Bruce Momjian <pgman@candle.pha.pa.us> writes:
> There is some code in gram.y that detects whether you are in a RULE so
> NEW/OLD can be detected. Seems the value is reset on parser start and
> set on RULE start, but not reset on rule and. A multi-query string
> could use NEW/OLD in the queries after the RULE even though they are
> invalid. The following patch fixes this by resetting the flag when the
> rule action happens.
I was about to say "fix ecpg's grammar too", but a quick look shows that
Michael was way ahead of the rest of us on this one ...
regards, tom lane