Hm, I found one problem with both my reimplementations, and this problem existed in the original implementation, too: there's a race at the point where the associated mutex is released and before the thread starts waiting for the trigger event. IIRC, POSIX promises this will be atomic. I'm not sure how to fix this, short of reverting to some sort of manual-reset model for the event.
Have to think about this a bit. -- Brane Äibej <[EMAIL PROTECTED]> http://www.xbc.nu/brane/
