Bootstrap miscomparison on mainline with checking disabled

[law@redhat.com]

In message < 469A0D4E-A9A6-11D8-A4A1-000393A6D2F2@physics.uc.edu >, Andrew Pinski
 writes:
 >
 >On May 19, 2004, at 10:48, law@redhat.com wrote:
 >
 >> In message < 1084974508.3630.240.camel@localhost.localdomain >, Diego 
 >> Novillo wri
 >> tes:
 >>>
 >>> --=-9XZfCcDdb7vvYOsYxIaK
 >>> Content-Type: text/plain
 >>> Content-Transfer-Encoding: 7bit
 >>>
 >>>
 >>> Folks,
 >>>
 >>> One of these patches (attached) is causing a bootstrap miscomparison 
 >>> of
 >>> mainline with checking disabled.  Could you look into it?
 >> It would help if you could at least mention what target.....
 >
 >Well dwarf2out.c is being miscompiled by phiopt:
 >
 >-  if (T.2057_36 != 0B) goto <L0>; else goto <L5>;
 >-
 >-<L0>:;
 >-
 >-  # rtl_335 = PHI <T.2057_36(1), 0B(0)>;
 >-<L5>:;
 >+  rtl_335 = 0B;
 >
 >I will look into why it is being replaced with 0 instead of rtl_335 
 >which
 >is the correct replacement.
Can you please send me the testcase (particularly if you've already 
trimmed it down to a particular function or sub-function of dwarf2out).

THanks,
jeff