[cl-plus-ssl-devel] Re: FW: bugs in asdf-installable version from cl-plus-ssl-devel-owner at common-lisp.net

[David Lichteblau]

Quoting rpgoldman at sift.info (rpgoldman at sift.info):
> I wanted to asdf-install Edi Weitz' Drakma, so got cl-plus-ssl as a
> dependency.  I found that doing this had problems with CFFI.
> 
> In particular, there are a number of calls to cffi:null-ptr and
> cffi:null-ptr-p.  These two are not the exported functions.  AFAICT,
> null-pointer and null-pointer-p are the correct functions.

I am afraid that is fixed in CVS, but not in the release tarball.

This whole tarballing, gpg-singing and symlinking all just for
asdf-install is not my favourite kind of free software work, so I don't
do it often enough.

If you or anyone else would like to take over, I would be glad to have
you added to the cl-plus-ssl project...

> I have then found the following other oddities.  I believe that the
> PTR error messages are artifacts of with-pointer-to-vector-data.....

Which Lisp is that?  I believe they are included for SBCL, not included
for CLisp, ...  When testing, I added a copying fallback implementation
as indicated in the CFFI proposal as needed.

But I am aware this is not a stable API.  CL+SSL has to be fixed here,
of course.  My excuse, other than general laziness, is that this API
might or might not get replaced by something else in CFFI, and I was
kind of waiting to find out what the CFFI developers end up offering
before changing CL+SSL accordingly.  See the thread on cffi-devel.[1][2]

And on Allegro, CL+SSL doesn't work for unknown reasons.  But Allegro
has its own OpenSSL wrapper anyway.  For drakma, it would probably be
best to just do #+allegro socket:make-ssl-client-stream instead.


d.
[1] http://common-lisp.net/pipermail/cffi-devel/2006-January/000499.html
[2] http://common-lisp.net/pipermail/cffi-devel/2006-January/000506.html