This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH] libffi support for CRIS
If we're refering to the patch in
as I think we are, then I'd prefer that gcconfig.h just leaves the
existing definition of CRIS in place, with a comment to that effect,
and tests CRIS later. The current convention doesn't use
architecture_os macros; it uses separate macros for each, i.e.
CRIS and LINUX in this case.
On Thu, 28 Oct 2004, Hans-Peter Nilsson wrote:
> On Wed, 27 Oct 2004, Hans Boehm wrote:
> > On Wed, 27 Oct 2004, Hans-Peter Nilsson wrote:
> > > >
> > > > > Same comment applies as to boehm-gc, CRIS32 is a bit confusing.
> > > > > I'd rather just use the predefined __CRIS__ and __linux__ (eh,
> > > > > make that __gnu_linux__ ;-) and not define anything in
> > > > > configury. Or maybe define CRIS_LINUX.
> > > > I choose CRIS_LINUX.
> > > >
> > If I understand ths correctly, please don't do this in the context of
> > the collector. It currently consistently defines its own macros for
> > each architecture in order to hide inconsistencies in macros predefined
> > by various compilers etc. These macros don't have underscores in them,
> > in that they are private to the collector implementation.
> > I would prefer CRIS to CRIS32, especially since I suspect there is no
> > CRIS64.
> CRIS is already defined (unless -iso -anso or -younameit-standard)
> in addition to __CRIS__.
> > (And if there is a CRIS16, it will never be supported by the
> > collector.) But in my view, this is minor.
> There's no CRIS16 and no plans for it. :-)
> > LINUX should of course also remain defined.
> It's not touched. Maybe there's a misunderstanding here?
> So, judging from your comments, defining CRIS_LINUX in boehm-gc
> configury to identify the port is fine by you?
> And you too prefer it to CRIS32? Then we're all set!
> brgds, H-P