On Sun, May 08, 2005 at 08:22:24AM +0200, Jens M Andreasen wrote: > If so, then pthread_cancel() should work just fine together with the > workaround described under BUGS in man pthread_cancel.
Using pthread_cancel() is exactly what I want to avoid at all cost. In my world, threads are both the thing managaed by the system + threads library, and a C++ object. If the thread is cancelled that can make it virtually impossible to cleanup the object in an orderly fashion. It may for example have to wait for a message confirming that others that depend on its existence know it's going to disappear, and are prepared for that. Also, the pthread itself is just a private implementation detail of the object and can not be accessed from the outside without breaking the abstraction. -- FA
