pSubstPoly() causes crash when called from Singular.jl

[steenpass]

The following julia code gives an error:

using Cxx
using Singular

function substitute_variable(p::Singular.spoly{T}, x::Singular.spoly{T},
        z::Singular.spoly{T}) where T
    R = parent(p)
    index_var = findfirst(Singular.gens(R), x)
    @assert index_var != 0
    p_ptr = p.ptr
    z_ptr = z.ptr
    R_ptr = R.ptr
    res_ptr = icxx"""
        const ring origin = currRing;
        rChangeCurrRing($R_ptr);
        poly res = pSubstPoly(pCopy($p_ptr), $index_var, $z_ptr);
        // poly res = pSubst(pCopy($p_ptr), $index_var, $z_ptr);
        rChangeCurrRing(origin);
        res;
    """
    R(res_ptr)
end

R = Nemo.ResidueRing(Nemo.FlintZZ, 32003)
S, (x_0, x_1) = Singular.PolynomialRing(R, ["x_0", "x_1"])
p = x_0
q = substitute_variable(p, x_0, S(5));
q   # this works
gc()
q   # error

Note that this works if pSubstPoly() is replaced by pSubst(), so most likely, Singular does not use the correct API to handle the Nemo coefficients. Therefore, this is probably a bug in the Singular kernel.

Thanks to Bill for tracking this down!

[ren]

Hi Andreas,

possibly stupid question: why are you not using p_Copy and p_Subst/p_SubstPoly instead of pCopy and pSubstPoly/pSubst? That way you don't have to change the current ring (I think the latter ones are just shortcuts for calling the former ones with currRing).

best regards, Yue.

[steenpass]

Hi Yue,

true, this would be possible and it's actually a good suggestion. But it doesn't solve the problem that p_SubstPoly() yields an error.

Best, Andreas