2025-11-18 18:05:51 <mps> riscv64 dev machine segfaults often
2025-11-18 18:39:38 <mps> Tue Nov 18 18:05:06 2025] make[34445]: unhandled signal 11 code 0x1 at 0x0000000000000000 in make[2adc712000+33000]
2025-11-18 21:14:22 <ncopa> One of the pioneers is very unreliable. It deadlocks or segfaults
2025-11-18 21:16:48 <ncopa> I’m hoping milk-v titan will be usable 
2025-11-20 12:43:56 <mps> gitlab.a.o is down?
2025-11-20 12:44:32 <mps> ah, it is online again
2025-11-20 12:53:24 <mps> this milk-v titan looks fine and cost is not too high
2025-11-21 11:45:21 <mps_> ncopa: btw, gitlab CI run on bpi-f3?
2025-11-21 12:06:18 <ncopa> one of the runners yes
2025-11-21 12:06:46 <ncopa> reminds me, we could set up a couple of CI runners in scaleway
2025-11-21 12:06:54 <mps> well, i meant riscv64
2025-11-21 12:09:15 <mps> I should check how fast riscv64 in qemu on loongarch64 host
2025-11-21 12:09:49 <ncopa> one of the riscv64 CI runners is a bpi-f3. We have multiple riscv64 CI runners
2025-11-21 12:10:29 <mps> aha
2025-11-21 12:11:28 <ncopa> looks liek we have 4 of them for aports
2025-11-21 12:12:14 <mps> which 4?
2025-11-21 12:12:39 <mps> 4 different machines?
2025-11-21 12:16:47 <mps> my bpi-f3 have only 2GB ram, very slow for build anything not small
2025-11-21 12:17:20 <ikke> 2 bpi-f3, 2 milk-v pioneer 
2025-11-21 12:23:29 <ncopa> I have a hifive premier p550 under my desk that I should set up for CI as well, but it is a bit noisy
2025-11-21 12:24:05 <ncopa> i have a starpro64 too that is "under construction", which is passively cooled and has 32G ram
2025-11-21 12:24:15 <ncopa> I just havent had time to set it up
2025-11-21 12:25:02 <ikke> Understood, getting 3.23 out has a higher priority atm
2025-11-21 12:25:09 <ncopa> yeah
2025-11-21 12:26:06 <mps> oh, when is release planned?
2025-11-21 12:26:54 <mps> and what version of linux-lts?
2025-11-21 12:26:55 <ikke> asap
2025-11-21 12:27:01 <ikke> No change 
2025-11-21 12:27:31 <ncopa> atm the plan is linux-lts 6.12
2025-11-21 12:27:51 <ncopa> but if 6.18 comes out today, we may be able to do 6.18, but its risky
2025-11-21 12:27:56 <ncopa> i'd live to have 6.18
2025-11-21 12:28:03 <ncopa> s/live/love/
2025-11-21 12:28:33 <mps> what a pity, 6.18 will be released soon and it will be next lts i think
2025-11-21 12:28:47 <ncopa> even if it comes out *now* i still need ~1 day to work on it
2025-11-21 12:28:56 <ncopa> yeah
2025-11-21 12:29:58 <mps> maybe you backport it later
2025-11-21 12:32:53 <ncopa> no
2025-11-21 12:33:32 <mps> :(
2025-11-21 12:45:05 <ikke> Are there RCs already we could test?
2025-11-21 12:48:02 <mps> yes, https://git.kernel.org/torvalds/t/linux-6.18-rc6.tar.gz
2025-11-21 12:48:38 <mps> i'm testing it on loongarch64 
2025-11-21 12:49:09 <mps> can test it on aarch64
2025-11-21 12:56:51 <mps> actually I run 6.17 on my machines - armv7, aarch64, riscv64 and loongarch64
2025-11-21 12:57:03 <mps> all works fine
2025-11-21 13:01:10 <ncopa> what needs testing is build the 3rd party modules
2025-11-21 13:02:46 <mps> oh, not my expertise
2025-11-21 13:04:52 <achill> ill look at zfs
2025-11-21 13:04:59 <achill> but usually it works with some backporting
2025-11-21 13:06:11 <achill> yeah there are patches available to build it with linux 6.18: https://github.com/openzfs/zfs/pull/17842#issuecomment-3424044265
2025-11-21 13:07:36 <achill> ncopa: if you need some help with 6.18 im here :3
2025-11-21 13:12:14 <ncopa> 1. compare the configs with the linux-stable. I suppose they are more or less aligned
2025-11-21 13:12:35 <ncopa> just make sure we got all recent kernel config changes in linux-lts with us
2025-11-21 13:12:52 <ncopa> 2. make sure all of the 3rd party modules build
2025-11-21 13:13:04 <ncopa> 3. test boot
2025-11-21 13:14:03 <ncopa> 0.5: check if we can drop any of the patches
2025-11-21 13:14:35 <ncopa> achill: would be nice with help with 6.18
2025-11-21 13:16:45 <mps> forgot tell, also latest stable is fine on loongarch64
2025-11-21 14:18:55 <achill> nice ill try that this evening or so
2025-11-23 08:32:48 <ikke> Does alpine target riscv64 with rvv extensions (can we assume them to be present)?
2025-11-23 09:02:36 <mps> isn't it what what gcc/llvm have set
2025-11-23 09:02:54 <mps> s/what//
2025-11-23 09:08:38 <ikke> https://gitlab.alpinelinux.org/alpine/aports/-/jobs/2109738#L509
2025-11-23 09:08:55 <ikke> undefined reference to `png_init_filter_functions_rvv'
2025-11-23 09:25:57 <ikke> It says we need to set -march=rv64gv1p0
2025-11-23 09:51:56 <ikke> So the question is, can we build with that set, orn ot?
2025-11-23 10:48:43 <mps> what is p0
2025-11-23 13:12:38 <iggy> it's gcc slang for vector extension 1.0
2025-11-23 13:21:56 <iggy> I don't think it's safe to assume we support that, but I'm not 100%
2025-11-23 13:22:59 <iggy> I think there are some things that only support a draft of rvv 1.0
2025-11-23 13:26:08 <ikke> I'm a bit confused whether it supports runtime detection or not
2025-11-23 13:26:11 <ikke> (libpng)
2025-11-23 13:26:55 <ikke> https://github.com/pnggroup/libpng/blob/libpng16/INSTALL#L282
2025-11-23 13:27:47 <ikke> ie, we don't provide --enable-hardware-optimizations, but the question is whether that's necessary or not
2025-11-23 14:43:55 <mps> iggy: isn't v1 for vector ext version 1
2025-11-23 14:45:25 <mps> runtime detection works in kernel but not sure it work in userspace
2025-11-23 15:25:11 <sorear_> it's V (the extension letter) 1p0 (1 "point" 0)
2025-11-23 15:29:11 <sorear_> runtime detection in userspace exists in general ("hwprobe", but you can also use getauxval for V) but libpng doesn't seem to support it
2025-11-23 15:39:09 <mps> aha, thanks
2025-11-23 15:44:35 <sorear_> if you include V in -march gcc will generate code using V for loops, struct assignments, etc with no runtime checks, so only use that if V is part of the documented hardware requirements
2025-11-26 09:30:37 <huajingyun> perl-json-webtoken aport test failed on riscv64 builder, so I quickly checked and found the cause was an upgrade of perl-crypt-openssl-rsa from version 0.35 to 0.37
2025-11-26 09:30:48 <huajingyun> So yes, now if you rebuild perl-json-webtoken based on perl-crypt-openssl-rsa 0.37 on any other architecture, you'll get the same error. I think it's some kind of compatibility issue
2025-11-26 09:31:09 <huajingyun> Just mentioning this because I'm about to go AFK, maybe someone will be interested in continuing to check it out
2025-11-26 11:59:18 <ncopa> thank you for sharing!
2025-11-26 14:16:39 <mio> as indicated in #17668, someone is looking into perl-json-webtoken issue, but it will take a few days
2025-11-26 14:20:51 <mio> as it probably involves rewriting a portion of the signing code that the test uses
2025-11-26 14:28:56 <mio> maintainer of perl-net-saml2 had encountered a related issue and has worked with perl-crypt-openssl-rsa code upstream, they kindly agreed to look into
2025-11-26 14:30:02 <mio> this one. unfortunately there hasn't been much activity over at perl-json-webtoken
2025-11-26 14:31:34 <mio> if it is a blocker for 3.23, maybe better to temporarily disable it (and its dependent, on the arch)
2025-11-26 14:31:58 <mio> s/, on the arch//
2025-11-26 14:38:15 <ncopa> do you have a ref/link for perl-net-saml2/perl-crypt-openssl-rsa work?
2025-11-26 14:38:40 <ncopa> this apparenly affects all arches, and affects only imapsync
2025-11-26 14:38:51 <ncopa> we could delete both for now
2025-11-26 14:39:40 <mio> yes, it affects all arches. perl-net-saml2 and perl-xml-sig was recently upgraded to be compatible with perl-crypt-openssl-rsa 0.37
2025-11-26 14:39:55 <ncopa> we could also revert 2a0dfbf63aa91772951818d027105697ac63db07
2025-11-26 14:40:09 <ncopa> but would be better to fix it
2025-11-26 14:42:19 <ncopa> more things are broken https://github.com/cpan-authors/Crypt-OpenSSL-RSA/issues/61#issuecomment-3581300444
2025-11-26 14:42:55 <mio> https://metacpan.org/dist/Net-SAML2 https://github.com/perl-net-saml2/perl-Net-SAML2/issues/239
2025-11-26 14:43:18 <mio> the changelog on net-saml2 summarises the issue
2025-11-26 14:43:20 <ncopa> thank you!
2025-11-26 14:45:24 <mio> you're welcome
2025-11-26 14:46:06 <mio> the other two blocks on 3.23 riscv64 are njs (failed test due to pcre 10.47 upgrade) and ostree-push
2025-11-26 14:47:11 <mio> the njs one may be safe to skip, guess it would be more a bug at worst and not critical
2025-11-26 14:47:38 <mio> opened an issue for maintainer feedback
2025-11-26 14:50:06 <mio> possibly related to https://github.com/PCRE2Project/pcre2/pull/817 "Add support for $+ to pcre2_substitute", not sure
2025-11-26 14:52:52 <mio> spent some time yesterday looking into ostree-push test but didn't find out why it fails only on riscv64. the function is called multiple times in other tests and does not fail then, only in one test
2025-11-26 14:57:20 <ncopa> i disabled imacysn and perl-json-webtoken. this could be a follow up https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/93786
2025-11-26 14:58:17 <mio> cool, thanks