Full Rebuild (Sync) of January 2026: bump ros2-distro-mutex to 0.13.0 and build_number to 14

[traversaro]

No description provided.

[wep21]

need to wait this migration?
https://conda-forge.org/status/migration/?name=fmt12_spdlog116

[traversaro]

Actually, indeed pcl and vtk have not migrated to spdlog 1.16, see conda-forge/vtk-feedstock#413 (I am trying to unblock this in conda-forge/vtk-feedstock#417). How was it possible for PCL to be installed here?

Ok, I think by demangling the problematic symbol, the issue is more clear:

spdlog::details::log_msg::log_msg(spdlog::source_loc, fmt::v12::basic_string_view<char>, spdlog::level::level_enum, fmt::v12::basic_string_view<char>)

rclpy expects spdlog with fmt 12, but somehow fmt 11 is installed instead, resulting in error. However, this should be handled at the level of runexports of fmt in the librcl_logging_spdlog library, perhaps there is some missing rosdep key or fmt dep.

[traversaro]

However, this should be handled at the level of runexports of fmt in the librcl_logging_spdlog library, perhaps there is some missing rosdep key or fmt dep.

Ok, this is indeed the output of:

2025-11-30T15:28:23.0502526Z  │ │ [lib/librcl_logging_spdlog.so] links against:
2025-11-30T15:28:23.0503013Z  │ │  ├─ lib/librcutils.so (ros-jazzy-rcutils)
2025-11-30T15:28:23.0503757Z  │ │  ├─ lib/libgcc_s.so.1 (libgcc)
2025-11-30T15:28:23.0504282Z  │ │  ├─ lib/librcl_logging_interface.so (ros-jazzy-rcl-logging-interface)
2025-11-30T15:28:23.0504798Z  │ │  ├─ libpthread.so.0 (system)
2025-11-30T15:28:23.0505139Z  │ │  ├─ libc.so.6 (system)
2025-11-30T15:28:23.0505490Z  │ │  ├─ lib/libstdc++.so.6.0.34 (libstdcxx)
2025-11-30T15:28:23.0505884Z  │ │  ├─ lib/libspdlog.so.1.16.0 (spdlog)
2025-11-30T15:28:23.0506330Z  │ │  └─ lib/librcpputils.so (ros-jazzy-rcpputils)

So fmt is not directly linked, but as spdlog ABI depends on fmt, we need anyhow to ensure that fmt is always a host dep if spdlog is a host dep, so in 7f5da0c I added fmt to spdlog rosdep key.

[traversaro]

Ok, now it fails at package installation time as expected. This should unblocked by conda-forge/vtk-feedstock#417 and conda-forge/vtk-feedstock#413 .

[traversaro]

Checked in .pixi folder. #136

6df054a

* [ament_clang_format (tidy?) doesn't work on Windows #98](https://github.com/RoboStack/ros-jazzy/issues/98)

I backported ament/ament_lint#563 in aeef077, fyi @tony-p

[traversaro]

#146
#147
#144
#127
#143
#145

05a9ed5

[traversaro]

• When ros-jazzy-desktop-full is installed, launching rviz2 command from terminal on Windows does not work, while calling ros2 run rviz2 rviz2 works #57

@gftabor if you want to check this pre-merge, I can also enable the SAVE_CACHE_AS_ARTIFACT option in the CI job.

[gftabor]

• When ros-jazzy-desktop-full is installed, launching rviz2 command from terminal on Windows does not work, while calling ros2 run rviz2 rviz2 works #57

@gftabor if you want to check this pre-merge, I can also enable the SAVE_CACHE_AS_ARTIFACT option in the CI job.

Ya I can test it. Presumably it would make a conda channel somewhere?

[traversaro]

• When ros-jazzy-desktop-full is installed, launching rviz2 command from terminal on Windows does not work, while calling ros2 run rviz2 rviz2 works #57

@gftabor if you want to check this pre-merge, I can also enable the SAVE_CACHE_AS_ARTIFACT option in the CI job.

Ya I can test it. Presumably it would make a conda channel somewhere?

No, the option itself just saves the cache of .conda files in a .zip artifacts. Then you can use it as a local conda channel, or upload it on a test channel (for example on prefix.dev).

[traversaro]

#141
#124

For this two, I would need to more time to investigate, probably we can investigate them post-build as anyhow they would probably not need full rebuild to add a patch.

[Tobias-Fischer]

I think for osx-arm64, we are just missing a full rebuild so the mavlink patch gets applied.

On osx-64 there is a weird iconv issue – I tried to find a solution, but failed. Personally, I think we can stop supporting – people still working on Intel Macs probably have other issues than getting ROS to run :P

[traversaro]

On osx-64 there is a weird iconv issue – I tried to find a solution, but failed. Personally, I think we can stop supporting – people still working on Intel Macs probably have other issues than getting ROS to run :P

I think the actual issues is:

2026-01-20T00:06:50.7862980Z  │ │ Undefined symbols for architecture x86_64:
2026-01-20T00:06:50.7864370Z  │ │   "GeographicLib::Geocentric::Geocentric(double, double)", referenced from:
2026-01-20T00:06:50.7866460Z  │ │       mavros::extra_plugins::FakeGPSPlugin::FakeGPSPlugin(std::__1::shared_ptr<mavros::uas::UAS>) in fake_gps.cpp.o
2026-01-20T00:06:50.7869200Z  │ │       mavros::extra_plugins::GuidedTargetPlugin::gp_origin_cb(std::__1::shared_ptr<geographic_msgs::msg::GeoPointStamped_<std::__1::allocator<void>>>) in guided_target.cpp.o
2026-01-20T00:06:50.7873180Z  │ │       mavros::extra_plugins::GuidedTargetPlugin::handle_position_target_global_int(mavlink::__mavlink_message const*, mavlink::common::msg::POSITION_TARGET_GLOBAL_INT&, mavros::plugin::filter::SystemAndOk) in guided_target.cpp.o
2026-01-20T00:06:50.7877570Z  │ │   "void tf2::fromMsg<geometry_msgs::msg::Transform_<std::__1::allocator<void>>, Eigen::Transform<double, 3, 2, 0>>(geometry_msgs::msg::Transform_<std::__1::allocator<void>> const&, Eigen::Transform<double, 3, 2, 0>&)", referenced from:
2026-01-20T00:06:50.7881160Z  │ │       mavros::extra_plugins::FakeGPSPlugin::mocap_tf_cb(std::__1::shared_ptr<geometry_msgs::msg::TransformStamped_<std::__1::allocator<void>>>) in fake_gps.cpp.o
2026-01-20T00:06:50.7884280Z  │ │       mavros::extra_plugins::FakeGPSPlugin::transform_cb(geometry_msgs::msg::TransformStamped_<std::__1::allocator<void>> const&) in fake_gps.cpp.o
2026-01-20T00:06:50.7886930Z  │ │   "GeographicLib::Geocentric::IntForward(double, double, double, double&, double&, double&, double*) const", referenced from:
2026-01-20T00:06:50.7889220Z  │ │       mavros::extra_plugins::FakeGPSPlugin::FakeGPSPlugin(std::__1::shared_ptr<mavros::uas::UAS>) in fake_gps.cpp.o
2026-01-20T00:06:50.7892620Z  │ │       mavros::extra_plugins::GuidedTargetPlugin::handle_position_target_global_int(mavlink::__mavlink_message const*, mavlink::common::msg::POSITION_TARGET_GLOBAL_INT&, mavros::plugin::filter::SystemAndOk) in guided_target.cpp.o
2026-01-20T00:06:50.7896090Z  │ │   "GeographicLib::Geocentric::IntReverse(double, double, double, double&, double&, double&, double*) const", referenced from:
2026-01-20T00:06:50.7898800Z  │ │       mavros::extra_plugins::FakeGPSPlugin::send_fake_gps(rclcpp::Time const&, Eigen::Matrix<double, 3, 1, 0, 3, 1> const&) in fake_gps.cpp.o
2026-01-20T00:06:50.7901750Z  │ │       mavros::extra_plugins::GuidedTargetPlugin::gp_origin_cb(std::__1::shared_ptr<geographic_msgs::msg::GeoPointStamped_<std::__1::allocator<void>>>) in guided_target.cpp.o
2026-01-20T00:06:50.7904220Z  │ │   "GeographicLib::Geoid::height(double, double) const", referenced from:
2026-01-20T00:06:50.7908440Z  │ │       mavros::extra_plugins::CameraPlugin::handle_camera_image_captured(mavlink::__mavlink_message const*, mavlink::common::msg::CAMERA_IMAGE_CAPTURED&, mavros::plugin::filter::SystemAndOk) in camera.cpp.o
2026-01-20T00:06:50.7913320Z  │ │       mavros::extra_plugins::FakeGPSPlugin::send_fake_gps(rclcpp::Time const&, Eigen::Matrix<double, 3, 1, 0, 3, 1> const&) in fake_gps.cpp.o

the path of least resistance is just to remove mavros_extras from osx-64. On dropping osx-64, I would like to do some announcments on that somehow, not sure what is the best place to do that (social media? website?).

[Tobias-Fischer]

On osx-64 there is a weird iconv issue – I tried to find a solution, but failed. Personally, I think we can stop supporting – people still working on Intel Macs probably have other issues than getting ROS to run :P

the path of least resistance is just to remove mavros_extras from osx-64. On dropping osx-64, I would like to do some announcments on that somehow, not sure what is the best place to do that (social media? website?).

Yes, that is a lot neater and works for me :)

[traversaro]

Perfect, now also osx-arm64 fails with:

2026-01-20T10:34:44.9939620Z  │ │ Undefined symbols for architecture arm64:
2026-01-20T10:34:44.9950980Z  │ │   "GeographicLib::Geocentric::Geocentric(double, double)", referenced from:
2026-01-20T10:34:44.9952200Z  │ │       mavros::extra_plugins::FakeGPSPlugin::FakeGPSPlugin(std::__1::shared_ptr<mavros::uas::UAS>) in fake_gps.cpp.o
2026-01-20T10:34:44.9953390Z  │ │       mavros::extra_plugins::GuidedTargetPlugin::gp_origin_cb(std::__1::shared_ptr<geographic_msgs::msg::GeoPointStamped_<std::__1::allocator<void>>>) in guided_target.cpp.o
2026-01-20T10:34:44.9955740Z  │ │       mavros::extra_plugins::GuidedTargetPlugin::handle_position_target_global_int(mavlink::__mavlink_message const*, mavlink::common::msg::POSITION_TARGET_GLOBAL_INT&, mavros::plugin::filter::SystemAndOk) in guided_target.cpp.o
2026-01-20T10:34:44.9958110Z  │ │   "void tf2::fromMsg<geometry_msgs::msg::Transform_<std::__1::allocator<void>>, Eigen::Transform<double, 3, 2, 0>>(geometry_msgs::msg::Transform_<std::__1::allocator<void>> const&, Eigen::Transform<double, 3, 2, 0>&)", referenced from:
2026-01-20T10:34:44.9978570Z  │ │       mavros::extra_plugins::FakeGPSPlugin::mocap_tf_cb(std::__1::shared_ptr<geometry_msgs::msg::TransformStamped_<std::__1::allocator<void>>>) in fake_gps.cpp.o
2026-01-20T10:34:44.9980540Z  │ │       mavros::extra_plugins::FakeGPSPlugin::transform_cb(geometry_msgs::msg::TransformStamped_<std::__1::allocator<void>> const&) in fake_gps.cpp.o
2026-01-20T10:34:44.9996210Z  │ │   "GeographicLib::Geocentric::IntForward(double, double, double, double&, double&, double&, double*) const", referenced from:
2026-01-20T10:34:44.9998730Z  │ │       mavros::extra_plugins::FakeGPSPlugin::FakeGPSPlugin(std::__1::shared_ptr<mavros::uas::UAS>) in fake_gps.cpp.o
2026-01-20T10:34:45.0002800Z  │ │       mavros::extra_plugins::GuidedTargetPlugin::handle_position_target_global_int(mavlink::__mavlink_message const*, mavlink::common::msg::POSITION_TARGET_GLOBAL_INT&, mavros::plugin::filter::SystemAndOk) in guided_target.cpp.o
2026-01-20T10:34:45.0004850Z  │ │   "GeographicLib::Geocentric::IntReverse(double, double, double, double&, double&, double&, double*) const", referenced from:
2026-01-20T10:34:45.0014360Z  │ │       mavros::extra_plugins::FakeGPSPlugin::send_fake_gps(rclcpp::Time const&, Eigen::Matrix<double, 3, 1, 0, 3, 1> const&) in fake_gps.cpp.o
2026-01-20T10:34:45.0018340Z  │ │       mavros::extra_plugins::GuidedTargetPlugin::gp_origin_cb(std::__1::shared_ptr<geographic_msgs::msg::GeoPointStamped_<std::__1::allocator<void>>>) in guided_target.cpp.o
2026-01-20T10:34:45.0019320Z  │ │   "GeographicLib::Geoid::height(double, double) const", referenced from:
2026-01-20T10:34:45.0020810Z  │ │       mavros::extra_plugins::CameraPlugin::handle_camera_image_captured(mavlink::__mavlink_message const*, mavlink::common::msg::CAMERA_IMAGE_CAPTURED&, mavros::plugin::filter::SystemAndOk) in camera.cpp.o

[traversaro]

• When ros-jazzy-desktop-full is installed, launching rviz2 command from terminal on Windows does not work, while calling ros2 run rviz2 rviz2 works #57

@gftabor if you want to check this pre-merge, I can also enable the SAVE_CACHE_AS_ARTIFACT option in the CI job.

Ya I can test it. Presumably it would make a conda channel somewhere?

The artifacts containg the all conda package built can found under "artifacts" in https://github.com/RoboStack/ros-jazzy/actions/runs/21168426529, (for win-64) can be found at https://github.com/RoboStack/ros-jazzy/actions/runs/21168426529/artifacts/5188060468 . Let me know if you can't access that files, I can upload them somewhere else.

[gftabor]

@traversaro , I confirmed this closes #57 and tested #125 while I was there. LGTM

[traversaro]

@traversaro , I confirmed this closes #57 and tested #125 while I was there. LGTM

Great! @Tobias-Fischer ready for me, feel free to merge if it is ok for you.

[Tobias-Fischer]

Yay!

[Tobias-Fischer]

Our new test repeatedly fails for osx-arm64, blocking package uploads:

│ Installing test environment
 │ ✔ Successfully updated the test environment
 │ Testing commands:
 │ [INFO] [launch]: All log files can be found below /Users/runner/.ros/log/2026-01-21-22-33-08-701961-sjc20-bb713-cfce5978-1427-4fa0-8c99-2d02f419a456-F632AC8EBA91.local-4588
 │ [INFO] [launch]: Default logging verbosity is set to INFO
 │ [INFO] [robot_state_publisher-1]: process started with pid [4590]
 │ [robot_state_publisher-1] Problem to set affinity of thread with id [0x16b543000,dds.log] to value 0. Error 'Protocol family not supported'
 │ [robot_state_publisher-1] 2026-01-21 22:33:08.913 [SYSTEM Error] Problem to set affinity of thread with id [0x16b4b7000,dds.shm.wdog] to value 0. Error 'Protocol family not supported' -> Function configure_current_thread_affinity
 │ [robot_state_publisher-1] 2026-01-21 22:33:08.916 [SYSTEM Error] Problem to set affinity of thread with id [0x16b5cf000,dds.ev.0] to value 0. Error 'Protocol family not supported' -> Function configure_current_thread_affinity
 │ [robot_state_publisher-1] 2026-01-21 22:33:08.916 [SYSTEM Error] Problem to set affinity of thread with id [0x16b65b000,dds.udp.7410] to value 0. Error 'Protocol family not supported' -> Function configure_current_thread_affinity
 │ [robot_state_publisher-1] 2026-01-21 22:33:08.916 [SYSTEM Error] Problem to set affinity of thread with id [0x16b6e7000,dds.udp.7400] to value 0. Error 'Protocol family not supported' -> Function configure_current_thread_affinity
 │ [robot_state_publisher-1] 2026-01-21 22:33:08.919 [SYSTEM Error] Problem to set affinity of thread with id [0x16b773000,dds.shm.7411] to value 0. Error 'Protocol family not supported' -> Function configure_current_thread_affinity
 │ [robot_state_publisher-1] 2026-01-21 22:33:08.919 [SYSTEM Error] Problem to set affinity of thread with id [0x16b7ff000,dds.udp.7411] to value 0. Error 'Protocol family not supported' -> Function configure_current_thread_affinity
 │ [robot_state_publisher-1] 2026-01-21 22:33:08.922 [SYSTEM Error] Problem to set affinity of thread with id [0x16b88b000,dds.asyn.0.0] to value 0. Error 'Protocol family not supported' -> Function configure_current_thread_affinity
 │ [robot_state_publisher-1] 2026-01-21 22:33:09.165 [SYSTEM Error] Problem to set affinity of thread with id [0x16b917000,dds.shm.7000] to value 0. Error 'Protocol family not supported' -> Function configure_current_thread_affinity
 │ [robot_state_publisher-1] 2026-01-21 22:33:09.166 [SYSTEM Error] Problem to set affinity of thread with id [0x16b9a3000,dds.udp.7000] to value 0. Error 'Protocol family not supported' -> Function configure_current_thread_affinity
 │ test_node_output (robot_state_publisher_smoke_test_launch.TestRobotStatePublisher.test_node_output) ... FAIL
 │ ======================================================================
 │ FAIL: test_node_output (robot_state_publisher_smoke_test_launch.TestRobotStatePublisher.test_node_output)
 │ ----------------------------------------------------------------------
 │ Traceback (most recent call last):
 │   File "$PREFIX/etc/conda/test-files/ros-jazzy-robot-state-publisher/0/robot_state_publisher_smoke_test_launch.py", line 45, in test_node_output
 │     proc_output.assertWaitFor(
 │   File "$PREFIX/lib/python3.12/site-packages/launch_testing/io_handler.py", line 146, in assertWaitFor
 │     assert success, 'Waiting for output timed out'
 │            ^^^^^^^
 │ AssertionError: Waiting for output timed out
 │ ----------------------------------------------------------------------
 │ Ran 1 test in 0.510s
 │ FAILED (failures=1)

[Tobias-Fischer]

I also don't understand why we don't use cyclonedds, we have https://github.com/RoboStack/ros-jazzy/blob/main/patch/ros-jazzy-rwm-implementation-cmake.osx.patch .. and that error should only happen with fastrtps?!?

[Tobias-Fischer]

Now I'm wondering if we need to backport eProsima/Fast-DDS#5971 which hasn't landed in fastrtps 2.14.5 which is used in jazzy