This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH: Darwin / Mac OS X native support
Franz Sirl wrote:
>
> The rs6000 directory was just cleaned from old-style includes recently, so
> this should read
>
> > + tm_file="${tm_file} darwin.h rs6000/darwin.h"
>
> and the corresponding #include's removed from your header files.
Oh yeah - I did this for the host stuff, but forgot about the target
bits.
> config/darwin.h also seems to contain a lot of unnecessary #undef's, rs6000.h
> got cleaned up a lot from OS dependent defines since gcc-2.95, so you should
> be able to clean it up a bit. Feel free to suggest some macros for move from
> rs6000.h to the OS headers too, if you think they prevent cleanly structured
> darwin OS headers.
The #undefs are mostly paranoia; some of them are probably historical; the
original theory was that generic Dar^H^H^HNeXT definitions would be useful
for all architectures, but in practice many of them had to be overridden.
> On the other hand I don't think PREFERRED_RELOAD_CLASS belongs in an OS
> header...
Yes, that one is ugly, but forced by the workings of the Mach-O PIC
code, otherwise symbol loads try to use r0, with unfortunate consequences.
Nobody here knows of a more elegant solution.
> Overall the patch is looking good at a first glance, and unless Geoff and
> David have major concerns with it, I guess it can go into 3.0?
It would be very convenient to have it in 3.0, but the mainline would be
fine too. I haven't tested the patched sources on LinuxPPC, wouldn't want
to cause any regressions on the branch...
> The
> FINALIZE_PIC stuff maybe controversial, I believe this hook is deprecated and
> no new targets should use it, so most probably you will have to find a
> different solution...
I didn't know it was deprecated. Perhaps the manual should say something?
Stan