This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [Bug 6358] Returning structs on Linux-i386
- To: Scott Snyder <snyder at d0sgibnl1 dot fnal dot gov>
- Subject: Re: [Bug 6358] Returning structs on Linux-i386
- From: Erik Corry <corry at arbat dot com>
- Date: Wed, 26 Apr 2000 13:08:39 +0200
- Cc: gcc at gcc dot gnu dot org, bugzilla at redhat dot com, Alex dot Buell at his dot spam dot proofed dot address
- References: <20000417134337.A1673513@daimi.au.dk> <yr8d7nm7jol.fsf@d0sgibnl1.fnal.gov>
On Wed, Apr 19, 2000 at 12:18:18AM -0500, Scott Snyder wrote:
>
> >I haven't seen any discussion of that change here so perhaps it
> >predates egcs.
>
> Well, there was
>
> http://gcc.gnu.org/ml/gcc/1998-11/msg01087.html
>
> but that didn't spark any discussion either...
I think the problem is that the fix is fine for other platforms, since
it brings gcc into line with the native compiler and the specification.
But on Linux, gcc _is_ the native compiler, and this change breaks
Linux binary compatibility. So it is reasonable enough for the egcs/gcc
people to allow this fix, but someone from the Linux community who cares
about binary compatibility should have made sure the behaviour didn't
change on Linux. Unfortunately such a person doesn't seem to exist.
When asked about it, the Red Hat people said they would defer to the
opinion of the egcs developers, and since they didn't have an opinion
(it's not their responsibility) nothing happened.
http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=2809
I note that when Alex Buell claimed on this list that egcs broke
binary compatibility for C programs he was generally disbelieved:
http://gcc.gnu.org/ml/gcc/1999-09n/msg00851.html
It seems to me he was right.
--
Erik Corry erik@arbat.com Ceterum censeo, Microsoftem esse delendam!