[RFC] Quad-float support in Fortran

[Tobias Burnus]

On October 27, 2010 23:47, FX wrote:
> The attached patch is the current state of my quad-float patch for gfortran.

Changes relative to FX's patch:
- libquad is now libquadmath and a separate libarary under the gcc top 
directory (including a LGPL copyright file)
- The symbols are now versioned
- libgfortran uses weak symbols (attribute weakref); however, it does 
not seem to work: I still get a dependence on libquadmath. Any idea why?

I was thinking of using something like:
       LIBQUADSPEC="%{static-libquadmath:--as-needed} -lquadmath 
%{static-libquadmath:--no-as-needed}"
However, this depends on HAVE_LD_AS_NEEDED. The check is done in 
gcc/configure.ac - and depends on variables being set regarding "ld". 
Any idea how to best handle this? An ugly brute force method would be to 
simply read ./gcc's config.h the other brute-force method is to copy and 
adapt a large chunk of ./gcc/configure.ac. Ideas?

Do you have other comments regarding the patch?

Tobias

PS: Given that there are many gfortran users, which do compile GCC 
themselves but do not use --enable-maintainer-mode, I have separately 
included the generated files.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: quad.diff.bz2
Type: application/x-bzip
Size: 190445 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20101106/11a5f3c7/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: quad_gen.diff.bz2
Type: application/x-bzip
Size: 98576 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20101106/11a5f3c7/attachment-0001.bin>