dgilmore changed the topic of #fedora-riscv to: Fedora on RISC-V https://fedoraproject.org/wiki/Architectures/RISC-V || Logs: https://libera.irclog.whitequark.org/fedora-riscv || Alt Arch discussions are welcome in #fedora-alt-arches
ahs3|afk has quit [Quit: Coyote finally caught me]
ahs3 has quit [Remote host closed the connection]
davidlt_ has joined #fedora-riscv
sorear has quit [Ping timeout: 252 seconds]
dgilmore has quit [Ping timeout: 265 seconds]
sorear has joined #fedora-riscv
dgilmore has joined #fedora-riscv
jcajka has joined #fedora-riscv
sharkcz has quit [Ping timeout: 264 seconds]
sharkcz has joined #fedora-riscv
<davidlt[m]> Kernel merge windows have one boring aspect. Stable queues are pretty much empty until merge window passes.
<davidlt[m]> And during rc1, rc2 large ones come in.
<rwmjones> davidlt[m]: remind me the command to kick off a koji build? koji -P riscv something?
<davidlt[m]> Depends on your configuration.
<davidlt[m]> Scratch build?
<rwmjones> a real build of dnf5
<davidlt[m]> But yeah, if you have a profile -p <profile_name>
<rwmjones> it builds fine for me
<rwmjones> plus there is a new version just today
<rwmjones> ok I don't think I have the profile set up
<davidlt[m]> ok
<rwmjones> yeah I'm mostly running F38 on nufive & nujive now
<rwmjones> but not the kernel
<davidlt[m]> Technically, I am not advising to use F38 now.
<davidlt[m]> There is somewhat a problem, which you can workaround by disabling SELinux and audit (or at least SELinux).
<rwmjones> it seems fine here, but selinux is disabled so ...
<davidlt[m]> Yeah, otherwise it wouldn't boot.
<davidlt[m]> Well, now it boots, but the actual issue isn't solved yet.
<davidlt[m]> Just a few workarounds to make it boot.
<davidlt[m]> Why... Why Fedora still don't have PC files fixed.
<davidlt[m]> ./pkgconfig/roc.pc:libdir=/usr/lib
<davidlt[m]> ./pkgconfig/imagequant.pc:libdir=${prefix}/li
<davidlt[m]> ./pkgconfig/imagequant.pc:libdir=${prefix}/lib
<rwmjones> so about doing a fedora.riscv.rocks build ... do I need an /etc/koji.conf.d/ configuration file?
<davidlt[m]> or ~/.koji
<davidlt[m]> I can send you instructions I sent to Nick
<rwmjones> I think I had this set up on the unleashed boards, but lost it when I moved to the unmatched
<rwmjones> ok thanks
<rwmjones> I'm pretty sure dnf5 ought to build fine, but I'll keep an eye on it once I can kick off the real build
<davidlt[m]> done
<rwmjones> hey davidlt[m], is koji_ca_cert.crt the one from /etc/kojid/koji_ca_cert.crt ?
<rwmjones> & also how to generate rjones.pem & rjones_browser_cert.p12
<davidlt[m]> Did you loose your certs?
<rwmjones> possibly, let me check the old server to see if it's there, one sec
<davidlt[m]> I most likely sent you those over the email.
<davidlt[m]> Check email from me on Tue, Jan 7, 2020, 12:13 PM
<davidlt[m]> Those still should be valid.
<davidlt[m]> Probably :)
<rwmjones> davidlt[m]: you couldn't just send me those again? seem to have lost the keys & having issues with email
<davidlt[m]> Done
<rwmjones> thanks
<rwmjones> hey, it works ...
<rwmjones> $ koji -p riscv hello
<rwmjones> bonjour, rjones!
<rwmjones> Authenticated via client certificate /home/rjones/.koji/rjones.pem
<rwmjones> You are using the hub at https://fedora.riscv.rocks/kojihub
<davidlt[m]> Just one thing to remember, your account has got powers most likely :) Ba careful :)
<rwmjones> davidlt[m]: when issuing a build do you use 'fedpkg build' (and some flag to select the profile which I cannot find)?
<rwmjones> or do you use koji -p riscv build directly?
<rwmjones> & i guess for koji we'd need to work out the target & scm url which is not trivial
<davidlt[m]> I use koji directly. I don't need -p because my main Koji profile is RISCV :)
<davidlt[m]> It's trivial.
<davidlt[m]> It's gonna be something like this:
<davidlt[m]> ```
<davidlt[m]> ```
<rwmjones> ok got it thanks
<davidlt[m]> I just have oneliner "build38" command in the shell.
<rwmjones> does it work out the scm url using git or something?
<davidlt[m]> If you are doing a scratch build, you are fine to send out SRPM.
<davidlt[m]> If you are building for F38 (not a scratch) build then let's do only SCM builds.
<rwmjones> of course
<davidlt[m]> SCM URL has a pattern.
<davidlt[m]> Just change "gnome-keyrig" to whatever is your package name.
<rwmjones> yup gotit
<rwmjones> just wondering if you had scripted it already
<davidlt[m]> # followed by commit bash.
<davidlt[m]> I have a bunch of stupid small things, like:
<davidlt[m]> ```
<davidlt[m]> ```
<davidlt[m]> koji build --nowait f38 "git+http://fedora.riscv.rocks:3000/rpms/$(basename $PWD).git#$(git rev-parse HEAD)"
<davidlt[m]> http://fedora.riscv.rocks:3000/ is our "dist-git overlay" where we can quickly patch/test things.
<rwmjones> ok
<rwmjones> if I fix this properly I mostly intend to push the fixes to fedora
<davidlt[m]> Yeah, that's the goal.
<davidlt[m]> I am off to eat something. Will be back later.
<davidlt[m]> rwmjones: do you need more tasks?
<rwmjones> dnf5 failed :-(
<rwmjones> so I'm still on that one
<rwmjones> it actually failed in a different and much more fun way through
<rwmjones> /builddir/build/BUILD/dnf5-5.0.9/redhat-linux-build/bindings/python3/libdnf5/CMakeFiles/python3_base.dir/basePYTHON_wrap.cxx: In function '_wrap_ResolveSpecSettings_group_with_name_set':
<rwmjones> /builddir/build/BUILD/dnf5-5.0.9/redhat-linux-build/bindings/python3/libdnf5/CMakeFiles/python3_base.dir/basePYTHON_wrap.cxx:21367:1: internal compiler error: Segmentation fault
<rwmjones> 21367 | } | ^
<rwmjones> Please submit a full bug report, with preprocessed source.
<davidlt[m]> Restart
<rwmjones> not reproducible locally obvs
<rwmjones> is that likely to be a temporary thing?
<davidlt[m]> This happens, especially in F38.
<rwmjones> ok
<davidlt[m]> I haven't figured it out, but ICE is sometimes generated if we run out of memory.
<davidlt[m]> I have confirmed that several times.
<davidlt[m]> But in F38 somehow there are more ICEs, but not more details why.
<davidlt[m]> So if it happens -> restart.
<rwmjones> ok
<davidlt[m]> I do plan to submit a new GCC build (the final one).
jcajka has quit [Quit: Leaving]
zsun has joined #fedora-riscv
gotmax23 has quit [Ping timeout: 246 seconds]
gotmax23 has joined #fedora-riscv
zsun has quit [Quit: Leaving.]
<rwmjones> davidlt[m]: soooo
<rwmjones> it failed in the same place I think
<davidlt[m]> Yeah
<davidlt[m]> Just in case I am rebuilding annobin plugin
<rwmjones> I'm going to try another local build but first make sure I have exactly the same compiler toolchain pkgs installed
<davidlt[m]> Oh, I forgot to disable LTO in annobin...
<rwmjones> I still have the older binutils (2.39) installed locally
<rwmjones> everything else looks the same
<rwmjones> but yeah LTO could be an issue simply because of memory usage
<davidlt[m]> Nah, Clang cannot do LTO
<davidlt[m]> There is no ld.gold support for riscv64.
<rwmjones> build failure was in g++
<rwmjones> & it was using lto
<davidlt[m]> So if your toolchain is Clang, you should disable LTO.
<davidlt[m]> Yeah, but I want to rebuild annobin GCC plugin just in case
<rwmjones> sure
<rwmjones> if it succeeds again locally I might try a scratch build with LTO disabled in the spec file
<davidlt[m]> Yeah, disabling LTO is relatively common "fix".
<davidlt[m]> I feel that on riscv64 it's still a bit more shaky than on other arches.
<davidlt[m]> That's not only about memory usage.
<rwmjones> agreed
<davidlt[m]> Also disabling LTO seems to allow to get rid of TEXTRELs
<davidlt[m]> I think I saw a patch for riscv64 + LTO for Clang using bfd linker.
<davidlt[m]> Is there a good System76 alternative in EU?
<davidlt[m]> Well, I used Tuxedo (German). There is another one in UK, another in Spain. Pretty much similar standard systems.
<davidlt[m]> rwmjones: annobin plugin was rebuilt, you might want to pull it in.
<davidlt[m]> Or just restart dnf5 build
<davidlt[m]> There is a newer version in F39, which at some point will land into F38 too.
<davidlt[m]> Are there any Fedora laptop deals around (that aren't focused on US market)?
<rwmjones> will do
<rwmjones> what are fedora laptop deals? you mean laptops that can run fedora?
<davidlt[m]> No, there used to be Fedora contributor deals
<rwmjones> oh, no idea about that
<nirik> lenovo has/had some... see https://communityblog.fedoraproject.org/expanded-availability-contributor-discount-for-lenovo-thinkpad-laptops-with-fedora-workstation/ (basically, login with your fedoraproject.org email alias address)
<davidlt[m]> Ah yes, I remember this.
<davidlt[m]> There was an email in 2022 about Linux systems being available in Europe.
<davidlt[m]> > I think the below is the full list currently:... (full message at <https://libera.ems.host/_matrix/media/v3/download/libera.chat/faede4853f43ae2a31bb90b15d13a94bb20c5869>)
<nirik> yeah, I think it changes tho... so no idea whats in there right now
ahs3 has joined #fedora-riscv
ahs3 has quit [Quit: ZNC 1.8.2+deb3+b4 - https://znc.in]
ahs3 has joined #fedora-riscv
davidlt_ has quit [Ping timeout: 276 seconds]
<rwmjones> gaaa new version fails in the same way .. http://fedora.riscv.rocks/kojifiles/work/tasks/5457/1405457/build.log