The ROS PMC meeting for this week was on Tuesday. The notes from the meeting are available at ROS PMC weekly meeting agenda
PMC Business
- No formal business
Ongoing Tasks
-
Buildfarm:
- Buildfarm issues triage sheet
- Buildfarm greenness report
- Issues:
- Regression introduced in packaging by enabling rosbag2_performance_benchmarking still affecting RHEL
- @mjcarroll to profile
rviz_default_pluginsbuild to attempt to reduce memory demands - @cottsay to artifically inject dependency to ensure that
rosbag2_performance_benchmarkingandrviz_default_pluginsdon’t run simultaneously
- @mjcarroll to profile
- [image_transport] Nightly test failure with rmw_zenoh due to SystemDefaultsQoS assumptions (Reported by Yadu) - Similar to a failure in
pointcloud_transport, there is a mismatch in test expectations that needs to be addressed
- Regression introduced in packaging by enabling rosbag2_performance_benchmarking still affecting RHEL
-
Waffle assignments:
- Waffle is a weekly meeting where maintainers triage incoming issues and pull requests that haven’t been otherwise acted upon. One person per week is assigned to be the “waffle note-taker” to both take notes in the meeting to keep the meeting as efficient as possible.
- July 31 + August 7: @MichaelOrlov
-
Rosdistro Assignments
- Rosdistro is a weekly assignment to monitor incoming issues and pull requests in GitHub - ros/rosdistro: This repo maintains a lists of repositories for each ROS distribution. Two people per week are assigned to be the rosdistro maintainers.
- August 12 - August 19: @nuclearsandwich + @mjcarroll
- Outgoing rosdistro issues:
- [@emersonknapp] Add pip version of docker package by bastianschildknecht · Pull Request #46946 · ros/rosdistro · GitHub
- We shouldn’t be adding
-pippackages that can conflict with system installed apt packages. - [@nuclearsandwich] The best way way to accopmlish this is to supplement your rosdep sources
- [@cottsay] We should also add a check to the rosdistro reviewer bot: Flag pip rules when there is already a system package rule · Issue #48 · ros-infrastructure/rosdistro-reviewer · GitHub
- We shouldn’t be adding
- [@emersonknapp] Series of 403 errors in rosdistro builds related to the
openembeddedpackage index.- [@cottsay] This check has been disabled, as we are getting throttled, but unclear from where (perhaps cloudflare?). Already looped in openembedded folks
- [@mjcarroll] add 'open3d' to rolling by christian-rauch · Pull Request #47284 · ros/rosdistro · GitHub - should this have the
_vendorsuffix?- [@MichaelOrlov] Internally, we call things vendor even if it’s being unconditionally built from upstream sources
- [@cottsay] This dependency, specifically, will be available in
plucky, and there are talks of it being included in Fedora. We want to be sure not to conflict with those dependencies - [@mjcarroll] I think the distinction here is that the OMPL package, referenced in the PR, actually has a
package.xmlin the repository, while theopen3dpackage is downloading and injecting apackage.xml(a vendored build)
- [@emersonknapp] Add pip version of docker package by bastianschildknecht · Pull Request #46946 · ros/rosdistro · GitHub
ROS Bosses
- ROS Humble (@christophebedard/@audrow)
- ROS Jazzy (@marcogg)
- ROS Kilted (@cottsay)
- ROS Rolling (@Yadunund)
- No updates
Agenda Items
- [@mjcarroll] Logging is kind of slow, even at levels below the level selected for the node. · Issue #2928 · ros2/rclcpp · GitHub
- Summary: When logging with the
/rosoutsink enabled, logging adds a lot of latency versus ROS 1, which is very noticeable when logging in critical sections. - [@christophebedard] We do disable
/rosoutlogging by default - [@JM_ROS] We do call
rcl_publishin the actual logging message, there isn’t any sort of buffering or queuing for log messages.- [@wjwwood] This can be further exacerbated by something like
rmw_zenoh, which (as far as I remember), doesn’t have a concept of async publishing, so all publish calls are blocking
- [@wjwwood] This can be further exacerbated by something like
- [@wjwwood] The intention here was to rely on the logging framework to do buffering/queuing of log messages, so it may be that the implementation doesn’t match the intention. Logging with
/rosoutshould likely not be blocking- [@JM_ROS] If we had a queue/buffer mechanism, it may also make sense to add a
LogMessageArraytype where the entire buffer could be flushed out in a single publish, rather than a sequence of individual messages.
- [@JM_ROS] If we had a queue/buffer mechanism, it may also make sense to add a
- @mjcarroll to check with @tomoyafujita who has recently been thinking about logging.
- Summary: When logging with the
- [@HemalShahNV] Native rcl::tensor type
- [@HemalShahNV] presented an overview and motivation for introducing a
rcl::tensortype, including why relying purely on type adaption may be insufficient to get the type of zero- or low- copy transport thatrcl::tensorwould solve - [@wjwwood] Proposed that it could also be used in conjunction with a custom typesupport generator. This custom typesupport could help prevent introducing parallel
tensor_msgstypes, which would maximize compatibility with existing code. - Team to read the design doc and provide feedback on discourse topic.
- [@HemalShahNV] presented an overview and motivation for introducing a
As a reminder, ROS PMC meetings are open to the public, though only committers and members may speak without being called on. If you have topics that you would like discussed, feel free to respond here.