Bug 68015: Variables/Watch Window closes on first view

[Vladimir Prus]

On Saturday 21 May 2005 03:48, jbb wrote:
> [snip]
>
> > > > I think this can be implemented by making variables and framestack
> > > > widget emit some other (new) signals, not 'slotSelectFrame'.
> > > > Currently, slotSelectFrame does many jobs -- it's switches the frame
> > > > for showing the frame in source view, it produces backtrace, and it
> > > > produces info about local vars. If more specific 'slotShowBacktrace'
> > > > is introduced, we can avoid accidental switching to different frame
> > > > in source view.
> > >
> > > Does anybody disagreed that this should be done?
> >
> > No, was it proposed previously.
>
> Actually, I remember now why I did it this way. Asking gdb for the local
> vars can take a long time. Once they've been asked for, I wanted to hold
> onto them, henece the tree view. So if you switch between frames and then
> back again the appropriate branch was expanded and they were not ask for
> again. At least that was the intention a few years ago now.
>
> Showing only the current frame in the variable list is better in that the
> interface is "cleaner" for the user. We should still cache the vars though
> and redisplay the cached version if they redisplay a frame.

Understand. Caching the vars would be indeed desirable, but luckily caching
independent from the interface. As soon as I get feedback on my current patch,
I'll see about implementing caching for vars and then making varview widget 
a simple list without frames.

- Volodya