On Fri, 1 Sep 2023 14:33:52 GMT, Andy Goryachev <ango...@openjdk.org> wrote:
>> Added automated test for 8262518:SwingNode.setContent does not close
>> previous content, resulting in memory leak
>
> tests/system/src/test/java/test/javafx/embed/swing/SwingNodeContentMemoryLeakTest.java
> line 94:
>
>> 92: //Lets throw in a little sleep so we can read the output
>> 93: try {
>> 94: Thread.sleep(100);
>
> I would suggest to use random delay here (random.nextInt(100)).
> But make sure to set a random seed at the beginning and print it so it can be
> reproduced. Although in this test, the outcome depends on many things that
> the test has no control over.
> What do you think?
No, let's not. While there may be some rare cases where randomness adds
something useful to the test, automated tests should be predictable.
> tests/system/src/test/java/test/javafx/embed/swing/SwingNodeContentMemoryLeakTest.java
> line 116:
>
>> 114: System.out.println("iteration " + count + " Panels in
>> memory: "
>> 115: + panelCount + " fail "
>> + fail);
>> 116: assertFalse(fail > 2);
>
> I actually seen the number to shoot up to 38 (albeit with Thread.sleep(1)).
> I wonder if there is a better way to detect failure? May be the fact that
> the 'panelCount` is ever increasing is the sign of failure, whereas if at
> least one time it's dropping we have succeeded.
> Or perhaps look at the count when the test complete and fail if panelCount <
> (attempts / 2) or something
I recommend using `JMemoryBuddy`, which is what we use for all of our newer
memory leak tests.
-------------
PR Review Comment: https://git.openjdk.org/jfx/pull/1228#discussion_r1313120071
PR Review Comment: https://git.openjdk.org/jfx/pull/1228#discussion_r1313120559
