Fedora Account System
Red Hat Associate
Red Hat Customer
openblas failed to build from source in Fedora rawhide/f44 https://koji.fedoraproject.org/koji/taskinfo?taskID=141172676 For details on the mass rebuild see: https://fedoraproject.org/wiki/Fedora_44_Mass_Rebuild Please fix openblas 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, openblas will be orphaned. Before branching of Fedora 45, openblas will be retired, if it still fails to build. For more details on the FTBFS policy, please visit: https://docs.fedoraproject.org/en-US/fesco/Fails_to_build_from_source_Fails_to_install/
Created attachment 2126409 [details] build.log file build.log too big, will only attach last 32768 bytes
Created attachment 2126410 [details] root.log file root.log too big, will only attach last 32768 bytes
Created attachment 2126411 [details] state.log
I can't reproduce the failure (test segfault) on Power9 based system, so perhaps this is hardware (Fedora builders are Power10) and/or compiler (GCC 16?) related.
Newer upstream releases > 0.3.29 contain some Power10 fixes, so perhaps just rebasing to newer upstream release could make the test failure go away.
My testing so far points to a combination of Power10 builder and GCC 16. Rebasing to newer upstream release won't (easily) help, because the new versions suffer from the same (0.3.30) or a different issue (0.3.31+).
Hi Jakub, please see the upstream issue on Github for my investigation so far.
Bisected to kernel/power/sbgemv_t_power10.c (everything from libopenblass*.a -O0 and just that one built normally works) and latest bisected to https://gcc.gnu.org/r16-1782 change (unfortunately, it causes significant RA choices in several functions in that file).