compiz team mailing list archive

[sam.spilsbury@xxxxxxxxxxxxx]

On Sat, 19 Nov 2011, Daniel van Vugt wrote:

> Can someone please point me to the exact code that was meant to fix this
> bug?
>
> If the fix did indeed make it into oneiric, then it still needs to be
> improved because it's still happening. Bug 806255 is evidence of this.

Hi,

I believe 
http://bazaar.launchpad.net/~compiz-team/compiz-core/oneiric/view/head:/src/event.cpp#L1023

is probably what you're looking for. That code ungrabs the keyboard if a 
plugin doesn't explcitly grab the keyboard (eg, using screen->pushGrab ()) 
while a key event is being handled and the action is dispatched. This 
means that if you held down the super key, the grab would persist from the 
time of the super key being pressed, to when the plugin handles the 
action.

Unfortunately that's not a perfect solution to this problem, and I haven't 
been able to find one. X Grabs are inherently racy, and when we have a 
keybinding set *only* on the super key, doing a passive grab on the super 
key also implicitly does a passive grab on every other modifier 
combination that uses the super key, and that grab remains until it is 
released. This means that if you were to hit super-space to trigger 
gnome-do really really quickly before compiz had a chance to dispatch the 
action and release the grab then that the other application wouldn't get 
the KeyPress event.

The grab is required in order to get the notification that the KeyPress 
happened on the super key (cannot register for KeyPress events on windows 
we don't own).

In terms of your other solution by not doing anything if another 
application has a grab, I don't think that will solve this problem. First 
of all, the key-action requires the passive grab to work (which is what's 
causing the problem here) and second of all, even if that were the case, 
there's still a race window where the application registers for the grab 
and its output buffer hasn't been flushed yet, so you'd get the dash + 
(whatever). Or worse, there would be unpredictable behaviour depending on 
which application gets there first.

I think that the best solution to this problem is to have a flag in the 
metadata which says "I won't require an active grab", so then you have a 
reduced amount of time for which the race can happen (since we can release 
the grab the moment the action gets dispatched). It might also be safe to 
release the grab in any event, but then there's the consequence of 
grabbing twice and dealing with things like, eg, focus changes that happen 
as a result.

> --
> You received this bug notification because you are a member of compiz
> packagers, which is subscribed to compiz in Ubuntu.
> https://bugs.launchpad.net/bugs/704231
>
> Title:
>  Unity steals <super> modifier key
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/unity/+bug/704231/+subscriptions
>
> _______________________________________________
> Mailing list: https://launchpad.net/~compiz
> Post to     : compiz@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~compiz
> More help   : https://help.launchpad.net/ListHelp