8.2 pl/pgsql crash bug (WAS: [pgadmin-support] Error craches pgAdmin)

This bug seems to be introduced by this recent change to avoid memory
leakage:

Log Message:
-----------
Prevent intratransaction memory leak when a subtransaction is aborted
in the middle of executing a SPI query. This doesn't entirely fix the
problem of memory leakage in plpgsql exception handling, but it should
get rid of the lion's share of leakage.

Modified Files:
--------------
pgsql/src/backend/executor:
spi.c (r1.164 -> r1.165)
(http://developer.postgresql.org/cvsweb.cgi/pgsql/src/backend/executor/spi.c.diff?r1=1.164&r2=1.165

I don't know that code too well, but somehow the tuptable memory context
gets messed up / not free'd properly etc.

Dave Page wrote:

The issue below was reported to us as a pgAdmin bug - it can be
recreated in psql on 8.2.0, and results in:

2006-12-12 09:06:50 LOG: server process (PID 4588) exited with exit
code -1073741819
2006-12-12 09:06:50 LOG: terminating any other active server processes
2006-12-12 09:06:50 WARNING: terminating connection because of crash of
another server process
2006-12-12 09:06:50 DETAIL: The postmaster has commanded this server
process to roll back the current transaction and exit, because another
server process exited abnormally and possibly corrupted shared memory.
2006-12-12 09:06:50 HINT: In a moment you should be able to reconnect
to the database and repeat your command.

In 8.1.5, it works as expected (ie. without crashing).

Regards, Dave.

-------- Original Message --------
Subject: [pgadmin-support] Error craches pgAdmin
Date: Mon, 11 Dec 2006 20:11:39 +0100
From: Paolo Saudin <paolo@ecometer.it>
To: <pgadmin-support@postgresql.org>

Hi,

I found a different pgAdmin behavior if I use it against Postgres
8.1.15 or 8.2.0. Here to try it out

I have a table filled with dates :

CREATE TABLE _master_h24 (
fulldate timestamp without time zone NOT NULL,
CONSTRAINT _master_h24_pkey PRIMARY KEY (fulldate)
) WITHOUT OIDS;

I have a function to fulfill it with dates :

CREATE OR REPLACE FUNCTION fillmastertable_h24()
RETURNS timestamp without time zone AS
$BODY$
declare
dt_now timestamp;
dt_last timestamp;
max_loops INTEGER;
v INTEGER;
BEGIN

-- gets the last update
SELECT fulldate INTO dt_last FROM _master_h24 ORDER BY fulldate DESC
LIMIT 1;

--RAISE NOTICE 'last : % ', dt_last;
-- gets the gmt - 1 hour date time
dt_now := to_timestamp(TIMEZONE('WAT',CURRENT_TIMESTAMP),
'YYYY-MM-DD');

--RAISE NOTICE 'dt_now : % ', dt_now;
max_loops := 100;
v := 0;

WHILE dt_last < dt_now AND v < max_loops loop
v := v + 1;
dt_last := dt_last + '24 HOUR'::INTERVAL;

/* execute query */
BEGIN
RAISE NOTICE 'Dt : % ', dt_last;
insert into _master_24 (fulldate) VALUES
(dt_last); /* ß <- HERE IS THE TABLE MISSPELLING
(_master_24 ) */

/* errors check */
EXCEPTION
/* in case of any error */
WHEN OTHERS THEN RAISE NOTICE 'ERROR in fillmastertable_h24';
END;
END loop;
return dt_last;
END;
$BODY$
LANGUAGE 'plpgsql' VOLATILE;

ALTER FUNCTION fillmastertable_h24() OWNER TO postgres;

I then insert the first date to begin with :

insert into _master_h24 (fulldate) VALUES ('2006-12-01'); -> OK

I run the script :

SELECT fillmastertable_h24();

And on Postgres 8.1.15 I get back ‘ERROR: “relation _master_24” does not
exist -> OK

While on Postgres 8.2.0 I get back only “:” and pgAdmin craches loosing
the connection to the server. If I click on the server node I get the
following message box :

An error has occurred:

Server closed the connection unexpectedly

This probably means the server terminated abnormally before or while
processing the request

I don’t know if depends on pgAdmin or the server itself.

Thanks,

Paolo Saudin

--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com

Heikki Linnakangas <heikki@enterprisedb.com> writes:

This bug seems to be introduced by this recent change to avoid memory
leakage:

I see no crash in CVS tip --- I believe it's same bug fixed here:

2006-12-07 19:40 tgl

* src/backend/executor/: spi.c (REL8_2_STABLE), spi.c: Avoid double
free of _SPI_current->tuptable. AtEOSubXact_SPI() now tries to
release it in a subtransaction abort, but this neglects possibility
that someone outside SPI already did. Fix is for spi.c to forget
about a tuptable as soon as it's handed it back to the caller. Per
bug #2817 from Michael Andreen.

regards, tom lane