[armedbear-devel] [PATCH] Problems with new JavaObject.javaInstance(Class c) and JSS

[Alessio Stalla]

On Wed, Aug 5, 2009 at 7:39 PM, Alessio Stalla<alessiostalla at gmail.com> wrote:
> On Wed, Aug 5, 2009 at 6:25 PM, Mark Evenson<evenson at panix.com> wrote:
>> On 8/5/09 5:33 PM, Alessio Stalla wrote:
>> […]
>>>
>>> Also, I have a question about your patch: do you have a specific
>>> reason to handle null by always signaling an error? In principle, null
>>> is assignable to any non-primitive type, so the type error ought to be
>>> signaled only for primitive types, imho.
>>
>> No good reason: I was just trying to refactor the logic, but misunderstood
>> your intent (that "null is assignable to any non-primitive type") so it is
>> incorrect as it stands.
>
> Ok. Looking at my commit again, I see that actually the change to
> javaInstance() is not particularly useful for the inspection of Java
> objects; it just makes code a little bit shorter in
> getInspectedFields(), where a type check is not done explicitly,
> relying on javaInstance() to do it. So the safest thing to do is to
> revert my change and do the type check, until we sort out precisely
> how to correctly implement javaInstance(Class). I also believe this
> method should be generified, in order to avoid an unnecessary cast of
> its return value from Object to the desired type.

Hopefully it's fixed on trunk now, sorry for the inconvenience. In
passing I also discovered & fixed a minor bug in inspection of Java
arrays - Lisp objects inside the array were incorrectly wrapped in a
JavaObject.

Bye,
Ale