On 02/03/2010 10:02 PM, Pete Zaitcev wrote:
As it turned out, event_loopbreak() does not awaken the thread that
exectutes event_dispatch(), but our code expected that it would.
One easily noticeable effect was that there was a noticeable delay
between the state transition to DB master and listening on sockets.
I knew the mysterious delay existed for a while, but never got around
to investigate. For ncld API, I moved the processing of CLD packets
to its own thread and suddenly everything else froze. Apparently the
existing code only works because of extra packets of CLD protocol.
For a fix, dispose of event_loopbreak and use a loopback pipe.
Also gone is state_tdb_new. That thing was just disgusting.
Notice that we still have one event_loopbreak remaining, because it
works correctly thanks to UNIX signal awakening the polling thread.
Signed-Off-By: Pete Zaitcev<[email protected]>
applied, with a few cosmetic changes (speling correction, use of 'switch')
--
To unsubscribe from this list: send the line "unsubscribe hail-devel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
