This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [Bug optimization/12268] New: Miscompilation of fortran/match.c
- From: Diego Novillo <dnovillo at redhat dot com>
- To: "gcc-bugzilla at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>
- Cc: "gcc-patches at gcc dot gnu dot org" <gcc-patches at gcc dot gnu dot org>
- Date: Sat, 13 Sep 2003 19:45:33 -0400
- Subject: Re: [Bug optimization/12268] New: Miscompilation of fortran/match.c
- Organization: Red Hat Canada
- References: <20030913153208.12268.pbrook@gcc.gnu.org>
On Sat, 2003-09-13 at 11:32, pbrook at gcc dot gnu dot org wrote:
> The following testcase fails when compiled with -O. Somewhere along the line
> the assignment in test() gets removed. Code derived from fortran/match.c.
>
This patch ought to fix it. We were not registering the fact that
call-clobbered pointers may end up pointing to global memory.
Bootstrap and tests in progress. Will commit afterwards.
Diego.
* tree-dfa.c (add_referenced_var): Call-clobbered pointers may
point to global memory.
Index: tree-dfa.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/Attic/tree-dfa.c,v
retrieving revision 1.1.4.158
diff -d -c -p -r1.1.4.158 tree-dfa.c
*** tree-dfa.c 13 Sep 2003 20:21:26 -0000 1.1.4.158
--- tree-dfa.c 13 Sep 2003 23:37:16 -0000
*************** add_referenced_var (tree var, struct wal
*** 2551,2556 ****
--- 2551,2558 ----
{
add_call_clobbered_var (var);
v_ann->is_call_clobbered = 1;
+ if (POINTER_TYPE_P (TREE_TYPE (var)))
+ v_ann->may_point_to_global_mem = 1;
}
}