Re: [Maxima-discuss] probable bug in sqrt

[Michel T. <ta...@lp...>]

Le 13/11/2025 à 18:24, Michel Talon a écrit :
>
>
> (%i5) soAT:linsolve(eqsAT, unksAT);
> CQUOTIENT: quotient is not exact
>  -- an error. To debug this try: debugmode(true);
>
> Commenting out domain:complex doesn't help. The message CQUOTIENT: 
> quotient is not exact is produced in rat3a.lisp, and means that the 
> remainder of an euclidean division does not vanish when it should. By 
> the way, the subresultant algorithm for computing gcds is supposed to 
> be absolutely sure if  slower than the standard one.
>
> Le 11/11/2025 à 14:53, Richard Fateman a écrit :
>> The earlier bug reports describe the situation in detail.  It is not that
>> there is a bug in sqrt, but a complication resulting from how 
>> programs deal with algebraic dependencies.


I think there is definitely a bug here: tracing the function cquotient 
shows there are a lot of calls to cquotient which work OK and then a 
last one which bombs out. This can be reproduced in a fresh maxima session:

 > maxima
Loading /home/michel/.maxima/maxima-init.mac
Maxima 5.48.1 https://maxima.sourceforge.io
using Lisp SBCL 2.5.8

(%i1) :lisp(MAXIMA::CQUOTIENT 188021706981212160 -32188114592695296)

Maxima encountered a Lisp error:

  attempt to THROW to a tag that does not exist: RAT-ERR

Automatically continuing.
To enable the Lisp debugger set *debugger-hook* to nil.
(%i1)

I have difficulty in understanding how this is not a bug in maxima or 
how this bug is related to algebraic dependencies. Maybe the numbers are 
too big for some reason.


-- 
Michel Talon