Red Hat Bugzilla – Bug 415301
possible bug in gcc-fortran
Last modified: 2007-12-07 06:31:47 EST
Description of problem:
My package does not compile with gcc in rawhide.
Version-Release number of selected component (if applicable):
Broken (buggy) version is 4.1.2-35
Working version is 4.1.2-33
Steps to Reproduce:
Try to compile it with gcc-gfortran-4.1.2-35.
Package fails to compile.
Package is build successfully and can be included in Astronomy spin :)
Failed task in koji, rawhide:
Correct build in koji, F8: http://koji.fedoraproject.org/koji/taskinfo?taskID=280596
Source RPM: http://mmahut.fedorapeople.org/reviews/munipack/
Worth noting that the package only fails on ppc and ppc64.
Another thing is that fortran compiler doesn't return with a failure -- it just
seems to skip generating .mod file output for some reason. It is rather
unpredictable which .mod file doesn't get built (ie varies between the builds).
Seems like it's due to -j8 and the modules building concurently, due to poorly
specified dependencies of makefile targets. I tried a scratch build without
mflags, and it succeeded.
That's just buggy Makefile. Whenever you USE some_module, you need a Makefile
dependency on .o file which creates the corresponding *.mod file.