Jochen Erwied <jochen@pgsql-bugs.erwied.eu> writes:
> Saturday, April 21, 2012, 7:49:55 PM you wrote:
>> So: some platform-specific misbehavior here. I have no time to poke at
>> it more now, though. Who else can reproduce this, on what platforms?
> One machine works, the other not:
I traced though it far enough to find that pg_next_dst_boundary indexes
off the end of an array when the given probe time is exactly the last
DST transition time for the zone. So what's surprising is not that
it fails weirdly, but that there seem to be numerous machines where
it doesn't (appear to) fail. The next array slot must chance to have
a sane value in some environments.
regards, tom lane