#archlinux32 | Logs for 2018-06-07

[00:29:05] -!- eduardoeae has quit [Ping timeout: 240 seconds]
[00:30:20] -!- eduardoeae has joined #archlinux32
[01:51:21] -!- NoobAlice has quit [Ping timeout: 264 seconds]
[02:51:40] -!- NoobAlice1 has joined #archlinux32
[04:24:02] -!- belanthor has quit [Quit: Leaving]
[05:54:27] -!- titus_livius has quit [Ping timeout: 240 seconds]
[05:55:38] -!- titus_livius has joined #archlinux32
[05:59:43] <buildmaster> girls, my database is dirty again ...
[06:26:03] * buildmaster resumes sanity.
[06:33:08] <buildmaster> girls, my database is dirty again ...
[06:43:46] -!- deep42thought has joined #archlinux32
[06:43:46] <buildmaster> Hi deep42thought!
[06:47:01] -!- titus_livius has joined #archlinux32
[06:54:46] -!- deep42thought has quit [Quit: Leaving.]
[07:21:22] -!- oaken-source has joined #archlinux32
[08:10:03] -!- NoobAlice1 has quit [Quit: Leaving.]
[08:19:00] -!- deep42thought has joined #archlinux32
[08:19:00] <buildmaster> Hi deep42thought!
[08:19:08] <deep42thought> Hi buildmaster, what's up?
[08:19:09] <buildmaster> up? I'm up for 1 hour, 34 minutes, load average: 0.84, 0.78, 0.72 ... and I'm insane :-D
[08:21:10] * buildmaster resumes sanity.
[08:33:56] -!- eduardoeae has quit [Ping timeout: 260 seconds]
[08:52:25] <buildmaster> tyzoid: Your buildslave "tyzoid-srv0-bs0" builds some outdated package.
[08:54:58] <deep42thought> indeed
[09:00:36] -!- oaken-source has quit [Ping timeout: 240 seconds]
[09:01:43] * buildmaster failed to execute a mysql query - can you have a look at "tmp.mysql-functions.query.stdin.2018-06-07T09:01:36.rjoWi7"?.
[09:27:18] -!- oaken-source has joined #archlinux32
[09:33:14] -!- abaumann has joined #archlinux32
[09:33:15] <buildmaster> Hi abaumann!
[09:33:19] <abaumann> morns. :-)
[09:33:21] <deep42thought> Hi abaumann!
[09:34:19] <buildmaster> girls, my database is dirty again ...
[09:48:01] * buildmaster resumes sanity.
[10:12:05] * buildmaster failed to execute a mysql query - can you have a look at "tmp.mysql-functions.query.stdin.2018-06-07T10:12:05.bsrnPM"?.
[10:15:35] <deep42thought> tyzoid: here, you go, the sanity check is now stricter: https://buildmaster.archlinux32.org
[10:15:36] <phrik> Title:sanity of the buildmaster's mysql database (at buildmaster.archlinux32.org)
[10:17:37] <deep42thought> gnaaaa ... delete-packages refuses to run, because the buildmaster is not sane, but I need to run it to make him sane again ...
[10:18:17] <deep42thought> we should introduce a "run-no-matter-what" switch
[10:25:18] <abaumann> henn-and-egg-buildmaster.
[10:26:13] <deep42thought> well, the current solution is as simple as commenting out that check :-)
[10:27:24] -!- alexandre9099 has quit [Quit: ZNC 1.7.0 - https://znc.in]
[10:29:31] * buildmaster resumes sanity.
[10:29:35] -!- alexandre9099 has joined #archlinux32
[10:30:15] <deep42thought> abaumann: If you get tired of fixing packages, todo#11 ("build other architectures") has now its dependencies solved - so you can think of / implement some way to build other architectures, too with build-packages (or a similar script)
[10:30:41] <deep42thought> tl;dr: the buildmaster now supports multiple architectures
[10:30:47] <abaumann> oh. that's good news. :-)
[10:31:03] <deep42thought> though, they're not yet "activated"
[10:31:06] <abaumann> I have the 486 builder almost ready on tyzoids vm
[10:31:33] <abaumann> I would like to change the way 486 diffs in PKGBUILDs are done from my bootstrap32 way to the official Archlinux32 way.
[10:31:47] <abaumann> this way it will be easier in the future to do a CARCH if in PKGBUILD diffs for instance
[10:32:00] <abaumann> as voidlinux is handling multiple architectures.
[10:32:19] <abaumann> (and archlinux did in the past ;-) )
[10:32:51] <deep42thought> do we want one source repository for each architecture or one for all together?
[10:33:12] <abaumann> I could do a fork for now?
[10:33:15] <abaumann> of the one repository.
[10:33:24] <abaumann> I don't expecting that many diffs actually
[10:33:35] <deep42thought> ok, so one repo would be better, then
[10:33:44] <abaumann> from a maintainability point of view, yes, I think so.
[10:34:28] <deep42thought> and the differences between i486 and i686 are made by CARCH switches?
[10:34:48] <abaumann> as an idea.
[10:34:53] <deep42thought> sounds reasonable
[10:35:18] <deep42thought> only downside: when we change a only-i486 thing in the PKGBUILD, all architectures will get rescheduled
[10:35:27] <abaumann> true.
[10:35:31] <deep42thought> but this should not be a big problem, I think
[10:35:39] <abaumann> and another drawback, what is CARCH for pentium3?
[10:35:45] <deep42thought> "pentium3"
[10:35:47] <deep42thought> ?
[10:35:50] <abaumann> ah?
[10:35:56] <deep42thought> why not?
[10:35:57] <abaumann> ok then. :-)
[10:36:08] <deep42thought> we need to patch pacman anyway
[10:36:22] <abaumann> for pentium3, not for i486
[10:36:24] <deep42thought> so a "i686" package can be installed on a "pentium3" pacman.conf system
[10:36:39] <deep42thought> well, also an i486 package should be installable on i686
[10:37:13] <deep42thought> how do we want to mark for which architecture packages should be built?
[10:37:15] <abaumann> What about having only two architecture: "i686-top-nock-SSE2-enabled" and "i486-runs-everywhere"
[10:37:35] <deep42thought> what's the advantage?
[10:37:41] <abaumann> yeah. we could add a variable like the ARM people.
[10:37:41] <deep42thought> 2/3 of build time?
[10:37:49] <abaumann> we have only two builds.
[10:37:55] <abaumann> and no pacman-patching.
[10:37:59] <deep42thought> re variable: but then we need to "modify" each package
[10:38:09] <abaumann> oh.
[10:38:41] <abaumann> having a separate file with a package/arch matrix?
[10:38:46] <deep42thought> my idea was, that we only build packages for different archs where it actually makes some difference
[10:39:03] <abaumann> but this we have to record again somewhere?
[10:39:06] <deep42thought> re matrix: sounds hard to maintain (think of changing names)
[10:39:11] <deep42thought> yes
[10:39:12] <deep42thought> :-(
[10:39:29] <deep42thought> I'm just brainstorming here
[10:39:37] <deep42thought> maybe these ideas are all terrible :-D
[10:39:39] <abaumann> I'm co-storming :-)
[10:40:06] <abaumann> let me see, if the slave can be made to run and let's think about some ideas..
[10:40:06] <deep42thought> we could explicitely mark packages as "makes no difference for arch $xy" in their pkgbuild
[10:40:31] <deep42thought> preferably _huge_ packages, that really make no difference
[10:40:41] <deep42thought> ah, and we need to add architectures to our black-list!
[10:40:58] <abaumann> oh. yes, that's another thing.
[10:41:02] <deep42thought> maybe some package cannot be built for i486 and i686, but for pentium3 it works
[10:41:37] <abaumann> yeah. the more gui-ish, company, etc. the less likely they will build on ancient architectures.
[10:42:47] <abaumann> Another issue is the broken package list, we may want to see architectures there..
[10:43:06] <deep42thought> that's hard
[10:43:10] <abaumann> ah. ok.
[10:43:15] <deep42thought> but not impossible
[10:43:22] <abaumann> priorization comes also to my mind.
[10:43:28] <deep42thought> because one build assignment may build multiple architectures
[10:43:29] <abaumann> firefox on 486 may not have a high priority.
[10:43:37] <deep42thought> (e.g. any doc and i686 binaries)
[10:43:42] <abaumann> ah. yes
[10:44:02] <deep42thought> well, there is a "priority" column for build assignments already
[10:44:14] <deep42thought> I may just need to expose it to the email interface
[10:44:14] <abaumann> ah. that's handy.
[10:47:05] <abaumann> mmh. I have to restore my personal forks from git.archlinux32.org instead of github.com first for some repos I hacked for i486.
[10:47:17] <abaumann> so, I'm busy for now. :-)
[10:47:57] <deep42thought> ok, have fun!
[10:48:28] <abaumann> I will, I have a terrible mess on that machine..
[10:52:49] <deep42thought> ok, showing the arch was easier than I thought: the build assignment also has an architecture assigned (which tells, what slave can build it)
[10:53:57] <abaumann> ah. so slaves also need a configuration option for arch?
[10:54:10] <deep42thought> or they auto-detect
[10:57:00] <deep42thought> but an option might be the better option, because we wil lack i686 slaves, because most slaves will be pentium3, I guess
[10:57:19] <abaumann> yes. and autodetection is not always reliable.
[10:57:53] <deep42thought> but that's all stuff, to be considerd under todo#11
[10:58:15] <abaumann> BTW jalv: the double lilv-0.22 and lilv-0.24 packages, you found the reason for that?
[10:58:26] <deep42thought> yes
[10:58:29] <deep42thought> it's a known issue
[10:58:33] <deep42thought> ... known to me
[10:58:45] <abaumann> ah. :-)
[10:58:50] <abaumann> good.
[10:58:54] <deep42thought> https://packages.archlinux32.org and https://packages.archlinux32.org
[10:58:56] <phrik> Title:Todos in the build scripts (at packages.archlinux32.org)
[10:59:07] <deep42thought> it is now included in the sanity-check
[10:59:09] <abaumann> ah. I remember: hopping renames
[10:59:12] <deep42thought> so we should notice it earlier
[11:04:51] -!- oaken-source has quit [Ping timeout: 256 seconds]
[11:05:30] -!- oaken-source has joined #archlinux32
[11:10:34] <abaumann> mmh. systemd is quite unstable lately, I always have to reboot my build host: Failed to attach 6114 to compat systemd cgroup /user.slice/user-0.slice/session-c1.scope/payload: No such file or directory
[11:10:59] <deep42thought> this happened to me, when I aborted a build with too many ^C's
[11:11:12] <abaumann> * abaumann feels guilty now
[11:11:18] <abaumann> got me.
[11:11:20] <abaumann> so, no Ctrl-C
[11:11:25] <deep42thought> only once!
[11:11:34] <deep42thought> and give it some time to clean up afterwards
[11:11:40] <abaumann> ok.
[11:11:52] <deep42thought> problem is: build-packages will try the next straw :-/
[12:00:37] -!- eduardoeae has joined #archlinux32
[12:38:27] -!- oaken-source has quit [Ping timeout: 240 seconds]
[12:53:01] -!- oaken-source has joined #archlinux32
[12:59:35] -!- oaken-source has quit [Ping timeout: 276 seconds]
[13:19:27] * buildmaster goes insane.
[13:22:38] * buildmaster resumes sanity.
[13:36:55] -!- oaken-source has joined #archlinux32
[14:14:52] <abaumann> systemd nightmare: error: unable to set user and group to '65534:78' on '/var/lib/libvirt/images/arch32-staging.qcow2': Success
[14:15:15] <abaumann> the next step will be that systemd stored the users in global database or what?
[14:15:40] <deep42thought> should I add a users and groups table to the buildmaster's database - for systemd?
[14:16:00] <abaumann> the funny part is: the group and the user exist in /etc/passwd, /etc/group
[14:16:16] <deep42thought> but not in the chroot?
[14:16:26] <abaumann> good point.
[14:16:38] <abaumann> but I'm trying to start a libvirtd managed machine.
[14:17:51] <abaumann> Jun 07 14:17:43 eurobuild3 libvirtd[12868]: 2018-06-07 12:17:43.141+0000: 12873: error : virSecurityDACSetOwnershipInternal:620 : unable to set user and group to '65534:78' on '/var/lib/libvirt/images/arch32-staging.qcow2': Success
[14:17:59] <abaumann> seems to be a libvirtd issue.
[14:18:32] <deep42thought> hmm, dunno - quantum physics don't tell me what the issue might be here ;-)
[14:18:35] <abaumann> I completly fail to see, why a virtual image should get group permissions for systemd-journal-remote
[14:19:02] <abaumann> maybe the user/groupids are now quantum-based and will eventually get right? ;-)
[14:19:15] <deep42thought> you just need to _observe_ them
[14:19:23] <abaumann> chmod 0666 /var/lib/libvirt/images/*
[14:19:36] <abaumann> yes, this helps. Exactly what the security responsible person should do. :->
[14:19:47] <abaumann> ah. I didn't observer enough :-)
[14:20:22] <abaumann> error: success usuaslly means programming mistake: errno of some function was 0 and most likely has nothing to do with the failing piece of code..
[14:21:10] <abaumann> aha: https://bugs.archlinux.org
[14:21:11] <phrik> Title:FS#58894 : [libvirt] Set user and group failed after upgraded to 4.4.0-1 (at bugs.archlinux.org)
[14:22:22] <abaumann> chapeau.. Archlinux has one of the best wiki, knowledge bases and bug hunting capabilities. :-)
[14:22:45] <deep42thought> and is most time first front when experiencing bugs :-)
[14:22:56] <abaumann> true. :-)
[14:24:59] <abaumann> ok. libvirtd is running as root and changes permissions of virtual machines.. in how many ways is this wrong? :-)
[14:25:13] <deep42thought> two
[14:25:19] <deep42thought> or was that a rethorical question?
[14:25:33] <abaumann> even when counted with quantum physics. lol. :-)
[15:12:30] -!- isacdaavid has joined #archlinux32
[15:37:40] <tyzoid> deep42thought: I think that just makes you uniquely qualified to hunt heisenbugs
[15:38:07] <deep42thought> :-D
[15:38:25] <deep42thought> well, there are other physisicst which can do that, too ;-)
[15:38:41] * deep42thought can't spell
[15:38:53] * tyzoid gives deep42thought a pass
[15:50:48] <deep42thought> abaumann: package_gcc() fails with "i686-pc-linux-gnu/libsanitizer/tsan: No such file or directory."
[15:50:59] <deep42thought> this sounds like some pretty fundamental thing is missing O.o
[16:02:42] -!- NoobAlice has joined #archlinux32
[16:22:26] -!- isacdaavid has quit [Quit: Leaving.]
[16:24:56] -!- deep42thought has quit [Quit: Leaving.]
[19:06:36] -!- AndrevS has joined #archlinux32
[19:27:16] -!- brosame has joined #archlinux32
[20:03:17] -!- brosame has quit [Ping timeout: 276 seconds]
[20:07:01] -!- brosame has joined #archlinux32
[20:15:22] -!- brosame has quit [Ping timeout: 264 seconds]
[20:32:42] -!- brosame has joined #archlinux32
[20:48:15] -!- deep42thought has joined #archlinux32
[20:48:15] <buildmaster> Hi deep42thought!
[20:55:18] <deep42thought> I really hate those haskell packages and what is needed to build them ...
[21:03:31] <tyzoid> Yup
[21:03:53] <tyzoid> You ever reach out to the haskell channels?
[21:04:03] <deep42thought> yes
[21:04:11] <tyzoid> I sent a message a while ago, but didn't hear anything back
[21:04:19] <deep42thought> they need to be built twice
[21:04:35] <deep42thought> gimme a sec to dig up the bug report while finishing my dinner
[21:04:47] <tyzoid> that's one late dinner xD
[21:05:01] <deep42thought> it's 9pm here
[21:05:04] <deep42thought> so not that late :-)
[21:05:38] <deep42thought> https://bugs.archlinux.org
[21:05:38] <phrik> Title:FS#58797 : [haskell-colour] [haskell-test-framework] [haskell-ansi-terminal] circular dependencies (at bugs.archlinux.org)
[21:06:12] <deep42thought> "This requires a two-pass build, the first pass changes --enable-tests to --disable-tests and builds with --nocheck, and the second pass changed back. Note that the soname will also change when flags were changed."
[21:06:51] <deep42thought> I think, we should somehow employ this automatically
[21:08:56] -!- brosame has parted #archlinux32
[21:11:28] <deep42thought> something like a "haskell-xy-nocheck" which is an artificial build_assignment, created from "haskell-xy" by doing the change mentioned by felixonmars
[21:11:42] <deep42thought> the built package would enter build-support (immediately)
[21:11:51] <deep42thought> and would provide "haskell-xy"
[21:15:25] <deep42thought> ok, I'll be off for the night - if anyone has any suggestions on how to handle the haskell nightmare, just say it and I'll read it in the logs :-)
[21:15:27] -!- deep42thought has quit [Quit: Leaving.]
[22:53:05] -!- AndrevS has quit [Quit: umount /dev/irc]
[23:04:25] -!- oaken-source has quit [Quit: leaving]