> Well, we have a bug report against 6.4, so my inclination is to change
> it in both trees. We know that the current code is broken on at least
> one major platform. I think that outweighs the unproven possibility
> that the other way might be broken somewhere...
OK, I applied the fix (confined to backend/utils/adt/nabstime.c) to both
CVS trees. This swaps the checking of HAVE_INT_TIMEZONE and HAVE_TM_ZONE
so that the TM_ZONE takes precedence.
This single-file patch should be a candidate for removal if others run
into trouble when testing.
One thing I wasn't certain about was the use of #error in non-gcc
compilers. Is that a standard feature?
- Tom