Add a manifold test case for custom "revolve".

[fire]

Add test.

[fire]

This might be as expected since it's not a crash and may or may not be manifold (probably non-manifold).

[elalish]

@pca006132 is it expected that about half our CI jobs (the ones with green checkmarks) just build but don't actually run any tests?

[elalish]

@fire - so it's working except that we're getting 16 triangles instead of your expected 48. Can you show the difference in renders so we can get an idea of what's wrong with the shape?

[fire]

Do you know if I can export GLTF2 from here? I can't preview the mesh in Godot Engine because the condition is that the manifold is manifold.

We could do empty mesh for (merge and intersection) and identity for subtraction.

[elalish]

Yeah, you can add this in like a bunch of our other tests have:

#ifdef MANIFOLD_EXPORT
  if (options.exportModels) {
    ExportOptions options2;
    ExportMesh("complex.glb", manifold.GetMeshGL(), options2);
  }
#endif

[elalish]

Then just run ./manifold_test --gtest_filter=Manifold.MeshWithCustomProps -e to export.

[pca006132]

@pca006132 is it expected that about half our CI jobs (the ones with green checkmarks) just build but don't actually run any tests?

Hmmm, weird. Will check.

[fire]

complex.glb.zip

Here's the complex glb.

[fire]

The failure points seem to be at the top and bottom.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 91.90%. Comparing base (3603542) to head (2c992b8).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1093      +/-   ##
==========================================
+ Coverage   91.87%   91.90%   +0.03%     
==========================================
  Files          30       30              
  Lines        5883     5883              
==========================================
+ Hits         5405     5407       +2     
+ Misses        478      476       -2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[elalish]

@fire I don't see a failure - can you highlight the difference? Also, your test passes, so....

And don't get me started on USDZ! Here you can drag and drop GLBs into my web viewer here: https://modelviewer.dev/editor/

[fire]

With this change, I cannot export.

% ./test/./manifold_test --gtest_filter=Manifold.MeshWithCustomProps -e
Note: Google Test filter = Manifold.MeshWithCustomProps
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from Manifold
[ RUN      ] Manifold.MeshWithCustomProps
manifold_test(58041,0x202c23840) malloc: Incorrect checksum for freed object 0x1537044b8: probably modified after being freed.
Corrupt value: 0x1000101000100
manifold_test(58041,0x202c23840) malloc: *** set a breakpoint in malloc_error_break to debug
zsh: abort      ./test/./manifold_test --gtest_filter=Manifold.MeshWithCustomProps -e

[fire]

These are non-manifold, according to Blender. See also the right side 3D-print fixer.

Edited:

Also, the custom mesh revolve's vertex are non-manifold at the far back.

Edited:

I think this is an input error, but I would like to know how to fix it.

[elalish]

These are non-manifold, according to Blender. See also the right side 3D-print fixer.

Um, I disagree? Your test is checking manifoldness and it passes. I trust my manifold checker better than anyone else's, but I especially don't trust file importers and exporters. Particularly since every mesh you've shown me doesn't even have shared verts, I have the feeling manifoldness is getting lost somewhere in Godot.

[elalish]

Remember - you cannot have a manifold without shared verts.

[pca006132]

Wonder if we can do an export-import round-trip in this case. If not, maybe the exporter is to blame here?

[elalish]

To export, you'd also need to compile Manifold with MANIFOLD_EXPORT=ON, which requires Assimp, so you may not be set up for that.

[pca006132]

@fire can you merge the current master branch? It seems that you do not have #1092 in your branch. I guess github somehow automatically merge stuff when running the CI...

And I tried a round-trip of export-import, it works fine.

[fire]

I suspect what is happening is that the listed mesh edge is not connected in Godot Engine's input Mesh. I will try to explore this hypothesis.

The CSGPolygon3D is relatively untested with hellish/manifold, but people depend on it, so we're trying to avoid crashing or empty scenes.

Also, these are failure cases, so I am a bit frustrated, but the MeshGL dumping tool has been a great way to communicate the failures and gives me hope.

This test should pass.

[fire]

I did some sleuthing, I tried printing after the mesh.Merge() operation and it gives this test result.

[fire]

!

If you force add a MeshExport64 it fails even with Merge();

[pca006132]

Merge is only best-effort and cannot guarantee the result to be manifold, so this is not really an issue?

[pca006132]

Try a larger tolerance value, this is probably because the tolerance value becomes too small when using MeshGL64.

[fire]

Thanks for the hint, I expect the test to pass.

[pca006132]

Btw, do note that if you need the merge operation to make the custom revolve operation produce a manifold mesh, it probably means you may want to modify the revolve operation to make sure it always outputs a manifold mesh. There is no guarantee that the merge operation can work in general.

[fire]

I did a test manifold's CSG revolve: godotengine/godot#99977, but I wanted to try using the existing godot engine CSG code first.

[fire]

@elalish Should I close this pull request? Any revisions I can make?