http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49023
Tobias Burnus <burnus at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |burnus at gcc dot gnu.org --- Comment #1 from Tobias Burnus <burnus at gcc dot gnu.org> 2011-05-18 16:09:54 UTC --- ==31261== Invalid read of size 8 ==31261== at 0x510673: gfc_resolve_expr (resolve.c:2674) ==31261== by 0x50E37C: resolve_code (resolve.c:9026) That's at first "if" block gfc_iso_c_func_interface. In this example, one has expr_type == EXPR_ARRAY. I think the following patch should be OK (for F2003, in F2008, the pointer returned by functions also counts as variable). Additionally, one needs to add a check for the second argument; the program crashes for C_ASSOCIATED(x,[y]) at gfc_conv_array_constructor_expr, at fortran/trans-expr.c:4120 --- resolve.c (revision 173865) +++ resolve.c (working copy) @@ -2666,9 +2666,8 @@ gfc_iso_c_func_interface (gfc_symbol *sy gfc_typespec *arg_ts; symbol_attribute arg_attr; - if (args->expr->expr_type == EXPR_CONSTANT - || args->expr->expr_type == EXPR_OP - || args->expr->expr_type == EXPR_NULL) + if (args->expr->expr_type != EXPR_VARIABLE + || args->expr->symtree->n.sym->attr.flavor != FL_VARIABLE) { gfc_error ("Argument to '%s' at %L is not a variable", sym->name, &(args->expr->where));