tf.Exception thrown while using waitForTransform when a frame doesn't exist

[marcoesposito1988]

As reported by other users as well in http://answers.ros.org/question/207039/tfexception-thrown-while-using-waitfortransform a bug seems to have appeared when using TF.

waitForTransform throws a tf.Exception if one of the frames doesn't exist yet (which is actually the point of waitForTransform).

It must be a bug in TF2. When I checked out geometry 1.11.4 the bug was still there, but when I also checked out geometry-experimental 0.5.7 the bug disappeared.

Any idea about the reason?

[sklaw]

Here are the combinations of TF and TF2 I've found so far that will cause this problem:
TF 1.11.6 with TF2 0.5.9
TF 1.11.6 with TF2 0.5.8

TF 1.11.4 with TF2 0.5.9
TF 1.11.4 with TF2 0.5.8

[tfoote]

What is the exception thrown? I added unit tests for wait for transform, in ros/geometry@9624370

[tfoote]

Related questions on answers.ros.org

• http://answers.ros.org/question/207582/does-tflookuptransform-python-take-time-to-fire-up/
• http://answers.ros.org/question/207052/rospy-tf-waitfortransform-not-waiting/

[tfoote]

I believe I've found the issue in this commit: 2f5ba1d

When running inside rospy, the roscpp client is not initialized and thus ros::ok is false. I'm going to have to think about how to make this work cleanly in both roscpp and rospy. .

[tfoote]

This should be fixed in geometry_experimental 0.5.11