I have two packages where a workaround of reducing the compilation optimization level
avoids runtime failures:
1 Use of -O1 to avoid a numerical test error "zdot wrong"
https://github.com/GlobalArrays/ga/issues/249
2 Use of -fno-lto to avoid segmentation faults https://gitlab.com/QEF/q-e/-/issues/460
It seems like the above problems are specific to gfortran-12 on f36.
Thanks for the %global macros.
How common in f36, and in general in fedora is the need for reducing the optimization
levels?
I don't have small examples to reproduce the failures, but both issues have a
Dockerfile that can be used to compile the programs and trigger the failures (it takes 10
- 30 minutes). They include the original FTBFS bugzilla links too, in case they should be
reassigned as gcc bugs instead.