Re: 8.1.x (tested 8.1.8) timezone bugs

"Joshua D. Drake" <jd(at)commandprompt(dot)com> writes:
> O.k., really trying to not be difficult here but... if we modified 8.1.9
> (to be released) to not fold the case of the unquoted timezone, doesn't
> that resolve the issue entirely? (well it is a band-aid but still).

No, it doesn't; the issue is what happens when there's no match in the
zic database for your TZ name.

Not to mention that we can't realistically put the case-folding change
into a minor release, both for compatibility reasons and because it was
part of a fairly major overhaul of the timezone code. I no longer
recall what-all it was connected to, but there were a number of
interrelated changes there.

Digging around some more in the zic code, I note that zic has an option
"-p" to select a specific known timezone as "posixrules", and that the
fallback code seems to prefer that to the hardwired string. That might
work for us, if what it takes from the zone is just the DST rules and
not the specific offset. Why don't you try
ln -s EST5EDT /usr/share/postgresql/timezone/posixrules
(adjust path as needed) and then see how it behaves with 'pst8pdt'?

regards, tom lane