Re: [Maxima] [Gcl-devel] Re: A bug in maxima-5.10.0.a

[Camm Maguire]

Greetings!

I take it maxima redefines toplevel or equivalent -- then why not call
(get-temp-dir) in the redefined version?

It has always been somewhat unclear to me the lines that govern the
relationship between lisp systems and programs built thereon.  The
latter can redefine any function in the former.  This is considered a
desired feature of lisp.  So what guarantees can the former make for
the latter?  I.e. even if we did 1. below, maxima might redefine
gazonk-name someday.  In fact, the gcl-alt-link code I submitted does
just that.

In practice, GCL can only go into (practically) untouchable areas by
writing in C and calling from main.  I'd like as much of GCL to be
written in lisp as possbile.

Everyone is area of *top-level-hook*, yes?  The way it appears to me
at the moment, get-temp-dir would be properly invoked were the maxima
toplevel to be invoked via this hook.

Take care,

"Vadim V. Zhytnikov" <address@hidden> writes:

> Hi!
> 
> I see from gcl code that *tmp-dir* is initialized by
> (gcl-temp-dir) only in one place - toplevel in gcl_cmpmain.lsp.
> And this works well for gcl itself but not for maxima
> and I really don't understand why.  I see only two ways to
> resolve the problem on gcl's side:
> 
> 1. Call get-temp-dir whem gazonk-name is called.
> Inefficient but reliable.
> 
> 2. Insert *tmp-dir* reset somewhere in the initialization
> code which will be executed for any image created by
> gcl (like maxima executable).
> 
> 
> >> I'm not sure this is really GCL bug.  The point that gcl
> >> always resets *tmp-dir* on start up.  But this doesn't
> >> happen with Maxima image.  Probably Maxima itself should
> >> take care of resetting *tmp-dir*.
> > 
> > Why is *tmp-dir* not assigned correctly in the Maxima image?
> > If Maxima is preventing GCL from assigning this value correctly,
> > or causing GCL to assign it incorrectly, it is a Maxima bug.
> > Otherwise it is a GCL bug.
> > 
> > On the whole I don't think Maxima should try to manage
> > internal variables for each Lisp implementation. There is
> > probably already some of that happening now, but I think
> > we should steer away from doing more of it.
> > 
> 
> 
> -- 
>       Vadim V. Zhytnikov
> 
>        <address@hidden>
>       <address@hidden>
> _______________________________________________
> Maxima mailing list
> address@hidden
> http://www.math.utexas.edu/mailman/listinfo/maxima
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah