Export creates non-manifold STLs [$15] about openscad HOT 15 CLOSED

Maybe better to join the openscad mailing list and send this issue there:

http://rocklinux.net/mailman/listinfo/openscad

from openscad.

After investigating this for a bit, I've come to the conclusion that the internal mesh handling in OpenSCAD is slightly broken when given geometry including very small triangles. The culprit seems to be some internal optimization done by the grid class. This will be tricky to clean up without breaking something else :/

from openscad.

thanks for the update!

I'm getting into the habit of running /everything/ through
cloud.netfabb.com these days.

I wonder if it would tell you anything useful to compare before and
after from that process?

from openscad.

I've got myself a minimal example, so I think I'm good for now:

cube(size = [10,10,5]);
translate([5, 5, 8.1-9+0.9]) cube(size = [5,10,10]);

from openscad.

Also, boolean operations result in a lot of overlapping facets! :)

from openscad.

For me, OpenSCAD (2011.09.30) produces an STL file that contains a bad facet:

facet normal 0 0 0
outer loop
vertex 9.5 104.991 9.26462
vertex 9.5 104.991 9.38249
vertex 9.5 104.991 9.5
endloop
endfacet

from openscad.

bwilson:

1. You're using a pretty old OpenSCAD
2. Could you post the .scad file producing this STL?

from openscad.

I am seeing this with http://goldnerlab.physics.umass.edu/~bgamari/battery_holder.scad (according to Slic3r resulting in very incorrect GCode). I haven't yet been able which part of the model is responsible for the non-manifoldness, but it occurs with the current git master (67eb2eb with OpenCSG 1.3.2).

from openscad.

It looks like at least the two separating walls are infinitely thin and would cause such an error.

from openscad.

Ultimately I was able to stumble upon the issue (being the separating walls, as you mentioned, as well as some problems with the gasket recess, if I recall correctly).

While I understand that handling these issues correctly is remarkably difficult, it would be nice if OpenSCAD could help a bit more in tracking down the source of the problem. Especially in large models it can be very difficult to determine exactly where the problem is originating as the symptoms are often quite vague.

from openscad.

I agree - my wish would be to be able to graphically highlight such issues.

from openscad.

Another minimal example:

rotate([90,0,180]) translate ([0, 0, 8]) cube(10, center=true); 
rotate([90,0,270]) translate ([0, 0, 8]) cube(10, center=true); 

from openscad.

A more involved example: http://forum.openscad.org/file/n6554/corner.scad (which is taken from http://www.thingiverse.com/thing:179597)

from openscad.

Is it possible to fix the STLs locally?

from openscad.

At some point between 2016 and today, this issue was fixed. Likely by a fix in upstream CGAL.
The tests also work using the current Manifold backend

from openscad.