[Xquartz-dev] [XQuartz] #241: xquarz uses 100% of CPU when not focused

[Jeremy Huddleston]

On Apr 12, 2009, at 17:56, Christof Wolf wrote:

> On Sunday 12 April 2009, Jeremy Huddleston wrote:
>> Ok,  I was able to reproduce it and fix it!  YAY!  So would you mind
>> trying 1.4.2-apple42 via MacPorts or download the X11.bin replacement
>> from http://xquartz.macosforge.org/downloads and verify that it's
>> fixed for you as well.
> Jeremy,
>
> I could not compile apple42 via MacPorts - but I downloaded it - and  
> it works
> fine!! no CPU issue - what caused it anyway? - if that is not too
> complicated :-)

Thanks for testing.

Could you tell me what error you're encountering with Macports  
building it?

It was a combination of two errors:

We were using the wrong values for ProximityIn and ProximityOut when  
generating the proximity events (pen going in or leaving the tablet  
area).  This resulted in num_events being 0 and we didn't queue up  
those events.   After enqueuing the events, we put a byte in a pipe  
that the server thread is select()ing on to wake it up (yes, hacky.).   
The server thread reads this byte out of that pipe after it dequeues  
the event.  Unfortunately, since there actually wasn't an event in the  
queue, it didn't get around to clearing the pipe and the select()  
turned into effectively a no-op in the while loop.

You didn't see this bad behavior when X11 was in the foreground  
because we were sending MotionNotify events as well (thus there was a  
"real" event in the queue).

So I fixed these two errors by checking num_events before the poke and  
making sure that we properly generate the proximity events.