axiom-developer
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Axiom-developer] RE: set function breakage


From: Mike Dewar
Subject: Re: [Axiom-developer] RE: set function breakage
Date: Thu, 24 Jul 2003 10:08:18 +0100

Hi Guys,

I wonder if the clue here is the message:

>    >> System error:
>    Value stack overflow.
> 
> protected-symbol-warn called with (NIL)

This is a CCL feature which I added for the following reason.  As some
of you know, CCL uses a byte-code interpreter for normal lisp code,
which is compact and portable but slower than compiled code in some (but
not all) cases.  Arthur Norman provided us with some simple but powerful
profiling tools and a mechanism for identifying perfomance-critical
functions, converting them to C and compiling them into the Lisp kernel.
As a result of this we were able to make the CCL version of Axiom not
only smaller and much more robust than the AKCL version, but at least as
fast if not faster.  However the drawback with this approach was that if
you wanted to take some existing algebra code and tinker with it (the
normal approach to Axiom development :-) ) then you might find that you
couldn't re-define a function beacuse it was compiled into the CCL
kernel.

To get round this I added some facilities to both Axiom and CCL to warn
users when they tried to re-define a kernel function and to allow them
to do so.  At the top level in Axiom you can go:
)set kernel
to see these options.  The defaults are chosen to prevent spurios
warings/re-definitions when algebra code is loaded from the default
library files (axiom.lib et al).

I believe that all the changes I made were to CCL and that the the
kernel options in Axiom simply toggle a couple of flags (in the usual
places setvars.boot and setvart.boot), but I might be wrong.  Sorry if
this is a red herring but it might help somebody track down the bug ...

Regards, Mike.

On Wed, Jul 23, 2003 at 10:17:33PM +0200, David MENTRE wrote:
> root <address@hidden> writes:
> 
> > If you copy the algebra subdirectory from the download version it
> > should work.
> 
> Err no.
> 
> I've tried to compiled the CVS xpoly.spad with your prebuild Axiom and
> it fails, in the same way as with GCL.
> 
> What I have done:
> 
> * In ~/pub/axiom-libre/axiom-cvs-2003-06-25-dm-i386/new/new/src/algebra:
> 
>  - notangle xpoly.spad.pamphlet > /tmp/xpoly.spad
> 
> * In the directory where I have untared axiom.linux.20030614.tgz:
> 
>  - export AXIOM=/home/david/00-poubelle/axiom/mnt/linux/
> 
>  - PATH=$AXIOM/bin:$PATH
> 
>  - axiom
> 
> * Under pre-build Axiom:
> 
> (1) -> )cd /tmp
>    The current AXIOM default directory is /tmp/ 
> (1) -> )co xpoly )con XPR
> [...]
>    compiling local outTerm : (R,E) -> OutputForm
> Time: 0.03 SEC.
> 
>    compiling exported coerce : $ -> OutputForm
> Time: 0.22 SEC.
> 
>  
>    >> System error:
>    Value stack overflow.
> 
> protected-symbol-warn called with (NIL)
> 
> 
> So the bug would be in the algebra???
> 
> Best regards,
> d.
> -- 
>  address@hidden
> 
> 
> _______________________________________________
> Axiom-developer mailing list
> address@hidden
> http://mail.nongnu.org/mailman/listinfo/axiom-developer
> 
> ________________________________________________________________________
> This e-mail has been scanned for all viruses by Star Internet. The
> service is powered by MessageLabs. For more information on a proactive
> anti-virus service working around the clock, around the globe, visit:
> http://www.star.net.uk
> ________________________________________________________________________

________________________________________________________________________
This e-mail has been scanned for all viruses by Star Internet. The
service is powered by MessageLabs. For more information on a proactive
anti-virus service working around the clock, around the globe, visit:
http://www.star.net.uk
________________________________________________________________________




reply via email to

[Prev in Thread] Current Thread [Next in Thread]