[Top][All Lists]

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

Re: [Axiom-developer] Error when converting to a set

From: root
Subject: Re: [Axiom-developer] Error when converting to a set
Date: Sun, 29 Jun 2003 12:37:16 -0400


Now THAT's interesting. I've built cmucl but have not yet built Axiom
on top of it. I'm trying to prepare two talks and a Rosetta CD for 
a conference so I'm rather lagged. 

I suspect that the problems we are encountering have to do with the
differences between the common lisps. The problem Bill Page found
( If R has IntegralDomain ) is also related to the underlying common
lisp as far as I can tell. It appears that SET-DIFFERENCE gives 
different (but set-equivalent) answers. For instance, in 

CCL (the NAG common lisp)

(setq a '(a))
(setq b '(a b c))
(set-difference b a) ==> (b c)


(setq a '(a))
(setq b '(a b c))
(set-difference b a) ==> (c b)

I'm not sure if Axiom depends on this but the compiler uses it to
decide which super-domains need to be processed. In theory it shouldn't
make any difference. In practice it means that the tree-walk will take
branches in a different order. 

I suspect the bug you've found and the bug related to the short-set
[1000] ==> [100] are related to subtle lisp differences.

Subtle differences in the common lisps are VERY hard to find since they
occur so deep within the processing stack. Ideally Axiom should not depend
on these differences.

Good catch, though. The input files will be used for regression testing
the results so it is important that people verify the answers given.

Tim Daly

(my address@hidden account seems to be VERY unreliable lately. please use
the address@hidden id instead).

reply via email to

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