I have had this problem in the past, and have seen documentation confirming it: OpenSCAD can't reliably make a correct manifold if vertices from different polyhedra exactly coincide. It gets confused about what belongs to what polyhedron, and what the faces are. Things get worse when any sort of subtractive geometry is involved.  (In the original this was stated with much better mathematical clarity, but I can't remember where that was.) But I find myself running into this repeatedly, as if into a brick wall! We tried making a beautiful spiral structure from hollow hexagons whose rims perfectly overlapped, extruded upward into a conical form using linear_extrude. Bam! The familiar failure! F5 display is missing some faces, and F6 warns of possible non-manifold. Attempt to 3D print shows most of the structure is lacking. This is painful! This suggests to me that some structural information is being lost between the parsing point and the CGAL point. Can we fix this? OpenSCAD would be a much better tool if we could! Attached is my example. Please let me know if my diagnosis is correct, or if this actually reflects some other problem. Thanks, Ken Brooks

F5 needs convexity setting in the linear_extrude to display correctly. scale = 0 makes a singularity at the top that is non-manifold. It works fine in F5 and F6 setting scale = 0.01 and convexity = 10. [image: image.png] On Wed, 25 Nov 2020 at 15:29, Ken Brooks <kb@joymaker.me> wrote: > I have had this problem in the past, and have seen documentation > confirming it: OpenSCAD can't reliably make a correct manifold if vertices > from different polyhedra exactly coincide. It gets confused about what > belongs to what polyhedron, and what the faces are. Things get worse when > any sort of subtractive geometry is involved. (In the original this was > stated with much better mathematical clarity, but I can't remember where > that was.) > > But I find myself running into this repeatedly, as if into a brick wall! > We tried making a beautiful spiral structure from hollow hexagons whose > rims perfectly overlapped, extruded upward into a conical form using > linear_extrude. Bam! The familiar failure! F5 display is missing some > faces, and F6 warns of possible non-manifold. Attempt to 3D print shows > most of the structure is lacking. > > This is painful! This suggests to me that some structural information is > being lost between the parsing point and the CGAL point. Can we fix this? > OpenSCAD would be a much better tool if we could! > > Attached is my example. Please let me know if my diagnosis is correct, or > if this actually reflects some other problem. > > Thanks, > Ken Brooks > _______________________________________________ > OpenSCAD mailing list > Discuss@lists.openscad.org > http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org >