This update has been unpushed.

@boycottsystemd1 I have been unable to reproduce this, so I'm planning to push this to stable unless you are able to provide more information about the failure.

@boycottsystemd1 Are you able to provide any more information about this failure?

I'm seeing the same EGL errors with the current version in f40: blender-4.2.2-2.fc40, but no segfault. Have you tested with blender-4.2.2-2.fc40 too ?

Can you upload the /tmp/blender.crash.txt file?

I can confirm that BZ#2294591 is fixed. Here is what I tested:

 sudo dnf install --enablerepo=updates-testing --refresh --advisory=FEDORA-2024-0b42137285 blender unzip
 curl -O -L https://mirrors.iu13.net/blender/demo/bundles/bundles-3.0/asset-demo-bundle-3.0-cube-diorama.zip
unzip asset-demo-bundle-3.0-cube-diorama.zip
cd cube_diorama/
blender  -b --debug-cycles -E CYCLES -f 20 cube_diorama.blend -- --cycles-device HIP --cycles-print-stats

Conflicting with compiler-rt < 19 is the next best thing we can do.

I think adding this conflict is the best solution for now given where we are at in the cycle. We were planning to move to a more python-like main/compat package solution, so maybe we can address this problem then.

New update with 18.1.6: FEDORA-2024-9f3b00db88

The compiler-rt test is failing due to https://bugzilla.redhat.com/show_bug.cgi?id=2271612.

If there are specific issues with the packages that are making it necessary to do rebuilds, please file bugs for these.

We were not anticipating that the ROCm packages would need to be rebuilt. When we discussed this issue back in December with the ROCm maintainers, I thought the conclusion was that ROCm would continue using llvm 17 via the compat packages and this would reduce the need for doing rebuilds. We added virtual provides to the compat packages specifically to support ROCm in this way, and we also added a completely new compat package (compiler-rt17) for this.

clang17-17.0.6-7.fc40 should fix the ROCm issue.

@tflink Can you file a bug so we are all on the same page?

The compiler-rt test failure looks like a bug in the test.

@vkadlcik are you looking into the baseos-qe.koji-build.scratch-build.validation failure? redhat-rpm-config updates are blocked until a fixed version of annobin lands in rawhide, so it would be good if we could get this update pushed soon.

I have a fix for the failing test: https://src.fedoraproject.org/rpms/annobin/pull-request/32#

@vkadlcik can these test failures be waived?

Is there some reason this hasn't been pushed to stable yet? It looks like the failing tests have been waived.

Here is the testing log I'm looking at: https://artifacts.dev.testing-farm.io/a2cb816f-8ce2-4c19-997f-01c421f7e9da/work-tests.ymlVu0RtR/ansible-output.txt