Can't instantiate type when a parent method uses a non public type in signature about byte-buddy

[raphw]

Hi, I already identified the actual issue. Refactoring gone wrong. I fixed the problem and just released version 0.6.10 where the visibility is appropriately resolved.

[bric3]

ok, great :)

[bric3]

It indeed fixes the issue :)

[raphw]

Great! This time I actually cloned Mockito and tested it and then reproduced the problem in my tests. I think I knew the issue and did not test it properly. Moved to fast.

I really hope there will not be too many of these corner cases coming up to the surface. Mockito is the first large-scale user of Byte Buddy, there will probably be one or two things I have not thought off.

[bric3]

No problem this is no small piece of development you crafted.

I'm excited by possibilities now :)
Just need time to work on it.

Also I think there's another issue with ByteBuddy or the way Mockito uses ByteBuddy.
The test in this commit hightlight an issue some are facing : bric3/mockito@7f2beb8

I didn't have time to dive in here yet, this probably have to do with a feature add in mockito lately that can generated partial mocks, so mocked types may be instantiated with the constructor instead of objenesis. So not sure it's a bytebuddy issue.

Must go now. Cheers

[raphw]

When you mock a class, you delegate all invocations to an instance field:

.method(any()).intercept(toInstanceField(MockMethodInterceptor.class, "mockitoInterceptor")
       .filter(isDeclaredBy(MockMethodInterceptor.class)))

If the field is not yet set (during a constructor call), a null pointer exception is thrown.

To avoid this, you can for example delegate to a static interceptor which then checks if the interceptor is already set and only delegates a method if the interceptor is already set: https://gist.github.com/raphw/8789f5ef08240008efc8

[bric3]

Yeah that's somehow what I suspected. Thanks for the tip.