It's not exactly a Y2K bug, because it started to fail at Dec 30 1999
23:00:00, your local time. Nonetheless, the abstime regression test
is now showing a "failure". It's not a code bug --- it's that the test
includes a query whose results depend on whether current time is before
or after 12-30-1999. Boo hiss.
Probably the best answer is to leave the test script alone and change
the expected output. Thomas, do you have a different opinion?
regards, tom lane