End of file error when running an extremely specific SQL call

Question

When I run the query below through my VB.NET app with certain table and column names as parameters on an Oracle 10.1 database (Connection.ServerVersion reports it as 10.1.0.4.0), it fails and throws the following error: Oracle.DataAccess.Client.OracleException: ORA-03113: end-of-file on communication channel. The query does not fail for most table/column names--only a few. The query does not fail with any set of parameters when I run it directly through PL/SQL Developer, nor does it fail with any set of parameters when I run it through VB against a 10.2 database. I'm using ODP.NET 10.2.0.2.20.
Here's the SQL:
SELECT search_condition check_constraint, cols.constraint_name
FROM user_constraints cons, user_cons_columns cols
WHERE constraint_type = 'C'
AND cols.owner = cons.owner
AND cols.table_name = cons.table_name
AND cols.constraint_name = cons.constraint_name
AND cols.table_name = UPPER(:tableName)
AND cols.column_name = UPPER(:columnName)But the weirdest part is that the query always works if I make any one of these changes:
*Change the WHERE line to: "WHERE cons.constraint_type = 'C'".
*Get rid of the UPPERs and pass in upper-case parameters.
*Remove one of the parameters
*Remove "constraint_type = 'C'"
*Change the last line to 'EXPECTED_VALUE' = UPPER(:columnName)
Is there something special about the Query of Death that I wrote? Is anyone even able to replicate this error, iterating in VB over all the columns in a good-sized database? I've never seen anything like this before.
Thanks in advance for your help,
-Justin

Answer 1

Hi Justin,
I tried to reproduce your complaint, but wasnt able to. I didnt have a 10104 db to test against though, but I did test against 10105, and 10202, using the following dumb table:
create table constr_test ( c1 number);
alter table constr_test add constraint const_1 check (c1 in (1,2,null));
insert into constr_test values (3);
commit;
Frequently a 3113 is the result of an ora-600 or ora-7445 on the database. The fact that it doesnt occur for you on 10.2 db leads me to believe that might be the case here. Have you checked your alert logs?
Cheers,
Greg