Object to control execution of background threads.
Encapsulates an activation loop that relies on pthread condition variables. The background thread calls waitForActivation() while it waits for an activation from an orchestrating thread. Before calling this method, however, it must acquire the mutex, in order to obey the pthread wait protocol. Once its work is complete, the activated state is unsed and the mutex must be released. In practice, therefore, waitForActivation() is usually invoked as follows.
while (true) {
bool activated = control.waitForActivation();
if (!activated) {
break;
}
}
The wait for an activation is interrupted either by receiving a signal or by setting a shutdown flag, in which case the method returns false. The flag is simply a pointer to some external boolean variable, which means that the background thread "watches for" a shutdown flag somewhere else. The orchestrating thread calls activate() to wake up the background thread, causing the waitForActivation() call to return with true. The wait parameter makes the activation wait to acquire the mutex, which guarantees that the signal was sent. Otherwise, if the mutex is already held (i.e., background thread is already running) the method returns false immediately.
Optionally, the activate method accepts an lsn, which is stored in endLSN, but only if it is greater than the currently set value. This class does not interpret this LSN value. It is only used by the background thread itself as a marker for the end of its job. This is useful for threads that consume its owrk units from the log, such as LogArchiver or ReaderThread. Other thread classes may completely ignore this variable.
- See also
- LogArchiver, LogArchiver::ReaderThread
- Author
- Caetano Sauer
