Le 19/10/2026 à 20:50, Jerry DeLisle a écrit :
See attached patch. When a pointer array function returns a pointer to
a component of a derived-type array, or to the real or imaginary part of
a complex array, the elements of the result are spaced by the span of
its descriptor, not by the element size. The caller indexed the result
by the element size, so expressions, ASSOCIATE names, SELECT TYPE on a
CLASS pointer associated with the result, and actual arguments all saw
the wrong elements. The fix makes each of these uses honor the span. An
actual argument for a dummy that does not accept a span is copied in and
out, but only when a run-time check finds the elements are not already
laid out as the dummy expects.
Regression tested on x86_64.
OK for mainline?
Yes, please go ahead.
Regards,
Jerry
