Re: [HACKERS] AbortTransaction and not in in-progress state

> > > I see a small problem in running VACUUM ANALYZE on the regression
> > > database which involves an entry in pg_operator left over from the
> > > char16 datatype. If the line containing "1004" is removed and the
> > > system is "make clean install; initdb"'d then vacuum completes
> > > successfully on that database.
> > Also, vacuum analyze; on regression database produces
> > regression-> vacuum analyze;
> > ERROR: type id lookup of 1004 failed
> > I don't know which file I need to edit as you suggested :-)
>
> Oh, sorry. src/include/catalog/pg_operator.h is the one to change. I've
> already made the change to the development cvs tree, so you could check
> out the newest code.
>
> I haven't had a chance to try your test case yet. Let me know how the
> regression test goes and then I'll look at your test if you would still
> like me to...

I have added contrib/findoidjoins/make_oidjoins_check which will take
the output of findoidjoins and create a set of SQL queries that will
find any bad oid's in the tables. I have run it on findoidjoins on the
template1 database, and have created template1_check.sql in
include/catalog to find any bad oid's. I found a few in pg_operator(the
oid was '6'), and fixed them too. We can run the query anytime to find
any of these problems in the future.

--
Bruce Momjian | 830 Blythe Avenue
maillist(at)candle(dot)pha(dot)pa(dot)us | Drexel Hill, Pennsylvania 19026
+ If your life is a hard drive, | (610) 353-9879(w)
+ Christ can be your backup. | (610) 853-3000(h)