Normative: Make HasCallInTailPosition for ImportCall always return false

[nicolo-ribaudo]

HasCallInTailPosition is currently implicitly defined on ImportCall via the chain rule: HasCallInTailPosition of import( xxx ) is HasCallInTailPosition of xxx. This is wrong, because xxx is not in tail position.

This PR explicitly defines it to be false.

For what is worth, JSC matches this PR (we can't talk about "web reality" in this case, it's just a spec bug):

[jmdyck]

ImportCall has a single RHS, and that has a single non-terminal (AssignmentExpression), so the chain rule handles it. (HasCallInTailPosition is explicitly defined on AssignmentExpression.)

[nicolo-ribaudo]

Applying the chain rule would return true for foo() in this case, but it's not a tail call:

function fn() {
  return import(foo());
}

[jmdyck]

Ah, so the problem is not that HasCallInTailPosition isn't defined for ImportCall, but rather that (via the chain rule) it's defined incorrectly?

[nicolo-ribaudo]

I edited the PR description. My original flow of thoughts was "why isn't this defined to be false", but I was thinking more about the part about it not being explicitly defined 😅

[bakkot]

LGTM.

This should be landed as normative, but does not require plenary approval because it clearly fixes a bug so that it will match committee intent.

[nicolo-ribaudo]

@ljharb I updated the PR title to be more descriptive, if you want to use it as the commit message