False positive and maybe-false negative with cascade_invocations with cascaded invocations on difference references

[Cyp]

Describe the issue
False positive with cascade_invocations on different references.

Also, a bonus maybe-false negative with uncascaded invocations on equal references, if allowed to assume that the getter's return value doesn't change.

To Reproduce

const a = 1;
const b = 2;
a.bitLength
  ..abs()
  ..ceil();
b.bitLength  // “Cascade consecutive method invocations on the same reference.”, false positive.
  ..abs()
  ..ceil();
b.bitLength.abs();  // No lint, maybe-false negative.
b.bitLength.ceil();  // No lint, maybe-false negative (if assuming that `abs()` didn't change the value of `b.bitLength`).

Expected behavior
No lint (although in this very specific example, a lint about abs and ceil having no side effects would be reasonable).

Additional context
The linter seems to think a.something and b.something are the same object, which is not in general even likely to be true.

[srawlins]

I fixed a very very related bug in #57855 many moons ago. In that fix, I added a TODO which is basically the "false negative" half of your bug.

I just addressed the "false positive" half of this bug here: dart-lang/linter#3713.

[karabanovbs]

any updates?

[srawlins]

I did try to fix this and found it extremely difficult. I mean to chat with @bwilkerson about it.

[srawlins]

Quick drive by: I see in the tests for join_return_with_assignment that they may have a mechanism for this in the impl there. The unJoinablePropertyAccessReturn and joinablePropertyAccessReturn test cases demonstrate this.