On Wed, 5 Apr 2006, Otto Moerbeek wrote:
> This has been spotted before. The PR was closed since we do not
> support non-standard options to gcc, so it does not affect OpenBSD base.
>
> On the other hand, I find this interesting, since it might be an
> indication of other bugs lurking. Good that you provided a simple test
> case!
Interesting is:
The problem disappears if the test program is compiled with
-O2 -fno-gcse -march=i686
So it's the common subexpression elimination in combination with i686
that's exposing/introducing/... the bug.
It could be debian has -fno-gcse in it's default config... but that's
just a wild guess.
-Otto
