RobotLocomotion / drake

Model-based design and verification for robotics.
https://drake.mit.edu
Other
3.35k stars 1.27k forks source link

High count collision geometry causes segfault #13818

Closed chenheamazon closed 2 years ago

chenheamazon commented 4 years ago

Before you post:

  • Have you reviewed the documentation on Getting Help? Yes
  • Have you reviewed existing GitHub issues and StackOverflow posts? Yes the existing segfault issues are not related to this current one.
  • Have you rerun install_prereqs.sh on your current commit of drake? Yes

If you have encountered a bug (installation, segfaults, etc.), please try to post reproduction steps and as much relevant output as possible. Consider posting long snippets as a Gist.

This is on commit 79200833ded29b88f01a28b74afacdd9d0323e87 from June 1, 2020

When I try to run the simple_gripper built in example, and change the number of collision geometry in the example https://github.com/RobotLocomotion/drake/blob/master/examples/simple_gripper/simple_gripper.cc#L67 (this line here) from 8 to 300, it causes a segfault. We have a similar use case where too many registered collision geometry causes a segfault and we suspect that this is the same issue.

sherm1 commented 4 years ago

Assigning to @SeanCurtis-TRI for initial disposition.

SeanCurtis-TRI commented 4 years ago

I suspect that this got resolved with #13676.

SeanCurtis-TRI commented 4 years ago

You should see the benefit in the next release (or in master). Based on what the issue says, you should be able to reduce the number of collision geometries to around 100 (to limit the number of contacts to <= 128) and be ok. Once you have access to master (or the next release), you won't have to live with that artificial constraint.

jwnimmer-tri commented 4 years ago

FYI #13676 is available as of v0.20.0 (and v0.21.0, etc.) tags.

SeanCurtis-TRI commented 4 years ago

I'm surprised; I went looking in the release notes and didn't see it. But I hadn't actually dug in the commit history to confirm it was included.

jwnimmer-tri commented 4 years ago

https://drake.mit.edu/release_notes/v0.20.0.html#multibody-dynamics

SeanCurtis-TRI commented 4 years ago

Oops...because it's here.

SeanCurtis-TRI commented 3 years ago

@chenheamazon Did you ever have the chance to confirm that the problem has gone away?

SeanCurtis-TRI commented 2 years ago

Closing because there's been no action for 1.5 years.