Subj : Re: [SpiderMonkey] JS_DefineObject returns null after moving to newest
To   : Brian Barnes <ggadwa@voyager.net>
From : Brendan Eich <brendan@meer.net>
Date : Sat Jul 19 2003 10:16 am

Brian Barnes wrote:

> It's not throwing an error (should have mentioned that), but I 
> figured  out what it is.  This is behavior I depended on in 1.4 to 
> make objects  you couldn't add properties to, so if you had mistyped 
> the name of a  property you'd get a error instead of adding a new 
> property.
>
> I did this by hooking up a resolving function, i.e.:
>
> JSClass            dim3_class={"dim3_class",JSCLASS_HAS_PRIVATE,
>                             
> JS_PropertyStub,JS_PropertyStub,JS_PropertyStub,JS_PropertyStub,
>                              
> JS_EnumerateStub,scripts_resolve_object,JS_ConvertStub,JS_FinalizeStub}; 


Ah, sorry I missed that resolve hook -- but I still wouldn't have known 
that was the problem, because there was no intentional change to cause 
defining a property (which should blow away any pre-existing property, 
unless you are passing in JSPROP_GETTER or JSPROP_SETTER) to lookup 
(which resolves).

> scripts_resolve_object just always threw an error, but for me, every  
> ambiguous property was an error.  Before it wasn't getting called on  
> JS_DefineObject, but now it is.  So I have to think of some other way  
> around this -- I probably shouldn't have depended on this, anyway.
>
> I get the idea my little hack work isn't the proper way to do this :) 


Can you do me a favor and show me the stack backtrace when 
scripts_resolve_object is called indirectly from your (doomed to fail, 
with the new version of the engine) call to JS_DefineProperty?

/be

.