Suppresses RLC non-final field overwrite warning for safe constructor field initialization

[iamsanjaymalakar]

Summary by CodeRabbit

• New Features

  • Enhanced resource leak checker to suppress false-positive warnings when a private field is assigned for the first time within a constructor.

• Tests

  • Added comprehensive test suite covering various constructor initialization scenarios, including inline initializers, instance initializers, method calls, and conditional branching.

[kelloggm]

Overall looks good, just a few minor comments. I'd like to take one more look before we merge, though.

[mernst]

@iamsanjaymalakar The typecheck job is not passing.

[kelloggm]

LGTM, I don't need to review again. Please address these comments and then request a review from Mike.

[mernst]

@iamsanjaymalakar Martin asked you to assign me when you had addressed his comments. Have you addressed them?

[iamsanjaymalakar]

@iamsanjaymalakar Martin asked you to assign me when you had addressed his comments. Have you addressed them?

Yes, I have. You can take a look.

[mernst]

Please clarify some documentation, and update the code if necessary. (I didn't read the code because there is no point in doing so until the documentation is clear.)

[iamsanjaymalakar]

@mernst CircleCI build is failing. I think it is not picking up the correct jdk annotations branch.

[mernst]

@iamsanjaymalakar I have fixed the CI problem.

[iamsanjaymalakar]

@iamsanjaymalakar I have fixed the CI problem.

Thanks. It's ready for review then.

[mernst]

@iamsanjaymalakar Could you take a look at "ResourceLeakFirstInitConstructorTest" which is failing in CI? Thanks.

[iamsanjaymalakar]

@iamsanjaymalakar Could you take a look at "ResourceLeakFirstInitConstructorTest" which is failing in CI? Thanks.

@mernst the test failure is happenning due to the same CI not picking up the latest code changes to the iamsanjaymalakar/jdk repo of 7049-dev branch. There in the latest commit I've added @SideEffectFree annotations to the FileInputStream's constructors.

In my local workspace, when I checkout to the typetools/jdks master branch I see the exact test failures in ResourceLeakFirstInitConstructorTest. And after switching to jdk's 7049-dev branch passes them. I'm seeing the exact same errors in the CI. So, I'm pretty confident that the new stubs not being picked up is the cause of the CI failures.

[mernst]

Is it a bug if this method is ever called? If so, I suggest replacing its body with a throw statement.

[mernst]

What if both REASSIGNMENT and FIRST_ASSIGNMENT appear in the list? For a switch statement, we would prefer to return FIRST_ASSIGNMENT. Maybe it would help to explain in what circumstances this method is called.