Fix #15883: Interpret inner static object access as this access

[liufengyun]

Fix #15883: Interpret inner static object access as this access

[olhotak]

I can't figure out how this works. I mean before this PR, the initialization checker does not process top-level vals at all. Where in the patch does it add top-level code to the code that the initialization checker processes? I can't find it.

[liufengyun]

I can't figure out how this works. I mean before this PR, the initialization checker does not process top-level vals at all. Where in the patch does it add top-level code to the code that the initialization checker processes? I can't find it.

The top-level vals will be inside a synthesized top-level wrapper object, thus it's transparent to the init checker.

The problem with the bug is how we interpret O.f. By default, the checker returns Hot for top-level object O, and selecting a field on it will also give Hot. That's the reason why we miss the warning for the issue.

Now, we change the logic, in interpreting a static object O:

• If O is accessed inside the class O$, it's interpreted as O.this
• If O is inside the entry class, it goes through this selection from the entry class.

[olhotak]

LGTM