[Patch,Fortran+Build] Use more complicated quad-math config check

[Tobias Burnus]

On some targets, __float128 exists but the support is insufficient to 
build libquadmath because some functionality is not provided by libc (or 
should it be by libgcc?). Thus, the configure test for libquadmath 
failed - and libquadmath is not build. (That's OK.)

In libgfortran, there is a similar check, but seemingly it uses fewer 
features and thus does not fail on some targets, which cannot build 
libquadmath. Result: The build fails as libquadmath is not available.

The solution is to use the same check as for libquadmath. (Additionally, 
one should, e.g., provide the missing bits in libgcc to enable a working 
quadmath support. But that's a separate issue.)

There are two related PRs to this issue:
a) PR 47716 for NetBSD 5.1 (x86-64); there the patch has been confirmed 
to work; unfortunately, there seem to be more issues on that platform. 
(Lots of test suite failures, cf. PR.)
b) PR 47648 for FreeBSD on ia64. (FreeBSD on x86-64 is OK.)

Build and regtested on x86-64 linux (and build on NetBSD, cf. PR).
OK for the trunk?

Tobias

PS: The result of the configure check is as if libgfortran were build 
with --disable-libquadmath-support. The front end will still believe 
that __float128 [or rather: REAL(16)] is available.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: quad-check.diff
Type: text/x-patch
Size: 3873 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20110215/bdedee6d/attachment.bin>