In the Fortran standard, I think it is best explained
in the description of the RANK intrinsic:

"Example. If X is an assumed-rank dummy argument and
its associated effective argument is an array of rank,
RANK(X) has the value 3."

That's already well tested in assumed_rank_16.f90;
however, as the PR shows, this should not be reset
to "-1" when passing it further on as actual argument to
another assumed-rank dummy argument.

OK for mainline?
Reported against GCC 10, not a regression but simple wrong-code fix;
does it make sense to apply there was well?

Tobias

-----------------
Mentor Graphics (Deutschland) GmbH, Arnulfstrasse 201, 80634 München Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Frank Thürauf