Unable to perform Boolean operations in mesh mode

[s_hemer]

Ok, I can report back that
a) Changing my install from freecad-python2 (which was chosen by default when installing freecad package on 18.04) to freecad-python3 (and of course removing freecad-python2 and all otherwise unused dependencies) works with the old OpenSCAD 2015.03-2+dfsg-2~18.04 from their PPA (which has not been updated yet, unfortunatelly)
b) it works with a simple example (cube/sphere difference) but with my initial meshes their seems to be another issue most likely with the mesh itself I assume (OpenSCADUtils.OpenSCADError: ' ERROR: CGAL error in CGAL_Nef_polyhedron3(): CGAL ERROR: assertion violation!\nExpr: e_below != SHalfedge_handle()\nFile: /usr/include/CGAL/Nef_3/SNC_FM_decorator.h\nLine: 426\nCurrent top level object is empty.\n'). Will further dig into that.

Two thing I'd suggest for improvement:

- Whether to fix the python2 variant or not (I understand that it might not be worth the effort) it would be nice that, if possible (depending on system), the python3 variant is installed by default.
- For the new error, it would be nice to not just see the same 'Unknown error' dialog in Mesh Design but a more telly one. OpenSCAD view shows the error on the output window which, I guess, was activate by myself otherwise it would have been quiet.

Just thoughts, trying to improve this cool piece of software.

[wmayer]

Added the fix to the 0.18 branch:
https://github.com/FreeCAD/FreeCAD/comm ... a1fbac0093

[wmayer]

- For the new error, it would be nice to not just see the same 'Unknown error' dialog in Mesh Design but a more telly one. OpenSCAD view shows the error on the output window which, I guess, was activate by myself otherwise it would have been quiet.

Can you offer some test data so that we can check the behaviour of OpenSCAD?

[s_hemer]

Sorry, I didn't find time in the last weeks to have a look on this problem. I checked another way as well: creating solids from the meshes and take the difference from them but it also fails (the one solid seems to have errors, removed them again as the file was to big).

Find the example attached. It is the CNC router base (from https://www.thingiverse.com/thing:3397013) where I want to remove the airchannel shape (Difference Base-Channel).

[TheMarkster]

When I open your file I see 2 different meshes, airchannel and makitabase. If you want to get rid of airchannel just delete it.

[s_hemer]

Did you mean that serious? How else do you say when you do a boolean difference operation: remove mesh B from mesh A, so of course two meshes. After that binary boolean operation I assume a single mesh C to be the result (C = A && !B).

[wmayer]

The suggestion is indeed not very helpful.

I checked the boolean cut with your test file and I can confirm the behaviour. However, the actual failure is something OpenSCAD or CGAL-specific which I cannot fix.

But the shown error message isn't very helpful either because it makes the user believe that OpenSCAD couldn't be found. This has been fixed with git commit 7415e3602

If you display the meshes in wireframe or flat lines mode you will see they have many skinny triangles which from a numerical point of view is very problematic. Maybe that's why CGAL fails to process the cut operation. So, what you can try is to re-mesh your meshes with an external application like OpenFlipper or maybe MeshLab.

[s_hemer]

Thanks for the clarification (and fix). The airchannel was meshed with (or from) FreeCAD itself while the other part was, as I already wrote, imported.

I ended up reworking that part as well and doing the substraction with the two FreeCAD parts. Worked for a first prototype but I came across a bunch of other issues (the funny renaming/renumbering of edges/faces/points, resulting in broken references in following steps (can't they be updated/derived?, Fillet can't be applied because of residual edges, refining shape removes them but Fillet still can't be applied, copying of a part with Shape binders creates also copies of the original shapes... ).

Really sad that all this lowers the very nice first impressions of FreeCAD. Surely, FreeCAD is an ambitious project and so on, but it would be really nice to see such basic things working more flawlessly. At least, I gained some experience (let's see where/how to share).

[vocx]

...(the funny renaming/renumbering of edges/faces/points, resulting in broken references...

This is the topological naming problem.

FreeCAD is pretty good at creating complex models. But users without experience are easily tripped by some quirks that are trying to be solved. If they were simple things they would have been solved a long time ago. They are complex issues.

Take a look at the users showcase subforum where you can see pretty impressive, real models.

[s_hemer]

Totally understand the problem, I just can't remember to have had this problem in a CAD software I once used at work. But thanks for the link and the correct term of that issue. I will definitively stay with FreeCAD.