[Bug fortran/36366] [4.3/4.4 Regression] ICE in gfc_conv_component_ref

["pault at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

------- Comment #2 from pault at gcc dot gnu dot org  2008-06-05 19:45 -------

> At the moment, I cannot see what draws these together. Nor can I see any
> difference in the .mod files for the cases that work....

Ah... go it!  gcc-4.2 had a derived type list for each namespace and care was
taken to apply the same backend_decl to each like derived type in all
namespaces, including formal namespaces.  gcc-4.3 onwards uses a single list,
generate in resolve.c.  This was failing to check derived type components, with
the result that they could be missed in formal namespaces.

This fixes the problem:

Index: gcc/fortran/resolve.c
===================================================================
*** gcc/fortran/resolve.c       (revision 136412)
--- gcc/fortran/resolve.c       (working copy)
*************** resolve_fl_derived (gfc_symbol *sym)
*** 7637,7642 ****
--- 7637,7647 ----
          return FAILURE;
        }

+       /* Ensure that all the derived type components are put on the
+        derived type list; even in formal namespaces.  */
+       if (c->ts.type == BT_DERIVED && c->ts.derived)
+       resolve_fl_derived (c->ts.derived);
+ 
        if (c->pointer || c->allocatable ||  c->as == NULL)
        continue;

It's mine!  If it regtests OK, I'll apply it as obvious.

Paul


-- 

pault at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |pault at gcc dot gnu dot org
                   |dot org                     |
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2008-05-29 12:05:35         |2008-06-05 19:45:48
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36366