Andrew Gierth <andrew@tao11.riddles.org.uk> wrote:
> Surely what matters is the value of the GUC at the time that you did
> the CREATE FUNCTION, not the value at the time you happen to be
> calling it?
Well, that's a change I'm arguing for. That would require both the
plpgsql parser change Tom is talking about, and a change to CREATE
FUNCTION such that there is an implied SET standard_compliant_strings
FROM CURRENT -- which is something I've suggested a couple times;
there's been no explicit response to that.
See back here in the thread for some behavior which surprised me:
http://archives.postgresql.org/pgsql-hackers/2009-04/msg00519.php
-Kevin