Huh, a non-zero XMAX is fine. You mark the XMAX when you _think_ you
are updating it. It is only expired when the XMAX on the tuple is
committed.
> Or perhaps MAINTAINED INDEX, meaning that it has always both tmin and tmax
> up-to-date.
> Btw 7.2 still has broken behaviour of xmax which by definition should
> not have a
> non-0 value for live tuples
>
> pg72b2=# create table parent(pid int primary key);
> NOTICE: CREATE TABLE / PRIMARY KEY will create implicit index
> 'parent_pkey' for table 'parent'
> CREATE
> pg72b2=# create table child(cid int, pid int references parent);
> NOTICE: CREATE TABLE will create implicit trigger(s) for FOREIGN KEY
> check(s)
> CREATE
> pg72b2=# insert into parent values(1);
> INSERT 16809 1
> pg72b2=# insert into child values(1,1);
> INSERT 16810 1
> pg72b2=# update child set pid=2;
> ERROR: <unnamed> referential integrity violation - key referenced from
> child not found in parent
> pg72b2=# select xmin,xmax,* from child;
> xmin | xmax | cid | pid
> ------+------+-----+-----
> 171 | 172 | 1 | 1
> (1 row)
-- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610)
853-3000+ If your life is a hard drive, | 830 Blythe Avenue + Christ can be your backup. | Drexel Hill,
Pennsylvania19026