Bad STL generated from relatively simple model

[Claghorn]

Using openscad-2013.06-5.fc19.x86_64 from the fedora linux repos, I created what I thought was a pretty simply model. The most complicated thing it did was subtract an imported STL file from the inside of a cylinder shape. When attempting to slice the resulting model, slic3r generated wacky gcode, and kisslicer generated different wacky gcode, but also reported issues like degenerate triangles and other problems.

The files in question are available here: http://home.comcast.net/~tomhorsley/hardware/solidoodle/solidoodle-pcbvise.html

and some additional discussion is found here: http://www.soliforum.com/topic/4473/just-had-my-first-slic3r-screwup/

Just thought I'd report it in case someone wanted to look at it. It did generate a problem free model when I was using a pure cylinder. When I tried to make the top half a very very slight cone shape is when the generated STL went haywire.

https://github.com/openscad/openscad/blob/5affb45eb36943c6b1e5909b3904b7426ab00d22/testdata/scad/bugs/issue541.scad

[kintel]

The STL appears to be the culprit. The following file crashes OpenSCAD when fed the STL in question:
render() import("/Users/kintel/Downloads/PCBviseBar.stl");

[Claghorn]

I can certainly believe that, but I guess I didn't think of it since I had no problems using the STL when I had just a cylinder. I now see that kisslicer does report problems with the bar when I open it. Thanks for taking a look at it.

[kintel]

We do have some robustness issues when importing STLs, so this is a good example of such a case. Not sure what can be done about it, but at least better reporting, perhaps light-weight automatic repair.
I'll eventually open a new issue on this..

[hroncok]

Maybe use admesh for reporting and/or repair?

[kintel]

Admesh is on the radar. I think Slic3r uses admesh-based repair of STL files.

[GilesBathgate]

It might be a good idea to put admesh in the testing pipeline and see if the export from openscads various tests produces stl with errors reported by admesh.

[kintel]

This test case displays the same behavior:
https://github.com/openscad/openscad/blob/master/testdata/scad/bugs/stl-cgal-convert_to_Polyhedron-crash.scad

[kintel]

From @nophead: NetFabb studio shows degenerate faces (of the example posted above)

[kintel]

Perhaps this can be improved by doing a delaunay re-evaluation before exporting, as discussed recently on the mailing list (http://forum.openscad.org/STL-export-and-triangulation-td6010.html).

[kintel]

The crash aspects of this issue was fixed in #410 - the remaining issue is about failure to export STL and render in preview mode due to the Nef_polyhedron -> Polyhedron3 conversion error.

[kintel]

This was fixed sometime during unstable refactoring. Should be fixed in the development snapshot builds