Can't declare empty rows/cols in declare_partials

empty_rowcol.py

import openmdao.api as om
import numpy as np

from openmdao.utils.array_utils import evenly_distrib_idxs
import mpi4py.MPI as MPI 

N = 3

class DistribComp(om.ExplicitComponent):

    def initialize(self):
        self.options['distributed'] = True

    def setup(self):
        self.add_input('x', shape=1)
        rank = self.comm.rank
        sizes, offsets = evenly_distrib_idxs(self.comm.size, N)
        self.add_output('y', shape=sizes[rank])
        rows = np.arange(0, sizes[rank])
        cols = np.zeros((sizes[rank],))
        self.declare_partials('y', 'x', rows=rows, cols=cols)
        self.declare_partials('y', 'x', rows=np.arange(0, sizes[rank]), cols=np.zeros((sizes[rank],)))


    def compute(self, inputs, outputs):
        rank = self.comm.rank
        sizes, offsets = evenly_distrib_idxs(self.comm.size, N)
        outputs['y'] = inputs['x']*np.ones((sizes[rank],))
        if rank == 0:
            outputs['y'][0] = 2.3

    def compute_partials(self, inputs, J):
        rank = self.comm.rank
        sizes, offsets = evenly_distrib_idxs(self.comm.size, N)
        if MPI.COMM_WORLD.rank == 0:
            J['y','x'] = np.zeros((sizes[rank],))
            J['y','x'][0] = 0.0
        else:
            J['y','x'] = np.ones((sizes[rank],))

class GroupThatDoesWork(om.Group):
    def setup(self):
        dvs = om.IndepVarComp()
        dvs.add_output('x', val=6.0)
        self.add_subsystem('dvs', dvs, promotes_outputs=['*'])
        self.add_subsystem('distcomp',DistribComp(), promotes_inputs=['*'])
        self.add_subsystem('execcomp',om.ExecComp('z = sum(y)', y=np.zeros((N,)), z=0.0))
        self.connect('distcomp.y', 'execcomp.y')

def mainfunc():
    prob = om.Problem()
    prob.model = GroupThatDoesWork()
    prob.model.add_design_var('x', lower=-1.0, upper=7.0)
    prob.model.add_objective('execcomp.z')
    prob.setup(mode='fwd')
    prob['dvs.x'] = 7.0
    prob.run_model()
    prob.model.list_outputs(print_arrays=True)
    prob.check_totals()


if __name__ == "__main__":
    # works if run with 3 or fewer procs
    # raises an error in _declare_partials if run with 4 or more (one proc has no output, raises ValueError)
    mainfunc()