Re: extended WM_TRANSIENT_FOR (was Re: _NET: Disabling shading)

[Havoc Pennington <hp redhat com>]

On Wed, 2003-10-01 at 12:37, Lubos Lunak wrote:
>  No, I was thinking about that one too. The window_group definition is 
> somewhat vague, but currently it seems to be set to the same value in all 
> windows of one app (Gtk does that and I submitted a patch for Qt to do it as 
> well).

Argh, no. ;-) That is probably a bug in GTK. ;-)

The correct behavior (IMHO, but we've discussed it here in the past) is
that the group refers to a single main window plus all associated
windows.

Though I'm not sure this idea of a group is all that useful, on the
other hand having the group identify the X client connection seems
pretty non-useful also. Perhaps we should simply extend WM_TRANSIENT_FOR
as you say, define group as implying something about transient
relationships, and then deprecate the group hint. e.g. the group could
mean that all not-marked-transient windows with type != NORMAL are
considered transient for all windows with type = NORMAL. Or something
like that.

Don't know. Really have to clarify the whole transiency/modality/group
thing as a whole.

Havoc