From: | Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> |
---|---|
To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Cc: | Peter Eisentraut <peter_e(at)gmx(dot)net>, PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: elog() proposal |
Date: | 2002-02-23 19:36:55 |
Message-ID: | 200202231936.g1NJauo29240@candle.pha.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Tom Lane wrote:
> Peter Eisentraut <peter_e(at)gmx(dot)net> writes:
> > So, basically, what this comes down to with respect to your patch:
>
> > 1. Renumbering the error codes breaks backward compatibility *silently*.
>
> Perhaps, but it doesn't bother me. We have *never* promised binary
> compatibility of server-side extensions across versions; usually,
> you should be happy if a recompile is sufficient ;-). (Structs,
> for example, are subject to field rearrangement all the time.)
I didn't think we had binary backward compatibility for server
functions. The switch statement to test *_min_messages levels is going
to look pretty bad, compared to a simple greater-than test for level
values.
One intestesting trick would be to start numbering the elog messages
levels at 10, and throw an error if any messages comes in with a value
less than that. That would properly warn people of old object files and
may be the best bet.
> In any case, we could maintain binary compatibility for the old-style
> codes (DEBUG, ERROR, etc); this does not force us to use matching
> codes for the new PG_ERROR etc. levels.
Yes, I think keeping the old symbols in for one release is a good idea.
Again, this would be done only just before beta, during pgindent run,
where there are no outstanding patches.
> > 2. CRASH doesn't seem like a good name to me.
>
> Why not? It's short, memorable, accurate, and what's wrong with
> a little levity?
>
> > 3. I agree with adding a LOG or INFO level between DEBUG and NOTICE.
>
> Both, I think; they're not the same thing. LOG = routine server
> operation notices (eg, "checkpoint starting now"). INFO =
> allegedly-helpful messages issued to client (eg, the one about
> truncating overlength identifiers). Normal configuration would
> be to put one but not the other into the postmaster log.
Sorry, truncating overly long identifiers is a NOTICE, not INFO. I
reserved info for only places where the information did not indicate any
kind of unusual situation, like SERIAL sequence creation. You can go
through the existing NOTICE's after I apply and salt to taste.
--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026
From | Date | Subject | |
---|---|---|---|
Next Message | Bruce Momjian | 2002-02-23 20:02:55 | Re: Patches split from 7.3 queue |
Previous Message | Bruce Momjian | 2002-02-23 19:26:39 | Re: elog() proposal |