Closed r-veenstra closed 7 months ago
This doesn't seem to happen for regular BasisCurves
.
Maybe a reason for CesiumCartographicPolygon
to have a rel
to a BasisCurves
prim instead of being a subclass.
Actually, it seems like both BasicCurves
and CesiumCartographicPolygon
are working in USD Composer 2023.2.2.
Still might be worth considering the rel
approach if that improves the workflow or fixes other bugs.
CC https://github.com/CesiumGS/cesium-omniverse/pull/630 and https://github.com/CesiumGS/cesium-omniverse/issues/636
This doesn't seem to happen for regular
BasisCurves
.Maybe a reason for
CesiumCartographicPolygon
to have arel
to aBasisCurves
prim instead of being a subclass.
Moving to a rel
would also let us query any prim that has a points
attr. I can image some hypothetical dataviz workflows where creating a BasisCurves
would be more work than needed. Users wouldn't necessarily have to even know about a BasisCurves
.
@corybarr @lilleyse I can confirm that the original repro steps from this issue no longer cause a crash in 2023.2.2 with latest main 57fdeeebf7c47ed08282d981056be98eb72c49cb
However, if I run the repro steps, save a USDA, and re-open the usda, I do get a crash
So it seems the issue is currently related to editing CesiumCartographicPolygons that have been loaded from a USDA, vs being freshly created.
That's puzzling. It's strong evidence that we won't be able to fix the issue in our codebase, but let me see if I can did up any more precise info in some tests.
I tested this with the absolute minimal inheritance example:
class CesiumBasisCurvesChildPrim "CesiumBasisCurvesChildPrim" (
doc = """Test prim for inheritance crashes"""
inherits = </BasisCurves>
customData = {
string className = "BasisCurvesChild"
}
) {
}
With a raw BasisCurves
I was able to create this with FSD enabled:
With a BasisCurvesChild
it crashed. Aside from executing the prim's Define via our UI, the example bypasses our codebase. I think we should provide this minimal example to NVIDIA so they can reproduce.
In the meantime, let's fall back to a rel
. Agreed? @lilleyse @r-veenstra
A rel
that points to a BasisCurve
with a GlobeAnchor
added, yes?
@r-veenstra Exactly
@corybarr makes sense. It should also resolve https://github.com/CesiumGS/cesium-omniverse/issues/636
Note: Does not crash when fabric disabled
cc @corybarr