Two-part patch for hppa64-hpux libstdc++ testsuite failures

[Mark Mitchell]

>This doesn't happen on the mainline; the relevant change is one small
>piece of a very large patch by Matt Austern which was applied to the
>mainline on March 12.  So the gcc part of this patch extracts that fix
>and backports it to the branch.
>
>  
>
How nice.

>The net result in the libstdc++ testsuite for PA64:
>
>-FAIL: 21_strings/basic_string/inserters_extractors/pod/10081-in.cc
>-FAIL: 21_strings/basic_string/inserters_extractors/pod/10081-out.cc
>-FAIL: 22_locale/numpunct/members/pod/1.cc
>-FAIL: 22_locale/numpunct/members/pod/2.cc
>-FAIL: 27_io/basic_istream/sentry/pod/1.cc
>-FAIL: 27_io/basic_ostream/sentry/pod/1.cc
>
>Bootstrapped i686-linux and hppa64-hpux11.11.  OK 3.4 branch?  The
>libstdc++ tweaks should go on mainline too.  (N.B. We can't use := for
>the Makefile variables that expand to $(shell ...) constructs because
>then they get run too early.)
>  
>
Yes, this is OK for 3.4, and the testsuite changes are OK for mainline.

However, I'd be happy to hear to comments from Benjamin or other V3 
experts.  So, please go ahead and check this in, but be prepared to make 
changes if the V3 folks think the testsuite change isn't quite right for 
some reason.

Thanks,

-- 
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com