-
Notifications
You must be signed in to change notification settings - Fork 1.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
uv 0.6.6
fails to build arm64 packages in the CI?
#12207
Comments
It looks like in the successful run faster-fifo is not built, perhaps because there's a cached build somewhere? I don't see logs showing it pulled from the cache. In the failing run, it looks like quite a few packages are being built from source. I'm not sure why yet. Does it pass with the cache disabled? It looks like there isn't an Regarding getting past the build error: it depends how |
The wheel cache version was last bumped in 0.6.4 #11738 |
Yes. On uv 0.6.5, with cache disabled, it seems to actually build and install faster-fifo just fine: https://github.com/NCSU-High-Powered-Rocketry-Club/AirbrakesV2/actions/runs/13888848816/job/38857347791?pr=179
In their readme, they only mention needing the basic gcc/g++/build-essential toolchain which should be available in the CI by default. I'm more confused on why the build is trying to use the x86_64 g++ in an arm64 environment ( |
The first link looks like the cache is used?
It might be clearer to run with It's interesting that the second succeeds. It looks like it's using |
I guess uv will still use leftover cache's even if I remove
So it's not a problem with the python with uv 0.6.6? The failing builds on 0.6.6 are trying to use Looks like
|
Confusing. Thanks for trying all those. So.. what's the failing case now? 0.6.6 when installing the whole project? |
Yep |
It looks like your original failing example is using a managed Python distribution
while the successful run you just shared is using
We won't be testing if there's a bug in the managed Python distributions if it's not being used. You can use Packages will pull the C compiler to use from the CPython build flags in |
Good catch! So the command So as I understand it, that command is using the latest python-build-standalone, but with uv 0.6.5? |
Yeah, that's the intent (and the logs seem to corroborate it) This looks like a regression upstream — I'll look into the exact cause and try to fix it on Monday. Thanks again! |
Okay I've confirmed this is in
but not for the
or in 0.6.5
This is because it's not in our expected replacements uv/crates/uv-python/src/sysconfig/mod.rs Lines 83 to 91 in 25e7209
And this indeed was caused by my pull request upstream |
This value can leak into user builds through sysconfig so it is not appropriate to use the `host_cxx`. See astral-sh/uv#12207 There's not a `musl-clang++` equivalent — so we skip this for those targets.
Can confirm that the latest uv |
Wonderful, thank you for following up. |
Summary
My github CI action failed for the github arm64 runner after updating from uv 0.6.5 to uv 0.6.6, with no changes in the lockfile between the two versions.
Here's the associated PR - NCSU-High-Powered-Rocketry-Club/AirbrakesV2#179
Example of a passing job, prior to that PR - https://github.com/NCSU-High-Powered-Rocketry-Club/AirbrakesV2/actions/runs/13743565044/job/38435717408
Example of a failing job, in that PR - https://github.com/NCSU-High-Powered-Rocketry-Club/AirbrakesV2/actions/runs/13888432712/job/38856321885?pr=179
Notice how the project successfully builds on x86_64, but doesn't on arm64, and the only thing which changed was updating
uv
to 0.6.6. If I pin the uv version to 0.6.5, the build and the job pass.I thought it could be a problem with
faster-fifo
, even though there is no new version released since then, so I tried debugging via a new job:but this passes successfully, so I'm not sure what the underlying issue is. Do I need to install gcc/g++ for the arm64 runner separately? Why does it work for uv 0.6.5 then?
Platform
Linux x86_64 and arm64 on github CI
Version
uv 0.6.6
Python version
Python 3.13.2
The text was updated successfully, but these errors were encountered: