Обсуждение: This script will crash the connection

Helo all,

    Please try this script, it will crash the current connection.
    I'm using the 01/18/2001 PostgreSQL v7.1 beta3 snapshot.

-- Script begin -------------------------------------
create table blah(
  var_field varchar(8),
  n1 integer default 23,
  n2 integer,
  arr_str varchar[],
  m money,
  s text
);

create rule blah_update as
 on update to blah
   do
     notify TestEvent;

INSERT INTO blah (var_field, n1, n2, arr_str, m, s) VALUES ('aaa', 1, 2,
NULL, NULL, NULL);
UPDATE blah SET n1=n1+1;  -- Won't crash the connection
UPDATE blah SET n1=2 WHERE var_field='aaa' AND n1=1 AND n2=2 AND arr_str IS
NULL AND m IS NULL; -- Will crash the connection

-- Script end -------------------------------------
    psql will print :

pqReadData() -- backend closed the channel unexpectedly.
        This probably means the backend terminated abnormally
        before or while processing the request.
The connection to the server was lost. Attempting reset: Failed.
!#

Any comments ?

I need this kind of code working for a demo for ZeosDBO users...

--

Best Regards,
Steve Howe

----- Original Message -----
From: "Steve Howe" <howe@carcass.dhs.org>


> Helo all,
>
>     Please try this script, it will crash the current connection.
>     I'm using the 01/18/2001 PostgreSQL v7.1 beta3 snapshot.
> UPDATE blah SET n1=2 WHERE var_field='aaa' AND n1=1 AND n2=2 AND arr_str
IS
> NULL AND m IS NULL; -- Will crash the connection

Confirmed - only when the update has a "where" clause - a rule on insert or
an update without where seems to work.

Doesn't seem to be the notify - backend crashes with

create rule blah_upd as on update to blah do insert into foo (dummy) values
('test');

version:
========
PostgreSQL 7.1beta3 on i586-pc-linux-gnu, compiled by GCC egcs-2.91.66

- Richard Huxton

"Richard Huxton" <dev@archonet.com> writes:
> Doesn't seem to be the notify - backend crashes with
> create rule blah_upd as on update to blah do insert into foo (dummy) values
> ('test');

Actually, that was a completely different bug :-(.  Thanks for the
report.

            regards, tom lane