Update to OpenJDK 15.

[raphw]

Use Java 15 in builds for most recent Java version.

[TimvdLippe]

Seems like we got some test failures :(

[raphw]

Needed to update Jacoco. Works now.

[raphw]

Strangely enough, it builds locally. Will investigate some more!

[raphw]

It's weird, if I build locally with the same command and Java version (build and major), I cannot reproduce this error...

[TimvdLippe]

I don't think I have Java 15 installed on any of my machines and not sure if I can get a hold of them. Looking at the test failures, many of them seem related to how we handle null with primitive matchers. Is there some way to check the changelog to see if we see any suspicious changes?

[raphw]

I ran the exact build version locally and it works, on a fresh clone even, also on Linux.

[Marcono1234]

I am seeing these failures for JDK 15 (build 15+36-1562) as well, but on Windows.
It looks like the cause for this is JDK-8233014 which enables "Helpful NullPointerExceptions" by default.

The tests which are failing apparently make assumptions about how NullPointerExceptions look like, which are now not true anymore:

• mockito/src/test/java/org/mockitousage/stubbing/StubbingWithDelegateTest.java

  Line 107 in 4d44781

  assertThat(e.toString()).doesNotContain("org.mockito");

  
  NullPointerException now mentions the name of the method which returned null:

  java.lang.NullPointerException: Cannot invoke "java.lang.Byte.byteValue()" because the return value of "org.mockitousage.IMethods.byteObjectReturningMethod()" is null

• mockito/src/test/java/org/mockitousage/matchers/InvalidUseOfMatchersTest.java

  Line 109 in 4d44781

  .hasMessage(null);

  
  Assumes that NullPointerException has no message.

• mockito/src/test/java/org/mockitousage/serialization/DeepStubsSerializableTest.java

  Line 85 in 4d44781

  assertThat(e).hasMessage(null);

  
  Assumes that NullPointerExceptions has no message.

[TimvdLippe]

Thanks for the great investigation @Marcono1234 Do you mind creating a PR with the necessary fixes to make it work on JDK 15?

[Marcono1234]

I fear that I am not familiar enough with Mockito to perform these changes while still keeping the intended functionality for the tests.
For StubbingWithDelegateTest it looks like this could be simplified to checking whether the returned Byte is null instead of provoking the NullPointerException, is that correct? I assume for DeepStubsSerializableTest it is similar, it might already suffice to check whether next() returned null.
But for InvalidUseOfMatchersTest maybe the only solution would be to drop the requirement that the message must be null.

[raphw]

The requirement does not seem meaningful to me, at least. I just fixed the tests to not require an empty message.

[codecov-io]

Codecov Report

Merging #2051 (b8bb0f0) into release/3.x (2e6c816) will not change coverage.
The diff coverage is n/a.

@@              Coverage Diff               @@
##             release/3.x    #2051   +/-   ##
==============================================
  Coverage          84.74%   84.74%           
  Complexity          2717     2717           
==============================================
  Files                325      325           
  Lines               8277     8277           
  Branches             989      989           
==============================================
  Hits                7014     7014           
  Misses               992      992           
  Partials             271      271           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2e6c816...b8bb0f0. Read the comment docs.

[Marcono1234]

I might not be familiar enough with Mockito, but wouldn't it be possible for StubbingWithDelegateTest and DeepStubsSerializableTest to not provoke the NullPointerException at all (as outlined above), but instead test for a null return value? That would probably make the tests more robust since they might otherwise catch unrelated NullPointerExceptions which are caused by bugs.

[Marcono1234]

Suggested change

	assertThatThrownBy(
	() ->
	when(deserialized_deep_stub.iterator().next().get(42))
	.thenReturn("no"))
	.isInstanceOf(NullPointerException.class);
	assertThat(deserialized_deep_stub.iterator().next()).isNull();

Due to Java expression evaluation order the enclosing when(...) and the following thenReturn(...) are not executed at all.

[Marcono1234]

Suggested change

	assertThatThrownBy(
	() -> {
	@SuppressWarnings("unused")
	byte b = methods.byteObjectReturningMethod();
	})
	.isInstanceOf(NullPointerException.class);
	assertThat(methods.byteObjectReturningMethod()).isNull();

NullPointerException here is not triggered by Mockito, but instead by unboxing Byte → byte so this can probably simplified to checking for a null return value.

[raphw]

Good point, I just adjusted the tests.

[TimvdLippe]

Test changes LGTM.