[Bug fortran/60302] [4.9 Regression] ICE with c_f_pointer and android cross compiler

[janus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60302

janus at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
                 CC|                            |burnus at gcc dot gnu.org
           Assignee|unassigned at gcc dot gnu.org      |janus at gcc dot gnu.org

--- Comment #4 from janus at gcc dot gnu.org ---
(In reply to janus from comment #3)
> The function 'gfc_check_c_f_pointer' was added in r197053

... and I think that's indeed the culprit. This patch fixes it:


Index: gcc/fortran/check.c
===================================================================
--- gcc/fortran/check.c    (revision 207988)
+++ gcc/fortran/check.c    (working copy)
@@ -3944,16 +3944,17 @@ gfc_check_c_f_pointer (gfc_expr *cptr, gfc_expr *f
   if (shape)
     {
       mpz_t size;
-
-      if (gfc_array_size (shape, &size)
-      && mpz_cmp_ui (size, fptr->rank) != 0)
+      if (gfc_array_size (shape, &size))
     {
+      if (mpz_cmp_ui (size, fptr->rank) != 0)
+        {
+          mpz_clear (size);
+          gfc_error ("SHAPE argument at %L to C_F_POINTER must have the same "
+            "size as the RANK of FPTR", &shape->where);
+          return false;
+        }
       mpz_clear (size);
-      gfc_error ("SHAPE argument at %L to C_F_POINTER must have the same "
-             "size as the RANK of FPTR", &shape->where);
-      return false;
     }
-      mpz_clear (size);
     }

   if (fptr->ts.type == BT_CLASS)



We should not free 'size' unless gfc_array_size is successful!