Follow up to PR 41807 - based on the comment at http://gcc.gnu.org/ml/fortran/2009-11/msg00181.html cf. also http://gcc.gnu.org/ml/fortran/2009-11/msg00208.html
"it is both unacceptable and unanticipated that gfc_is_constant_expr have side effects like that. Should we, perhaps copy the expression within gfc_is_constant_expr, test that and free the new expression before exiting?" and in http://gcc.gnu.org/ml/fortran/2009-11/msg00185.html "I think we have some other bugs lurking here. For example, why could we simplify the original expr in get_array_index and not it's copy. gfc_copy_expr is missing something. "The gfc_is_constant_expr issue is down in gfc_constant_ac. I think we may be trying to expand the constructors in more than one place, in gfc_constant_ac and somewhere else." -- Summary: gfc_is_constant_expr has unacceptable side effects Product: gcc Version: 4.5.0 Status: UNCONFIRMED Keywords: wrong-code Severity: normal Priority: P3 Component: fortran AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: burnus at gcc dot gnu dot org http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42189