Broken synchronization for event handling in ClientPooledConnection40

• Key: DERBY-3308
• Type: Bug
• Status: Closed
• Resolution: Fixed
• Priority: Minor
• Assignee: Kristian Waagan
• Reporter: Kristian Waagan
• Component/s: JDBC, Network Client
• Affects Version/s: 10.2.2.0, 10.3.2.1, 10.4.1.3
• Fix Version/s: 10.2.2.1, 10.3.3.0, 10.4.1.3

Description

Access to the the list of event listeners is not synchronized properly.

Taken the rather infrequent use of the relevant methods and the small critical sections, I mean it is sufficient to add synchronization to all the methods that access it at the method level. The same approach is taken in ClientPooledConnection (although not followed through consistently, some of the methods are unsynchronized).

Comments

[Kristian Waagan]

'derby-3308-1a-eventlisteners_synchronization.diff' synchronizes all methods accessing the list (a java.util.Vector) of statement event listeners.

I ran derbyall/suites.All without errors. Patch ready for review.

The patch looks fine to me.

Since the list is now guarded by synchronization on the pooled connection, couldn't it be changed from Vector to ArrayList? (Should be safe to do with listeners_ in the parent class as well.)

[Kristian Waagan]

Thanks for commenting on the patch Knut Anders.

I have incorporated you suggestion about replacing Vector with ArrayList in revision 1b.

Regarding the suggested changes in the parent class, there are some methods that are synchronized and some that are not. A patch could be attached to the subtask of this issue. Feel free to add it, or maybe I will if I get around to it.

[Kristian Waagan]

Committed 'derby-3308-1b-eventlisteners_synchronization.diff' to trunk with revision 614536.

I don't expect there will be more work on this issue, will close in a few days.

[Kristian Waagan]

Backported the fix (revision 614536):
10.2 -> revision 618639
10.3 -> revision 618640

I ran all tests for both branches without errors, and both merges were clean.
Updated the fix versions accordingly.

[Kristian Waagan]

Haven't seen any problems with the patch in the regression tests. Closing issue.