On Saturday 07 February 2009 00:11:27 Tom Tromey wrote:
> >>>>> "Vladimir" == Vladimir Prus <vladimir@codesourcery.com> writes:
> 
> Vladimir> This patch fixes this by making MI observer print frame
> Vladimir> again, into MI uiout, if necessary. It passes all MI tests
> Vladimir> in (sync,async)x(native,gdbserver) combinations.  How does
> Vladimir> this look, and are non-MI changes OK? If approved, I'll add
> Vladimir> a test that CLI commands result in proper *stopped.
> 
> I don't understand all the implications of the core change, but I do
> like how it moves some MI logic out of the core and into MI.
> 
> Vladimir> -@deftypefun void normal_stop (struct bpstats *@var{bs})
> Vladimir> +@deftypefun void normal_stop (struct bpstats *@var{bs}, int @var{print_frame})
> Vladimir>  The inferior has stopped for real.
> Vladimir>  @end deftypefun
>  
> I would like to ask for documentation describing the meaning of the
> new argument.

The attached patch does so. Eli, can you see if the doc change is fine?

I've also added testcase, as promised. One detail I forgot to say in the
original email is that this patch does not make *stopped emitted for CLI
identical to *stopped for the corresponding MI commands. In particular,
output for CLI does not include 'reason' field sometimes, and for 'finish'
it will not include the return value. I think this is fine -- CLI commands
are meant to be issued by the user in the console, and therefore user will
see the CLI output from such commands. The only purpose of MI *stopped is
to let MI frontend know that the program is not stopped in a different place.

- Volodya