https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68560
--- Comment #10 from Dominique d'Humieres <dominiq at lps dot ens.fr> --- > Created attachment 36902 [details] > Make the intrinsic procedure declarations less context-dependant > >I can't test this right now (bootstrap broken). Sorry about that. Anything serious? > It basically unconditionally clears the ignore_optional flag (and removes it). > So that optional arguments are always passed. This make LTO happy but causes several regressions on the tests for (MIN|MAX)(VAL|LOC). I have looked at the failures for gfortran.dg/scalar_mask_2.f90 and I think they come from < _gfortran_smaxloc0_4_r4 (&atmp.30, D.3514, 0B, D.3516); --- > _gfortran_smaxloc0_4_r4 (&atmp.30, D.3511, &C.3512); '<' with the patch '>' without it.