Re: How to Delay, Wait, Pause...

FROM : Jens Alfke
DATE : Sun May 25 18:53:13 2008

On 25 May '08, at 2:40 AM, Thomas Davie wrote:

> I hate to say this, but any form of delay here is the *wrong* way to 
> do this.  You already know you have a race condition, all you're 
> doing is making it so the race condition will work out in your favor 
> 99.99% of the time.  There are still exceptional cases where the OS 
> will be busy doing something and your 0.1 second sleep will not be 
> enough to sort it out.  Instead of doing this, do your threading 
> properly, and get your locking right.


This is correct advice when it comes to interactions between multiple 
threads; but Steve's problem involves only a single thread. The timing 
here is measured in iterations of the thread's runloop, not in clock 
time. As Dmitri put it:

> From 10.4 on fetch: results are delayed until the next iteration of 
> the run loop


So I believe that any technique that cranks the runloop through at 
least one iteration will solve the problem, and that includes any sort 
of timer or perform-after-delay. (I often use performAfterDelay: 0.0, 
though it looks nonsensical on first glance, to delay some action 
until the very next runloop iteration.)

—Jens