> That's just a workaround, to avoid the freezing. It would be
> nice if we
> could find the underlying cause of the SIGALRM crash, because
> it should
> never happen.
I also had earlier made the same exact change to deliver.c as A.Wik, but as Philip mentions above it only works around the freezing.
If connect_timeout option is not set (default of course) then exim behaves normally, ie when a timeout occurs, it is properly deferred and not frozen, log shows:
defer (110): Connection timed out
but if connect_timeout is set, the log shows:
defer (-1): smtp transport process returned non-zero status 0x000e: terminated by signal 14
(and without below patch the message is frozen)
My system is linux, kernel 2.4.20, glibc 2.2.4
But for the real problem, I guess the alarm handler set by sigaction() isn't being set?
Thanks,
--- exim-4.31/src/deliver.c.orig Tue Apr 13 20:27:34 2004
+++ exim-4.31/src/deliver.c Tue Apr 13 20:27:52 2004
@@ -3136,7 +3136,7 @@ set_process_info("delivering %s", messag
addrlist = parlist[poffset].addrlist;
/* If the process did not finish cleanly, record an error and freeze (except
-for SIGTERM, SIGKILL and SIGQUIT), and also ensure the journal is not removed,
+for SIGTERM, SIGKILL, SIGQUIT and SIGALRM), and also ensure the journal is not removed,
in case the delivery did actually happen. */
if ((status & 0xffff) != 0)
@@ -3153,7 +3153,7 @@ if ((status & 0xffff) != 0)
(msb == 0)? "terminated by signal" : "exit code",
code);
- if (msb != 0 || (code != SIGTERM && code != SIGKILL && code != SIGQUIT))
+ if (msb != 0 || (code != SIGTERM && code != SIGKILL && code != SIGQUIT && code != SIGALRM))
addrlist->special_action = SPECIAL_FREEZE;
for (addr = addrlist; addr != NULL; addr = addr->next)
