Re: [GD-General] Two issues: __attribute__((cdecl)) and MSVC++ Toolkit 2003

[Ignacio <cas...@ya...>]

On Wednesday 07 July 2004 20:05, Brian Hook wrote:
> On Wed, 7 Jul 2004 22:14:09 -0400, Brian Hook wrote:
> > Here it is enumerated.  The prefix form of __attribute__((cdecl)):
> >
> > gcc 2.95, Linux/x86 -- YES
> > gcc 3.3.4, Linux/A64 -- NO
> > gcc 3.3, OS X/PowerPC -- NO
> > gcc 3.3.1, Cygwin/x86 -- YES
> >
> > No rhyme nor reason, other than possibly needing to be a pure x86
>
> ACtually, there is a rhyme -- it's supported on x86, and not supported
> on non-x86, where A64 is considered non-x86 for some reason.  This
> makes sense given that calling conventions are by and large an x86
> kind of thing (at least, these specific ones).
>
> So it's not a positioning thing, it's an availability thing.  Duh,
> shoulda thought of that a while ago.

Additionally I can also confirm that the followings also work:

gcc 3.2.3, Linux/x86
gcc 3.3.1, Linux/x86
gcc 3.4.0, Linux/x86

I don't have access to a non-x86 machine, but your conclusions make a lot o=
f=20
sense. To discard any othe possibility, you could also try with any other=20
attribute on a non-x86, like noreturn for instance.

=2D-=20
Ignacio Casta=F1o
cas...@ya...