This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/71691] [6/7 Regression] wrong code at -O3 in both 32-bit and 64-bit modes on x86_64-linux-gnu (Floating point exception)
- From: "law at redhat dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Fri, 05 Aug 2016 22:03:41 +0000
- Subject: [Bug tree-optimization/71691] [6/7 Regression] wrong code at -O3 in both 32-bit and 64-bit modes on x86_64-linux-gnu (Floating point exception)
- Auto-submitted: auto-generated
- References: <bug-71691-4@http.gcc.gnu.org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71691
Jeffrey A. Law <law at redhat dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |law at redhat dot com
--- Comment #7 from Jeffrey A. Law <law at redhat dot com> ---
Damn interesting test.
I really wonder if what uncprop is doing is valid.
Essentially if i_48 has a potentially undefined value, but there are no paths
where its used undefined, then we're OK. Uncprop comes along and introduces a
use of i_48 on a path where it didn't exist before, now causing us to use the
uninitialized value. And boom.
And note this is a potential issue for all the transformations made by uncprop.