Re: GtkNotebook accelerators.

[Havoc Pennington <hp redhat com>]

Tim Janik <timj gtk org> writes: 
> who has this "our general idea"?

Owen and I discussed it.

> simply relying on the mapped state of
> a widget will not cut it, e.g. menus are for the most part unmapped
> and still you want their accelerators to take effect.

We're aware of that, that's why I said "with some special-casing to
handle menus"

The mapped state of a widget controls the underline accelerators, not
the Ctrl+S shortcuts.

> also that doesn't account for accelerator change negotiation, i.e.
> when accelerators are being loaded from a config file or changed
> on menu items.

Well yes, that has to be considered. But the basic deal is that on all
other toolkits (Swing, Qt, WinForms), you simply set the accelerator
on a widget, and set up a connection between the displaying label and
the accelerator, and things Just Work. i.e. accelerators are
automatically installed when the widget is visible, and there's no
need to manually manage accel groups. This is the behavior we should
also have.

> besides that, there are other yet unresolved issues wrt accelerators
> pending.

There are several issues, yes.
 
> we can't sufficiently resolve this for 2.0, there's simply too
> little time left for a complete revamp of the way accelerators
> work, and that's what we need long term.
> so this issue definitely will be punted to post-2.0, untill then,
> people can use different accel groups for notebook pages (we migth
> want to introduce convenience functions to a) block whole accel
> groups, and b) couple an accel groups blocked state to notebook
> page visibility, but that's about it).

Well, there's still the issue of plug/socket which also requires accel
changes; at that point a revamp may be almost as easy as hacking
around things in plug/socket and notebook. The thoughts we had on how
to revamp things don't make it a huge project, since it should be kept
pretty simple, and we'd keep accel group around as implementation
detail and/or power-user feature, but it would still take a week or so
to finish up presumably.

I do tend to agree that this should not be done in 2.0, I was just
telling Alex what we discussed the other day.

Havoc