PreparedStatement loses negative sign?!?: setInt()/setObject()

Question

I have 3 sql statements in my application (1 select/1 update and 1 insert) where I've noticed that after some time of working properly, all of a sudden these statements stop handling negative integers and instead process them as positive integers.
i.e. the select statement would return result as if the where clause parameter was positive instead of negative, the insert statement would actually insert the number as a positive, etc... The only way for me to fix the problem is the restart the iPlanet application server (connecting to an oracle 8.1.7 db using the oracle 8.1.7 jdbc)
I log the parameters from the PreparedStatement setInt() and setObject() comments just before execution and clearly the values are logged as negatives.
e.g. (not the actual source code, but the same idea)
PreparedStatement pstmt =
conn.prepareStatement ("insert into testTable (anInt, name) values (?,?)");
int anInt = -1;
String name = "test";
System.out.println("Before insert: " + anInt + " " + name); // outputs Before insert: -1 test
pstmt.setInt(1, anInt);
pstmt.setString (2, name);
pstmt.execute ();
====================
select * testTable;
anInt name
===========
1 test
Any ideas?
regards,
Rob

Answer 1

Did the author of the original posting (Rob Terro) ever learn anything about the nature of this problem and the appropriate fix?
We are observing what sounds like the same behavior: after some time of working properly, all of the sudden stored procedures to which we pass negative parameters (in particular, -1) in Java start receiving these as positive parameters at the database.
We are using the Oracle JDBC 8.1.6 thin driver, running on Solaris 5.9 generic_112233-03, using the Sun HotSpot JVM, version 1.3.1_07.
We had been running with this configuration for some time without apparent problems. Recently, we added to the command line the JVM "-server -Xconcurrentio" options. These greatly improved performance, but apparently lead to this problem of losing negative signs when talking to Oracle. However, the problem appears erratically, so we are not 100% certain that these options cause the problem. Also, we would ideally like to hang on to the performance gains produced by using these options.
How does this configuration compare to the configuration used by the original poster when he encountered this problem?
Does anyone have insight into the possible nature of this problem?