[CM] quicklisp clm

Ralf Mattes rm at seid-online.de
Sun Dec 11 11:41:42 PST 2016


On Sun, Dec 11, 2016 at 06:16:21PM +0100, anders.vinjar at nmh.no wrote:
> >>>>> "R" == Ralf Mattes <rm at seid-online.de> writes:
>     >> Good luck.
> 
>     R> Looking at the state of the code I think that's really needed.
> 
> I use the CL version of CLM from time to time, using sbcl and lw, and i
> beleive some others also use CL/CLM at times.  What is it with the state
> of the code you find problematic?

Just as a quick reply, not neccessarily in order of importance (and
_not_ meant as a critique):

 - a rather baroque way to build/install (that makes distribution via 
   quicklisp rather unölikely). 
 - build-customization via *features* (again, not really working for
   Quicklisp)
 - implementation specific loading of libraries (there's no real need to
   maintain a list of *shared-object-extension* any more. Delegate all
   of this to cffi). In general, a lot of the code could be factored out
   to "general" libraries provided by Quicklisp (btw, the same is true 
   for CM as well).
 - That extra strange mixture of autoconf and (c source) compilation
   from lisp. I was trying to load CLM from the git repository but
   compilation of the library failed (missing symbols). With a proper
   makefile I would have been able to start to compile outside of lisp
   and debug the build.

Now, non of this is impossible or even hard to fix, but it needs to be
done and it's quite some work. I'm glad to see people work on that code,
but, in my very humble, personal and probably grummpy opinion, it needs
to be done by systematically cleaning up the code and not by adding
another layer of fixups - that only makes the system even more brittle.
Just one quick example: dir-setup.lisp uses
quicklisp:*local-project-directories* and concatenates it with a string
"/clm". Hmm, now we've lost the possibility to load CLM from asdf. And
we can't install it with Quicklisp either (iff it's picked up by
quicklisp at some point in the future). There's fine support to find the
source code location in ASDF/UIOP. Why not use that?
And don't assume quicklisp directories are writeable ... ;-)
I hope I've made my poit clear. I did not intend to lessen anyones
valuable work.

 Cheers, Ralf Mattes


  


 - 
> -anders
> 
> _______________________________________________
> Cmdist mailing list
> Cmdist at ccrma.stanford.edu
> https://cm-mail.stanford.edu/mailman/listinfo/cmdist
> 


More information about the Cmdist mailing list