Last update: 23 February,1998
Eastwood #13 _____________________________________________________________________________ Topic: EDEADLCK Relevant Sections: pthread_rwlock_rdrlock Spec: XSH Issue 5 Resolution Request: ------------------- There appears to be a contradiction in pthread_rwlock_rdlock (Note Sanity draft page numbers quoted) Page 647 ll 21609-21610 states: (pthread_rwlock_rdlock())... "Results are undefined if the calling thread holds a write lock on rwlock at the time the call is made." It then states on ll 21632: The pthread_rwlock_rdlock() and ... functions may fail: ll 21635 [EDEADLCK] The current thread already owns the read-write lock for writing This would seem to be a contradiction, between saying its undefined and then defining a may fail situation. Similarly pthread_rwlock_wrlock() has a problem: page 651 ll 21717-21718 states: Results are undefined if the calling thread holds a read-write lock (whether a read or a write lock) at the time the call is made. And ll 21740 pthread_rwlock_wrlock may fail: [EDEADLCK] The current thread already owns the read-write lock for writing or reading. Resolution Response ------------------- There is no conflict as written, the results are undefined. No change is required to the specification. Rationale ------------- None. Forwarded to Base group: Dec 11 1996 Approved: Feb 1998