This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [3.3 branch] IA64 bootstrap failure
- From: "Zack Weinberg" <zack at codesourcery dot com>
- To: Hans-Peter Nilsson <hp at bitrange dot com>
- Cc: "H. J. Lu" <hjl at lucon dot org>, "'gcc at gcc dot gnu dot org '" <gcc at gcc dot gnu dot org>
- Date: Tue, 08 Jul 2003 21:52:28 -0700
- Subject: Re: [3.3 branch] IA64 bootstrap failure
- References: <Pine.BSF.4.44.0307090017220.26903-100000@dair.pair.com>
Hans-Peter Nilsson <hp@bitrange.com> writes:
> On Tue, 8 Jul 2003, H. J. Lu wrote:
>> On Tue, Jul 08, 2003 at 04:30:11PM -0700, Jim Wilson wrote:
>> > On Tue, 2003-07-08 at 15:56, S. Bosscher wrote:
>> > > Doesn't valgrind check for uninitialized/out-of-bounds memory refs?
>> > Ah, yes, I keep forgetting about valgrind. Maybe I should try using it
>> > someday.
>> I tried valgrind with a cross compiler to ia64 from ix86. I didn't
>> find anything serious:
>>
>> http://gcc.gnu.org/ml/gcc/2003-07/msg00084.html
>
> It looks by "tried valgrind" you man running just plain valgrind
> instead of configure --enable-checking=valgrind. The latter
> would instrument the memory management inside GCC (and properly
> handle the mmap end+1 thingy), so without it you would miss that
> reused-pool-entries are actually uninitialized memory accesses.
For a bug like this, suggest also using --param ggc-min-expand=0
--param ggc-min-heapsize=0 on the file being miscompiled - this
forces a garbage collection run at every opportunity, which will
expose more uses of freed memory. --enable-checking=gcac,valgrind
would have the same effect, but on everything, and the build would
probably run into August.
zw