#archlinux32 | Logs for 2020-02-14

[01:04:01] -!- isacdaavid has quit [Quit: Leaving.]
[02:48:23] -!- DepositePirate has quit [Ping timeout: 240 seconds]
[02:50:49] -!- DepositePirate has joined #archlinux32
[07:27:11] -!- abaumann has joined #archlinux32
[07:27:11] <buildmaster> Hi abaumann!
[07:27:11] <buildmaster> !rq abaumann
[07:27:12] <phrik> buildmaster: <abaumann> my "not" key on the keyboard is broken ;-)
[07:29:03] -!- deep42thought has joined #archlinux32
[07:29:04] <buildmaster> Hi deep42thought!
[07:29:04] <buildmaster> !rq deep42thought
[07:29:04] <phrik> buildmaster: <deep42thought> you should regularly clean and shower your slaves
[07:29:10] <abaumann> hi deep42thought
[07:29:11] <deep42thought> good morning, abaumann!
[07:29:26] <deep42thought> I think, we hit levi with the glibc bug :-/
[07:29:37] <abaumann> oh.
[07:29:54] <deep42thought> but nothing critical, it seems, re could recover by downgrading
[07:30:15] <deep42thought> https://bbs.archlinux32.org
[07:30:17] <phrik> Title: My eeepc failed to boot after updating on Tuesday / Pacman / Pacman Upgrades / Arch Linux 32 Forum (at bbs.archlinux32.org)
[07:30:22] <abaumann> yep.
[07:31:50] <abaumann> mmh. still on 3999 packages todo
[07:31:58] <abaumann> come back (shortly) later - I cannot lock build list.
[07:32:04] <deep42thought> btw: encrypted swap on i486 works now
[07:32:11] <abaumann> ah, yes.
[07:32:12] <abaumann> cool
[07:32:21] <deep42thought> afk, breakfast
[07:32:27] <abaumann> ok. cu
[07:35:22] <trotz> 2020/02/14 07:34 OK buildmaster OS updates 0 updates, 0 ignored
[07:57:11] -!- deep42thought has quit [Quit: Leaving.]
[08:10:40] <abaumann> return-assignment are running and the buildmaster is blocked: come back (shortly) later - I cannot lock build list
[08:10:50] <abaumann> tried some restarts and cleanups, no help..
[08:48:50] -!- deep42thought has joined #archlinux32
[08:48:50] <buildmaster> Hi deep42thought!
[08:48:50] <buildmaster> !rq deep42thought
[08:48:51] <phrik> buildmaster: <deep42thought> I have the impression, their main operating area is television via cable
[08:48:57] <deep42thought> abaumann: I'll take a look
[08:51:35] <deep42thought> it's an interpret-mail running for almost 2 h :-/
[08:51:46] <abaumann> uh. long mail. :-)
[08:51:51] <deep42thought> lol
[08:52:09] <deep42thought> more likely, that I broke it ;-)
[08:53:24] <deep42thought> ah, there are intentions left
[08:53:39] <abaumann> good intentions, I hope ;-)
[08:54:00] <deep42thought> a good load of intentions, at least
[08:54:36] <deep42thought> empty intentions :-D
[09:00:12] <deep42thought> the buildmaster's find has a different default regextype than my find ...
[09:00:23] <abaumann> a better one? ;-)
[09:00:31] <abaumann> that's weird..
[09:00:46] <deep42thought> one that finds less packages
[09:06:07] <deep42thought> hehe, now we have a few succesful returns with execution durations of ~20 min again :-D
[09:07:45] <abaumann> --system-call-filter="clock_nanosleep_time64"
[09:07:46] <abaumann> clock_nanosleep_time64(CLOCK_REALTIME, 0, {tv_sec=5, tv_nsec=0}, 0xfffe226c) = -1 EPERM (Operation not pe
[09:07:49] <abaumann> --system-call-filter="~clock_nanosleep_time64"
[09:07:51] <abaumann> write(2, "cannot read realtime clock", 26cannot read realtime clock) = 26
[09:08:05] <abaumann> this is weird. I'm getting a different strace behaviour based on whether I set the seccomp filter
[09:08:24] <abaumann> but both go into the direction of not allowing sleep to call nanosleep
[09:08:57] <abaumann> this whole idea of securing a monolithic kernel with a userlang-thing like systemd/capabilities/seccomp drives me crazy..
[09:09:05] <abaumann> ..instead of doing proper micro-kernel architectures
[09:09:20] <abaumann> sadly all major Linux distributions go into that direction
[09:10:08] <deep42thought> one can do micro kernels with linux?
[09:10:25] <abaumann> when you rewrite the linux kernel, yes. :-;2~>
[09:10:29] <deep42thought> lol
[09:10:53] <abaumann> they instist to write it in C with a monotithic architecture for performance reasons..
[09:10:57] <abaumann> ..that's so 1990
[09:11:38] <abaumann> CAP_SYS_RESOURCE: allow more than 64hz interrupts from the real-time clock;
[09:11:39] <abaumann> mmh.
[09:13:39] * abaumann reads glibc 2.31 news and system-nspawn source code
[09:13:53] -!- deep42thought has quit [Remote host closed the connection]
[09:14:18] -!- deep42thought has joined #archlinux32
[09:14:18] <buildmaster> Hi deep42thought!
[09:14:19] <buildmaster> !rq deep42thought
[09:14:19] <phrik> buildmaster: <deep42thought> doesn't one usually try to avoid rust on metals?
[09:14:39] <deep42thought> there was also one patch dropped upstream (but which seems unrelated)
[09:14:44] <deep42thought> https://git.archlinux.org
[09:14:45] <phrik> Title: svntogit/packages.git - Git clone of the 'packages' repository (at git.archlinux.org)
[09:15:11] <abaumann> yeah, this is about misreading some ELF-stuff
[09:15:38] <abaumann> in the glibc news there are many many changes around time functions, so..
[09:16:26] <abaumann> * System call wrappers for time system calls now use the new time64 system calls when available. On 32-bit targets, these wrappers attempt to call the new system calls first and fall back to the older 32-bit time system calls if they are not present. This may cause issues in environments that cannot handle unsupported system calls gracefully by returning
[09:16:31] <abaumann> -ENOSYS. Seccomp sandboxes are affected by this issue.
[09:16:53] <abaumann> we where asked some time ago on IRC about whether we bother about changing :-)
[09:16:55] <deep42thought> you found it :-)
[09:17:04] <deep42thought> right
[09:17:06] <deep42thought> I remember
[09:17:20] <abaumann> we get EPERM though, not ENOSYS
[09:18:25] <buildmaster> girls, please have a look at my dirty database
[09:18:25] * buildmaster goes insane.
[09:18:35] <girls> buildmaster: sure
[09:18:39] <abaumann> the funny thing is: when I use clock_nanosleep_time64 in a seccomp filter for systemd-nspawn, it sort of doesn't have an effect..
[09:21:28] <abaumann> clock_nanosleep_time64 is in @default, so "System calls that are always permitted"
[09:21:32] <abaumann> mmh.
[09:23:11] -!- T`aZ has quit [Remote host closed the connection]
[09:25:29] <abaumann> https://sourceware.org
[09:25:31] <phrik> Title: Szabolcs Nagy - Re: glibc in master is incompatible with systemd-nspawn (at sourceware.org)
[09:26:01] <abaumann> ah, this at least explains why there is an EPERM suddenly..
[09:27:12] <abaumann> ok, a system call not being available can have two reasons: a) it's not there in the kernel b) some filter system doesn't allow us to use it, but it's there.
[09:27:26] <abaumann> I'm actually more in the EPERM-fraction in this case. :-)
[09:28:36] <abaumann> exec ${CARCH:+setarch "$set_arch"} systemd-nspawn
[09:28:41] <abaumann> in arch-nspawn
[09:28:50] <abaumann> why not set a --personality=x86 ?
[09:28:55] <abaumann> and does it have the same effect
[09:29:08] * abaumann had a free retorical question of the day for free
[09:29:11] * buildmaster resumes sanity.
[09:29:30] <deep42thought> abaumann: would it set CARCH, too?
[09:29:38] <abaumann> ah. no.
[09:29:40] <deep42thought> that variable is queried inside makepkg
[09:29:49] <abaumann> but uname would report i686 or so
[09:29:50] <deep42thought> ah, wait
[09:29:51] <deep42thought> nevermind
[09:30:04] <abaumann> ah.
[09:30:11] <deep42thought> CARCH *is* already set
[09:30:12] <abaumann> I meant additionally set a personality
[09:30:15] <deep42thought> it's just queried
[09:30:43] * abaumann starts to use strace and gdb on a debug version of systemd-nspawn
[09:33:28] <abaumann> https://bugzilla.redhat.com
[09:33:30] <phrik> Title: 1770154 – systemd: Default seccomp filter in systemd-spawn breaks userspace ABI (at bugzilla.redhat.com)
[09:33:38] <buildmaster> i686/vis are broken (says nlopc46): https://archlinux32.org
[09:34:38] <abaumann> "systemd-nspawn in its current state is not useful for building packages."
[09:34:41] <abaumann> ahem.
[09:35:37] <abaumann> policy-driven security like SElinux or seccomp have the problem that things get out of sync sometimes :-)
[09:36:08] <deep42thought> "The failure is visible with and without pkeys because applications expect ENOSYS and apply a fallback, but not EPERM. glibc will use utime if utimensat_time64 fails with ENOSYS, but not in the case of EPERM."
[09:36:09] <abaumann> the "problem" with systemd-nspawn is, that you cannot white list a unknown system call.
[09:36:53] <deep42thought> so we might get away by switching from EPERM to ENOSYS in glibc?
[09:36:54] <abaumann> I'm currently checking on systemd itself, there must be pull-request for this
[09:37:05] <abaumann> aeh. this is not a good idea.
[09:37:14] <deep42thought> :-D that's me :-D
[09:37:20] <abaumann> uff :-)
[09:37:30] <abaumann> I'm used to fight stupid ideas on a daily basis..
[09:37:43] <abaumann> ..so I thought the comment was an actualy suggestion.. :-)
[09:37:44] <deep42thought> (the suggestion was real, but it might still have been totally stupid)
[09:38:04] <abaumann> no, I think we need a patch for the host systemd
[09:38:11] <abaumann> which might already exist
[09:38:15] <abaumann> "utimensat_time64\0"
[09:38:20] <abaumann> at least this one got in:
[09:38:24] <abaumann> https://github.com
[09:38:26] <phrik> Title: Add more syscalls to the seccomp lists by keszybz · Pull Request #13975 · systemd/systemd · GitHub (at github.com)
[09:38:37] <abaumann> this was about not being able to set time of files when installing glibc 2.31
[09:39:06] <abaumann> I'm always amazed why people are not able to simply grep all system calls from the kernel sources and add them to their packages..
[09:39:21] <abaumann> ..on the other hand there should be _one_ header file listing all those call names
[09:39:29] <abaumann> per architecture
[09:39:52] <abaumann> "hmm, i figure we might miss some more i386 syscalls, since noone of us ever runs "systemd-analyze syscall-filters" on any host that's not x86-64, and we need to add these syscalls manually..."
[09:39:59] <abaumann> original comment by Mr. systemd
[09:40:11] <deep42thought> yeah, I just read it :-D
[09:40:26] <abaumann> that's why he gets some hate sometimes :-)
[09:41:04] <deep42thought> so what he's saying is "systemd is developed and tested on x86_64 only"
[09:41:13] <abaumann> exactly
[09:41:37] <abaumann> and if systemd gets into a distro, this defacto makes linux a x86_64-only ecosystem
[09:42:03] <deep42thought> well, this sounds *exactly* like the piece of software, that I would put into charge of almost all linux systems on all architectures :-)
[09:42:04] <abaumann> to be fair, if you fix it and add a pull request, it gets added to the master
[09:43:40] <deep42thought> I still don't understand what systemd-nspawn currently does wrong which should be corrected
[09:43:49] <deep42thought> (sry for my slowness)
[09:43:59] <abaumann> It has a list of known system calls
[09:44:11] <abaumann> in the code
[09:44:32] <abaumann> then there is a default whitelist of system calls, which are fine to be called inside an spwanned child
[09:44:47] <abaumann> now, this list is not complete, it lacks some system calls from the 32-bit kernel
[09:45:39] <abaumann> glibc 2.31 moves into the direction of 64-bit time functions, so the 32-bit kernel somehow has to provide such functions along to the old 32-bit time functions
[09:45:52] <abaumann> the glibc developers then call differnt syscalls
[09:46:01] <abaumann> depending on the architecture
[09:46:10] <abaumann> now, we run on a 64-bit kernel
[09:46:14] <deep42thought> and systemd-nspawn needs to understand them in order to correctly handle them
[09:46:19] <abaumann> exactly
[09:46:31] <deep42thought> it cannot just pass all system calls to the kernel
[09:46:34] <abaumann> you cannot whitelist something the systemd seccomp filter doesn't know about
[09:46:49] <deep42thought> this is a strange design
[09:46:49] <abaumann> that would render whitelisting a little bit pointless :-)
[09:47:15] <abaumann> well, the design per-se is not so bad.
[09:47:25] <deep42thought> so each syscall of the kernel needs to be advertised in systemd-nspawn, too
[09:47:26] <abaumann> syscalls are different per architecture in Linux, that's just a fact
[09:47:37] <abaumann> yes.
[09:47:37] <deep42thought> ok, and they need translation
[09:47:41] <deep42thought> ok, got it
[09:47:43] <deep42thought> thanks :-)
[09:47:47] <abaumann> np :-)
[09:48:12] <deep42thought> should I teach you something about nonlinear optics in return?
[09:48:13] <abaumann> strange design, yes
[09:48:36] -!- T`aZ has joined #archlinux32
[09:48:41] <abaumann> but as long as Linux is monolithic and you try to jail the containers, this is the way to go.
[09:48:57] <abaumann> I'm not confident we get into an easy-to-understand and secure system this way
[09:48:57] <deep42thought> yeah, I see
[09:49:31] <abaumann> oh dri300 broken for radeon in mesa, cool. My mac is text-only now. :-)
[09:50:18] <abaumann> https://github.com
[09:50:19] <phrik> Title: Add more syscalls to the seccomp lists by keszybz · Pull Request #13975 · systemd/systemd · GitHub (at github.com)
[09:50:47] <abaumann> should be merged *puzzle8
[09:51:05] <deep42thought> in master
[09:51:09] <deep42thought> but is it released?
[09:51:09] <abaumann> yes
[09:52:28] <abaumann> git+https://github.com/systemd/systemd-stable#tag=${_tag}?signed git+https://github.com/systemd/systemd#tag=v${pkgver%.*}?signed with b7ed902b2394f94e7f1fbe6c3194b5cd9a9429e6 and 244.3
[09:52:29] <phrik> Title: GitHub - systemd/systemd-stable: Backports of patch from systemd git to stable distributions (at github.com)
[09:53:00] <abaumann> a little bit hard to see.
[09:53:22] <abaumann> grep -r clock_nanosleep_time *
[09:53:22] <abaumann> systemd/src/shared/seccomp-util.c: "clock_nanosleep_time64\0"
[09:53:39] <abaumann> yeah, should be in the @default list
[09:53:51] * deep42thought clones systemd
[09:54:08] <abaumann> but, if glibc 2.31 is compiled for a 32-bit kernel on a 64-bit chroot, maybe things go really wrong..
[09:54:37] <abaumann> I personally don't feel comfortable with build slaves on 64-bit and systemd-nspawnd
[09:54:44] <abaumann> I rather like the approach of archlinux ARM
[09:54:59] <abaumann> they don't have a choice, they must run it on original hardware
[09:55:27] <abaumann> besided. cmake and some CPU sniffing scripts (autoconf,meson) more and more make wrong guesses about the target architecture
[09:55:43] <abaumann> and as libvirt itself is not emulating some things correctly..
[09:57:30] <deep42thought> yeah, but building on 32-bit hardware has two major disadvantages: 1. speed, 2. availability
[09:57:41] <abaumann> true true
[09:57:51] <abaumann> and an also not so nice setup with ccache and distcc
[09:57:54] <deep42thought> and just consider the enormous memory requirements of some packages
[09:58:03] <abaumann> another good point
[09:58:22] <abaumann> and long-term: nobody would care anymore about cross-compilation and correct CPU-feature detection..
[09:58:29] <deep42thought> if glibc built in a setarch'ed chroot is broken, it should be fixed
[09:58:33] <abaumann> ..making the Linux land even more x86_64-ish
[09:58:47] <abaumann> I doubt that, actually
[09:58:57] <deep42thought> yes, me too, I was just saying
[09:59:00] <abaumann> usually you tell glibc the architecture you want to build for..
[09:59:05] <abaumann> ..and then it does the right thing.
[09:59:26] <abaumann> I just wanted to raise the possibility that also glibc could have been wrongly built.
[09:59:38] <deep42thought> yes
[09:59:47] <deep42thought> but I even built it in a vm and it was also broken
[09:59:49] * abaumann considers building a glibc 2.31 on a real 32-bit machine, just for comparision
[10:00:21] <abaumann> yeah, the simpler hypothesis is systemd-nspawn is not up-to-date
[10:00:23] <deep42thought> though, I think, I used devtools to build it inside the 32-bit vm
[10:04:14] <deep42thought> the commit should be in the systemd package
[10:04:21] <deep42thought> maybe yet another syscall is missing?
[10:06:11] <abaumann> mmh. the sleep fails in clock_nanosleep_time64 with EPERM, no matter whether I all --system-call-filter="clock_nanosleep_time64" or not
[10:07:21] <abaumann> clock_gettime64?
[10:07:24] <buildmaster> i686/signal-desktop is broken (says eurobuild6-2): https://archlinux32.org
[10:08:45] <buildmaster> i486/broadcom-wl is broken (says eurobuild6-7-i486): https://archlinux32.org
[10:08:52] <abaumann> shared/seccomp-util.c lists both of then in @default
[10:09:37] <abaumann> gpg: can't connect to the agent: IPC connect call failed
[10:09:45] <abaumann> this is inter process communication
[10:09:59] <abaumann> @ipc
[10:10:37] -!- deep42thought has parted #archlinux32
[10:10:40] -!- deep42thought has joined #archlinux32
[10:10:41] <buildmaster> Hi deep42thought!
[10:10:41] <buildmaster> !rq deep42thought
[10:10:41] <phrik> buildmaster: <deep42thought> computers ranked by dirt: smokers, car mechanics, bedroom, ...
[10:13:07] <buildmaster> any/python-aioresponses are broken (says rechenknecht): https://archlinux32.org
[10:21:43] <deep42thought> regarding build back-log: I removed gcc along with glibc (maybe that was wrong) from the build-list, now libtool cannot be built, because it depends on gcc>=9.2.1 (which is yet to be built, but I removed it)
[10:22:31] <deep42thought> suggestion: we build a new gcc, get things rolling now, and once we (you) sort out the glibc issue, we build glibc and gcc once more
[10:24:33] <abaumann> we hope gcc doesn't depend on glibc
[10:24:44] * deep42thought checks this
[10:25:10] <deep42thought> not according to its PKGBUILD
[10:25:46] <abaumann> then that should work find
[10:25:47] <abaumann> *dine
[10:25:50] <abaumann> *fine
[10:26:02] <buildmaster> i486/nvidia-390xx is broken (says eurobuild6-7-i486): https://archlinux32.org
[10:26:02] <abaumann> damn those fingers..
[10:27:52] <buildmaster> pentium4/python-cffi is broken (says eurobuild6-4): https://archlinux32.org
[10:29:08] <buildmaster> any/python-oslo-i18n is broken (says eurobuild6-6): https://archlinux32.org
[10:31:55] -!- z3ntu_ has joined #archlinux32
[10:32:03] <buildmaster> any/python-oslo-context is broken (says eurobuild6-4): https://archlinux32.org
[10:36:47] <buildmaster> any/python-mox3 is broken (says eurobuild6-4): https://archlinux32.org
[10:39:34] <buildmaster> any/python-os-service-types are broken (says eurobuild6-1): https://archlinux32.org
[10:43:37] <buildmaster> any/python-os-client-config is broken (says rechenknecht): https://archlinux32.org
[10:45:13] <deep42thought> abaumann: your build slaves are hanging in the "checking-source" state of gcc for quite some time - is everything all-right, there?
[10:45:20] <abaumann> mmh.
[10:45:22] <abaumann> I'll check..
[10:45:44] <abaumann> Cloning gcc git repo
[10:45:48] <deep42thought> ok
[10:45:58] <abaumann> the disks are not the fastest ones..
[10:46:06] <deep42thought> ok, that's a valid excuse
[10:46:07] <abaumann> ..and cloning gcc three times is also not helping :-)
[10:46:16] <deep42thought> twice ;-)
[10:46:28] <abaumann> ah. yes
[10:46:34] <abaumann> one is on nlopc46
[10:47:13] <abaumann> yeah. both are just happily cloning.. and 5 other slaves are on the same machine doing stuff.
[10:47:25] <abaumann> I might have to think again about an SSD cache or some more platters
[10:47:26] <deep42thought> :-D
[10:47:28] <deep42thought> no trouble
[10:47:44] <deep42thought> I just wanted to make sure, that they're not unlawfully stuck
[10:48:18] <buildmaster> any/python-keystoneauth1 is broken (says eurobuild6-4): https://archlinux32.org
[10:48:54] * buildmaster goes insane.
[10:48:59] <deep42thought> welllll
[10:49:04] <abaumann> oups
[10:49:11] <deep42thought> was that you?
[10:49:16] <abaumann> nope
[10:49:27] <abaumann> I'm not aware of anything
[10:49:33] <deep42thought> !grab abaumann
[10:49:34] <phrik> deep42thought: Tada!
[10:49:34] <abaumann> I reported i486 packages from testing
[10:50:34] <deep42thought> why-don't-you did not mark its mysql-query as unimportant
[10:50:46] <deep42thought> so the sanity-check thought, there was some unfinished mysql query
[10:51:16] * buildmaster resumes sanity.
[10:51:36] <buildmaster> any/python-sqlalchemy-migrate is broken (says rechenknecht): https://archlinux32.org
[10:52:26] <buildmaster> any/python-oslo-utils are broken (says eurobuild6-1): https://archlinux32.org
[10:53:24] <buildmaster> any/python-stestr is broken (says eurobuild6-6): https://archlinux32.org
[10:53:44] <buildmaster> any/repo is broken (says nlopc43): https://archlinux32.org
[10:54:24] <buildmaster> any/python-oslo-config is broken (says eurobuild6-4): https://archlinux32.org
[10:58:55] <abaumann> gcc builds now
[10:59:00] <deep42thought> :-)
[10:59:34] <deep42thought> my slave has a head-start of 19k logged lines - might yours still win?
[10:59:50] <abaumann> something positive about systemd: the repo is well-organized and the C code is very readable. :-)
[10:59:59] <abaumann> slave racing? ;-)
[11:00:13] <abaumann> ..so wrong again on so many levels..
[11:08:01] <buildmaster> any/python-openstacksdk is broken (says nlopc43): https://archlinux32.org
[11:08:46] <deep42thought> circular dependencies yay \o/
[11:09:53] <abaumann> yeah, python will need a haskell treatment soon
[11:10:31] <deep42thought> [ -z "${pkgname##python*}" ] && unset check && unset checkdepends
[11:11:06] * abaumann nods
[11:11:47] <deep42thought> :haskell_without_check: does that, already
[11:11:56] <deep42thought> "try with :without_check: iff this is a haskell-*, python-* or python2-* package"
[11:12:42] <abaumann> afk for shoppinge and early lunch
[11:12:47] <deep42thought> cu
[11:45:05] -!- T`aZ has quit [Remote host closed the connection]
[11:47:47] -!- T`aZ has joined #archlinux32
[12:52:21] <buildmaster> pentium4/libarchive is broken (says eurobuild6-6): https://archlinux32.org
[12:52:58] <buildmaster> pentium4/systemd is broken (says eurobuild6-4): https://archlinux32.org
[12:53:38] <buildmaster> pentium4/js52 is broken (says eurobuild6-6): https://archlinux32.org
[12:54:17] <buildmaster> pentium4/python-ruamel.yaml.clib is broken (says eurobuild3): https://archlinux32.org
[12:54:20] <buildmaster> i686/cjs are broken (says eurobuild6-5): https://archlinux32.org
[12:54:53] <buildmaster> i686/js60 is broken (says eurobuild6-4): https://archlinux32.org
[12:55:06] <buildmaster> pentium4/js60 is broken (says eurobuild6-6): https://archlinux32.org
[12:55:33] <buildmaster> i686/s3fs-fuse is broken (says eurobuild6-4): https://archlinux32.org
[13:09:17] <abaumann> that's me, I accidently enabled glibc 2.31 in the build slaves..
[13:09:31] <abaumann> ..aka as hacking on the productive machine running the slaves
[13:17:07] <deep42thought> but now we know, that it still does not work ;-)
[13:18:40] <abaumann> straceing systemd-nspawn is not a good idea, btw..
[13:18:47] <deep42thought> why?
[13:18:55] <abaumann> too much output
[13:18:59] <abaumann> epoll_wait
[13:19:04] <abaumann> thousands per second
[13:19:09] <deep42thought> oh
[13:19:21] <deep42thought> debugger-unfriendly binary?
[13:19:29] <abaumann> I don't know, but my experience tells me, that things you cannot debug with strace have a problem (like the Java VM)
[13:19:36] <deep42thought> or is it, because systemd-nspawn executes stuff underneith
[13:19:46] <abaumann> no, the stuff is a bash
[13:19:53] <deep42thought> :-/
[13:19:56] <abaumann> so I don't understand why polling 1000/s is a good idea
[13:20:16] <abaumann> systemd-nspawn -D /var/lib/archbuild/staging-pentium4/abaumann, sleep 5, sleep: cannot read realtime clock: Operation not permitted
[13:20:20] <abaumann> that's a good test case
[13:20:29] <deep42thought> :-)
[13:21:21] <abaumann> systemd-nspawn --system-call-filter='clock_nanosleep_time64' -D /var/lib/archbuild/staging-pentium4/abaumann
[13:21:32] <abaumann> clock_nanosleep_time64(CLOCK_REALTIME, 0, {tv_sec=5, tv_nsec=0}, 0xffea480c) = -1 EPERM (Operation not permitted)
[13:21:44] <abaumann> either I don't understand something here or this is royally borked
[13:22:36] <deep42thought> do you have a setarch inbetween there somewhere?
[13:22:37] <abaumann> 244.3 trunk, 244.2 in core
[13:22:42] <abaumann> systemd
[13:23:05] <abaumann> I rather think the missing clock_nanosleep_time64 is missing in-between
[13:23:28] <abaumann> so, I'll build a patches systemd for 64-bit to see
[13:23:43] <deep42thought> you have some idea *what* to patch?
[13:23:49] <abaumann> https://github.com
[13:23:51] <phrik> Title: Add more syscalls to the seccomp lists by keszybz · Pull Request #13975 · systemd/systemd · GitHub (at github.com)
[13:23:58] <abaumann> at least
[13:24:04] <deep42thought> this is in systemd
[13:24:08] <abaumann> yes.
[13:24:15] <deep42thought> e.g. in the built systemd on x86_64
[13:24:21] <abaumann> yep
[13:24:33] <buildmaster> i686/nfs-utils are broken (says nlopc46): https://archlinux32.org
[13:25:08] * abaumann makes a diff of two huge systemd git repos
[13:28:00] <buildmaster> i686/js68 is broken (says eurobuild6-4): https://archlinux32.org
[13:57:57] <abaumann> nah. both are the same and have the syscalls in seccom-util.c
[13:58:19] <deep42thought> that's what I meant
[13:58:24] <abaumann> ah :-)
[13:58:39] <abaumann> so completely wrong track?
[13:58:48] <deep42thought> dunno
[13:58:51] <deep42thought> it sounds so right
[13:59:18] <abaumann> more debugging :-)
[14:02:33] <deep42thought> maybe open a bug against systemd and let upstream solve it?
[14:02:41] <deep42thought> or at least: let upstream *know* of the problem
[14:03:21] <deep42thought> other question: do the syscalls work on pure i686 hardware (or a vm) with the new glibc?
[14:04:11] <abaumann> good question
[14:04:33] <abaumann> I try a real machine.. I have some of them :-)
[14:04:37] <deep42thought> I guess, it does, but we should make sure before complaining
[14:04:43] <deep42thought> :-D
[14:04:56] <deep42thought> just make sure not to close your ssh connections!
[14:05:11] <abaumann> yeah. :-)
[14:05:27] <abaumann> I also have to pick a machine I can easily reinstall (aka has a CDROM-drive)
[14:05:33] <abaumann> there are two questions actually:
[14:05:47] <deep42thought> no, why?
[14:05:50] <abaumann> does glibc 2.31 show any problems on the real system
[14:05:59] <abaumann> and what is systemd-nspawn doing from 32-bit to 32-bit
[14:06:08] <deep42thought> you just do a `pacman -U /var/cache/pacman/pkg/glibc-2.30-3.0*`
[14:06:30] <abaumann> for recovering, yes
[14:06:34] <deep42thought> keep some root terminal open during experimentation
[14:08:51] <deep42thought> buildmaster: wtp happy
[14:08:52] <buildmaster> deep42thought: i686/happy: i686/community (1.19.12-1.0)
[14:08:52] <buildmaster> pentium4/happy: pentium4/community (1.19.12-1.0)
[14:15:57] <buildmaster> dirty! girls, my database - so dirty :-(
[14:15:57] * buildmaster goes insane.
[14:16:54] <deep42thought> dammit, I thought, I had fixed that issue ...
[14:26:36] <deep42thought> I don't understand, why this still occures ...
[14:28:31] * buildmaster resumes sanity.
[15:23:07] <buildmaster> pentium4/findomain is broken (says nlopc46): https://archlinux32.org
[15:27:30] <buildmaster> i686/cargo-crev is broken (says eurobuild6-6): https://archlinux32.org
[15:28:24] <buildmaster> i686/intel-graphics-compiler is broken (says eurobuild6-5): https://archlinux32.org
[15:28:48] <buildmaster> pentium4/intel-graphics-compiler is broken (says eurobuild6-4): https://archlinux32.org
[15:33:52] <buildmaster> i686/findomain is broken (says eurobuild6-1): https://archlinux32.org
[15:34:12] <buildmaster> pentium4/cargo-crev is broken (says eurobuild6-2): https://archlinux32.org
[15:41:30] <buildmaster> i686/virtualbox-modules-arch is broken (says eurobuild6-2): https://archlinux32.org
[15:43:10] <buildmaster> pentium4/haskell-http-client is broken (says nlopc46): https://archlinux32.org
[15:43:52] <buildmaster> i686/haskell-http-client is broken (says eurobuild6-4): https://archlinux32.org
[15:44:27] <buildmaster> pentium4/haskell-network-uri is broken (says eurobuild6-5): https://archlinux32.org
[15:44:50] <deep42thought> hey buildmaster, what makes you think, you can build haskell-* packages before ghc???
[15:45:50] <buildmaster> i686/haskell-network-uri is broken (says rechenknecht): https://archlinux32.org
[15:45:56] <deep42thought> ah, they're in a loop
[15:46:10] <buildmaster> i686/haskell-http is broken (says eurobuild6-1): https://archlinux32.org
[15:46:42] <deep42thought> abaumann: btw, my archlinux32 test vm seems broken, not sure if this is the glibc issue
[15:47:14] <deep42thought> ssh fails and the emulated window is really slow and does not even release the keyboard/focus at some point
[15:47:28] -!- bill-auger has quit [Ping timeout: 265 seconds]
[15:48:16] <buildmaster> pentium4/haskell-http is broken (says eurobuild6-6): https://archlinux32.org
[15:48:19] <buildmaster> i686/haskell-warp is broken (says eurobuild6-2): https://archlinux32.org
[15:48:41] <buildmaster> pentium4/haskell-warp is broken (says nlopc46): https://archlinux32.org
[15:51:01] <deep42thought> anyway: gotta go - have a nice weekend!
[15:51:05] -!- deep42thought has quit [Quit: Leaving.]
[15:51:16] <buildmaster> pentium4/haskell-criterion is broken (says eurobuild6-5): https://archlinux32.org
[15:53:59] <buildmaster> i686/haskell-criterion is broken (says rechenknecht): https://archlinux32.org
[16:07:55] <abaumann> you to, have a nice weekend. :-)
[16:37:47] <abaumann> just short: on a 32-bit host I get the same with glibc 2.31 when calling systemd-nspawn
[16:38:45] <abaumann> sleep 5 sleep: cannot read realtime clock: Operation not permitted
[17:33:50] -!- abaumann has quit [Quit: leaving]
[17:52:49] <trotz> 2020/02/14 17:52 CRIT jeti100 Master Mirror Certificate HTTP CRITICAL - Invalid HTTP response received from host on port 443: cURL returned 28 - Timeout was reached
[18:02:49] <trotz> 2020/02/14 18:02 OK jeti100 Master Mirror Certificate OK - Certificate 'arch.eckner.net' will expire on Sun 12 Apr 2020 04:02:41 AM GMT +0000.
[18:48:30] <buildmaster> pentium4/ghc is broken (says eurobuild6-3) - already flagged out-of-date upstream on 2020-02-12: https://archlinux32.org
[20:09:52] <buildmaster> i686/haskell-storable-complex is broken (says nlopc46) - I rescheduled: haskell-base-orphans: https://archlinux32.org
[20:11:30] <buildmaster> i686/haskell-fclabels are broken (says eurobuild6-2): https://archlinux32.org
[20:16:46] <nit-picker> key A0B250C0FC9FC079EC04ADB7A50C0F20AEC3AF00 (from Polichronucci (Arch Linux 32 Master Key) <polichronucci@archlinux.gr>) in package archlinux32-keyring-transition-20200113-1.0-any.pkg.tar.zst expires on 2020-05-05 (in 80 < 100 days).
[20:18:43] -!- torv has quit [Ping timeout: 240 seconds]
[20:21:51] -!- torv has joined #archlinux32
[20:32:12] <buildmaster> i686/haskell-chunked-data is broken (says eurobuild6-4) - I rescheduled: haskell-vector: https://archlinux32.org
[20:35:22] <buildmaster> i686/haskell-critbit is broken (says eurobuild6-3): https://archlinux32.org
[20:37:52] <buildmaster> i686/haskell-chasingbottoms are broken (says eurobuild6-4) - I rescheduled: haskell-quickcheck, haskell-splitmix: https://archlinux32.org
[20:37:53] <buildmaster> i686/haskell-microspec is broken (says nlopc46): https://archlinux32.org
[20:38:22] <buildmaster> i686/haskell-data-clist is broken (says eurobuild6-5): https://archlinux32.org
[20:38:36] <buildmaster> i686/haskell-dlist is broken (says eurobuild6-2): https://archlinux32.org
[20:39:51] <buildmaster> i686/haskell-non-negative is broken (says eurobuild6-1): https://archlinux32.org
[20:40:17] <buildmaster> i686/haskell-quickcheck-io is broken (says eurobuild6-3): https://archlinux32.org
[20:48:40] <buildmaster> i686/haskell-data-dword is broken (says nlopc46): https://archlinux32.org
[20:49:05] <buildmaster> i686/haskell-spdx is broken (says eurobuild6-5) - I rescheduled: haskell-tasty-quickcheck: https://archlinux32.org
[20:49:07] <buildmaster> i686/haskell-filepath-bytestring is broken (says eurobuild3): https://archlinux32.org
[20:50:36] <buildmaster> i686/haskell-stmonadtrans are broken (says eurobuild6-3): https://archlinux32.org
[20:50:42] <buildmaster> i686/haskell-hourglass are broken (says eurobuild6-1): https://archlinux32.org
[20:53:03] <buildmaster> i686/haskell-gnuidn is broken (says nlopc46): https://archlinux32.org
[20:53:04] <buildmaster> i686/haskell-ghc-typelits-knownnat is broken (says eurobuild6-6): https://archlinux32.org
[21:00:02] <buildmaster> i686/haskell-say is broken (says eurobuild6-2): https://archlinux32.org
[21:03:30] <buildmaster> i686/haskell-fold-debounce is broken (says eurobuild3): https://archlinux32.org
[21:04:38] <buildmaster> i686/haskell-alsa-mixer is broken (says eurobuild6-5): https://archlinux32.org
[21:07:41] <buildmaster> i686/wireguard-arch is broken (says nlopc46): https://archlinux32.org
[21:08:08] <buildmaster> i686/wireguard-lts are broken (says eurobuild6-2): https://archlinux32.org
[21:12:16] <buildmaster> i486/ndiswrapper-arch is broken (says eurobuild6-7-i486): https://archlinux32.org
[21:12:39] <buildmaster> pentium4/wireguard-arch is broken (says eurobuild6-1): https://archlinux32.org
[21:14:13] -!- bill-auger has joined #archlinux32
[21:15:04] <buildmaster> pentium4/wireguard-lts are broken (says eurobuild6-4): https://archlinux32.org
[21:19:06] <buildmaster> i486/pigeonhole is broken (says eurobuild6-7-i486): https://archlinux32.org
[21:50:23] <buildmaster> pentium4/gitea is broken (says nlopc46): https://archlinux32.org
[21:51:56] <buildmaster> i686/gitea is broken (says eurobuild6-2): https://archlinux32.org
[22:01:34] <buildmaster> i686/haskell-vector is broken (says eurobuild6-5): https://archlinux32.org
[22:05:23] <buildmaster> i686/haskell-enclosed-exceptions are broken (says eurobuild6-6): https://archlinux32.org
[22:07:16] <buildmaster> i686/haskell-tasty-quickcheck is broken (says eurobuild6-5): https://archlinux32.org
[22:28:30] <buildmaster> i486/qcad is broken (says nlopc46-i486bs0): https://archlinux32.org
[22:39:58] <buildmaster> i686/haskell-shake is broken (says rechenknecht) - I rescheduled: haskell-js-jquery: https://archlinux32.org
[22:43:44] <buildmaster> i486/intel-graphics-compiler is broken (says eurobuild6-7-i486): https://archlinux32.org
[22:45:51] <buildmaster> i686/haskell-js-jquery is broken (says rechenknecht): https://archlinux32.org
[22:47:53] <buildmaster> i486/libretro-beetle-psx is broken (says nlopc46-i486bs0): https://archlinux32.org
[22:49:28] <buildmaster> i486/libretro-flycast is broken (says eurobuild6-7-i486): https://archlinux32.org
[22:56:15] <buildmaster> i486/wireguard-arch is broken (says eurobuild6-7-i486): https://archlinux32.org
[23:01:23] <buildmaster> i486/kresus are broken (says eurobuild6-7-i486): https://archlinux32.org
[23:05:14] <nit-picker> key A0B250C0FC9FC079EC04ADB7A50C0F20AEC3AF00 (from Polichronucci (Arch Linux 32 Master Key) <polichronucci@archlinux.gr>) in package archlinux32-keyring-20200113-1.0-any.pkg.tar.zst expires on 2020-05-05 (in 80 < 100 days).
[23:35:40] <buildmaster> i486/go is broken (says eurobuild6-7-i486): https://archlinux32.org