Am 09.09.2013 10:08, schrieb Paul J Stevens: > On 09-09-13 08:27, Thomas Raschbacher wrote: > >>> write(13, "Q", 1) = -1 EAGAIN (Resource >>> temporarily unavailable) >>> read(25, 0x1843e70, 11) = -1 EAGAIN (Resource >>> temporarily unavailable) > > Thomas, > > could you please test 4c23432cc270554557f9e130331214d81164131b > > I've put a mutex lock around the file-handle involved here, since read > and write are not thread-safe, and both actions here can (and will) > happen in different threads. > > The only scenario I could come up with is: > > Worker thread starts writing to filehandle 13 (the selfpipe). This > triggers a read-event in the main thread even before the worker has > finished writing, causing both of them to be unable to finish the call. > > If this hypothesis is correct, this should be solved be mentioned commit > ID. And anyway, I doesn't hurt to protect the self-pipe handles with a lock
great - while i could never reproduce this problem my load-testing on 4c23432cc270554557f9e130331214d81164131b is fine as well we tested Apple Mail, Outlook 2003/2007/2010 and mixed load POP3/IMAP looks like a good 3.1.6 candidate for now but let's wait for others with problems in the past
signature.asc
Description: OpenPGP digital signature
_______________________________________________ DBmail mailing list [email protected] http://mailman.fastxs.nl/cgi-bin/mailman/listinfo/dbmail
