[BUG] orphaned function - Mailing list pgsql-hackers

From Drouvot, Bertrand
Subject [BUG] orphaned function
Date
Msg-id 5a9daaae-5538-b209-6279-e903c3ea2157@amazon.com
Whole thread Raw
Responses Re: [BUG] orphaned function
List pgsql-hackers

Hi hackers,

here is a scenario that produces an orphaned function (means it does not belong to any namespace):

Session 1:

postgres=# create schema tobeorph;
CREATE SCHEMA

postgres=# BEGIN;
BEGIN

postgres=*# CREATE OR REPLACE FUNCTION tobeorph.bdttime() RETURNS TIMESTAMP AS $$
DECLARE  outTS TIMESTAMP;
BEGIN  perform pg_sleep(10);  RETURN CURRENT_TIMESTAMP;
END;
$$ LANGUAGE plpgsql volatile;
CREATE FUNCTION

=> Don't commit

Session 2: 

postgres=# drop schema tobeorph;
DROP SCHEMA

Session 1:

postgres=*# END;
COMMIT

Function is orphaned:
postgres=# \df *.bdttime               List of functions Schema | Name  |   Result data type    | Argument data types | Type
--------+---------+-----------------------------+---------------------+------    | bdttime | timestamp without time zone |           | func
(1 row)

It appears that an AccessShareLock on the target namespace is not acquired when the function is being created (while, for example, it is acquired when creating a relation (through RangeVarGetAndCheckCreationNamespace())).

While CreateFunction() could lock the object the same way RangeVarGetAndCheckCreationNamespace() does, that would leave other objects that belong to namespaces unprotected. Locking the schema in QualifiedNameGetCreationNamespace() will protect those objects.

Please find attached a patch that adds a LockDatabaseObject() call in QualifiedNameGetCreationNamespace() to prevent such orphaned situations.

I will add this patch to the next commitfest. I look forward to your feedback.

Bertrand
Attachment

pgsql-hackers by date:

Previous
From: Andres Freund
Date:
Subject: Re: Printing LSN made easy
Next
From: Pavel Stehule
Date:
Subject: Re: proposal: unescape_text function