On Tue, Aug 23, 2005 at 09:12:10PM -0400, Andrew Dunstan wrote:
> The attached patch completes (I hope) the work begun by Michael Fuhr in
> an earlier unapplied patch, and makes strict mode work as recently
> discussed. I moved the embedded strings out of the calling functions
> into global macros to try to make the code a little more readable.
>
> Unfortunately we can't have regression tests for this because it relies
> on a custom variable class.
Hmmm...even if the "plperl" custom variable class isn't defined in
postgresql.conf, plperl.use_strict springs into existence when the
interpreter is initialized:
test=> SET plperl.use_strict TO on;
ERROR: unrecognized configuration parameter "plperl.use_strict"
test=> CREATE FUNCTION foo() RETURNS void AS $$$$ LANGUAGE plperl;
CREATE FUNCTION
test=> SET plperl.use_strict TO on;
SET
test=> CREATE OR REPLACE FUNCTION foo() RETURNS void AS $$ $x = 1234; $$ LANGUAGE plperl;
ERROR: creation of Perl function failed: Global symbol "$x" requires explicit package name at (eval 8) line 1.
test=> SET plperl.use_strict TO off;
SET
test=> CREATE OR REPLACE FUNCTION foo() RETURNS void AS $$ $x = 1234; $$ LANGUAGE plperl;
CREATE FUNCTION
Is such automatic creation of a GUC variable intended? If so,
couldn't you exploit that in regression tests?
--
Michael Fuhr