gdb-mi current status

[Vladimir Prus]

On Thursday 11 May 2006 15:41, Vladimir Prus wrote:
> On Wednesday 10 May 2006 15:44, Hamish Rodda wrote:
>
> Hi Hamish,
>
> > From my experience with the -mi branch, three things currently stand out
> > as important issues:
> >
> > 1) User interface delay, and more importantly, apparent forgetting of
> > requests issued by the user during debugger busy periods.
> >
> > I find that the delay from having to re-issue step commands (and having
> > to realise that the command was lost) far outweighs the delay of the
> > debugger actually doing processing :(
>
> I must admit I'm not yet sure what the problem is. In my testing,
> repeatedly pressing F10 steps the debugger just fine, and pretty quickly
> too. The step action gets disabled when command is sent to gdb, but then
> becomes enabled again and handles next press of F10. I'm not sure if I'm
> using wrong testcases, or there some other differences.
>
> I can cause different kind of problem -- when pressing F10 too fast,
> KDevelop can try to create variable object for a variable that's no longer
> in scope, getting error. I'll try to look into that.

Well, since revision 539660, as soon as you emit "step" command, all "state 
reloading" command (like "list stack frames") are removed from the queue. Now 
I can step though the code much faster than I can understand what's going on, 
and don't get any errors.

Can you please try this version and let me know if it works for you?

- Volodya