Bug
import-mesh documents scale 1 1 1 as "no scaling", but the loader unconditionally multiplies coordinates by 1000 (URDF metres→millimetres convention), so importing a millimetre STL at documented-neutral scale produces a 1000× part. Workaround in the wild: import-mesh-scaled 0.001.
Found during #844's rana-60c assembly example (see PR #852, which documents the workaround in the example rather than changing behavior mid-PR).
Ask
Pick one and document it loudly:
- make
scale 1 1 1 actually neutral and add an explicit :units m|mm option for the URDF path, or
- keep the ×1000 default but fix the doc comment and warn when no units are stated.
Either way, add a regression test importing a known-dimension mm STL and asserting its bounding box.
(Related doc-vs-behavior drift, noted in PR #854: cylinder/prism doc comments claim "centered at origin" but both build from z=0 — the docs are wrong, the behavior is relied upon.)
Bug
import-meshdocumentsscale 1 1 1as "no scaling", but the loader unconditionally multiplies coordinates by 1000 (URDF metres→millimetres convention), so importing a millimetre STL at documented-neutral scale produces a 1000× part. Workaround in the wild:import-mesh-scaled 0.001.Found during #844's rana-60c assembly example (see PR #852, which documents the workaround in the example rather than changing behavior mid-PR).
Ask
Pick one and document it loudly:
scale 1 1 1actually neutral and add an explicit:units m|mmoption for the URDF path, orEither way, add a regression test importing a known-dimension mm STL and asserting its bounding box.
(Related doc-vs-behavior drift, noted in PR #854:
cylinder/prismdoc comments claim "centered at origin" but both build from z=0 — the docs are wrong, the behavior is relied upon.)