Debbuger

[Hamish Rodda]

On Wednesday 04 February 2009 04:32:33 Andreas Pakulat wrote:
> On 03.02.09 20:06:20, Vladimir Prus wrote:
> > On Tuesday 03 February 2009 19:42:26 David nolden wrote:
> > > Am Dienstag 03 Februar 2009 07:48:48 schrieb Vladimir Prus:
> > > > Oh, competition! That's always good :-)
> > > >
> > > > - Volodya
> > >
> > > Would have been more useful if you actually said something about
> > > Hamish's points. ;-)
> > >
> > > What do you think about the thing? I think Hamish's points are valid.
> > > We're coming closer to a release, and we do need a working debugger. If
> > > we had one that was working a year ago, then it's worth considering
> > > going back there.
> >
> > Well, it is worth pointing out that in HEAD, to the best of my knowledge
> > we have working thread/frame view, breakpoint widget, watches, and
> > variable tooltips. We don't have display of local variables, and we don't
> > have breakpoint markers in source. So, the most straightforward way to
> > get a working debugger is to fix the missing thing. Branching from a Qt4
> > port of KDevelop 3 debugger which is known not to be very great seems
> > like an equally complex way to get something worse.

Issues with head:
1) No toggle on breakpoints in the editor - you have to type in the location.  
I tried to fix this but the new breakpoint framework is complex and seemingly 
unable to accept a programmatic insertion of a breakpoint. IIRC I asked for 
help with fixing it, and was told it would be worked on, but that was last 
year.

2) No local variables display .
The variables display that is there appeared to be completely new code, 
throwing out everything that was done for 3.x; why was this?
I also tried to fix this here, but because the variables code was new it was 
hard for me to see how to port the frameitem and threaditem code.

3) No meaningful commit since the hackathon.

Whereas with the branch, which took me only two hours, I now have everything 
back to a working state, including QString display.

> Hmm, so it seems fixing the current plugin up shouldn't be too much work.
> Maybe I can do a little exercise this weekend, as I was planning to fix
> that up anyway (just couldn't decide on wether I should start with this or
> designer) :)

If I thought that was the case I would have done that instead.

Don't get me wrong, I don't want maintainership of the debugger part, and I 
value everything that Vladimir is/has done for it including the improvements 
to gdb; but I think we need to find an interim solution for 4.0 or have 
someone step up to fix the problems with the head version.  If what I have 
done spurs someone on to fix head, then I'll have achieved my aims.

Cheers,
Hamish.