I think I found a bug in the gfortran compiler related to compiling legacy code. (such as F77 specified by -std=legacy or -std=f90  or anything older than f95 )
This relates to floating point math. I can try to report it, though I think I won't be taken seriously, as it relates to a very old code base that became archived in 2022.
yeah, looking into that, I cant raise issues with it, because it has been archived.

Affected files:
../sysdeps/ieee754/flt-32/math_config.h
 ../sysdeps/ieee754/flt-32/e_expf.c
(and everything else in ../sysdeps/ieee754/flt-32/ of libc)

this effects floating point numbers where code is compiled with flags for versions of fortran older than F95
What I can do is try building from a source before the archival.