#archlinux32 | Logs for 2022-12-09

[00:14:24] -!- drathir_tor has joined #archlinux32
[03:28:55] -!- lithium_pt has joined #archlinux32
[03:31:29] -!- T`aZ_ has joined #archlinux32
[03:32:00] -!- finstern1s has joined #archlinux32
[03:34:49] -!- lithiumpt has quit [Ping timeout: 256 seconds]
[03:34:49] -!- KillerWasp has quit [Ping timeout: 256 seconds]
[03:34:50] -!- T`aZ has quit [Ping timeout: 256 seconds]
[03:34:50] -!- finsternis has quit [Ping timeout: 256 seconds]
[03:34:54] finstern1s is now known as finsternis
[04:42:35] -!- bill-auger has joined #archlinux32
[05:35:55] -!- finsternis has quit [Remote host closed the connection]
[07:02:44] -!- T`aZ_ has quit [Ping timeout: 248 seconds]
[07:04:53] -!- T`aZ has joined #archlinux32
[07:18:21] -!- abaumann has joined #archlinux32
[07:18:22] <buildmaster> Hi abaumann!
[07:18:22] <buildmaster> !rq abaumann
[07:18:22] <phrik> buildmaster: <abaumann> we break things - boldly - then we unbreak them again :-)
[07:18:59] <abaumann> KitsuWhooa: ffmpeg, ok SIGILL indicates that SIMD CPU operations are laeaking into the build.. Let me see if I can find out where they come from..
[07:19:43] <abaumann> protobuf fails to build with gcc 12, no matter what versions, I tried to build it with clang with even worse effects.. that's what I feared, it indicates problems in the toolchain..
[07:28:45] <abaumann> yep, gdb hit stable, libguile got stuck, I'll push that..
[07:42:50] <abaumann> so, gdb is ok.
[07:49:20] <abaumann> so, you are on 686: there I see => 0xb5f315e7: f3 0f 10 83 d0 0b 00 00 movss 0xbd0(%ebx),%xmm0
[07:49:35] <abaumann> on the SIGILL. on pentium4 it works fine.
[07:50:46] <abaumann> </code>
[07:50:49] <abaumann> https://bugs.archlinux32.org
[07:50:51] <phrik> Title: FS#311 : [ffmpeg] fails with SIGILL on i686 (at bugs.archlinux32.org)
[07:51:20] -!- drathir_tor has quit [Remote host closed the connection]
[07:52:58] <abaumann> I cannot make a shim package for protobuf without maybe an older gcc compiler, those got deleted upstream (and also in our cleanup action :-) ), so I'll have to build a gcc-10 or/and gcc-11 first in build-support..
[08:01:51] -!- drathir_tor has joined #archlinux32
[08:06:06] <abaumann> https://archlinux.org
[08:06:07] <phrik> Title: Arch Linux - gcc11 11.3.0-5 (x86_64) (at archlinux.org)
[08:06:14] <abaumann> mmh, strange, we should have a gcc 11
[08:07:09] <abaumann> I'm not optimistic you can build gcc 11 with gcc 12 though..
[08:10:37] <abaumann> 9428209822a8f0ef0f20b6117c9144635c849821: blacklisted gcc11 (needed only for cuda, which is also blacklisted)
[08:10:45] <abaumann> aha, I have to blame myself (as usual) :-)
[08:11:11] <abaumann> If it would have built at the time I would not have blacklisted it probably..
[09:04:18] <KitsuWhooa> abaumann: yeah, I've helped with sigill before so I am aware. It's just that gdb not working stopped me dead on my tracks and then I distracted
[09:04:49] <KitsuWhooa> Also technically it's usually sse2 that leaks into i686 :p
[09:58:43] <abaumann> That's why I fixed gdb first. ;-)
[09:59:32] <abaumann> The minimal ISA is with sse2, this has the consequence that many packages do not really bother anymore to do the checks for MMX, SSE, SSE2 individually.
[09:59:46] <abaumann> This is a logical consequence and will create more and more problems in the future.
[10:00:28] <abaumann> Patches for SSE and so are usually extremely brittle, not really of any interest to the public (and the authors) anymore. This creates a bigger and bigger patch set to maintain locally.
[10:00:59] <KitsuWhooa> I agree
[10:01:27] <KitsuWhooa> When you say a protobuf shim, what do you mean?
[10:01:35] <KitsuWhooa> A newer version implementing the api differences while talking to an old one?
[10:02:10] <abaumann> I thought that there were differences in protobuf failing it to build. In this case I would have made a protobuf211 package with the last known good version.
[10:02:35] <abaumann> This would only have been a short term solution anyway (as with icu69 and such shim packages).
[10:02:36] <KitsuWhooa> If I have time this weekend I'll look into trying to get it to build properly. How do you build packages? I assume it's on arch64
[10:02:49] <abaumann> Yes, mainly.
[10:02:50] <KitsuWhooa> How as in, 32 bit chroot?
[10:03:00] <KitsuWhooa> hm, okay
[10:03:12] <KitsuWhooa> Any instructions on setting one up?
[10:03:19] <abaumann> Just to finish the thought above: turns out, gcc 12 has problems to build any protobuf version, so I'll try now with gcc-11
[10:03:27] <abaumann> aeh.
[10:03:40] <abaumann> we are mainly building in arch64 devtools chroots, yes.
[10:03:41] <KitsuWhooa> I've been building 32 bit kernels on amd64 without a chroot, and while the kernel ends up being 32 bit, the tools are all 64 bit and obviously don't run on a P3 :p
[10:03:49] <abaumann> But for i486, which are building on 486 VMs.
[10:04:10] <abaumann> The kernel is one of the best packages, if it comes to cross-compilation.
[10:04:10] <KitsuWhooa> I only have an i686 installation so I'm focusing on that
[10:04:16] <abaumann> So, no wonder. :-)
[10:04:23] <KitsuWhooa> although you're right, I could probably run an i686 VM
[10:04:26] <KitsuWhooa> it'd be easier
[10:05:05] <abaumann> Especially a 64-bit chroot can build packages, which probe for /proc/cpuinfo flags or similar stuff, so that's how we end up with ffmpeg not running on i686 for instance.
[10:05:18] <abaumann> Manually building a package is easy with 'asp32'
[10:05:25] <abaumann> and with 'devtools32'
[10:05:34] <abaumann> both are on git.archlinux32.org
[10:05:40] <KitsuWhooa> Awesome, I'll have a look
[10:05:48] <abaumann> they are forks of the upstream versions and they should work exactly the same way.
[10:06:26] <abaumann> So, install the packages and a 'asp32 update', 'asp32 export core/which', 'staging-i686-build' sequence should build a package for the 'i686' subarchitecture.
[10:06:42] <abaumann> the chroot is in /var/lib/archbuild, so you can chroot into that one and do some debugging.
[10:06:46] <KitsuWhooa> cool!
[10:10:55] <abaumann> # of expected passes 25518
[10:10:55] <abaumann> # of unexpected failures 226
[10:10:55] <abaumann> # of expected failures 179
[10:10:55] <abaumann> # of unresolved testcases 1
[10:10:55] <abaumann> # of unsupported tests 602
[10:10:58] <abaumann> /build/gcc11/src/gcc-build/gcc/xgcc version 11.3.0 (GCC)
[10:11:08] <abaumann> mmpf.. How much do I trust that toochain, I wonder?
[10:26:18] -!- epony has quit [Quit: QUIT]
[10:37:35] <KitsuWhooa> It'll be fineā„¢
[11:06:02] -!- GNUtoo has quit [Ping timeout: 255 seconds]
[11:06:29] -!- drathir_tor has quit [Ping timeout: 255 seconds]
[11:12:47] -!- GNUtoo has joined #archlinux32
[11:13:30] -!- drathir_tor has joined #archlinux32
[11:25:34] -!- drathir87 has joined #archlinux32
[11:27:11] -!- drathir_tor has quit [Ping timeout: 255 seconds]
[11:27:12] drathir87 is now known as drathir_tor
[12:37:03] -!- epony has joined #archlinux32
[13:10:17] -!- epony has quit [Ping timeout: 268 seconds]
[13:15:44] -!- epony has joined #archlinux32
[15:39:29] -!- T`aZ has quit [Ping timeout: 260 seconds]
[15:41:12] -!- T`aZ has joined #archlinux32
[16:05:56] -!- abaumann has quit [Quit: leaving]
[17:05:48] -!- finsternis has joined #archlinux32
[19:21:49] -!- bill-auger has quit [Ping timeout: 256 seconds]
[19:27:34] <girls> abaumann: I'm pondering to create arch32 docker images for build slaves - do you think, we could actually make use of something like that?
[22:00:20] -!- GNUtoo has quit [Ping timeout: 255 seconds]
[22:22:21] -!- GNUtoo has joined #archlinux32