By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

Issue Link, if available

#72

Description

The bug is in the following lines:

if (phaser.getPhase() == ExecutionPhase.RUNNING.getValue()) {
    // Operation not done yet
    phaser.register();

After checking the phaser status (is RUNNING) and before a new party is registered, the phaser could have changed its status (advanced to the next phase). This is a typical TOCTOU bug

This bug is reproduced with an additional delay between them:

if (phaser.getPhase() == ExecutionPhase.RUNNING.getValue()) {
    try {
        Thread.sleep(2000);
    } catch (InterruptedException ignored) {
    }
    // Operation not done yet
    phaser.register();

With the delay, some test cases will hang forever because the phaser of the operation is registered but never notified (arrived).

The proposed fix here is first calling register to lock the phase state, then check the state and deregister immediately if it's completed

phaser.register();
if (phaser.getPhase() != ExecutionPhase.RUNNING.getValue()) {
    phaser.deregister();
} else {
    ...
}

Demo/Screenshots

added 2 seconds delays in different places in waitForOperationCompletion and the tests always succeeded (but slower)

Checklist

• I have filled out every section of the PR template
• I have thoroughly tested this change

Testing

Unit Tests

Have unit tests been written for these changes? Unable to cover the race condition in a unit test

Integration Tests

Have integration tests been written for these changes? Existing integration tests should run smoother than before

Examples

Has a new example been added for the change? (if applicable)