Unexpected behaviour of `removehredundancy`

[ma-b]

I'm getting some unexpected errors when removing redundancy from a polyhedron as opposed to removing it from its H-representation.
Here is an example (using Julia v1.11.3, Polyhedra v0.8.1, and GLPK v1.2.1):

A = [1 1; 1 1; -1 0; 0 -1]
b = [1, 1, 0, 0]
h = hrep(A, b, BitSet([1,2]))
lib = DefaultLibrary{Rational{Int}}(GLPK.Optimizer)
p = polyhedron(h, lib)

This produces the following (the H-representation of p contains a duplicate hyperplane):

Polyhedron DefaultPolyhedron{Rational{Int64}, MixedMatHRep{Rational{Int64}, Matrix{Rational{Int64}}}, MixedMatVRep{Rational{Int64}, Matrix{Rational{Int64}}}}:
2-element iterator of HyperPlane{Rational{Int64}, Vector{Rational{Int64}}}:
 HyperPlane(Rational{Int64}[1, 1], 1//1)
 HyperPlane(Rational{Int64}[1, 1], 1//1),
2-element iterator of HalfSpace{Rational{Int64}, Vector{Rational{Int64}}}:
 HalfSpace(Rational{Int64}[-1, 0], 0//1)
 HalfSpace(Rational{Int64}[0, -1], 0//1)

When I do

removehredundancy!(p)

all works fine and the duplicate hyperplane is removed (probably in the process of implicitly calling detecthlinearity!(p)). But when I just do

removehredundancy(h, GLPK.Optimizer)

then I get ERROR: The index MOI.VariableIndex(1) is invalid. Note that an index becomes invalid after it has been deleted.

(The same error occurs if I add a redundant halfspace, so that there actually is some redundancy in the halfspace part of the H-representation.)

Interestingly, if I lift p to one dimension higher and add a (non-redundant) hyperplane at index 1, as in the following code, then the variable index in the error message above becomes MOI.VariableIndex(2).

A2 = [0 0 1; A zeros(Int, size(A,1))]
b2 = [0; b]
h2 = hrep(A2, b2, BitSet([1,2,3]))
p2 = polyhedron(h2, lib)
removehredundancy(h2, GLPK.Optimizer)  # fails
removehredundancy!(p2)  # works fine

Could this have something to do with a dual variable for the redundant hyperplane that is being deleted too early?

Thanks!