Bug in contrib/spi/refint.c

From: Anand Surelia <anand(at)bytekinc(dot)com>
To: "pgsql-general(at)postgreSQL(dot)org" <pgsql-general(at)postgreSQL(dot)org>, pgsql-hackers(at)postgreSQL(dot)org
Subject: Bug in contrib/spi/refint.c
Date: 1998-10-02 21:57:23
Message-ID: 36154C43.122C2BFE@bytekinc.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general pgsql-hackers

There is a bug in check_foreign_key of refint.c which is bundled with
the standard distribution. It occurs when a trigger calling this
function recursively fires another trigger which calls the same
function. The calling check_foreign_key loses its plan informantion and
when it tries to use it the backend closes its channel. You can check it
with the sql script I am attaching below.
The solution to this is to do a find_plan again before executing it at
line 483 of refint.c.
Therefore two more lines should be added before line 483:

sprintf(ident, "%s$%u", trigger->tgname, rel->rd_id);
plan = find_plan(ident, &FPlans, &nFPlans);

before line 483 which is:
ret = SPI_execp(plan->splan[r], kvals, NULL, tcount);

Here is the script to check it:
create table a (
a int4 primary key,
b int4,
c int4,
d int4
);

create table b (
a int4 primary key,
b int4,-- foreign key references a
c int4,
d int4
);

create table ab (
a int4 primary key,
b int4,
c int4,
d int4
);

create table bc (
a int4 primary key,
b int4,
c int4,
d int4
);

create function check_primary_key()
returns opaque
as '/home/bye2data/postgresql-6.3.2/contrib/spi/refint.so'
language 'c';

create function check_foreign_key()
returns opaque
as '/home/bye2data/postgresql-6.3.2/contrib/spi/refint.so'
language 'c';

create trigger t1 before delete or update
on a for each row execute procedure
check_foreign_key(1,'cascade','a','b','b');

create trigger t2 before insert or update
on b for each row execute procedure
check_primary_key('b','a','a');

create trigger t3 before delete or update
on ab for each row execute procedure
check_foreign_key(2,'cascade','a','a','a','bc','a');

create trigger t4 before insert or update
on a for each row execute procedure
check_primary_key('a','ab','a');

create trigger t5 before insert or update
on bc for each row execute procedure
check_primary_key('a','ab','a');

insert into ab values(1,1,1,1);
insert into bc values (1,1,1,1);
insert into a values (1,1,1,1);
delete from ab where a=1;

Sorry for the confusing names of tables and attributes.

- Anand.

Responses

Browse pgsql-general by date

  From Date Subject
Next Message Taral 1998-10-02 22:04:24 RE: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL] CNF vs. DNF)
Previous Message Bruce Momjian 1998-10-02 21:57:15 Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL] CNF vs. DNF)

Browse pgsql-hackers by date

  From Date Subject
Next Message Bruce Momjian 1998-10-02 21:58:00 Open 6.4 items
Previous Message Bruce Momjian 1998-10-02 21:57:15 Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL] CNF vs. DNF)