[display] nonsense field completion subject in nested switch

[Gama11]

class Main {
	static function main() {
		switch (nodeName) {
			case "p"
				switch (a) {}
		}
	}
}

When typing the : after the "p" here, Haxe returns field completion (?) with 0 items and this range as the subject:

"range": {
    "start": {
        "line": 3,
        "character": 3
    },
    "end": {
        "line": 4,
        "character": 17
    }
},

This then leads to some rather strange postfix completion:

When the inner switch is commented out, Haxe doesn't give a completion response at all (as expected, since completion was auto-triggered with : after a case).

Doesn't seem to be a regression (on the Haxe side at least). Original issue at vshaxe/vshaxe#337. Should be able to work around it by not allowing postfix completion if the completion was auto-triggered with a :.

[Simn]

Has nothing to do with nesting, the problem is this code in matcher.ml:

haxe/src/typing/matcher.ml

Lines 574 to 578 in 1ce52fa

	if ctx.is_display_file && DisplayPosition.display_position#enclosed_in p then begin match eo,eo_ast with
	| Some e,Some e_ast -> ignore(TyperDisplay.display_expr ctx e_ast e DKMarked with_type p)
	| None,None -> ignore(TyperDisplay.display_expr ctx (EBlock [],p) (mk (TBlock []) ctx.t.tvoid p) DKMarked with_type p)
	| _ -> assert false
	end;

This came from fixing #5656. I cannot reproduce that anymore after removing these lines though, so I guess I'll just optimistically do that.

[Simn]

Not sure how to test this because it depends on wasAutoTriggered: true.

[Simn]

Simplest test case is this:

class Main {
	static function main() {
		switch ("p") {
			case "p":
				"foo";
		}
	}
}

This gave string completion from the "foo" before my fix.