sxemacs-devel
[Top] [All Lists]

Re: Do we need to re-evaluate?

From: Zajcev Evgeny <zevlg@xxxxxxxxx>
Subject: Re: Do we need to re-evaluate?
Date: Fri, 12 Nov 2004 09:19:40 +0300
User-agent: Gnus/5.1002 (Gnus v5.10.2) XEmacs/21.4 (Reasonable Discussion, berkeley-unix)
Steve Youngs <steve@xxxxxxxxxxx> writes:

>
> Well considering that DSO is already there (even in 21.4) I disagree.
> XEmacs is not likely to drop DSO any time soon so we have to support
> it, especially as we'll do a much better job of it anyway.
>
> No, this doesn't mean that FFI is out.  It means that we either have
> both (at least initially) or just FFI if it can load the XEmacs
> DSO's.

DSOs can be easily loaded.  Main difference from FFI module is that
DSO will not have any FFI arguments translations, because they are
already translated by means of DEFUN.  Another difference is that DSO
will need to do job to exporting/kill its symbols, functions, etc
to/from elisp for itself (currently `load-module' does that job).  But
it is easily achieved using _init()/_fini() functions controled by
dlopen()/dlclose().  However i'm not awared about situation for M$,
but who cares about M$? :)

Also, to be fully compatible it is possible to create special
`ffi-load-module-dso' function, which will do most of the job what
`load-module' do.

-- 
lg

<Prev in Thread] Current Thread [Next in Thread>