Re: Re: BUG #11014: Postgres can be put into an error state by setting invalid timezone.

David G Johnston <david(dot)g(dot)johnston(at)gmail(dot)com> writes:
> Though pondering this additionally the actual coding to make -0400 be
> interpreted as "-4:00" would indeed be a feature enhancement - though
> accepting a result of 400 I would argue is a bug even if it is one we are
> going to choose to live with in back branches for compatibility reasons. We
> seem to already agree that it should at least error out on a going-forward
> basis.

The fact that it crashes is indisputably a bug (fixed as of a few minutes
ago). However, whether SET TIMEZONE = '-0400' should be read as setting
the UTC offset to 400 hours, or 400 minutes, or 400 seconds, or 4 hours,
or who knows what else, is not a bug, it's a definitional disagreement.

As a comparison point, the syntax SET TIME ZONE INTERVAL '-0400' is
specified in the SQL standard, and unless I'm totally misreading it, the
standard requires that to be interpreted as 400 minutes. (The standard's
examples of useful values look more like '-04:00'.) Now that might look
silly on its face, but if the string is '-180' then reading it as UTC
minus 3 hours doesn't seem so silly.

Anyway, we could certainly have a discussion about changing the
interpretation, but it would not be a back-patchable bug fix IMO.
There's a non-negligible risk of breaking apps that worked before.

regards, tom lane