> > I have been studying DeadLockCheck for most of a day now,
> > and I doubt that this is the only bug lurking in it.
> > I think that we really ought to throw it away and start
> > over, because it doesn't look to me at all like a standard
> > deadlock-detection algorithm. The standard way of
> I have been studying DeadLockCheck for most of a day now, and I doubt
> that this is the only bug lurking in it. I think that we really ought
> to throw it away and start over, because it doesn't look to me at all
> like a standard deadlock-detection algorithm. The standard way of doing
Go ah
Create three tables and start four transactions, then do:
XACT 1: LOCK TABLE A;
XACT 2: LOCK TABLE B IN ROW SHARE MODE;
XACT 3: LOCK TABLE B IN ROW EXCLUSIVE MODE;
XACT 4: LOCK TABLE C;
XACT 2: LOCK TABLE C;
XACT 3: LOCK TABLE C;
XACT 1: LOCK TABLE B IN SHARE MODE;
<< wait at least
