Stephan Schroevers

@delanym that could be implemented using a flag. But I'm not an Error Prone maintainer, so I'm not sure such a PR would be accepted. Let's see whether any of...

Since the nullability analysis isn't perfect, perhaps we should emit both suggested fixes in either case, but in a different order. Just like in [`FluxFlatMapUsage`](https://github.com/PicnicSupermarket/error-prone-support/blob/d45682143dc3c1b032c7a795415112bf3e4a175c/error-prone-contrib/src/main/java/tech/picnic/errorprone/bugpatterns/FluxFlatMapUsage.java#L92-L101).

Hey @BLasan! I believe so; I don't readily find code that would already cover this. As mentioned, we worry a bit that Refaster solutions to this issue may produce too...

> So the Windows build machines don't have this version yet. This is _still_ the case. #964 should provide a solution, after which we can merge #434.

~We merged #434 instead; Renovate should close this PR tomorrow.~ Nope, there's of course still version 3.9.6, which we're not adopting until `mvnd` supports it.

Thanks for the feedback @msridhar! > Would it be possible for the check to not require these blank lines, and for the auto-fix not to insert them? I suppose we...

Ack! Then I'll look into adding this feature in the near future. :+1:

Hi @brunacunha! Good point about `serialVersionUID`; that's a special case which should be ignored by the hypothetical check described here, as it is [part of Java's serialization functionality](https://www.baeldung.com/java-serial-version-uid). Looking at...

I looked into this a few months ago, but for some reason this doesn't seem to work with Refaster as one would expect. My code is [here](https://github.com/PicnicSupermarket/error-prone-support/compare/sschroevers/test-refaster-with-protected-methods); perhaps I overlooked...

> The rewrite from `TestMode.AST_MATCH` to `TestMode.TEXT_MATCH` is not a behavior preserving one. Likewise for `.doTest()`.