This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
RE: core dumps in optimized case with gcc 3.0 and 3.0.1 (regression a gainst 2.95)
- To: "'Andreas Jaeger'" <aj at suse dot de>
- Subject: RE: core dumps in optimized case with gcc 3.0 and 3.0.1 (regression a gainst 2.95)
- From: "Nuesser, Wilhelm" <wilhelm dot nuesser at sap dot com>
- Date: Mon, 27 Aug 2001 17:06:57 +0200
- Cc: "'gcc at gcc dot gnu dot org'" <gcc at gcc dot gnu dot org>, "Hiller, Thomas" <thomas dot hiller at sap dot com>
Hi,
first of all thanks for your answer.
>
> I would play around with the various compiler flags and check whether
> one of the new optimizations causes the problem.
>
> A usual suspect is that your program breaks the strict aliasing rules,
> I would try -fno-strict-aliasing first.
>
> Andreas
Well, I tested some combinations and since our makes take some time
only a few results are already available:
1) compiled only the library which cores with -fno-strict-aliasing
-> core
2) compiled the whole bunch with -fno-strict-aliasing
-> core
3) compiled averything with -O1
-> no more cores
Seems like we have to test some other combinations. I'll start from
-O1 and include more and more options, following the sequence given
in toplev.c to get at least a little bit systematics into it.
Any better idea would be greatly appreciated ;-)
Willi