[PATCH] Optionally use -mlong-double-128 by default on {powerpc{,64},s390{,x},sparc,alpha}-linux (take 3)

[Benjamin Kosnik]

I'm going to drop cc's on this. It was getting ridiculous....

> There are actually 2 changes to testsuite/testsuite_abi.cc that would be
> desirable:
> 1) accept @@ -> @ symver changes (or perhaps better, @->@@ changes, from
> what is stored in regenerated abi *.txt files)
> 2) treat *_LDBL* symvers specially, in that we put abi *.txt files with
> the *_LDBL* symvers into SVN, but if the libstdc++.so being checked
> doesn't have any *_LDBL* symbol versions at all, they aren't reported
> as failure.

Here's what I've tested. It seems to work. I just decided to ignore
'@'  and '@@' in symbol bits, and compare on symbol names not "raw" ie
symbol name + version. I think this will be fine.

I would suggest this, and not checking in new baseline files with the
long double bits. At least until we can check these versioning bits
into an actual release. We'll have a lot of compatible symbols when
LDBL active, but no complexity when not.

-benjamin

-------------- next part --------------
A non-text attachment was scrubbed...
Name: p.20060207-1.bz2
Type: application/x-bzip2
Size: 10248 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20060207/68d4b201/attachment.bz2>