Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> Wow, that is bizarre. However, I think the tests are run with US
> timezones.
The regression tests run in PST8PDT zone, and therefore the US
transition days are the ones that count, not your local ones.
This is even documented ;-) ... see section 26.2.3 in the 7.4 docs.
regards, tom lane