[GCC-XML] fails to parse __float128 on linux 32-bits, OK in 64-bits

[Steve M. Robbins]

Brad,

Thanks for the note.  I don't think I described my problem clearly
enough.

I'm building ITK, compiled to use fftw.  The newest fftw3.h header
contains code similar to the following

On Mon, Dec 19, 2011 at 11:05:11AM -0500, Brad King wrote:
> On 12/18/2011 5:58 PM, Steve M. Robbins wrote:
> >	#if (__GNUC__>  4 || (__GNUC__ == 4&&  __GNUC_MINOR__>= 6)) \
> >	&&  (defined(__i386__) || defined(__x86_64__) || defined(__ia64__))
> >	    typedef __float128 quadcomplex[2];
> >	#endif

... that surrounds code using __float128.


> Treat gccxml like any other compiler that pretends to be __GNUC__
> (such as Intel) and test for it with __GCCXML__ and these macros that
> tell you the internal GCC parser version:
> 
>  -D__GCCXML_GNUC__=4
>  -D__GCCXML_GNUC_MINOR__=2
>  -D__GCCXML_GNUC_PATCHLEVEL__=1

I think you are suggesting to add gccxml-specific code to the
problematic header.  Unfortunately, the header is fftw3.h which I do
not have the ability to change.


> Unfortunately gccxml will not be able to dump an xml interface description
> that includes __float128.

OK.  So if I cannot get anything sensible, I may as well hack around
this by using "gccxml -D__float128=double ...".

I am curious, however, why the 64-bit version of gccxml does not seem to
have a problem, while 32-bits does.  Any ideas?

Thanks,
-Steve

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 190 bytes
Desc: Digital signature
URL: <http://public.kitware.com/pipermail/gccxml/attachments/20111220/027c6454/attachment.sig>