[Bug fortran/100136] [11/12/13 Regression] ICE, regression, using flag -fcheck=pointer

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100136

anlauf at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-valid-code

--- Comment #6 from anlauf at gcc dot gnu.org ---
(In reply to José Rui Faustino de Sousa from comment #3)
> Only handles the ICE...
> 
> Patch posted:
> 
> https://gcc.gnu.org/pipermail/fortran/2021-April/055949.html

Hi Jose,

unfortunately it appears that your patch was never reviewed,
and you did not ping it.

I think the patch is mostly fine, but that it should use
POINTER_TYPE_P (TREE_TYPE (tmp)), so that it reads (adjusted to mainline):

diff --git a/gcc/fortran/trans-expr.cc b/gcc/fortran/trans-expr.cc
index 850007fd2e1..e35ea2fc790 100644
--- a/gcc/fortran/trans-expr.cc
+++ b/gcc/fortran/trans-expr.cc
@@ -7220,16 +7220,15 @@ gfc_conv_procedure_call (gfc_se * se, gfc_symbol * sym,
              else
                goto end_pointer_check;

+             tmp = parmse.expr;
              if (fsym && fsym->ts.type == BT_CLASS)
                {
-                 tmp = build_fold_indirect_ref_loc (input_location,
-                                                     parmse.expr);
+                 if (POINTER_TYPE_P (TREE_TYPE (tmp)))
+                   tmp = build_fold_indirect_ref_loc (input_location, tmp);
                  tmp = gfc_class_data_get (tmp);
                  if (GFC_DESCRIPTOR_TYPE_P (TREE_TYPE (tmp)))
                    tmp = gfc_conv_descriptor_data_get (tmp);
                }
-             else
-               tmp = parmse.expr;

              /* If the argument is passed by value, we need to strip the
                 INDIRECT_REF.  */

Are you still working on gfortran?  Or should I pick up your patch and resubmit
for you?