Oops, wrong message with the other patch...this was the patch for the other

Oops, wrong message with the other patch...this was the patch for the other
comment, so here is the comment for the other patch *grin*

> > You are right.  I checked the gramar and saw the ability to use the
> > parameter.  I looked at the manual pages, and saw no reference to it.  I
> > tried running it, and found vacuum does nothing when you give it a table
> > name.
> >
> > I checked a debug version of postgres, and the table name is passed to
> > vacuum() in the variable (char *vacrel).  The problem is that the vacuum
> > spans transactions, and the vacrel name gets changed to '<vacuum>',
> > which is the name of the portal that gets created in
> > vacuum.c::_vc_vacuum().  vacuum.c::_vc_init() does a
> > CommitTransactionCommand() which frees the memory allocated to vacrel.
> >
> > Should I change vacuum.c to copy the relation name to a local string
> > variable of vacuum(), or do you recommend we allocate the table name in
> > a different fashion?  You are the man who knows the most about this.
>
> static NameData VacRel;

Done.  Attached is the patch.  I have already applied it to the 2.0
tree. (Marc!)

I tested it and it works.  I also applied documentation patches to go
with it.

So now vacuum can be run for only one table if you wish.

Submitted by: Bruce Momjian <maillist@candle.pha.pa.us>