I have seen one report of the "new" thread library not honoring an unlock causing a deadlock. We actually traced it and saw an call to unlock yet all subsequent calls to lock failed, because it wasn't released.
2 main points: this was with the Sol9 thread libs on Sol8 (as suggested by Sun), and (2) this was seen actually in the SSL mutex stuff, not in the accept mutex loop.
Humm...
I have seen a problem with the SSL mutex stuff segfaulting on shutdown on Windows. I've not had time to debug it but a mutex_unlock in an SSL callback (from SSLeay) fails (bad handle) then the segfault happens when logging the message because the mutex in the apr_file_t has been whacked. May be a pool cleanup issue...
Bill
