[HappySamuel]

Backport

• Fix Task::Active::rewind() so it actually rewinds #154

Fixes #146 on jazzy: rewind_task never rewinds. With the apt Jazzy binaries (ros-jazzy-rmf-task-sequence 2.5.1), rewinding to a later completed phase hangs the fleet adapter.

The cherry-pick applies cleanly. Before #154, Task.cpp and test_Sequence.cpp were identical on jazzy, lyrical and main, so this is the same change as on main. The commit carries (cherry picked from commit c9f7201...).

Testing

• colcon build --packages-select rmf_task rmf_task_sequence of this branch on Ubuntu 24.04 / Jazzy, then test_rmf_task_sequence: all tests pass (359 assertions in 6 test cases), including the new "Rewind a task to an earlier phase" scenario.
• The rmf_demos results in Fix Task::Active::rewind() so it actually rewinds #154 were run against main.

GenAI Use

We follow OSRA's policy on GenAI tools

• I used a GenAI tool in this PR.
• I did not use GenAI

Generated-by: Claude Code (Claude Opus 5.5) was used to cherry-pick the change, run the tests and draft this description. I reviewed the changes and results.

🤖 Generated with Claude Code

[HappySamuel]

The asan/tsan failures are the existing jazzy CI issue: the rolling job can't find eigen3_cmake_module when configuring rmf_task, the same failure as on #152, so nothing gets built.

The rewind change and its tests are the same in #156 (lyrical), whose CI passes on all four distros, including the jazzy/noble asan and tsan jobs. This branch also builds and passes all test_rmf_task_sequence tests locally on Ubuntu 24.04 / Jazzy.

[HappySamuel]

@aaronchongth @mxgrey A gentle ping on the backports of #154 (the rewind_task fix for #146): this PR for jazzy and #156 for lyrical. Both are clean cherry-picks of the merged commit.

• Backport #154 to lyrical #156 (lyrical): all checks pass.
• Backport #154 to jazzy #155 (jazzy): the only failures are the two rolling asan/tsan jobs, which can't find eigen3_cmake_module. That's the existing jazzy CI issue, and Backport #130, #133, #134, #135 to jazzy #152 merged with the same failure. The jazzy jobs pass.

If you'd rather fold this into your next batch backport like #152, that works for me too.

[aaronchongth]

@HappySamuel sorry for the late reply, I just learnt that our CI was disabled due to inactivity for 6 months+ (commits), even scheduled workflows were disabled automatically. I just enabled them back up, can you run an empty commit to trigger CI again? Thanks!

edit: this applies to the backport to lyrical PR as well

[HappySamuel]

@aaronchongth Thanks for re-enabling CI! I pushed an empty commit to both PRs. #156 (lyrical) is now all green.

On this one, all three jobs fail before rmf_task is compiled. The cause is the jazzy branch's workflows: they still build against rolling on noble, and since rolling moved to resolute, rosdep can't resolve anything on noble. That's also why the build job failed this time, after passing on Nov 3. (Correcting my earlier comment: there was never a jazzy job here, only rolling ones.)

I opened #157 to switch the jazzy branch workflows to jazzy / noble, like rmf_ros2's jazzy branch. Once that's in, CI on this PR should pass.

[aaronchongth]

that makes sense, thanks for setting that up. I've merged #157, you would probably need to rebase, to get this CI to run properly

[aaronchongth]

LGTM