https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128608
Kris Fabmoos <kris.fabmoos at gmail dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |RESOLVED
Resolution|--- |DUPLICATE
--- Comment #2 from Kris Fabmoos <kris.fabmoos at gmail dot com> ---
(In reply to Drea Pinski from comment #1)
> I think this is a dup of bug 128092 or bug 128171 .
Thanks for the quick look - yes, it is the same bug as PR128092.
I ran the testcase from PR128092 comment 0 with our 15.2.0 build. It aborts
at -O2 and passes with -fno-ipa-bit-cp, just like ours, and
-fdump-ipa-cp-details shows the same picture: a PASS THROUGH jump function
with max_expr, and the callee's bits coming out as a constant. The testcase
from PR128171 fails here too (-O3 -fwhole-program -fno-inline) and passes
with -fno-ipa-bit-cp.
Sorry for the noise. I did search before filing, but clearly not well enough.
One small thing that might be useful: our test goes through a signed 64-bit
parameter with max (x, 1), while the tests in 128092 go from int to unsigned
with max (x, 0). I can't build trunk here, so I have not been able to try
the patch from attachment 65929 on it. If someone who has the patch applied
wants one more testcase, the bug.c attached here is small and has no
includes.
*** This bug has been marked as a duplicate of bug 128092 ***
