Re: Writeable CTEs and empty relations
Marko Tiikkaja <marko.tiikkaja@cs.helsinki.fi>
From: Marko Tiikkaja <marko.tiikkaja@cs.helsinki.fi>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2010-02-09T05:32:44Z
Lists: pgsql-hackers
On 2010-02-09 01:09 +0200, Tom Lane wrote: > Marko Tiikkaja <marko.tiikkaja@cs.helsinki.fi> writes: >> I traced this down to heapam.c, which has this: >> ... >> This doesn't exactly work anymore since we modify the snapshot after >> calling ExecInitScan(). > > So don't do that. The executor is generally entitled to assume that > parameters given to ExecutorStart are correct. In particular, changing > the snapshot after the query has started to run seems to me to ensure > all sorts of inconsistent and undesirable behavior. You do remember that the whole patch in its current form depends on modifying the snapshot? Regards, Marko Tiikkaja