[clbuild-devel] RfC: SBCL with rlwrap and sb-aclrepl

[Tobias C. Rittweiler]

David Lichteblau <david at lichteblau.com> writes:

> Quoting Tobias C. Rittweiler (tcr at freebits.de):
> > I just had the idea that it might make sense if clbuild built its sbcl
> > such that it includes rlwrap and automatically uses sb-aclrepl.
> > 
> > How do people feel about this?
>
> I'm interested in usability improvements.
>
> My design choices would be sligthly different in the following respects:
>
>   - I've recently published a portable fork of sb-aclrepl, called PREPL
>     (for "Portable REPL").  I'd prefer using this portable library over
>     an SBCL contrib.  (The unportable bits are handled by a lower-level
>     library: conium, which is a feature-reduced fork of swank-backend
>     for use in hemlock and prepl).
>
>   - Although rlwrap is a nice trick, there is also linedit, which
>     provides the same capabilities written in portable Lisp, which I
>     think is more interesting.
>
>   - The SBCL built my 'clbuild compile-implementation sbcl' should be an
>     upstream SBCL.  (The only non-upstream thing we do is to enable
>     threads, and I would argue that since upstream ships its *binaries* with
>     that feature and just hasn't enabled it in the sources, that a
>     reasonable deviation.)
>
>     It might go a bit far to load extra ASDF systems.
>
> Since I don't want to override the normal repl in "clbuild lisp", I
> have instead added a command "clbuild prepl".
>
> By default prepl is loaded at run time.  Users who want it built into
> their image for improved start up speed can dump a monster.core.
>
> Proposal:
>
> Perhaps it would be nice to enable linedit in "clbuild prepl" by
> default?  (Note that "clbuild prepl" also works on other Lisps, in
> particular Clozure.  We should see to it that the linedit feature is
> either supported by those Lisps, or enabled only if running on SBCL).

Sounds good. In fact I might actually be tempted to use clbuild. :-)

  -T.