нашел в исходниках ioloop.c такой кусок кода:........................
if (ioloop_time > ioloop_timeval.tv_sec) {
time_t diff = ioloop_time - ioloop_timeval.tv_sec;
/* Note that this code is here only because this is the easiest
place to check for this. The I/O loop code itself could be
easily fixed to work with time moving backwards, but there's
really no point because there are a lot of other places
which may break in more or less bad ways, such as files'
timestamps moving backwards. */
if (diff > IOLOOP_MAX_TIME_BACKWARDS_SLEEP) {
i_fatal("Time just moved backwards by %ld seconds. "
"This might cause a lot of problems, "
"so I'll just kill myself now. "
"http://wiki.dovecot.org/TimeMovedBackwards",
(long)diff);
} else {
i_error("Time just moved backwards by %ld seconds. "
"I'll sleep now until we're back in present. "
"http://wiki.dovecot.org/TimeMovedBackwards",
(long)diff);
/* Sleep extra second to make sure usecs also grows. */
diff++;
while (diff > 0 && sleep(diff) != 0) {
/* don't use sleep()'s return value, because
it could get us to a long loop in case
interrupts just keep coming */
diff = ioloop_time - time(NULL) + 1;
}
/* Try again. */
io_loop_handle_timeouts(ioloop, TRUE);
}
}
.........................
что если убрать функцию i_fatal?
насколько это опасно?
ЗЫ смущают коменты...
Вариант для распечатки
OpenNET: Виртуальная конференция (Public)
on 29-Янв-08, 12:14 
