This is the mail archive of the
mailing list for the GCC project.
RE: Defining __powerpc__ in rs6000.h
- To: "'Fergus Henderson'" <fjh at cs dot mu dot oz dot au>, David Edelsohn <dje at watson dot ibm dot com>
- Subject: RE: Defining __powerpc__ in rs6000.h
- From: "Boehm, Hans" <hans_boehm at hp dot com>
- Date: Mon, 21 May 2001 09:56:52 -0700
- Cc: Franz Sirl <Franz dot Sirl-kernel at lauterbach dot com>, boehm at acm dot org, "'gcc at gcc dot gnu dot org'" <gcc at gcc dot gnu dot org>
I just changed gcconfig.h in my tree to test for either powerpc or
__powerpc__ under Linux. I'm happy to remove the first test if someone
tells me that it's never needed.
I second Fergus' statement: The garbage collector can't entirely play by
ANSI C rules. However, I would like to remove spurious problems such as
> -----Original Message-----
> From: Fergus Henderson [mailto:firstname.lastname@example.org]
> On 20-May-2001, David Edelsohn <email@example.com> wrote:
> > It sounds like the Boehm GC is not as pedantically ISO
> C clean (or
> > whatever the relevant standard is). It should be using
> __powerpc__. As
> > Boehm is cooperating with GCC, we should work with the developers to
> > correct this.
> I agree that it would be good to change the Boehm GC to use
> and the like, so that it works correctly when compiled by GCC
> in strict
> ISO C mode. However, I think GCC should continue to define `powerpc'
> when `-ansi' is not specified, at least for a considerable transition
> period, so that old versions of the Boehm GC and other software which
> depends on this will continue to compile correctly.
> Note that the Boehm GC will never be strictly conforming ISO
> C89 / C99,
> even if the code is changed to use `__powerpc__' instead of `powerpc'.
> It inherently relies on the use of operations which have undefined
> behaviour in ISO C.