Russ Tedrake

*[`geometry/optimization/test/graph_of_convex_sets_test.cc` line 2558 at r5](https://reviewable.io/reviews/RobotLocomotion/drake/21402#-O12n8u03sEUxfZE5d2O:-O137rzS5c7doGf_WAJ2:b-z6ub96) ([raw file](https://github.com/RobotLocomotion/drake/blob/ebf930c19d3f1c9d73c325a13680a1223dc103b1/geometry/optimization/test/graph_of_convex_sets_test.cc#L2558)):* Previously, cohnt (Thomas Cohn) wrote… No preference from me -- up to you if you want to wait. I think people probably won't...

Here's a slightly more minimal reproduction: ``` import numpy as np from pydrake.all import ( StartMeshcat, DiagramBuilder, AddMultibodyPlantSceneGraph, Parser, MeshcatVisualizer, ) directives = """ # Model directive for iiwa with...

I made a quick c++ reproduction (which can run in `meshcat_visualizer_test.cc`): ``` GTEST_TEST(MeshcatVisualizerTest, Issue21219) { std::string directives = R"""( directives: - add_model: name: iiwa file: package://drake/manipulation/models/iiwa_description/urdf/iiwa14_spheres_collision.urdf - add_weld: parent: world...

Good call @xuchenhan-tri . That's exactly what's happening. I'll look for the next offender.

Ok. The segfault is in `fcl::DynamicAABBTreeCollisionManager`, called from `SceneGraph::CalcPoseUpdate`. Maybe we should actually throw in the kinematics call, even in release mode?

The `DRAKE_ASSERT_VOID` that made this a debug-only throw is [here](https://github.com/RobotLocomotion/drake/blob/2b325f9b38c07774df423e73b332b3c0a86e50b8/math/rotation_matrix.h#L328). Do I understand your comment as saying that we should _not_ change this line to a release-build throw; e.g. directly...

+@rpoyner-tri for disposition (and resolution?) since our normal component owner is ooo.

yes. I have both the `/sys/devices/system/cpu/cpufreq/boost` and the `.../policyX` directories.

Thanks. Even for the primitives, this would be very useful to me. I *could* write the OBBs myself, but I don't want to (in every application)!