This is currently introducing a regression on rmf_visualization in Rolling, which pops up all the way to rmf_demos https://discourse.openrobotics.org/t/preparing-for-rolling-sync-2026-01-16/51785.

I tested this PR in Rolling + testing and it seems to build fine, if anyone else wants to test it the procedure would be:

• Try it under Jazzy to make sure there are no regressions.
• Try it under the upcoming Rolling binary packages by enabling ros2-testing as documented here

Without the PR Rolling should fail, with it should be OK

Tested with the two approaches suggested:

• with a Jazzy distrobox
• with a Rolling distrobox, then enabling ros2-testing

Can confirm that the build works for both! I'm able to run the usual demos in Jazzy. However gz keeps crashing if I try to run demos in Rolling - can't tell if that is expected behavior or not. Either way the build issue for rviz seems to be fixed, as it was able to launch properly.

However gz keeps crashing if I try to run demos in Rolling - can't tell if that is expected behavior or not. Either way the build issue for rviz seems to be fixed, as it was able to launch properly.

Ah good find, it seems to be a very hairy issue with launch substitutions for the headless parameter, I'll try to figure out whether it can be fixed either in rmf_demos or upstream

I ran into a build issue while testing this out, although it's not the fault of this PR, nor anything we can fix within rmf_visualization. It seems to be a faulty config-file in rviz.

I've made a comment describing the problem on the issue ticket related to dropping Qt5. If we see users complaining that rmf_visualization is no longer building, we will probably need to advise them to uninstall Qt5 until this gets resolved upstream.