JG> This is NOT correct.

This is a confrontational way of starting a reply,  that puts the receiver
in a sour mood right away (I do it on purpose all the time :-).


 JG> int     thread_finished;

 JG> In this case, access to the thread_finished variable 
 JG> does NOT have to be protected by a semaphore because 
 JG> both reading and writing integers are atomic 
 JG> operations. Read-modify-write is NOT an atomic

Under uni-processor OS/2,  for sure.  But what about OS/2 SMP?


--- Maximus/2 2.01