Re: Bug in RI

At 10:18 21-9-00 -0700, Stephan Szabo wrote:
>Odd, it looks like it had trouble doing the heap_openr
>on the relation, although I don't immediately see why...
>
>What does
> select * from pg_trigger where
> tgconstrname='fk_folders__parent'
>give you?

First it didn't give me anything (0 rows). After I recreated the constraint
triggers:

CREATE CONSTRAINT TRIGGER "fk_folders__parent" AFTER INSERT OR UPDATE ON
"folders" NOT DEFERRABLE INITIALLY IMMEDIATE FOR EACH ROW EXECUTE
PROCEDURE "RI_FKey_check_ins" ('fk_folders__parent', 'folders', 'folders',
'UNSPECIFIED', 'parent', 'nr');

CREATE CONSTRAINT TRIGGER "fk_folders__parent" AFTER DELETE ON
"folders" NOT DEFERRABLE INITIALLY IMMEDIATE FOR EACH ROW EXECUTE
PROCEDURE "RI_FKey_noaction_del" ('fk_folders__parent', 'folders',
'folders', 'UNSPECIFIED', 'parent', 'nr');

CREATE CONSTRAINT TRIGGER "fk_folders__parent" AFTER UPDATE ON
"folders" NOT DEFERRABLE INITIALLY IMMEDIATE FOR EACH ROW EXECUTE
PROCEDURE "RI_FKey_noaction_upd" ('fk_folders__parent', 'folders',
'folders', 'UNSPECIFIED', 'parent', 'nr');

the above query returned three rows:

jeroenv=> select * from pg_trigger where tgconstrname='fk_folders__parent' ;
tgrelid | tgname | tgfoid | tgtype | tgenabled |
tgisconstr
aint | tgconstrname | tgconstrrelid | tgdeferrable | tginitdeferred |
tgna
rgs | tgattr | tgargs

---------+----------------------------+--------+--------+-----------+-----------
-----+--------------------+---------------+--------------+----------------+-----
----+--------+------------------------------------------------------------------
-----------
20152 | RI_ConstraintTrigger_21856 | 1644 | 21 | t | t
| fk_folders__parent | 0 | f | f |
6 | |
fk_folders__parent\000folders\000folders\000UNSPECIFIED\000parent
\000nr\000
20152 | RI_ConstraintTrigger_21858 | 1654 | 9 | t | t
| fk_folders__parent | 0 | f | f |
6 | |
fk_folders__parent\000folders\000folders\000UNSPECIFIED\000parent
\000nr\000
20152 | RI_ConstraintTrigger_21860 | 1655 | 17 | t | t
| fk_folders__parent | 0 | f | f |
6 | |
fk_folders__parent\000folders\000folders\000UNSPECIFIED\000parent
\000nr\000
(3 rows)

But the same query (update folders set title='Sitemap' where nr=43) still
crashes the backend at exactly the same spot.

So, still no clue.

Thanks,

Jeroen