pmix failed to build from source in Fedora rawhide/f33 https://koji.fedoraproject.org/koji/taskinfo?taskID=48349804 For details on the mass rebuild see: https://fedoraproject.org/wiki/Fedora_33_Mass_Rebuild Please fix pmix at your earliest convenience and set the bug's status to ASSIGNED when you start fixing it. If the bug remains in NEW state for 8 weeks, pmix will be orphaned. Before branching of Fedora 34, pmix will be retired, if it still fails to build. For more details on the FTBFS policy, please visit: https://fedoraproject.org/wiki/Fails_to_build_from_source
Created attachment 1708906 [details] build.log file build.log too big, will only attach last 32768 bytes
Created attachment 1708907 [details] root.log file root.log too big, will only attach last 32768 bytes
Created attachment 1708908 [details] state.log
I don't know why the post-build tests are now failing only on armv7hl (they worked fine prior to the recent f33 mass rebuilds) -- for now I am going to exclude that arch.
Excluding armv7hl breaks the whole mpi stack.
Now it passed on armv7hl and failed on aarch64: https://koji.fedoraproject.org/koji/taskinfo?taskID=48671697 I guess the tests are just unstable.
Koschei has been repeatedly rebuilding pmix on any dependency change for years now, so my feeling is that something changed in rawhide to create this new problem.
Can we do another build with disabled tests to get the arm build back? Like Christoph said - the openmpi stack is broken now and other packages can't be built.
Or build openmpi without external pmix. I am happy with either solution.
I never intended the armv7hl exclusion to be permanent, but since that arch exclusion is creating more problems than it solves ("breaks the mpi stack") I'll disable the tests for that arch. This would also be a temporary solution of course until upstream or someone else finds the real problem.
All supported arches are back in rawhide. No promises as to whether or not pmix actually works on arm -- the tests have been disabled.
Thanks, let's work with upstream to figure this out.
The test timeouts were too brief. Fixed upstream (timeouts lengthened) and patched in Fedora pending next pmix release. Closing.