Re: [Usability] Re: Error dialogs (among others?) and focus stealing prevention

[Dan Winship <danw novell com>]

On Wed, 2004-10-27 at 01:28 -0400, Havoc Pennington wrote:
> We're kind of in gtk-devel territory rather than usability@ territory
> here, but I guess my question is: how many things would get a USER_TIME
> of 0 if we said the timestamp on a window comes from the *current* event
> instead of the *most recent* event.
> 
> i.e. how often do we open a window asynchronously from the event that
> causes it to open, for example if you queued an idle to open the window,
> or something of that nature.

One thing this wouldn't work for would be a "could not load page" error
in a web browser. The decision to pop up the error would happen from a
network I/O handler, far far away from the click or keypress handler
that resulted in the original HTTP request being made.

But in this case, calling "gtk_window_set_focus_on_map (w, FALSE)" would
be wrong. You need to set the focus time of the window to the timestamp
of that click/keypress event, because if the user is just sitting and
waiting for the web page to load, then the dialog should get the focus.

Very few windows are actually completely "out of the blue".

-- Dan