Think this is simple enough for a StarterBug? cc @xedin too

Yes, I think it makes sense to mark it as StarterBug, the problem is related to `PreCheckExpression`, we don't properly recognize all of the cases there when checking `InOutExpr` position.

@swift-ci create

Also agreed that it makes a good starter bug. If you're interested in fixing, as Pavel says, you'll want to start here, which is where the "Use of extraneous '&'" diag gets emitted before the expression is type checked.

The condition is checking for TupleExpr and ParenExpr, which are currently parent expressions of an argument, and then checking for the parent expression of that, which should be the expression making the call. You'd want to add a condition checking if that parent is a SubscriptExpr, and if so emit a custom diagnostic (adding the message to DiagnosticsSema.def, and using a call similar to diag::extraneous_address_of).

Note this won't handle obscure things like s1[takesPtr: (&i)], you'd need to keep walking up the parent chain to handle that (and that would also be a good opportunity to emit a better diagnostic for applies with inout arguments wrapped in parentheses, which we currently also reject with "Use of extraneous '&'").

I'm going to push a trivial change to make it so type-checker gets past PreCheckExpression in situations like this, but improving diagnostic to be aware of implicit conversions would be less trivial, logic in MiscDiagnostics would have to be adjusted to make it work.

Actually I think we should be able to diagnose this directly if we see call like that in `PreCheckExpression`.

Should be fixed by #20336

Please verify using next master snapshot.

Thanks Pavel!