Ilya Zakharevich on Fri, 6 Sep 2002 11:42:33 -0700


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

Re: yet another rnfkummer() posting


On Fri, Sep 06, 2002 at 10:40:53AM +0200, Bill Allombert wrote:
> > BTW, can you check with Math::Pari before release(s)?  It would
> > immediately report the mismatches in the signature (when the string
> > signature is changed to "a complicated one", but the numeric one is
> > not changed to 99).
> > 
> > Unfortunately, the current version does not report the function *names*, only suspicious signatures.  >>TODO.
> 
> I have completly get rid of signatures in the description system.

What does it mean?  That the CVS version has your updates?

> Instead I use a perl script that compute signatures. This is imbedded
> in the make_proto822.pl script.
> 
> Some of the function with prototype "GL" have valence 21, some others 
> valence 23. I suppose it is a bug in init.c
> 
> Could you check the script ? The only mandatory rule for PARI/GP are:

Math::Pari nowadays has a sanity check - it would work better than me.

>  * Currently the following functions have no code word:
>  * 'O' 50, 'if' 80, 'until' 82, 'while' 81, 'global' 88,
>  *
>  * Valence 0 reserved for functions without mandatory args.
>  * Valence 99 reserved for codes which do not correspond 1-to-1 to valences.
>  * Any other valence (what to do with 0?) should correspond to exactly
>  *  one code.
>  
>  So you can change the other valence as you want them.

Math::Pari knows about many numeric valences; it uses them during
compilation.  The unknown numeric valences (more or less 99 only) are
"dispatched" at runtime, using the string valence.

Ilya