[Bug fortran/125762] Segfault in gfc_conv_scalarized_array_ref for a DATA statement

[mikael at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125762

--- Comment #6 from Mikael Morin  ---
(In reply to Jerry DeLisle from comment #5)
> Created attachment 64791 [details]
> Patch with a better approach
> 
> Regression tested OK

I still find it unbelievable the compiler has been working with non-resolved
expressions  for so long, but looks good.
OK to push with the real PR number in the testcase and a suitable ChangeLog
entry.

[Bug fortran/125762] Segfault in gfc_conv_scalarized_array_ref for a DATA statement

[jvdelisle2 at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125762

Jerry DeLisle  changed:

   What|Removed |Added

  Attachment #64775|0   |1
is obsolete||

--- Comment #5 from Jerry DeLisle  ---
Created attachment 64791
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=64791&action=edit
Patch with a better approach

Regression tested OK

[Bug fortran/125762] Segfault in gfc_conv_scalarized_array_ref for a DATA statement

[jvdelisle2 at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125762

--- Comment #4 from Jerry DeLisle  ---
The new fix:

diff --git a/gcc/fortran/resolve.cc b/gcc/fortran/resolve.cc
index 57561e8686f..526755fef42 100644
--- a/gcc/fortran/resolve.cc
+++ b/gcc/fortran/resolve.cc
@@ -19787,10 +19787,15 @@ resolve_data_variables (gfc_data_variable *d)
 static void
 resolve_data (gfc_data *d)
 {
+  gfc_data_value *v;

   if (!resolve_data_variables (d->var))
 return;

+  for (v = d->value; v; v = v->next)
+if (v->expr && !gfc_resolve_expr (v->expr))
+  return;
+
   values.vnode = d->value;
   if (d->value == NULL)
 mpz_set_ui (values.left, 0);

[Bug fortran/125762] Segfault in gfc_conv_scalarized_array_ref for a DATA statement

[jvdelisle2 at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125762

--- Comment #3 from Jerry DeLisle  ---
(In reply to Mikael Morin from comment #2)
> How is it possible the array ref remains AR_UNKNOWN?
> resolve_data_variables should call gfc_resolve_expr on every value.

Very good point. Its related to the ordering of gfc_resolve_values and
gfc_resolve_data.
The resolve_values runs before resolve_data in the same pass. So when
resolve_values traverses the namespace, p1->value is still NULL. Then
resolve_data stores the unresolved t(3) (with AR_UNKNOWN) into p1->value — and
resolve_values never runs again.

I am testing a different patch.

[Bug fortran/125762] Segfault in gfc_conv_scalarized_array_ref for a DATA statement

[mikael at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125762

Mikael Morin  changed:

   What|Removed |Added

 CC||mikael at gcc dot gnu.org

--- Comment #2 from Mikael Morin  ---
How is it possible the array ref remains AR_UNKNOWN?
resolve_data_variables should call gfc_resolve_expr on every value.

[Bug fortran/125762] Segfault in gfc_conv_scalarized_array_ref for a DATA statement

[jvdelisle2 at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125762

Jerry DeLisle  changed:

   What|Removed |Added

 CC||jvdelisle2 at gmail dot com

--- Comment #1 from Jerry DeLisle  ---
Created attachment 64775
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=64775&action=edit
Proposed patch

[Bug fortran/125762] Segfault in gfc_conv_scalarized_array_ref for a DATA statement

[jvdelisle at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125762

Jerry DeLisle  changed:

   What|Removed |Added

   Last reconfirmed||2026-06-12
   Assignee|unassigned at gcc dot gnu.org  |jvdelisle at gcc dot 
gnu.org
 Status|UNCONFIRMED |NEW
 Ever confirmed|0   |1