What if you export obj/3mf?

On Thu, Feb 12, 2026, 20:53 Adrian Mariano via Discuss <
discuss@lists.openscad.org> wrote:

So, you're using some dev version for the manifold export, and the old
release version for the CGAL export?  There's a lot to pick at between
those two versions.  Using the same version for both exports would be
more telling.

Yeah, using STL is problematic, got to dig into that writing software to
do stone walls.  The Manifold library won't take non-manifold meshes,
but that doesn't mean the export process to STL doesn't mung something
up.  3MF is a much better alternative to preserve all the information
needed to define a manifold mesh, recommended if your workflow supports it.

On 2/12/2026 5:52 AM, Adrian Mariano via Discuss wrote:

I didn't realize CGAL changed in the dev versions.  I tested in 2026.02.07
and found that if I select CGAL, half my model is missing, like it was
non-manifold.  If I use manifold backend and save as 3mf then I don't get
errors.  I'm not sure if I can post designs e.g. on printables with a 3mf
instead of an stl?

I investigated the reason CGAL doesn't like the model in the dev version.
The model is a union of four rectangles that gets extruded.  Apparently
having the rectangles touch but not overlap is confusing CGAL.  If I
overlap one pair in particular then CGAL in the dev version works.  And
also there are no errors if I use manifold and save an STL.

On Thu, Feb 12, 2026 at 11:55 AM Glenn Butcher via Discuss <
discuss@lists.openscad.org> wrote:

Sometimes I can't readily figure out why my OpenSCAD exports generate
models with errors.  In those cases after Prusa Slicer auto corrects the
errors I export the corrected objects to have clean files for sharing.
This can be scripted with Prusa Slicer's command line mode to automate the
process for projects that will have revisions.

On Thu, Feb 12, 2026 at 12:45 PM Adrian Mariano via Discuss <
discuss@lists.openscad.org> wrote:

At printables you can upload just about any format model, including .scad.
Since Manifold has been dumping STLs with errors for the past several
months, I've been uploading .obj for single color models and .3mf for
multicolor models.
Dan

On Thu, Feb 12, 2026 at 6:46 PM Adrian Mariano via Discuss <
discuss@lists.openscad.org> wrote:

On 2/12/26 11:09 AM, Todd Allen via Discuss wrote:

There are "errors" and there are errors. See
https://github.com/elalish/manifold?tab=readme-ov-file#3d-formats

It's the saving as STL that creates some errors, often real errors.

The other is, IIUC, the definition of manifold for Manifold (i.e. see
https://github.com/KhronosGroup/glTF/blob/main/extensions/2.0/Vendor/EXT_mesh_manifold/README.md
) is not the same as the simplistic check used to report errors in your
slicer. But if you export as OBJ or 3mf, you should find it always
slices just fine despite the errors. If you don't and you're not
creating your own shapes (with vnf/polyhedron, like a lot of BOSL2
does), then file a bug. I give that exception because, in all
likelihood, it's the less-tested and complicated shapes being created in
VNF that are causing the errors and not the low-level libraries which
create primitives and do the union/intersection/difference operations.

• Cory

Hi Adrian,

Cgal has a lot of geometry repair features built in. Manifold has only
one, afaik. a forcecleanup, if the flag is set. (I'm  not sure if
openscad gives access to this level, anyway.)

Best wishes,

Ray

On 12/02/2026 12:52, Adrian Mariano via Discuss wrote:

I filed a bug report.  I also noticed that when render runs I only see this
in the console:

Parsing design (AST generation)...

Which was confusing because it makes it look like render hasn't finished.
No other messages appear even though the render is complete.

On Thu, Feb 12, 2026 at 3:23 PM Raymond West via Discuss <
discuss@lists.openscad.org> wrote:

Pertinent to this thread, I recently wrote a missive on why STL files
are deficient:

https://glenn.pulpitrock.net/blog/posts/2025-12-21-whats-so-bad-about-stl-files/

Didn't think much about it until I wrote a heightmap library for a
one-off program, spent a lot of time dorking with the logic to make a
manifold mesh of the height array.

On 2/12/2026 6:44 PM, Adrian Mariano via Discuss wrote: