Re: Bug in autovacuum.c?

Bruce Momjian wrote:
> > > I am not so concerned about this case but about other cases where we are
> > > computing xid distances across the invalid range.
> >
> > Such as?
>
> Not sure. I have not had time to research this, but there might be
> cases where this backward movement matters --- remember our XIDs are
> valid only within about a 2 billion range, and we do less/greater
> comparisons in that range (using a macro). That macro is not going to
> cover over backward xid movement.

OK, I am done training for the day, and found this macro:

/* advance a transaction ID variable, handling wraparound correctly */
#define TransactionIdAdvance(dest) \
do { \
(dest)++; \
if ((dest) < FirstNormalTransactionId) \
(dest) = FirstNormalTransactionId; \
} while(0)

which seems OK, but we the -= all over varsup.c

/*
* We'll refuse to continue assigning XIDs in interactive mode once we get
* within 1M transactions of data loss. This leaves lots of room for the
* DBA to fool around fixing things in a standalone backend, while not
* being significant compared to total XID space. (Note that since
* vacuuming requires one transaction per table cleaned, we had better be
* sure there's lots of XIDs left...)
*/
xidStopLimit = xidWrapLimit - 1000000;
if (xidStopLimit < FirstNormalTransactionId)
xidStopLimit -= FirstNormalTransactionId;

Now I am not sure where to add a C comment. :-(

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

+ It's impossible for everything to be true. +