debug mode performance patch

[Jonathan Wakely]

On 21 November 2010 20:55, Paolo Carlini <paolo.carlini@oracle.com> wrote:
> .. first blush looks pretty good to me. But I see now an unconditional
> use of <mutex>, which isn't always available, see:
>
> ...
>
>  #if defined(_GLIBCXX_HAS_GTHREADS) && defined(_GLIBCXX_USE_C99_STDINT_TR1)
>
> ...
>
> Thus if we commit this kind of change to debug.cc the bootstrap breaks
> for some targets. Please figure out a solution not relying
> unconditionally on <mutex>.

I still don't see what's wrong with moving the body of swap into a
separate function, as suggested in
http://gcc.gnu.org/ml/libstdc++/2010-11/msg00063.html

That doesn't need any C++0x features, just what's available in
<ext/concurrence.h>