Re: API bug in DetermineTimeZoneOffset()

I wrote:
> The second attached patch, to be applied after the first, removes the
> existing checks of HasCTZSet in the backend.  The only visible effect of
> this, AFAICT, is that to_char's TZ format spec now delivers something
> useful instead of an empty string when a brute-force timezone is in use.
> I could be persuaded either way as to whether to back-patch this part.
> From one standpoint, this to_char behavioral change is clearly a bug fix
> --- but it's barely possible that somebody out there thought that
> returning an empty string for TZ was actually the intended/desirable
> behavior.

Any opinions about whether to back-patch this part or not?  It seems
like a bug fix, but on the other hand, I don't recall any complaints
from the field about to_char's TZ spec not working with brute-force zones.
So maybe the prudent thing is to leave well enough alone.
        regards, tom lane