Heiko Schlittermann wrote:
> Hello,
>
>
> Philip Hazel <ph10@???> (Do 09 Nov 2006 10:41:54 CET):
>> On Wed, 8 Nov 2006, Heiko Schlittermann wrote:
>>
>>> If each log line could have some pid=4711 (or whatever the key is named)
>>> it could help to reproduce what happend with sessions if they didn't get
>>> the 'spool id' assigned yet.
>> If you turn on log_selector=+smtp_connection the log will contain a
>> start and end connection line. Using the IP address this should help
>> sort out an individual connection, though I guess if a client makes two
>> or more simultaneous connections it won't be clear.
ACK - and a common occurence.
>
> I patched 4.63 (mainly in log.c) a bit to achieve the following:
>
> log_selector = +pid
>
> now creates log lines as follows:
>
> 2006-11-09 12:45:58 [12382] 1Gi8M2-0003Di-0w <= root@??? U=root P=local S=336
>
> But I'm afraid that it may break some tools!?
>
Possibly. But I don't think it would bother eximstats. Easly checked.
> Some other option would be to append something like PID=12382 the each
> log line?
>
> Comments?
>
One of the reasons I suggested generating a message ID 'early' instead of using
PID, was that *AFAIK* all the 'tools' already accomodate that format.
PH pointed out that it would not be wise to do that.
Though not specified, reflection says that one of the reasons might be that the
'convenience' of being able to track the same ID from 'connect' through final
delivery may be offset by making it a bit more difficult to separate messages
that traversed the entire cycle from mere arrivals that were rejected, early-on,
yet carry the same sort of number as if a message had been handled.
TANSTAAFL.
Bill
