> This appears to make quite a lot of assumptions about
> glibc implementation details.  I'd except just about
> any glibc update to break this test ...
> 
> As this is a PowerPC-specific case anyway, can you not
> use a test that explicitly has the atomic sequence in
> the code provided with the test case?

Testcase updated, making use of explicit atomic sequences for 64-bit
binaries.

This testcase handles stepping through atomic sequences only. The
related problem with threads and atomic sequences will be handled in a
different testcase.

OK to commit?

Regards,
Luis