[Bug middle-end/46488] [4.5 regression] server/core_filters.c from apache httpd 2.2.17 miscompiled at -O3

["pcpa at mandriva dot com.br" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46488

--- Comment #39 from Paulo CÃsar Pereira de Andrade <pcpa at mandriva dot com.br> 2010-12-01 18:25:50 UTC ---
(In reply to comment #38)
> (In reply to comment #37)
> > > If you'd re-use the same type the issue would go away as well (I think).
> > 
> > Yes, this works for the testcase, but probably only because the problematic
> > fields are now volatile too (APR_RING_ENTRY has volatile fields, APR_RING_HEAD
> > doesn't).  The warning is still emitted for the more natural:
> > 
> > #define APR_RING_SENTINEL(hp, elem, link)                \
> >     (struct elem *)((char *)(hp) - APR_OFFSETOF(struct elem, link))
> > 
> > with -Wstrict-aliasing so the underlying aliasing issue is probably still
> > there.
> > 
> > Closing as "invalid" since the original SPARC issues were fixed elsewhere and
> > the -O3 issue is an aliasing violation in the source code.
> 
>   Maybe this should this be considered a problem in the sense that it
> does not warn with -Wstrict-aliasing ? As in this case, httpd source
> was patched to "correct" the warning, but the problem remained.
>   I mean, otherwise it is a kind of "please use -fno-strict-aliasing"
> for everything you do not want to risk having incorrect code generation.

Damn, sorry for quickly replying to myself. It warns with both
-Wstrict-aliasing=1 and -Wstrict-aliasing=2