deferred constraints failing on commit - Mailing list pgsql-sql

From Michael Richards
Subject deferred constraints failing on commit
Date
Msg-id 3A6418E0.0000CB.62512@frodo.searchcanada.ca
Whole thread Raw
Responses Re: deferred constraints failing on commit  (Stephan Szabo <sszabo@megazone23.bigpanda.com>)
List pgsql-sql
Hi.

I'm having trouble with committing a transaction. Intuitively it 
should work but does not.

I've got a table with 2 foreign keys, minrev and maxrev. They refer 
to a the revisionid value in another table. I need to update the 
second table to reduce the revisionid, then insert a new row. At the 
end of this all the keys match up yet the commit fails.

urdr=> begin;
BEGIN
urdr=> update objects set revisionid=2 where id=2 and 
revisionid=99999999;                                                                                
UPDATE 1
urdr=> insert into objects 
(id,typeid,repositoryid,parentid,deleted,permissions,revisionid,name) 
values (2,2,1,NULL,'f',NULL,99999999,'test.sql');         
INSERT 246107 1
urdr=> select id,revisionid from objects;                             id | revisionid 
----+------------ 1 |   99999999 2 |          1 2 |          2 2 |   99999999
(4 rows)
urdr=> select * from objcatalog ;objectid | repositoryid |  minrev  |  maxrev  |   key    |  data    
----------+--------------+----------+----------+----------+----------       2 |            1 | 99999999 | 99999999 |
mimetype|text/plain
 
(1 row)

urdr=> commit;
ERROR:  <unnamed> referential integrity violation - key in objects 
still referenced from objcatalog

At commit all the keys check out properly. minrev and maxrev both 
point to the same revisionid in the row we just inserted.

Is this a bug or me just misreading how things should work again?

-Michael
_________________________________________________________________    http://fastmail.ca/ - Fast Free Web Email for
Canadians

pgsql-sql by date:

Previous
From: Johann Spies
Date:
Subject: Re: pg_dump error - further information
Next
From: Jan Wieck
Date:
Subject: Re: Question #4 about PL/PGSQL