Re: [Jython-dev] ctypes...

[John S. <jo...@sz...>]

On Wed, Apr 1, 2009 at 12:21 PM, John Szakmeister <jo...@sz...> wrote:
[snip]
> So, I'm at an impasse.  Despite Samuele's insistence, I'm thinking
> that implementing ctypes directly is more feasible, assuming that it
> really doesn't allow you to craft abitrary pointers.  Additionally, we
> could get more tests around ctypes that makes it's api more fully
> specified.  Thoughts?  Oh, and where should I put what I have?

I knew I should have waited to send that email.  You can indeed craft
an arbitrary pointer from the public api:

>>> import ctypes
>>> p = ctypes.c_voidp(1)
>>> dir(ctypes)
>>> cp = ctypes.cast(p, ctypes.c_char_p)
>>> cp.value
Bus error

So, we probably either need to get the jna guys open the door for us
to do this, or create our own set of bindings by bringing in libffi
directly via jna (IOW, use jna only to get us access to libffi).

This is definitely a big task though.

-John