[Openmcl-devel] Semaphore troubles

[Ron Garret]

On May 9, 2012, at 11:10 PM, Gary Byers wrote:

> As I said, I think that it's surprising that you see the problem when one
> thread sleeps for 10 seconds, but it occurred to me that if I could reproduce
> that I might be able to poke around during that 10-second interval and see
> what was going on.  I tried running your code (the first version you sent
> today) and couldn't get it to fail at all (several thousand dots and no
> assertion failures.)

I was able to get it to fail, but only by changing the sleep time to 0.01 seconds, and even then it took a few hundred iterations.  At 0.02 seconds (or more) I could not get it to fail at all.

The fact that James sees MORE frequent failures with LONGER delays is exceeding strange.  I'll give very long odds that he is not accurately reporting his experimental conditions.

rg