This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
Re: mmix ICEs on mainline during building of libc
- From: Andreas Jaeger <aj at suse dot de>
- To: Hans-Peter Nilsson <hp at bitrange dot com>
- Cc: gcc-bugs at gcc dot gnu dot org, <dalej at apple dot com>
- Date: Sun, 02 Mar 2003 14:12:56 +0100
- Subject: Re: mmix ICEs on mainline during building of libc
- References: <Pine.BSF.4.44.0303020726170.50573-100000@dair.pair.com>
Hans-Peter Nilsson <hp at bitrange dot com> writes:
> On Sat, 1 Mar 2003, Andreas Jaeger wrote:
>> Bootstrapping mmix in a ueberbaum fails in string/wcsspn.c:
>
>>From the locus of the failure, it looks as if this change:
> 2003-02-28 Dale Johannesen <dalej at apple dot com>
That's what I also just figured out.
> * loop.c (struct movable): Add insert_temp, shrink savemode.
> (scan_loop): Accept invariants that require copying; mark as
> insert_temp.
> (combine_movables): Don't combine insert_temp movables.
> (move_movables): Insert copies for insert_temp movables.
> Don't record the info based on regno for insert_temp's.
>
> uncovered the need for moves between CCmodeish operands.
Oh, I see. So, it was a latent bug on mmix?
> I see other ports have such patterns, so I'll add one to mmix.md.
>
> Strange that it hasn't been needed until now.
Thanks for looking into it, I'll retry when your patches are in.
Andreas
--
Andreas Jaeger
SuSE Labs aj at suse dot de
private aj at arthur dot inka dot de
http://www.suse.de/~aj