This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: continuing egcs-1.1 problems
- To: Fred Bacon <bacon at aerodyne dot com>
- Subject: Re: continuing egcs-1.1 problems
- From: David Edelsohn <dje at watson dot ibm dot com>
- Date: Tue, 10 Nov 1998 11:10:06 -0500
- Cc: Jeffrey A Law <law at cygnus dot com>, Franz Sirl <Franz dot Sirl-kernel at lauterbach dot com>, Elgin Lee <ehl at funghi dot com>, Amal Phadke <phadke at kewalo dot eng dot hawaii dot edu>, bacon at mediaone dot net, egcs at cygnus dot com
>>>>> Fred Bacon writes:
Fred> I'll rebuild my copy of egcs today with the first patch and without the
Fred> second. If the test code works, then I'm willing to remove the second
Fred> patch.
That is not a basis for including a patch in EGCS. We need a
definitive answer about the SVR4 PPC ABI and what is required as far as
alignment and trampolines. Just because making a change prevents a
failure does not mean that the change is the correct solution.
I am certain that removing the alignment part of the patch and
only reversing the trampoline test ordering will solve the symptom. I
thought that someone already did that. The question is whether the
trampoline test ordering truly is the source of the problem.
I thought that we previously received a confirmation a couple of
months ago that the alignment part of the patch violated the SVR4 PPC ABI
and was incorrect.
David