#archlinux32 | Logs for 2022-09-18

Back
[02:21:07] -!- lithiumpt has quit [Ping timeout: 248 seconds]
[02:50:59] -!- lithiumpt has joined #archlinux32
[09:37:04] -!- abaumann has joined #archlinux32
[09:37:04] <buildmaster> Hi abaumann!
[09:37:05] <buildmaster> !rq abaumann
[09:37:05] <phrik> buildmaster: <abaumann> the Linux kernel goes the Firefox way a) create a foundation, b) let everybody code on it, c) build times exceed the 20 minute margin on modern hardware d) reprogram everything in rust
[09:37:11] <abaumann> https://bugs.archlinux32.org
[09:37:13] <phrik> Title: FS#281 : Firefox compilation (at bugs.archlinux32.org)
[09:37:23] <abaumann> I think, the differences are there because I build against staging..
[09:37:56] <abaumann> The build is almost completely working, but in the end failing with obscure errors in the BFD ld..
[09:38:22] <abaumann> Oh, rechenknecht is really efficient and spits out signing errors.. :-)
[09:38:35] <abaumann> And my 486 build slaves does "Process 8050 (python) of user 1000 dumped core."
[09:39:01] <abaumann> That would be one problem with buildbot: it requires python, and python currently starts to depend on rust and is not exactly stable on i486..
[09:39:28] <abaumann> that said, python _should_ probably be stable for completely different reasons like meson and such..
[09:40:34] <abaumann> I disabled rechenknecht for now..
[09:46:25] <abaumann> The python thingy is python itself heavily segfaulting on i486 in almost all tests..
[09:50:04] <bill-auger> abaumann: i left you a note with memosrv
[09:51:08] <bill-auger> FYI firefox 104.0.1 is a "go", with these caveats:
[09:51:08] <bill-auger> [[ "${CARCH}" == 'i686' ]] && makedepends=( ${makedepends[*]/wasi-compiler-rt/wasi-compiler-rt=13.0.1-1.0} ) # dustbin
[09:51:08] <bill-auger> [[ "${CARCH}" == 'i686' ]] && makedepends=( ${makedepends[*]/wasi-libc++*/} )
[09:51:08] <bill-auger> [[ "${CARCH}" == 'i686' ]] && makedepends+=( wasi-libc++=13.0.1-1.0 wasi-libc++abi=13.0.1-1.0 ) # dustbin
[09:51:08] <bill-auger> [[ "${CARCH}" == 'i686' ]] && makedepends+=( llvm13 )
[09:57:05] <abaumann> ah, thanks. I saw it.
[09:57:18] <abaumann> you are building against stable, testing or staging?
[10:00:59] -!- ejjdhfjsu has joined #archlinux32
[10:06:17] -!- deep42thought1 has joined #archlinux32
[10:06:23] <deep42thought1> good morning!
[10:06:41] deep42thought is now known as girls
[10:06:42] <abaumann> morning :-)
[10:06:47] deep42thought1 is now known as deep42thought
[10:07:03] <deep42thought> let me check, what's wrong with rechenknecht and just reenable it
[10:07:19] <deep42thought> btw: I think, it does not hurt to keep it enabled, even though, if it mis-signs packages
[10:07:30] <deep42thought> the buildmaster will just reject them and I get notified for it on irc
[10:07:39] <abaumann> it just builds things and the buildmaster sends funny iirc messages, I reckon.
[10:07:42] <deep42thought> in contrast, I do *not* get notified, if it's disabled :p
[10:07:48] <abaumann> ah :-)
[10:08:10] <abaumann> reenabled :-)
[10:08:54] <deep42thought> thx
[10:09:29] <deep42thought> additionally, *broken* packages still work the way, they should: we get the error logs of the build :D
[10:09:40] <bill-auger> only the standard repos, except for those two packages from the archive
[10:10:46] <bill-auger> archarm builds without the wasi compiler - that is probably another option
[10:10:55] <abaumann> mmh, I got mixed results using an old gcc10 from the archive, so I tries to go ahead.
[10:11:14] <bill-auger> gcc10 has not worked since v102
[10:11:14] <abaumann> wasi is not causing the linking issues, I hope, it's unrelated..
[10:11:25] <abaumann> ah. good to know.
[10:11:51] <abaumann> https://bugs.archlinux32.org
[10:11:51] <phrik> Title: FS#281 : Firefox compilation (at bugs.archlinux32.org)
[10:12:08] <abaumann> my last build went well past all problems only to hit a different linking error in the end.
[10:12:17] <abaumann> I'm positive, that this is one of the last ones. :-)
[10:12:26] <abaumann> (TM: famous last words)
[10:13:53] <bill-auger> what i remember is that is would not build without both llvm and llvm13
[10:14:10] <abaumann> llvm13 most likely because of an older rust version (1.61)
[10:14:21] <abaumann> It looks like it builds fine now with ruste 1,63
[10:14:34] <bill-auger> yes i probably have all of the error messages - it took several iterations to get it right
[10:14:53] <abaumann> *sigh*
[10:15:39] <abaumann> yes, we are supposed to save energy this winter, but building open source software needs more and more electrons somehow..
[10:19:57] <bill-auger> this is the .BUILDINFO of the succesful package https://termbin.com
[10:20:43] <abaumann> aha, so all against gcc 12.
[10:20:52] <abaumann> which is good, because gcc11 never built on Arch32
[10:21:43] <bill-auger> that was build in a clean chroot; so that is the minimal package set
[10:24:21] <bill-auger> mozilla is a rube goldberg contraption - IIRC, it needs 9 different compilers/interpreters
[10:26:20] <abaumann> lol. yes. :-)
[10:26:20] -!- epony has quit [Read error: Connection reset by peer]
[10:26:34] <abaumann> what firefox is in small, linux is in big. :-)
[10:29:01] <abaumann> installed = rust-1:1.61.0-1.0-i686
[10:29:13] <abaumann> ah, yes. so then this caused the need for llvm13
[10:29:24] <abaumann> installed = gcc-12.1.0-2.1-i686
[10:29:24] <abaumann> installed = gcc-libs-12.1.0-2.1-i686
[10:29:31] <abaumann> good. I'll try again.. :-)
[10:30:48] <bill-auger> notice that llvm 14 also needed to be installed - so technically it was 10 compilers this time
[10:31:15] -!- ejjdhfjsu has quit [Remote host closed the connection]
[10:31:19] <abaumann> https://releases.llvm.org
[10:31:28] <abaumann> oh, good. another one is coming..
[10:31:39] -!- ejjdhfjsu has joined #archlinux32
[10:31:51] <abaumann> at least they are honest about their version numbering and they increase the major version..
[10:31:58] <deep42thought> abaumann: you won't need any gas for heating this winter - because you can heat with a lot of failing builds
[10:32:10] <abaumann> things tend to be highly incompatible between major versions in llvm..
[10:32:37] <abaumann> deep42thought: true :-)
[10:33:02] <deep42thought> but with the current gas prices, it's almost cheaper heating this way ;)
[10:35:43] <abaumann> I optimized builds around high and low energy tarifs for now..
[10:35:55] <abaumann> ..so builds are off from 18:00 to 22:00..
[10:36:05] <abaumann> shutdownasap came in really handy here :-)
[10:36:16] <deep42thought> :D
[10:36:39] <abaumann> all the rest of the time its low tarif, so heating it is.. ;-)
[10:40:57] <deep42thought> \o/ woohoo \o/ rechenknecht is back on track :)
[10:41:17] -!- epony has joined #archlinux32
[10:42:57] <deep42thought> cu, have a nice weekend!
[10:45:05] <abaumann> cool. thanks :-)
[10:48:32] <abaumann> bill-auger, I'll have a look at freecad, but it has quite some dependencies for Intel special or 64-bit only libraries..
[10:50:18] -!- deep42thought has quit [Quit: Leaving.]
[10:55:37] <bill-auger> i got freecad to build for armv7h
[10:56:03] <abaumann> oh. cool.
[10:57:57] <bill-auger> i only needed to remove ospray from makedepends
[10:58:34] <abaumann> ah, ok. openimagedenoise and rkcommon have trouble anyway down the dependency line.
[11:00:59] -!- drathir_tor has quit [Ping timeout: 258 seconds]
[11:24:17] <abaumann> I'll also have to remove med-openmpi for IA32, then I get med.h not found, please install development header-files for libmedc.
[11:24:30] <abaumann> I really hope, there still is the non-MPI version of libmed around..
[11:25:33] <abaumann> ah 'med'
[11:34:36] <abaumann> ninja: error: '/usr/lib/libdl.so', needed by 'lib/libSMDS.so', missing and no known rule to make it
[11:34:39] <abaumann> mmh.
[11:38:20] <abaumann> mmh. glibc has done away with -ldl being shared for a long time. There is a /lib/libdl.a stuf, so they are adding libdl.so explicitely?
[11:38:23] <abaumann> nice
[11:39:32] <abaumann> I can put 'sudo ln -s /usr/lib/libdl.so.2 /usr/lib/libdl.so' somewhere in PKGBUILD, but this feels rather dirty.. :-)
[11:42:53] <bill-auger> it is - ive done something similar though manually in the chroot, and made a note of my shame in the PKGBUILD
[11:43:02] <abaumann> lol :-)
[12:19:57] -!- russell-- has joined #archlinux32
[12:22:27] <russell--> hi, i just picked up a OLPC box (/proc/cpuinfo says celeron m 900MHz) which i've had archlinux on since 2016-ish, and now i'm having trouble updating, not least because packages use zst instead of bz2 and zstd isn't installed
[12:31:22] <bill-auger> pacman and libarchive IIRC, need to be upgraded together, before upgrading the system
[12:32:03] <bill-auger> that was quite a long time ago though - it is not a good idea to wait so long between upgrades
[12:38:18] -!- bill-auger has quit [Ping timeout: 265 seconds]
[12:39:22] -!- bill-auger has joined #archlinux32
[12:51:55] <russell--> there was some helpful instruction here: https://www.archlinux32.org
[12:51:56] <phrik> Title: Arch Linux 32 (at www.archlinux32.org)
[12:52:43] -!- drathir_tor has joined #archlinux32
[13:24:16] <russell--> are there special instructions for install archlinux32 (i couldn't find them on the website)?
[13:26:40] <abaumann> https://bbs.archlinux32.org
[13:26:41] <phrik> Title: Archlinux32 specific installation notes / Installation / Arch Linux 32 Forum (at bbs.archlinux32.org)
[13:26:53] <abaumann> but usually it's exactly the same as upstream.. with some caveats..
[13:27:11] <abaumann> I didn't teset archinstall (the python script) lately, that one might not work..
[13:36:45] <abaumann> I have a patched version of freecad now..
[13:37:05] <abaumann> I'm patching build.ninja after cmake before ninja build
[13:37:33] <abaumann> I cannot add a libdl.so link in a PKGBUILD, the builduser has no NOPASS sudo right in the build chroot (which is probably a good idea)
[13:52:03] <russell--> pacstrap is giving me a segfault at about the time of importing pgp keys after downloading 120 packages
[13:52:51] <russell--> downloading required keys ...
[13:53:09] <abaumann> Yeah, either refresh those keys, or install pacman without checking (SigLevel=Never).
[13:53:34] <abaumann> there was more than bug and expired key since 2017 :-)
[13:54:23] <abaumann> by refresh I meant: wipe the keys, pacman-key --init, pacman-key --populate and then try a refresh again..
[13:54:31] <russell--> no, this is from a fresh install from the 2022-09-01
[13:55:24] <russell--> step 2.2: pacstrap /mnt base linux linux-firmware
[13:56:07] <abaumann> oups
[13:56:34] <abaumann> then I have to see if I can reproduce what creates this segfault..
[13:57:37] <russell--> it doesn't seem to be memory pressure, i have 1GB and 2GB of swap and top isn't showing any substantial swap usage
[13:57:57] <abaumann> segfault is either a problem in pacman, pgp or a problem with the keys.
[13:58:14] <abaumann> you are on the ISO, I suppose?
[13:58:19] <russell--> yeah
[13:58:25] <abaumann> try another iso.
[13:58:38] <abaumann> most likely the current iso has an inssue with some of the software..
[13:58:55] <russell--> archlinux32-2022.09.01-i686.iso
[13:59:10] <abaumann> ok, lemme try that one..
[14:10:50] <abaumann> mmh. cannot reproduce this in libvirt vms..
[14:11:13] <russell--> i'm downloading the 06.01 iso
[14:11:36] <abaumann> ok. let's see..
[14:14:44] <russell--> 06.02 rather
[14:14:52] <russell--> still downloading ...
[14:19:44] <abaumann> Celeron has only SSE, no SSE2, (check /proc/cpuinfo), make sure, pacstrap installs a i686 subarchitecture..
[14:19:59] <abaumann> ..though this would result in SIGILL usually rather than a SIGSEGV
[14:22:57] <russell--> booting...
[14:24:23] <abaumann> segfaults also have a probable cause in bad memory dimms sometimes..
[14:25:07] <russell--> it has always run fine in the past
[14:25:26] <abaumann> ok. so let's assume it's the ISO..
[14:25:35] <bill-auger> that is defintely a good idea - the best reason to avoid mutating the build system in the PKGBUILD, is because "the build system" may actually be the working system of a user, who takes the PKGBUILD and runs `makepkg` on it
[14:26:02] <abaumann> exactly.
[14:26:33] <bill-auger> its probably why you can not run makepkg as root; so your build system is simulating that
[14:27:11] <abaumann> at least make install does it with fakeroot, I suppose..
[14:27:28] <bill-auger> we build everything in a chroot with fakeroot; so there no such safety mechanism preventing naughtiness
[14:28:18] <bill-auger> but i just keep in mind that the distro packager is not the only user of the PKGBUILD
[14:36:08] <russell--> 06.02 gives me: error: keyring is not writable
[14:43:41] <russell--> arch-chroot'ing and doing the pacman-key --init and --populate helped
[14:45:07] <abaumann> ah, ok. interesting.
[14:45:09] <russell--> well, sort of. i still get a bunch of invalid or corrupt package messages
[14:45:20] <abaumann> SigLevel=Never then
[14:45:36] <abaumann> The PGP infrastrucure is really brittle..
[14:46:06] <russell--> i'm downloading the august iso
[14:46:48] <abaumann> We should really do some more testing of the ISO like autoamated installs.. :-(
[14:51:10] <abaumann> lemme check the 09.02 iso on a real machine.. :-)
[14:55:26] <russell--> how do you use SigLevel=Never?
[14:58:05] <abaumann> /etc/pacman.conf
[14:58:33] <abaumann> on the iso, or you have to copy /etc/pacman.conf to /tmp/pacman.conf, then edit that one and use it with -c /tmp/pacman.conf in pacstrap
[14:59:17] <russell--> trying the august iso
[15:00:52] -!- bill-auger has quit [Ping timeout: 265 seconds]
[15:01:02] -!- bill-auger_ has joined #archlinux32
[15:01:47] -!- ejjdhfjsu has quit [Remote host closed the connection]
[15:02:12] -!- ejjdhfjsu has joined #archlinux32
[15:06:10] <russell--> /proc/cpuinfo says i have sse and sse2
[15:06:56] <abaumann> ah, good. :-)
[15:06:59] <abaumann> pentium4 then
[15:07:18] -!- ejjdhfjsu has quit [Ping timeout: 268 seconds]
[15:09:23] <russell--> the august iso fails with keyring is not writable again
[15:09:52] <russell--> i don't see anything in the install instructions about running pacman-key before pacstrap
[15:14:22] <russell--> it's -C not -c, fwiw
[15:22:42] <abaumann> oi, I get some really weird errors on the ISO 09.01: missing shared libraries of systemd subprocesses and it doesn't problem the cdrom drive correctly (thus the ISO is not even mounted).
[15:23:11] <abaumann> My problem slowly becomes: I don't have running old hardware anymore - I already lost 2 machines (dea power supply, unsupported RAID controller)..
[15:29:52] <abaumann> mmh. I think this is my DVD-RW burner, something doesn't work here..
[15:40:30] <abaumann> freecad: error while loading shared libraries: libpyside2.cpython-310-i386-linux-gnu.so.5.15: cannot open shared object file: No such file or directory
[15:40:34] <abaumann> interesting..
[15:41:23] <abaumann> this sounds like pyside2 has to be rebuild against puthon 3.10..
[15:44:42] <abaumann> libGL error: MESA-LOADER: failed to open swrast: libLLVM-13.so: cannot open shared object file: No such file or directory (search paths /usr/lib/dri, suffix _dri)
[15:45:05] <abaumann> but it starts..
[15:45:06] <abaumann> :-)
[15:51:38] <abaumann> russell--: usually the keys should all be on the ISO, but sometimes they expire exactly one day _after_ releasing the ISO, then you have to manually update them.. :-)
[16:04:30] <russell--> what's the command to import them manually?
[16:05:27] <abaumann> pacman-key --init, pacman-key --populate archlinux, pacman-key --populate archlinux32, pacman-key --refresh IIRC
[16:05:57] <abaumann> but you have to have /etc/pacman.d/gnupg writeable, so you have to boot and load the ISO into a ramdisk..
[16:06:07] <abaumann> normal boot might not work
[16:06:33] <abaumann> (there are boot options on the ISO for "boot from ISO" and "boot from ISO and load into RAM" or so)
[16:07:30] <russell--> i managed to install with SigLevel=Never and booted into the normal system
[16:07:37] <abaumann> ah. good. :-)
[16:07:48] <abaumann> I agree, this is not the recommended way..
[16:07:54] <abaumann> ..but, yeah..
[16:08:24] <abaumann> turns out, one DVD-RW burner burns only half of a cdrom and then puts garbage onto the disk.. *grmpf*
[16:10:24] <russell--> the --refresh is throwing some errors
[16:10:37] <abaumann> about servers and keys not found, I suppose.
[16:10:44] <abaumann> Sad to say: this is normal
[16:11:07] <russell--> of the form "error retrieving '<foo>' via WKD: No data'
[16:11:12] -!- bill-auger_ has quit [Ping timeout: 264 seconds]
[16:11:44] <abaumann> web key directories, yeah. I usually take the one from Ubuntu :-)
[16:12:13] <abaumann> keyserver keyserver.ubuntu.com
[16:12:25] <abaumann> gpg.conf
[16:15:19] <russell--> where does gpg.conf live?
[16:15:41] <abaumann> /etc/pacman.d/gnupg, usually it is not there, you have to create one..
[16:17:15] <russell--> it is there
[16:20:03] <abaumann> ok, I can reproduce now the segmentation fault in pacstrap, just after showing the two archlinux32 signing keys.
[16:20:16] <abaumann> It only happens on real hardware, not in qemu (which makes me extremely happy)
[16:20:25] <abaumann> thanks for finding this one :-)
[16:23:26] <abaumann> pacstrap is a shell script, so it's quite easy to find out, where a called command segfaults (I guess, it's gpg)
[16:31:57] <KitsuWhooa> a bit late, but <abaumann> It only happens on real hardware, not in qemu <-- that's how I ended up installing arch32
[16:32:10] <KitsuWhooa> I ran the installer in a vm
[16:36:52] <abaumann> ah ok
[16:56:20] -!- drathir_tor has quit [Ping timeout: 258 seconds]
[16:58:22] -!- drathir_tor has joined #archlinux32
[18:11:33] -!- abaumann has quit [Quit: leaving]
[18:26:03] -!- abaumann has joined #archlinux32
[18:26:04] <buildmaster> Hi abaumann!
[18:26:04] <buildmaster> !rq abaumann
[18:26:05] <phrik> buildmaster: <abaumann> mmh.. my build slaves are on strike? Maybe I should pay them? ;-)
[18:26:42] <abaumann> So, pacstrap fetches the keys for arch32 from an unknown server or so, so we get a segfault.
[18:27:30] <abaumann> keyserver keyserver.ubuntu.com
[18:27:45] <abaumann> in /etc/pacman.d/gnupg/gpg.conf on the ISO
[18:27:53] <abaumann> gets copied to the pacstrap chroot
[18:27:57] <abaumann> and then it works..
[18:28:58] <abaumann> we should add a keyserver onto the ISO, just I don't think ubuntu is the right one :-)
[18:29:03] <abaumann> we could take our own?
[18:33:58] <abaumann> mmh. not enough, ok, we have to update the keys on the iso too?
[18:44:33] <girls> we don't run our own keyserver ;)
[18:44:42] <girls> just take ubuntu's keyserver - I think, it's fine
[18:48:02] <abaumann> ah, good :-)
[19:05:29] <girls> actually, we should just put *recent* keys onto the iso
[19:05:57] <girls> ah, I think, the problem is, that *my* build infrastructure has some quite-outdated archlinux32-keyring package - and that's, what lands on the iso ...
[19:06:04] <girls> I'll try to look into this (harder)
[19:20:39] <abaumann> I'm also not completely sure, this is the problem. gnupg segfaults seem to be a little bit random..
[19:21:41] <girls> this is the cause, though the segfault is an additional problem ;)
[19:22:48] <girls> wtf, some script runs amok and rebuilds perl-wx over and over again on my machine - I'm already at pkgrel=2626 O.o
[19:34:44] <abaumann> oh. that's a well build package ;-)
[19:36:37] <abaumann> I hat this: gpg-agent. background processes spawned with the old configuration..
[19:36:49] <abaumann> *hate
[20:46:53] -!- abaumann has quit [Quit: leaving]
[22:45:53] -!- GNUtoo has quit [Remote host closed the connection]
[22:56:07] -!- GNUtoo has joined #archlinux32