Re: [HACKERS] Bogus "Non-functional update" notices

> With fairly current sources (last cvs update on 7/20), I am seeing
> occasional occurrences of
> NOTICE: Non-functional update, only first update is performed
> I think this is a bug. The update commands that are triggering this
> message *are* getting executed. I looked at the sources and couldn't
> even understand what condition was being tested to generate the
> message.
> The source code looks like it's trying to disallow more than one
> update to the same tuple within a transaction, which is so silly that
> I have to be misreading it...

I recall seeing this in the past when two conditions in an update are
such that the second condition will never be significant. Can't remember
an example, but the regression test has at least one case which provokes
this.

- Tom