Asus akmods builds failing

hey folks - noticed the akmods f40 asus builder is failing. I am pretty sure its because the rtl8814au-kmod package is out of date. I would appreciate some eyes on a few PRs and some COPR builds since I don't have the permissions.
GitHub
chore(ci): Update builder to 24.04 · ublue-os/akmods@7fecd52
A caching layer for pre-built Fedora akmod RPMs. Contribute to ublue-os/akmods development by creating an account on GitHub.
62 Replies
Noel
Noel•6mo ago
@Kyle Gospo do you handle these coprs? I don't have access to them.
mulderje
mulderje•6mo ago
pulling into the thread for reference: I think we will need to do a few things: 1) https://github.com/ublue-os/packages/pull/25 2) rtl8814au-kmod COPR package & rtl8814au COPR package builds 3) kick the akmods 40 build
GitHub
chore: update srccommit in rtl8814au-kmod.spec by mulderje · Pull R...
Update rtl8814au to latest commit to hopefully fix the akmods f40 asus builder Will likely need someone to kick the COPR build? Potentially closes: ublue-os/akmods#185
mulderje
mulderje•6mo ago
looks like there are a few people just requested build perms as well (feel free to ignore that last if preferred)
Noel
Noel•6mo ago
I will include @EyeCantCU and @bsherman on this thread.
mulderje
mulderje•6mo ago
thanks!
Noel
Noel•6mo ago
no problem! Thanks for pointing this out!
EyeCantCU
EyeCantCU•6mo ago
Can take a look at this in half an hour or so
mulderje
mulderje•6mo ago
thanks! lmk if you have any thoughts/questions
mulderje
mulderje•6mo ago
FYI just noticed this is also hitting bluefin asus 40. eg. https://github.com/ublue-os/bluefin/actions/runs/9099556968/job/25012367053
GitHub
bluefin 40 · ublue-os/bluefin@6b38313
An interpretation of the Ubuntu spirit built on Fedora technology - bluefin 40 · ublue-os/bluefin@6b38313
mulderje
mulderje•6mo ago
so that'll also need a kick at the end
EyeCantCU
EyeCantCU•6mo ago
Yeah. World will have to be rebuilt
mulderje
mulderje•6mo ago
yep. fingers crossed that spec update should fix it all. the upstream diff seems to indicate it should
EyeCantCU
EyeCantCU•6mo ago
Sorry. Sidetracked. Pulling up now
mulderje
mulderje•6mo ago
no worries. lmk if anything i can do
EyeCantCU
EyeCantCU•6mo ago
Thank you! Updating the package might help resolve the failure. I'm thinking we probably need to bump the version and reset the epoch too. I'll push a commit to your existing PR 🙂
mulderje
mulderje•6mo ago
sounds good. will take a look. fyi the version upstream didn't change at all, so maybe just epoch? or actually, wouldn't that just be release
EyeCantCU
EyeCantCU•6mo ago
Apparently there aren't any releases. Needs at least an epoch bump Bumped the 'release' and merged (not epoch sorry, in Wolfi mode)
mulderje
mulderje•6mo ago
yeah exactly the spec "release" field is what i meant 🙂
EyeCantCU
EyeCantCU•6mo ago
Kicking off a Copr build now
mulderje
mulderje•6mo ago
perfect. thanks thanks for catch if bumping the release in rtl8814au-kmod does it also need to be bumped to match in rtl8814au? nvm you already got it 🙂
EyeCantCU
EyeCantCU•6mo ago
No, you don't. It looks like rtl8814au was just building from HEAD. Just good to know when something was updated
Kyle Gospo
Kyle Gospo•6mo ago
Speaking of AK mod changes ryzen_smu and I'm sure some others need to be x86 only As they fail to build for arm
EyeCantCU
EyeCantCU•6mo ago
Can have a look at that too 🙂 Still failing. About to sit down for dinner but will look after
mulderje
mulderje•6mo ago
hmm ok thanks will take a quick look before dinner its the other one - rtl88xxau the source in that files just pulling head of the "v5.6.4.2" branch at build time. so if i'm reading that right, it looks like it'd have the source from the build from 22 days ago. maybe just hit those two builds again for now to pick up HEAD?
%global srcversion 5.6.4.2
# ... snip ...
URL: https://github.com/aircrack-ng/rtl8812au
Source0: %{url}/archive/refs/heads/v%{srcversion}.zip
%global srcversion 5.6.4.2
# ... snip ...
URL: https://github.com/aircrack-ng/rtl8812au
Source0: %{url}/archive/refs/heads/v%{srcversion}.zip
looks like there have been a few commits since then, but haven't looked beyond that also heading for dinner now though should probably look at adding in something like packit to auto bump these yeah, @EyeCantCU - can you just bump the rev number for those two packages and build them in COPR? i think that'll fix for now
mulderje
mulderje•6mo ago
could set up something like https://packit.dev/posts/monorepos to auto-build the COPRs to get them building
Introducing monorepository support | Packit
We are very happy to announce a major enhancement to Packit! We have now added support for
mulderje
mulderje•6mo ago
ive never used before, but could try to mock up a PR if interest but that may not do anything without there being releases upstream could also maybe do something like a cron to pull the latest commit revs in a GH action and then edit the files, and kick the builds either way a rev bump/build for now will hopefully fix
EyeCantCU
EyeCantCU•6mo ago
I'll give it a try
mulderje
mulderje•6mo ago
Thanks! I’d try myself but don’t have permissions
EyeCantCU
EyeCantCU•6mo ago
Still failing. Looks like zenergy now
mulderje
mulderje•6mo ago
@Kyle Gospo - any thoughts there? looks like you own the repo. current build failure on ASUS 40 in the c code
2024/05/16 14:32:03 akmodsbuild: /tmp/akmodsbuild.32dTszn9/BUILD/zenergy-kmod-0.0.git.30.98b5115f/_kmod_build_6.9.0-667.rog.fc40.x86_64/zenergy.c: In function ‘amd_create_sensor’:
2024/05/16 14:32:03 akmodsbuild: /tmp/akmodsbuild.32dTszn9/BUILD/zenergy-kmod-0.0.git.30.98b5115f/_kmod_build_6.9.0-667.rog.fc40.x86_64/zenergy.c:229:27: error: ‘struct cpuinfo_x86’ has no member named ‘x86_max_cores’
2024/05/16 14:32:03 akmodsbuild: 229 | sockets = cpus / c->x86_max_cores;
2024/05/16 14:32:03 akmodsbuild: | ^~
2024/05/16 14:32:03 akmodsbuild: make[2]: *** [scripts/Makefile.build:244: /tmp/akmodsbuild.32dTszn9/BUILD/zenergy-kmod-0.0.git.30.98b5115f/_kmod_build_6.9.0-667.rog.fc40.x86_64/zenergy.o] Error 1
2024/05/16 14:32:03 akmodsbuild: /tmp/akmodsbuild.32dTszn9/BUILD/zenergy-kmod-0.0.git.30.98b5115f/_kmod_build_6.9.0-667.rog.fc40.x86_64/zenergy.c: In function ‘amd_create_sensor’:
2024/05/16 14:32:03 akmodsbuild: /tmp/akmodsbuild.32dTszn9/BUILD/zenergy-kmod-0.0.git.30.98b5115f/_kmod_build_6.9.0-667.rog.fc40.x86_64/zenergy.c:229:27: error: ‘struct cpuinfo_x86’ has no member named ‘x86_max_cores’
2024/05/16 14:32:03 akmodsbuild: 229 | sockets = cpus / c->x86_max_cores;
2024/05/16 14:32:03 akmodsbuild: | ^~
2024/05/16 14:32:03 akmodsbuild: make[2]: *** [scripts/Makefile.build:244: /tmp/akmodsbuild.32dTszn9/BUILD/zenergy-kmod-0.0.git.30.98b5115f/_kmod_build_6.9.0-667.rog.fc40.x86_64/zenergy.o] Error 1
GitHub
ublue akmods 40 · ublue-os/akmods@9f48cad
A caching layer for pre-built Fedora akmod RPMs. Contribute to ublue-os/akmods development by creating an account on GitHub.
mulderje
mulderje•6mo ago
re-run the build or think you need to make changes? could also block on the asus build
Kyle Gospo
Kyle Gospo•6mo ago
if it's only failing on asus, they might have broken something in ther kernel probably best to pull it from there
M2
M2•6mo ago
It's only asus + nvidia that has failures on 40
mulderje
mulderje•6mo ago
is zenergy being used by any of the images?
Kyle Gospo
Kyle Gospo•6mo ago
bazzite atm
mulderje
mulderje•6mo ago
ok
Kyle Gospo
Kyle Gospo•6mo ago
but that's fsync
mulderje
mulderje•6mo ago
should we only build for fsync? ie 1) only build fsync or 2) turn off for asus
Kyle Gospo
Kyle Gospo•6mo ago
I think zenergy is useful, others may want it proper power monitoring for AMD CPUs
mulderje
mulderje•6mo ago
no preference here. will put it on no asus to start and can limit further later if it causes issues pr up shortly
mulderje
mulderje•6mo ago
GitHub
fix(zenergy): remove from ASUS build to fix builder by mulderje · P...
Thank you for contributing to the Universal Blue project! Please read the Contributor's Guide before submitting a pull request.
mulderje
mulderje•6mo ago
will need a few acks there and then someone to do the builder dance for me with that though it should unbreak the 40 build since zenergy was the last on the list being built @M2 - that should unbreak the asus build and everything downstream once in fyi
EyeCantCU
EyeCantCU•6mo ago
Change lgtm, but you're missing a backslash. Added a suggestion to the PR 🙂
mulderje
mulderje•6mo ago
doh added in from GH UI
EyeCantCU
EyeCantCU•6mo ago
Sounds good 🙂
mulderje
mulderje•6mo ago
thanks for the catch. hate it when i save after the git commit and dont pay attention i've never used that UI feature before. super handy!
EyeCantCU
EyeCantCU•6mo ago
It's great for quick edits. If you don't have it set up yet, I'd highly recommend VS Code. Tons of extensions that are helpful. I use a combination of that, nano, and Zed (my absolute favorite)
mulderje
mulderje•6mo ago
i've been toying around with vs code and helix lately. historically have liked vim for console editing since its what i'm used to. heard great things about zed been trying to figure out a good dev container workflow with vs code very much a git novice for anything more advanced than a basic merge, so the UI stuff is helpful for learning all that is there a way for me to kick off a PR build somehow? i've seen some projects where people can type in things like "/build" or whatever in a PR comment if they dont have commit access yet
EyeCantCU
EyeCantCU•6mo ago
I approved the runs. Let's see how this goes...
mulderje
mulderje•6mo ago
thanks! looks all green now
EyeCantCU
EyeCantCU•6mo ago
Merging!
mulderje
mulderje•6mo ago
great! hopefully we see the green trickle through everything thanks for going down the rabbit hole we need something to keep those specs up-to-date or will likely just happen again soon @EyeCantCU - is it intentional to not have a GH action/job for a push on main to kick the build? wondering if that should be added to the workflow
EyeCantCU
EyeCantCU•6mo ago
GitHub
ublue akmods 40 · ublue-os/akmods@2ead2b3
A caching layer for pre-built Fedora akmod RPMs. Contribute to ublue-os/akmods development by creating an account on GitHub.
mulderje
mulderje•6mo ago
yep the merge queue run worked, but once merged on main there isn't a build for "main" branch and 40 still shows as failing overall so far as i can tell
mulderje
mulderje•6mo ago
No description
mulderje
mulderje•6mo ago
ie not like the one on the bottom
mulderje
mulderje•6mo ago
No description
mulderje
mulderje•6mo ago
i think there needs to be something in the workflow file for bazzite. eg on --> push --> branches --> main
No description
mulderje
mulderje•6mo ago
think being the operative word there 🙂 and assuming you want that vs just the daily cron
EyeCantCU
EyeCantCU•6mo ago
Oh, I see what you mean. I believe it's set up this way to prevent us from having a gazzillion concurrent queued builds. So if we introduce a change that's small, we aren't wasting runners rebuilding everything each time
mulderje
mulderje•6mo ago
got it. makes sense
Want results from more Discord servers?
Add your server