Hi FX, > While this is fresh in your memory, could I suggest you have a look at this > FINDLOC issue, which seems possibly related: > https://gcc.gnu.org/pipermail/fortran/2020-September/055016.html > and further messages from Thomas Koenig?
I briefly checked this, but the issue with FINDLOC seems to have a different origin than MINLOC/MAXLOC. Let's see what Thomas makes of it. There are also PRs with ICEs when the KIND argument is present, e.g. PR87711 with len_trim. The non-uniformness of the internal treatment of intrinsics w.r.t. to this optional argument within gfortran is a bit frustrating. Cheers, Harald