117849 - NativeJavaMethod do not store "this" pointer.

Status: VERIFIED INVALID

(Rhino Graveyard :: Core, enhancement)

Description

[Yang Liu]

Try the following scenario:
java -classpath js.jar org.mozilla.javascript.shell.Main
js> x = new java.util.Vector()

js> y = x.addElement

js> y('10')

Then I get:
js: "<stdin>", line 3: Java method "addElement" was invoked with a 'this' value 
that was not a Java object.

I understand this is because the NativeJavaMethod do not store the "this" 
pointer in it, and expect the "thisObj" to be passed through the "call" 
function, while when I call "y('10')" the "thisObj" parameter is null.

I think maybe it can be enhanced to pass in the "this" pointer when 
constructing the NativeJavaMethod method. Since I believe the scenario I write 
is totally legal.

Thanks

Comment 1

[Phil Schwartau]

Confirming reported behavior on WinNT, JDK 1.3

Comment 2

[Phil Schwartau]

I don't know if this helps, but here is a similar example using
only JavaScript objects. The array value doesn't get added, but
there is no error: 

java org.mozilla.javascript.tools.shell.Main
js> var x = new Array();
js> x.toString(); // showing that it has no elements
js>
js> var y = x.push;
js> y('10');
1
js> x.toString(); // it still has no elements...
js>

Comment 3

[Norris Boyd]

I get the same behavior in SpiderMonkey as in additional comment #2.

Comment 4

[Norris Boyd]

ECMA specifies ways in which the 'this' pointer is resolved during a call. We
aren't at liberty to change this behavior. See "11.2.3 Function Calls". The key
here is the GetBase operation, which must resolve to the global object since
y is a property of the global object. 

Resolving as INVALID.

Comment 5

[Phil Schwartau]

Marking Verified -

Yang Liu, thank you for this report -

Comment 6

[Igor Bukanov]

Targeting as resolved against 1.5R3