Hi,

This is a modified version of Paul's patch for the single stepping over
atomic instruction sequences.

The main modification is the removal of the portion of code that was
previously added to ppc-linux-tdep.* files. As some people on the
mailing list noticed, it was not really needed. The contents of that
file were transferred to the rs6000-tdep.c file. 

rs6000-tdep.c now handles all the functionality for the single stepping
over locking instruction sets, and a suggested modification was included
in the code in order to fix possible internal errors due to inconsistent
handling of breakpoint creation/removal.

Follows the testsuite runs with and without the patch for HEAD.

Testsuite Summary with the patch applied

# of expected passes            11071
# of unexpected failures        356
# of expected failures          42
# of known failures             41
# of unresolved testcases       3
# of untested testcases         11
# of unsupported tests          18

Testsuite Summary without the patch

# of expected passes            11071
# of unexpected failures        357
# of expected failures          42
# of known failures             41
# of unresolved testcases       3
# of untested testcases         10
# of unsupported tests          18

It appears to have no regressions.

I am aware that Emi is currently working on a similar patch, based on
Paul's also. But since this is an issue that's been rolling for long and
that negatively impacts on more specific debugging activities, wouldn't
it be possible to include this patch as a fix while we wait for Emi's
copyright assignment to be ready? After his copyright assignment is
ready, it would be possible for him to incorporate additional changes
that he wrote on top of this one.

Regards,
Luis