On Mon, 2007-04-23 at 17:00 +0100, Mike Dransfield wrote: > David Reveman wrote: > > On Mon, 2007-04-23 at 16:14 +0100, Mike Dransfield wrote: > > > >> David Reveman wrote: > >> > >>> On Mon, 2007-04-23 at 14:28 +0100, Mike Dransfield wrote: > >>> > >>> > >>>> While working on the python plugin, I realised that the > >>>> timeout system does not allow for notification when a timeout > >>>> is removed. This has been fine up until now because he closures > >>>> never needed to be cleaned up. > >>>> > >>>> In the python plugin, I pass through a custom structure which > >>>> needs to be cleaned up if the timeout is removed. This is almost > >>>> certainly the cause of some horrible memory leaks for me. > >>>> > >>>> The attached patch is a simple solution to the problem. I think > >>>> it should be good enough for most situations. > >>>> > >>>> I could maintain some sort of lookup inside my plugin, but I think > >>>> other plugins would benefit from this. > >>>> > >>>> It is OK to go in or were there any comments? > >>>> > >>>> > >>> Timeouts are only removed as a result of the timeout call-back function > >>> returning FALSE. So unless I'm missing something, all cleanup can be > >>> done in the regular call-back function before returning FALSE. > >>> > >>> > >>> > >> The pyplane (and the original plane) use a timeout which gets > >> canceled if the user quickly switches back to another viewport. > >> It is not a problem for plane because the closure is a CompScreen > >> which does not need to be cleaned up. > >> > >> The plugins call compRemoveTimeout to stop it. I'm not sure if this > >> is right in this case, but it seems that it would be useful to stop a > >> timeout before it expires, and it makes it easy to clean up the closure > >> without having to maintain a lookup table of handles to closures. > >> > > > > OK, I see. > > > > 1. compRemoveTimeout can return the closure pointer. > > > > 2. We add a function like this: > > > > void * > > compTimeoutGetUserData (CompTimeoutHandle handle); > > > > 3. We add a remove call-back function (your patch). > > > > > > I'm in favor of solution 1, it's simple and efficient. What do you > > think? > > > > Yes, number 1 sounds like a better solution. > > Ill prepare some patches for that. Should I also increase the > ABIVERSION when I do this?
Great. I don't think that change can cause any ABI breakage but you can bump the version if you want. - David _______________________________________________ compiz mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/compiz
