Re: [HACKERS] temp table oddness?

Bruce Momjian <maillist@candle.pha.pa.us> writes:
> I have added temp invalidation code for aborted transactions:

> New behavour:

>     test=> begin;
>     BEGIN
>     test=> create temp table test (x int);
>     CREATE
>     test=> create index i_test on test(x);
>     CREATE
>     test=> abort;
>     NOTICE:  trying to delete a reldesc that does not exist.
>     NOTICE:  trying to delete a reldesc that does not exist.
>     ABORT
>     test=> create temp table test(x int);
>     CREATE

OK, cool.  I think I know where to fix those "NOTICES", too:
the relcache indexes temp relations by their real names, so
RelationNameGetRelation() ought to substitute the real name before
probing the cache.  As it stands you wind up with two relcache entries
for the temp table, which is bad.  Working on it now.
        regards, tom lane