Re: CharSequence support

[Stuart Ballard]

Brian Jones wrote:
> 
> "Nic Ferrier" <address@hidden> writes:
> 
> > Hacking CNI is pretty easy compared to JNI, are there any guidelines
> > about how to make things portable between GCJ and Classpath?
> >
> > Do the GCJ people keep their CNI interfaces and link with abstracted
> > C modules which are also linked by Classpath's JNI?
> >
> > Or do you guys just not bother merging the native code?
> 
> To date, we haven't tackled it in a big way.  I've started, but I
> don't have the details worked out so I'm pecking at the problem as I
> have time.  Indeed, once the mechanics are worked out... what is
> common, and what is not, and how to compile either; we should be in
> good shape.

I just found the following from Paul Fisher on the mailing list archive
while looking for the messages referenced in another thread:

> The JNI/CNI issue can be temporarily postponed.  I suspect that we'll
> end up using my C++ layer, because it'll get the job done, and it's
> rather trivial to implement.  It's not the optimal solution, but it
> will permit us to generate JNI code from the majority of CNI code.

(reference: http://www.mail-archive.com/address@hidden/msg02085.html)

I couldn't find a description of this "C++ layer" or any information
about what exactly Paul was proposing. Paul, if you're still around,
could you elaborate on exactly what you had in mind here? From the brief
description in this paragraph, it sounds ideal: Hack CNI, get JNI out.

Anyone else have any thoughts on this? A solution that let us use CNI
code wholesale would certainly be a great thing for the merging process.

Stuart.