On Tuesday 23 March 2004 7:13 pm, John Levon wrote: > On Tue, Mar 23, 2004 at 07:06:30PM +0000, Angus Leeming wrote: > > > Maybe I misunderstood what you wanted to change in Qt. A "busy > > > loop" would be one that chewed CPU despite no events arriving. > > > > This is what seems to happen in the Qt-equivalent. We get a > > never-ending stream of calls to 'someNonTrivialFunction()' when > > everything else is totally quiet. > > Then we definitely want to continue using a timer for this. > > > Strategy 1 is to modify the main GUI loop, as above. > > > > Strategy 2 is to use a Timeout as we do currently, but use it > > only to post the callbacks of these 'reaped_children' rather > > than, as now, using it to both reap the zombies AND post the > > callbacks. > > I prefer the latter.
Me too, if only because hacking into the Qt event loop turns out to be a right PITA. Nonetheless, playing devil's advocate, why is Strategy 2 better than the existing strategy of using a Timeout to reap the children also? (Genuine question.) > > Basically, I'm at a bit of a loss. The reaping-children bit > > appears to be optimized as the handler is invoked only when a > > SIGCHLD event is emitted. It seems to be a real pity that we > > can't be similarly efficient when posting the callbacks. > > Well, if Qt had support for reaping children... You're the Qt guru here. If you tell me it doesn't then it doesn't ;-) > What routines need to know about children that have died? 'I'd rather say 'Completed' than 'Died'. 'Died' is so emotive ;-) The graphics loading code needs to know that an image file has been converted to a loadable format so that it can begin the process of loading the thing. > regards > john Angus
