Allylic Radicals: Open Babel Error in Do 3D coordinate generation failed

Environment Information

Open Babel version: Open Babel 3.1.0 -- Jun 13 2022 -- 01:49:04
Operating system and version: MacOS Ventura 13.2.1

Expected Behavior

I expect to generate a 3D geometry for the allylic radical below:

obabel -:"C/C=C(C)/[CH2]" --gen3D --ff MMFF94 -h -oxyz

Actual Behavior

When I attempt to generate the 3D geometry for the allylic radical, I obtain an error instead:

obabel -:"C/C=C(C)/[CH2]" --gen3D --ff MMFF94 -h -oxyz
==============================
*** Open Babel Error  in Do
  3D coordinate generation failed
14

C          0.00000        0.00000        0.00000
C          0.00000        0.00000        0.00000
C          0.00000        0.00000        0.00000
C          0.00000        0.00000        0.00000
C          0.00000        0.00000        0.00000
H          0.00000        0.00000        0.00000
H          0.00000        0.00000        0.00000
H          0.00000        0.00000        0.00000
H          0.00000        0.00000        0.00000
H          0.00000        0.00000        0.00000
H          0.00000        0.00000        0.00000
H          0.00000        0.00000        0.00000
H          0.00000        0.00000        0.00000
H          0.00000        0.00000        0.00000
1 molecule converted

Changing the alkene geometry (obabel -:"C/C=C(C)\[CH2]" --gen3D --ff MMFF94 -h -oxyz) or removing the alkene geometry entirely (obabel -:"CC=C(C)[CH2]" --gen3D --ff MMFF94 -h -oxyz) has no impact; I get the same error.

The corresponding anion seems fine (including with the other alkene geometry):

obabel -:"C/C=C(C)/[CH2-]" --gen3D --ff MMFF94 -h -oxyz
14

C          0.93745        0.00500        0.09536
C          2.42550       -0.10483        0.03668
C          3.12936       -1.25052        0.00992
C          2.51958       -2.62452        0.04605
C          4.57019       -1.21118       -0.07364
H          0.64233        1.06012        0.10304
H          0.54410       -0.45593        1.00624
H          0.47811       -0.45886       -0.78226
H          2.95568        0.84895        0.00129
H          1.42593       -2.60389        0.06982
H          2.85452       -3.16324        0.93934
H          2.81401       -3.20125       -0.83735
H          5.10498       -0.27298        0.02505
H          5.10141       -2.05996       -0.48980
1 molecule converted

I can get a structure for the original allylic radical by enumerating the Hs on the radical explicitly, like so, with no alkene geometry specified:

obabel -:"CC=C(C)[C]([H])[H]" --gen3D --ff MMFF94 -h -oxyz
14

C          0.96609       -0.13361        0.10476
C          2.44475       -0.13741       -0.11041
C          3.09441        0.53782       -1.07856
C          2.42731        1.41491       -2.10109
C          4.53079        0.44165       -1.18629
H          5.08925       -0.26795       -0.58627
H          5.08002        1.18435       -1.75434
H          0.71893       -0.72475        0.99267
H          0.44834       -0.58446       -0.74721
H          0.59124        0.88077        0.26961
H          3.01264       -0.74417        0.59270
H          1.33640        1.40950       -2.01873
H          2.67599        1.07752       -3.11348
H          2.76034        2.45275       -1.99086
1 molecule converted

Any idea what might be going on here? The closest related issue I could find was #2317.

Thanks so much for any assistance!

Start by reproducing the reported obabel --gen3D --ff MMFF94 command for the allylic radical and compare it with the working anion and explicitly hydrogenated forms. Read the handling of radical input and 3D coordinate generation, using issue #2317 as the related reference. Done means the original radical command generates valid, nonzero 3D coordinates without the reported error.