Bokhan Artem wrote:
I would suggest that buffered writes in client/server communications are
necessary too.
Network writes have been buffered for some time already. Yesterday I
added an optimization to them to reduce the number of syscall by not
resizing the buffer each time data had been written to the client.
But these are not about that:
[pid 29764] 02:06:40.056515 <... read resumed> "Q", 1) = 1
[pid 29774] 02:06:40.056570 <... write resumed> ) = 1
[pid 29764] 02:06:40.056618 write(18, "* 1 FETCH (FLAGS () UID 2)\r\n",
28 <unfinished ...>
We're using a self-pipe with single-char writes in the worker threads to
notify the main thread a message is waiting in the async message queue.
That is what you are seeing here.
[.....]
[pid 29774] 02:06:40.057614 write(16, "Q", 1) = 1
[pid 29774] 02:06:40.057707 write(16, "Q", 1) = 1
[pid 29764] 02:06:40.057774 write(18, "* 2 FETCH (FLAGS () UID 4)\r\n",
28 <unfinished ...>
--
________________________________________________________________
Paul Stevens paul at nfg.nl
NET FACILITIES GROUP GPG/PGP: 1024D/11F8CD31
The Netherlands________________________________http://www.nfg.nl
_______________________________________________
Dbmail-dev mailing list
[email protected]
http://twister.fastxs.net/mailman/listinfo/dbmail-dev
