age(xid) function bug

I see we recommend using the age(xid) function to check for XID
wraparound:

http://www.postgresql.org/docs/8.3/static/routine-vacuuming.html#VACUUM-FOR-WRAPAROUND

However, the function only does a subtraction with no adjustment for
overflow, and considering it returns a signed int, it can't even display
the full range of valid values:

test=> \df age
List of functions
Schema | Name | Result data type | Argument data types
------------+------+------------------+----------------------------------------------------------
...
pg_catalog | age | integer | xid

For example:

test=> select txid_current();
txid_current
--------------
397
(1 row)

test=> select age('10000'::xid);
age
-------
-9602
(1 row)

Seems this is a bug. I think the fix is to do the arithmetic in int8
and return an int8.

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

Bruce Momjian <bruce(at)momjian(dot)us> writes:
> However, the function only does a subtraction with no adjustment for
> overflow, and considering it returns a signed int, it can't even display
> the full range of valid values:

Say again? The possible range of ages is only 2 billion.

regards, tom lane

Tom Lane wrote:
> Bruce Momjian <bruce(at)momjian(dot)us> writes:
> > However, the function only does a subtraction with no adjustment for
> > overflow, and considering it returns a signed int, it can't even display
> > the full range of valid values:
>
> Say again? The possible range of ages is only 2 billion.

If we assume that only xid stored in actual tables are used, yes:

test=> select txid_current();
txid_current
--------------
402
(1 row)

test=> select age('4294967290'::xid);
age
-----
409
(1 row)

OK, so it does work based on the limitations we place on stored xids.

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +