Dana Burd <DBurd@cpr.org> writes:
> to_char(timestamp, 'SSSS') has a Daylight Savings bug. It includes the
> missing 2am hour in it's "seconds from midnight count" for DST begin day
> (4/6/2003), and does not include the extra hour for DST end day
> (10/26/2003).
It's not readily apparent to me that this is a bug. The SSSS value is
evidently being calculated as HH*3600+MM*60+SS, so it corresponds to the
nominal time of 3:00:00 AM. I think it would be rather surprising if
SSSS *didn't* correspond to HH:MM:SS.
However, to_char() is intended to slavishly imitate Oracle's to_char(),
warts and all. What does Oracle do with this situation?
regards, tom lane