running 4.4 libstdc++ testsuite using 4.5's libstdc++ shared library

[Jonathan Wakely]

I'll just repeat that these changes weren't to experimental features,
they were to utterly non-functional stubs which have gone from
unusable to working.

Can anyone point to a single program in the wild which uses
std::tr1::regex/std::regex or std::condition_variable_any and which
was built with GCC 4.4?  How many packages in the Debian repo were
affected by this issue?  I suspect it's exactly zero.

Users would have to use -std=c++0x for condition_variable_any, and for
that or regex they would have to be pretty stupid to think they could
use what amounted to stubs.

Those stubs probably shouldn't have been there until we were in a
position to provide implementations, but that doesn't change the fact
that anyone who introduced a dependency on std::condition_variable_any
when it didn't work was pretty stupid.

Bumping the soname for these changes would have been a big problem for
users who don't use -std=c++0x or care about it yet. I believe the
vast majority of libstdc++ users would rather have a stable soname
than *any* c++0x features.

I'd prefer to see us ship two libs, so the C++0x stuff can evolve
independently. I have no idea how hard it would be to build both
libstdc++.so.6 and libstdc++.so.7 at the same time, but it would seem
to solve a lot of problems.