Re: Maxima rtestint failures with GCL 2.6.14

I've chased this bug into MTOSC in src/defint.lisp, where it seems to
have something to do with the functions RIB, CSEMIUP, and CSEMIDOWN.
There are a couple of special variables BPTU and BPTD which are
supposed to tie those together. Could be a bug in the handling of
special variables, or in the assignments to those.

I would like to note that this bug is only observed with GCL 2.6.14.
Other Lisp implementations don't show it.

best

Robert


On Mon, Apr 8, 2024 at 8:36 AM Robert Dodier <[email protected]> wrote:
>
> The current version of Maxima (00fcc04) bumps into several failures in rtestint when compiled with GCL 2.6.14.
>
> GCL advertises itself as:
> GCL (GNU Common Lisp)  2.6.14 Fri Jan 13 10:47:56 AM EST 2023  ANSI    git: Version_2_6_14
>
> The failures start with
>
> Running tests in rtestint:
> ********************** Problem 56 ***************
> Input:
>           cos(x)
> integrate(------, x, minf, inf)
>            2
>           x  + 1
>
>
> Result:
>   - 1    2
> %e    (%e  + 1) %pi
> -------------------
>          2
>
> This differed from the expected result:
>   - 1
> %e    %pi
>
> and then there are about a dozen more in a similar vein. There aren't any unexpected failures in any other test scripts.
>
> I don't remember seeing failures in rtestint before, although I don't check regularly. I don't know if it's due to a change in Maxima or in GCL.
>
> I built the same Maxima version w/ other Lisps (SBCL, Clisp, Clozure CL, ECL) and none of them report errors in rtestint.
>
> I haven't taken the time to investigate, can anyone hazard a guess?
>
> best
>
> Robert
>
>
>