#archlinux32 | Logs for 2019-01-25

[00:09:07] -!- woshty has quit [Ping timeout: 246 seconds]
[00:46:42] -!- bill-auger has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
[00:47:17] -!- bill-auger has joined #archlinux32
[00:53:20] -!- AndrevS has quit [Remote host closed the connection]
[01:26:59] -!- isacdaavid has joined #archlinux32
[01:31:11] -!- MrBIOS has quit [Quit: MrBIOS]
[01:50:20] <buildmaster> i686/libgda is broken (says nlopc46).
[02:00:40] <buildmaster> i686/deepin-daemon is broken (says nlopc43).
[02:03:58] <elibrokeit> So there was a musl update and now I'm waiting on i686 musl in order to roll out a new pacman-static build :D
[02:05:09] <elibrokeit> ooh it is in staging
[02:16:25] <buildmaster> i686/vegeta is broken (says nlopc46).
[02:50:10] -!- isacdaavid has quit [Quit: Leaving.]
[05:32:46] -!- yans has quit [Quit: chaos is the only true answer]
[06:49:45] <girls> hmm, I wonder, why musl is not being moved from staging to testing ...
[06:49:55] <girls> our database dumps will show it, for sure!
[07:01:33] -!- oaken-source has joined #archlinux32
[07:16:26] <guys> girls: especially given it is only used as a makedepends for other things
[07:17:26] <girls> yeah
[07:17:31] <girls> this should be nice to debug
[08:00:15] <girls> ok, I think, I found it :-)
[08:01:06] <girls> anyone knows that feeling when you find some critical bug and you think: "how did this ever work?"
[08:06:30] -!- oaken-source has quit [Ping timeout: 268 seconds]
[08:35:27] -!- thePiGrepper has quit [Ping timeout: 240 seconds]
[08:36:34] -!- thePiGrepper has joined #archlinux32
[09:13:09] -!- oaken-source has joined #archlinux32
[09:16:44] -!- abaumann has joined #archlinux32
[09:16:45] <buildmaster> Hi abaumann!
[09:16:45] <buildmaster> !rq abaumann
[09:16:46] <phrik> buildmaster: <abaumann> We should really not build a package more than 99 times. ;-)
[09:24:50] -!- deep42thought has joined #archlinux32
[09:24:50] <buildmaster> Hi deep42thought!
[09:24:50] <buildmaster> !rq deep42thought
[09:24:51] <phrik> buildmaster: <deep42thought> computers ranked by dirt: smokers, car mechanics, bedroom, ...
[09:24:56] <abaumann> morning deep42thought
[09:25:01] <deep42thought> Good morning, abaumann!
[09:25:43] <deep42thought> db-update should move more packages now (hopefully the right ones)
[09:25:48] <abaumann> :-)
[09:25:51] <deep42thought> as a downside, it also takes much more time now :-/
[09:26:06] <abaumann> fast and wrong is not an option ;-)
[09:26:57] <abaumann> Is devtools32 master rolled out currenctly to all i486 slaves?
[09:27:11] <abaumann> Because, then I could make my changes in there..
[09:32:01] <deep42thought> yes, it is
[09:32:37] <deep42thought> but the architecture field needs to be hand-edited after upgrading
[09:34:37] <abaumann> architecture field in pacman.conf?
[09:35:54] <deep42thought> yes
[09:35:55] <abaumann> There are too many files there.. it's hard to see, what gets actually executed.. hence we build a i486 identical to i686 since the beginning.
[09:36:12] <deep42thought> what dou you refer to?
[09:36:19] <abaumann> devtools32 aka devtools
[09:36:21] <deep42thought> you mean all the pacman-*.confs?
[09:37:08] <abaumann> More in general. The more gets done in the scripts.. the likelier it fails.
[09:37:48] <deep42thought> true
[09:38:08] <deep42thought> but the diff between the files is small
[09:38:34] <abaumann> yeah. this is a sign of problems: copy-paste with just a little bit of difference means lack of generalisation.
[09:38:53] <abaumann> OTOH I don't want to deviate from upstream too much..
[09:39:02] <deep42thought> yeah, problem is, that pacman needs a config - not a config + patch ;-P
[09:39:24] <deep42thought> we could create the pacman-32*.confs on the fly via Makefile
[09:39:34] <deep42thought> at least during build of devtools32
[09:39:39] <abaumann> s) nosetarch=1 ;;
[09:39:40] <abaumann> mmh.
[09:40:17] <abaumann> in mkarchroot.in
[09:40:51] <abaumann> yeah. but this should also be used during the build, not just wehn creating the chroot
[09:41:16] <abaumann> arch-nspawn is particullarly nasty, as it sets all kind of weird limits
[09:41:36] <abaumann> you couldn't bootstrap with those scripts.
[09:41:45] <deep42thought> well, we have the :without_systemd_nspawn: straw
[09:41:54] <deep42thought> which executes makepkg directly
[09:42:04] <abaumann> yeah. but this only helps, if previous builds fail, and they don't
[09:42:13] <abaumann> otherwise you have to run all i486 slaves with that straw
[09:42:29] <deep42thought> no - you can simply pass '-s :without_systemd_nspawn:' and it will try only that
[09:42:34] <deep42thought> ah, ok
[09:42:37] <deep42thought> right :-/
[09:42:56] <deep42thought> so you think the problem is systemd-nspawn introducing stuff in glibc?
[09:43:23] <abaumann> no. setarch is causing config.guess to think it's on a i686 inside a virtual machine explicitely set up to report a i486
[09:43:35] <abaumann> config.guess is the autoconf sniffer in glibc
[09:43:52] <abaumann> and based on that it is choosing optimizations for C library functions like strcmp.S with CMOV
[09:44:08] <deep42thought> maybe we should hack, whatever config.guess uses?
[09:44:11] <deep42thought> uname, probably
[09:44:47] <abaumann> uname is working correctly. setarch has a documented bug, 'The execution domains currently only affects the output of uname -m.'
[09:44:48] <deep42thought> or run the build slave with full virtualization (I think, the root cause is trying to optimize the virtualization by executing directly on the x86_64 host)
[09:44:56] <abaumann> uname -p (for processor) reports 'unknown'
[09:45:16] <deep42thought> uname is _not_ working correctly (in the sense that it _is_ reporting i686)
[09:45:19] <abaumann> no. it happens also on tyzoids 486 slave, which reports correctly uname as i486
[09:45:28] <abaumann> Linux abaumann-i486 4.20.0-arch1-1.0-ARCH #1 SMP PREEMPT Tue Jan 1 00:40:57 CET 2019 i486 GNU/Linux
[09:45:43] <deep42thought> run 'setarch i486 uname -m'
[09:45:50] <abaumann> no, uname works correctly. setarch is part of coreutils
[09:46:03] <deep42thought> ok
[09:46:07] <abaumann> has nothing to do with reporting the architecture - well, it does half of the job, sort of.
[09:46:09] <deep42thought> setarch + uname does not work correctly
[09:46:32] <abaumann> uname -m technically is i686/i386 as machine architecture. it says nothing about the specific processor.
[09:46:46] <deep42thought> $ setarch i486 uname -m
[09:46:46] <deep42thought> i686
[09:46:56] <deep42thought> while:
[09:46:57] <deep42thought> $ uname -m
[09:46:57] <deep42thought> i486
[09:47:02] <abaumann> exatly.
[09:47:10] <abaumann> uname -p is unusable, as it reports nothing.
[09:47:21] <abaumann> so config.guess relies on uname -m and some other trickery.
[09:47:47] <abaumann> There is no way you can change half of the autoconf-packagages to have a better CPU sniffer, I'm afraid
[09:48:11] <abaumann> another option is to cross-compile critical packages: glibc: --build=i486 --host=i486
[09:48:20] <abaumann> but this might trigger unwanted effects..
[09:48:45] <abaumann> but it feels really bad to cross-compile just because some core utilities are broken..
[09:48:49] <deep42thought> such as?
[09:48:57] <deep42thought> yeah, ok
[09:49:05] <abaumann> well, cross-compilation may trigger non-normal paths..
[09:49:10] <abaumann> ..and less tested paths.
[09:49:14] <abaumann> that's what I'm afraid.
[09:49:17] <abaumann> of.
[09:49:57] <abaumann> I think, I'll add some flags to staging-i486, so that setarch is simply not used.
[09:50:12] <abaumann> setarch is in devtools, becuase 64-bit people want to build lib32- packages for i686.
[09:50:20] <abaumann> That's all this is good for..
[09:50:45] <abaumann> I'll make a test on my build slave :-)
[09:51:13] <abaumann> 99.9% of all i486 packages should not rely on setarch and trigger funny optimisations - but ohoh, who knows :-)
[09:52:23] <deep42thought> no, setarch is in devtools, because x86_64 people want to build i686 packages (so do we)
[09:52:54] <deep42thought> but turning setarch off for i486 seems reasonable
[09:54:27] <abaumann> what I meant was: for all normal cases the setarch thingy works just fine, becuase nobody expects to build packages for below i686. :-)
[09:54:46] <deep42thought> I even think, setarch was optional one
[09:54:48] <deep42thought> *once
[09:54:59] <deep42thought> e.g. when building for x86_64 on x86_64 it was not used
[09:55:05] <deep42thought> but I might be wrong there
[09:56:00] <abaumann> 5b9c06d275dd1a2dca09514e39f853bd9c57be3a: mkarchroot, arch-nspawn: Add an `-s` flag to inhibit `setarch`. for ARM.
[09:56:07] <abaumann> So there the -s flags comes from. :-)
[09:57:04] <deep42thought> where is this -s flag?
[09:57:08] <deep42thought> in which executable?
[09:57:15] <abaumann> mkarchroot
[09:57:43] <abaumann> the problem is: there is also a makechrootpkg
[09:57:46] <deep42thought> hmm, so it is missing from archbuild/makechrootpkg
[09:57:55] <deep42thought> mkarchroot only sets up the chroot
[09:57:57] <abaumann> yes, I think so too.
[09:58:05] <deep42thought> and makechrootpkg calls makepkg inside that chroot
[09:58:12] <abaumann> yep.
[09:58:26] <deep42thought> and archbuild (a.k.a staging-*-build, ...) does the right thing of those
[09:59:03] <abaumann> there is a problem when ommiting the setarch: you cannot build anymore on a x86_64
[09:59:26] <abaumann> for ARM this is not an issue, because chrooting doesn't work :-)
[09:59:36] <deep42thought> exec ${CARCH:+setarch "$CARCH"} systemd-nspawn
[09:59:41] <abaumann> they build on ARM machines using distcc/ccache heavily on faster machines, I reckon.
[09:59:51] <deep42thought> so an unset CARCH leads to no setarch being invoked
[10:00:23] <deep42thought> I would simply skip setarch if $CARCH == $(uname -m)
[10:00:38] <abaumann> outside, you mean?
[10:00:43] <deep42thought> everywhere
[10:01:10] <deep42thought> or we ship our own setarch binary which does nothing if `uname -m` reports the architecture that is the target
[10:01:30] <deep42thought> ^might be the most thorough and least hackish solution
[10:02:15] <abaumann> *abaumann checks out coreutils and has a look at setarch.c
[10:02:59] <abaumann> aeh, I mean util-linux
[10:03:12] <deep42thought> :-D
[10:03:28] <deep42thought> I think, I even opened a bug against util-linux exactly for this problem
[10:03:44] <abaumann> ah.
[10:04:01] <abaumann> there is also a Debian bug..
[10:04:18] <abaumann> https://bugs.debian.org
[10:04:23] <phrik> Title: #506632 - util-linux: setarch does not provide i386, i486, or i586 information - Debian Bug report logs (at bugs.debian.org)
[10:04:39] <abaumann> It's a limitation of the kernel interface is the answer..
[10:04:44] <deep42thought> no, _that_ I understand
[10:04:50] <abaumann> ..so, you utsname already reports the wrong data.
[10:05:03] <abaumann> setarch.c is only a front to report data delivered by the kernel.
[10:05:09] <deep42thought> what I don't understand is why it _raises_ from i486 to i686 (e.g. why it changes at all!)
[10:05:30] <abaumann> If I remember correctly, there was once a LFS hack to change uname information as a loadable module, but I cannot find any traces of it..
[10:05:48] <deep42thought> https://github.com
[10:05:49] <phrik> Title: setarch on qemu i486 raises to i686 · Issue #707 · karelzak/util-linux · GitHub (at github.com)
[10:05:51] <abaumann> nobody cares about those architectures?
[10:06:00] <deep42thought> that's my bug report
[10:06:25] <deep42thought> though, this does not look like some official repo to me anymore
[10:06:32] * deep42thought wonders how he found that repository ...
[10:06:39] <abaumann> "I do not see anything to fix in setarch command, maybe improve docs. Closing."
[10:06:42] <abaumann> cool.
[10:06:44] <abaumann> document the bugs.
[10:07:04] <deep42thought> documented => the bug becomes a feature
[10:07:14] <abaumann> !grab deep42thought
[10:07:15] <phrik> abaumann: Tada!
[10:07:21] <abaumann> yeah. exactly.
[10:07:33] <abaumann> so, I'll hunt this uname module hack of LFS, or I write a new one.
[10:07:55] <deep42thought> why not make setarch do nothing if you call "setarch $(uname -m)"?
[10:07:56] <abaumann> This leaves devtools, setarch in peace and fixes the problem in the right place. :-)
[10:08:08] <abaumann> It might have other consequences..
[10:08:14] <deep42thought> uname _does_ work - if not called inside setarch
[10:08:28] <deep42thought> so what do you want to fix about it?
[10:08:33] <abaumann> mmh.
[10:08:36] <abaumann> good point.
[10:08:43] <abaumann> more reading.. more debuuging. :-)
[10:10:18] <deep42thought> setarch is only called at two points in devtools - maybe we should really just patch that
[10:11:20] <deep42thought> ah, no it's 3
[10:12:18] <deep42thought> I have to leave, maybe already for the weekend
[10:12:20] <deep42thought> cu later!
[10:12:35] <abaumann> ok. thanks for discussing the issue with me. :-)
[10:12:36] <abaumann> cu
[10:13:38] -!- deep42thought has quit [Quit: Leaving.]
[10:16:25] -!- abaumann has quit [Quit: leaving]
[11:16:47] -!- oaken-source has quit [Ping timeout: 246 seconds]
[11:40:03] -!- oaken-source has joined #archlinux32
[12:05:36] <buildmaster> dirty! girls, my database - so dirty :-(
[12:05:36] * buildmaster goes insane.
[12:18:43] <bill-auger> that always me me laugh
[12:39:34] -!- KitsuWhooa has quit [Quit: Unable to handle kernel NULL pointer dereference at null]
[12:39:48] -!- KitsuWhooa has joined #archlinux32
[13:16:52] -!- ofara has quit [Remote host closed the connection]
[13:18:03] -!- abaumann has joined #archlinux32
[13:18:04] <buildmaster> Hi abaumann!
[13:18:04] <buildmaster> !rq abaumann
[13:18:04] <phrik> buildmaster: <abaumann> Never understood why there is a fallback ramdisk, but no fallback kernel image.
[13:18:12] <abaumann> buildmaster made a boo boo :-)
[13:18:59] <abaumann> package difference between mysql db and repo (mariadb, mariadb-clients)
[13:19:15] <abaumann> The following packages from the master mirror are not tracked in the database or vice versa:
[13:19:18] <abaumann> package-file i486/extra/mariadb-10.1.37-1.0-i486.pkg.tar.xz
[13:19:21] <abaumann> package-file i486/extra/mariadb-clients-10.1.37-1.0-i486.pkg.tar.xz
[13:19:23] <abaumann> package-file i486/extra/mytop-10.1.37-1.0-i486.pkg.tar.xz
[13:19:26] <abaumann> package-file i686/extra/mytop-10.1.37-1.0-i686.pkg.tar.xz
[13:19:35] -!- ofara has joined #archlinux32
[13:22:11] -!- oaken-source has quit [Ping timeout: 244 seconds]
[13:22:25] -!- abaumann has quit [Client Quit]
[14:12:12] * buildmaster resumes sanity.
[14:12:46] <girls> abaumann: I just removed these packages
[14:13:20] <girls> however, I have the feeling, we will see this happening more often in the future - until I find and fix the corresponding bug ;-)
[14:39:30] -!- oaken-source has joined #archlinux32
[14:59:13] -!- MrBIOS has joined #archlinux32
[15:10:49] <elibrokeit> girls: btw that repository would be official since karelzak is a kernel Dev and the maintainer of util-linux to boot
[15:12:32] <elibrokeit> Difference between GitHub mirror vs. lkml mailing list and kernel.org cgit => not really that important
[15:12:56] -!- T`aZ has quit [Ping timeout: 240 seconds]
[15:15:13] -!- T`aZ has joined #archlinux32
[15:22:33] -!- yans has joined #archlinux32
[15:30:02] -!- davor_ has joined #archlinux32
[15:30:30] -!- davor has quit [Ping timeout: 268 seconds]
[15:30:30] davor_ is now known as davor
[15:52:31] <buildmaster> i686/prometheus are broken (says nlopc46).
[16:04:05] <elibrokeit> English, buildmaster, English!
[16:15:56] -!- AndrevS has joined #archlinux32
[16:47:19] -!- oaken-source has quit [Ping timeout: 246 seconds]
[17:41:59] <buildmaster> i686/gitlab is broken (says nlopc46).
[17:56:04] <buildmaster> i686/deepin-file-manager is broken (says nlopc46).
[17:59:40] -!- AndrevS has quit [Remote host closed the connection]
[18:14:41] <buildmaster> i686/deepin-session-ui is broken (says eurobuild3).
[18:17:44] <buildmaster> i686/deepin-daemon is broken (says eurobuild3).
[18:25:37] <buildmaster> i686/startdde is broken (says eurobuild3).
[18:35:26] <buildmaster> i686/certbot-dns-cloudflare is broken (says nlopc43).
[19:15:11] <buildmaster> i686/vkd3d is broken (says eurobuild3).
[20:18:00] -!- alexandre9099 has quit [Ping timeout: 252 seconds]
[20:22:14] -!- alexandre9099 has joined #archlinux32
[20:37:39] <buildmaster> i686/deepin-control-center is broken (says nlopc43).
[20:38:16] <buildmaster> i686/deepin-session-ui is broken (says buildknecht).
[21:26:51] -!- AndrevS has joined #archlinux32
[21:32:32] <buildmaster> i686/percona-server is broken (says nlopc46).
[21:36:51] <buildmaster> i686/startdde is broken (says nlopc46).
[22:05:13] -!- oaken-source has joined #archlinux32
[23:25:23] <buildmaster> i686/python-js2py is broken (says buildknecht).
[23:32:19] -!- MrBIOS has quit [Quit: MrBIOS]