[Axiom-developer] Re: Sources of the port of Axiom to SBCL/CMUCL/Clisp

[Gregory Vanuxem]

Le dimanche 25 février 2007 à 03:30 +0100, Waldek Hebisch a écrit : 

[...]

> I suspect that the use of compiled-function-p is bogus: Axiom probably
> assumes that the function is either given by a lambda list or is
> a compiled function.  But it looks that such assumption was true in
> the past, but now both compiled and interpreted functions are (represented
> by) function objects.  I think that the case you quote above just
> tests for archaic representation of interpreted functions.

This is what I think too.

> 
> I hope that Lisp implementation that is capable of finding name of a function
> object will return this name as a third value of function-lambda-expression.
> However, I noticed that sbcl returns funny things for functions obtained
> from lambda expresions so I added a little samity check.  I wonder
> how well such implementation works with other Lisps -- in principle
> it may be quite portable.

It is definitely a better implementation.

[...]

> I had apply the following diff to build with  sbcl 0.9.16.  The first
> hunks just corrects a typo, the second chunk is a quick hack around the
> problem that sb-unix::unix-mkdir accepts only base-string, while we have
> a string as an argument.  I did not try building using sbcl 1.0, but
> I did check that in sbcl 1.0 sb-unix::unix-mkdir works just with
> strings, so probably 1.0 needs no workaround.

I'm working with 1.0.2 (in fact pre 1.0.3 since I use a file from CVS (because 
of some changes
in the FOP compiler in 1.0.2, in some circumstances, undeclared variables are 
not "accepted",
the "real" compiler treats them as specials)).
I'll keep my implementation since the unix feature and the function unix-mkdir 
is available on
Windows, though I'm hesitant.

Thanks,

Greg