2016-06-01 04:45:30 Now that alpine 3.4 is released, does anyone know when we might expect a tag to be made for the alpine Docker image? 2016-06-01 05:35:16 Is it safe to upgrade from 3.3 to 3.4? I have quite a few apps that need PHP5 2016-06-01 06:03:03 nVitius: none of developers handles docker image, ask andyshinn 2016-06-01 06:03:10 Sachiru: should be 2016-06-01 06:03:22 there is a switch for a dry run so you can see what will happen 2016-06-01 06:03:34 Thanks. 2016-06-01 06:03:41 I'll make an image of the system, just in case. 2016-06-01 06:25:01 hi all 2016-06-01 06:33:06 hello 2016-06-01 06:36:02 Hello guys, is there any chance that you would port Alpine to powerpc (ppc) architecture? 2016-06-01 06:39:29 ODra: are you the IBM guy who was here a while ago? 2016-06-01 06:39:51 if not then, most likely 2016-06-01 06:40:27 there was some IBM representative here talking about it recently 2016-06-01 06:44:29 @Tsutsukakushi no, I'm didn't know that somone already asked the same Q. 2016-06-01 06:45:17 he didn't ask 2016-06-01 06:45:23 he asked what IBM could do to get it 2016-06-01 06:45:37 how they could help 2016-06-01 06:45:38 :3 2016-06-01 06:45:50 That would be great, when could we expect ppc port? : ) 2016-06-01 06:46:06 i don't know more than this :p 2016-06-01 06:47:10 Probably I don't have enough knowledge, either way I would gladly help you. 2016-06-01 06:47:36 OK, I see. 2016-06-01 06:48:59 ODra: we have no hardware, that's the first obstacle for porting Alpine to anything 2016-06-01 06:52:20 @barthalion Ah, what a pity. Well I'm just student and fan of this project and I do have only a single piece of PPC. 2016-06-01 06:54:10 Anyway someone offered Mac Mini Model A1103 https://bugs.alpinelinux.org/issues/3596 2016-06-01 07:06:52 morning 2016-06-01 07:15:26 Good morning! 2016-06-01 07:15:33 (or afternoon in my case) 2016-06-01 07:17:03 good mornoon 2016-06-01 07:44:20 ncopa: morning. http://tpaste.us/GPax 2016-06-01 07:45:35 I can't see tree/_scripts/generate_releases.lua in the alpine-mksite repo 2016-06-01 07:46:16 got it 2016-06-01 07:49:54 hmm 2016-06-01 07:49:54 (1/1) Replacing python3 (3.5.1-r3 -> 3.5.1-r3) 2016-06-01 07:49:54 ERROR: python3-3.5.1-r3: bin/2to3-3.3: no dirent in archive 2016-06-01 07:49:54 ERROR: python3-3.5.1-r3: BAD archive 2016-06-01 07:49:54 1 errors; 970 MiB in 229 packages 2016-06-01 07:49:56 on armhf 2016-06-01 08:33:48 :( 2016-06-01 08:36:22 ScrumpyJack: 2016-06-01 08:36:24 Applying: alpine-mksite/Alpine-3.4.0-released.md: Text cleanup 2016-06-01 08:36:24 error: posts/Alpine-3.4.0-released.md: does not match index 2016-06-01 08:36:24 Patch failed at 0001 alpine-mksite/Alpine-3.4.0-released.md: Text cleanup 2016-06-01 08:36:32 do you think you can rebase it? 2016-06-01 08:36:59 arg, i thought the patch looked dodgey 2016-06-01 08:40:06 ncopa: GNU make upstream seems to have a fix for the parralell build issue now, do you want me to upgrade GNU make to 4.2 with the upstream patch applied? 2016-06-01 08:46:28 nmeum: yes please 2016-06-01 08:57:10 ncopa: sorry about that. should work http://tpaste.us/GVk1 2016-06-01 08:57:54 error: posts/Alpine-3.4.0-released.md: does not match index 2016-06-01 08:57:55 Patch failed at 0001 alpine-mksite/posts/Alpine-3.4.0-released.md: text update 2016-06-01 08:57:55 The copy of the patch that failed is found in: .git/rebase-apply/patch 2016-06-01 08:58:06 ScrumpyJack: which branch do you use? 2016-06-01 08:58:08 master? 2016-06-01 08:58:21 here is what i suggest 2016-06-01 08:58:24 yes, that was on a fresh git clone 2016-06-01 08:58:42 how did you create the patch? 2016-06-01 08:58:45 git show? 2016-06-01 08:58:55 git format-patch 2016-06-01 08:59:08 git format-patch --stdout -1 | tpaste 2016-06-01 08:59:10 hm 2016-06-01 08:59:13 correct 2016-06-01 08:59:14 yip 2016-06-01 08:59:42 bah 2016-06-01 08:59:47 i think its my tree thats dirty 2016-06-01 09:00:16 http://tpaste.us/GmPJ 2016-06-01 09:00:32 you were right 2016-06-01 09:00:43 problem was at my side sorry 2016-06-01 09:00:53 hey np 2016-06-01 09:00:55 http://wwwtest.alpinelinux.org/posts/Alpine-3.4.0-released.html 2016-06-01 09:00:59 looks good? 2016-06-01 09:01:53 still looking 2016-06-01 09:02:54 ok cool. i can't see any more typos 2016-06-01 09:03:15 great. i merge it for production then 2016-06-01 09:03:31 thank you 2016-06-01 09:04:03 you ok with the raspberry pi bit? i think there is a userbase 2016-06-01 09:04:21 yes 2016-06-01 09:05:20 for generate_releases.lua, the desc are mouse overs, but i think they should be more obvious 2016-06-01 09:05:31 yes 2016-06-01 09:05:37 the downloads page can be better 2016-06-01 09:05:49 i am open for suggestions 2016-06-01 09:06:50 a table colomn between iso and size containing the text from desc 2016-06-01 09:07:17 can you build the site locally? 2016-06-01 09:07:35 http://git.alpinelinux.org/cgit/alpine-mksite/tree/README 2016-06-01 09:07:47 i think you may need apk add lua5.3 too 2016-06-01 09:25:35 ncopa: would have been nice if changes link to some kind of info page(s) 2016-06-01 09:26:01 clandmeter: how do you mean? 2016-06-01 09:26:02 Improved support for the Raspberry Pi ecosystem? 2016-06-01 09:26:09 bugs fixed mainly 2016-06-01 09:26:16 New iso image for virtual machines: alpine-virt. 2016-06-01 09:26:21 right 2016-06-01 09:26:27 a page with more info 2016-06-01 09:26:30 understand 2016-06-01 09:26:36 those 2 should have some more info imho 2016-06-01 09:26:48 i see what you mean 2016-06-01 09:29:37 nice to see the committers list is getting bigger. 2016-06-01 09:53:25 hey 2016-06-01 10:05:49 ncopa: a bit of fun, possibly not what you expected :) http://tpaste.us/AR65 2016-06-01 10:06:49 nice! 2016-06-01 10:06:50 i like that 2016-06-01 10:07:05 ncopa, how long does it take having offical docker 3.4 alpine image? 2016-06-01 10:08:55 https://github.com/gliderlabs/docker-alpine/pull/179 2016-06-01 10:10:32 gr8 2016-06-01 10:14:04 i can't look at the tables in download page at the moment, but would be cool to make those mouse overs more bovious 2016-06-01 10:14:14 s/bovious/obvious 2016-06-01 10:14:51 mouseovers does not show for tablet/mobile users 2016-06-01 10:28:26 or console users w3m/elinks/lynx etc :) 2016-06-01 10:29:25 change pushed to wwwtest 2016-06-01 10:37:32 someone wrote me an email about busybox vs suckless ... very interresting stuff 2016-06-01 10:37:43 i dont know who it was, but thanks for that email 2016-06-01 10:44:19 I'm trying to build something with clang -fsanitize=address but I get `err: /usr/bin/ld: cannot find /usr/bin/../lib/clang/3.6.2/lib/linux/libclang_rt.asan-x86_64.a: No such file or directory`. 2016-06-01 10:44:32 Next thing I tried to do is to build clang with abuild and include these files in the package. Would this be accepted? 2016-06-01 10:44:44 (as a patch) 2016-06-01 10:56:48 clang 3.6.2? 2016-06-01 10:58:11 ncopa: Yes. I'm just trying edge now in case I get a different result... 2016-06-01 11:04:35 ncopa: same on 3.8: `err: /usr/bin/ld: cannot find /usr/bin/../lib/clang/3.8.0/lib/linux/libclang_rt.asan-x86_64.a: No such file or director` 2016-06-01 11:05:01 I just built it manually with `abuild unpack`, ran cmake and make 2016-06-01 11:05:05 and I see no sign of the .a file. 2016-06-01 11:05:15 maybe the runtime lives in a different place? 2016-06-01 11:05:56 hm, under debian it lives in libclang-common-3.5-dev 2016-06-01 11:06:05 (or, 3.8) 2016-06-01 11:08:01 and getting the source for that package I get llvm-toolchain-snapshot-3.5 2016-06-01 11:11:34 I guess it's http://llvm.org/releases/3.6.2/compiler-rt-3.6.2.src.tar.xz which I need to build 2016-06-01 11:13:01 Aha, I see that in edge there is a compiler-rt package. 2016-06-01 11:13:52 sadly that does not appear to solve the problem 2016-06-01 11:15:19 Aha! -DCOMPILER_RT_BUILD_SANITIZERS=OFF http://git.alpinelinux.org/cgit/aports/tree/main/compiler-rt/APKBUILD?id=a86c04789aeec7cac077704f93423e34a3bfd638#n46 2016-06-01 11:15:52 ncopa: is there a reason for that to be off and any chance it could get switched on? Is it inline with the alpine philosophy?# 2016-06-01 11:17:33 Ah. I see that it's off because it breaks the build. http://pastebin.com/euNZQD5Q 2016-06-01 11:17:40 error: 'dlvsym' was not declared in this scope 2016-06-01 11:19:24 I see that dlvsym is a glibc non-posix thing. 2016-06-01 11:23:19 iri: i spent some days on it 2016-06-01 11:23:25 ah.. 2016-06-01 11:23:30 ACTION started going down the rabbithole 2016-06-01 11:23:32 basically, sanitizers does not work with musl 2016-06-01 11:23:37 ah :( 2016-06-01 11:23:47 please do, but i would recommend talk with #musl ppl 2016-06-01 11:23:58 interesting 2016-06-01 11:24:04 probably beyond the scope of my time budget sadly 2016-06-01 11:24:18 I'd love to see the sanitizers on alpine 2016-06-01 11:24:27 Is there any existing public discussion to be aware of? 2016-06-01 11:28:12 Most of the failures I've seen so far actually are guarded by #if !SANITIZER_ANDROID. Is there any leeway in pretending to be android, I wonder... 2016-06-01 11:29:17 hm, no. Then it starts trying to pull in android crap. 2016-06-01 11:31:20 I'm having a medium amount of success here. Looks like some of the bits and pieces are building. No idea if they work properly after this hackery though, and there are still roadblocks. 2016-06-01 11:39:26 so now I'm blocked on having a lack of sys/ustat.h, which isn't present in alpine (and is marked obsolete in favour of sys/statfs.h) but simply importing that doesn't help. 2016-06-01 11:40:32 http://pastebin.com/MvKbBjLS error: invalid application of 'sizeof' to incomplete type '__sanitizer::ustat' 2016-06-01 11:49:18 woot. I got it to build. 2016-06-01 11:49:24 thanks algitbot ;p 2016-06-01 11:49:39 I'm amused that a bot replied to that. 2016-06-01 11:51:24 algitbot: please behave 2016-06-01 11:51:52 ncopa: so my strategy was to take advantage of #if SANITIZE_ANDROID which turns off a lot of unsupported stuff. I added SANITIZE_MUSL. 2016-06-01 11:51:59 not sure if it works properly or not yet though. 2016-06-01 11:52:15 I can assume you got this far when you looked at it? 2016-06-01 11:52:28 i dont remember 2016-06-01 11:52:40 ncopa: was it a while ago? I guess the world may have moved on 2016-06-01 11:52:49 (SANITIZE_ANDROID might also be new) 2016-06-01 11:53:04 i suppose it could be turned into: #if defined(__linux__) && !defined(__GLIBC__) 2016-06-01 11:53:31 and turn it into SANITIZE_NONGNU 2016-06-01 11:53:33 or similar 2016-06-01 11:54:01 it was this year, a couple of months ago think 2016-06-01 11:54:41 some things I think we should do relatively soonish: 2016-06-01 11:54:50 - upgrade gcc to 6 2016-06-01 11:55:14 - replace libtool with sblibtool 2016-06-01 11:55:27 - migrate to libressl 2016-06-01 11:55:44 sblibtool? 2016-06-01 11:56:00 libtool replacement 2016-06-01 11:56:12 by midipix 2016-06-01 11:56:27 slibtool 2016-06-01 11:56:32 https://github.com/midipix-project/slibtool 2016-06-01 11:57:06 I see 2016-06-01 12:00:06 - use -Wformat -Werror=format-security in cflags by default 2016-06-01 12:00:10 woohoo, I now have the .a files I need installed via an APK. Let's see if the sanitizer works. 2016-06-01 12:00:20 iri: congrats! 2016-06-01 12:00:59 oh that's strange, the linker still doesn't see them 2016-06-01 12:01:15 oh it's looking in slightly the wrong place 2016-06-01 12:01:36 clang is looking at /usr/lib/clang/3.8.0/linux/libclang_rt.asan-x86_64.a but they got installed to /usr/lib/clang/3.8.0/libclang_rt.asan-x86_64.a 2016-06-01 12:01:44 (i.e, /linux/ is missing) 2016-06-01 12:02:24 sorry, /lib/linux/ 2016-06-01 12:02:31 (the first path I stated was missing /lib) 2016-06-01 12:03:24 ok.. now I'm missing a load of undefined references from the code which I patched out. I guess I need to do a more thorough analysis of where #if SANITIZE_ANDROID is used. 2016-06-01 12:03:35 Doesn't look horrendous. 2016-06-01 12:04:14 mention it in #musl, i think those guys have some experience with it 2016-06-01 12:09:14 have done. Is there a way to avoid `abuild` from deleting the src/ directory? I'm currently developing in there. Is there a better way of developing? 2016-06-01 12:09:23 (boy am I glad I exported the patch first..) 2016-06-01 12:10:44 I'm running `git diff` inside a docker container with `--terminal` but the pager shows lots of control characters instead of colour. Any ideas? 2016-06-01 12:11:03 `git diff --no-color` helps. 2016-06-01 12:20:33 iri: you have to use a pager that supports colors 2016-06-01 12:20:36 apk add most 2016-06-01 12:20:40 export PAGER=most 2016-06-01 12:20:45 then you have nice colors :) 2016-06-01 12:34:02 ah thanks leo-unglaub 2016-06-01 12:34:14 simple but not obvious :) 2016-06-01 12:35:14 uh, most appears to be missing from main/community 2016-06-01 12:35:51 `apk add less` was actually sufficient as it happens 2016-06-01 12:36:38 iri: its in testing 2016-06-01 12:45:50 down to one missing symbol! 2016-06-01 12:47:53 I'm loving ccache, gotta say. 2016-06-01 12:48:18 holy shit, the address sanitizer just worked 2016-06-01 12:48:29 amusingly, it spotted a fail inside ASAN itself. 2016-06-01 12:48:46 or rather, an assertion failure.. 2016-06-01 12:49:02 maybe I was too hasty, it's actually broken :) 2016-06-01 12:49:28 looking pretty good though, it gave me a backtrace complete with symbols and line numbers. 2016-06-01 12:51:16 in answer to my own question earlier: I can avoid deleting the src directory by using `apk rootpkg`. 2016-06-01 12:51:22 *`abuild rootpkg` 2016-06-01 12:56:17 OK! Now it ran hello world correctly. 2016-06-01 12:56:19 Woohoo. 2016-06-01 12:58:23 awesome, it produces an error report, complete with line numbers 2016-06-01 13:47:11 hello 2016-06-01 13:47:16 anyone still here 2016-06-01 13:48:11 ncopa: I had a medium amount of success. 2016-06-01 13:48:38 I got everything to build and even a few examples working. But then someone on #musl pointed out that this might never work because it's not possible to interpose internal malloc 2016-06-01 13:55:30 i am almost done getting alpine working on intel edison 2016-06-01 13:55:31 http://paste.ubuntu.com/16891902/ 2016-06-01 13:56:00 however i get Alpine Init 3.0.4-r2 mount: mounting shm on /dev/shm failed: Invalid argument 2016-06-01 13:56:02 and mount: mounting PARTUUID=333a128e-d3e3-b94d-92f4-d3ebd9b3224f on /sysroot failed: No such file or directory failed. initramfs emergency recovery shell launched. Type 'exit' to continue boot 2016-06-01 13:56:26 its in internal emmc ...is there anyway to make alpine look for apks in a partition mmc 0:5 2016-06-01 14:22:47 oneinsect: i'm not sure that's what you need 2016-06-01 14:39:02 there actually is 2016-06-01 14:39:08 and ncopa wrote me how 2016-06-01 14:39:42 ask someone to grep for my nick and 'rpi' in one sentence, it's somewhere there 2016-06-01 14:40:01 it was not long before 3.3.0 2016-06-01 14:40:14 (or grep yourself if you have archives) 2016-06-01 14:51:56 hi! sorry regarding adduser improvement fail. :( my only (and poor) line of defense is that I explicitly requested review and no one caught the problem until release happened. well, I could also somewhat blame terribleness of adduser/addgroup, but blaming doesn't improve a thing. I'll go later through all systemized adduser calls and enhance them with -G and prepend them with addgroup call. 2016-06-01 14:52:57 barthalion: he said apks, but he means apkovl? 2016-06-01 14:53:06 I believe so 2016-06-01 14:57:05 isn't there a file you can add paths to for nlplug-findfs to search for apkovls? 2016-06-01 14:58:23 .boot_repository ? 2016-06-01 14:58:39 maybe i'm mixing things up 2016-06-01 15:18:00 ScrumpyJack: 2016-06-01 15:18:02 hello 2016-06-01 15:18:21 did you check the pastebin? 2016-06-01 15:22:37 Initramfs unpacking failed: junk in compressed archive 2016-06-01 15:22:50 it says for initramfs-vanilla 2016-06-01 15:22:54 i wonder why 2016-06-01 21:40:26 today I’ve met a guy from BigClown – new startup that creates open-source electronics for DIY home automation projects – and told him about Alpine, that it would be great option for their RPi3 based hubs; he just sent me an email and he’s really excited about Alpine! 2016-06-02 05:46:28 Query: Is there any way to install Ruby 2.2.x on 3.4? 2016-06-02 05:56:25 Repository pinning or a rebuild 2016-06-02 06:33:20 I see. Thanks. 2016-06-02 07:31:48 morning 2016-06-02 07:56:39 anyone got an example of a recent APKBUILD file that creates mulitple sub packages? 2016-06-02 07:57:30 coreutils? 2016-06-02 07:57:36 php? 2016-06-02 07:57:49 php has multiple builds too 2016-06-02 08:00:33 php5 looks good. so... a function, with any variables that i want changed from main vars? 2016-06-02 08:08:56 is git server a bit overloaded ? 2016-06-02 08:09:34 seems to take a hell of a time to compress object on a git pull 2016-06-02 08:09:40 very very slow 2016-06-02 08:20:13 oom 2016-06-02 08:22:10 Jun 2 08:05:30 git daemon.err git-daemon[23796]: error: git-upload-pack died of signal 13 2016-06-02 08:30:25 ok 2016-06-02 08:49:08 ncopa: out of curiosity what's the specs of the srv ? 2016-06-02 08:52:56 48GB ram iirc 2016-06-02 08:53:13 49358592 2016-06-02 08:53:21 12 cores 2016-06-02 08:53:47 i think its dual hexacore 2016-06-02 09:16:11 ok 2016-06-02 09:26:44 pkgs.a.o has v3.4 added now. 2016-06-02 09:35:10 ncopa: would you agree with this? (fixes #5673) 2016-06-02 09:35:11 http://sprunge.us/KafO 2016-06-02 09:35:56 iptables modules directory: /usr/lib/xtables 2016-06-02 09:35:56 libc has setns: yes 2016-06-02 09:35:56 SELinux support: no 2016-06-02 09:35:56 ELF support: yes 2016-06-02 09:45:08 check what implementation of libelf they use too 2016-06-02 09:46:37 it's bpf (it this is what you mean) 2016-06-02 09:46:57 i have the feeling htat there are more than on e libelf 2016-06-02 09:47:36 arch linux's libelf is version 0.166 2016-06-02 09:47:38 documentation does not says much about that 2016-06-02 09:47:41 alpine is 0.8 something 2016-06-02 09:48:47 looks that busybox ip route is going to add elf support 2016-06-02 09:48:51 http://lists.busybox.net/pipermail/buildroot/2016-February/153307.html 2016-06-02 09:49:23 thats not busybox 2016-06-02 09:49:42 its iproute2 package for buildroot 2016-06-02 09:49:51 oh, ok 2016-06-02 09:49:53 sorry 2016-06-02 10:36:25 would there be any reason for vim to not - at least not completely - honor commands in .vimrc ? 2016-06-02 10:45:09 any suggestions for building multiple python packages from one source tarball? 2016-06-02 10:45:48 shall I make one APKBUILD file for each package? 2016-06-02 10:49:21 ScrumpyJack: subpackage maybe ? 2016-06-02 10:52:48 doesn't my function name needs to match my sub package name, which is py-something (with a hyphen)? 2016-06-02 10:54:06 no it doesn't :) 2016-06-02 11:07:49 ncopa: I've sent complementary fix to lighttpd to alpine-aports ML for now. later I'll prepare mass patch bringing status quo to 56 packages (or 55 if excluding unmaintained) regarding user's primary group. sorry for the mess. 2016-06-02 11:17:54 przemoc: i think i already fixed lighttpd 2016-06-02 11:19:17 only partially. I explained it in the commit message. 2016-06-02 11:21:10 it was explained also in #5666 my comment. you only added group, but did not make it a primary group as it was before my mass changes. 2016-06-02 11:23:44 but yes, your fix was good enough to make lighttpd work again 2016-06-02 11:24:00 hey :) 2016-06-02 11:24:16 hi 2016-06-02 11:35:58 do i have to be in a speciffic group to use pm-is-supported from pm-utils? 2016-06-02 12:45:32 is there an archive of debugging symbols for .apk packages? 2016-06-02 12:46:05 archive? 2016-06-02 12:46:22 for packages that are no longer on our mirror? 2016-06-02 12:47:34 no, for current packages. 2016-06-02 12:48:05 if there are not in the mirror, it means we currently dont have them. 2016-06-02 12:48:12 but we can add it if needed. 2016-06-02 12:48:50 some have a -dbg subpkg. 2016-06-02 12:48:58 do thy install with apk, or... 2016-06-02 12:49:30 sure 2016-06-02 12:49:35 like any other pkg 2016-06-02 12:50:11 leo-unglaub: pm-is-supported --suspend ; echo $? 2016-06-02 12:51:56 what would the package name be? 2016-06-02 12:52:36 (I'm not an alpine dev, I'm just trying to debug a crash that doesn't occur elsewhere) 2016-06-02 13:15:10 I can't find the debug symbols in the repos, and I don't see any mention of them in docs. Looking for the symbols for python-2.7.11-r3 2016-06-02 13:23:30 which repo? 2016-06-02 13:25:01 http://dl-cdn.alpinelinux.org/alpine/v3.3/main 2016-06-02 13:27:18 I can add it to edge if you like. 2016-06-02 13:30:46 not sure that it would help with the core file I have now. but might if he could replicate on edge instead of 3.3 2016-06-02 14:02:00 any python setuptools gurus lurking about? 2016-06-02 14:06:41 I’m not setuptools guru, but you can try me, maybe I’ll know the answer 2016-06-02 14:06:54 cool, let me pm you 2016-06-02 14:07:00 ok 2016-06-02 14:35:20 I am open to ideas how to make it super simple to build 3rd party repos 2016-06-02 14:37:04 I’ve described one approach here https://github.com/jirutka/aports#how-to-setup-your-own-repository 2016-06-02 14:37:04 it’s very simple, but it requires an HTTP server with SSH access 2016-06-02 14:38:43 the problem is that I didn’t find a simple way how to build packages on a machine that doesn’t have direct file-system access to the target directory with packages and apk index 2016-06-02 14:39:28 if we solve this issue, then it can be simplified a bit 2016-06-02 14:40:17 hm 2016-06-02 14:40:26 so locally build a single apk 2016-06-02 14:40:31 push to remote repo 2016-06-02 14:40:37 and update index on remote repo 2016-06-02 14:40:40 it may be interesting to use a service like https://bintray.com, but the integration for our use case seems to be very complicated 2016-06-02 14:40:45 exactly 2016-06-02 14:41:38 it could be done with 2 different signing keys 2016-06-02 14:41:53 could sign the locally built package with one key 2016-06-02 14:42:08 and sign the index on the remote with another 2016-06-02 14:42:24 you could need add both .pub keys to your target /etc/apk/keys/ 2016-06-02 14:45:44 is this a huge problem right now that people cannot build there own repos? 2016-06-02 14:45:53 because its not that complicated 2016-06-02 14:46:23 they already can build their own repo, but it’s maybe not so easy as it can be 2016-06-02 14:46:46 yes, what i want is making it simpler than it is 2016-06-02 14:47:34 do you want to go the launchpad route? 2016-06-02 14:47:49 i dont know what the launchpad route is :) 2016-06-02 14:48:08 but i want make it easy for third parties to create official precompiled binaries 2016-06-02 14:48:10 me neither… it’s somehow related to Ubuntu, isn’t it? 2016-06-02 14:48:25 uh, official precompiled binaries? 2016-06-02 14:48:32 for example 2016-06-02 14:48:41 you can get mysql.deb packages 2016-06-02 14:48:54 chromium.deb packages 2016-06-02 14:49:12 ubuntu launchpad -> they made it that every project has there own repo that you have to add to your source.list 2016-06-02 14:49:12 spotify.deb 2016-06-02 14:49:18 think about the reasons why they do their own packages instead of relaying on distribution 2016-06-02 14:49:45 because if you build an awesome product 2016-06-02 14:49:48 and you ship it 2016-06-02 14:49:55 hmmm, the only reason those companies provide precompiled binaries is because they want to screw with there users 2016-06-02 14:49:58 then you want people to start using it immediately 2016-06-02 14:50:00 IMO it’s just a nasty workaround, the way how to solve the problem of museum-like repos of Debian, RHEL etc. 2016-06-02 14:50:04 and force more shit at them 2016-06-02 14:50:06 not wait til next distro release 2016-06-02 14:50:16 > hmmm, the only reason those companies provide precompiled binaries is because they want to screw with there users 2016-06-02 14:50:16 exactly! 2016-06-02 14:50:35 > and force more shit at them 2016-06-02 14:50:35 totally my words! 2016-06-02 14:51:06 I really don’t think that this is the apparoch that we should support 2016-06-02 14:51:31 it has huge consequencies 2016-06-02 14:51:43 well 2016-06-02 14:51:50 look at this: https://github.com/docker-library/php/blob/master/7.0/alpine/Dockerfile 2016-06-02 14:52:08 some docker crap… 2016-06-02 14:52:15 the official docker image for php is built from source 2016-06-02 14:52:27 instead of using the distro package 2016-06-02 14:52:33 yes, the official docker image… exactly… 2016-06-02 14:53:17 what i'd like is, they build from source, but spit out an .apk, or a small repo 2016-06-02 14:53:27 ncopa: when you say "third parties", you are really saying "companies". Do I need to remind you why making it easy for companies to distribute their precompiled binaries is not such a good idea? 2016-06-02 14:53:29 when I use a distribution X, then the only “official” packages are from the distributuon X, not from a random vendors of some software… 2016-06-02 14:53:50 ^ 2016-06-02 14:53:55 ^ 2016-06-02 14:54:04 skarnet: the example at hand, php, only ships sources 2016-06-02 14:54:10 the docker approach there is horrible and extremely messy 2016-06-02 14:54:28 and about Docker… this is exactly one of the reasons why I hate Docker ecosystem… those so called official images full of crap 2016-06-02 14:54:30 i dont know any reason why i would use the docker image over the current official alpine package 2016-06-02 14:54:40 > the docker approach there is horrible and extremely messy 2016-06-02 14:54:40 exactly! 2016-06-02 14:55:05 well, they hired me to fix it 2016-06-02 14:55:23 It's practical to allow 3rd parties to ship precompiled binaries *if* they also ship the source and the binaries are entirely reproducible 2016-06-02 14:55:31 so i'm not thinking "if". I'm thinking "how" 2016-06-02 14:55:37 i.e. there's no possible way of bundling crapware 2016-06-02 14:55:45 which *they will do* if unchecked 2016-06-02 14:55:52 skarnet: some does 2016-06-02 14:55:59 i think nginx ships precompiled binaries 2016-06-02 14:56:11 I don't mind as long as it's reproducible 2016-06-02 14:56:17 but it HAS TO be 2016-06-02 14:56:19 mariadb ships precompiled binaries 2016-06-02 14:56:32 yes, but no one used them 2016-06-02 14:56:39 because there are bad 2016-06-02 14:57:10 i dont agree that no one uses them 2016-06-02 14:57:14 You gotta ask yourself: do I want to make a high-quality distro or do I want to be a redistributor for your average Internet company crap0? 2016-06-02 14:57:16 the only reason people use docker images in the way you want to do it is because the build procress of moders software is so horrible that no one knows how to build it anymore 2016-06-02 14:57:28 ^ 2016-06-02 14:57:52 there are a lot of reports where companies shipped there own containers that also did more than expected 2016-06-02 14:58:05 thats why stuff like docker is a nogo for security areas 2016-06-02 14:58:11 an alpine package is verifyable 2016-06-02 14:58:20 but the 3party container is not 2016-06-02 14:58:56 well 2016-06-02 14:59:05 people still want build their own package 2016-06-02 14:59:21 More than ever, the future of free software is threatened by companies. You know it. One of them is paying you, but that's not a good enough reason to entirely flip your coat. 2016-06-02 14:59:49 The danger isn't Microsoft and Apple anymore. The danger is companies pretending to do free software but bundling in their agenda. 2016-06-02 15:00:05 Distributions are the main line of defense. 2016-06-02 15:00:06 ^ exactly! RedHat 2016-06-02 15:00:13 skarnet: +1 2016-06-02 15:00:33 like debian? 2016-06-02 15:00:51 Yes, like Debian. Debian sucks for other reasons. 2016-06-02 15:01:19 still there are many 3rd party deb packages/repos 2016-06-02 15:01:49 mysql, mariadb, nodejs, nginx, ... 2016-06-02 15:01:52 those are "unofficial". And Official Debian won't take a hard stance on them because Debian is too big and disorganized. 2016-06-02 15:01:53 all from vendor 2016-06-02 15:02:09 yeah. Not an example to follow. 2016-06-02 15:02:33 reporting bugs on debian that are caused by an unofficial package relase a lot of anger, there are not happy about it 2016-06-02 15:02:49 and they een cought google on cheating on the chrome .deb package 2016-06-02 15:03:00 I came to Alpine because I couldn't trust Debian anymore. Don't make me go to Void for the same reason. 2016-06-02 15:03:05 du you guys remember the story of the hidden data loader in the deb? 2016-06-02 15:03:11 chrome.deb was a bad example 2016-06-02 15:03:56 i still thing there are valid reasons for making your own 3rd party repo 2016-06-02 15:04:13 what exactly is the problem? if vendor would like to have more up-to-date package, they always can contribute to the distribution; we have edge in Alpine, so users are not forced to wait until next release; the problem may be a compatibility, but I’m sure that there are other ways how to solve it 2016-06-02 15:04:29 of course there are. Doesn't mean you have to endorse the 3rd-party repo, or lift a finger to help it. 2016-06-02 15:04:39 If you do, then it better be verifiable. 2016-06-02 15:05:00 because if you do, it will involve Alpine's reputation. 2016-06-02 15:05:08 maybe we can add some repo to branch v3.x that will be updated more often and always built against v3.x, not against edge 2016-06-02 15:05:42 ncopa: i even think its important to have the posibility to have your own repo. i have my own repo that has one package in it that deploys my server configs to all alpine servers, but i would not have it a "3 clicks to your own repo" thing where all of the sudden we have 200 repos and people have to seach what will come from where on the system 2016-06-02 15:06:19 jirutka: the problem is that a software vendor tend to think that their app is the only in the universe. so they want backports 2016-06-02 15:06:34 i dont care if they maintain their set of backports 2016-06-02 15:06:39 ACTION plays the world's smallest violin for those vendors 2016-06-02 15:06:42 5 versions of php or whatever 2016-06-02 15:06:50 as long as i dont need to maintain it 2016-06-02 15:06:56 but if a company thinks that way fuck them and get rid of them 2016-06-02 15:07:16 You don't need to maintain it but if you distribute it you STILL need to check that the company isn't screwing users 2016-06-02 15:07:20 and THEY WILL DO IT 2016-06-02 15:07:49 yep ... they WILL DO IT ... they always do 2016-06-02 15:08:15 not always, there are counter-examples, but the majority will, sooner or later 2016-06-02 15:08:23 i'm ok with them distributing it too 2016-06-02 15:08:34 i dont want alpine distro to ship garbage 2016-06-02 15:08:39 and not even this, users will report bugs caused by these vendor craps to your bug tracker, because users often don’t know or even can’t know what exactly is wrong 2016-06-02 15:09:21 jirutka: they often cann't even know why it is failing because they just get a 200mb docker thing that is impossible to debug even for the best 2016-06-02 15:09:28 ^ support resources is another big point too 2016-06-02 15:09:45 leo: exactly 2016-06-02 15:09:48 "oh, it's a .apk that fails, so I guess I'll complain to Alpine devs" 2016-06-02 15:13:21 i mean i understand ncopa here to, docker is hired him because he has a great dristro and now they want him to solve there docker problem and he is not between camps ... the docker "i ship messy stuff" fraction and the clean and secure alpine fraction. combining them is not easy 2016-06-02 15:13:38 i fully understand him there that it is hard, but going that route is really dangerous! 2016-06-02 15:13:49 exactly 2016-06-02 15:13:50 and i would destroy the good reputation alpine has 2016-06-02 15:14:58 the challenge is to allow people do their messy stuff, while still providing the tools/guidance to do it clean and "correct" 2016-06-02 15:15:05 I understand ncopa’s point of view too, but I really don’t think that this is the only way how to solve that problem, moreover, I’m quite convinced that this is one of the worst, very dangerous way 2016-06-02 15:15:07 and at the same time protect alpine 2016-06-02 15:15:18 i mean 2016-06-02 15:15:32 you cannot prevent people from doing stupid things 2016-06-02 15:15:33 protecting Alpine should be your FIRST priority, that's the value you have to Docker too 2016-06-02 15:15:47 if they want a Docker lackey, they can find one easily 2016-06-02 15:15:50 i think thats the problem because is things get insecure its not docker that people remembeer ... i will be the "alpine security problem" 2016-06-02 15:16:00 what they want is the Alpine maintainer 2016-06-02 15:16:01 and then docker does not care and jumps to the next distro 2016-06-02 15:16:03 what i want to do is: make it easy to do it correct and make it hard to do it wrong 2016-06-02 15:16:30 ncopa: then don't cater to the messy stuff! 2016-06-02 15:16:40 best way to make it hard to do it wrong :) 2016-06-02 15:16:44 the reason i use alpine is because it always seamed to me that we all follow the same approach: if its not secure we are not doing it 2016-06-02 15:17:17 if someone would come to alpine and demand that we make it less secure we would all kick him out of the channel 2016-06-02 15:17:37 well, my boss but me in a big pile of mess and tell me: clean up! 2016-06-02 15:17:38 I want Adobe Flash and antiviruses packaged in Alpine! 2016-06-02 15:17:45 so thats what i'm doing 2016-06-02 15:17:58 skarnet: you actually have adobe flash :) 2016-06-02 15:18:00 ncopa: tell him that the only way to fix docker is: rm -rf / 2016-06-02 15:18:10 ^ :) 2016-06-02 15:18:32 ugh, we have Flash? *deletes Alpine* 2016-06-02 15:18:39 ncopa: you are an excelent dev, if he hires you as an expert he should listen to you and not just use your good name because alpine is secure and hipp currently 2016-06-02 15:18:56 docker does not care about a solution, they jump to the next distro as soon as it gets hip 2016-06-02 15:19:06 alpine will be here longer than docker 2016-06-02 15:19:08 skarnet: http://git.alpinelinux.org/cgit/aports/tree/non-free/adobe-flashplayer 2016-06-02 15:19:16 there are no precompiled binary though 2016-06-02 15:19:40 good :P 2016-06-02 15:20:36 tbh, cleaning up the mess means: forcing companies to provide reproducible, automatable recipes to build their packages and forcing them to publish them 2016-06-02 15:20:49 it's hard, but it puts the burden where it needs to be 2016-06-02 15:21:00 +1 2016-06-02 15:21:09 +1 2016-06-02 15:21:20 if the way to build the binary is available along with the binary, then there's no problem in allowing binaries to be distributed 2016-06-02 15:21:48 +1 2016-06-02 15:21:54 that is the case i have at hand 2016-06-02 15:23:00 i need to go 2016-06-02 15:23:08 thanks for you input 2016-06-02 15:23:23 skarnet: well, but this will not prevent companies to publish packages full of crap 2016-06-02 15:23:29 you are welcome :) 2016-06-02 15:23:58 i am still open to ideas how to make this simpler: https://github.com/jirutka/aports#how-to-setup-your-own-repository 2016-06-02 15:24:00 ncopa: yw, now just make sure to take it into account :) 2016-06-02 15:24:18 jirutka: a company can always distribute a crappy .apk but then its the users fault if he uses it and not alpines 2016-06-02 15:24:31 ^ 2016-06-02 15:25:05 you can't prevent anyone from publishing crap, what you can do is ensure people have a way to know that it's crap 2016-06-02 15:25:08 ubuntu had the same problem with launchpad, now every program got there own repository and people started shipping wired stuff and then all of the sudden someone put a lib in there that was not in the main repo and it all crashed 2016-06-02 15:27:58 skarnet: and always make it easy to avoid the garbage 2016-06-02 15:28:38 once thing more that i need to lift of my chest, re docker 2016-06-02 15:28:57 docker has been very nice with me - and alpine 2016-06-02 15:29:14 if it was not for docker there would likely not have been any v3.4 release 2016-06-02 15:29:38 well, at least not in May 2016-06-02 15:32:20 they have probably done more alpine than most ppl in alpine community realizes 2016-06-02 15:32:45 but they also got a lot out of it! they got the name and the security 2016-06-02 15:33:33 i think thats reasonable 2016-06-02 15:36:28 again, i am out of a job myself and i know how great it is to find one and i am really glad your found one at docker 2016-06-02 15:36:42 but you have to see that they got a lot more than you or alpine did 2016-06-02 15:36:57 even if they would pay you 10 000 a month they still would be the big winner 2016-06-02 15:37:33 the got the name, a working secure distro, the reputation, the press, ... and soooo much docker support in irc here that they did not hve to do themself 2016-06-02 15:37:49 so its not that they did this because they are soo nice and did not get anything in return 2016-06-02 15:40:23 yes. You can be thankful to Docker for employing you and allowing to work on Alpine, but don't be *too* thankful, because if it wasn't beneficial to them, they wouldn't have done it. 2016-06-02 16:02:15 ncopa: I don't know how much simpler it can be to run a custom 3rd party repo, its pretty trivial, the documentation is just lacking. 2016-06-02 16:02:25 The only one I think that is simpler is Arch... 2016-06-02 16:02:49 at least from what I've seen. 2016-06-02 16:03:10 Maybe having a list / submission / approval system for 3rd party repos (ppa?) 2016-06-02 16:23:31 hi, I have improved the api.a.o and updated the db with v3.4 pkgs 2016-06-02 16:23:31 pls free to check and let me know any new uri you are expecting. 2016-06-02 16:23:31 I have add/improved /flagged too ;) 2016-06-02 16:23:31 now working on filters, better error handlings 2016-06-02 16:23:56 code is at github.com/insteps/aport-api 2016-06-02 16:24:10 and congrats for v3.4 !! 2016-06-02 16:44:33 for fun one can do, 2016-06-02 16:44:35 new=$(curl api.a.o/aport-api/flagged/new | jq .data[].relationships.packages.links.self | sed -e 's/"//g'); curl $new | jq .data[].attributes.origin | uniq 2016-06-02 17:16:39 ncopa: that's lovely how this discussion escalated from technical topic to lecturing you what to do with project you're leading… 2016-06-02 17:17:12 ncopa: so while this is a bad idea in many ways, what if user could point abuild to a directory, without messing with build() and package() functions 2016-06-02 17:17:32 ncopa: and then it would just output an .apk file, with all dynamic dependencies detected? 2016-06-02 17:18:01 barthalion: point to directory of what? 2016-06-02 17:18:19 ncopa: that would solve the "spit out a package the easy way" part of the problem 2016-06-02 17:19:05 jirutka: of anything 2016-06-02 17:19:28 barthalion: I don’t understand how to magically build a package from _anything_ 2016-06-02 17:20:10 because you apparently never reverse engineered a package for distribution 2016-06-02 17:20:13 repacked, anything 2016-06-02 17:20:23 if you had done this, that would give you the general idea :) 2016-06-02 17:21:23 ncopa: wrt repo management, I actually find the way Alpine handles indices confusing; I like how Arch or Debian clearly draws a line between building package and handling repository by using different tools for either 2016-06-02 17:21:54 ncopa: so besides packaging, if I want a repository, I just do repo-add myrepo.db.tar.xz *.pkg.tar.xz and I'm done 2016-06-02 17:22:08 barthalion: the lecture means we care. You should appreciate that there are users who like what Alpine is today and don't want it to change. 2016-06-02 17:22:17 barthalion: sorry, but your very fuzzy… „what if user could point abuild to a directory“… this really doesn’t say anything… to a directory of my personal photos? or to a directory of APKBUILDs? to a directory of arbitrary sources files? 2016-06-02 17:23:11 *your → you’re 2016-06-02 17:23:32 jirutka: I don't care about APKBUILD in this scenario; it can be just a stub with pkgname and pkgver; directory of personal photos is a good example if someone wishes to have it packaged 2016-06-02 17:24:30 skarnet: well, that's none of my concern, as I never had a reason to tell ncopa what Alpine should be, before or after Docker hired him 2016-06-02 17:24:57 skarnet: and I'm one of the few people here that like when devel part of this channel has also some connection to the topic we talk about 2016-06-02 17:25:54 barthalion: this still doesn’t make damn sense; most of the software must be compiled, how can abuild know how to build an arbitrary software? 2016-06-02 17:26:26 "without messing with build() and package() functions" 2016-06-02 17:26:49 these functions aren’t there because it’s fun, there are kinda needed, aren’t they? 2016-06-02 17:27:21 you don't get my point and I don't really have time now to write little steps explanations… 2016-06-02 17:27:26 you skip the build part 2016-06-02 17:27:33 barthalion: ncopa is a big boy, he will do what he wants to do; but he's mostly getting the Docker perspective every day, so it's not necessarily a bad idea to also remind him the non-Docker, non-company, free-software-advocate perspective. 2016-06-02 17:27:54 yeah, exactly, I still don’t get your point and you’re refusing to explain it… 2016-06-02 17:28:02 Also, if you have a better idea for a channel where we could best discuss those issues, please share 2016-06-02 17:28:09 ^ exactly 2016-06-02 17:28:19 #alpine-devel sounded like a good place to do si 2016-06-02 17:28:21 so* 2016-06-02 17:29:02 sigh, nevermind, this is going offtopic again 2016-06-02 17:29:09 he decided to open this topic on IRC, maybe because he wants to know opinions inside a community… so what’s your proiblem? 2016-06-02 17:29:50 maybe we should have #alpine-philosophy 2016-06-02 17:30:25 jirutka: mess with Dockerfiles proves that people don't care about formal packaging process; if you give them possibility to skip writing entire APKBUILD, it will accomplish some of "ease" goal of ncopa 2016-06-02 17:30:29 anyway, ncopa pointed to some problem and suggested one solution, many of us don’t like this solution, so the best what we can do is to propose him a better solution… 2016-06-02 17:31:12 barthalion: yes, people don’t care about doing things right, we know that 2016-06-02 17:31:41 and anyway, I left my feedback, no problem on my side, I rarely keep track of all that huge backlog anyway 2016-06-02 17:31:43 have a nice evening 2016-06-02 17:32:03 barthalion: so your suggestion is to give them more guns for shooting into their feets? 2016-06-02 17:36:33 APKBUILD is very simple format, abuild makes it very easy (except requirement for security keys) to build and immediately install the package (abuilr -r -i), if it’s too hard for someone, maybe (s)he just should not create packages… 2016-06-02 17:37:33 figuring out how to actually build some software and what damn C dependencies it needs is much harder then writing few lines of shell script called APKBUILD 2016-06-02 17:50:58 isn't the bigger problem care? 2016-06-02 17:51:04 "this works for me" done! 2016-06-02 17:51:18 I've seen so many people viewing docker files as package management 2016-06-02 17:52:30 IMO the problem is that writing a package in mainstream distros like Debian or RedHat is really pain in ass, so most users doesn’t bother with it 2016-06-02 17:53:15 and the reason why they don’t use distro package is often that these mainstream distros ships too old versions 2016-06-02 17:53:54 except this is alpine 2016-06-02 17:53:58 exactly 2016-06-02 17:54:00 it has stable but lives on the edge 2016-06-02 17:54:24 but users comes from these shitty distros like Debian and tries to do things in the same bad way as before 2016-06-02 17:55:12 they writes crazy dockershits and don’t even think about writing an abuild 2016-06-02 17:55:32 because they don't have to provide maintenance 2016-06-02 17:55:42 not just this 2016-06-02 17:55:44 its fixed with a always building version 2016-06-02 17:55:54 you can write an abuild without pushing it into the official distro 2016-06-02 17:56:37 i'm aware 2016-06-02 17:56:45 i have my own customized alpine i maintain 2016-06-02 17:56:59 own repo and everything because git versions of packages are kinda frowned upon 2016-06-02 17:58:56 I’m fan of custom repositories like Gentoo Overlays, for personal usage 2016-06-02 18:00:13 however, it’s a two-edged sword 2016-06-02 18:02:41 if it’s significantly easier to maintain own repository and use someone else personal repository than contributing to the official distro repository, then it may lead to a hell of gazillion repositories and even so called “official” vendor’s repositories… 2016-06-03 06:39:49 morning. Happy Friday! 2016-06-03 07:57:52 would someone be kind enough to look at the AVR patches in patchwork? patches 2034 to 2038 2016-06-03 09:14:00 ScrumpyJack: I'm doing avr stuff now 2016-06-03 09:14:34 im doing patchwork stuff ;-) 2016-06-03 09:14:59 seems working now, so be prepered for an upgrade :) 2016-06-03 09:16:57 ok 2016-06-03 09:16:59 goody 2016-06-03 09:17:01 btw 2016-06-03 09:42:56 yay! 2016-06-03 09:50:04 can I request some patchwork for 2064 and 2065 as well ? :) 2016-06-03 09:50:23 I'm too lazy to setup my own repo :D 2016-06-03 10:05:05 %2064 2016-06-03 10:18:03 thanks ncopa 2016-06-03 10:20:37 np 2016-06-03 10:39:20 ncopa: want me resend lmdb ? 2016-06-03 10:39:32 would be nice thanks 2016-06-03 10:39:42 coredumb: maybe you can do: 2016-06-03 10:39:59 git format-patch -1 --stdout | tpaste 2016-06-03 10:40:03 and pate the url here 2016-06-03 10:40:35 ok wait 2016-06-03 10:44:17 http://tpaste.us/GZko 2016-06-03 10:44:20 ncopa: ^ 2016-06-03 10:57:24 Hello guys :) My name is Francesco and I'm trying to make an alpine distro which is capable of running inside an unpriviledged linux container 2016-06-03 10:57:45 I'd have a couple of questions before starting my research, would you help me? 2016-06-03 10:57:46 ncopa: I messed up a bit the dep for py3-msgpack, I'm sending an update right now 2016-06-03 10:57:57 cescus: feel free to ask 2016-06-03 10:58:09 Great! Thank you, ncopa :) 2016-06-03 10:58:24 unpriviledged? 2016-06-03 10:58:38 ACTION sucks air through his teeth 2016-06-03 10:58:54 ScrumpyJack: root inside container is not root for real 2016-06-03 10:58:55 only kidding :) 2016-06-03 10:59:02 how do you usually build your entire alpine distro (the one shipped inside the .iso)? 2016-06-03 10:59:20 :P 2016-06-03 10:59:48 using buildrepo from alpine-aports to build the entire repo(s) 2016-06-03 10:59:49 ncopa: i was being idiomatic and idiotic :) 2016-06-03 11:00:06 then alpine-iso to build the is 2016-06-03 11:00:09 the iso 2016-06-03 11:00:26 correction: buildrepo is from lua-aports 2016-06-03 11:00:53 http://git.alpinelinux.org/cgit/lua-aports/ 2016-06-03 11:01:40 nice! I'll take a look to that repo, ncopa :) 2016-06-03 11:02:02 Do I need some other stuff to build the distro? 2016-06-03 11:02:15 or just the basic toolchain? 2016-06-03 11:02:18 abuild 2016-06-03 11:02:20 apk 2016-06-03 11:02:39 http://git.alpinelinux.org/cgit/abuild/ 2016-06-03 11:03:00 http://git.alpinelinux.org/cgit/apk-tools/ 2016-06-03 11:03:16 hey friends :) 2016-06-03 11:03:16 Cool, thank you! 2016-06-03 11:03:24 how is everyone? 2016-06-03 11:03:30 some of the scripts are Lua scripts 2016-06-03 11:03:36 so you need lua 2016-06-03 11:03:37 very last question: does that procedure builds the kernel, too? 2016-06-03 11:03:44 yes 2016-06-03 11:04:06 http://git.alpinelinux.org/cgit/aports/tree/main/linux-grsec 2016-06-03 11:04:12 amazing, thank you very much for your willingness :) 2016-06-03 11:04:13 thats the recipy for the default kernel 2016-06-03 11:04:43 cescus: i suppose you were lucky today :) 2016-06-03 11:04:47 ncopa: %2075 2016-06-03 11:04:48 I'll mess for some time with these stuff and I'll make you know if something useful comes out :P 2016-06-03 11:04:56 sorry about that :$ 2016-06-03 11:05:11 skarnet: how good is your knowledge of dht's ? 2016-06-03 11:05:37 thank you, ncopa, have a nice day!! :D 2016-06-03 11:06:05 leo-unglaub: it's not looking good, since I have no idea what dht means ^^ 2016-06-03 11:06:38 (maybe I know the thing and not the TLA) 2016-06-03 11:07:29 distributed hash tables 2016-06-03 11:07:36 oh. 2016-06-03 11:08:51 well, not good. Network is complex - as I say, I'll do network when I'm done with the local machine. ;) 2016-06-03 11:09:12 i just installed Alpine on a latop here at the local hackerspace and the internet connection was very slow and downloading all packages took around 30 minutes ... but then i thought ... why download them again? i already have them on my laptop in the same network. why not make apk use dht as well and use the fastest path 2016-06-03 11:12:17 lbu? 2016-06-03 11:14:08 nah, no lbu 2016-06-03 11:14:19 apk cache? 2016-06-03 11:15:27 ncopa: my laptop was running alpine and i installed alpine on the laptop of another person. i would have setup the repos on my laptop first before installing it on his and then adding my laptop as a repos. that is a lot of work, with dht apk would have picked the fastest mirror by themself 2016-06-03 11:15:41 to my surprise i noticed that microsoft windows 10 is doing something similar 2016-06-03 11:15:52 they also ship there updates from normale computers as well 2016-06-03 11:15:58 in my opinion a good idea 2016-06-03 11:16:26 interesting idea indeed 2016-06-03 11:16:34 or: how to make your base system updater depend on a P2P infrastructure 2016-06-03 11:16:40 or fetching packages via bittorrent 2016-06-03 11:16:50 right 2016-06-03 11:16:53 p2p infra 2016-06-03 11:17:21 ncopa: yes, but torrent would require a tracker of some sorts, with dht it would work without one as far as i understand it 2016-06-03 11:17:42 also dht handles adding and removing of new data better than torrent 2016-06-03 11:18:00 you would need to have a bootstrap point of the p2p network 2016-06-03 11:19:28 a quote from wikipedia: Responsibility for maintaining the mapping from keys to values is distributed among the nodes, in such a way that a change in the set of participants causes a minimal amount of disruption. This allows a DHT to scale to extremely large numbers of nodes and to handle continual node arrivals, departures, and failures. 2016-06-03 11:19:47 i think that would be perfect for this case because not every computer is there all the time 2016-06-03 11:20:00 the discovery of new temporary nodes would be very quick 2016-06-03 11:20:21 i would asome that would take a lot of load of the mirrors 2016-06-03 11:20:33 and speed up downloads as well 2016-06-03 11:20:53 you had an experience with a slow network in which a cache was the ideal solution for you, you decided not to implement it because it would have been too much work, and instead you decided that DHT would have solved your problem? 2016-06-03 11:21:21 drugs are bad for you! 2016-06-03 11:22:18 :) 2016-06-03 11:26:46 lawl 2016-06-03 11:31:26 well, i am a sucker for everything decentralized :) 2016-06-03 11:31:35 also today is a free day for me ... 2016-06-03 11:31:41 so i am playing around with some code 2016-06-03 12:16:47 nice, zfs works again. 2016-06-03 12:17:22 ncopa: can we move it to main? 2016-06-03 12:17:48 ok 2016-06-03 12:18:20 i think i messed up one of the patches. 2016-06-03 12:20:56 tank 15.4G 0 15.4G 0% /mnt/tank :) 2016-06-03 13:00:10 shame we couldn't put "added zfs support" in the release notes for 3.4 :) 2016-06-03 13:06:51 ScrumpyJack: that was the idea 2016-06-03 13:07:00 i was in a rush, but couldnt make it work before release. 2016-06-03 13:30:19 <^7heo> Hey there. 2016-06-03 13:57:22 hi 2016-06-03 14:04:41 hi 2016-06-03 14:04:50 hi 2016-06-03 14:18:55 hi 2016-06-03 14:41:21 hi 2016-06-03 14:45:04 algibot: tell us, what’s the current state of Rust on Alpine? :P 2016-06-03 14:51:44 hatseflats 2016-06-03 14:51:50 zfs in main :) 2016-06-03 14:52:04 \o/ 2016-06-03 14:52:20 lets get us some lawyers :) 2016-06-03 14:53:07 hehe 2016-06-03 15:06:30 how do one disable ipv6 ? 2016-06-03 15:06:59 I've added net.ipv6.conf.all.disable_ipv6 = 1 2016-06-03 15:07:21 in my sysctl.d/00-alpine.conf config but seems like something is re-enabling it afterward 2016-06-03 15:11:26 seems like it's the kernel module loading that does that 2016-06-03 15:24:40 ncopa: quick question. Alpine -> libressl? 2016-06-03 15:29:45 i am asking because on that it will depend it i use http://man.openbsd.org/OpenBSD-current/man3/tls_init.3 or the old openssl api 2016-06-03 15:37:01 leo-unglaub1: yes. libressl is the current plan 2016-06-03 15:41:01 yay 2016-06-03 15:41:24 ugh 2016-06-03 15:41:54 ncopa: when will this come down the pipeline? 2016-06-03 15:42:12 i have to rebuild my stuff, and I don't know if it'll actually work - since libressl isn't backward compat fully 2016-06-03 15:42:20 ncopa: what was the reason again to not have a more hardened flavor of the grsec kernel ? 2016-06-03 15:42:53 seems like it would only be an additional config file in the kernel package right ? 2016-06-04 09:04:33 hey 2016-06-04 09:33:48 hmm, i found a bug in the php package 2016-06-04 09:33:59 php is not cleaning up the sessions on the server 2016-06-04 09:35:00 also i think we should change the default session directory from /tmp 2016-06-04 15:22:26 Hello. I would like to submit a diff on APKBUILD for the libxml2 package. How do I do it? 2016-06-04 15:23:02 is it the "patch workflow" i have to follow? 2016-06-04 15:54:11 oreo: so, there are a couple of ways 2016-06-04 15:54:27 oreo: you can put your patch on pastebin and leave it here 2016-06-04 15:54:38 oreo: you can send it to alpine-aports mailing list with git send-email 2016-06-04 15:54:44 oreo: or make a pull request on github 2016-06-04 15:54:49 oreo: it's up to you 2016-06-04 16:17:12 it's only a one-line change 2016-06-04 16:17:21 removal actually 2016-06-04 16:17:40 currently libxml2 has --disable-static, but it's useful to include the static libraries 2016-06-04 16:24:20 how so? I guess it's rather desired to have no stati c libraries 2016-06-04 16:24:40 I can disable it once I'm back home; you want to build something with static lib? 2016-06-04 16:25:44 yes 2016-06-04 16:26:17 a go app that i want to able to run on non-musl distros 2016-06-04 16:26:34 and it uses gokogiri, which itself uses libxml2 2016-06-04 16:30:13 also, libz includes a static version, so it doesn't seem like there is a global policy on static/dynamic packages 2016-06-04 16:30:45 i think it's preferable to include both whenever possible -- gives more flexibility to users 2016-06-04 20:01:11 [28750.528189] make[31603]: segfault at 0 ip 00007f635fbc4cb3 sp 00007ffccd9e1588 error 4 in ld-musl-x86_64.so.1[7f635fb76000+88000] 2016-06-04 20:01:33 does that ring a bell? chroot centos7 alpine 3.4 x86_64 2016-06-04 20:04:23 gcc -o hello hello.c works, it's a "make" issue :( 2016-06-04 20:04:36 make -n segfault too 2016-06-04 20:07:46 basic 2 lines Makefile: hello:hello.c\n\tgcc -o hello hello.c\n is enough 2016-06-04 20:16:40 there is no issue when I try on a virtual machine (as expected), it's only inside my chroot that I get a segfault 2016-06-04 20:18:16 never mind, /proc was not mounted ... 2016-06-05 00:14:46 hi all, I am trying to run alpine linux 3.4.0 as domU on xen 4.6.0 and I am getting this errors: `libxl: error: libxl_dom.c:37:libxl__domain_type: unable to get domain type for domid=35` `xl: unable to exec console client: No such file or directory` `libxl: error: libxl_exec.c:118:libxl_report_child_exitstatus: console child [14207] exited with error status 1`. thanks! http://lists.alpinelinux.org/alpine-user/0075.html#start75 2016-06-05 00:15:14 I been trying for 4 days, any help would be really appreciated! 2016-06-05 00:27:24 ironpillow: is xenconsoled running 2016-06-05 00:31:31 kaniini: yep 2016-06-05 00:31:51 I am able to console into other alpine-3.3.3 VM find 2016-06-05 00:31:54 *fine 2016-06-05 00:40:32 kaniini: the problem is with all of 3.4.0 version, regular, xen, and virt. And all of the 3.3.3 versions are fine. 2016-06-05 01:33:12 i'm getting an "unable to be delivered to the list" when trying to send to -devel 2016-06-05 05:22:54 hi all, I am trying to run alpine linux 3.4.0 as domU on xen 4.6.0 and I am getting this errors: `libxl: error: libxl_dom.c:37:libxl__domain_type: unable to get domain type for domid=35` `xl: unable to exec console client: No such file or directory` `libxl: error: libxl_exec.c:118:libxl_report_child_exitstatus: console child [14207] exited with error status 1`. thanks! http://lists.alpinelinux.org/alpine-user/0075.html#start75 2016-06-05 05:23:26 the problem is with all of 3.4.0 versions: regular, xen, and virt. And all of the 3.3.3 versions are fine. 2016-06-05 05:56:38 ironpillow: we got it the first time, but most Alpine devs are EU-based, so they'll probably answer during their daytime. Also it's Sunday, so don't be in a hurry. ;) 2016-06-05 07:12:33 skarnet: cool. thanks. just trying to find a solution for four days now. you are right, Sunday. I'll take a break too :) 2016-06-05 10:51:20 friends anyone here 2016-06-05 10:55:16 can someone please tell me how does alpine look for a partition containing apks folder when booting from an internal emmc 2016-06-05 23:45:13 Hello guys! This is Valdecir from São Paulo, Brazil. I would like to say thanks for the Alpine Linux project 2016-06-05 23:45:48 I would like to contribute to the project translating the documentation to Brazilian Portuguese. 2016-06-05 23:47:34 I'm not sure if here is the right place to ask this kind of question. If any of you could please point me the right direction, I will be more than happy to help 2016-06-06 07:26:08 Is there a problem with the repos today? I'm getting the following errors: http://pastebin.com/t05D4QLs 2016-06-06 07:33:15 Sachiru: did you try another mirror? 2016-06-06 07:33:51 Yes 2016-06-06 07:34:01 Other mirrors spit out errors with /community 2016-06-06 07:34:05 Or /main 2016-06-06 07:34:13 But pretty much all have errors 2016-06-06 07:34:23 Gonna try upgrading apk tools 2016-06-06 08:11:19 fabled: morning. could you look at #2066 when you have a moment? 2016-06-06 08:11:31 nope, that's not it 2016-06-06 08:11:41 this one http://patchwork.alpinelinux.org/patch/2066/ :) 2016-06-06 09:06:48 hey :) 2016-06-06 09:19:28 ho 2016-06-06 09:20:35 it's work from home we go 2016-06-06 09:34:13 I wish 2016-06-06 09:36:18 ncopa: there are a couple of xfce4 panel plugins that got released 2016-06-06 09:36:25 but i am not sure it we should upgrade them 2016-06-06 09:36:36 because they all ported to gtk3 and ... of course break all themes 2016-06-06 09:56:28 leo-unglaub: i saw. i think we probably need rebuild xfce with gtk3 2016-06-06 09:56:32 sooner or later 2016-06-06 09:58:22 yeah, and thats what i am afraid off 2016-06-06 09:58:31 i tryed some components locally against gtk3 2016-06-06 09:58:33 its horrible 2016-06-06 09:58:52 its around 10% slower, uses more memory and ... it looks ugly because all themes break 2016-06-06 09:59:17 with xfce going gtk3 i think the last usable linux desktop environment took a deep hit 2016-06-06 10:01:21 Query: Ran apk add ruby2.2 on Alpine Linux 3.4, how come running any sort of "gem " errors out with: -ash: gem: not found 2016-06-06 10:47:51 i found the issue that made syslinux not pass the cmdline options at boot prompt \o/ 2016-06-06 10:54:28 sounds great, ncopa! it should help checking stuff without rebuilding ISO. I still have to dig into #5479, but didn't have time for it this weekend. :( 2016-06-06 10:55:03 ncopa: \o/ is there a ticket on bugs.a.o which explains the impacts? or could you? 2016-06-06 11:07:19 stwa-: http://repo.or.cz/syslinux.git/patch/8dc6d758b564a1ccc44c3ae11f265d43628219ce is the patch. impact was simple and kind of severe: you couldn't alter cmdline, e.g. disable default quiet with "noquiet" 2016-06-06 11:09:30 przemoc: thanks, just asking because i was trying to get resume=… to work without success last week 2016-06-06 13:35:34 ncopa, does it work option makedepends in subpkg() function? 2016-06-06 13:36:30 no 2016-06-06 13:36:53 umh 2016-06-06 13:36:58 the subpkg() function is only executed after build() 2016-06-06 13:37:13 makedepends is needed before we start bduil() 2016-06-06 13:37:14 i need a subpkg who pick up the package-dev 2016-06-06 13:37:33 the only option is having a different package then. 2016-06-06 13:37:48 what? 2016-06-06 13:38:27 $package should create a subpackage with python bindings. But this python bindings needs $package-dev 2016-06-06 13:38:39 it's i2c-tools 2016-06-06 13:39:22 i2c-tools has python bindings called py-smbus, so i was wondering to use py-smbus as subpackage. 2016-06-06 13:39:26 so you add a py-* subpackage with depends="$pkgname-dev" 2016-06-06 13:39:34 it should work 2016-06-06 13:39:42 depends rather than makedepends 2016-06-06 13:39:51 correct 2016-06-06 13:39:58 its a runtime dependency, right? 2016-06-06 13:40:03 yes 2016-06-06 13:40:23 what odes it need from -dev? 2016-06-06 13:40:35 i2c-dev.h 2016-06-06 13:41:10 if its a single file, then it might be worth moving that file to i2c-tools main package 2016-06-06 13:41:28 so you dont need to pull in all the -dev stuff 2016-06-06 13:41:34 i dont know who big the -dev package is 2016-06-06 13:41:52 looks anyway that depends in subpkgdir is not honored. 2016-06-06 13:41:55 it sounds like i2c-dev.h becomes a runtime file 2016-06-06 13:42:20 it should work 2016-06-06 13:42:44 py() { 2016-06-06 13:42:44 cd "$_builddir"/py-smbus 2016-06-06 13:42:44 arch=all 2016-06-06 13:42:44 pkgdesc="$pkgname python bindings" 2016-06-06 13:42:44 depends="python-dev py-setuptools i2c-tools-dev" 2016-06-06 13:42:45 python setup.py build || return 1 2016-06-06 13:42:47 python setup.py install --prefix=/usr --root="$subpkgdir" || return 1 2016-06-06 13:42:49 } 2016-06-06 13:42:52 this is the subpkgdir 2016-06-06 13:43:20 you know what 2016-06-06 13:43:22 i've removed python-dev in the "main" part 2016-06-06 13:43:29 you should do that in build() 2016-06-06 13:43:31 not in py() 2016-06-06 13:43:37 oh 2016-06-06 13:43:54 then move the file in pu() 2016-06-06 13:43:56 *py() 2016-06-06 13:44:02 and python setup.py install.... should be done in package() 2016-06-06 13:44:27 and yes, finallly py() should just move the built file 2016-06-06 13:44:32 but still i need i2c-tools-dev. 2016-06-06 13:44:38 can i add that in makedepends? 2016-06-06 13:44:45 isn't a circle? 2016-06-06 13:44:48 <^7heo> isn't there a dev depends? 2016-06-06 13:45:35 you will probably need to set CFLAGS="-I$pkgdir/usr/incelu" or similar 2016-06-06 13:45:56 set the search path to a place it can find it 2016-06-06 13:46:22 <^7heo> oh wait, makedepends is the one using the dev packages. My bad. 2016-06-06 13:46:35 <^7heo> Sorry for the noise folks. 2016-06-06 14:12:06 ncopa: http://sprunge.us/iMdK 2016-06-06 14:12:16 this build correctly. 2016-06-06 14:12:35 But before commit, i'd like you review that 2016-06-06 14:12:45 + CFLAGS="$CFLAGS -I${_builddir}/include" 2016-06-06 14:13:04 i move that line to right over: + cd "$_builddir"/py-smbus 2016-06-06 14:13:16 or change it to: 2016-06-06 14:13:28 CFLAGS="$CFLAGS -I${_builddir}/include" python setup.py build || return 1 2016-06-06 14:13:42 with this: 2016-06-06 14:13:43 CFLAGS="$CFLAGS -I${_builddir}/include" python setup.py build || return 1 2016-06-06 14:13:45 build fails 2016-06-06 14:14:00 interesting 2016-06-06 14:14:07 nope 2016-06-06 14:14:13 i was wrong. 2016-06-06 14:14:28 it was failing before 'cause the CFLAGS path was wrong 2016-06-06 14:14:42 now build fins 2016-06-06 14:14:44 *fine 2016-06-06 14:14:53 i think thats better 2016-06-06 14:15:06 http://sprunge.us/WIUM 2016-06-06 14:15:09 like that ^ 2016-06-06 14:15:37 yes, but set pkgrel=2 2016-06-06 14:15:50 yes 2016-06-06 14:16:01 patch file also is useless 2016-06-06 14:16:01 tgtm 2016-06-06 14:16:04 ok 2016-06-06 14:16:08 thx 2016-06-06 14:16:18 lgtm* 2016-06-06 14:20:30 friends anyone are here 2016-06-06 14:20:53 can anyone please enlighten me on how alpine looks for apks in x86? 2016-06-06 14:21:04 apks folder? 2016-06-06 14:28:55 when and where? 2016-06-06 14:28:57 during boot? 2016-06-06 14:29:27 it wil try all block devices, which blkid says has a filesystem 2016-06-06 14:29:40 and it will for for a file .boot_repository 2016-06-06 15:11:01 hi great ncopa himself is here @ncopa: well this my scenario 2016-06-06 15:11:48 i am trying to get alpine running in Intel Edison, install u-boot and formatted the first partition to be fat32 however it doesnt seem to find it 2016-06-06 15:11:53 installed* 2016-06-06 15:12:22 i have put all files include apks folder from x86 iso and initramfs and modloop files including vmlinuz kernel 2016-06-06 15:13:12 is there a way we can ask alpine to look for apks folder at a particular partition or particular emmc by specifiying them in the bootargs 2016-06-06 15:14:24 does the partition needs to be a first fat32 partition? can it be a second partition or nth partition with ext4 filesystem? and it looks for a .boot_repository within the apks folder in the root directory? 2016-06-06 15:15:10 does the partition needs to be bootable flagged via fdisk??? 2016-06-06 15:15:48 please guide... this will help port alpine to edison ... i have already ported it to orange pi and nanopi and other alwinner boards 2016-06-06 15:16:43 i am preparing one more wiki entry on these lines https://wiki.alpinelinux.org/wiki/DIY_Fully_working_Alpine_Linux_for_Allwinner_and_Other_ARM_SOCs 2016-06-06 15:16:51 for edison and similar type boards 2016-06-06 15:42:21 <^7heo> jirutka: quid about PR 108? 2016-06-06 15:43:08 ^7heo: sorry, I’m quite busy now, I’ll respond to it later 2016-06-06 15:44:22 tl;dr; I have too look at the first thing, I don’t know what the problem is now; the second thing is not a bug, this is intentional 2016-06-06 15:44:27 <^7heo> jirutka: ok thanks :) 2016-06-06 15:44:49 <^7heo> jirutka: I understood that it is intentional, but I consider it to be faulty design. 2016-06-06 15:45:02 <^7heo> thanks for your answer, though :) 2016-06-06 15:45:04 do you have a better solution? 2016-06-06 15:45:23 <^7heo> yes, not failing on not having a package to build. Failing only when the package fails. 2016-06-06 15:45:52 <^7heo> unless I missed something, we don't need to fail if there is no package to be built. 2016-06-06 15:46:00 then if someone forgot to bump pkgrel or pkgver, then the build will silently succeed… 2016-06-06 15:46:20 <^7heo> Ok, that's another problem. 2016-06-06 15:46:26 99,9 % of the pull requests are for packages 2016-06-06 15:46:28 <^7heo> This has to be solved differently. 2016-06-06 15:46:47 <^7heo> Yes, but hardcoding that every pull request is for a package isn't the solution. 2016-06-06 15:46:58 I don’t think it’s a big deal when 0,1 % of PRs will fail because it doesn’t change any package, but something in .travis directory ;) 2016-06-06 15:47:10 <^7heo> Thanks for your explanation, I will edit my PR to solve that problem as well. 2016-06-06 15:47:21 <^7heo> But, at the same time... 2016-06-06 15:47:37 btw `local files="$(ls -Ap | sed '/\/$/d; s/$/\\|/' | tr -d '\n' | sed 's/\\|$//')"` this is VERY bad solution, never do that 2016-06-06 15:47:45 <^7heo> How can a package change AND git not return its diff? 2016-06-06 15:48:01 <^7heo> because if git returns a diff, then the package will be built. 2016-06-06 15:48:02 simply use find -type d 2016-06-06 15:48:37 no, if you don’t bump pkgrel or pkgver, then the package is not built IIRC 2016-06-06 15:49:05 but I’m not sure now, I’ll check it later 2016-06-06 15:49:35 <^7heo> find works too. It's posix, so you're right, it's better. 2016-06-06 15:51:00 moreover, `find -type d` is less cryptic and safe 2016-06-06 15:51:02 <^7heo> jirutka: I forgot to bump pkgrel in testing/gogs and it failed for that. 2016-06-06 15:51:11 and that’s right 2016-06-06 15:51:39 <^7heo> jirutka: it sure is less cryptic. I wrote that ls-sed-fu because I was focusing on something else; it's a honest mistake. 2016-06-06 15:51:44 it should fail in this case, because it may cause trouble if this is merged to upstream 2016-06-06 15:51:45 <^7heo> thanks for picking it up. 2016-06-06 15:52:06 np 2016-06-06 15:52:44 <^7heo> and btw 2016-06-06 15:53:41 <^7heo> it's so cryptic that you advise me to use find -type d while that expression does find -type f 2016-06-06 15:53:54 <^7heo> So indeed, you seem to be right ;) 2016-06-06 15:58:00 <^7heo> jirutka: btw, I didn't change the www-data dep 2016-06-06 15:58:04 <^7heo> I wonder why it didn't fail before. 2016-06-06 16:00:43 jirutka: not built package should be a success IMHO 2016-06-06 16:01:00 jirutka: packages with pkg{ver,rel} unchanged are just skipped on our infra 2016-06-06 16:04:33 barthalion: what is the point of sending patch that will be skipped? ;) 2016-06-06 16:04:57 bartahalion: in most cases it’s a mistake, not intention, so it’s better to fail build 2016-06-06 16:23:19 <^7heo> barthalion: that's my point. 2016-06-06 16:24:02 <^7heo> and jirutka you can fail this in a different way than hardcoding "PR is package diff" 2016-06-06 16:24:32 in what way? Travis/GitHub doesn’t have any other option 2016-06-06 16:24:44 the build is success, or failed, nothing between 2016-06-06 16:26:53 <^7heo> you don't get my point. 2016-06-06 16:27:19 <^7heo> we can fail the build, still; but not because there is a change that doesn't involve a package change. 2016-06-06 16:27:41 <^7heo> we can fail the build, because there is a change that involve a package change AND no package successfuly compiled. 2016-06-06 16:28:10 <^7heo> currently, the travis build is hardcoded so that any change that doesn't involve a package change fails. 2016-06-06 16:28:15 yes, sure, but that would be a little more complicated 2016-06-06 16:28:22 <^7heo> yes, and that is my intent. 2016-06-06 16:28:29 I’m not sure if it’s worth to additional complexity 2016-06-06 16:28:34 <^7heo> ?! 2016-06-06 16:28:51 <^7heo> First, I don't see that (made up?) complexity. 2016-06-06 16:29:03 <^7heo> Second, of course it is worth it not to hardcode edge cases. 2016-06-06 16:29:20 <^7heo> we're not writing a missile guidance system here. 2016-06-06 16:29:24 I dunno, I’m really busy with https://github.com/alpinelinux/aports/pull/112 and other things right now 2016-06-06 16:29:30 <^7heo> no worries. 2016-06-06 16:29:41 <^7heo> I will try to remember all the point you mentionned 2016-06-06 16:29:46 <^7heo> and comment/modify my PR accordingly. 2016-06-06 16:47:28 jirutka: style changes for example 2016-06-06 16:47:43 jirutka: or another change that has no effect on resulting package 2016-06-06 16:47:55 hm, that’s a good point 2016-06-06 16:48:08 sure, in majority of cases you're right 2016-06-06 16:48:24 but there is a possibility of legimate change with no pkgrel bump 2016-06-06 16:48:37 and it's easy to catch during review without failing the build 2016-06-06 16:49:05 when the build fail, then it brings your attention 2016-06-06 16:49:06 but you are maintaining travis-ci cfg for aports, so it's up to you :) 2016-06-06 16:49:15 well, not really 2016-06-06 16:49:42 <^7heo> jirutka: I assumed you've seen my README.me PR 2016-06-06 16:49:44 it doesn't bring more of my attention that finished build 2016-06-06 16:49:49 <^7heo> but it looks like you didn't ;) 2016-06-06 16:49:51 s/that/than/ 2016-06-06 16:49:58 I do review first 2016-06-06 16:50:01 then check if it builds 2016-06-06 16:50:03 not the other way 2016-06-06 16:50:40 and I don't bother to test patches from alpine-aports if they clearly don't change the package 2016-06-06 16:51:16 <^7heo> so, to make it clear: the PR 108 is because of the failing PR 107 2016-06-06 16:51:36 <^7heo> I closed it but essentially; it was failing because I didn't submit a package. 2016-06-06 16:51:46 <^7heo> And I didn't want to submit one, because the change was about the README.md file. 2016-06-06 16:52:07 how many of such PRs do we have? 2016-06-06 16:52:25 one per 1000 of build packages? 2016-06-06 16:52:26 <^7heo> no idea, but is that relevant? 2016-06-06 16:52:31 yes, it is 2016-06-06 16:52:35 <^7heo> no, it isn't. 2016-06-06 16:52:44 sorry, I really don’t have a time for that now 2016-06-06 16:52:48 I will discuss it later 2016-06-06 16:52:50 <^7heo> ok. 2016-06-06 16:57:27 jirutka: https://travis-ci.org/alpinelinux/aports/builds/135646856 here we go 2016-06-06 16:57:32 this is why it's simply wrong 2016-06-06 16:59:03 +1 2016-06-06 16:59:16 deleting package for example 2016-06-06 16:59:18 heh, you’ve merged broken package into upstream and now you’re telling me that travis build script is wrong? funny :) 2016-06-06 16:59:24 sigh 2016-06-06 16:59:29 you don't even look at git log 2016-06-06 16:59:34 or the diff 2016-06-06 16:59:41 it we me who pushed it 2016-06-06 16:59:47 Revert "testing/aspcud: new aport" 2016-06-06 16:59:48 barthalion: reverted it 2016-06-06 16:59:58 reverted package is not a build failure 2016-06-06 17:00:03 <^7heo> jirutka: ypi 2016-06-06 17:00:06 <^7heo> oops 2016-06-06 17:00:13 however, you’re right that it definitely should not fail when deleting an abuild 2016-06-06 17:00:15 <^7heo> jirutka: you're too busy/tired to be objective. 2016-06-06 17:00:25 <^7heo> I propose to delay this discussion before it becomes a troll. 2016-06-06 17:00:42 jirutka: and cut that "funny" passively aggressive comment please 2016-06-06 17:00:48 <^7heo> +1 2016-06-06 17:00:49 yes, because I’ve spent two days trying to build Julia on Alpine and it still fails on Illegal instruction and I don’t know why :( 2016-06-06 17:00:51 this is now how we work here 2016-06-06 17:01:05 there could be commits that are typo fixes in comments 2016-06-06 17:01:15 <^7heo> jirutka: that's fine, we're all failing on stuff. But let's keep this community nice and making progress ;) 2016-06-06 17:01:21 <^7heo> +1 ncopa 2016-06-06 17:01:46 <^7heo> jirutka: I think I did a fair job with fixing the segfault issue in the travis-ci linked above in 108 2016-06-06 17:01:50 you surely can send a PR and fix it, but silently passing a build when someone forget to bump pkgrel or pkgver isn’t a good solution IMO 2016-06-06 17:02:02 <^7heo> if I will fix that horrible ls command I used instead of find -type f 2016-06-06 17:02:03 don't merge if drunk… 2016-06-06 17:02:06 <^7heo> as you suggested. 2016-06-06 17:02:21 <^7heo> barthalion: unfortunately, there's no BAC probe for github. 2016-06-06 17:02:26 because if nothing is actually build, then it doesn’t bring any value 2016-06-06 17:02:29 <^7heo> maybe we should get that? 2016-06-06 17:02:31 wat 2016-06-06 17:02:40 so code quality os not a value? 2016-06-06 17:02:42 <^7heo> jirutka: if nothing is actually built, it means that the tests are not covering enough code. 2016-06-06 17:02:44 s/os/is/ 2016-06-06 17:03:02 <^7heo> jirutka: don't assume that your code coverage is 100%... it's FAR from that. 2016-06-06 17:03:15 you sometimes sound like you are trying to ignore all examples people tell you 2016-06-06 17:03:17 code quality is definitely a value, that’s not what i said 2016-06-06 17:03:24 there is always some but 2016-06-06 17:03:45 <^7heo> But hey, guys... jirutka is tired and we're all arguing against his opinion. 2016-06-06 17:04:00 <^7heo> I've been in that position too; it's normal he becomes a tad passive/aggressive. 2016-06-06 17:04:10 <^7heo> again, I propose we delay this discussion, if we're all okay with that. 2016-06-06 17:04:26 sure, but when he's not, I'm most likely at work, where I can't do two things at once 2016-06-06 17:04:27 <^7heo> Let's use the github comments to have persistency 2016-06-06 17:04:39 what if someone fix a typo in APKBUILD and maybe accidently miss a parenthesis? he don’t bump pkgver/pkgrel, so Travis will not build it, yet say that build succeeded… 2016-06-06 17:04:43 <^7heo> so we can get rid of that synchronization requirement. 2016-06-06 17:04:49 jirutka: that's life 2016-06-06 17:04:56 our builders will catch it 2016-06-06 17:05:04 <^7heo> jirutka: that's why PR exist. 2016-06-06 17:05:17 <^7heo> if we didn't want to review code, we would grant write rights to @. 2016-06-06 17:05:21 Travis is for help us with review 2016-06-06 17:05:35 <^7heo> Travis is to catch the obvious. 2016-06-06 17:05:35 it doesn’t push to production, it doesn’t force you to not merge something if build fail 2016-06-06 17:05:54 sure, and it spams my inbox because I reverted bad change 2016-06-06 17:05:54 <^7heo> jirutka: again, I have the impression that you're not reading all we write. 2016-06-06 17:05:59 I am 2016-06-06 17:06:00 but it suddenly knows better what's good 2016-06-06 17:06:08 <^7heo> jirutka: it's fine, you already said that you're busy. 2016-06-06 17:06:13 I agree that we should fix it, e.g. to not fail when abuild is deleted 2016-06-06 17:06:21 <^7heo> barthalion: let's use the Github comments to discuss that? 2016-06-06 17:06:23 <^7heo> I know I will. 2016-06-06 17:06:34 <^7heo> IRC is only gonna be too much trouble atm. 2016-06-06 17:06:35 does 'abuild sanitycheck' source APKBUILD? 2016-06-06 17:07:11 it goes 2016-06-06 17:07:13 does even 2016-06-06 17:07:42 jirutka: so what about running just 'abuild sanitycheck' if pkg{rel,ver} is the same? 2016-06-06 17:08:00 and just exit 0 if package has been deleted in given commit? 2016-06-06 17:08:14 that’s a good idea 2016-06-06 17:09:24 and same for no APKBUILD in list of changes files, exit 0 2016-06-06 17:09:52 <^7heo> also, jirutka, I think you missed the point of my PR. 2016-06-06 17:10:12 <^7heo> my PR won't fail if there is no package changed, but if there is one; and pkg{rel,ver} hasn't been bumped, it will fail. 2016-06-06 17:10:18 <^7heo> as desired. 2016-06-06 17:51:53 <^7heo> I'm not sure if replacing `ls -Ap | sed '/\/$/d'` by `find -type f -maxdepth 1 | sed 's:^\./::'` makes the code any better... 2016-06-06 17:51:56 <^7heo> but it's done. 2016-06-06 18:00:17 hi all, I am trying to run alpine linux 3.4.0 as domU on xen 4.6.0 and I am getting this errors: `libxl: error: libxl_dom.c:37:libxl__domain_type: unable to get domain type for domid=35` `xl: unable to exec console client: No such file or directory` `libxl: error: libxl_exec.c:118:libxl_report_child_exitstatus: console child [14207] exited with error status 1`. thanks! http://lists.alpinelinux.org/alpine-user/0075.html#start75 2016-06-06 18:01:23 I am sorry to post this many times. But I've been trying to solve this for 5 days now and am not able to see what is actually going wrong. thanks, really appreciate the help. 2016-06-06 18:02:03 ironpill_: been busy sorry 2016-06-06 18:02:19 ncopa: no problem :) 2016-06-06 18:04:00 files I am using and `xl create` log: https://gist.github.com/SunilObj/4c32f1007301b896d3127bd1316a283b 2016-06-06 18:07:48 <^7heo> barthalion, jirutka: I hope that the comments I left in the code of 108 now make it clear. 2016-06-06 18:09:59 ironpill_: did the same config work in xen 4.5? 2016-06-06 18:11:44 ncopa: I did not check with xen 4.5. I started using xen recently so started with 4.6. But alpine 3.3.3 is working on xen 4.6. 2016-06-06 18:12:02 ok 2016-06-06 18:12:31 and alpine 3.4 does not 2016-06-06 18:13:06 yeah. alpine 3.4, alpine-xen 3.4 and alpine-virt 3.4 all do not work 2016-06-06 18:13:21 while all alpines (3.3.3) work 2016-06-06 18:14:20 i should test it on my old xen 4.5 box 2016-06-06 18:14:32 its on my todo list 2016-06-06 18:15:02 cool 2016-06-06 18:17:40 let me know if you need me to do more tests. I am running xen 4.6, debian jessie as dom0, alpine 3.3.3 as domU1 and trying to run alpine 3.4 as domU2. x86_64 intel processor. 2016-06-06 18:52:47 sigh, obviously libspotify is linked against glibc 2016-06-06 18:54:07 there is no enough sighs I can make 2016-06-06 18:55:22 Try facepalms 2016-06-06 18:55:45 they're the advanced version of sighs 2016-06-06 19:01:01 now I'm not sure if I should do the blasphemy of this https://github.com/sgerrand/alpine-pkg-glibc – but built for armv6, or just use glibc based distro… 2016-06-06 19:01:51 barthalion: did you try the libc-compat package? 2016-06-06 19:02:06 I didn't know we have one 2016-06-06 19:02:19 adobe flash is linked against glibc but works (kind of) with musl 2016-06-06 19:02:35 musl is somewhat ABI compatible with glibc 2016-06-06 19:02:47 <^7heo> isnt' the musl ABI more or less equivalent to the glibc one? 2016-06-06 19:02:56 <^7heo> (I have no idea, I just read something about this some time ago) 2016-06-06 19:02:59 so you only need symlink to musl libc.so 2016-06-06 19:03:10 <^7heo> yeah, that % 2016-06-06 19:03:13 <^7heo> that ^* 2016-06-06 19:03:24 yes they try be abi compat with glibc, at least partially 2016-06-06 19:03:27 do we have package for that? 2016-06-06 19:03:30 yes 2016-06-06 19:03:33 libc-compat 2016-06-06 19:03:40 <^7heo> yeah, the musl libc is cool for that. 2016-06-06 19:03:51 why I don't see it in aports :( 2016-06-06 19:03:51 libc6-compat 2016-06-06 19:03:51 <^7heo> it's like the glibc, but good. 2016-06-06 19:04:03 not everything works 2016-06-06 19:04:04 libc6-compat, that's the reason 2016-06-06 19:04:10 awesome, will check it out 2016-06-06 19:04:30 for example the binary .so for netflix support in chromium does not work 2016-06-06 19:05:10 if you do ldd /path/to/libspotify.so 2016-06-06 19:05:29 you'll know if musl has all symbols it needs 2016-06-06 19:09:41 thanks 2016-06-06 19:22:44 skarnet, are you here? 2016-06-06 19:32:04 I am, what's up? 2016-06-06 19:32:58 jirutka ^ 2016-06-06 19:33:31 skarnet: do you have any idea what can cause Illegal instruction in https://github.com/alpinelinux/aports/pull/112#issuecomment-224059497 ? 2016-06-06 19:34:36 illegal instruction is pretty bad. It generally means you're running on an arch that doesn't support all the instructions your binary is using. 2016-06-06 19:35:00 I’ve checked julia libs and dependencies with https://github.com/alexxy/analyze-x86 and it seems to be okay, except libcrypto that contains 3dnow! (not available on Intel), but libcrypto is used in many packages and there’s no problem, so it’s probablyl not the cause 2016-06-06 19:35:11 julia itself and most of the deps are build on the same machine 2016-06-06 19:35:17 jirutka: do you know if gitlab-runner instance should be accessible from Internet or is it fine to have it behind NAT? 2016-06-06 19:35:43 barhalion: it SHOULD be behind firewall (or NAT…) 2016-06-06 19:36:10 well, sure, I'm not asking for sane design, I want to know what they did 2016-06-06 19:36:21 jirutka: if no other package is using 3dnow! instructions but julia happens to be for some reason, then that might very well be it 2016-06-06 19:36:24 but that's what I wanted to know, thanks 2016-06-06 19:36:36 I understand, it should work without any problem behind NAT 2016-06-06 19:37:34 of course gitlab-runner must be able to connect to GitLab and vice versa 2016-06-06 19:37:44 yeah 2016-06-06 19:37:46 wrt julia, you said "most of the deps"… did you try to build them all on one machine 2016-06-06 19:38:15 not yet, I can’t build llvm on my machine, it fails on something (don’t have logs right now) 2016-06-06 19:41:06 is there some way how to find what exact instruction it tries to use? 2016-06-06 19:41:41 when I try to build llvm: 2016-06-06 19:41:41 ``` 2016-06-06 19:41:41 [ 13%] Built target LLVMDebugInfoPDB 2016-06-06 19:41:41 [ 13%] Linking CXX static library ../libLLVMMC.a 2016-06-06 19:41:42 [ 13%] Built target LLVMMC 2016-06-06 19:41:42 make: *** [Makefile:150: all] Error 2 2016-06-06 19:41:43 >>> ERROR: llvm: all failed 2016-06-06 19:41:43 ``` 2016-06-06 19:41:56 jirutka: yes there is, but I don't have the knowledge for that. Last time I had to do that, nsz walked me step by step holding my hand, and I don't remember all the steps. 2016-06-06 19:42:07 :) 2016-06-06 19:42:15 You could try to find him on #musl and ask. It may work with you too. :) 2016-06-06 20:03:58 sigh, llvm fails to build even on different machine, with totally different error… 2016-06-06 20:04:23 why is llvm always so problematic? 2016-06-06 21:45:33 omfg can someone just destroy all the GNU code?! 2016-06-06 21:45:50 especially these GNU specific extensions that breaks compatibility everywhere 2016-06-06 21:47:00 find me a flamethrower and I'll gladly do it for you 2016-06-06 21:59:49 heh, julia fails on instruction UD2… wtf XD 2016-06-06 22:00:19 something intentionally generates undefined instruction? o.O 2016-06-06 22:03:29 Hi there, and thank you for this wonderful linux distro. I am a python developer and was pleasantly surprised there is some support for binary bindings (eg: py-psycopg2) 2016-06-06 22:05:14 There is already python3 (3.5.1) but couldnt see any suport for bindings. I am primarely interested in gevent. 2016-06-06 22:08:36 Maybe I can help to make some of this packages available? Any idea who should I chat with? 2016-06-06 22:08:49 jirutka: if UB is detected, UD2 can be generated. 2016-06-06 22:09:28 mzdaniel: ncopa but also this https://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package 2016-06-06 22:09:42 not sure whether julia has her own set of UBs 2016-06-06 22:10:03 UBs? 2016-06-06 22:10:09 undefined behavior 2016-06-06 22:10:14 aha 2016-06-06 22:10:20 not sure whether you've seen my previous message 2016-06-06 22:10:24 jirutka: if UB is detected, UD2 can be generated. 2016-06-06 22:11:08 I haven’t 2016-06-06 22:11:33 clang was first doing so, but apparently gcc followed suit 2016-06-06 22:11:36 mzdaniel: we are currently discussing how to add support for Python 3 modules 2016-06-06 22:12:51 Thank you Adran. I was a little hesitant to contact him directly. 2016-06-06 22:15:56 hi jirutka1, yes I saw in the mailing list briefly about python3. Is the idea to make python3 the default python? 2016-06-06 22:16:06 would hope so 2016-06-06 22:16:14 considering that python2 is eol soon enough 2016-06-06 22:16:38 2020 is not that soon, I would say 2016-06-06 22:17:51 i would say that Python 3.x has been released 8 years ago… 2016-06-06 22:18:20 it’s really a shame that it takes so extremely long for Python community to finally migrate 2016-06-06 22:18:48 red hat 2016-06-06 22:18:58 Gentoo and Arch Linux already uses Python 3 as default 2016-06-06 22:19:06 and also Fedora 2016-06-06 22:19:12 yeah, RedHat 2016-06-06 22:21:45 python3 would be great to be the default. Though renaming packages is not exactly fun. Wouldn't py3-package_name be enough to start with. 2016-06-06 22:23:40 it’s more complicated… most of the python packages are compatible with py2 and py3, so we should make two subpackages: py2-pkgname and py3-pkgname 2016-06-06 22:25:12 Sure. I was thinking specifically for bindings. 2016-06-06 22:29:30 I'll check the package page and give it a try. Thank you Adran & jirutka1! 2016-06-06 22:29:47 you’re welcome :) 2016-06-06 22:43:01 could please someone look at https://github.com/alpinelinux/aports/pull/110 ? is adding `options="libtool"` to this abuild okay? 2016-06-06 22:43:05 uh, the default is something I did not mention 2016-06-06 22:43:34 I see no reason not to follow PEP394 2016-06-06 22:43:47 which says that python should link to python2 2016-06-06 22:43:54 yes 2016-06-06 22:44:29 I'm long enough with Arch to remember python3 migration 2016-06-06 22:44:41 we still have plenty of find+sed combinations to fix shebangs 2016-06-06 22:44:47 I'd like to avoid it here 2016-06-06 22:44:59 that PEP was specifically written after Arch did migration… 2016-06-06 22:45:22 agree, one friend of mine works with Python in Fedora and he complained a lot about this decision in Arch 2016-06-06 22:45:49 that was a mistake, but no one expected it python3 adoption will take so long… 2016-06-06 22:45:55 jirutka: wrt that patch, it makes sense to me 2016-06-06 22:46:20 what exactly does that this option do? 2016-06-06 22:46:33 prevents *.la removal 2016-06-06 22:46:54 why are by default removed? 2016-06-06 22:47:21 because most likely there is no need to keep them 2016-06-06 22:48:11 these are useful during build as far as I know 2016-06-06 22:48:24 okay, thanks for explanation 2016-06-06 22:49:19 https://blog.flameeyes.eu/2008/04/what-about-those-la-files looks like answer is not that simple 2016-06-06 22:50:11 hmm, it seems that Travis already builds even modified abuilds without bumped pkgver/pkgrel… I thought that `ap builddirs` ignores these 2016-06-06 22:55:47 what the f… julia-debug works! but julia (release) does not 2016-06-07 06:36:39 Error relocating libspotify.so.12.1.51: backtrace: symbol not found 2016-06-07 06:36:41 Error relocating libspotify.so.12.1.51: backtrace_symbols: symbol not found 2016-06-07 06:36:43 :( 2016-06-07 06:43:58 anyone managed to run alpine arm in qemu? 2016-06-07 07:09:12 ncopa, for appstream-glib upgrade a new lib is needed, libgcab 2016-06-07 07:09:28 i've built it and fixed appstream-glib 2016-06-07 07:09:37 but libgcab should be pushed directly to main 2016-06-07 07:09:46 thats ok 2016-06-07 07:09:54 ok, thanks. 2016-06-07 07:10:01 Before commit i'll send you the diff 2016-06-07 07:14:17 ncopa, this is libgcab: http://sprunge.us/faSV 2016-06-07 07:14:22 and this is appstream-glib: http://sprunge.us/jOWe 2016-06-07 07:20:52 pushit 2016-06-07 08:22:52 ncopa: so what I did with my libspotify issue: I made a shared library with backtrace and backtrace_symbols stubs and linked it against musl 2016-06-07 08:22:56 everything seems to be happy 2016-06-07 08:23:04 algitbot: gimme a cookie 2016-06-07 08:23:42 ACTION blink 2016-06-07 08:24:04 algitbot has run out of cookies 2016-06-07 08:25:34 ha! 2016-06-07 08:25:49 barthalion: thats great 2016-06-07 08:26:21 i wonder if we should hack the libc6-compat libs to do the same 2016-06-07 08:26:58 there were a few symbols for glibc fortify implementation __stack_chk something 2016-06-07 08:27:53 that was needed for libwidevinecdm*.so (for netflix support in chromium) 2016-06-07 08:27:58 http://www.openwall.com/lists/musl/2016/02/19/2 2016-06-07 08:30:47 barthalion: what app do you use wants to use libspotify? 2016-06-07 08:46:27 ncopa: mopidy-spotify 2016-06-07 08:46:40 which uses pyspotify, which loads libspotify via ffi 2016-06-07 08:47:20 ncopa: yeah, stubs like that would make sense to me in libc6-compat 2016-06-07 10:13:46 ncopa: could you please merge https://github.com/alpinelinux/aports/pull/117 (it’s for the main repo) and backport it to v3.4 (#5688)? 2016-06-07 10:51:53 hey :) 2016-06-07 12:13:55 hi leo 2016-06-07 12:50:41 jirutka: ping wrt. the prayer aport 2016-06-07 12:51:22 Should I make the suggested changes another commit, or should I make a new PR for it? 2016-06-07 12:51:34 CcxCZ: please change the commit 2016-06-07 12:51:57 CcxCZ: btw CZ in your nickname means Czech Republic? 2016-06-07 12:52:45 Yes. We've seen each other few times I think. I was at the Gentoo stand last LinuxDays. 2016-06-07 12:53:05 CcxCZ: ahoj! :) 2016-06-07 12:53:32 nazdar :-) 2016-06-07 13:02:28 CcxCZ: there’s a chance that we will have Alpine stand and talks on this year’s LinuxDays :) 2016-06-07 13:18:22 <^7heo> jirutka: the point of a community is to have more people participate. 2016-06-07 13:18:40 <^7heo> jirutka: if you want to develop alpine linux by yourself, please do so; but you'll soon find yourself swamped in work. 2016-06-07 13:18:48 ^7heo: ? 2016-06-07 13:19:05 <^7heo> You essentially took my PR, did another one with the same content tweaked, merged it. 2016-06-07 13:19:10 <^7heo> it's nice that you fixed the issue tho. 2016-06-07 13:19:26 ^7heo: have you read the code? 2016-06-07 13:19:36 <^7heo> I have. 2016-06-07 13:19:46 ^7heo: do you think that your solution was better? 2016-06-07 13:19:49 <^7heo> No. 2016-06-07 13:19:58 ^7heo: so? 2016-06-07 13:20:14 <^7heo> So, since you already made me change some stuff... It could have been done in the same PR as before. 2016-06-07 13:20:17 ^7heo: maybe I’ve merged it too fast, I should give you more time for comments 2016-06-07 13:20:29 <^7heo> That would have been nice, yes; but it's done anyway. 2016-06-07 13:20:39 <^7heo> The most important fact is that the issue is now fixed. 2016-06-07 13:20:48 <^7heo> You did a good job on the issue, no question about that. 2016-06-07 13:20:54 the problem is that I cannot add other commits to some else’s PR 2016-06-07 13:21:03 <^7heo> Yeah I know, that's a limitation. 2016-06-07 13:21:09 <^7heo> github isn't perfect. 2016-06-07 13:21:16 I it was far more easy for me to just fix it myself and open new PR 2016-06-07 13:21:24 <^7heo> I can totally understand that. 2016-06-07 13:21:55 <^7heo> But I took time to submit my PR, and even tho you found a more elegant way to solve the problem; I felt a bit 'put aside' 2016-06-07 13:21:59 <^7heo> I'm only reporting that. 2016-06-07 13:22:15 <^7heo> It's possible that this behavior would drive contributors away. 2016-06-07 13:22:27 I’m sorry for that, I appreciate your effort and it helped me to understand the problem 2016-06-07 13:22:30 <^7heo> again, it's only my opinion, I'm reporting that because I think I can help. 2016-06-07 13:22:34 <^7heo> s/I can/it can/ 2016-06-07 13:22:43 <^7heo> Anytime. :) 2016-06-07 13:24:41 <^7heo> Anyway, now that this if fixed, I can re-submit 107. 2016-06-07 13:24:45 I thought that Travis doesn’t build abuilds that have been modified, but pkgver/pkgrel haven’t been bumped, but from recent commits I see that it builds even those 2016-06-07 13:25:01 <^7heo> yeah it does, afaik 2016-06-07 13:25:17 there’s a —force in https://github.com/alpinelinux/aports/blob/master/.travis/build-pkgs#L61 2016-06-07 13:25:47 so it should, but IIRC ap builddirs didn’t consider abuilds without bumped pkgver/pkgrel before; https://github.com/alpinelinux/aports/blob/master/.travis/build-pkgs#L31 2016-06-07 13:27:34 anyway, I hope that all the reported bugs should be fixed now; if you find another one, do not hesitate to tell me; I hope that I’ll be in better mood next time :) 2016-06-07 13:28:26 <^7heo> yeah I'll ping you here if I do. 2016-06-07 13:28:36 <^7heo> first I'll merge #115 2016-06-07 13:29:02 <^7heo> yeah, no. 2016-06-07 13:29:34 https://github.com/alpinelinux/aports/pull/115 is already merged 2016-06-07 13:30:04 <^7heo> nah, locally. 2016-06-07 13:30:31 <^7heo> I assumed that since I don't have the rights on the alpine repo, it was obvious that I would do that locally ;) 2016-06-07 13:31:12 well, I think that you should pull from master and rebase your PR on top of it 2016-06-07 13:31:28 instead of merging 115 into your local clone 2016-06-07 13:31:31 <^7heo> jirutka: also, if we know each other better, mood matters less. 2016-06-07 13:31:40 <^7heo> So, next time might be easier already. 2016-06-07 13:31:55 <^7heo> I want to merge 115 locally to see if I forgot something. 2016-06-07 13:32:09 <^7heo> but sure, I'll do that for the other branges. 2016-06-07 13:32:11 <^7heo> branches( 2016-06-07 13:32:50 <^7heo> damn, s/\(/*/ 2016-06-07 13:36:13 <^7heo> also, jirutka; git merge is actually clever enough to discard a commit when it's empty. 2016-06-07 13:36:50 I know 2016-06-07 13:37:20 <^7heo> So all happened right :) 2016-06-07 13:37:30 <^7heo> and I saw that we didn't forget anything 2016-06-07 13:41:21 <^7heo> Now I opened 118 and if it doesn't fail, it means that 115 works just fine. 2016-06-07 13:43:17 could you please mention in the readme that this repository is a read-only mirror of git://git.alpinelinux.org/aports? 2016-06-07 13:45:28 and also add address of the mailing list for patches: alpine-aports@lists.alpinelinux.org 2016-06-07 13:46:54 <^7heo> jirutka: I didn't know which email to mention. Thanks for telling me. 2016-06-07 13:47:40 also I think that we should not assume that people will read this readme only on GH 2016-06-07 13:48:16 it will be merged into the upstream repository, so it may be misleading for people who clone directly the upstream repository and not the GitHub mirror 2016-06-07 13:49:50 <^7heo> jirutka: how is the read-only property of github relevant to contributors, though? 2016-06-07 13:50:05 <^7heo> for them, it doesn't matter, since they can contribute via github anyway. 2016-06-07 13:50:12 hmm… you’re right, it’s not 2016-06-07 13:50:31 <^7heo> I have put the mail of the ML, now. 2016-06-07 13:50:42 it’s relevant only for maintainers, we must not merge PRs directly wia GH 2016-06-07 13:50:46 *via 2016-06-07 13:51:17 <^7heo> wait, how do you merge then? 2016-06-07 13:51:39 eh, using git? :) 2016-06-07 13:51:44 <^7heo> do you take the commits in the PR and push them from the git.a.o? 2016-06-07 13:51:52 yeah 2016-06-07 13:51:55 <^7heo> aah ok. 2016-06-07 13:52:00 <^7heo> now I get it... 2016-06-07 13:52:23 <^7heo> it's quite some work... 2016-06-07 13:52:38 I’m using https://github.com/github/hub, so for me it’s just: 2016-06-07 13:52:38 git am https://github.com/alpinelinux/aports/pull/118 2016-06-07 13:52:38 git pull 2016-06-07 13:53:14 <^7heo> yeah, I see. 2016-06-07 13:53:18 I don’t use GH’s Merge button even on my own projects 2016-06-07 13:53:27 <^7heo> it's not so hard if you configure the GH remote; you're right 2016-06-07 13:53:28 because it creates a merge commit that I don’t like 2016-06-07 13:53:30 <^7heo> I didn't think of that. 2016-06-07 13:54:03 and ncopa has the same opinion about merge commits, we always rebase patches on top of the master 2016-06-07 13:54:26 <^7heo> yeah I don't like rebasing too much. 2016-06-07 13:54:36 <^7heo> I mean, I do that when my patches are simple. 2016-06-07 13:54:45 <^7heo> otherwise, git is merge-based for a reason. 2016-06-07 13:55:15 I prefer standard merge when there are some more complicated conflicts 2016-06-07 13:55:38 <^7heo> yeah sure, it's easier to read, later on. 2016-06-07 13:56:35 <^7heo> and I'm gonna work on the gogs branch as soon as my readme's merged. 2016-06-07 13:58:01 fyi, we should ask ncopa about the readme, I think that I should not merge it before he reviews it 2016-06-07 13:58:29 So are testing php-* packages supposed to be renamed to php5-* ? 2016-06-07 13:59:00 and if so, what is the appropriate way to handle that re commit message, package bump etc? 2016-06-07 13:59:20 pkgrel bump I mean 2016-06-07 13:59:55 <^7heo> jirutka: agreed. 2016-06-07 14:00:17 when you rename the package, there’s no need to bump pkgrel… for the Alpine build system, it’ll be basically a new abuild 2016-06-07 14:00:31 okay, that makes sense. 2016-06-07 14:00:40 maybe you should add `replaces="php-NAME"`, but I’m not sure about it 2016-06-07 14:00:45 okay 2016-06-07 14:01:12 so if it's already non-zero, probably reset to zero? 2016-06-07 14:03:53 yes 2016-06-07 14:12:06 ^7heo: btw “folder” is a windowsism, it’s traditionally called “directory” 2016-06-07 14:12:35 ^7heo: you don’t find the term “folder” in any unix software 2016-06-07 14:18:12 <^7heo> I never paid attention to that. 2016-06-07 14:18:16 <^7heo> to me they're equivalent. 2016-06-07 14:18:26 <^7heo> But, alright :) 2016-06-07 14:19:03 <^7heo> corrected. 2016-06-07 14:20:22 <^7heo> http://superuser.com/a/247396/179313 < Interesting. 2016-06-07 14:27:10 ncopa: regarding http://bugs.alpinelinux.org/issues/5679 i'm not sure if i can reply as i submitted the bug via email, but I can confirm here that the issue has been solved. 2016-06-07 14:27:17 i first downgraded my custom Xen build to 4.6.1-r1 and Xorg would now segfault, i then upgraded to Xen 4.6.1-r2 from edge and Xorg no longer segfaults 2016-06-07 14:27:27 thank you very much for including the patch 2016-06-07 14:43:40 <^7heo> I'm seraching for a specific file in an apk package 2016-06-07 14:43:44 <^7heo> is there a way to find that? 2016-06-07 14:44:02 ^7heo: yes, http://pkgs.alpinelinux.org/contents 2016-06-07 14:44:51 <^7heo> jirutka: thanks a lot. 2016-06-07 14:45:07 <^7heo> should we add an apk action to query that? 2016-06-07 14:45:26 <^7heo> I mean, it's not necessary great but at least people won't ask for it then. 2016-06-07 14:47:43 ^7heo: it would be very useful, but these information are not in the APKINDEX 2016-06-07 14:50:29 <^7heo> I know, that's why I'm saying: querying that. 2016-06-07 14:50:34 <^7heo> via curl or whatnot. 2016-06-07 14:51:02 yes, once we will have a RESTful API for pkgs, then we can query it online 2016-06-07 14:51:04 we have it already 2016-06-07 14:51:17 <^7heo> btw, texlive doesn't work in @testing. 2016-06-07 14:51:28 https://github.com/jfrazelle/apk-file 2016-06-07 14:51:48 barthalion: sorry, but this is not a solution, this is a nasty hack… it parses HTML 2016-06-07 14:52:02 I wait for your RESTful API 2016-06-07 14:52:10 until then, let me use what works for me 2016-06-07 14:52:26 unless you say missing solutions are better than existing 2016-06-07 14:52:30 sure you can use it, I’m just saying that it’s not _done_ 2016-06-07 14:52:40 someone alreadys works on RESTful API 2016-06-07 15:17:42 art 2016-06-07 15:19:23 I like it 2016-06-07 15:22:41 ncopa: with this switch to libressl? how will that affect if a package depends on openssl? 2016-06-07 15:34:30 shouldn't package file listing be integrated in the package ? 2016-06-07 15:35:12 coredump: it is in the package, but apk works only with APKINDEX when searching packages 2016-06-07 15:35:42 and I’m not sure if it’s a good idea to include file listing in APKINDEX, because it may be huge 2016-06-07 15:35:49 sure 2016-06-07 15:36:23 so there's searching based on pkg name 2016-06-07 15:36:35 and searching on file name 2016-06-07 15:37:06 is it a real issue having biggerish APKINDEX ? 2016-06-07 15:39:29 I dunno, that’s probably a question for fabled and ncopa 2016-06-07 15:46:20 I really miss rpm -qf and yum whatprovides :'( 2016-06-07 15:53:30 coredumb: apk info --who-owns ? 2016-06-07 15:53:51 is it possible to submit just via pull requests? 2016-06-07 15:54:06 this mailing list thing is such a pain in the ass. (packages) 2016-06-07 15:54:22 coredump: btw do you know that you can find e.g. so libraries by name? e.g. so:libcurl.so 2016-06-07 15:54:25 coredumb: apk info --help 2016-06-07 15:54:27 mostly self induced though from my email locking 2016-06-07 15:54:40 boingolov: rofl I should read it more often :D 2016-06-07 15:55:20 but doesn't resolve the case for non installed packages ;) 2016-06-07 15:55:44 Adran: I have a VM with a local postfix relay just for doing the mailing list, then submit via git send-email 2016-06-07 15:55:56 boingolov: Probably going to have to do that 2016-06-07 15:56:26 but I agree, I prefer the github pull request workflow 2016-06-07 16:25:50 Adran: there are some api incompatibilites between libre and open, so we will have to port some packages 2016-06-07 16:28:55 i know the package I've been working on won't work. 2016-06-07 16:29:35 it will be our responsibility if it makes it to aports… 2016-06-07 16:29:57 lol 2016-06-07 16:30:53 cool, in git there is support for libre 2016-06-07 16:41:51 wogiz: thank you 2016-06-07 16:48:45 Adran: re libressl switch, all packages depending on openssl will have to be rebuilt 2016-06-07 16:49:28 coredumb: we did evalulate the possibility to include the filelisting in the APKINDEX 2016-06-07 16:49:45 btw what about migrating from gcc to cmake as a default compiler? is it currently possible and useful? 2016-06-07 16:49:48 i think we even calculated the size of how big it would be 2016-06-07 16:50:15 jirutka: you mean clang 2016-06-07 16:50:15 the conclusion was that it is not worth it 2016-06-07 16:50:20 :p 2016-06-07 16:50:26 yeah, clang, sorry 2016-06-07 16:50:52 gives a little extra value but the cost is big for everyone 2016-06-07 16:50:55 why it’s not worth it? too many packages that hard depend on gcc? 2016-06-07 16:51:08 i'm talking about filelist in apkindex still 2016-06-07 16:51:23 sorry for beeing too slow on switching contenxt 2016-06-07 16:51:27 i would say less GNU, less trouble :) but gcc is still de facto standard on Linux, so it may be indeed a hard path 2016-06-07 16:51:36 and i was mainly talking to coredumb 2016-06-07 16:51:37 ncopa: rebuilt is fine, I was wondering more about the incompat issues 2016-06-07 16:51:39 aha, ok 2016-06-07 16:52:06 Adran: thost we will have to work out. I dont know what the incompat issues are yet 2016-06-07 16:52:33 some security tools might have some issues, because tehy rely on the finer points of openssl - the stuff "nobody uses" 2016-06-07 16:52:53 Adran: if you have an overview of what the incompat issues are, then feel free to write to alpine-devel 2016-06-07 16:52:56 would be helpful 2016-06-07 16:53:16 i think in worst case, we'd have to provide openssl as separate package 2016-06-07 16:53:18 i'll go test and see 2016-06-07 16:53:20 provide both 2016-06-07 16:53:22 ncopa: thats not easily possible 2016-06-07 16:53:44 libre/openssl are incompat since they use the same api headers (forgive me for not using the right terms) its super early morning 2016-06-07 16:53:44 well, its the openssl tools that conflict, right? 2016-06-07 16:53:50 yes 2016-06-07 16:54:09 more like some stuff was gutted without regard. but I'll go test and post results if its worth metnioning on devel 2016-06-07 16:54:13 i know the API changed in libressl so some packages will need to be patched 2016-06-07 16:54:21 i think ssl3 was removed 2016-06-07 16:54:30 oh dang 2016-06-07 16:54:33 :| 2016-06-07 16:54:42 there are some packages that will need patches i think 2016-06-07 16:55:14 my initial thought is that we add libressl side by side 2016-06-07 16:55:30 lol 2016-06-07 16:55:33 it breaks snort and bro 2016-06-07 16:55:34 ha 2016-06-07 16:55:35 even if you cannot install them both at the same time 2016-06-07 16:55:43 guys, I'm thinking about ways of fixing UID/GID mess for 3.4.1, because recent patches fix only stuff for new package installations, not already installed (primary GID won't change). do you think that one shell script having knowledge of all potentially "broken" passwd entries and trying to fix them all (assign proper primary GID, remove user from nogroup) would be good enough? 2016-06-07 16:56:21 przemoc: i doubt its worth it 2016-06-07 16:56:29 ncopa: https://wiki.freebsd.org/OpenSSL/No-SSLv3 2016-06-07 16:56:44 przemoc: we could make post-upgrade symlink 2016-06-07 16:56:58 so it would run the trigger on upgrade too 2016-06-07 16:57:14 przemoc: what happend? I saw huge patches in the repo around adding users/groups 2016-06-07 16:57:48 but you can always work around it by apk del pkg && deluser foo && delgroup foo && apk add pkg 2016-06-07 16:58:19 przemoc: do you think many are affected? 2016-06-07 17:00:39 I haven't inspected yet how all users are used within packages, just wanted to bring status quo (i.e. using group named the same as user) in the first place, as it looked the safest and quickest way to do to fix most of the mess. 2016-06-07 17:01:34 jirutka: http://lists.alpinelinux.org/alpine-aports/3059.html tl;dr in my previous pre-3.4 patch set I overlooked that: system user creation doesn't add same named group and uses nogroup as primary group unless explicitly specified via -G 2016-06-07 17:02:43 dl-cdn is the preferred update method now right? 2016-06-07 17:02:44 so "improvement" turned into "deprovement" 2016-06-07 17:03:07 przemoc: I think that I’ve already read this, but could you please send me a link to the patch that has introduced this problem? 2016-06-07 17:04:18 ccc056dbf9d3 2016-06-07 17:04:36 thanks 2016-06-07 17:04:38 http://git.alpinelinux.org/cgit/aports/commit/?id=ccc056dbf9d3 2016-06-07 17:05:31 you’ve just added `-S`… this changed behaviour so much? 2016-06-07 17:05:52 yes, system users are created differently than non-system users 2016-06-07 17:08:35 Arch, Debian and Fedora disabled SSLv3 in their openssl packages 2016-06-07 17:08:49 so the work about it is already done 2016-06-07 17:09:46 jirutka: i think adduser -S will not automatically create the group 2016-06-07 17:09:50 "I would be more concerned about our certificate parsing code - we use a lot of internal apis there. well, they are exposed, but really not frequently used" 2016-06-07 17:10:36 some downstream packages have dropped ssl3 support already i think. iirc ngircd did so 2016-06-07 17:11:01 Adran: yes i saw https://wiki.freebsd.org/OpenSSL/No-SSLv3 2016-06-07 17:11:24 what is good about it is that there are already patches for most of the affected stuff 2016-06-07 17:11:34 having adduser -S ... NAME in pkgs is good, bad was not adding addgroup -S NAME before and not extending adduser with -G NAME to preserve user:group as was used before (non-S adduser is meant for human users and thus it automatically creates group for them, unless -G is used) 2016-06-07 17:11:35 and someoneelase has already tried this before us 2016-06-07 17:12:27 przemoc: i think the chage was good and it was needed. I calculated the risk, and i dont regret pushing the patch. 2016-06-07 17:12:37 and i am very happy that you are eager to fix it 2016-06-07 17:12:58 i would not be so happy if i would have to clean up ;) 2016-06-07 17:13:58 that's more than understandable 2016-06-07 17:21:37 <^7heo> damn 2016-06-07 17:21:51 <^7heo> how does one uses DHCP and specifies what DNS servers to use? 2016-06-07 17:22:30 ^7heo: you want to ignore DNS servers provided by DHCP? 2016-06-07 17:22:33 <^7heo> yes. 2016-06-07 17:22:43 wait a sec 2016-06-07 17:22:44 <^7heo> telekom sucks. 2016-06-07 17:23:38 <^7heo> it's quite annoying not to find any doc' about that. 2016-06-07 17:24:00 ^7heo: hm, I’m not sure if this will work on Alpine, but try to add `dns-nameservers a.b.c.d` to /etc/net/interfaces 2016-06-07 17:24:17 * /etc/network/interfaces 2016-06-07 17:24:41 although, resolv.con shouldnt change 2016-06-07 17:24:55 ``` 2016-06-07 17:24:55 iface eth0 inet dhcp 2016-06-07 17:24:55 ``` 2016-06-07 17:24:55 dns-nameservers a.b.c.d 2016-06-07 17:25:20 <^7heo> damn 2016-06-07 17:25:28 <^7heo> resolv.conf is really sticky. 2016-06-07 17:25:44 <^7heo> dns-nameservers doesn't do it. 2016-06-07 17:25:50 resolv.conf is dynamic, I always prefer to set DNS servers in /etc/network/interfaces 2016-06-07 17:25:54 hm, that’s bad 2016-06-07 17:26:27 so then I dunno; I’ve done this only on Gentoo that uses netifrc, there it’s easy 2016-06-07 17:27:41 what makes it dynamic? 2016-06-07 17:27:51 my experience if you didn't have resolv-conf packages it'd stay pretty dull 2016-06-07 17:28:08 it’s generated by if-up or how it is called on alpine 2016-06-07 17:31:12 <^7heo> jirutka: oh, I never tried to edit /etc/resolv.conf directly. 2016-06-07 17:31:19 <^7heo> I only tried to edit the /etc/network/interfaces 2016-06-07 17:31:26 <^7heo> but unfortunately, it doesn't work. 2016-06-07 17:31:42 as I said, /etc/resolv.conf is generated 2016-06-07 17:31:49 <^7heo> ? 2016-06-07 17:32:23 <^7heo> Did you read only the last line, or? ;) 2016-06-07 17:32:51 IIRC Busybox/Alpine uses the same format of interfaces as Debian, try to google it 2016-06-07 17:33:14 <^7heo> okay I'll do that again, maybe I mistyped something. 2016-06-07 17:35:08 <^7heo> and btw, s/google/duckduckgo/ 2016-06-07 17:35:24 I make it more clear: 1. /etc/resolv.conf is genereated, so when you edit it by hand, it will be rewrited when you start an interfaces 2016-06-07 17:35:24 2. I don’t know how to specify in /etc/network/interfaces to not get DNS from DHCP; I tought that dns-nameserves would work, but as you tried, it doesn’t 2016-06-07 17:35:36 ^7heo: +1 2016-06-07 17:35:58 ^7heo: however, I’ve tried using DuckDuckGo and I was not happy with it, so I use it just as a gateway to Google :( 2016-06-07 17:36:10 ^7heo: maybe I should get a second chance to startpage 2016-06-07 17:36:31 <^7heo> jirutka: I never edited /etc/resolv.conf 2016-06-07 17:36:38 <^7heo> not since, I dunno, 2008 2016-06-07 17:37:21 <^7heo> jirutka: if you want a gateway to google, maybe you want startpage.com / ixquick.eu 2016-06-07 17:37:34 I’ve already mentioned startpage ;) 2016-06-07 17:37:37 <^7heo> oh, you wrote it already 2016-06-07 17:37:44 <^7heo> yeah, you only read the last line, I only read the rest :P 2016-06-07 17:37:52 <^7heo> together we can have a 100% coverage. 2016-06-07 17:37:54 <^7heo> ACTION hides 2016-06-07 17:37:57 I’ve tried it like a year or two ago and it was quite slow 2016-06-07 17:38:02 XD 2016-06-07 17:38:04 <^7heo> it's still slow sometimes. 2016-06-07 17:38:50 <^7heo> jirutka: btw, are you in Prague? 2016-06-07 17:38:55 I’d like to somehow outsource writing review for thesis; I’m out of inspiration now :/ 2016-06-07 17:38:56 yes 2016-06-07 17:39:02 <^7heo> Ok, how about a beer, soon? 2016-06-07 17:39:10 you’re from Prague? 2016-06-07 17:39:13 <^7heo> Berlin. 2016-06-07 17:39:16 <^7heo> It's not *too* far. 2016-06-07 17:39:53 that’s good idea :) 2016-06-07 17:40:19 <^7heo> it would only take me 90h by foot. 2016-06-07 17:40:27 lol 2016-06-07 17:40:27 <^7heo> just over two weeks of full time ;) 2016-06-07 17:40:42 <^7heo> by bike, I can probably come in 2 day 2016-06-07 17:40:43 <^7heo> s 2016-06-07 17:40:44 maybe you can try a horse ;) 2016-06-07 17:40:57 <^7heo> Nah, my horse is too small for that. 2016-06-07 17:41:06 <^7heo> (I have a domestic horse) 2016-06-07 17:41:20 <^7heo> (aka dog) 2016-06-07 17:41:30 <^7heo> But I always wanted to visit prague. 2016-06-07 17:41:45 I’ve also heard that some ppl tries to use dragons as a means of transport 2016-06-07 17:41:51 <^7heo> yeah 2016-06-07 17:41:57 <^7heo> But the firebreath is annoying in summer. 2016-06-07 17:42:09 <^7heo> In winter, it's convenient, esp. for sausages... 2016-06-07 17:42:12 <^7heo> but in summer... 2016-06-07 17:42:19 the only problem is a damage they often cause when landing… you know… like fire… 2016-06-07 17:42:36 <^7heo> yeah that too. 2016-06-07 17:42:42 <^7heo> just land in a giant bbq. 2016-06-07 17:42:49 <^7heo> then everybody's happy. 2016-06-07 17:42:55 <^7heo> until the dragon poops. 2016-06-07 17:43:06 XD 2016-06-07 17:43:32 <^7heo> hey dude 2016-06-07 17:43:34 <^7heo> you know what? 2016-06-07 17:43:39 <^7heo> I owe you an apology. 2016-06-07 17:43:46 <^7heo> I didnt test that 'dns-nameservers' 2016-06-07 17:43:54 <^7heo> I placed it in the wrong iface section v_v 2016-06-07 17:43:59 <^7heo> and its' not friday yet... 2016-06-07 17:44:26 <^7heo> maybe I got too many coffees 2016-06-07 17:44:34 <^7heo> maybe not enough 2016-06-07 17:44:53 <^7heo> but still, doesn't work v_v 2016-06-07 17:45:36 <^7heo> this time I'm trying something more aggressive. 2016-06-07 17:45:48 <^7heo> still doesn't work... 2016-06-07 17:46:31 <^7heo> So yeah, nothing changed in the end. 2016-06-07 17:47:29 I should probably go to the second coffee, I can’t focus; or maybe it’s cause by that noise outside, some drums or what is it 2016-06-07 20:26:44 hi, I filed PR for new package, does it makes sense to update usage of "_builddir"? https://github.com/alpinelinux/aports/pull/119#issuecomment-224400907 2016-06-07 20:43:53 andypost: sure 2016-06-07 20:44:16 w00t, just build the first alpine package on my new xps15 :) 2016-06-07 20:44:42 algitbot: thank you so very much. 2016-06-07 21:00:21 clandmeter, I can't find in docs about pkgrel so reverted that to 0, is that right? 2016-06-07 21:00:48 yes, pkgrel is 0-based 2016-06-07 21:01:15 jirutka, thanx! 2016-06-07 21:01:21 np 2016-06-07 21:01:50 reset it to 0 when pkgver is updated 2016-06-07 21:02:52 clandmeter, so 0 is valid cos new package is new, old one was php5-apcu 2016-06-07 21:03:17 yes, thats right. 2016-06-07 21:03:46 thanx a lot 2016-06-07 21:03:54 ACTION will wait for reviews 2016-06-07 22:13:44 Thank you again for this wonderful distro! I am giving Alpine a try in the desktop. It is sort of working! I am following https://wiki.alpinelinux.org/wiki/XFCE_Setup 2016-06-07 22:14:01 except that apk add alpine-desktop says unsatisfiable constraints: so:libetpan.so.17 missing. Any idea why? 2016-06-07 22:14:33 There is also "required by: claws-mail-3.13.2-r1[so:libetpan.so.17]" . I suspect there is a dependency issue on edge. 2016-06-07 22:26:12 Anybody? 2016-06-07 22:46:01 So what is the usualy process from moving an aport over from testing into community? 2016-06-07 22:48:12 boingolov: IIRC, packages in the community repository should be supported at least 6 months 2016-06-07 22:48:42 boingolov: before moving to the community, the package should be verified that it really works 2016-06-07 22:49:12 also all the package dependencies must be at least in the community repo or main 2016-06-07 22:56:35 hey jirutka. I am trying alpine desktop now with some unpolished pieces. It is alpine-desktop ready for users? 2016-06-07 22:57:17 mzdaniel: I dunno, I use Alpine only on servers 2016-06-07 22:57:28 so as to the "verified that it really works", we're using libcouchbase and the language bindings for it, but I suppose you mean verified by someone who isn't the maintainer? 2016-06-07 22:58:22 boingolov: if you’re actually using that package and it works well, then it’s sufficient 2016-06-07 22:58:30 okay 2016-06-07 22:59:38 boingolov: some maintainers makes packages for other people for software that they don’t know, so they can’t verify if it really works as expected 2016-06-07 23:00:33 I also doing the same, though if it works reasonable well on the desktop I'd like to jump in. 2016-06-07 23:02:10 sure, I am not a libcouchbase contributer, and I don't claim to know it forward and backwards, but we do use it as a consumer and I have verified that all of their tests pass (including the ones with the mock server). We do use it as a consumer quite a bit though 2016-06-07 23:02:26 we're slowly migrating all of our images to alpine 2016-06-07 23:02:48 then it’s perfectly fine :) 2016-06-07 23:04:50 I certainly can offer to do testing and maybe patches if developers are interested 2016-06-07 23:05:59 mzdaniel: that would be great :) as I said, I’ve never tried Alpine on desktop, but it’s surely supported and I know people how use Alpine on desktop, so it should work fine 2016-06-07 23:07:51 That's what I thought too. My experience today with edge is not quite confirming this. release 3.4.0 is not even able to find alpine-desktop 2016-06-07 23:08:42 What is the best channel for support as a developer? This one? 2016-06-07 23:08:49 edge may be broken 2016-06-07 23:08:57 yeah, this one 2016-06-07 23:09:42 what do you mean with that 3.4.0 is not even able to find alpine-desktop? 2016-06-07 23:10:29 I'll do the full process again. Give me a sec 2016-06-07 23:18:15 usb is not cooperating. I am reimaging my usb drive 2016-06-07 23:26:00 Ok I have a fresh booted release. Plenty of packages are missing and required by world. eg: xf86-video-vesa when doing setup-xorg-base 2016-06-07 23:26:42 this package is in v3.4 https://pkgs.alpinelinux.org/package/v3.4/main/x86_64/xf86-video-vesa 2016-06-07 23:27:04 what repositories do you have in /etc/apk/repositories? 2016-06-07 23:29:13 only the /media/sdb/apks which explains why. I should run setup-alpine for effimeral usb boot? 2016-06-07 23:30:20 yeah, you must add some remote repository, e.g. http://dl-cdn.alpinelinux.org/alpine/v3.4/main 2016-06-07 23:30:49 yes, setup-alpine populates repositories IIRC 2016-06-07 23:30:55 btw https://wiki.alpinelinux.org/wiki/Installation 2016-06-07 23:34:41 Thank you Jirutka. Yes, I saw that. I am looking at http://wiki.alpinelinux.org/wiki/XFCE_Setup 2016-06-07 23:36:11 After doing setup-xorg-base, apk add alpine0desktop xfce4 raises: ERROR: unsatisfiable constraints: alpine-desktop (missing): required by: world(alpine-desktop) 2016-06-07 23:54:24 have you run `apk update` after adding repo into /etc/apk/repositories? 2016-06-07 23:54:25 I see. alpine-desktop is in v3.3 and edge but not in v3.4 for x86_64 2016-06-07 23:54:55 huh, this looks like a bug 2016-06-07 23:55:20 This is why am in this channel :) 2016-06-07 23:55:58 it’s definitely a bug 2016-06-07 23:57:05 this could be related to "alpine-desktop says unsatisfiable constraints: so:libetpan.so.17 missing" on edge? 2016-06-07 23:57:46 I’ve created a bug report http://bugs.alpinelinux.org/issues/5691 2016-06-07 23:58:18 nope, alpine-desktop is just a metapackage 2016-06-07 23:58:57 it seems that libetpan has been removed, because it’s not in edge 2016-06-07 23:59:34 pardon, not 2016-06-08 00:00:09 it is in the repository, but greater version, so:libetpan.so.20 2016-06-08 00:00:22 Understand. libetpan seems to be a dependency of claws-mail. It requires version 17 but edge provides v20 2016-06-08 00:00:25 yep 2016-06-08 00:01:12 oh crap, claws-mail is yet another package that is not available for x86_64 but should be :( 2016-06-08 00:01:16 in v3.4 2016-06-08 00:01:32 this is bad, we should check it 2016-06-08 00:04:16 #5692 2016-06-08 00:08:57 Perfect. thank you again jirutka. I'll follow these issues. 2016-06-08 00:09:46 this is not perfect, it’s a shame that we have not noticed this problem :( 2016-06-08 00:10:12 thank you for pointing to it! 2016-06-08 00:13:43 The alpine community is doing an awesome job. Sure, processes can be improved and more things can be automated, but lots of the work is done by volunteers, isn't it? 2016-06-08 00:13:59 yes, it is 2016-06-08 00:15:22 Glad I'm able to help. I'll keep playing with it. 2016-06-08 00:55:10 so curious if there is anything i need to do with my ghc port, or if its not going to get merged/etc... or if there is any action I need to take? 2016-06-08 00:56:17 also as a note the prior ghc 7.10.3 port can probably be excised, it was x86_64 only and i'll probably not worry much about it now that 8.0 is out 2016-06-08 02:13:57 mitchty: I wouldn't be too surprised if everyone has been too busy chasing their tails with the 3.4 release. It's been 3 weeks, maybe worth a followup email? 2016-06-08 02:48:43 boingolov: sure, not in a rush but just curious what needs to happen 2016-06-08 07:04:31 morning all 2016-06-08 07:04:53 morning 2016-06-08 08:04:26 how do I report bug for alpine? RFE a missing requirement for dpkg 2016-06-08 08:06:14 bugs.alpinelinux.org 2016-06-08 08:07:29 thx 2016-06-08 08:16:15 https://bugs.alpinelinux.org/issues/5693 :P 2016-06-08 09:18:38 tru_tru, once you apk add xz, debootstrap works? 2016-06-08 09:44:42 tru_tru: should be fixed by a3a2b4ccb7b7fde68310627529df0120a1cc468f 2016-06-08 09:45:02 wrong commit checksum I meant fa88e39a4bc3aa7b8ff1cced8176d9911335f73b 2016-06-08 11:18:22 fcolista: yes, nmeum: thx 2016-06-08 11:29:10 hey :) 2016-06-08 11:29:18 o/ 2016-06-08 11:35:34 clandmeter, please comment on https://github.com/alpinelinux/aports/pull/119 2016-06-08 11:36:01 is there any guides about maintainership? 2016-06-08 11:37:27 hi 2016-06-08 11:37:40 yes, simple and short, you maintain it. 2016-06-08 11:37:43 :) 2016-06-08 11:38:06 in testing its ok without maintainer 2016-06-08 11:38:51 if you want it in main/community, you need a maintainer. 2016-06-08 11:40:50 apk add i3status returns an error because libconfuse.so.0 is missing 2016-06-08 11:41:06 does that package need a rebuild or is another package missing all of the sudden? 2016-06-08 11:42:27 who is sören tempel in irc? 2016-06-08 11:45:11 leo-unglaub: edge? 2016-06-08 11:47:07 clandmeter: yes 2016-06-08 11:48:04 also a minor detail, the description url got changed to https://i3wm.org/i3status/ 2016-06-08 11:48:15 too late 2016-06-08 11:49:03 try the rebuild now. 2016-06-08 11:50:13 too late? oh noooo. we are all going to die !!!1! 2016-06-08 11:50:33 i am trying the rebuild as soon as my wifi works again ... fucking wpa_supplicant 2016-06-08 11:50:38 yeah, we really have some problem, there are missing packages for x86_64 in v3.4 2016-06-08 11:50:59 see #5691 and #5692 2016-06-08 11:51:22 missing as in failed to build? 2016-06-08 11:51:28 leo-unglaub: I am sören tempel, I will take a look at it 2016-06-08 11:51:28 I dunno 2016-06-08 11:51:45 nmeum: ah, good to know :) 2016-06-08 11:51:52 there are built in edge 2016-06-08 11:51:52 nmeum: i just rebuild it. 2016-06-08 11:51:56 looks like a missing rebuild 2016-06-08 11:51:58 yeah 2016-06-08 11:52:00 thanks 2016-06-08 11:52:10 so I think that there’s no problem with packages itself, but build infrastructure 2016-06-08 11:52:57 clandmeter: the rebuild is not available on the master mirror nl. 2016-06-08 11:53:04 clandmeter, thanx for idea, I'm glad to maintain it in testing til php7 will be ready for community 2016-06-08 11:53:05 nl is not master 2016-06-08 11:53:22 clandmeter: not? did you guys change thatß 2016-06-08 11:53:43 leo-unglaub: ncopa recommended me dl-cdn.alpinelinux.prg 2016-06-08 11:54:08 yes use cdn 2016-06-08 11:54:13 it should be up2date 2016-06-08 11:54:41 oh, never heared about that one 2016-06-08 11:54:47 leo: you can also try https://repository.fit.cvut.cz/mirrors/alpine/, that’s my mirror, synced each hour 2016-06-08 11:55:00 it supports both HTTP and HTTPS and both IPv4 and IPv6 2016-06-08 11:55:08 most mirrors are synced every hour 2016-06-08 11:55:16 except leaseweb 2016-06-08 11:55:25 rsync.a.o is master mirror 2016-06-08 11:55:35 my mirror is synced every hour as well, but i thought nl. is always pushed when a build is done 2016-06-08 11:55:45 but cdn should be a direct up2date mirror of it. 2016-06-08 11:55:53 it does not cache the index. 2016-06-08 11:55:58 pardon, not every hour, but every 15 minutes :) 2016-06-08 11:56:19 so it should always get the latest packages from rsync.a.o 2016-06-08 11:56:32 and it’s synced against upstream, rsync://rsync.alpinelinux.org/alpine/ 2016-06-08 11:57:16 hmm, i see 2016-06-08 11:57:28 clandmeter: I think that we should somehow check all the packages in v3.4, there are probably more missing packages 2016-06-08 11:57:33 but also dl-cdn.alpinelinux.org does not contain the rebuild clandmeter 2016-06-08 11:57:39 did you put it in edge? 2016-06-08 11:57:51 it does not? 2016-06-08 11:57:56 not in the index? 2016-06-08 11:58:05 no sir, apk upgrade -U does not find it 2016-06-08 11:58:09 hmm 2016-06-08 11:58:22 then just try rsync.a.o 2016-06-08 11:58:37 or fr.a.o (its the same) 2016-06-08 11:59:08 maybe my understanding of the cdn is incorrect 2016-06-08 11:59:15 i didnt set it up. 2016-06-08 11:59:32 jirutka: i can take a look 2016-06-08 11:59:49 jirutka: regarding missing packages: every now and then a package build fails for some reason (e.g. failure to fetch the source tarball) rebuilding claws-mail and alpine-desktop should probably fix it 2016-06-08 11:59:51 clandmeter: could you please merge https://github.com/alpinelinux/aports/pulls?q=is%3Aopen+is%3Apr+label%3Abackport-it and backport them to v3.4 (→ #5688 and #5690)? 2016-06-08 11:59:55 sadly there are now buildlogs for 3.4 2016-06-08 12:00:14 s/now/no/ 2016-06-08 12:00:40 clandmeter: yes, rsync.a.o contained the correct package 2016-06-08 12:00:47 the package works now! 2016-06-08 12:01:09 ncopa: shouldnt cdn use an incashed apkindex? 2016-06-08 12:01:14 nmeum: this is bad, because Alpine v3.4 is now broken for many desktop users 2016-06-08 12:01:29 noncached... 2016-06-08 12:01:46 brainfart 2016-06-08 12:01:46 jirutka: it would be nice to check if all package where build succesfully before making a new release I guess… 2016-06-08 12:02:06 nmeum: exactly 2016-06-08 12:02:15 it should 2016-06-08 12:02:30 its an option on abuild to stop at failure. 2016-06-08 12:02:47 maybe ncopa forgot to set it correctly 2016-06-08 12:03:12 hmm, we really have to cleanup our mirrors 2016-06-08 12:03:14 I’ll eventually write some script for that, to mass check if all the packages that should be available are there 2016-06-08 12:03:29 finding stuff and mirrors has become very hard 2016-06-08 12:03:48 leo-unglaub: it is? 2016-06-08 12:04:06 if you want something first minute, you just have to use the master mirror. 2016-06-08 12:04:28 clandmeter: i think so, yes. understanding where the build server pushes to and what mirrors synces from what mirror is hard to understand 2016-06-08 12:04:47 we should have a page where we list all mirrors and there parents so people have a better overview 2016-06-08 12:04:56 leo-unglaub: I think that all mirrors are synced against rsync.a.o, aren’t they? 2016-06-08 12:05:22 jirutka: i have no idea. i think most of them are, but do they really have to all go to the master? 2016-06-08 12:05:36 i think ncopa once had the idea to use a tier system 2016-06-08 12:05:45 leo-unglaub: it may be useful to monitor mirrors, if there are up-to-date, I’ve seen that in some distribution, don’t remember which one 2016-06-08 12:05:47 t1 mirros, t2 mirrors and then t3 as the community mirrors 2016-06-08 12:06:17 if rsync.a.o can handle it, then I don’t see a reason for that… 2016-06-08 12:06:36 thats why we advise to use our cdn 2016-06-08 12:06:41 something like that here: https://paste.debian.net/plain/726357 2016-06-08 12:07:37 leo-unglaub: just use the cdn, but dont complain if a package doesnt arrive first minute. 2016-06-08 12:08:02 the cdn does what? redirect to a random mirror? 2016-06-08 12:08:29 from my understanding it has multiple cached servers 2016-06-08 12:08:39 but it *should not* cache the apkindex 2016-06-08 12:09:02 so when a pkg is not yet in the cache, it will fetch it because the index is already updated. 2016-06-08 12:09:42 but i didnt set it up, and i dont know who did. 2016-06-08 12:11:08 you just made exactly my point 2016-06-08 12:11:29 there is no way knowing who runs what mirror and what exactly it does 2016-06-08 12:11:35 and i think that should be changed 2016-06-08 12:11:44 maybe we can extend mirrors.yml 2016-06-08 12:12:25 add contact, source of the mirror and also if it supports IPv6 :) 2016-06-08 12:12:32 and how often it sync 2016-06-08 12:13:06 https://github.com/alpinelinux/aports/blob/master/main/alpine-mirrors/mirrors.yaml 2016-06-08 12:13:07 that is even better than my idea to do a wiki page about it! 2016-06-08 12:13:30 I’m not sure, where is mirrors.yml actually used? 2016-06-08 12:13:38 maybe we should generate some HTML page from it 2016-06-08 12:14:31 i think the MIRRORS.txt is generated from it? 2016-06-08 12:14:36 or am i wrong? 2016-06-08 12:16:43 looks like latest-stable is also still wrong 2016-06-08 12:19:23 leo-unglaub: create a pr with your ideas, im sure we can work out something. 2016-06-08 12:21:51 clandmeter: yeah, we will find a solution. 2016-06-08 12:22:04 but i am not sure myself what the best and cleanest solution is 2016-06-08 12:22:14 leo-unglaub: btw, i just checked cdn, it does have the new pkg in index, but http listing still shows old pkg. 2016-06-08 12:22:50 leo-unglaub: thats what a pr is for, atleast it doesnt get lost like we do here. 2016-06-08 12:23:00 yes :) 2016-06-08 12:23:32 http://dl-cdn.alpinelinux.org/alpine/edge/testing/x86_64/i3status-2.10-r1.apk 2016-06-08 12:23:41 and trying to download htat one works. 2016-06-08 12:25:36 yes, apk does find the new package now and the bug is fixed 2016-06-08 12:35:21 clandmeter: afternoon 2016-06-08 12:35:54 hi 2016-06-08 12:36:33 when you have some spare time, could you give some though to creating a url shortening service that uses a spiffy alpine linux style domain? 2016-06-08 12:36:57 what does this error mean when package() function is run? 2016-06-08 12:36:57 >>> ERROR: logstash*: libc.so.6: path not found 2016-06-08 12:37:38 try to install libc6-compat 2016-06-08 12:38:17 logstash provides glibc binaries? 2016-06-08 12:38:17 why this error come up when package() ? 2016-06-08 12:38:33 aha… hm… 2016-06-08 12:38:34 clandmeter, might be. 2016-06-08 12:38:35 I dunno 2016-06-08 12:38:48 logstash definitely doesn’t provide glibc binaries 2016-06-08 12:39:16 logstash works fine even with that error though 2016-06-08 12:39:58 fcolista: this might be useful for you http://haste.fit.cvut.cz/nabiyar 2016-06-08 12:40:30 it’s not complete, because I found that logstash is quite a mess so I don’t wanna use it anymore 2016-06-08 12:40:43 ScrumpyJack: you mean the service itself? 2016-06-08 12:40:51 but before that I wrote package for jruby and its dependencies :) 2016-06-08 12:41:22 clandmeter: yeah, so we can have a nice alpine specific shortening domain hack 2016-06-08 12:41:23 you use java-lz4 2016-06-08 12:41:33 i'm using openjdk-8 2016-06-08 12:41:34 ScrumpyJack: i think we already have it. 2016-06-08 12:41:45 neat, where is it? 2016-06-08 12:41:47 fcolista: and? 2016-06-08 12:41:50 http://dup.pw 2016-06-08 12:41:53 And it works fine 2016-06-08 12:41:55 but it has no interface 2016-06-08 12:42:05 fcolista: logstash runs on jruby 2016-06-08 12:42:07 without the workaround you have in build() 2016-06-08 12:42:10 ncopa created it for our builders 2016-06-08 12:42:18 sorry, prepare() 2016-06-08 12:42:44 ncopa: how does one use dup.pw ? 2016-06-08 12:42:55 ScrumpyJack: but it should be very easy to create one. 2016-06-08 12:42:58 fcolista: maybe it can work 2016-06-08 12:43:20 or you can also use already implemented solutions. 2016-06-08 12:43:21 so far a simple log works 2016-06-08 12:43:54 just run: 2016-06-08 12:43:58 fcolista: java-lz4 is https://github.com/jpountz/lz4-java, this is dependency of logstash, it’s not replacement for openjdk-8… 2016-06-08 12:44:13 /usr/share/logstash&bin/logstash -e 'input { stdin { } } output { stdout {} }' 2016-06-08 12:44:27 try to use something that utilizes lz4 compressions 2016-06-08 12:44:30 compression 2016-06-08 12:44:39 clandmeter: sure, but i like the idea of an alpine linux specific domain hack for geek points 2016-06-08 12:45:11 ScrumpyJack: check my paste solution 2016-06-08 12:45:16 tpaste.us 2016-06-08 12:45:25 fcolista: the fact is that logstash uses lz4-java that bundles binaries for glibc, so it may not work on musl libc; musl is partially binary compatible with glibc, so there’s a change that it will work 2016-06-08 12:45:40 clandmeter: i use that regularly 2016-06-08 12:46:08 the code to make a shortening service should be similar. 2016-06-08 12:46:15 jirutka, ok 2016-06-08 12:46:25 we have logstash in testing repo 2016-06-08 12:46:31 so i wanted to upgrade it 2016-06-08 12:46:34 clandmeter: can you add it to tpaste? 2016-06-08 12:46:46 ? 2016-06-08 12:46:55 this: cd "$builddir"/vendor/bundle/jruby/1.9/gems/jruby-kafka-*-java/lib/net/jpountz/lz4/lz4/ || return 1 2016-06-08 12:46:55 rm 1.* || return 1 2016-06-08 12:46:55 local lz4ver=$(readlink -f /usr/share/java/lz4.jar \ 2016-06-08 12:46:55 | sed -E 's/.*lz4-([0-9.]+).jar$/\1/') 2016-06-08 12:46:55 [ -n "$lz4ver" ] || return 1 2016-06-08 12:46:57 mkdir "$lz4ver" 2016-06-08 12:46:59 cp /usr/share/java/lz4-$lz4ver.jar $lz4ver/ || return 1 2016-06-08 12:47:01 url shortening 2016-06-08 12:47:01 aarh 2016-06-08 12:47:03 sorry 2016-06-08 12:47:03 i mean you can use parts of it to create a new app/service 2016-06-08 12:47:05 http://sprunge.us/jMJg 2016-06-08 12:47:14 jirutka, ^ 2016-06-08 12:47:32 fcolista: uh 2016-06-08 12:47:33 but it would make sense to extend dup.pw 2016-06-08 12:47:37 with this, logstash works with the command i pasted before 2016-06-08 12:47:52 fcolista: logstash bundles jruby… I thought that it’s not acceptable on Alpine… 2016-06-08 12:48:24 fcolista: also we have proper jruby pacakge now, so it’s not even needed to use the bundled one 2016-06-08 12:48:45 that's right 2016-06-08 12:48:53 that APKBUILD should be rewritten 2016-06-08 12:54:57 fcolista: btw logstash also uses java-snappy that also bundles glibc binaries, with this one I’m sure that it doesn’t work on musl, because I’ve already tried it with different application; at least java-snappy is able to use system-provided *.so, so it’s sufficient to just install package java-snappy-native and no hacks are needed 2016-06-08 12:56:44 fcolista: correction, snappy prefers bundled binary by default, you must set `-Dorg.xerial.snappy.use.systemlib=true`… but I’ve already added it to the jruby launch script that is installed with the jruby package 2016-06-08 12:58:52 fcolista: you can read more about problems with Java native extensions on https://github.com/jruby/jruby/wiki/JRuby-on-Alpine-Linux :) 2016-06-08 13:17:08 yeah I'd like to see someone fork logstash in another language 2016-06-08 13:17:36 It pains me to see java everywhere -_- 2016-06-08 13:43:50 coredump: once again, Logstash is written in Ruby, not Java; it runs on JRuby, that is Ruby on JVM 2016-06-08 13:48:09 coredump: Java Virtual Machine has Java in its name, but nowdays it’s mostly for just a historical reasons; there are plenty of languages that runs on JVM and that has nothing common with Java as a language; there are dynamic languages like Groovy, functional languages like Scala and also many languages that uses JVM just as yet another runtime, like Ruby (JRuby), Python (Jython), JavaScript (Nashorn) and many others 2016-06-08 14:05:23 coredump: btw, actually, Logstash can run even on MRI (aka CRuby), but the reason why they prefer JRuby is much better concurrency and performance in general (not talking about memory usage, that is traditionally bad on JVM, and not about startup time, that is really very bad, but that doesn’t matter too much for long running applications) 2016-06-08 14:18:56 jirutka: that's the problem, what pains me the most in Java is definitely the JVM 2016-06-08 14:19:13 coredump? why? 2016-06-08 14:19:42 bad security, bad performances, leaks memory everywhere etc etc 2016-06-08 14:20:22 and most of the big apps require you to run the Oracle version of it 2016-06-08 14:20:45 Elasticsearch to only name it as we were talking about logstash 2016-06-08 14:20:49 bad security? most of the security issues are not in JVM itself, but Java applets which are horrible and not recommended to use for years 2016-06-08 14:22:02 bad performance is partially a myth – JVM itself has great performance (as I wrote that’s the reason why Logstash uses JRuby instead of MRI…), the problem are extremly complex shitty enterprise applications written in Java 2016-06-08 14:23:02 yeah yeah yeah 2016-06-08 14:23:08 nothing new 2016-06-08 14:23:19 JVM itself doesn’t leak memory from what I know, the prbolem with high memory footprint is caused by garbage collection from principle and also a lot of bloat in JVM 2016-06-08 14:23:33 running java stuff is still the main PITA of my carreer 2016-06-08 14:23:41 I agree 2016-06-08 14:24:00 but that’s mostly not a problem of JVM, but applications 2016-06-08 14:24:05 who's doing the GC ? isn't it the JVM itself ? 2016-06-08 14:24:15 Java ecosystem is horrible 2016-06-08 14:24:27 JVM does GC 2016-06-08 14:24:44 GC needs more memory from principle 2016-06-08 14:25:06 needing more memory is something 2016-06-08 14:25:47 using more memory than what is set and swapping like crazy when there's no reason to is another 2016-06-08 14:26:06 > and most of the big apps require you to run the Oracle version of it 2016-06-08 14:26:06 you mean most shitty mostly proprietary enterprise craps? like WebSphere that doesn’t run even on Oracle JVM, but needs IBM JVM 2016-06-08 14:26:07 And sure apps can do some crazy stuff 2016-06-08 14:26:34 but what's the point of the JVM if it can't contain them ? 2016-06-08 14:26:50 oh don't even start with websphere please 2016-06-08 14:27:11 sadly I have to run some quite big enterprishy apps on faculty, like Alfresco and Liferay (the worst shit I’ve ever seen) and they run fine on OpenJDK… however, Liferay needs some patches for it 2016-06-08 14:28:09 got full stack Oracle crappy suite here 2016-06-08 14:28:14 + all atlassian shit 2016-06-08 14:28:31 it’s not fault of JVM that some craps doesn’t run on any anything but Oracle JVM or others… if some app doesn’t run on OpenJDK, then in the most cases it’s because it uses internal APIs that should not be used at all and even Oracle warns devs that they should not use them 2016-06-08 14:28:43 + the nice ElasticSearch you don't get any support if you use OpenJDK 2016-06-08 14:29:00 yeah, politics… 2016-06-08 14:29:04 oh I never said it was the JVM fault 2016-06-08 14:29:18 but you know trends these days 2016-06-08 14:29:47 let's fix this weblogic shit by going full stack on microservices with docker!!! 2016-06-08 14:29:52 it’s like saying: “hey, our product is crap, we can’t guarantee you that it will run on any other JVM than X in version Y” 2016-06-08 14:30:05 uh, don’t even start with docker :( 2016-06-08 14:30:09 yeah microservices in Java! running only on Oracle JVM!! 2016-06-08 14:30:12 wuuuuut 2016-06-08 14:30:16 O_o 2016-06-08 14:30:24 microservices in Java EE… the best joke from Oracle XD 2016-06-08 14:30:41 people need to be hit harder 2016-06-08 14:30:47 it’s extreme example of hidden complexity 2016-06-08 14:31:11 it's my daily crappy working life 2016-06-08 14:31:32 sure, your JAR has only few megabytes, that’s great, isn’t it? you should not care that it needs few hundreds of JAR dependencies on the server side… 2016-06-08 14:32:36 as I said, Java ecosystem is really horrible and ill :( 2016-06-08 14:33:30 they call themselves developpers when they only drag and drop shit in eclipse ... 2016-06-08 14:33:34 my god 2016-06-08 14:33:36 yeah 2016-06-08 14:33:40 exactly 2016-06-08 14:34:08 last time I asked a dev to solve a bug by removing a single chmod in his code 2016-06-08 14:34:25 I have several Java projects on GitHub and I’m sometimes shocked what kinds of developers writes me and how they totally don’t understand what they are doing 2016-06-08 14:34:44 "Oh no can't do, it will take at least 6h of work with major risk of breaking another instance of the same codebase" 2016-06-08 14:34:50 XD 2016-06-08 14:35:01 they don’t have a button in their shitty IDE to do that! 2016-06-08 14:35:19 god damn it if you don't even know how to use the branch feature of your SCM, I can't do anything for you 2016-06-08 14:35:58 because they are afraid of command line, so they use git only via really horrible UI in their IDE 2016-06-08 14:36:11 oh no they don't use Git 2016-06-08 14:36:13 come on 2016-06-08 14:36:18 you are you kidding 2016-06-08 14:36:21 I know git very well, but I’m not able to do anything in these crappy Git GUIs in IDEs 2016-06-08 14:36:25 at most they use SVN 2016-06-08 14:36:40 it’s impossible to understand Git when using only these horrible GUIs 2016-06-08 14:36:41 took them 10y to move out of CVS 2016-06-08 14:36:51 uh 2016-06-08 14:36:56 now SVN > Perforce 2016-06-08 14:37:06 what’s Perforce?! 2016-06-08 14:37:23 I guess the biggest proprietary SCM there is 2016-06-08 14:37:30 even Google is using it 2016-06-08 14:37:46 Nvidia and a lot of _big_ companies 2016-06-08 14:38:53 it's actually really good at handling very big binary files 2016-06-08 14:39:03 heh, it’s funny how much people are still convinced that Google is the best company with the best developers making the best software, so they use it as an example :) 2016-06-08 14:39:08 so it's used by most of the big game studios and the likes 2016-06-08 14:39:44 oh no it was just an example of _big_ company 2016-06-08 14:40:20 I understand, this was only a note aside 2016-06-08 14:44:35 so to sum it up 2016-06-08 14:44:44 I hate java \o/ 2016-06-08 14:45:27 well, Java as an ecosystem, Java as a language or Java as a platform, i.e. JVM? :P 2016-06-08 14:46:30 the whole mix :D 2016-06-08 14:55:51 the JVM as a platform is actually pretty decent 2016-06-08 14:56:15 ^ that are my words :) 2016-06-08 14:57:00 I'm too lazy too read the whole history ;) 2016-06-08 14:57:54 chris: the main part about JVM from me is this: 2016-06-08 14:57:54 Java Virtual Machine has Java in its name, but nowdays it’s mostly for just a historical reasons; there are plenty of languages that runs on JVM and that has nothing common with Java as a language; there are dynamic languages like Groovy, functional languages like Scala and also many languages that uses JVM just as yet another runtime, like Ruby (JRuby), Python (Jython), JavaScript (Nashorn) and many others 2016-06-08 14:57:54 btw, actually, Logstash can run even on MRI (aka CRuby), but the reason why they prefer JRuby is much better concurrency and performance in general (not talking about memory usage, that is traditionally bad on JVM, and not about startup time, that is really very bad, but that doesn’t matter too much for long running applications) 2016-06-08 15:01:04 agreed, once you get rid of the Java, life on the JVM is pretty fun 2016-06-08 15:01:19 yeah 2016-06-08 15:02:40 I'm currently doing Scala at work, the java legacy is not great and that is sort of the JVMs fault, but other than that, it make life quite easy 2016-06-08 15:31:24 jirutka: still around? 2016-06-08 15:31:34 clandmeter: yeah 2016-06-08 15:31:56 which pkgs did you miss? 2016-06-08 15:31:59 on 3.4 2016-06-08 15:32:29 there are missing packages for x86_64 in v3.4 2016-06-08 15:32:29 see irc://chat.freenode.net:6697/#5691 and irc://chat.freenode.net:6697/#5692 2016-06-08 15:32:52 sorry for mess in links, but algibot parsed it correctly :) 2016-06-08 15:34:18 jirutka: did you check the repos? 2016-06-08 15:34:28 cause i can see alpine-desktop-2.5-r1.apk 2016-06-08 15:34:29 yes, yesterday 2016-06-08 15:34:45 in /var/www/localhost/htdocs/alpine/v3.4/community/x86_64 2016-06-08 15:35:01 see here https://pkgs.alpinelinux.org/packages?name=alpine-desktop&branch=&repo=&arch=&maintainer= 2016-06-08 15:35:16 alpine-desktop x86_64 is missing in v3.4 2016-06-08 15:35:24 ah ok 2016-06-08 15:35:30 but its on the repo 2016-06-08 15:35:42 could be tar has issues with it 2016-06-08 15:36:32 ah i got it 2016-06-08 15:36:35 its a meta pkgs 2016-06-08 15:36:38 pkg 2016-06-08 15:36:43 so it wont list a thing 2016-06-08 15:37:15 but then why does it add it for other arch.... makes no sense 2016-06-08 15:38:25 and also user reported that he’s not able to install it 2016-06-08 15:38:40 and the second package is not meta, but the same issue http://bugs.alpinelinux.org/issues/5692 2016-06-08 15:40:10 here is a small list of missing pkgs on v3.4 2016-06-08 15:40:12 http://sprunge.us/UAgV 2016-06-08 15:40:27 i need to go, would be nice if you can check it this list is correct. 2016-06-08 15:40:42 okay, I’ll do it later :) 2016-06-08 15:40:46 then we can try to fix it tomorow 2016-06-08 15:41:04 nice evening 2016-06-08 15:41:09 see ya 2016-06-08 19:10:00 should -dev depend on ? 2016-06-08 19:39:19 jirutka: depends on the details of , but most of the time it's a given 2016-06-08 19:39:38 skarnet: given? like automatically discovered by abuild? 2016-06-08 19:39:57 yes, and "it's a given" means "it goes without saying" as in "yes" 2016-06-08 19:41:19 when you compile software from a tarball, you get a bunch of files; when you package that, some files go into "foo", others into "foo-dev", but since you can't make the files in "foo" depend on the ones in "foo-dev", often it's the other way 2016-06-08 19:41:34 it's pretty rare that the set of files in foo-dev can be totally independent from the set of files in foo 2016-06-08 19:42:59 but if we wanted to be clean, I guess we could have "foo-common", "foo" and "foo-dev" :/ 2016-06-08 19:43:00 if abuild didn’t discover this dependency, shoul I add it explicitly? 2016-06-08 19:43:52 to be more specific, this package https://github.com/alpinelinux/aports/pull/112/commits/49332c9586c987c810f57514271840f03a7aa55e 2016-06-08 19:44:20 try without the dependency, see if stuff breaks when you're compiling something with julia-dev :D 2016-06-08 19:45:04 well, when I add just dsfmt-dev to the julia makdeps, then it fails to build 2016-06-08 19:45:20 … julia fails to build 2016-06-08 19:45:26 …itself XD 2016-06-08 19:51:20 ACTION just noticed that 'who' doesn't show any results, even if you run 'who -a' 2016-06-08 19:51:30 this applies to busybox and to coreutils 2016-06-08 19:51:40 any idea how to get 'who' working? 2016-06-08 19:52:16 did some looking at old images, and it seems 'who' hasn't worked since alpine 2.7 2016-06-08 19:53:39 tdtrask: it's by design. who relies on utmp. musl doesn't implement utmp, because it's a security nightmare. 2016-06-08 19:54:00 ok, makes sense that it's musl related 2016-06-08 19:54:20 any idea how I can get similar info without the security nightmare? 2016-06-08 19:54:43 ps | cut -f1 -d' ' | uniq 2016-06-08 19:54:52 ... if you're root, else grsec won't let you see other people XD 2016-06-08 19:55:21 in other words: no. 2016-06-08 19:55:48 thing is, a grsec-enabled kernel isn't meant to let you see who else in on the machine. 2016-06-08 19:55:48 acf-openssh used to be able to show info about the current ssh sessions 2016-06-08 19:55:53 now it can't 2016-06-08 19:56:12 well, at least not to the same level of detail 2016-06-08 19:57:17 ACTION will play with it to see what I can still display without 'who' 2016-06-08 19:57:25 thanks skarnet 2016-06-08 19:57:43 yw 2016-06-08 19:59:40 skarnet: sorry, I’m doing too many things in parallel now… about dsfmt, if julia needs dsfmt and not just dsfmt-dev during build phase, then should I explicitly declare dependency of dsfmt-dev → dsfmt in the dsfmt abuild? 2016-06-08 20:01:17 not if dsfmt-dev doesn't depend on dsfmt intrinsically. 2016-06-08 20:01:38 in which case you should just make the julia build depend on dsfmt. 2016-06-08 20:02:02 but julia needs dsfmt-dev in the build phase 2016-06-08 20:02:13 it needs both, dsfmt and dsfmt-dev 2016-06-08 20:02:26 then make the julia build depend on both 2016-06-08 20:02:51 okay, thanks for avice :) 2016-06-08 20:03:27 and another question, how much do Alpine care about *.pc? if package doesn’t provide *.pc, should I add it? 2016-06-08 20:03:52 only if packages that depend on it use pkg-config 2016-06-08 20:06:28 what flags should be in *.pc’s `Cflags:`? specifically in the case of dsfmt https://github.com/jirutka/alpine-aports/blob/49332c9586c987c810f57514271840f03a7aa55e/testing/dsfmt/APKBUILD (it doesn’t have proper Makefile) 2016-06-08 20:08:04 sorry, I have to go, the building is closing; I’ll read your response from irc log, thanks in advance! :) 2016-06-08 20:16:55 skarnet: I can determine the currently connected ssh clients using 'netstat' and I can see user and tty from 'ps | grep sshd', but I don't know how to correlate them 2016-06-08 20:17:07 ACTION used to get that info from 'who' :( 2016-06-08 20:20:28 I miss 'last' as well 2016-06-08 20:21:14 ACTION tried that too :( 2016-06-08 20:21:58 most useful command to me 2016-06-08 20:22:13 last | head and I instantly know who broke the server 2016-06-08 20:22:16 :D 2016-06-08 20:22:40 wouldn't that just point at you? :) 2016-06-08 20:24:06 depend 2016-06-08 20:24:19 but I usually don't complain when I break my own servers 2016-06-08 20:24:21 ^^ 2016-06-08 20:33:43 ah ha! 2016-06-08 20:34:11 netstat can output the PID, which can be matched up with ps output 2016-06-08 20:34:23 ACTION learned something new 2016-06-08 21:15:45 jirutka: I don't know the .pc format by heart, I tend to avoid it whenever I can 2016-06-08 21:16:03 less so now that I know pkgconf is a thing, but still... it's unneeded 2016-06-08 21:16:48 (the original pkg-config is a nightmare to build) 2016-06-08 21:17:51 Cflags: The compiler flags specific to this package and any required libraries that don't support pkg-config. If the required libraries support pkg-config, they should be added to Requires or Requires.private. 2016-06-08 21:18:43 the question is what is semantic of this keyword, i.e. when and for what is it used? 2016-06-08 21:20:04 i think that these flags are used when building A that depends on B, when B has .pc file 2016-06-08 21:20:15 *has → provides 2016-06-08 21:20:42 in configure/make scripts, the CFLAGS variable usually has stuff like -I/opt/foobar/include 2016-06-08 21:20:46 but what cflags are actually relevant in this case? just -l, -L, or some others? 2016-06-08 21:20:55 -I 2016-06-08 21:21:02 -L/-l are LDFLAGS 2016-06-08 21:21:10 unless pkg-config doesn't know ldflags 2016-06-08 21:21:34 so all of these: -DNDEBUG -DDSFMT_MEXP=19937 -DDSFMT_DO_NOT_USE_OLD_NAMES -fPIC -O3 -finline-functions -fomit-frame-pointer -fno-strict-aliasing --param max-inline-insns-single=1800 -Wmissing-prototypes -Wall -std=c99 -shared 2016-06-08 21:21:35 are irrelevant here? 2016-06-08 21:22:29 I think that I should omit -W*, -O3, -fPIC… but not really sure about e.g. -DNDEBUG -DDSFMT_MEXP=19937 -DDSFMT_DO_NOT_USE_OLD_NAMES 2016-06-08 21:22:44 those are used to compile dsfmt 2016-06-08 21:22:50 not to compile things that depend on dsfmt 2016-06-08 21:23:04 packages that will use the .pc will provide their own set of flags 2016-06-08 21:23:12 you only need to specify what's specific to dsfmt 2016-06-08 21:24:37 what do you mean, what’s specific to dsfmt? 2016-06-08 21:24:54 sorry, juggling several things at once 2016-06-08 21:25:07 no problem :) 2016-06-08 21:25:20 foo.pc contains the info necessary for packages that depend on foo to find the foo headers/libraries/etc. 2016-06-08 21:25:22 that's all 2016-06-08 21:25:44 aha, that explains a lot! :) 2016-06-08 21:25:45 those packages don't care about the internal details of how foo was built 2016-06-08 21:26:07 I think that I understand it know, thanks! 2016-06-08 21:26:16 *know -> now 2016-06-08 21:27:40 am I the only to see nginx failing to be correctly stopped by its service file ? 2016-06-08 21:28:31 oh indeed the pid file not at the correct place 2016-06-08 22:07:36 clandmeter: I can confirm that all of that packages listed in http://sprunge.us/UAgV are missing, according to https://pkgs.alpinelinux.org 2016-06-09 07:17:40 morning all 2016-06-09 07:20:26 morning 2016-06-09 08:04:27 morning climber(s) 2016-06-09 08:51:20 hey :) 2016-06-09 08:53:53 ho 2016-06-09 09:28:41 quick question folks is there is a rootfs.tar.gz for alpine x86? that can run from memory only? 2016-06-09 09:32:47 it was a strange question, but to K-Line somebody for it... 2016-06-09 09:49:41 clandmeter: hahaha 2016-06-09 11:54:34 hi, I have add basic search in api.a.o, 2016-06-09 11:54:34 try curl https://api.alpinelinux.org/search/packages/category/v3.4:all:x86/name/bas* | jq . | less 2016-06-09 11:55:48 any inputs are welcomed 2016-06-09 18:38:22 hi, i just installed? alpine in a VM. after selecting the disk "vda", and how to use it "sys", an ERROR: libuuid-2.28-r1: network error was printed 2016-06-09 18:38:55 then something like a copying or formatting took place and now i'm back on the prompt without an indication if the install was successful or not 2016-06-09 18:40:31 aha, after a reboot the hd was not detected as bootable so i guess it failed 2016-06-09 18:45:52 so formatting the disk only works if your internet connection is up... interesting concept 2016-06-09 19:07:55 sh4rm4^bnc: which iso did you use? 2016-06-09 19:08:19 alpine-3.4.0-x86_64.iso 2016-06-09 19:08:43 sounds like a bug, it should be self-contained for base install 2016-06-09 19:08:50 although never tried it offline honestly 2016-06-09 19:09:04 I'll take a look tomorrow 2016-06-09 19:09:06 i'm on-and-offline 2016-06-09 19:09:23 usually online, but the wifi conn gets disconnected randomly 2016-06-09 19:09:47 i did select a mirror though when asked 2016-06-09 19:10:06 http://git.alpinelinux.org/cgit/alpine-iso/tree/alpine.packages hm 2016-06-09 19:10:14 and now the install is incredibly slow, it seems to be fetching updated pkgs rather than using stuff from the iso 2016-06-09 19:11:09 so yeah, something during hdd setup is trying to use internet and fails completely if its off in that very moment 2016-06-09 19:11:10 ncopa: ^ you might want to check it 2016-06-09 19:11:31 sh4rm4^bnc: what's in yours /etc/apk/repositories? 2016-06-09 19:11:59 i cannot tell right now, since the installation seems to be stuck at 15% 2016-06-09 19:12:46 probably its trying to fetch stuff from the internet, the connection went off for a few secs, and now its waiting forever for a download to succeed 2016-06-09 19:13:42 it's ca 30 mins since i restarted the install 2016-06-09 19:14:04 probably a bad idea to use any non-iso packages during install 2016-06-09 19:14:24 and do a sysupdate on first boot when internet is on 2016-06-09 19:14:30 I really wish I had any interesting tips for you, but yeah, that's pretty much it 2016-06-09 19:14:44 you will at least have functional base installation 2016-06-09 19:15:00 i retry and see if i can select "none" when asked for mirrors 2016-06-09 19:24:42 another improvement suggestion: after selecting the keyboard layout (before selecting the variant) already load the first layout that matches 2016-06-09 19:25:17 because entering the name of the variant can be a PITA if for example your dash is not where you're used to 2016-06-09 19:27:30 or, print a 1,2,3 style selection 2016-06-09 19:27:43 so you dont have to enter the name 2016-06-09 19:28:55 ok, when selecting "done" in the mirror selection, the hdd formatting fails due to unsatisfiable constraints 2016-06-09 19:29:53 probably i should have selected "e" (edit) and put a magic repo name that points to the iso into the repositories file.. 2016-06-09 19:34:27 this time i edited, and left "/media/cdrom/apks" in the file, but it fails the same way 2016-06-09 20:43:03 this happens when you try to break this stupid systemd vs. SysVinit dichotomy… https://github.com/Yelp/dumb-init/pull/84 2016-06-09 20:43:50 using systemd and SysVinit as only two examples of init system is actully systemd propaganda 2016-06-09 20:47:28 absolutely 2016-06-09 20:47:44 have just a little more patience... just one more year 2016-06-09 20:47:50 and then things will change :) 2016-06-09 20:48:00 how? 2016-06-09 20:48:17 because I'll have the time and money to devote to changing it 2016-06-09 20:49:01 do you really believe that you can break systemd dominance? 2016-06-09 20:49:13 yes. 2016-06-09 20:49:19 It won't happen overnight, or even in a year. 2016-06-09 20:49:26 But slow and steady wins the race. 2016-06-09 20:49:35 I would like to believe that, really 2016-06-09 20:49:52 well, you're using Alpine, that's a good start. 2016-06-09 20:49:57 but historically bad solutions almost always wins :( 2016-06-09 20:50:06 nope, they only win temporarily. 2016-06-09 20:50:34 Linux won the server market against Microsoft, in a crushing way. It only took 10 years. 2016-06-09 20:51:04 Android will win against Apple eventually (I'm not saying Android is good, it's just less closed than iOS). It will just take time. 2016-06-09 20:51:18 systemd will crumble in the end, we just need to be patient and work like ants towards it. 2016-06-09 20:51:25 Android is a good example of very bad solution that wins 2016-06-09 20:51:49 if you know a better alternative for the smartphone world, feel free to advertise it. 2016-06-09 20:52:05 It's technically bad, but it's the least closed-source we have. 2016-06-09 20:52:06 currently all of them are basically dead :( 2016-06-09 20:52:35 WebOS was very interesting 2016-06-09 20:53:06 BlackBerry, closed-source but great 2016-06-09 20:54:02 anyway I don't care about phones... no keyboard, not easy to have a terminal, and it's a client through and through. Not my cup of tea. :P 2016-06-09 20:54:12 Sailfish OS isn’t technically dead, but well, no one uses it 2016-06-09 20:54:20 I agree, I hate phones 2016-06-09 20:54:35 I hate them since I can’t find any phone with decent HW keyboard on the whole shitty marker 2016-06-09 20:54:53 but that’s irrelevant for this topic, I used it only as an example 2016-06-09 20:55:31 My point is that with dedication, patience and technical quality, we can make better solutions win. 2016-06-09 20:55:38 Because ugly solutions annoy *everyone* 2016-06-09 20:55:40 Debian and his child Ubuntu are the most popular Linux distribution… I can’t call them in other name than uter crap 2016-06-09 20:55:50 and they only maintain themselves with inertia. 2016-06-09 20:56:09 That’s the problem, there are a lot of companies that loves ugly solutions 2016-06-09 20:56:14 because they can make money from it 2016-06-09 20:56:33 they could make as much money with good solutions. 2016-06-09 20:56:46 Simplicity is a great virtue but it requires hard work to achieve it and education to appreciate it. And to make matters worse: complexity sells better. ~ Dijkstra (1984) 2016-06-09 20:57:21 Don't give in to conspiracy theories. Sure complexity sells, but today simplicity sells too. 2016-06-09 20:57:34 skarnet: not without corp support 2016-06-09 20:57:40 no, you don’t need consultants, enterprise certificates and others for simple functional solutions 2016-06-09 20:57:43 You don't need to be malicious to write bad software. You just need to be incompetent. 2016-06-09 20:57:56 And companies have a tendency to be extremely good at being incompetent. 2016-06-09 20:58:00 yes 2016-06-09 20:58:17 I’m not saying that there’s always intentions behind it, the most often it’s just incompetent 2016-06-09 20:58:44 but from company perspective, it’s quite clear 2016-06-09 21:00:38 Show a company a solution that's simpler to maintain, that requires less resources (human resources, for them to handle the complexity), and that they can still sell consulting about, and they'll love it 2016-06-09 21:01:16 they want to make money, they don't care how they make it 2016-06-09 21:01:27 they only sell complex solutions because they don't know how to design simple ones 2016-06-09 21:02:10 BTW is it okay to use `&>` in abuild’s pre-install script? it’s a bashism, but ash knows this syntax 2016-06-09 21:02:55 this to be more specific https://github.com/myENA/aports/blob/e03b6321ae5f94feaaa4908280377c2b40f0ab31/testing/gearmand/gearmand.pre-install 2016-06-09 21:04:38 I find it uncanonical to use &>, just write >sth 2>&1 and it will work everywhere 2016-06-09 21:04:58 it’s not my code, I’m coding a code-review 2016-06-09 21:05:02 *doing 2016-06-09 21:05:04 (even windows) 2016-06-09 21:06:12 be harsh in your code reviews, you're not there to please the one you're reviewing, you're there to ensure code quality 2016-06-09 21:06:34 so, firmly suggest >blah 2>&1 instead 2016-06-09 21:06:56 especially if shebang is sh only 2016-06-09 21:07:28 I think that ncopa will not by happy if I’d turn on my pedantic mode when doing code reviews XD 2016-06-09 21:08:19 is he the one you're reviewing? in this case, turn on your pedantic mode full throttle, he will love it :P 2016-06-09 21:08:43 nope 2016-06-09 21:09:07 it’s pull requests from myENA 2016-06-09 21:09:22 re style: I prefer to keep > next to redirection sink 2016-06-09 21:09:38 (without space) 2016-06-09 21:09:43 show contributors that Alpine is strict with code quality 2016-06-09 21:09:52 better define yourself than defend yourself 2016-06-09 21:10:10 przemoc: me too; `>/dev/null` instead of `> /dev/null` 2016-06-09 21:11:04 and I would drop this getent stuff 2016-06-09 21:11:17 it brings nothing, just a noise 2016-06-09 21:12:10 agree 2016-06-09 21:13:09 ACTION haven't sent report about breakage regarding commit adding -S to adduser but not adding group and using -G in adduser call, but it will be sent somewhat this week 2016-06-09 21:13:52 https://wiki.alpinelinux.org/wiki/APKBUILD_Reference#.24pkgname.pre-install shows boilerplate 2016-06-09 21:14:20 and I would advise sticking to it, even regarding arguments orders 2016-06-09 21:14:46 it helps processing stuff 2016-06-09 21:15:05 and reading them in mass 2016-06-09 21:15:07 my comment https://github.com/alpinelinux/aports/pull/122/files#r66523063 2016-06-09 21:24:27 ok, I added my own https://github.com/alpinelinux/aports/pull/122/files#r66524576 2016-06-09 21:24:39 hope you're okay with that, jirutka 2016-06-09 21:25:23 sure I’m okay with that :) 2016-06-09 21:27:13 we redirect stderr to get away with adding user/group already existing, so user is not bothered, but I don't see a point in redirecting stdout 2016-06-09 21:30:37 I only now noticed Cage face :D 2016-06-09 21:36:44 do you believe me that until now I didn’t now that it’s face of Nicolas Cage, yet I’m using this meme often? 2016-06-09 21:37:01 no, I don't believe you 2016-06-09 21:37:08 how you can not now it?! 2016-06-09 21:37:13 s/now/know/ 2016-06-09 21:37:14 I dunno 2016-06-09 21:37:31 maybe I was just not thinking about it 2016-06-09 21:39:01 well, I only very recently learned that one of used-to-be-popular gifs is from Italian Spiderman and that is the title I never heard of 2016-06-09 21:39:23 which gif? 2016-06-09 21:40:05 https://www.youtube.com/watch?v=DJdfca1yf4U gif is the excerpt from this scene 2016-06-09 21:40:24 aha, this one 2016-06-09 21:40:31 I see it in 9GAG comments quite often 2016-06-10 06:21:22 hi all 2016-06-10 06:22:16 hi 2016-06-10 06:41:14 morning climbers. Happy Friday! 2016-06-10 06:53:52 morning 2016-06-10 08:22:30 Hey folks 2016-06-10 08:22:46 I just discovered that the grsec GRKERNSEC_PROC_GID is 30 2016-06-10 08:23:20 ok forget about it 2016-06-10 08:29:02 that's totally awesome 2016-06-10 08:29:16 did you also discover that 2 and 2 are 4 ? :P 2016-06-10 08:29:56 you meany :( 2016-06-10 08:30:22 ACTION laughs manically 2016-06-10 11:19:53 friends is there a way we can create a rootfs for x86 and boot using that instead of an ISO??? 2016-06-10 11:20:34 a read only rootfs with kernel in another partition? Any howtos or suggestions on that? 2016-06-10 11:21:13 depends hugely on your bootloader 2016-06-10 11:21:33 qemu can do that natively, but with real hardware, that's another story 2016-06-10 11:24:18 well one of my boards has an internal emmc with u-boot x86...i created two partitions: one with vmlinuz and another with extracted initramfs 2016-06-10 11:24:33 where do i place the apks??? 2016-06-10 11:25:22 can i place within the second ext4 partition? in /? can i just ask alpine to pick the apks up from that partition using PARTUUID? 2016-06-10 11:43:26 https://wiki.alpinelinux.org/wiki/Installation - look at diskless mode "You'll boot from read-only medium such as the installation CD, a USB drive, or a Compact Flash card" 2016-06-10 11:59:38 Let me try BitL0G1c 2016-06-10 14:20:40 ncopa, are you here? 2016-06-10 14:22:18 jirutka: hi 2016-06-10 14:22:56 sh4rm4^bnc: you should use the alpine-extended.iso if you need offline install 2016-06-10 14:23:14 ncopa: could you please merge https://github.com/alpinelinux/aports/pulls?q=is%3Apr+is%3Aopen+label%3Abackport-it 2016-06-10 14:23:14 and backport them (#5688, #5690)? 2016-06-10 14:24:37 ok 2016-06-10 14:25:45 ncopa: thanks :) 2016-06-10 14:26:18 builder will be busy with the 4.4.13 kernels for a while though... 2016-06-10 14:26:31 i thought i pushed the kernels yesterday 2016-06-10 14:26:36 but apparetnly i didnt 2016-06-10 14:27:22 ncopa: could you please take a look at this https://github.com/alpinelinux/aports/pull/122/files#diff-431da93e83b41d67f44e17212f21322fR15 ? is it okay to generate subpackages libgearman and libgearman-dev for package gearmand? 2016-06-10 14:27:47 I mean, is this the preferred way? 2016-06-10 14:28:08 we normally would do: germand-dev 2016-06-10 14:28:42 are there more libs than libgerman.so.*? 2016-06-10 14:29:20 source="https://launchpad.net/${pkgname}/1.2/${pkgver}/+download/${pkgname}-${pkgver}.tar.gz 2016-06-10 14:29:31 what is "1.2" in there? 2016-06-10 14:29:38 very good question! 2016-06-10 14:30:02 take a look at this https://launchpad.net/gearmand/+download?memo=10&start=10 2016-06-10 14:30:09 “series” 2016-06-10 14:30:59 0.41 is from 1.0 series, 1.0.1 is from 1.0 series, 1.1.0 is from 1.2 series… this makes sense, doesn’t it? XD 2016-06-10 14:31:19 I wince everytime I see a main link to launchpad.net 2016-06-10 14:31:24 i think its the API or ABI 2016-06-10 14:31:42 skarnet: me too 2016-06-10 14:32:05 https://launchpad.net/gearmand 2016-06-10 14:32:25 but still… 2016-06-10 14:32:31 ACTION winces 2016-06-10 14:32:37 ncopa, why you so cruel 2016-06-10 14:32:57 :) 2016-06-10 14:33:28 looks like the do odd/even releases, where odd=development and even=stable 2016-06-10 14:33:38 similar to gnome stuff 2016-06-10 14:34:04 or like Windows? 2016-06-10 14:34:14 ah, I see the source of their inspiration 2016-06-10 14:34:23 in windows i think everything is just "odd" :) 2016-06-10 14:34:29 yeah 2016-06-10 14:34:32 it was the Linux model for a long while :P 2016-06-10 14:34:38 yes 2016-06-10 14:34:47 but odd versions was more odd then even versions (or opposite)? 2016-06-10 14:35:16 lol 2016-06-10 14:35:22 what I’ve just written 2016-06-10 14:35:46 this punc wasn’t even intended 2016-06-10 14:35:53 :) 2016-06-10 14:36:10 one option is to call the package "germand1.2" 2016-06-10 14:36:21 but i dont know if we bother 2016-06-10 14:36:32 I dunno, I don’t even know this software 2016-06-10 14:36:40 i dont think we bother 2016-06-10 14:37:08 it just scares me that it’s on launchpad, but at least it’s not written in Go 2016-06-10 14:38:18 what about libgearman and libgearman-dev? personally I’d name the package gearman, the *.so lib gearman-lib and then standard gearman-dev 2016-06-10 14:38:47 * s/gearman/gearmand/g 2016-06-10 14:38:53 its the name of the tarball 2016-06-10 14:39:16 i'd call it germand, germand-libs and germand-dev 2016-06-10 14:39:47 agree :) 2016-06-10 14:40:18 it might make sense to do: germand, germand-client-libs, germand-libs and germand-dev 2016-06-10 14:40:32 i think i do that with mariadb and postgresql 2016-06-10 14:40:32 their naming is also odd, the sofware is named Gearman, without “d”, but they have gearmand in URL and also tarball 2016-06-10 14:41:08 yes. it seems to be the project name 2016-06-10 14:41:40 if it's a daemon, it makes sense 2016-06-10 14:42:01 gearman = project name, libgearman = client library, and gearmand = server 2016-06-10 14:42:13 +1 2016-06-10 14:42:52 i suppose when they started the project it was only a server 2016-06-10 14:43:00 or they didnt know what the 'd' means :) 2016-06-10 14:45:50 jirutka: i pushed llvm 2016-06-10 14:45:56 i hope you tested it... 2016-06-10 15:36:24 ncopa: yes, I’ve tested it with Julia 2016-06-10 15:37:35 ncopa: btw, we have finally fully working Julia package on Alpine! it was quite a lot of work, but Julia devs helped us a lot :) 2016-06-10 15:37:35 https://github.com/alpinelinux/aports/pull/112 2016-06-10 15:43:42 fcolista, are you here? 2016-06-10 15:48:52 i am giving up slowly on intel edison ...such a shame i cant run alpine on it 2016-06-10 15:49:04 :( 2016-06-10 15:49:04 these days arm sbcs are so much better... 2016-06-10 15:49:30 what’s the main problem with running alpine on intel edison? 2016-06-10 15:50:04 i tried so many way to boot...it just doesnt boot...crashes.. 2016-06-10 15:50:20 kernel crashes? or userspace? 2016-06-10 15:50:26 the kernel is too old 3.10.x for alpine to work.... kernel crashes 2016-06-10 15:50:38 squashfs etc arent so good... 2016-06-10 15:50:56 the latest kernel doesn’t support intel edison? 2016-06-10 15:51:23 no it doesnt...they have a huge patched version of kernel 3.10 running ....yocto build 2016-06-10 15:51:35 uh 2016-06-10 15:51:39 sometimes i feel its worse than arm ecosystem these days 2016-06-10 15:52:39 so their patched kernel works, but not with Alpine? 2016-06-10 15:53:23 yes with Yocto only it works...somehow it doesnt boot alpine....i have tried all kinds of ways... 2016-06-10 15:53:37 how exactly does it crash? 2016-06-10 15:53:42 i tried similar to my guide https://wiki.alpinelinux.org/wiki/DIY_Fully_working_Alpine_Linux_for_Allwinner_and_Other_ARM_SOCs 2016-06-10 15:53:46 well 2016-06-10 15:54:09 let me show you a couple of errors 2016-06-10 15:54:54 http://paste.ubuntu.com/16884163/ 2016-06-10 15:55:08 http://paste.ubuntu.com/16884145/ 2016-06-10 15:56:07 http://paste.ubuntu.com/16860696/ 2016-06-10 15:56:58 those are few of 100s of combinations i tired...old kernels, new ones, what not 2016-06-10 15:57:22 internal emmc ...has a custom u-boot ... 2016-06-10 15:58:59 hm, it fails to mount rootfs 2016-06-10 15:59:47 initramfs is yours or it’s from Yocto? 2016-06-10 16:00:11 its mine as well I am trying initramfs-vanilla from alpine linux...yocto doesnt have any initramfs 2016-06-10 16:01:04 the only saving grace is this guide however https://edison.internet-share.com/wiki/Using_a_vanilla_Linux_kernel_with_Intel_Edison 2016-06-10 16:01:34 vanilla linux doesnt have any patches so neither it can recognize the internal emmc neither external sdio card hence i cant load my apks 2016-06-10 16:02:43 i hate to give up but i have no choice...i havent slept for like days because of this 2016-06-10 16:02:59 apks folder i meant* 2016-06-10 16:03:30 have ever you managed to mount rootfs with this setup? 2016-06-10 16:03:56 meaning? 2016-06-10 16:03:58 Alpine Init 3.0.4-r2 mount: mounting shm on /dev/shm failed: Invalid argument * Loading boot drivers: ok. * Mounting root: [ 3.162712] nlplug-findfs (877) used greatest stack depth: 6688 bytes left [ 3.245491] mdev (1212) used greatest stack depth: 6564 bytes left [ 3.773542] clocksource: tsc: mask: 0xffffffffffffffff max_cycles: 0xe642f98297, max_idle_ns: 881590439301 ns [ 3.790296] clocksource: Switched to clocksource tsc 2016-06-10 16:04:24 mount: mounting PARTUUID=333a128e-d3e3-b94d-92f4-d3ebd9b3224f on /sysroot failed: No such file or directory failed. 2016-06-10 16:04:39 u-boot very much shows that PARTUUID 2016-06-10 16:05:00 http://paste.ubuntu.com/16891902/ 2016-06-10 16:05:11 thats the complete paste bin when i tried vanilla kernel 2016-06-10 16:05:59 I mean, is this the last step when it fails, or have you find some combination to get further? 2016-06-10 16:06:25 i tried lots of them however this is only combination where atleast i get a prompt 2016-06-10 16:06:32 all other combinations crash 2016-06-10 16:06:46 what does /dev/shm failed: invalid argument mean? 2016-06-10 16:06:58 the point is if the main problem is really in mounting (or finding) the rootfs or if the problem lies elsewhere 2016-06-10 16:07:15 hmmmm 2016-06-10 16:08:06 the /dev/shm is mountpoint for shared memory, this is not blocking, you can run system without it 2016-06-10 16:08:22 http://ark.intel.com/products/70103 that is Intel cpu 2016-06-10 16:09:28 here is the https://github.com/01org/edison-linux/ development 2016-06-10 16:09:58 there is still work in progress from 3.10 to 3.19 2016-06-10 16:10:12 wait a moment, since it’s trying to mount /dev/shm, it means that you’re running Alpine’s init so you must have rootfs already mounted, this is not in initramfs (?) 2016-06-10 16:11:23 no actually since u-boot recognizes the mmc...i am loading the initramfs file into memory along with kernel 2016-06-10 16:11:31 that is where it is find its init 2016-06-10 16:12:03 lines 33 and 36 from here http://paste.ubuntu.com/16891902/ 2016-06-10 16:12:20 so the init in your initramfs is actually Alpine’s init, i.e. OpenRC? o.O 2016-06-10 16:12:29 yes 2016-06-10 16:12:31 ofcours 2016-06-10 16:12:35 ofcourse 2016-06-10 16:13:22 its from alpine iso 2016-06-10 16:13:22 init in initramfs is usually just a simple script that mounts rootfs and then execute the OS’ init… 2016-06-10 16:13:38 also note line 226 Initramfs unpacking failed: junk in compressed archive 2016-06-10 16:14:00 hmmm 2016-06-10 16:14:36 for example this is how init inside initramfs can look like https://gist.githubusercontent.com/jirutka/990d25662e729669b3ce/raw/479a7b1a91829f2c5a3ca0c84b2ce0cef04a69aa/init 2016-06-10 16:14:36 that initramfs is actually from latest alpine-linux iso 2016-06-10 16:14:45 hmm 2016-06-10 16:15:45 the cpu performance of this humble dual core atom is less than 1watt ...0.7 watts roughly at peak and matches the cpu performance of raspberry pi 3 in some cases 2016-06-10 16:15:53 TDP* 2016-06-10 16:16:36 should i open a ticket? will we support edison? 2016-06-10 16:17:07 give me a sec 2016-06-10 16:17:39 I’ll take a look at the Alpine’s initramfs 2016-06-10 16:17:43 sure 2016-06-10 16:18:23 for your info ...this how centos built a working rootfs 2016-06-10 16:18:23 https://seven.centos.org/2015/08/a-flashable-centos-image-for-the-intel-edison/ 2016-06-10 16:21:50 hmm, this is init from Alpine’s initramfs-virtgrsec http://haste.fit.cvut.cz/ugeyuju.sh 2016-06-10 16:22:45 i think i seen it before 2016-06-10 16:23:04 have you tried boot without initramfs? I don’t think that you actually need to use initramfs with u-boot 2016-06-10 16:23:28 what i tried was creating an ext4 file using dd 2016-06-10 16:23:38 u-boot natively supports ext4, doesn’t it? 2016-06-10 16:23:39 filling it with extracted initramfs 2016-06-10 16:23:42 yes 2016-06-10 16:24:04 and it does boot however there are lots of errors where it says read only 2016-06-10 16:24:06 cannot write 2016-06-10 16:24:37 instead of copying do i need to "install" to an ext4? 2016-06-10 16:25:24 i want a diskless install basically so i extracted initramfs and filled it in ext4 2016-06-10 16:26:36 I suggest you to try the most simple setup to just boot the Alpine with writable rootfs 2016-06-10 16:26:49 and how does it then pick up the apks folder??? where do i place the apks folder in that case 2016-06-10 16:27:38 to be honest, I’ve never tried diskless mode 2016-06-10 16:28:02 but the bigger problem is to actually boot the system 2016-06-10 16:28:11 then you can solve diskless 2016-06-10 16:28:44 the rootfs is writable jirutka: 2016-06-10 16:30:08 one noob question 2016-06-10 16:30:16 how do i prepare a writable rootfs?? 2016-06-10 16:30:25 in an ext4 file 2016-06-10 16:31:02 try this: create ext4 or other FS compatible with u-boot on the MMC card, copy content of the Alpine’s ISO image to that FS, replace alpine’s kernel with your kernel, and configure u-boot to mount that FS and load the kernel 2016-06-10 16:31:26 aha 2016-06-10 16:31:31 let me try right away 2016-06-10 16:31:38 I belive that this is the most simple setup you can start with 2016-06-10 16:32:06 initramfs, diskless setup etc. just adds another complexity that you can solve later 2016-06-10 16:32:56 are you here for a sometime? 2016-06-10 16:33:03 i will quickly test and report? 2016-06-10 16:33:10 yes, I’m here quite often 2016-06-10 16:33:48 excellent i just prepared an ext4 hdd file...let me burn it to you intel edison and report back 2016-06-10 16:34:17 burn it? you have MMC, don’t you? 2016-06-10 16:34:36 i mean its emmc 2016-06-10 16:34:42 internal 2016-06-10 16:34:51 it doesnt support microsd card etc 2016-06-10 16:35:01 so i need to flash it ... 2016-06-10 16:35:11 huh, maybe I should really look what Intel Edison actually is XD 2016-06-10 16:35:45 is this internal memory normally writable? 2016-06-10 16:36:29 yes yes 2016-06-10 16:36:49 hmm, very nice board, it’s even smaller than Arduino Uno 2016-06-10 16:37:11 amazing performance/watt i have come across in a long long time 2016-06-10 16:37:41 http://srlm.io/2015/03/02/intel-edison-benchmarks/ 2016-06-10 16:37:47 hmm, I’ll ask me colleague to buy one of these for experiments, then I’ll be able to try it myself 2016-06-10 16:38:42 it’s much more expensive than RPi 2016-06-10 16:38:42 http://www.microcenter.com/product/458394/Edison_Breakout_Board_Kit 2016-06-10 16:38:57 it could be cheaper in amazon 2016-06-10 16:53:20 okie flashing the board now as we speak 2016-06-10 16:57:38 uh, it costs a lot 2016-06-10 17:01:08 jirutka: http://paste.ubuntu.com/17176544/ 2016-06-10 17:01:09 well I guess if you need the form factor and performances 2016-06-10 17:01:15 seems not that bad 2016-06-10 17:05:25 okie i have added init=/init to the bootargs and it does go forward however take a look here what it says 2016-06-10 17:05:32 oneinsect: why do you have this systemd crap in kernel cmdline? `systemd.unit=multi-user.target` 2016-06-10 17:06:05 shucks i can remove it 2016-06-10 17:06:32 http://paste.ubuntu.com/17176875/ 2016-06-10 17:06:43 jirutka: can you take a look at that pastebin 2016-06-10 17:07:06 I’d also try to replace PARTUUID=012b3303-34ac-284d-99b4-34e03a2335f4 with device name/path 2016-06-10 17:07:19 ooh 2016-06-10 17:08:34 let me try and the apks folder is within the / how do i tell alpine to look there? 2016-06-10 17:08:39 alpine_dev=? 2016-06-10 17:09:02 have you copied content of Alpine’s ISO image to MMC? 2016-06-10 17:10:32 let me paste bin everything in one 2016-06-10 17:10:53 also remove: init=/ini 2016-06-10 17:11:16 actually, try to boot it just with: root=PARTUUID=012b3303-34ac-284d-99b4-34e03a2335f4 rootfstype=ext4 console=ttyMFD2 2016-06-10 17:11:32 just to try what it will do 2016-06-10 17:12:50 maybe add init=/sbin/init 2016-06-10 17:13:15 it should be the default although 2016-06-10 17:15:24 can you please take a look here and tell me what should be my device dev ? 2016-06-10 17:15:25 http://paste.ubuntu.com/17177203/ 2016-06-10 17:15:44 coredumb: yeah, but for fooling around with some community, not really 2016-06-10 17:17:42 oneinsect: do you really have /init? Alpine’s init is in /sbin/init 2016-06-10 17:17:53 oooh but i have init 2016-06-10 17:18:23 may be symbolic link? i havent checked 2016-06-10 17:21:53 hmm, try to set "root=/dev/mmcblk0p8" 2016-06-10 17:22:42 barthalion: agreed :) 2016-06-10 17:23:15 the problem is that busybox’s blkid doesn’t recognize names like PARTUUID etc. 2016-06-10 17:23:25 will test right away yes i noticed that ... 2016-06-10 17:23:25 that’s why it fails here 2016-06-10 17:23:39 from the log it seems that device path is /dev/mmcblk0p8 2016-06-10 17:23:46 I hope that I’ll work with that 2016-06-10 17:23:54 trying 2016-06-10 17:25:40 got a strange error http://paste.ubuntu.com/17177700/ 2016-06-10 17:26:05 interesting 2016-06-10 17:26:19 some Usage: switch_root [-c /dev/console] NEW_ROOT NEW_INIT [ARGS] 2016-06-10 17:26:21 whats that 2016-06-10 17:27:44 this shouldn’t be here 2016-06-10 17:28:02 have you deleted initramfs from MMC? 2016-06-10 17:28:17 I’ll take a look what exactly is on Alpine’s ISO 2016-06-10 17:28:54 yesss there is not initramfs-vanilla file or anything...everything deleted 2016-06-10 17:29:03 infact this is how it looks 2016-06-10 17:29:07 give me a sec 2016-06-10 17:30:03 http://paste.ubuntu.com/17177888/ this will be helpful 2016-06-10 17:30:53 I’ve just booted Alpine’s ISO on my infra 2016-06-10 17:32:02 where the heck nlplug-findfs comes from 2016-06-10 17:33:05 hm, could you send me ls /etc/init.d? 2016-06-10 17:34:37 well its just hanging after that prompt ...let me try 2016-06-10 17:35:38 with PARTUUID atleast i get a prompt ..hold on please i will pastebin it 2016-06-10 17:37:43 ls: /etc/init.d: No such file or directory 2016-06-10 17:40:54 ah, I’m stupid 2016-06-10 17:41:26 it was nonsense to copy content of Alpine’s ISO 2016-06-10 17:41:36 it’s not done as usual 2016-06-10 17:41:42 which means?? 2016-06-10 17:42:18 however generic arm image is different 2016-06-10 17:42:19 http://dl-cdn.alpinelinux.org/alpine/v3.4/releases/armhf/alpine-uboot-3.4.0-armhf.tar.gz 2016-06-10 17:42:20 why so? 2016-06-10 17:42:39 ahaa 2016-06-10 17:43:42 okay, different way 2016-06-10 17:44:17 yes please i am all ears 2016-06-10 17:45:02 what we need is to get standard alpine rootfs, but that’s not what this tarball, nor ISO image contains 2016-06-10 17:45:21 where do we get it from? 2016-06-10 17:46:40 oneinsect: nowhere 2016-06-10 17:46:43 I usually do this like https://github.com/jirutka/aports/blob/edge/.travis/install-alpine 2016-06-10 17:46:49 you can use apk or apk-static directly 2016-06-10 17:46:57 and install base packages to a directory 2016-06-10 17:47:04 that's I did every Alpine install since 2.6 2016-06-10 17:47:06 but will it work for different arch? 2016-06-10 17:47:10 me too 2016-06-10 17:47:16 this board is x86, right? 2016-06-10 17:47:32 so it's not a problem, just tell apk to use x86 or x86_64 repo 2016-06-10 17:47:43 wait a moment, yes, it seems to be x86, so why is oneinsect using armhf image? 2016-06-10 17:47:52 he is not 2016-06-10 17:47:53 i am not using armhf 2016-06-10 17:47:58 he points to an example of rootfs 2016-06-10 17:48:02 aha 2016-06-10 17:48:08 okay, then it’s simple 2016-06-10 17:48:16 yes i am showing an example how they are packed the armhf images 2016-06-10 17:48:20 it is an x86 cpu 2016-06-10 17:48:42 ah, so you do the same for building in travis 2016-06-10 17:48:49 but it doesnt even boot barthalion: so cant tell it 2016-06-10 17:48:50 download http://dl-cdn.alpinelinux.org/alpine/v3.4/main/x86_64/apk-tools-static-2.6.6-r1.apk 2016-06-10 17:49:17 okie let me get my alpine vm up 2016-06-10 17:49:29 no need to, if you run any linux 2016-06-10 17:49:39 ubuntu 14.04 x64 2016-06-10 17:49:40 yes 2016-06-10 17:49:50 apk-tools-static works very well from any distribution 2016-06-10 17:49:51 run: apk.static --arch x86 --root /tmp/alpine --allow-untrusted --update-cache --initdb --no-progress add alpine-base 2016-06-10 17:50:00 this should install base Alpine rootfs into /tmp/alpine 2016-06-10 17:50:18 apk file is just gzipped tarball 2016-06-10 17:50:27 404 error 2016-06-10 17:50:36 let me try finding another mirror 2016-06-10 17:50:45 s/dl-cnd/nl/ 2016-06-10 17:51:01 what the hell is with repository? 2016-06-10 17:51:27 http://nl.alpinelinux.org/alpine/v3.4/main/x86_64/apk-tools-static-2.6.7-r0.apk 2016-06-10 17:51:47 saved it 2016-06-10 17:51:53 that cdn never works for me… 2016-06-10 17:52:30 should ungzip it? 2016-06-10 17:52:32 i 2016-06-10 17:53:28 okie extracted and running the above command 2016-06-10 17:53:51 holy..... ./apk.static: cannot execute binary file: Exec format error 2016-06-10 17:54:20 Linux EDISON 3.13.0-24-generic #46-Ubuntu SMP Thu Apr 10 19:08:14 UTC 2014 i686 i686 i686 GNU/Linux 2016-06-10 17:54:38 may be i need the x86 package ? 2016-06-10 17:54:47 apt-get install ia32-libs 2016-06-10 17:55:38 E: Unable to locate package ia32-libs 2016-06-10 17:55:57 okay, I got my beard now 2016-06-10 17:56:11 lol 2016-06-10 17:56:33 try to install libc6:i386 and libstdc++6:i386 2016-06-10 17:56:50 forgot that Ubuntu is mutliarch now 2016-06-10 17:56:57 multi even 2016-06-10 17:57:27 you can actually use x86_64 package as well 2016-06-10 17:59:33 ERROR: unsatisfiable constraints: alpine-base (missing): required by: world[alpine-base] 2016-06-10 18:00:19 jirutha: any help? 2016-06-10 18:00:32 he timeouted 2016-06-10 18:00:54 add --repository http://nl.alpinelinux.org/alpine/v3.4/main to parameters 2016-06-10 18:03:36 great it works 2016-06-10 18:03:49 it installed ..what would be the next step? 2016-06-10 18:04:00 so this is your rootfs, I guess 2016-06-10 18:04:18 ooh great i will try it out then 2016-06-10 18:04:19 copy it to where you copied iso before 2016-06-10 18:04:33 sure i will try and report back 2016-06-10 18:04:35 not sure what else jirutka was going to recommend 2016-06-10 18:05:27 clandmeter: did you fix that missing packages issue? 2016-06-10 18:26:38 so if you already have rootfs, copy it to MMC and try to boot it ;) 2016-06-10 18:31:56 yesss 2016-06-10 18:32:18 doing it now as we speak jirutka1: 2016-06-10 18:47:36 oneinsect: and does it blend? i mean, boot :) 2016-06-10 18:51:09 hello here is the pastebin http://paste.ubuntu.com/17181005/ 2016-06-10 18:51:27 it seems to hang here .... * Caching service dependencies ... [ ok ] 2016-06-10 18:51:42 does it depend on overlayfs, squashfs ??? 2016-06-10 18:54:13 after hanging ...i think after 30 seconds an internal watch dog resets the board...no console yet 2016-06-10 18:58:58 be back in a while folks 2016-06-10 19:57:26 oneinsect: no, base system doesn’t depend on overlayfs or similar 2016-06-10 19:58:36 this looks much better than previous! your FS is mounted RW and OpenRC is running 2016-06-10 20:00:23 hm, no, it seems that FS is RO 2016-06-10 20:01:27 there’s some problem with adjusting hwclock 2016-06-10 23:47:30 jirutka, didn't even think to check if you all had an irc channel 2016-06-10 23:48:54 I'm rebuilding gearman now with the new package names and added another patch to fix a warning about undefined __WORDSIZE ... found an example in a patch ncopa committed to tvheadend a few years ago to fix a similar issue 2016-06-11 00:01:21 pushed 2016-06-11 00:23:33 it now builds "gearmand" "gearman-libs" and "gearman-dev" 2016-06-11 02:07:38 barthalion: i am struck here http://paste.ubuntu.com/17181005/ 2016-06-11 02:07:40 any ideas? 2016-06-11 02:09:50 is this your video ? https://asciinema.org/a/3361 2016-06-11 13:17:00 hey :) 2016-06-11 13:31:20 clandmeter: firefox 47 got released 2016-06-11 15:33:20 clandmeter, filed update for your package https://github.com/alpinelinux/aports/pull/127 2016-06-11 21:25:55 Hello, there are couple of LLVM patches for 3.8 branch. Can those be upstreamed? 2016-06-11 21:27:22 three patches in total: http://git.alpinelinux.org/cgit/aports/tree/main/llvm?id=d9b69fee3569993ce5f5981c523098a2e428b03b 2016-06-11 21:27:52 four (i miscounted) 2016-06-11 23:46:51 hey :) 2016-06-11 23:46:54 how are you guys? 2016-06-11 23:46:56 hi 2016-06-11 23:47:12 i am sooo close to get rust working on alpine 2016-06-11 23:47:16 ooh 2016-06-11 23:47:23 really? that’s awsome! 2016-06-11 23:47:26 tell me more! 2016-06-11 23:47:35 well, i hope so ... it looks promissing 2016-06-11 23:47:42 algobot is also interesting, isn’t he? \o/ 2016-06-11 23:47:47 *interested 2016-06-11 23:47:47 but i need clean build environment 2016-06-11 23:48:02 what is better, lxc or that docker crap? 2016-06-11 23:48:08 LXC 2016-06-11 23:48:16 i thought so 2016-06-11 23:48:44 or you can try Vagga, it looks very interesting and it’s written in Rust! ;) https://github.com/tailhook/vagga 2016-06-11 23:49:03 hahahahaha 2016-06-11 23:49:03 its author is tailhook, this person is very active in Rust 2016-06-11 23:49:19 i think that would be a chicken <-> egg problem *g* 2016-06-11 23:49:26 eh, right XD 2016-06-11 23:50:05 it would be even worse, because Rust itself is a chicken – egg problem, isn’t it? XD 2016-06-11 23:50:23 so what’s your approach? cross compilation? 2016-06-11 23:51:58 hardcore editing and patching of stuff i dont understand *g* 2016-06-11 23:52:12 that sounds very familiar to me XD 2016-06-11 23:53:05 but I’ve managed to bring Julia to Alpine with the same approach, so it works! XD 2016-06-11 23:53:15 julia? 2016-06-11 23:53:19 yeah! 2016-06-11 23:53:24 you haven’t heard about it yet? 2016-06-11 23:53:33 https://pkgs.alpinelinux.org/package/edge/testing/x86_64/julia 2016-06-11 23:53:42 actually no, i was busy with real life crap the last couple days 2016-06-11 23:54:13 it was quite a lot work, but Julia devs helped us a lot! :) https://github.com/alpinelinux/aports/pull/112 2016-06-11 23:54:42 however, it was still much easier than Rust, because Julia doesn’t need Julia to compile :) 2016-06-11 23:54:59 looks a lot like lua from the syntax 2016-06-11 23:55:10 yeah, it’s inspired by Lua 2016-06-11 23:55:34 hehe, thought so 2016-06-11 23:55:38 but a good choice 2016-06-11 23:55:43 lua is a good language 2016-06-11 23:55:52 it’s a lang designed for mathematics 2016-06-11 23:56:10 like r? 2016-06-11 23:56:22 it’s aim is to replace shitty R and also Python, that is quite slow for numeric calculations 2016-06-11 23:56:33 python is always slow ... 2016-06-11 23:56:51 Julia has built-in many math libraries for numerical calculations 2016-06-11 23:56:52 some people blame the GIL, some agree that the langue just sucks 2016-06-11 23:57:01 julia looks very clean 2016-06-11 23:57:11 yeah 2016-06-11 23:57:18 but the package is very huge 2016-06-11 23:57:23 yes, it is :/ 2016-06-11 23:57:34 are all the /usr/share/*.jl files needes? 2016-06-11 23:57:44 or could they go into a docs folder? 2016-06-11 23:57:49 these are stdlib 2016-06-11 23:57:57 in /usr/share?? 2016-06-11 23:58:20 yes, because they are not platform depent, so it should be in /usr/share 2016-06-11 23:58:41 ah, its noarch 2016-06-11 23:58:45 i understand 2016-06-11 23:58:54 but maybe you have a good point 2016-06-11 23:59:10 because Julia compiles a lot stuff to sys.so 2016-06-11 23:59:17 I’m not sure if it contains even stdlib 2016-06-11 23:59:22 I’ll ask some Julia dev 2016-06-11 23:59:53 thats why i got confused, because it ships /usr/lib/julia/libjulia.so, /usr/lib/julia/sys.so and /usr/lib/julia/libRmath-julia.so 2016-06-12 00:00:07 but if its noarch stuff you are correct sir 2016-06-12 00:00:36 they are also some tests in /usr/share/julia 2016-06-12 00:00:41 hmmm, this lxc crap is going on my nerves ... 2016-06-12 00:00:46 thy did it override my tar? 2016-06-12 00:00:53 now its gnutar ... wtf??? 2016-06-12 00:01:29 well, I was happy that it finally works and doesn’t throw Illegal instruction or segfault, so maybe I’ll try to optimize the pacakge later… I should ask Julia devs about what files are really needed 2016-06-12 00:01:42 what version of LXC do you use? 2016-06-12 00:01:49 but well done job !! 2016-06-12 00:01:53 I’m still on 1.1.x 2016-06-12 00:01:59 packaging a language is the hardest job there is! 2016-06-12 00:02:02 you did good 2016-06-12 00:02:06 :) 2016-06-12 00:02:45 I’ve also fixed R package before this one 2016-06-12 00:02:52 R is really messy 2016-06-12 00:03:01 NOT AS MESSY AS LXC 2016-06-12 00:03:03 ..... 2016-06-12 00:03:10 why is there so much bad software out there? 2016-06-12 00:03:30 I dunno :( the problem is that there are only bad alternatives in container virtualization 2016-06-12 00:03:43 but what problem do you have? LXC works quite well for me 2016-06-12 00:05:12 hmm, I have still running a testing container on OpenVZ, I’ve created it for Julia devs for debugging problem on Alpine 2016-06-12 00:05:32 i have no network adapters 2016-06-12 00:05:38 if you give me your SSH key, I can give you access to it 2016-06-12 00:05:40 and i cannot use a bridge 2016-06-12 00:05:47 why you cannot use a bridge? 2016-06-12 00:06:03 you will not beleave it if i tell you 2016-06-12 00:06:11 ? 2016-06-12 00:06:24 you’ve burdned all bridges behing you? XD 2016-06-12 00:07:29 my ethernet chip thinks its loosing a link as soon as i do a bridge on it with a linux kernel ... i have to remove my pc power for a few secounds to get the ethernet chip working again 2016-06-12 00:07:35 i have absolutly no idea why 2016-06-12 00:07:45 must be something with that onboard chip 2016-06-12 00:07:49 just create a standalone bridge 2016-06-12 00:07:56 do not connect it to a physical network card 2016-06-12 00:08:14 create a separate network on this bridge and route the traffic to your network card 2016-06-12 00:09:02 I almost always prefer this way, it’s imho better manageable and clean 2016-06-12 00:10:21 how do you link your aports repos into your lxc root? 2016-06-12 00:12:02 something like this: 2016-06-12 00:12:02 `lxc.mount.entry = /var/cache/distfiles var/cache/distfiles none bind,create=dir 0 0` 2016-06-12 00:12:27 and how does it handle the chmod? 2016-06-12 00:12:37 who would own a file created in the lxc on the outside? 2016-06-12 00:12:43 it depends… 2016-06-12 00:13:13 if you’re using user-ns, so you have shifted uids/gids in the container, then it’s a little complicated 2016-06-12 00:13:44 uids/gids are not mapped for these binds 2016-06-12 00:15:13 however, you can create a hole in your mapping, e.g. map 0…999 with offset 100000, 1000…1999 with offset 0 and then 2000…65536 with offset 1000000 2016-06-12 00:15:27 that’s what I’m doing for user homes 2016-06-12 00:16:16 but if you need to just test something, then it’s imho not worth the trouble, just disable uid/gid shifting 2016-06-12 00:16:35 another way is to use NFS or similar to mount disk into countainer 2016-06-12 00:17:09 or if the disk is used only from countainers with the same offset, then you can just leave it 2016-06-12 00:24:31 leo, are you still here? 2016-06-12 09:31:38 in aports, usually how long it takes for a package to get promoted from testing repo to main? 2016-06-12 09:34:16 peterj_: as sson as you report it as working 2016-06-12 09:36:28 barthalion: bpye ported libunwind, and i can confirm that CoreCLR builds on Alpine with that. context: https://github.com/dotnet/coreclr/pull/5731 2016-06-12 09:39:11 sure, I'll do some minor changes to APKBUILD and move it to main 2016-06-12 09:41:13 Thanks! also, lib lttng-ust master branch builds on Alpine but there is no package. I am currently building lldb 3.8 and it is going fine so far 17%.. that is the last piece of puzzle, once lldb builds i will run coreclr PAL tests and can confirm the status here all three libs: libunwind (available in testing repo), lldb (unavailable) and lttng-ust (unavailable). :) 2016-06-12 09:50:08 barthalion: this person sent a patch for lttng-ust to aports: http://lists.alpinelinux.org/alpine-aports/3028.html, which is pending review. But now all those fixes are upstreamed: https://github.com/lttng/lttng-ust/commits/master?author=mjeanson, hence so no patch required to build master branch on Alpine. perhaps that APKBUILD file in that proposed patch can be used to land it in testing? 2016-06-12 09:51:02 are there any chances for a release instead? 2016-06-12 09:51:49 ah, I can just backport patches 2016-06-12 09:51:53 yeah, that makes sense 2016-06-12 09:54:00 barthalion: apparently, his fixes are also available in rel-2.8 branch https://github.com/lttng/lttng-ust/commits/stable-2.8?author=mjeanson 2016-06-12 09:54:10 libunwind is in main now 2016-06-12 09:54:24 great! thank you! :) 2016-06-12 09:54:50 bpye: FYI ^ :) 2016-06-12 09:55:52 I need to run an errand now, so I'll take a look at adapting lttng-ust later today 2016-06-12 09:58:19 barthalion: no problem. meanwhile I will continue to lookafter lldb build :D 2016-06-12 15:47:14 skarnet, are you here? 2016-06-12 15:47:50 I'm here for 5 more minutes, so hurry :D 2016-06-12 15:48:23 skarnet: I just wanna know your opinion about https://github.com/JuliaLang/julia/pull/16881#issuecomment-225444288 :) 2016-06-12 15:51:46 Thats interesting 2016-06-12 15:52:07 not sure why you're asking me, I just happen to be here :D 2016-06-12 15:52:10 to provide you more context, Julia’s build system automatically fetches and builds ~18 dependencies! hopefully, it can be disabled and forced it to use system-provided dependencies, that’s what I’ve done https://github.com/alpinelinux/aports/blob/master/testing/julia/APKBUILD#L94-L112 2016-06-12 15:52:17 if you want an official Alpine answer, wait for ncopa's 2016-06-12 15:52:37 as usual, both sides are kinda right 2016-06-12 15:52:48 yeah, I’m asking you about technical merit of the thing 2016-06-12 15:53:24 I agree with tkelman that a packaging system should make it possible and easy to have several version of a piece of software side by side 2016-06-12 15:53:37 of course 2016-06-12 15:53:40 and that's one of the reasons why I advocate slashpackage and dislike deb/rpm/... 2016-06-12 15:54:07 but bundling dependencies is generally a no-no 2016-06-12 15:54:20 you should document your dependencies, with the precise versions of each 2016-06-12 15:54:38 but bundling them duplicates maintenance 2016-06-12 15:54:57 I’d also prefer static linking and bundles instead of traditional overcomplicated and hard-to-maintain linux approach, but Alpine is also about small size and bundling all dependencies is against this 2016-06-12 15:55:23 static linking is orthogonal to this 2016-06-12 15:56:29 they’re not bundling it in their code base, but when you build Julia from sources, it downloads source tarballs of these ~18 dependencies, builds them and somehow install with julia 2016-06-12 15:56:38 the really correct solution is to document exactly what the dependencies are; doing the build is a packager's job 2016-06-12 15:56:50 they have it 2016-06-12 15:57:07 if they want to provide working binaries/tarball/whatever, they can put on a packager's hat for a moment 2016-06-12 15:57:08 the problem is that they’re still using LLVM 3.7 2016-06-12 15:57:14 and we have LLVM 3.8 in v3.4 2016-06-12 15:57:38 well it's a flaw of Alpine not to be able to provide both 2016-06-12 15:58:01 maybe we can provide packages llvm3.7, llvm3.8…? 2016-06-12 15:58:32 but that’s kinda insane that llvm isn’t backward compatible even between minor versions :( 2016-06-12 15:58:45 that would be good, but ideally it would be llvm-3.7 and llvm-3.8, organized in a generic way $package-$version 2016-06-12 15:59:14 and you'd have an automatic constraint satisfier on $package and $version for dependencies 2016-06-12 15:59:16 currently we have e.g. ruby2.2, lua5.1, … 2016-06-12 15:59:35 yeah, that's the current way to make it work with apk 2016-06-12 15:59:57 but that kind of duplication will become more and more common 2016-06-12 16:00:02 and it will be uglier and uglier 2016-06-12 16:00:20 it's still ok for now, but at some point we'll have to make a clean pass on this 2016-06-12 16:00:25 yeah 2016-06-12 16:00:59 I'll have to modify apk next year anyway, I'll talk with Alpine people and maybe enjoy the opportunity to clean that up 2016-06-12 16:01:07 and now I have to go, have a nice evening 2016-06-12 16:01:29 you too :) 2016-06-12 19:00:58 brathalion: lldb successfully built with aports/main/llvm ! :) 2016-06-12 19:01:08 had to apply this three-liner patch: https://gist.github.com/jasonwilliams200OK/c84bdd51ede5390b30c7dadcbcad1665 (see the comment below the diff about how to apply and build it with aports llvm v3.8) 2016-06-12 19:25:22 peterj: can you drop me an e-mail what needs to be done on our side? I tend to forget about things from IRC; b@bpiotrowski.pl 2016-06-12 19:28:53 barthalion: sure will do that today (shortly). Also tagged you on dotnet/coreclr PR :) 2016-06-12 21:12:30 barthalion: I have sent the email. :) 2016-06-12 22:24:33 jirutka: llvm basically is unique each version, the ir isn't guaranteed to be protable for anything but the version its built with, thats why my ghc port ported/uses its own llvm 2016-06-12 22:25:16 i've already done most of the work for 3.7.1 for ghc, and it would be nice to have a shard llvm people could use 2016-06-12 22:25:22 shared even 2016-06-12 22:26:02 what exactly does each version mean? is 3.8.0 is compatible with 3.8.2? 2016-06-12 22:26:12 generally yep 2016-06-12 22:26:17 barring bugs 2016-06-12 22:26:32 like 3.5.0 had some weird bugs that were fixed with 3.5.2 so even major versions are iffy 2016-06-12 22:27:02 okay, so the solution might be to make e.g. llvm3.7, llvm3.8… similar to ruby and lua 2016-06-12 22:27:15 yeah, its likely you want the latest either way 2016-06-12 22:27:30 well for a major version at least 2016-06-12 22:27:51 I don’t like when people doesn’t follow semver :( 2016-06-12 22:28:10 well llvm doesn't really advertise their IR format as being an api 2016-06-12 22:28:23 doesn't stop people from trying to treat it that way 2016-06-12 22:29:38 time to ping the ML about my ghc port, been about a month 2016-06-12 22:29:50 ML? 2016-06-12 22:30:54 mailing list 2016-06-12 22:30:59 aha 2016-06-13 04:43:40 friends my alpine linux is struck waiting for Caching service dependencies ... [ ok ] 2016-06-13 04:43:47 http://paste.ubuntu.com/17181005/ 2016-06-13 04:44:10 i had used this to build a base linux and am trying to boot 2016-06-13 04:44:10 ./apk.static --repository http://nl.alpinelinux.org/alpine/v3.4/main --arch x86 --root /tmp/alpine --allow-untrusted --update-cache --initdb --no-progress add alpine-base 2016-06-13 04:44:39 barthalion: any help? 2016-06-13 05:59:23 oneinsect: yes 2016-06-13 06:00:08 Install Alpine in VM and check rc-update output 2016-06-13 06:00:15 aha 2016-06-13 06:00:38 You need to have same or similar on your board 2016-06-13 06:01:09 Btw, what timezone you live in? 2016-06-13 06:01:34 the kernel is fine...it boots yocto...its an x86... also 2016-06-13 06:01:37 India 2016-06-13 06:01:40 https://wiki.alpinelinux.org/wiki/Installing_Alpine_Linux_in_a_chroot 2016-06-13 06:01:58 according to that url Before you chroot in you will probably want to mount /proc and /sys in the chroot: 2016-06-13 06:02:05 do i need to mount proc and sys? 2016-06-13 06:02:26 and setup devices? 2016-06-13 06:02:31 You can live without it 2016-06-13 06:02:44 it also mentions To make the system actually bootable, we need to add some initscripts to appropriate runlevels 2016-06-13 06:02:51 are those needed? 2016-06-13 06:03:11 On my road to office now 2016-06-13 06:03:18 But yeah 2016-06-13 06:03:35 ooh which timezone are you in? 2016-06-13 06:03:45 I don't know if the list on the wiki is up to date 2016-06-13 06:03:52 ooooh 2016-06-13 06:04:04 CET 2016-06-13 06:04:37 was this your video ? https://asciinema.org/a/3361 2016-06-13 06:05:06 Yeah, but it's old and mostly unrelated to booting Alpine 2016-06-13 06:05:16 let me check install it in vm and check rc-update output 2016-06-13 06:05:33 hmmmm 2016-06-13 06:05:46 do you any instructions (latest) that can get it to work? 2016-06-13 06:06:52 My bet is you only need these daemons 2016-06-13 06:07:14 The rest you did before is fine 2016-06-13 06:07:40 okie let me try and report back 2016-06-13 06:51:04 barthalion: it shows 2016-06-13 06:51:04 apk-tools 2.6.7, compiled for x86_64. 2016-06-13 06:51:18 when i run rc-update and a bunch of options 2016-06-13 06:51:20 nothin else 2016-06-13 07:10:17 Out of curiosity, why is OpenRC the init system for Alpine Linux? Just want to know the rationale and history for this (and yes, I hate systemd too) 2016-06-13 07:11:54 Sachiru, it's the gentoo one, and we have quite a bit of gentoo influence due to historical reasons 2016-06-13 07:12:08 we've had some interest on looking at s6 2016-06-13 07:13:07 Is there any effort towards moving, or is everyone happy with OpenRC? 2016-06-13 07:13:27 (I am in no way advocating for a move. I just want to get a sense for what the community wants.) 2016-06-13 07:42:50 oneinsect: uhm, that's weird? 2016-06-13 07:43:22 oneinsect: can you take a screenshot or something? 2016-06-13 07:43:50 Sachiru: well, it works, so no big reasons to complain 2016-06-13 07:44:30 Ah. 2016-06-13 07:44:33 True 2016-06-13 07:45:01 One good thing about OpenRC is that it knows that it's an init system, and only tries to be an init system, and nothing more 2016-06-13 07:45:14 It doesn't pretend to be a logger, dns system, dhcp, web browser, word processor, etc. 2016-06-13 07:49:09 Sachiru: openrc has some issues 2016-06-13 07:49:22 with parallel boot, hotplugging etc 2016-06-13 07:49:49 ncopa: Yes, but better serialized boot, than an init system that has to do dhcp, dns, word processing, SETI@Home before it boots. 2016-06-13 07:49:56 Due to bloat 2016-06-13 07:50:23 systemd is not an option if that is what you are thinking 2016-06-13 07:50:51 i agree that embedding dhcp in the init system is not a brilliant idea 2016-06-13 07:51:37 openrc is okish imho 2016-06-13 07:51:41 but non-optimal 2016-06-13 07:52:06 i would be interested in process supervision 2016-06-13 07:52:25 s6 might be an option 2016-06-13 07:52:28 How about runit or s6? 2016-06-13 07:52:44 i'm not 100% happy with runit 2016-06-13 07:52:46 its okish too 2016-06-13 07:52:51 Granted, this is low-hanging fruit, an Alpine Linux VM boots on my i3 potato PC in 3 seconds. 2016-06-13 07:53:06 So clearly OpenRC is only micro-optimizations. 2016-06-13 07:53:17 What issues do you see with runit? 2016-06-13 07:53:31 i dont like that the poll-approach 2016-06-13 07:53:57 i would prefer doing things when the uevent happens 2016-06-13 07:54:06 (eg like how systemd does it) 2016-06-13 07:54:45 i like the ideas around s6 though: http://skarnet.org/software/s6/why.html 2016-06-13 07:55:10 morning 2016-06-13 07:55:50 I never understood why the init system had to do everything 2016-06-13 07:56:08 Logging, dhcp, dns, udev, and all that 2016-06-13 07:56:45 It lived fine as a collection of scripts, if we could paralellize that and handle hotplugging gracefully from the start, nobody would be whining about sysvinit 2016-06-13 07:56:53 i suppose part of the reason is political/economic strategy 2016-06-13 07:57:22 IMHO it's a case of your car needing a paint job, and "while we're here, might as well do X" feature creep that results in buying 10 new cars. 2016-06-13 07:58:06 shell scripts are non-optimal from performance view 2016-06-13 07:58:20 logging actually makes some sense, and I prefer journalctl a lot over walking around /var/log to find file I need 2016-06-13 07:58:26 sure, you can discuss binary format and so on 2016-06-13 07:58:30 but the general idea is fine 2016-06-13 07:59:07 I know service name, so I am one command away to get its log 2016-06-13 08:01:15 Unless of course your log service is coupled with so much crap that it bugs whenever you so much as glance at it. 2016-06-13 08:01:31 Need a logging system during boot? Write one, make init launch it first. 2016-06-13 08:01:39 Don't write it into init. 2016-06-13 08:08:29 I don't know why peole want to "paralellize" processes starting. For the boot speed? Is it worth adding complexity of design just to speed things up? I've never foud "serial" booting to be slow, but that's a value judgement: I rarely need to boot, and I'm patient. 2016-06-13 08:09:52 and I don't buy the need for supervision either. 2016-06-13 08:11:57 perhaps there is an industry that wants it that I don't have experience in 2016-06-13 08:20:22 Desktops. 2016-06-13 08:20:46 Those who want to replace Windows boxes with "The One True OS (TM)" 2016-06-13 08:21:19 (And do so by making Linux work exactly the same way as Windows, despite saying "The Windows Way is Inferior!") 2016-06-13 08:27:44 Sachiru: and this is exactly what systemd does 2016-06-13 08:27:49 Sachiru: journald is not part of init... 2016-06-13 08:28:07 the fact that init starts it unconditionally is different story, but don't spread FUD 2016-06-13 08:28:28 If it's started unconditionally, it might as well be considered as part of it. 2016-06-13 08:28:30 IMHO 2016-06-13 08:28:42 considered != is 2016-06-13 08:29:14 Fair enough. 2016-06-13 08:29:15 Sachiru: I'm running AL as a desktop here and now. I neither need paralisation, nor supervision. 2016-06-13 08:29:41 but I got used to being minority here with my fondness of systemd, so ignore me 2016-06-13 08:30:39 ScrumpyJack: "But 3 seconds to boot on a hard disk is too slow!" (According to Poettering) 2016-06-13 08:31:06 "2.5 seconds is faster and should be done! No matter if we break 99% of server configs in the process!" 2016-06-13 08:31:36 moar fud 2016-06-13 08:31:52 Sachiru: I have plans to add s6 support to Alpine next year. It requires quite a lot of work. 2016-06-13 08:32:21 skarnet: wrt this, any plans for openrc compatibility or is it going to be Alpine 4.0? 2016-06-13 08:32:45 barthalion: the whole plan always was to have an alternative. That's why it takes so much work. 2016-06-13 08:32:55 I see 2016-06-13 08:33:38 there are openrc assumptions and hardcoding all over the place, and the biggest part of the work will be to list them and reimplement them in a rc-agnostic way 2016-06-13 08:33:57 to be able to have several different implementations of the init/rc system 2016-06-13 08:36:16 Sachiru, ScrumpyJack: one of the big problems with systemd is that it offers parallelization and speed at the cost of reliability and simplicity, and one of its big fallacies is that it pretends there's no other choice. 2016-06-13 08:36:29 But having both is possible and I can prove it. :) 2016-06-13 08:40:37 I have a talk about init/rc systems design and the s6 approach. Slides done, I did a presentation to my current customers, and I'm going to translate and give it in French in July. I'm very interested in giving this talk to anyone who will listen. :D 2016-06-13 08:42:04 skarnet: when/where is the talk? 2016-06-13 08:42:28 skarnet, i'm interested. 2016-06-13 08:43:33 ScrumpyJack: the talk in French happens in Montreuil (near Paris) on Friday July 29th. 2016-06-13 08:44:34 barthalion: sorry was away 2016-06-13 08:44:44 let me quickly get back to desk and check 2016-06-13 08:45:00 fcolista: if more people are interested, we could set up an Internet meeting for some time this summer for instance. I'm totally down with it. (But I need some time to rework the talk a bit after feedback from my customers) 2016-06-13 08:45:27 skarnet, why don't you rec the video and post it on youtoube/vimeo? 2016-06-13 08:46:47 fcolista: because I've never done video editing and it would probably be horrible. I will definitely do that the next time I do a presentation in English in a real venue (last time was over Skype - not my choice) with real video recording. 2016-06-13 08:47:08 that's holiday season :) 2016-06-13 08:47:54 ScrumpyJack: and for good reason - I only get free time during the holidays. :P 2016-06-13 08:48:58 +1 to recording the talk in Montreuil 2016-06-13 08:51:39 ScrumpyJack: It probably won't be (I'm not the decision maker and this is some kind of weird community-organized event) but I'll ask. 2016-06-13 08:52:36 do you have a URL? 2016-06-13 08:52:42 (for the event) 2016-06-13 08:54:25 ScrumpyJack: I have a URL for the place, but the calendar hasn't been input yet. https://radar.squat.net/fr/montreuil/jardin-dalice 2016-06-13 09:01:48 I would be interested in such talk too (in English) 2016-06-13 09:07:58 ScrumpyJack: process supervision is useful for instance when you need to provide continuous service, while acknowledging that your current software is not bug-free. it's often more important to process some data despite occasional crashes in some corner cases. 2016-06-13 09:12:00 przemoc: that's one of my slides. You already know the talk! ;) 2016-06-13 09:14:09 the gotcha is: process supervisors have to be bug-free. ;) 2016-06-13 09:14:22 przemoc: I've never run software that was bug ridden to the extent that it required automatic restart because it crashes more often than my tolerance thresholds will allow 2016-06-13 09:15:30 booting in 3 seconds is slow for a modern computer 2016-06-13 09:15:38 for a desktop its no problem 2016-06-13 09:15:53 but if you want boot N vms 2016-06-13 09:16:09 ScrumpyJack: the next slide explains that automatic restarts are actually not the main benefit of supervision systems. :D 2016-06-13 09:16:11 or do thinkgs like qubes, where you run 1 app per "machine" 2016-06-13 09:20:04 ScrumpyJack: maybe because you were using mature COTS. when you develop stuff in-house processing lots of data and talking with different systems along the way (and specs of them are provided by 3rd party and not always complete), bugs are simply inevitable and without automatic restarts, log supervision (so they won't eat your whole disk), etc. you would simply fail to meet customer expectations. 2016-06-13 09:21:49 przemoc: wow! don't take this the wrong way, but IMHO, that is wrong on so many levels. 2016-06-13 09:22:15 welcome to reality 2016-06-13 09:23:10 (I also have a slide that explains how supervisors are not a crutch that allows admins to be lazy. If admins are lazy, that's their own damn fault, that's not on supervision.) 2016-06-13 09:23:11 it all depends on your costomer's requirements (what you call expectations) 2016-06-13 09:25:09 usually the expectation is that the service runs 2016-06-13 09:25:11 Have you heard of Design Assurance Level? Try convincing me that I need supervision now ;) 2016-06-13 09:25:33 come to Montreuil and I will! 2016-06-13 09:27:38 If i'm delivering a sevice that requires DAL A, where a failure results in the loss of life, no amount of supervision is going to help meet that requirement 2016-06-13 09:31:26 At the moment I'm delivering a DAL D project. The requirements around the software development, testing and running mean that supervision is superflous. 2016-06-13 09:32:03 if a failure results in the loss of life, you damn well should use supervision because you do not want your service to rely on a single process remaining alive, and that's exactly what supervision does 2016-06-13 09:32:33 the lifetime of a process is not something to rely on. 2016-06-13 09:34:28 skarnet: yes it is, it's called Product Service History. 2016-06-13 09:36:57 I'm talking about unix processes. A single unix process is not something to rely on. If your service is critical, you want a supervision system, plus a monitoring and alerting system, plus lots and lots of redundancy. 2016-06-13 09:37:14 don't take me wrong, ScrumpyJack. it's not like I like to write bug-having software, really. there are often some time constraints making you unable to write code as polished as it should be. and even if you write yourself perfectly safe code, you're often not the only one on the team and your team members may not be as great coders as you are, stuff may pass code review and still be unsafe and yo 2016-06-13 09:37:20 u end up with segfault somewhere. unacceptable, I agree, wrong on many levels, indeed, but if the service remains dead and customer is not getting what it pays for, you're screwed. 2016-06-13 09:37:56 skarnet: in DAL, there is *never* any relience on a single anything. 2016-06-13 09:38:14 Static analysis is all fine and dandy, and guarantees on paper are nice to have. Nevertheless, when you run on a Unix system, crashes may, and at some point will, happen. (You need to control the process, and you also need to control the environment, which is the main cause for crashes.) 2016-06-13 09:38:30 ScrumpyJack: of course, and I very much hope it's so ;) 2016-06-13 09:39:35 so yeah, you need a whole set of guards to make sure your service will remain alive. One of those tools is static analysis. Another one of those tools is process supervision, no matter how you call it or how it's implemented. 2016-06-13 09:42:37 skarnet: a process that supervises a process? who watches the watchmen? :) 2016-06-13 09:44:38 that's why supervisors should be extremely carefully designed and implemented. you do it once and good. it doesn't have to make you coffee, have limited set of needed features and resources, so it can be done. 2016-06-13 09:46:40 i don't understand what it adds 2016-06-13 09:47:30 it can only hurt you 2016-06-13 09:50:29 as long as your services are written in a way that they can resume work properly upon restart (and it's a good design), possibly marking properly input that caused failure for later inspection, then it shouldn't hurt you. 2016-06-13 09:51:32 I can agree, though, that it may not fit all services out there 2016-06-13 09:51:36 why would i need a supervisor for that? 2016-06-13 09:54:13 oh you mean if a service crashes? i should have something to restart it automatically? 2016-06-13 09:55:32 ScrumpyJack: I answer all your objections and interrogations in the talk. Right now you have preconceived ideas and are prejudiced, so you won't be convinced; I'm not going to waste any more time. 2016-06-13 10:00:01 skarnet: Harsh. I have made design desicions based on requirements. I'm still interested in other people opinion. 2016-06-13 10:02:29 ScrumpyJack: supervision is more than just "automatic restarts", but that can't be explained in a few lines of IRC and the other benefits are a lot more subtle, much higher-level. That's why I have a whole talk about it. 2016-06-13 10:05:55 so what else is your idea/implementation of supervision? which begs the question, why are you giving more to supervision than supervision? :) 2016-06-13 10:08:15 I'm not giving it more, it comes with the design. Come to Montreuil, or wait until I can release a video or organize a presentation with more people. :P 2016-06-13 10:12:23 for instance you may set some restrictions on process and want to be notified/killed/whatever when they are crossed. it's another kind of self-defense against bugs (like resource leak). mind that bugs are not always in your particular software, but also in rest of the stack - it can be libc, it can be kernel (you use watchdog, right?). because beefy machines are often used for software, unrestrict 2016-06-13 10:12:29 ed and not supervised service's issue detection can be unnecessarily delayed in time. 2016-06-13 10:16:23 ok, sorry, don't have time to dwell on it any further. maybe skarnet will convince you in the future. ;) 2016-06-13 10:24:28 hey :) 2016-06-13 10:24:33 przemoc: That's not Supervise. That's Monitor & Control (check thresholds, restrict, take action or whatever). I already have M&C, why do I need a supervisor for that? 2016-06-13 10:25:33 perhaps you're telling me that i need *one* supervisor to do *all* of that? 2016-06-13 10:25:53 ScrumpyJack: how about stopping the passive aggression and fud? 2016-06-13 10:25:58 Nobody ever said that. 2016-06-13 10:26:53 What I *am* saying, though, is that unless your M&C system includes a supervisor, it *will* be unreliable and buggy. That's not me, that's just how Unix works. A supervisor is an integral part of any decent M&C system. 2016-06-13 10:27:16 sorry, I don't mean to come accross as passive aggressive. I was trying to make systemd appear somewhere behind that 2016-06-13 10:28:47 yeah, the systemd approach is "let's replace everything with our integrated implementation" indeed. And everyone here agrees it's a terrible approach. 2016-06-13 10:30:02 if supervising is availability, then again, i do that elsewhere. 2016-06-13 10:30:10 maybe i'm not getting it 2016-06-13 10:30:37 It's obvious you're not. :P 2016-06-13 10:30:44 lol 2016-06-13 10:31:51 If you are already getting availability from some software, then the software you are using to get availability *already includes* process supervision, or *should*, else it's buggy; and my audience is not you, it's the people who design the software you're using. 2016-06-13 10:39:32 that's not supervising to me, that's using a whole bunch of tools that when put together, they provide, say, HA clustering. 2016-06-13 10:41:47 for me the supervisor is a process that runs on a box, and restarts a crashed process. I don't want that. if it's doing anything else, i probably already have a tool for it, and i wouldn't call that tool a supervisor. 2016-06-13 10:42:10 you can consider that process supervision is the lowest level in a stack of HA tools. 2016-06-13 10:42:17 I'm beginning to sound religous, which is annoying 2016-06-13 10:42:53 no, you're not sounding religious, you just keep saying "I already have that". Which is great: you already have process supervision, you just don't recognize it yet. :P 2016-06-13 10:43:27 so just stop protesting that you don't need it, because you already have it. 2016-06-13 10:44:19 Unless, again, your services are not started in a supervision tree by your HA system - in which case your HA system is buggy.. 2016-06-13 10:47:15 i have nothing that restarts a dead process on the same host 2016-06-13 10:52:08 hat starts a process is seperate to what monitors a process which is seperate to what might signal a process and seperate to happens after that. no single one of those things is supervision in my mind. the sum could be called something. 2016-06-13 10:52:28 oh yeah, it has a name: a half-assed, ugly hack 2016-06-13 10:54:02 hahaha 2016-06-13 10:54:09 i have to agree with skarnet on this one 2016-06-13 10:54:21 you do somethign similar, but propobly in a very hacky way 2016-06-13 10:55:49 likely with a lot of polling and /proc scanning and race conditions, but nobody cares because the machines are powerful and the system needs to overscale anyway 2016-06-13 11:01:38 that was like a logical overview. i think we are talking about different things. perhaps i'm glued to the fact that for me a supervisor (like supervise, or smf, or whatever systemd does) *restarts*. i have never wanted that. 2016-06-13 11:22:03 ScrumpyJack: one of the goals of a supervisor is to restart, if the admin so wishes. This is generally what people use a supervisor for, but it isn't mandatory; its benefits go way beyond that. Supervision is mostly about reliability (absence of race conditions and reproducible starting environment), death/readiness notification, and other things you need for easy monitoring. Yeah, it restarts, but only if you 2016-06-13 11:22:03 want it to and it's far from the only thing it does. 2016-06-13 11:23:33 (It's still a good idea to have the restarts, so if something kills your process, or even all the processes in your system, the supervision tree comes alive again as if nothing happened.) 2016-06-13 11:35:24 skarnet: "good idea to have the restarts" is debatable, but i'm really interested in what you say about "absence of race conditions and reproducible starting environment" as examples of what a super could do. We all want that, and implement that in different ways. Why do you think it's a supervisors job to provide a "clean and safe" environment to start and signal a process? There has to be some 2016-06-13 11:35:30 good reason to implement that in a super. What is it? Perhaps some other time though, I'm clogging up the channel. Hopefully your talk might answer that 2016-06-13 11:35:59 it does. 2016-06-13 11:36:33 with examples, i love examples :) 2016-06-13 11:37:04 I bet he have them. customers love examples too. 2016-06-13 11:37:28 and the thing is, a supervisor's architecture (the supervisor is the daemon's parent process) is the only architecture that gives you those things. It's just how Unix works. 2016-06-13 11:38:47 it's not "adding that to a supervisor", it's "when you have a proper process supervision infrastructure, you naturally get that". Come to #s6 if you want to talk about it more without us flooding this channel with OT even more. 2016-06-13 11:39:25 there is an #s6? I'm there! 2016-06-13 11:39:39 oh, didn't know s6 has its own channel. I'm not using it yet, but I'll gladly idle there at least 2016-06-13 12:21:16 ncopa: could you please take a look at my patches in the abuild repo? https://github.com/alpinelinux/abuild/pulls 2016-06-13 12:31:13 ok 2016-06-13 12:38:08 jirutka: i think i tag abuild 2.27.2 release 2016-06-13 12:38:20 those are all fixes 2016-06-13 12:42:19 ncopa: not all of them 2016-06-13 12:43:26 this one https://github.com/alpinelinux/abuild/pull/6 is imho more an improvement than a fix 2016-06-13 12:44:36 well 2016-06-13 12:45:01 if there are not dev files and -dev subpackage is created 2016-06-13 12:45:09 abbuild will fail 2016-06-13 12:45:22 so, i 'm kinda okish to consider it as a "fix" 2016-06-13 12:46:10 okay :) 2016-06-13 12:46:20 jirutka1: have you tested it? or can i just tag it? 2016-06-13 12:46:41 i pushed them all 2016-06-13 12:46:48 so you can test from git 2016-06-13 12:47:13 just basic smoke test to verify that its not completely useless 2016-06-13 12:47:33 to be honest, I didn’t yet, but these are very little changes and I’m quite confident that it will work… sounds like the last words before death, so I’ll rather test it today :) 2016-06-13 12:48:08 thanks 2016-06-13 12:48:18 about https://github.com/alpinelinux/abuild/pull/6/files, I’ve tested it separately, the find command and pipe 2016-06-13 12:48:53 just give the other things a simple test run 2016-06-13 12:49:07 to make sure there are no typos or similar 2016-06-13 12:49:31 just verify that they generate an APKBUILD 2016-06-13 13:20:19 ncopa: what does 0x10001 mean at this place? shouldn’t it be a single char option as in other cases? 2016-06-13 13:20:19 ``` 2016-06-13 13:20:19 { 'd', "description", "Print description for PACKAGE" }, 2016-06-13 13:20:19 + { 0x10001, "license", "Print license for PACKAGE" }, 2016-06-13 13:20:20 ``` 2016-06-13 13:23:24 it means that there are no short opt for it 2016-06-13 13:23:41 you cannot do apk info -l ... 2016-06-13 13:23:51 but you have to do: apk info --license 2016-06-13 13:23:51 aha 2016-06-13 13:24:21 but why 0x10001? it looks like some magic number, a constant would be better IMHO 2016-06-13 13:24:38 it is a magic number 2016-06-13 13:25:19 i used that over constant for consistency 2016-06-13 13:25:44 using magic numbers in code is generally considered as a bad practice… 2016-06-13 13:25:46 it should have such values exported at least as defines, to not be retyped across the source code and to de-magic-ify it 2016-06-13 13:26:01 "abvoid constant for consistency" sounds like a slef contradiction :) 2016-06-13 13:26:11 exactly 2016-06-13 13:26:58 retyping magic values easily leads to typos 2016-06-13 13:27:30 i would normally agree with you both, but in this case i think it doesnt matter that much 2016-06-13 13:28:08 well, I guess you'll accept PR improving that :) 2016-06-13 13:28:25 jirutka is possibly already working on it :) 2016-06-13 13:28:44 you'll have to send those PRs via fabled 2016-06-13 13:28:57 I’m not, C is not my cup of tea… 2016-06-13 13:29:20 however, I’m thinking about rewriting apk into Rust! :) leo, are you in? XD 2016-06-13 13:32:29 ACTION prefers having basic distro tools buildable with C compiler, or even C++, but not with stuff that needs its own bootstrapping 2016-06-13 13:32:40 +1 2016-06-13 13:33:06 uh, C doesn’t need its own bootstrapping? ;) 2016-06-13 13:33:20 well, C is kind of necessary bootstrapping 2016-06-13 13:33:38 why? just because it’s so widespread? 2016-06-13 13:34:52 yes, you build most Unixes with C 2016-06-13 13:35:56 s/Unixes/Unices/ 2016-06-13 13:36:12 C is historically the language for Unix; all the Unix API has been built with C in mind. And lots of effort has been put into it, lots of bugs and misdesigns have been fixed, etc. etc. 2016-06-13 13:36:34 There are still a lot of things to fix, obviously, but we're in a much better place than 40 years ago 2016-06-13 13:37:11 you mean that lots of bugs have been introduced, because programming in C is like dancing on minefield with boths eyes closed? ;) 2016-06-13 13:38:20 when you'll have complete Rust-based toolchain and all the basic utils (and I'm know there are such attempts to port them to Rust, at least to some extent), we can talk about Rust being a basic block. but before such implementations will mature, some time has to pass first. 2016-06-13 13:38:47 jirutka1: not dancing in a minefiled with both eyes closed, no. It's a science, *and* an art. https://www.youtube.com/watch?v=mr834Cs9ncs 2016-06-13 13:38:56 there’s also an effort to rewrite Musl to Rust https://github.com/dikaiosune/rusl :) 2016-06-13 13:39:19 sorry, I have to finish review for thesis now 2016-06-13 13:39:41 somehow Russell Crowe immediately came to my mind ;) 2016-06-13 13:41:02 ncopa: forgot to ask in my email, but for the llvm-3.7 bit should I put llvm in an isolated prefix or try to figure out how to suffix binaries after they're built? 2016-06-13 13:41:27 przemoc: you mean Vincent Cassel? ;) 2016-06-13 13:44:39 that one I don't get, please explain 2016-06-13 13:45:20 mitchty: i dont know if we need to be able to support having llvm-3.7 *and* 3.8 installed in parallel 2016-06-13 13:45:46 i think the package name should be "llvm3.7" 2016-06-13 13:46:24 (13:29:20) jirutka: however, I’m thinking about rewriting apk into Rust! :) leo, are you in? XD 2016-06-13 13:46:24 lets get rustc working first 2016-06-13 13:46:27 ncopa: presumably you'd want to, example, you could have llvm/clang 3.8 installed for things like c ffi/etc.. and then ghc/julia could use 3.7 for the things they can't use 3.8 for 2016-06-13 13:46:28 then we can do that as well 2016-06-13 13:46:29 przemoc: the guy in the video I linked is Vincent Cassel, not Russel Crowe. If there's a movie where Russell Crowe does similar things (or dances in a minefield) then I missed it. 2016-06-13 13:49:18 ah, I was referring to rusl name 2016-06-13 13:51:27 oh, sorry XD 2016-06-13 13:57:03 mitchty: hum. yes it would probably be the best 2016-06-13 13:59:29 ncopa: ok, so i'll try getting llvm built with moving things around so it can coexist, aka llc-3.7 llvm-config-3.7 etc... 2016-06-13 14:00:08 but testing this all will take a while, the arm stuff takes a day to build 2016-06-13 14:00:30 well 19 hours to be exact 2016-06-13 14:01:28 got it 2016-06-13 14:01:39 i wonder if i should just push what we have to testing for now 2016-06-13 14:01:42 and take it from there 2016-06-13 14:02:21 its not a big deal, besides if ghc-llvm... doesn't need to exist in general no point adding it then deleting it 2016-06-13 14:02:30 ok 2016-06-13 14:02:56 i've got it all mostly automated anyway, just arm takes ages to build 2016-06-13 14:03:39 i'd be happy to getting 3.4.1 out meanwhile 2016-06-13 14:29:24 mitchty: building anything on arm takes ages 2016-06-13 14:29:34 mitchty: we should be more concerned about memory usage 2016-06-13 14:29:52 barthalion: indeed, even more so when llvm is invoked at -O3 2016-06-13 14:30:22 so for ghc, i think the biggest process size I saw was around 2 gigabytes 2016-06-13 14:30:32 I don't remember the spec of our arm builders 2016-06-13 14:30:33 that might have been virtual however so grain of salt 2016-06-13 14:30:37 ncopa: ^ ? 2016-06-13 14:30:55 i build it on a 2g box with 1g of swap 2016-06-13 14:31:43 so we should be fine 2016-06-13 14:32:37 yep, i think llvm itself was bigger 2016-06-13 14:33:10 well in linking rather 2016-06-13 14:33:41 let me login and look at the logs quick 2016-06-13 14:34:36 gah didn't save em, eh when i rebuild i'll look at memory use 2016-06-13 15:21:28 mitchty: we use wandboard quad 2016-06-13 15:21:42 iirc they have 2G ram and 4G swap 2016-06-13 17:39:14 hi guys 2016-06-13 18:45:31 sup 2016-06-13 18:52:50 guys im looking for some help, i need to make an apk to install an already compiled bin file into /etc/bin, could anyone point me to an example please ? 2016-06-13 18:54:03 joagt: binary files should not be inside /etc 2016-06-13 18:54:12 this directory is intended for configuration files etc. 2016-06-13 18:54:26 sorry my bad inside /bin/ 2016-06-13 18:54:31 joagt: put that binary into source= 2016-06-13 18:54:40 joagt: build() { return 0 } 2016-06-13 18:55:02 joagt: package() { install -Dm "$srcdir/binfile" "$pkgdir"/bin/binfile } 2016-06-13 18:55:20 pretty much it, you obviously need some other fields but you can get these from any APKBUILD 2016-06-13 18:55:22 thanks barthalion 2016-06-13 18:55:40 -Dm755 of course 2016-06-13 18:56:23 np 2016-06-13 19:30:12 barthalion: Hi, I've been working on packaging lttng for alpine and to be able to build the tracing kernel modules, I would need some config options added to the kernel, would that be doable? 2016-06-13 20:00:35 mjeanson: most likely yes, this is not something I can do without discussion with ncopa though 2016-06-13 20:02:12 barthalion: ok, I was just wondering if it was possible before investing time 2016-06-13 20:03:30 I'll start working on patch then 2016-06-13 20:05:03 lttng had its own out-of-the-tree kernel modules in the past IIRC. how it is nowadays? and what probe techniques it supports nowadays beside tracepoints? kprobes and maybe uprobes too? 2016-06-13 20:12:52 mjeanson: yeah, external modules might be some serious part of the discussion… 2016-06-13 20:21:42 przemoc: kprobes, kretprobes, ftrace but most are from tracepoints 2016-06-13 20:22:29 I'm not really a kernel guy tough, you might find more useful information in https://github.com/lttng/lttng-modules/tree/master/probes 2016-06-13 20:23:04 after all Desnoyers introduced tracepoints, so I am sure they remain the main probing technique 2016-06-13 20:24:35 barthalion: as long as the config options are enabled in the kernel, the modules can be built manually by a user 2016-06-13 20:24:46 but it would be nice to have it in aports 2016-06-13 20:24:59 there seem to be quite a few oot modules already 2016-06-13 20:36:05 mjeanson: does lttng-ust also rely on the said kernel module? or that can be ported separately first? :) 2016-06-13 20:36:34 peterj: the modules are only required for kernel tracing 2016-06-13 20:37:28 I'm currently working on porting the cli tools to musl 2016-06-13 20:37:52 mjeanson: meaning lttng-ust would need to wait until the decision is main? 2016-06-13 20:38:07 mjeanson: is it lttng cli tools you are referring to? :) 2016-06-13 20:38:50 peterj: the cli tools control both kernel and userspace tracing, but they don't require the modules to be present 2016-06-13 20:39:31 as long as we have liblttng-ust and the cli tools, we will be able to to userspace tracing 2016-06-13 20:40:12 cool. for coreclr, i guess the basic lib lttng-ust is required by their configure script. 2016-06-13 20:41:02 yeah, with lttng-ust you'll be able to build applications with tracing 2016-06-13 20:41:22 but then if you want to do actual tracing, you'll need tools 2016-06-13 20:42:09 in this scenario, the kernel modules is just nice to have 2016-06-13 20:42:24 barthalion: (a friendly question) regarding lldb38, is there any blocker to add it with this patch: https://gist.github.com/jasonwilliams200OK/c84bdd51ede5390b30c7dadcbcad1665 ? Just wanted to manage my hopes/expectations/targets for this week 2016-06-13 20:42:30 :) 2016-06-13 20:44:29 I think we don't even need a separate addition for lldb, it just need to be added to llvm APKBUILD script. 2016-06-13 20:45:01 (lldb is just one of many tools in llvm-src/tools dir) 2016-06-13 20:52:03 ncopa: I’ve tested newapkbuild and apkbuild-cpan and both works fine :) and I’ve also opened another pull request, just a cosmetic change (but important for my OCD XD) https://github.com/alpinelinux/abuild/pull/7 2016-06-13 20:53:53 mjeanson: I never followed kernel maintenace in Alpine too closely, but the general trend was reducing number of 3rd party modules 2016-06-13 20:54:06 mjeanson: but in that case, enabling options shouldn't be an issue 2016-06-13 20:54:24 peterj: not at all, I'll handle this tomorrow morning before work 2016-06-13 21:01:28 barthalion: thank you so much! :) 2016-06-13 21:02:04 well, after further testing, the way grsec is configured doesn't allow enabling tracing support in the kernel 2016-06-13 21:02:19 mjeanson: we have -vanilla kernel though 2016-06-13 21:02:43 yeah and it seems to have the required options 2016-06-13 21:03:09 is switching as simple as installing the vanilla kernel package? 2016-06-13 21:03:54 and probably updating bootloader, yes 2016-06-13 21:07:19 barthalion: can the user space trace lttng-ust be ported first (since it requires no kernel package)? 2016-06-13 21:08:16 yeah, that's the plan for now 2016-06-13 21:12:57 can't quite figure the proper way to switch kernel between kernel flavours 2016-06-13 21:14:10 update-extlinux regens the bootloader config 2016-06-13 21:14:19 but always in the same order 2016-06-13 21:14:56 do I have to remove the grsec flavour kernel or can I specify the flavour I want enabled by default? 2016-06-13 21:17:34 mjeanson: I'm sorry, no idea – I never liked any bootloader magic so I have manual config on my personal machines 2016-06-13 21:21:22 mjeanson: haven't used vanilla myself, but I think you need to change default= in /etc/update-extlinux.conf 2016-06-13 21:22:07 or set hidden=0, so you'll see the menu 2016-06-13 21:22:39 przemoc: that's the file I was looking fo, thanks 2016-06-13 21:23:17 and after changes run update-extlinux, of course 2016-06-13 21:26:29 you could also change /boot/extlinux.conf directly, but it wouldn't survive any kernel update in future, so it's often a bad idea, unless you want to really quickly test some tuning 2016-06-13 21:27:21 and don't blame me if you'll make your system unbootable ;) 2016-06-13 21:28:58 is it normal that the grsec flavor is named kver-pkgver-flavour and the vanilla one is just kver 2016-06-13 21:31:20 default=vanilla is what you should put there (conclusion after quickly looking into update-extlinux script code), if you're concerned about it 2016-06-13 21:47:08 I was just wondering why the _abi_release for the grsec kernel is ${pkgver}-${pkgrel}-$_flavor while for the vanilla it's just ${pkgver} 2016-06-13 21:47:59 oversight 2016-06-13 21:48:49 anyway, I have an apk for lttng-modules that builds successfuly against the vanilla kernel 2016-06-13 22:33:15 anyone know what this is called in debian by chance: mipsel-openwrt-linux-musl-cpp 2016-06-13 22:33:22 err stupid copy paste 2016-06-13 22:33:35 musl-ldd* 2016-06-14 07:17:11 barthalion: 2016-06-14 07:17:16 are you here 2016-06-14 07:18:46 i am made chroot and now i am trying to copy it to an ext4. file 2016-06-14 07:19:03 however it does seem to copy proc/misc etc 2016-06-14 07:19:10 NOT* 2016-06-14 07:19:13 oneinsect: how do you copy? 2016-06-14 07:19:35 it shouldnt copy /proc? 2016-06-14 07:19:44 here the steps 2016-06-14 07:19:47 # dd if=/dev/zero of=alpine-image.ext4 bs=1M count=300 2016-06-14 07:19:54 mkfs.ext4 alpine-image.ext4 2016-06-14 07:20:04 oneinsect: uhm 2016-06-14 07:20:08 didn't you do that already? 2016-06-14 07:20:10 cp -avr /tmp/alpine /mnt/ 2016-06-14 07:20:30 why copy to ext4-formatted img file? 2016-06-14 07:20:44 because i need to burn to an internal emmc on Intel edison 2016-06-14 07:21:31 rsync --one-file-system 2016-06-14 07:21:34 thing is if i just use ./apk.static --repository http://nl.alpinelinux.org/alpine/v3.4/main --arch x86 --root /tmp/alpine --allow-untrusted --update-cache --initdb --no-progress add alpine-base 2016-06-14 07:21:38 it doesnt boot 2016-06-14 07:21:40 will exclude proc/* 2016-06-14 07:21:47 okie let me try 2016-06-14 07:22:09 you probably also want --numeric-ids 2016-06-14 07:22:16 and --devices 2016-06-14 07:22:38 i mean alpine-base doesnt boot...hangs... so i included mount -t proc none ${chroot_dir}/proc mount -o bind /sys ${chroot_dir}/sys 2016-06-14 07:23:06 uh 2016-06-14 07:23:10 and mount -o bind /dev ${chroot_dir}/dev 2016-06-14 07:23:10 it hangs where? 2016-06-14 07:23:18 here is the pastelog 2016-06-14 07:23:28 did you do rc-update? 2016-06-14 07:24:46 http://paste.ubuntu.com/17181005/ 2016-06-14 07:25:09 yes rc-update did not return anything so what I did was 2016-06-14 07:25:12 rc-update add devfs sysinit 2016-06-14 07:25:15 i added all of them 2016-06-14 07:25:24 listed here 2016-06-14 07:25:24 https://wiki.alpinelinux.org/wiki/Installing_Alpine_Linux_in_a_chroot 2016-06-14 07:25:30 now trying to copy 2016-06-14 07:25:53 let me try @ncopa: 2016-06-14 07:26:04 except this article is terribly old so that's why I asked you to compare the list with installation on virtual machine 2016-06-14 07:26:19 hmmm 2016-06-14 07:27:42 here is my 2016-06-14 07:27:44 rc-update 2016-06-14 07:27:45 http://paste.ubuntu.com/17319529/ 2016-06-14 07:27:54 barthalion: 2016-06-14 07:33:40 okie trying this 2016-06-14 07:33:41 rsync -aAXv --numeric-ids and --devices /tmp/alpine /mnt 2016-06-14 07:34:03 it seems to hang at alpine/proc/kcore....waiting for it 2016-06-14 07:34:19 you forgot --one-file-system 2016-06-14 07:34:37 shucks 2016-06-14 07:36:07 thanks guys i am gonna burn this ext4 with vmlinuz to edison right away and try it out 2016-06-14 07:36:14 it copied successfully 2016-06-14 07:38:02 fingers crossed and praying to god 2016-06-14 08:00:20 http://paste.ubuntu.com/17319826/ 2016-06-14 08:00:26 barthalion: 2016-06-14 08:00:37 @ncopa: any ideas on the above pastebin 2016-06-14 08:01:05 its hangs there and resets itself after sometime...dont get cmd prompt 2016-06-14 08:08:35 with slightly modified bootargs ..here the paste bin however it still hangs 2016-06-14 08:08:36 http://paste.ubuntu.com/17319891/ 2016-06-14 11:40:01 whee! i managed to create an UEFI boot usb which booted on a macbook \o/ 2016-06-14 11:46:42 going to do some maintenance on patchwork.a.o 2016-06-14 11:46:54 hold on to your seats 2016-06-14 11:54:19 ncopa: I’ve tested newapkbuild and apkbuild-cpan and both works fine  and I’ve also opened another pull request, just a cosmetic change (but important for my OCD XD) https://github.com/alpinelinux/abuild/pull/7 2016-06-14 11:57:43 pushed thanks! 2016-06-14 11:58:53 ncopa: I’ve tested it, so I think that you can now release new version :) 2016-06-14 12:06:35 that question may sound weird, but what is alpine expected debug workflow? setting pax/softmode to 1 and rebuilding package with -dbg subpackage (as only a few have them defined out-of-the-box) and installing them? I got segfault when using git grep -P, but I still see mostly ??. both /usr/lib/debug/usr/bin/git.debug and /usr/lib/debug/usr/libexec/git-core/git-grep.debug seem suspiciously small 2016-06-14 12:06:41 (2520 bytes), so maybe -dbg isn't building as expected (git's Makefile doing some shenigans?) 2016-06-14 12:08:14 possible 2016-06-14 12:11:51 it's `git grep -P '^adduser ((?!-G).)*$'` (I know the regex is inefficient, it's not the point here) performed in aports git repo, just for the record 2016-06-14 12:16:57 I think it will be a good thing in future to make sure that simply adding -dbg in any existing package produces proper debug stuff. (having ready to use -dbg subpkgs for all packages would be nice thing too, maybe they could be kept on debug dedicated apk repository) 2016-06-14 12:19:06 Oracle distributes JDK with restricted cryptography and with bundled adware, but that doesn’t matter, because it’s dead simple to install Unlimited Cryptography and adware is optional, so what… OMG http://stackoverflow.com/questions/37796958/gradle-build-fails-in-jenkinsalpine-container-unsatisfiedlinkerror-linux-amd6/37811227?noredirect=1#comment63087173_37811227 2016-06-14 12:19:46 patchwork should be online again. 2016-06-14 12:19:53 please test if you can send patches. 2016-06-14 12:20:43 patchwork is still insecure… #5713 2016-06-14 12:24:58 jirutka: how does it redirect to http? 2016-06-14 12:25:14 i just logged in with https and it kept me on https. 2016-06-14 12:25:29 I’ve tried to know for the second time 2016-06-14 12:26:01 say what? 2016-06-14 12:26:11 However, it doesn’t solve the whole problem. When user authentication is involved, it should not be available over insecure HTTP at all. 2016-06-14 12:26:46 now it doesn’t redirect me to HTTP :) 2016-06-14 12:27:53 but the session cookie is not limited to https only, so when I open it via HTTP, I’m stil logged and my session cookie is transfered in plain text. 2016-06-14 12:28:42 can't wait for all these poor public patches to be hacked 2016-06-14 12:29:18 barthalion: Alpine is about security, isn’t it? 2016-06-14 12:29:49 well, this is about infra, not about alpine per se 2016-06-14 12:30:01 and public patchwork instance holding public patches is hardly a target 2016-06-14 12:30:02 barthalion: your point is totally irrelevant, it’s dead simple to make it secure, so there’s no excuse to not do it 2016-06-14 12:30:28 let me decide about relevance of my points 2016-06-14 12:31:06 well 2016-06-14 12:31:14 and there are also other concerns… some poor public WiFis injects ads to the pages 2016-06-14 12:31:19 nobody wants to get laughed at. :/ 2016-06-14 12:31:40 if it did end up being compromised or used negatively, Alpine ends up being a punchline. :( I'd be sad to see that personally. 2016-06-14 12:31:50 And I'm not even devy that much 2016-06-14 12:32:17 ? 2016-06-14 12:34:03 the non https only stuff. alone it isn't a big deal, but if anything else appeared, it'd be a punchline for "more proof". plus its an easy fix. 2016-06-14 12:34:44 clandmeter: sorry, I’ve somehow mixed to sentences into one; “I’ve tried to know for the second time” – I wanted to say „I’ll try it for the second time and let you know” 2016-06-14 12:35:05 ok 2016-06-14 12:35:07 got it 2016-06-14 12:35:14 most important is, it works. 2016-06-14 12:35:24 now i need to know patches still get in 2016-06-14 12:35:46 we have had a discussion about non http last time. 2016-06-14 12:35:54 i think the decission was not really made. 2016-06-14 12:36:01 actually I didn’t, how is that possible that I wrote such a nonsense and realized that it’s nonense after few minutes later? :( maybe I need coffee 2016-06-14 12:36:34 lol 2016-06-14 12:36:40 dont worry 2016-06-14 12:36:44 I do it all the time 2016-06-14 12:36:51 and some times on purpoused 2016-06-14 12:39:29 ACTION Coffee time! 2016-06-14 12:59:44 ncopa: could you please look at https://github.com/JuliaLang/julia/pull/16881#issuecomment-225444288? 2016-06-14 13:22:22 friends any idea how much will a tiny desktop occupy 2016-06-14 13:22:24 in alpine 2016-06-14 13:22:42 i wanted to install one on arm based board 2016-06-14 13:52:44 minimalist, about 600mb 2016-06-14 13:59:40 darn 2016-06-14 15:54:35 friends has lbu ci change between 3.3 and 3.4v ? it doesnt seem to commit anymore even if i give /media/mmcblk0p1/ 2016-06-14 15:55:20 oneinsect: any progress with Intel Edison? 2016-06-14 15:56:01 cmd prompt wasnt coming so i left it....i progressed a bit where it was loading sshd...and then waiting for eternity 2016-06-14 15:56:17 i was frustrated jirutka1: 2016-06-14 15:56:28 oneinsect: have you log? 2016-06-14 15:56:34 hmm wait 2016-06-14 15:57:13 http://paste.ubuntu.com/17319891/ 2016-06-14 15:58:07 what i moment, what is „SELinux filesystem“ doing here? we don’t have SELinux in Alpine…? 2016-06-14 15:58:09 this is what i did.... i live booted a cd and then rsync -aAXv --one-file-system --numeric-ids and --devices to ext4 file 2016-06-14 15:58:49 however, this is a huge progress! 2016-06-14 15:59:04 and flashed the ext4 file...with only change is i use edison vmlinuz file...and their modules 2016-06-14 15:59:15 yes i dont know what selinux is doing... 2016-06-14 15:59:24 is it inside the kernel? 2016-06-14 15:59:45 oneinsect: now edit /etc/rc.conf, set `rc_interactive="YES"`, then press I (if I remember it correctly) when OpenRC is started 2016-06-14 16:00:06 it should ask you before starting each service, so you can easily find what’s wrong 2016-06-14 16:00:42 okie wait ...i deleted everything out of frustration ...let me start the live cd boot and copy rync to an ext4 2016-06-14 16:00:46 and reflash edison 2016-06-14 16:01:19 oneinsect: it seem that you’re really very frustated from it :( 2016-06-14 16:01:41 i have been on it from past one month... i tried boot initramfs a lot but it could not 2016-06-14 16:02:03 arm based old kernels 3.4 and mainline boot so well with arm boards.... 2016-06-14 16:02:09 i could not take intel not booting 2016-06-14 16:02:22 arm boards with alpine i meant 2016-06-14 16:02:46 lucky i have the ext4 file...let me edit /etc/rc.conf 2016-06-14 16:02:55 i really want edison to have alpine based distro 2016-06-14 16:03:11 and probably will push it then to Intel galileo 2 2016-06-14 16:03:13 oneinsect: have you find something for “clock skew detected”? 2016-06-14 16:03:24 rtc clock isnt functioning 2016-06-14 16:03:42 but that is okie...i get it all the time on raspberry pi etc which dont have rtc clock 2016-06-14 16:04:04 so maybe disable /etc/conf.d/hwclock? 2016-06-14 16:04:08 they still will work past through 2016-06-14 16:05:01 yes i will disable it now 2016-06-14 16:05:11 btw I’ve told about Intel Edison with me coworker, he was interested, but he’s not gonna buy them until I get a job in his company and I’m still not decided XD 2016-06-14 16:06:14 should i buy and send you one? 2016-06-14 16:06:40 its easy if you are in usa 2016-06-14 16:06:46 its cheap 2016-06-14 16:06:46 I’m not 2016-06-14 16:06:50 oooh 2016-06-14 16:06:53 I’m from Czech Republic, Europe 2016-06-14 16:07:04 aaaaah 2016-06-14 16:07:18 nevertheless i can still send one but later on 2016-06-14 16:08:25 it would be quite a commitment for me, I’m not sure how much time and patience I’ll have with it 2016-06-14 16:09:21 great then i will post it soon...let me get this thing running ... 2016-06-14 16:09:50 when you get time please send me your postal address to this atlury@gmail.com 2016-06-14 16:10:15 is fedex/dhl/ups good? 2016-06-14 16:12:56 yes, but as I said, I really don’t know if I’ll be able to get it run and if I’ll have time and patience for it 2016-06-14 16:13:56 it doesnt matter, i am only interested in getting alpine run on it... 2016-06-14 16:14:17 and that which anyways we are doing it here now 2016-06-14 16:14:49 okie in rc.conf i only have rc_tty_number=12 enabled 2016-06-14 16:15:03 should i now include rc_interactive="YES" 2016-06-14 16:15:19 yes 2016-06-14 16:15:46 done and etc/conf.d/ does contain hwclock.... how do i remove it? 2016-06-14 16:15:53 this is not chroot... 2016-06-14 16:15:54 and also add rc_logger="YES" 2016-06-14 16:16:51 done included the logger and should i rename hwclock to some other name?? 2016-06-14 16:17:16 I think that you can just remove symlink /etc/runlevels/boot/hwclock 2016-06-14 16:18:41 link in etc/runlevels/boot/hwclock points to hwclock -> /etc/init.d/hwclock and removed it 2016-06-14 16:19:00 anything else? 2016-06-14 16:20:13 hmm, could you send me ls /etc/runlevels/{sysinit,boot,default}? I don’t remember what is here by default 2016-06-14 16:20:52 right away 2016-06-14 16:21:36 http://paste.ubuntu.com/17329341/ 2016-06-14 16:22:20 okay, this looks fine, try to boot it 2016-06-14 16:22:40 aye aye captain!! 2016-06-14 16:22:42 OpenRC should log something like ”Now press I for interactive mode” 2016-06-14 16:23:01 then press the key, it should ask you before starting each service 2016-06-14 16:23:15 yes i will now flash it into edison 2016-06-14 16:31:28 okie gonna take a few minutes...waiting on it 2016-06-14 16:32:19 kk 2016-06-14 16:44:35 should i press I when it says interactive setup?...here is the pastelog http://paste.ubuntu.com/17330415/ 2016-06-14 16:44:43 yes 2016-06-14 16:45:56 next time also remove symlink /etc/runlevels/sysinit/modloop 2016-06-14 16:46:09 let me try 2016-06-14 16:46:15 not needed now 2016-06-14 16:46:21 what does it do after the last line? just hang? 2016-06-14 16:46:40 just hang 2016-06-14 16:47:00 at 30 seconds it still hanging so the watchdog kicks in and reboots 2016-06-14 16:47:01 the board 2016-06-14 16:47:06 now i am in interactive 2016-06-14 16:47:10 About to start the service devfs... 2016-06-14 16:47:15 * 1) Start the service 2) Skip the service * 3) Continue boot process 4) Exit to shell 2016-06-14 16:47:39 start all but modloop 2016-06-14 16:47:43 sure 2016-06-14 16:48:14 okie the board reset itself after snd_intel_sst: request fw failed 2016-06-14 16:48:26 while i was running the interace mode 2016-06-14 16:48:44 what is it, some watchdog? 2016-06-14 16:49:28 hmm, sshd generates certificates at the first start, so it needs enough entropy to do that… this may be a problem on such board 2016-06-14 16:50:08 there are bin files in the firmware folder namely fw_sst_119a.bin and intel_mcu.bin 2016-06-14 16:50:17 i wonder what is the first one... 2016-06-14 16:50:36 yes its the watchdog 2016-06-14 16:51:11 so it needs some userspace program that will periodically inform board that it’s still alive…? 2016-06-14 16:51:21 or should kernel do this? 2016-06-14 16:52:10 hmmmm could be? let me check if there is a way to disable watch dog 2016-06-14 16:52:11 https://seven.centos.org/2015/08/a-flashable-centos-image-for-the-intel-edison/ 2016-06-14 16:52:22 I’m just googling it 2016-06-14 16:52:52 Starting Disable the watchdog device on the Intel Edison… 2016-06-14 16:52:52 [ 7.665241] intel_scu_watchdog_evo: watchdog_stop 2016-06-14 16:53:15 The disable-watchdog service starts late in the boot process, so touching /.autorelabel will result in a bootloop (the watchdog timer times out before the selinux relabel is finished) 2016-06-14 16:53:32 that was note on centos blog post 2016-06-14 16:53:51 try to extract their the script for disabling watchdog 2016-06-14 16:54:40 it seems that you can also disable it from u-boot, http://forum.osdev.org/viewtopic.php?f=1&p=242715 2016-06-14 16:55:08 exactly 2016-06-14 16:55:18 mw.l 0xff009000 0x10f8 1 2016-06-14 16:55:25 let me intercept u-boot 2016-06-14 16:55:28 disable it 2016-06-14 16:55:47 and hope that these magic numbers are correct for your revision of the board :P 2016-06-14 16:55:52 yeaaa 2016-06-14 16:59:19 okie gonna retry 2016-06-14 17:02:36 jirutka: http://paste.ubuntu.com/17331257/ 2016-06-14 17:03:22 does it hang or restart? 2016-06-14 17:03:23 the ones that are not started are skippped 2016-06-14 17:03:27 still hangs 2016-06-14 17:03:34 give it some time 2016-06-14 17:03:51 the board resets itself 2016-06-14 17:04:00 i wonder the watchdog doesnt stop 2016-06-14 17:04:06 hm, so watchdog is still active 2016-06-14 17:04:48 maybe it would be easier to extract the script for disabling watchdog from https://seven.centos.org/2015/08/a-flashable-centos-image-for-the-intel-edison/ 2016-06-14 17:04:57 i will check 2016-06-14 17:05:07 before that you can try to start it and exit to shell before starting ssh 2016-06-14 17:05:20 and just try if the shell works etc. 2016-06-14 17:06:55 sure 2016-06-14 17:06:58 let me try 2016-06-14 17:08:15 i am in shell 2016-06-14 17:09:05 let me see when in shell the watchdog doesnt kick up or not 2016-06-14 17:09:44 okie it does kick in and reboot so let me find a script that can disable it 2016-06-14 17:11:01 echo 1 >/sys/devices/virtual/misc/watchdog/disable 2016-06-14 17:14:09 let’s create a simple runscript for it: http://haste.fit.cvut.cz/opisuxu.sh 2016-06-14 17:14:28 https://github.com/01org/edison-u-boot/blob/edison-v2015.10/include/configs/edison.h 2016-06-14 17:14:38 i can disable watch dog there and recompile u-boot 2016-06-14 17:14:41 what say? 2016-06-14 17:14:57 name it e.g. disable-watchdog, add this file to /etc/init.d/ and create symlink /etc/runlevels/sysinit/disable-watchdog → /etc/init.d/disable-watchdog 2016-06-14 17:15:00 also it says 2016-06-14 17:15:06 can't create /sys/devices/virtual/misc/watchdog/disable: nonexistent directory 2016-06-14 17:15:11 there’s also an option 2016-06-14 17:15:30 ah, pardon, it obviously must run after sys is mounted 2016-06-14 17:15:56 never mind let me quickly recompile u-boot with edison watch dog disabled 2016-06-14 17:16:09 http://haste.fit.cvut.cz/uxohema.sh 2016-06-14 17:16:11 ok 2016-06-14 17:22:39 jirutka 2016-06-14 17:22:45 yeah? 2016-06-14 17:23:16 i think i will head back to nanopi m1 (arm based allwinner h3 quad core board)... alpine works so beautifully with it 2016-06-14 17:23:25 its hardly 11$ 2016-06-14 17:23:43 it still doesn’t work? 2016-06-14 17:23:48 naa it doesnt 2016-06-14 17:24:06 so it still restarts itself? 2016-06-14 17:24:10 yesss 2016-06-14 17:24:15 i mean 2016-06-14 17:24:24 i tried changing u-boot and recompile 2016-06-14 17:24:29 but cannot flash it 2016-06-14 17:24:39 some invalid signature issue 2016-06-14 17:24:43 hmm 2016-06-14 17:24:51 then try the init script instead 2016-06-14 17:24:58 this should work 2016-06-14 17:25:00 i cant even format the internal emmc...there is no tool at all 2016-06-14 17:25:09 init script? 2016-06-14 17:25:37 http://haste.fit.cvut.cz/uxohema.sh 2016-06-14 17:26:17 # echo 1 >/sys/devices/virtual/misc/watchdog/disable /bin/ash: can't create /sys/devices/virtual/misc/watchdog/disable: nonexistent directory / # [ 64.319436] snd_intel_sst: request fw failed 2016-06-14 17:26:39 there no such device? at all? 2016-06-14 17:26:53 when you have run this? 2016-06-14 17:27:05 just sometime back 2016-06-14 17:27:10 is /sys even mounted? 2016-06-14 17:27:12 ls /sys ? 2016-06-14 17:27:27 okie wait let me wait till it mounts and then intercept it... 2016-06-14 17:27:28 is /sys even mounted? 2016-06-14 17:27:31 is /sys even? 2016-06-14 17:27:35 is /sys? 2016-06-14 17:27:37 is? 2016-06-14 17:27:43 haaahaaa ...so funny 2016-06-14 17:27:44 skarnet! 2016-06-14 17:28:17 nonexistent directory / <- well, there is even no /, so /sys cannot be there either ;) 2016-06-14 17:28:17 skarnet: what’s wrong, another grammar mistake or you just don’t like /sys? :P 2016-06-14 17:28:44 nothing's wrong, I just found your "ls /sys?" line funny right after your previous one 2016-06-14 17:28:49 przemoc: ? / is mounted 2016-06-14 17:29:03 aha XD 2016-06-14 17:29:08 but the fact is that /sys is not supposed to be a stable interface, and kernel folks *will* break it 2016-06-14 17:29:27 skarnet: let’s give us a better solution ;) 2016-06-14 17:29:31 so userspace should not use it, if possible, the only exceptions being mdev/udev/etc. 2016-06-14 17:30:08 skarnet: it’s kinda hard to debug something that reboots itself after 30 seconds, you know? 2016-06-14 17:30:23 no shit, Sherlock 2016-06-14 17:30:44 skarnet: so the most imortant right now is to disable this stupid dog by any means, then we can figure out how to do it better 2016-06-14 17:31:02 intel may split into two companies 2016-06-14 17:31:05 thats the talk 2016-06-14 17:31:07 sure, sure, I was just talking in general 2016-06-14 17:31:41 ISTR reading something in the kernel doc about watchdog options 2016-06-14 17:31:55 like, disable_watchdog=1 on the kernel or modprobe command line 2016-06-14 17:32:06 (not exactly that, but something similar) 2016-06-14 17:32:21 okie turned off the stupid watch dog ... 2016-06-14 17:32:31 sit! staaaaay. 2016-06-14 17:33:03 let me see if it still kicks in... 2016-06-14 17:33:20 i want to break a few things around...i dont know which ones 2016-06-14 17:33:36 Try people! They make funny noise when you break them. 2016-06-14 17:33:40 noises* 2016-06-14 17:33:41 XD 2016-06-14 17:34:08 jirutka: it seems stable now 2016-06-14 17:34:14 hurray! 2016-06-14 17:34:19 \o/ 2016-06-14 17:34:24 \o/ 2016-06-14 17:34:41 \o/ 2016-06-14 17:34:57 so what's your final solution? 2016-06-14 17:35:13 should we go back to the ext4 file and disable modloop and put in your script? 2016-06-14 17:35:15 now you can try to start sshd and see if it just needs some time to generate certificates or what’s going on here 2016-06-14 17:35:54 service sshd start restarted in less than a second 2016-06-14 17:35:59 hardly take a sec 2016-06-14 17:36:04 took* 2016-06-14 17:36:10 great! 2016-06-14 17:36:40 about modloop… you have a writable rootfs now, so you can do just `rc-update del modloop sysinit` 2016-06-14 17:37:09 let me do it at the ext4 level 2016-06-14 17:37:18 that way i can flash many other edison with it 2016-06-14 17:37:31 edisons*.... 2016-06-14 17:38:12 how do i make it live ...can i convert this to initramfs? 2016-06-14 17:38:22 then remove the symlink /etc/runlevels/sysinit/modloop 2016-06-14 17:38:38 let me move it and reflash it 2016-06-14 17:38:55 http://haste.fit.cvut.cz/uxohema.sh 2016-06-14 17:38:59 where do i keep this script? 2016-06-14 17:39:22 put it into /etc/init.d/disable-watchdog (or some other name) 2016-06-14 17:39:33 how do i update rc-update? 2016-06-14 17:39:34 and then create a symlink in /etc/runlevels/sysinit 2016-06-14 17:39:38 alrite 2016-06-14 17:58:55 jirutka: http://paste.ubuntu.com/17333781/ 2016-06-14 17:59:08 disabled watch dog timer...waiting on sshd 2016-06-14 17:59:42 when you started it manually from shell, it started in less than second, right? 2016-06-14 17:59:49 yes 2016-06-14 18:00:01 i think the bootargs are still old.... 2016-06-14 18:00:02 hm, that’s weird 2016-06-14 18:00:04 let me change it 2016-06-14 18:03:26 http://paste.ubuntu.com/17333983/ 2016-06-14 18:03:39 the bootargs doesnt matter...can be solved... 2016-06-14 18:03:45 it still hangs at sshd 2016-06-14 18:04:37 okay, try to remove sshd from the runlevel, let’s see what will happen 2016-06-14 18:04:44 sure 2016-06-14 18:14:57 still hangs here 2016-06-14 18:14:58 http://paste.ubuntu.com/17334531/ 2016-06-14 18:15:11 jirutka: is it time to say good bye to edison? 2016-06-14 18:15:20 it depends on you 2016-06-14 18:15:36 per aspera ad astra 2016-06-14 18:15:46 I think that it looks promising 2016-06-14 18:15:56 any other things we could try? 2016-06-14 18:16:12 i am not able to understand what could be the issue here.... 2016-06-14 18:17:06 me neither, the next and last step is just display login prompt 2016-06-14 18:17:15 what can be wrong here 2016-06-14 18:17:18 skarnet, any idea? 2016-06-14 18:19:12 clock skews all over the place 2016-06-14 18:19:32 I'm not sure how openrc handles its dependencies but it could have something to do with it 2016-06-14 18:19:42 if it relies on file dates at some point 2016-06-14 18:19:58 skarnet: what can be done with that clock skew problem? 2016-06-14 18:20:11 the rtc clock needs access to internet 2016-06-14 18:20:26 unfortunately without login i cant get to internet 2016-06-14 18:20:29 boot with init=/bin/sh, then edit the init scripts to add a very early "date -s 2016-06-14" 2016-06-14 18:20:43 alrite 2016-06-14 18:20:45 just set the system clock to something reasonable as early as possible 2016-06-14 18:20:55 doesn't need to be accurate, just reasonable 2016-06-14 18:20:55 skarnet: init=/bin/sh? this will skip OpenRC… 2016-06-14 18:21:10 that's the point, since openrc hangs :P 2016-06-14 18:21:25 try it now 2016-06-14 18:21:28 shall* 2016-06-14 18:21:28 I don’t think that this is a solution 2016-06-14 18:21:39 what about adding a runscript that will set the date? 2016-06-14 18:21:52 that's exactly what I was suggesting 2016-06-14 18:22:20 sure, you can just throw init system away, or throw away whole Edison board, this will also solve the problem XD 2016-06-14 18:22:32 lol 2016-06-14 18:22:57 my init=/bin/sh suggestion was only to have access to the machine in order to edit scripts >.> 2016-06-14 18:23:25 oh, but you have gettys? forget it then, if you can log in normally 2016-06-14 18:23:43 meaning? 2016-06-14 18:24:09 if you have gettys (on tty2 for instance) then you can log in 2016-06-14 18:24:26 well http://haste.fit.cvut.cz/acirale.sh 2016-06-14 18:24:27 if you don't, you can't get a shell unless you reboot with init=/bin/sh 2016-06-14 18:24:33 also snd_intel_sst: request fw failed kicks in once i go via /bin/sh 2016-06-14 18:24:50 its better i try the interactive mode 2016-06-14 18:24:52 who cares about snd_intel_sst 2016-06-14 18:25:01 its reboots the board 2016-06-14 18:25:02 skarnet: we have enabled interactive mode, so he can just tell OpenRC to exit to the shell at any time during boot 2016-06-14 18:25:18 ... oh, openrc can do that. Yay. 2016-06-14 18:25:32 yeah, s6 can’t do that?! 2016-06-14 18:25:44 btw 2015-10-21 is more reasonable time imo, it’s the time of The Future!! 2016-06-14 18:25:44 no, because you don't need it 2016-06-14 18:26:18 well there are files with a later time than that, so no, try something more recent 2016-06-14 18:26:20 at least when everything works… 2016-06-14 18:26:36 okay, so totally boring 2016-06-14… 2016-06-14 18:26:42 yup 2016-06-14 18:26:57 booooring! XD 2016-06-14 18:27:01 gentlemen let me try 2016-06-14 18:27:28 jirutka: with s6-linux-init, you have a getty very early on so you can always log in even when the service manager fails 2016-06-14 18:27:48 oneinsect: you could enable rc_verbose in /etc/rc.conf, so we would see a bit more 2016-06-14 18:28:00 not sure whether it will tell us anything useful, though 2016-06-14 18:28:30 yeah, that’s a good idea! I forgot about this option 2016-06-14 18:28:34 shall do so now for rc_verbose 2016-06-14 18:30:11 rc_verbose="YES" is that it? 2016-06-14 18:30:23 yes 2016-06-14 18:30:28 or even "YES" 2016-06-14 18:30:51 maybe "y" or "1" should work as well :P 2016-06-14 18:31:04 be consistent 2016-06-14 18:31:09 yeah 2016-06-14 18:32:43 well, even default /etc/rc.conf isn't consistent... values are quoted, or not quoted 2016-06-14 18:32:52 http://paste.ubuntu.com/17335291/ 2016-06-14 18:33:00 there are new errors that have shown up 2016-06-14 18:33:43 well, they're not new, we just didn't see them before ;p 2016-06-14 18:33:52 yes indeed 2016-06-14 18:34:19 should disable those services? 2016-06-14 18:35:13 /var/run is not present in your rootfs? 2016-06-14 18:35:42 ln -s /run /var/run 2016-06-14 18:36:00 \o/ pid files \o/ 2016-06-14 18:36:13 skarnet, please behave! ;) 2016-06-14 18:36:35 we all know that PID files are just stupid workaround 2016-06-14 18:36:42 let me try.... 2016-06-14 18:36:57 sorry, when I see start-stop-daemon and pid files... on Alpine... I can't help but point and laugh 2016-06-14 18:37:10 and dance the schadenfreude dance 2016-06-14 18:37:25 well, it’s still better than what other distributions have… 2016-06-14 18:37:56 like SysVinit crap, broken-by-design Upstart crap, totally shitty systemd plague… 2016-06-14 18:37:56 void linux has runit IIRC, so it's not that bad 2016-06-14 18:38:37 run -> /run is present in /var 2016-06-14 18:39:05 and is /run mounted? 2016-06-14 18:39:10 it should be tmpfs 2016-06-14 18:39:17 on the start-stop-daemon and pid files aspect, no, there's nothing worse out there: it's either as bad or correct. (systemd is horrible, but it does eliminate the start-stop-daemon ugliness.) 2016-06-14 18:39:44 (and yes, Void does it better) 2016-06-14 18:39:51 (for now) 2016-06-14 18:39:53 start-stop-daemon is still much better than nothing (like in SysVinit) 2016-06-14 18:40:05 tmpfs 96.0M 44.0K 96.0M 0% /run 2016-06-14 18:40:18 I disagree, but let's stop the OT right here because I can ramble on those subjects for days 2016-06-14 18:40:21 try touch /var/run/foo… 2016-06-14 18:40:45 lock/ openrc/ utmp.... which ones should i touch 2016-06-14 18:40:58 just /var/run/foo 2016-06-14 18:41:04 just to know if it’s really writable 2016-06-14 18:41:12 it is 2016-06-14 18:41:26 i can see foo 2016-06-14 18:41:54 hm, may we have just wrongly interpreted this warning message 2016-06-14 18:42:06 yes, I guess start-stop-daemon fopens always 2016-06-14 18:42:12 OpenRC checks if some pid file already exists 2016-06-14 18:42:21 honestly I don't think the problem comes from the messages we see 2016-06-14 18:42:23 so, ignore it, it’s irrelevant 2016-06-14 18:42:39 things actually succeed 2016-06-14 18:42:49 well, we still don’t know why it hangs there 2016-06-14 18:43:11 there's something openrc doesn't like right after syslog 2016-06-14 18:43:24 you need to check the dependency tree and see what the next service would be 2016-06-14 18:43:37 gentlemen please find the pastebin for dmesg 2016-06-14 18:43:38 http://paste.ubuntu.com/17335724/ 2016-06-14 18:43:44 can we find any clues there? 2016-06-14 18:44:19 it looks like a wrong log… 2016-06-14 18:44:24 this is definitely not from Alpine 2016-06-14 18:44:53 hmmmmm 2016-06-14 18:45:27 shucks sorry sorry ...i paste my vm log 2016-06-14 18:45:34 schizophrenic machine. The kernel thinks it's running systemd, userland thinks it's running openrc 2016-06-14 18:45:35 crucial question, how are you interacting with this board? some serial port perhaps? is proper ttyS* in /etc/inittab then? is it added to /etc/securetty too? 2016-06-14 18:45:36 please wait while i paste my correct log 2016-06-14 18:45:40 yes 2016-06-14 18:45:42 serial ports 2016-06-14 18:46:24 the correct paste bin 2016-06-14 18:46:25 http://paste.ubuntu.com/17335847/ 2016-06-14 18:46:39 skarnet: maybe Intel just injected systemd directly to the chip… same as Dell or Lenovo with some Windows craps? 2016-06-14 18:46:54 przemoc: i am using putty via serial console 2016-06-14 18:46:58 nah, systemd wouldn't fit on a chip 2016-06-14 18:47:31 I think that przemoc had a good point 2016-06-14 18:47:42 so ttyMFD2 needs to be put in /etc/inittab 2016-06-14 18:47:54 http://paste.ubuntu.com/17335902/ 2016-06-14 18:47:55 and in /etc/securetty too 2016-06-14 18:47:58 is my inittab 2016-06-14 18:48:04 yes rightaway 2016-06-14 18:48:21 remove those ttys on VTs you're not using 2016-06-14 18:48:32 and uncomment the serial port one, with the right tty name 2016-06-14 18:48:53 so... the machine is not hanging, you're just not getting a getty on your terminal. XD 2016-06-14 18:49:13 ttyMFD2::respawn:/sbin/getty -L ttyS0 115200 vt100 2016-06-14 18:49:27 is that correct gentlemen 2016-06-14 18:49:29 ttyMFD2, not ttyS0 2016-06-14 18:49:37 after the -L 2016-06-14 18:49:42 ooh yes yes 2016-06-14 18:49:43 ofcourse 2016-06-14 18:50:29 with that, normally you'll get the getty early on, and the openrc messages will still be written there so it will be ugly 2016-06-14 18:50:51 but after it's done, will bring you back to the beloved login: prompt. 2016-06-14 18:51:34 trying now...fingers crossed 2016-06-14 18:52:03 did you comment the tty1...tty6 lines? you won't need it if you don't have a screen with VTs 2016-06-14 18:52:04 holy moly.....i got the prompt 2016-06-14 18:52:09 \o/ 2016-06-14 18:52:14 Welcome to Alpine Linux 3.4 Kernel 3.10.98-poky-edison+ on an i686 (/dev/ttyMFD2) localhost login: 2016-06-14 18:52:25 yaaaay!!! 2016-06-14 18:52:26 5 euros will be enough 2016-06-14 18:52:31 lol 2016-06-14 18:52:31 /jk 2016-06-14 18:52:43 so in short: everything was working fine, and pebkac. 2016-06-14 18:52:48 :P 2016-06-14 18:52:51 now how do i make an initramfs out of this? 2016-06-14 18:53:04 the internal mmc has very limited rw cycles 2016-06-14 18:53:16 btw you guys rock!!! 2016-06-14 18:53:23 [bow in humbleness] 2016-06-14 18:54:03 oneinsect: congratz! you have just deflowered Intel Edison board! XD 2016-06-14 18:54:22 ROFL 2016-06-14 18:54:36 now rename it to Intel Tesla and we can talk 2016-06-14 18:54:47 ^ my point! XD 2016-06-14 18:54:49 its been so much harder than ARM boards...seriously ...without you guys...i wonder 2016-06-14 18:55:15 cat /proc/cpuinfo? 2016-06-14 18:55:17 i am gonna write a wiki entry and release it to public now 2016-06-14 18:55:22 yes hold on 2016-06-14 18:56:10 processor : -1 2016-06-14 18:56:17 http://paste.ubuntu.com/17336221/ 2016-06-14 18:56:24 thats the cpuinfo 2016-06-14 18:57:43 nice, it has aes! 2016-06-14 18:57:53 yeah, my bad - it's an Intel board, what was I thinking, of course it's an Intel CPU 2016-06-14 18:58:24 looks like I missed important milestone 2016-06-14 18:58:35 it could have Outtel CPU 2016-06-14 18:58:48 jirutka: now the question remains how do we get an initramfs working for this? 2016-06-14 18:58:53 but it seems that it doesn’t have VTx… I definitely wanna run KVM virtual machines on SoC! XD 2016-06-14 18:59:20 it has vtx 2016-06-14 18:59:22 oneinsect: I guess that you don’t want just initramfs, but basically a diskless Alpine installation? 2016-06-14 18:59:23 or diskless system 2016-06-14 18:59:34 s/vtx/vmx/ 2016-06-14 18:59:35 yes basically a diskless alpine linux 2016-06-14 18:59:39 sorry 2016-06-14 18:59:42 aha 2016-06-14 18:59:52 jirutka: I was actually disappointed to see that the VMWare Workstation VM doesn't have VTx, so I can't run KVM under VMWare 2016-06-14 19:00:12 vmx, yes 2016-06-14 19:00:34 the cpu of edison is exactly this 2016-06-14 19:00:34 http://www.intel.in/content/dam/www/public/us/en/documents/product-briefs/atom-z34xx-smartphones-tablets-brief.pdf 2016-06-14 19:01:48 oneinsect: well, I’ve never done a diskless installation; I’m sure that there are guys with more experience in this… skarnet, przemoc? 2016-06-14 19:02:11 skarnet: I run vmware fusion on osx, and it does have vmx, so I run alpine in windows10/hyper-v in osx/vmware 2016-06-14 19:02:15 nope, I don't like diskless, it's a hack. 2016-06-14 19:02:40 I haven't used diskless yet either, sorry 2016-06-14 19:02:45 hmmmm 2016-06-14 19:02:50 okay, what I fool was thinking XD 2016-06-14 19:03:02 so, ncopa? could you help oneinsect with custom diskless installation? 2016-06-14 19:03:04 ncopa: maybe fusion on osx, but not Workstation Pro on Windows 10 ;) 2016-06-14 19:03:44 lord @ncopa: please show your grace on this humble being 2016-06-14 19:03:58 XD 2016-06-14 19:03:59 oneinsect: maybe you could poke around rpi image to see how it's done 2016-06-14 19:04:28 i infact use rpi image a lot...its so elagent...yes have to dig in.... 2016-06-14 19:04:39 oneinsect: that will require a sacrifice of two full bags of coffee and one Docker Prayer 2016-06-14 19:05:22 why Docker? better pray to Copa directly 2016-06-14 19:05:34 thyself is willing to offer himself instead of coffee and prayer...what is more better than offering himself...to the lord 2016-06-14 19:05:43 skarnet: just one? :( 2016-06-14 19:06:08 it's to appease the Lord's surroundings when the Lord stops catering to them for a moment and descends to grace mortals 2016-06-14 19:06:31 good one :) 2016-06-14 19:06:35 oneinsect: but very briefly: you need kernel, initramfs and directory with '.boot_directory' (or smth like that) file where all apks are stored 2016-06-14 19:07:14 err, that directory has both .boot_directory file AND apk files, because I fail at putting commas in correct places 2016-06-14 19:07:26 yes i tried extracting the iso and tried building an initramfs but somehow it doesnt work...may be i need to inject the correct tty and disable watch dog 2016-06-14 19:07:44 the iso is kinda strange 2016-06-14 19:08:11 pray can someone please tell me 2016-06-14 19:08:28 the initramfs in x86 iso is similar in structure as rpi initramfs 2016-06-14 19:08:33 the structure i mean? 2016-06-14 19:08:47 the sbin/init etc? 2016-06-14 19:08:52 well, you don't need special initramfs 2016-06-14 19:08:59 hmmm 2016-06-14 19:09:00 the one from mkinitfs is good 2016-06-14 19:09:10 I don't know why do you insist so much on making your own 2016-06-14 19:09:26 diskless install is just the matter of correct kernel parameters 2016-06-14 19:09:33 (mostly) 2016-06-14 19:09:44 find . | cpio -H newc -o | gzip -9 > /root/initramfs 2016-06-14 19:10:09 mkimage -n initramfs -A x86 -O linux -T ramdisk -C none -d initramfs initramfs-edison 2016-06-14 19:10:15 i generally build that way 2016-06-14 19:10:21 is that correct? 2016-06-14 19:10:29 is this what mkinitfs do? 2016-06-14 19:10:51 the above works for me for alpine linux in arm 2016-06-14 19:10:52 https://wiki.alpinelinux.org/wiki/DIY_Fully_working_Alpine_Linux_for_Allwinner_and_Other_ARM_SOCs 2016-06-14 19:11:17 barthalion: just a note, he’s not running Alpine’s kernel, but some old kernel with intel patches 2016-06-14 19:11:23 correct 2016-06-14 19:11:25 oneinsect: dod you know what module you need to disable? 2016-06-14 19:11:37 meaning? 2016-06-14 19:11:39 oh, a custom kernel 2016-06-14 19:11:47 no custom kernels… 2016-06-14 19:11:54 ncopa: this is intel edison board 2016-06-14 19:11:58 so x86, but with uboot 2016-06-14 19:12:04 they havent mainlined the patches yet 2016-06-14 19:12:09 ah, ok 2016-06-14 19:12:10 they are still at 3.10.x 2016-06-14 19:12:18 still, no magic involved 2016-06-14 19:12:41 oneinsect: did you manage to make it boot the kernel? 2016-06-14 19:12:46 i enable all options in kernel config like squashfs, initramfs etc etc 2016-06-14 19:12:48 yes ofcourse 2016-06-14 19:12:53 let me pastbin it 2016-06-14 19:13:00 ok good. i dont have that much experience with uboot 2016-06-14 19:13:29 http://paste.ubuntu.com/17336980/ 2016-06-14 19:13:43 that is a successful boot@ncopa: 2016-06-14 19:14:08 do you know if the initramfs in x86 iso is similar as the initramfs in raspberry pi 2016-06-14 19:14:43 creator knows everything, by definition 2016-06-14 19:14:54 im not creator of rpi stuff... 2016-06-14 19:14:58 so i dont know really 2016-06-14 19:15:06 but i would believe its the same initramfs 2016-06-14 19:15:12 hmmm 2016-06-14 19:15:14 looks like you manage to boot it? 2016-06-14 19:15:19 yesss 2016-06-14 19:15:34 its just that internal mmc has limited rw cycles 2016-06-14 19:15:43 i have to run it from ram only 2016-06-14 19:15:46 aha 2016-06-14 19:15:50 you boot it from disk 2016-06-14 19:15:53 yes 2016-06-14 19:16:03 booted from ext4 partition from internal mmc 2016-06-14 19:17:15 i suppose you could use mkinitfs to build an initramfs image 2016-06-14 19:17:28 i have no clue where to start 2016-06-14 19:17:29 you need make sure the kernel modules for the mmc is included 2016-06-14 19:17:43 you can start with: apk add mkinitfs 2016-06-14 19:17:48 aha 2016-06-14 19:17:50 then mkinitfs -h 2016-06-14 19:18:02 so i can build from within alpine linux? 2016-06-14 19:18:07 yes 2016-06-14 19:18:10 lovely 2016-06-14 19:18:18 from you booted edison 2016-06-14 19:18:23 let me try right away 2016-06-14 19:18:46 then you need make a modloop image 2016-06-14 19:19:28 its a squashfs image right? 2016-06-14 19:20:01 mksquashfs squashfs-temp/ modloop-sunxi -b 1048576 -comp xz -Xdict-size 100% 2016-06-14 19:20:07 something on those lines? 2016-06-14 19:20:26 is modloop necessary @ncopa: ? 2016-06-14 19:20:41 yes 2016-06-14 19:20:45 http://git.alpinelinux.org/cgit/alpine-iso/tree/Makefile#n59 2016-06-14 19:20:55 there is how the official modloop is built 2016-06-14 19:21:04 i need to go. 2016-06-14 19:21:06 see u 2016-06-14 19:21:15 and back up the Lord goes 2016-06-14 19:21:41 thank you Lord @ncopa 2016-06-14 19:21:52 i will come back tomorrow to worship you again 2016-06-14 19:21:55 i will get some sleep 2016-06-14 19:22:11 good fellow sub-lords 2016-06-14 19:22:32 good night* 2016-06-14 19:22:59 we're mere minions 2016-06-14 19:23:20 and i am but a mere insect 2016-06-14 19:23:51 that deflored Edison, so be proud! 2016-06-14 19:24:05 lol 2016-06-14 19:24:09 gone now 2016-06-14 19:24:29 so now for something completely different 2016-06-14 19:25:35 shouldn't builddir definition be officially put in abuild? 2016-06-14 19:26:05 it's used once in default_prepare() as "${builddir:-$srcdir/$pkgname-$pkgver}" 2016-06-14 19:27:23 but I think it would be sensible to have it globally defined with such default (if not already defined), just like srcdir or pkgbasedir are defined 2016-06-14 19:30:27 przemoc: I’m actually planning to improve builddir support in abuild 2016-06-14 19:31:02 we can get rid of `cd "$builddir"` at the start of almost every function in all APKBUILDs 2016-06-14 19:38:30 building out of tree with aports on ro bind also proved to be more complex than it should be 2016-06-14 19:43:02 APKBUILD=/home/przemoc/git/git.alpinelinux.org/aports/main/git/APKBUILD srcdir=$PWD/src pkgbasedir=$PWD/pkg REPODEST=$PWD/repo abuild -r 2016-06-14 19:43:05 just for the record 2016-06-14 20:00:15 jirutka: what happens when you don't build anything? 2016-06-14 20:00:19 you're just packaging up? 2016-06-14 20:00:30 Adran: ? 2016-06-14 20:00:36 your comment above 2016-06-14 20:00:54 I don’t understand…? 2016-06-14 20:01:39 maybe i've been doing it wrong, but I have stuff that is a package, but there's nothing being built. its just unzipping an archive and moving it to the proper directories 2016-06-14 20:05:22 that's fair point 2016-06-14 20:05:32 Adran: I guess you could set builddir to srcdir then 2016-06-14 20:05:46 Adran: most of the packages needs more than just copy bunch of files 2016-06-14 20:06:08 yes, then you don’t need builddir at all, just srcdir 2016-06-14 20:06:16 and you can always redefine builddir 2016-06-14 20:06:45 but I don't see much gain in builddir as starting point 2016-06-14 20:06:52 default builddir would make me more happy 2016-06-14 20:07:01 we need both 2016-06-14 20:07:16 why? 2016-06-14 20:07:18 well, as shown by long history of Alpine, you don't need any 2016-06-14 20:07:27 just grep APKBUILDs for `cd "\$_?builddir"`… 2016-06-14 20:07:40 okay, not need, but it would be very convenient 2016-06-14 20:07:56 removal of "cd $builddir" isn't really huge improvement to packager's life 2016-06-14 20:08:04 the same as default_prepare, it cleaned APKBUILDs a lot, it’s quite silly to copy&paste the same code for applying patchaes all over again 2016-06-14 20:08:07 but I agree that it wouldn't hurt 2016-06-14 20:08:13 yeah 2016-06-14 20:08:57 there's lot of copy-paste e.g. in python and perl packages 2016-06-14 20:09:21 yeah, this too 2016-06-14 20:09:23 maybe we should have default_py_install etc. 2016-06-14 20:09:33 but I remember ncopa complaining about getting to close to ebuilds 2016-06-14 20:09:36 and magic behind 2016-06-14 20:12:18 I totally understand this concern 2016-06-14 20:12:47 I wrote a lot of ebuilds and still don’t understand all the eclasses 2016-06-14 20:12:54 I wasn't proposing removing cd "$builddir", but moving builddir default definition to abuild itself. but I can see why removing cd would be somewhat useful to avoid accidental omissions. 2016-06-14 20:12:56 it’s too much magic 2016-06-14 20:13:52 btw I’m talking just about quite trivial changes that will _not_increase abuild complexity 2016-06-14 20:19:14 uh, it’s already over 22.00 and I’m still in the office 2016-06-14 20:50:02 ncopa: "Error installing php, unsatisfiable dependencies. Provided by php5" 2016-06-14 21:15:09 anyone know if julia needs to link against libllvm or if it just uses the llvm opt/llc/etc.. binaries for its llvm integration? 2016-06-14 21:15:35 it links against libllvm 2016-06-14 21:19:22 ah poop, makes this harder, was hoping not to have to deal with the shared library 2016-06-14 21:19:29 ok cool danke 2016-06-14 21:24:18 you’re working on llvm3.7, llvm3.8…? 2016-06-14 21:25:50 3.7 atm 2016-06-14 21:26:08 cmake is not my most favorite thing atm 2016-06-14 21:29:47 ACTION never liked cmake. another layer of build indirection, but not necessarily improving things nor making them easier to maintain and reason about 2016-06-14 21:31:28 nobody likes cmake 2016-06-14 21:32:02 apparently somebody does and even use it! 2016-06-14 21:32:12 llvm loves it 2016-06-14 21:32:17 they ditched autoconf 2016-06-14 21:32:32 I hate both cmake and make 2016-06-14 21:32:34 autohell is not my cup of tea either 2016-06-14 21:32:37 Using a tool doesn't mean you have to like it. Not at all. 2016-06-14 21:32:52 llvm simply hated autotools more than cmake. 2016-06-14 21:32:59 whole C build system is totally horrible, autohells is probably the worst thing ever invented 2016-06-14 21:33:17 plain old Makefile works. It's not perfect, but it works. 2016-06-14 21:33:33 you can write readable, working and extensible Makefile without autocrap 2016-06-14 21:33:37 ^ 2016-06-14 21:33:43 what about .PHONY? 2016-06-14 21:33:49 Makefile is just crap 2016-06-14 21:34:10 what about .PHONY? 2016-06-14 21:34:18 it's ugly, that's all. 2016-06-14 21:34:34 why the hell I need to add task to .PHONY, when it doesn’t generate a file? 2016-06-14 21:34:37 make is ugly, but it can be made to work. 2016-06-14 21:34:45 Never said it was elegant. 2016-06-14 21:34:50 everything can be made to work, if you’re insane enough… 2016-06-14 21:34:50 Makefiles aren't perfect, true, but they're acceptable IMHO, contrary to alternatives 2016-06-14 21:35:02 I prefer plain shell scripts 2016-06-14 21:35:15 jirutka: the real quote is "everything is a dildo if you're brave enough" 2016-06-14 21:35:20 but obivously just to very simple things 2016-06-14 21:35:44 the best make-like thing I’ve tried is Rake from Ruby 2016-06-14 21:38:13 ok, shell scripts are fine too, but Makefile have some build-related boilerplate built-in that make them more suited than plain shell scripts. I do use shell scripts to automate building many projects in one go, like preparing some sysroot, calling Makefiles in subdirectories and so on. 2016-06-14 21:39:00 this is one example when I used plain shell script instead of Makefile https://github.com/jirutka/one-context/blob/master/install 2016-06-14 21:39:12 as I said, just for very simple things 2016-06-14 21:43:42 I like generating ninja files with shell scripts 2016-06-14 21:47:53 ninja files? 2016-06-14 21:49:31 I hate with passion when building C/C++ project requires some tool in some interpreted languages like python or ruby. in such cases cmake seems a lesser evil (libgit2 case for instance) 2016-06-14 21:50:22 why? 2016-06-14 21:50:43 Ruby is much better for writing more complex scripts than shell or Makefile 2016-06-14 21:51:09 primarily because most people is not able to write clean and readable shell script… 2016-06-14 21:51:43 ninja is basically just a low-level build system for resolving dependencies. kind of the idea of make, but boiled down 2016-06-14 21:51:55 hmm, I’ll take a look at it 2016-06-14 21:52:54 it suits my one use case of a scientific fortran code with ridiculous interdepedencies 2016-06-14 21:52:56 btw todays is not needed to use fullblown MRI Ruby for simple scripts, you can use mruby which is very minimal (re)implementation of Ruby language; it has just 448 kB (https://pkgs.alpinelinux.org/package/edge/testing/x86_64/mruby) 2016-06-14 21:53:13 why? it's such a PITA, requires me to obtain some full-blown runtime just to build something. it's ugly and inconvenient. 2016-06-14 21:54:15 it's a question of dependencies. The point of C/C++ is to not depend on any runtime (except libc/c++); depending on some additional runtime to *build* the project is a pain. 2016-06-14 21:54:22 mruby is very interesting, it’s designed to be embeddable like Lua and intended for building whole application on it as a single executable with no dependencies 2016-06-14 21:54:56 do you know what’s much worse pain? reading awfully writen Makefiles or shell scripts with thousnds of unreadable lines… 2016-06-14 21:55:10 and total hell is reading anything that has something to do with autohells 2016-06-14 21:55:38 pardon, but installing Ruby, mruby or Lua to build the project is NOTHING in comparisinon with this insane mess that C/C++ programmers often calls a build system 2016-06-14 21:56:07 indeed, Ruby and Lua are not build systems. :P 2016-06-14 21:57:04 Higher-level languages do not solve the problem make solves. All they do is make scripts easier to write. 2016-06-14 21:57:06 Rake is 2016-06-14 21:57:16 but that’s the exact point 2016-06-14 21:57:30 not just it 2016-06-14 21:58:32 it’s not just about easier to write, the point is that average programmers are able to write more readable code in it 2016-06-14 21:58:58 like coding monkeys from Google… eh… 2016-06-14 21:59:23 Google people are far from the worst monkeys 2016-06-14 21:59:33 I’m not so sure… 2016-06-14 22:00:00 they're actually pretty smart. The problem with Google is that they have a lot of resources so they underestimate the cost of complexity. 2016-06-14 22:00:07 yeah, you’re right, because popps of the worst monkeys are in proprietary code 2016-06-14 22:00:53 I think that Makefile problem comes from the fact, that not that much programmers pay attention to cleanliness, minimalism, simplicity and moreover they treat building as necessary evil, quickly cooking up some works-for-me build solution (be it terrible Makefile or equivalents in cmake or other crap) 2016-06-14 22:01:11 ^ 2016-06-14 22:01:19 Google, as basically every big company, creates complexity, it’s not in their business interest to decrease it 2016-06-14 22:01:30 also, tbh, the Makefile language doesn't make it easy to write clean and simple stuff. :) 2016-06-14 22:01:53 you both are very right! 2016-06-14 22:03:03 jirutka: no, that conspiracy theory needs to die. Companies do not increase complexity on purpose. Complexity is just a side effect of what they do, because they do not primarily focus on keeping things simple. 2016-06-14 22:03:28 the reasult is the same 2016-06-14 22:05:42 do you know that it’s not a coincidence or inability that every next Windows until Win7 has much higher HW requirements? it used to be a conspiracy that they do it for purpose, to increase sales of hardware… well, todays we know that they really had a deal with Intel… 2016-06-14 22:06:24 but yeah, I don’t think that every company do it for purpose, more likely it’s just a result of incompetence 2016-06-14 22:07:08 I also think that incompetence is the main source of many problems 2016-06-14 22:07:33 maybe I would tell that most of the problems 2016-06-14 22:07:57 look up Hanlon's Razor :) 2016-06-14 22:07:58 but also lack of devotion, because lot of people want to find the quickest solution, not necessarily the best 2016-06-14 22:08:21 yeah, this is very sad 2016-06-14 22:09:39 it was very sad for me to find out that most people don’t care about findin the best solution and they just pick the first one and stick to it forever 2016-06-14 22:10:36 that should be a great filter for the people you choose to associate yourself with and work with :) 2016-06-14 22:11:00 because I’m quite opposite, I almost always want to find not just somehow working solution, but the best one… it’s quite diffucult to me to stop myself, because I need to do also another work 2016-06-14 22:13:22 jirutka: it's double edged sword, and bright side (for company, not for the project) is that people lacking devotion look like quick solvers, bringing solutions faster, so they seem "better". the maintenance problems are in near future, but company is usually happy, because it already has something... 2016-06-14 22:15:29 and that's why company-produced software usually sucks :) 2016-06-14 22:15:38 yeah 2016-06-14 22:17:42 well, there are deadlines and everything in realworld, so giving best solutions is not always possible in short timeframe, that much I understand. what saddens me greatly is often even lack of attempt to do something at least decently and in thought-out manner. 2016-06-15 04:57:45 so is the mailing list pretty much ignored now for aports? 2016-06-15 04:58:03 looks like most of the action is happening with github PRs, no? 2016-06-15 04:59:16 I have four pending patches on the mailing list, three are just minor updates to existing, one is new. none of them have been merged, should I redo those as github PRs? 2016-06-15 04:59:22 or is Timo just on vacation? 2016-06-15 06:39:16 boingolov: no, it's just jirutka being super active 2016-06-15 06:39:28 I can go through patchwork on friday, if it can wait 2016-06-15 06:53:08 barthalion: 2016-06-15 06:53:16 do you have some spare time 2016-06-15 06:53:26 yes 2016-06-15 06:55:04 should we test a diskless for intel edison 2016-06-15 07:05:45 morning 2016-06-15 07:07:22 ScrumpyJack: 2016-06-15 07:07:24 good morning 2016-06-15 07:07:28 how is life today? 2016-06-15 07:10:24 its sucking with my exams 2016-06-15 07:10:46 hmmmm...you still write exams? 2016-06-15 07:10:54 Less depressing than yesterday, the closer we get to Friday 2016-06-15 07:11:20 what's that in regex 2016-06-15 07:12:30 yes but 2016-06-15 07:40:16 Query: I installed ruby2.2 from edge_community repo, why am I getting "gem- not found" errors when I try "gem install" 2016-06-15 07:40:18 ? 2016-06-15 08:07:27 Thanks, it was a DNS issue, nameservers were incorrect 2016-06-15 08:18:47 clandmeter why pkgs.alpinelinux.org/flagged does not update? 2016-06-15 08:21:02 dunno 2016-06-15 08:21:06 probably an sql issue 2016-06-15 08:21:17 the pager is also broken on that page. 2016-06-15 08:21:27 did you flag something? 2016-06-15 08:22:12 nope, i've upgraded packages a while but is still not notified. I notice that a while ago, but always forgot to mention 2016-06-15 08:22:39 explain? what do you expect? 2016-06-15 08:22:58 you upgraded a pkg, but it still lists in flagged? 2016-06-15 08:23:39 that after pushed the upgrade, the package should not be there anymore. 2016-06-15 08:25:03 which pkg? 2016-06-15 08:28:40 nah, nevermind. 2016-06-15 08:28:50 now looks that it works 2016-06-15 08:28:55 probably it was fixed 2016-06-15 08:29:10 sorry for the noise 2016-06-15 08:33:08 it takes max 15 min 2016-06-15 08:56:47 ncopa, are you ok with: http://sprunge.us/aICf 2016-06-15 08:56:55 clandmeter, and you with: http://sprunge.us/hdfQ ? 2016-06-15 09:23:57 sure push it 2016-06-15 09:24:13 maybe check git changelog for big changes 2016-06-15 09:24:24 git changelog.txt :) 2016-06-15 09:29:50 fcolista: pushit 2016-06-15 09:31:17 boingolov: sorry for not having time for the alpine-aports 2016-06-15 09:32:25 wow we are really lagging behind 2016-06-15 12:36:05 jiruthka: good morning 2016-06-15 12:36:14 jirutka: 2016-06-15 12:44:31 good afternoon :) 2016-06-15 12:44:40 I’m in GMT+02 2016-06-15 12:45:33 great...shall we begin 2016-06-15 12:45:39 our last assault 2016-06-15 12:50:28 ACTION sells heavy armor 2016-06-15 12:54:13 so diskless... basically the iso boots up and then looks up for the apks folder and extract them and then you will find the /etc/inittab and /etc/securetty 2016-06-15 12:54:59 however without ttyMFD2 the terminal will not come up...and without the terminal i cant put it in /etc/inittab etc 2016-06-15 12:55:06 chicken and egg 2016-06-15 12:57:09 similar issue for the watchdog timer 2016-06-15 15:38:03 oneinsect: sorry, I was busy at work 2016-06-15 15:38:28 no issues there is always time 2016-06-15 15:38:31 :-D 2016-06-15 15:39:06 not sure if you guys saw this 2016-06-15 15:39:08 http://arstechnica.com/information-technology/2016/06/goodbye-apt-and-yum-ubuntus-snap-apps-are-coming-to-distros-everywhere/ 2016-06-15 15:39:16 but it directly relates to the stuff that was being discussed earlier 2016-06-15 15:40:12 Adran: this looks way more interesting https://www.habitat.sh/ 2016-06-15 15:40:33 Adran: I’m just reading documentation and it looks very promising 2016-06-15 15:40:44 ACTION which habitat or snaps? 2016-06-15 15:41:01 ACTION votes for Habitat! 2016-06-15 15:41:17 it looks like apks 2016-06-15 15:41:22 formatwise 2016-06-15 15:41:29 but that isn't the point of what I'm getting at 2016-06-15 15:41:38 yes, the format of plan.sh is inspired from Arch’s PKGBUILD 2016-06-15 15:42:00 both of those habitat and snaps bundle everything into a "package" 2016-06-15 15:42:05 yes 2016-06-15 15:42:07 which means you get a lot of duplication and such 2016-06-15 15:42:16 yeah 2016-06-15 15:44:23 how does that improve system security or system size? 2016-06-15 15:44:30 it seems like this makes things worse 2016-06-15 15:45:15 Habitat takes security seriously from what I have read 2016-06-15 15:45:34 it’s not oriented to small system size 2016-06-15 15:46:08 both of them "security seriously" 2016-06-15 15:46:21 But you have specific versions of libraries, if those libraries turn out vulnerable 2016-06-15 15:46:21 it’s just different tradeoff… application bundles (i.e. app with all dependencies) solves a lot of problem of traditional approach, but the price is that bundles are larger 2016-06-15 15:46:36 then you're leaving that entire application vulnerable 2016-06-15 15:46:36 then you can rebuild the package 2016-06-15 15:46:40 e.g. what happened with OS X 2016-06-15 15:46:46 I still have 8 apps on my mac that are vulnerable 2016-06-15 15:46:56 and this process is automated, including configuration updates 2016-06-15 15:47:26 it’s actually easier to update bundle, because you don’t care about other applications linked against different versions, there’s just one application 2016-06-15 15:47:43 so you can atomically update each application 2016-06-15 15:47:52 but if someone leaves maintaining then you're stuck 2016-06-15 15:47:57 indeed 2016-06-15 15:48:10 but how is conventional approach different? 2016-06-15 15:48:20 you always must care about the system 2016-06-15 15:48:41 well things like openssl, etc are probably updated separately 2016-06-15 15:49:09 again, when new version of openssl is released, you can update all the pacakges that uses it and all the running instances automatically 2016-06-15 15:50:31 Habitat basically takes a role of the distribution, with all its pros and cons 2016-06-15 15:53:29 it solves quite the same problem as Docker + configuration management tools, but IMO in fundamentally better and cleaner way 2016-06-15 15:54:35 but I haven’t tried it yet… at the first sight it looks good, we will see later 2016-06-15 16:00:19 ACTION chimes in and advertise https://github.com/gmkurtzer/singularity 2016-06-15 16:00:46 tru_tru: thanks! I didn’t know about singularity, I’ll take a look 2016-06-15 16:02:22 it builds and run on alpine/voidlinux, ie you can run (centos/debian) containers on alpine 2016-06-15 16:03:08 tru_tru: > “Singularity does not provide a pathway for escalation of privilege… so you must be able to become root on the host system … in order to modify the container. 2016-06-15 16:03:08 so you can’t be a root in the container without being a root on the host system? 2016-06-15 16:03:24 i.e. it doesn’t use user namespace? 2016-06-15 16:04:13 you need to loop mount, ie needs root priv. 2016-06-15 16:04:58 ? 2016-06-15 16:05:00 it is design for end-user running software anywhere, but building these containers needs root 2016-06-15 16:05:21 what if some software running inside the container needs root privileges? 2016-06-15 16:05:50 anothrer: > Additionally relevant file systems on your host are automatically shared within the context of your container. 2016-06-15 16:05:50 what are “relevant file systems”? how does it work? 2016-06-15 16:06:12 singularity is not designed for that use case for the moment, afaik 2016-06-15 16:06:19 aha :( 2016-06-15 16:07:10 so it’s not quite usable for me… however, still a good thing to be added to my lists of container solutions better than Docker (okay, if we pretend that Docker is a container XD) 2016-06-15 16:08:19 also I like its name :P 2016-06-15 16:09:46 it looks like unshare with few extra features 2016-06-15 16:10:03 mount -> /etc/singularity/singularity.conf reads: 2016-06-15 16:10:10 mount proc = yes ## Should we automatically mount /proc within the container? 2016-06-15 16:10:21 bind path = /home 2016-06-15 16:10:27 etc 2016-06-15 16:10:27 aha 2016-06-15 16:10:53 ACTION looking at habitat 2016-06-15 16:20:27 main grief about habitat: no automatic dependency setup, no yet conflict management (not seen yet) 2016-06-15 16:20:48 what do you mean with dependency setup? 2016-06-15 16:21:13 it’s very similar to e.g. Alpine package, i.e. you explicitly declare its dependencies 2016-06-15 16:21:22 it automatically resolves transitive dependencies 2016-06-15 16:21:32 you write the dependancy manually... 2016-06-15 16:22:02 yes, exactly the same as in all linux packages… 2016-06-15 16:22:34 hm, actually apk is able to autodiscover dynamically linked dependencies 2016-06-15 16:22:57 not sure if there will be such an option 2016-06-15 16:23:39 rpm does it automatically, although you can specify name-version-release for deps 2016-06-15 16:24:21 it can’t resolve all the dependencies, because there’s no such mechanism, the best it can do is the same as apk does 2016-06-15 16:24:35 i.e. scan binaries using ldd 2016-06-15 16:26:56 I am still looking at complicated setup, see easybuild recipes for toolchains hpc low levels libs (blas/lapack/atlas/openblas/..), or perl/python/R nigthmare libs 2016-06-15 16:27:54 what easybuild recipes? 2016-06-15 16:28:35 https://github.com/hpcugent/easybuild 2016-06-15 16:29:10 http://easybuild.readthedocs.io is quite nice 2016-06-15 16:29:18 I don’t know this; however, I’m the maintainer of the Alpine’s R and Julia packages, both uses OpenBLAS and other math libs 2016-06-15 16:31:15 easybuild tries to answer: how would you rebuild R with the intel compilers? or anothe gcc version and let them live at the same time on the same machine 2016-06-15 16:32:09 with Habitat, there’s no such problem as how to use multiple versions or variant of some lib on single machine… 2016-06-15 16:32:49 it’s something like building a statically linked binary with zero dependencies 2016-06-15 16:33:03 but I’m not sure if I understand your concern 2016-06-15 16:34:40 I was just wondering how habitat would scale when one adds compilers versions, openblas versions in the loop 2016-06-15 16:34:54 in the loop? 2016-06-15 16:35:15 for each compiler ; for each openblas version; rebuild R 2016-06-15 16:36:13 hm, you can use just single set of dependencies per package, so you must decide which to use; or build multiple packages for each variant 2016-06-15 16:36:18 but habitat does look interesting 2016-06-15 16:36:32 but that’s not much different from what we currently have e.g. on Alpine 2016-06-15 16:36:48 obviously, it’s be design less flexible than Gentoo 2016-06-15 16:36:56 by 2016-06-15 16:37:23 I like the ring stuff 2016-06-15 16:39:36 Hi folks - I'm trying to build a package for phusion passenger, and am having trouble with compilation and looking for any pointers you all might have. Things build fine if I run them in prepare(), but the same commands in package() blow up 2016-06-15 16:40:29 ACTION waves bye 2016-06-15 16:40:45 From what I've been able to figure out, it's the fakeroot environment that's causing the issue - LD_LIBRARY_PATH is unset outside of fakeroot, but inside it is set, and the package list from 'apk info' is different in the fakeroot as well. 2016-06-15 16:40:50 tru_true: bye and thanks for tip to the singularity 2016-06-15 16:41:14 phusion passanger is a module for nginx, isn’t it? 2016-06-15 16:41:15 ryanschwartz: prepare() is for applying patches etc. use build() for actual compile 2016-06-15 16:42:04 package() which runs in fakeroot (as you noticed) should only copy the compiled stuff to the staging area, eg `make install DESTDIR="$pkgdir"` 2016-06-15 16:42:19 jirutka1: yes, it is - it actually grabs nginx and builds a custom binary with the module compiled in 2016-06-15 16:43:45 ok, so package() should just grab the stuff that compiled in build() - I didn't find the documentation on the whole process very easy to follow. 2016-06-15 16:44:47 the thing about passenger is there's no "make install" - building things creates an /opt/nginx with the compiled nginx in it 2016-06-15 16:45:07 i.e. it's a one-shot build/install 2016-06-15 16:45:23 (which is why I was running it in package() in the first place, I guess) 2016-06-15 16:45:32 sounds like the didn't design it for package managers 2016-06-15 16:45:38 they* 2016-06-15 16:46:46 i suppose you could create a temp dir in "$srcdir" when it build/sintalls it 2016-06-15 16:46:56 and then you could move it to $pkgdir 2016-06-15 16:50:01 no, it's not very packager friendly - it wraps a bunch of stuff up in a rake task and hides it so the average "not interested in how it works, just gimme the bits" user 2016-06-15 16:55:09 I guess I'll see if I can figure out exactly what they're doing in the install process and tease things apart 2016-06-15 20:26:27 barthalion: any news about lldb? :) 2016-06-16 02:39:48 Anyone ever get therubyracer working on Alpine 3.4? 2016-06-16 04:36:47 barthalion: are you guys here 2016-06-16 04:37:01 any info on howto generate an initramfs file in alpine chroot? 2016-06-16 04:37:16 and a modloop file 2016-06-16 06:58:20 morning. 2016-06-16 07:03:52 what's the make target for alpine-rpi to end up with the archive? 2016-06-16 07:21:24 overlays seem to be missing from alpine-rpi-3.4.0-armhf 2016-06-16 07:38:46 hmm 2016-06-16 07:41:16 fabled: i got one of these, for fun. http://clusterhat.com/ 2016-06-16 07:42:06 cool 2016-06-16 07:42:15 the makefile should copy the overlays 2016-06-16 07:42:15 so now i have a reason to get usb gadget to work :) 2016-06-16 07:42:35 could it be becuase of this? http://git.alpinelinux.org/cgit/alpine-iso/commit/?id=56c1f95f8b694e5a8ee823fd3d422e8a47dc0f02 2016-06-16 07:42:49 http://git.alpinelinux.org/cgit/alpine-iso/tree/Makefile#n452 2016-06-16 07:43:00 oh 2016-06-16 07:46:16 yes, looks like thekernel image is missing the overlays now 2016-06-16 07:49:02 the one from raspberry pi foundation? 2016-06-16 07:52:58 smells like they split the dtb install to separate make target 2016-06-16 07:54:19 oh, we copy them by hand 2016-06-16 07:54:35 should probably change that to dtb_install 2016-06-16 07:54:56 dtbs_install that is 2016-06-16 07:55:40 ah 2016-06-16 07:55:45 they renamed them from .dtb -> .dtbo 2016-06-16 07:55:52 that's why they got left out 2016-06-16 07:59:56 so they should end up in linux-rpi package? 2016-06-16 08:00:17 yes 2016-06-16 08:01:04 http://sprunge.us/aiCF is probably the minimal fix 2016-06-16 08:01:15 linux-rpi is one version behind too 2016-06-16 08:01:23 but i think we should fix that to use dtbs_install make target 2016-06-16 08:03:20 right, so you pull them into http://dev.alpinelinux.org/~tteras/rpi/linux-*.patch 2016-06-16 08:03:55 then apply that patch 2016-06-16 08:06:28 do you cherry pick what goes into http://dev.alpinelinux.org/~tteras/rpi/linux-*.patch ? 2016-06-16 08:08:59 no, it's always plain diff of release vs. github rpi tree 2016-06-16 08:22:17 hey :) 2016-06-16 08:22:21 how is everyone? 2016-06-16 08:24:11 clandmet1r, too bad new patchwork.alpinelinux.org still not support comments on patch :( 2016-06-16 08:24:24 nice job btw 2016-06-16 08:24:52 ScrumpyJack, you around? 2016-06-16 08:27:11 yup 2016-06-16 08:27:33 hi. re: py-skywriter-hat 2016-06-16 08:27:47 you've set arch="armhf" 2016-06-16 08:28:11 it shoubl be set to "noarch", since no binary compile are done 2016-06-16 08:31:42 good spot, i'll change it 2016-06-16 08:31:54 nevermind 2016-06-16 08:31:59 just FYI 2016-06-16 08:32:19 i've merged your patch and changed it 2016-06-16 08:38:41 ncopa: could you please trigger a claws-mail rebuild so it finds its shared libraries again, currently its impossible to install it 2016-06-16 08:38:44 big thanks :) 2016-06-16 08:48:51 fcolista: thanks dude. I'm flattered, but there are older patches on patchwork that could probably do with attention :) 2016-06-16 09:20:57 fcolista: it doesnt, and probably never will. 2016-06-16 09:21:34 is still in their feature request...but yes, i agree with you that probably never will. 2016-06-16 09:22:11 fcolista: Patchwork isn't intended to replace a community mailing list; that's why you can't comment on a patch in Patchwork. If this were the case, then there would be two forums of discussion on patches, which fragments the patch review process. Developers who don't use Patchwork would get left out of the discussion. 2016-06-16 09:24:03 the only way would be if patchwork would send to the ML on behalf of the logged in person. 2016-06-16 09:33:01 clandmeter: that's the idea :) 2016-06-16 09:56:29 clandmeter: patchwork can display e-mails replies as comments 2016-06-16 09:56:33 it doesn't seem to work like that now 2016-06-16 09:57:38 any thoughts on this? http://openbuildservice.org/about/ 2016-06-16 11:23:51 barthalion: http://patchwork.alpinelinux.org/patch/2033 2016-06-16 11:25:00 apparently I was talking about previous version 2016-06-16 11:28:03 barthalion: i dont get it 2016-06-16 11:28:15 so it works now? or it didnt work before? 2016-06-16 11:37:26 yeah 2016-06-16 12:55:58 fabled: any chance you could apply your http://sprunge.us/aiCF and rebuild linux-rpi? 2016-06-16 13:12:28 i'd like to use dtbs_install instead 2016-06-16 13:12:35 building on arm is slow 2016-06-16 18:38:04 I'm creating a package that creates a user in pkgname.pre-install . This user does not exist on the build host. If I do chown of files in pkgname.post-install , I get a warning that I should not use chwon in the post-install, but instead use the package() function inthe APKBUILD file 2016-06-16 18:38:09 seems kind of chicken and egg, no? 2016-06-16 18:38:14 is there an easy way to work around this? 2016-06-16 18:42:00 or does it actually run the pre and post install scripts? 2016-06-16 18:42:05 when building? 2016-06-16 18:43:47 no 2016-06-16 18:43:54 I get unknown user/group blah 2016-06-16 18:45:01 I guess I'll just ignore the warnings 2016-06-16 18:49:58 boingolov: use pkgusers and pkggroups. do chown in package(), not in scripts. 2016-06-16 18:50:08 przemoc: okay 2016-06-16 18:50:15 https://wiki.alpinelinux.org/wiki/APKBUILD_Reference 2016-06-16 18:50:39 ahh, thanks! 2016-06-16 18:58:16 please use builddir instead of _builddir, as _ prefix is reserved for APKBUILD 2016-06-16 19:17:14 przemoc: the documentation should probably be adjusted: https://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package#builddir 2016-06-17 06:09:58 ACTION don't know how to see whether barthalion replied while I was away .. 2016-06-17 06:10:26 peterj: he doesn't, lldb is on my today schedule 2016-06-17 06:11:56 barthalion: that's a great news! :D 2016-06-17 07:26:40 ncopa, i need to add python3 support on py-psutils. This is what i've done (not yet pushed): 2016-06-17 07:26:40 http://sprunge.us/OYMB 2016-06-17 07:26:46 what do you think? 2016-06-17 07:50:59 that looks pretty good 2016-06-17 07:51:15 i've verified that it works. 2016-06-17 07:51:51 so i need to rename py-psutil package in py2-psutil and push it 2016-06-17 07:51:52 i wonder what we should cann the APKBUILD 2016-06-17 07:52:01 call* 2016-06-17 07:52:08 yeah 2016-06-17 07:52:11 we can do: 2016-06-17 07:52:23 py-psutil as metapackage that install both py2 and py3 2016-06-17 07:52:25 or 2016-06-17 07:52:35 py2-psutil with py3 as subpkacage 2016-06-17 07:52:50 dunno what is pros/cons of both approach though 2016-06-17 07:53:18 i prefer first option 2016-06-17 07:53:33 because at some point we will probably want to get rid of py2-* 2016-06-17 07:54:27 python3.5 2016-06-17 07:54:35 so py-psutil will not have any file 2016-06-17 07:54:39 we should also have a _py2ver=2.7 and _py3ver=3.5, or a way to figure out automatically. 2016-06-17 07:54:52 instead of hardcode that, I'd like to get that dynamic 2016-06-17 07:54:59 exactly 2016-06-17 07:55:08 eg execyte python and make it print version, or directory 2016-06-17 07:55:32 we can run python first 2016-06-17 07:55:37 because is not yet installed 2016-06-17 07:56:14 should we use apk info python and strip the version? 2016-06-17 07:57:05 i think we can make python print the libdir 2016-06-17 07:57:08 or similar 2016-06-17 08:00:39 python -m sysconfig contains all the info 2016-06-17 08:00:49 What is the best way to distribute an Alpine custom repo key ? 2016-06-17 08:01:34 I thought about doing a package for it, but you would still have to --allow-untrusted the package to get the key 2016-06-17 08:01:35 coredumb, i create package with the key + /etc/apk/repositories.d/repo.list file 2016-06-17 08:01:49 initial time you need allow-untrusted 2016-06-17 08:01:52 after that it's in 2016-06-17 08:02:02 on diskless installs lbu picks up the keys for overlay 2016-06-17 08:02:03 yep OK that's what I thought then 2016-06-17 08:02:38 fabled: before I get to read abuild code, how do I manage my APKINDEX file ? 2016-06-17 08:03:20 this is how i build mine: https://github.com/rameplayerorg/rameplayer-alpine/blob/master/build.sh#L37 2016-06-17 08:09:46 thx fabled gonna check that 2016-06-17 08:12:49 oh apk index 2016-06-17 08:12:54 easier than I thought :P 2016-06-17 08:17:08 fcolista: something like this: http://tpaste.us/3lXa 2016-06-17 08:17:26 i think we need rename python to python2 2016-06-17 08:17:55 yes 2016-06-17 08:17:58 umh 2016-06-17 08:18:15 sounds like a template we should apply to all py[23]packages 2016-06-17 08:18:23 correct 2016-06-17 08:18:28 thats what i'm trying to do :) 2016-06-17 08:18:48 yes, i guessed :) 2016-06-17 08:18:49 what i found encouraging was that it looks simpler than I feared 2016-06-17 08:19:06 yes 2016-06-17 08:19:10 lua is much worst 2016-06-17 08:19:26 this because we can run setup.install in package() for each py version 2016-06-17 08:19:27 there we also have more versions to deal with 2016-06-17 08:19:33 this makes the things simpler 2016-06-17 08:19:37 yes 2016-06-17 08:19:59 i think barthalion might have an opinion there too 2016-06-17 08:20:06 there are some discussion on it on the ml 2016-06-17 08:20:24 yes, I saw that, but we didn't reach any conclusion so far 2016-06-17 08:20:27 iirc 2016-06-17 08:21:03 I did reach my conclusion 2016-06-17 08:21:20 I don't see any reasons not to remove old, unmaintained or unpopular python libraries 2016-06-17 08:22:42 ncopa: why install both py2 and py3 from package()? 2016-06-17 08:23:03 ah, I see why 2016-06-17 08:23:49 ncopa: I'd rather run it in py2 and py3 functions though, if package() can be left empty 2016-06-17 08:25:04 ScrumpyJack, i have rpi dtbs fix + refreshed patch test building. will push later today if it builds ok. 2016-06-17 08:26:28 http://sprunge.us/iDGI 2016-06-17 08:27:18 ncopa, python2 and python3 syntax are not the same to get stdlib 2016-06-17 08:28:27 i would not rely on a library to get this info 2016-06-17 08:29:05 python might change libs and break things (like in this case) 2016-06-17 08:32:20 can we wait with this discussion a couple of days? just do whatever you need to package it now, I'm going to walk through all py apkbuilds anyway 2016-06-17 08:35:44 fabled: awesome. and i had just raised an issue. #5729. quickest fix of the week! 2016-06-17 08:40:17 we can wait with the discussion 2016-06-17 09:11:10 ncopa, anyway, i'm going to commit this: http://sprunge.us/NQje 2016-06-17 09:11:57 small changes to the lib functions, set arch to subpkgdir, improved python stdlib detection for py2 and py3 2016-06-17 09:13:07 guys and gals (are there any here?), what would you say about improving favicons for bugs.a.o, lists.a.o, patchwork.a.o and wiki.a.o, so they would have small letter (maybe red for better visibility?) in the bottom-right corner on top of AL logo (B, L, P or W respectively). when I'm on Fx I have it configured that tabs are as short as displayed favicon, so it would help a bit in visual navigation. 2016-06-17 09:13:13 just a random idea that came to me just now. 2016-06-17 09:23:55 ncopa: what would be your stance regarding patches that would be aimed at 3.4 only, like adding links .pre-upgrade -> .pre-install and adding additional `addgroup NAME NAME` there, so people that have wrong primary groups for users created in 3.4.0 will get at least proper membership? changing primary group is a no go, because it could have been altered by users on their own or they could have adapt 2016-06-17 09:24:01 ed already used one. they rather shouldn't go to master, as these are 3.4 fixes. 2016-06-17 09:32:43 przemoc: good idea re favicon 2016-06-17 09:33:17 przemoc: we can push 3.4 specific fixes to 3.4-stable branch 2016-06-17 09:33:20 that no problem 2016-06-17 09:33:41 but it would be nice to have a corresponding fix for master too 2016-06-17 09:36:00 is that normal that php7 doesn't have any -cli subpackage ? 2016-06-17 09:36:20 coredumb: i dont know. i didnt do the php7 package 2016-06-17 09:36:21 pkg.a.o gives me no result for phpize binary on php7 2016-06-17 09:36:35 its maybe in php7-dev? 2016-06-17 09:36:45 i think we should have a -cli package too 2016-06-17 09:36:53 was actually fixing php-geoip => php5-geoip 2016-06-17 09:37:04 check git log who did the php7 stuff 2016-06-17 09:37:06 thought I could do php7 as well while at it 2016-06-17 09:37:07 and ask hum 2016-06-17 09:37:10 ask him 2016-06-17 09:38:40 ncopa: will check 2016-06-17 09:42:07 my point is that master is used to branch out new release (3.5) in future, so having band-aids (like additional addgroup to do what adduser already does in current form [as it has proper primary group specified, but only for packages that haven't been installed yet]) that are necessary only for people upgrading from 3.4 only would be not nice. the only problem I can see with such solution would be p 2016-06-17 09:42:13 eople keeping 3.4 not upgraded to 3.4.1 and in future upgrading directly to 3.5. so the question is whether we really want to support that too? 2016-06-17 09:45:02 (I assumed that people living on the edge can remove package, remove old user entry and reinstall the package, that's why I find leaving such changes out of master acceptable) 2016-06-17 09:47:54 or would you prefer to introduce all band-aids to master and simply phase them out in master after 3.5 release? 2016-06-17 09:49:19 (pre-upgrade -> pre-install links count as band-aid too) 2016-06-17 09:53:06 clandmeter: does patchwork support showing cover letters? 2016-06-17 09:54:22 lists.a.o also breaks formatting of mails. some new lines are lost making them tad less readable. 2016-06-17 10:01:17 I noticed also that abuild lacks user/group removal of those that have not been present before building a package, so it's advised to not test the package on the same host you build it using abuild, unless you do the passwd/group cleanings yourself. 2016-06-17 10:04:20 to mitigate it I think that adduser should never create its "home" directory (-H), well, whatever dir is used by package binaries/libraries later, but it should always be created in APKBUILD's package() 2016-06-17 10:05:27 and it's a good guideline regardless or lack of user/group removal after abuild run. 2016-06-17 10:05:44 s/(regardless) or/\1 of/ 2016-06-17 12:54:06 is api.alpinelinux.org usage documented somehow or somewhere ? 2016-06-17 13:11:25 fcolista: AFAIK the only documentation is the source code 2016-06-17 13:11:27 https://github.com/insteps/aport-api 2016-06-17 13:13:23 thx przemoc 2016-06-17 13:14:55 you can also `curl https://api.alpinelinux.org/packages | jq . | less` and look at the links in the json to get some idea 2016-06-17 15:01:06 is there some documentation or guide or git repo how i can setup a custom build server for personal alpine packages? 2016-06-17 15:03:17 mosez: when you do abuild 2016-06-17 15:03:20 it generates a repository directory 2016-06-17 15:03:26 you can host that and download from it 2016-06-17 15:10:08 ok... currently i only have an arm server running natively with alpine. the rest is running as docker images on amd64... 2016-06-17 15:10:36 mostly i want to build packages for jitsi-videobridge and jitsi-meet. 2016-06-17 15:13:26 oh god... damn java packaging -.- https://github.com/jitsi/jitsi-videobridge 2016-06-17 15:13:54 https://github.com/jitsi/jitsi-meet the node stuff shouldnt be that complicated. 2016-06-17 15:14:03 luckily prosody is already packaged 2016-06-17 15:41:46 barthalion: when you have a minute, can you look at my babeltrace fix and lttng-tools new aport, merging those would make lttng userspace tracing fully functionnal 2016-06-17 18:04:48 mjeanson: yeah, saw these, sorry for the delay 2016-06-17 18:04:50 tough week 2016-06-17 18:05:01 jirutka: is travis capable to test-build llvm? 2016-06-17 18:06:20 barthalion: do you have a concept of maintainer like debian, where I could push updates to my packages without having to bother you? 2016-06-17 18:06:45 barthalion: I dunno, Travis has some limitations – IIRC max 4 MiB console log and 60 min (or 120 min?) runtime 2016-06-17 18:07:23 mjeanson: kind of, a trusted user, who can push to testing and community 2016-06-17 18:07:24 mjeanson: you can push it to https://github.com/alpinelinux/aports/pulls and we will take care of it ;) 2016-06-17 18:08:37 mjeanson: I have no reasons to complain about quality of your patches, I can talk with ncopa about it 2016-06-17 18:10:33 jirutka: hm, so you tested julia/llvm locally? 2016-06-17 18:10:44 what's the preferred method, patches on the ML or pull requests on GH? 2016-06-17 18:10:45 barthalion: yes 2016-06-17 18:12:24 mjeanson: ML, but recently jirutka has been hyper active so pull requests tend to get merged sooner 2016-06-17 18:12:57 ACTION being hyperactive XD 2016-06-17 18:14:28 mjeanson: both are supported. pull requests are automatically built on Travis CI, so you’ll get quick response if package is okay (i.e. at least if it can be build); and I’m also doing code-review for pull requests 2016-06-17 18:14:31 barthalion: I'd be interested in commit rights to testing/community whenever you think it's appropriate 2016-06-17 18:16:18 jirutka: one PR per package? 2016-06-17 18:16:55 mjeanson: if the packages depends one each other, then it’s better to send it in single PR 2016-06-17 18:18:49 ok great, the CI job will build multiple packages. Ordered by commits? 2016-06-17 18:19:18 ordered by dependencies 2016-06-17 18:21:11 barthalion: could you please take care of https://github.com/alpinelinux/aports/pull/123 ? 2016-06-17 18:22:12 moving source code sounds a little bit ugly, but LGTM 2016-06-17 18:22:33 go binaries cannot be stripped? 2016-06-17 18:23:01 I don’t know 2016-06-17 18:23:52 Go has extremly stupid handling of dependencies, so I don’t know if this nonsense he’s doing is the only solution to build this Go crap, or if he’s doing it wrong 2016-06-17 18:25:59 mjeanson: okay, your patches are in, thank you! 2016-06-17 18:26:16 jirutka: I'd be really happy if travis could stop e-mailing me about broken build 2016-06-17 18:26:31 barthalion: okay, I’ll disabled it 2016-06-17 18:26:46 thanks 2016-06-17 18:26:55 barthalion: well, thank you 2016-06-17 18:28:43 mjeanson: does it all work in grsec kernel? 2016-06-17 18:29:20 przemoc: per our previous discussion, no 2016-06-17 18:29:39 paxmark…? 2016-06-17 18:29:45 przemoc: kernel tracing won't work but userspace tracing should be fine 2016-06-17 18:30:26 ok, sorry, I was switched to other stuff and forgot about our prior discussions 2016-06-17 18:30:27 though I forgot to run the test suite with the grsec kernel 2016-06-17 18:30:38 I should try that 2016-06-17 18:30:44 definitely 2016-06-17 18:35:32 I don't see any information in the commit introducing lttng regarding not working with grsec. it's good to add such stuff. 2016-06-17 18:35:37 mjeanson: maybe you could create some wiki page regarding lttng on AL? 2016-06-17 18:38:10 przemoc: sure, I just have to register and I can create pages? 2016-06-17 18:38:16 exactly 2016-06-17 18:38:19 yes 2016-06-17 18:41:19 test 2016-06-17 18:41:57 functions are executed in order specified in $subpackages? 2016-06-17 18:42:06 yes 2016-06-17 18:45:47 mjeanson: and you're right, wiki should be updated regarding builddir. but it's not a simple s/_//, because abuild's default_prepare in AL 3.3 may not support it (have to check that, maybe I'm wrong, don't remember when exactly it was introduced), so such edit should possibly mention that. I haven't done it yet simply because there is more important stuff I have to do, like patches for upcoming 3.4 2016-06-17 18:45:53 .1 (band-aiding user:group mess my ccc056dbf9d3 introduced) 2016-06-17 18:51:00 I'll leave testing for the very last, because to check whether something works apparently you have to fix it first in some cases and by fix I mean fixing stuff unrelated to user:group mess. 2016-06-17 18:51:13 przemoc: the rule of thumb is AL >= 3.4 use $builddir? otherwise $_builddir 2016-06-17 18:51:41 peterj: fyi, I'm testing the compilation of llvm locally now before I push 2016-06-17 18:52:10 mjeanson: yes, + there is default prepare() now, as you noticed I changed this in your APKBUILDs 2016-06-17 18:53:10 barthalion: yeah, I was wondering because I used the template from the wiki to start those packages 2016-06-17 18:54:02 we have serious problem with documentation, but no one is eager to fix it 2016-06-17 18:54:15 it's not easy job, and boring for almost everyone 2016-06-17 18:56:56 barthalion: thanks! :) 2016-06-17 18:57:00 in case of APKBUILD documentation I believe it should be moved to be part of (not existent yet) README.md in abuild's repository, so whenever abuild improves something appropriate change in doc lands too. 2016-06-17 18:57:10 would it land in main with LLVM? 2016-06-17 18:58:10 peterj: yes 2016-06-17 18:58:31 Great! 2016-06-17 18:58:34 przemoc: I don’t think that README.md is a good place for APKBUILD documentation 2016-06-17 18:58:38 then we could simply link to http://git.alpinelinux.org/cgit/abuild/about/ from wiki (like we link to http://git.alpinelinux.org/cgit/awall/about/ for instance) 2016-06-17 18:59:02 przemoc: however, it would be helpful to add links to particular pages on alpine wiki to the readme 2016-06-17 19:00:02 I’d like to rewrite documentations from wiki to AsciiDoc (Asciidoctor), so we can generate e.g. man pages from it :) 2016-06-17 19:00:21 and also PDF, ebook, … and HTML of course 2016-06-17 19:00:41 maybe reStructuredText? equally powerful 2016-06-17 19:00:48 NO 2016-06-17 19:00:52 RST is totall mess 2016-06-17 19:01:06 it’s IMHO the worst syntax ever invented 2016-06-17 19:01:29 strong opinion 2016-06-17 19:01:38 yes, because i really hate it 2016-06-17 19:01:58 AsciiDoc is designed for writing even complex documents like books and not necessary just for technical writing 2016-06-17 19:02:19 for example documentation of git is written in AsciiDoc 2016-06-17 19:02:30 Linux kernel is currently moving to AsciiDoc 2016-06-17 19:02:44 Groovy, Activiti, some Spring projects uses AsciiDoc 2016-06-17 19:03:10 O’Reilly uses AsciiDoc in their publishing platform Atlas 2016-06-17 19:05:44 I haven't built any asciidoc docs lately, but I remember that generating man-pages in git from it was painfully slow 2016-06-17 19:06:00 that’s one of the reasons why Asciidoctor has been born 2016-06-17 19:06:12 you’re probably used the old asciidoc implementation 2016-06-17 19:06:16 that was really very slow 2016-06-17 19:06:34 Asciidoctor is very fast, even kernel devs was surprised how fast it is 2016-06-17 19:06:44 citation needed ;) 2016-06-17 19:06:55 give me a minute 2016-06-17 19:07:02 I'd read it too ;) 2016-06-17 19:07:37 it's in ruby, right? 2016-06-17 19:07:41 yes 2016-06-17 19:07:55 https://twitter.com/mojavelinux/status/691785986167197700 2016-06-17 19:07:55 https://twitter.com/mojavelinux/status/697929580254900224 2016-06-17 19:07:59 I would prefer it to be C/C++ project, but well 2016-06-17 19:08:06 I’ll try to find the conversation he’s referring in the tweet 2016-06-17 19:08:44 you really don’t want write such a thing in C ;) ask skarnet how he hates parsing even mere configs in C 2016-06-17 19:08:46 seems legit 2016-06-17 19:08:47 https://www.spinics.net/lists/linux-doc/msg33934.html 2016-06-17 19:09:59 the discussion is pretty long, I’ll ask Dan to send me link to the mail he referred in his tweet 2016-06-17 19:09:59 what does asciidoctor do better than regular asciidoc? 2016-06-17 19:10:04 or am I mixing things? 2016-06-17 19:10:14 where is the praise regarding speed? 2016-06-17 19:10:32 apparently speed 2016-06-17 19:10:55 it’s more modern implementation, syntax has been slightly improved, there are more convertors (e.g. convertor directly to PDF) 2016-06-17 19:11:26 syntax has been improved? I thought it was somewhat standardized already 2016-06-17 19:11:56 yes, it’s still backward compatible, but there are some new elements 2016-06-17 19:12:19 reST is a bit verbose, true, but you can write there everything. IIRC asciidoc wasn't as flexible 2016-06-17 19:12:49 jirutka1: the problem with parsing in C isn't C, it's parsing :P 2016-06-17 19:12:55 it’s not jsut verbose, it’s totally __nonsense`` 2016-06-17 19:13:18 you can write almost anything also in AsciiDoc, it’s very flexible 2016-06-17 19:13:34 but it's possibly true also that asciidoc's utter slowness drove me far from it rather early 2016-06-17 19:13:38 as I said, many big projects uses it for their documentation 2016-06-17 19:13:46 reST is also used 2016-06-17 19:14:04 so what are those? yet another rich text format? 2016-06-17 19:14:54 przemoc: please just open reST and Asciidoctor writing guide and honestly tell us what syntax is more reasonable and simpler to use :) 2016-06-17 19:15:00 plain-text formats expressing rich-text format intents 2016-06-17 19:15:05 skarnet: ^ 2016-06-17 19:16:06 how is asciidoc support nowadays for e.g. latext fragments? 2016-06-17 19:16:10 przemoc: there are a dozen of those coming up each decade, is asciidoc the 2016 flavour? 2016-06-17 19:16:20 you mean LaTeX? 2016-06-17 19:16:24 sorry 2016-06-17 19:16:26 typo 2016-06-17 19:16:27 yes 2016-06-17 19:16:29 LaTeX 2016-06-17 19:16:36 skarnet: haha 2016-06-17 19:16:48 skarnet: no, asciidoc is a bit older 2016-06-17 19:17:08 no problem, http://asciidoctor.org/docs/user-manual/#activating-stem-support 2016-06-17 19:17:21 but apparently asciidoctor, new implementation, is more recent and allowing to "compile docs" in reasonable time 2016-06-17 19:17:36 we’re using it a lot for writing documentation that includes math 2016-06-17 19:18:00 ASCII Doctor! everyone loves it! :D 2016-06-17 19:18:15 jk, i just opened http://www.sloganizer.net/en/ and it made it up for me :P 2016-06-17 19:18:19 nroff wasn't good enough, SGML wasn't good enough, DocBook wasn't good enough, Texinfo wasn't good enough, wiki wasn't good enough, and now it looks like Markdown isn't good enough 2016-06-17 19:18:25 what makes you think asciidoc will be good enough? 2016-06-17 19:18:59 AsciiDoc is based on DocBook semantics… and you really don’t wanna write documents in XML, do you? ;) 2016-06-17 19:19:23 you’re also mixing different things… wiki isn’t syntax, wiki is… eh… wiki… 2016-06-17 19:19:44 nobody does, but nobody wants to write TeX either 2016-06-17 19:20:03 Markdown is good for very simple formatting like blog comments etc., it’s unusable for anything more complex; trust me, I know Markdown very well 2016-06-17 19:20:03 replace wiki with whatever language is used to write wikis 2016-06-17 19:20:19 TeX is acceptable 2016-06-17 19:20:35 not to GNU folks 2016-06-17 19:20:47 only Ph.D. students crave for LaTeX 2016-06-17 19:21:00 skarnet: because I’m very interested in this topic, know these syntaxes, limitations etc. 2016-06-17 19:21:28 texinfo is the most dreadful thing, though 2016-06-17 19:21:32 I hate it so much 2016-06-17 19:21:38 TeX is excellent typesetting tool, not a syntax for writing documents by normal humans… 2016-06-17 19:22:05 those tools always have the same common limitation, which is human laziness 2016-06-17 19:22:11 and that one will be the same no matter the tool 2016-06-17 19:22:18 it will *always* be a pain to format text 2016-06-17 19:22:37 of course, AsciiDoc will not write document for you… ;) 2016-06-17 19:22:46 what's the secret to adding external URLs to a wiki page? it's blocking me, I suppose it thinks I'm a spammer 2016-06-17 19:23:05 it won't!! w00t.. 2016-06-17 19:23:08 it just helps you to format it without going crazy from Word or other insane WYSIWY(N)G editors 2016-06-17 19:23:47 algibot: I won? :) 2016-06-17 19:23:52 yay! 2016-06-17 19:24:55 you can _sugarcoat_ it \bold{however} the fuck you .I want\n, expressing
rich text
in plain text will *always* be painful 2016-06-17 19:25:40 what I also like about TeX is that it concentrates lot of great people. e.g. pgf/TikZ are super drawing tools within LaTeX, but it also pains me that there are within LaTeX, i.e. not standalone utilities, which would make them even greater... 2016-06-17 19:26:16 you may be interested in https://github.com/asciidoctor/asciidoctor-latex 2016-06-17 19:26:43 however, I should warn you that the implementation is… well… quite… horrible… James it mathematician, he’s not good programmer 2016-06-17 19:27:29 I *think*, standard markdown is the answer to `all` the formatting issues 2016-06-17 19:27:42 but if you just want to write math in AsciiDoc, it’s supported by default 2016-06-17 19:27:51 peterj: you mean CommonMark? 2016-06-17 19:27:54 markdown wasn't good enough from the very beginning. original spec was very loose. commonmark attempts to fix that, but there is not much you can do to not break markdown completely. 2016-06-17 19:28:04 how do you create a table? or maybe something more simple, footnote? 2016-06-17 19:28:34 well, markdown couldn't even do right *this* _simple_ thing 2016-06-17 19:28:40 it’s great for very simple formatting, not for writing more complex texts 2016-06-17 19:29:05 jirutka1: that's the thing, with markdown you don't do too much of a fancy stuff at all! :D 2016-06-17 19:29:29 everyone stick to plain and simple with little backticks and such.. 2016-06-17 19:29:46 I mean asterisk was used for bold and underscore for underscoring for ages, so why go with double ** or __? insane 2016-06-17 19:29:47 that said, I wrote my master thesis in Markdown… using Pandoc to translate it into LaTeX and then generate PDF… 2016-06-17 19:30:00 sounds awfully awful :) 2016-06-17 19:30:14 but I wan’t using “Markdown”, whatever this means, but Pandoc dialect of Markdown with a lot of hacks… 2016-06-17 19:30:30 but if it worked for you, then well, it worked 2016-06-17 19:30:36 it was actually more comfortable than wraiting it directly in (La)TeX, but yeah… 2016-06-17 19:30:59 correction: Xe(La)TeX… i don’t use (La)TeX, it doesn’t support even Unicode 2016-06-17 19:31:00 jirutka1: obligatory https://xkcd.com/763/ 2016-06-17 19:31:09 I wrote once LaTeX rant: http://abyss.przemoc.net/post/28208254393/getting-things-done-in-latex-or-not 2016-06-17 19:31:11 nice with GH markdown, `` you can `do` this to keep the do enclosed in backtick `` 2016-06-17 19:31:47 I’ll read it later 2016-06-17 19:33:39 let the windows narrator read it :D 2016-06-17 19:34:20 I’ve never finished my conversion scripts and improved XeLaTeX template, but I gave it to my colleague and he published it on GitHub, so if you want some nightmares, take look at these insane sed regexp we wrote XD https://github.com/hroncok/diplomka/blob/master/bin/convert#L13-L157 2016-06-17 19:35:00 some of these are mine… filter_hotfix_1 is definitely his, this is too much even for me XD 2016-06-17 19:35:18 I'd really love metrics about the ratio of time used to make tools supposed to help you / time used to actually do productive things 2016-06-17 19:35:29 yeah 2016-06-17 19:35:35 functors 2016-06-17 19:35:42 not even counting the time spent TALKING about those wonderful tools, including on IRC 2016-06-17 19:35:50 :D 2016-06-17 19:36:01 ^ you’ve killed me XD 2016-06-17 19:36:19 yeah, I’ve spent a lot of time talking about it, just to avoid actually writing my thesis XD 2016-06-17 19:37:16 przemoc: the full userspace tracing test suite runs just fine on the grsec kernel 2016-06-17 19:37:25 great! 2016-06-17 19:38:17 I took 5 minutes to write this: https://wiki.alpinelinux.org/wiki/LTTng 2016-06-17 19:38:37 thank you from all Alpine Linux users 2016-06-17 19:38:47 the wiki won't let me add http URLs, I've just added spaces in the links for now 2016-06-17 19:38:53 barthalion: mjeanson: great to see that lttng-ust is upgraded and without the patch! Thank you so much for your efforts <3 2016-06-17 19:40:20 mjeanson: I just used lttng-ust-dev to build coreclr two days ago, is -tools package necessary now? 2016-06-17 19:40:25 peterj: the changes to lttng-tools were more extensive, it'll take a little while to get it upstreamed 2016-06-17 19:40:43 peterj: just if you want to do some tracing 2016-06-17 19:41:35 coreclr shouldn't depend on it, only ust 2016-06-17 19:42:27 at least kprobes & co should be working on grsec 2016-06-17 19:42:57 they're enabled in -grsec 2016-06-17 19:42:58 mjeanson: i haven't run their test suites now (just successfully compiled it), its like they have one very lightweright version and the other suite is over 2GBs. Once lldb is in, i will create a new VM start building from scratch and run end-to-end testing on Alpine. :) 2016-06-17 19:45:10 mjeanson: thanks, I’ve fixed the links for you; https://wiki.alpinelinux.org/wiki/LTTng 2016-06-17 19:45:31 the current modules won't build without CONFIG_TRACEPOINT enabled, maybe they could be reworked so only the kprobes part can be built, but I'm not sure it's something we would like to support 2016-06-17 19:46:18 no, I mean lttng was about tracepoint from the beginning, I just thought it could work with kprobe partially 2016-06-17 19:47:12 samsung guys ran the test on some arm32 and it took about 25 hours to only run the tests (https://github.com/dotnet/coreclr/issues/3977) good luck to me on alpine-x64 i suppose .. :/ 2016-06-17 19:49:12 przemoc: there is a subset of lttng-modules that uses kprobes, but it came after tracepoints and I think the build system could be tweaked so you can only build this part if you don't have tracepoints. I'm just not sure how useful it would be 2016-06-17 19:49:47 maybe we could add tracepoints to config-grsec. we'll have to check whether it works and maybe convince later ncopa 2016-06-17 19:50:20 jirutka1: thanks, will I be able to edit the page again with the links in it? 2016-06-17 19:50:32 mjeanson: I hope so 2016-06-17 19:50:58 mjeanson: to be honest, I have no clue why it didn’t work for you 2016-06-17 19:51:34 mjeanson: but I don’t know MediaWiki well, maybe there’s some protection that you must first edit few pages to be able to add links or something like that 2016-06-17 19:51:57 wiki.a.o gets many spammers, so filters may be quite restrictive 2016-06-17 19:52:09 s/MediaWiki/WikiMedia ? :D 2016-06-17 19:52:20 jirutka1: it really looked like some kind of anti-spamming system 2016-06-17 19:52:48 I think that it’s MediaWiki https://www.mediawiki.org/wiki/MediaWiki 2016-06-17 19:53:14 jirutka1: you mentioned mruby recently. does it allow easy building of "native" apps, like e.g. asciidoctor? being able to use it without obtaining ruby runtime would be superhandy. 2016-06-17 19:53:38 ops, I confused it with https://wikimedia.org/ 2016-06-17 19:54:04 przemoc: config_tracepoint conflicts with at least one grsec config option, it think it was /dev/mem filtering 2016-06-17 19:54:36 mjeanson: oh, I see, then it's rather out of question 2016-06-17 19:55:03 that's what I tought 2016-06-17 19:55:22 przemoc: yes, mruby is designed for building single zero-deps binary; it’s very interesting idea to use it for Asciidoctor, I’ll try it 2016-06-17 19:56:12 by zero-deps you mean libc-only or completely static? 2016-06-17 19:56:19 libc-only 2016-06-17 19:56:37 https://pkgs.alpinelinux.org/package/edge/testing/x86_64/mruby 2016-06-17 19:56:40 I made an aport to build lttng-modules against the vanilla kernel, would that have any chance to be accepted? 2016-06-17 19:58:13 mruby package is just mruby itself (obviously), its binary is also static; when you build your application with mruby, the result is the same 2016-06-17 19:59:58 you'll have to convince ncopa. you could send mail to alpine-devel showing awesome and useful things that can be done with these modules - that should ease the convincing task. ;) 2016-06-17 20:04:24 barthalion: could you please merge https://github.com/alpinelinux/aports/pull/127 ? it’s into main, I don’t have rights to the main repository 2016-06-17 20:19:44 mjeanson: ncopa is your man, I don't feel brave enough to push it and lose any trust I gained 2016-06-17 20:20:47 jirutka: tomorrow, I want to finish lldb today 2016-06-17 20:20:54 okay 2016-06-17 21:14:46 is it possible to specify depends for a subpackage? 2016-06-17 21:14:52 of course 2016-06-17 21:15:05 declare depends= in the split function 2016-06-17 21:15:24 or if it’s dev package, then just use depends_dev= on top level 2016-06-17 21:16:46 jirutka: thanks 2016-06-17 21:16:53 you’re welcome 2016-06-17 21:17:38 I really don’t understand why people use Ubuntu on server; I’m serious, I don’t get it; this system is total mess, everything I touch is broken 2016-06-17 21:18:18 the thing is, everything is broken 2016-06-17 21:18:26 it's just matter of getting used to 2016-06-17 21:18:35 no, Alpine and Gentoo works pretty well 2016-06-17 21:18:51 but on Ubuntu even basic things are broken 2016-06-17 21:18:53 both have their annoyances 2016-06-17 21:19:07 But Ubuntu has Long Term Support!!1 2016-06-17 21:19:21 srsly, I’m not even able to restart one fucking service! 2016-06-17 21:19:29 upstart simple do nothing 2016-06-17 21:19:29 sounds like PEBKAC 2016-06-17 21:19:53 slackware was ok back in the day too 2016-06-17 21:20:07 maybe I’m typing command `restart` with wrong fingers or what can I do wrong? 2016-06-17 21:20:24 oh I hear trolling around 2016-06-17 21:20:34 (I wrote "back in the day" only because I haven't used any recent version) 2016-06-17 21:21:10 coredumb: your hearing is flawless, sir 2016-06-17 21:21:12 jirutka: did you turn it off and on again? 2016-06-17 21:21:44 \o/ 2016-06-17 21:21:54 no, I should not restart this system 2016-06-17 21:21:59 so, while we're trolling, why don't we switch topic to docker or systemd? two favorite things to cry about here recently 2016-06-17 21:22:10 I need very very simple thing… just restart dnsmasq to load new configuration… 2016-06-17 21:22:23 oh barthalion 2016-06-17 21:22:31 but dnsmasq is started with lxc-net 2016-06-17 21:22:47 did I hear PEBKAC? 2016-06-17 21:22:49 when i stop lxc-net, nothing happen 2016-06-17 21:22:56 I'd have so much things to say about systemd 2016-06-17 21:23:05 no, this is pre-systemd version 2016-06-17 21:23:20 or is it just blaming Ubuntu for poor lxc UX? 2016-06-17 21:23:23 I'm not so sure! 2016-06-17 21:23:37 I use LXC on Gentoo without problem 2016-06-17 21:23:41 cool 2016-06-17 21:23:45 I was wondering if it was common to have openrc deps in the conf.d file ? 2016-06-17 21:23:46 but LXC scripts on Ubuntu are totally broken 2016-06-17 21:23:47 I don't use Gentoo without problem 2016-06-17 21:23:57 I use Gentoo for many years 2016-06-17 21:23:58 like I'd want my nginx start _after_ php-fpm 2016-06-17 21:24:00 ACTION goes away 2016-06-17 21:24:02 good night 2016-06-17 21:24:16 but I'd prefer not modifying the init script 2016-06-17 21:24:28 at least not everytime there's an upgrade ^^ 2016-06-17 21:24:35 so I'd send a patch 2016-06-17 21:24:39 but is it ? 2016-06-17 21:24:43 coredump: what distro you’re talking about? that nginx thing 2016-06-17 21:25:02 are we not on #alpine-devel ? 2016-06-17 21:25:05 okay 2016-06-17 21:25:13 so you don’t need to modify init script 2016-06-17 21:25:21 just define rc_depend in the /etc/conf.d 2016-06-17 21:25:31 ah brilliant 2016-06-17 21:25:38 not like systemd 2016-06-17 21:25:41 \o/ 2016-06-17 21:25:53 pardon, rc_after in your case 2016-06-17 21:26:00 you can put some stuff in /etc/rc.conf too, e.g. rc_dropbear_need="!net" which is useful on lxc containers 2016-06-17 21:26:02 read /etc/rc.conf for examples 2016-06-17 21:26:24 yes, but rc.conf is for global config 2016-06-17 21:26:56 if you wanna just configure nginx to start after php-fpm, then the right place to do it si in /etc/conf.d/nginx : rc_after="php-fpm" 2016-06-17 21:27:09 perfect 2016-06-17 21:27:30 exactly… and Ubuntu is exactly opposite! 2016-06-17 21:29:31 yeah 2016-06-17 21:29:43 what's the point even trying to use ubuntu 2016-06-17 21:29:46 ? 2016-06-17 21:30:02 unotrightinurmindorwhat? 2016-06-17 21:30:18 I’m not trying it… it’s a legacy server installed by my former coworker 2016-06-17 21:30:18 please have some decency 2016-06-17 21:30:30 I just need to install openvpn client here and setup dns 2016-06-17 21:30:53 then just wipe it and find the guy back and slap him in the face when he leaves his appartment 2016-06-17 21:31:13 I can’t do the first part 2016-06-17 21:31:22 but I’m really thinking about the second part… 2016-06-17 21:31:42 that's usually what I do 2016-06-17 21:32:28 devs like: "hell yeah we need you to configure the server so that we can use docker in full production, you know like the bosses we are" 2016-06-17 21:32:37 so I slap them hard in the face 2016-06-17 21:32:52 coredump: I like you! XD 2016-06-17 21:33:07 then they like: "pfff OK we gonna do it ourselves" 2016-06-17 21:33:16 "suits you kind sir" 2016-06-17 21:33:29 2 weeks after they get back crawling for help 2016-06-17 21:33:49 jirutka: you like me but are unable to use your tab key 2016-06-17 21:33:51 :D 2016-06-17 21:34:00 see it's dumb ... not dump 2016-06-17 21:34:02 :D 2016-06-17 21:34:03 or: "Hey, install Ubuntu on this server for us, we know it and can administrate it by ourselvs”… It’s a TRAP! 2016-06-17 21:34:18 nah they stopped doing that 2016-06-17 21:34:26 cause I spit on them if they do 2016-06-17 21:34:29 holly shit 2016-06-17 21:34:31 they don't like that :D 2016-06-17 21:34:36 OMFG! 2016-06-17 21:34:38 I’m idiot! 2016-06-17 21:34:51 I’ve tried to restart LXC… 2016-06-17 21:35:00 and forgot that I’m connected through VPN that runs on it! 2016-06-17 21:35:06 my boss is gonna kill me 2016-06-17 21:35:08 fuck 2016-06-17 21:35:09 FUCK 2016-06-17 21:35:14 \o/ 2016-06-17 21:38:59 jirutka: that's cool 2016-06-17 21:39:06 last time I messed up 2016-06-17 21:39:18 I was ready to do sepuku 2016-06-17 21:39:31 boss is cool, I‘m just talking with him 2016-06-17 21:39:33 my boss was like "meh who cares... shit happens" 2016-06-17 21:39:55 then it's my business customer's team lead that took the blame 2016-06-17 21:39:57 \o/ 2016-06-17 21:39:59 lol 2016-06-17 21:40:10 I "love" my company 2016-06-17 22:59:43 barthalion: any luck building llvm with lldb? :) 2016-06-18 03:14:20 jirutka you around? 2016-06-18 09:08:17 Hello jirutka: 2016-06-18 09:14:52 hi there! I filed one more PR https://github.com/alpinelinux/aports/pull/130 2016-06-18 09:15:07 does it makes sense to rename package? 2016-06-18 09:25:53 peterj: not yet, trying to fix cmake 2016-06-18 09:26:24 andypost: it does to me 2016-06-18 09:28:16 barthalion, so here's another pr https://github.com/alpinelinux/aports/pull/131 2016-06-18 09:45:34 barthalion: is there some patch required for lldb cmake script? 2016-06-18 09:55:46 peterj: possibly not, I didn't work much with cmake; testing the build now before commiting 2016-06-18 09:56:09 peterj: after all it will be separate package, first in testing and when you confirm it works, I'll move it to main 2016-06-18 10:06:05 barthalion: i thought lldb will be part of llvm package (like llvm-ar and other tools that live in llvm/tools dir). No? 2016-06-18 10:20:45 peterj: lldb is shipped in different tarball, and llvm maintainer also built clang separately (in opposite to Arch which builds everything from llvm) 2016-06-18 10:20:55 peterj: I guess you haven't seen this build error? https://paste.xinu.at/dE3zr/ 2016-06-18 10:31:50 barthalion: i did, it was fixed with my patch https://gist.github.com/jasonwilliams200OK/c84bdd51ede5390b30c7dadcbcad1665 2016-06-18 10:38:02 weird, I applied it 2016-06-18 10:38:29 oh, nvm 2016-06-18 10:39:47 so it should be done in 20 minutes 2016-06-18 10:40:16 andypost: I’m quite confused, is https://github.com/alpinelinux/aports/pull/130 still relevant when there’s https://github.com/alpinelinux/aports/pull/131 ? 2016-06-18 10:42:46 jirutka, I think 130 should be superseded by 131 as above barthalion said 2016-06-18 10:43:05 andypost: so 130 can be closed? 2016-06-18 10:43:13 yep 2016-06-18 10:43:44 I just don't know procedure - it's version update + fix 2016-06-18 10:44:52 jirutka, personally I think better to rename and update version in 131 2016-06-18 10:47:54 jirutka, what is a proper commit message "abuild rename"? 2016-06-18 13:40:30 I have added basic docs api-json.md, usage.md in docs folder, https://github.com/insteps/aport-api 2016-06-18 13:42:40 fcolista: ^^^ 2016-06-18 13:50:13 follow rel link for more data, pls let me know if you expect some uri or data currently not there or shows incorrectly 2016-06-18 14:34:03 barthalion: any luck with cmake issue? 2016-06-18 14:54:09 Hello 2016-06-18 20:16:53 peterj: http://git.alpinelinux.org/cgit/aports/commit/?id=f697997042cf98dcfedae1fad22b0c669b5044c9 2016-06-18 20:17:52 barthalion: shouldn’t every new package has a maintainer? 2016-06-18 20:20:42 jirutka: even if, who you gonna call? 2016-06-18 20:21:02 ? 2016-06-18 20:21:02 it's just testing 2016-06-18 20:21:21 and there is no rule saying "you shall set maintainer' 2016-06-18 20:22:29 so what did you want me to merge? 2016-06-18 20:23:24 this https://github.com/alpinelinux/aports/pull/127, but change the commit message prior merging 2016-06-18 20:23:39 it should be: main/php5-apcu: upgrade to 4.0.11 2016-06-18 20:26:50 barthalion: great news! and thank you so much! :) 2016-06-18 20:27:19 does every package build by aports build as compiled binary? 2016-06-18 20:27:43 or do they get compiled from source per install (like happens in pkgsrc) ? 2016-06-18 20:27:44 peterj: yes – you can just do 'apk add -U lldb' 2016-06-18 20:27:59 should be on all mirrors in two hours 2016-06-18 20:28:48 looks like I shouldn't have separated python bindings 2016-06-18 20:31:53 barthalion: I will build CoreCLR with lldb now. Incidentally, when lldb is verified with CoreCLR, can it be promoted to main repo.? 2016-06-18 20:32:33 peterj: sure thing, I can't imagine better validation than by using different project 2016-06-18 20:34:47 Yup and now i have over compiled the whole test suite for dotnet core, its `4.08 GB (4,388,339,712 bytes)` in size, i will run it over night on Alpine to validate the whole framework :D 2016-06-18 21:00:37 barthalion: (minor) can the project URL be set to http://lldb.llvm.org/ instead of http://llvm.org/? 2016-06-18 21:01:21 https://pkgs.alpinelinux.org/package/edge/testing/x86_64/lldb 2016-06-18 21:16:20 yeah, I'll change it 2016-06-18 21:18:06 jirutka: I don't understand why should we ask contributors to squash commits 2016-06-18 21:18:33 barthalion: https://github.com/alpinelinux/aports/pull/136/commits 2016-06-18 21:18:46 you don't get it 2016-06-18 21:18:52 it takes 20s to do it locally 2016-06-18 21:18:55 the commit message: “Fix patch to apply without offset” is against our rules 2016-06-18 21:18:58 leaves better impression that you care 2016-06-18 21:19:23 I’m doing it myself when merging commits 2016-06-18 21:19:24 for the same reason we don't reject all patches from mailing list that doesn't match some unwritten rules we follow 2016-06-18 21:19:34 okay, then nevermind 2016-06-18 21:19:50 136 is ready to merge? 2016-06-18 21:20:01 but since I’m not the one who will be merging, then I’ve asked him to squash it so it’ll be ready 2016-06-18 21:20:14 I think so 2016-06-18 21:20:31 and Andy just merged it ;) 2016-06-18 21:20:36 it doesn't make any difference to me, when I review patches from ML I also have to build these locally 2016-06-18 21:20:50 so squashing is relatively easy and quick task 2016-06-18 21:20:54 okay 2016-06-18 21:21:10 I'll wait for travis and merge it 2016-06-18 21:21:48 I know, as I said, when I’m the one who will be merging the PR, then I don’t ask for squashing and other trivial things, just do it myself before pushing 2016-06-18 21:22:04 https://github.com/alpinelinux/aports/pull/96 what about this? :P 2016-06-18 21:22:29 I dunno, ncopa wasn’t sure if we want this in the offical repo, 2.1 is old version 2016-06-18 21:22:43 no sec upgrades? 2016-06-18 21:22:55 no, this is a new abuild 2016-06-18 21:23:09 I know, I ask about upstream 2016-06-18 21:23:14 is it EOL'ed? 2016-06-18 21:23:17 we have ruby2.2 and ruby (that is 2.3) in the repo 2016-06-18 21:23:42 if it still receives security fixes, I'm fine with having this in testing/community 2016-06-18 21:23:44 hmm, not yet 2016-06-18 21:24:02 okay, then I’m gonna merge it 2016-06-18 21:26:58 btw php5-pear-auth_sasl2 is failing… 2016-06-18 21:28:08 just on arm 2016-06-18 21:28:24 please read the log http://build.alpinelinux.org/buildlogs/build-edge-armhf/main/php5-pear-auth_sasl2/php5-pear-auth_sasl2-0.1.0-r0.log 2016-06-18 21:28:29 it’s failing because of bad checksum 2016-06-18 21:28:31 looks like tarball is corrupted 2016-06-18 21:28:34 yeah 2016-06-18 21:28:42 nothing I can do, sorry 2016-06-18 21:28:46 it's fabled's turf 2016-06-18 21:29:07 the problem is that the tarball’s name is not unique 2016-06-18 21:29:12 IMHO 2016-06-18 21:29:16 hm, true 2016-06-18 21:29:26 it's main? 2016-06-18 21:29:39 yes 2016-06-18 21:29:43 people should care about this 2016-06-18 21:30:06 I’ve fixed the same problem today in other php package 2016-06-18 21:31:45 it's not naming clash though 2016-06-18 21:31:54 both packages are using same tarball 2016-06-18 21:31:57 hmm, so upstream has modified tarbal? 2016-06-18 21:32:00 (so it's kinda okay) 2016-06-18 21:32:07 yes, apparently 2016-06-18 21:32:41 anyway, Auth_SASL2 may clash with other package quite easy 2016-06-18 21:33:22 looks like we could remove the testing one 2016-06-18 21:34:29 btw https://github.com/alpinelinux/aports/pull/136 is ready for merge… 2016-06-18 21:36:48 jirutka: travis didn't catch bad checksum 2016-06-18 21:37:14 that’s weird 2016-06-18 21:38:28 how is it possbile? 2016-06-18 21:38:48 it had to use gd-iconv patch from first commit apparently 2016-06-18 21:39:18 that’s not possible 2016-06-18 21:39:22 each travis build is isolated 2016-06-18 21:40:23 it’s not checking checksums at all 2016-06-18 21:41:10 it looks like a bug in abuild 2016-06-18 21:43:04 hm, abuild skips checksums check if run with -f 2016-06-18 21:43:21 -f Force specified cmd, even if they are already done 2016-06-18 21:45:25 I’m not sure why I’ve used -f in the travis script 2016-06-18 21:56:52 I didn't even know about this switch :D 2016-06-18 21:57:14 abuild -h is your friend ;) :P 2016-06-18 22:02:35 I never cared, -r is all I need 2016-06-18 22:02:47 -i is also very useful 2016-06-18 22:11:02 it's kind of misleading, at least for testing, because depending on what's in .pre-install script, you won't get exactly same state as if you installed pkg in the host that was not abuilding this package before. but it's more a problem of abuild not really bringing status quo after build 2016-06-18 22:16:18 jirutka: for repo mantainers, there is a new feature in GitHub when merging the PR, see https://github.com/blog/2141-squash-your-commits 2016-06-18 22:16:47 I know about it, but we can’t merge directly via GitHub 2016-06-18 22:17:01 GitHub is just a mirror of http://git.alpinelinux.org/cgit/aports/ 2016-06-18 23:08:59 is there any difference in sscanf formatters on Alpine? 2016-06-18 23:09:26 this test is failing: https://github.com/dotnet/coreclr/blob/775003a/src/pal/tests/palsuite/c_runtime/sscanf/test15/test15.c#L33 2016-06-18 23:10:17 (that DoFloatTest function is here: https://github.com/dotnet/coreclr/blob/775003a4c72f0acc37eab84628fcef541533ba4e/src/pal/tests/palsuite/c_runtime/sscanf/sscanf.h#L211-L243) 2016-06-18 23:19:07 ERROR: scanned float incorrectly from "1234567890.0123456789E" using "%E". Expected "1234567936.000000", got "0.000000". 2016-06-19 12:32:24 ncopa: why did you change kdb's datadir? This breaks third party font packages like terminus-font when using kdb 2016-06-19 12:32:55 peterj: about that scanf error, #musl is probably the best place for this kind of questions 2016-06-19 12:35:13 barthalion: already on it, made this steps to repro https://ideone.com/ReFzeO (see code comments) 2016-06-19 13:17:25 bathalion: sscanf is investigated, that is a bug in glibc, and musl is conforming with POSIX/ISO std. I have updated the issue. 2016-06-19 16:33:16 barthalion: for the VirtualAlloc and VirtualProtect failures, if i paxmark those test binaries, the tests pass. For example, i first ran paxctl -c bin/obj/Linux.x64.Debug/src/pal/tests/palsuite/filemapping_memmgt/VirtualAlloc/test5/paltest_virtualalloc_test5 2016-06-19 16:33:31 then ran paxctl -psm 2016-06-19 16:33:38 then ran the test and it passes this one. 2016-06-19 16:33:55 same goes for all the six failing virtualallloc tests 2016-06-19 16:34:36 this feels like circumventing the mem-protection of Alpine Linux platform. 2016-06-19 16:36:22 is there a more genuine way to achieve it to make it legit? or is paxmarking the package something a common trait for programming frameworks with just-in-time compilation (i.e. execute and run privilege) 2016-06-19 16:43:49 yes, you need to paxmark anything that uses JIT 2016-06-19 16:44:37 this is circumventing memory protection, but you can't eat cake and have cake 2016-06-19 18:15:45 barthalion: isn't eating the cake having it ? 2016-06-19 18:16:04 ^ +1 2016-06-19 18:19:03 coredumb: I presume its like eating and keeping it simultaneously.. you gotta lose some to gain some 2016-06-19 18:19:28 when you eat the cake, you still have the cake, in your stomach… ;) 2016-06-19 18:21:04 "The body is the best place to keep your cake." (with apologies to Frank Herbert) 2016-06-19 18:26:25 :P 2016-06-19 18:26:26 (cowsay) 2016-06-19 18:26:57 cowsay $(fortune) :)) 2016-06-19 19:01:17 just bake two cakes 2016-06-19 19:05:36 the cake is a lie! don't eat.. said the spirit 2016-06-19 19:08:58 https://youtu.be/36reZ9-3VK0 2016-06-19 20:05:17 is there any reason for the default syslog daemon to always log in UTC ? 2016-06-19 20:21:56 musl's implementation choice 2016-06-19 20:22:20 better than localtime: the logs are consistent across timezones 2016-06-19 20:22:38 it's the client syslog() that creates the timestamp, not the syslogd server 2016-06-19 20:22:56 (the standard says so) 2016-06-19 20:23:56 ok 2016-06-19 20:23:57 anyway, syslog() is horrible so who cares what it does :P 2016-06-19 20:23:58 localtime may have DST. every time I see a system storing timestamp using local time I want to murder the one behind it. always use UTC for storing and show local time in presentation layer, if needed 2016-06-19 20:24:22 przemoc: even better, use TAI ;) 2016-06-19 20:35:51 TAI? 2016-06-19 20:38:02 maybe http://www.timeanddate.com/time/international-atomic-time.html ? 2016-06-19 20:59:44 barthalion: do you know of any difference in mutex handling on Alpine compared to other Unices? 2016-06-19 21:00:15 this mutext stress test https://github.com/dotnet/coreclr/blob/4c474ea/src/pal/tests/palsuite/threading/NamedMutex/test1/namedmutex.cpp is failing on Alpine with: 'paltest_namedmutex_test1' failed at line 703. Expression: WaitForSingleObject(m, FailTimeoutMilliseconds) == WAIT_ABANDONED_0 'paltest_namedmutex_test1' failed at line 875. Expression: AbandonTests_Parent() 2016-06-19 21:01:18 this is the last test from PAL suite failing that probably needs to be fixed, before I can run the managed .NET assemblies test suite. 2016-06-19 22:27:47 added advanced search to api.a.o 2016-06-19 22:27:52 eg. curl -X POST 'http://api.alpinelinux.org/search/packages' -d '{"name":"q_","category":"edge:main:x86","flagged":"yes","maintainer":"Nat_"}' | jq . | less 2016-06-19 22:37:23 why you’ve used POST body for search? 2016-06-19 22:37:37 it doesn’t even support logical operators, so you can just use URI parameters 2016-06-19 22:37:44 …query string 2016-06-19 23:42:14 now add same variables in GET request too 2016-06-19 23:42:28 added* 2016-06-19 23:43:58 would wait to see if things are simple, then remove POST method 2016-06-19 23:45:05 remember the consumer app may/maynot be a web interface 2016-06-19 23:48:58 it doesn’t matter 2016-06-19 23:50:18 get eg. curl 'https://api.alpinelinux.org/search/packages/category/v3.4:all:x86/name/q_/flagged/no/maintainer/Nat_' | jq . | less 2016-06-19 23:50:33 wtf is this?! 2016-06-19 23:51:07 ? 2016-06-19 23:51:50 query string, not path… http://api.alpinelinux.org/search/packages?name=q_&category=edge:main:x86&flagged=yes&maintainer=Nat_ 2016-06-19 23:52:28 sry, path is wrong 2016-06-19 23:52:40 http://api.alpinelinux.org/packages?name=q_&category=edge:main:x86&flagged=yes&maintainer=Nat_ 2016-06-19 23:52:49 and this curl 'https://api.alpinelinux.org?_url=/search/packages/category/edge:main:x86/name/q_/flagged/yes/maintainer/Nat_' | jq . | less 2016-06-19 23:52:52 ?? 2016-06-19 23:53:08 this URL doesn’t make much sense for me 2016-06-19 23:53:33 the part: flagged/yes/maintainer/Nat_ doesn’t make any sense 2016-06-19 23:53:49 its key/value pair 2016-06-19 23:54:25 wait a moment 2016-06-19 23:54:39 so https://api.alpinelinux.org?_url=... is acceptable ? 2016-06-19 23:54:48 this is really a query string? _url=/search/packages/category/edge:main:x86/name/q_/flagged/yes/maintainer/Nat_ 2016-06-19 23:54:50 why? 2016-06-19 23:55:05 yes 2016-06-19 23:55:14 omg 2016-06-19 23:55:32 dont't just curse, rationale ?? 2016-06-19 23:55:38 it’s just totally wrong 2016-06-19 23:55:53 or quote some discussion/site 2016-06-19 23:56:10 well, read any tutorial about RESTful APIs 2016-06-19 23:56:19 site url pls 2016-06-19 23:57:14 I’ll check my thesis, I’ve referenced some articles here 2016-06-19 23:57:21 ok 2016-06-19 23:57:55 http://apigee.com/about/resources/ebooks/web-api-design 2016-06-19 23:58:09 Mark Masse. REST API Design Rulebook. O’Reilly Media, Inc., 2011, s. 1– 94. : 9781449310509 2016-06-19 23:58:50 thanks, onto it 2016-06-19 23:59:38 https://www.infoq.com/articles/webber-rest-workflow 2016-06-19 23:59:40 that's why I been poking around for URI suggestion 2016-06-20 00:00:22 I thought that I have more links in bookmarks :/ 2016-06-20 00:01:55 what does this mean? "name": "q_" – is it a filter for packages with name "q_"? 2016-06-20 00:02:12 by any chance do you have the ebook lying around to share ? 2016-06-20 00:02:32 what ebook? REST API Design Rulesbook? 2016-06-20 00:02:46 hi, i've an issue with cross compile alpine kernel for arm.. i download vanilla kernel v 4.4.11, apply the kernel patch for vanilla kernel, apply the grsecurity patch for this kernel from alpine and when i try to compile default sunxi_defconfig with grsecurity i get this error - kernel/module.c:2564:27: error: ‘struct module’ has no member named ‘core_size’ 2016-06-20 00:02:46 info->symoffs = ALIGN(mod->core_size, symsect->sh_addralign ?: 1); 2016-06-20 00:02:46 .. any idea how to solve this? (i use fedora and as compiler i use gcc-arm-linu-gnu v 5.1.11) 2016-06-20 00:03:51 yes 2016-06-20 00:05:27 q_ => name=q and _ at prefix/suffix is used for wildcard 2016-06-20 00:06:11 asterisk is more common char for wildcard 2016-06-20 00:06:12 q_ gets converted to "name LIKE 'q_%'" 2016-06-20 00:06:38 yes I reverted that, _ more URL friendly 2016-06-20 00:07:01 asterisk is not a reserved character in query string, so it should be okay 2016-06-20 00:07:28 wget URL/a/b* gets saved to file named b* 2016-06-20 00:07:33 kinda not very nice 2016-06-20 00:08:51 let me know when you find that apigee ebook 2016-06-20 00:08:58 give me a sec 2016-06-20 00:14:19 vkris: I’ve sent you direct message 2016-06-20 00:14:41 ok , thanks 2016-06-20 00:14:47 already downloaded 2016-06-20 07:09:29 ncopa: why did you change kdb's datadir? This breaks third party font packages like terminus-font when using kdbs setfont utility 2016-06-20 07:10:14 it was a mistake 2016-06-20 07:11:38 in that case: would you mind if I revert that change? 2016-06-20 07:12:01 i will 2016-06-20 07:12:19 i started on a bigger refactor 2016-06-20 07:12:24 ah, alright 2016-06-20 07:12:27 so i need clean it up a bit 2016-06-20 07:12:34 before i can revert 2016-06-20 07:12:50 i kinda thought i already did revert it 2016-06-20 07:12:54 but apparetnly i didnt 2016-06-20 07:12:56 sorry 2016-06-20 07:13:13 no problem 2016-06-20 07:18:15 i worked on switching to the x11 keymaps last friday 2016-06-20 07:18:37 i think thats the way forward 2016-06-20 07:33:13 ncopa: I want to remove nginx-lua 2016-06-20 07:33:35 how can I make apk install main/nginx as the replacement? 2016-06-20 07:38:21 does main/nginx provide same functionality? 2016-06-20 07:38:46 if so, then replaces="nginx-lua"; provides="nginx-lua" 2016-06-20 07:39:37 ncopa: yes, it does 2016-06-20 07:39:55 I remember nginx was meant to be vanilla and nginx-lua to have some of openresty modules, but that's not the case anymore 2016-06-20 07:39:58 thanks 2016-06-20 07:40:29 I would like to do the same change for 3.4, if it doesn't make you cringe too much – nginx-lua is completely out of date 2016-06-20 07:40:34 morning everybody! happy summer solstice! 2016-06-20 07:41:22 barthalion: ok 2016-06-20 07:41:36 k, will push it today 2016-06-20 08:07:01 morning 2016-06-20 09:01:16 turns out main/nginx literally provides nginx-lua subpackage 2016-06-20 09:43:23 barthalion: yeah it's even easier now that 1.10 support external modules 2016-06-20 10:05:01 yeah, I packaged two for Arch in past, didn't know someone refactored Alpine package 2016-06-20 10:05:29 I'll probably add some more modules there 2016-06-20 10:06:00 any preferences about naming? it's nginx-$module now, while I'd prefer nginx-mod-$module at least for new subpackages 2016-06-20 10:13:01 ncopa: ^ 2016-06-20 10:19:51 any chance of a linux-rpi re-build? version has been bumped and there's a fix I'm interested in :) 2016-06-20 10:20:16 ScrumpyJack: you need to ping fabled about it, unless it's straightforward pkgver gump 2016-06-20 10:21:39 barthalion: i dont have any strong opinion re nginx naming 2016-06-20 10:22:03 re rpi, i think there is a blocking package 2016-06-20 10:22:06 i'll look at it now 2016-06-20 10:22:20 I'll take it as a wild card 2016-06-20 10:25:12 I think it's just a matter of triggering bulldozer for linux-rpi. A patch to linux-rpi has been applied 2016-06-20 10:25:54 ah, yes, then as ncopa said, builder is blocked 2016-06-20 10:25:54 but i could be wrong, i've no experience of bulldozer 2016-06-20 10:26:04 tried to workaround this, but to no avail 2016-06-20 10:26:43 ah, i see. blocked on building a package that fails? 2016-06-20 10:26:48 subversion? 2016-06-20 10:29:49 yes 2016-06-20 10:29:52 i'm looking at it now 2016-06-20 10:35:27 barthalion: I prefer nginx-$module 2016-06-20 10:36:23 why is that? I like that nginx-mod-$module clearly says that this is a module, not a subpackage coming from upstream tarball 2016-06-20 10:37:49 barthalion: what about modules that come from upstream tarball ? 2016-06-20 10:38:07 they are built-in into nginx binary 2016-06-20 10:38:14 they shouldn't be 2016-06-20 10:38:20 that's the whole point 2016-06-20 10:38:49 that's the point for external modules, not official ones, developed as part of nginx 2016-06-20 10:39:39 not really 2016-06-20 10:39:52 and anyway, that's how it works now 2016-06-20 10:40:06 it means now you don't need to have GeoIP as a dep when you just want to do reverse proxying 2016-06-20 10:40:40 would be nice to have modsecurity as well now that it's possible to integrate 2016-06-20 10:42:20 oh now I see 2016-06-20 10:42:22 official modules default to static, not dynamic, and I'm talking about 3rd party now anyway 2016-06-20 10:42:32 there's a lot missing from the official ones you can externalize 2016-06-20 10:43:39 I'm not really interested in splitting these for now, and it still would make sense to name it nginx-mod-$module imho 2016-06-20 10:45:21 so you wanna change the actual nginx-$modname-module to nginx-mod-$modname ? 2016-06-20 10:45:46 ah no my bad I missread 2016-06-20 10:45:49 no, no 2016-06-20 10:45:56 we have currently nginx-$mod 2016-06-20 10:46:04 yep true 2016-06-20 10:48:19 you know what, lets do nginx-mod-* 2016-06-20 10:48:20 for modules 2016-06-20 10:48:53 nginx-$mod can be used for nginx binary variants 2016-06-20 10:49:06 where you compiled in the module in the main binary 2016-06-20 10:49:33 where with nginx-mod-$mod, you dont get nginx binary, only the plugin/module 2016-06-20 10:49:57 ok but for example 2016-06-20 10:50:05 barthalion: while at it, i think we want a -debug build (which is not the same as -dbg) 2016-06-20 10:50:05 the perl module 2016-06-20 10:50:34 If tomorrow you enable it, would be nice to put it as nginx-mod-perl 2016-06-20 10:50:49 and i think we should try follow upstream configure options if possible 2016-06-20 10:50:53 cause you don't want - at least I don't - fetch perl if not using the module 2016-06-20 10:52:05 https://github.com/nginxinc/docker-nginx/blob/master/mainline/alpine/Dockerfile 2016-06-20 10:52:36 https://www.nginx.com/blog/dynamic-modules-nginx-1-9-11/ < there's the complete list which apart from mail don't seem to be enabled on Alpine's build 2016-06-20 10:53:26 yep quite 2016-06-20 12:25:39 does alpine work on pi zero now? 2016-06-20 13:13:24 yes it does :) 2016-06-20 13:18:06 any takes for this? https://bugs.alpinelinux.org/issues/5758 2016-06-20 13:41:18 oneinsect: i haven't really thought that providign .xz and .gz and .bz2 of same content would provide much value compared to the extra bw cost 2016-06-20 13:42:02 you can use 7zip to extract files fro iso similar to tar or zip 2016-06-20 13:42:27 ncopa: i have struggled and gave up... alpine works but only in rw mode when using chroot based ext4 2016-06-20 13:42:35 there are 3 ways (that i know) you can extract the files 2016-06-20 13:42:51 however the initramfs from armhf simply works by extracting and making an ext4 2016-06-20 13:42:53 7z x 2016-06-20 13:43:05 in ro only mode 2016-06-20 13:43:06 mount -o loop file.iso && cp -r 2016-06-20 13:43:11 yes 2016-06-20 13:43:14 ncopa: 2016-06-20 13:43:24 and uniso < file.iso 2016-06-20 13:44:28 @ncopa: is there a way we can build an ext4 file-system based on alpine but that retains the in memory nature of alpine 2016-06-20 13:44:38 as i understand #5758 it is just like: mount -o loop file.iso /mnt && tar -czf file.tar.gz mnt/ 2016-06-20 13:44:51 diskless but that doesnt use initramfs 2016-06-20 13:45:06 that will be a huge huge huge help 2016-06-20 13:45:19 diskless but without initramfs just using ext4 2016-06-20 13:45:27 @ncopa: ? 2016-06-20 13:45:28 for example, to boot in uefi mode 2016-06-20 13:46:01 its just to create the contents of the iso to a uefi boot partition and set up the boot manager 2016-06-20 13:47:00 to create a boot able ext4 image: 2016-06-20 13:47:13 alpine works in ext4 image 2016-06-20 13:47:16 dd if=/dev/zero of=file.img ... 2016-06-20 13:47:23 but it writes to the emmc 2016-06-20 13:47:31 mkfs.ext4 file.img 2016-06-20 13:47:41 yes i did that 2016-06-20 13:47:44 and booted 2016-06-20 13:47:44 mount -o loop file.img /mnt 2016-06-20 13:47:48 it works 2016-06-20 13:47:55 7z x file.iso -> /mnt 2016-06-20 13:47:59 umount /mnt 2016-06-20 13:48:10 well you need run extlinux there too 2016-06-20 13:48:15 and fix mbr 2016-06-20 13:48:49 my main point is that the alpine.iso is not much different than alpine.zip or alpine.tar.gz 2016-06-20 13:48:57 hmmm 2016-06-20 13:49:01 one question 2016-06-20 13:49:18 never mind 2016-06-20 13:49:46 i dont think its worh the extra disp space and bandwith 2016-06-20 13:49:52 diskspace 2016-06-20 13:49:55 never problem 2016-06-20 13:49:57 i understand 2016-06-20 13:50:16 the only benefit it provides is that you can use tar -zxf instead of 7z x 2016-06-20 13:50:33 eitherway i tried the above... creating an ext4....its just that the legacy kernel doesnt support the initramfs file somehow...even when enabled in config 2016-06-20 13:50:40 but thats okie 2016-06-20 13:51:01 huh? 2016-06-20 13:51:11 its Intel Edison 2016-06-20 13:51:29 yes but intel edison is special 2016-06-20 13:51:33 i know 2016-06-20 13:51:42 if you want different kernel modules in there by default 2016-06-20 13:51:48 then thats other question 2016-06-20 13:51:59 do you know what kernel modules you need? 2016-06-20 13:52:21 yes squashfs, ext4, fat, snd etc etc... 2016-06-20 13:52:38 the good part is this ... it works ...alpine linux boots if i create an ext4 via chroot method...since there is no initramfs file involved 2016-06-20 13:52:46 you need modules to find the mmc and ext4 2016-06-20 13:52:55 yes ofcourse all of them compiled in 2016-06-20 13:53:07 the problem is this 2016-06-20 13:53:16 i dont think providing a .tar.gz image would help 2016-06-20 13:53:34 it would just end up as: the tar.gz does not work on edison, please provide a tar.bz2 2016-06-20 13:53:38 :) 2016-06-20 13:53:39 the default serial is ttyMFD in intel and the iso excepts i think ttys1 2016-06-20 13:54:18 neither can i get to change it in intel edison or neither can i change it in the iso since inittab etc are already packed 2016-06-20 13:54:33 they are packed in the apks 2016-06-20 13:54:36 folde 2016-06-20 13:54:46 if you extract the iso 2016-06-20 13:54:51 to ext4 2016-06-20 13:54:57 yes i dont that 2016-06-20 13:55:00 done* 2016-06-20 13:55:03 ok 2016-06-20 13:55:06 but you dont use syslinux 2016-06-20 13:55:10 its uboot right? 2016-06-20 13:55:15 no syslinux 2016-06-20 13:55:18 only U-Boot 2016-06-20 13:55:31 the future is going to be only u-boot in many places 2016-06-20 13:55:32 how do you edit boot cmdline in uboot? 2016-06-20 13:55:39 yes yes yes 2016-06-20 13:55:45 its a file you can edit riight? 2016-06-20 13:55:51 yes 2016-06-20 13:55:57 then you add console=ttyMFD 2016-06-20 13:56:15 the default alpine initramfs will edit the inittab for you 2016-06-20 13:56:25 oh 2016-06-20 13:56:26 it did not pickup so had to manually change in etc/inittab 2016-06-20 13:56:28 maybe it doesnt 2016-06-20 13:56:32 right 2016-06-20 13:56:34 its not picking up 2016-06-20 13:56:35 thats a bug 2016-06-20 13:56:41 and i think i know why 2016-06-20 13:56:47 can you please help 2016-06-20 13:57:01 going forward i see many boards from intel on their roadmap with only u-boot 2016-06-20 13:57:03 no bios 2016-06-20 13:57:11 no syslinux 2016-06-20 13:57:32 should i file a bug??? 2016-06-20 13:58:04 yes please 2016-06-20 13:58:20 in any case i think #5758 is bogus. sorry :) 2016-06-20 13:58:35 unless ofc 2016-06-20 13:58:40 please discard it 2016-06-20 13:58:43 i am sorry 2016-06-20 13:59:00 unless it says: please provide x86/x86_64 image with uboot 2016-06-20 13:59:52 may be that should have the correct ticket 2016-06-20 13:59:56 you can edit it? 2016-06-20 14:02:17 please you do 2016-06-20 14:03:53 oneinsect: do you have a console=ttyMFD line example? 2016-06-20 14:04:11 bootargs rootwait root=/dev/mmcblk0p5 rw rootfstype=ext4 console=ttyMFD2 level=8 init=/sbin/init console=ttyMFD2 earlyprintk=ttyMFD2 2016-06-20 14:04:12 is it just console=ttyMFD ? 2016-06-20 14:04:27 sorry 2016-06-20 14:04:35 ok 2016-06-20 14:05:04 bootargs rootwait root=/dev/mmcblk0p5 rw rootfstype=ext4 console=ttyMFD2 earlyprintk=ttyMFD2 level=8 init=/sbin/init 2016-06-20 14:05:30 EXPECTED ttyMFD2 needs to be put in /etc/inittab and in /etc/securetty ttyMFD2::respawn:/sbin/getty -L ttyMFD2 115200 vt100 2016-06-20 14:05:36 not happening 2016-06-20 14:05:44 filing bug? 2016-06-20 14:05:46 should i go ahead? 2016-06-20 14:06:20 http://bugs.alpinelinux.org/issues/5759 2016-06-20 14:07:19 i have no permissions to edit the title and change to please provide x86/x86_64 image with uboot for http://bugs.alpinelinux.org/issues/5758 2016-06-20 14:08:38 and should hte inittab line look like? 2016-06-20 14:08:40 ncopa: I can boot Alpine from UEFI boot partition? That's awesome 2016-06-20 14:09:01 barthalion: some interesting details about the mutex test: https://github.com/dotnet/coreclr/issues/5872#issuecomment-227115826. Is it a known behavior/issue on Alpine Linux, the way OS is handling mutext and issuing sigEOWNERDEAD? 2016-06-20 14:10:48 peterj: your questions are far beyond my knowledge, you should ask skarnet or on #musl, sorry 2016-06-20 14:11:57 barthalion: np; sure will ask on #musl and ping skarnet too :) 2016-06-20 14:12:36 let me show you one moment @ncopa: 2016-06-20 14:13:34 don't ping me for questions about mutexes, I suck at multithreading :D 2016-06-20 14:13:42 #musl will have the answers though 2016-06-20 14:16:31 ncopa: this line should be there in inittab! ttyMFD2::respawn:/sbin/getty -L ttyMFD2 115200 vt100 2016-06-20 14:17:04 and in /etc/securetty it should have ttyMFD2 2016-06-20 14:17:41 why do people have to change the names of serial devices 2016-06-20 14:17:46 it makes everything harder 2016-06-20 14:18:00 if vendors could just agree on ttyS0 for everyone >.> 2016-06-20 14:20:39 well stupid intel edison 2016-06-20 14:20:55 intel has used a different non standard naming 2016-06-20 14:21:19 also The alpine initramfs is not picking up the bootargs and changing the etc/inittab and etc/securetty 2016-06-20 14:21:21 barthalion: https://wiki.alpinelinux.org/wiki/Create_UEFI_boot_USB 2016-06-20 14:21:32 this is supposed to happen ...hence ncopa 2016-06-20 14:23:20 arent there any ttyUSB* kind of serial too? 2016-06-20 14:23:33 ncopa: I'll test it on my laptop today, thx 2016-06-20 14:24:36 barthalion: i have tested on my lappy, and on a macbook 2016-06-20 14:24:48 it boots. it gives some warning/error during boot 2016-06-20 14:24:58 i think it fails to mount efivars or similar 2016-06-20 14:25:10 but it boots 2016-06-20 14:25:47 i also verified that it work to copy missing firware to firmware/ on boot media 2016-06-20 14:31:07 oneinsect: do you have a custom initramfs? 2016-06-20 14:32:19 well i tried with custom initramfs ....i just extracted from initramfs-vanilla and added lib/modules/3.9.xxx folder and firmware folder and repacked it 2016-06-20 14:32:23 no big changes 2016-06-20 14:32:32 can you please try this patch: http://tpaste.us/2NkB 2016-06-20 14:33:17 that is modify /init in initramfs 2016-06-20 14:33:46 let me check ncopa: its gonna take a while ...i have to head home and check it 2016-06-20 14:33:56 i will report to the bug directly of the results 2016-06-20 14:40:46 ncopa: does alpine_dev exist? does it every pick up the apks folder if given in bootargs? 2016-06-20 14:40:49 variable* 2016-06-20 14:41:26 or alpine_dev is now ignored?? 2016-06-20 14:41:37 alpine_dev is ignored now 2016-06-20 14:42:06 so nlplug-findfs should find boot repos and apkovl?? what if they cannot find ? 2016-06-20 14:42:16 then it gives up 2016-06-20 14:42:32 with error 2016-06-20 14:42:42 same if it does not find the root=... 2016-06-20 14:42:54 aaaah i wish alpine_dev existed ...it is so much easier for internal eMMc based boards 2016-06-20 14:43:23 never mind 2016-06-20 14:43:45 blkidid needs to find it 2016-06-20 14:44:01 it should be easier now 2016-06-20 14:44:11 as you should not need to specify the device 2016-06-20 14:44:28 so if you use /dev/sr0 or /dev/sda1 2016-06-20 14:44:32 it should not matter 2016-06-20 14:44:35 harder with intel edison ....yeaaa 2016-06-20 14:45:00 not really, you need kernel to detect it, and have a file system that blkid recognizes 2016-06-20 14:45:02 thats all 2016-06-20 14:45:08 hmmmm 2016-06-20 14:45:14 just stuff in the kernel modules in initramfs and it should work 2016-06-20 14:46:20 yeaa i will try with your tty patch as well later today 2016-06-20 14:46:37 i already pushed it 2016-06-20 14:46:46 i tested it in qemu 2016-06-20 14:48:57 ncopa: where you successful with subversion? 2016-06-20 14:49:17 i think so 2016-06-20 14:49:34 its still building 2016-06-20 14:53:40 ncopa: one question what are the bootargs to helping kernel find initramfs file and modloop files as well apks folder in /dev/mmcblk0p5 2016-06-20 14:53:54 this will not work 2016-06-20 14:53:55 initrd=/boot/initramfs-grsec modloop=/boot/modloop-grsec modules=loop,squashfs,sd-mod,usb-storage quiet usbdelay=1 2016-06-20 14:54:16 my fat is the 5th partition in the eMMc 2016-06-20 14:55:20 the first is u-boot0 and then uboot0-env and then uboot1 and then uboot1-env and finally fat partition ...with redundancy for u-boot 2016-06-20 14:55:37 blkid only needs to find it 2016-06-20 14:56:25 how will find it without any args? i mean what are correct args ? the above are right assuming those file exist in /dev/mmcblk0p5? 2016-06-20 15:06:04 ncopa: where is this file initramfs-init.in 2016-06-20 15:06:36 not there in initramfs-vaniila 2016-06-20 15:07:06 got it 2016-06-20 15:07:23 setup_inittab_console() 2016-06-20 15:50:04 ncopa: it boots, but it didn't load apks 2016-06-20 15:50:23 despite .boot_repository file 2016-06-20 15:50:37 does it find the repo? 2016-06-20 15:50:53 what does /tmp/repositories say? 2016-06-20 15:51:09 I ran apk update and it didn't fail, so I think so 2016-06-20 15:51:16 I installed packages from it 2016-06-20 15:51:48 and it was enough to actually fix my screwed Arch initramfs so I'm happy anyway 2016-06-20 15:51:56 :) 2016-06-20 15:52:36 exactly what I expected from Alpine :) 2016-06-20 15:53:10 barthalion: is your initramfs custom made? 2016-06-20 15:53:31 and can you please tell me what are your bootargs? 2016-06-20 15:53:37 oneinsect: no, but I thought it's going to work fine after compressing with pbzip2 2016-06-20 15:53:45 hmmmm 2016-06-20 15:53:50 also I'm talking about UEFI scenario 2016-06-20 15:53:51 not yours 2016-06-20 15:53:58 i know 2016-06-20 15:54:11 i just applied ncopa: patch let me know if i get any console now 2016-06-20 15:54:17 https://wiki.alpinelinux.org/wiki/Create_UEFI_boot_USB this is all I did to make it boot 2016-06-20 15:54:18 see* 2016-06-20 16:07:52 guys 2016-06-20 16:07:53 UDF-fs: warning (device ram0): udf_fill_super: No partition found (2) 2016-06-20 16:08:03 any idea what that error means? 2016-06-20 18:14:47 ncopa: wrt nginx-debug, what did you mean? 2016-06-20 18:15:18 ncopa: --with-debug? 2016-06-20 18:34:01 Debian has funny naming scheme… libnginx-mod-$modname 2016-06-20 19:23:19 barthalion: yes, --with-debug 2016-06-20 19:23:43 and rename the binary to nginx-debug 2016-06-20 19:24:26 similar to what upstream does: https://github.com/nginxinc/docker-nginx/blob/master/mainline/alpine/Dockerfile#L81 2016-06-20 20:36:15 ncopa: yeah, that's how I figured out what you mean 2016-06-20 21:34:59 good evening 2016-06-20 21:54:42 <^7heo> moin 2016-06-20 22:06:10 looks like naxsi can also be a dynamic module 2016-06-20 22:06:49 and Valery even adapted main/nginx for those, which got my -1 for bumping to development branch 2016-06-20 22:13:43 philosophical question: should installed module be loaded automatically? 2016-06-20 22:14:05 it would make sense imho 2016-06-20 22:14:32 but nginx works differently, it must be explicitly enabled in config file, right? 2016-06-20 22:16:27 yeah, but I'll modify default config 2016-06-20 22:16:43 how? 2016-06-20 22:16:49 using post-install hook? 2016-06-20 22:17:01 no cookies for existing installations 2016-06-20 22:17:13 ? 2016-06-20 22:17:44 I'll just sed things in /etc/nginx/nginx.conf 2016-06-20 22:17:56 when/how? 2016-06-20 22:18:26 well, there aren't many options, are there? 2016-06-20 22:18:34 in either prepare or package 2016-06-20 22:19:04 that can’t work… 2016-06-20 22:19:14 packages are built on the build server, not end-user system 2016-06-20 22:19:18 … 2016-06-20 22:19:40 so the only option is using post-install script imho 2016-06-20 22:20:26 eh, what modules do you mean? built-in or dynamically loaded that will be in separate packages? 2016-06-20 22:22:35 to recap, because you misunderstand me 2016-06-20 22:22:42 no changes will be made to existing installations 2016-06-20 22:22:57 new installs of nginx will have 'include /etc/nginx/modules.d/*.conf' line 2016-06-20 22:23:25 aha 2016-06-20 22:23:55 and modules will provide /etc/nginx/modules.d/$modname.conf files, so they get autoloaded after service restart 2016-06-20 22:24:06 gotcha 2016-06-21 06:30:40 morning 2016-06-21 06:30:44 barthalion: you about? 2016-06-21 06:30:55 ScrumpyJack: at work, but yes 2016-06-21 06:31:36 just wondering if you are still using that owncloud account, I'm thinking of either upgrading to owncloud9 or moving to baikal. 2016-06-21 06:32:29 I moved to notebook a couple of weeks ago 2016-06-21 06:32:38 I used baikal in past and it was quite good 2016-06-21 06:34:09 notebook? 2016-06-21 06:34:27 aka. pen and paper 2016-06-21 06:34:53 :) 2016-06-21 07:59:30 <^7heo> I guess you guys saw http://kmkeen.com/maintainers-matter/ ? 2016-06-21 08:22:26 morning 2016-06-21 08:47:22 morning 2016-06-21 11:01:38 morning fellas 2016-06-21 11:10:49 afternoon, technically 2016-06-21 11:15:10 rather afternoon 2016-06-21 11:27:54 linux-rpi 4.4.13 still hasn't build. is buildozer blocked again? 2016-06-21 11:28:20 looks like building docker on armhf is failing 2016-06-21 11:51:08 any ideas why i get UDF-fs: warning (device ram0): udf_fill_super: No partition found (2) 2016-06-21 11:51:16 it hangs there on intel edison 2016-06-21 11:51:26 during booting 2016-06-21 12:25:31 somehow your kernel isn't creating the ram disks on boot? 2016-06-21 12:37:32 ScrumpyJack: 2016-06-21 12:37:37 what could be the reason 2016-06-21 12:40:31 does alpine initramfs depend on sysfs? 2016-06-21 12:41:51 i have enabled Initial RAM filesystem and Ram disk (initramfs/initrd) support in kernel 2016-06-21 12:50:15 have a look at Documentation/initrd.txt 2016-06-21 12:54:31 if you find out why your ramdisk aren't created, would be nice to know 2016-06-21 12:55:37 reading initrd 2016-06-21 12:55:48 he kernel has to be compiled with RAM disk support and with support for the initial RAM disk enabled. Also, at least all components needed to execute programs from initrd (e.g. executable format and file system) must be compiled into the kernel 2016-06-21 12:56:09 i am doing that....executable format and file system....yes there are in the kernel 2016-06-21 13:44:49 ncopa: hi dude. 2016-06-21 13:45:14 i think docker is blocking buildozer on armhf 2016-06-21 13:45:27 hi 2016-06-21 13:45:49 yes, but I'm not sure how to fix it 2016-06-21 13:46:32 can we remove docker from the armhf build? 2016-06-21 13:48:13 (we still don't have a working alpine-rpi-3.4.0, as overlays are missing) 2016-06-21 13:50:20 its building now 2016-06-21 13:50:23 :) 2016-06-21 13:50:24 i disabled pax 2016-06-21 13:50:44 its the go memory allocator that does not work well with pax on arm 2016-06-21 13:51:44 ScrumpyJack: note that the 3.4-stable builder is not the same as the git master builder 2016-06-21 13:51:59 ScrumpyJack: what package are you waiting for? 2016-06-21 14:05:08 linux-rpi 2016-06-21 15:56:27 ScrumpyJack: isnt that in main? 2016-06-21 15:56:32 it should not block packages in main 2016-06-21 16:13:14 sorry, (bloddy) meetings 2016-06-21 16:15:13 in that case, maybe it's my mirror that's behind 2016-06-21 16:16:00 edge/main that is 2016-06-21 17:21:07 would someone be kind enough to create a new alpine-rpi-3.4.0 file to replace the one on the download page which is missing overlays? 2016-06-21 17:21:14 or should we wait for 3.4.1? 2016-06-21 17:33:22 i am publishing the alpine distro for intel edison as we speak 2016-06-21 17:33:28 anyone wants to test 2016-06-21 17:41:16 ScrumpyJack: the alpine-rpi-3.4.0.tar.gz? 2016-06-21 17:41:22 or the linux-rpi package 2016-06-21 17:41:37 the releases only happen when we tag 2016-06-21 17:42:38 http://bugs.alpinelinux.org/projects/alpine/issues?fixed_version_id=108 2016-06-21 17:59:38 hey girls :) 2016-06-21 17:59:50 girls? where?! :) 2016-06-21 18:02:00 ley, what about Rust on Alpine? any progress here? 2016-06-21 18:04:45 barthalion: the branch for v3.4 is 3.4-stable not 3.4/stable 2016-06-21 18:07:06 jirutka: you have to ask ncopa about that :) he is the man! 2016-06-21 18:07:38 leo: about Rust or girls? 2016-06-21 18:07:44 ncopa: uh, I thought I configured git to push to 3.4-stable despite local name 2016-06-21 18:07:46 ncopa: sorry 2016-06-21 18:07:55 np 2016-06-21 18:08:03 nothing bad happened except confusion 2016-06-21 18:08:21 barthalion: can you fix it or do you want me to do so? 2016-06-21 18:08:31 if you delete wrong branch bad things will happen... 2016-06-21 18:09:13 ncopa: after your last sentence, I guess I'd rather you to do it… 2016-06-21 18:09:31 ok. i'll do :) 2016-06-21 18:10:31 'bad things will happen' is quite nice way to describe the disaster 2016-06-21 18:25:03 i will cherry-pick your commit first 2016-06-21 18:28:07 done 2016-06-21 18:28:32 <3 2016-06-21 22:06:26 search now supports traditional queryies 2016-06-21 22:06:37 try -> curl 'https://api.alpinelinux.org/search/packages?category=edge:main:x86&name=q_&flagged=yes&maintainer=Nat_&sort=fid' | jq -C . | less 2016-06-21 22:07:01 jirutka: ^^^ happy ;) 2016-06-21 22:07:03 why /search/packages and not just /packages ? 2016-06-21 22:07:43 /packages/search ?? 2016-06-21 22:07:56 no, why do you need “search”? 2016-06-21 22:08:07 URI should contain just nouns, not verbs 2016-06-21 22:08:18 sometimes there’s no other option, but that’s definitely not that case 2016-06-21 22:08:48 you have resources /packages and just wanna filter it 2016-06-21 22:10:26 you mean no keyword ? 2016-06-21 22:10:43 . /packages should return (paginated) collection of all packages, right? when you wanna filter that collection, why to switch to another resource? 2016-06-21 22:10:44 like search|edit|delete 2016-06-21 22:10:51 yes 2016-06-21 22:11:10 then /packages/search ?? 2016-06-21 22:11:23 once again, no verbs! 2016-06-21 22:11:54 hmm... 2016-06-21 22:12:28 have you read some of the books I’ve shared you before? 2016-06-21 22:13:32 nope, but just tried to skip to part where search by POST is ok for complex queries 2016-06-21 22:13:42 ;) 2016-06-21 22:13:54 yes, for complex queries 2016-06-21 22:14:02 this is not a complex query, but quite simple query 2016-06-21 22:14:21 that is why its still in GET 2016-06-21 22:16:17 implementing is not an issue, arguments after implementing is 2016-06-21 22:16:35 you could come up with all URI, and I can implement it 2016-06-21 22:16:55 since the api is still WIP 2016-06-21 22:17:27 let me know the URIs 2016-06-21 22:18:30 if it breaks your previous existing regious parsing scripts, that is no argument 2016-06-21 22:18:37 well, my suggestion is design the API first (e.g. using raml.org) 2016-06-21 22:18:46 the api would come with documents 2016-06-21 22:19:07 and some hints on how use it 2016-06-21 22:20:05 since you have issue with URI, just suggest me all, I am ready to implement 2016-06-21 22:20:32 otherwise I am ok with current set 2016-06-21 22:21:37 sry, but I currently don’t have a time to design whole API :( if you give me some document with API design (URIs, representations, etc), then I can take a look at it and add comments 2016-06-21 22:24:27 however, as I already said, it’d be better if you read some books/articles how to design RESTful APIs, then actually design it, and after that start implementing; otherwise you’re building a house from the roof 2016-06-21 22:33:27 /search/{where} seems ok to me, I might think of adding /packages/..filters.. too 2016-06-21 22:33:37 as a re-route to /search 2016-06-21 22:34:16 /search/ for kinds of search 2016-06-21 22:34:25 it doesn’t make sense 2016-06-21 22:34:40 don't user /search then 2016-06-21 22:35:16 REST is not RPC 2016-06-21 22:35:18 just follow your previous URI, both would be there 2016-06-21 22:39:40 how do you rationale, http://api.example.restapi.org/france/paris/louvre/leonardo-da-vinci/mona-lisa 2016-06-21 22:40:02 ? 2016-06-21 22:41:03 that’s okay, all of the path’s components make sense as resources 2016-06-21 22:42:40 btw do you have version number in the path? like api/v1/ 2016-06-21 22:42:59 you are more concerned with scripts oo object/resource that handles it 2016-06-21 22:43:17 ? 2016-06-21 22:43:44 i don’t understand what you mean 2016-06-21 22:44:45 meaning, /packages/edge/main/x86?...filter... should be ok then ? 2016-06-21 22:44:57 yes 2016-06-21 22:45:04 hmmm ... 2016-06-21 22:45:39 actually… 2016-06-21 22:46:22 sounds more like JAVA userland 2016-06-21 22:46:36 ?? 2016-06-21 22:47:49 nm, would thing over /packages/edge/main/x86 2016-06-21 22:48:11 is the order important ? 2016-06-21 22:48:19 it depends what actually the package is; /packages/edge/main/x86 implies that the package is uniquely identified by branch, repo and arch 2016-06-21 22:50:26 so api user needs to know the order 2016-06-21 22:50:36 branch/repo/arch 2016-06-21 22:50:52 this order is already set, http://dl-cdn.alpinelinux.org/alpine/ 2016-06-21 22:51:37 well, the api in gererall 2016-06-21 22:51:59 it’d be probably reasonable to split package into something like a class (APKBUILD) and instances (.apk files) 2016-06-21 22:53:25 APKBUILD’s name should be unique across all the official repos (or at least I hope so) 2016-06-21 22:54:36 no, this may not be always true and we already uses / as id 2016-06-21 22:58:14 hm, I think that /packages/{branch}/{repo}/{arch} is both correct and pragmatic solution 2016-06-21 23:03:40 so the docs need to specify the order 2016-06-21 23:05:19 as hierarchy seem important 2016-06-21 23:05:25 yes 2016-06-21 23:08:58 was thinking of api/v1/, good to bring it out 2016-06-21 23:09:37 since we have api.a.o, api.a.o/api/v1 OR api.a.o/v1 ? 2016-06-21 23:10:13 api.a.o/v1 makes more sense imo 2016-06-21 23:10:44 I am going nuts, let me get 'tea' 2016-06-21 23:12:57 :-) 2016-06-21 23:13:39 It's tea-time in Arizona in 7 minutes ;-) 2016-06-21 23:21:06 :) 2016-06-21 23:21:24 ok, I think the basic code for /packages/flagged is done 2016-06-21 23:21:57 you have a conflict here 2016-06-21 23:22:21 this would have data similar to http://pkgs.alpinelinux.org/flagged landing page 2016-06-21 23:22:27 how do you disquish between /packages/{branch} and /packages/flagged? 2016-06-21 23:22:57 hmm.. so how to get data similar to http://pkgs.alpinelinux.org/flagged ? 2016-06-21 23:23:08 just the URI ? 2016-06-21 23:23:29 its pkgs.a.o 2016-06-21 23:23:31 http://pkgs.alpinelinux.org/package/edge/community/x86_64/wine 2016-06-21 23:23:31 http://pkgs.alpinelinux.org/flagged 2016-06-21 23:23:50 flagged can be just a filter 2016-06-21 23:24:23 something like /packages/edge/main/x86?flagged=true 2016-06-21 23:24:33 ok I thought a identifying resource needs to be there 2016-06-21 23:24:45 ok 2016-06-21 23:25:12 and /packages?flagged=true to get flagged packages across branches, repositories and arch 2016-06-21 23:25:57 so http://pkgs.alpinelinux.org/package/v3.4/main/armhf/zabbix is superfluous 2016-06-21 23:26:00 should be http://pkgs.alpinelinux.org/v3.4/main/armhf/zabbix 2016-06-21 23:26:25 hmm.. ok kinda getting now 2016-06-21 23:26:36 pkgs.a.o is a website, not API… 2016-06-21 23:27:17 ok 2016-06-21 23:30:09 api.a.o/package/edge is same as api.a.o/package?edge=true 2016-06-21 23:30:10 and not api.a.o/package?branch=edge 2016-06-21 23:30:23 or code for both ? 2016-06-21 23:30:35 branch=edge 2016-06-21 23:30:46 ok 2016-06-21 23:30:48 you really don’t wanna add parameters for all the repos ;) 2016-06-21 23:30:58 ;) 2016-06-21 23:30:59 correction: branches 2016-06-21 23:31:15 like edge=true, v3.4=true… moreover these are mutually exclusive 2016-06-21 23:31:17 resources = plural 2016-06-21 23:31:20 ok 2016-06-21 23:34:48 so if someone is familiar with aports and basic distro packaging + RESTFul API, 2016-06-21 23:35:03 designing the URI would be sufficient and take 1 or 2hrs 2016-06-21 23:35:24 backend coding guide is not important 2016-06-21 23:43:46 hmmm.., ok, I would go with my current set of routes, 2016-06-21 23:43:46 as long as "Representational State Transfer" URIs are clear and dintinct in what "state it represents" and "exposes a set of data and functions to facilitate interactions between computer programs and allow them to exchange information" 2016-06-21 23:43:50 Would try to only expose in DOCs URI get universally accepted as RESTFul (cannot keep everybody happy, or can I) 2016-06-21 23:44:13 URIs that* 2016-06-21 23:44:58 the Framework (phalconphp) seems flexible enough to handle it 2016-06-21 23:47:34 would try to avoid words like "design patterns", "resource archetypes" "resource" where possible 2016-06-21 23:48:01 resource is esential… 2016-06-21 23:48:59 we re not building billion dollar company code, that cannot be of < 2000 lines 2016-06-21 23:49:14 it would not look nice on tax papers 2016-06-21 23:49:19 ?? 2016-06-21 23:49:31 REST is all about resources 2016-06-21 23:49:45 nw, just a pun/remark 2016-06-21 23:49:49 kk 2016-06-21 23:49:55 yes, agreed 2016-06-21 23:56:27 irrespective of that, those words make my tea go sour 2016-06-22 06:35:34 ncopa: got it 2016-06-22 07:32:21 fabled: morning dude, linux-rpi 4.4.13-r0 still doesn't contain the overlays :( 2016-06-22 07:36:54 unless i'm missing something 2016-06-22 07:42:55 they are in the linux-rpi package but they don't get installed up update-kernel 2016-06-22 09:51:52 fixed in patch 2134 2016-06-22 09:52:35 perhaps that should have gone to the alpine-devel mailing list, as technically not an aport? 2016-06-22 11:41:39 hmm, overlays aren't loading at boot :( 2016-06-22 18:14:54 ncopa, are you here? :) 2016-06-22 18:22:22 jirutka: hi. im here 2016-06-22 18:22:44 ncopa: I’ve created GitHub webhook handler for closing pull requests https://github.com/jirutka/github-pulls-closer :) 2016-06-22 18:23:18 ncopa: just setting it up. and also welcome algitbot on GitHub https://github.com/algitbot :) 2016-06-22 18:23:37 ncopa: how can I send you password for the algitbot account securely? 2016-06-22 18:24:12 gpg encrypted email? 2016-06-22 18:24:34 using the pubkey used for signing the iso images 2016-06-22 18:24:47 okay 2016-06-22 18:54:25 For RestURIs, I formalized my set of keywords to follow - "simple, consistent, meaningful, intuitive, somewhat-hierarchal, distinct, unambiguous" 2016-06-22 18:54:28 I am trying to find as who cooked up 'verb', 'plural' thingies into the RESTApi 2016-06-22 18:54:39 looking into original http://www.ics.uci.edu/~fielding/pubs/dissertation/top.htm now 2016-06-22 18:55:35 http://pkgs.alpinelinux.org/packages?name=&branch=&repo=&arch=&maintainer= 2016-06-22 18:55:35 http://pkgs.alpinelinux.org/package/v3.4/main/armhf/zabbix 2016-06-22 18:55:35 hmmm... both singular+plural resources 2016-06-22 18:55:51 jirutka: implement both resource ? 2016-06-22 18:56:34 yes, both makes sense; the first one is for filtering by some attributes, the second one is fully qualified package name 2016-06-22 18:57:13 making sense is different issue, its programatic, not end user ease 2016-06-22 18:57:22 Ok, would stick to the above keywords and try follow the dissertation or any only if it conveys such ideas. 2016-06-22 18:57:26 In absense of any standards 2016-06-22 18:57:27 I am moving all the api uri's to singular and only use plural where it doesn't convey the above keywords/ideas properly 2016-06-22 18:57:44 http://jsonapi.org/format/1.1/#document-resource-objects 2016-06-22 18:57:44 Note: This spec is agnostic about inflection rules, so the value of type can be either plural or singular 2016-06-22 18:58:01 I am moving all 'type' to singular, including any references in URIs 2016-06-22 18:58:01 eg. api.alpinelinux.org/packages becomes api.alpinelinux.org/package 2016-06-22 18:58:14 singular is what we use in verbal communication 2016-06-22 18:58:14 it is also consistent with page/2 and not pages/2 2016-06-22 18:58:24 In verbal communication do we say "pls get group 2 here" or "pls get groups 2 here" ? 2016-06-22 18:58:38 makes sense _conceptually_ 2016-06-22 18:59:14 why singular? it’s de facto standard to use plural 2016-06-22 18:59:22 for resources that returns collections 2016-06-22 18:59:30 "standard" where ? 2016-06-22 18:59:35 de facto standard… 2016-06-22 18:59:40 hmmm 2016-06-22 18:59:48 most of the APIs uses plural 2016-06-22 18:59:54 and it makes the most sense when you think about it 2016-06-22 18:59:55 I dont't care 2016-06-22 19:00:08 what /packages returns? a single package? or multiple packages? 2016-06-22 19:00:23 unless it becomes FORMALIZED standard, singular is what I am doing 2016-06-22 19:00:40 there’s no formalized standard in RESTful APIs… 2016-06-22 19:01:02 there are just conventions that most of the best APIs follows 2016-06-22 19:01:10 exactly, so I use JSONAPI + singular 2016-06-22 19:01:39 JSONAPI also recommends plural 2016-06-22 19:01:43 look at their examples 2016-06-22 19:01:52 they have articles, people, … 2016-06-22 19:02:06 also relationships as a special word, not relationship 2016-06-22 19:02:34 when someone qualifies any collection/noun with a id or group it intuitively creates an impression of collection, 2016-06-22 19:02:37 and thus use plural does not make sense 2016-06-22 19:03:14 when I say mobile 2016-06-22 19:03:30 its intuitively creates an imprssion of ALL mobiles 2016-06-22 19:03:43 I am not sayin your mobile or mine 2016-06-22 19:03:58 and when I say /mobile/android 2016-06-22 19:04:17 it creates an impression AH! there must must be other types too 2016-06-22 19:04:35 so no need to pluralize here 2016-06-22 19:05:29 it goes well with how humans communicate 2016-06-22 19:05:48 api is all about easy communication 2016-06-22 19:06:15 avoid ambiguity 2016-06-22 19:06:44 using /package /packages means extra learning curve 2016-06-22 19:07:00 I am not creating a grammer book in my DOCs 2016-06-22 19:07:47 so singular for now 2016-06-22 19:07:50 most of the RESTful APIs uses plural, so you’re argument is invalid 2016-06-22 19:08:11 most thought earth was flat... till ! 2016-06-22 19:08:20 for example I’m just using GitHub API, also plurals https://developer.github.com/v3/ 2016-06-22 19:08:27 that is no argument, create Standard 2016-06-22 19:08:42 if you want a standard, then don’t use REST, but SOAP… 2016-06-22 19:08:56 again that is not argument 2016-06-22 19:09:08 there are plenty of standards around SOAP, you’ll be very happy 2016-06-22 19:09:28 most people thinks that there are too much standards, but maybe you’re not one of them 2016-06-22 19:09:31 if it conveys these "simple, consistent, meaningful, intuitive, somewhat-hierarchal, distinct, unambiguous" 2016-06-22 19:09:40 then its all I want for now 2016-06-22 19:09:51 can it also make coffee? 2016-06-22 19:09:59 and cook with olive oil? 2016-06-22 19:10:30 I am trying hard, it rejects to heatup 2016-06-22 19:10:53 SOAP and happy in the same sentence... wuuuut ? 2016-06-22 19:10:55 how are these words related to singular vs. plural? you’re just consciously breaking a widespread convention 2016-06-22 19:11:02 coredump: he wants STANDARDS 2016-06-22 19:11:22 jirutka: coredumb with a B 2016-06-22 19:11:22 please don't use /package unless it is actual slashpackage 2016-06-22 19:12:10 coredump: I hate SOAP, but maybe it’s better for people how needs standards to make something usable that will be familiar for others 2016-06-22 19:12:11 would agree for other verbs like edit/delete 2016-06-22 19:12:19 but "search" is ok 2016-06-22 19:12:24 jirutka: still coredumb with a B 2016-06-22 19:12:26 :D 2016-06-22 19:12:33 what B? 2016-06-22 19:12:39 you seem to not like it :( 2016-06-22 19:12:54 it's coredumB not coredumP 2016-06-22 19:13:16 you always do that it makes me sad you know :'( 2016-06-22 19:13:28 eh, sorry 2016-06-22 19:13:42 #ThatsTheJoke 2016-06-22 19:13:44 I should use a better client for IRC 2016-06-22 19:14:01 Adium doesn’t have autocomplete for names 2016-06-22 19:14:25 does it have a font that makes you tell the p's from the b's? 2016-06-22 19:14:31 because that's all you need :P 2016-06-22 19:14:32 :) 2016-06-22 19:14:38 lol 2016-06-22 19:15:15 pls get me packages/2, eh!!, do you want 2 packages or package with IDnum 2 2016-06-22 19:16:10 singular it stays, blow all trumpets ! 2016-06-22 19:17:10 singular, lowercase, simple-as-possible 2016-06-22 19:17:42 no lowerCamel case anywhere 2016-06-22 19:18:28 and confusion-inducing for people who already use /package 2016-06-22 19:18:41 granted, in a totally different context 2016-06-22 19:18:44 but still 2016-06-22 19:19:17 if it breaks you existing religious script, not acceptable 2016-06-22 19:19:22 ACTION is too lazy to scroll back 2016-06-22 19:19:37 we are moving with http2 and newer standards, update ! 2016-06-22 19:20:00 Is it Release The Trolls Day already? 2016-06-22 19:20:07 but summer has barely begun! :( 2016-06-22 19:20:09 and wait !! the api is still oblivious to them 2016-06-22 19:21:43 skarnet: it's always friday! 2016-06-22 19:21:53 I wish 2016-06-22 19:22:05 but if wanna wait http2 to get going in another decade, thats different story 2016-06-22 19:22:50 http1 took more than a decade, and crap http servers still cannot agree on how to handle headers 2016-06-22 19:23:07 irrespective of standards 2016-06-22 19:23:38 uff, GitHub Pulls Closer is online, pull requests in alpinelinux/aports will be automatically closed after merging, even when the commits are rebased and/or squashed :) 2016-06-22 19:23:38 https://github.com/jirutka/github-pulls-closer 2016-06-22 19:24:55 GitHub Pulls Closer sounds abusive 2016-06-22 19:25:04 what if I don't want it to pull me closer 2016-06-22 19:25:17 well, I can’t figure out a better name :/ 2016-06-22 19:25:23 but I’m open for suggestions :) 2016-06-22 19:25:35 agreed with skarnet 2016-06-22 19:25:35 well its a verb, cannot say ! 2016-06-22 19:25:51 btw any suggestions for better profile picture for Mr. Algitbot? https://github.com/algitbot : 2016-06-22 19:25:52 ) 2016-06-22 19:25:56 any second word that cannot be interpreted as a verb will do 2016-06-22 19:26:25 github-pr-closer for instance 2016-06-22 19:26:46 that was my first idea, but PR also means Public Relations… 2016-06-22 19:26:58 Public Relations Closer… not good 2016-06-22 19:27:08 it's github you're talking about 2016-06-22 19:27:17 not good is their bread and butter 2016-06-22 19:27:48 or simply github-pullrequests-closer? that's explicit and fine 2016-06-22 19:27:58 and also long as hell :) 2016-06-22 19:27:58 github-close-pulls 2016-06-22 19:28:09 github-closed-pulls 2016-06-22 19:28:17 hmm, that sounds indeed better; github-close-pulls 2016-06-22 19:28:26 closed 2016-06-22 19:28:30 longer is better 2016-06-22 19:28:38 most girls will tell you that 2016-06-22 19:28:42 LOL 2016-06-22 19:29:10 well, then we can name it github-webhook-handler-for-closing-pull-requets… 2016-06-22 19:29:23 I wouldn't mind 2016-06-22 19:29:28 as long as it's explicit 2016-06-22 19:29:45 of course, you'd always refer to it as ghwhhfcpr 2016-06-22 19:29:50 XD 2016-06-22 19:30:12 which is kinda the sound I make when I'm forced to use github 2016-06-22 19:34:12 :D 2016-06-22 19:36:47 hello 2016-06-22 19:38:22 I saw there's no php7-ssh2 package among the apk's, I'm looking now at the official ssh2 repo with v1.0 supporting php7. Question is, how could I add it in the Alpine apk's (build from source, deploy etc...) ? 2016-06-22 19:39:44 dminca: hi there 2016-06-22 19:39:58 there is a reference on APKBUILDs on wiki 2016-06-22 19:40:29 https://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package I mean this article 2016-06-22 19:40:37 ah oke, checking it. Thank you barthalion 2016-06-22 19:41:21 APKBUILDs for all official packages are here: https://github.com/alpinelinux/aports 2016-06-22 19:41:40 there are plenty of php modules, so you can base on these 2016-06-22 19:41:53 I was searching for apks in https://pkgs.alpinelinux.org/packages 2016-06-22 19:41:56 we accept patches either via github, mailing list or on IRC 2016-06-22 19:42:13 ah I see 2016-06-22 19:42:38 someone's usually around CET working hours if you need any further help 2016-06-22 19:46:04 the first real commit closed by algitbot https://github.com/alpinelinux/aports/pull/138#issuecomment-227855512 :) 2016-06-22 19:47:06 (y) 2016-06-22 19:47:17 \o/ 2016-06-22 19:50:34 would go and watch next episode of 'Ancient Aliens' while change code to use singular keywords 2016-06-22 19:50:42 and have tea 2016-06-22 19:59:09 why the pkgdesc="PHP extension couchbase" ? 2016-06-22 19:59:34 testing/php5-amqp: new aport 2016-06-22 20:01:02 people made mistakes 2016-06-23 02:29:20 barthalion: coreclr has this check for libuuid header: https://github.com/dotnet/coreclr/blob/af687f46/src/pal/src/configure.cmake#L37, it fails with `libuuid` package installed but passed when I installed `util-linux-dev`. 2016-06-23 02:29:50 I have updated the build instructions accordingly: https://gist.github.com/jasonwilliams200OK/7d6f5594d3bf697a27c9c1036d349fce, but just wanted to confirm if this is the smart thing to install plathora of linux utilities for just a header when we already have libuuid pacakge? 2016-06-23 02:30:18 Query: Currently speccing out for Alpine Linux Router/UTM, would this box be sufficient? https://www.olx.ph/item/intel-quad-core-pfsense-firewall-4-gigabit-lan-ID6Zo8E.html?p=3&h=5e4a8e15c9#5e4a8e15c9 2016-06-23 02:30:26 seems like we are probably missing libuuid-dev? 2016-06-23 02:30:47 And can Alpine Linux run on that? Ideally it should be capable of running the full DMVPN stack (OpenNHRP, quagga, etc.) 2016-06-23 02:31:25 For those in the know, this will be evaluated if it can be used as endpoint for data/VoIP in a SKE facility. 2016-06-23 04:19:41 peterj: yes, that's fine 2016-06-23 09:46:07 ncopa: could you please merge https://github.com/alpinelinux/aports/pull/140 (main/qemu-openrc: upgrade to 0.4.1)? 2016-06-23 10:26:07 <^7heo> Hey 2016-06-23 10:26:26 <^7heo> irssi behaves weird. 2016-06-23 10:26:33 <^7heo> /script doesn't work 2016-06-23 10:26:48 <^7heo> it prints: -!- Irssi: Unknown command: script 2016-06-23 10:27:00 <^7heo> I have installed irssi-perl ofc. 2016-06-23 10:27:04 <^7heo> and I have restarted irssi. 2016-06-23 10:27:12 <^7heo> and it still does that... 2016-06-23 10:27:16 <^7heo> any hint? 2016-06-23 10:32:50 ^7heo: are there any option to print compile flags? 2016-06-23 10:33:22 <^7heo> nah, it's apparently that I was missing /load perl 2016-06-23 10:33:30 <^7heo> but... shouldn't it load perl by default if it is there? 2016-06-23 10:53:32 <^7heo> ncopa: sorry, I had to quit a few times 2016-06-23 10:53:35 <^7heo> what did you write? 2016-06-23 11:45:01 ^7heo: yeah, I've been there, but haven't investigated origin behind lack of automation. I keep that kind of notes on my wiki page: https://wiki.alpinelinux.org/wiki/User:Przemoc/Notes (there aren't many more, though) 2016-06-23 11:52:25 <^7heo> przemoc: for now, my main issue isn't the lack of automation, it's bugs. 2016-06-23 11:52:31 <^7heo> przemoc: such as geary not working. 2016-06-23 11:58:05 sorry, cannot help you there as I'm not using X in alpine. (and I'm generally rather avoiding looking under the hood of glib-based stuff.) 2016-06-23 12:00:22 <^7heo> yeah, same. 2016-06-23 12:00:35 <^7heo> but I don't know any sane MUA anyway. 2016-06-23 12:00:39 <^7heo> so I take a simple one. 2016-06-23 12:00:39 ^7heo: geary? what is geary? 2016-06-23 12:00:44 <^7heo> jirutka1: a MUA. 2016-06-23 12:01:20 <^7heo> but, the problem behind it is that there is no unicodesn tokenizer support with alpine's SQLite. 2016-06-23 12:01:28 <^7heo> not geary in itself. 2016-06-23 12:02:42 <^7heo> $ sqlite3 -cmd '.load unicodesn.sqlext' geary.db 2016-06-23 12:02:45 <^7heo> works ^ 2016-06-23 12:03:11 <^7heo> but then, creating the table with: tokenize=unicodesn "stemmer=english", gives: Error: unknown tokenizer: unicodesn 2016-06-23 12:35:09 <^7heo> what's the page to find alpine linux packages by content already? 2016-06-23 12:35:20 https://pkgs.alpinelinux.org/contents 2016-06-23 12:35:50 <^7heo> thanks jirutka1 2016-06-23 12:36:55 <^7heo> damn, dbus is a pain 2016-06-23 12:37:14 <^7heo> geary requires gnome-keyring. 2016-06-23 12:37:15 <^7heo> dafuq. 2016-06-23 12:40:59 <^7heo> Well, geary works well enough, now that I have installed gnome-keyring and dbux-x11 2016-06-23 12:41:07 <^7heo> missing a few icons but that's all 2016-06-23 12:46:26 why not mutt? 2016-06-23 13:24:14 <^7heo> I hate mutt with all my guts 2016-06-23 13:24:20 <^7heo> but, I'll try it again 2016-06-23 13:24:22 <^7heo> and tell you why ;) 2016-06-23 13:47:15 <^7heo> Okay; here's why 2016-06-23 13:47:22 <^7heo> I'm 20 minutes into the setup of that bs program 2016-06-23 13:47:25 <^7heo> I installed 3 other programs 2016-06-23 13:47:31 <^7heo> and I still can't read my emails. 2016-06-23 13:48:31 you hate the kiss principle, not mut 2016-06-23 13:49:06 <^7heo> wtf? 2016-06-23 13:49:11 <^7heo> do you even know what you're talking about? 2016-06-23 13:49:27 yes 2016-06-23 13:49:30 <^7heo> installing 4 programs, writing 4 configuration files, and not having it working; you call that kiss? 2016-06-23 13:49:38 <^7heo> you'd better reboot yourself, dude. 2016-06-23 13:49:48 <^7heo> your memory is obviously corrupt. 2016-06-23 13:50:04 every tool does one thing, and does it well instead off one tool doing everything badly 2016-06-23 13:50:19 <^7heo> for your information, kiss means "keep it stupid simple" not "keep it stupidly separated" 2016-06-23 13:50:34 <^7heo> mail is one thing. 2016-06-23 13:50:43 <^7heo> no matter what I'm using. 2016-06-23 13:50:44 you want to perform a complex operation. 2016-06-23 13:50:50 <^7heo> I dont' want a tool to add 'line endings' to the stream 2016-06-23 13:50:55 <^7heo> another tool to talk imap 2016-06-23 13:50:58 <^7heo> another tool to write my name 2016-06-23 13:51:03 <^7heo> another tool to write my password 2016-06-23 13:51:10 <^7heo> another tool to write ... 2016-06-23 13:51:19 <^7heo> (I take that you got it now) 2016-06-23 13:51:33 <^7heo> ideally 2016-06-23 13:51:34 aah, passwords, yeah, i want them compartmentalized 2016-06-23 13:52:05 and yes i want one program to send, and one to receive mails. and one to read them, and one to edit them (in my editor, not some gui component) 2016-06-23 13:52:11 <^7heo> I would have a tool that creates a virtual filesystem reflecting my emails. 2016-06-23 13:52:27 <^7heo> and I could do all the rest with the binaries which are coming with my userland. 2016-06-23 13:52:30 <^7heo> done. 2016-06-23 13:52:34 <^7heo> THAT would be kiss. 2016-06-23 13:52:43 <^7heo> and it doesn't have to be complex to do that. 2016-06-23 13:52:55 the original mail(1) also is just a mua 2016-06-23 13:53:09 <^7heo> write data into a file on that virtual fs, done, you've sent your email 2016-06-23 13:53:12 not a mua+mta+mda+passwordstore+etc 2016-06-23 13:53:30 but i agree on a mail filesystem. maildirs are pretty good. 2016-06-23 13:53:45 <^7heo> in the end 2016-06-23 13:54:02 <^7heo> those programs aren't even good, they are full of ncurse when not needed 2016-06-23 13:54:23 <^7heo> (esp. mutt) 2016-06-23 13:54:32 <^7heo> I need curses for my editor. 2016-06-23 13:54:36 <^7heo> *IF* I want so. 2016-06-23 13:54:39 <^7heo> and that's about it 2016-06-23 13:58:03 i think you could do that with mailx(1) 2016-06-23 14:04:05 <^7heo> yeah possibly. 2016-06-23 14:05:08 <^7heo> but honestly, I dunno why it doesn't work; and that's annoying 2016-06-23 14:09:15 <^7heo> and again, kiss is something that does simple things, tells you how, why, and when. 2016-06-23 14:09:39 <^7heo> from those tools (fetchmail, procmail, msmtp and mutt) I don't see ANY tool that behaves that way. 2016-06-23 14:09:49 <^7heo> it's all automagic and weird. 2016-06-23 14:34:50 ^7heo: anything involving user interaction will be complex, there's no way around it. Because it intertwines policy and mechanism without any good way of separating the two. 2016-06-23 14:35:35 <^7heo> well, mounting a virtual file system by providing the parameters necessary for reaching the mail 2016-06-23 14:35:40 <^7heo> that would be a good way to start. 2016-06-23 14:35:50 <^7heo> and then operating on that file system would be a matter of having normal userland tools. 2016-06-23 14:36:20 so yes, a MUA will be complex. Either it leaves to you, the user, as many policy decisions as it can, in which case configuration is complex. Or it decides for you, and you get smartphone apps that you hate because they think they're cleverer than you. 2016-06-23 14:36:53 the complexity of a MUA has nothing to do with ways to reach the mail. Maildirs are simple. 2016-06-23 14:37:54 If you want to unify an interface for local folders and imap folders, then you're doing two things. 2016-06-23 14:38:01 Of course it's going to be complex. 2016-06-23 14:38:16 in particular, it involves the network. The network sucks. 2016-06-23 14:38:36 And the classic e-mail infrastructure and protocols suck even more. 2016-06-23 14:50:39 hey :) 2016-06-23 14:50:44 how is everyone? 2016-06-23 14:51:40 ncopa: i had to think about you while reading this: http://flatpak.org/ 2016-06-23 14:51:48 maybe this helps you with your docker thing 2016-06-23 14:52:06 personally it gave me a heart attack, but maybe its helpful for you 2016-06-23 14:52:44 skarnet: if you dont want to ruin your evening, dont click the link *g* 2016-06-23 14:53:49 lol 2016-06-23 14:54:11 getting a bit jaded about that kind of thing... it's only the n+1th attempt 2016-06-23 14:54:51 its a messy container approach again ... where everything ships there own copy of libs ... 2016-06-23 14:55:01 and therefore it makes stuff hell of a lot more insecure! 2016-06-23 14:55:11 application packagers are like social networks: they all suck, and one day for no reason at all, one of them gets traction and is successful 2016-06-23 14:55:24 and a few years later, people start realizing that there are security issues 2016-06-23 14:55:31 hehe, indeed 2016-06-23 14:55:51 i mean i understand why big companies want to fuce push those things 2016-06-23 14:56:06 its sooo easy to hide bad code in there a have a fucked up build process 2016-06-23 14:56:11 "doing things easily" is more important for them than "doing things right" 2016-06-23 14:57:46 hehe 2016-06-23 14:57:49 sadly yes 2016-06-23 14:57:58 but i guess big companies are always like that 2016-06-23 15:09:45 i think systemd has a solution for that too 2016-06-23 15:11:19 ncopa: its both red hat ... so i would not be surprised if they merge at some point 2016-06-23 15:11:51 lol 2016-06-23 15:12:29 spf = systemd package format 2016-06-23 15:12:31 *g* 2016-06-23 15:12:47 Red Hat is so schizophrenic it has to develop several solutions to a problem that distributions (i.e. the thing they're supposed to be) is meant to solve in the first place 2016-06-23 15:13:56 refusing that job offer from Red Hat 16 years ago: best decision I ever made :P 2016-06-23 15:14:17 hehe, indeed! 2016-06-23 15:52:26 ScrumpyJack: you got pinged in #opensmtpd 2016-06-23 15:52:27 ;) 2016-06-23 15:52:31 just a reminder *g* 2016-06-23 16:23:46 ncopa: pidgin 2.11 is available with some security fixes 2016-06-23 16:24:41 leo-unglaub: send a pr then 2016-06-23 16:26:34 scadu: the patch is larger than the ping in irc 2016-06-23 16:26:36 http://dpaste.com/3DRXPP6.txt 2016-06-23 17:31:19 leo-unglaub: providing simple patch is faster that poking someone 2016-06-23 17:31:24 also, update checksums 8) 2016-06-23 17:32:14 not really 2016-06-23 17:32:24 because i have no packaging environment on my laptop 2016-06-23 17:32:36 :[ 2016-06-23 17:34:01 leo-unglaub: embrace containers 2016-06-23 17:34:05 it takes about 15 minutes 2016-06-23 17:34:27 also I update checksums directly on my laptop OS, without development env 2016-06-23 17:35:03 nah, i dont want that shit on my computer 2016-06-23 17:35:39 XD 2016-06-23 17:46:25 leo-unglaub: "do as I say, don't do as I do" 2016-06-23 17:49:20 <^7heo> skarnet: a shorter version is "do as I say, not as I do" 2016-06-23 17:50:02 it may be shorter, but it's not canonical 2016-06-23 17:51:39 leo-unglaub: well, looks like you are the problem then )) 2016-06-23 17:53:21 if not using containers is wrong ... then i am proud of beeing wrong 2016-06-23 17:54:16 <^7heo> leo-unglaub: not using containers is right, if I may. 2016-06-23 17:55:09 I love how the number of proud people here multiplied 5 or 6 times since my first commit =) 2016-06-23 17:55:55 I'm proud no matter the issue and my position or absence of position on the issue, does that count? 2016-06-23 17:57:12 leo-unglaub: barthalion provided you with the simplest and quickest solution to just simply bump this package, rather than pinging people around to do this. 2016-06-23 17:57:51 I don't know what is the problem. If this solution is not suitable for you then choose another one. 2016-06-23 17:57:53 kthxbye 2016-06-23 17:58:13 there is no problem 2016-06-23 17:58:22 i informed the maintainer that there is a new version with security fixed in it 2016-06-23 17:58:30 cool. 2016-06-23 17:58:44 maybe i am oldschool, but i provide patches only for things a developer cannot do themself quickly 2016-06-23 17:59:01 every maintainer around can bump a package faster themself than he can pull and merge a patch 2016-06-23 17:59:08 if i had a buxfix, i would submit the patch 2016-06-23 17:59:22 but please stop with all the merging and pull requests for a single letter replacement 2016-06-23 17:59:34 what? you don't want to do the maintainer's job for them? wtf are you doing reporting bugs in a distro channel XD 2016-06-23 17:59:35 i dont care if i am listed in the git log or not 2016-06-23 18:04:06 skarnet: this reminds me of one guy in one of my projects sending me a pull request that changed one type. there was litterally 1 character changed. we did this pull request thing on github and then i did not use the gui to merge it. i simply did the change myself and pushed and he totally enraged because i did not merge it and it was not his commit in the timeline and in the github statistics *g* 2016-06-23 18:05:01 you should have replied "your patch was not optimal, I had a few changes to make to it" 2016-06-23 18:06:02 leo-unglaub: well, maybe you are right. ncopa would do it quicker if he were here. 2016-06-23 18:06:29 scadu: dont get me wrong, if i had changed something i had debugged i of course would have send a diff 2016-06-23 18:08:13 leo-unglaub: I'm idle. what do you want done? 2016-06-23 18:09:22 ScrumpyJack: ? you mean about your ping before? 2016-06-23 18:09:31 i was just reading your question in #opensmtpd 2016-06-23 18:09:40 and interrested in the return code thing as well 2016-06-23 18:10:36 no, i meant pidgin. just testing build on package bump 2016-06-23 18:11:08 ah, about pidgin. the new version contains a lot of security fixed found by an intel employeed if i remember corrcetly 2016-06-23 18:11:33 and the php7 port also got released a couple of minutes ago ... some security problems in there as well 2016-06-23 18:27:23 leo-unglaub: do you have a CVE or something for the commit message? (no browser here) 2016-06-23 18:27:36 ScrumpyJack: pidgin or php? 2016-06-23 18:27:42 pidgin 2016-06-23 18:28:08 yes, sure 2016-06-23 18:29:38 TALOS-CAN-0120, TALOS-CAN-0140, TALOS-CAN-0138, TALOS-CAN-0135, TALOS-CAN-0118, TALOS-CAN-0119, TALOS-CAN-0123, TALOS-CAN-0128, TALOS-CAN-0133, TALOS-CAN-0134, TALOS-CAN-0136, TALOS-CAN-0137, TALOS-CAN-0142, TALOS-CAN-0141, TALOS-CAN-0139, TALOS-CAN-0143 2016-06-23 18:29:52 as i sayed, there are a lot of fixes :) 2016-06-23 18:29:53 hehehe 2016-06-23 18:29:54 Talos can a lot of things 2016-06-23 18:30:12 what a man 2016-06-23 18:30:20 sod that. 2016-06-23 18:31:18 :) 2016-06-23 18:31:25 done 2016-06-23 18:31:30 awesome! 2016-06-23 18:31:32 thanks! 2016-06-23 18:31:57 i don't have push so you'll need to bug someone else for that. 2016-06-23 18:34:07 does anyone has some experience with remote console for HP Proliant servrers? 2016-06-23 18:34:47 jirutka: nope 2016-06-23 18:34:52 i.e. how to use it from OS X or Linux without Java applet 2016-06-23 18:36:53 well, he left 2016-06-23 18:36:57 no cookies for him 2016-06-23 18:37:06 but the funny answer is, you can't 2016-06-23 18:37:15 can I get the cookies instead? 2016-06-23 18:37:33 sure, be my guest 2016-06-23 18:37:41 thanks! 2016-06-23 18:37:47 ACTION stuffs himself with cookies 2016-06-23 18:37:52 for Windows you need terribly outdatated java 6 and firefox 22 or 24 2016-06-23 18:38:04 on OS X any java/fx combo will do 2016-06-23 18:38:15 irclogger_com: enjoy my monologue 2016-06-23 18:38:48 those applets are horrible 2016-06-23 18:38:58 my datacenter has one of them as well for remote maintainance 2016-06-23 18:39:05 every time a big pain in the ***** 2016-06-23 18:45:47 a question about the libpng package 2016-06-23 18:46:22 i think we can remove the libpng.la file and i am also not sure if we should extract the pngfix binary from it 2016-06-23 18:46:40 because its technically not part of the lib and a standalone tool 2016-06-23 20:00:26 oh no ! my api in under the spell of ReWrite, 2016-06-23 20:00:26 api.a.o/packages is same as api.a.o?_url=/packages 2016-06-23 20:00:26 if admin decides to remove the spell will the api become unrestFul? 2016-06-23 20:00:55 what? 2016-06-23 20:02:28 api.a.o/packages is same as api.a.o?_url=/packages 2016-06-23 20:09:19 just don’t use the second one… 2016-06-23 20:10:37 ahem cough ! ok, would bribe the admin with beer now and then 2016-06-23 20:11:49 so he does not switch server or remove the spell 2016-06-24 08:19:29 anyone has an idea about #5768 ? 2016-06-24 08:19:33 /usr/bin/qemu-system-x86_64 wants MPROTECT disabled on /lib/ld-musl-x86_64.so.1 2016-06-24 08:19:40 smells of pax 2016-06-24 08:23:01 nope 2016-06-24 08:48:20 just pushed fix 2016-06-24 08:48:24 (i think) 2016-06-24 08:51:11 you did 2016-06-24 08:51:50 it works 2016-06-24 09:33:35 skarnet: you about dude? 2016-06-24 09:34:00 what do you need? 2016-06-24 09:38:05 if you have a spare moment, would you be kind enough to respond to http://forum.alpinelinux.org/forum/general-discussion/how-change-init-system-s6-instead-openrc ? It might help others too 2016-06-24 09:39:28 i mean with a "coming soon to an alpine near you" sort of thing - as the s6 author - if that makes sense 2016-06-24 09:45:45 uhhh... first I would need to register and that hurts my religion 2016-06-24 09:46:05 (I have already sinned several times, but try to limit the amount of sin) 2016-06-24 09:46:21 I'll post a blurb for you 2016-06-24 09:47:36 then I don't like to respond to broken English messages that ask a question while making it sound like the answer is easy when it's not, and say "I want easy". Why not "kthxbye" while they're at it 2016-06-24 09:48:05 and then, what would I answer? "please come back in a year and you'll be able to do it" ? 2016-06-24 09:48:24 whatever you feel is appropiate 2016-06-24 09:49:22 "English motherfucker, do you speak it?" 2016-06-24 09:49:47 not sure if it's appropriate, but it's what I feel like answering :P 2016-06-24 09:49:55 ACTION moves on 2016-06-24 09:50:14 ask me anytime when you need help with something! 2016-06-24 09:52:42 what i meant was, you don't seem in the best of moods today, so best probably best i left you to it ... but i could be wrong 2016-06-24 09:55:34 It's just that some messages - especially on web forums - have a tendency to rub me the wrong way, and I feel it's best for me to ignore them. 2016-06-24 09:57:12 But please, if you feel it's useful/important/... to answer it, by all means do so; the heart of the answer is "it's really difficult and we'll be working to make it easy next year". 2016-06-24 09:57:19 lemme pm you 2016-06-24 10:38:47 ncopa: thank you for supporting and closing the tty ticket 2016-06-24 10:39:34 http://bugs.alpinelinux.org/issues/5759 2016-06-24 10:44:20 thank you for helping fix it 2016-06-24 11:19:43 dont mention 2016-06-24 11:34:31 why dont we have opencv in alpine apks? 2016-06-24 11:42:09 maybe nobody asked for it? 2016-06-24 11:44:23 hmmmm 2016-06-24 11:44:33 can i ask for it? 2016-06-24 12:18:22 sure 2016-06-24 12:22:24 http://bugs.alpinelinux.org/issues/5795 2016-06-24 12:30:15 it will not be for 3.4.1 2016-06-24 12:55:26 that is alrite 2016-06-24 16:25:12 there’s a security upgrade of main/php5 ready to be merged: https://github.com/alpinelinux/aports/pull/142 2016-06-24 18:05:49 jirutka: ideas for good place with awesome knedliky in the centre of Prague? 2016-06-24 18:16:18 barthalion: hmm, I don’t like Czech cuisine too much, so I dunno 2016-06-24 18:17:42 well, it's pretty much like Polish, but we want to check if your dumplings are different from ours 2016-06-24 18:19:06 barthalion: http://www.restauraceandel.cz/en/ is quite good, but it’s quite a long time since I’ve been there 2016-06-24 18:20:35 thanks, we'll check it :) 2016-06-24 18:21:26 barthalion: this one is also quite favourite http://www.cestacasem.cz/ 2016-06-24 18:27:09 barthalion: not sure about this one, I don’t remember names, but if I’m correct, then this one should be very good: http://www.upinkasu.com/ 2016-06-25 23:42:29 barthalion: i guess this issue can be closed as resolved: https://bugs.alpinelinux.org/issues/4966 :) 2016-06-26 19:14:29 so as a note, the 3.4 iso appears to fail booting on a usb stick in the same way as the vmware post to alpine-devel notes 2016-06-26 19:15:07 i bought a skull canyon nuc and tried to boot it with alpine linux 3.4 and that failed, i'll try 3.3 in a bit 2016-06-26 19:15:36 that said, having 2x nvme drives makes for a little speedy machine 2016-06-27 06:57:49 morning all 2016-06-27 06:58:27 i finally got through my $ rush weeks. so i'm back to my regular timetable regarding AL development. 2016-06-27 07:46:20 splendid news :) 2016-06-27 07:47:33 so i can bug you for things like http://patchwork.alpinelinux.org/patch/2134/ ? :) 2016-06-27 08:45:42 hey :) 2016-06-27 08:45:47 how is everyone? 2016-06-27 08:57:07 hi 2016-06-27 08:57:49 good. how are you leo-unglaub? 2016-06-27 08:58:27 thats nice to hear. i am okay. i hurt my hand during work so typing is a litle bit hard 2016-06-27 08:58:44 ncopa: there is a bunch of stuff in testing that can be moved out. what's the general consensus behind chosing main or community? 2016-06-27 09:02:08 ScrumpyJack: for things we want maintain for 2 years: move to main 2016-06-27 09:02:35 for things we are ok maintain just for 6months - 1 year, we move to community 2016-06-27 09:02:48 things in main cannot depend on anything in community 2016-06-27 09:03:15 so unless you have reasons to want 2 years support, move it to community 2016-06-27 09:09:31 morning 2016-06-27 09:14:10 i think community is in most cases the better location 2016-06-27 09:14:22 main should be as small as possible 2016-06-27 09:14:25 +1 2016-06-27 09:16:09 yeah, something I was crying about for past few releases - unless you find something crucial, it belongs to community 2016-06-27 09:18:22 ok 2016-06-27 09:24:52 >>> go-cross*: Tracing dependencies... 2016-06-27 09:24:53 >>> go-cross*: Package size: 1.2 GB 2016-06-27 09:25:00 go is 1.2GB 2016-06-27 09:25:11 its insane 2016-06-27 09:26:19 well, it's for cross compilation, you can't get that for free 2016-06-27 09:26:34 there are many other insane things about this language 8) 2016-06-27 09:27:00 they have some good ideas 2016-06-27 09:27:05 but the implementation 2016-06-27 09:27:15 i suppose we should get rust built 2016-06-27 09:28:31 +10000000000000000000000000000000000000000000000 2016-06-27 09:28:38 :) 2016-06-27 09:28:40 go is typical google bloat ... 2016-06-27 09:28:55 "bloat" is understatement 2016-06-27 09:28:58 they have excelent ideas ... but its soo huge that no one can validate it or check it 2016-06-27 09:29:22 but yeah, rust woud be the way to go *g* 2016-06-27 09:29:36 i think it should be possible to build it on alpine now 2016-06-27 09:29:40 with the latest fixes 2016-06-27 09:29:44 ok? 2016-06-27 09:29:58 is itpossible to bootstrap it now? 2016-06-27 09:31:47 according to one dev it should 2016-06-27 09:31:55 but i did not try it myself 2016-06-27 10:07:03 i just installed a new network card to give my lxc container a own interface 2016-06-27 10:07:19 then i have to configure it to send git patches 2016-06-27 10:07:23 brb 2016-06-27 10:11:11 you could just bridge it or put it behind veth 2016-06-27 10:11:51 nah, not possible in my setup ... i have there already some wired stuff configured 2016-06-27 10:12:29 sure, if you have possibility to add new network card 2016-06-27 10:12:51 I'm laptop-only for past 3 years, so this wasn't really an option to me 2016-06-27 10:13:30 hehe 2016-06-27 10:13:36 i love my two big towers :) 2016-06-27 10:14:05 i have two of them :) http://gaming.coolermaster.com/en/products/cases/Trooper/ 2016-06-27 10:14:20 guys, re #5805, any suggestions on how to fix? 2016-06-27 10:14:49 12:12 (root@2ua4020qdl) ~# ldd /usr/sbin/arping 2016-06-27 10:14:50 /lib/ld-musl-x86_64.so.1 (0x706a3d1b1000) 2016-06-27 10:14:50 libpcap.so.1 => /usr/lib/libpcap.so.1 (0x706a3cd6d000) 2016-06-27 10:14:50 libnet.so.1 => /usr/lib/libnet.so.1 (0x706a3cb55000) 2016-06-27 10:14:50 libc.musl-x86_64.so.1 => /lib/ld-musl-x86_64.so.1 (0x706a3d1b1000) 2016-06-27 10:15:20 LIBNET_LINK <- is this function missing? 2016-06-27 10:17:59 ScrumpyJack: where did you post the pidgin upgrade? 2016-06-27 10:18:08 because it is still not in edge 2016-06-27 10:26:19 fcolista: configure script fails to detect link-layer method 2016-06-27 10:26:26 due to linux-headers missing 2016-06-27 10:26:35 aha 2016-06-27 10:26:37 there is a configure option: --with-link-layer=linux 2016-06-27 10:26:38 Linker.........: -Wl,--as-needed -lpcap -lnet -lrt -lm 2016-06-27 10:26:40 should do it 2016-06-27 10:26:52 no, i dont think its a linker 2016-06-27 10:27:04 its for network link layer (layer 2) 2016-06-27 10:27:04 i was thinking that was a linker issue 2016-06-27 10:27:17 ok 2016-06-27 10:27:19 got it 2016-06-27 10:27:49 seems like just --with-link-layer=linux might work 2016-06-27 10:28:06 where did you find ? 2016-06-27 10:28:14 i looked in configure.in 2016-06-27 10:28:32 arping 2.16 ? 2016-06-27 10:28:40 no 2016-06-27 10:28:43 sorry 2016-06-27 10:28:47 its libnet 2016-06-27 10:28:51 ah... 2016-06-27 10:28:53 arping just uses libnet 2016-06-27 10:28:59 yes. 2016-06-27 10:29:03 the error message indicates that its libnet problem 2016-06-27 10:29:23 libnet is used also from other sw 2016-06-27 10:29:35 likely, yes 2016-06-27 10:29:48 this should fix it: http://tpaste.us/G590 2016-06-27 10:29:54 so we might expect the same issue with rsyslog, sambaa, netfilter etc 2016-06-27 10:30:07 anything that needs link-layer functionality yes 2016-06-27 10:30:44 fcolista: feel free to fix libnet if you wan 2016-06-27 10:30:54 i can cherry-pick it for 3.4-stable afterwards 2016-06-27 10:31:17 yes, doing it right now 2016-06-27 10:31:18 i think we can sneak in the fix for 3.4.1 2016-06-27 10:32:25 pushed 2016-06-27 10:32:53 leo-unglaub: the pidgin upgrade is in the patchwork queue 2016-06-27 10:34:02 thx 2016-06-27 10:42:15 does APKBUILD support conflicts option? 2016-06-27 10:42:41 fcolista, depends="!conflict" 2016-06-27 10:43:00 if $package conflicts with $package2 i mean 2016-06-27 10:43:16 e.g.: arping and iputils conflicts 2016-06-27 10:43:34 arping/APKBUILD: depends="!iputils" 2016-06-27 10:43:44 aha 2016-06-27 10:43:45 k 2016-06-27 10:49:13 ncopa: could you please look at https://github.com/alpinelinux/aports/pulls?q=is%3Aopen+is%3Apr+label%3Amain ? 2016-06-27 10:52:08 fcolista: thanks for fixing arping 2016-06-27 10:54:56 ScrumpyJack, nm..ncopa found the solution :) 2016-06-27 10:55:16 afk 2016-06-27 11:34:44 does patchwork 2146 make sense? 2016-06-27 11:36:48 fabled: would you have any idea about #5773? 2016-06-27 12:04:06 ScrumpyJack, does "/etc/init.d/hwdrivers start" help? 2016-06-27 12:46:29 nope :( has already been started at boot since day one 2016-06-27 12:52:21 do the overlays work for you? can you do ls /sys/kernel/config/device-tree/overlays/ 2016-06-27 12:52:56 something like: apk add raspberrypi && /opt/vc/bin/dtoverlay -d /media/mmcblk0p1/overlays/ sdio 2016-06-27 12:53:29 should fail 2016-06-27 13:03:45 ncopa thats about the same size as ghc cross compiler :) 2016-06-27 13:04:02 ugh 2016-06-27 13:06:35 though for ghc the reason is there are 3 different runtimes all built, and debug versions of each and profiled, and etc... 2016-06-27 13:06:56 ghc-8.0.1-r0 installed size: 2016-06-27 13:06:56 1169625088 2016-06-27 13:07:01 ugh 2016-06-27 13:07:19 so i finally got llvm 3.7 to build and install so llvm-system and llvm3.7 work (for ghc at least) 2016-06-27 13:07:42 nice! 2016-06-27 13:07:58 it works in the sense that ghc works with it, i can't promise its "right" 2016-06-27 13:08:30 i haven't tried for example to build julia with it yet 2016-06-27 13:08:31 well, if ghc works it is at least not totally broken/useless 2016-06-27 13:10:05 yep, so i'll clean up the apkbuild (took longer than i planned) and push that to testing maybe? 2016-06-27 13:10:09 or patch for testing rather 2016-06-27 13:29:27 ncopa: I’m back, what about https://github.com/alpinelinux/aports/pull/145? I see that I forgot to bump pkgrev, I’ll fix it in a second, any other issues? 2016-06-27 13:30:51 are you sure? 2016-06-27 13:30:57 about what? 2016-06-27 13:31:10 are you sure that kernel devtmpfs create them as net/* 2016-06-27 13:31:48 netdev is correct group 2016-06-27 13:31:51 so that part is ok 2016-06-27 13:32:41 I can confirm that on Gentoo Hardened 3.18.9 and OpenVZ container with kernel 2.6.32-042stab113.11 (RHEL) running Alpine 2016-06-27 13:32:56 I haven’t tested it on Alpine kernel, but it should be the same 2016-06-27 13:33:26 …and also hardened 4.1.6 2016-06-27 13:33:39 hmm, why I have such old kernel on that hypervizor… 2016-06-27 13:34:14 I’ll try it on Alpine kernel today, I’m just installing it 2016-06-27 13:36:08 i verified it 2016-06-27 13:36:11 its correct 2016-06-27 13:36:45 good 2016-06-27 13:37:12 :) 2016-06-27 13:38:00 thanks! 2016-06-27 13:38:44 i'll cherry-pick that for 3.4.1 2016-06-27 13:39:06 i hoped for v3.4.1 release today but then kernel 4.4.14 came out 2016-06-27 13:40:03 I wonder if we should also automate testing patches from ml 2016-06-27 13:40:31 for example, ML -> github "translation", so travis is re-used 2016-06-27 13:40:57 probably more complicated than just applying patch locally by bot though 2016-06-27 13:41:10 I was already thinking about that, but not sure if it’s a good idea to mirror patches from ML to GH pull requests 2016-06-27 13:41:37 the discussion could be kept on mailing list 2016-06-27 13:42:35 the problem is that when we mirror patches from ML, what if someone add comments to these pull requests? it may be very confusing 2016-06-27 13:42:53 well, these would be meaningless 2016-06-27 13:43:13 the bot could put the description of pull request that it's just for build test purporses 2016-06-27 13:43:16 purposes even 2016-06-27 13:43:17 meanmeaningless? 2016-06-27 13:43:54 I mean, with that description and let's say a label, we could ignore such comments with clear conscience 2016-06-27 13:44:04 another option would be just create a branch with the applied patches and let it test on Travis, without creating a PR 2016-06-27 13:44:39 yes, but that doesn't give clear overview which patches are good to merge 2016-06-27 13:44:42 (IMHO) 2016-06-27 13:44:42 it would be awesome to synchronize comments between GH and ML :) 2016-06-27 13:45:57 and youtube comments while you're at it 2016-06-27 13:46:12 that’s obviously nonsense 2016-06-27 13:47:41 no, comment synchronization is not something I care about 2016-06-27 13:47:59 automated testing is awesome 2016-06-27 13:48:09 if patch is proposed on mailing list, discussion belongs there 2016-06-27 13:48:17 code-review is also awesome… 2016-06-27 13:48:36 what about reddit 2016-06-27 13:48:45 skarnet, you’re not helping 2016-06-27 13:49:02 we don’t accept patches via reddit, do we? 2016-06-27 13:50:30 I'm not helping? You're making me very sad. I thought you liked me. 2016-06-27 13:50:37 code review is done on mailing list as well 2016-06-27 13:50:42 you don't need github for that 2016-06-27 13:51:44 barthalion: GH is much more convenient for code-review, you can add comments directly to the lines in patches 2016-06-27 13:51:56 I can do the same using e-mail… 2016-06-27 13:52:07 show me… 2016-06-27 13:52:23 sure, go to alpine-aports and see any mine or ncopa's message 2016-06-27 13:54:58 okay, how do you track what issues has been fixed? ;) 2016-06-27 13:55:33 I compare patches? 2016-06-27 13:55:48 really, I thought we're already past this pointless topic 2016-06-27 13:56:06 I know that you can copy&paste patch, add comments between etc. but it’s really very inconvenient 2016-06-27 13:57:37 we already have better tools for code-review than plain mails; and even GH is far from ideal for code-reviews, it quickly become messy in bigger pull requests with a lot of comments (but still way better than messing with mails) 2016-06-27 13:57:43 okay 2016-06-27 13:59:15 skarnet: yes, I do, but not when you start complaining against making things more user friendly :) 2016-06-27 13:59:46 or you could use the google method https://github.com/google/git-appraise 2016-06-27 14:00:07 oh no, yet another crap written in Go… 2016-06-27 14:00:20 okay, when I have some time I’ll think about integration between ML and GH for automated testing 2016-06-27 14:02:22 maybe I’ll wait for someone to take good idea from git-appraise and implement them in simple way and in some sane language 2016-06-27 14:02:50 hm, I think that it’s just based on git notes (https://git-scm.com/docs/git-notes), isn’t it? 2016-06-27 14:03:51 yes 2016-06-27 14:04:00 its just adding lines into git-notes 2016-06-27 14:04:07 and using that to synchronize state 2016-06-27 14:05:49 and yet they needed 2k5 LOC because Google and Go… 2016-06-27 14:12:20 well it exists at least, note i'm not a go fan but it is a working example of something that say fossil has 2016-06-27 14:36:00 according to http://patchwork-freedesktop.readthedocs.io/en/latest/rest.html patchwork provides a RESTful API, but http://patchwork.alpinelinux.org/api/1.0/projects returns 404… what is the correct address? 2016-06-27 15:13:16 We don't use freedesktop fork 2016-06-27 15:13:29 jirutka: ^ 2016-06-27 15:13:50 aha, pardon, I’ve overlooked that it’s a fork 2016-06-27 15:13:59 so just RPC interface? 2016-06-27 15:19:14 Afaik yes 2016-06-27 15:19:59 uh, well… but still better than SOAP :) 2016-06-27 15:44:26 jirutka: there's a character that you type often and that I can't see, I suspect it's "..." in just one character. What is that character and what encoding are you using? It's possible, even likely that you're simply using UTF-8 and my client sucks, but I'd like to be sure. 2016-06-27 15:46:03 skarnet: yes, it’s probably elipsis that is a single character …; often incorrectly written as ... (three dots, that’s not the same) 2016-06-27 15:46:11 jirutka: we use patchwork.alpinelinux.org to track patches 2016-06-27 15:46:32 skarnet: I hope that my client sends UTF-8 2016-06-27 15:46:36 ncopa: I know 2016-06-27 15:46:44 commenting lines in email is what linux kernel devs has done for a decade 2016-06-27 15:48:23 jirutka: can you check what it sends? my client definitely can't read your 1-char ellipsis, but it's a very basic non-configurable Windows thingy so it may very well be using a windows-1252 font - I'd just like to know. 2016-06-27 15:48:51 skarnet: did you really wrote Windows?! 2016-06-27 15:49:12 we had this discussion before, I said yes, and I explained that was intentional 2016-06-27 15:49:18 it is utf-8 ... 2016-06-27 15:50:01 kaniini: thanks. 2016-06-27 15:51:21 skarnet: I know, but this really doesn’t mean that it’s the best option, kernel devs are very conservative… but doesn’t matter, I’m not forcing anyone to use GH, I was just thinking about syncing comments between ML and GH, so everyone would have a choice what to use for commenting 2016-06-27 15:52:16 skarnet: you know, there are things that one would rather forget :) 2016-06-27 15:53:37 jirutka: I wasn't talking about that ML/GH thing at all. Earlier, I was making fun of you, that's all (and it took you some time to realize it! :P) but I wasn't serious. idc what your workflow is as long as mine keeps sailing smoothly. 2016-06-27 16:03:05 skarnet: it's UTF-8, take one more: … 2016-06-27 16:03:20 embrace pl and cz keymaps in xorg 2016-06-27 16:03:25 resistance is futile 2016-06-27 16:03:31 stop torturing my dumb client 2016-06-27 16:04:21 we have more funny characters if you want it to suffer 2016-06-27 16:05:54 yeah, like √ 2016-06-27 16:06:07 or do you want some emoji? 😸 2016-06-27 16:07:16 I can actually see that cat emoji 2016-06-27 16:07:51 and what about ಠ_ಠ ? :) 2016-06-27 16:08:29 yeah, I can see those too. So I guess my client understands utf-8 correctly but uses a font that doesn't know ellipsis. 2016-06-27 16:13:37 that’s really weird 2016-06-27 16:14:05 I’d be kinda ᕦ(ò_óˇ)ᕤ from such client (¬‿¬) 2016-06-27 17:22:32 Hi, I begin to using Docker and it says that runs with a lightweight containers. But when I see e.g., jboss/wildfly image I can see that it relies on the CentOS 7 image. I suppose that it run the entire distro with all the default running services and default OS file footprints, so I think that all these set should be reduced in order to lightweight OS. Do you know if exist another way to get good performance of a lightweight OS? 2016-06-27 17:57:12 I really doubt it runs entire distro 2016-06-27 17:57:30 Docker is really designed for single process containers 2016-06-27 17:57:44 it doesn't depend on underlying operating system 2016-06-27 17:57:52 panzon: ^ 2016-06-27 17:58:31 so, while centos isn't known for being the most lightweight distribution ever, it doesn't imply image author runs entire init there 2016-06-27 19:32:22 barthalion: ok, so the Docker's lightweight is embedded in its single process container execution, right ? 2016-06-27 19:33:18 or Can't we assert that Docker is really lightweight? 2016-06-27 19:33:47 panzon: have you really used “docker” and ”lightweight” in one sentence? 2016-06-27 19:37:24 jirutka: I am trying to learn Docker correctly; its approach to run a lightweight OS, I just cannot understand why Docker assert that its solution is lightweight 2016-06-27 19:37:42 because marketing… 2016-06-27 19:38:29 really, Docker is anything but lightweight 2016-06-27 19:38:53 for example, you usually don’t need >6 GiB RAM to compile a lightweight program… 2016-06-27 19:42:09 so openshift lies too? e.g., they args that using "atomic host" they are building/running lightweight OS tools/libraries inside a container 2016-06-27 19:43:54 maybe I misunderstood you; of course you can run a lightweight OS inside a Docker “container”, but Docker itself isn’t lightweight 2016-06-27 19:44:31 panzon: clearly openshift lies 2016-06-27 19:44:36 you can run e.g. Alpine inside a docker and Alpine is surely a lightweight OS, but that’s not docker’s merit 2016-06-27 19:44:38 everywhere 2016-06-27 19:44:40 :D 2016-06-27 19:45:03 because when I inspect their Dockerfiles I don't see any OS footprint reduction or in system services. I suppose that the reason behind is that mentioned by barthalion 2016-06-27 19:46:16 yes, it’s quite common practice to put fullblown and bloated ubuntu into a docker container and pretend that it’s “lightweight”… 2016-06-27 19:46:25 what's lightweight in a sense, is what you end up running: a containerized service 2016-06-27 19:46:43 compared to a full OS in a VM, yes it's lightweight 2016-06-27 19:47:10 but docker in itself as jirutka said, is not lightweight 2016-06-27 19:50:31 thank you for clarify! 2016-06-27 19:51:23 now ask yourself 2016-06-27 19:51:36 what problem are you trying to solve with docker ? 2016-06-27 20:19:13 jirutka: lol, that FUD of yours is really amazing 2016-06-27 20:19:25 what FUD? 2016-06-27 20:20:12 panzon: you can't assert docker lightweightness, it's just a daemon that helps you to build/fetch and run images as containers 2016-06-27 20:20:46 panzon: I'm not sure what openshift atomic host is about, but it really depends on the image 2016-06-27 20:21:00 panzon: also depends about what kind of being light we're talking about 2016-06-27 20:21:42 panzon: image can be small, can eat less memory than one would expect, or simply run one process, but the last one is really de facto assumption nowadays 2016-06-27 20:21:59 panzon: it's hard to talk about this without wider context 2016-06-27 20:22:11 barthalion: and that daemon is anything but lightweight, that’s the point 2016-06-27 20:24:04 panzon: that atomic host you mentioned is pretty much different thing, something similar to coreos 2016-06-27 20:25:59 panzon: that is, it's specialized distribution with clever way of upgrading, shipping with docker + networking solution + some container scheduler 2016-06-27 20:26:07 jirutka: it's possible that docker itself is small, and the big-ass amount of heap it allocates is just the Go runtime :) 2016-06-27 20:26:19 skarnet: +1 XD 2016-06-27 20:26:48 skarnet: but it’s more likely both 2016-06-27 20:26:50 panzon: if you are just starting with docker, don't think about all of it, just focus on using docker efficiently and making good quality images 2016-06-27 20:27:42 panzon: and don’t assume that docker “containers” are secure, they are not, not at all 2016-06-27 20:27:45 (full disclosure: I like Go, I think it's a much saner language than most alternatives, if I was using a high-level language I'd definitely use Go. What I absolutely do not like in Go is its insane runtime. Not sure if that's a problem of the language or the reference implementation.) 2016-06-27 20:28:31 skarnet: +1 2016-06-27 20:29:09 hm, Go is better than PHP… and maybe better than perl XD 2016-06-27 20:29:22 time to go home, I’m still at works 2016-06-27 20:29:23 work 2016-06-27 20:29:41 jirutka: thank you 2016-06-27 20:30:20 no thanks for me ;( 2016-06-27 21:17:24 barthalion: should i do $(apk update && apk upgrade) to get latest greatest musl-libc? 2016-06-27 21:47:17 barthalion: i built the binaries and ran the tests after the apk fetch && apk update, that test is still failing. could it be i am still not getting latest version of musl-libc? 2016-06-27 21:54:14 ah, i just enabled the edge/main in /etc/apk/repositories and now i am on r11. :D 2016-06-27 22:25:03 peterj: you should be warned that edge is not stable… 2016-06-27 22:25:43 peterj: there might be problems with version incompatibilities etc. 2016-06-27 22:57:25 jirutka: thanks! actually ncopa added a patch in edge/main for the issue https://github.com/dotnet/coreclr/issues/5872. So i wanted to test it. :) 2016-06-27 23:50:18 barthalion: since all the coreclr PAL tests are passing, would be nice to promote lttng-ust to main as it is proven to be genuine. :) 2016-06-28 06:41:17 good morning climbers! 2016-06-28 06:41:26 im back 2016-06-28 06:41:30 i know you missed me! 2016-06-28 06:42:05 ACTION returns back to idle mode 2016-06-28 06:49:02 hi clandmeter 2016-06-28 06:49:14 hi fcolista 2016-06-28 06:49:23 Guys, i need some help in fixing #5806 2016-06-28 06:49:37 I just received a present from GIGABYTE for alpine linux 2016-06-28 06:49:52 really? What kind of present? 2016-06-28 06:50:53 :) 2016-06-28 06:50:57 let me show you a photo 2016-06-28 06:50:59 one sec 2016-06-28 06:52:59 my phone seems to have issues with office wifi. 2016-06-28 06:53:26 Seems that the gift you received is not a phone :) 2016-06-28 07:00:44 fcolista: https://goo.gl/photos/3trSpGgeuuikEmAP9 2016-06-28 07:01:12 clandmeter, oh...nice... 2016-06-28 07:01:18 i thought it was a gadget! 2016-06-28 07:01:35 o_o 2016-06-28 07:01:38 its a large gadget 2016-06-28 07:01:41 how are you planning to use it? 2016-06-28 07:01:50 what kind of server is that? 2016-06-28 07:02:08 its an ARMv8 server 2016-06-28 07:02:15 aha 2016-06-28 07:03:32 https://www.scan.co.uk/products/gigabyte-r120-p30-single-socket-1u-rackmount-armv8-24ghz-pcie-30-8x-ddr3-slots-ecc-unbuffered-2x-10g 2016-06-28 07:03:52 1300 pound, that around 500 euro now ;-) 2016-06-28 07:04:14 :-D lol 2016-06-28 07:05:01 1500 euro :> 2016-06-28 07:05:26 scadu he's koking about brexit consequences ;) 2016-06-28 07:05:29 *joking 2016-06-28 07:06:30 sorry, bad joke. 2016-06-28 07:06:56 morning ncopa 2016-06-28 07:07:10 ncopa: https://goo.gl/photos/3trSpGgeuuikEmAP9 :) 2016-06-28 07:07:26 nice!! 2016-06-28 07:07:29 anyway, if someone is available for fixing #5806, please ping me 2016-06-28 07:08:45 fcolista: did you use replaces? 2016-06-28 07:08:54 clandmeter, no 2016-06-28 07:09:24 what is the problem? 2016-06-28 07:09:30 different packages provides same file? 2016-06-28 07:09:35 it looks like you splitted the pkg? 2016-06-28 07:09:44 i splitted the package, yes 2016-06-28 07:10:12 sqlite subpkg cannot overwite the main pkg 2016-06-28 07:10:17 the subpkacakge does a symlink to libbareoscats 2016-06-28 07:11:02 i've the same issue with the other subpkgs: -mysql and -pgsql 2016-06-28 07:11:24 the issue is in this function: 2016-06-28 07:11:25 _mv_backend() { 2016-06-28 07:11:25 mkdir -p "$subpkgdir"/usr/lib 2016-06-28 07:11:25 mkdir -p "$subpkgdir"/etc/bareos/scripts 2016-06-28 07:11:25 mv "$pkgdir"/usr/lib/libbareoscats-${1}-${pkgver}.so \ 2016-06-28 07:11:25 "$subpkgdir"/usr/lib || return 1 2016-06-28 07:11:27 ln -s libbareoscats-${pkgver}.so "$subpkgdir"/usr/lib/libbareoscats.so 2016-06-28 07:11:29 ln -s libbareoscats-${1}-${pkgver}.so \ 2016-06-28 07:11:31 "$subpkgdir"/usr/lib/libbareoscats-${pkgver}.so 2016-06-28 07:11:33 } 2016-06-28 07:14:58 clandmeter: awesome, is it going to be aarch64 builder or armv7? 2016-06-28 07:15:27 well, i dont think im the right person to do the porting. 2016-06-28 07:16:02 so it will probably depend on fabled and ncopa. 2016-06-28 07:16:12 but first lets get it booted :) 2016-06-28 07:16:26 fcolista: can you show me the whole apkbuild? 2016-06-28 07:16:56 clandmeter, http://sprunge.us/GKIG 2016-06-28 07:18:46 why are you creating the symlinks? 2016-06-28 07:22:08 morning climbers 2016-06-28 07:22:14 any icelandic about? 2016-06-28 07:22:36 :) 2016-06-28 07:22:42 they are still sleeping 2016-06-28 07:23:22 I'm so pleased they won. 2016-06-28 07:23:41 not so please abuot the other thing though 2016-06-28 07:24:08 ScrumpyJack: you are from the UK? 2016-06-28 07:24:18 the last match: Iceland vs Poland 2016-06-28 07:24:27 lol 2016-06-28 07:24:52 i can imagine iceland in the final but... ;-) 2016-06-28 07:24:57 clandmeter: i'm dual. I have a UK passport, but I'm renewing my french passport now 2016-06-28 07:25:46 since, you know, the other thing that happend :( 2016-06-28 07:26:01 ACTION should not make jokes about football coming from holland.. 2016-06-28 07:26:10 ScrumpyJack: understand 2016-06-28 07:26:17 it does suck 2016-06-28 07:30:31 anyone has 4 ssd's lying around? 2016-06-28 07:30:40 i have 1 2016-06-28 07:30:44 :) 2016-06-28 07:39:45 clandmeter: not lying around sadly :D 2016-06-28 07:39:50 morning o/ 2016-06-28 07:40:50 too bad. but we will manage :) 2016-06-28 07:48:27 Morning guys. I am having some issues with apk not respecting proxy settings. I posted on the alpine-linux channel but if there is anybody who might be able to answer a question or 2 I'd appreciated it 2016-06-28 07:50:36 is this when you su? 2016-06-28 07:52:07 No 2016-06-28 07:52:16 I am running the alpine docker image 2016-06-28 07:52:18 So I am root 2016-06-28 07:52:29 Running wget works fine with the proxy 2016-06-28 07:52:47 I tried wget the tar.gz alpine index file that apk tried to download and that worked fine 2016-06-28 07:52:55 But running apk fails saying there is no internet 2016-06-28 07:53:24 I googled and found people asking about apk issues with proxy settings however answers were that it respects those. Not my experience at the moment :( 2016-06-28 07:54:01 i need to set up a docker image, but i have no knowledge of docker :( 2016-06-28 07:54:16 Its alpine official docker image 2016-06-28 07:54:22 ScrumpyJack: the docker docs are pretty good 2016-06-28 07:54:29 But I am guessing it should be no different than normal apk 2016-06-28 07:54:42 ByteFlinger: set http_proxy env var 2016-06-28 07:54:50 ncopa: I'm sure they are, i just haven't taken the time to go through them. i ought to really 2016-06-28 07:54:50 Yes that what I did 2016-06-28 07:55:01 for example: export http_proxy=http://proxy:8080 2016-06-28 07:55:12 I first ran this command 2016-06-28 07:55:14 export http_proxy=x.x.x.x:yyyy && export https_proxy=x.x.x.x:yyyy && wget http://dl-cdn.alpinelinux.org/alpine/v3.4/community/x86_64/APKINDEX.tar.gz 2016-06-28 07:55:28 I tested untaring the downloaded file and it works 2016-06-28 07:55:50 but apk update does not? 2016-06-28 07:55:53 Nope 2016-06-28 07:56:05 When I replace the wget with apk update in the same command I got this 2016-06-28 07:56:15 fetch http://dl-cdn.alpinelinux.org/alpine/v3.4/community/x86_64/APKINDEX.tar.gz ERROR: http://dl-cdn.alpinelinux.org/alpine/v3.4/community: network error (check Internet connection and firewall) 2016-06-28 07:56:52 I'll try here 2016-06-28 07:57:01 i'll set up a local tinyproxy 2016-06-28 07:58:19 seems to work for me 2016-06-28 07:58:25 :/ 2016-06-28 07:58:44 Not sure what else I could be doing wrong. Seems pretty straight forward to me 2016-06-28 07:58:52 let me check my proxy log 2016-06-28 07:58:59 just to verify that it actually uses it 2016-06-28 07:59:55 yes 2016-06-28 07:59:56 it does 2016-06-28 08:00:40 ByteFlinger: i think i know whats wrong 2016-06-28 08:00:48 Oh? Please tell :) 2016-06-28 08:00:53 export http_proxy=x.x.x.x:yyyy && export https_proxy=x.x.x.x:yyyy && ... 2016-06-28 08:01:12 Whats the issue with that? 2016-06-28 08:01:17 do: `export http_proxy=http://x.x.x.x:yyyy` 2016-06-28 08:01:24 note the http:// in there 2016-06-28 08:01:28 You mean with backticks? 2016-06-28 08:01:32 Ah 2016-06-28 08:01:35 Let me check 2016-06-28 08:02:32 Yepp. That did it 2016-06-28 08:02:47 It shouldnt matter but I guess alpine expects http:// 2016-06-28 08:03:02 Is that a bug? Of sorts 2016-06-28 08:04:04 Thank you very much ncopa 2016-06-28 08:13:18 clandmeter, re: bareos symlink. That part was dumbly copy/pasted from bacula's APKBUILD 2016-06-28 08:14:35 anyway, i've removed the symlinks and tried to install the apkbuild manually 2016-06-28 08:14:38 ByteFlinger: not a bug, that's how http_proxy is used by libcurl, libfetch and possibly everything else 2016-06-28 08:14:51 pig:~# apk add ./bareos-sqlite-15.2.4-r1.apk --allow-untrusted 2016-06-28 08:14:51 ERROR: unsatisfiable constraints: 2016-06-28 08:14:51 bareos-15.2.4-r0: 2016-06-28 08:14:51 conflicts: bareos-sqlite-15.2.4-r1[so:libbareoscats-15.2.4.so=0] 2016-06-28 08:14:51 satisfies: bareos-sqlite-15.2.4-r1[bareos] 2016-06-28 08:14:52 bareos-sqlite-15.2.4-r1: 2016-06-28 08:14:56 masked in: cache 2016-06-28 08:14:58 conflicts: bareos-15.2.4-r0[so:libbareoscats-15.2.4.so=0] 2016-06-28 08:15:00 satisfies: world[bareos-sqlite> pig:~# apk add ./bareos-sqlite-15.2.4-r1.apk ./bareos-15.2.4-r1.apk --allow-untrusted 2016-06-28 08:15:04 ERROR: unsatisfiable constraints: 2016-06-28 08:15:06 bareos-15.2.4-r1: 2016-06-28 08:15:08 masked in: cache 2016-06-28 08:15:10 conflicts: bareos-sqlite-15.2.4-r1[so:libbareoscats-15.2.4.so=0] 2016-06-28 08:15:10 barthalion: It did work with wget but sure. I'll bite :) 2016-06-28 08:15:11 sigh 2016-06-28 08:15:12 satisfies: world[bareos> bareos-sqlite-15.2.4-r1: 2016-06-28 08:15:15 fcolista: use pastebin 2016-06-28 08:15:16 masked in: cache 2016-06-28 08:15:17 Just happy to get it out of the way 2016-06-28 08:15:18 conflicts: bareos-15.2.4-r1[so:libbareoscats-15.2.4.so=0] 2016-06-28 08:15:20 satisfies: world[bareos-sqlite> fcolista: what is your apkbuild now? 2016-06-28 08:54:37 could someone apply 2147 and 2148. i need to find a better way for test/move stuff 2016-06-28 09:20:20 clandmeter, http://sprunge.us/VcfT 2016-06-28 09:21:59 that's the result: http://pastebin.com/rG7vzUAM 2016-06-28 09:26:18 fcolista: thats not right 2016-06-28 09:26:25 all subpkgs have the same file 2016-06-28 09:26:35 ? 2016-06-28 09:26:49 ah wait 2016-06-28 09:26:55 im not reading it correctly 2016-06-28 09:27:03 there's ${1} 2016-06-28 09:27:50 http://sprunge.us/SUDZ 2016-06-28 09:28:01 this the content of bareos-sqlite ^ 2016-06-28 09:28:37 and this is bareos-mysql: http://sprunge.us/PCfe 2016-06-28 10:32:12 ncopa, i'm not able to fix #5806. 2016-06-28 10:32:50 i don't understand why it complains about a conflict, while the subpackage does not provide the lib apk is claiming 2016-06-28 10:33:22 http://pastebin.com/2sGBZKrS 2016-06-28 10:33:46 but bareos-sqlite does not provide libbareoscats-15.2.4.so 2016-06-28 10:54:55 i'll check 2016-06-28 11:35:16 fcolista 2016-06-28 11:35:23 yes 2016-06-28 11:35:30 both bareos and bareos-sqlite has the following file: 2016-06-28 11:35:39 usr/lib/libbareoscats-15.2.4.so 2016-06-28 11:35:53 in bareos, it is a real file 2016-06-28 11:36:05 in bareos-sqlite it is a symlink 2016-06-28 11:36:32 ncopa, i don't have that file in pkg/bareos-sqlite dir 2016-06-28 11:36:33 s -l 2016-06-28 11:36:33 total 24 2016-06-28 11:36:33 -rwxr-xr-x 1 fcolista fcolista 22320 Jun 28 11:28 libbareoscats-sqlite3-15.2.4.so 2016-06-28 11:36:33 lrwxrwxrwx 1 fcolista fcolista 31 Jun 28 11:28 libbareoscats-sqlite3.so -> libbareoscats-sqlite3-15.2.4.so 2016-06-28 11:37:04 http://pastebin.com/U1MuMihd 2016-06-28 11:37:48 that's the content of bareos-sqlite dir: 2016-06-28 11:37:49 http://pastebin.com/Y0ehLTvs 2016-06-28 11:37:58 http://git.alpinelinux.org/cgit/aports/commit/?id=d679b301817b0d4f04db071b8609d67374215955 that commit fixes it 2016-06-28 11:38:10 - ln -s libbareoscats-${1}-${pkgver}.so \ 2016-06-28 11:38:10 - "$subpkgdir"/usr/lib/libbareoscats-${pkgver}.so 2016-06-28 11:38:40 this line created usr/lib/libbareoscats-${pkgver}.so 2016-06-28 11:38:43 as a symlink 2016-06-28 11:39:15 so you fixed it earlier today 2016-06-28 11:43:46 so you mean that it's enough re-adding ln -s libbareoscats-${1}-${pkgver}.so \ 2016-06-28 11:43:50 "$subpkgdir"/usr/lib/libbareoscats-${pkgver}.so 2016-06-28 11:43:54 to fix that? 2016-06-28 11:46:54 no 2016-06-28 11:47:00 i mean that it is already fixed 2016-06-28 11:47:28 this doesn't fix actually 2016-06-28 11:47:56 seems that the issue is: 2016-06-28 11:47:57 provides = so:libbareoscats-15.2.4.so=0 2016-06-28 11:48:07 this is in bareos-sqlite .control file 2016-06-28 11:48:16 the same is in the bareos package 2016-06-28 11:48:35 cat .control.bareos-sqlite/.provides-so 2016-06-28 11:48:35 libbareoscats-15.2.4.so 0 2016-06-28 11:49:10 and this: cat .control.bareos-sqlite/.provides-so 2016-06-28 11:49:10 libbareoscats-15.2.4.so 0 2016-06-28 11:49:16 sorry: http://sprunge.us/VFAE 2016-06-28 11:49:20 (wrong paste) 2016-06-28 11:52:32 $ readelf -d pkg/bareos-sqlite/usr/lib/ 2016-06-28 11:52:32 libbareoscats-sqlite3.so | grep soname 2016-06-28 11:52:32 0x000000000000000e (SONAME) Library soname: [libbareoscats-15.2.4.so] 2016-06-28 11:53:32 so, the name is libbareoscats-sqlite3.so, but actually is libbareoscats-15.2.4.so 2016-06-28 11:55:47 so i can't split those packages 2016-06-28 11:56:29 (also --disable-rpath is not honored) 2016-06-28 11:59:14 could someone install cadaver and tell me if they get a dav:> prompt? i get nothing (type help at that point) 2016-06-28 12:09:16 ok, i think i'm going to remove the package split ncopa 2016-06-28 12:09:28 ScrumpyJack: if you install cadavers, the thing you're likely to get is a disease 2016-06-28 12:12:09 everyone's a comedian ;) 2016-06-28 12:26:41 rpath is other thing 2016-06-28 12:26:54 fcolista: yes seems like you'll have to remove the package split 2016-06-28 12:27:05 or disable the automatic dependency tracing 2016-06-28 12:27:26 ncopa: funny as aur has it with splitting 2016-06-28 12:27:53 they dont automatically trac soname deps then 2016-06-28 12:29:14 for simplicity now i'm going to remove the package split 2016-06-28 12:29:43 and yes, rpath is another thing 2016-06-28 12:29:50 that is not honored, though 2016-06-28 12:35:54 thx for this hints ncopa 2016-06-28 12:36:43 two questions: how can I disable automatic dependecy tracking? And how to fix the rpath issue? 2016-06-28 12:37:49 ScrumpyJack, can you please let me know if bareos-webui works? And also bareos package itself? 2016-06-28 12:37:54 When you can/want 2016-06-28 12:41:39 fcolista: options="!tracedeps" 2016-06-28 12:41:51 then you will have to do the deps manually 2016-06-28 12:42:04 what is the rpath issue? 2016-06-28 12:42:58 >>> bareos-pgsql*: Stripping binaries 2016-06-28 12:42:58 >>> WARNING: bareos-pgsql*: Redundant /usr/lib in rpath found 2016-06-28 12:43:00 this ^ 2016-06-28 12:43:13 i think they ad /usr/lib to rpath 2016-06-28 12:43:14 even if in configure there's the --disable-rpath option 2016-06-28 12:43:16 its just a warning 2016-06-28 12:43:37 readelf -d $path/to/file 2016-06-28 12:43:42 seems that the configure option is not honored. Not a big deal 2016-06-28 12:43:54 will probably show that there is a rpath is set in the file 2016-06-28 12:43:55 readelf -d $path/to/file there's RPATH 2016-06-28 12:43:56 yes 2016-06-28 12:43:58 is there 2016-06-28 12:44:21 correct fix is to fix the makefile or configure script to not add it 2016-06-28 12:44:40 a workaround is to remove the rpath after its built 2016-06-28 12:44:46 i dont remember the tool for that 2016-06-28 12:45:04 but there are some tool that can edit the elf 2016-06-28 12:45:25 ok, good to know. 2016-06-28 12:46:39 chrpath 2016-06-28 13:10:33 I'm writing release notes for 3.4.1 2016-06-28 13:10:53 is this good? http://wwwtest.alpinelinux.org/posts/Alpine-3.4.1-released.html 2016-06-28 13:11:12 i dont know if we should mention that it fixes syslinux boot prompt issue 2016-06-28 13:11:21 boot issues on some machines 2016-06-28 13:11:43 http://bugs.alpinelinux.org/versions/108 2016-06-28 13:44:29 ncopa: “travis: disable email notifications” is probably not useful in release notes 2016-06-28 13:49:32 its a raw git shortlog 2016-06-28 13:49:57 jirutka: WRT that "prayer" abuild I was working on, it appears it can't work with the imaplib as it is packaged on alpine. There is a patch in aports that made it build by including wrong file, but it obviously doesn't work at run-time. 2016-06-28 13:50:44 I've decided to run it on gentoo instead where it works fine, not sure if I'll have time in near future for investigating what's wrong with imaplib on alpine. 2016-06-28 13:53:12 fcolista: you got a minute? could you apply 2147 and 2148? 2016-06-28 13:55:06 ScrumpyJack, 2147 is invasive 2016-06-28 13:55:39 have you tried to rebuild all of those packages, to be sure that they don't have dependencies in testing? 2016-06-28 13:57:48 aha and remind does not have maintainer 2016-06-28 13:58:07 and aha source is missing 2016-06-28 15:00:52 hey :) 2016-06-28 15:01:08 is someone of you running android studion on alpine linux? 2016-06-28 15:17:32 fcolista: i installed and tested the binaries in the packages. i then ran a script to check whether the dependancies for each packages were met in edge/main or edge/community 2016-06-28 15:18:49 ah source is in the APKBUILD file 2016-06-28 15:19:56 I have *not* corrected the APKBUILD files unless the packages was broken, so you'll see various poops in there. 2016-06-28 15:20:21 like no maintainer, but that doesn't break a package. 2016-06-28 15:21:49 I'm trying to install and exec each [a-z] package in testing/ and I'm still on letter a 2016-06-28 15:23:37 so, what do you mean that 2147 is invasive? 2016-06-29 07:11:54 fabled: hi dude. have you had any thoughts about the overlay issues? 2016-06-29 07:13:12 the first issue is that update-kernel doesn't copy them to the destdir (fixed in patch 2134) 2016-06-29 07:13:36 the second issue is the overlays they fail to load :( 2016-06-29 07:47:51 ScrumpyJack, sorry, still catching up backlog. i'll try to look at that later today or tomorow 2016-06-29 07:56:46 cool. backlog of what? 2016-06-29 07:57:39 mail, etc 2016-06-29 10:52:21 right, time to upgrade to 3.4 :) 2016-06-29 11:42:23 reboot 2016-06-29 11:46:35 back! yay! 2016-06-29 20:52:09 ncopa or barthalion, are you here? 2016-06-30 05:50:37 Is it possible to compile i386 binaries on the x86_64 root? I just found that if I set -m32 for gcc, sizeof(uint64_t) returns 32 (unsigned long). 2016-06-30 06:21:51 fatal: No names found, cannot describe anything. < what's the meaning of this when finishing abuild ? It's successful though 2016-06-30 06:26:19 avagin_: you need use the 32 bit libc headers for that 2016-06-30 06:26:31 i dont know what the proper way to solve it is 2016-06-30 06:26:51 we had same issue with xen hvmloader 2016-06-30 06:26:54 which is 32bit 2016-06-30 06:27:13 we patched it to use its own stdint.h 2016-06-30 06:29:03 avagin_: this is what we did with xen: http://git.alpinelinux.org/cgit/aports/commit/main/xen?id=bcf7b52774f1b0a3e405a207c3c4a5342b951f40 2016-06-30 06:31:18 @ncopa: do you run a compiled hvmloader in the same root? Is it possible to install musl-i386 in x86_64 root? 2016-06-30 06:32:29 avagin_: it is currently not possible to install musl-i386 in x86_64 root 2016-06-30 06:32:39 well, you can do it manually 2016-06-30 06:32:55 wget musl.apk && tar -zxf musl.apk 2016-06-30 06:33:29 @ncope: Ok. Thank you for the help. 2016-06-30 06:33:35 hvmloader is statically linked (afaik) 2016-06-30 06:34:05 just curious, what are you trying to gcc -m32? 2016-06-30 06:34:44 i have been thinking of installing/shipping the 32 bit headers in 64bit musl 2016-06-30 06:35:11 and maybe patch gcc to update the include search if -m32 is passed 2016-06-30 06:35:30 but so far i dont think it has been worth it 2016-06-30 06:36:00 in glibc 32bit and 64bit use same headers. in musl they differ 2016-06-30 06:55:57 @ncopa: https://github.com/xemul/criu/issues/181 2016-06-30 06:57:21 @ncopa: In criu, we have a few blobs which have to be compiled for i386. We need them to inject code into i386 processes. 2016-06-30 07:08:34 I sent two new aports to the mailing list the other day and updated another related one, namely: plzip, lzlib & lzip. Do I just have to wait until they get merged/built or is there anything else I can/have to do? Just being curious. 2016-06-30 07:11:08 Also, have there been any efforts so far to provide -libre flavored kernels? It seems to me that a blob free kernel is the last missing thing in Alpine to make it secure and blob free by default, which are arguably related. 2016-06-30 14:03:06 ncopa, are you ok with http://sprunge.us/bRKh ? 2016-06-30 14:04:09 checkapk returns : http://pastebin.com/M3C8HG7X 2016-06-30 14:32:22 fatal: No names found, cannot describe anything. < what's the meaning of this when finishing abuild ? It's successful though 2016-06-30 14:34:19 coredumbbb: I’m also interested what does this mean :) 2016-06-30 14:36:57 fatal: cannot describe anything, not even what this error message means 2016-06-30 15:07:07 skarnet: +1 XD 2016-06-30 15:56:24 well I guess I could read abuild sources .... 2016-06-30 15:56:26 :D 2016-06-30 16:00:25 ncopa or barthalion, anyone of you here? 2016-06-30 16:01:16 kind of 2016-06-30 16:01:22 going to eat something in 3 minutes 2016-06-30 16:02:01 barthalion: I think that https://github.com/alpinelinux/aports/pull/146 is good to merge, except `pkgrel` should be changed to 2 2016-06-30 16:04:31 sure, will do it later today 2016-06-30 16:04:40 okay 2016-06-30 23:20:56 <^7heo> does anyone has a copy of musl-dbg-1.1.14-r9.apk? 2016-06-30 23:21:04 <^7heo> I'm running a program and need to gdb into it 2016-06-30 23:21:12 <^7heo> and I can only obtain musl-dbg-1.1.14-r11.apk 2016-06-30 23:21:15 <^7heo> =/ 2016-06-30 23:21:20 <^7heo> or musl-dbg-1.1.14-r10.apk 2016-06-30 23:22:06 I don’t have it 2016-06-30 23:22:55 I’d like to start making snapshosts of testing branch, but still doesn’t have it 2016-06-30 23:23:23 <^7heo> yeah =/ 2016-06-30 23:23:33 <^7heo> any way to obtain that? 2016-06-30 23:23:55 you need exactly r9, right? 2016-06-30 23:23:59 <^7heo> yeah 2016-06-30 23:24:15 why not just build it yourself? 2016-06-30 23:24:19 <^7heo> if I could afford anything else I would just go for r11 2016-06-30 23:24:30 <^7heo> well, I could 2016-06-30 23:28:47 <^7heo> building... 2016-06-30 23:28:51 <^7heo> let's hope it doesn't take forever. 2016-06-30 23:29:02 <^7heo> at least it's parallelizing 2016-06-30 23:29:49 <^7heo> installed. 2016-06-30 23:29:49 <^7heo> wow. 2016-06-30 23:29:53 <^7heo> that was FAST 2016-06-30 23:29:57 :) 2016-06-30 23:29:57 <^7heo> thanks jirutka 2016-06-30 23:30:00 <^7heo> I love apk. 2016-06-30 23:30:11 you’re welcome 2016-06-30 23:30:14 yeah, apk rocks! 2016-06-30 23:30:17 <^7heo> really, git + apk is the easiest part of debugging python with gdb :P 2016-06-30 23:30:45 <^7heo> if only I had been clever enough to implement signal handling before in my python code... 2016-06-30 23:30:57 <^7heo> I wouldn't have to reload everything from a running context... 2016-06-30 23:31:08 what are you building? 2016-06-30 23:31:29 <^7heo> well 2016-06-30 23:31:33 <^7heo> I'm bruteforcing something 2016-06-30 23:31:38 <^7heo> hence the non-stopping 2016-06-30 23:31:46 <^7heo> but 2016-06-30 23:31:56 like cracking some password? 2016-06-30 23:31:57 <^7heo> since I didn't implement anything for output 2016-06-30 23:31:59 <^7heo> yeah 2016-06-30 23:32:05 <^7heo> I'm stuck on waiting 2016-06-30 23:32:07 <^7heo> and... 2016-06-30 23:32:17 okay, maybe it’d be better to not ask more :) 2016-06-30 23:32:18 <^7heo> since I don't trust myself not to have made mistakes 2016-06-30 23:32:28 <^7heo> nah it's just me being curious 2016-06-30 23:32:36 <^7heo> on how feasible it is to guess an email from a gravatar 2016-06-30 23:32:41 <^7heo> (because md5, etc) 2016-06-30 23:32:55 heh, that’s interesting task :) 2016-06-30 23:33:07 <^7heo> it takes only a few seconds to find my email if you get the host part 2016-06-30 23:33:14 <^7heo> with my gravatar 2016-06-30 23:33:27 <^7heo> but since complexity goes exponentially... 2016-06-30 23:33:38 <^7heo> if you don't get the host part, you wait MUCH longer. 2016-06-30 23:33:50 <^7heo> also, my email is short, 7heo@ is pretty short ;) 2016-06-30 23:33:56 yeah, beware of eating all the universe resources, that would be bad… XD 2016-06-30 23:34:02 <^7heo> ;D 2016-06-30 23:34:11 <^7heo> anyway 2016-06-30 23:34:15 <^7heo> I wonder how fast it goes 2016-06-30 23:34:16 <^7heo> so 2016-06-30 23:34:38 <^7heo> I wonder if I'm still checking for 0902ng0wngosgn0@domain.tld 2016-06-30 23:34:39 do you some special HW or at least something like CUDA? 2016-06-30 23:34:53 <^7heo> or if I'm probing strings that are 1000 characters long or something 2016-06-30 23:34:58 <^7heo> nah 2016-06-30 23:35:03 <^7heo> Just a 4810MQ 2016-06-30 23:35:06 <^7heo> and even 2016-06-30 23:35:11 <^7heo> I didn't implement multithreading 2016-06-30 23:35:17 <^7heo> so it's using ONE of my eight threads 2016-06-30 23:35:19 <^7heo> so... 2016-06-30 23:35:31 <^7heo> as you might guess, I *might* want to reload my code. 2016-06-30 23:35:38 <^7heo> and given how far it went 2016-06-30 23:35:45 yeah, you definitely should implement paralelism 2016-06-30 23:35:47 <^7heo> I might even want to stop it now to save time by optimization. 2016-06-30 23:35:51 <^7heo> yeah 2016-06-30 23:36:08 <^7heo> I wanted to implement opportunistic and adaptative parallalism 2016-06-30 23:36:22 <^7heo> (ie. grab all you can, all the time) 2016-06-30 23:36:41 <^7heo> (and share the load while constantly re-balancing) 2016-06-30 23:36:47 <^7heo> also, over network too, if possible. 2016-06-30 23:36:55 <^7heo> that would be sick 2016-06-30 23:37:02 <^7heo> but well, it's far from where I am 2016-06-30 23:37:27 <^7heo> I'm sure that by cleaning up my code, I'd already gain a few (dozen of?) percents of performance 2016-06-30 23:38:05 <^7heo> dang 2016-06-30 23:38:07 <^7heo> warning: the debug information found in "/usr/lib/debug//lib/ld-musl-x86_64.so.1.debug" does not match "/lib/ld-musl-x86_64.so.1" (CRC mismatch). 2016-06-30 23:38:26 <^7heo> I have to find the EXACT version of the r9 I have... 2016-06-30 23:39:43 I’ve discussed the “serverless” buzzword with a colleague of mine today… I said him: “this is bullshit, more real serverless would be to use users’ computers… i.e. make a webpage that starts many JS workers on background and start computing…” and he said “dude, this is not called serverless, this is called botnet!” 2016-06-30 23:42:52 <^7heo> hahaha 2016-06-30 23:42:54 <^7heo> yeah 2016-06-30 23:43:17 <^7heo> buzzwords 2016-06-30 23:43:25 <^7heo> or how marketing poisons our lives 2016-06-30 23:43:42 yeah 2016-06-30 23:43:55 <^7heo> okay, so there is only one r9 version 2016-06-30 23:44:10 <^7heo> which is logical since you got to change the rev if you do a change 2016-06-30 23:44:11 I hate this culture of marketing buzzwords, hypes and start-ups so much 2016-06-30 23:44:21 <^7heo> startups are okay 2016-06-30 23:44:26 <^7heo> the rest... 2016-06-30 23:45:24 <^7heo> I don't GET IT! 2016-06-30 23:45:26 <^7heo> :'( 2016-06-30 23:45:29 <^7heo> whhhhhyyy 2016-06-30 23:45:45 <^7heo> the debug symbols don't match the binary 2016-06-30 23:45:48 Slac, the company that just reinvented IRC, has higher value on markets than SpaceX, mindblowing company of Elon Musk which managed to create a space program in just few years… this is sick, absolutely sick 2016-06-30 23:45:52 <^7heo> wrong CRC 2016-06-30 23:46:12 <^7heo> because I have a different computer than the one that built the original musl, I presume? 2016-06-30 23:46:33 <^7heo> yeah but everybody uses slack 2016-06-30 23:46:35 <^7heo> because 2016-06-30 23:46:52 <^7heo> marketing lusers want shiny smilehs on smartphones 2016-06-30 23:47:00 <^7heo> and it guys want irssi/weechat/whatnot 2016-06-30 23:47:07 not necessarily, not every software allows really reproducible builds 2016-06-30 23:47:19 <^7heo> yeah, gcc powa 2016-06-30 23:47:35 <^7heo> but then I'm fucked 2016-06-30 23:47:46 <^7heo> because my software's runnin 2016-06-30 23:47:52 <^7heo> with a musl version 2016-06-30 23:48:02 <^7heo> and I can't ever import the right symbols... 2016-06-30 23:48:18 <^7heo> :'( 2016-06-30 23:48:34 <^7heo> I so wish I could get a copy of the musl-dbg package I need =/ 2016-06-30 23:48:39 in my new job we use Slack… and our crazy security guy disabled XMPP and IRC gateway, so I’m forced to their sh!ny so called native app that is just packaged Chromium with their JS :( 2016-06-30 23:48:52 <^7heo> oh 2016-06-30 23:48:58 <^7heo> I would be like 2016-06-30 23:49:05 <^7heo> I'm sorry 2016-06-30 23:49:08 <^7heo> I don't have a smartphone 2016-06-30 23:49:16 at least they provides an API, but I cannot find any actually usable alternative client 2016-06-30 23:49:20 <^7heo> I can't use slack on my Nokia 3310 2016-06-30 23:49:27 I mean on my computer, not phone 2016-06-30 23:49:36 <^7heo> meh 2016-06-30 23:49:58 <^7heo> I would like to see how having the IRC gateway is a security flaw 2016-06-30 23:50:04 <^7heo> but whatever 2016-06-30 23:50:34 <^7heo> if only an alpine mirror had the old files... 2016-06-30 23:51:45 he disabled even IMAP (on Gmail)… I said to me coworker that I can understand why he disabled XMPP/IRC gateway, but disabling IMAP is totally absurd and I’m definitely not gonna use this shitty webmail 2016-06-30 23:52:25 <^7heo> dude 2016-06-30 23:52:29 <^7heo> that'd be a reason for me to quit 2016-06-30 23:52:30 <^7heo> really 2016-06-30 23:52:35 <^7heo> I would just quit and be like 2016-06-30 23:52:39 the problem isn’t in XMPP/IRC protocols (obviously), but bad apperantly bad implementation on the side of Slack 2016-06-30 23:52:45 <^7heo> "there's no way I will ever put up with that level of absurd" 2016-06-30 23:52:56 he eventually enabled IMAP just for me 2016-06-30 23:53:03 <^7heo> "the simple fact that you put this person in charge tells me that we cannot work together" 2016-06-30 23:53:23 <^7heo> doesn't matter 2016-06-30 23:53:28 <^7heo> slack is used to send lolcats 2016-06-30 23:53:37 <^7heo> how is that a fucking security problem? 2016-06-30 23:53:55 <^7heo> in any case, their app isn't supporting End to End encryption I presume 2016-06-30 23:53:57 <^7heo> so.... 2016-06-30 23:54:03 <^7heo> fts 2016-06-30 23:54:08 <^7heo> (fuck that shit) 2016-06-30 23:54:14 well, I don’t even have a contract yet, it’s just two days I’ve joined, but yeah, I’m quite dissapointed from this approach… 2016-06-30 23:54:45 <^7heo> yeah 2016-06-30 23:54:56 I’m looking forward to meet him in personal and ask him about this stuff 2016-06-30 23:55:03 <^7heo> are we short on space on the mirrors? 2016-06-30 23:55:12 to find if he’s really a security expert or just Google fanboy doing security by obscurity 2016-06-30 23:55:24 <^7heo> because, maybe we could keep a few versions of packages? 2016-06-30 23:55:39 the repository is not so large currently 2016-06-30 23:55:49 <^7heo> you mean, the content, or the container? 2016-06-30 23:55:56 <^7heo> I believe you will find the latter, for your security guy 2016-06-30 23:56:16 content I guess? 2016-06-30 23:56:16 28G ./edge 2016-06-30 23:56:16 17G ./v3.0 2016-06-30 23:56:16 18G ./v3.1 2016-06-30 23:56:17 15G ./v3.2 2016-06-30 23:56:17 21G ./v3.3 2016-06-30 23:56:18 24G ./v3.4 2016-06-30 23:56:36 <^7heo> couldn't we keep a few weeks of packages 2016-06-30 23:56:37 <^7heo> ? 2016-06-30 23:56:50 we must ask ncopa about this 2016-06-30 23:56:53 <^7heo> I'm not even asking for r2 2016-06-30 23:56:59 <^7heo> it's only TWO revs behind 2016-06-30 23:57:01 <^7heo> on edge 2016-06-30 23:57:03 <^7heo> it's not even old ;) 2016-06-30 23:57:11 <^7heo> yeah 2016-06-30 23:57:17 <^7heo> but that doesn't help me right now 2016-06-30 23:58:00 <^7heo> I have musl dbg on my server... 2016-06-30 23:58:14 <^7heo> 1.1.11-r2 2016-06-30 23:58:19 <^7heo> not helpinh. 2016-06-30 23:58:25 <^7heo> helping* 2016-06-30 23:58:56 the job is really interesting, so I’m gonna give it a chance, but I’m starting to be afraid of dictatorship… this guy loves Chrome, so we all should (?) use Chrome… the product has nothing to do with web browser, so it’s nonsense… 2016-06-30 23:59:15 <^7heo> yeah 2016-06-30 23:59:27 <^7heo> that's commonplace in the business world 2016-06-30 23:59:32 maybe try to look at other mirrors, if some is out-of-date 2016-06-30 23:59:38 <^7heo> people don't want to fight a stupid dickhead 2016-06-30 23:59:46 <^7heo> so they just let him be the dictator