Java: document nullness false negative as qltest

[aschackmull]

The correlated conditions logic doesn't properly account for loops.

[Copilot]

Pull Request Overview

This PR documents a false negative in the nullness analysis by adding two test cases to the Java test suite. The issue involves correlated conditions logic not properly accounting for loops, where control flow can iterate and create scenarios where null pointer exceptions should be detected but aren't.

• Adds test cases demonstrating false negative behavior in nullness analysis within loops
• Documents the difference between loop-scoped and iteration-scoped variable behavior

Comments suppressed due to low confidence (2)

java/ql/test/query-tests/Nullness/B.java:412

• [nitpick] The parameter name 'a' is not descriptive. Consider renaming it to 'conditions' or 'flags' to better indicate its purpose in testing correlated conditions.

  public void corrCondLoop1(boolean a[]) {

java/ql/test/query-tests/Nullness/B.java:426

• [nitpick] The parameter name 'a' is not descriptive. Consider renaming it to 'conditions' or 'flags' to be consistent with the test's purpose.

  public void corrCondLoop2(boolean a[]) {

[hvitved]

I would have expected some changes to a .expected file?

[aschackmull]

I would have expected some changes to a .expected file?

It's not an inline expectations test, and the two new cases are FN and TN, resp., so that means no .expected changes.

[hvitved]

It's not an inline expectations test, and the two new cases are FN and TN, resp., so that means no .expected changes.

Ah, yes, I mistakenly read the second example as a TP.