No, I am confident that we are not passing sockets through threads or processes. Could this have anything to do with closing and re opening connections?
On Mon, Jul 9, 2012 at 2:52 AM, Pieter Hintjens <[email protected]> wrote: > Hi Thomas, > > Can you definitely exclude that sockets end up in the wrong thread? I > mean, are you at any time passing socket handles between threads? > > -Pieter > > On Mon, Jul 9, 2012 at 1:53 PM, Thomas S Hatch <[email protected]> wrote: > > Hi, > > > > I have a report of an assertion in Salt. It is very rare and I cannot > > reliably reproduce it. The user that has reported it gave me a strace > output > > which makes it look like the assertion is happening in a multiprocess. I > am > > aware that you cannot have the same socket call from another thread or > > process, but I see no indication that this is the case. > > > > Here is the trace: > > > > 14:51:29.164045 epoll_wait(9, {{EPOLLIN, {u32=44624800, u64=44624800}}}, > > 256, 4294967295) = 1 > > 14:51:55.130374 poll([{fd=8, events=POLLIN}], 1, 0) = 1 ([{fd=8, > > revents=POLLIN}]) > > 14:51:55.130428 write(2, "Assertion failed: ok (mailbox.cpp:84)\n", 38) > = -1 > > EBADF (Bad file descriptor) > > 14:51:55.130466 rt_sigprocmask(SIG_UNBLOCK, [ABRT], NULL, 8) = 0 > > 14:51:55.130500 tgkill(5969, 5980, SIGABRT) = 0 > > 14:51:55.130530 --- SIGABRT (Aborted) @ 0 (0) --- > > > > Any ideas? > > > > _______________________________________________ > > zeromq-dev mailing list > > [email protected] > > http://lists.zeromq.org/mailman/listinfo/zeromq-dev > > > _______________________________________________ > zeromq-dev mailing list > [email protected] > http://lists.zeromq.org/mailman/listinfo/zeromq-dev >
_______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
