Robert Haschke

I think, the culprit is the abi-dumper as a diff of both dumps reveals that these method are considered private in the newer version: ```sh $ grep -B 4 -A...

I'm also affected by this issue. Is there a reason to not apply this patch?

Having thought more about it, the main reason to get more fine-grained access to build steps is not for caching, but for accessing the results of individual steps, on which...

@tylerjw: this thread will be interesting for you :wink:

No, I don't want to reimplement the ROS buildfarm and actually the buildfarm is rather poor at caching... > We just have to find a way to break the cache,...

Thanks for your initiative! I need to have a look later this week... > > This is trivial on GHA: Just use `${{ hashfiles(workspace.sh) }}`. I'm sure other CI providers...

To simplify reproducing your issue, could you please create a new test in [subframes_test.cpp](https://github.com/ros-planning/moveit/blob/03bd588bcec36fc012bf5de56fa6c362288a579e/moveit_ros/planning_interface/test/subframes_test.cpp) using `computeCartesianPath()` and the Panda robot from `moveit_resources_panda_moveit_config`? Thanks.

Please just file a PR with your unit test illustrating the issue and refer to this issue.

I think you are just missing the definition of the `grasp` pose in your `GenerateGraspPose` stage. Comparing with the Baxter example, you are missing this line: [`stage->setGraspPose("closed");`](https://github.com/ubi-agni/mtc_demos/blob/master/src/baxter.cpp#L63) This grasp pose...

> I'm also seeing a failed build here https://travis-ci.com/github/ros-planning/moveit_tutorials/jobs/497517586 This was due to the cpp subfolder introduced with your cpp + python doc config. It was fixed in https://github.com/ros-planning/moveit_tutorials/pull/628 >...