Bruce Momjian <pgman@candle.pha.pa.us> writes:
> If there is a problem, maybe we can fix it, or perhap have the kill
> function use SIGINT, then wait for the query to cancel, then SIGTERM.
Well, if someone could prove that the SIGTERM path is equivalent to
a transaction-aborting error followed by normal client disconnect,
I'd feel a lot better about its reliability. We know those two code
paths work well.
Right now I do not think they are equivalent, but maybe they could be
made so.
regards, tom lane