Re: [RFA] Resubmit reverse debugging [4/5]

[Pedro Alves <pedro@codesourcery.com>]

A Thursday 09 October 2008 02:18:08, Michael Snyder wrote:
> Sorry, it's an artifact of the fact that I've been on a
> fork for so long.  When I copied this code from finish_command,
> the code that I copied had a similar call to internal_error.
> 
> In fact, finish_command_continuation still does.

Yeah, the continuation has a check for `function != NULL',
though.

> In fact, it's the same call that used to be in "finish_command".
> 
> So what should it be?  Just "error"?

Ah, I think I see what's going on.  "finish" is not meaningful
in the outermost frame, so, you'd get an error before reaching
here, if you had no symbols.  

 (gdb) finish
 "finish" not meaningful in the outermost frame.
 (gdb) reverse-finish
 "finish" not meaningful in the outermost frame.

Is it possible to be at frame != #0 and not find a function?

Wait, what does your remark about the frame #0 special
case mean then?

(I have to admit I got confused again by what finish in reverse
means.  I'd personaly trade it easily for "begin", and I'd
do the reverse step to go to the caller myself.  Then, you
could do begin at frame #0 too...)

> I think I understand that you think it would be more "local"
> to put the error here -- but is it worth it if it makes us
> add complexity?
> 
> finish_command already tests a number of things, including
> whether we are async and (now) whether we are reverse, and
> contains a number of error calls already.

No biggie with me.  Just thought you had signed up to do
the function split.  ;-)

-- 
Pedro Alves