Full rebuild August 2026 II : bump ros2-distro-mutex to 0.20.0 and build number to 25 + Add RoboPlan packages

[sea-bass]

Adds all the roboplan / roboplan-ros packages, except for roboplan_ros_franka which requires mujoco_ros2_control.

Also does a full rebuild to update Pinocchio and Coal, bumping the build number to 25.

Closes #28

[traversaro]

Note to self/other reviewers: in general we try to avoid doing this for abi compat version, but as this is an header only library isn ot a problem.

[sea-bass]

I'd be fine with proxsuite 0.7.2 as well, though, if it's better

[traversaro]

I guess if you install ros2-proxsuite 0.7.2 , you end up with proxsuite 0.7.3 being installed, as ros2-proxsuite has a proxsuite >=0.7.2,<0.8.0a0 dep on proxsuite, see https://prefix.dev/channels/robostack-rolling/packages/ros2-proxsuite (this derive from max_pin: 'x.x') used here.

[sea-bass]

It does appear the macos and Windows jobs blew past the 6 hour timeout 🤣

[traversaro]

It does appear the macos and Windows jobs blew past the 6 hour timeout 🤣

The cache is there and active, so just restarting should work fine.

[sea-bass]

The cache is there and active, so just restarting should work fine.

Even without removing packages and letting the job run to completion?

(I reverted the change, but will need help restarting the cancelled jobs since I don't have permissions)

[traversaro]

The cache is there and active, so just restarting should work fine.

Even without removing packages and letting the job run to completion?

Yes, the cache is saved also on jobs with timeout.

[traversaro]

The cache is there and active, so just restarting should work fine.

Even without removing packages and letting the job run to completion?

Yes, the cache is saved also on jobs with timeout.

Ok, apparently not, but it will be again soon: #40 .

[sea-bass]

I just rebased on main to include #40 , saw that it was already merged!

[sea-bass]

I thin the macos arm run failed due to a race with the recipes checker, probably just needs a re-run? thanks!

[mini-1235]

I see that the build timed out, but the cache was uploaded properly this time. I think we just need to close and reopen the PR a few times and it should eventually pass, assuming there are no actual build errors. From my previous experience, a full Windows rebuild take around 18 hours

[sea-bass]

So it appears there was an issue on Windows. The cached build for moveit_core seems to have been empty/incomplete, which then failed the build on the first dependent package (chomp_motion_planner).

I've modified the workflow YAML temporarily to remove the bad cache since there seems to be precedent for that. But maybe we can kick off this rerun the 2nd macos job completes so we don't accidentally damage that one 🙃

[sea-bass]

ok the macos jobs have all passed, this should be good to approve the workflow and get the windows build passing now -- thanks, all!

[sea-bass]

@traversaro this seems ready to go!

[traversaro]

@traversaro this seems ready to go!

Thanks a lot! I would just de-vendor mujoco as done in the other distros to avoid confusion, let me do that change.

[traversaro]

@traversaro this seems ready to go!

Thanks a lot! I would just de-vendor mujoco as done in the other distros to avoid confusion, let me do that change.

Actually mujoco was already devendored, I just added an option to ensure it was never accidentally vendored again.