https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63674
janus at gcc dot gnu.org changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |accepts-invalid
Status|UNCONFIRMED |NEW
Last reconfirmed| |2014-12-10
CC| |janus at gcc dot gnu.org
Ever confirmed|0 |1
--- Comment #1 from janus at gcc dot gnu.org ---
(In reply to Valery Weber from comment #0)
> the following code is compiling fine with 4.9.1, but shouldnt gcc complain
> about calling a nonpure procedure from a pure one?
It definitely should, and the problem also occurs with the 5.0 trunk.
The part that checks for pureness in 'resolve_function' is apparently not
applied to procedure-pointer components, but could e.g. be transferred to
'resolve_expr_ppc' like this:
Index: gcc/fortran/resolve.c
===================================================================
--- gcc/fortran/resolve.c (Revision 218598)
+++ gcc/fortran/resolve.c (Arbeitskopie)
@@ -6048,6 +6048,7 @@ static bool
resolve_expr_ppc (gfc_expr* e)
{
gfc_component *comp;
+ const char *name = NULL;
comp = gfc_get_proc_ptr_comp (e);
gcc_assert (comp != NULL);
@@ -6074,6 +6075,32 @@ resolve_expr_ppc (gfc_expr* e)
if (!update_ppc_arglist (e))
return false;
+ if (!pure_function (e, &name))
+ {
+ if (forall_flag)
+ {
+ gfc_error ("Reference to non-PURE function '%s' at %L inside a "
+ "FORALL %s", comp->name, &e->where,
+ forall_flag == 2 ? "mask" : "block");
+ return false;
+ }
+ else if (gfc_do_concurrent_flag)
+ {
+ gfc_error ("Reference to non-PURE function '%s' at %L inside a "
+ "DO CONCURRENT %s", comp->name, &e->where,
+ gfc_do_concurrent_flag == 2 ? "mask" : "block");
+ return false;
+ }
+ else if (gfc_pure (NULL))
+ {
+ gfc_error ("Reference to non-PURE function '%s' at %L "
+ "within a PURE procedure", comp->name, &e->where);
+ return false;
+ }
+
+ gfc_unset_implicit_pure (NULL);
+ }
+
gfc_ppc_use (comp, &e->value.compcall.actual, &e->where);
return true;
This is sufficient to trigger the right error for the test case, but more work
needs to be done: 'pure_function' needs to be taught to handle PPCs and one
needs to add a pureness check for procedure-pointer assignments. Finally it
should be checked if the same problem also applies to 'ordinary' procedure
pointers (which are not a type component) and to the case of subroutines
(instead of functions).
