Re: relcache leak warnings vs. errors - Mailing list pgsql-hackers

From Andres Freund
Subject Re: relcache leak warnings vs. errors
Date
Msg-id 20200413200025.3rnw7r5ssbsqntvv@alap3.anarazel.de
Whole thread Raw
In response to Re: relcache leak warnings vs. errors  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: relcache leak warnings vs. errors  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
Hi,

On 2020-04-11 10:54:49 -0400, Tom Lane wrote:
> Peter Eisentraut <peter.eisentraut@2ndquadrant.com> writes:
> > How about a compile-time option to turn all the warnings in resowner.c 
> > into errors?  This could be enabled automatically by --enable-cassert, 
> > similar to other defines that that option enables.
> 
> [ itch... ]  Those calls occur post-commit; throwing an error there
> is really a mess, which is why it's only WARNING now.

> I guess you could make them PANICs, but it would be an option that nobody
> could possibly want to have enabled in anything resembling production.
> So I"m kind of -0.5 on making --enable-cassert do it automatically.
> Although I suppose that it's not really worse than other assertion
> failures.

I'd much rather see this throw an assertion than the current
behaviour. But I'm wondering if there's a chance we can throw an error
in non-assert builds without adding too much complexity to the error
paths. Could we perhaps throw the error a bit later during the commit
processing?

Greetings,

Andres Freund



pgsql-hackers by date:

Previous
From: Robert Haas
Date:
Subject: Re: documenting the backup manifest file format
Next
From: Andrew Dunstan
Date:
Subject: Re: documenting the backup manifest file format