>>
>> It looks right to me.  With a ChangeLog entry and testsuite results
>> from some target with threads, it would be OK to commit.
>>
>
I propose a new implementation that does not change the old behavior of 
gdb with respect to the printing of the last frame in case we changed of 
thread during "thread apply" command execution.

This implementation does not make any regression in the testsuite for a 
i386-linux target.
To me the part that prints the stack frame at the end of execution of 
"thread apply" could be entirely removed but this will imply some 
changes in the testsuite. I can do that it's up to you.

About the bug initially fixed by this patch, I include some new tests in 
threadapply.exp.

-- 
Denis