03 18:48:47 pavlix: Drop an iso to a dvd, stick in server, boot :) 2017-03-03 18:48:48 TemptorSent: I can probably just look at the patch and wait for the published branch if that's going to happen. 2017-03-03 18:49:12 TemptorSent: Boot and then what? 2017-03-03 18:49:30 pavlix: The patch makes the directory-structre refactor rather less than clear I'm afraid. 2017-03-03 18:49:33 1. Boot, 2. ???, 3. Profit! XD 2017-03-03 18:50:13 pavlix: Run from ram, mounting storage and running services, with all configuration stored seperately. 2017-03-03 18:50:16 jirutka: Yep. :) TemptorSent described the technical part of booting the system from media (which you can already do somehow) and left out the actual focus. 2017-03-03 18:50:37 TemptorSent: Just configuration or installed packages and stuff like that? 2017-03-03 18:53:06 pavlix: My application is a zfs-based storage/KVM server, so it will boot to the initfs, mount up the storage, export it, and run the VMs. Configuation is handled with overlays only, just like the media-based environment currently does. 2017-03-03 18:53:54 TemptorSent: I'm not going to have a similar use case, though. 2017-03-03 18:53:55 pavlix: Kernel upgrades look just like building the original image. 2017-03-03 18:54:02 TemptorSent: hmm, interesting, this reminds me the project at vpsFree I’m involved in 2017-03-03 18:54:31 jirutka: vpsFree also somewhat favors ZFS over Btrfs and anything else... :) 2017-03-03 18:54:32 pavlix: No problem, the point of modularizing the build system was to allow for significant flexability in the profiles. 2017-03-03 18:54:43 TemptorSent: +1 2017-03-03 18:55:25 pavlix: The other side project is a rpi based WX station and camera :) 2017-03-03 18:55:57 TemptorSent: :) 2017-03-03 18:56:22 TemptorSent: I have one „side“ project to get ready to freelance after five years of employment. :D 2017-03-03 18:56:32 pavlix: The fun part is the VM images will be built by the same tool, so I can package an iso with the running fs AND the initial VM image. 2017-03-03 18:56:40 pavlix: yes, unfortunately, I don’t like ZFS… but that’s not THE interesting part on this project 2017-03-03 18:56:54 jirutka: I see, the FS is just a backend, right? 2017-03-03 18:57:04 pavlix: yeah 2017-03-03 18:57:20 jirutka: What's your flavor? I intend to make FS support essentially transparent going forward. 2017-03-03 18:57:38 TemptorSent: Btrfs and ext4 2017-03-03 18:58:09 TemptorSent: I used to like RaiserFS, but unfortunately its developer is in jail and it seems that he’s not allowed to code here :( 2017-03-03 18:58:16 jirutka: I haven't played with btrfs in a number of years now -- how is it looking these days? 2017-03-03 18:58:56 and RaiserFS is currently outdated, can’t use multiple cores for parallelization on operations 2017-03-03 18:59:30 TemptorSent: it looks good, userland tools are quite powerful; but RAID 5/6 is horribly broken 2017-03-03 18:59:34 it eats data 2017-03-03 18:59:47 jirutka: Yeah, it's been broken for many years now, and even in it's hayday it only worked well for specific loads (think nntp server) 2017-03-03 18:59:48 other RAIDs should be okay, at least I hope… 2017-03-03 19:00:03 TemptorSent: The patch seems to include a lot of renaming. When on a branch, the renaming would occur in a separate commit and I could see more detailed code changes then. 2017-03-03 19:00:11 jirutka: Ouch! Yeah -- that's why I'm sticking with ZFS. I like my data intact. 2017-03-03 19:01:02 jirutka: RAIDs broken that doesn't sound good. 2017-03-03 19:01:35 pavlix: I did my best to clean up the commit history so it's clear what's happening, but git doesn't let me explicitly copy files, it just detects them by content. 2017-03-03 19:03:05 pavlix: I think only mkimage.sh and mkimage-yaml.sh have stayed in their original location, and the rest have been split apart. 2017-03-03 19:03:34 TemptorSent: I know Git doesn't have any metadata for that, but I simply `git mv` files, commit the moves + any changes needed to keep it working, and only then do more modifications in separate commits. 2017-03-03 19:03:34 and mkimage-yaml.sh's days are numbered. 2017-03-03 19:04:03 pavlix: Believe it or not, that's what I did in most cases! 2017-03-03 19:04:32 TemptorSent: But then you have multiple commits and not just a big diff. 2017-03-03 19:05:07 pavlix: I think part of it is that git is puking on files which only a small portion is retained and the rest deleted while splitting apart. 2017-03-03 19:05:41 pavlix: There should be 19 commits there... 2017-03-03 19:07:03 TemptorSent: Splitting is a slightly different story than just moving. 2017-03-03 19:08:02 TemptorSent: Nice. Then I'll be happy to access those as a branch once it's there. 2017-03-03 19:08:07 pavlix: Right, which is where git lost it.. I copied the source file to be split to each destination file, commited that, then changed the results. 2017-03-03 19:08:53 pavlix: If you apply it to a branch on your tree, it SHOULD have all commits available to peruse. 2017-03-03 19:09:38 ...in the mean time, let's see if I can figure out my github login... :) 2017-03-03 19:10:14 TemptorSent: :) 2017-03-03 19:12:13 Right about now, I'm ready to shoot whoever killed the gpm package... 2017-03-03 19:12:39 Ever try retyping every url by hand between terminals? 2017-03-03 19:12:46 TemptorSent: RAID 1 should be okay; Btrfs is already used in production, not so long as ZFS though 2017-03-03 19:14:00 jirutka: Understood. I actually use RAID-Z2 for my important data, with additional duplicates per file if I'm really paranoid. 2017-03-03 19:14:29 jirutka: The real win for me on zfs is snapshotting/zfs send :) 2017-03-03 19:16:46 jirtuka: The only thing really lacking for me is strong support for flash, especially in direct mode. 2017-03-03 19:17:21 jirutka: What drives your preference for Btrfs? 2017-03-03 19:21:35 TemptorSent: ZFS is like an alien in Linux, it’s not well integrated 2017-03-03 19:21:51 TemptorSent: ZFS is very inflexible, it doesn’t know anything like rebalancing 2017-03-03 19:22:36 TemptorSent: I don’t usually have spare data storage where I can move all the data just b/c ZFS volumes are like in concrete, you can’t change anything 2017-03-03 19:23:09 jirutka: mimi_vx (usually in #gentoo.cs) is also a big fan of btrfs over zfs 2017-03-03 19:23:23 jirutka: With basically the same arguments, I'd say. 2017-03-03 19:23:49 jirutka: Generally, I just add pools as needed or migrate to larger disks. 2017-03-03 19:24:43 jirtuka: But yes, it would be nice to easily add storage to an existing zpool and use it without restriping everything. 2017-03-03 19:33:53 I’ve easily expanded RAID1 with single disk to 5 disks and double capacity per disk 2017-03-03 19:35:12 b/c I needed to migrate from ext4 and didn’t have spare space where to copy all data 2017-03-03 19:35:43 jirutka: ZFS would have no problem restripng to the added disks transparently actually. 2017-03-03 19:36:31 jirutak: The only problem is adding discs to an existing pool without restriping. 2017-03-03 19:45:29 TemptorSent: Why's gpm not available? 2017-03-03 19:46:24 pavlix: Not sure, it probably got dropped because everyone is assumed to be running X? 2017-03-03 19:47:00 TemptorSent: Are they going to drop the VT support as well? 2017-03-03 19:47:45 :) 2017-03-03 19:51:23 pavlix: Yeah, actually there's been noise about that. 2017-03-03 19:52:34 pavlix: Okay, I think I fixed the bugs I introduced in that last refactor... One sec and I'll dump a new patchset 2017-03-03 19:53:02 Making sure it actually builds the stuff that worked before :) 2017-03-03 19:54:48 My network connection is a slightly glorified set of tin-cans on a wet string :) 2017-03-03 19:55:20 Refetching packages takes ages. 2017-03-03 20:00:01 TemptorSent: I was going to fix NetworkManager but haven't got back to it since my initial experiments. 2017-03-03 20:00:20 But that's another story. 2017-03-03 20:00:58 TemptorSent: But I'm tempted to cherry-pick a couple of packages to fix, so it might be something that helps you as well. 2017-03-03 20:05:17 pavlix: Yeah, the fix is replacing the half mile of burried copper with ugly splices with fibre. 2017-03-03 20:07:47 pavlix: Downlink speed is almost usable at times (as high as 1M/s when it's doing well), but uplink is UGLY at 100K/s and nasty buffer bloat somewhere, probably the DSLAM. 2017-03-03 20:08:28 pavlix: Downloading the kernel and firmware repeatedly takes inordinately long. 2017-03-03 20:10:06 TemptorSent: I'm pretty much fortunate that in my country you can get a Wifi link in most places. 2017-03-03 20:20:51 pavlix: Yeah, I don't even get a cell signal at my home. I'm a little out in the sticks ;) 2017-03-03 20:22:40 TemptorSent: Yeah, plus now I get LTE as a backup for urgent stuff. 2017-03-03 20:27:19 pavlix: Okay, fixing up a couple of bugs so I get the existing builds working right. 2017-03-03 20:42:10 pavlix: This actually works (for at least some value of works) - http://termbin.com/uwab 2017-03-03 20:56:59 pavlix: Now for the fun, making the overlays work right :) 2017-03-03 21:02:06 TemptorSent: Were you successful with your GitHub account? 2017-03-03 21:02:45 TemptorSent: Anyway, what do you mean by overlays here? Overlay file systems? 2017-03-03 21:06:28 pavlix: Yes, I found my account, but don't have it configured yet. 2017-03-03 21:07:19 pavlix: And yes, overlay filesystems, allowing all configuration to be done in the overlay. 2017-03-03 21:11:12 TemptorSent: I love the layering possibilities. 2017-03-03 21:11:25 TemptorSent: I first work with such a system when using OpenWRT on a router. 2017-03-03 21:14:04 pavlix: Yes, it makes life MUCH easier in many cases. 2017-03-03 21:23:15 TemptorSent: I'd be happy to use it as an integral part of the system. With tools. You have packages that install stuff to /etc, right, let's install stuff to the system layer. Did I make a change? Let's put that into the config layer on top of the system layer. And now please give me the diff for that file between the overlays. Then you woudn't need all the lennartish /usr/lib and /etc configuration 2017-03-03 21:23:21 directories which is basically a poor man's version of the described use case. 2017-03-03 21:23:57 TemptorSent: Now that I'm leaving my job and more or less leaving Fedora development and stuff like that, it feels so tempting to explore some distros that actually differ in the basic concepts. 2017-03-03 21:25:10 pavlix: Yeah, my *other* system is funtoo :) 2017-03-03 21:26:00 pavlix: I was using the earliest versions of gentoo back in the late '90s, moved to funtoo when drobbins forked it. 2017-03-03 21:26:01 TemptorSent: I've been using Gentoo even at Red Hat. :) 2017-03-03 21:26:19 TemptorSent: http://www.abclinuxu.cz/blog/pavlix/2014/2/fedora-packaging-in-gentoo 2017-03-03 21:26:36 TemptorSent: Is Funtoo that much different from Gentoo after all? 2017-03-03 21:29:40 pavlix: In some regards, yes -- it has a nice MODULAR profile system (sound familiar?) and some better design decisions. 2017-03-03 21:31:09 TemptorSent: What does that mean? I have a couple of Gentoo overlays, would you be able to use them with Funtoo? 2017-03-03 21:32:38 pavlix: Yes, the overlays work fine between the two. The big difference is the profile system, which has the ability to have modular inheratance. 2017-03-03 21:33:26 pavlix: So you can define your own custom profile with your configuration needs and still track updates to the portage tree. 2017-03-03 21:38:04 TemptorSent: Sounds interesting. 2017-03-03 21:38:21 TemptorSent: I'm using some default profile plus a couple of tweaks. 2017-03-03 21:38:54 not much custom stuff 2017-03-03 21:47:45 pavlix: Okay, I have it building overlays and everything... Now I'm just trying to convince tar to strip the leading ./ from the archive. 2017-03-03 21:51:54 TemptorSent: For what purpose? 2017-03-03 21:57:13 DeepStack, the AI that beat top Poker players, is written in Lua and runs on LuaJIT! \o/ They use framework Torch7. 2017-03-03 21:58:31 pavlix: No actual need, just trying to keep everything clean. 2017-03-03 21:59:01 TemptorSent: So you have an existing tar archive and you want to modify it? 2017-03-03 22:00:28 pavlix: No, I was just trying to get tar to strip the leading "./" when generating an archive with that as the filespec. 2017-03-03 22:01:39 tar -c -C "$srcdir" "./" | gzip -9n > "$outfile" 2017-03-03 22:07:24 pavlix: It works now, I'll leave it alone unless there's a reason to mess with it I guess :) 2017-03-03 22:13:40 pavlix: Okay, this appears to be functional, including building overlays: http://termbin.com/s7hc 2017-03-03 22:15:05 jirutka: Take a look if you're interested, this is beginning to demonstrate the capabilities. 2017-03-03 22:16:06 TemptorSent: can you show me some scripts using this stuff or something like that? 2017-03-03 22:16:42 jirutka: mkimage.sh runs everything! 2017-03-03 22:16:44 also it’d be better to open pull request for it, so we can easily review it 2017-03-03 22:17:08 jirutka: Working on getting my github account talking to my dev box. 2017-03-03 22:17:49 jirutka: Apparently people have forgotten that some of us don't use any GUI on our development boxes :) 2017-03-03 22:18:09 TemptorSent: no, they don’t… 2017-03-03 22:18:23 TemptorSent: everything you need is just to add your public SSH key from the box to your GH account, that’s all 2017-03-03 22:18:47 you can do it even from CLI, but IMO it’s easier to just use it from web from some computer with GUI 2017-03-03 22:19:34 jirutka: Hmm, any docs on how to do it without a browser? Otherwise I get to have fun playing the 4-way shuffle to get a key to github. 2017-03-03 22:20:03 do you have SSH access to that box? 2017-03-03 22:20:54 jirutka: Not currently ;) It's a totally stand-alone machine on my network, no connections to any other devices. 2017-03-03 22:21:25 TemptorSent: WAT, so how do you work with it? you’re connected to it directly and looking into a HW terminal? 2017-03-03 22:21:42 is it at least connected to the Internet? 2017-03-03 22:22:33 jirutka: Yup, sitting in front of an actual piece of hardware! Yes, it has a live connection to the outside world, but not to any of my other machines. 2017-03-03 22:23:04 TemptorSent: Your configuration is unusual, at the least. :) 2017-03-03 22:23:15 then just copy your SSH pubkey to e.g. to http://tpaste.us/ 2017-03-03 22:23:25 TemptorSent: Why don't you just publish out your authorized_keys and then use it from wherever you need. 2017-03-03 22:23:38 jirutka: Whoever depreciated gpm should be shot -- retyping everything is a PITA. 2017-03-03 22:24:04 TemptorSent: then just connect via damn SSH and use a modern terminal emulator ;) 2017-03-03 22:24:04 pavlix: Untrusted machines on the network :/ 2017-03-03 22:24:37 TemptorSent: Your needs are not to be judged by me. ;) 2017-03-03 22:25:13 jirutka: No machines with a terminal emulator that even comes close to the usability of a nice big monitor on a hardware console. 2017-03-03 22:25:18 jirutka: When gpm is in unmaintained, what does it mean? That it doesn't get built at all? 2017-03-03 22:26:12 jirutka: I flip between a half-dozen VTs instantly. 2017-03-03 22:27:02 TemptorSent: this sounds like that the only terminal emulator you know is cmd.exe or Putty, otherwise it’s just non-sense… or terminal emulator or Linux are MUCH worse than I’ve expected (I use iTerm2 on macOS and it cannot be compared with (un)usability of HW terminal) 2017-03-03 22:29:01 jirutka: I could work with XTerm with a small number of configuration tweaks, to be honest. Now I'm using urxvt. I'd even cope with gnome-terminal, it has some nice features, but I don't really use that much of them when in i3. 2017-03-03 22:29:02 anyway, for example http://unix.stackexchange.com/questions/136894/command-line-method-or-programmatically-add-ssh-key-to-github-com-user-account, there is described how to upload your pubkey to GH using curl 2017-03-03 22:29:15 but not sure if you can read StackOverflow on your device :P 2017-03-03 22:29:31 jirutka: I've been working from the command line for 30+ years now. I HATE having to touch a mouse, and frankly, term emulators are SLOW. 2017-03-03 22:30:10 TemptorSent: then you’ll love https://github.com/jwilm/alacritty 2017-03-03 22:30:14 A cross-platform, GPU-accelerated terminal emulator 2017-03-03 22:30:20 jirutka: Yeah, I get to retype it into my browser machine (ANCIENT Windoze if you must know) 2017-03-03 22:30:39 I didn’t understand this project until now :) b/c I don’t have any problems with speed of my terminal emulator 2017-03-03 22:31:18 jirutka: Looking at it now... looks like it might at least have a chance! 2017-03-03 22:31:23 ah, so I should probably copy the description so you don’t have to retype the URI: Alacritty is the fastest terminal emulator in existence. Using the GPU for rendering enables optimizations that simply aren't possible in other emulators. Alacritty currently supports macOS and Linux, and Windows support is planned before the 1.0 release. 2017-03-03 22:33:07 Hmm, it's rusty too.. that makes it even more interesting, thanks! 2017-03-03 22:34:13 I guess now I get to figure out how to get X working on the fire pro card in this old dell. 2017-03-03 22:37:15 It's a FirePro W5000, which happens to be a Pitcarin card -- not supported in the new series AMD drivers and apparently the older drivers suffer from bit rot and don't like the modern kernels so much. 2017-03-03 22:37:56 Sadly, this machine is probably going to end up being relegated back to a drafting workstation soon. 2017-03-03 22:40:23 jirutka: does it have any value on intel based laptops without dedicated graphics? 2017-03-03 22:41:26 pavlix: definitely, every laptop has GPU, just some integrated in processor, but these are quite powerful nowadays and supports even OpenCL 2017-03-03 22:42:05 actually, GPUs in Intel Core supported OpenCL like 6 years ago, maybe even sooner 2017-03-03 22:42:21 jirutka: Very cool project! It's about time someone started thinking about usability again. 2017-03-03 22:42:41 yup 2017-03-03 22:42:57 jirutka: Will give it a try on X230 then. 2017-03-03 22:42:59 but still, http://iterm2.com/features.html is winner for me :) 2017-03-03 22:43:49 but I’ll try alacritty, maybe I’ll be surprised and eventually realize that my current terminal emulator is slow 2017-03-03 22:44:58 iterm looks pretty cool, but I don't think it would fit my workflow well. 2017-03-03 22:45:56 probably it wouldn’t if you’re so strongly used to HW terminals 2017-03-03 22:46:25 but I wonder how you can use it productively… 2017-03-03 22:46:30 jirutka: I have plenty of experience working in GUIs, I just hate it! 2017-03-03 22:47:02 jirutka: In my opinion i3 with some terminal app can be used very closely like a bunch of VTs but better. 2017-03-03 22:47:08 GUIs quite often sucks 2017-03-03 22:47:19 pavlix: agree 2017-03-03 22:47:45 jirutka: GUIs a great for doing research, reading pdfs, etc, but when it comes to text, they generally suck. 2017-03-03 22:48:26 I'm very quickly switching workspaces by number and very quickly open a fullscreen terminal there. Plus a second one if I don't want to use tmux and stuff like that. 2017-03-03 22:48:53 jirutka: I can have 6 vts going with different users logged in, flip between them instantly, and need essentially nothing for hardware to do it 2017-03-03 22:49:07 TemptorSent: I'm very much of the same blood and use UIs mostly when I'm lacking a CLI/TUI alternative. 2017-03-03 22:49:15 pavlix: i3 is very simple and minimalistic, no bullshits and fancy stuff, so it feels closer to hw terminals, but provides essential multi-window features 2017-03-03 22:49:36 TemptorSent: But still I prefer i3+urxvt or a similar combination over Linux VTs. 2017-03-03 22:50:23 Back in the old days, I actually used compiz-fusion to manage a whole crapload of virtual desktops, then set some of those up as full-screen terminals. 2017-03-03 22:50:27 TemptorSent: And yep, I had a really hard time finding something that would fit... I used Gnome, people advised me to try Enlightenment, Awesome, OpenBox, .... 2017-03-03 22:51:23 TemptorSent: All of those projects felt wrong to me in a way. I would say i3 is the first WM that doesn't stand in my way. There are possibly others but I'm happy for now. 2017-03-03 22:51:24 pavlix: Enlightment actually was my preferred wm for quite a while because I could get rid of anything I didn't want and still have some nice features. 2017-03-03 22:52:47 TemptorSent: I used e20 or git versions close to that and it felt like Gnome with a lot more options but undocumented. It didn't help me much and it was rather buggy. 2017-03-03 22:52:58 pavlix: I'll hve to poke at i3 a bit more, it's been a while since id did the WM-search last. 2017-03-03 22:53:16 TemptorSent: And the binary registry-like config database went down for me once or twice. 2017-03-03 22:53:37 pavlix: Wow, up to e20? I think e17 was bleeding edge last I checked. 2017-03-03 22:53:47 uh, binary registry-like config on Linux, this must be killed with fire! 2017-03-03 22:54:03 jirutka: That's Gnome and Enlightenment. :) 2017-03-03 22:54:03 pavlix: Yeah, enlightment has gotten very bloated since the early days. 2017-03-03 22:54:25 jirutka: And both projects are surprisingly so wrong in so many other areas. 2017-03-03 22:54:55 jirutka: I actually have nothing against binary format to be honest, as long as they translate to something I can edit in vim... but they must not break. 2017-03-03 22:55:54 jirutka: Gnome database was stable but inconvenient, Enlightenment one was not safe, and as for journald I lost logs so many times that I just wanted to get rid of it. 2017-03-03 22:56:01 pavlix: I have everything against binary formats *for configuration* 2017-03-03 22:56:03 pavlix: Yeah, I can live with parsible binaries in some cases, but there's no reason to keep the entire configuration in binary rather than just a cached version. 2017-03-03 22:56:22 or logs in binary, similarly bad idea 2017-03-03 22:56:24 TemptorSent: Sure. 2017-03-03 22:57:03 jirutka: Logs in binary have so many theoretical advantages that I sort of like the idea. But journald didn't go well for me. 2017-03-03 22:57:27 jirutka: I hate some attributes of line based text logs. 2017-03-03 22:57:43 Binary logging -- why? 2017-03-03 22:58:12 TemptorSent: because Lennart… 2017-03-03 22:59:17 TemptorSent: Because each record can have all sorts of metadata. And because with all sorts of metadata, a binary form could be easily accessed very quickly, and because text representation of date is almost always wrong. :) 2017-03-03 22:59:33 TemptorSent: this is what happens when some arrogant incompetent idiot get backed by a huge company like RedHat… 2017-03-03 22:59:46 TemptorSent: If there was a *good* implementation, I would gladly use that. 2017-03-03 23:00:57 TemptorSent: You said it right. He got backed and he was doing everything to get where he is. 2017-03-03 23:01:08 ^ sorry jirutka 2017-03-03 23:01:26 pavlix: yes, it makes sense… until FS corruption… then you can’t read anything… 2017-03-03 23:01:39 pavlix: the exact situation when you need the logs most 2017-03-03 23:02:40 jirutka: With FS corruption you (1) might not be able to read text logs either and (2) having binary logs doesn't prevent you from having selected data exported to text logs as well. 2017-03-03 23:03:08 pavlix: there’s much greater chance that you can read at least parts of the logs 2017-03-03 23:03:11 jirutka: That's why I say that the theoretical concept sounds good to me if it's implemented and deployed in ways that suit your needs. 2017-03-03 23:03:19 pavlix: when they are textual 2017-03-03 23:04:06 jirutka: Anyway, what Lennart achieved is amazing. He now has a lifelong job and he can choose among multiple companies to sponsor that job. 2017-03-03 23:04:06 pavlix: yeah, you’re right, the concept is actually okay, the problem is the current implementation 2017-03-03 23:04:39 jirutka: It is that sort of achievement that is amazing for the achiever, not for the rest of the world. 2017-03-03 23:04:45 pavlix: yes, he helped to destroy Linux, really amazing… of course for him it is… 2017-03-03 23:05:03 pavlix: but let’s stop this topic here 2017-03-03 23:05:09 jirutka: He never cared about Linux and he always hated the open source community. 2017-03-03 23:05:16 jirutka: And he was pretty much open about that. 2017-03-03 23:05:20 pavlix: uh, really?! 2017-03-03 23:05:49 pavlix: omfg, so this may not be just a result of his arrogance and incompetence, but his plan! 2017-03-03 23:05:54 jirutka: Or maybe not always but at least all the time I watch is career and read his blogposts and stuff like that. 2017-03-03 23:06:15 *lol* Where's my popcorn? ;) 2017-03-03 23:06:24 jirutka: I wouldn't say plan, rather attitude. Maybe he was capable of planning something, but many things just happened. :) 2017-03-03 23:07:18 doesn’t matter… I really hate him, but didn’t think about the option that he’s actually evil, I just thought that all of this is result of arrogance and incompetency 2017-03-03 23:07:47 Anyway, dev question: When setting up my git repo to add my branch to, should I start by forking the alpinelinux/ports head on github? 2017-03-03 23:07:49 of all the mother of offtopics 2017-03-03 23:07:51 okay, switch the topic 2017-03-03 23:07:54 jirutka: Anyway Lennart showed us how weak the open source community was comparing to what we thought. 2017-03-03 23:08:07 pavlix: well, that’s very true, unfortunately :( 2017-03-03 23:08:29 TemptorSent: yes 2017-03-03 23:08:33 I'm currently tracking git.alpinelinux.org master for my branch 2017-03-03 23:08:34 jirutka: And Gentoo, Alpine, AntiX and others show that we are still there. :D 2017-03-03 23:10:00 TemptorSent: just click on Fork, add this to your remotes, push into it and use e.g. https://github.com/Idnan/github-pr or similar tool to open PR from the branch 2017-03-03 23:10:09 TemptorSent: then you don’t have to touch web interface anymore 2017-03-03 23:10:39 jirutka: Cool, thanks. 2017-03-03 23:11:12 jirutka: And when I think about myself in particular, the events in the past years actually strenghtened my opinion that we need to stay somewhat independent of companies, especially the larger ones, even when we work for them. 2017-03-03 23:11:13 TemptorSent: when someone add a comment, you can reply via email; but you’ll not see a context in email (the line of code relevant to the comment) :/ 2017-03-03 23:11:33 TemptorSent: but there are some CLI tools that can help even with that 2017-03-03 23:11:58 pavlix: totally agree! 2017-03-03 23:11:58 jirutka: I don't so much mind the web for that sort of stuff, it's the actual coding that kills me. 2017-03-03 23:12:04 jirutka: And there are many people like that working for the companies. 2017-03-03 23:13:54 pavlix: I have just people who don’t understand my distrust towards big companies like Amazon or Google :( all ppl are like “I need to deploy this chat app for few people, I definitely want to use AWS or GCP for that, because they are the biggest, so the most reliable, so the best at all!” 2017-03-03 23:14:32 jirutka: I might be using AWS for some stuff actually. :/ :) 2017-03-03 23:15:08 and I’m always like: “WTF, this doesn’t make sense, these services are quite expensive for this use case and there are plenty of other companies that provides various hostings, why the hell you so insist on supporting only the big ones to create a monopoly?!” 2017-03-03 23:16:05 jirutka: +1 2017-03-03 23:16:43 jirutka: It doesn't make sense to use AWS for non-cloudish use any more than any classic VPS provider. 2017-03-03 23:19:05 yeah 2017-03-03 23:32:53 pavlix, jirutka : Have it on github now :) https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts/scripts/mkimage 2017-03-03 23:33:51 great! 2017-03-03 23:34:05 what time is now in USA? :) 2017-03-03 23:34:23 Middle of the afternoon - 15:34 2017-03-03 23:36:16 wow, you’ve basically rewritten it all! 2017-03-03 23:36:31 jirutka: Yeah, pretty much :) 2017-03-03 23:36:52 looks like huge amazing work 2017-03-03 23:37:16 jirukta: Thanks! I'm curious to see what fabled and ncopa think as well :) 2017-03-03 23:38:23 jirutka: There's a fair bit of work left to be done just in terms of cleaning up and documenting everything consistently still. 2017-03-03 23:38:44 I’d like to eventually move these scripts into a separate repository 2017-03-03 23:38:56 jirutka: But the bones are mostly there and so is some of the flesh. 2017-03-03 23:39:53 jirutka: Agreed, they don't belong in the aports tree so much, as they have significant functionality of their own. If mkinitfs is it's own repo, mkimage probably should be as well. 2017-03-03 23:40:44 jirutka: In fact, I could easily see merging mkinitfs, update-kernel, and mkimage into one repository, since they all are tied together. 2017-03-03 23:41:05 now please open a pull request from that branch against alpinelinux/aports:master 2017-03-03 23:41:43 you should see a button for it on https://github.com/TemptorSent/aports 2017-03-03 23:41:44 jirutka: With only a little more effort, I could probably split the load/build logic off from the image functions entirely. 2017-03-03 23:42:56 TemptorSent: +1! 2017-03-03 23:43:39 jirutka: Note -- currently it has 22 commits and isn't smashed. Will that be okay for a pull request, or should I finish the clenaup work before smashing and sending the PR? 2017-03-03 23:44:05 TemptorSent: please don’t smash it, keep the commits 2017-03-03 23:44:37 TemptorSent: Maybe you can create a pull request with a note that it's RFC for now + never squash commits if it actually made sense for you to split them. 2017-03-03 23:44:56 ^ +1 2017-03-03 23:44:59 jirutka: Okay, maybe a update of the dev doc is in order then, because the docs seem vehement on squashing from what I was seeing. 2017-03-03 23:45:07 TemptorSent: It will make sense for others as well. I only squash commits when some of them are fixups or some of them simply don't work with each other. 2017-03-03 23:45:29 TemptorSent: Makes sense. 2017-03-03 23:45:50 TemptorSent: I’ll talk about it with fabled and ncopa, but I’d prefer to extract all commits from aports repo scripts directory into a separate repo and then merge your work into it, unsquashed 2017-03-03 23:45:57 TemptorSent: to keep all history 2017-03-03 23:46:15 jirutka: Sounds good, thanks! 2017-03-03 23:49:08 TemptorSent: my bad memory… I said before that you can’t see context of comments in mails from GH, that’s not true, it includes even context! 2017-03-03 23:49:51 TemptorSent: so you can reply to code review comments even from email exclusively 2017-03-03 23:56:31 jirutka: Very cool! Okay, pull request sent, let me know if you need anything revised right off. 2017-03-03 23:58:38 jirutka: I'm guessing that .travis doesn't have any idea about building images? 2017-03-03 23:58:57 TemptorSent: no, there are no tests for building images 2017-03-03 23:59:33 jirutka: That's one of the things I'd like to add going forward -- at least minimal functional testing and possilby regression testing. 2017-03-03 23:59:57 TemptorSent: +100! 2017-03-04 00:00:25 jirutka: What I really need is for someone on arm to take a look and let me know what I need to do to clean things up there. 2017-03-04 00:01:25 jirutka: Also, the archs are the next thing I'm eyeballing at extracting to their own playground, which would pretty much take care of any remaining hard-coded bits. 2017-03-04 00:02:49 jirutka: I know almost nothing about the rest of the alpine ecosystem beyond the core of packages involved in building images and booting, so any help with design decisions much appreciated. 2017-03-04 00:03:40 jirutka: Oh, and any idea why the hell the vanilla kernel isn't just named vmlinuz-vanilla? It would make life MUCH simpler! 2017-03-04 00:04:40 jirutka: There's some voodo kernel/module land that I haven't even looked at. 2017-03-04 00:06:00 jirutka: But, for instance, you can now easily build your own xen profile and have it actually work now that the hardcoaded check for $profile="xen" is gone and it's implemented as a feature instead. 2017-03-04 00:08:27 pavlix: Anyway, getting back to your original question several hours ago -- it's ready for playing with, and shouldn't be too far from meeting your specific needs. 2017-03-04 00:12:17 Before it actually goes into a live tree, it needs several things: Better documentation of functions, overall user documentation, addition of a couple more calling options to allow injecting nonce data, keys, etc. 2017-03-04 00:14:22 It also needs the features to be refactored out better and overlays deduplicated of common configurations. Every profile with package that depends on udev is going to need it's suite of runlevels and such in the overlay, there's no reason every overlay should try to create the same links. 2017-03-04 00:14:48 Oh, and the build-order code has had absolutely NO testing yet :) 2017-03-04 00:16:12 I just rigged up a simple _before _after _needs _conflicts resolving system (and yes, it does have a watchdog counter in case of infinite loops). 2017-03-04 00:16:25 TemptorSent: That sounds great. I'm going to bed now. Will keep in touch. 2017-03-04 00:17:41 pavlix: Alright, have a great night and I'll talk to you soon. Feel free to drop me an email - chrisgiorgi at gmail. 2017-03-04 00:18:27 jirutka: So, does seeing it in-use help explain the reasoning behind the fkrt utility? ;) 2017-03-04 00:20:53 jirutka: The multiple-instance support isn't used as of yet, but I can easily see the need arrising when we start building multiple outputs at once.. 2017-03-04 00:37:25 Hmm, I wonder when the new apk is going to drop... I REALLY need the cache dir config option! 2017-03-04 00:45:42 jirutka: I should be able to tackle the parallel dispatch soon so we can burn on several cores at once. 2017-03-04 01:12:07 Wow! Was someone listening to me? New apk tools just popped in on apk upgrade :) 2017-03-04 01:12:14 Thank you! 2017-03-04 05:25:02 Hmm, --cache-dir doesn't seem to be doing what I expected.. 2017-03-04 05:27:08 Never mind, I'm an idiot, don't mind me... It helps if you define APK_CACHE_DIR from the getopts BEFORE you try to use it :) 2017-03-04 05:51:20 Anyone have any idea how do I provide a package that needs either python2 or python3? 2017-03-04 05:52:11 pickfire_ : Sorry, haven't gotten too deep into the apk/ abuild system itself yet. 2017-03-04 05:52:27 Needs either or both? 2017-03-04 05:52:32 TemptorSent: So what do you mean? 2017-03-04 05:52:37 TemptorSent: One of them. 2017-03-04 05:53:02 So let's say if the user have python2, it depends on python2, if the user have python3, it depends on python3. 2017-03-04 05:53:04 Hmm, can you evaluate an expression inline? 2017-03-04 05:53:19 But if the user have none, let the user install either python2 or python3 2017-03-04 05:53:31 I don't want the user to install python2 if they have python3 just for this. 2017-03-04 05:53:57 check for a python interpreter of any sort? 2017-03-04 05:53:59 Or I just put depends="python2 python3"? 2017-03-04 05:54:03 Yeah 2017-03-04 05:54:09 any python interpreter will work. 2017-03-04 05:57:26 By the way, how do I enable the user to view /sys/class stuff? 2017-03-04 05:57:35 I need to read from there to be able to get acpi info. 2017-03-04 06:00:35 Hmm, haven't had to do that one yet... probably mdev/udev? 2017-03-04 06:00:59 What about a pre-inst if you can't do it directly in the apk? 2017-03-04 06:10:27 fabled would be the one to ask. 2017-03-04 06:18:51 TemptorSent: Nevermind, I had sent it to the mailing lists. 2017-03-04 06:19:40 Shitty gnupg2 everytime can't decrypt, blame their bad agent. 2017-03-04 06:28:42 Okay, current revision pushed to https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts , diff history at http://termbin.com/aina 2017-03-04 06:30:08 Hmm, do I need to do anything to the PR to update it to the new HEAD of my branch? 2017-03-04 06:31:44 It looks like the answer is no -- it was automagical. 2017-03-04 08:23:38 before I reinvent the wheel: is there any sort of fancy command or API that can determine how much size a package graph will add to an installed system? 2017-03-04 08:24:05 for instance, when you 'apt-get install build-essential' on debian it says "After installation, 46.2 MiB of additional space will be used." 2017-03-04 08:25:02 can apk provide that sort of information? I know it can do individual packages (installed size) in the index, but I am curious if it can do an entire graph (for instance, if you pick a gtk app, include gtk lib itself if not already installed) 2017-03-04 08:27:33 apparently apk add -i does that, but not in machine friendly way 2017-03-04 08:29:24 guess I can look at code for apk add -i... nevermind, sorry for noise :) 2017-03-04 08:31:09 awilfox - sorry, haven't delved into that yet -- fabled or ncopa should know for sure. 2017-03-04 08:46:09 http://git.alpinelinux.org/cgit/apk-tools/tree/src/commit.c#n274 found :) 2017-03-04 08:55:56 Cool! 2017-03-04 09:21:58 Wtf? Did apk just use mv -i or something? Interactive prompt in the middle of my script, that's not good... 2017-03-04 09:40:59 fabled_ : Check your PR review requests -- I've been busy : ) 2017-03-04 10:50:52 'morning SWF 2017-03-04 12:01:24 pickfire: https://wiki.alpinelinux.org/wiki/APKBUILD_examples:Python 2017-03-04 12:02:45 jirutka: Why are you giving me that? 2017-03-04 12:03:01 pickfire: you’ve asked "Anyone have any idea how do I provide a package that needs either python2 or python3?" 2017-03-04 12:03:09 jirutka: Yes 2017-03-04 12:03:23 But the package name isn't py2-ydcv or py3-ydcv 2017-03-04 12:03:25 It's not in pypi 2017-03-04 12:03:42 There is no setup.py 2017-03-04 12:03:45 Just a python file. 2017-03-04 12:03:49 pickfire: can you explain what exactly you need? 2017-03-04 12:04:03 pickfire: create an abuild, install some pkg, or…? 2017-03-04 12:04:07 jirutka: https://github.com/felixonmars/ydcv 2017-03-04 12:04:10 Create a package. 2017-03-04 12:04:28 jirutka: That repo only have a zsh-completion file and a python script 2017-03-04 12:04:41 That is not a pypi project 2017-03-04 12:05:05 So it will be named ydcv instead of py-ydcv 2017-03-04 12:05:53 Should I call it ydcv-py2 and ydcv-py3? 2017-03-04 12:06:39 pickfire: aha, this is not a python library, but a tool that just happens to be written in python; create a package ydcv and make it depend on python3, it’s not needed to support python2 for this 2017-03-04 12:07:16 jirutka: But it can use python2 as well. 2017-03-04 12:07:32 pickfire: and maybe also pypy or any other python interpreter… 2017-03-04 12:07:32 So if others installed python2 and not python3, I see no reason for them to install python3 just for this. 2017-03-04 12:07:42 Oh, yeah 2017-03-04 12:08:06 Or I just create ydcv without any depends? 2017-03-04 12:08:24 pickfire: unfortunately we don’t have support for such situations in apk yet 2017-03-04 12:08:30 Oh 2017-03-04 12:08:39 But doesn't apk have something like provides? 2017-03-04 12:08:43 pickfire: and not adding any python into depends is definitely not right solution 2017-03-04 12:08:44 pacman have that too 2017-03-04 12:09:04 it has, but it doesn’t work for this case 2017-03-04 12:09:13 So something like python2, python3, pypi all provides python 2017-03-04 12:09:30 As well as lua5.1, lua5.2, lua5.3 where all of them provides lua as well 2017-03-04 12:09:49 sry, i don’t have time to explain it now 2017-03-04 12:09:59 this is known issue and I hope that we’re gonna solve it soon 2017-03-04 12:10:16 it’s more complicated than it seems to be though 2017-03-04 12:10:49 if you want to create a package for ydcv, then just make it depend on python3 (b/c this should be the default python) and don’t overcomplicate it 2017-03-04 12:11:13 Huh? 2017-03-04 12:11:14 How come? 2017-03-04 12:11:19 The default python is python2. 2017-03-04 12:11:31 Try `python`, it's python2 2017-03-04 12:12:02 python3 should be the default but it isn't 2017-03-04 12:27:13 pickfire: sry for interruption, I’m at restaurant now and I got a meal 2017-03-04 12:27:43 pickfire: yes, /usr/bin/python is python2, as PEPxxx (don’t remember number) specifies 2017-03-04 12:28:58 pickfire: but default target Python for tools should be python3; python2 is legacy, not under active development and hopefully will be finally retired in 2020 2017-03-04 12:31:00 pickfire: the first paragraph: https://fedoraproject.org/wiki/Packaging:Python#Python_Version_Support, we currently follow the same strategy, at least for new packages 2017-03-04 12:33:30 pickfire: PEP 394 https://www.python.org/dev/peps/pep-0394/; this PEP was published after Arch switched python → python3 that caused a lot of issues 2017-03-04 12:58:44 Ah 2017-03-04 17:08:38 awilfox: if you request interactive-mode by `touch /etc/apk/interactive` it will literally work just like apt :) 2017-03-04 17:08:42 awilfox: may be a good thing for adelie 2017-03-04 17:52:51 kaniini, yeah I was thinking that but my question was more for a packagekit/muon type thing I am thinking/planning related to horizon 2017-03-04 19:46:19 BitL0G1c: let me know on logstash review and i will take care of it 2017-03-04 19:47:07 Good morning kaniini. 2017-03-04 19:56:31 kaniini - yes looking at it now 2017-03-05 06:53:05 Good evening -- for anyone interested I just pushed the lated revision of my rewrite of mkimage to github. See https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts/scripts/mkimage 2017-03-05 06:53:45 Now supports generating ssh keys and autostart features/overlays! 2017-03-05 07:03:35 I am planning to create ranger APKBUILD, should I separate rifle as another package? 2017-03-05 07:30:23 *shrugs* Um, what's ranger? ;) 2017-03-05 07:32:40 Oh, just looked that up -- kinda cool :) 2017-03-05 07:37:57 Huh TemptorSent you really didn't know ranger? 2017-03-05 07:38:08 It's a cool file manager? I feel like it's better than vifm. 2017-03-05 07:38:12 :) 2017-03-05 07:38:19 I seldom use it but it is very useful. 2017-03-05 07:39:23 Yeah, I've been a bit out of the loop. 2017-03-05 07:40:46 TemptorSent: Should I separate the ranger.desktop as well? 2017-03-05 07:40:58 Because *.desktop are totally useless for me. 2017-03-05 07:41:44 *lol* Agreed, but the number of packages might get as long as the number of installed files at that rate :) 2017-03-05 07:42:22 just blackhole the target directory perhaps? 2017-03-05 07:43:12 ln -s /dev/null /path/to/desktop/target ;) 2017-03-05 07:43:41 preceeded by rm -rf /path/to/destktop/target if necessary. 2017-03-05 07:44:06 mount --bind /dev/null /path/to/target works too. 2017-03-05 07:44:36 Oh 2017-03-05 07:44:44 I didn't know I can do that. 2017-03-05 07:45:01 I might just as well ln -s /dev/null /usr/share/doc/ 2017-03-05 07:45:08 *lol* It's a horrible, horrible thing to do. 2017-03-05 07:45:12 How come? 2017-03-05 07:45:16 I only read man pages. 2017-03-05 07:45:36 I never cared /usr/share/doc (unless it is used by the applications like krita) 2017-03-05 07:45:42 But it is a rare case 2017-03-05 07:45:52 Taskwarrior uses them as well. 2017-03-05 07:46:00 Mostly because when you're TRYING to install a doc and you can't figure out why it's not showing up, you'll bash your head into the wall for hours until you remember what you did. 2017-03-05 07:46:11 Desktop files doesn't cost much space but /usr/share/doc sure does 2017-03-05 07:46:29 TemptorSent: I install package-doc just for the man pages. 2017-03-05 07:46:31 nothing else 2017-03-05 07:46:33 :D 2017-03-05 07:46:43 Try mounting a tmpfs to /usr/share/docs, that way things don't complain on install that they can't make the file/directory. 2017-03-05 07:47:27 Oh 2017-03-05 07:47:31 Yeah, I'd really like to see man pages separated from all the other docs. I NEED man pages, especialy on alpine where I need to figure out what the version of a tool I'm using actually supports. 2017-03-05 07:47:40 I wanted that as well. 2017-03-05 07:47:50 I bet I won't ever install -doc if there are -man 2017-03-05 07:48:15 Lots of ways to deal with obnoxious directories you don't care about, depending on what's trying to write to them. 2017-03-05 07:48:33 TemptorSent: But one thing, alpine had definitely loss a lot of man-pages. 2017-03-05 07:48:49 TemptorSent: Maybe we do a feature request for -man? 2017-03-05 07:49:10 mounting/unmounting a tmpfs does a good job of letting things work "normally", then making them disappear. 2017-03-05 07:49:20 As well as that, .1.bz2 would be nice as well 2017-03-05 07:49:36 TemptorSent: No, it doesn't 2017-03-05 07:49:37 Yeah, the documentation situation needs some help. 2017-03-05 07:49:40 Applications can't read 2017-03-05 07:49:54 What do you mean applications can't read? 2017-03-05 07:50:16 open("/usr/share/doc/blah/shit-is-gone", "r"); 2017-03-05 07:50:52 fprintf(stderr, "/usr/share/doc/blah/shit-is-gone cannot be found!"); 2017-03-05 07:51:05 Right, the tmpfs trick is mostly for handling things that install a bunch of crap in a dirctory you don't want. 2017-03-05 07:51:24 The best thing -man 2017-03-05 07:51:24 so you mount it, install the package, and unmount it. 2017-03-05 07:51:39 No 2017-03-05 07:51:56 That's troublesome, I wouldn't wanna mount tmpfs for /usr/share/doc 2017-03-05 07:54:37 Add to .profile: apk-nodoc() { mount -t tmpfs tmpfs /usr/share/doc && apk $@ ; umount /usr/share/doc } 2017-03-05 07:54:59 er, quotes on the $@ may be desirable. 2017-03-05 07:55:56 Oh 2017-03-05 07:55:58 Nice 2017-03-05 07:56:24 *lol* Like I said, it's evil, but it works! 2017-03-05 07:56:32 I would just alias that but it's not the correct way 2017-03-05 07:56:53 Yeah, the alias doesn't do so well with the parameter in the middle of the string. 2017-03-05 07:57:15 but you can alias apk=apk-nodoc 2017-03-05 07:57:58 Then you're set. If you really want to install a package with docs intact, run /sbin/apk 2017-03-05 07:58:53 A nice trick might be to allow a config option/variable that lets you set the tar transform/exclusion list. 2017-03-05 07:59:30 Then apk could let you filter out / transform whatever you want pretty easily. 2017-03-05 08:04:49 That's a project to delve into later -- maybe fabled to could take a quick look at the feasiblity? It should be pretty straightforward to pass through to tar. 2017-03-05 08:17:39 Does anyone know if Bjoren Schilberg hangs out on here? 2017-03-05 08:19:09 TemptorSent: I have found a problem. 2017-03-05 08:19:19 TemptorSent: ranger have doc 2017-03-05 08:19:32 How? https://wiki.alpinelinux.org/wiki/APKBUILD_examples:Python 2017-03-05 08:19:40 That site doesn't show how to do it with docs? 2017-03-05 08:19:55 Or should I just not care about python2? Just go with python3. 2017-03-05 08:27:41 Hmm, I really don't know the abuild format yet... 2017-03-05 08:28:08 Can you easily make a seperate _docs() function to grab them? 2017-03-05 08:28:24 Or are they intermeshed with the rest of the output? 2017-03-05 08:29:18 is py-serial similar in structure by chance? 2017-03-05 08:30:00 That looks like it brute-forces it reasonably well. 2017-03-05 08:33:31 pickfire: Looking at it, it might be as easy as running "make DOCDIR="$subpkgdir/usr/share/docs/ranger" doc" 2017-03-05 08:34:16 Nevermind, I had submitted that package. 2017-03-05 08:34:27 I won't want troubles from py2 and py3 2017-03-05 08:34:33 Just use python3 2017-03-05 08:35:01 TemptorSent: If you want, add your name to the "Contributor" 2017-03-05 08:35:04 ^^ 2017-03-05 08:35:11 Agreed. If you want to use python2, go ahead. 2017-03-05 08:35:18 ?? Contributor on? 2017-03-05 08:36:49 I don't see much reason to have a python2-ONLY system, so anyone using python should assume python3 is supported by default and python2 requires explicit dep. 2017-03-05 10:14:43 Alright, time for me to call it -- g'night all! 2017-03-05 10:20:30 sorry but py3 is a failed academic attempt which is now very similar to systemd is being pushed down the throats of everyone. 2017-03-05 10:26:47 thats an impressive amount of nonsense in a single sentence 2017-03-05 10:31:49 agreed 2017-03-05 10:35:12 why you think? 2017-03-05 10:41:23 yGweSm1OzVHe: No 2017-03-05 10:41:30 yGweSm1OzVHe: python 3 is not being pushed down anyone's throats 2017-03-05 10:41:36 both 2 and 3 are readily available on pretty much any distro 2017-03-05 10:41:37 python2 is going to die in 2020 2017-03-05 10:42:36 Patchwork moderator: Please kill #3103 2017-03-05 10:42:43 Patchwork moderator: Please kill %3103 2017-03-05 10:42:49 So confusing 2017-03-05 10:43:55 TemptorSent: I mean just contribute by adding python2 subpackage to ranger and your name as well. I am currently using python3 only. 2017-03-05 10:43:59 asie: maybe your throat is less sensitive than mine 2017-03-05 10:44:05 No need extra 100MB for nothing 2017-03-05 10:44:26 yGweSm1OzVHe: https://pythonclock.org/ 2017-03-05 10:44:27 yGweSm1OzVHe: feel free to fork python 2 and maintain it if that's a problem for you 2017-03-05 10:44:29 at least one person has done tht 2017-03-05 10:44:31 that* 2017-03-05 10:44:38 Yes 2017-03-05 10:44:40 that pythonclock is part of my throatpain 2017-03-05 10:44:47 Haha :D 2017-03-05 10:45:01 yGweSm1OzVHe gives me stomach pain 2017-03-05 10:45:08 ACTION laugh too much 2017-03-05 10:45:19 pickfire: Oh, perhaps if I end up actually finding a use for python I'll worry about packaging for it ;) 2017-03-05 10:45:39 ? 2017-03-05 10:46:02 TemptorSent: So you want to package it with python2 only if you find that python2 is useful? 2017-03-05 10:46:41 By 2020, I bet there is python4 and there will be as much pain as now. 2017-03-05 10:46:49 I have hardly written a line of python. As far as packaging, I was thinking a bit further upstream, in the tool rather than the package. 2017-03-05 10:47:27 Just like lua5.1, lua5.2, lua5.3. Pickfire forsee python2, python3, python4 by 2020. The difference is that lua uses 1M and python uses 100M 2017-03-05 10:47:53 Yeah, python is horribly bloated IMHO. 2017-03-05 10:48:03 Yes and I still write in python. 2017-03-05 10:48:16 Why? Pandas have no alternative in C. 2017-03-05 10:48:40 TemptorSent: Are you a patch maintainer? 2017-03-05 10:48:46 moderator* 2017-03-05 10:48:54 patchwork moderator* 2017-03-05 10:49:38 pickfire: Nope, I'm just getting involved in Alpine dev, and haven't really been active on the scene in general in a number of years now. 2017-03-05 10:49:42 Note: docker is now uses monthly releases, abump docker-17.03.0-ce fails 2017-03-05 10:49:55 TemptorSent: How to get involve in Alpine dev? 2017-03-05 10:50:41 I want to fix that broken wget -s in abuild, I have sent a patch to alpine-devel@lists.alpinelinux.org, is that correct? 2017-03-05 10:50:47 I saw a need in mkimage that I wanted to fill for my own purposes and brought it up here about two weeks back. 2017-03-05 10:51:16 I haven't actually even subscribed to the lists yet myself :) I shold probably do that. 2017-03-05 10:51:18 I want to fix nlplug-findfs as well, if not everytime I boot to recovery shell. 2017-03-05 10:51:37 That's more up the line I'm working on right at the moment. 2017-03-05 10:51:42 Huh? 2017-03-05 10:51:52 TemptorSent: You're working on that? 2017-03-05 10:52:03 I'm rebuilding the entire image system, including overlays. 2017-03-05 10:52:34 I haven't actually started work on the guts of mkinitfs and friends, but that's next on the list. 2017-03-05 10:52:55 TemptorSent: Oh, that's even worse. I don't even know how those overlays stuff works. 2017-03-05 10:53:05 By tomorrow I should have the whole structre working 2017-03-05 10:53:18 I mean I am just interested to know how alpine bootloop, like you can plug out usb drive after booting. 2017-03-05 10:53:22 No, I mean I have an overlay BUILDER to go with th image builder :) 2017-03-05 10:53:34 overlay BUILDER? 2017-03-05 10:53:49 hang on... 2017-03-05 10:53:56 Just bump me 2017-03-05 10:54:46 pickfire: See https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts/scripts/mkimage 2017-03-05 10:55:52 I have just a little more surgery to do and I'll have a nice, stand-alone plugin loader utility that can be used for whatever you want :) 2017-03-05 10:56:31 Huh? 2017-03-05 10:56:49 Take a look at the code :) 2017-03-05 10:57:39 No doc? 2017-03-05 10:57:41 Basically, it will be able to handle making a proper initfs with whatever you want enabled from a simple profile. 2017-03-05 10:57:58 *lol* Nope, not yet. 2017-03-05 10:58:00 ACTION think it is something to build the alpine iso 2017-03-05 10:58:25 Currently, yes - that is the original use. 2017-03-05 10:58:31 I really hope that there is doc for alpine. 2017-03-05 10:59:16 I'll spend some time documenting once everything has stabilized a bit. I'm not sure what's actually left of the original code at this point... probably not much :) 2017-03-05 10:59:28 kaniini: What do you think if I add apk.1? 2017-03-05 11:00:06 Of course, I will write it in mdoc(1) instead of roff(1). 2017-03-05 11:01:17 I just got it to the point of building a complete image with overlays and all, including autogenering ssh keys. It knows how to handle xen in any profile... 2017-03-05 11:01:38 ZFS/NFS/iSCSI root fs support coming soon :) 2017-03-05 11:02:29 :) 2017-03-05 11:02:45 TemptorSent: How is that useful? 2017-03-05 11:03:00 ...so yeah, I'll be poking at the guts of the initfs code. 2017-03-05 11:03:25 Write profile, build image, run image :) 2017-03-05 11:04:45 Where profile has the ability to include other profiles, but more importantly, can be built with features that specify functionality rather than lists of packages. 2017-03-05 11:05:10 And the overlays are built based on the profiles/features specified. 2017-03-05 11:05:33 11:46:40 pickfire │ By 2020, I bet there is python4 and there will be as much pain as now. 2017-03-05 11:05:37 http://www.curiousefficiency.org/posts/2014/08/python-4000.html 2017-03-05 11:05:57 So you need a dozen differnt rpi machines all with their own setup? No problem :) 2017-03-05 11:06:10 Weird, I am not bumped when Shiz message me. 2017-03-05 11:06:57 Anyway, I need sleep -- I can't read the screen anymore without having to stick my face in it. 2017-03-05 11:07:14 Good night, sleep tight and don't let the bed bugs bite. 2017-03-05 11:07:26 i'm a long time python dev, and i did my share of porting py2 to py3 in the end it did only cause extra work and no extra benefits, from my experience neither people still get confused and fuck up raw string/unicode handling as they did with py2. as far as i see py3 is an attempt to fix things they fucked up in py2 (like /bin/init) and now they're pushing their "solution" (which dose not make things any 2017-03-05 11:07:28 better, but worse in some regards) instead of trying again to fix the things correctly 2017-03-05 11:07:58 Yeah.. the spiders ar the ones I worry about.. they've been trying to share the bed with the constant storms out lately. 2017-03-05 11:08:22 please instead of going adhominem and laugh at someone try to understand and argue about the topic, no personal attacks, lets leave that to the systemd maintainers, let's not have the same style 2017-03-05 11:09:35 yGweSm1OzVHe: imagine if they released py4 now which broke string handlig again 2017-03-05 11:09:41 they won't 2017-03-05 11:09:46 the situation is a bit different because you can't have ten pythons the way you have ten init systems 2017-03-05 11:09:53 well, you can, but they're just ten languages 2017-03-05 11:09:57 just ask perl, or lua 2017-03-05 11:10:05 (lua 5.1 is not 5.2 is not 5.3) 2017-03-05 11:11:00 also, porting is always the raw end of th estick 2017-03-05 11:11:03 exactly, i consider py3 a different language from py2 but they share a lot in common, but then i also wrote function bodies that are also c and python at the same time. 2017-03-05 11:11:19 yup, this is the correct way of thinking. 2017-03-05 11:11:28 my only issue is your characterisation of python 3 as being pushed down anyone's throats 2017-03-05 11:11:30 if so, why hate py3 for being "pushed on"? you don't HAVE to use it like you HAVE to use systemd 2017-03-05 11:11:33 on a systemd-based distro 2017-03-05 11:11:33 it reeks of weird conspiracy theories 2017-03-05 11:11:39 people write python3 because they prefer python3 2017-03-05 11:11:45 the only problem people have with systemd is that you can't remove it 2017-03-05 11:11:48 not sure about that 2017-03-05 11:11:49 python3 is one apk command away 2017-03-05 11:11:51 i only write python3 when i write python code because i find python2 unpleasant to write i 2017-03-05 11:11:53 there's social pressure 2017-03-05 11:11:53 n 2017-03-05 11:12:00 yGweSm1OzVHe: there's always social pressure 2017-03-05 11:12:05 there's social pressure to use windows 2017-03-05 11:12:12 or steam 2017-03-05 11:12:16 or skype, or facebook 2017-03-05 11:12:17 eeek 2017-03-05 11:12:24 there's always social pressure to varying degrees, but that's not relevant 2017-03-05 11:12:31 social pressure is normal - people tell people to use what works for them 2017-03-05 11:12:35 what's relevant is the characterisation of it being pushed down people's throats 2017-03-05 11:12:43 social pressure is not necessarily malicious 2017-03-05 11:12:45 and is not automatically malicious 2017-03-05 11:12:47 it's human. 2017-03-05 11:12:54 you wrote "people write py3 because they like it" not all of them like it, som eof them only do it because of the pressure 2017-03-05 11:12:59 then they're weak 2017-03-05 11:13:02 which is both unpleasant and creates an unnecessarily hostile environment for people who like and prefer python3 2017-03-05 11:13:06 i contest your generalization from your own perspective 2017-03-05 11:13:11 and i contest yours 2017-03-05 11:13:26 if they let other people influence their decisions, then they shouldn't do that. simple 2017-03-05 11:13:31 unless you have numbers of people who cave in to python 3 because of peer pressure, a conclusion of "being pushed down people throats" is both hasty and toxic 2017-03-05 11:13:32 as in 2017-03-05 11:13:34 if it bothers them 2017-03-05 11:13:53 if it bothers you that other people want you to use py3, start hanging out with other people. there's literally billions to choose from. 2017-03-05 11:14:01 s/that other people/that people/ 2017-03-05 11:14:06 how about having both? 2017-03-05 11:14:11 both py2 and py3? 2017-03-05 11:14:13 i have no problems with coexistence 2017-03-05 11:14:14 (also, there's nothing "academic" about python3) 2017-03-05 11:14:20 that's currently the case 2017-03-05 11:14:23 one doesn't conflict with the other 2017-03-05 11:14:26 except for /usr/bin/python 2017-03-05 11:14:34 replacing py2 with py3 is not coexistence 2017-03-05 11:14:38 who is replacing it? 2017-03-05 11:14:40 where? 2017-03-05 11:14:54 people move on from py2 to py3 because they want to do it; same reason people move on from lua 5.2 to 5.3 2017-03-05 11:14:59 (and same reason mike pall of luajit doesn't...) 2017-03-05 11:15:08 scroll back to the discussion aboug gdb depending on py2 2017-03-05 11:15:11 about 2017-03-05 11:15:20 it's not like python 2 software is retroactively disappearing from the internet 2017-03-05 11:15:33 yGweSm1OzVHe: oh, yes; people will generally pick the more supported option for default packaging 2017-03-05 11:15:35 until it is, "replacing" is hardly a honest way of putting it 2017-03-05 11:15:45 if you can choose an older, unmaintained version of a software 2017-03-05 11:15:47 or a newer, long-term one 2017-03-05 11:15:52 which would you pick, as a system maintainer? 2017-03-05 11:16:00 if you want an OS which bends to your will, use Gentoo 2017-03-05 11:16:18 can we not serve all needs? 2017-03-05 11:16:20 no? 2017-03-05 11:16:21 no 2017-03-05 11:16:24 do we have to pick sides? 2017-03-05 11:16:29 why? 2017-03-05 11:16:29 yes, unless you want bloat 2017-03-05 11:16:33 if you want to complain about picking sides 2017-03-05 11:16:34 why bloat? 2017-03-05 11:16:40 including both py2 and py3? 2017-03-05 11:16:42 complain to gdb which apparently only allows ONE python version to be used 2017-03-05 11:16:43 (and py4 and py5...) 2017-03-05 11:16:44 this is not alpine's issue 2017-03-05 11:16:45 also, ^ 2017-03-05 11:16:47 it's gdb's 2017-03-05 11:16:53 why not gdb-nopy, gdb-py2, gdb-p3? 2017-03-05 11:16:59 because that's 3x the build server effort 2017-03-05 11:17:07 help pay for the build server and i'm sure you'll be given an OK 2017-03-05 11:17:13 but the build server is a limited resource 2017-03-05 11:17:20 are you paying for it? 2017-03-05 11:17:24 no, but someone is 2017-03-05 11:17:28 oh. 2017-03-05 11:17:32 it's not a free resource 2017-03-05 11:17:36 packages need a server to be built on 2017-03-05 11:17:40 more packages = slower builds on updates 2017-03-05 11:17:44 PHP had the same discussion 2017-03-05 11:17:46 the issue with option packages like that is that they exponentially grow with the number of options 2017-03-05 11:17:49 we argued which PHP versions to keep 2017-03-05 11:17:52 out of 5.6, 7.0, 7.1 2017-03-05 11:17:57 at some point, we as a distro have to draw a line 2017-03-05 11:17:59 but each version is effectively 10-15 additional packages 2017-03-05 11:18:02 so? 2017-03-05 11:18:05 x 5-10 architectures 2017-03-05 11:18:07 alpine packaging is fairly flexible, but it's not portage 2017-03-05 11:18:09 that's HOURS of building effort 2017-03-05 11:18:13 it will never have something like USEflags 2017-03-05 11:18:16 for the low resources alpine has 2017-03-05 11:18:21 also, hard disk space 2017-03-05 11:18:27 servers have a limited supply of that as well 2017-03-05 11:18:35 i'm not saying draw the line at gdb/gdb-py2/gdb-py3 btw 2017-03-05 11:18:43 we're talking in general 2017-03-05 11:18:44 just making a general picture that we'll never be as flexible as e.g. gentoo 2017-03-05 11:18:51 the question of supporting multiple user configuration 2017-03-05 11:18:55 versus trying to pick a few common use cases 2017-03-05 11:19:05 as i said, if you want to bend a system to your will, you need a source-based distribution 2017-03-05 11:19:11 (also, just saw gdb installs /usr/bin/run; eww) 2017-03-05 11:19:12 or at least a package overlay on top of alpine, which isn't hard to make 2017-03-05 11:19:18 alpine is one of the easiest distros to make a repo for 2017-03-05 11:19:54 e.g., taking a random gentoo package 2017-03-05 11:20:06 yGweSm1OzVHe: gdb-py2 and gdb-py3 would mean someone at alpine has to test both to ensure they both work correctly, too 2017-03-05 11:20:11 tester effort is even more "expensive" than buildserver effort 2017-03-05 11:20:17 gdb: Installed versions: 7.10.1(22:31:03 17/02/16)(client -expat -lzma -multitarget -nls -python -server -test -vanilla PYTHON_SINGLE_TARGET="python3_3 -python2_7 -python3_4" PYTHON_TARGETS="python3_3 -python2_7 -python3_4") 2017-03-05 11:20:42 if we were to embed all the possible options here, we'd have 2**12 gdb packages 2017-03-05 11:20:43 and as you said 2017-03-05 11:20:49 (4096) 2017-03-05 11:20:49 it's not about gdb-py2 and gdb-py3 2017-03-05 11:20:56 the same applies to every scriptable piece of software in the world, period 2017-03-05 11:21:25 pleasing everyone in this regard would cost alpine gigabytes of hard drive space, hours of buildserver effort and potentially days of testing effort 2017-03-05 11:21:36 for the benefit of a very small fraction of the userbase vs. things like adding new features or packages 2017-03-05 11:21:50 py[23]-latest would be enough i guess 2017-03-05 11:21:56 but py2 and py3 is available 2017-03-05 11:22:03 it's just that gdb forces you to choose one or the other 2017-03-05 11:22:11 and alpine goes with the more maintained one of the two 2017-03-05 11:23:11 as i said, if we make an exception for python 2017-03-05 11:23:14 do we also make an exception for lua? 2017-03-05 11:23:17 or php? 2017-03-05 11:23:18 what about java? 2017-03-05 11:23:24 or ruby? 2017-03-05 11:23:30 or node.js? or web browser versions? 2017-03-05 11:23:49 the reason we have hundreds of Linux distributions is because they all cater to a different audience 2017-03-05 11:23:50 those are not different languages ,but different versionsof the same language 2017-03-05 11:23:55 yGweSm1OzVHe: lua 5.1 is not 5.2 is not 5.3 2017-03-05 11:23:59 they're all incompatible with each other 2017-03-05 11:24:01 in some ways 2017-03-05 11:24:06 luajit is also not lua 5.1 or is it 5.2 2017-03-05 11:24:07 i dont consider py3.2 a differnt language from py3.7 2017-03-05 11:24:16 php 5.5, 5.6, 7.0, 7.1, etc. are all incompatible as well 2017-03-05 11:24:26 do they have backward compatible changes without any benefits? 2017-03-05 11:24:29 java 7 and 8 also have behaviour differences. i ran into one just recently 2017-03-05 11:24:38 yGweSm1OzVHe: NONE of them are 100% compatible with each other 2017-03-05 11:24:43 ALL of them have some breaking changes, even if very very small 2017-03-05 11:24:51 Ruby might not, I suppose. 2017-03-05 11:24:55 yes, but do the backward incompatible changes actually have a benefit? 2017-03-05 11:25:02 Benefit is subjective. 2017-03-05 11:25:05 not really 2017-03-05 11:25:11 yes, really. 2017-03-05 11:25:16 Mike Pall considers Lua 5.3 a bad design and made LuaJIT cater to Lua 5.2. 2017-03-05 11:25:17 show me the benefits. the subjective ones please 2017-03-05 11:25:22 Most other Lua developers consider 5.3 a good design. 2017-03-05 11:25:24 Look into it. 2017-03-05 11:25:35 let's focus on py first before we shift goalposts 2017-03-05 11:25:44 No, because you're thinking from YOUR personal problem's perspective 2017-03-05 11:25:48 no 2017-03-05 11:25:49 whereas I'm trying to explain why THIS problem cannot be solved 2017-03-05 11:26:01 if we make an exception for Python , we either make an exception for every such case 2017-03-05 11:26:07 ...or we're a bunch of hypocrites 2017-03-05 11:26:10 i actually postulate that py3 is not an improvement to py2 2017-03-05 11:26:22 according to some key figures, Lua 5.3 is not an improvement to Lua 5.2 2017-03-05 11:26:26 and this is what i call a failed academic experiment 2017-03-05 11:26:27 according to other key figures, it is 2017-03-05 11:26:49 i'm trying to tell you that Python is not the only case in the world where two incompatible versions created a schism 2017-03-05 11:26:51 https://up.shiz.me/M2I2MmU3.png 2017-03-05 11:26:53 here's a 2013 survey 2017-03-05 11:27:06 can you drop the point now 2017-03-05 11:27:08 again, no problem with having a new language. 2017-03-05 11:27:25 No, you're clearly having a problem with it. 2017-03-05 11:27:33 Because gdb only lets you be compatible with one or with the other. 2017-03-05 11:27:40 And there's a lot of other software which does this as well. 2017-03-05 11:28:11 Shiz: can you tell me the how was this sampled? 2017-03-05 11:28:13 >Note that "python" plugin in WeeChat can support only one version, so it can be a 2.x or 3.x, not both at same time. 2017-03-05 11:28:15 This is an example, yGweSm1OzVHe 2017-03-05 11:28:34 this is another piece of software which would require two separate builds 2017-03-05 11:28:58 "The survey was publicized via posts to comp.lang.python, python-dev and hacker news." 2017-03-05 11:29:48 While WeeChat is a bit different in that it recommends Python 2 for compatibility reasons, does that mean we should not support Python 3? I mean, gdb, in your hypothetical scenario, has versions for both. 2017-03-05 11:30:25 also relevant: 2017-03-05 11:30:30 http://ianozsvald.com/wp-content/uploads/2016/06/work_python.png vs http://ianozsvald.com/wp-content/uploads/2016/06/home_python.png 2017-03-05 11:30:34 peer pressure, huh :-) 2017-03-05 11:30:45 Shows a clear split. 2017-03-05 11:30:51 More Python 2 at work due to older codebases. 2017-03-05 11:31:04 python-dev and python.lang seems to be biased as a crowd, no? 2017-03-05 11:31:18 Every human being is biased. 2017-03-05 11:31:23 python-dev, maybe 2017-03-05 11:31:30 comp.lang.python is a user's community 2017-03-05 11:31:51 python-dev doesn't have enough active distinct users to shift the results of the survey that much 2017-03-05 11:31:54 on its own 2017-03-05 11:35:14 btw do you know who did this survey and what is their bias? 2017-03-05 11:36:01 maybe the solution is to code up a a py-shim ;) which you can plug in either py you want without recompiling. 2017-03-05 11:36:14 the python C API doesn't work that way. 2017-03-05 11:36:19 meh 2017-03-05 11:36:31 what's the problem with it? 2017-03-05 11:36:50 significant architectural changes between python 2 and 3's C API 2017-03-05 11:40:00 yGweSm1OzVHe: if you're willing to put in the effort to make one 2017-03-05 11:42:15 Looking from here, Shiz and asie looks like the same person talking. 2017-03-05 11:42:35 pickfire: Clearly, it's a conspiracy 2017-03-05 11:42:55 An agreement to perform together an illegal, wrongful, or subversive act. 2017-03-05 11:43:20 Subversive sounds about right. 2017-03-05 11:44:05 Sex and creativity are often seen by dictators as subversive activities 2017-03-05 11:44:13 But yeah, I think I understand it now. 2017-03-05 11:44:44 In a nutshell, gdb should default to python2 and should never use python3 right? 2017-03-05 11:45:04 pickfire: In a nutshell, we should provide a package for "gdb with python2" and "gdb with python3". 2017-03-05 11:45:21 Ah, that's what I want. But I thought you all are against that? 2017-03-05 11:45:23 Which makes sense on first glance, but then we'd have to apply the same logic to /everything/ 2017-03-05 11:45:30 That's my only problem with it. 2017-03-05 11:45:37 It's a great idea, but Gentoo does it far better than Alpine ever will. 2017-03-05 11:45:46 ACTION don't quite like gentoo 2017-03-05 11:45:52 Need to compile everything myself. 2017-03-05 11:47:31 whats the usecase for a gdb with python2 2017-03-05 11:47:38 any pre-existing commonly-used scripts that break? 2017-03-05 11:50:10 there's a bunch of reverse engineering tools based on this 2017-03-05 11:50:18 I don't use gdb with python personally. 2017-03-05 11:51:19 right 2017-03-05 12:55:24 pickfire: "Patchwork moderator: Please kill #3103" … kill as resolved? 2017-03-05 12:57:21 fyi, I’ll rather ignore all this bullshit about python2 vs. python3; python2 is dead, not under active development anymore, period. 2017-03-05 12:58:12 and no, there will definitely not be Python 4 at 2020, this is yet another total bullshit 2017-03-05 13:03:16 maybe py2 is feature complete? an doesn't need more development? 2017-03-05 13:03:43 you can’t be serious… 2017-03-05 13:04:09 why? what is missing from py2? 2017-03-05 13:04:20 i mean besides the stuff they messed up again in py3? 2017-03-05 13:04:20 maybe broken unicode support? 2017-03-05 13:04:26 also broken in py3 2017-03-05 13:04:33 any many other things that are fixed fixed in python3 2017-03-05 13:05:10 it’s not very good even in py3, but still much better than py2 2017-03-05 13:05:25 people still fuck it up like in py2 2017-03-05 13:06:23 sorry, but I don’t wanna discuss this topic; this is the most annoying thing in entire Python universe 2017-03-05 13:07:32 Python is represented by Python developers, they’re goal is very clear, if you disagree, then fork Python; the only reason why Python 2 is still somehow maintained are big companies that has written piles of shit in it and are not willing to even try to port it into Python3 2017-03-05 13:08:01 bec 2017-03-05 13:08:09 ause there is no benefit in porting to py3 2017-03-05 13:08:17 the bugs they fixed are not fixed well 2017-03-05 13:08:21 of course there are 2017-03-05 13:08:29 just read what’s new in python 3 2017-03-05 13:08:37 you can do this yourself 2017-03-05 13:09:06 actually unicode support is decent but not intuitive 2017-03-05 13:09:49 well, yes, as probably everything in Python it’s kinda weird and half-working… but that’s for a different discussion 2017-03-05 13:09:56 agreed ;) 2017-03-05 13:10:23 a new version should make things better not different 2017-03-05 13:10:46 they did it better, just not great 2017-03-05 13:11:20 if they would fixed it properly, then it would create even more incompatibilities and more hate among conservative devs 2017-03-05 13:12:21 maybe, we don't know 2017-03-05 13:12:52 yeah 2017-03-05 14:40:14 jirutka: No, that's duplicate. 2017-03-05 14:41:40 jirutka: Wait, I am talking about %3103 2017-03-05 14:54:03 I have a question regarding the new soon to be policy of check() in APKBUILD, any framework you plan on using to implement unit tests for shell projects? 2017-03-05 14:54:56 In alpine I found testing/bats unmaintained/roundup, but nothing in main or community... 2017-03-05 17:19:42 HRio: that’s a good suggestion, we should move bats at least into community 2017-03-05 17:24:27 please, check first that there are no vampire bats among them 2017-03-05 17:26:33 TemptorSent: I don’t know about many Python projects that provides man pages (actually no one I can remember) and -doc subpkg with HTML pages are IMHO useless, you can just read them online w/o messing your system with it; maybe that’s why there are no examples for py- pkgs with -doc :/ 2017-03-05 17:30:20 pickfire: ranger? IIRC it’s in unmaintained, it was horribly broken 2017-03-05 17:31:17 jirutka: actually, it's not THAT uncommon for some projects to only have .html docs but none of those online 2017-03-05 17:31:25 i've countered a few of those in the past yearss, sadly 2017-03-05 17:31:49 also, can't expect the users to always have internet connectivity, so i'd advocate for keeping .html in docs 2017-03-05 17:32:51 okay 2017-03-05 17:33:38 example: https://github.com/CoreSecurity/pcapy 2017-03-05 17:33:56 docs only in repo, not online 2017-03-05 17:33:57 (pcapy.html) 2017-03-05 17:34:06 I personally don’t mind it, b/c I almost never install -doc subpkgs; I have dozens of system with Alpine and prefer not to pollute it with something I can easily find online or on some non-production system that I use like a playground 2017-03-05 17:34:21 sure, but then you can just not install -doc :) 2017-03-05 17:34:24 i do the same on prod systems 2017-03-05 17:34:25 aha, that’s okay, it’s just single html 2017-03-05 17:34:45 but some packages generates many HTML pages, including styles and images… sometimes it has many megabytes… 2017-03-05 17:35:05 yeah, but even stuff like that... 2017-03-05 17:35:07 idk, it can be useful 2017-03-05 17:35:19 luckily we split -doc packages up in the first place :) 2017-03-05 17:36:56 yeah, if the pkg’s build script atuomatically installs relevant docs, then I usually keep it and only move it into /usr/share/doc/$pkgname, if it does not install there by default (and add -doc) 2017-03-05 17:51:10 what's up? 2017-03-05 18:03:15 jirutka: The package pickfire was working on (ranger) did happen to have both pydoc and man pages, but my wish to have man pages able to be installed WITHOUT installing the rest of the docs in /usr/share/docs is not python specific. 2017-03-05 18:04:22 TemptorSent: we currently don’t have any -man subpkgs, just -doc 2017-03-05 18:04:33 jirutka: Exactly :) 2017-03-05 18:05:13 TemptorSent: there’s also a special package "docs" that pulls -doc for very installed pkg 2017-03-05 18:05:20 jirutka: As a general solution, setting an environment variable to pass to tar on extraction would probably kill many birds with one stone. 2017-03-05 18:05:36 TemptorSent: no 2017-03-05 18:05:39 nonooo 2017-03-05 18:05:50 ?? 2017-03-05 18:06:16 TemptorSent: I agree that in same cases it would make sense to differentiate man pages and rest of the docs 2017-03-05 18:06:30 What is so horrible about filtering tar? 2017-03-05 18:07:39 I'd love to see a "man" package similar to docs that ONLY installs the man pages. 2017-03-05 18:08:15 It could use the -doc apks, but filter the file list on extraction. 2017-03-05 18:10:10 the distinction between man pages and other docs makes sense for me, but definitely not via filtering tar 2017-03-05 18:10:15 but as normal subpkg 2017-03-05 18:13:09 jirutka: Filtering tar or explicitly listing files for extraction is a simple way of allowing users more flexabilty in what they install on a minimal system wihout needing to break packages down to infintessimals. 2017-03-05 18:14:03 it’d generate many problems and increase complexity 2017-03-05 18:16:01 jirutka: It would probably be better than my current solution, which is to blackhole the output directory on extract :) 2017-03-05 18:16:46 that’s your own solution on your own computer, you can do whatever you want with your computer ;) 2017-03-05 18:16:47 jirutka: Current solution: in .profile: apk-nodoc() { mount -t tmpfs tmpfs /usr/share/doc && apk $@ ; umount /usr/share/doc } 2017-03-05 18:17:30 jirutka: Yeah, it's ugly as sin, but it does work. 2017-03-05 18:17:40 that’s good for you :) 2017-03-05 18:18:40 yeah, ranger is a good example of pkg that puts bloat of some examples and other things into -doc 2017-03-05 18:18:41 jirutka: I guess what I'd like to see is the ability to use a tar exclude file to specify files/directories to NEVER be extracted by apk. 2017-03-05 18:19:13 I’d like to see if anyone actually use these… 2017-03-05 18:20:01 in this case, it’d make sense to extract these examples into a -examples subpkg 2017-03-05 18:20:16 jirutka: It doesn't make man packages magically happen, but it would let users set /usr/share/docs as an excluded directory, which is just as good for mnay practical applications (think rpi) 2017-03-05 18:20:42 or maybe just provide something like -src subpkg with full sources… 2017-03-05 18:21:06 no, it’s not a solution, it’s just a dirty hack 2017-03-05 18:21:18 jirutka: Filtering becomes almost a necessity when you start dealing with network file systems and shared resources. 2017-03-05 18:22:12 https://hastebin.com/iwexipomix.txt … most of the files in /usr/share/doc are actually not a documentation, that’s the problem 2017-03-05 18:22:32 jirutka: For instance, /usr/share/doc on a RO NFS mount or CD. 2017-03-05 18:24:09 jirutka: The other REALLY useful tool is --xform, which would make setting up configurations with odd mount points much easier. 2017-03-05 18:24:15 TemptorSent: you can suggest this to fabled, maybe he’ll have different opinion about it than me; and more importantly he’s the person who is competent to decide this, not me 2017-03-05 18:25:16 jirutka: Fair enough. I'm looking at the types of configurations *I* need to deal with, which may not be as common as most. 2017-03-05 18:27:08 TemptorSent: I agree with distinction between man pages and rest of the docs, this makes sense for me too, and we already have a mechanism for this – subpackages, not needed to implement some new hacks for it 2017-03-05 18:28:30 personally I’d just remove anything but man pages or plain text files from -doc, but understand that especially desktop users have quite different needs 2017-03-05 18:28:32 jirutka: Here's a use case: I run postgresql as my database server and need to relocate it's root directory to /datastore/postgresql on the machines that are running in media mode. 2017-03-05 18:29:02 you can do that in /etc/conf.d/postgresql 2017-03-05 18:29:15 jirutka: Yeah, I don't much care about the images, examples, etc. When I need docs, I want docs. 2017-03-05 18:29:16 if you mean data directory… 2017-03-05 18:29:45 jirutka: Not just data directory, but the whole install. 2017-03-05 18:30:09 ranger segfaults 2017-03-05 18:30:24 the same problem as in the previous package that is currently in unmaintained 2017-03-05 18:30:30 just run ranger and it segfaults 2017-03-05 18:30:38 IIRC ^7heo_ have tried to fix it before…? 2017-03-05 18:31:02 jirutka: The reason being that postgresql version needs to match database version, and our boot media may have an old version. 2017-03-05 18:31:24 and…? 2017-03-05 18:31:56 aha, I see where it’s going 2017-03-05 18:32:25 but pgsql dbs are already installed in a version-segregated directory ;p 2017-03-05 18:32:36 I have probably the same problem… I still have PostgreSQL cluster on Gentoo and here I can install multiple versions in parallel which is needed when upgrading between minor versions 2017-03-05 18:33:32 I don’t know how to simply do this on Alpine… IIRC I’ve already suggested to maintain add version suffix into postgresql pkg and always mantain at least two versions, so people can upgrade 2017-03-05 18:35:07 Shiz: yes, but for example runscript is not version-segregated and you cannot install two different version of PgSQL on Alpine, can you? 2017-03-05 18:39:59 jirutka: Yeah, it's a real problem for me anyway, and it sounds like it's bit you too. 2017-03-05 18:40:35 TemptorSent: yes and I’m gonna migrate our PgSQL instances soon, so I should somehow solve it asap :/ 2017-03-05 18:41:06 jirutka: That's why tar filter/xform on extract would be very useful on my end. If nothing else, I could version the pgsql binaries names directly. 2017-03-05 18:41:26 TemptorSent: I don’t see how you can solve this issue with tar filtering… 2017-03-05 18:41:36 jirutka: I throw around some databases with many gigs of data in them. 2017-03-05 18:42:24 btw how can I respond to a patch in Patchwork without subscribing to aports ML? 2017-03-05 18:43:05 I don’t know what magic did Patchwork use to match emails into threads, if only subject is sufficient… 2017-03-05 18:43:50 s/respond/reply/ 2017-03-05 18:43:50 jirutka: tar --xform='s|postmaster|postmaster-$version|' :) 2017-03-05 18:44:14 no, this would not actually help 2017-03-05 18:44:36 b/c apk does not allow you to install two versions of the same pkg 2017-03-05 18:44:53 and even if it would allow it, this is a dirty hack, not a systematic solution 2017-03-05 18:45:33 actually the only truly systematic solution is abandoning this damn FHS… 2017-03-05 18:47:32 +1 2017-03-05 18:47:52 I'm glad you're bumping against it again and again 2017-03-05 18:48:10 with your voice maybe we can convince the big bosses 2017-03-05 18:48:36 ha, skarnet, could you please reply to that it’s broken and add link to http://bugs.alpinelinux.org/issues/6839 ? 2017-03-05 18:48:45 (your voice is the cheapest torture implement I can think of) 2017-03-05 18:48:51 damn 2017-03-05 18:48:52 rude 2017-03-05 18:49:02 what?! my voice?! 2017-03-05 18:49:03 I should have put a smiley 2017-03-05 18:49:29 jirutka: Essentially, I just throw the FHS out for my databases and dump them in self-contained directories. 2017-03-05 18:49:40 skarnet: oh crap, reply to this http://patchwork.alpinelinux.org/patch/3102/ 2017-03-05 18:49:55 Shiz: context: last time we talked about convincing people to ditch FHS, I was up for waterboarding people and jirutka insisted on talking to them first 2017-03-05 18:50:05 ah :p 2017-03-05 18:50:08 aha XD 2017-03-05 18:50:13 *lol* 2017-03-05 18:50:19 im pro-cleaned-up-FHS 2017-03-05 18:50:27 :p 2017-03-05 18:50:34 cleaned up? 2017-03-05 18:50:37 which involves no such nonsense as /root 2017-03-05 18:50:40 but I didn’t mean to torture them by talking, but to agree on it, like civilized people! XD 2017-03-05 18:50:41 or /boot 2017-03-05 18:50:52 you can't clean that up, it needs to be burnt down to the ground and rebuilt upon 2017-03-05 18:51:06 I somewhat agree 2017-03-05 18:51:16 but I also fear that you're gonna put forward djb's filesystem layout 2017-03-05 18:51:18 which i dislike even more 2017-03-05 18:51:20 :p 2017-03-05 18:51:25 FHS is great for the basic system tools, but applications that have a bunch of files with interdependencies and version requirements spread across the FS are nuts. 2017-03-05 18:51:29 I'm not a /package fan 2017-03-05 18:51:37 Shiz: do you know Homebrew at least a little? 2017-03-05 18:51:41 I like some of the guarantees it provides 2017-03-05 18:51:42 jirutka: i use it daily 2017-03-05 18:51:44 :p 2017-03-05 18:52:11 but I'd be just as happy with a similar layout that provide the same guarantees and maybe more 2017-03-05 18:52:12 Shiz: great, then that’s how I personally imagine that it may work, just with better names than Cellar etc :P 2017-03-05 18:52:23 ' /opt has the sanest structure for applications actually. 2017-03-05 18:52:24 homebrew's systme is interesting, i agree 2017-03-05 18:52:28 Shiz: still use FHS, but just for compatibility, by putting symlinks into it 2017-03-05 18:52:53 no, /opt is defined in FHS, so it’d be better to avoid it 2017-03-05 18:53:01 to not confuse people too much 2017-03-05 18:53:24 but /opt is the exact right place to have a directory-based hierarchy 2017-03-05 18:53:29 maybe something like /pkgs/$pkgname-$pkgver (and symlink /pkgs/$pkgname) 2017-03-05 18:53:32 I mean that currently in FHS, only opt is actually intended to encapsulate. 2017-03-05 18:53:40 not following the FHS doesn't mean we can't have /bin 2017-03-05 18:53:41 yes, that’s true 2017-03-05 18:53:47 jirutka: OK, if bats is in community or main I can use that. Thanks 2017-03-05 18:53:49 so it doesn't mean we can't have /opt, either 2017-03-05 18:54:08 yeah, but not for creating directories for packages inside it 2017-03-05 18:54:34 oh? 2017-03-05 18:54:45 I don't remember anything in FHS forbidding it 2017-03-05 18:54:50 but I may have misread 2017-03-05 18:54:50 I donjt like /opt/$pkgname/ … 2017-03-05 18:55:05 /opt actually usually DOES have a directory for each package. 2017-03-05 18:55:19 TemptorSent: that's what I thought, too 2017-03-05 18:55:24 b/c FHS defines /opt as a directory for stuff that is not managed by package manager 2017-03-05 18:55:32 and ppl are used to use it like that 2017-03-05 18:55:36 ah 2017-03-05 18:55:49 you'd like homebrew's equivalent 2017-03-05 18:55:51 so it's ok for me to use it like that on my machines, but for a distro it's more problematic. I see. 2017-03-05 18:55:53 for its secondary set of symlinks 2017-03-05 18:55:55 /usr/local/opt 2017-03-05 18:55:57 :p 2017-03-05 18:55:59 ugh 2017-03-05 18:56:27 why? isn’t more clear to just made up some actually reasonable name right under / ? 2017-03-05 18:56:40 /opt is intended for packages which don't fit the FHS and are locally installed. 2017-03-05 18:56:42 ~ » ls -lP /usr/local/opt/ffmpeg 2017-03-05 18:56:43 lrwxr-xr-x 1 mark admin 22 21 Feb 02:13 /usr/local/opt/ffmpeg -> ../Cellar/ffmpeg/3.2.4 2017-03-05 18:56:49 jirutka: honestly idc, the argument you gave is good 2017-03-05 18:56:52 instead of burying it into /usr/local/whatever/insanely/long/crap 2017-03-05 18:57:01 jirutka: i was being sarcastic, obviously 2017-03-05 18:57:21 homebrew is forced to usr something under /usr/local because that's the only /usr place OSX lets you install stuff 2017-03-05 18:57:22 I like /opt, but if a distro shouldn't use /opt, then let's use something else 2017-03-05 18:57:24 Shiz: wat a moment, /usr/local is used on macOS for very specific reasons, it doesnjt make any sense for us 2017-03-05 18:57:32 For instance /opt/google/chrome is the target location for chrome. 2017-03-05 18:57:32 jirutka: that's why i was being sarcastic 2017-03-05 18:57:33 /pkgs sounds good, or even /p for shortness 2017-03-05 18:57:45 i prefer /pkg if we're going that way 2017-03-05 18:57:49 /bin is not /bins either 2017-03-05 18:58:04 isn't /pkg used in some BSD? 2017-03-05 18:58:11 yeah, or /pkg, that’s maybe better than /pkgs 2017-03-05 18:58:20 not that we care, but avoiding any possible confusion would be useful 2017-03-05 18:58:41 skarnet: not seeing anything in man hier 2017-03-05 18:58:51 jirutka: what's wrong with http://patchwork.alpinelinux.org/patch/3102/ and why should I reply to it? 2017-03-05 18:58:55 I was not very clear with FHS and symlinks… I’d like to keep all configs in one place, e.g. /etc, and also logs and application data 2017-03-05 18:59:00 I'm not sure pkg is the right abstraction, since a given application may use several packages. 2017-03-05 18:59:20 skarnet: netbsd has /usr/pkg 2017-03-05 18:59:34 "packages maintained by groups other than the NetBSD Project." 2017-03-05 18:59:37 actually I think about /pkg as immutable, it should be possible to mount it read-only 2017-03-05 18:59:43 absolutely 2017-03-05 19:00:02 that's not gonna play nicely with a lot of stuff 2017-03-05 19:00:08 but you should then remount it rw everytime you apk upgrade :) 2017-03-05 19:00:15 yes 2017-03-05 19:00:17 jirutka: That could work for most of my use cases I guess. 2017-03-05 19:00:25 so datafiles would be scattered throughout the FS 2017-03-05 19:00:26 ? 2017-03-05 19:00:28 :p 2017-03-05 19:01:01 no, just in /var/lib/$pkgname as now, or some other directory with the same semantics 2017-03-05 19:01:08 you could have a /rw/$package, iow /var/lib/$package 2017-03-05 19:01:12 and /var/db 2017-03-05 19:01:14 it can be just /var/$pkgname, doesn’t matter 2017-03-05 19:01:17 and whatever else is under /var 2017-03-05 19:01:44 anything under /var is fine, by definition it's rw 2017-03-05 19:02:06 . /var and /etc makes sense for me, (/usr)/bin, (/usr)/lib, (/usr)/share is the problem 2017-03-05 19:02:33 . /bin etc. make sense for symlinks, to simplify PATH 2017-03-05 19:02:46 who are you and what have you done with jirutka 2017-03-05 19:02:53 ? 2017-03-05 19:02:53 you make a ton of sense lately :P 2017-03-05 19:03:02 /lib also makes sense, probably 2017-03-05 19:03:08 for LD_LIBRARY_PATH... 2017-03-05 19:03:15 gaaah 2017-03-05 19:03:19 my client doesn't know /say 2017-03-05 19:03:28 just type two slashes 2017-03-05 19:03:28 we’ve already talked about this on FOSDEM and I said the same as now… 2017-03-05 19:03:28 jirutka: What I was using on some previous machines was something like /app/postgresql/9.6.2/* 2017-03-05 19:03:31 that's how mine escapes it 2017-03-05 19:03:35 nope 2017-03-05 19:04:08 ah, nettalk 2017-03-05 19:04:10 beyond saving 2017-03-05 19:04:28 :p 2017-03-05 19:04:38 I'll take any suggestion for a simple, free, Windows client 2017-03-05 19:04:57 hexchat? 2017-03-05 19:04:58 skarnet: you’re sentence is semantically invalid 2017-03-05 19:05:10 and yours is grammatically invalid :P 2017-03-05 19:05:12 most common relatively simple windows client i see used 2017-03-05 19:05:17 https://hexchat.github.io/ 2017-03-05 19:05:19 skarnet: it’s internally inconsistent 2017-03-05 19:05:27 Hmm, does BitchX work on windoze? 2017-03-05 19:05:29 Shiz: thanks, I'll check 2017-03-05 19:05:34 skarnet: eh, well, that’s quite possible, my English still sucks :/ 2017-03-05 19:05:46 bx has a rather unfortunate name still... 2017-03-05 19:05:54 btw Textual for macOS is great ;) 2017-03-05 19:05:55 TemptorSent: I hated BitchX when I ran it on a Linux client 2017-03-05 19:06:25 *lol* Yeah, the name is unfortunate these days I guess.. I've been using it for 20 years though. 2017-03-05 19:06:33 Shiz: "Highly scriptable with Python and Perl". I said simple! 2017-03-05 19:06:35 i'll still be stuck on weechat in 2025 i think 2017-03-05 19:06:41 skarnet: the UI is simple 2017-03-05 19:06:46 :p 2017-03-05 19:06:52 nettalk's code probably isn't all that jazzy either 2017-03-05 19:07:13 I don't think any IRC client's code is jazzy 2017-03-05 19:07:13 skarnet: Why did you hate BitchX? It feels better than mIRC IMHO. 2017-03-05 19:07:35 bx's name was always unfortunate 2017-03-05 19:07:50 A rusty IRC client would be interesting indeed... 2017-03-05 19:08:27 TemptorSent: couldn't say. Maybe it was the GUIness of it on a Linux client and my general discomfort. 2017-03-05 19:08:37 Shiz: Yeah, it started in the days before anyone cared about political correctness. 2017-03-05 19:08:51 nothing wrong with female dogs 2017-03-05 19:08:56 doesn't make it any less appropriate back then 2017-03-05 19:09:42 skarnet: GUI? It has a two line status bar and that's about it to clutter the screen. 2017-03-05 19:10:11 it runs in a terminal? I really don't remember. 2017-03-05 19:10:21 but I couldn't get used to it anyway. 2017-03-05 19:10:36 skarnet: Yup! I don't think I've ever even tried to use it in X. 2017-03-05 19:10:45 bx, irssi and weechat are all TUI clients yes 2017-03-05 19:11:44 relatedly: every time jirutka (re)tweets i feel like i'm learning a little bit of czech 2017-03-05 19:11:46 very educational 2017-03-05 19:11:56 :) 2017-03-05 19:11:57 anyway I didn't like it, personal preference, and you can't generalize on that anecdotal evidence because I'm a very weird data point when it comes to clients. 2017-03-05 19:11:58 At the time I started using it, one of the big draws was it was much less vulnerable to script-kiddie attack. 2017-03-05 19:12:31 the only thing i know bx for is having auto-away messages on by default 2017-03-05 19:12:38 skarnet: Okay, I was just curious if it had some major pain point that I'd missed. 2017-03-05 19:12:41 this weekend was InstallFest, that’s why so many tweets in Czech 2017-03-05 19:13:12 ACTION is away (Auto-Away after 15 mins) [BX-MsgLog On] 2017-03-05 19:13:14 like that 2017-03-05 19:13:35 Shiz: Auto nick completion is nice, multiple windows for different channels if desired, etc. 2017-03-05 19:13:52 mitchty: https://twitter.com/jakubjirutka/status/838437341400268800 (ghc) 2017-03-05 19:14:21 Shiz: DCC that actually works was a real boon too back in the day -- xdcc and friends were twitchy. 2017-03-05 19:14:24 the only one i understood was about some vps provider offering colorful alpine images 2017-03-05 19:14:25 :p 2017-03-05 19:16:20 My IRC-foo is rusty these days, I had a haitus of several years (>10?!? - where do the years go?). 2017-03-05 19:18:46 ...speaking of VPS images, would anyone be up for building some various test images using my branched mkimage? I don't know about production ready, but it's getting close -- which is good, because my client needs images a week ago :) 2017-03-05 19:19:40 jirutka: yep, but at the same time ghc is as old as python, its got its warts, but it was a fun port, once i got past having to learn 10 things at once 2017-03-05 19:20:39 mitchty: but according to what you said and how insanely big the pkg is, it’s definitely not SIMPLE as Don Stewart said, is it…? 2017-03-05 19:20:58 mitchty: also, read https://twitter.com/mimi1vx/status/838465059923836929 2017-03-05 19:22:55 don't think he knows what he's on about, the debian ghc install is 700Mib with profiled libraries 2017-03-05 19:23:11 300Mib without them 2017-03-05 19:23:14 mitchty: I think that he lies… I’ve tried it on Fedora now: https://termbox.io/term/pod-LwEunaZBBQ 2017-03-05 19:23:23 mitchty: 986 MiB 2017-03-05 19:23:41 i'd ask for proof, sounds like bullshit or ignorance of the actual size, maybe compressed 2017-03-05 19:23:49 mitchty: yeah, he probably doesn’t count all the dependencies… classic error 2017-03-05 19:24:16 Holy crap: A gig for a compiler, tools, and libs? What is this? vizual studio? 2017-03-05 19:24:24 with xz compression you can get a ghc dist install to about 80Mib 2017-03-05 19:24:52 hm, but download size 89 MiB… this looks like a very good compress ratio… 2017-03-05 19:24:58 TemptorSent: its 3 runtimes that the size gets up with, mostly profiled/debugging libraries 2017-03-05 19:25:13 i hope at least in the alpine those prof/dbg libs are separate 2017-03-05 19:25:15 :p 2017-03-05 19:25:17 alpine package* 2017-03-05 19:25:37 Shiz: yes, ghc is now split with ghc/ghc-dev the latter being the profiled libraries 2017-03-05 19:26:09 https://pkgs.alpinelinux.org/package/edge/testing/armhf/ghc-dev 2017-03-05 19:27:28 jirutka: and as for simple, its not TOO bad, but coming into the ghc internals i won't say its easy, depends on how we're defining simple 2017-03-05 19:27:49 https://www.infoq.com/presentations/Simple-Made-Easy 2017-03-05 19:28:34 the make system is byzantine though 2017-03-05 19:29:04 and being rewritten https://github.com/snowleopard/hadrian 2017-03-05 19:30:18 as for his original comment, i have done the same with some simulated annealing stuff 2017-03-05 19:31:00 great 2017-03-05 19:31:11 i had a toy version that used cpus to run 500 000 steps on my i7 laptop cpu in about 8 hours, added the accelerate framework and Acc in two spots, took 20 seconds using cuda 2017-03-05 19:31:12 a build system in the same language of the compiler youre compiling... 2017-03-05 19:31:17 people don't really think this through do they 2017-03-05 19:31:39 its no different than Make then not? 2017-03-05 19:32:00 skarnet: ad Patchwork, I don’t know how to reply to the patch without having the email in mailbox and so subscribing to that ML; the ranger is broken, it segfaults and there’s already issue for it http://bugs.alpinelinux.org/issues/6839 2017-03-05 19:32:05 at some point every compiler had to be bootstrapped 2017-03-05 19:32:15 make doesn't need make to compile 2017-03-05 19:33:23 omg, I really love when lang must be bootstrapped AND also use build system written in the same lang so you have double chicken-egg problem… the same problem with Rust :( 2017-03-05 19:34:01 can probably blame the rewrite on rust actually :) 2017-03-05 19:36:20 mitchty: there’s the answer: https://twitter.com/mimi1vx/status/838472613668212736 2017-03-05 19:36:37 you can build GNU make by doing ./configure && ./build.sh 2017-03-05 19:36:38 :) 2017-03-05 19:36:38 mitchty: he don’t know what he’s speaking about… 2017-03-05 19:37:02 jirutka: thought so, unless you get strip crazy its almost always ~1GiB 2017-03-05 19:37:22 this is much bigger even than OpenJDK! 2017-03-05 19:37:31 i can get it down to 200MiB but it doesn't work then 2017-03-05 19:37:34 I cannot find anything more bloated than JDK… 2017-03-05 19:37:47 well, of course, except ghc :P 2017-03-05 19:38:24 they inline a lot, and i don't think theres eve been any attention to code size 2017-03-05 19:38:56 can it be compiled with -Os ? 2017-03-05 19:39:31 the rts yes 2017-03-05 19:39:39 its default -O2 2017-03-05 19:39:45 for the c side of the rts 2017-03-05 19:40:02 but there are three of them, default, threaded, and profiled/profiled+threaded 2017-03-05 19:40:05 why do you build it with -O3? 2017-03-05 19:40:34 noticeably reduces compile times of libraries 2017-03-05 19:40:46 rustc is pretty big too 2017-03-05 19:41:26 I'm afraid the art of easy bootstrapping has been dead for a decade or more 2017-03-05 19:41:29 and not a huge impact on the resulting size of things tbh, it just tries harder to inline things 2017-03-05 19:42:22 that guy wrote me in #vpsfree that -O3 and perf-llvm is the most stupid idea that he seen… but I really doubt that he know what’s he talking about… 2017-03-05 19:42:37 I asked him to join our discussion here, but he didn’t… 2017-03-05 19:42:54 ironically the llvm backend produces the smallest code size 2017-03-05 19:43:05 i compared for x86_64 and the native code generator 2017-03-05 19:43:18 llvm backend and -O3/-O2 was the smallest 2017-03-05 19:43:47 so... i guess have him tell me why its stupid with numbers 2017-03-05 19:44:43 its easy enough to change 2017-03-05 19:48:03 also sounds like a likely troll tbh, i'd just ignore them 2017-03-05 19:48:24 i prefer results to hot air 2017-03-05 19:49:09 yes, we’ve just “discussed” in #vpsfree… I asked him once again to talk with you about it if he have some constructive critique 2017-03-05 19:50:07 -O3/-O2 was smaller than -Os? -Os should optimize for size, so I’d expect to produce smaller binaries… 2017-03-05 19:50:36 jirutka: ghc only uses -ON, -Os would have to be in SRC_CC_OPTS as it would only impact the c rts 2017-03-05 19:50:41 also that high compress ratio for *binary* files suggests that there’s extreme duplication in these binaries 2017-03-05 19:50:46 aha 2017-03-05 19:50:53 there likely is 2017-03-05 19:51:18 once i get more of a handle on the core -> llvm stuff i wanted to have some fun pruning the low hanging fruit 2017-03-05 19:51:53 i think its inlining more than it needs to generally 2017-03-05 19:52:42 i could kick off a build of ghc with -Os and compare 2017-03-05 19:55:28 oh another thing that might make the alpine ghc bigger, splitsections is being used, which gives less decrease in binary size than splitobjects, but the latter is a hack, and it only works on x86 2017-03-05 20:02:19 I think that it’d be also better to separate haskell-* into subpkgs instead of just defining provides 2017-03-05 20:02:42 thats a lot harder, the compiler uses those libraries 2017-03-05 20:03:57 ghc package can depends on them… 2017-03-05 20:04:32 sure, but to what end 2017-03-05 20:04:47 also it means that mimi1vx lied b/c he told me in #vpsfree that it’s a stdlib, so it doesn’t make to count it into size… 2017-03-05 20:05:17 if compiled needs them for itself, then these are not *just* stdlib… 2017-03-05 20:05:26 but regular dependencies of the compiler 2017-03-05 20:05:43 basically 2017-03-05 20:05:55 well you can argue they're both 2017-03-05 20:06:20 and there is talk about splitting them out into subpkgs on their own right, but its a lot of thankless work 2017-03-05 20:07:58 i'm also only one person working on this in my part time, mostly because I wanted a static pandoc, and also to use idris on alpine linux 2017-03-05 20:09:18 ok, I understand that 2017-03-05 20:10:49 also, i still see people doing stuff like this https://github.com/fpco/pid1/blob/c11322bbdd3e894e8d35c8db7ee1e348a2978459/static-base/Dockerfile#L16 2017-03-05 20:11:01 which will just bloat the size of every executable 2017-03-05 20:11:21 even though the fix is to just specify ld-options: -static in the cabal file for the executable 2017-03-05 20:11:38 you see people applying glibc fixes for ubuntu everywhere 2017-03-05 20:13:52 i was going to open a ton of issues and fix pr's once cabal gets merged into testing 2017-03-05 20:15:50 well, you know, docker users usually don’t understand what are they doing… 2017-03-05 20:16:45 so i'll do a quick look at the size of ghc with perf-llvm -O3, c -Os, perf-llvm no -O, -Os, perf-llvm w/o any -O settings, and stock without any, and stock with c -Os 2017-03-05 20:16:50 what are these .hi files? 2017-03-05 20:16:58 haskell interfaces 2017-03-05 20:17:08 sources or compiled code? 2017-03-05 20:17:15 its compiled 2017-03-05 20:17:36 they're used by the compiler in one of its passes 2017-03-05 20:19:08 https://ghc.haskell.org/trac/ghc/wiki/Commentary/Compiler/IfaceFiles 2017-03-05 20:21:47 are these files ghc-specific or usable even for some other haskell compilers? 2017-03-05 20:27:32 ghc specific 2017-03-05 20:27:50 realistically ghc is the only haskell compiler of note, there is hugs but that hasn't seen an update for 5 years 2017-03-05 20:27:57 and jhc stopped development 2017-03-05 20:29:00 uhc? 2017-03-05 20:29:18 Fedora uses prefix ghc- for haskell pkgs; if these binaries are ghc-specific, then it’s maybe better than haskell-prefix 2017-03-05 20:29:52 i was borrowing from arch there 2017-03-05 20:29:55 and all of them have also suffix -devel 2017-03-05 20:31:38 so i honestly don't care what we call them, theres a ton of examples to pick from or invent a new name for them 2017-03-05 20:33:28 uhc looks to also be about the only other one that is actively maintained, not sure if it supports anything past haskell 98 though 2017-03-05 20:40:34 I have server at academic network with 1 Gbps connectivity and downloading from that f*cking amazon s3 at speed 500 kiB/s… grr 2017-03-05 20:57:05 ok so in ~3 hours i'll have numbers on the disk size of all the permutations of ghc with -O... with the c rts, and the haskell side, and with/without llvm 2017-03-05 20:57:36 which is more productive and scientific than arguing which is best imo 2017-03-05 20:58:08 based on that we can change it or not 2017-03-05 21:37:41 apk installed size output is in bytes correct 2017-03-05 21:39:31 -Os doesn't change the profiled library sizes at all, ghc by 12.28KiB 2017-03-05 21:40:05 i suppose i could try compiling against llvm with -Oz 2017-03-05 21:40:50 but at this point that points to the c side not contributing a hell of a lot to overall code size 2017-03-05 22:14:16 and size with -O3 and -O2 is no different 2017-03-05 22:14:26 will compare without perf-llvm 2017-03-05 22:14:45 note this is all x86_64 2017-03-05 22:16:46 so why you’ve used -O3 instead of default -O2? 2017-03-05 22:17:03 i haven't compared how fast these compilers build things 2017-03-05 22:17:20 last i tried though this cut off a good 30 minutes on build times for stack and cabal on armhf 2017-03-05 22:18:06 but given it takes 12 hours to compile those entirely from scratch i'm not inclined to test each of these permutations to validate someones point, this is ONLY on disk install size 2017-03-05 22:18:43 i'm willing to change whatever, but i'd like actual numbers to base a decision on 2017-03-05 22:22:02 i guess i can compare the build times of cabal and stack with the resulting compilers too 2017-03-05 22:22:40 I’m just asking why you’ve changed the defaults at the first place, what was the initial reason 2017-03-05 22:23:17 to make the armhf builds faster 2017-03-05 22:23:24 in general 2017-03-05 22:23:26 aha 2017-03-05 22:23:59 i'll just remove it, tbh i don't really give a shit, and i'm spending more time talking about it than its worth, i'll just submit a pr to remove it 2017-03-05 22:27:16 https://github.com/alpinelinux/aports/pull/980 2017-03-05 22:32:19 maybe you misunderstood me; I don’t have any problem with it, I’m just asking to understand the decisions that you’ve made 2017-03-05 22:34:36 nah its fine i just don't have the numbers to back it up right now 2017-03-05 22:35:37 i'll go setup a jenkins job to go compare it all on x86_64 and armhf 2017-03-05 22:35:50 but it'll take like a week or more to get the actual reality 2017-03-05 22:36:10 and then it actually has something tangible rather than my horrible memory of what i did 2017-03-05 22:36:12 pickfire: i am generally supportive of a manpage! 2017-03-05 22:51:09 so anyway, is there a chance the cabal apk could get merged? or if i need to change anything in it let me know, but I hacked the cabal bootstrap.sh script to download all its dependencies and use the cached dependencies https://github.com/alpinelinux/aports/pull/915 2017-03-05 22:51:31 if we want it to work any different let me know but not having cabal limits ghc's usefulness 2017-03-05 22:53:50 or if it should get broken apart into each pkg getting its own discreet package I can do that too, but it'll take a while to work, and there is... 19 pkgs that cabal depends on for builds (not runtime) 2017-03-05 23:01:51 all depends on how we want to do ghc pkgs or not to save on any recompilation time for any dependent packages 2017-03-05 23:44:11 and jirutka sorry if i'm coming off as angry or mad, i'm not really, and for the provides, do you want me to split the packages into subpackages named ghc-$library-dev? 2017-03-05 23:46:59 looking at it now 2017-03-05 23:48:44 what’s this error: "sh: /usr/lib/gcc/x86_64-alpine-linux-musl/6.3.0/../../../../x86_64-alpine-linux-musl/bin/ld-: unknown operand" 2017-03-05 23:53:07 the ld detection failing 2017-03-05 23:53:34 and the reason for https://github.com/alpinelinux/aports/pull/915/commits/52f4806caadd3a91eed56a1e0661a0e364424f2f#diff-15f7e2fe859b5cc21fb2498121943aeaR7 2017-03-05 23:53:47 aka force it to use ld.gold for linking 2017-03-05 23:57:12 i could update that patch to rip out the entire ld detection logic in the cabal bootstrap.sh script 2017-03-05 23:58:40 that’d be probably better if it doesn’t affect too many lines; and most importantly write a comment why you did it, so future you or other maintainers would know 2017-03-05 23:59:56 uff, it takes forever to build 2017-03-06 00:02:02 yep, no parallelism in it 2017-03-06 00:02:33 let me see how much it impacts 2017-03-06 00:07:24 no, no, no, NO! https://github.com/alpinelinux/aports/pull/972 2017-03-06 00:07:57 he must really hate his users that he moved all sources to his own repo with git-web and *disabled* downloading tarballs! 2017-03-06 00:16:15 btw if anyone interested in this, read https://www.earth.li/~noodles/blog/2017/03/github-tos-change.html 2017-03-06 00:58:39 jirutka: its not too bad, i pushed up an autosquash commit with the changes 2017-03-06 00:59:08 ah whoops, didn't see it was merged already 2017-03-06 00:59:59 this is the changes https://github.com/mitchty/aports/commit/bfe6061074d15f8bb0947ec2bba8077cd28cb8df 2017-03-06 01:04:26 jirutka: thanks for merging cabal! now i can open a ton of pr's to tell people to stop using my old port 2017-03-06 02:08:48 jirutka: Huh? ranger is unmaintained? 2017-03-06 02:12:57 kaniini: Yeah, I support manpages as well! :) 2017-03-06 02:14:17 help https://github.com/alpinelinux/aports/pull/952 2017-03-06 02:18:06 pickfire: i mean for apk :) 2017-03-06 02:19:54 kaniini: Yes, I support man. 2017-03-06 02:19:59 -man* 2017-03-06 02:20:25 dlintw: I don't find that pull request ince. 2017-03-06 02:20:26 nice* 2017-03-06 02:23:33 kaniini, could you help me check why CI failed in https://github.com/alpinelinux/aports/pull/952 2017-03-06 02:24:57 kaniini, what's things should I do on https://github.com/alpinelinux/aports/pull/940 and https://github.com/alpinelinux/aports/pull/939, I can not see them on testing after several days. 2017-03-06 02:25:38 pickfire, what's you mean what's pull request? 2017-03-06 03:56:48 dlintw: I mean https://github.com/alpinelinux/aports/pull/952 2017-03-06 05:15:04 Hi, I am currently packaging a non-free application for alpine (wps-office) which provides multiple arch, how do I separate the x86_64 arch from x86? 2017-03-06 07:46:07 Bloody hell -- I've been fighting this IFS problem all day trying to make the plugin loader prettier. 2017-03-06 07:52:38 IFS? If you're writing shell, don't expect it to be pretty. "Working" is more or less the best you can do. :P 2017-03-06 08:52:55 IFS hell solved! *YAY* Talk about wasting a day though! 2017-03-06 09:06:21 grats 2017-03-06 09:18:02 skarnet: I wasn't worried about the pretty with the IFS, I was worried about the working part. 2017-03-06 09:19:19 skarnet: What I really should do is convert EVERYTHING to \n separated everywhere and convert on output only if needed. 2017-03-06 09:19:49 yes. Even better if you can make your words null-separated, but that's difficult to do in shell. 2017-03-06 09:19:53 skarnet: Anyway, I finally got it working. 2017-03-06 09:20:24 skarnet: Yeah, I use xargs -0 and printf %s\0 where appropriate. 2017-03-06 09:20:29 nice. 2017-03-06 09:21:28 skarnet: But I haven't even started trying to go through and change ALL the logic in the code to use \n separated values, especially since it will require further work on the means of setting/getting them. 2017-03-06 09:21:57 all in good time. 2017-03-06 09:22:00 Give me a sec and I'll finish cleaning up the work and push it. 2017-03-06 09:53:06 Major revision of mkimage pushed to https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts/scripts/mkimage 2017-03-06 09:55:37 (and breakage it seems, grr. More testing, less pushing :) ) 2017-03-06 09:58:01 I love it when totally unrelate parts of code break at random :/ 2017-03-06 09:59:18 Well, that proved to be simple... a stray semicolon in place of &&. 2017-03-06 10:01:01 Unbroken version pushed :) 2017-03-06 10:02:27 Hmm... need to talk to fabled still about why apk --cache-dir isn't actually doing what it's supposed to. 2017-03-06 12:22:02 mitchty: IMO it’d be better to wait until we move it into community repo 2017-03-06 12:23:10 pickfire: yes, b/c it’s broken, http://bugs.alpinelinux.org/issues/6839 2017-03-06 12:24:16 skarnet: TemptorSent: shell can be pretty… 2017-03-06 12:24:57 TemptorSent: but If you’re dealing with spaces in items, then shell is not a good tool… 2017-03-06 12:25:00 murder can be pretty, too 2017-03-06 12:25:27 skarnet: ofc it can ;) 2017-03-06 12:26:22 TemptorSent: it’d be probably better to rewrite mkimage (at least part of) to Lua, it seems to be already too much complex for shell 2017-03-06 13:06:35 <_errm> I just opened a PR to subpackage npm on the nodejs apk https://github.com/alpinelinux/aports/pull/981anyone have any thoughts about this? 2017-03-06 13:07:20 <_errm> I think its usefull, since npm is not always needed & there is an alternative package manager now - yarn 2017-03-06 13:33:25 _errm: I agree with that, but please change one abuild per commit 2017-03-06 13:37:46 <_errm> ok … 2017-03-06 13:39:03 <_errm> its ok to keep all the changes in the same PR on github? 2017-03-06 13:48:48 of course 2017-03-06 13:48:51 it is 2017-03-06 13:49:15 <^7heo> jirutka: btw, we might have to go away from github. 2017-03-06 13:53:58 has there been clarifications on the new TOS ? 2017-03-06 13:55:18 <^7heo> not sure. 2017-03-06 13:55:26 <^7heo> but it still stands that "we might" 2017-03-06 13:55:34 <^7heo> and therefore it "might" be a good idea to have an alternative. 2017-03-06 13:55:46 ^7heo: no, read https://www.earth.li/~noodles/blog/2017/03/github-tos-change.html 2017-03-06 13:56:08 yes, it’s a good thing to have a *working* alternative… 2017-03-06 13:56:40 <^7heo> right. 2017-03-06 13:56:47 <^7heo> thanks for the link, alos 2017-03-06 13:56:50 <^7heo> also 2017-03-06 13:57:25 <_errm> gitlab.com is nice … 2017-03-06 13:57:36 <^7heo> nope. 2017-03-06 13:57:41 <^7heo> not in my experience. 2017-03-06 13:57:51 <^7heo> But maybe I should give it another shot. 2017-03-06 13:57:57 oh hey, it's that conversation again 2017-03-06 13:58:13 how are you my old friend? long time no see, it's been at least two... weeks 2017-03-06 13:58:30 <^7heo> :P 2017-03-06 14:02:57 <_errm> well I say nice, I mean somewhat useable… 2017-03-06 14:05:16 <^7heo> yeah 2017-03-06 14:05:19 <^7heo> but then gogs also is. 2017-03-06 14:05:29 <^7heo> and gitea (from mosez and cie) 2017-03-06 14:05:48 <^7heo> and some other clones I guess, too. 2017-03-06 14:06:52 NoGo… 2017-03-06 14:07:20 <^7heo> jirutka: stop being a NoGo :P 2017-03-06 14:07:28 but well, everything’s better than Patchwork… but GitHub works very well for us now 2017-03-06 14:07:42 <^7heo> yeah 2017-03-06 14:08:04 <^7heo> jirutka: so, it IS to see if we have something in case github would "break" for us. 2017-03-06 14:08:08 <^7heo> jirutka: thanks for confirming. 2017-03-06 14:08:48 that also reminds me that I should finally write the proposal for thesis about minimalistic GitHub/GitLab/Gogs/whatever alternative… something very simple with pull requests support and hooks 2017-03-06 14:08:59 <^7heo> yeah 2017-03-06 14:09:18 <^7heo> but the challgenge is: 2017-03-06 14:09:27 <^7heo> the interface for those PRs and hooks 2017-03-06 14:09:35 <^7heo> if you do hooks, people will want webhooks 2017-03-06 14:09:42 <^7heo> if you do PRs, people will want reviews 2017-03-06 14:09:50 <^7heo> and then you end up like github. 2017-03-06 14:09:57 with hooks I mean both 2017-03-06 14:10:23 and with support for PR I mean primarily for reviews, that’s the whole point 2017-03-06 14:10:34 It still can be very simple 2017-03-06 14:10:48 I’d like to avoid traditional user accounts… 2017-03-06 14:10:50 also 2017-03-06 14:10:57 so everything you need is your SSH pubkey 2017-03-06 14:11:29 and probably some token that you get by SSH so you can login into web interface to add comments, if you want to use web interface 2017-03-06 14:11:52 but it should also provide usable remote CLI interface 2017-03-06 14:12:28 <^7heo> jirutka: well, user accounts could be based on git. 2017-03-06 14:12:51 yes, but the problem is that you cannot use SSH pubkey to login into web page :( 2017-03-06 14:13:00 <^7heo> jirutka: also, it technically is possible to use git branches or tags or whatnot to "push reviews" 2017-03-06 14:13:01 (securely) 2017-03-06 14:13:15 no, git notes 2017-03-06 14:13:19 <^7heo> or that. 2017-03-06 14:13:21 <^7heo> right. 2017-03-06 14:13:21 or how is that called 2017-03-06 14:13:25 <^7heo> yeah gotchat. 2017-03-06 14:13:27 <^7heo> s/chat/cha/ 2017-03-06 14:13:41 yes, that’s what I’m thinking about, but there are still some unresolved questions 2017-03-06 14:13:42 <^7heo> why would you need to auth to the web? 2017-03-06 14:14:04 for users who’d like to add comments via web interface 2017-03-06 14:14:39 to be honest, I don’t know how to make a simple CLI interface that would allow you to add comments to the code (to specific lines), this is much easier via web interface 2017-03-06 14:15:01 anyway, I need to work now 2017-03-06 14:15:19 <^7heo> jirutka: you can definitely start the git note (or whatever) with a line identifier 2017-03-06 14:15:25 but you can kick into me once a day to remind me to write the proposal for this topic :) 2017-03-06 14:15:26 <^7heo> jirutka: and that solves the problem 2017-03-06 14:15:39 <^7heo> I'll try to poke gently 2017-03-06 14:15:41 <^7heo> not kick :D 2017-03-06 14:29:06 what's the biggest change in current ToS? sorry, didn't have time to look at this yet 2017-03-06 14:29:22 scadu: read https://www.earth.li/~noodles/blog/2017/03/github-tos-change.html 2017-03-06 14:30:05 jirutka: oh thanks. I missed this one. 2017-03-06 14:30:05 unfortunately there are some people who create a panic b/c of their interpretation of ToS… 2017-03-06 14:30:23 and now we have to deal with some sources on personal git with no tarballs… 2017-03-06 14:31:29 no that git-web doesn’t provide support for this, but the developer has disabled it… 2017-03-06 14:34:26 according to twitter they (GitHub) got a lot of questions about ToS and are currently refining it to be more clear, so we’ll see 2017-03-06 15:29:50 one reason to always ask a lawyer to read legalese 2017-03-06 15:35:30 <^7heo> if you got the cash, sure. :P 2017-03-06 15:35:59 well you could also ask the laywers working for the fsf* 2017-03-06 15:37:24 <^7heo> and risk to have a tainted answer. 2017-03-06 15:37:30 <^7heo> no matter how... 2017-03-06 15:37:37 <^7heo> ... it's best to pay. 2017-03-06 15:51:00 the fsf* would have complained if the panicking people were right about their misunderstandings. 2017-03-06 15:54:34 <^7heo> don't assume that the complaining people will always complain in case of fuckup. 2017-03-06 15:54:59 <^7heo> or rather, don't assume you can trust an 1:1 equivalent of trigger-warning people and fuckups. 2017-03-06 15:55:12 <^7heo> because the day they're tired, crap will hit the fan ;)( 2017-03-06 16:22:36 yes, and also don’t trust non-lawyer’s interpretation of a legal documents; there are many non-senses in laws and especially in USA companies must protect themselves, b/c anyone can sue them for total bullshits; so they may not be any bad intention behind it, just various interpretations 2017-03-06 16:23:59 huh, someone did the change in abuild I’m talking about for a year and still didn’t actually do it; I’m ashamed now :( I wanted to start with refactoring and then implement this change and that’s not an easy work so… I postponed it 2017-03-06 16:25:03 I went just without refactoring :) 2017-03-06 16:26:48 yeah, that’s probably more reasonable, b/c it can be done in shorter time… or let’s say, be actually done :) 2017-03-06 16:29:24 but let's see how people react 2017-03-06 16:30:59 this is definitely good change, I hate writing `|| return 1` on almost every line; abuild should be run with -e since the beginning 2017-03-06 16:31:25 I just worry if it really works as expected, b/c… read my email 2017-03-06 16:43:43 i am tempted to just push that to git and fix things as they break ... 2017-03-06 16:43:50 the set -e abuild thingy 2017-03-06 16:45:50 agree, but we should review the patches first… 2017-03-06 16:46:15 and most importantly verify that it really aborts on non-zero status inside all functions 2017-03-06 16:47:01 the ${foo:0:1} is bashism isnt it? 2017-03-06 16:47:13 seemed to work anyway :/ 2017-03-06 16:47:16 well, yes, but ash supports it 2017-03-06 16:47:25 and we already use it in APKBUILDs 2017-03-06 16:47:44 simialr to `${foo/o/x}` 2017-03-06 16:47:50 ans we use ${foo/../..} 2017-03-06 16:47:51 yes 2017-03-06 16:48:06 my thinking is that we should try avoid when possible 2017-03-06 16:48:12 hm, not sure if dash supports this syntax 2017-03-06 16:48:13 and i think we can avoid in this case 2017-03-06 16:48:17 i dont think it deso 2017-03-06 16:48:58 $ dash -c 'foo=foo; echo ${foo:0:1}' 2017-03-06 16:48:58 dash: 1: Bad substitution 2017-03-06 16:49:49 I’ve tried it too :) 2017-03-06 16:49:55 just about to write the same result 2017-03-06 16:58:19 leitao: seems like static build is broken on ppc64le? 2017-03-06 16:58:43 ncopa, with musl, right? 2017-03-06 16:58:46 yes 2017-03-06 16:58:52 gromero investigated a little bit about it. 2017-03-06 16:58:59 not sure how far he went 2017-03-06 16:59:20 $ gcc -static hello.c && ./a.out 2017-03-06 16:59:20 Segmentation fault 2017-03-06 17:05:02 right, it fails on every static compilation 2017-03-06 17:05:22 let's see how further gromero went, and I can dig further 2017-03-06 17:48:13 leitao: ncopa just far to get gdb compiled on alpine, but nothing after discovering that if is passed the problem is gone -fno-pie... I have to investigate further so... 2017-03-06 17:49:01 *if "-fno-pie -static" is passed 2017-03-06 17:49:40 gromero, that is why I do not reproduce this problem on Debian. 2017-03-06 17:50:03 I am able to run static compiled binaries on Debian with musl. But I checked and musl-gcc is using no-pie 2017-03-06 17:51:09 leitao: makes sense. -fpie is enabled by default on alpine to improve security, but I'm not sure if it's musl issue or something also related to the loader, for instance... 2017-03-06 18:00:54 *yawn* Good morning all. Has anyone seen fabled around lately? 2017-03-06 18:47:46 WTF? Why does qemu-system-x86_64 depend on mesa/wayland? 2017-03-06 18:48:43 TemptorSent: OpenGL guest acceleration 2017-03-06 18:52:22 And why is SDL support disabled? 2017-03-06 18:52:41 idk policy, only mechanism 2017-03-06 18:53:15 *lol* Yeah... methinks a lightweight qemu may be needed. 2017-03-06 18:54:04 qemu does not lend itself to having multiple versions installed easily, nor does it lend itself to having split deps 2017-03-06 18:54:51 mesa is required increasingly by most X11 apps because ~compositing~ so really it is not that big of a deal to have that dep in /my/ opinion, but I am not a developer here so that is not my judgement call to make. 2017-03-06 18:55:43 awilfox: Because my database server needs x11 compositing, right? 2017-03-06 18:56:16 why does your database server need x11 at all 2017-03-06 18:56:26 qemu is an X11 application (unless you use serial console and no vga, but I'm not sure that is even in tree any more) 2017-03-06 18:57:04 and qemu is designed to be used to run virtual machines, some of which run windows, which needs opengl acceleration to even boot these days (thinking 7/8/10) 2017-03-06 18:57:19 so there is actually a very good reason to have mesa as a qemu dep 2017-03-06 18:57:56 if you are not happy with it, perhaps you could use a different virtualisation environment, such as xen pv (which does not even use qemu) or bochs (which does not support opengl) 2017-03-06 18:58:20 not sure if alpine has a bochs package, but I know it runs a damn good xen pv host because it's doing that for me right now :) 2017-03-06 19:04:52 awilfox: Yeah, I have xen working, but I happen to need qemu/kvm for several applications. 2017-03-06 19:05:29 awilfox: A qemu-light package might be the easiest solution. 2017-03-06 19:24:52 kaniini: Exactly -- when was the last time I actually started X on any of my servers? Granted, at some point I might end up with server-hosted applications and such, but my db server, web server, and storage server sure as hell don't need X! 2017-03-06 19:26:58 Query: Is there any reason NOT to generate isolinux.cfg, syslinux.cfg, and extlinux.cfg for all images which use any flavor of syslinux on x86(_64)? 2017-03-06 20:12:52 TemptorSent: why does your db server need qemu then 2017-03-06 20:13:53 kaniini: for the VMs. 2017-03-06 20:14:09 then it's a VM server, that's a different story altogether :) 2017-03-06 20:15:16 kaniini: Actually, more of databases running on a VM layer -- it makes migration and such MUCH easier. 2017-03-06 20:15:58 but sure, a vnc-only qemu may be useful 2017-03-06 20:17:01 kaniini: Why do I need VNC for TEXT? 2017-03-06 20:17:15 kaniini: ssh-console would be more like it :) 2017-03-06 20:18:27 well after looking at that compiler in dyalog that runs on a gpu, be handy to have gpus generally available. 2017-03-06 20:18:50 kaniini: the curses interface actualy suits nicely - being able to use the full terminal size would be a nice touch. 2017-03-06 20:19:36 TemptorSent: 99% of the world does not use serial console 2017-03-06 20:19:44 kaniini: and if we need some sort of graphics support, install a full-fledged package. 2017-03-06 20:20:03 kaniini: Actually, they probably do, they just don't know it :) 2017-03-06 20:20:05 ACTION is generally in favor of a lightweight package still being useful to 99% of interested users 2017-03-06 20:20:21 ACTION does not maintain qemu though, so has no opinion 2017-03-06 20:21:28 kaniini: Can you think of any vm server that really needs all the weight of supporting graphics unless it's specifically for a desktop or scientific viz? 2017-03-06 20:22:54 kaniini: Most of the VM appliances I see these days are more likely than not to have a web config tool or be totally lights-out and just take config files to get them booted until you can ssh in. 2017-03-06 20:24:20 kaniini: On a desktop, hopefully you'll be able to use PCI passthrough and have the guest use the graphics card directly if you're going to sit in front of it. 2017-03-06 20:35:22 TemptorSent: 99% of people using kvm even just boot normal boot media and install using the normal GUI installer for their distro 2017-03-06 20:48:49 ncopa, fabled : Hi, I sent an email earlier to the list about making Alpine s390x packages available on Alpine's official repo. How do you think about it ? 2017-03-06 20:49:35 Hi fabled 2017-03-06 20:52:35 Good afternoon fabled! 2017-03-06 20:54:03 tmh1999, TemptorSent: hi, sorry i've been away few days, and came to just check few things and going again in a minute. hope to catch up on things tomorrow/the day after. 2017-03-06 20:56:16 fabled: No problem, when you've got a chance I ran into a couple issues I could use a hand on. 2017-03-06 20:57:27 ...and I'm currently trying to figure out why my iso is bailing out to the recovery shell in the initrd.. 2017-03-06 20:58:39 With a complaint of no such file or directory and invalid argument, I'm guessing a driver loading? Trying under qemu 2017-03-06 21:00:35 dying on: umount: can't unmount /sysroot/proc: Invalid argument 2017-03-06 21:02:36 Hmmm... is DEVICETREEDIR really supposed to be in there I wonder? 2017-03-06 21:10:07 Probably not -- that's for u-boot it looks like 2017-03-06 21:13:42 fabled : sure we all have busy days 2017-03-06 21:15:32 Bloody hell this is getting annoying. Has anyone run into the initrd bailing when trying to start using isolinux under qemu for no apparent reason? 2017-03-06 21:30:55 Question: Where was 'modloop' being injected into the kernel command line again? I'm not finding it anywhere obvious. 2017-03-06 21:44:26 Hmm, time to start comparing the FS against a known release I guess. 2017-03-06 22:11:32 *facepalm* It *might* be a key problem... Where's my verbose option? 2017-03-06 22:24:09 Okay, got it to at least tell me the complaint is an UNTRUSTED signature for APKINDEX.tar.gz ... what's missing here? 2017-03-06 22:33:38 TemptorSent: signing key must be in /etc/apk/keys/ 2017-03-06 22:34:03 (public key, obviously) 2017-03-06 22:36:48 tdtrask: Noted - but the other odd thing I'm seeing is an interactave request to overwrite APKINDEX.tar.gz 2017-03-06 22:36:59 tdtrask: And it's not part of my code :) 2017-03-06 22:37:31 tdtrask: I just went on a cleaning spree removing stale/orphan keys... let's see if that helped. 2017-03-06 22:41:35 TemptorSent: never seen that before, but never played in the code your replacing before 2017-03-06 22:41:38 Hmm, still no joy. 2017-03-06 22:42:12 tdtrask: Nothing I actually touched should have changed the signing that I can think of... 2017-03-06 23:07:12 jirutka: when you get a chance i'm curious to have a look at your cargo integration stuff and looking into how I can get cabal doing similar 2017-03-06 23:12:26 Bloody hell.. don't you hate it when variables get redefined because someone left out a check and just overwrote with the default? Doesn't show up until things run in a different order. 2017-03-06 23:39:16 okay, where is this 'mv -i /path/to/output/APKINDEX.tar.gz' coming from? 2017-03-06 23:46:31 skarnet: where is fcntl.h located on various systems? POSIX specifies fcntl.h (top-level), musl and macOS respects this, gnu has sys/fcntl.h, not sure about OpenBSD and NetBSD 2017-03-06 23:46:54 always fcntl.h top-level 2017-03-06 23:47:01 there's also one in gnu 2017-03-06 23:47:25 the one in sys is probably included by the top-level one 2017-03-06 23:47:42 aha, yes, you’re right 2017-03-06 23:47:42 probably a legacy from before everything got standardized 2017-03-06 23:47:45 Weird.. keys straightend out, but still invalid argument unmounting /proc. 2017-03-06 23:47:59 and wait.h? 2017-03-06 23:48:07 sys/wait.h 2017-03-06 23:48:42 I don't think there's any logic to "is this header in top-level or in sys/" 2017-03-06 23:49:10 you have to look at POSIX, typically http://pubs.opengroup.org/onlinepubs/9699919799/idx/is.html 2017-03-06 23:49:25 which tells you what's supposed to be in sys/ 2017-03-06 23:52:56 thanks! 2017-03-06 23:54:13 np :) 2017-03-06 23:55:31 Bugger, I seem to have found myself another project. Is anyone else running into problems with proc not wanting to unmount in an initfs? 2017-03-07 00:06:56 you can hardly ever umount /proc 2017-03-07 00:07:33 what else are you mounting, and what processes are you running at the time of the unmounting you want? 2017-03-07 00:14:39 skarnet: Trying to make the iso boot! 2017-03-07 00:15:59 I'm not asking you for a functional spec, but for an operational one :) 2017-03-07 00:16:11 (iow, the low-level details) 2017-03-07 00:16:16 Testing unde qemu and getting to the point of unpacking the inital root fs, then puking trying to unmount proc before getting ready to pivot root. 2017-03-07 00:17:08 yes. That's why I'm asking what exactly is running when you're trying to unmount. 2017-03-07 00:17:12 It appears somethign in the init script isn't getting along with the tools. 2017-03-07 00:18:32 take a look at /usr/share/mkinitfs/initramfs-init line 610 2017-03-07 00:18:47 online link? 2017-03-07 00:18:51 That is where it's bombing out. 2017-03-07 00:19:33 um, any alpine box or the git repo for mkinitfs I guess. I don't have a browser on this box. 2017-03-07 00:20:09 you want me to help you, you gotta help me help you. Sorry, can't boot an Alpine box atm. 2017-03-07 00:20:39 *lol* Okay, let me find a link and type it :) 2017-03-07 00:21:03 I see http://git.alpinelinux.org/cgit/mkinitfs/ but are your contributions pushed? 2017-03-07 00:21:23 skarnet: That's just it, I'm using the stock one! 2017-03-07 00:22:09 I suspect recent kernel changes or busybox change may have an impact? Odd, to put it mildly. 2017-03-07 00:22:39 holy fucking shell script batman 2017-03-07 00:23:23 I can't believe we need ALL of this in an initramfs 2017-03-07 00:24:31 skarnet: http://git.kernel.org/cgit/boot/dracut/dracut.git/tree/ 2017-03-07 00:24:34 skarnet : Yeah, that's why I'm trying to modularize things a bit :) 2017-03-07 00:24:58 especially http://git.kernel.org/cgit/boot/dracut/dracut.git/tree/modules.d 2017-03-07 00:25:04 skarnet: Include the portions we need, not the rest. 2017-03-07 00:25:47 though, dracut /is/ modular and only includes what you specify 2017-03-07 00:25:47 awilfox: "the rest of the world sucks even more" has never comforted me 2017-03-07 00:26:10 if I was comforted because other people suck more, I'd feel like a god all the time 2017-03-07 00:26:19 bahaha 2017-03-07 00:26:41 Do we actually need to bind-mount proc for that little operation, or can we just throw in a couple symlinks then kill them after? 2017-03-07 00:27:00 you'd have to ask fabled, at a more EU-friendly time 2017-03-07 00:27:28 my job is to write initramfs scripts that are 20 lines long 2017-03-07 00:27:31 tios 2017-03-07 00:27:34 tops* 2017-03-07 00:27:43 not to analyze giant moussakas 2017-03-07 00:28:28 hmph, ours is 329 lines :( https://code.foxkit.us/adelie/image/blob/master/cdinit.c 2017-03-07 00:28:50 apparently that's the job *you*'ve decided to take, and in the wise words of Mr. T: I pity the fool 2017-03-07 00:28:52 *Grin* Yeah, this is a bit painful.. 2017-03-07 00:29:23 I'm ALMOST tempted to just rewrite it, but I'm already a week late on a project for a client with this. 2017-03-07 00:29:59 awilfox: ah, because initramfs scripts aren't hard enough to debug, so you want initramfs *binaries* instead :P 2017-03-07 00:30:25 skarnet: not needing to pull in an entire busybox or bash or zsh or ash made it much smaller 2017-03-07 00:30:27 skarnet: Actually, that might be less painful at this point ;) 2017-03-07 00:30:36 I'm talking 2-7 MiB smaller, depending on which one of those I picked 2017-03-07 00:30:47 awilfox: what the hell do you think execline is for 2017-03-07 00:30:57 hm 2017-03-07 00:31:06 didn't think to try execline. never have written a script in it before 2017-03-07 00:32:31 skarnet: please, you can’t be serious with execline… 2017-03-07 00:32:53 you need to say "surely you can't be serious" 2017-03-07 00:33:05 so I can answer "I am serious, and don't call me Shirley" 2017-03-07 00:33:24 what about Laverne? 2017-03-07 00:33:50 Laverne? 2017-03-07 00:35:33 jirutka: to clarify: I'm not suggesting using it for Alpine - you'd need to napalm the current initramfs first and spread ammonia on the ashes - but to replace Adélie's C program it's perfectly well-suited 2017-03-07 00:35:34 awilfox : I guess we're the geezers in the room *lol* 2017-03-07 00:35:56 wat? In file included from /usr/include/tclap/CmdLine.h:27: 2017-03-07 00:35:56 #include 2017-03-07 00:36:05 In file included from /usr/include/tclap/CmdLine.h:27: 2017-03-07 00:36:05 . /usr/include/tclap/SwitchArg.h:27:10: fatal error: 'string' file not found 2017-03-07 00:36:05 #include 2017-03-07 00:36:17 clang complains, g++ does not 2017-03-07 00:36:44 TemptorSent: or just the Americans. Foreigners don't know all the American sitcoms that come out. :P 2017-03-07 00:37:03 string.h ? 2017-03-07 00:37:34 maybe g++ automatically completes the extension while clang doesn't. 2017-03-07 00:38:45 jirutka: clang++ 2017-03-07 00:39:00 awilfox: https://youtu.be/ixljWVyPby0?t=66 2017-03-07 00:39:18 skarnet: like that usually means C++ std::string, which is in header 2017-03-07 00:39:32 oh crap: error: unknown warning option '-Wno-maybe-uninitialized'; did you mean '-Wno-uninitialized'? 2017-03-07 00:39:37 ah, C++ 2017-03-07 00:39:57 skarnet: lol, yes, I've seen Airplane! :P 2017-03-07 00:40:48 skarnet: Yeah, "Laverne & Shirley" was somewhat less widespread than some of the other shows of the time I guess. 2017-03-07 00:53:00 Bingo! Looks like a change in busybox that doesn't like umount having multiple arguments! 2017-03-07 00:54:19 Damn, nope... something still strange. 2017-03-07 01:00:15 a non-trivial chunk of the alpine initramfs is for boot from ram setup 2017-03-07 01:07:20 kaniini: Which is exactly what I'm trying to use... 2017-03-07 01:08:01 kaniini: The frustrating part is the rather less than clear codepaths and options, with random behaviours and NO debugging output. 2017-03-07 01:09:31 awilfox: but why it can’t find it? I have installed clang, clang-libs, libstd++… 2017-03-07 01:12:22 kaniini: Do you have any thoughts as to why I'm getting kicked to a rescue shell when trying to boot an iso that uses that standard initramfs? 2017-03-07 01:13:52 :D 2017-03-07 01:13:58 TemptorSent: i have no idea sorry 2017-03-07 01:17:23 kaniini: Thanks :) 2017-03-07 01:20:09 Okay, got it to boot after modifying the init script to use 'umount -l -r -f $mntpt 2> /dev/null' for each mountpoint. 2017-03-07 01:20:31 Something is weird in mount-land somewhere, but I'm not seeing where yet. 2017-03-07 01:23:14 Okay, that's cool... now where is my world file? 2017-03-07 01:24:15 Hard for the overlay to run daemons that aren't installed :P 2017-03-07 01:27:54 /etc/apk/world 2017-03-07 01:32:04 Yeah, it's looking awfully empty :) 2017-03-07 01:32:14 jirutka: are you running the binary 'clang++' 2017-03-07 01:32:23 yes 2017-03-07 01:32:31 I don't believe the build system previously installed anything other than base in the image. 2017-03-07 01:32:39 when g++ is installed on the system, then there’s no problem 2017-03-07 01:33:01 jirutka: that is because clang++ is built against the G++ headers on linux 2017-03-07 01:33:06 jirutka: unless you enable libc++ 2017-03-07 01:33:11 i thought that for clang build I can install only clang and libdstc++ 2017-03-07 01:33:16 aha 2017-03-07 01:33:47 TemptorSent: it didn't 2017-03-07 01:34:10 TemptorSent: because the idea is to install the packages you need :P 2017-03-07 01:34:18 kaniini: A bit of a probem when trying to use a run-from-ram box. 2017-03-07 01:34:25 no? 2017-03-07 01:34:39 you are supposed to use lbu(1) to commit the packages after installing them 2017-03-07 01:34:57 kaniini: Autogenerating and installing ssh keys is nice and all, but it helps to actually have ssh install at boot. 2017-03-07 01:34:59 the 'profiles' just determine what packages are available on the CD repository 2017-03-07 01:35:36 sure, but that does not mean that you should be installing asterisk and starting it with a default configuration on first boot 2017-03-07 01:35:52 for example 2017-03-07 01:35:57 kaniini : Yeh, like I said, it needs a bit of help :) 2017-03-07 01:36:27 what if the admin does not want SSH enabled by default? 2017-03-07 01:36:38 kaniini: Actually, that's EXACTLY the type of thing I need. Services running by sticking media in a drive, rebooting, and walking away. 2017-03-07 01:37:21 yes, but the current system intentionally isn't like that 2017-03-07 01:37:31 it does not "need a bit of help", it is working as designed 2017-03-07 01:37:33 kaniini - Yeah, for hands-on application it makes sense, but it's hard to ssh into a virtual host that doesn't have sshd running :) 2017-03-07 01:38:04 if you need 'services running by sticking media in a drive' why not just run something like turnkey linux 2017-03-07 01:38:45 kaniini - I understand it's well beyond the original scope of the project, but it is seriously useful. 2017-03-07 01:39:15 I'm building custom, one off images with distinct configruation per image. 2017-03-07 01:39:29 i mean, the difference here is simply that what turnkey linux does would be officially supported by alpine 2017-03-07 01:39:45 not personally convinced its worth that :p 2017-03-07 01:41:04 Current project: GIS datastore and DB, and contanerized applications sitting on a box a thousand miles away that has nobody that even remotely speaks unix. 2017-03-07 01:41:41 Side project: rpi based weather stations/cameras that get a physical sd card per unit. 2017-03-07 01:42:56 On the server side, I actually want to be able to build nested images, with the vm host running in ram and the vms running on whatever storage you want. PXE boot most likely. 2017-03-07 01:44:21 Anyway, it shouldn't be too hard to convince the build system to make a world file to go with it when desired. 2017-03-07 01:45:23 TemptorSent: I think the point is more that if you know it is well beyond the original scope of the project, why are you expecting the project to just work for your use case :P 2017-03-07 01:49:28 don't get me wrong, i think adding more capability to customize the ISOs is a good thing 2017-03-07 01:49:45 i am just concerned that people will start demanding lots of respins 2017-03-07 01:53:22 kaniini - they can spin their own! 2017-03-07 01:53:44 if the framework is flexible enough, sure 2017-03-07 01:53:48 and it has good documentation 2017-03-07 01:53:49 kaniini: in fact, an image for building images wouldn't be a bad thought. 2017-03-07 01:54:24 As far as doing things it wasn't intended to do, the probems I'm running into are more peripheral than that.. I can spit out a world file easily enough :) 2017-03-07 01:56:17 awilfox : Take a look at https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts/scripts/mkimage 2017-03-07 01:57:19 I've at least put a pretty good start into documenting the code, and documenting the rest should probably wait until it's somewhat stable I guess... 2017-03-07 01:58:27 Just pushed revision which I'm working with. 2017-03-07 01:59:09 jirutka: I didn't know that people tried packaging ranger. 2017-03-07 01:59:35 Building a new profile is childs play. For that matter, it's pretty easy to add new boatloaders, image types, and features too. 2017-03-07 01:59:36 pickfire: well, always look into unmaintained first ;) 2017-03-07 02:00:13 :) 2017-03-07 02:01:03 pickfire: but to be honest, I also rarely looks into unmaintained :/ I remember ranger, b/c it’s one of the pkgs took over from one ex-contributor 2017-03-07 02:02:01 Ah 2017-03-07 02:03:25 Me too, I just look into those package useful for me. 2017-03-07 03:14:40 Question: If it makes sense to put a local repository on a cd or usb stick but not install it in the world file, does it make sense to do the same on a rootfs image? 2017-03-07 05:55:29 Alright, just about there on something actually usable for actually building bootable media! 2017-03-07 06:17:09 Eyes are bleadinng, but it appears to actually work! I'm going to have to call it an early night I'm afraid, so I'll miss the guys in europe. 2017-03-07 06:17:31 <_ikke_> TemptorSent: night! 2017-03-07 06:17:49 Not dead yet, just not even going to try for 4 in the morning again. 2017-03-07 06:18:54 Anyway, mkimage now builds custom images, including ready-to-run overlays. 2017-03-07 06:19:40 Still have to figure out why ssh it bitching about host pub key formats, but that's because I'm too tired to care about that kind of detail right now. 2017-03-07 06:20:56 Oh, and it would help if I actuall populated authorized_keys, not just created the public keys for the root login :) 2017-03-07 06:50:20 Okay, much better now. 2017-03-07 08:43:58 Tried sleeping, not happing yet - might as well be productive... Anyone know who maintains the zfs package? 2017-03-07 08:50:27 TemptorSent: zfs is maintained by clandmeter 2017-03-07 08:50:52 ACTION hides 2017-03-07 08:50:57 ;v 2017-03-07 08:51:05 clandmeter: run! 2017-03-07 08:59:29 Hi clandmeter! 2017-03-07 09:00:14 So I'm working on getting a rewrite of mkimage done, have it working, bringing up zfs, and everything... but it's bitchign about the initscript format :) 2017-03-07 09:01:07 Thanks for outing clandmeter scadu! 2017-03-07 09:02:03 clandmeter: So, do you have a working ZFS root setup? 2017-03-07 09:02:10 TemptorSent: I have no doubts he's the happiest man in the world now :f 2017-03-07 09:02:30 *lol* 2017-03-07 09:02:48 scadu: He may have taken your advice and headed for the hills ;) 2017-03-07 09:04:20 Anyway, as far as I can tell the only missing link is probably the initfs/init 2017-03-07 09:05:10 Oh, and a nice little auto-installer at some point if I'm on raw hardware. 2017-03-07 09:51:57 TemptorSent, hi 2017-03-07 09:52:00 i was in meeting 2017-03-07 09:52:08 no im currently not using it. 2017-03-07 10:06:55 clandmeter: No problem, I'm just looking to know what I'm getting into and if anyone else has already done the work :) 2017-03-07 10:07:09 its working 2017-03-07 10:07:13 ncopa uses it 2017-03-07 10:07:25 but he did have some issues with dual booting 2017-03-07 10:08:27 Got it... I have it happily booting up into the ramfs, and running through import, mount, export, and zed. 2017-03-07 10:09:02 The only thing it's complaining about so far is the format of the init.d scripts being unsupported. 2017-03-07 10:09:35 I haven't thrown it on a real pool yet to see how it acts, just testing the startup process thus far. 2017-03-07 10:10:09 hmm, i never had issues with initd, but thats some time ago 2017-03-07 10:10:12 what is the error? 2017-03-07 10:11:39 it's just saying that it doesn't like the runscript header I believe. 2017-03-07 10:12:38 ah that one. 2017-03-07 10:12:50 Simply adding a new first line with an appropriate #! should do it. 2017-03-07 10:13:13 just update it to #!/sbin/openrc-run 2017-03-07 10:14:42 Yup. It doesn't actually break anything, but it makes the otherwise clean startup look a bit crufty :) 2017-03-07 10:15:09 i think zfs provides the openrc files theselves 2017-03-07 10:20:48 I believe that's the case -- it should be easy enough to do some surgery on them. 2017-03-07 10:21:54 (echo '#!/sbin/openrc-run' ; head -n -1 $file ) | cat - > $file 2017-03-07 10:32:04 clandmeter: Here's what I'm using right now to bring zfs into an overlay and autostart it: https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts/scripts/mkimage/features/zfs 2017-03-07 10:37:10 I need to work on the nfs/iscsi features too, but it currently builds images with working overlays that populate tools and run services. 2017-03-07 10:38:53 Anyway, I need to try again on the sleep thing. It's not a good idea to contniue coding when you can no longer see the screen clearly... it's worse to try to do anything with root :) 2017-03-07 12:48:38 The patch that I have sent is superseeded? What does that mean? %3103 2017-03-07 12:55:48 pickfire, maybe there is something related on github? 2017-03-07 12:56:40 clandmeter: ok 2017-03-07 12:59:21 doesnt look like it. does it mention who superseded it? 2017-03-07 12:59:37 No 2017-03-07 12:59:52 changes are it was jirutka 2017-03-07 13:14:42 pickfire: I’ve told you it three times: this package is broken 2017-03-07 13:14:55 pickfire: http://bugs.alpinelinux.org/issues/6839 2017-03-07 13:15:05 Oh 2017-03-07 13:15:31 pickfire: I’ve tested it and the same problem as reported in #6839 persists 2017-03-07 13:15:37 Wai 2017-03-07 13:15:38 t 2017-03-07 13:16:00 wait a moment 2017-03-07 13:16:05 sry, you’re talking about some other patch 2017-03-07 13:16:21 jirutka: How is that related to ranger. 2017-03-07 13:16:33 it’s not 2017-03-07 13:16:42 sorry about typing slowly, still training on dvp 2017-03-07 13:16:52 omg i hate patchwork so much 2017-03-07 13:17:19 Why? 2017-03-07 13:17:34 doesn’t matter now 2017-03-07 13:17:48 this is right, http://patchwork.alpinelinux.org/patch/3103/ has been superseded by http://patchwork.alpinelinux.org/patch/3104/ 2017-03-07 13:18:41 Github workflow is the one that is complicated. 2017-03-07 13:18:58 XD 2017-03-07 13:19:00 Oh 2017-03-07 13:19:30 ACTION my bad 2017-03-07 13:20:20 complicated?! 2017-03-07 13:20:36 it’s not complicated at all, you just create a branch and open PR for it, that’s all 2017-03-07 13:20:42 then you can keep adding new commits into it 2017-03-07 13:20:49 and it’s very clear that it’s related to the same thing 2017-03-07 13:20:54 jirutka: No cli 2017-03-07 13:20:59 there is CLI 2017-03-07 13:21:22 actually there are multiple CLIs that you can use 2017-03-07 13:21:41 patches are easier 2017-03-07 13:21:51 patches via email may be easier for you, but it’s a nightmare for us 2017-03-07 13:21:51 Just no CI 2017-03-07 13:22:10 oh 2017-03-07 13:23:12 jirutka: How's the github workflow? 2017-03-07 13:24:08 Now, I just git send-email 2017-03-07 13:24:33 pickfire: 1. create “Fork” of https://github.com/alpinelinux/aports, 2. clone it locally, 3. create new topic branch, 4. add/change whatever you want, commit and push, 5. open PR 2017-03-07 13:24:49 pickfire: you do 1 and 2 just once 2017-03-07 13:25:18 pickfire: maybe you misses the review process 2017-03-07 13:25:35 jirutka: I had already did it. 2017-03-07 13:26:20 pickfire: when you send a patch, someone should review it, write comments and then you should fix the issues, if any… with GH you just make changes in your branch, commit and push, it’s still in the same PR, very clear what have you changed 2017-03-07 13:26:29 I just didn't submit PR 2017-03-07 13:28:01 you’ve just sent two patches, when the second replaces the first, but on Patchwork they are two separate patches and we must figure out what the heck is that 2017-03-07 13:28:21 this does not happen on GH, you can just fix your patch and force push into your branch 2017-03-07 15:48:55 can I somehow force make to not rebuild targets based on timestamp of files? 2017-03-07 15:56:09 how woudl make know what to rebuild then? 2017-03-07 15:56:15 only missing files? 2017-03-07 15:56:42 <_ikke_> checksums :P 2017-03-07 16:03:29 checksums is what i need and what I would expect from decent build system (that make is definitely not) 2017-03-07 16:03:43 or even missing files would be suitable for one case i need 2017-03-07 17:50:56 'mornning all. Does anyone happen to have any idea why the passwd file at /usr/share/mkinitfs/passwd is so extensive? 2017-03-07 17:54:41 Do we really need 43 entries in the passwd file for the initfs? 2017-03-07 17:56:35 Is it just so we can resolve usernames in the early boot? 2017-03-07 18:08:42 :D 2017-03-07 18:09:24 i think most likely somebody just took their /etc/passwd and used that 2017-03-07 18:09:26 :P 2017-03-07 18:11:03 kaniini: I'm not sure about that.... Take a look at it. 2017-03-07 18:12:01 portage? 2017-03-07 18:12:24 mkinitfs predates modern alpine, when it was ncopa trying to make a gentoo binary distro 2017-03-07 18:12:29 so yes, likely :P 2017-03-07 18:12:36 Ahh, got it! 2017-03-07 18:12:50 I was going to say, it looks more like gentoo than alpine :) 2017-03-07 18:13:33 Okay, it can probably be trimmed to just a couple of lines then safely. 2017-03-07 18:14:37 Or replace it with one that has every known account installed by any package on alpine so the userid->user resolving works. 2017-03-07 18:14:53 Right now it's just confusing. 2017-03-07 18:15:05 minimal mkinitfs file seems fine 2017-03-07 18:15:20 initfs should not start any services really 2017-03-07 18:15:31 Do we have a template passwd file with all the uids mapped? 2017-03-07 18:15:56 No, but when you're in a rescue shell trying to figure out what broke, it can be useful to resolve the uids. 2017-03-07 18:15:56 it only needs 1 uid in it 2017-03-07 18:15:58 for root 2017-03-07 18:16:00 :p 2017-03-07 18:16:14 everything in the initfs 2017-03-07 18:16:17 runs as root 2017-03-07 18:16:41 Right, I mean when you get dumped to a rescue shell because the boot sequence failed. 2017-03-07 18:17:37 It's nice to figure out that something isn't starting because it's the ownership on it's key is wrong. 2017-03-07 18:18:16 just use the host's passwd file then 2017-03-07 18:18:27 And especially useful if you have to use it as a rescue tool. 2017-03-07 18:19:12 kaniini: Yeah, if your rootfs is mounting correctly. 2017-03-07 18:20:05 When you're trying to recover a zpool with a crashed os drive, it can be not so easy :) 2017-03-07 18:21:21 But that can be fscked with later if needed. For now, I'll plan on trimming the passwd and group file in mkinitfs 2017-03-07 18:21:57 We can allways annotate it later if we find it needful. 2017-03-07 18:24:48 i mean 2017-03-07 18:24:49 copy it 2017-03-07 18:24:55 when you generate the image 2017-03-07 18:29:20 hi 2017-03-07 18:29:49 TemptorSent: How are you doing? 2017-03-07 18:33:08 pavlix: Still a bit bleary eyed, but doing pretty well otherwise. Yourself? 2017-03-07 18:33:52 pavlix: Late start on the coffee :) 2017-03-07 18:34:40 TemptorSent: I had quite some coffee and teas today and will have more during the week. 2017-03-07 18:37:30 pavlix: I'm out of my preferred beans, so I need to take a trip down the hill into suburbia and visit my favorite roaster ASAP. 2017-03-07 18:39:09 TemptorSent: Where are you from btw? 2017-03-07 18:40:57 I live in Foresthill, California, U.S.A. up in the foothills of the Sierra. 2017-03-07 18:42:05 pavlix: Just a bit below the normal snow-line - but we've been seeing quite a bit of white this year. 2017-03-07 18:48:54 TemptorSent: Anyways, I'm in a pub... was waiting for a friend. He's arrived, so closing laptop.. :) 2017-03-07 18:53:10 pavlix: Have a great evening then. 2017-03-07 19:19:06 Query regarding default work/output directories: Is it safe to assume that somewhere in /tmp is a good default location for building rather than $PWD unless specified? Output can default to $PWD as long as $PWD isn't the script's dir... 2017-03-07 19:22:33 Once I figure out how to get apk to actually use the package cache specified by --cache-dir, that should be sane, right? 2017-03-07 21:44:19 TemptorSent: avoid building in /tmp if you can. /tmp is most likely a tmpfs, i.e. every file you create there eats up RAM. 2017-03-07 21:44:37 that's not a problem for small files, but for a build... not good. 2017-03-07 21:44:41 Use /var/tmp if anything. 2017-03-07 21:48:54 skarnet: what do you think, is it a good idea to mount /tmp as tmpfs at all? 2017-03-07 21:50:55 I actually use tmpfs for building intentionally for both speed and auto-cleanup :) 2017-03-07 21:51:21 on machines with large amounts of RAM, it's actually fabulous; my 24 GB build box chews up packages in seconds using tmpfs :P 2017-03-07 21:52:18 That's my thinking as well. I use /tmp for exactly that on my machines usually. 2017-03-07 21:53:19 skarnet: In low memory uses, specify a work dir somewhere with space to spare explicitly. 2017-03-07 21:54:18 skarnet: What I don't like about /var/tmp is it often doesn't get cleaned up for long periods and can grow excessively on something like a flash card, not to mention thrashing. 2017-03-07 21:54:44 that's all fine and dandy, but please bear in mind that Alpine has low-resource environments as an intended target 2017-03-07 21:55:08 skarnet: So on things like a rpi, I'd probably want to run the build on a thumb drive or extenal device rather than the sd root. 2017-03-07 21:55:45 if you have a build process that doesn't clean up its temporary files, it's a bug and you own it 2017-03-07 21:56:12 skarnet: Actually, in the case of mkimage, it's intentional. 2017-03-07 21:56:56 skarnet: We really don't want to clean up after a run when we're going to use the same artifacts for several more images. 2017-03-07 21:57:23 I guess you're not going to do this on a low-spec machine then. Have at it. 2017-03-07 21:57:54 jirutka: yes, it's a good idea, because large builds blowing up /tmp are the exception, not the rule 2017-03-07 21:57:55 skarnet: It should work fine on low-spec machines, just pass --workdir /var/tmp 2017-03-07 21:58:45 skarnet: I suppose I could even check arch and possibly parse free to decide. 2017-03-07 21:59:09 absolutely not. KISS please. 2017-03-07 21:59:37 smart heuristics are the first step on the slippery slope leading to moussaka hell. 2017-03-07 21:59:51 (aka GNU programs) 2017-03-07 22:00:39 skarnet: *lol* I mean setting the default based on the arch and throwing a warning if we don't have much free space. 2017-03-07 22:01:25 skarnet: It actually would be good to at least estimate space needs before building to avoid filling up filesystems accidentally on rpis and such with small sd cards. 2017-03-07 22:01:46 that's more complex than it needs to be. If it's user-configurable then you're good, and you should focus on more important things until users come screaming at you that they blew up their /tmp. 2017-03-07 22:02:25 if someone attempts to build 24 GB of images on a RPi then they have more problems than you can fix (i.e. pebkac) 2017-03-07 22:02:32 skarnet : Of course it's user configurable! I'm just looking at where to default it, and where to default the output if the person happens to try to run it in the script dir. 2017-03-07 22:03:16 I'd use /var/tmp as the default, but if you don't like it, /tmp is fine. Just add a warning in the mkimage man page and you're set. 2017-03-07 22:03:19 skarnet: *lol* Yeah, noted -- but I'm also trying to make this something a user could use. 2017-03-07 22:04:40 skarnet: Yeah, I've added significantly more informative text to the build process so it's easier to debug profiles and such. Uses msg, so it can be squelched if desired. 2017-03-07 22:04:42 don't assume users are stupid and need pampering. Of course they are and they do, but you can't assume it - doing so leads to bad design. 2017-03-07 22:05:05 XD 2017-03-07 22:05:48 skarnet: Yeah, mostly I'm worried about making it obvious WHY something fails, since I wasted a couple days trying to track down stupid bugs I couldn't even find the realm of in the initfs stuff. 2017-03-07 22:06:20 learning to print good error messages is a life's work ;) 2017-03-07 22:06:26 When something FAILS, the quiet flag shouldn't inhibit the output! 2017-03-07 22:06:50 yeah, warnings != errors. "quiet" should suppress warnings, but not errors. 2017-03-07 22:06:59 to suppress errors, users can always 2>/dev/null 2017-03-07 22:07:24 I couldn't find the stupid minor bug in my code because of error suppression in init. 2017-03-07 22:08:15 When it fails to load the boot repository due to untrusted keys, I really DO want to know about it before getting unceremoniously dumped to a recovery shell. 2017-03-07 22:09:21 init (and initramfs' pid 1 even more so) is a special case because you don't have a logging infrastructure in place so verbosity management is a bit complicated. If it feels raw, it's because it has to be. 2017-03-07 22:09:23 skarnet: I prefer the precept of be verbose by default and put a gag on it if you want to. 2017-03-07 22:09:32 that's not how Unix works. 2017-03-07 22:10:21 Nope, take a look -- there's no reason why it shouldn't have thrown the warning! I had to force "noquiet" to get it to tell me what crapped out. 2017-03-07 22:10:42 If noquiet let's me see the error, the verbosity management is broken. 2017-03-07 22:11:10 that's very possible. Again, errors shouldn't be silenced by a "normal" quiet mode. 2017-03-07 22:11:30 Actually, for the most part tools are (reaonably) verbose by default, and you have to pass a flag to shut them up. 2017-03-07 22:12:04 Traditional Unix works the other way - terse by default, verbose with a flag. 2017-03-07 22:12:25 but none of this matters when there are actual errors. 2017-03-07 22:12:47 hey leo-cutie 2017-03-07 22:13:02 flirting? in MY #alpine-devel? 2017-03-07 22:13:12 skarnet: take a look at make, cc, etc.. they tend to give you info unless you ask them not to. 2017-03-07 22:13:30 skarnet: If you want EXTRA verbose, you pass the -v flag 2017-03-07 22:13:39 Shiz: anywhere is fair game! Unless you passed regulations? 2017-03-07 22:13:52 i was just gonna say "it's more likely than you think" 2017-03-07 22:14:42 *lol* What is this? Rusty & Eddies? ;) 2017-03-07 22:15:09 TemptorSent: build tools are kind of an exception, especially GNU tools. GNU's not Unix and it shows a lot when you run make+gcc. 2017-03-07 22:15:30 (arguably when you run a build, you *want* reasonable verbosity by default.) 2017-03-07 22:16:16 skarnet: In the OLD days, things like init scripts used to give you a fair bit of info, or at least status as they went. 2017-03-07 22:16:39 what init scripts? sysv-rc? 2017-03-07 22:16:42 skarnet: In fact, they would often actually echo commands. 2017-03-07 22:16:47 older? 2017-03-07 22:16:57 skarnet: sys-v? Yeah that was the NEW init system! 2017-03-07 22:17:55 bsd init was common, before that, it was often brewed-to-taste. 2017-03-07 22:18:05 and again, init scripts are not a very good example because it's just lines scrolling on a terminal at boot time 2017-03-07 22:18:44 and the verbosity is partly intentional, so it looks cool and techy 2017-03-07 22:18:49 Right, which is exactly what I'm bitching about having NO output, even with noquiet it's very minimal and not useful for debugging. 2017-03-07 22:19:38 Shiz: with the gender ratio around here, I doubt it's likely at all. https://plus.google.com/u/0/+LaurentBercotSka/posts/BuK7KxZ7y4U 2017-03-07 22:19:51 skarnet: Wow, the world has changed since I got started in the unix world.. we didn't give a crap what anything looked like then, we just wanted to be able to make it work and keep it working. Readability was a plus. 2017-03-07 22:20:16 skarnet: nothing wrong with a lil gay 2017-03-07 22:21:17 Shiz: obviously, but I don't know that many gay tech guys. (It's not something you usually say out of the blue.) 2017-03-07 22:21:30 skarnet: If you wanted pretty, you did ascii-art or went really out there and did ansi+IBM extended character drawing (remember slackware). 2017-03-07 22:21:35 for what it's worth, i'm bi 2017-03-07 22:22:05 TemptorSent: I know. I always hated that. Some people even used figlet. I wanted to kill them. 2017-03-07 22:23:16 Shiz: that certainly opens up the possibilities in a 99% male world :D 2017-03-07 22:23:24 skarnet: I still had a form-feed line-printer running into the late '90s for logging and even terminal use. 2017-03-07 22:23:38 well, luckily you don't *always* have to combine work with pleasure 2017-03-07 22:23:39 :p 2017-03-07 22:23:50 (my last gf was an artist...) 2017-03-07 22:24:25 of course you don't, but my point is that for us straight guys, we basically *cannot* combine 2017-03-07 22:24:39 TemptorSent: lp1 on fire? :D 2017-03-07 22:24:44 skarnet: What's the old line, "bis are better with statistics"? :) 2017-03-07 22:25:20 and anyway leo silently rejected me 2017-03-07 22:25:38 it is a curse 2017-03-07 22:25:39 now I need to drink my heartbreak away 2017-03-07 22:26:04 skarnet: Yeah, we used to have fun logging into each others BBSs and spamming the logs at 4 in the morning. 2017-03-07 22:26:50 skarnet: It was a great way of pining someone without ringing the phone. 2017-03-07 22:26:51 teenage mischievousness 2017-03-07 22:27:01 with a geek tint 2017-03-07 22:27:37 skarnet: Sometimes, more often than not it was to let them know a UUCP session failed or something and get them to log in and fix it while the phones were still cheap. 2017-03-07 22:28:48 *lol* or FIDO in some cases. 2017-03-07 22:30:28 even if you were well-intentioned, if you found a way to wake me up at 4 am by sending my noisy printer into overdrive, I'd delete your account right away, then disconnect, axe the printer and go back to sleep, and fuck the UUCP session 2017-03-07 22:30:46 that's how friendly I am when something wakes me up 2017-03-07 22:31:44 when I'm oncall, I get paid extra for it. :P 2017-03-07 22:32:06 skarnet: Yeah, this was back in the days when that's the ONLY time you could actually tie up the phone lines for hours at a time, usually 3-6 sessions with different nodes each night on a fixed schedule. 2017-03-07 22:32:53 I'm glad we have decent connections now :) 2017-03-07 22:33:17 skarnet: So if a session died and didn't repop for some reason, an entire days traffic to a major hub might be delayed. 2017-03-07 22:33:29 speak for yourself 2017-03-07 22:33:34 3G is the new UUCP 2017-03-07 22:33:48 TemptorSent: and that's why we need supervision. #myworkhereisdone 2017-03-07 22:34:15 TemptorSent: I had a great evening as you instructed me. :) 2017-03-07 22:34:21 *lol* The difference between 3G and UUCP is I can actually use UUCP from home :) 2017-03-07 22:35:11 skarnet: in re that G+ post 2017-03-07 22:35:24 skarnet: from my mother's experience in hospital work, yes. 2017-03-07 22:35:51 interesting data point, thank you :) 2017-03-07 22:35:56 the hospital's CEO was a woman, the entire executive team were women, and the entire records division was women + 1 gay guy 2017-03-07 22:36:33 not surprised one bit 2017-03-07 22:39:05 Woah -- apk segfaulting.. that's not cool. 2017-03-07 22:39:58 TemptorSent: is this by any chance during a removal? 2017-03-07 22:43:00 awilfox Nope, bad options being passed. 2017-03-07 22:43:29 My bug, but apk dumping core rather than just puking gracefully on the input. 2017-03-07 22:45:27 ACTION sobs 2017-03-07 22:49:00 Bug found -- was simple transpose of $ and { when concatenating ${i}{$needle} is what I had, accidentally resulting in things like linux{-grsec}, which crahed apk when passed for some reason. 2017-03-07 22:50:00 huh 2017-03-07 22:50:29 TemptorSent: exact line passed to apk? 2017-03-07 22:57:58 ug, I already fixed my bug :) it looks something like apk --cache-dir /tmp/mkimage.cache fetch --root $WORKDIR --link --recursive --output "$archdir" wit ha long line of apks, but some of the flavored ones malformed as {-grsec} for the suffix rather than -grsec. 2017-03-07 22:59:32 Anyway, now that that's taken care of, everything is running happily again. 2017-03-07 23:10:15 what G+ post 2017-03-07 23:11:13 Google+ I'm guessing? 2017-03-07 23:11:28 oh, i see 2017-03-07 23:12:14 awilfox: Also, if you know anything about it, I can't seem to get apk --cache-dir to actually use the cache dir I specify! 2017-03-07 23:12:43 humm, should be working 2017-03-07 23:12:54 try apk fetch --cache-dir 2017-03-07 23:13:01 I don't know anything about --cache-dir, sorry 2017-03-07 23:13:32 kaniini: Oh, hope to hell I don't have to specify it after each command... 2017-03-07 23:13:40 :) 2017-03-07 23:13:54 TemptorSent: unless you link /etc/apk/cache to some place yes 2017-03-07 23:13:58 I'm getting a cache dir created, but it's not actually caching anything but indexes to it. 2017-03-07 23:15:40 I have four APKINDEX.tar.gz files and an empty 'installed' file in the specified cache dir. 2017-03-07 23:17:40 Well drat, if the options can't preceed the command, the easy fix of changing the name of the apk command doesn't work. 2017-03-07 23:17:58 that shouldn't be an issue afaik 2017-03-07 23:18:08 apk --arch $ARCH -X "https://distfiles.adelielinux.org/adelie/1.0-alpha/$EXTRA_MIRROR" -U --root squashroot-$ARCH --initdb add $PACKAGES $ARCH_PKGS 2017-03-07 23:18:16 in fact, I put it at the very end 2017-03-07 23:18:18 but 2017-03-07 23:18:22 I don't use cache-dir in this script 2017-03-07 23:18:29 so I don't know if that option specifically is different 2017-03-07 23:18:42 Yeah, --cache-dir is a new functionality. 2017-03-07 23:19:10 Wait a sec, I think I know what the bug is in apk! 2017-03-07 23:19:36 Does root just blindly set the cache dir to $apkroot/etc/apk/cache? 2017-03-07 23:20:13 If so, having it parse AFTER the --cache-dir option may cause it to overwrite the explicitly specified location. 2017-03-07 23:20:54 Subsequent --cache-dirs overwriting each other is fine, but if the root overwrites it, there's a problem. 2017-03-07 23:22:03 kaniini: Can you verify that? 2017-03-07 23:24:43 no... 2017-03-07 23:24:47 Anyone up for taking a look at the current state of my mkimage branch and seeing if it works for you? 2017-03-07 23:24:48 the default is no cache dir is used 2017-03-07 23:26:34 kaniini: Understood, what I mean is that we have a command the specifies an explicit cache directory with --cache-dir, then later explicitly specifies a root with --root. What I suspect happens in the code is the explicit path specified by the cache dir is overwritten by the one inferred from the root. 2017-03-07 23:33:14 cachedir is set using "dbopts->cache_dir = optarg;" on line 134 of apk-static... 2017-03-07 23:34:24 apk.c as well. 2017-03-07 23:45:03 It looks like the --cache-dir is getting referenced in absolute terms in some places and relative to root_fd in others. 2017-03-07 23:45:50 Hopefully ncopa or fabled can take a look at it. 2017-03-07 23:48:50 fd = openat(db->root_fd, dbopts->cache_dir, ... 2017-03-07 23:49:54 Maybe check for leading slash before prepending the --root root? 2017-03-08 00:21:04 TemptorSent: if it is inconsistent, that is a bug, but --root changes the root everywhere. including cache dir 2017-03-08 00:25:09 awilfox: Right, which essentially makes the --cache-dir feature useless. 2017-03-08 00:25:57 I'm installing in four different --roots in the proces of building ONE image. 2017-03-08 00:26:36 Same goes for building chroot environments, it's not useful if I have to go set up a link for every single one. 2017-03-08 00:27:52 maybe --cache-root would make sense, idk, you'll have to wait for fabled. 2017-03-08 00:28:10 If is pass " --cache-dir /media/usb/apkcache --root /mnt/mynewalpine", it really should use /media/usb/apkcache as the cache directory! 2017-03-08 00:29:14 If is pass " --cache-dir tmp/apkcache --root /mnt/mynewalpine" then I might reasonably expect to get /mnt/mynewalpine/tmp/apkcache. 2017-03-08 01:42:57 Alright, in other news, let's see if we can get postgis and all of its deps bumped successfully... 2017-03-08 02:15:06 you talk a lot in absolutes 2017-03-08 02:15:16 XYZ is useless because it doesn't do what i want 2017-03-08 02:17:28 kaniini: Okay, can you explain how you would expect it to act and when you would use the --cache-dir option with a --root option and with it appending the cache dir to the root dir? 2017-03-08 02:20:08 have you considered that they aren't meant to be used together? 2017-03-08 02:21:47 and what happens by the way, is that it chroot() into the --root directory and then --cache-dir processing comes later 2017-03-08 02:23:00 (well that is an oversimplification, in reality openat(2) is used to accomplish it) 2017-03-08 02:23:27 kaniini: Right, I looked at the code. It's currently treating the --root option as the absolute root of the fs, not the root of the apk tree. 2017-03-08 02:23:44 that's the intended use of it 2017-03-08 02:23:47 Try asking tar to do the same thing and see what happens.. 2017-03-08 02:23:55 apk isn't tar 2017-03-08 02:24:14 tar -c -C /some/path / 2017-03-08 02:24:28 that doesn't change the fact that apk isn't tar and has a different usecase 2017-03-08 02:25:06 TemptorSent: consider this 2017-03-08 02:25:07 Like I said, can you explain when you would use --cache-dir on the command line if you're NOT trying to foce it to an absolute directory? 2017-03-08 02:25:29 fuck this, i'm out 2017-03-08 02:25:49 Sorry, I'm just trying to understand how the current behavior actually provides a benefit. 2017-03-08 02:26:11 I can't figure out how I would use it in practice as it sits. 2017-03-08 02:26:28 i've explained to you already, why it is why it is. 2017-03-08 02:26:53 it is not foreseen that one would want to use both --cache-dir and --root at the same time 2017-03-08 02:27:16 That wast the explicit reason I requested it of fabled a couple weeks ago. 2017-03-08 02:28:04 If one wanted to set the cache dir globally, a config file or env variable would be the place to do it, not the command line. 2017-03-08 02:29:02 okay, let me try once more 2017-03-08 02:29:08 Using --root /some/new/root is a huge thrash of cache. 2017-03-08 02:29:23 EVERY OTHER OPTION THAT SPECIFIES A PATH ALSO DOES IT RELATIVE TO --ROOT 2017-03-08 02:29:36 HAVE YOU CONSIDERED JUST USING A FUCKING HARDLINK 2017-03-08 02:30:10 EVERY time I call apk --root --init? 2017-03-08 02:30:15 YES 2017-03-08 02:30:25 EVERY TIME YOU CALL ANY APK COMMAND 2017-03-08 02:30:32 AS I EXPLAINED HOURS AGO 2017-03-08 02:30:34 I'm talking DOZENS of calls. 2017-03-08 02:30:55 then make /etc/apk/cache a hardlink to your actual cache dir 2017-03-08 02:31:06 not a symlink; a hardlink, so it can escape the chroot 2017-03-08 02:31:22 Okay, so basicaly go back to what I was doing before and creating a whole directory structre, creating sym links, etc? 2017-03-08 02:31:37 oh my god 2017-03-08 02:32:25 apk --root /foo --initdb 2017-03-08 02:32:48 ln /your/path/to/cache /foo/etc/apk/cache 2017-03-08 02:32:52 Right, and my apk cache is on /var, and my builds are on /tmp 2017-03-08 02:32:57 2017-03-08 02:33:01 different FS 2017-03-08 02:33:03 okay 2017-03-08 02:33:12 mount -o bind /your/path/to/cache /foo/etc/apk/cache 2017-03-08 02:33:12 Hardlinks across FS no bueno :) 2017-03-08 02:33:26 Agan, that's for EVERY call to apk. 2017-03-08 02:33:37 oh gnoes 2017-03-08 02:33:49 Hundreds of them to build all profiles. 2017-03-08 02:34:03 whatever shall we do with the 2 to 3 extra seconds this takes in aggregate 2017-03-08 02:34:10 even for hundreds of them 2017-03-08 02:34:33 Um, it's a bit more than that... 2017-03-08 02:34:43 bind mounting? not really 2017-03-08 02:34:56 And this happens as non-root HOW btw? 2017-03-08 02:35:02 make a suid helper 2017-03-08 02:35:18 Complexity much? 2017-03-08 02:35:28 your entire project is complex 2017-03-08 02:35:33 what's a little more 2017-03-08 02:35:40 It's hardly just mkimage. 2017-03-08 02:36:19 I want a new alpine root at /mnt/alpine, right? 2017-03-08 02:36:36 I do apk --root /mnt/alping --initdb 2017-03-08 02:37:05 20:25 <@kaniini> fuck this, i'm out 2017-03-08 02:37:11 Now I install a dozen or so apks: 'apk --root /mnt/alpine add $apks' 2017-03-08 02:37:28 you could flatten them down to a single transaction 2017-03-08 02:37:33 however, see above 2017-03-08 02:38:20 your average chroot deployment does not care about the cache dir 2017-03-08 02:38:21 Now I want to do it again for another root, so 'apk --root /mnt/alpine2 --initdb add $apks' 2017-03-08 02:38:28 yeah 2017-03-08 02:38:31 this is a solved problem 2017-03-08 02:38:33 they call it docker 2017-03-08 02:38:36 good luck, have fun 2017-03-08 02:39:15 you could also just use a local repository 2017-03-08 02:39:20 then caching doesn't matter 2017-03-08 02:39:27 Now, I want to have it use the cache, so I do: 'apk --cache-dir /media/usb/mycache /mnt/alpine3--initdb add $apks' 2017-03-08 02:39:34 Now, what happens? 2017-03-08 02:40:01 Trying again attempting to use the same cache: 'apk --cache-dir /media/usb/mycache /mnt/alpine4--initdb add $apks' 2017-03-08 02:40:04 literally you get two syntax errors (/mnt/alpine3--initdb is not a valid command) 2017-03-08 02:40:15 *lol* okay, my fingers are tired. 2017-03-08 02:40:31 again, just run your own local repository 2017-03-08 02:40:32 Trying again attempting to use the same cache: 'apk --cache-dir /media/usb/mycache --root /mnt/alpine4 --initdb add $apks' 2017-03-08 02:40:45 That's what I'm trying to BUILD 2017-03-08 02:40:56 serve it using SimpleHTTPServer on localhost:8888 or something 2017-03-08 02:41:11 you're trying to build a local repository? that is not what cache-dir is for 2017-03-08 02:41:23 the pain is that apk knows how to use links, but WON'T and I can't force it to still. 2017-03-08 02:41:28 awilfox: don't even bother :D 2017-03-08 02:41:35 apk fetch -r $apks -o /your/stuff, then apk index it and serve it 2017-03-08 02:41:38 simple 2017-03-08 02:41:41 so simple 2017-03-08 02:41:43 lol 2017-03-08 02:41:51 How many copies do I REALLY need on my system? 2017-03-08 02:42:01 apk does not store a copy by default 2017-03-08 02:42:06 if it's in the cache, it just links it rather than copies it. 2017-03-08 02:42:09 i have been telling you that for hours too 2017-03-08 02:42:19 when you apk add a package 2017-03-08 02:42:35 it just streams the package file straight into the processing routines 2017-03-08 02:42:38 I have /etc/apk/cache -> /var/cache/apk 2017-03-08 02:42:41 does not store it on disk 2017-03-08 02:42:43 yes 2017-03-08 02:42:51 but you do not need it 2017-03-08 02:43:07 You must have a hell of a lot faster connection than I do! 2017-03-08 02:43:26 I already solved your problem 2017-03-08 02:43:27 It takes me 20 minutes to download a full set of apks for a reasonable system. 2017-03-08 02:43:38 by the way, if you do apk --repository /foo/bar it will reference the repo outside of the chroot 2017-03-08 02:43:40 fetch once, make your own server, done 2017-03-08 02:43:49 you do not even have to make your own server 2017-03-08 02:45:02 I need a package cache that acts like a cache because that's what it's being used for. If 4 different profiles need the same package on the same arch, it should just ln it. 2017-03-08 02:45:28 so implement an apk cache server 2017-03-08 02:45:31 What is the intent of the --cache-dir option otherwise? That's what I can'f figure out. 2017-03-08 02:45:38 When would you use it? 2017-03-08 02:45:57 the apk cache itself is a badly named feature sir 2017-03-08 02:46:36 it is for 'run from ram' instances to lazily build a local repo on a local disk for storing updates + added packages so the system can be reassembled upon reboot 2017-03-08 02:46:46 It seems to work just fine when I can use it by having multiple runs with the same root. 2017-03-08 02:47:14 i am just telling you what it is for 2017-03-08 02:47:20 no need to be combative 2017-03-08 02:47:25 So people like me that have slow networks are basically SOL? 2017-03-08 02:48:10 I'm not trying to be combatative, I'm trying to figure out why a feature works the way it does when such behaviour seems to have no particular use. 2017-03-08 02:48:52 seems to have a particular use to the people who run from ram 2017-03-08 02:49:02 (it's not for building chroots, really, it's not!) 2017-03-08 02:49:24 what you really want to do is 2017-03-08 02:49:28 I read through the source of apk and I get what it's doing with the database -- the apk cache basically stores versioned APKINDEXes that allow multiple sources to coexist. 2017-03-08 02:49:29 install something like nginx 2017-03-08 02:49:54 and have it cache the packages at your gateway 2017-03-08 02:49:56 I don't really want multiple copies of every package needed, especially the kernel and firmware! 2017-03-08 02:49:57 then you use 2017-03-08 02:50:02 right 2017-03-08 02:50:04 so 2017-03-08 02:50:06 listen 2017-03-08 02:50:09 and i will tell you what you actually want to do 2017-03-08 02:50:33 what you want to do is implement your own local package cache server 2017-03-08 02:50:35 using nginx 2017-03-08 02:50:36 maintain a local fork of apk if I have to ;) 2017-03-08 02:50:52 which is very easy to do 2017-03-08 02:51:27 Does that work with 'apk fetch -L' ? 2017-03-08 02:51:48 Because *THAT* is what I'm trying to accomplish 2017-03-08 02:52:12 ONE set of copies where I'm building and that's it. 2017-03-08 02:52:43 you want something like this 2017-03-08 02:52:54 http://turtle.dereferenced.org/~kaniini/nginx-cache.txt 2017-03-08 02:53:02 it will accelerate your apk fetches 2017-03-08 02:53:04 done 2017-03-08 02:54:13 well 2017-03-08 02:54:16 really more like this 2017-03-08 02:54:28 http://turtle.dereferenced.org/~kaniini/nginx-cache2.txt 2017-03-08 02:54:32 then you use 2017-03-08 02:54:42 http://some.ip:8080/alpine/edge/main 2017-03-08 02:54:44 as your repo 2017-03-08 02:54:50 and voila it's cached 2017-03-08 02:54:56 with only one copy downloaded 2017-03-08 02:55:42 that approach also works with debian repos too 2017-03-08 02:57:10 Okay, thats's cool and all, but when I have say 10 apkroots with the same packages, how many actual files do I have? 2017-03-08 02:59:51 just hardlink the duplicates https://liw.fi/dupfiles/ 2017-03-08 03:00:35 mitchty: That's the point, apk already knows how to use hardlinks where it can, but I can't force the directory to an absolute path. 2017-03-08 03:01:16 I'm not IN a chroot jail, I'm outside trying to get my cache path to referr to the location I specify. 2017-03-08 03:02:24 In fact, there's no actual chroot involved, just a lot of initilizing repositories, indexing them, signing them, and dumping them to an image. 2017-03-08 03:04:53 then maybe ask fabled nicely to add literal path support to cache-dir 2017-03-08 03:04:55 My real question is is there a use case for the current behavior to stand? 2017-03-08 03:05:16 yes, run from ram setups 2017-03-08 03:06:14 Okay, more to the point, if the --cache-dir option is passed and the --root option is passed, is there any reason to not have the --cache-dir option refer to the absolute root as long as it has a leading slash? 2017-03-08 03:07:42 Can you think of any case where that would NOT work? 2017-03-08 03:08:01 Or where that would not be the expected behavior? 2017-03-08 03:10:04 The help text says '-p, --root DIR Install packages to DIR' 2017-03-08 03:10:58 And '--cache-dir CACHEDIR Override cache directory' 2017-03-08 03:12:09 Nowhere does --root say it treats that as the FS root, only the root for installing, nor does --cache-dir say anything about CACHEDIR being relative to --root. 2017-03-08 03:13:13 So all I'm asking is for the behavior to match the documentation, or explain why it behaves in the manner it does and change the documentation (and add a way to ACTUALLY override the cache dir) 2017-03-08 03:14:40 It's a bug either way you slice it when the actual behavior is ambiguous or contrary to docuemtation. 2017-03-08 03:15:37 I can parse C pretty well, so I was able to suss out what was taking place, but it was non-obvious to say the least. 2017-03-08 03:29:03 kaniini: On a run from ram system, where my apk root is the same as my fs root, --cache-dir referrs to whatever location relative to absolute by accident. In a pre-pivot situation, if I actually expected to access the cache in the sysroot, I would expect to prepend $sysroot, otherwise loading from media, etc, fails. 2017-03-08 03:29:42 So I believe even for that use case, the current behavior is in fact broken. 2017-03-08 03:37:21 IMHO all specified files/directories should be handled as follows: Begins with /,./,../, refer to actual root, unadorned leading paths refer to whatever the current root is. 2017-03-08 03:39:30 Sorry, it's just this happens to be a very real and immediate pain point for me and is severly impacting my productivity. 2017-03-08 03:40:14 And all the FS thrashing actually IS causing problems, as my SD card is starting to hang up while remapping. 2017-03-08 03:55:06 SD card? :D 2017-03-08 04:02:55 kaniini: What can I say, it's what I had on hand at the time :) 2017-03-08 04:03:09 Micro-sd no less, in one of those tiny readers. 2017-03-08 04:05:06 At this point, I'm too far in the middle of a project to take the time to build a proper sytem.. I have 32g of memory sitting next to me, but I don't want to take the tiem or risk some random damage taking me out when I'm already behind. 2017-03-08 04:07:03 I'm working of 24G of a 32G sd card! 2017-03-08 04:07:29 I have more ram sitting next to me than my entire storage space. 2017-03-08 04:08:10 So when it comes to wasting storage and bandwith resources I feel the pain worse than most! 2017-03-08 12:30:24 fedora hosted retired: https://fedoraproject.org/wiki/Infrastructure/Fedorahosted-retirement 2017-03-08 12:30:36 we need fix those: $ grep fedorahosted main/*/APKBUILD community/*/APKBUILD testing/*/APKBUILD | tpaste 2017-03-08 12:30:36 http://tpaste.us/1vO0 2017-03-08 13:29:34 is this an abi change? 2017-03-08 13:29:35 -usr/lib/libosinfo-1.0.so.0.3.0 2017-03-08 13:29:35 +usr/lib/libosinfo-1.0.so.0.1000.0 2017-03-08 13:44:59 holy version bump batman :D 2017-03-08 13:46:25 fcolista, only if the soname changes 2017-03-08 13:47:01 this means that only if from usr/lib/libosinfo-1.0.so becames usr/lib/libosinfo-1.1.so 2017-03-08 13:47:05 ? 2017-03-08 13:47:55 depends on how the library is linked 2017-03-08 13:48:25 by default, yes, your interpretation is correct and this is not an ABI change 2017-03-08 13:48:40 but it would still be wise to check the way the library is built 2017-03-08 13:50:39 skarnet, how ? 2017-03-08 13:52:14 does the configure/make/$buildsystem pass weird flags to ld? typically, grep for "-soname" 2017-03-08 13:53:05 if no -soname is given to ld (no -Wl,-soname to the link step of gcc) then the soname is still libosinfo-1.0 and you're fine 2017-03-08 13:53:39 if it's a regular gnu configure then it's likely the case 2017-03-08 13:54:25 or, a quicker test would be to try building the stuff that uses it and see if it breaks ;) 2017-03-08 13:54:52 skarnet, thx! 2017-03-08 13:55:06 np :) 2017-03-08 15:52:00 fabled: i'm struggling with a segfault. This is what i get with valgrind: 2017-03-08 15:52:00 https://dpaste.de/B0U3/raw 2017-03-08 15:52:16 is this related to a stack corruption? 2017-03-08 15:53:11 might be related to an issue with stack (that in musl is small, while in glibc is bigger)? 2017-03-08 15:54:28 I've this issue opened since a while, and openvas-dev are ignoring this issue: 2017-03-08 15:54:29 http://lists.wald.intevation.org/pipermail/openvas-devel/2016-November/003769.html 2017-03-08 16:05:52 ncopa, during my musl debug, I found an issue on ppc64le that is not solved. http://www.openwall.com/lists/musl/2017/03/08/2 2017-03-08 16:29:36 *yawn* 'morning all. 2017-03-08 16:39:12 fabled: Are you still on by chance? 2017-03-08 20:18:41 Hi, though it is beyond devel chat, I am asking because I could not find alpine-linux chat group. Could not find any documents on installing alpinelinux's root in zfs (which reportedly is possible since version 3.5 (https://www.alpinelinux.org/posts/Alpine-3.5.0-released.html). Can anyone elaborate or point to any document? Already tried with ROOTFS=zfs setup-alpine -m sys /mnt, but got installed into ext4. Even trying to load 2017-03-09 06:52:00 Hello fabled, how are you? 2017-03-09 06:52:51 TemptorSent, hey, back online. catch backlog. 2017-03-09 06:53:00 catching up on backlog* 2017-03-09 07:08:18 fabled: Oh.. see you in couple days then I guess ;) 2017-03-09 07:36:53 fabled, i know this is up on my head, but any hint from you can be enlightening: https://dpaste.de/UgF3#L8 2017-03-09 07:37:16 i've greenbone-security-assitant crash when I login 2017-03-09 07:37:33 this collected the core dump and run gdb to get the backtrace 2017-03-09 07:37:45 *I collected 2017-03-09 07:38:17 i've sent those info to openvas devel since one year, never got a reply. So I contacted them over irc 2017-03-09 07:38:25 still waiting for reply 2017-03-09 07:38:49 I wondering what else i should send to help them in troubleshooting 2017-03-09 07:39:42 Seems that nobody uses greenbone security assistant on musl.. 2017-03-09 08:28:31 fcolista, do we have alpine or upstream bug report on it? 2017-03-09 08:29:36 fabled, yes. 2017-03-09 08:29:37 http://lists.wald.intevation.org/pipermail/openvas-devel/2016-November/003769.html 2017-03-09 08:29:59 I've rebuilt greenbone with symbols 2017-03-09 08:30:07 and this is the backtrace i got: https://dpaste.de/7qb3 2017-03-09 08:38:21 thsi is with full symbols (from oepnvas-libraries too): https://dpaste.de/VzR8 2017-03-09 08:46:49 Hey fabled, can you take a look at the behavior of --cache-dir when --root is also passed and let me know if that is what you intended? 2017-03-09 08:47:37 It's forced to a relative (to --root) root even when it was passed an absolute 2017-03-09 08:49:11 TemptorSent, hum, i did not test that. it should probably be absolute. 2017-03-09 08:50:01 Okay, that's what I thought -- it's being set based on root_fd,db->cache_dir IIRC. 2017-03-09 08:50:16 Does the same apply to host keys? 2017-03-09 08:50:55 If so, the easy thing would be to check the leading / and force absolute if present. 2017-03-09 08:51:39 hum 2017-03-09 08:51:46 i thought leading / makes it absolute 2017-03-09 08:51:49 that's how openat() should work 2017-03-09 08:52:11 man page says: 2017-03-09 08:52:12 If pathname is absolute, then dirfd is ignored. 2017-03-09 08:52:12 Hmm, not acting right for me at least then.. I get a rather empty cache dir. 2017-03-09 08:52:43 I'm wondering if it may be actually overwritten by --root processign then. 2017-03-09 08:53:20 should not be 2017-03-09 08:53:24 What I'm trying to do is replace all apk calls with $APK calls, which I have set as APK="abuild-apk --cache-dir=$cachedir" 2017-03-09 08:53:50 ...thus the root command and --root follow the --cache-dir spec. 2017-03-09 08:53:58 yeah, that makes sense and was on my planned changes. but even better if you get it done :) 2017-03-09 08:54:56 You might want to take a look at the PR ;) 2017-03-09 08:56:23 Also, I have a question, would it be absolutely horrific to get rid of the separate _flavored lists and append an artifical "flavor" of "FLAVOR" to the packages that need the special help instead? 2017-03-09 08:57:07 ...and let a quick run through sed | xargs take care of the damage :) 2017-03-09 08:59:33 In fact, take a look at the most recent push just a few seconds ago and you can see where mkinitfs can now be dovetailed in with almost no effort beyond a call. 2017-03-09 09:00:31 At this point, I could use some input regarding the rest of the boot architecture and where you want to go with things. 2017-03-09 09:01:47 It's almost to the point I can use it for my real-world needs, but I won't pretend to have all the use-cases covered. 2017-03-09 09:04:13 Oh, and a bunch of people have been on asking about how to do zfs-root, so it might be good to document what's actually supported, as I think people expect it to work out of the box and setup with setup-alpine. 2017-03-09 09:12:12 Basically what's left is a couple calls to the plugins loader, a bit of startup code and argument passing, an iterator or two, and really not much more that's not implemented as a plugin. 2017-03-09 09:13:14 I have a few more things that bug me that are on the TODO list, but I hit many of the major ones already, and hopefully it's fairly readable. 2017-03-09 11:15:08 fcolista, yes, looks out-of-stack 2017-03-09 11:18:44 fabled, how do we workaround this issue with musl? 2017-03-09 11:21:37 i suspect it's libmicrohttpd creating threads, and gsad not requesting large enough stack size 2017-03-09 11:26:40 fcolista, try something like http://sprunge.us/jdKS 2017-03-09 11:27:04 great 2017-03-09 11:27:07 let me patch 2017-03-09 11:32:00 ok, i've applied the patch. 2017-03-09 11:32:01 https://dpaste.de/N2a8 2017-03-09 11:32:07 it still segfaults 2017-03-09 11:32:46 guys from openvas says that it's related to a malloc() which fails (maybe due to too much local variables declared) 2017-03-09 11:32:59 *might be related, actally 2017-03-09 11:51:50 fcolista, try 4MB ? 2017-03-09 11:52:03 fabled, i tested a HUGE amount 2017-03-09 11:52:08 oh 2017-03-09 11:52:09 512*1024*1024 2017-03-09 11:52:21 that's not good 2017-03-09 11:52:40 but the crash location is different now too 2017-03-09 11:53:05 https://dpaste.de/makz 2017-03-09 11:53:12 this is the new crash 2017-03-09 11:54:51 this is the crash with stack 512*1024*1024 2017-03-09 11:55:17 strace? 2017-03-09 11:55:42 doing right now 2017-03-09 11:56:33 https://dpaste.de/2g8D 2017-03-09 11:56:35 here we go 2017-03-09 11:57:41 fcolista, thread size did not change, i think your value was too large or something 2017-03-09 11:58:32 you need to strace -f to show what created threads are doing, those are the ones that are segfaulting 2017-03-09 11:58:49 the main thread is just blocking in pselect() 2017-03-09 12:00:59 yeah, that too. but mmap() before clone() says it's allocating only ~80k stack 2017-03-09 12:01:22 yeah 2017-03-09 12:01:39 so either the patch is not working, or something else is wrong 2017-03-09 12:01:59 does the thing actually call pthread_attr_setstacksize() with the value it's given in your patch? 2017-03-09 12:02:06 it should 2017-03-09 12:03:23 I can't help but read "gsad" as "gee... sad." 2017-03-09 13:03:52 fabled, how did you see the 80k stack? 2017-03-09 13:14:14 fcolista, in the end of the strace: 2017-03-09 13:14:14 mmap(NULL, 90112, PROT_NONE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x3353d255d000 2017-03-09 13:14:14 mprotect(0x3353d255e000, 86016, PROT_READ|PROT_WRITE) = 0 2017-03-09 13:14:14 clone(child_stack=0x3353d2572a98, flags=CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|CLONE_CHILD_CLEARTID|0x400000, parent_tidptr=0x3353d2572b20, tls=0x3353d2572ae8, child_tidptr=0x3353d2572b20) = 3989 2017-03-09 13:14:27 the second argument (first non-null) to mmap() 2017-03-09 13:15:19 ah 2017-03-09 13:15:19 90112 2017-03-09 13:15:54 so the patch actually didn't change the stack size 2017-03-09 13:16:16 anwyay, the segfault happens in openvas-libraries 2017-03-09 13:19:10 yes, but in a thread created from libmicrohttpd 2017-03-09 13:19:56 ok 2017-03-09 13:20:05 I've tested libmicrohttpd 2017-03-09 13:20:13 with doc/examples it is shipped 2017-03-09 13:20:25 (wondering if it's libmicrohttpd having issue) 2017-03-09 13:20:27 it works 2017-03-09 13:20:33 with all examples 2017-03-09 13:20:40 oh, there's a bug in my patch 2017-03-09 13:20:51 so this is defenetly a openvas bug 2017-03-09 13:20:56 really? 2017-03-09 13:20:57 it's https? 2017-03-09 13:21:01 affects https only 2017-03-09 13:21:07 i tried both 2017-03-09 13:21:18 http and https 2017-03-09 13:21:20 hmm, should've worked for http 2017-03-09 13:21:22 http://sprunge.us/ASMi 2017-03-09 13:21:40 the https side needs the parameter before dh_params due to it being conditional 2017-03-09 13:22:20 applied 2017-03-09 13:22:23 testing now 2017-03-09 13:25:12 http://sprunge.us/MNLB strace output 2017-03-09 13:26:06 still 90112 2017-03-09 13:44:46 strange 2017-03-09 13:44:52 sounds like the patch is not working then 2017-03-09 13:45:52 interestingly. fabled 2017-03-09 13:45:53 http://sprunge.us/GZAj 2017-03-09 13:46:00 https shows that stack is augmented 2017-03-09 13:46:15 532480 2017-03-09 13:46:32 this is when i run gsad via https 2017-03-09 13:56:24 well, it seems to crash regardless 2017-03-09 13:57:15 right 2017-03-09 13:57:33 so it's not due (only) to the stack size 2017-03-09 13:58:47 oh 2017-03-09 13:58:52 you have newer gsad than me 2017-03-09 13:59:40 yes. Is the one in edge 2017-03-09 14:02:11 fcolista, seems there's unix listener now, so http://sprunge.us/gQOX 2017-03-09 14:02:18 not sure if that is it or not 2017-03-09 14:03:34 building with this patch 2017-03-09 14:03:50 but i suspect it might be something different, since it crashed with https too 2017-03-09 14:04:11 oh 2017-03-09 14:04:17 gsad does not build with this patch 2017-03-09 14:04:21 /home/fcolista/aports/community/greenbone-security-assistant/src/greenbone-security-assistant-7.0.2/src/gsad.c: In function 'start_unix_http_daemon': 2017-03-09 14:04:33 error: 'MHD_OPTION_STACK_SIZE' undeclared (first use in this function) 2017-03-09 14:04:56 should be declared 2017-03-09 14:05:43 typo, they should be MHD_OPTION_THREAD_STACK_SIZE 2017-03-09 14:06:06 ah ok 2017-03-09 14:06:36 as in http://sprunge.us/gFJE 2017-03-09 14:06:42 http://sprunge.us/JYOI 2017-03-09 14:06:51 but i think there's something more going on 2017-03-09 14:06:57 any simple way to reproduce? 2017-03-09 14:07:29 apk add openvas-manager 2017-03-09 14:07:29 apk add openvas-cli openvas-scanner 2017-03-09 14:07:29 apk add greenbone-security-assistant python 2017-03-09 14:07:29 apk add redis 2017-03-09 14:07:47 sed -i -e "s/# \(unix.*\)/\1/" /etc/redis.conf 2017-03-09 14:07:52 rc-service redis start 2017-03-09 14:07:52 rc-update add redis 2017-03-09 14:07:57 greenbone-nvt-sync 2017-03-09 14:07:57 greenbone-scapdata-sync 2017-03-09 14:07:58 greenbone-certdata-sync 2017-03-09 14:08:02 openvas-manage-certs -a 2017-03-09 14:08:07 rc-service openvassd create_cache 2017-03-09 14:08:07 rc-service openvassd start 2017-03-09 14:08:12 urgh :) 2017-03-09 14:08:13 openvasmd --create-user=admin --role=Admin 2017-03-09 14:08:21 openvasmd --rebuild 2017-03-09 14:08:30 rc-service openvasmd restart 2017-03-09 14:08:54 then, as last: 2017-03-09 14:08:57 /usr/sbin/gsad --listen=0.0.0.0 --port=80 --mlisten=127.0.0.1 --mport=9390 --http-only -f 2017-03-09 14:08:59 :D 2017-03-09 14:09:40 I like your simple ways to reproduce 2017-03-09 14:09:59 jsut forgot the "simple" part 2017-03-09 14:10:01 :) 2017-03-09 14:18:10 fcolista, urgh, the openvas-libraries seem to allocate 1MB on stack 2017-03-09 14:18:12 try: http://sprunge.us/Vigd 2017-03-09 14:20:08 it affects directly scalability and memory usage in high-traffic situations 2017-03-09 14:20:19 it allocates thread per connection 2017-03-09 14:20:42 i think it would be appropriate to file a bug against openvas-libraries on the huge stack variable 2017-03-09 14:20:49 it's not good in embedded platforms 2017-03-09 14:21:50 fabled, tried this patch : http://sprunge.us/Vigd 2017-03-09 14:21:53 stil lsegfault 2017-03-09 14:30:52 fcolista, try http://sprunge.us/EBJK 2017-03-09 14:30:59 it seems to need 2 1MB buffers 2017-03-09 14:31:04 so 2mb stack is not enough 2017-03-09 14:31:13 that's kinda lame code 2017-03-09 14:39:12 fcolista: i think it should be reported upstream 2017-03-09 14:39:51 ncopa, with 4*1024*1024 patch gsad works 2017-03-09 14:40:08 i think you should report it upstream 2017-03-09 14:40:17 ncopa, yes of course 2017-03-09 14:40:24 i was on openvas channel 2017-03-09 14:40:29 irc channel 2017-03-09 14:40:35 since this bug is there since 1yr 2017-03-09 14:40:37 its not good to expect 4MB stack 2017-03-09 14:40:41 and nobody cared 2017-03-09 14:40:57 code is gnu style 2017-03-09 14:41:02 ... 2017-03-09 14:41:36 yeah, it's crappy xml parser requiring two 1MB buffers on stack 2017-03-09 14:41:51 its an xml parser? 2017-03-09 14:41:52 ugh 2017-03-09 14:41:56 thats security issue then 2017-03-09 14:42:12 it's thread per-connection, with 4MB stack; not great for scalability... 2017-03-09 14:42:34 urgh 2017-03-09 14:42:51 well 2017-03-09 14:43:00 on gnu libc its 8MB per connection then :) 2017-03-09 14:43:14 oh, i thought glibc has 4mb stack 2017-03-09 14:43:26 glibc has 8 by default 2017-03-09 14:43:40 or whatever you set with ulimit 2017-03-09 14:43:56 ok 2017-03-09 15:35:42 fcolista: did you have a backtrace of the openvas-library problem? 2017-03-09 15:37:24 hum 2017-03-09 15:37:27 i think i found it 2017-03-09 15:38:51 well its multiple places 2017-03-09 15:38:59 i wonder if they have some recursive thing 2017-03-09 15:39:54 ncopa, they fixed the issue in their svn repo 2017-03-09 15:41:18 " svn 2017-03-09 15:41:18 we did allocate several mb for the xml parser on the stack instead of using the heap" 2017-03-09 15:43:57 so they did that on purpose? 2017-03-09 15:44:45 apparently yes 2017-03-09 15:46:26 ACTION slow claps. 2017-03-09 15:48:16 skarnet: do you have any url or reference i can use that explains why it is a bad idea 2017-03-09 15:48:41 its a thing that pops up once in a while 2017-03-09 15:48:54 using the stack rather than the heap? 2017-03-09 15:49:02 yes 2017-03-09 15:49:03 using the *thread* stack 2017-03-09 15:49:27 allocating bigger buffers on the stack instead of on heap 2017-03-09 15:49:31 it's a bad idea precisely because of that very use case 2017-03-09 15:49:32 why that is a generally bad idea 2017-03-09 15:49:52 systems generally have limited stack space 2017-03-09 15:49:56 the "fix" i guess is to allocate bigger stack space 2017-03-09 15:50:03 *thread* stack space 2017-03-09 15:50:08 it doesn't matter in a single-threaded program 2017-03-09 15:50:20 but in a multi-threaded program, you have one address space for N threads 2017-03-09 15:50:33 so N stacks have to share the address space 2017-03-09 15:51:09 so you need to decide in advance how much vsz you'll give each stack 2017-03-09 15:51:33 doesnt it also have a perfomance impavt? 2017-03-09 15:51:40 impact* 2017-03-09 15:51:44 it does unless you overcommit 2017-03-09 15:51:49 i mean, if you have your big data on large stack 2017-03-09 15:52:00 the thing is, stack overflow can't be caught like heap overflow 2017-03-09 15:52:15 *nod* 2017-03-09 15:52:21 im thinking cache misses 2017-03-09 15:52:27 so in order to avoid crashes, if you don't overcommit, then you gotta commit all your stack when you create a thread 2017-03-09 15:52:46 if you have large buffer on stack 2017-03-09 15:52:48 (maybe not, but in theory you should) 2017-03-09 15:52:54 data is populated over various functions 2017-03-09 15:53:13 again, I don't think it matters for single-threaded programs with only 1 stack 2017-03-09 15:53:20 then you'd get cpu cache misses on every call? 2017-03-09 15:53:37 that I don't know 2017-03-09 15:53:54 but the problem I'm aware of is definitely virtual space explosion 2017-03-09 15:54:46 and I don't have a good reference about that, but maybe #musl does 2017-03-09 15:55:05 for the record, I do allocate buffers and large structures on the stack 2017-03-09 15:55:27 but 1. my programs are single-threaded and 2. "large", to me, means 1-2 MB at most 2017-03-09 15:55:54 I think the biggest amount of stack I've ever used was around 800 kB 2017-03-09 15:56:20 big applications with megabytes of data should definitely use the heap 2017-03-09 15:58:53 svn log https://svn.wald.intevation.org/svn/openvas/branches/openvas-libraries-9.0/ -l 1 --diff 2017-03-09 16:03:35 i think i need to backport this to 3.5 2017-03-09 16:08:21 fcolista: that was the solution i was hoping for 2017-03-09 16:08:51 good 2017-03-09 16:09:01 good job 2017-03-09 16:09:36 I'm building gsad without the previous patch 2017-03-09 16:09:48 and with the upstream patch applied to openvas-libraries 2017-03-09 16:09:59 let me check if it works 2017-03-09 16:10:47 they should probably check if the malloc succeeded to handle low memory situations 2017-03-09 16:11:39 ah glib handles that 2017-03-09 16:13:07 yup. It worked! 2017-03-09 16:13:26 ah 2017-03-09 16:13:27 no 2017-03-09 16:13:32 it crashes after a while 2017-03-09 16:13:33 :/ 2017-03-09 16:13:49 /0\ 2017-03-09 16:14:14 they probably allocate more big buffers 2017-03-09 16:14:58 fcolista: can you build with debugging symbols enabled 2017-03-09 16:15:09 and generate a backtrace? 2017-03-09 16:15:13 sure 2017-03-09 16:15:18 but now the issue is different 2017-03-09 16:15:22 [107821.227773] MHD-connection[24436]: segfault at 73bce643aef0 ip 000073bce405dd94 sp 000073bce643aee0 error 6 in libpcre.so.1.2.8[73bce404d000+25a000] 2017-03-09 16:15:29 it crashes on libpcre 2017-03-09 16:15:57 would be nice with a backtrace 2017-03-09 16:16:02 sure 2017-03-09 16:16:20 i need to run now, i'll prepare a clean env tomorrow 2017-03-09 16:16:23 thx! 2017-03-09 16:16:26 ok 2017-03-09 16:16:29 have a ncie evening 2017-03-09 16:43:17 question on test suites, if there are multiple levels of tests, aka a short, medium, and silly long running version, what should go into check()? 2017-03-09 16:44:17 reason I ask is i was adding the ghc testsuite stuff and there is a short test suite good for a run on say travis to stay within its hour timeout, a normal run, which takes about 45 minutes on my skylake box, and a long run, which is still going 2017-03-09 16:44:44 i'd go for shorter test 2017-03-09 16:45:19 if we do sillly long then people will just disable all the testing 2017-03-09 16:52:11 Annapurna Labs Alpine platform (ARCH_ALPINE) [N/y/?] (NEW) 2017-03-09 16:52:15 new kernel config option 2017-03-09 16:52:19 i checked up what it was 2017-03-09 16:52:28 http://www.annapurnalabs.com/ 2017-03-09 16:52:37 the logo looks somewhat familiar 2017-03-09 16:52:51 and they apparently have a product line called "alpine" 2017-03-09 16:56:18 humm 2017-03-09 16:56:39 not sure what we should do about that 2017-03-09 16:59:44 i suppose we simply dont support their hardware :) 2017-03-09 17:00:08 seriously, i dont think we care too much 2017-03-09 17:24:50 <^7heo> I can't believe it... 2017-03-09 17:25:00 <^7heo> it's not possible to use WinSCP to connect to sshd... 2017-03-09 17:25:10 <^7heo> (with sshd from alpine) 2017-03-09 17:25:22 <^7heo> I believe it's a mismatch of security protocols. 2017-03-09 17:26:51 yes, not the same as when some company tried to release "GNOME OS" 2017-03-09 17:26:53 :p 2017-03-09 17:29:05 <^7heo> ? 2017-03-09 17:30:21 ^7heo: I connect WinSCP to dropbear on a regular basis. 2017-03-09 17:30:52 ^7heo: https://arstechnica.com/tech-policy/2014/11/gnome-open-source-project-fights-groupon-over-gnome-trademark/ 2017-03-09 17:32:06 <^7heo> skarnet: using openssh here. 2017-03-09 17:32:14 <^7heo> skarnet: it's default in alpine. 2017-03-09 17:32:17 <^7heo> skarnet: can't change it no. 2017-03-09 17:32:18 <^7heo> now* 2017-03-09 17:32:39 I also use WinSCP to connect to openssh's sshd on a regular basis. 2017-03-09 17:32:49 <^7heo> dude, I dunno what's happening. 2017-03-09 17:32:57 <^7heo> I have negative time before deadlines 2017-03-09 17:33:08 <^7heo> I have 100938410985yu105 things on my plate left until tomorrow 2017-03-09 17:33:20 <^7heo> I have NO fucking time to debug any of this. 2017-03-09 17:33:22 yu105 yourself 2017-03-09 17:33:27 <^7heo> yeah I know. 2017-03-09 17:33:31 <^7heo> I don't care actually, sorry. 2017-03-09 17:33:34 <^7heo> I just want this to WORK. 2017-03-09 17:33:41 <^7heo> I didn't expect *ssh* to fuck up. 2017-03-09 17:33:44 <^7heo> first time I have problems ith it. 2017-03-09 17:33:47 <^7heo> with it. 2017-03-09 17:33:52 <^7heo> (unlike my keyboard) 2017-03-09 17:34:33 <^7heo> the real issue is that I don't have any log. 2017-03-09 17:34:35 <^7heo> it just fails. 2017-03-09 17:34:47 <^7heo> and ofc no change in the sshd_config increases the log amount. 2017-03-09 17:34:48 <^7heo> v_v 2017-03-09 17:36:43 <^7heo> ok debug worked. 2017-03-09 17:36:50 <^7heo> skarnet: just one question, are you using windows 10? 2017-03-09 17:36:54 <^7heo> skarnet: or windows 7? 2017-03-09 17:38:38 <^7heo> ok 2017-03-09 17:38:47 <^7heo> auhorized_keys isn't the right name. 2017-03-09 17:39:41 <^7heo> I wasted 1h on a typo. 2017-03-09 17:39:42 <^7heo> as usual. 2017-03-09 17:40:39 <^7heo> Sorry for the noise. 2017-03-09 17:46:18 <^7heo> also, wtfismyip.com \o/ 2017-03-09 19:19:12 have -devel swapped name with -offtopic last time? 2017-03-09 19:19:57 scadu: Quite possible :) 2017-03-09 19:22:59 mmlb: Okay fixed the root cause of xtables-addons even getting considered by making the base profile only set the kernel if it's not already set. Now if kernel_flavors isn't empty, it won't overwrite it. 2017-03-09 21:34:30 clandmeter: new version of gitea released... 1.1.0 8) 2017-03-09 21:34:58 https://github.com/go-gitea/gitea/releases/tag/v1.1.0 2017-03-09 21:35:46 the pr for the blog post have to be merged, but we got 348 merged pull requests :D 2017-03-09 21:58:53 Hi. There is a problem with Alpine Wiki: on page https://wiki.alpinelinux.org/wiki/Upgrading_Alpine#Upgrading_Alpine_Linux_on_CD there are code examples that look like 'setup-bootable -u ERROR in latestalp.php: failed to read /var/www/dl-3.alpinelinux.org/htdocs/alpine/.latest.x86_64.txt /media/$LBU_MEDIA' 2017-03-09 21:59:10 Does anybody know how to fix it? 2017-03-09 22:10:23 laskin: I'm guessing but I think "ERROR in latestalp.php: failed to read /var/www/dl-3.alpinelinux.org/htdocs/alpine/.latest.x86_64.txt" is meant to be a version 2017-03-09 22:10:58 so `setup-bootable -u 3.5 /media/$LBU_MEDIA` maybe? 2017-03-09 22:13:49 (Or maybe v3.5) 2017-03-09 22:19:13 is there a switch with abuild to make sure it calls the check() function in the apkbuild? i'm currently using abuild -r but don't see a switch that screams i'll run checks as well 2017-03-09 22:35:19 Trying to debug cross-arch build of images, but mmlb is reporting a failure by mkinitfs, mostly likely originating from lddtree doing the wrong thing. 2017-03-09 22:36:17 Has anyone else run into issues with lddtree resolving libraries when in a foreign arch root? 2017-03-09 22:40:13 mosez, nice 2017-03-09 22:40:46 i had the minor update still in by git stashed 2017-03-09 22:47:44 mosez, looks like govendor is not happy 2017-03-09 23:25:45 clandmeter: why govendor? you don't need to execute it. the deps are already bundled. 2017-03-09 23:26:19 then why provide the vendor file? 2017-03-09 23:26:21 https://github.com/go-gitea/gitea/tree/master/vendor everything should be there 2017-03-09 23:26:34 because we need to update the deps somehow :) 2017-03-09 23:26:55 sure, so its broken now 2017-03-09 23:27:23 i mean even if i dont need it it should work right? 2017-03-09 23:29:44 would be nice if somebody could fix the makefile to allow setting a version number if building from tarball 2017-03-09 23:31:28 mosez, I moved it to community now. 2017-03-09 23:31:30 Well this is an entertaing rabbit hole -- any sage advice on getting apk to do the right thing when cross-building? 2017-03-09 23:32:16 clandmeter: https://cl.ly/3M0F0T1G2U2M builds fine for me on alpine 2017-03-09 23:32:31 clandmeter: damn... i forgot the version thing :( 2017-03-09 23:33:59 mosez, yes it builds fine. 2017-03-09 23:34:05 its on the repo now 2017-03-09 23:34:12 awesome 2017-03-09 23:34:14 but i had to remove govendor 2017-03-09 23:34:28 and fix the version patch 2017-03-09 23:34:31 you mean the vendor.json file? 2017-03-09 23:34:39 yes 2017-03-09 23:34:59 ok 2017-03-09 23:35:11 on 1.0.x i could use govendor just fine 2017-03-09 23:35:21 not sure how you guy's pull in vendor files 2017-03-09 23:35:44 but we just updated deps, otherwise there had been no changes to the vendor.json 2017-03-09 23:36:42 when i find one i usually use it to make sure the deps are uptodate 2017-03-09 23:37:56 ^7heo, you are now the official maintainer of gitea :p 2017-03-09 23:38:02 I was too lazy to remove it 2017-03-09 23:38:04 i will create an issue for the version 2017-03-09 23:38:40 mosez, did any of the performance fixes get in? 2017-03-09 23:38:54 regarind flat repo's in particular 2017-03-09 23:39:02 regarding... 2017-03-09 23:39:25 like aports... 2017-03-09 23:44:30 bedtime, gnite ppl 2017-03-09 23:45:31 clandmeter: https://github.com/go-gitea/gitea/issues/1173 2017-03-09 23:45:40 clandmeter: i have not checked latest results 2017-03-09 23:50:22 migrating linux and aports to try.gitea.io now :) 2017-03-09 23:53:05 g'night clandmeter 2017-03-10 00:02:12 Hmm, segfaulting apk again, this time no apparent reason. 2017-03-10 00:06:23 strace shows it dying following openat etc/apk/repositories.d 2017-03-10 00:10:17 forcing --repositories-file causes it to crash trying to find keys, specifying --keys-dir leads to crash trying to read APKINDEX caches... 2017-03-10 00:15:21 is adding a test suite worth bumping the release field? 2017-03-10 00:44:50 <^7heo> clandmeter: why not mosez? AFAIK he does the job of leading it 2017-03-10 00:51:02 so… what license is Alpine available under? asking for a friend. 2017-03-10 00:51:40 <^7heo> the microsoft EULA 2017-03-10 00:51:43 Weird - apk throwing "BAD FILE DESCRIPTOR" 2017-03-10 00:55:04 Found I think -- attempting to use --output to a new, non-existent directory. 2017-03-10 00:55:41 Should either throw a "directory not found" error or create the directory. 2017-03-10 00:58:19 mikeee_: it's a distro, there's no single unified license 2017-03-10 00:58:27 it's the various licenses its packages consist of 2017-03-10 00:58:49 all of the alpine-specific stuff like apk and alpine-conf etc are bsd/mit iirc 2017-03-10 00:59:45 oh, apk is gpl2 :) 2017-03-10 01:06:46 and there's no "umbrella" license for the pre-built images on alpinelinux.org/downloads or anything like that? 2017-03-10 07:18:34 morning 2017-03-10 07:22:13 'morning! 2017-03-10 07:22:29 Well, evening for me for another 40 minutes or so :) 2017-03-10 07:24:45 ...so I've been beating my head against the combination of mkinitfs, lddtree, and building for aarch on x86_64. Has anyone already solved this, or shall I continue cussing at it until one of us wins? 2017-03-10 07:28:38 ncopa, i'm going to build openvas with the new patch and symbols 2017-03-10 07:36:25 A spurious libc.musl-x86_64.so.1 dep is showing up when trying to build the image, but a manual run from the shell doesn't come up with the same dep. 2017-03-10 08:33:30 TemptorSent, it's fixed in new lddtree in github 2017-03-10 08:33:34 should apply the patch 2017-03-10 08:34:08 TemptorSent, fix is https://github.com/ncopa/lddtree/commit/a4c4b3ae3ef7ba6f601317e5eb5b92e33fe2f5d4 2017-03-10 08:36:11 fabled: Thank you! I just wasted the entire day trying to work around what looked like a path issue. 2017-03-10 08:36:59 ncopa, please tag lddtree 2017-03-10 08:37:02 That's my only holdup on having aarch64 profiles building on x86_64 hosts at this point. 2017-03-10 08:37:59 yeah, one major reason why we revamped alpine-iso away was to support cross-building images properly 2017-03-10 08:38:44 fabled: Right now, it's doing everything but being happy with cpio's return status because it can't find the spurious x86_64 file. 2017-03-10 08:38:59 yeah, it's the lddtree 2017-03-10 08:39:03 I'm taking a WAG that the environment changes, since I can run it cleanly on the cmdline. 2017-03-10 08:39:05 i fixed it when writing mkimage.sh 2017-03-10 08:39:15 i wanted to also cross-build images :) 2017-03-10 08:40:12 Well, mkimage.sh can now do all sorts of fun things, like dump unique ssh keys on your image, (theoretically cross-arch) 2017-03-10 08:54:45 fabled: Did you get a chance to take a look at apk --cache-dir w/ --root and reproduce the odd behavior? 2017-03-10 08:55:38 fabled: The firmware isn't so small :) 2017-03-10 08:57:53 TemptorSent, yeah, firmware is pretty big; that's why modloop creation script tries to delete unneeded ones 2017-03-10 08:58:09 i had idea that the .apk could be trimmed down too a bit 2017-03-10 08:58:24 but it would require keeping list of the kernel configs in linux-firmware aport 2017-03-10 08:58:39 or more likely the union of produced .ko object's and the firmwares mentioned in them 2017-03-10 08:59:17 would be especially useful if the linux-firmware package was made arch specific and keeping only the arch required subset of firmwares 2017-03-10 08:59:21 I think we can actually pull out what we need pretty easily by splitting the features out better. 2017-03-10 08:59:53 Agreed - there is a huge installed overhead just by having that package in the apk repo 2017-03-10 09:00:29 I actually setup a seperate list for apks ONLY needed for building the initfs so it can be a bit more sane. 2017-03-10 09:01:28 I'd suggest at least splitting the firmware by arch and the modules by function/driver 2017-03-10 09:02:39 So if you need iscsi, the feature pulls in the iscsi deps and nothing else. Need a LSI driver, include the driver_lsi feature. 2017-03-10 09:03:43 It actually should be pretty simple to integrate the function of mkinitfs in the framework of mkimage directly. 2017-03-10 09:03:58 Then just let them both be applets :) 2017-03-10 09:05:47 The modules included on the initfs boot cmdline are now configurable, so we should be able to add everything needed to completely automate the initfs generation and overlays. 2017-03-10 09:21:33 fabled there was a pull request: https://github.com/ncopa/lddtree/pull/5 2017-03-10 09:21:37 apparently he gave up 2017-03-10 09:22:23 Good morning ncopa. 2017-03-10 09:22:31 morning TemptorSent 2017-03-10 09:22:43 sorry for not reviewing your work 2017-03-10 09:22:51 it looks overwhelming 2017-03-10 09:22:53 :) 2017-03-10 09:23:35 ncopa, pull request should prevent tagging release now; you can make new release later when PR is applied :) 2017-03-10 09:24:17 yeah 2017-03-10 09:25:36 ncopa: No problem, it's a bit of a change ;) 2017-03-10 09:26:53 I'm trying to make sure it actually works for all expected functionality at least before even considering moving it towards a production status. 2017-03-10 09:27:10 that would be required :) 2017-03-10 09:27:22 well, at least without breaking existing stuff 2017-03-10 09:27:34 *lol* It seems to be an unusual requirement among many sofware projects these days. 2017-03-10 09:27:57 Right up with eating your own dogfood. 2017-03-10 09:29:13 Was mkimage previously building for all archs correctly from x86_64? 2017-03-10 09:29:59 i think so 2017-03-10 09:30:07 well 2017-03-10 09:30:11 no, i dont know 2017-03-10 09:30:32 *lol* Got it :) 2017-03-10 09:30:53 so, for now, its enough if it does not break building native arch 2017-03-10 09:31:11 fabled: have you tested lddtree from git master? 2017-03-10 09:31:43 ncopa: I had that yesterday AFAIK :) 2017-03-10 09:32:27 What I need to do soon is get a clear idea of what each release profile should contain and what modules are supported for which kernel flavors/archs. 2017-03-10 09:33:16 I'd like to make the base profile a true base with only what's required to bring up a working environment, then have the rest add from there. 2017-03-10 09:34:42 Also, what would you like to see in the overlays, since they can now be quite granular. 2017-03-10 09:37:07 the smallest release profile is the minirootfs 2017-03-10 09:37:32 which should contain the minimal required stuff for a minimal chroot/container image 2017-03-10 09:37:41 this should be used for docker image 2017-03-10 09:37:53 this does not even need openrc 2017-03-10 09:38:06 which you otherwise need for a "working environment" 2017-03-10 09:39:48 then we have a "netinstall" image, which is the alpine-standard 2017-03-10 09:40:07 which is a live cd with kernel etc 2017-03-10 09:40:14 but is minimalistic 2017-03-10 09:40:28 you have only the tools needed to get the network up 2017-03-10 09:40:36 so you can install system from network 2017-03-10 09:41:05 the thinking there is 2017-03-10 09:41:34 if you download a full live-dvd on several GB 2017-03-10 09:41:44 then install that to disk 2017-03-10 09:41:58 i mean if the purpose is to install to disk 2017-03-10 09:42:33 then there is likely alot of stuff on that full live dvd that you will never use 2017-03-10 09:42:57 and there is also a big change that some of the packages are outdated 2017-03-10 09:43:18 so on system update you'll get newer versions of the same packages that you already downloaded 2017-03-10 09:44:01 this means that you transfer things you dont need over the wire to get that big install live dvd 2017-03-10 09:44:41 so on our alpine-standard (which would be renamed to netinstall) 2017-03-10 09:44:50 we only include the things needed to bring up the network 2017-03-10 09:44:57 and you install the rest from there 2017-03-10 09:45:41 that way we reduce the amount of bytes transferred over the wire that you dont really need 2017-03-10 09:46:02 you only install the packages that you intend to use 2017-03-10 09:46:02 Okay, makes sense. 2017-03-10 09:46:12 and you get the updates on first download 2017-03-10 09:46:39 I think we can get away from using the one-off builder for the minirootfs and just make it a profile 2017-03-10 09:47:03 thats good 2017-03-10 09:47:08 i felt it was a bit hackish 2017-03-10 09:47:10 The overlay builder handles pretty much everythign else that I can think of right now. 2017-03-10 09:47:40 because the other profiles were focusing on live cd/dvd style 2017-03-10 09:47:42 with kernel etc 2017-03-10 09:47:56 with kernel and bootrepo 2017-03-10 09:48:02 The worst hackish thing left is the split of apks between flavored and unflavored so they can parse right for multiple kernel builds. 2017-03-10 09:48:45 Would you object horribly to handling all flavored packages as pkgname-FLAVOR and parsing them out at the end? 2017-03-10 09:49:24 not following 2017-03-10 09:49:34 you mean linux-grsec zfs-grsec? 2017-03-10 09:50:53 we dont build all 3rd party modules for all kernel flavors 2017-03-10 09:51:06 for example it makes no sense to build dahdi-virtgrsec 2017-03-10 09:53:06 xtables-addons is the main one. 2017-03-10 09:53:50 we have both xtables-addons and xtables-addons-flavor 2017-03-10 09:54:17 So we have to handle the flavored one specially if we want to have multiple kernels installed in an image. 2017-03-10 10:01:42 i think xtables-addons is userspace or common and xtables-addone-$flavor is the kernel modules 2017-03-10 10:02:02 kernel modules may or may not need userspace tools 2017-03-10 10:02:47 iirc dahdi-linux is the firmware which is common for all supported kernel flavors 2017-03-10 10:04:20 ncopa Right, I just want to stick the logic in the features rather than buried through every tool on the system 2017-03-10 10:04:32 do we really want xtables-addons hard-coded in mkinitfs? 2017-03-10 10:05:42 im not sure why we'd want xtables-addons in the mkinitfs at all 2017-03-10 10:05:58 So if I can put in each feature which kernels it supports, we don't have to worry about adding/removing packages so much. 2017-03-10 10:06:12 I believe so the modules are there for routers and such. 2017-03-10 10:06:29 but thats never needed in the initramfs stage 2017-03-10 10:06:30 Run-from-ram. 2017-03-10 10:06:52 They won't get baked into the .modloop if they're not present when building the initfs. 2017-03-10 10:07:24 that sounds wrong 2017-03-10 10:08:08 That's the way update-kernel/mkinitfs currently work. 2017-03-10 10:08:16 initfs and modloop are different things solving different problems 2017-03-10 10:08:49 The modloop is build as part of the update-kernel process, which just takes apks and mkinitfs features. 2017-03-10 10:09:26 yeah that makes sense 2017-03-10 10:09:36 mkinitfs unpacks the apks, then runs the requested feature filters against that, and runs lddtree to pull all deps for libs, and depmod for kernel modules. 2017-03-10 10:10:08 right 2017-03-10 10:10:15 So ideally each feature would only include the kernel modules it actually needs, not, say, the entire scsi driver tree! 2017-03-10 10:10:31 yes 2017-03-10 10:10:46 so the xtables-addons kernel module gets into the modloop 2017-03-10 10:10:51 but not into the initramfs 2017-03-10 10:11:07 Right, that's what I'm working on cleaning up. 2017-03-10 10:11:26 It will mean changes to update-kernel and mkinitfs as well to do it correctly. 2017-03-10 10:11:39 right 2017-03-10 10:11:45 i think the problem was zfs right? 2017-03-10 10:12:09 we currently need both kernel modules *and* user space tools in the initramfs 2017-03-10 10:12:44 so we need both the zfs-$flavor apk and the zfs.apk (with userspace tool) 2017-03-10 10:13:09 similar with the lvm initramfs feature 2017-03-10 10:13:23 where we need the lvm2 package to be able to build the initramfs 2017-03-10 10:14:37 Right, zfs I have taken care of actually :) 2017-03-10 10:15:18 I just want to get rid of the duplicate variables for tracking the flavored and unflavoed apks. 2017-03-10 10:16:14 where xtables-addons and dahdi-linux are "unflavored" apk and xtables-addons-grsec and dahdi-linux-grsec are the flavored? 2017-03-10 10:16:17 Calling the flavored packages $pkgname-FLAVOR shouldn't conflict with anything (Hopefully?!), and is easy for sed to che up and spit out. 2017-03-10 10:16:25 exactly. 2017-03-10 10:17:10 what happens if there is a flavored apk which does not have any unflavored .apk? 2017-03-10 10:17:19 And zfs is fun, with spl-$flavor, zfs-$flavor, and zfs 2017-03-10 10:17:31 It works currently. 2017-03-10 10:18:20 i suppose spl is a good example of a package that does not have unflavored apk 2017-03-10 10:18:27 the flavored packages are ONLY treated with flavored suffix, so it won't try including the unflavored version (which breaks a few things othewise) 2017-03-10 10:19:13 Exactly. That's what actually started me down this whole rabbit hole with the old alpine-iso (which, by the way, needs to be mentioned as DEPRECIATED on the wiki :) ) 2017-03-10 10:19:13 actually there is a spl userland 2017-03-10 10:19:32 its a wiki ;) 2017-03-10 10:19:56 Noted. 2017-03-10 10:20:12 virtualbox-guest-modules 2017-03-10 10:20:18 has no unflavored apk 2017-03-10 10:20:32 Right, but the initfs needs just the spl & zfs kernel modules and zfs user space. 2017-03-10 10:20:37 right. 2017-03-10 10:21:09 Currently, it would be handled by doing "add_initfs_apks_flavored virtualbox-guest-modules" 2017-03-10 10:21:44 makes sense 2017-03-10 10:21:47 I'd like to change that to "add_initfs_apks virtualbox-guest-modules-FLAVOR" 2017-03-10 10:22:09 then just parse initfs_apks once for each flavor. 2017-03-10 10:22:42 and replace the FLAVOR keyword 2017-03-10 10:22:50 Correct. 2017-03-10 10:23:13 so you want introduce a variable (or macro) and expand it 2017-03-10 10:23:20 Realistically, we should probably have kernel_apks and initfs_apks seperated out. 2017-03-10 10:23:28 Already have automatic expansions for those :) 2017-03-10 10:23:51 var_list_* are aliased for each list variable, so it's already done. 2017-03-10 10:24:26 take a look at one of the profiles in my branch. 2017-03-10 10:24:44 All of the accessors are set up by doing "var_list_alias " 2017-03-10 10:25:04 So you get all the syntactic sugar for relatively little overhead. 2017-03-10 10:35:42 ok 2017-03-10 10:36:14 only thing i think is slightly confusing is the "feature" concept 2017-03-10 10:36:31 which can be mixed with mkinitfs feature 2017-03-10 10:39:13 In theory, the zfs feature will end up enabling everything needed, from the kernel modules, to the kernel command line, initfs binaries, and overlay. 2017-03-10 10:39:53 enabling iscsi will enable the scsi drivers needed for iscsi and that's it, not the whole pile of hardware drivers. 2017-03-10 10:40:04 ncopa, i have the backtrace of yesterday's segfault 2017-03-10 10:40:16 https://dpaste.de/w8tj 2017-03-10 10:41:29 this is the patch i've applied to openvas-libraries: http://sprunge.us/DYSa 2017-03-10 10:45:00 ncopa: Also, not sure if you saw earlier, but I managed to crash apk in all sorts of entertaining ways with misformed input strings such as "package-{grsec}" from a bad "var-{$var2}" instead of $var-${var2} 2017-03-10 10:45:42 If also croked when fed a non-existent directory rather than creating it or complaining. 2017-03-10 10:47:03 thast interesting :) 2017-03-10 10:48:46 Funny what you feed programs when you screw up your script :) 2017-03-10 11:14:44 I'm getting really sick of staring at this thing right now... I can't tell WHAT'S breaking. 2017-03-10 11:15:03 maybe you just need some sleep 2017-03-10 11:15:08 its late over there isnt it? 2017-03-10 11:15:57 Yeah, 3:15... 2017-03-10 11:16:34 I spent 18+ hours fighting with trying to get the damn cross-arch build working, now I'm just going in circles 2017-03-10 11:16:48 get some sleep 2017-03-10 11:16:50 it helps 2017-03-10 11:16:52 And it's prbably a stupid typo somewher e:) 2017-03-10 11:17:12 thank you for working on it 2017-03-10 11:17:23 the general impression of it is good 2017-03-10 11:17:29 looks modular and nice 2017-03-10 11:18:24 Yeah, as soon as I finish seperating functionality, it should actually be easy to debug. 2017-03-10 11:18:48 I seem to be chasing the tail of a random return value. 2017-03-10 11:21:31 Alright, time to give up.. 2017-03-10 11:22:03 and by the way do you have ANY idea where the mv -i for signing the index is coming from? Its seriously fubaring my debugging! 2017-03-10 11:23:06 hum 2017-03-10 11:23:07 dunno 2017-03-10 11:24:13 abuild-apk, abuild-sign, any other place it could be? 2017-03-10 11:26:21 its interactive? 2017-03-10 11:26:41 so it stops the execution? 2017-03-10 11:26:47 and prompt for input? 2017-03-10 11:26:53 try using ps 2017-03-10 11:29:26 Got it! 2017-03-10 11:29:48 Something is popping up interactive mv after signing the index. 2017-03-10 11:30:00 I'm suspecting abuild-sign at the moment... 2017-03-10 11:30:13 Okay, we have it building aarch64! 2017-03-10 11:30:21 Give me a sec and I'll push. 2017-03-10 11:34:53 ncopa, that's the strace, if you want to look at : https://dpaste.de/k4Ud 2017-03-10 11:37:07 fcolista: need to tag the lddtree release first 2017-03-10 11:37:15 im making a simple test uitre 2017-03-10 11:37:18 test suite 2017-03-10 11:38:55 Okay, pushed - should work for any arch now (I think) 2017-03-10 11:42:13 ncopa: Okay, so the release images will be something along the lines of: bare_fs, net_install, minimal, basic, standard, extended, right? 2017-03-10 11:43:20 corresponding to minirootfs, , base, vanilla, standard, extended 2017-03-10 11:43:32 you have the release images here: https://alpinelinux.org/downloads/ 2017-03-10 11:44:06 also 2017-03-10 11:44:32 the text description show on that page should be specified in the profile 2017-03-10 11:44:38 image profile 2017-03-10 11:44:56 Right, those are basically flavors of the other profiles through, right? 2017-03-10 11:45:17 Most of them use standard as their base 2017-03-10 11:45:38 right 2017-03-10 11:45:40 I'll have to look at exactly what was in which for the original profiles. 2017-03-10 11:46:01 i think that standard, vanilla and virtual are the same, with different kernels 2017-03-10 11:46:05 But the way it's put together now, we should be able to make the profiles much cleaner. 2017-03-10 11:46:10 i think maybe i enable serial console on virtual too 2017-03-10 11:46:37 Right, that can all be done with just a couple lines in a profile now. 2017-03-10 11:46:48 extended is same as "standard" but with extra packages added 2017-03-10 11:46:52 you can add AND REMOVE from lists. 2017-03-10 11:47:00 ok 2017-03-10 11:47:28 the arm are slightly different because they are not .iso but .tar.gz images 2017-03-10 11:47:38 And I have it so you can force the kernel from a profile by setting it prior to including the base profile 2017-03-10 11:47:50 I have that working too AFAIK 2017-03-10 11:48:05 ok 2017-03-10 11:48:08 i will test it later 2017-03-10 11:48:14 need clean up other stuff first 2017-03-10 11:48:17 At least it's generating a uboot image for me now for aarch64 :) 2017-03-10 11:49:09 I figured if I can get it building images for x86, x86_64 and arm, we're good for now. 2017-03-10 11:49:29 yeah 2017-03-10 11:50:08 Once I can force update-kernel and mkinitfs to use an explicit apk-cache, it will actually be reasonably quick to generate the whole set, as only minor parts change. 2017-03-10 11:50:39 And get it using -L for copy... 2017-03-10 11:51:20 The repeted copy and extract of the kernel/firmware is killing my poor sd-card based root fs! 2017-03-10 11:52:13 Also, it's sufficiently broken down now that I believe we can toss a multi-worker dispatcher in there and let it rip :) 2017-03-10 11:53:59 Let me know if the automatic ssh features work for you -- add "feature_ssh autostart autokeygen" to a profile and look in the keys dir of the media for the matching set. 2017-03-10 11:54:44 Oh, and is bootchart REALLY necessary on arm? 2017-03-10 11:55:33 i think no 2017-03-10 11:55:46 i think we should disable it by default btw 2017-03-10 11:55:58 it was only to optimize things at boot 2017-03-10 11:56:06 so its a development tool 2017-03-10 11:56:30 okay, good - then I can rip acct out of the initfs! 2017-03-10 11:56:52 Things will look a lot smaller when I'm done with thm. 2017-03-10 11:58:00 Running a build that I'm actually going to be using the profile for on a client's box now... let's see if I broke x86_64 in the process of fixing aarch64 cross-build :) 2017-03-10 11:59:14 Should be able to get the uboot image down by almost half! 2017-03-10 12:01:26 Hmm, only issue I'm seeing with the tarball is it ended up with my username/perms, probably because it's not wrapped in fkrt yet. 2017-03-10 12:05:55 Looks like it still makes native images too. 2017-03-10 12:06:16 Alright, bedtime, before I break something :) 2017-03-10 12:06:31 Goodnight ncopa. Thanks for the input! 2017-03-10 12:06:47 goodnight 2017-03-10 14:15:30 is this correct for adding a testsuite and check() function? https://github.com/mitchty/aports/commit/ab8f3e0e0d7aadbf9dcfec5ae118c4d52842c5ff 2017-03-10 14:16:50 noting that if we're cross compiling there is no way the cross compiled ghc will be testable reliably 2017-03-10 14:33:06 mitchty, when cross compiling, check() is not called 2017-03-10 14:33:32 http://git.alpinelinux.org/cgit/abuild/tree/abuild.in#n71 2017-03-10 14:36:32 ah cool, wasn't sure will nuke that then thanks fabled! 2017-03-10 14:38:24 should i be bumping the release field for this or is it ok to skip it? 2017-03-10 14:47:53 something iw weird with the aarch64 toolchain 2017-03-10 14:48:29 linux-vanilla build fails with 2017-03-10 14:48:30 ld: unrecognized option '-Wl,--as-needed' 2017-03-10 14:48:31 ld: use the --help option for usage information 2017-03-10 14:48:31 make[3]: *** [/home/ncopa/aports/main/linux-vanilla/src/linux-4.9/scripts/Makefi 2017-03-10 14:48:31 le.build:293: arch/arm64/mm/ioremap.o] Error 1 2017-03-10 14:48:47 does not happen on x86_64 2017-03-10 14:49:27 hum 2017-03-10 14:49:32 seems its export LDFLAGS 2017-03-10 15:00:05 and slightly unrelated question, for the idris port https://github.com/alpinelinux/aports/pull/684 I set that up originally to use cabal to download and install all of idris' dependencies at build time https://github.com/alpinelinux/aports/pull/684/files#diff-13750fba489569338de862b1bd266832R31 is that ok or should I work on a similar cabal bridge to what jirutka has for cargo? 2017-03-10 15:02:04 i set it up to use a frozen version constraints file so things shouldn't break in that the versions in the cabal.config will be what cabal uses to build 2017-03-10 15:08:03 but i'm not sure how exactly we want to do build time dependencies for things that get built, I setup idris so that it can run on its own 2017-03-10 15:08:23 though it now needs to have ghc-dev for makedepends 2017-03-10 15:09:53 not sure how to deal with ghc. maybe jirutka has opinion 2017-03-10 15:10:53 yep no rush, iris though is about 50% of why I ported ghc :) 2017-03-10 15:11:01 idris, too early to type 2017-03-10 15:27:10 mitchty: ncopa: I don’t have a solution for cargo/rust yet, just an idea how to do it :) I’ll write more about it later 2017-03-10 15:27:59 jirutka: my mistake, thought you had it ready 2017-03-10 15:28:05 not yet :/ 2017-03-10 15:28:56 anyway, I had started down the path of doing what arch linux does which amounts to having a one way bridge from cabal to pkgbuild files, but its super hacky and i'm not a fan of trying to recreate cabal's pkg management in another package manager 2017-03-10 15:30:01 mitchty: i totally agree with the last sentence; recreating pkg management in another package manager is nonsense 2017-03-10 15:32:01 mitchty: the benefit for us of ghc and rust is that they link ghc/rust dependencies statically, so we must deal just with build dependencies, that simplifies the situation 2017-03-10 15:32:05 jirutka: agreed, i have something that kinda sorta works, but its a ton of work/effort that i'm not sure is worth it in the end 2017-03-10 15:32:39 mitchty: btw I’ve tried upx to minimize size of binaries ghc produces, it has great compress ratio, but doesn’t work with grsecurity :( 2017-03-10 15:33:14 jirutka: heh, i use upx to build a static pandoc and compress it so i can run it anywhere 2017-03-10 15:33:31 mitchty: not everywhere, b/c upx do nasty things in memory… 2017-03-10 15:33:51 jirutka: well, in this case, out of date suse/redhat machines 2017-03-10 15:34:20 mitchty: even when I disabled memory protection for the binary, it still doesn’t work (just segfauts), so maybe it has problem even with running in LXC container 2017-03-10 15:34:42 jirutka: upx is a bit of a hack imo 2017-03-10 15:35:06 I still don’t understand why are binaries produced by ghc so huge 2017-03-10 15:35:16 imo ghc is not doing its work well 2017-03-10 15:35:28 afk 2017-03-10 15:47:11 its hard to say specifically, functional compilers are a bit new to me and i don't yet understand the simplifiers 2017-03-10 16:07:16 ncopa: i noticed that when crosscompiling ppc too 2017-03-10 16:10:01 mitchty: what is in ghc-dev? ghc doesnt seem like something that should have a -dev (since it is probably files you need to run the compiler) 2017-03-10 16:11:37 mitchty: in general -- we want to avoid downloading files on the builders at buildtime, but i am not sure what the solution is for these particular cases 2017-03-10 16:15:11 kaniini: libraries for debugging, its basically the profiled libraries you use if you need to debug something 2017-03-10 16:15:33 you can build things without them 2017-03-10 16:16:13 there are 3 different runtimes, the default, threaded, and profiled/profiled+threaded runtimes, the -dev basically encompasses the latter 2017-03-10 16:17:09 generally unless you're building/debugging libraries you don't need the latter 2017-03-10 16:21:27 got it 2017-03-10 16:22:06 it was also an attempt at making the base ghc a bit smaller 2017-03-10 20:22:43 mmlb : Are you around? 2017-03-10 20:33:35 fabled: one thing for apk-tools 3 i think we should have is multi-line descriptions 2017-03-10 20:36:03 kaniini : That would be nice, but preferable done so that the first line of the description shows for short, and the extened continues naturally, unlike the way several do short/long 2017-03-10 20:36:13 yes, precisely. 2017-03-10 20:36:34 like Debian 2017-03-10 20:36:52 first line is short description, then you continue it below 2017-03-10 20:37:06 vs the RPM way 2017-03-10 20:37:10 And maybe even get people to standardize on the layout of the short description! 2017-03-10 20:38:06 Actually, it would be nice to have the project name show up in the description automagically. 2017-03-10 20:38:30 ACTION is working on gtk+ apk frontend 2017-03-10 20:40:01 add "projname="MyProject"", then have desc automatically prepend that. 2017-03-10 20:40:22 why? the package name is shown in apk search 2017-03-10 20:40:24 :p 2017-03-10 20:40:30 It would make the searches a lot easier where package names don't match the names of the projects they're under. 2017-03-10 20:41:03 for instance "projname="dropbear"" in dbclient 2017-03-10 20:41:54 or projname="ISC DHCP" in dhclient and dhcp 2017-03-10 20:42:35 and the many instances where you want to find all packages related to a project. 2017-03-10 20:44:01 We're grabbing the URL, it would make sense to grab a cananonical package name. 2017-03-10 20:44:29 Then the description wouldn't have to include the project name every time to clarify. 2017-03-10 20:46:18 pkgdesc would be printed as "$projname - $pkgdesc", which makes automagically creating descs for docs and dev packages quite simple. 2017-03-10 20:47:53 It would also allow sorting by project, which would group packages you expect to use together in the same place. 2017-03-10 21:03:40 What's the chance of alpine-baselayout getting split into -baselayout and -baselayout-host? We really only need hosts, group, protocols, profile, TZ, shells, hostname, services, shadow, passwd, and profile.d in the baselayout, as the rest may or may not bee desired. 2017-03-11 09:03:24 fabled, ncopa http://i.imgur.com/75C8u0S.png 2017-03-11 09:04:52 kaniini: sanity check: is the GUI separate from the package manager? (iow can I get apk without X?) 2017-03-11 09:05:23 yes 2017-03-11 09:05:32 it just runs apk as a child process 2017-03-11 09:05:57 ah, so "package manager" is a gui wrapper 2017-03-11 09:05:59 ok, thanks :) 2017-03-11 09:06:16 oh 2017-03-11 09:06:25 you can actually set the text of the window via CLI 2017-03-11 09:06:30 --status-str, --title-str 2017-03-11 09:06:44 i have reasons for this :D 2017-03-11 09:06:54 in the wrapper, not in apk, I guess 2017-03-11 09:07:01 yes 2017-03-11 09:07:11 for example, you can do 2017-03-11 09:07:32 apk-gtk --title-str 'Updating' --status-str 'Applying system updates' upgrade --available 2017-03-11 09:08:02 the new nifty options such as --available are passed as is to apk? 2017-03-11 09:08:14 yes 2017-03-11 09:08:27 but --available is not new :p 2017-03-11 09:23:46 Good evening. 2017-03-11 09:28:46 Slick kaniini. 2017-03-11 09:45:27 Alright, I'm too tired to continue coding -- still need to finish profile and mkinitfs surgery. 2017-03-11 09:48:44 G'night all. 2017-03-11 20:15:31 https://github.com/kaniini/apk-gtk 2017-03-11 20:17:13 sometimes, i think about working on a desktop environment repo overlay for alpine (based on xfce, with properly configured GUI-based configuration management, a selection of lightweight software and themes, etc.) 2017-03-11 22:28:43 anyone running alpine x86_64 on x86_64 machine ? Simple command like this $ qemu-system-x86_64 -cdrom alpine_x86_64/alpine-vanilla-3.5.2-x86_64.iso -m 1024 -nographic gives me no output. 2017-03-11 22:30:31 Tried -curses, boot ok then stuck at "1024 x 768 Graphic mode" 2017-03-11 22:49:07 tmh1999 : passing the noautodetect kernel flag should get you past the vid output issue. 2017-03-11 22:49:49 tmh1999 : It's running a display server by default, and isn't built with SDL support. Yeah - it bit me too :) 2017-03-11 23:03:54 TemptorSent : Thank you ! 2017-03-11 23:05:34 TemptorSent : how do you pass the kernel flag to qemu ? 2017-03-11 23:06:31 TemptorSent : do you run qemu with other options ? Something like just the kernel image and the initrd image 2017-03-11 23:06:46 thus, without the whole ISO image 2017-03-11 23:11:15 tmh1999 just boot the iso as normal, then when it comes to the boot prompt type 'grsec (or virtgrsec) noautodetect' 2017-03-11 23:12:14 well, I cannot even get to the boot prompt 2017-03-11 23:12:15 hum 2017-03-11 23:12:38 you use -nographic or -curses ? 2017-03-11 23:12:41 tmh1999 : the pxe boot loader the qemu starts 2017-03-11 23:12:44 using -curses 2017-03-11 23:12:59 using -nographic you'll need to connect using VNC or spice 2017-03-11 23:14:40 hum, it goes straight to OpenRC without a pause for me to enter any kernel argument 2017-03-11 23:16:06 I use vanilla, is it the cause ? 2017-03-11 23:16:10 tmh1999 : hold down space as it boots. 2017-03-11 23:16:27 oh, in that case, you'll type 'vanilla noautodetect' 2017-03-11 23:17:18 oh right I forgot the space stuffs ... Thanks 2017-03-11 23:25:36 TemptorSent : it returns "Loading vanilla...failed: No such file or directory" 2017-03-11 23:28:05 TemptorSent : ha it's good now :) 2017-03-11 23:29:59 Cool, good deal. 2017-03-11 23:33:47 tmh1999: are you using virtio drivers? 2017-03-11 23:34:51 TemptorSent : I am running without it first then try it afterwards. 2017-03-11 23:36:15 tmh1999: Okay, let me know how it works for you... I'm getting random 'illegal instructions' for some reason, but I haven't identified the source. 2017-03-11 23:36:36 you run other arch or x86_64 ? 2017-03-11 23:36:45 x86_64 at current 2017-03-11 23:37:01 building env to cross-build to aarch64 2017-03-11 23:37:45 I thought fabled has done cross-build for aarch64 ? 2017-03-11 23:38:23 well I am trying to make mkinitfs work with s390x 2017-03-11 23:38:42 tmh1999: Ahh, then you're in a special level of hell like me :) 2017-03-11 23:38:58 Trying to delve the depths of mkinitfs/update-kernel. 2017-03-11 23:39:29 I'm trying to get it to build minimal initrds, which I've got down now... 2017-03-11 23:39:37 The next part is getting modloop trimmed to size. 2017-03-11 23:39:44 TemptorSent : my hell-level fellow :) 2017-03-11 23:40:04 right 2017-03-11 23:40:06 Currently, I'm down to a 50mb iso... most of that is the bloody modules! 2017-03-11 23:40:29 how does aarch64 boot ? from ISO, or just kernel + initrd images ? 2017-03-11 23:40:53 aarch64 profiles I was building were uboot based. 2017-03-11 23:41:00 yeah modules take lots of spaces 2017-03-11 23:41:16 ..essentially, I was trying to get it to build the existing profiles under the new infrastruture and do it across archs. 2017-03-11 23:41:46 Right now, I'm working on trimming up the profiles a bit. 2017-03-11 23:42:14 Next trick is to make the profiles able to be built for various base-types by just changing a flag. 2017-03-11 23:43:05 that seems tough. I haven't got there ... 2017-03-11 23:43:37 I have it implemented in mkimage now actually, but mkinitfs is the bottleneck. 2017-03-11 23:44:05 I'm runnining a fairly modified version to play with minimal modules required to boot. 2017-03-11 23:45:31 tmh1999: Where I really need the help is building more granual features for mkinitfs. 2017-03-11 23:47:05 For instance, I created a scsi-base with just scsi_mod, sd_mod, sr_mod, and sg and ata-ahci with just ahci, libahci, and libata 2017-03-11 23:50:22 12.5mb of 50 is modloop, 26mb for all packages. The kernel is 3.5, and the initramfs is only 2.8! 2017-03-11 23:51:13 And that's with alpine-base, not a skeleton layout too. 2017-03-11 23:51:40 What's the status on the s390 stuff? 2017-03-11 23:51:58 I just started reading mkinitfs code ... 2017-03-11 23:52:09 I don't have much exp in this 2017-03-11 23:53:01 *lol* mkinitfs isn't exactly transparent either. 2017-03-11 23:53:17 Did you look at my mkimage tree yet? 2017-03-11 23:53:41 where's that ? 2017-03-11 23:54:12 https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts/scripts/mkimage 2017-03-11 23:54:53 Still in a state of major flux refactoring everything, but getting usable. 2017-03-11 23:55:36 Oh you are in CA ? 2017-03-11 23:56:22 Yeah, about half way between sacramento and tahoe essentially 2017-03-11 23:57:25 Up in the sierra foothills, but below the main snow-line. 2017-03-11 23:58:11 cool I am currently in CA now. About to catch a flight back to NY. Good to have someone with close timezone 2017-03-11 23:58:19 it must be tough for you 2017-03-11 23:58:27 other dev mostly in Europe 2017-03-11 23:59:26 tmh1999: Yeah, the all-nighter coding sessions aren't nearly so easy as they once were, that's for sure! 2017-03-12 00:00:30 alright, I'm going to leave the base layout as it is for now and push the beginnings of my profile-surgery. 2017-03-12 00:01:32 Sorry I can't be helpful for now. I just started picking up on building the boot image. Lots of stuffs to read. But I marked your things above. 2017-03-12 00:01:36 So, do we really need the keymaps in the base system? 2017-03-12 00:02:34 tmh1999: Yeah, the boot image building is currently about half voodo... that needs to be fixed, but I don't have time to do that this week. 2017-03-12 00:04:06 tmh1999: Right now, mkimage actually seems to do the right thing as far as its handling of the boot stuff. 2017-03-12 00:04:24 tmh1999: What boot scheme does s390 want? 2017-03-12 00:05:34 TemptorSent : typically it just need a kernel image and and initrd image. That's what I know till now. 2017-03-12 00:06:10 Okay, and how does it configure the boot? 2017-03-12 00:07:48 I am still figuring it out 2017-03-12 00:08:51 Okay, let me know what it needs and we'll see if we can't add a generator in mkimage 2017-03-12 00:19:43 tmh1999: Most recent revision pushed has profile surgery -- take a look, I think it might be straightforward to add the s390 target once you have a viable kernel. 2017-03-12 00:27:18 Thanks for the hint 2017-03-12 00:46:57 It looks like someone tagged lddtree 1.26 *YAY* Now, let's see if it works. 2017-03-12 00:50:49 Well, it doesn't break mkinitfs, so I think we're good. Thank you! 2017-03-12 03:51:27 Is anyone aware of a current means of doing a 'run-once' script on first startup currently, or shall I hack one together? 2017-03-12 03:52:51 (The preceeding question was brought to you by the department of redundancy department) 2017-03-12 05:25:40 there isnt really any support for firstrun 2017-03-12 05:35:05 kaniini: Oh well, guess I get to make something. 2017-03-12 05:35:32 Or at least a runonce that offs itself... 2017-03-12 05:35:52 I'm thinking /var/local/runonce.d 2017-03-12 05:38:29 TemptorSent: linux on s390 runs inside a VM usually fwiw 2017-03-12 05:38:45 if youve ever used Xen it is a lot like that 2017-03-12 05:39:10 kaniini: Okay, got it -- anything special needed there in terms of config file generation/driver loading/ etc? 2017-03-12 05:39:58 kaniini: Yeah, xen is giving me fits... I'm likely to go back to KVM for most applications I think. 2017-03-12 05:40:30 virtio+9p+kvm is looking pretty good right now. 2017-03-12 05:41:12 I don't need the extras of xen for most of my workloads. 2017-03-12 05:41:54 kaniini: If someone wants to give me some basics for a s390 profile, I'll add it to mkimage 2017-03-12 05:42:47 Right at the moment, I'm working on getting postgres to auto-start and auto-load a dump. 2017-03-12 05:46:47 I think I have everything in there now that I need to load a database from media on the first run. 2017-03-12 05:49:10 TemptorSent: i am not sure 2017-03-12 05:50:15 i used to have a s390 mainframe (appexpress or something like that), but i got rid of it when i moved several years ago 2017-03-12 05:50:41 you have to use another system to actually control it 2017-03-12 05:50:51 the machine that came with mine ran OS/2 2017-03-12 05:51:02 so that should give you an idea of how weird this stuff is :P 2017-03-12 05:54:11 *lol* Was it WarpDesktop? 2017-03-12 05:55:40 idk 2017-03-12 05:55:55 it had something that looked like windows 3.1 program manager 2017-03-12 05:56:03 i dont think it was a full OS/2 install 2017-03-12 06:06:49 Flashbacks. 2017-03-12 06:38:45 kaniini : Anyway, it'd be nice to support it if we can without much pain 2017-03-12 07:48:12 Hmm, replacing config files after they're installed is a PITA when you have to edit the ones provided in the APKS... 2017-03-12 08:20:00 Comments on installing self-destructing run-once hooks into /var/local/hooks/? 2017-03-12 08:21:43 The post-extraction/pre-rc hook will probably require minor changes to init, but the remainder should be doable through standard /etc/runlevel entries 2017-03-12 08:23:48 It's fairly necessary for run-from-ram setups or automated installs. 2017-03-12 08:38:56 Would hacking inittab be cleaner? 2017-03-12 09:07:22 no 2017-03-12 09:07:34 hacking inittab is never clean 2017-03-12 09:08:25 whatever you do, please make sure it's as independent from the init system and service manager as possible 2017-03-12 09:09:56 the init scripts should end with a oneshot that looks into a run-once hooks directory, runs stuff there, and deletes it if the subprocess exits 0 2017-03-12 09:10:15 skarnet: It seems like it may be in this instance, we can do a mount --bind /our/tmp/inittab /etc/inittab before pivot, and have the last thing it does before starting openrc is unmounting the bind and hupping itself. 2017-03-12 09:10:16 so you would just have to add your run-once hooks to that directory. 2017-03-12 09:10:37 what part of "never clean" don't you understand 2017-03-12 09:10:46 skarnet fair enough... 2017-03-12 09:11:49 hupping init is a nest of vipers. I wanted to do it in my own apk, to get something supervised by init instead of launched by openrc 2017-03-12 09:12:07 Hmm, it works for me... 2017-03-12 09:12:09 for every problem it solved, it created 2 2017-03-12 09:12:30 What sort of issues did you run into? 2017-03-12 09:13:09 first, obvious thing is that it won't work anymore when we replace init :P 2017-03-12 09:13:21 There is that :) 2017-03-12 09:13:30 and the day may be closer than you think 2017-03-12 09:14:02 One problem is that from the initrd init process, the sysroot doesn't get pivoted to until the exec call to init. 2017-03-12 09:14:44 So prepending a "preinit" runlevel would be an easy, and relatively clean fix. 2017-03-12 09:15:10 you want to run a script at the *end* of the init sequence, don't you? 2017-03-12 09:15:24 If there's anything in that directory, it runs. 2017-03-12 09:15:44 it sounds like you want to run stuff *before* init 2017-03-12 09:15:58 I want after apks extracted/overlays applied, but before init on the sysroot. 2017-03-12 09:15:58 please clarify exactly when you want to run your stuff 2017-03-12 09:16:03 I usually use a normal init script for oneshot tasks 2017-03-12 09:16:34 then store flag file e.g. in /var/lib to indicate completion 2017-03-12 09:17:12 Essentially, I want to apply modifications to the root fs AFTER the initrd init completes and pivots root, but BEFORE the services start with /sbin/init. 2017-03-12 09:17:14 with this approach, you can leverage the init systems dependency management 2017-03-12 09:18:05 I need a "run-me-first" essentially for it to work that way. 2017-03-12 09:18:11 TemptorSent: what prevents you from adding your stuff at the end of the initramfs's /init before the exec /sbin/init ? 2017-03-12 09:18:49 skarnet: The pivot-root takes place as the exec call to /sbin/init. 2017-03-12 09:19:01 wat 2017-03-12 09:19:39 skarnet: take a look at /usr/share/mkinitfs/initramfs-init 2017-03-12 09:21:09 yeah, so what 2017-03-12 09:21:32 skarnet it does an "exec /bin/busybox switch_root $sysroot $chart_init /sbin/init $KOPT_init_args" 2017-03-12 09:21:32 if you want to perform something between switch_root and /sbin/init, insert a script there 2017-03-12 09:22:39 True, not impossible... but it only kills one bird with the stone, and might break things worse. 2017-03-12 09:23:24 I'll have to see how the logistics for that would work out. 2017-03-12 09:23:59 the question is, do you have a rw filesystem after you switch_root 2017-03-12 09:24:13 where you can actually write your runonce script 2017-03-12 09:24:38 with a tmpfs rootfs, it's trivial 2017-03-12 09:24:38 skarnet: That's why the bind mount (or tmpfs) 2017-03-12 09:25:37 yeah, you can create a tmpfs in $sysroot/runonce, write your script there, and unmount it afterwards 2017-03-12 09:25:40 What I'm tryign to do is modify config files and add users between the extraction of the rootfs and the execution of the services. 2017-03-12 09:26:04 oh, then your rootfs is rw no matter what 2017-03-12 09:26:31 skarnet: Right -- what I was hoping to do was mount it to /etc/runleves/preinit and let inittab handle the magic. 2017-03-12 09:26:43 if you're gonna modify the rootfs on the fly anyway (which is bad but I guess it's ok for the first run) then there's no problem 2017-03-12 09:27:00 skarnet: Think run-from-ram systems. 2017-03-12 09:27:11 I'm precisely thinking *other* systems 2017-03-12 09:27:17 run-from-ram systems are easy 2017-03-12 09:27:26 skarnet: They'll ALWAYS have a fresh image needing modification. 2017-03-12 09:27:47 "let inittab handle the magic" = delegating the work to the init system, which isn't nice to me 2017-03-12 09:27:59 the clean solution is 2017-03-12 09:28:04 The problem is when you have a system that mounts root from an external source. 2017-03-12 09:28:36 ...you only want to do your config once, and once you've used lbu or whatnot, you really don't want to be overwriting things again. 2017-03-12 09:29:09 mount -t tmpfs tmpfs $sysroot/runonce; cp mypreinit /runonce/preinit ; exec /bin/busybox switch_root $sysroot %chart_init /runonce/preinit 2017-03-12 09:29:13 more to the point, let openrc do it's thing just like it normally would. 2017-03-12 09:29:24 and /runonce/preinit does its stuff and unmounts the tmpfs 2017-03-12 09:30:02 I meant cp mypreinit $sysroot/runonce/preinit obviously 2017-03-12 09:30:22 and preinit execs into /sbin/init at the end 2017-03-12 09:30:33 Right, but how do I inject the preinit in the initrd init cleanly with that approach? 2017-03-12 09:30:33 you want to insert stuff between initramfs and init, that's where you do it 2017-03-12 09:31:15 it's a script, you can have it as a heredoc document in your initramfs /init, or anywhere else in the initramfs 2017-03-12 09:31:17 What I really want a pre-sysinit runlevel I guess. 2017-03-12 09:31:40 that's exactly what I'm giving you 2017-03-12 09:31:51 a script running as process 1 in the real rootfs 2017-03-12 09:31:56 that execs into init at the end 2017-03-12 09:32:17 stop thinking in terms of runlevels 2017-03-12 09:32:26 skarnet: Understood, like I said, that kills the preinit bird.. 2017-03-12 09:32:27 think in terms of what's really happening 2017-03-12 09:33:04 skarnet: Then I need to run another set of scripts once the system is started (think database loading) 2017-03-12 09:33:29 that's a normal service you can let the service manager run at the end 2017-03-12 09:33:30 Hence the reason I was eyeballing inittab. 2017-03-12 09:33:50 consider inittab a hot poker, unsuitable for your eyeballs 2017-03-12 09:34:12 Right - does OpenRC give you a first/last dep? 2017-03-12 09:34:31 there should be a rc.local service or something like that 2017-03-12 09:34:35 if there isn't, we should add one 2017-03-12 09:35:04 run your "early" stuff before executing /sbin/init, run your "late" stuff as a normal part of the service manager 2017-03-12 09:35:25 so you'll be entirely independent from the init system *and* the service manager 2017-03-12 09:37:04 If there was a pre-sysinit runlevel, that would take care of it, and possibly be useful elsewhere. 2017-03-12 09:37:27 PLEASE stop thinking in terms of runlevels 2017-03-12 09:38:05 runlevels are semantically vague 2017-03-12 09:38:06 Okay, service manager dep then. 2017-03-12 09:38:35 "before sysinit" means "the first thing the system should run at boot time" 2017-03-12 09:38:39 In terms of how busybox's init works, it's pretty simple. 2017-03-12 09:39:11 before sysinit means the first thing the system should do after initrd has given over control 2017-03-12 09:39:21 yes 2017-03-12 09:39:55 and that's exactly what you do in a pid 1 script that execs into your real init when it's done 2017-03-12 09:39:58 Before starting system services, but from within openrc would be ideal. 2017-03-12 09:42:28 "::sysinit:/sbin/openrc initfirst" would do it. 2017-03-12 09:43:04 You're designing something at a *lower level* than the service manager and even init, and you're still thinking of terms of init and openrc 2017-03-12 09:43:15 skarnet: Part of the problem is this needs to be fairly general. 2017-03-12 09:43:25 ?? What do you mean at a lower level? 2017-03-12 09:43:59 The overlays work by adding the links for the runlevels. 2017-03-12 09:44:01 you're doing stuff in the initramfs and system installation. 2017-03-12 09:44:24 This is more bare-metal, has more privileges, than even init. 2017-03-12 09:45:25 trying to contort your stuff in order to fit in the init/openrc boxes is reversing the abstraction. 2017-03-12 09:45:28 skarnet: Think apkovl. 2017-03-12 09:45:49 skarnet: It configures a completely normal alpine system. 2017-03-12 09:46:23 The only problem with an overlay is you need the files you want to operate on to exist first. 2017-03-12 09:46:53 I'm trying to talk software design and approach to system stuff and you keep dragging me back to specifics. 2017-03-12 09:46:54 Say I want to add a set of users based on a file retrieved from the web. 2017-03-12 09:47:48 Can I even expect things like su and adduser to work before I'm in my real init under busybox? 2017-03-12 09:48:08 If you're in your real rootfs, yes you can 2017-03-12 09:48:48 What happens with the exec/pivot root then? Are my open files orphaned or ?? 2017-03-12 09:49:04 That's where it gets sticky from where I'm looking at it. 2017-03-12 09:49:06 um, don't have open files when you switch_root 2017-03-12 09:49:12 that's a given 2017-03-12 09:49:29 pivot_root wouldn't be a problem, but switch_root is more complicated 2017-03-12 09:50:14 skarnet: Then I'm going to have to figure out how to setup a context that I can get away with it. 2017-03-12 09:50:23 also, I don't want to get into specifics, but "Say I want to add a set of users based on a file retrieved from the web" elicits a strong "why on earth would I want to do that" reaction. 2017-03-12 09:50:28 As it stands, proc refuses to unmount from the sysroot. 2017-03-12 09:51:07 skarnet: Grabbing configs for an automatically provisioned system. 2017-03-12 09:51:39 skarnet: They all have unique keys, but the inidivual configurations may not be entirely determined at deployment time. 2017-03-12 09:51:53 yeah, no, you don't do that before init. 2017-03-12 09:52:12 you boot a minimal system and THEN grab a config. 2017-03-12 09:52:18 skarnet: So they grab a signed package, add the users, modify settings, etc. 2017-03-12 09:52:56 Right, the problem is I may need the users existing before booting the system in some cases. 2017-03-12 09:53:29 those cases are broken, sorry. 2017-03-12 09:54:12 a minimal system really doesn't need that many users. 2017-03-12 09:54:42 skarnet: They need to work for SSH in my cases. 2017-03-12 09:55:20 skarnet: In some cases, it's even required for them to establish ANY network connection. 2017-03-12 09:56:35 you're talking about users and ssh and config - all that's fine, but it has nothing to do with initramfs and running stuff before sysinit. 2017-03-12 09:56:41 skarnet: So unless I want to hack the ssh server as well, my best option is to add the users home directory/keys/etc to the overlay, then add the user to the system at the earliest opportunity. 2017-03-12 09:56:54 You have several classes of problems and they need to be addressed separately. 2017-03-12 09:57:18 skarnet: Um, if I need it to actually mount what I need for services, it becomes a catch-22. 2017-03-12 09:59:05 skarnet: What I have right now is a client that I'm two weeks late on getting automated images to, and the main thing holding me up is the inability to alter the config before the services I need to run. 2017-03-12 10:00:23 ... you mean you're doing this because you have a job to do for a client ? 2017-03-12 10:00:24 skarnet: Modifying inittab and placing the modified copy in the overlay will take me a few minutes, and doesn't require modifications to any other packages to work. 2017-03-12 10:00:38 skarnet: Among other things, yes. 2017-03-12 10:01:23 how about you make something work for you and your client, and you don't try to rush it into Alpine mainline before we get the time to fully think it over and properly design it? 2017-03-12 10:01:31 skarnet: I have unix-inept users that I'm distributing database servers to that need fully automated configuration. 2017-03-12 10:02:12 skarnet: I intend to, but I'd rather put my efforts towards something that works for the general case, not just this particular project. 2017-03-12 10:02:29 you can't do that when you're late. 2017-03-12 10:02:34 I also need much the same functionality for a rpi based WX station project. 2017-03-12 10:02:37 Douse your fire. 2017-03-12 10:02:57 We'll think about the general case when you don't have external pressure. 2017-03-12 10:03:03 skarnet: Yeah, most of the being late was my back going out and not being able to sit at a terminal for a couple weeks. 2017-03-12 10:04:08 skarnet: The project has been up and down on their priority list, so I don't feel entirely horrible and I'd rather have it done right than done fast and have to pick up the pieces later. 2017-03-12 10:05:27 skarnet: I've already dedicated a fair bit of time to testing other configurations and making sure I wasn't breaking things, and hopefully have more working. 2017-03-12 10:06:14 if you'd rather have it done right than done fast, why are you working on it on a Saturday night and why am I spending a good chunk of my Sunday morning discussing it when I'd rather be doing something else? 2017-03-12 10:06:25 I'm tempted to just ship a static passwd file for now, since I have that working already. 2017-03-12 10:06:42 Because this is when I can get work done :) 2017-03-12 10:07:10 Ship your static passwd file, it's a KISS solution. 2017-03-12 10:07:31 And I still need soon, just not rushed. 2017-03-12 10:08:21 working on the week-ends screams rushed. 2017-03-12 10:08:31 skarnet: Yeah, the database can get a version of the current config file to modify for the moment. 2017-03-12 10:08:57 skarnet: I don't have a regular work week at all, so I tend to sit down and work when I feel like it. 2017-03-12 10:09:31 skarnet: Insomnia lends itself to late-night coding. 2017-03-12 10:09:48 so do I, but you need to understand most Alpine people have a schedule, and you can't take them hostage for live design discussions and reviews. 2017-03-12 10:10:17 You also can't force them to keep pace with your schedule - "soon" may be too soon. 2017-03-12 10:10:21 Understood. 2017-03-12 10:11:27 I don't need anything done on that end really, I'm fine working out of my own tree -- I just want to make sure I'm actually making something useful. 2017-03-12 10:11:35 that's why I'm advising you to work on a solution for your client, and come back to discuss integration when you don't have any external pressure and it's easier for other people to sync with you. 2017-03-12 10:12:43 skarnet: The client's solution is essentially done as of this evening, excepting the configs. 2017-03-12 10:14:07 skarnet: mkimage can build me one-off images with unique ssh keys, install all the apks I need, build the overlay, and get it all packaged as a bootable iso. 2017-03-12 10:15:03 That solves my immediate need, but isn't quite fulfilling my long-term support goals of dumping a new image and walking away. 2017-03-12 10:15:50 Oh, and it does zfs nicely, which is where the data will live, while the system itself runs from ram. 2017-03-12 10:17:06 What I really want to work on once I get the client settled is getting the size of the image trimmed down by removing all but the required set of modules, esepcially for VM usage. 2017-03-12 10:18:10 The virt image is currently 50MB, and 25% is modules. 2017-03-12 10:18:39 don't mind my objections: my approach is one of minimalism (so, no zfs), having the rootfs on disk (not in ram), and I don't like the idea of initramfs in the first place (I'd do away with it entirely if I could), so we come from different places, and making Alpine work for all those cases is a challenge. ;) 2017-03-12 10:19:51 skarnet: I'm using the run-from-ram setup for this application because it can't be broken beyond simply replacing the disc. 2017-03-12 10:20:51 my pov is that a minimal read-only rootfs can't be broken either (and doesn't need an initramfs). But obviously not everyone agrees. 2017-03-12 10:21:06 skarnet: I don't use initrds on most of my own installations because I have complete controll over them and can build a tailor-fit kernel with every driver I need and none I don't. In some cases, I run no modules at all! 2017-03-12 10:21:47 yup, that's what I do too. 2017-03-12 10:21:59 For installations you don't entirely control, it's harder. 2017-03-12 10:22:17 skarnet: In ths case, the machine may not even have a SATA/SCSI hdd that I use, it may be ALL usb attached. 2017-03-12 10:23:16 ...and that's where ZFS saves my ass -- they can stick a pair of usb2+ drives in, stick a CD in, and let it go. 2017-03-12 10:23:55 ...I can zfs send the image to backup, and when they kill something, I just rebuild it from scratch and restore. 2017-03-12 10:24:27 Even if the backup happens to be some windoze share. 2017-03-12 10:26:04 skarnet: Yeah, if I wasn't a thousand miles away and dealing with people with more experience digging in dirt than in databases, it would be much easier. 2017-03-12 10:27:09 As it is, I give them an iso to burn, they stick it in the drive, boot the system (with me talking them through the bios if necessary), and walk away. 2017-03-12 10:27:54 Then they just drop reports in an incomming directory via sftp and get the result files back the same way. 2017-03-12 10:29:01 All behind their firewall without needing any outside connections hopefully - that's one of the reasons I'm trying to get it right the first time. 2017-03-12 10:29:46 I'll have provisions for establishing a remote-tunnel if needed, as a backup plan. 2017-03-12 10:30:40 But yes, real life applications, coming soon. 2017-03-12 10:32:59 Playing with code is fun and all, but I tend to go for the practical applications to make my life easier first. 2017-03-12 10:35:37 You know what may solve the problem in a cleaner and more flexible way long term? Adding the ability for APK to run hook scripts stored in the apkovl before/after applying them. 2017-03-12 10:37:18 In fact, that would solve MANY birds with one stone by keeping the apkovls split rather than one big combined overlay. 2017-03-12 10:38:51 you'd need to talk to fabled and/or kaniini about this. I don't know much about apkovls. My first reaction is a red blinking light and a buzzer honking "warning: security risk" though. 2017-03-12 10:39:05 Just put the initial stuff in the overlay loaded by the initrd, then have a service apply any overlays you want to do later. 2017-03-12 10:39:31 skarnet: How so? 2017-03-12 10:39:52 apks already have hooks that run on installation. 2017-03-12 10:40:00 apks are signed. 2017-03-12 10:40:08 Are apkovls signed too? 2017-03-12 10:40:29 No reason you couldn't refuse to run scripts if they're not signed. 2017-03-12 10:41:33 Or at least matching a pre-stored hash. 2017-03-12 10:42:30 I'm not sure if the encrypted overlays are also signed or not, haven't looked at that yet. 2017-03-12 10:45:16 The other intersting thing is that overlay mounts can have MULTIPLE lower members, so some of this may be simplifed by just adding each layer in sequence at the appropriate time. 2017-03-12 10:45:34 (for run-from-ram, mostly) 2017-03-12 10:48:25 This would let us easily build configuration deltas that could be backed up and replayed to a clean install. 2017-03-12 10:49:28 (using lbu + whatever scripting we need for state information) 2017-03-12 10:50:22 Anyway, I'm going to get some sleep with any luck, thank you for your time! 2017-03-12 10:51:05 sweet dreams! 2017-03-12 11:09:41 Yes they are also signed 2017-03-12 18:44:47 skarnet: apkovl are signed and optionally encrypted 2017-03-12 19:21:55 kaniini: What are your thoughts regarding adding the ability to add pre/post scripts to apkovls? 2017-03-12 19:23:08 And I'm still looking for a way to get abuild-sign to stop prompting me with mv -i for the index every time it gets regenerated. 2017-03-12 21:35:54 ncopa: any ideas how can finally fix %2961? I encountered yet another package affected by this. I think we should just revert the change for now… 2017-03-13 00:57:30 TemptorSent: seems reasonable 2017-03-13 01:00:36 kaniini: Okay, I'll add that to the hit list once I'me done with my current project (refactoring the functions I need of update-kernel into mkimage and scrapping the rest of the mess I don't. 2017-03-13 05:16:46 Question - what is the use case for NOT including host-keys when building an image? 2017-03-13 06:10:27 Okay, project reimplement update-kernel appears to be working so far... 2017-03-13 06:11:11 Trimming of modules included in modloop should be straightforward to implement now, leading to significant image size reductions. 2017-03-13 07:15:02 fabled / ncopa : Any ETA on having an extract function in apk? 2017-03-13 07:19:08 TemptorSent, tar 2017-03-13 07:20:30 clandmeter: Not fun. 2017-03-13 07:20:53 clandmeter: recursive get can't pipe to tar. 2017-03-13 07:21:46 clandmeter: And we don't have a flat list of packages fetched (Downloading ...) 2017-03-13 07:22:34 clandmeter: Essentially a target-tree without setting root. 2017-03-13 07:23:33 clandmeter: Otherwise, the option is to download recurseive to a new output dir, iterate files in that dir and untar them. 2017-03-13 07:25:04 I didnt follow what you are trying to accomplish. You should check out the draft of new features/fixes from fabled for new apk-tools. 2017-03-13 07:25:46 i dont think fabled will put a lot of effort in current apk-tools. 2017-03-13 07:28:07 clandmeter: Understandable.. I'll have to take a look and see what it would take. I hate hacking around things like this. 2017-03-13 07:30:00 Piping the kernel package to tar to extract /boot is painful at best. 2017-03-13 07:30:26 its how i made pkgs.a.o :) 2017-03-13 07:31:46 clandmeter: I'm already wasting literally hours per day waiting on things to extract (and redownload still in some cases :/) 2017-03-13 07:33:20 To fix it, I have to track down every apk fetch --stdout and hack them. 2017-03-13 07:33:47 And the fix is far more fragile. 2017-03-13 07:44:17 What really kills me is that apk will just naievely fetch the file again rather than checking the checksum of an existing one.. 2017-03-13 08:00:35 Anyway, the update-kernel functionality is in, working on testing it now. 2017-03-13 08:12:08 TemptorSent: setup-apkcache && apk cache download? 2017-03-13 08:12:49 tru_tru: LOL Oh, how I wish it was that simple! 2017-03-13 08:13:25 tru_tru: I'm working on the image builder, which needs at least one repo of its own per arch. 2017-03-13 08:14:49 https://wiki.alpinelinux.org/wiki/Local_APK_cache -> "Using the Local Cache with tmpfs volumes" might be ok? one file based cache per repo/arch 2017-03-13 08:14:53 tru_tru: The problem is that apk --cache-dir isn't currently doing what's expected. 2017-03-13 08:15:59 tru_tru: No dice -- the problem is that apk will fetch to stdout or outdir and not cache anything in the process. 2017-03-13 08:21:18 even if you use the --no-network flag? 2017-03-13 08:48:41 tru_tru: I need it to fetch packages from the network as needed, then cache them and NOT go to the network unless I really don't have a copy. 2017-03-13 09:17:31 hackish -> apk --no-network add wget || apk add wget, apk del wget && apk cache -v sync --purge 2017-03-13 09:17:57 otoh, ymmv :P 2017-03-13 09:34:31 tru_tru : The apk add isn't where I'm running into problems as much, it's the apk fetches that are really fubaring things. 2017-03-13 09:35:38 tru_tru : at least once I add it to the repo, it doesn't re-fetch the next time I do something (fix, say) 2017-03-13 09:36:16 tru_tru: Or worse yet, when the package happens to be in two different lists that end up getting merged at the end, but it pulls two copies in the mean time. 2017-03-13 10:08:24 could "apk policy XXX" help? -> apk policy wget| grep 'etc/apk/cache' || apk fetch wget (but download locally, not in /var/cache/apk) 2017-03-13 10:54:24 ncopa : Any idea why abuild-sign insists on using 'mv -i' when signing a repository index, totally buggering debugging by piping stdout/stderr to less :) 2017-03-13 11:05:53 do you have coreutils installed? 2017-03-13 11:06:40 there is an mv in do_sign 2017-03-13 11:06:44 but there are no -i 2017-03-13 11:09:11 TemptorSent: do you know if you have GNU coreutils mv or busybox mv? 2017-03-13 11:09:40 ncopa good question.... I'm assuming busybox, but lemme see. 2017-03-13 11:09:48 which mv 2017-03-13 11:10:19 Right, mv is /bin/mv --> busybox 2017-03-13 11:12:34 oh 2017-03-13 11:12:38 this is nasty 2017-03-13 11:13:31 https://git.busybox.net/busybox/tree/coreutils/mv.c#n109 2017-03-13 11:15:15 seems like it has been like that for a long time 2017-03-13 11:16:32 Ouch.. hmm I guess I could try blinding abuild by sticking it in a subshell with env -i and a null redirect. 2017-03-13 11:16:57 looks like it happens if target exists 2017-03-13 11:17:09 weird that we havent seen this before 2017-03-13 11:19:12 TemptorSent: i think we need to use --force in abuild-sign 2017-03-13 11:19:16 Not happy-making when it pops up in the middle of a 20 minute scratch session. 2017-03-13 11:19:37 yes, its stupid 2017-03-13 11:21:18 So, any thoughts on getting apk to directly spew out the contents of apks into a directory structure without setting up a new root? MAJOR bonous points if it lets you filter the extracted files, even more points for using the cache :) 2017-03-13 11:22:58 i assume including dependencies 2017-03-13 11:23:55 Even better :) 2017-03-13 11:24:07 TemptorSent: can you check if this solves your interactive problem? http://tpaste.us/NRmP 2017-03-13 11:24:51 to extract individual package: 2017-03-13 11:25:11 apk fetch --stdout --quiet $pkg | tar -zx 2017-03-13 11:26:16 ncopa Yeah, not a good solution... especially when the cache isn't doing it's job. 2017-03-13 11:26:34 Worse when you want /boot out of linux-grsec. 2017-03-13 11:27:25 I want a means of caching EVERY package that apk touches until I tell it to go on a cleaning spree. 2017-03-13 11:28:05 I'll test the mv -f later, right now I'm too tired to see straig. 2017-03-13 11:28:48 Running into an irritation that's crashing mkinitfs with a spurious /path/* not found 2017-03-13 11:30:03 Has something changed recently that causes globbing to fail on strings like "$var/"* vs "$var"/* works? 2017-03-13 11:30:04 you want to rename a file? use rename, not mv. 2017-03-13 11:30:14 busybox and coreutils both have rename. 2017-03-13 11:30:28 avoids problems with the mv -i alias. 2017-03-13 11:31:23 skarnet: It's in abuild-sign. 2017-03-13 11:31:25 (I normally don't suggest that because it's not posix, but here it's not about posix, it's about making things work on a system we control.) 2017-03-13 11:31:35 skarnet: rename? I’ve never heard about this command 2017-03-13 11:31:52 Anyway, I'm going to sleep, must finish debuggin when I can see. 2017-03-13 11:32:00 ...and thinking straight might help too. 2017-03-13 11:32:05 jirutka: for good reason - it's not a standard one. 2017-03-13 11:32:08 TemptorSent: gn 2017-03-13 11:32:14 G'night all. 2017-03-13 11:32:20 skarnet: aha, that’s why it’s not on macOS/FreeBSD 2017-03-13 11:32:37 indeed, it's a GNU thing that bb later implemented 2017-03-13 11:32:50 re: perl and pod2man any objection to applying this patch http://sprunge.us/FIZf ? 2017-03-13 11:32:53 ncopa: ^ 2017-03-13 11:32:56 jirutka: ^ 2017-03-13 11:33:11 skarnet: but why actually? how it’s different from mv? 2017-03-13 11:33:32 it's just mv -f without the possibility of aliasing 2017-03-13 11:33:47 nmeum: what, Valery haven’t fixed it yet?! 2017-03-13 11:33:49 and, I suppose, ensuring that it's a rename() syscall 2017-03-13 11:34:13 because mv -f will work across filesystems, and of course you can't rename() across filesystems, so in that case mv won't be atomic 2017-03-13 11:34:19 whereas rename will just fail 2017-03-13 11:34:36 nmeum: someone wrote him an email that he broke a lot of abuilds with his changes in perl pkg, I forgot about it then 2017-03-13 11:35:01 jirutka: he proposed a patch but it wasn't merged so far 2017-03-13 11:35:11 see the link in the commit description 2017-03-13 11:35:23 aha, patchwork again >_< 2017-03-13 11:35:41 his patch just moves a subset of these scripts back into the original package 2017-03-13 11:35:50 I would prefer to move all of them back for now just to be sure... 2017-03-13 11:36:45 anyways: unless you or ncopa dislike my proposed patch I would just commit it to finally fix this annoying issue 2017-03-13 11:37:06 nmeum: I agree with you 2017-03-13 11:37:19 hum 2017-03-13 11:37:33 i think the intention was to make the perl runtime package as small as possible 2017-03-13 11:37:50 nmeum: can you check the size difference? 2017-03-13 11:38:14 sure 2017-03-13 11:38:19 yes, the intention was good, but it breaks a lot of abuilds and no one was willing to find which and fix them 2017-03-13 11:38:23 ACTION is currently building perl 2017-03-13 11:38:31 i suppose the other alternative is to try fix the abuilds as they pop up 2017-03-13 11:39:05 vakartel: do you have any opinion on reverting the perl pod* thingy? ^^^ 2017-03-13 11:39:31 that's going to be pretty annoying. gdb for instance just had an empty man page as a result of this change it didn't output any errors during the build 2017-03-13 11:39:45 oh 2017-03-13 11:39:50 silent breakages 2017-03-13 11:39:52 thats bad 2017-03-13 11:39:53 yep 2017-03-13 11:40:10 I would prefer the following approach: 2017-03-13 11:40:15 1. revert the change 2017-03-13 11:40:22 2. find packages which depend on pod2man 2017-03-13 11:40:40 3. after finding all packages include perl-dev in their makedepends and apply the change again 2017-03-13 11:41:54 if packages breaks silently then i dont think we have any option but revert 2017-03-13 11:42:56 here is the size difference messured with du(1): 2017-03-13 11:42:56 8368perl-5.24.1-r1.apk 2017-03-13 11:43:15 8416 packages/main/x86_64/perl-5.24.1-r2.apk 2017-03-13 11:44:15 so this is the compressed package not the space it actually takes on disk but well…I don't believe that difference justifies the impact the change has 2017-03-13 11:46:20 http://sprunge.us/MeNJ y/n? 2017-03-13 11:46:39 also reverted changes I made to fix some packages like gdb for instance 2017-03-13 11:46:54 sry, afk 2017-03-13 11:48:34 nmeum: dpkg reduces pkgrel, either leave pkgrel untouched or increase it 2017-03-13 11:48:50 ok 2017-03-13 11:48:59 we should probably alos mention silent breakages in commit message 2017-03-13 11:49:28 which i think is the major issue 2017-03-13 11:51:31 ok, added that the commit message 2017-03-13 11:51:34 *to 2017-03-13 11:53:27 any other suggestion? otherwise I will push this now 2017-03-13 12:02:04 pushed 2017-03-13 12:03:05 i just added a commen to ml 2017-03-13 12:05:44 <^7heo> moin leute 2017-03-13 15:35:51 Have anyone used any video conferencing tools on alpine? 2017-03-13 15:36:34 I tried packaging skype and hangouts but both needs glibc. 2017-03-13 15:47:29 ncopa: the chromium 57 update is segfaulting again because of the pthread stacksize we use the same as glibc now, not sure how much is really necessary https://github.com/Duncaen/void-packages/blob/75be4272f602d26e9c7b3163d8bda6ca71f58535/srcpkgs/chromium/files/musl-patches/default-pthread-stacksize.patch 2017-03-13 16:54:56 duncaen: are you sure its due to thread stacksize? 2017-03-13 16:59:58 its not the GetDefaultThreadStackSize size, i think 2mb is still god there 2017-03-13 17:00:15 the additional changes to kShutdownDetectorThreadStackSize fixed it 2017-03-13 17:01:35 we tried a lot of builds, ~10 hours to find a working solution 2017-03-13 17:02:26 somehow our gdb stops at the wrong place and says that it received an unknown signal, not sure if this is related to chromium 2017-03-13 17:04:03 with a coredump we could see that the failing instruction moves something into a non writable area 2017-03-13 17:06:14 ncopa: http://github.com/kaniini/apk-gtk 2017-03-13 17:09:56 the crashing thread uses kShutdownDetectorThreadStackSize/PTHREAD_STACK_MIN which is just 2048 with musl, instead of the stacksize returned by GetDefaultThreadStackSize 2017-03-13 17:15:03 kaniini: nice ::) 2017-03-13 17:15:48 duncaen: good work, i will look at that when i upgrade for alpine 2017-03-13 17:17:06 i updated a few other patches, some code moved from webkit to base, this release fixes many cves 2017-03-13 17:18:25 and gold linking with gold seems to be broken again, at least on void 2017-03-13 19:06:43 Anyone know wtf the firmware 2017-03-13 19:07:03 'carl9170fw' is for? 2017-03-13 19:09:04 It has a full source tree included in the kernel-grsec package -- is that intentional, or did something break in the kernel build spewing it? 2017-03-13 19:09:18 sorry, firmware package. 2017-03-13 20:04:41 fabled/ncopa: Proposal for apk that may solve many problems when scripting: Add flags to echo the bare package-name/filename/output-file-path for fetch to allow a construct like: 2017-03-13 20:08:01 apk --list-full-path fetch -R $pkgs | xargs -N 1 tar -C "$dest" -xzf 2017-03-13 21:26:32 what would be the recommended way to store (redundand) 60TB of data? 2017-03-13 21:26:54 i am thinking of a raid10, but i am not a mass storage expert 2017-03-13 21:29:34 leo-unglaub - It largely depends on your access patterns. Personally, I'd use zfs raid-z2 or -z3 with a big fronting cache. 2017-03-13 21:30:08 leo-unglaub: But thats with my typical usage scenarious, yours may vary greatly. 2017-03-13 21:30:40 so you would let the fs do all the mirroring logic, ... ? 2017-03-13 21:31:28 leo-unglaub: Definitely -- at that point, it's the only way you can sanely handle write gaps and possible silent data corruption. 2017-03-13 21:32:18 hmmm, thats what i thought .... because the kernel internal tools for raids suck very hard at this huge level 2017-03-13 21:32:23 leo-unglaub: At 60TB, chances are you'll see drives failing more during restripe. 2017-03-13 21:32:59 i am calculating a disc failure every 4 month 2017-03-13 21:33:08 does this sound reassonable? 2017-03-13 21:33:15 leo-unglaub: It depends on what the data and access patterns look like, esp hot / warm / cold data access times. 2017-03-13 21:33:58 leo-unglaub: In my experience, you tend to have them go in batches unless you've done a good job of distributing your devices across manufacturing periods. 2017-03-13 21:34:37 very interresting point! never thought about it that way 2017-03-13 21:34:42 leo-unglaub: I'd probably try for different vendor or at least diferent batch disks in my primary/backup. 2017-03-13 21:34:59 See the 'click-of-death' 2017-03-13 21:36:15 Whatever you do, when you're looking at 60TB of data, you'll want to distribute it across multiple redundant nodes if possible. 2017-03-13 21:37:13 Your restoration times from backup are prohibitive to say the least, so you really want to be able to hot-fail, or at least warm-fail to another running machine. 2017-03-13 21:38:26 ZFS has a few inflexiblities in terms of pool upgrades, but it's definitely well tested and designed for storing large volumes of data safely. 2017-03-13 21:39:22 The next step up would be a clustering FS or object store, which may be appropriate, depending on your application. 2017-03-13 21:41:05 leo-unglaub : how do you calculate disc failure ? 2017-03-13 21:41:53 tmh1999: statistics from the last 10 years in our datacenter 2017-03-13 21:42:20 TemptorSent hmmm, good pointt 2017-03-13 21:42:35 RAID 10 requires 4x required storage, and call it 50% overhead, that's 4x120TB or 480TB disk required and can still fail with certain pairs of disks going at once. 2017-03-13 21:45:18 yeah ... hmmm 2017-03-13 21:46:18 all this storage stuff sucks ...its clearly not my field of expertise ... ususally i am the guy who works in the stuff that runs in memory ... as soon as the data is on the disc its not my problem anymore ... but that changed today ... 2017-03-13 21:46:28 RAId Z-3 requires N+3 DRIVES to provide safe 2 disk failure protection (and not so safe beyond that) -- so you can do pools of 12 10TB data + 3 10TB for parity + 1 10TB hot spare and require 160TB of drives for the same or better protection. 2017-03-13 21:47:13 The tradeoff is speed, but some tuned caching and extra memory can usually compensate unless you have very specific workloads. 2017-03-13 21:48:28 memory is not an issue ... i have 512 GB in every node 2017-03-13 21:48:29 So what I would do is run two mirrored servers with a 160TB arrays in each and load them up with memory/inifiniband with the extra money. 2017-03-13 21:49:17 leo-unglaub : Okay, you might be talking the realm of needing a high performance clustering solution... what kind of workload are you dealing with? 2017-03-13 21:50:17 If you're pushing big bandwidth as well, you're getting out of my realm of daily experience and into the HPC crowd. 2017-03-13 21:51:26 its just for emails 2017-03-13 21:51:33 maildir storge format 2017-03-13 21:53:04 leo-unglaub: Then ZFS would probably be good for your needs with some tuning, or possibly something lighter and a heavy backup program. 2017-03-13 21:54:21 Do some research on zfs, your average filesize, and your transaction rate and compare it to xfs or btrfs or even gluster. 2017-03-13 21:56:10 i will, thanks for the help 2017-03-13 21:56:12 brb 2017-03-13 22:21:29 kaniini, nice :) 2017-03-13 23:07:01 leo-unglaub: if you need to store 60 TiB of data, forget about RAID, you need some distributed FS with redundancy, like Ceph and similar 2017-03-13 23:36:29 I have an odd one here -- why would mkinitfs be puking when invoked on its own (appers to be giving lddtree bogus args, making cpio vomit), but run perfectly when run under sh -x /sbin/mkinitfs? 2017-03-14 04:55:00 Weird, how come /usr/sbin/sendmail is symbolic link to /bin/busybox and /bin/busybox is a dynamic binary, I thought /sbin stuffs are static? 2017-03-14 04:57:44 Oh, I think it's my mistake. 2017-03-14 04:58:14 It's system binary instead of static binary 2017-03-14 06:16:11 'evening fabled. 2017-03-14 06:16:41 lddtree is giving me fits again and I don't know why. 2017-03-14 06:17:14 Any recent changes that would have impacted it? 2017-03-14 06:17:50 hi 2017-03-14 06:17:52 not that i know of 2017-03-14 06:18:45 Hmm, well it had been working happily with mkinitfs, and now, not so much -- I'm trying to see if I'm triggering something or what. 2017-03-14 06:19:34 Hmm, looks like there's a bump to both scanelf and pax-utils available.. 2017-03-14 06:23:54 I *think* I may have found it, mkinitfs is passing it a null value for the target elf. 2017-03-14 06:24:45 Which probably means something in the input directory structure is fubar. 2017-03-14 06:33:03 Same error on the command line. 2017-03-14 06:49:29 https://github.com/alpinelinux/aports/pull/1019 2017-03-14 06:49:37 one has to ponder the implications of this 2017-03-14 06:50:05 audit is something that is only really effective if it is part of the base system (so that things can build-dep on it and make use of it) 2017-03-14 07:06:24 *facepalm* I don't think we really want to use the features directory of the basedir by default in mkinitfs, especially if it doesn't exist! 2017-03-14 07:37:30 features_files is returning nothing... 2017-03-14 07:46:56 Damn, that's two complete days wasted trying to figure this out so far and nothing jumping out at me still. 2017-03-14 08:08:57 I just want it to work now, I'll rewrite that part later! 2017-03-14 08:10:01 ncopa: as a docker person, do you know why docker wants us to ship support for auditd? 2017-03-14 08:10:35 ACTION is mostly concerned about it being (a) done right, and (b) not rotting 2017-03-14 08:16:14 hi 2017-03-14 08:16:40 kaniini: where they want us to ship auditd? 2017-03-14 08:18:08 <_ikke_> ncopa: https://github.com/alpinelinux/aports/pull/1019 2017-03-14 08:18:35 ah 2017-03-14 08:18:36 audit 2017-03-14 08:18:43 for selinux 2017-03-14 08:18:54 docker uses alpine userspace 2017-03-14 08:18:57 with their own kernel 2017-03-14 08:19:10 and they are looking into selinux 2017-03-14 08:20:24 so they will support this work in upstream alpine? 2017-03-14 08:20:39 yes 2017-03-14 08:20:51 well, they want the tools in upstream 2017-03-14 08:21:01 it will not work with the alpine kernel 2017-03-14 08:21:16 auditd on its own should function with the alpine kernels 2017-03-14 08:21:28 ok thats good 2017-03-14 08:21:32 i can do maintenance on that 2017-03-14 08:21:37 it wont do much of anything useful since it would need selinux or some other containment system 2017-03-14 08:21:57 i need to go now, will be afk for half day at least 2017-03-14 08:22:00 but it would log the events no problem 2017-03-14 08:22:05 okay 2017-03-14 08:22:15 well #1019 needs an initscript anyway 2017-03-14 08:22:38 we can push the binary to testing and add initscript later 2017-03-14 08:22:46 see u later 2017-03-14 08:23:15 kk 2017-03-14 08:29:07 ncopa: pushed 2017-03-14 08:29:40 i think it should go to main/, but indeed, selinux is probably not something we are likely to support in alpine proper 2017-03-14 08:29:57 apparmor makes more sense for us on linux-vanilla 2017-03-14 08:30:06 (and possibly on grsec too) 2017-03-14 08:37:23 ACTION zzz 2017-03-14 09:17:48 What would cause apk to throw a bad file descriptor on fetch? 2017-03-14 09:18:55 Ahh, never mind -- non-existent output directory. 2017-03-14 12:08:10 ls 2017-03-14 12:09:33 <^7heo> pwd 2017-03-14 12:10:49 kill -9 -1 2017-03-14 12:12:03 <^7heo> v_v 2017-03-14 12:12:07 <^7heo> savage. 2017-03-14 12:12:50 *lol* Sorry, flipping terms trying to track down the last bugglets. 2017-03-14 13:27:33 https://github.com/ers35/luastatic this allows to compile Lua program, including native extensions, into single binary, both dynamically linked with libc or completely statically linked against musl; hello world compiled against lua5.3 and dynamically linked against libc has just 220 kiB (compare it with Go… ;) ) :) 2017-03-14 13:27:45 Lua is so awesome! 2017-03-14 13:27:59 <^7heo> yeah Lua is pretty cool. 2017-03-14 13:28:05 <^7heo> You know what sucks tho? 2017-03-14 13:28:08 <^7heo> My connection. v-v 2017-03-14 13:28:10 <^7heo> v_v* 2017-03-14 13:28:12 heh 2017-03-14 13:28:38 I’m on academic network now, 1 Gbps :P 2017-03-14 13:28:40 <^7heo> (I typed all of this in one go without any printing back to my screen, sorry for the botched smiley) 2017-03-14 13:28:56 <^7heo> I'm on consumer-DSL network, 1kbs. 2017-03-14 13:29:01 <^7heo> kbps even. 2017-03-14 13:29:13 almost direct connection into NIX.CZ (main peering point in CZ) 2017-03-14 13:29:21 <^7heo> nice. 2017-03-14 13:29:54 but it has also downside, the network is monitoring, for example you can’t download any torrents 2017-03-14 13:30:05 otherwise you’ll be banned pretty soon 2017-03-14 13:30:32 but never mind, I don’t download movies at work 2017-03-14 13:30:44 <^7heo> you mean "monitored" 2017-03-14 13:30:47 s/monitoring/monitored/ 2017-03-14 13:30:50 <^7heo> yeah 2017-03-14 13:31:02 <^7heo> in .de you just don't download torrents. 2017-03-14 13:31:15 <^7heo> it's just a bad idea 2017-03-14 13:45:58 hm, abuild -K does not work in HEAD 2017-03-14 13:51:19 <^7heo> try abuild -D HEAD 2017-03-14 13:51:27 <^7heo> ACTION hides 2017-03-14 13:53:55 what? 2017-03-14 13:54:28 <^7heo> it was a stupid joke, nevermind. 2017-03-14 15:36:26 ^7heo : lol 2017-03-14 17:54:12 ^7heo: thats why you live in some place like norway where you can basically do whatever you want 2017-03-14 17:54:14 :p 2017-03-14 17:57:16 like shiver in the winter? 2017-03-14 17:57:50 you can even own a giant pickup truck with a rebel flag on it in norway and not get looked at like a weirdo 2017-03-14 17:57:59 that's more free than america 2017-03-14 17:58:10 don't get me wrong, scandinavia is great 2017-03-14 17:58:43 but what do you do in the winter when it's dark at 3pm and freezing? 2017-03-14 17:58:52 <^7heo> kaniini: :D 2017-03-14 17:59:02 <^7heo> skarnet: you drink. 2017-03-14 17:59:05 <^7heo> skarnet: all you find. 2017-03-14 17:59:10 <^7heo> skarnet: and eat lots of shit 2017-03-14 17:59:10 skarnet: you go to barbados for the entire winter 2017-03-14 17:59:26 c/o norway government 2017-03-14 17:59:33 michael moore told me this so it must be true 2017-03-14 17:59:53 <^7heo> yeah a LOT of people here go to the south for the whole winter. 2017-03-14 18:00:00 <^7heo> it's barely okay to stay here inside. 2017-03-14 18:00:01 oh wait, you guys probably dont know who michael moore is 2017-03-14 18:00:08 ... 2017-03-14 18:00:11 it's Europe here 2017-03-14 18:00:15 he is a political commentator who makes movies 2017-03-14 18:00:22 where he basically BSes half of it 2017-03-14 18:00:23 we actually know who American people are 2017-03-14 18:00:48 you're thinking the other way around: Americans don't know shit outside of their own country 2017-03-14 18:01:03 that largely is true 2017-03-14 18:01:18 also norway prisons look pretty posh 2017-03-14 18:01:30 even the high security one looks way nicer than any i've seen on TV in america 2017-03-14 18:01:31 I don't intend on visiting one 2017-03-14 18:02:04 yeah, the whole you cant leave aspect does seem problematic 2017-03-14 18:08:05 skarnet, ^7heo so what is your opinion on europe's main cultural export to america: DJ Bobo? 2017-03-14 18:08:32 ok ok, of the 1990s anyway 2017-03-14 18:10:02 if he really is the main cultural export, then we're in trouble 2017-03-14 18:10:35 /o/ 2017-03-14 18:10:40 \o\ 2017-03-14 18:11:22 anyway he's Swiss, that's not Europe 2017-03-14 18:11:48 kaniini: when you merge some pull request, you can modify the patches, like fixing whitespaces or squashing, github-pr-closer is able to detect even modified PRs 2017-03-14 18:13:02 jirutka: that was more a test of how committed docker is to seeing this go in 2017-03-14 18:13:38 jirutka: as in, "hey there's an initscript that is needed, but while youre at it, please fix the indentation" :p 2017-03-14 18:13:56 aha 2017-03-14 18:14:45 and there i added init.d script 2017-03-14 18:15:00 :P 2017-03-14 18:15:20 audit was already on my todo list anyway, because i want to integrate apparmor 2017-03-14 18:15:35 i just copied the gentoo init.d script 2017-03-14 18:15:42 should be fine 2017-03-14 18:15:50 btw I’d like to add check for whitespaces into tests run on Travis, b/c I’m quite bored of telling that people over and over again… programmers should be intelligent people and still they can’t handle such stupid thing like tabs and spaces :( 2017-03-14 18:15:50 maybe i could ask vakartel to clean it up 2017-03-14 18:16:00 jirutka: good idea 2017-03-14 18:16:12 ncopa: why apparmor? 2017-03-14 18:16:14 and we could make some basic whitespace tests in abuild sanitycheck too 2017-03-14 18:16:21 isn't vakartel 2017-03-14 18:16:21 apparmor? 2017-03-14 18:16:32 the guy who made that sshd init script you hate 2017-03-14 18:16:34 lol 2017-03-14 18:16:40 yeah 2017-03-14 18:16:42 it is 2017-03-14 18:16:47 well 2017-03-14 18:17:11 i think much of his stuff is good 2017-03-14 18:17:18 well… 2017-03-14 18:17:19 probably most of it 2017-03-14 18:17:24 I’m not so sure 2017-03-14 18:17:29 the problem is big risky changes 2017-03-14 18:17:30 that initscript seems bad reading through the PR 2017-03-14 18:17:43 and break stuff without good reason 2017-03-14 18:17:49 and break stuff way too often 2017-03-14 18:17:53 he doesn’t know consequences of some of his actions 2017-03-14 18:17:53 i would be irritated if i restarted sshd and got kicked out of my box 2017-03-14 18:18:16 and that type of irritation would be something that would probably cause many people to reconsider using alpine 2017-03-14 18:18:24 exactly 2017-03-14 18:18:27 "they cant even make their SSH init work right" 2017-03-14 18:18:43 ssh is special in that regard 2017-03-14 18:18:53 needs to take extra care there 2017-03-14 18:19:03 i didnt look too close to the package split parts 2017-03-14 18:19:17 i think the 3 commits i did from his work should be more readable 2017-03-14 18:19:19 and that’s why ncopa reviewed it thoroughly, isn’t it? sshd must work well 2017-03-14 18:19:39 jirutka: as for why apparmor -- it provides similar hardening as grsecurity RBAC, and could be used on both grsec and vanilla kernels for that purpose 2017-03-14 18:19:56 jirutka, kaniini: it would be great if you could review the 3 rebase commits 2017-03-14 18:20:11 i dont want push without atleast one more (skilled) reviews it 2017-03-14 18:20:17 kaniini: why not just use SELinux then? 2017-03-14 18:20:37 i think apparmor is simpler 2017-03-14 18:20:40 jirutka: apparmor is a lot easier to use and fits alpine philosophy way better than selinux :p 2017-03-14 18:20:41 selinux is complicated 2017-03-14 18:21:07 isn’t AppArmor Canonical project? 2017-03-14 18:21:11 no 2017-03-14 18:22:08 it comes from immunix (which was a debian distribution which had similar goals to alpine) 2017-03-14 18:22:27 the apparmor team is mostly at suse now as i understand it 2017-03-14 18:22:27 aha 2017-03-14 18:22:54 i thought i filed a bug for apparmor and musl 2017-03-14 18:23:06 it fails on scandirat 2017-03-14 18:24:16 selinux makes more sense for docker than apparmor though 2017-03-14 18:24:22 because you can label the containers 2017-03-14 18:24:26 for further confinement 2017-03-14 18:25:58 the launchpad.net bugtracker is worse than our redmine... 2017-03-14 18:26:03 i cannot find the bug i filed 2017-03-14 18:26:19 launchpad.net is horrible 2017-03-14 18:26:22 https://bugs.launchpad.net/apparmor/+bug/1671857 2017-03-14 18:26:24 not just as bug tracker 2017-03-14 18:26:32 it feels sluggish too 2017-03-14 18:26:35 yes pretty much everything about launchpad is crap 2017-03-14 18:26:48 kaniini: ^ +1 2017-03-14 18:27:27 apparmor$ tpaste < APKBUILD 2017-03-14 18:27:27 http://tpaste.us/7BgB 2017-03-14 18:27:37 apparmor$ tpaste < musl-fixes.patch 2017-03-14 18:27:38 http://tpaste.us/1vJ0 2017-03-14 18:27:45 in case someone want to continue on it 2017-03-14 18:28:15 i will just pay dalias to implement scandirat() as i did rtld_lazy ;) 2017-03-14 18:28:32 you paid him for that 2017-03-14 18:28:37 yes 2017-03-14 18:28:39 thats awesome 2017-03-14 18:29:00 thank you for doing so 2017-03-14 18:29:05 kaniini: btw how it goes with vanilla kernel? I’d like to have linux-virtvanilla 2017-03-14 18:29:52 jirutka: i am hoping to get a virt profile for vanilla into 3.6 :) 2017-03-14 18:29:54 but 2017-03-14 18:30:04 the way we do kernels is really, i don't know 2017-03-14 18:30:09 we need to clean it up somehow 2017-03-14 18:30:13 agree 2017-03-14 18:30:18 the current state is total mess :/ 2017-03-14 18:30:23 talked with fabled about kernel config the other day 2017-03-14 18:30:34 managing kernel configs is not that easy 2017-03-14 18:30:40 and yes its a mess 2017-03-14 18:30:58 debian uses one source package for all of it 2017-03-14 18:31:03 maybe this could help https://github.com/crossdistro/kernel-tools (by pavlix) 2017-03-14 18:31:06 which may be how we ultimately have to do it 2017-03-14 18:31:08 i dont know 2017-03-14 18:31:08 but I haven’t looked into it yet 2017-03-14 18:31:33 i think gentoo also has a source package 2017-03-14 18:31:38 at least the used to 2017-03-14 18:31:50 uff, linux-vanilla 8.5 MiB… I’d probably just build my own kernel 2017-03-14 18:32:02 grsec is same size no? 2017-03-14 18:32:05 :-o 2017-03-14 18:32:11 8.5MB! 2017-03-14 18:32:13 thats huge 2017-03-14 18:32:23 kaniini: no, b/c I use virtgrsec 2017-03-14 18:32:36 and I’m not even talking about dozens of useless modules… 2017-03-14 18:32:40 how large is virtgrsec? 2017-03-14 18:32:55 3.8 MiB 2017-03-14 18:33:06 i am pretty sure we can shrink vanilla down like that 2017-03-14 18:35:37 btw, i wonder if someone has experience with dnssec-root 2017-03-14 18:35:44 the root-anchors.xml thingy 2017-03-14 18:35:54 kernel-3.7.5-hardened-quest-v1 … 3.7 MiB – that’s my custom kernel I used in VMs, no modules 2017-03-14 18:36:20 http://tpaste.us/pQrW 2017-03-14 18:36:26 icann publised new .xml 2017-03-14 18:36:56 the cmp root-anchors.txt check we have in the APKBUILD fails now 2017-03-14 18:37:02 my ultimate goal is to replace linux-grsec with linux-hardened that is just upstream PaX + AppArmor, while linux-vanilla is without PaX 2017-03-14 18:37:18 this solves 'the grsec problem' 2017-03-14 18:37:31 can you get the PaX patches separately? 2017-03-14 18:37:39 yes, he is releasing separately again 2017-03-14 18:37:43 aha 2017-03-14 18:38:05 i saw the vmware mini distro uses it 2017-03-14 18:38:07 * Starting networking ... 2017-03-14 18:38:07 * br0 ... 2017-03-14 18:38:07 ip: RTNETLINK answers: File exists [ !! ] 2017-03-14 18:38:07 * ERROR: net.br0 failed to start 2017-03-14 18:38:11 photon or what they call it 2017-03-14 18:38:13 what the heck is wrong with that :/ 2017-03-14 18:38:40 it creates bridge, I can see it in brctl show, but still it exists with this 2017-03-14 18:39:12 jirutka: try --debug 2017-03-14 18:39:20 file exists usually means there is a duplicate route or ip 2017-03-14 18:39:56 that’s not the case 2017-03-14 18:40:35 aah, I’m stupid 2017-03-14 18:42:15 obviously, I should not set gateway when creating a standalone bridge… 2017-03-14 18:43:35 kaniini: where are the pax patches released? 2017-03-14 18:43:55 ncopa: https://www.grsecurity.net/~paxguy1/pax-linux-4.9.13-test6.patch 2017-03-14 18:44:59 right 2017-03-14 18:45:10 they just dont publish the stable pax patch 2017-03-14 18:45:46 seems like there are fragments of it here: https://github.com/vmware/photon/tree/master/SPECS/linux 2017-03-14 18:46:38 from what i have been told, it is basically the same as what we publish now 2017-03-14 18:47:16 they take the free patch and rebase? 2017-03-14 18:47:23 yes 2017-03-14 18:47:34 ok 2017-03-14 18:49:11 what i'd like is feature to disable pax protections per container 2017-03-14 18:51:09 yes, that would be nice for docker hosts and lxc/lxd hosts 2017-03-14 18:51:30 looks like adding scandirat(3) to musl is pretty easy 2017-03-14 18:51:39 ah nice 2017-03-14 18:52:32 i was thinking adding #ifndef HAVE_SCANDIRAT 2017-03-14 18:52:59 and inthere have a copy of musl scandir with the "at" fd 2017-03-14 18:53:20 that is another way we could go 2017-03-14 18:53:25 dalias said it might make sense to add scandirat to musl 2017-03-14 18:54:11 what exactly requires scandirat? 2017-03-14 18:54:20 apparmor 2017-03-14 18:54:46 sounds like a good reason not to use it 2017-03-14 19:24:44 skarnet: while i do not personally like scandir()/scandirat(), apparmor + pax is a viable replacement to what we currently ship as grsec 2017-03-14 19:49:31 ncopa: please look into #alpine-commits, linux-grsec failed on armhf 2017-03-14 19:50:27 looks like config needs updating 2017-03-14 20:07:45 Speaking of kernel configs, I think the ability to roll vm-host-specific kernels would yield the smallest containers reasonably possible, especially if the modules could be avoided entirely for base configs. Specifically I'm working on a virtio-only target with no other drivers that aren't necessary. 2017-03-14 20:42:12 ncopa: Gentoo has a source-only package but I have a full Gentoo package building a binary kernel and initramfs. 2017-03-14 20:42:40 ncopa: Plus I was somewhat involved in Fedora DNSSEC initiatives. 2017-03-14 21:13:15 kaniini: ^ 2017-03-14 21:16:44 TemptorSent: you mean VMs, not containers…? b/c you don’t need (and actually can’t have) own kernel inside a container 2017-03-14 21:21:19 jirutka: Sorry, you're correct -- a VM PROVIDING containers to be correct. 2017-03-14 21:22:30 jirutka: LXC on KVM being the concept. 2017-03-14 21:23:54 jirutka: So a single VM may contain several interrelated contanerized apps, but be isolated from any other VM 2017-03-14 21:25:27 TemptorSent: yeah, I have this setup on vpsFree :P 2017-03-14 21:25:42 TemptorSent: LXC containers inside QEMU/KVM machine that runs inside OpenVZ container 2017-03-14 21:25:53 because, you know, we must to go deeper :P 2017-03-14 21:29:35 jirutka: For me, the interesting thing is the ability to migrate VMs while keeping the actual containerization environment as light as possible. 2017-03-14 21:30:23 jirutka: Plus the ability to have control over the kernel, which is sometimes necessary. 2017-03-14 21:38:35 so i preordered one of these https://www.solid-run.com/marvell-armada-family/armada-8040-community-board/ hoping it should be fairly easy to get alpine linux running on it 2017-03-14 21:43:01 or does the aarch64 stuff already support the uefi arm boot stuff already and I just never noticed? 2017-03-14 21:46:45 mitchty: I'm not sure on the arm uefi stuff, it should be straightforward to add if the standard config doesn't work. 2017-03-14 21:47:36 TemptorSent: cool, it looks like the new arm server type boards have a pretty standard setup though, so supporting this should mean the others work from what i've read 2017-03-14 21:49:39 but 16GiB of ram ecc even on an arm 64 board would be ideal 2017-03-14 21:51:53 mitchty: Let me know when you've got HW to test on. 2017-03-14 22:08:51 jirutka: Please take a look at the latest rev on my PR - it integrates the update-kernel functionality and will make it easy to trim the modloop. 2017-03-14 22:09:07 what PR? 2017-03-14 22:18:10 https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts/scripts/mkimage 2017-03-14 22:22:03 jirutka: The one based on that ^^^ 2017-03-14 23:57:21 pavlix: what are your thoughts on cleaning up kernel APKBUILDs at alpine? 2017-03-15 02:44:35 note: apk is throwing a spurious warning "Ignoring /home/chrisgio/packages/testing/aarch64/APKINDEX.tar.gz: No such file or directory" for an arch I haven't built packages on, but I'm pulling apks for. 2017-03-15 02:45:02 This is breaking things rather badly that rely on piping the output from apk. 2017-03-15 02:46:11 'apk fetch $pkg --stdout | tar -xzC $dir' for instance 2017-03-15 02:49:37 TemptorSent : I guess it happens for the first built package only. After one package is built, it's gone. 2017-03-15 02:51:29 tmh1999: Too bad I'm not setup for actually BUILDING on aarch64 2017-03-15 02:52:53 tmh1999: I built x86_64 apks, and now I'm pulling aarch64 packages for an image - no building, no aarch64 directory in ~/packages. 2017-03-15 02:54:03 tmh1999: So now I'm dead in the water testing the aarch64 building until I can figure out why APK is throwing a fit (and --quiet doesn't help) 2017-03-15 02:55:03 I see 2017-03-15 02:55:06 Hum.. 2017-03-15 02:55:24 tmh1999: Give it a shot -- maybe it's something local to me... build a package for your native arch, then try to fetch one for a foreign arch. 2017-03-15 02:55:26 Is the message on stderr or stdout ? 2017-03-15 02:56:03 It's being handled as if it's stdout, but I haven't tried splitting hairs with it yet. 2017-03-15 02:57:06 ...needless to say having to do a redirected stderr while consuming stdout in every location apk fetch --stdout is used is not going to be pretty. 2017-03-15 02:57:14 $ apk fetch --arch aarch64 musl downloads the musl package for me to $PWD 2017-03-15 02:57:52 Do you have a local packages repo as that user for your native arch? 2017-03-15 02:58:40 and none built for aarch64? 2017-03-15 02:58:58 hold on it downloads the native package not aarch64 2017-03-15 02:59:17 Yeah, you need to pass the --arch and probably a different root 2017-03-15 03:00:52 tmh1999: If you can, please test against https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts/scripts/mkimage 2017-03-15 03:01:22 using profile uboot and arch aarch64 2017-03-15 03:03:10 I'm trying to shoot down any last big-ugly showstopper bugs in mkimage and keep hitting pain-points in other tools (apk, lddtree, mkinitfs, etc -- I replaced update-kernel entirely) 2017-03-15 03:04:44 I am not sure you are able to fetch packages from different arch. 2017-03-15 03:05:17 Before I built packages and setup my own local repo, it worked FINE! 2017-03-15 03:06:05 The warning is new, spurious IMHO (the entire arch directory doesn't exist, a non-existent APKINDEX is to be expected!), and breaks things that worked before. 2017-03-15 03:06:36 The only change was the addition of a repository that only has my native arch, not foreign archs that I want to pull. 2017-03-15 03:08:03 do 'mkdir /tmp/apktst && apk add --arch aarch64 --root /tmp/apktst --initdb -R alpine-base' 2017-03-15 03:10:06 it doesn't work for me :( 2017-03-15 03:10:26 alpine-base missing required by world[alpine-base] 2017-03-15 03:10:49 option -R doesn't even exist 2017-03-15 03:12:14 oops, sory - the -R is on my fetch 2017-03-15 03:12:51 run the --initdb, they try a fetch command 2017-03-15 03:13:45 do 'mkdir /tmp/apktst && apk add --arch aarch64 --root /tmp/apktst --initdb && apk fetch --arch aarch64 --root /tmp/apktt -R alpine-base' 2017-03-15 03:15:06 Right now, I'm going to have to punt and let someone else test it I guess. 2017-03-15 03:15:39 I need to finish cleaning up a couple other issues I need sooner. 2017-03-15 04:25:20 Working one it :) 2017-03-15 04:25:32 s/one/on/ 2017-03-15 05:08:56 o/ ello. Trying to build libstrophe on Alpine, found https://raw.githubusercontent.com/faradayio/aports/master/testing/libstrophe/APKBUILD but it uses an old version; decided to build from scratch as described on https://github.com/strophe/libstrophe but getting an error running ./configure "configure: error: C compiler cannot create executables" 2017-03-15 05:10:02 Most likely it is related to autoreconf putting wrong arguments for gcc into ./configure (e.g. it uses flags -qversion and -V instead of --version and -v, ergo the error) 2017-03-15 05:22:19 First thought I solved it by changing "--version -v -V -qversion" to "--version -v" but then realized that I was missing the libc-dev apk on my system; somehow now it ran ./configure with no errors, the thing got compiled just fine 2017-03-15 05:23:51 Could somebody please create a request for that apk to be updated for libstrophe version 0.9.1 and brought back into the repos? It's a good library and is available on most distros. I just don't have the account on the bugtracker, a github person. 2017-03-15 05:24:01 Sorry for the bother, have a good one! 2017-03-15 07:41:38 kaniini: I don't have any details, I haven't actually seen the kernel builds at all. But I would help if kernel-tools proved useful to the project. 2017-03-15 07:45:57 pavlix: can kernel-tools work across kernel versions? for example, raspberry pi kernel is different version frequently than what we normally ship (but we want to keep the base features in sync) 2017-03-15 08:54:18 kaniini: it is a young project and it can basically do whatever you wish; in other word, now it's the best time to define requirements for the project 2017-03-15 08:55:05 kaniini: currently it's very basic and it serves me as a Gentoo user to reliably and reproducibly build a kernel for my system and actually *check* the configuration 2017-03-15 08:56:23 kaniini: I had some multiversion consideration but as long as it is just a personal project, I'm going to address them *as they come*, but if there are specific requirements in the issues, I can answer them and address them 2017-03-15 10:11:48 mozilla further selfsabotaging firefox: https://bugzil.la/1345661 - can we disable this in alpine? 2017-03-15 10:13:18 <^7heo> we have to 2017-03-15 10:13:55 <^7heo> no question 2017-03-15 10:14:01 <^7heo> thanks for reportinv 2017-03-15 10:14:04 <^7heo> reporting 2017-03-15 10:15:31 wut, I didn't know that Firefox depends on PA 2017-03-15 10:16:36 it's the new default 2017-03-15 10:16:43 fortunately it can still be worked around 2017-03-15 10:16:49 <^7heo> Yeah 2017-03-15 10:17:14 <^7heo> but you know what can not be worked around? 2017-03-15 10:17:31 <^7heo> firefox's legal requirements 2017-03-15 10:17:41 I'm curious how long it could be worked around 2017-03-15 10:17:45 <^7heo> we should change the name at the very lesadt 2017-03-15 10:17:54 <^7heo> least* 2017-03-15 10:18:21 <^7heo> scadu: let's see. 2017-03-15 10:18:35 let's change to Snow bear! 2017-03-15 10:18:39 Firefox 54 deletes ALSA completely, no? 2017-03-15 10:19:02 oops. 2017-03-15 10:19:58 https://bugzilla.mozilla.org/show_bug.cgi?id=1345661#c64 2017-03-15 10:20:10 distros pathcing alsa back 2017-03-15 10:20:23 <^7heo> yeah 2017-03-15 10:27:04 <^7heo> after firefox and iceweasel, what about badger? 2017-03-15 10:43:35 for anyone not following #alpine-linux http://tpaste.us/rqax 2017-03-15 10:47:31 clandmeter: this had been a long time coming 2017-03-15 10:47:52 mm, nice. so we are going into apparmor now? I didn't check yesterday's​ backlog from -devel 2017-03-15 10:48:02 does it make me a bad person if I find that hilarious? 2017-03-15 10:48:08 well, at least we will have lazy loading... 2017-03-15 10:48:18 this grsecurity thing going on is bad :-( 2017-03-15 10:48:33 they are making all grsecurity patches private 2017-03-15 10:50:42 ncopa, it is and it has been since some time... 2017-03-15 10:50:58 good thing is, you will have to spend less time on it. 2017-03-15 10:52:07 join me and dance on grsec's corpse 2017-03-15 10:52:26 xD 2017-03-15 10:53:06 skarnet: let's dance then 2017-03-15 10:55:30 \o\ /o/ \o/ \o_ _o/ -o- _/o\_ *^^* 2017-03-15 14:52:19 i am curious to find out why this person on the ML wants to find out the alpine linux licence ... 2017-03-15 14:52:43 this person is working for Raytheon, they build bombs and all that war shit that kills people .. 2017-03-15 14:53:07 i would asume they want to use alpine linux on those bombs in the future ... 2017-03-15 14:58:44 that's the price of success: people get interested in you, even the ones you'd rather they didn't. 2017-03-15 14:59:29 that's also the price of freedom - if software is to be used by everyone, then "everyone" includes arms manufacturers. 2017-03-15 15:20:11 leo-unglaub: hm, that’s the risk… we may specify in license Alpine must not be used for evil purpose, but then it would not be truly free software anymore 2017-03-15 15:20:42 you guys are looking at it the wrong way 2017-03-15 15:20:47 they are using free software 2017-03-15 15:20:58 to drop free (as in freedom) bombs on "the terrorists" 2017-03-15 15:21:00 just buy their hardware to request the source 2017-03-15 15:21:07 to bomb ISIS with terrorists 2017-03-15 15:21:10 er 2017-03-15 15:21:12 with freedom 2017-03-15 15:21:17 leo-unglaub: and also it’s quite a problem how to define “evil”… well, we may directly specify “military”, but that’s not black and white 2017-03-15 15:21:17 <^7heo> whatever works. 2017-03-15 15:21:19 Freedom! 2017-03-15 15:21:41 or socalled cyber 'weapons' 2017-03-15 15:24:09 i just hope they dont use gpsd 2017-03-15 15:25:27 well, the answer for raytheon is the same answer as everyone else 2017-03-15 15:25:38 the software itself is dependent on the license of the software itself 2017-03-15 15:27:11 "Open-source software enables low-cost maintenance, stability, and most importantly, continuous improvement to forecast accuracy and timeliness" 2017-03-15 15:27:15 I’m afraid that some programs may recompile binaries with different options when you run `make check` and some maintainers may not notice it 2017-03-15 15:27:46 yay optimizing for the test bench 2017-03-15 15:28:07 ./configure --volkswagen 2017-03-15 15:28:10 I’m always checking what the check task do, but I don’t expect that everyone do it 2017-03-15 15:29:08 do the check after the install target 2017-03-15 15:31:57 yeah, that’s one of the possible solutions how to prevent it 2017-03-15 15:34:58 do you have many aports that use check? 2017-03-15 15:36:29 not yet 2017-03-15 15:38:22 clandmeter: I think that you’ll welcome this change :) https://github.com/lxc/lxc/commit/72ead1c05401789ced73d9e3f47d1c11aa6dd951 (I’ve already backported it to main/lxc@edge) 2017-03-15 15:38:41 ah ok, we have a open task for this in void too, but we are still not sure how we handle it in case it needs more dependencies etc, and if we enable them by default 2017-03-15 15:39:09 duncaen: you’re Void developer? :) 2017-03-15 15:39:39 yes 2017-03-15 15:39:47 duncaen: :) 2017-03-15 15:41:47 duncaen: how do you handle the problem when more packages provides same executables and you want to be able to install them in parallel and switch between them, like various versions of JVM, py2 or py3 variant of some tool etc. Debian, Fedora and some others use update-alternatives for this, Gentoo eselect 2017-03-15 15:42:08 we have xbps-alternatives which creates symlinks 2017-03-15 15:42:26 duncaen: is this written from scratch or based on update-alternatives? 2017-03-15 15:42:33 from scratch afaik 2017-03-15 15:42:34 http://sprunge.us/VHVR 2017-03-15 15:42:54 there are packages, and alternative groups which can be in multiple packages 2017-03-15 15:43:03 great, where can I find source code of this tool? 2017-03-15 15:43:21 https://github.com/voidlinux/xbps/tree/master/bin/xbps-alternatives 2017-03-15 15:43:22 jirutka: apk-tools 3 will have alternatives 2017-03-15 15:43:36 and https://github.com/voidlinux/xbps/blob/master/lib/package_alternatives.c 2017-03-15 15:43:41 duncaen: thanks! 2017-03-15 15:43:53 there are some things that are not so nice 2017-03-15 15:44:07 e.g. creating a target directory to avoid dangling symlinks 2017-03-15 15:44:19 kaniini: I’m afraid that we can’t wait until apk-tool 3 for this… and also I don’t think that it should be integrated directly into apk…? 2017-03-15 15:44:51 dont look at me, it is what fabled plans to do :P 2017-03-15 15:45:18 kaniini: ok, I’ll discuss it with him :) 2017-03-15 15:45:45 we shoot ourself by allowing to choose /bin/sh, when readline was updated users with bash as /bin/sh broke the install scripts at installation time :s 2017-03-15 15:46:08 kaniini: maybe I can write a prototype in Lua, try how it works, and he can eventually rewrite it in C as part of apk-tools 2017-03-15 15:46:38 yes 2017-03-15 17:29:31 anyone has a copy of lua-xml package in your build machine ? looks like the host providing the package is down 2017-03-15 17:29:55 LuaXML_130610.zip is the package name 2017-03-15 17:41:42 nvm archive.org is awesome 2017-03-15 18:03:37 tmh1999: you could always use different mirror. see http://rsync.alpinelinux.org/alpine/MIRRORS.txt 2017-03-15 18:06:31 scadu : I mean the source package, not .apk 2017-03-15 18:53:09 looks like perl-test2-suite has a new version, and require new perl-sub-info and perl-term-table. 2017-03-16 04:12:35 Found bug in 'mount' that may be responsible for odd behavior at boot. 2017-03-16 04:13:38 /dev/sda2 is mounted on /mnt (ext4), then /mnt/tmp bind mounted over /tmp, which is working as expected. 2017-03-16 04:15:31 however both invoking 'mount' and checking /proc/mounts show '/dev/sda2 /mnt ext4 rw,relatime,data=ordered 0 0' '/dev/sda2 /tmp ext4 rw,relatime,data=ordered 0 0', and no indcation of the bind source directory. 2017-03-16 04:16:41 I suspect this is what's breaking the unmounting of the bind-mounted /proc in the initramfs. Any thoughts on to where this behavior is coming from? I suspect it is recent considering the lack or breakage reported. 2017-03-16 04:17:30 libc musl problem perhaps? 2017-03-16 04:18:39 Or is it a kernel issue on the revision I happen to be running (4.9.10) 2017-03-16 06:27:45 ncopa: i should have PaX roughly 40% split out by this weekend 2017-03-16 06:30:13 *me 2017-03-16 06:30:22 that's amazing 2017-03-16 06:34:46 as an aside, i am also auditing the PaX source as i split it out 2017-03-16 06:35:54 that audit has lead me to believe that PaX only really works correctly on x86/arm -- my ventures outside of that have observed logic errors in the implementation in functions like copy_to_user() 2017-03-16 06:36:05 lols 2017-03-16 07:26:25 I am wondering normally how do we do the testing procedures ? 2017-03-16 08:25:58 interesting 2017-03-16 08:26:02 bwrap --version 2017-03-16 08:26:09 on x86_64 in lxc container works 2017-03-16 08:26:30 on aarch64 i get the error: 2017-03-16 08:26:30 bwrap --version 2017-03-16 08:26:31 Dropping capability 0 from bounds: Operation not permitted 2017-03-16 08:26:39 with sudo it works 2017-03-16 08:28:49 running it in strace also works 2017-03-16 08:30:07 <_ikke_> heisenbug 2017-03-16 08:30:13 ah 2017-03-16 08:30:21 its suid root 2017-03-16 08:30:27 -rwsr-xr-x 1 root root 51280 Mar 14 14:50 /usr/bin/bwrap 2017-03-16 08:31:59 !? 2017-03-16 08:32:32 $ ls -l $(which bwrap) 2017-03-16 08:32:32 -rwsr-xr-x 1 root root 38616 Mar 14 18:18 /usr/bin/bwrap 2017-03-16 08:32:47 $ ls -l ./bwrap 2017-03-16 08:32:47 -rwxr-xr-x 1 ncopa ncopa 50656 Mar 16 08:29 ./bwrap 2017-03-16 08:33:02 $ ./bwrap --version 2017-03-16 08:33:02 bubblewrap 0.1.7 2017-03-16 08:46:54 its missing CAP_SETPCAP 2017-03-16 11:55:55 fabled : I forgot to add YAMA to linux-vanilla patch ;) 2017-03-16 12:05:13 <^7heo> do we have a script that spawns a copy of the alpine base, and runs a chroot on it? 2017-03-16 12:20:35 ^7heo: docker run --rm -it alpine ... :-p 2017-03-16 12:54:38 <^7heo> ncopa: without having 10 metric tons of bloatware to do weird stuff I don't need :P 2017-03-16 12:55:18 <^7heo> I thought that maybe jirutka had done something like that. 2017-03-16 12:58:24 i think he had 2017-03-16 12:58:42 <^7heo> Right? 2017-03-16 13:03:08 ^7heo: this? https://github.com/jirutka/alpine-chroot-install/ 2017-03-16 13:03:12 <^7heo> maybe. 2017-03-16 13:03:21 <^7heo> Lemme have a quick look. 2017-03-16 13:04:12 <^7heo> I think that is it yes. 2017-03-16 13:04:16 <^7heo> I'll try it today. 2017-03-16 13:04:47 <^7heo> jirutka: I have to give you kudos for that. 2017-03-16 13:04:56 <^7heo> The readme is excellent. 2017-03-16 13:05:08 thanks :) 2017-03-16 13:05:35 <^7heo> The documentation is a bit verbose in the source. 2017-03-16 13:05:42 i should move it under alpinelinux org 2017-03-16 13:05:50 <^7heo> Definitely. 2017-03-16 13:05:55 <^7heo> That is a very interesting tool to have. 2017-03-16 13:06:17 <^7heo> the only advantage docker would have over your script is that docker does network. 2017-03-16 13:06:25 <^7heo> but your script has several advantages over docker :P 2017-03-16 13:07:32 <^7heo> I don't know the syntax `: ${...}` 2017-03-16 13:07:34 <^7heo> What does it do? 2017-03-16 13:08:29 <^7heo> I get that ${ALPINE_BRANCH:="v3.4"} is returning ALPINE_BRANCH and defaulting to "v3.4" 2017-03-16 13:08:34 ^7heo, ':' is basically no-op, so the effect is ${}'s side effect 2017-03-16 13:08:46 <^7heo> fabled: I know that : is basically noop 2017-03-16 13:08:58 for := it is assign variable if it's not defined 2017-03-16 13:09:06 <^7heo> assigning? 2017-03-16 13:09:12 <^7heo> I thought it was returning. 2017-03-16 13:09:15 <^7heo> like 2017-03-16 13:09:25 <^7heo> ${foo:-bar} 2017-03-16 13:09:26 <^7heo> oh wait 2017-03-16 13:09:29 <^7heo> :- is returning 2017-03-16 13:09:34 <^7heo> := might be assigning. 2017-03-16 13:09:39 <^7heo> that might be what I was missing there. 2017-03-16 13:09:41 If parameter is unset or null, the expansion of word is assigned to parameter. The value of parameter is then substituted. Positional parameters and special parameters may not be assigned to in this way. 2017-03-16 13:10:14 <^7heo> thanks dog, they wrote "substituted", not "interpolated". 2017-03-16 13:10:26 <^7heo> (shouldn't it be "subsitued" btw?) 2017-03-16 13:10:40 <^7heo> Ok 2017-03-16 13:11:33 <^7heo> the other thing I don't understand about this syntax is: as far as I have understood, from what I read, `:` is equivalent to a comment. Everything coming after will be ignored until newline. 2017-03-16 13:11:49 <^7heo> Does it have a different behavior? 2017-03-16 13:12:02 it's needed to ignore the substituted value 2017-03-16 13:12:24 it's not comment, the stuff is still evaluated, but after that it's not executed 2017-03-16 13:12:35 thus it's different from '#' which treats remainder as comment 2017-03-16 13:13:01 <^7heo> evaluated but not executed... 2017-03-16 13:13:08 yeah 2017-03-16 13:13:21 <^7heo> man eval 2017-03-16 13:13:46 <^7heo> The eval utility shall construct a command by concatenating arguments together, separating each with a character. The constructed command shall be read and *executed* by the shell. 2017-03-16 13:13:51 <^7heo> Now I'm lost :P 2017-03-16 13:13:59 <^7heo> What do you mean by evaluated? 2017-03-16 13:14:16 <^7heo> (I got that it's not the meaning you wanted to put in it, but I don't understand how you can evaluate something without executing it) 2017-03-16 13:23:37 <^7heo> Well, does not matter, actually. 2017-03-16 13:23:47 <^7heo> It's not like I care much about shell. 2017-03-16 13:35:28 <^7heo> fabled: if you feel like explaining what "evaluated but not executed" means, I'm really interested, tho. 2017-03-16 13:35:51 <^7heo> (even if I said I didn't care much - I do care about understanding stuff and that's gonna bug the hell outta me for hours otherwise) 2017-03-16 13:37:12 http://unix.stackexchange.com/questions/31673/what-purpose-does-the-colon-builtin-serve 2017-03-16 13:39:09 <^7heo> fabled: right, I could have used my favorite search engine to get that answer. Sorry and thanks for your time. 2017-03-16 13:42:50 <^7heo> Oh, so that is an exploit of an undocumented implementation specific detail... 2017-03-16 13:42:55 <^7heo> not so neat =/ 2017-03-16 13:43:10 <^7heo> ie. http://unix.stackexchange.com/a/39768/28394 2017-03-16 13:51:51 <^7heo> "Let's further suppose that you are a cretin and insist upon programming csh scripts." 2017-03-16 13:51:56 <^7heo> (source: http://www.faqs.org/faqs/unix-faq/faq/part3/section-16.html) 2017-03-16 13:51:59 <^7heo> <3 2017-03-16 13:58:36 <^7heo> But it still does not explain how ':' works. 2017-03-16 13:59:01 : is basically just a shorthand for true 2017-03-16 13:59:16 <^7heo> not according to what I've found. 2017-03-16 13:59:58 <^7heo> according to the user Gilles on SO, it's supposed to be (like in C) the start of a label 2017-03-16 14:00:01 <^7heo> for a goto. 2017-03-16 14:00:08 my script just sets up plain old chroot, without any namespaces, so you can use the host’s network 2017-03-16 14:00:17 <^7heo> But since bash doesn't have goto's, it doesn't have use for labels 2017-03-16 14:00:25 <^7heo> and just ignores all the things coming after `:`. 2017-03-16 14:00:47 <^7heo> That still does not explain why the rest of the line is read and interpreted. 2017-03-16 14:01:02 <^7heo> jirutka: yeah I like your script. 2017-03-16 14:01:11 it does not do any isolation, so it’s not comparable with docker, but totally fine for CI or similar use cases 2017-03-16 14:01:26 <^7heo> jirutka: I just stumbled upon that `:` operator used in a different way than I've seen it used so far. 2017-03-16 14:01:39 <^7heo> which is raising my curiosity to insane levels. 2017-03-16 14:01:41 it do the same as our script in alpinelinux/aports that builds pkgs on Travis 2017-03-16 14:01:54 <^7heo> "does not do any isolation" 2017-03-16 14:02:00 imo `: ${foo:=bar}` is almost idiomatic use of : :) 2017-03-16 14:02:00 <^7heo> not true, chroot is some kind of isolation. 2017-03-16 14:02:09 well, yes… 2017-03-16 14:02:24 <^7heo> But I got what you meant: it's not doing cgroups/network isolation. 2017-03-16 14:02:26 but just a FS isolation 2017-03-16 14:02:28 <^7heo> yeah 2017-03-16 14:02:29 yes 2017-03-16 14:02:47 <^7heo> but FS isolation is the most important one. 2017-03-16 14:02:47 heh, however, it can be very simply added ;) 2017-03-16 14:02:53 <^7heo> indeed. 2017-03-16 14:03:00 using unshare and ip 2017-03-16 14:03:53 <^7heo> What bugs me is: 2017-03-16 14:04:07 <^7heo> why is : behaving differently than # and where is this documented? 2017-03-16 14:04:29 <^7heo> and behaving differently than true, too. 2017-03-16 14:04:39 well, why true behaves differently than # ? 2017-03-16 14:04:49 <^7heo> ah no it's behaving like true. 2017-03-16 14:04:53 maybe because there are totally different? :) `:` has nothing in common with a comment 2017-03-16 14:04:57 <^7heo> I just have overlooked it. 2017-03-16 14:05:26 <^7heo> I think I start to understand. 2017-03-16 14:05:27 `: ${foo:=bar}` is a shorthand for `foo="${foo:-bar}"` 2017-03-16 14:05:43 <^7heo> True being an external program (not a necessarily a builtin) 2017-03-16 14:05:55 <^7heo> (depending on the shell) 2017-03-16 14:06:13 <^7heo> the shell needs to FIRST process the arguments 2017-03-16 14:06:27 <^7heo> and THEN invoke the binary with said arguments 2017-03-16 14:06:50 <^7heo> because it has no knowledge that true is true, it could be /bin/echo. 2017-03-16 14:07:00 yes, it can 2017-03-16 14:07:20 <^7heo> I see now. 2017-03-16 14:07:22 echo is special, it behaves quite strange, but try e.g. ls ${foo:=42} 2017-03-16 14:07:44 <^7heo> It's not an operator, it's just a shorthand for something that shouldn't be assumed to be a builtin. 2017-03-16 14:08:05 actually, it works even with echo 2017-03-16 14:08:06 <^7heo> and therefore, cannot rely on the fact that it knows that the execution has to be stopped there. 2017-03-16 14:08:09 <^7heo> but 2017-03-16 14:08:15 <^7heo> yeah because echo can be /bin/echo 2017-03-16 14:08:20 <^7heo> echo isn't necessarily a builtin either. 2017-03-16 14:09:05 <^7heo> first, the origins of `:` == `true` seem to be extremely shady. 2017-03-16 14:09:20 <^7heo> and second, the it isn't specified anywhere, as far as I've been able to see. 2017-03-16 14:09:28 <^7heo> That bugs the hell out of me. 2017-03-16 14:11:41 you are looking for logic in the shell. That's a recipe for failure and frustration. 2017-03-16 14:12:10 <^7heo> I'll take your educated word for it. 2017-03-16 14:12:10 The shell isn't made of logic. It's made of historical cruft that has fossilized. 2017-03-16 14:12:17 fcolista: community/tint2 get different checksum on downloads 2017-03-16 14:12:20 <^7heo> Seems like it yeah. 2017-03-16 14:12:29 <^7heo> ncopa: it's tinted. 2017-03-16 14:12:34 <^7heo> ACTION hides 2017-03-16 14:12:59 ncopa, let me check 2017-03-16 14:14:51 ncopa, i've checket the tag: 2017-03-16 14:14:51 https://gitlab.com/o9000/tint2/tags 2017-03-16 14:14:59 *s/checked/downloaded 2017-03-16 14:17:15 abuild cleancache 2017-03-16 14:17:26 and it will give different checksum 2017-03-16 14:17:45 might be that the tar.gz is generated each time even thoug there's a tag? 2017-03-16 14:17:52 ^7heo: ${foo:=42} is a shell expansion, shell processes these before executing commands 2017-03-16 14:17:56 *thought 2017-03-16 14:18:12 fcolista: yes i think it does so 2017-03-16 14:18:21 ^7heo: just read how shell processes the command line, it’s not so complicated ;) 2017-03-16 14:18:23 .gz might have a timestamp field 2017-03-16 14:18:24 <^7heo> jirutka: I got that yes. 2017-03-16 14:18:40 .bz2 should work, but i dont think it does either 2017-03-16 14:18:41 liek github 2017-03-16 14:18:53 <^7heo> jirutka: but since : is a builtin, and has no spec, and is there because it is a reminescence of the "goto labels"... 2017-03-16 14:18:59 but what's the purpose of the tag at this point... 2017-03-16 14:19:05 <^7heo> jirutka: It's hard to tell that it *has* to work that way. 2017-03-16 14:19:11 <^7heo> jirutka: catch my drift? 2017-03-16 14:19:12 its a tag in git 2017-03-16 14:19:19 ^7heo: maybe I confused you with `echo` that it behaves differently, my fault, of course it doesn’t have any effect to expansions, it’s unrelated 2017-03-16 14:19:31 ^7heo: well, actually now 2017-03-16 14:19:35 s/now/no/ 2017-03-16 14:20:20 as I said, you can write even `ls ${foo:=42}` and it will also set foo to 42 if empty or unset 2017-03-16 14:21:24 ncopa, yeah..but those tag helped in having a fixed checksum in github. Apparentrly not in gtlab 2017-03-16 14:21:42 do you want me to upload it on dev.alpinelinux.org ? 2017-03-16 14:21:54 i think github caches it 2017-03-16 14:22:12 because shell processes expansions first, so it become `ls 42` and then executes `ls 42`; the important thing is that ${foo:=42} has a side effect, so when the expansion happen, it assigns 42 to foo (if foo is empty is unset) 2017-03-16 14:22:13 I'm looking at tint2 git, but does not seems to have a "fixed" package 2017-03-16 14:23:06 main/procps probaby have same issue 2017-03-16 14:23:34 and mdds1.0 2017-03-16 14:23:35 hm, GitLab… I thought that they provide stable tarballs 2017-03-16 14:23:52 might be they do 2017-03-16 14:23:58 but maybe they updated git 2017-03-16 14:24:16 i had the same issue with generated tarballs on cgit 2017-03-16 14:24:22 but to be honest I’m not very surprised that it doesn’t work, b/c they way of providing tarballs is crappy 2017-03-16 14:24:48 .gz has a timestamp in there 2017-03-16 14:24:50 hi 2017-03-16 14:24:53 .bz2 has not 2017-03-16 14:24:56 hi kaniini 2017-03-16 14:25:44 ACTION is hoping to chew through a good chunk of PaX this afternoon 2017-03-16 14:26:35 I feel for your teeth 2017-03-16 14:26:52 gzip -N is the only sane way to use gzip 2017-03-16 14:27:14 kaniini: do you think we will be able to maintain pax patch? 2017-03-16 14:27:26 i know they are upstreaming bits of it with KSPP 2017-03-16 14:27:28 sorry -n rather 2017-03-16 14:27:39 -N is the opposite/default of saving timestamp 2017-03-16 14:27:41 ncopa: do you want me to update the checksum and commit the updated APKBUILD ? 2017-03-16 14:27:43 which is the reason spender closes the public patch 2017-03-16 14:27:50 fcolista: yeah pleas do so 2017-03-16 14:28:38 ncopa: i think at the very least, it will be much easier for KSPP to upstream stuff from PaX after i am done with this 2017-03-16 14:29:26 it will annoy spender 2017-03-16 14:29:31 i dont know if its wise 2017-03-16 14:29:50 uff 2017-03-16 14:30:03 <^7heo> jirutka: sorry, went to make food 2017-03-16 14:30:19 do we really care if it’d annoy spender? 2017-03-16 14:30:29 <^7heo> jirutka: I think poeple would be amused. 2017-03-16 14:30:41 <^7heo> it's mean but it's true. 2017-03-16 14:31:06 he closed even unstable patches, so we and Gentoo are quite in trouble with grsecurity now… 2017-03-16 14:31:10 <^7heo> jirutka: essentially my point is: `:` is just existing because `:foo` was label foo in some previous shell, and isn't specified anywhere. 2017-03-16 14:31:19 i kind of care. still trying to talk into some kind of solution 2017-03-16 14:31:22 ncopa, i think you should delete the cached tar.gz 2017-03-16 14:31:27 and if I haven’t missed something, he didn’t get us know in advance 2017-03-16 14:31:29 <^7heo> jirutka: it could very well be the `else` operator in some future shell. 2017-03-16 14:32:09 ncopa: annoying people is unwise when you expect things from them, when you plan to work with them in the future 2017-03-16 14:32:13 <^7heo> jirutka: and because it has no specification, one cannot assume it is "a shorthand for true" at all times. 2017-03-16 14:32:31 <^7heo> skarnet: untrue, there are a lot of human relationships based on annoyance. 2017-03-16 14:32:41 <^7heo> skarnet: for further information, contact your administrations. 2017-03-16 14:32:54 so if we will not violate the license, it should be totally okay 2017-03-16 14:33:04 <^7heo> speaking of violating licenses 2017-03-16 14:33:10 <^7heo> can we pick a name for firefox? 2017-03-16 14:33:20 ncopa: but annoying someone who has shown a complete inability to work with others is a. not an issue, and b. cathartic 2017-03-16 14:33:29 <^7heo> snow-wolf or something 2017-03-16 14:33:54 skarnet: are we going to work with grsecurity in feature? it doesn’t look like, since it become a private project 2017-03-16 14:34:09 colon is in posix though 2017-03-16 14:34:22 <^7heo> mitchty: ah? 2017-03-16 14:34:26 <^7heo> mitchty: where? 2017-03-16 14:34:27 jirutka: while true, spender is ... very destructive when annoyed 2017-03-16 14:34:28 what’s the problem with Firefox? 2017-03-16 14:34:33 ^7heo: I suggest foxling, because alpine foxling (look it up) 2017-03-16 14:34:37 i am still hoping we can find a solution to work with grsecurity 2017-03-16 14:34:38 ^7heo: http://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html 2017-03-16 14:34:39 kaniini: hm, what can he do? 2017-03-16 14:34:43 search for colon 2017-03-16 14:34:47 and i think we should respect his wish 2017-03-16 14:34:48 jirutka: DDoS 2017-03-16 14:34:49 <^7heo> jirutka: it's disallowed to distribute any mozilla products that hasn't been compiled by mozilla, with the mozilla name on it. 2017-03-16 14:34:51 or colon - null 2017-03-16 14:34:54 <^7heo> jirutka: same goes for firefox. 2017-03-16 14:35:00 <^7heo> mitchty: many thanks. 2017-03-16 14:35:01 if nothing else for respect of his work 2017-03-16 14:35:13 ^7heo: really? how other distributions handle it? 2017-03-16 14:35:24 <^7heo> they ship it under a different name. 2017-03-16 14:35:27 <^7heo> see iceweasel/ 2017-03-16 14:35:30 <^7heo> or they do not build it. 2017-03-16 14:35:31 <^7heo> see ubuntu. 2017-03-16 14:35:34 how it’s named in Fedora? 2017-03-16 14:35:38 <^7heo> No idea. 2017-03-16 14:35:40 <^7heo> I don't use it. 2017-03-16 14:35:52 <^7heo> they might also take the binary from mozilla servers directly. 2017-03-16 14:35:56 if someone DDoSses others when he's annoyed, how come he's not widely named and shamed? 2017-03-16 14:36:02 idk 2017-03-16 14:36:16 the package is definitely named firefox http://pkgs.fedoraproject.org/cgit/rpms/firefox.git/tree/firefox.spec?h=f26 2017-03-16 14:36:27 <^7heo> skarnet: because nobody wants to talk first and be DDoSed 2017-03-16 14:36:40 <^7heo> jirutka: right, so it's a binary taken from mozilla 2017-03-16 14:36:44 no 2017-03-16 14:36:45 <^7heo> jirutka: or they're illegal. 2017-03-16 14:36:46 they compile it 2017-03-16 14:36:49 since when do we negotiate with terrorists? 2017-03-16 14:36:49 if you read the specfile 2017-03-16 14:36:52 <^7heo> then they're illegal. 2017-03-16 14:37:13 and no, mozilla allows distros to do this as long as you work with them to fix concerns they have with the packaging 2017-03-16 14:37:13 no it’s not 2017-03-16 14:37:14 <^7heo> skarnet: since IRC isn't a democracy. 2017-03-16 14:37:24 which we do 2017-03-16 14:37:25 <^7heo> kaniini: ah? 2017-03-16 14:37:32 <^7heo> kaniini: any source for this info? 2017-03-16 14:37:39 mozilla is quite aware of the firefox package in alpine, they have not complained 2017-03-16 14:37:41 meh, I liked alpine foxling :( 2017-03-16 14:38:06 ^7heo: the reason why debian has iceweasel is because the requirement to address concerns mozilla has with the package violates DFSG 2017-03-16 14:38:34 i think they were more aggresive on the trademark earlier 2017-03-16 14:39:20 <^7heo> mitchty: http://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18_16 :) 2017-03-16 14:39:25 <^7heo> mitchty: nice link, thanks a lot. 2017-03-16 14:39:44 <^7heo> kaniini: ah ok. 2017-03-16 14:39:59 <^7heo> kaniini: I'll take your word for it. I don't have the time to check that now. But thanks for correcting me. :) 2017-03-16 14:40:13 ^7heo: no problem, generally its good to always search through that site, half the time i learn something is posix that i never knew about, stupidly old spec 2017-03-16 14:40:25 <^7heo> skarnet: me too, I like snowwolf or something alike. 2017-03-16 14:40:40 ^7heo: i would say our relationship with mozilla is one of the better relationships we have with upstreams 2017-03-16 14:40:51 <^7heo> mitchty: the issue atm for me is that aside from typing/reading on IRC, I am rushing at work 2017-03-16 14:40:54 definitely better than our current relationship with grsecurity hahahaha 2017-03-16 14:41:10 <^7heo> mitchty: so I really appreciate that the people on IRC are helpful and minimize my www-interaction. 2017-03-16 14:41:21 <^7heo> so thanks again. 2017-03-16 14:41:24 if you’re afraid of spender, then maybe kaniini can just release his patches under his name, with no relation to Alpine, and we will just find it by accident and use :) 2017-03-16 14:41:45 jirutka: that is kind of what KSPP does 2017-03-16 14:41:51 <^7heo> kaniini: does anyone have a good relationship with grsecurity at this point? 2017-03-16 14:42:02 i am sure the people who pay $500/mo/server do 2017-03-16 14:42:04 <^7heo> ok. 2017-03-16 14:42:06 money talks etc 2017-03-16 14:42:13 <^7heo> I see. 2017-03-16 14:42:19 according spender its not about money 2017-03-16 14:42:27 oh really? so what is this about? 2017-03-16 14:42:42 just tired of being disrespected and exploited 2017-03-16 14:42:47 ncopa: it's about being shown up by a woman, really 2017-03-16 14:42:57 the main driver of KSPP is a woman (laura abbott) 2017-03-16 14:43:02 I would understand if it is about money, it’s his own decision and he has full right for it, but if it’s not, then what the heck is all this about…? 2017-03-16 14:43:28 she is doing what he would not, and that is getting under his skin 2017-03-16 14:43:31 that's the reality of it 2017-03-16 14:43:35 heh :) 2017-03-16 14:44:07 he is annoyed that they present the security features as their inventions 2017-03-16 14:44:15 that is a lie 2017-03-16 14:44:27 every single KSPP commit (and i looked) that derived code from grsecurity says so 2017-03-16 14:44:39 without giving proper credit 2017-03-16 14:44:49 well, thats how he feels it 2017-03-16 14:44:55 what is proper credit in spender's opinion? 2017-03-16 14:45:04 that i dont know 2017-03-16 14:45:36 i mean, i don't think the kernel guys are going to give spender top billing in their README etc 2017-03-16 14:46:07 but the KSPP website and all of their commits say it comes from previous work grsecurity & pax 2017-03-16 14:46:39 so, i think it's really one of two things 2017-03-16 14:46:50 (a) he is pissed that a woman is doing most of the driving on KSPP 2017-03-16 14:47:02 (b) he is pissed that KSPP is a threat to his new revenue stream 2017-03-16 14:47:20 and lets be honest here 2017-03-16 14:47:32 KSPP would not exist if he had not closed the stable patches to begin with 2017-03-16 14:47:48 so all of this is really self-inflicted 2017-03-16 14:47:56 <^7heo> kaniini: I'm not saying this is wrong, but I'd really like some evidence about that, instead of going for the "witch hunting feminist integrist theory" 2017-03-16 14:48:18 i suspect he is annoyed that his patches no longer applies when there comes new version 2017-03-16 14:48:24 <^7heo> (which I'm naming that way because without a doubt, convicting without proofs just calls for MORE convicting without proofs) 2017-03-16 14:48:31 and he gets more work to figure out how to rebase 2017-03-16 14:48:39 its a bit confusing 2017-03-16 14:49:00 because he is also annoyed that they claim it is grsecurity - eg exploiting the good name "grsecurity" 2017-03-16 14:49:01 well, if he did not close the stable patches, nobody would have cared 2017-03-16 14:49:09 while it is a watered out version of it 2017-03-16 14:49:29 where do they claim that though? 2017-03-16 14:49:37 the website says: 2017-03-16 14:49:50 These kinds of protections have existed for years in PaX, grsecurity, and piles of academic papers. For various social, cultural, and technical reasons, they have not made their way into the upstream kernel, and this project seeks to change that. Our focus is on kernel self-protection, rather than kernel-supported userspace protections. The goal is to eliminate classes of bugs and eliminate methods of 2017-03-16 14:49:51 exploitation. 2017-03-16 14:50:14 How come I can't use wacom tablet? 2017-03-16 14:50:34 Is it because of CONFIG_INPUT_TABLET? 2017-03-16 14:50:46 kaniini: yeah i find it a bit confusing 2017-03-16 14:50:50 at any rate, i would like to continue shipping PaX 2017-03-16 14:51:04 so if they will no longer supply patches, clearly it is time to get to work 2017-03-16 14:51:23 if it makes spender happy, we can call it 2017-03-16 14:51:32 -not-really-grsec-no-really-its-not 2017-03-16 14:52:24 <^7heo> speaking about renaming shit 2017-03-16 14:52:32 <^7heo> can we say Alpine Linuk from now on? 2017-03-16 14:52:38 <^7heo> 'cause Linux now is GNU/Linux. 2017-03-16 14:52:46 wot 2017-03-16 14:52:47 <^7heo> And the Linux Kernel does not exist alone per se. 2017-03-16 14:52:53 who says that 2017-03-16 14:53:00 me too, i'd like continue ship grsecurity or pax if possible 2017-03-16 14:53:01 <^7heo> wikipedia, the linux foundation, etc. 2017-03-16 14:53:16 <^7heo> And so Linu-K (linu(x) k(ernel)) would be better. 2017-03-16 14:53:17 Yeah 2017-03-16 14:53:23 <^7heo> So people understand it's not GNU/Linux. 2017-03-16 14:53:24 and doing it while giving proper respect and credit to the people who deserves it 2017-03-16 14:53:38 i sitll havent figured out how 2017-03-16 14:53:48 https://guiodic.files.wordpress.com/2009/04/android_not_linux.png 2017-03-16 14:54:32 <^7heo> thanks pickfire for providing with evidences supporting what I'm going for :) 2017-03-16 14:54:38 ncopa: we can document that the -hardened patch is originally derived from grsecurity/PaX features but is explicitly not grsecurity/PaX 2017-03-16 14:54:45 that is about all we can do, really 2017-03-16 14:54:55 yeah 2017-03-16 14:55:25 also, do know that we have an unofficial port of grsecurity 2017-03-16 14:55:37 we asked that we changed the wording on out web page 2017-03-16 14:55:52 it used to say that it was grsecurity (which it was earlier) 2017-03-16 14:56:33 ^7heo: Do you compile custom kernel? 2017-03-16 14:56:54 <^7heo> nope, no time for that. 2017-03-16 14:57:00 <^7heo> but I will eventually come to that point yes. 2017-03-16 14:57:05 <^7heo> (I need that for some projects) 2017-03-16 14:57:12 ^7heo: Everything is working fine? 2017-03-16 14:57:16 <^7heo> yep 2017-03-16 14:57:24 I can't use wacom bamboo tablet here. 2017-03-16 14:57:28 <^7heo> But I'm exclusively using lenovos 2017-03-16 14:57:34 <^7heo> without peripherals 2017-03-16 14:57:35 <^7heo> so... 2017-03-16 14:57:44 I am using x220 here as well 2017-03-16 14:57:45 <^7heo> pickfire: very possible. 2017-03-16 15:00:05 <^7heo> Anyway, ncopa, kaniini, jirutka and others: as pickfire pointed out (by linking that very revealing slide: https://guiodic.files.wordpress.com/2009/04/android_not_linux.png ), we might have to think about changing the "Linux" in our distro name. 2017-03-16 15:00:20 wtf are you talking about 2017-03-16 15:00:21 <^7heo> to something that reflects that we're using the Linux KERNEL, not Linux. 2017-03-16 15:00:36 <^7heo> skarnet: that Linux != the kernel anymore. 2017-03-16 15:00:42 <^7heo> skarnet: Linux == GNU/Linux. 2017-03-16 15:00:46 sorry but no. 2017-03-16 15:01:01 <^7heo> Sorry but read wikipedia, gnu.org and the linux foundation website. 2017-03-16 15:01:10 <^7heo> or linux.com 2017-03-16 15:01:28 <^7heo> Officially, Linux == GNU/Linux, as of 2016 already. 2017-03-16 15:01:54 <^7heo> you can deny that all you want, you'll be a very rare person. 2017-03-16 15:01:54 is this a troll? 2017-03-16 15:01:58 <^7heo> No, it is real. 2017-03-16 15:02:00 because corporate has always been successful in imposing terminology, amirite 2017-03-16 15:02:05 i cant always tell with you ;) 2017-03-16 15:02:18 <^7heo> that's fine. 2017-03-16 15:02:21 <^7heo> but this isn't a troll. 2017-03-16 15:02:31 <^7heo> Maybe you've been under a rock for the last years about that terminology, but it has changed. 2017-03-16 15:02:32 well I'm going to keep treating it as a troll 2017-03-16 15:02:46 <^7heo> skarnet: just to hurt people? 2017-03-16 15:02:49 even if GNU are the trolls - which they are 2017-03-16 15:02:59 <^7heo> nah it's not GNU trolling here. 2017-03-16 15:03:02 <^7heo> it's Linus too. 2017-03-16 15:03:16 <^7heo> GNU is actually asking for the name to be GNU/Linux (with GNU first ofc) 2017-03-16 15:03:27 so? 2017-03-16 15:03:53 <^7heo> So maybe, for the sake of clarity and correctness, we should use what the rest of the world population uses. 2017-03-16 15:03:54 Even Linus can't change the fact that there's a lot of userland to go with the kernel and not all of it is GNU. 2017-03-16 15:03:58 <^7heo> to communicate. 2017-03-16 15:04:10 ^7heo: https://www.gnu.org/gnu/gnu-linux-faq.html#linuxsyswithoutgnu 2017-03-16 15:04:14 <^7heo> because we ALL KNOW, here, that we're using the Linux Kernel, and a very limited (if any) amount of GNU software. 2017-03-16 15:04:20 I've been communicating real fine with the rest of the world so far, you being a rare exception. 2017-03-16 15:04:27 <^7heo> thanks pickfire for doing the search as I'm talking here; I needed that link. 2017-03-16 15:04:36 https://itvision.altervista.org/why.linux.is.not.ready.for.the.desktop.current.html 2017-03-16 15:04:37 <^7heo> skarnet: your world is very limited too. 2017-03-16 15:04:56 <^7heo> skarnet: you are surrounded by clever people, who are educated. 2017-03-16 15:05:04 You don't get to judge the extent of my world. 2017-03-16 15:05:08 <^7heo> right. 2017-03-16 15:05:11 <^7heo> But let's say. 2017-03-16 15:05:16 No, let's not say. 2017-03-16 15:05:26 <^7heo> I really STRONGLY assume that you've surrounded yourself with clever people. 2017-03-16 15:05:34 <^7heo> because you're clever too. 2017-03-16 15:05:39 I read pickfire's link. It's the exact same thing as it was yesterday, as it was for 10 years. 2017-03-16 15:05:45 Nothing has changed here. 2017-03-16 15:05:49 ^7heo: Can you just trim those unused modules? 2017-03-16 15:05:56 <^7heo> pickfire: which? 2017-03-16 15:06:05 The only thing that has changed is somehow you became aware of it and started trying to educate people. 2017-03-16 15:06:07 <^7heo> skarnet: the thing that has changed is about communication with most of the population. 2017-03-16 15:06:10 Who don't need it. 2017-03-16 15:06:23 I read those today. 2017-03-16 15:06:28 at any rate, i would be more inclined to just call it Alpine OS 2017-03-16 15:06:33 <^7heo> skarnet: look, if you get people to understand "The Kernel" when you say Linux, please tell me how your magic works. 2017-03-16 15:06:38 <^7heo> kaniini: +1 2017-03-16 15:06:53 Then just say Alpine 2017-03-16 15:06:56 <^7heo> pickfire: no. 2017-03-16 15:07:01 <^7heo> pickfire: that is a MUA. 2017-03-16 15:07:07 <^7heo> skarnet: because then it would solve my problem: i.e. communicating with others is a pain currently. 2017-03-16 15:07:10 + Linux Kernel 2017-03-16 15:07:11 pickfire: then people go "you mean the car radio manufacturer?" 2017-03-16 15:07:32 "Linux" has always colloquially meant "the Linux kernel plus some userland software, parts of which may be GNU and parts of which may not" 2017-03-16 15:07:39 that's how everyone understands it. 2017-03-16 15:07:42 kaniini: We can take their name as well. 2017-03-16 15:07:52 skarnet: No 2017-03-16 15:08:03 What about Void Linux? 2017-03-16 15:08:05 <^7heo> skarnet: it goes as: "what is your OS?" - "Alpine Linux" - "Ah so you can run ." - "No I'm not sure I can." - "But it's built FOR LINUX!" 2017-03-16 15:08:10 <^7heo> skarnet: "it's complicated." 2017-03-16 15:08:21 in a way theo has a point 2017-03-16 15:08:21 <^7heo> skarnet: "ok you actually refuse to, right?" 2017-03-16 15:08:33 Then Alpine Musl/Linux 2017-03-16 15:08:41 <^7heo> skarnet: you have NO idea how much trouble this has got me in the past. 2017-03-16 15:08:46 We show them muscle 2017-03-16 15:08:57 <^7heo> skarnet: so again, I am all okay to be nice and respectful to you; because I respect you. 2017-03-16 15:09:05 I have a pretty good idea of your own ability to get into trouble as soon as you engage conversation, no matter the subject. :P 2017-03-16 15:09:10 I have trouble with musl. 2017-03-16 15:09:14 <^7heo> skarnet: but I won't accept something I don't agree with, just because it comes from you :) 2017-03-16 15:09:19 Yes 2017-03-16 15:09:31 No wonder this is what happened back then. 2017-03-16 15:09:42 https://www.gnu.org/gnu/gnu-linux-faq.html#finishhurd 2017-03-16 15:09:44 lol 2017-03-16 15:09:46 why not finish GNU hurd 2017-03-16 15:09:46 <^7heo> huhu 2017-03-16 15:09:48 hahahahahaha 2017-03-16 15:09:51 ^7heo: Is this how you get kickban back then? 2017-03-16 15:10:00 <^7heo> pickfire: from suckless? 2017-03-16 15:10:02 we would never kickban theo 2017-03-16 15:10:06 Haha 2017-03-16 15:10:07 <^7heo> pickfire: nah that was from trolling with quinq. 2017-03-16 15:10:07 he is such a great guy 2017-03-16 15:10:08 finish hurd so people can still call it Linux and piss RMS off even more 2017-03-16 15:10:14 <^7heo> kaniini: come on stop with the harmful sarcasm. 2017-03-16 15:10:16 Who? 2017-03-16 15:10:21 <^7heo> pickfire: quinq. 2017-03-16 15:10:23 kaniini: ^ 2017-03-16 15:10:25 sometimes I tell friends that one of the advantages of Alpine Linux is that they don’t have to call it GNU/Linux without upsetting Stallman, because there’s no any GNU software in base installation :) 2017-03-16 15:10:26 Ah 2017-03-16 15:10:27 <^7heo> pickfire: but we solfed it out. 2017-03-16 15:10:31 <^7heo> solved even. 2017-03-16 15:10:41 ^7heo: i am serious! 2017-03-16 15:10:52 <^7heo> kaniini: thanks then. 2017-03-16 15:10:56 <^7heo> skarnet: you misjudge me. 2017-03-16 15:11:00 theo is easy to type on dvorak. 2017-03-16 15:11:04 <^7heo> skarnet: some people actually appreciate my personality. 2017-03-16 15:11:09 ^7heo is a mess 2017-03-16 15:11:10 <^7heo> skarnet: outside of Frankreich. 2017-03-16 15:11:23 <^7heo> skarnet: at least I'm honest - if not politically correct. 2017-03-16 15:11:34 I'm one of those. I would have muted you a long time ago if I didn't appreciate your personality. :P 2017-03-16 15:11:35 hey it worked for trump 2017-03-16 15:11:39 <^7heo> pickfire: sorry about that nick, I' wish I could use 7heo 2017-03-16 15:11:46 he just blurted out whatever he thought at the time 2017-03-16 15:11:48 and look at him now 2017-03-16 15:11:50 <^7heo> skarnet: v_v 2017-03-16 15:12:21 7 needs shiyt 2017-03-16 15:12:22 <^7heo> kaniini: are you comparing me with Trump now? 2017-03-16 15:12:26 shift* 2017-03-16 15:12:36 <^7heo> kaniini: if you are, I'm not gonna believe that this isn't meant as a harmful comment ;) 2017-03-16 15:12:52 no 2017-03-16 15:12:56 <^7heo> good :D 2017-03-16 15:13:06 <^7heo> I'm not that old. 2017-03-16 15:13:09 <^7heo> or that orange. 2017-03-16 15:13:26 apparently the orange is from using tanning bed 2017-03-16 15:13:30 who’s pushing to aports now? there are a lot of failures on build servers 2017-03-16 15:13:51 kaniini: I don't quite like patching on github. 2017-03-16 15:14:03 tint2 push came from fcolista 2017-03-16 15:14:06 <^7heo> pickfire: the ML can also be used. 2017-03-16 15:14:07 Not even a single comment until now. 2017-03-16 15:14:14 <^7heo> pickfire: link? 2017-03-16 15:14:24 jirutka bringing back the serious in the channel :D 2017-03-16 15:14:32 i asked ncopa to delete the cached tint2 tar.gz 2017-03-16 15:14:38 ^7heo: But I heard that github is easier for maintainers. 2017-03-16 15:14:42 pickfire: why you don’t like patching on GH? 2017-03-16 15:14:54 pickfire: they are both good 2017-03-16 15:15:01 Because often, github is slow 2017-03-16 15:15:05 pickfire: coming in here and saying "yo review this " also works 2017-03-16 15:15:16 do what you want to do 2017-03-16 15:15:23 <^7heo> pickfire: github is definitely easier for maintainers. 2017-03-16 15:15:28 we have many contribution channels for that reason 2017-03-16 15:15:30 ;) 2017-03-16 15:15:31 I had sent patches and no one review it and then I sent more patches 2017-03-16 15:15:37 and I’m very sure that even for contributors 2017-03-16 15:15:49 And git pull -f remove my old patches 2017-03-16 15:15:57 once they used to it and stop complaining that they must use different command… 2017-03-16 15:15:57 <^7heo> pickfire: my experience is: if github is slow, the ML is probably slower. 2017-03-16 15:16:02 https://github.com/alpinelinux/aports/pull/1020#issuecomment-287002457 2017-03-16 15:16:09 <^7heo> thanks for the link tho 2017-03-16 15:16:13 <^7heo> saving us time to click many times. 2017-03-16 15:16:23 ^7heo: i check patchwork for things usually each day 2017-03-16 15:16:26 fcolista: i did delete it 2017-03-16 15:16:40 still failing? 2017-03-16 15:16:42 <^7heo> kaniini: yeah, that's cool; but github notifes quite fast if you're subscribed. 2017-03-16 15:16:49 ncopa: What about you? github or mailing lists is easier? 2017-03-16 15:16:59 pickfire: what the hell is that?! https://github.com/alpinelinux/aports/pull/1020/files#diff-e74cce27af5d72760a36be525d08364eR17 2017-03-16 15:17:01 seems the current failure is libwebsockets 2017-03-16 15:17:44 jirutka: Oh, I don't know how to write that part. 2017-03-16 15:17:44 pickfire: for individual patches i prefer mailing list, for a long set of patches i prefer git pull 2017-03-16 15:17:56 uff 2017-03-16 15:17:59 git pull from where? 2017-03-16 15:18:13 pickfire: anywhere. github works fine for that 2017-03-16 15:18:17 ACTION have libwacom patches 2017-03-16 15:18:26 Where should I sent? 2017-03-16 15:18:30 that way i can pull in more than one patch in one go 2017-03-16 15:18:39 ncopa: the main problem with Patchwork is that the review process is complicated and unclear on Patchwork 2017-03-16 15:18:40 ncopa: What about git.pickfire.tk? 2017-03-16 15:18:59 I see tor traffic there everyday. 2017-03-16 15:19:29 git.pickfireywcq2wf2.onion 2017-03-16 15:19:44 pickfire: good enough for me if there is no need to review 2017-03-16 15:19:49 ncopa: when some bad patch is accepted nowadays, it’s very often via Patchwork, not GH 2017-03-16 15:19:54 ncopa: https://github.com/alpinelinux/aports/pull/1020/files#diff-e74cce27af5d72760a36be525d08364eR17 2017-03-16 15:20:00 I don't know how to write that part. 2017-03-16 15:20:31 How to pass [start|stop|reload] to tlp init [start|stop|reload]? 2017-03-16 15:20:39 jirutka: i just pushed broken libwebsockets from GH 2017-03-16 15:20:50 :D 2017-03-16 15:21:27 i decided to not build test it myshelf and jus trust travis... 2017-03-16 15:22:00 hu 2017-03-16 15:22:01 ncopa: it’s broken on arm, not on x86_64… b/c we don’t test arm on Travis (yet) 2017-03-16 15:22:17 qemu-user-arm duh 2017-03-16 15:22:23 aha, x86 2017-03-16 15:22:23 come on 2017-03-16 15:22:27 x86 too 2017-03-16 15:22:28 but works on x86_64 2017-03-16 15:22:32 its probably only 32bit 2017-03-16 15:22:33 I hate these errors 2017-03-16 15:22:35 yeah 2017-03-16 15:22:39 why we still must support x86? 2017-03-16 15:22:50 I haven’t seen i686 for years 2017-03-16 15:23:00 I ran a i486 two months ago 2017-03-16 15:23:01 i would have pushed it broken if i'd taken it from alpine-aports too 2017-03-16 15:23:24 i asked about it earlier and skarnet still uses x86 2017-03-16 15:23:30 jirutka: well the smaller pointer widths saves on ram to start with 2017-03-16 15:24:18 ncopa: How should I do git send-email? 2017-03-16 15:24:29 pickfire: how many patches is it? 2017-03-16 15:24:30 if you are in a small vps, alpine 32-bit is much more useful than alpine 64-bit 2017-03-16 15:24:34 2 2017-03-16 15:24:40 Wacom related 2017-03-16 15:24:51 Still not usable because of kernel 2017-03-16 15:25:03 pickfire: you can use github if you are ok with that. otherwise git send-email to alpine-aports works too 2017-03-16 15:25:14 <^7heo> pickfire: you're sure it's not due to some GNU userland? :D 2017-03-16 15:25:38 <^7heo> ncopa: AFAIR the question was more about what you personally prefer. 2017-03-16 15:25:46 I can't 2017-03-16 15:25:51 i dont have strong opinion either way 2017-03-16 15:25:52 https://github.com/alpinelinux/aports/pull/1020 2017-03-16 15:25:56 On going pull 2017-03-16 15:26:08 you have to use separate branches for each change 2017-03-16 15:26:11 that you want 2017-03-16 15:26:49 kaniini: But can I have multiple on going pull request on github? 2017-03-16 15:28:06 yes, one per branch 2017-03-16 15:29:14 Ah 2017-03-16 15:29:29 I personally prefer ML 2017-03-16 15:29:38 <^7heo> pickfire: then go for it ;) 2017-03-16 15:29:52 <^7heo> pickfire: the big advantage of gh is that jirutka added testing to it. 2017-03-16 15:30:01 <^7heo> pickfire: so it tends to be preferred for that reason, when it is. 2017-03-16 15:30:05 But if I feel like something needs to be changed, I will throw it to github. 2017-03-16 15:30:10 Ah 2017-03-16 15:30:45 ^7heo: Is the tester (ci) our package builder? 2017-03-16 15:30:53 actually the review process is pretty nice too with GH 2017-03-16 15:31:00 <^7heo> pickfire: no it's travis. 2017-03-16 15:31:14 <^7heo> pickfire: when using the ML you actually use the builder as CI 2017-03-16 15:31:23 Why can't we use it to build? 2017-03-16 15:31:24 <^7heo> but then it can lead to builders being blocked and stuff 2017-03-16 15:31:26 <^7heo> and it sucks. 2017-03-16 15:31:33 How? 2017-03-16 15:31:37 <^7heo> How what? 2017-03-16 15:31:43 when using the ML you actually use the builder as CI 2017-03-16 15:32:01 ACTION is going offline soon. 2017-03-16 15:32:26 <^7heo> is that a threat? :D 2017-03-16 15:32:48 <^7heo> pickfire: in the ML we don't have testing. 2017-03-16 15:32:55 how can I “emulate” x86 on x86_64 machine with x86_64 kernel? maybe a silly question, but really I’ve never needed this… 2017-03-16 15:32:57 <^7heo> pickfire: so if your patch is accepted, after review... 2017-03-16 15:33:06 <^7heo> pickfire: it'll go to the builders. 2017-03-16 15:33:12 <^7heo> pickfire: if it breaks, it'll break there. 2017-03-16 15:33:16 <^7heo> pickfire: that's how. 2017-03-16 15:33:25 :) 2017-03-16 15:33:30 <_ikke_> jirutka: qemu? 2017-03-16 15:33:39 <^7heo> jirutka: by compiling the code in 32 bit? 2017-03-16 15:33:40 jirutka: But I thought our buildbot is x86_64? 2017-03-16 15:33:50 Cross compile 2017-03-16 15:34:03 <^7heo> _ikke_: isn't any x86_64 CPU actually able to run x86 code natively? 2017-03-16 15:34:21 <^7heo> (hence the name) 2017-03-16 15:34:21 in practice yes 2017-03-16 15:34:29 Then we need not to worry about having not enough processing power. 2017-03-16 15:34:30 it is 2017-03-16 15:34:36 <^7heo> wow, if you tell me that the practice works, but not the theory... 2017-03-16 15:34:40 <^7heo> I'll be blown away. 2017-03-16 15:34:44 + we can use github as cdn 2017-03-16 15:34:50 in theory, you could make a x86-64 cpu that booted in x86-64 mode and did not have it 2017-03-16 15:34:50 rawgit.com 2017-03-16 15:35:10 <_ikke_> github is not a cdn, they'll throttle you if they notice too much traffic to resources 2017-03-16 15:35:38 pickfire: i grabbed 2 of your patches 2017-03-16 15:35:45 Thanks 2017-03-16 15:35:59 thank you 2017-03-16 15:36:21 you might be able to copy the .initd from gentoo 2017-03-16 15:36:30 ncopa: Can I send patches for linux-grsec to support wacom as well or you don't want it to bloat? 2017-03-16 15:36:41 ncopa: I can't find that from gentoo. 2017-03-16 15:36:42 if its a module then no problem 2017-03-16 15:36:46 That's why. 2017-03-16 15:37:27 pickfire: I’ll write the runscript later, if no one else do it, but currently I’m too busy 2017-03-16 15:37:56 What run script? 2017-03-16 15:38:02 init script 2017-03-16 15:38:09 jirutka: For? 2017-03-16 15:38:27 tlp 2017-03-16 15:38:33 Ah 2017-03-16 15:38:37 Thanks a lot. 2017-03-16 16:29:45 > sudo 2017-03-16 16:29:45 Failed to execute process '/usr/bin/sudo'. Reason: 2017-03-16 16:29:45 exec: unknown error (errno was 1) 2017-03-16 16:29:51 Oh no, how come? 2017-03-16 16:33:23 https://transfer.sh/p9cik/2017-03-17-003311-1366x768-scrot.png 2017-03-16 16:33:28 1 is EPERM 2017-03-16 16:33:36 Those with dark red have the same issue. 2017-03-16 16:33:47 I can't execute them for no reason. 2017-03-16 16:34:29 What do I do nom? 2017-03-16 16:34:32 nonewprivs for any reason? 2017-03-16 16:34:43 those are suid programs. 2017-03-16 16:34:51 duncaen: What is nonewprivs? 2017-03-16 16:35:04 That means I can't suid? 2017-03-16 16:35:23 sounds like grstupidity 2017-03-16 16:35:23 if you run something in a namespace, seccomp container or with ptrace afaik, you cant regain privilegues 2017-03-16 16:36:02 ah, seccomp will also do that 2017-03-16 16:36:09 Oh 2017-03-16 16:36:13 I ran strace 2017-03-16 16:36:24 yes strace needs higher prives than the target 2017-03-16 16:36:24 What nowZ 2017-03-16 16:36:39 you can't strace suid programs indeed 2017-03-16 16:36:43 different error 2017-03-16 16:37:05 Now I can't use any of those suid stuff 2017-03-16 16:37:13 kill your strace 2017-03-16 16:37:49 I can't find it 2017-03-16 16:37:58 I think I ran it with sudo 2017-03-16 16:57:08 <^7heo> pickfire: thanks for contributing to packets 2017-03-16 16:57:14 <^7heo> packages* 2017-03-16 17:11:44 ^7heo: :) 2017-03-16 17:11:53 Anyone got sway working here? 2017-03-16 17:12:07 I tried running sway but looks like input isn't working. 2017-03-16 17:14:09 is libinput working? 2017-03-16 18:38:00 <^7heo> jirutka: I'm gonna fork and open a PR to your tool 2017-03-16 18:38:18 ^7heo: okay; what you’d like to change? 2017-03-16 18:38:30 <^7heo> the requirement for root at the start 2017-03-16 18:38:39 <^7heo> add a few checks (dependencies) 2017-03-16 18:38:51 <^7heo> add some resilience for curl/wget with check 2017-03-16 18:40:08 <^7heo> (i.e. if it has curl but not wget, use curl, if it has only wget use wget, etc.) 2017-03-16 18:40:22 <^7heo> AFAIK, it does not need root if the mounting is done somewhere else. 2017-03-16 18:40:24 why curl? 2017-03-16 18:40:26 <^7heo> right? 2017-03-16 18:40:34 <^7heo> beacuse some systems have curl and no wget. 2017-03-16 18:40:40 you need root for chroot 2017-03-16 18:40:44 <^7heo> ah that. 2017-03-16 18:40:47 what systems have curl and not wget? 2017-03-16 18:40:50 <^7heo> no idea. 2017-03-16 18:40:53 <^7heo> I don't remember. 2017-03-16 18:40:57 <^7heo> I just remember having that. 2017-03-16 18:41:01 <^7heo> curl -O does the same as wget 2017-03-16 18:41:15 then I’d prefer to first find what systems actually don’t have wget 2017-03-16 18:41:16 <^7heo> so it'd be worth it supporting wget AND curl imho. 2017-03-16 18:41:23 <^7heo> really?! 2017-03-16 18:41:28 <^7heo> I mean, REALLY?! 2017-03-16 18:41:51 <^7heo> you're RATHER search for proof than you're not ignoring some stupid system rather than support an hypothetical maximum of system? 2017-03-16 18:42:00 <^7heo> O.o 2017-03-16 18:42:09 <^7heo> (yes I'm trying new smileys) 2017-03-16 18:42:17 <^7heo> s/you're/you'd/ 2017-03-16 18:42:37 <^7heo> let's be clear here: I'm gonna do those changes no matter what. 2017-03-16 18:42:46 <^7heo> the only question is: will you accept them upstream or not. 2017-03-16 18:43:06 <^7heo> s/than/that/ 2017-03-16 18:43:08 <^7heo> damn I'm tired. 2017-03-16 18:43:27 <^7heo> s/system\ 2017-03-16 18:43:30 yes, because it adds complexity 2017-03-16 18:43:33 <^7heo> s/system\?/systems?/ 2017-03-16 18:43:55 <^7heo> well, YAGNI is a beautiful ideal. 2017-03-16 18:44:15 <^7heo> but you know, BSD has less complexity than Linux, if you go for that. 2017-03-16 18:44:31 <^7heo> and in Linux, there is a HUGE number of features/firmwares/etc. you're never gonna use, hence being YAGNI. 2017-03-16 18:44:40 <^7heo> yet, they are here. 2017-03-16 18:44:55 with the same logic what about rewriting it to some very old shell, don’t remember name, that one where you must write craps like `[ "x$foo" = x42 ]`, just for a chance that someone may have a very old system that doesn’t have normal POSIX compatible shell…? 2017-03-16 18:44:56 <^7heo> because it is what allows the system to work on the maximum of architectures. 2017-03-16 18:45:07 <^7heo> actually 2017-03-16 18:45:29 <^7heo> we SHOULD be writing `[ "x$foo" = "x42" ]` if we wanted to do it right. 2017-03-16 18:45:38 no, definitely not 2017-03-16 18:45:41 it’s non-sense 2017-03-16 18:45:49 <^7heo> well, it's non-sense, no question about that. 2017-03-16 18:45:59 <^7heo> but it's unfortunately necessary non-sense. 2017-03-16 18:46:02 <^7heo> like a lot of IT non-sense. 2017-03-16 18:46:04 I don’t wanna support ancient systems that maybe 1 person on Earth may use 2017-03-16 18:46:14 no, it’s not necessary at all 2017-03-16 18:46:18 <^7heo> Your point of view. 2017-03-16 18:46:32 <^7heo> The problem with bloat doesn't come from supporting a maximum of systems. 2017-03-16 18:46:37 this script can work only on Linux 2017-03-16 18:46:52 what Linux system does not have any normal POSIX compatible shell, hm? 2017-03-16 18:46:54 <^7heo> It comes from intellectual masturbation and the unnecessary level of astraction(s) that comes with it. 2017-03-16 18:47:04 no ash, dash, bash, zsh, … 2017-03-16 18:47:16 doesnt quoting alrady make the x stuff useless? 2017-03-16 18:47:20 <^7heo> jirutka: does POSIX actually defines that the shell has to consider "" different than nothing? 2017-03-16 18:47:22 of course it does 2017-03-16 18:47:34 thix "x" is totall non-sense that was maybe required 50 years ago… 2017-03-16 18:47:50 <^7heo> duncaen: no the whole problem is that this shell (POSIX AFAIK) consider an empty quoted string equivalent to nothing 2017-03-16 18:48:02 too much of backward compatibility is road to hell 2017-03-16 18:48:03 <^7heo> duncaen: and therefore crashes on [ "somestring" = ] 2017-03-16 18:48:17 [ "$foo" = "" ] 2017-03-16 18:48:21 oh its related to /bin/[, and argv 2017-03-16 18:48:25 jirutka: The reason for the [ x"$foo" =x"val" ] is that some broken implementations of test look through the quotes and see the option '-d' in foo="-d -r $file" 2017-03-16 18:48:28 <^7heo> duncaen: yes afaik 2017-03-16 18:48:34 however, this should be corectly written [ -n "$foo" ] 2017-03-16 18:48:40 <^7heo> jirutka: true dat. 2017-03-16 18:49:09 TemptorSent: do you remember what broken test, on what systems? 2017-03-16 18:49:12 <^7heo> jirutka: problem is, instead of doing `-n "$foo" -a "$foo" = "..."` 2017-03-16 18:49:18 <^7heo> people just write the latter part. 2017-03-16 18:49:20 In theory, any recent, sane shell will in fact see the quotes as introducing a value. 2017-03-16 18:49:43 i think its just for shells without builtin [, there you cant know if its quoted or not 2017-03-16 18:49:50 speaking of compatibility: there are ncursis-terminfo-base and ncurses-terminfo. while -base is sufficient today, many pkgs still depend und full terminfo . is that on purpose 2017-03-16 18:49:50 jirutka: I've been on a couple where it was an issue, but it was years ago. 2017-03-16 18:50:06 s/purpose/purpose?/ 2017-03-16 18:50:16 <^7heo> TemptorSent: are you 100% certain this is only about the possibility of the presence of flags and not also empty strings? 2017-03-16 18:50:42 <^7heo> anyway, jirutka, wget is gnu, curl isn't. 2017-03-16 18:50:53 nonempty variable would result in null which would result in no argv 2017-03-16 18:50:59 ^7heo: sry, I don’t have time now to discuss it; and as you said, you can send whatever PRs you want, it’s up to me to accept it or not; and I’ll definitely NOT accept any change of type “maybe this may not work on some system that I’ve never heard of” 2017-03-16 18:51:04 ^7heo: The empty strings were a gotcha in broken shells sometimes, but a quoted empty string worked in all intances I recall. 2017-03-16 18:51:18 <^7heo> jirutka: fine. 2017-03-16 18:51:22 <^7heo> jirutka: I'll still fork it. 2017-03-16 18:51:27 At least posix-like ones -- weird embedded stuff and VMS aside. 2017-03-16 18:51:38 <^7heo> jirutka: I need to be sure it works on modern systems and not only on GNU systems. 2017-03-16 18:51:55 ^7heo: if you tell me what real system doesn’t have wget but have curl by default, then I’m willing to accept it; otherwise it’s irrational change 2017-03-16 18:52:14 Alpine is not GNU and we have wget by default, provided by busybox 2017-03-16 18:52:25 wget is simpler tool than curl 2017-03-16 18:52:32 void has no curl or wget by default :P 2017-03-16 18:52:32 <^7heo> jirutka: I'm too tired and stressed to search for it now 2017-03-16 18:52:38 <^7heo> jirutka: but I'll try to find it for the PR 2017-03-16 18:52:47 <^7heo> duncaen: yeah that might be one. 2017-03-16 18:52:57 <^7heo> duncaen: but I know I used a system that had no wget 2017-03-16 18:52:59 <^7heo> duncaen: only curl. 2017-03-16 18:53:03 duncaen: well, what can be used on Void instead of wget or curl? 2017-03-16 18:53:09 <^7heo> nc :D 2017-03-16 18:53:20 ^7heo: What are you targeting as host systems? 2017-03-16 18:53:21 we have libfetch in xbps, there is a xbps-uhelper fetch command 2017-03-16 18:53:28 ^7heo: see! you must implement your own wget alternative in da script, b/c there’s at least one system that doesn’t have wget and curl by default! 2017-03-16 18:54:49 TemptorSent: any up-to-date linux system, but primarily debian-based b/c that’s what usually runs on public CIs 2017-03-16 18:56:28 ^7heo : What I've been doing with the featurs in mkimage is setting up functional fetures where it makes sense, such as ssh or ntp, then allowing the profile to use whichever flavor they like and the feature will configure it appropriately. 2017-03-16 18:56:36 <^7heo> jirutka: I will then. 2017-03-16 18:56:47 <^7heo> jirutka: I'm fine with implementing them in separate files 2017-03-16 18:56:50 <^7heo> jirutka: and sourcing what I need. 2017-03-16 18:57:00 ^7heo: no, this was a fucking joke! 2017-03-16 18:57:08 <^7heo> this kind of modular implementation is actually easy to extend 2017-03-16 18:57:19 <^7heo> if someone needs to use scp or something else. 2017-03-16 18:57:33 <^7heo> or git 2017-03-16 18:57:34 <^7heo> or I dunno 2017-03-16 18:57:41 <^7heo> (you never know what ports are open) 2017-03-16 18:57:43 ^7heo: alpine-chroot-install should be single shell script without any includes and bullshits 2017-03-16 18:57:50 <^7heo> yeah 2017-03-16 18:58:15 <^7heo> I kinda like that too. 2017-03-16 18:58:23 <^7heo> but a choice has to be made, a line has to be drawn. 2017-03-16 18:58:37 ^7heo: this is how I usually use it https://github.com/bigclownlabs/bc-bridge/blob/master/script/ci-install#L29 2017-03-16 18:58:38 <^7heo> I'm not sure I want to draw the line at "no curl support" 2017-03-16 18:58:55 <^7heo> and also 2017-03-16 18:59:07 <^7heo> it sucks that it needs root for the actual chrooting (even if it's logical) 2017-03-16 18:59:13 <^7heo> but yeah 2017-03-16 18:59:22 KISS 2017-03-16 18:59:22 <^7heo> I'd rather use root only for the necessary operations 2017-03-16 18:59:28 <^7heo> than for all the things. 2017-03-16 18:59:29 ^7heo: If you need more general purpose modular, take a look at base code of the mkimage fork at https://github.com/TemptorSent/aports/tree/mkimage-refactor-scripts/scripts/mkimage. 2017-03-16 18:59:36 <^7heo> thanks TemptorSent 2017-03-16 18:59:40 <^7heo> checking it now. 2017-03-16 18:59:59 too much modularity is also road to the hell 2017-03-16 19:00:09 <^7heo> modularity is abstraction. 2017-03-16 19:00:27 ^7heo: It's flexible enough that you can use the plugin loader for whatever you want and throw away the rest actually. 2017-03-16 19:00:30 <^7heo> it has obviously to be done with moderation in mind, eys. 2017-03-16 19:00:33 <^7heo> s/ey/ye 2017-03-16 19:01:28 <^7heo> this whole discussion gives ncopa's proposition a lot more seriousness. 2017-03-16 19:01:53 <^7heo> ie. <@ncopa> ^7heo: docker run --rm -it alpine ... :-p 2017-03-16 19:02:59 I have it down to basically a plugin loader, a bit of code to handle the options, and a section of the builder that handles the actual image generation. 2017-03-16 19:03:07 <^7heo> damn, why can't things be simple, AND complete, AND without tons of abstraction? 2017-03-16 19:03:30 ^7heo: Choose two and maybe :) 2017-03-16 19:03:39 <^7heo> yeah 2017-03-16 19:03:42 <^7heo> chose any two. 2017-03-16 19:04:01 <^7heo> like: cheap, fast, reliable 2017-03-16 19:04:08 Although simple, complete, with tons of abstraction seems self-inconsistent. 2017-03-16 19:04:21 <^7heo> well, that's basically IT in a nutshell 2017-03-16 19:04:59 ^7heo: if you want to support even ancient broken systems, then you can’t achieve simplicity 2017-03-16 19:05:16 jirutka: Just ask autotools! 2017-03-16 19:05:24 you mean autohells? 2017-03-16 19:05:24 <^7heo> curl == ancien && broken? 2017-03-16 19:05:44 <^7heo> by that standard let's assume busybox also is. 2017-03-16 19:06:02 no 2017-03-16 19:06:10 ^7heo: No, my Atari from the late '80s runing a unix environment is broken :) 2017-03-16 19:06:14 I mean that x"" bullshit 2017-03-16 19:06:18 <^7heo> aaah that 2017-03-16 19:06:19 <^7heo> yeah ok. 2017-03-16 19:06:43 <^7heo> TemptorSent: sorry to read that. You could run a recent NetBSD and have a good system. 2017-03-16 19:07:22 jirutka: Actually, I've seen broken '[' implementations crop up more recently that DID see through qoutes, so it's not a dead issue. 2017-03-16 19:07:35 <^7heo> also that. 2017-03-16 19:07:43 <^7heo> long story short 2017-03-16 19:08:13 TemptorSent: where did you see it? 2017-03-16 19:08:19 <^7heo> I won't curl http://github.com/jirutka/alpine-chroot.sh | sudo sh 2017-03-16 19:08:21 TemptorSent: some HP-UX? 2017-03-16 19:08:33 <^7heo> and it is very bad to ask people to do that. 2017-03-16 19:08:39 ^7heo: I do a bunch of work on embedded systems (no, not rpis, REAL embedded systems with a couple hundred K or so.... you wouldn't believe the hacks you see there.! 2017-03-16 19:08:49 definitely, this would give you HTML page… 2017-03-16 19:09:10 <^7heo> jirutka: aha. You got what I meant. 2017-03-16 19:09:20 and I’m not advising that 2017-03-16 19:09:30 <^7heo> TemptorSent: I would believe it. I graduated embedded development @uni 2017-03-16 19:09:35 jirutka: *lol* HP-UX used to be a nightmare for compatability, picking at random which flags to support for standard tools and sometimes adding their own that completely contradicted standard practice... 2017-03-16 19:09:39 <^7heo> jirutka: well, you are. 2017-03-16 19:09:43 and not using that, I’ve sent you example of my usage and it includes verification of checksum 2017-03-16 19:09:45 <^7heo> jirutka: wget