2008/8/29, Tor Arntsen wrote:

> Segmentation fault
>
> o The above was with the Intel compiler, icc 9.1
> (--with-ssl (or --without-ssl) --disable-debug --disable-ipv6)
>
> o It does _not_ fail with gcc 4.2.3.
>
> o It fails the same way with both icc 9.1 and icc 10.0)
>
> o It does _not_ fail if I use --enable-debug instead of --disable-debug
> o It does _not_ fail if I build with CFLAGS="-O0 -g" (i.e. optimizations off)
> o In those cases where it doesn't fail (e.g. -O0), a valgrind run doesn't
> indicate any problems either.
>
> Tricky.

Eek, yeah!

In order to view the differences in both build logs, compiler options
which are being used, etc...

Maybe you could run runtests.pl with --runtestopts="-p 557" in both
cases '--enable-debug' and '--disable-debug'.

This would only execute test case #557, while allowing the build logs
to be sent to the autobuild page.

If that test case core-dumps, then running runtests.pl with
--runtestopts="none" in both cases '--enable-debug' and
'--disable-debug' which will prevent any test from running.

This seems like a configuration issue, but changes to configure.ac or
acinclude.m4 near the jul-10 date you mention are dated: jul-02,
jul-04 and jul-16.

-- 
-=[Yang]=-