On Friday 14 March 2008 18:11:15 Daniel Jacobowitz wrote:
> On Fri, Mar 14, 2008 at 11:07:32AM -0400, Daniel Jacobowitz wrote:

> >        /* That switch put us at the top of the stack (leaf frame).  */
> >        switch_to_thread (tp->ptid);
> >        print_stack_frame (get_selected_frame (NULL), 0, LOCATION);
>
> I think Nick was right about printing the frame level.  It's not
> necessarily going to be zero in the future.  For instance, GDB might
> automatically select the nearest user frame if you stop when a C++
> exception is thrown, like we already do for Ada; and then we could
> teach GDB not to lose the selected frame whenever it switches threads.
> That wants doing anyway.

Ok, I've enabled printing of stack level.

> > > I have realized that while the patch is mostly about MI, it also
> > > touches generic code -- thread.c. Is that part of patch (attached again
> > > for convenience) OK?
> 
> Oh, one other thing.  I went back to look at Denis's patches.  He made
> -thread-info without an argument describe the current thread only.
> Doesn't that seem useful, especially before we get notifications
> completely hammered out?  I do not see any other MI command that
> reports the current thread.
> 
> So that would mean we needed -thread-list-all-threads back again.

Hmm, I'd rather not. Currently, the current thread is reported by 
*stopped, and is not supposed to randomly change -- because the
target is fully stopped while frontend talks with gdb. Therefore,
using -thread-info as a roundabout way to get the current thread
does not really give you anything.

For non-stop mode, I have long and evil plans that I'll post soon,
but they don't require this either.

I attach the patch revised per your comments. How does it look?

- Volodya