This is the mail archive of the
mailing list for the GCC project.
Re: AIX bootstrap failure
- To: David Edelsohn <dje at watson dot ibm dot com>, kenner at vlsi1 dot ultra dot nyu dot edu
- Subject: Re: AIX bootstrap failure
- From: Geoff Keating <geoffk at geoffk dot org>
- Date: 11 Nov 2001 00:12:20 -0800
- CC: gcc at gcc dot gnu dot org
- References: <200111110543.AAA27580@makai.watson.ibm.com>
- Reply-to: Geoff Keating <geoffk at redhat dot com>
David Edelsohn <email@example.com> writes:
> >>>>> Richard Kenner writes:
> Richard> So I think this has to be debugged "from scratch".
> Richard> What, precisely, is going wrong?
> What do you want as an answer? I posted the internal error
> message. stage1 is miscompiling stage2 which blows up when compiling
> unwind-dw2.c. I don't know exactly *how* stage1 is miscompiling stage2.
Did someone say this was CONST_DOUBLE related?
I remember thinking (when my attention was drawn to those changes by
their breaking the x86 bootstrap) that they'd also be trouble on rs6000,
but couldn't remember where. This discussion finally tickled my memory.
Richard, did you look at rs6000_hash_constant and toc_hash_eq in rs6000.c?
I think they might need updating. For instance, there's the comment
/* Gotcha: One of these const_doubles will be in memory.
The other may be on the constant-pool chain.
So rtx_equal_p will think they are different... */
and I'm not sure all the related code is still correct. This could
easily cause a hard-to-find bootstrap problem.
- Geoffrey Keating <firstname.lastname@example.org> <email@example.com>