Pitest not filtering try-with-resources correctly in Java 11

[johnsterken]

Summary

Pitest is trying to mutate bytecode used for closing resources when using the try-with-resources in Java 11

Expected: No mutations on the generated parts of the bytecode, this behaviour works correctly in Java 8
Actual: Mutations happening on the generated parts in the bytecode used for closing the resources that are used in the try block

Steps to reproduce

• Check out the example project
• Run mvn clean test pitest:mutationCoverage
• Look at the pitest report, it shows mutations on the try block

Other info

Probably this is related to #689

Tested with OpenJDK

Java 11

openjdk version "11.0.7" 2020-04-14
OpenJDK Runtime Environment (build 11.0.7+10-post-Ubuntu-2ubuntu218.04)
OpenJDK 64-Bit Server VM (build 11.0.7+10-post-Ubuntu-2ubuntu218.04, mixed mode, sharing)

Java 8

openjdk version "1.8.0_252"
OpenJDK Runtime Environment (build 1.8.0_252-8u252-b09-1~18.04-b09)
OpenJDK 64-Bit Server VM (build 25.252-b09, mixed mode)

I also checked out the pitest repository locally. When I create a test in TryWithResourcesFilterTest with the Java 11 bytecode from my local JDK I don't see any mutation being filtered. When I run the same test on a Java 8 class file it does filter out mutations

java8_pit_report.zip
java11_pit_report.zip

[ketan2505]

I am using pitest results for quality checks. The same problem occured with me. I had to disable pitest from my project. I would like to see this working with Java 11 as well.

[TomerFi]

Same here.
Working with Java11, using try-with-resources, I always end up with a bunch of surviving mutants.
e.g. a mutant removing the close() statement inside the auto-generated finally block.

[hcoles]

Sorry for the delay fixing this folks, it is very annoying. Should have something out early next week.

[hcoles]

This has gone a little easier than I'd exepected, so I'll push a release out later today.

New filtering catches the issues in @johnsterken 's example, plus some sample real world projects. I expect there will be further issues with more complex examples, but current code seems to give 90% of the benefit.

[johnsterken]

Looking good! I just tried your 1.6.3 version in the project that triggered this issue. It's working like a charm. Thanks @hcoles!