Iron Tutorial builds are consistently failing in robotiq_driver tests

[TSNoble]

Description

e.g. https://github.com/moveit/moveit2/actions/runs/11748057912/job/32731324687

May be related to PickNikRobotics/ros2_robotiq_gripper#60

The fix there seemed to be that a user was trying to build for one version of ros, but was pulling in a different version of robotiq_driver, leading to errors due to the interface changing.

I'm a little confused as to why this is an issue on the tutorial workflows, since it looks like all of them use the main version of robotiq_driver, and the tests in question are internal to that repo, so I'd expect both the source and test code of robotiq_driver to be in sync.

Is it possible that the tutorial workflows are installing two separate versions of robotiq_driver, and the tests are running against a different version of the source?

ROS Distro

Iron (moveit2 tutorials workflow)

OS and version

Github Workflows

Source or binary build?

Source (via moveit2 tutorials workflow)

If binary, which release version?

N/A

If source, which branch?

All (as far as I can tell)

Which RMW are you using?

None

Steps to Reproduce

Raise a pull request in this repo and wait for the workflow to fail

Expected behavior

When making moveit changes, these tests should pass

Actual behavior

When making any changes, the pipeline consistently fails for reasons unrelated to a PR

Backtrace or Console output

No response

[TSNoble]

I think the issue is that the moveit2 tutorial docker image is based on the moveit2 source image, which appears to pull in an iron-compatible iron version of robotiq_driver for the iron builds. The tutorial docker image then pulls in the main version of robotiq_driver via rosdep, and the main tests attempt to run against the iron source.

[rr-tom-noble]

I tried downloading the moveit:moveit2:iron-source Docker image, but couldn't find any versions of robotiq_driver under /opt/ros/iron. That being said, the build was from 4 months ago (presumably when the pipeline started failing). I wonder if that image has started pulling in a robotiq_driver binary, hence the failures from that point onwards?

[rr-tom-noble]

Additionally, I'm unsure if it even makes sense to be running tests for build-from-source dependent packages of the tutorial image, since they're not covered by moveit code.

Perhaps a potential fix would be to set --cmake-args -DBUILD_TESTING=OFF for dependencies?

Then again, if they're all needed for the tutorials to work, I suppose running them provides guarantees that the tutorials as a whole will work.

[rr-tom-noble]

@sea-bass I think it'd make sense to close this given your changes removing the jobs

[sea-bass]

Closing by ignoring the problem (i.e., not building Iron 😆)