sum(1/((2*n+1)^2-4)^2, n, 0, oo, algorithm='maxima') is wrong

[seblabbe]

From ask.sagemath.org:

sage: n = var('n')
sage: sum(1/((2*n+1)^2-4)^2, n, 0, oo)
1/64*pi^2 - 1/12
sage: sum(1/((2*n+1)^2-4)^2, n, 0, oo, algorithm='maxima')
1/64*pi^2 - 1/12

but correct answer is 1/64*pi^2. SymPy (with #22004) and Mathematica do it right:

sage: sum(1/((2*n+1)^2-4)^2, n, 0, oo, algorithm='mathematica')
1/64*pi^2
sage: sum(1/((2*n+1)^2-4)^2, n, 0, oo, algorithm='sympy')
1/64*pi^2

I am using version:

$ sage -standard | grep maxima
maxima..................................5.35.1.p2 (5.35.1.p2)     

See #18920 for the ticket updating maxima version.

Depends on #18920

Upstream: Fixed upstream, in a later stable release.

Component: symbolics

Keywords: maxima

Issue created by migration from https://trac.sagemath.org/ticket/22005

[seblabbe]

comment:3

Seems to be fixed upstream or "perhaps an interaction with some flags that Sage is setting" according to Robert Dodier:

https://sourceforge.net/p/maxima/bugs/3236/#7b13/aab3

[seblabbe]

Upstream: Fixed upstream, in a later stable release.

[dimpase]

comment:4

this is to be fixed in #18920; this bug does not appear in the currently tested configuration of ECL-16.1.3+Maxima 5.39.0; this ticket should be closed after #18920 is done.

[dimpase]

Dependencies: #18920

[seblabbe]

comment:6

This is fixed in 8.0.beta11.

[seblabbe]

comment:8

Just to say that a doctest is already testing this issue:

https://github.com/sagemath/sage-prod/blob/develop/src/sage/calculus/calculus.py#L569

It was added in #22004 and updated during the upgrade to maxima 5.39.