Re: Syslog Facility Patch - Mailing list pgsql-hackers

From Tom Lane
Subject Re: Syslog Facility Patch
Date
Msg-id 5839.974228037@sss.pgh.pa.us
Whole thread Raw
In response to Re: Syslog Facility Patch  (Peter Eisentraut <peter_e@gmx.net>)
Responses Re: Syslog Facility Patch  (Tom Lane <tgl@sss.pgh.pa.us>)
Re: Syslog Facility Patch  (Peter Eisentraut <peter_e@gmx.net>)
List pgsql-hackers
Peter Eisentraut <peter_e@gmx.net> writes:
> Larry Rosenman writes:
>> In looking at this some more, it appears that *SOMETHING* is not
>> allowing messages from set_config_option() in
>> /src/backend/utils/misc/guc.c out WHEN WE ARE DEALING WITH syslog type
>> stuff and we are reading it from the postgresql.conf file.  I'm not
>> sure how to deal with this..  I set syslog=2 then syslog_facility to 
>> some invalid value, and all I get is the syntax error message, not the 
>> "option rejects value" message. 

> I can't reproduce that.  I set 'syslog_facility = local97' and got the
> right error message.

I'm surprised you get any error message at all (as seen by a client,
that is, not as seen in the postmaster log).  AFAICT, backend libpq
is not fired up until well down inside PostmasterMain --- look at the
call to pq_init.  Until that's done, it's impossible to send error
messages to the client.  I've been meaning to look at whether the
pq_init call couldn't be moved up to before (instead of after) option
and switch processing ...
        regards, tom lane


pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: Re: [PATCHES] PostgreSQL virtual hosting support
Next
From: Larry Rosenman
Date:
Subject: Re: Syslog Facility Patch