Re: [cp-patches] RFC: New JCL helper functions for RawData handling

[Chris Burdess]

Tom Tromey wrote:
> Guilhem> +      rawDataClass = (*env)->FindClass (env, 
> "gnu/classpath/RawData64");
>
> > > At FOSDEM we talked about renaming these ... want to take it on?  :-)
>
> Guilhem> Could you be more specific ? A link to the discussion ? I may
> Guilhem> take it on but I need to know what was decided. ;-)
>
> Maybe the discussion never got written down.  Or maybe I just made it
> all up :-)
>
> libgcj had a class named RawData before Classpath, but with different
> semantics.  libgcj's RawData is not a wrapper for a pointer but is
> instead a pointer itself, and it is specially treated by the compiler
> and the garbage collector.
>
> I've always thought it was confusing that Classpath had a class with
> the same name but different semantics.

Yes; it was my fault, I introduced this as part of porting libxmlj to 
64-bit architectures - not that I have any experience in this at all 
but it was the only thing I tried that worked. I did say at FOSDEM that 
it was a bad name choice. If the libgcj way is better we should use 
that instead since as I understand it it would save a layer of 
indirection, but perhaps it needs VM-specific support to work? Failing 
that, just renaming the class Pointer or PointerWrapper would also be 
better than the current state of affairs.
--
Chris Burdess