https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90329
--- Comment #32 from Jakub Jelinek <jakub at gcc dot gnu.org> --- BTW, as can be seen on https://fortran.godbolt.org/z/ckMt1t it is a pure luck that these bogus C/C++ wrappers seem to work, if one tries simple subroutine foo (a, b, c, d, e, f, g, h) integer :: b, c, d, e, f, g, h character(len=1) :: a call bar (a, b, c, d, e, f, g, h) end subroutine foo then at -O2 both older gfortran and ifort emit a tail call, which will overwrite the hidden string length argument, so if one prototypes this as void foo_ (char *a, blas_int b, blas_int c, blas_int d, blas_int e, blas_int f, blas_int g, blas_int h); and calls it with say foo_ ("1", 2, 3, 4, 5, 6, 7, 8); then it will overwrite some variables in the caller both with older gfortran and with ifort.
