On Wed, 20 May 1998, Ben Smithurst wrote:
> > This could perhaps be changed to
> >
> > if (strncmp(message_id, last_message_id, 6) > 0)
> > message_sequence_number = 0;
>
> Well, I've applied that, and the patch, and I think it's fixed now:
You shouldn't need that as well as the patch.
> Evem when the clock jumped back six seconds, the time portion of the message
> wasn't incremented, and the count of messages per second per process was. A
> quick check reveals no more duplicate IDs.
Good. I think that confirms the cause of the duplication.
> Hmm... I'll have a fiddle around with that, but it might not be needed
> now I've applied Philip's fixes. I suppose it might create confusing
> entries in other logs, but I would have thought that would happen
> however the time was adjusted.
If the time is always adjusted forwards it should never cause problems.
(i.e. if it's slow you have some short seconds and if it's fast you have
some long seconds) then nobody should ever get confused. I would imagine
there are programs other than Exim that won't be happy with clocks that
go backwards. Think about timestamps on files.
--
Philip Hazel University Computing Service,
ph10@??? New Museums Site, Cambridge CB2 3QG,
P.Hazel@??? England. Phone: +44 1223 334714
--
*** Exim information can be found at
http://www.exim.org/ ***
