Suboptimal JS code with full DCE

[Jozef Chúťka]

I noted this issue while running closure complier for haxe generated code:

class Test {
    static var a(default, set):Int = 0;
   
    static function main() {
        a--;
    }
   
    static function set_a(value:Int):Int {
        return a = value;
    }
}

would produce ( http://try.haxe.org/#f002d ) :

(function (console) { "use strict";
var Test = function() { };
Test.main = function() {
 var _g = Test;
 var _g1 = _g.a;
 _g.set_a(_g1 - 1);
 _g1;
};
Test.set_a = function(value) {
 return Test.a = value;
};
Test.a = 0;
Test.main();
})(typeof console != "undefined" ? console : {log:function(){}});

google closure complains:

JSC_USELESS_CODE: Suspicious code. This code lacks side-effects. Is there a bug? at line 7 character 1
 _g1;
 ^

The problem is the extra line with _g1; , this will get removed with analyzer but I thought the compiler itself should not produce it anyway.

[Simon Krajewski]

Yes, this actually gave us trouble with our own diagnostics checker as well: https://github.com/HaxeFoundation/haxe/issues/5581

I closed this issue after ignoring non-user variables, but now that I think about it again it might indeed be prudent to not generate the value in the first place. This requires keeping track of whether or not we expect a value when typing the unop.

Simon