orb 5.1.0 breaks the Semigroups package

[james-d-mitchell]

I'm now seeing ci failures:

https://github.com/libsemigroups/libsemigroups/actions/runs/23948328161/job/69849495049?pr=942

I'm investigating if there's a workaround.

[james-d-mitchell]

It looks like the "wrong" hash function (ORB_HashFunctionFor8BitVectors) is being chosen somewhere in Semigroups, then applied to a vector that isn't in Is8BitVectorRep and so returns fail when an integer is expected. This must be happening because of the changes in a27a89a, but I need to investigate further.

[james-d-mitchell]

I didn't understand exactly what changed to make the tests start to fail but it seems that removing:

  SEMIGROUPS.HashFunctionForPlistRowBasisOverFiniteField := function(x, data)
    if Rank(x) = 0 then
        return 1;
    fi;
    Assert(1, IsRecord(data));
    return data.func(x!.rows, data.data);
    end;

    InstallMethod(ChooseHashFunction, "for plist row basis over finite field",
    [IsPlistRowBasisOverFiniteFieldRep, IsInt],
    function(x, hashlen)
    local data;
    if Rank(x) <> 0 then
        data := ChooseHashFunction(x!.rows, hashlen);
    else
        data := hashlen;
    fi;
    return rec(func := SEMIGROUPS.HashFunctionForPlistRowBasisOverFiniteField,
                data := data);
  end);

from Semigroups resolves the issue, I've opened a PR which resolves this locally for me at least:

semigroups/Semigroups#1171

[fingolfin]

Indeed I saw the semigroups test failed, hence 5.1.0 was not merged into the package distro, see gap-system/PackageDistro#1310.

I've not yet had time to investigate this, but it is on my TODO list.

[james-d-mitchell]

Thanks @fingolfin I'm pretty sure I fixed it already,and will make a release of Semigroups probably tomorrow, so it should be okay to do nothing also.