Re: PGC_S_DEFAULT is inadequate

Greg Stark <gsstark(at)mit(dot)edu> writes:
> On Wed, May 11, 2011 at 3:20 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> So this is fine if the
>> current value was from the file or was the boot_val, but if we'd
>> overridden the boot value with a "replacement" default value using
>> PGC_S_DEFAULT, that code would cause the value to revert to the boot_val
>> not the replacement value. Not desirable.

> Doesn't this mean if you had a setting of timezone in your config file
> and remove it and reload you'll get a different value than you would
> have if you had actually booted without the line originally?

Yes, except it's timezone_abbreviations (and client_encoding) that are
at risk. I think the latter is probably observably broken in existing
releases, though I've not tried to test. Given the minuscule
use-case for setting client_encoding in postgresql.conf, I'm not
concerned about back-patching a fix, but I'd like to make it work better
going forward.

regards, tom lane