I am still investigating this issue, but a work around could be:
The interface definition of "xyz" in interfaces module can be renamed to a module local name and this error can be avoided as shown below.
use ... *** code ***
end subroutine xyz
Thanks for your answer, the renaming works indeed. However, applying the workaround to all the files would require to edit quite a bit. I will think on a python script, maybe.
What I was wondering: Is the openf95 behaviour standard conform and ifort/gfortran just a bit relaxed or is it the other way around? I would guess that ifort/gfortran are correct as our approach seems to be the most obvious one if dealing with legacy code. (Unfortunately, ifort does not check whether the interface in interfaces.f90 conforms with interface of the routine itself, so lots of chances to make erros... I do not know about gfortran).
Anyway, I also encountered a
Assertion failure at line 3637 of /local/home/qa/nightly_build_avx/sandbox/open64/osprey/be/lno/simd.cxx:
### Compiler Error in file xyz..f90 during Loop Nest Optimizer phase
when using -O3 optimizer flag with version 22.214.171.124. Flag -O2 does not give this error. I will try to reduce the large source file to reproduce the error with a smaller example.