Re: [Predef-contribute] Predefined macro updates and fixes

[Guillem J. <gu...@ha...>]

On Sun, 2012-12-30 at 16:42:08 +0100, Bjorn Reese wrote:
> Thanks for your detailed reply. Most of the information has been added
> now. More replies below.

Thanks!

> On 2012-12-10 12:24, Guillem Jover wrote:
> >On Fri, 2012-11-23 at 19:25:45 +0100, Guillem Jover wrote:
> >>For GNU/kFreeBSD both __FreeBSD_kernel__&&  __GLIBC__ should be
> >>checked, because FreeBSD has started defining __FreeBSD_kernel__
> >>(since FreeBSD 8.3, FreeBSD 9.1 and FreeBSD 10.0 on those branches).
> 
> I have added this to the FreeBSD and GNU/kFreeBSD entries.

It seems the __FreeBSD_kernel__ variable was typoed in the wiki, it's
missing the two trailing underscores.

> Just out of curiosity, do you know why it doesn't define something like
> a __gnu_freebsd__ macro?

My guess is that these __gnu_*__ macros are not widely used at all? I
think the first time I've seen them mentioned anywhere has been from
the predef documentation and then after checking into gcc source code.

> >>For GNU/Hurd, __gnu_hurd__ is also defined, and because it's currently
> >>based on the GNU Mach kernel, so is __MACH__.
> 
> I have added the __gnu_hurd__ macro. I assume that this is sufficient to
> identify GNU/Hurd.

Sure, although __GNU__ is the customary macro being checked, I just
mentioned it for completelness. Would be nice to add a reference to
__MACH__ too, so that people is aware that is defined there as well.

> >>The new x32 architecture (or different ABI for AMD64, depending how
> >>one looks at it), can be distinguished with __amd64__&&  __ILP32__.
> 
> This has been tentatively captured on:
> 
>   http://sourceforge.net/p/predef/wiki/DataModels/

Ah right, but I'm not sure if that might be clear enough when looking
for an x32 reference, though? As not all current architectures will
have such data models readily available. Maybe a entry could still be
added to the Architectures page, with references to the DataModels
page, instead of detailing that out there?

Thanks,
Guillem