This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

gcc-4_1-branch backports of the long double changes


Hi!

I'm attaching backports of the long double patches to gcc-4_1-branch.

The first one (together with some form of the ppc32 -mlong-double-128
support patch, http://gcc.gnu.org/ml/gcc-patches/2006-01/msg01907.html
and followups) is something I'd really like to see in GCC 4.1,
it is something where it is quite obvious from the patches that
it will not affect the (so far default) -mlong-double-64 operation in
any way.  But in -mlong-double-128 mode both this one and the msg01907.html
patch affect a) mangling b) symbols exported from libgcc_s.so.1, so
it is IMHO desirable to get this codified already in GCC 4.1.
The rs6000 part doesn't have to look exactly as in this patch,
it depends on what David Edelsohn wants to do with AIX etc.
But the Linux -mlong-double-128 -mabi=ibmlongdouble behavior
should be the same as in the patch.

The second and third patches are for optionally enabling TFmode
long double as default.  The second one is for everything but
libstdc++, the last one is libstdc++ only.

The patches have been tested on {ppc,ppc64,s390,s390x}-linux,
in all cases with a --without-long-double-128 build + make check,
--with-long-double-128 build + make check and then for libstdc++
another 2 make check-target-libstdc++-v3 runs for each target
- in both cases with --without-long-double-128 configured compiler
- once tested linking against --without-long-double-128 libstdc++.so.6
and running against --with-long-double-128 libstdc++.so.6 and
once tested with both linking and running against --with-long-double-128
libstdc++.so.6.
On sparc-linux I have tested so far --enable-languages=c build
(double checking the exported ABI list), full bootstrap/regression
testing still in progress.  alpha-linux is unfortunately untested,
I don't have access to it.
On {ppc,ppc64,s390,s390x,sparc,sparcv9,sparc64}-linux also glibc
has been built and regression tested.

	Jakub

Attachment: gcc41-ldbl-mangle-as-g.patch
Description: Text document

Attachment: gcc41-ldbl-default.patch
Description: Text document

Attachment: gcc41-ldbl-default-libstdc++.patch
Description: Text document


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]