Make pastDateExpression test more stable

[snuyanzin]

No description provided.

[kingthorin]

Is this a timezone issue?

[snuyanzin]

i suspect kind of NTP...
there are 3 times when new Date() is called via datafaker's api and then a check that second is between first and last...

[bodiam]

You triggered a new build?

Seems the code isn't completely repeatable:

  ["name.nameWithMiddle"="Gavin Durgan Bauch Jr."]
but the following map entries had different values:
  ["name.nameWithMiddle"="Gavin Durgan Bauch and Sons" (expected: "Gavin Durgan Bauch Jr.")]

[snuyanzin]

yes, somehow it appeared today not only for this PR

[bodiam]

Well, at least it's nice that it's consistently failing...

[bodiam]

I just tried it locally, with Java 20 on a Mac, and it seems to work fine.

[bodiam]

Correction: I can reproduce this quite often.

This never fails though:

    @RepeatedTest(10000)
    void name() {
        Faker faker1 = new Faker(new Random(0));
        Faker faker2 = new Faker(new Random(0));

        assertThat(faker1.name().nameWithMiddle()).isEqualTo(faker2.name().nameWithMiddle());
    }

But having a RepeatedTest(100) on the shouldCreateRepeatableValues makes it fail (in)consistently: around 30-60% of the tests fail.

Could it somehow be a caching or concurrency issue?

[snuyanzin]

looks like the issue is with concurrent tests
#837 should improve it

[bodiam]

I just went through the git history, and git d9c2cc262d29deb3552c9f49ab71edf0b3f6dc30 caused it. Before this, the tests worked fine, but after this one, the test started breaking (ironically called: "Make FakeValuesService thread safe")