Indicate missing support for unique network flows

[anamud]

This PR indicates missing support in rmw_connextdds for unique network flows for publishers and subscribers in communicating nodes. Such indication is required for graceful error handling in the ROS2 RMW and above layers.

Associated pull-requests:

• design
• examples
• rclcpp
• rcl
• rmw
• rmw_implementation
• rmw_fastrtps
• rmw_cyclonedds
• rmw_connext (NB: This repository is soon to be deprecated)
• Fast DDS

Initial contributions stem from Ericsson and eProsima.

Signed-off-by: Ananya Muddukrishna ananya.x.muddukrishna@ericsson.com

[asorbini]

Hi @anamud, thank you for your submission!

I left a couple of comments to make sure this code works with both rmw_connextdds and rmw_connextddsmicro, and with older versions of ROS2.

Once those are addressed, this PR will be ready to be merged, although it will need to be held off until ros2/rmw#294 is merged.

[asorbini]

Thank you for the updates @anamud!

The PR looks good now. I think two feature flags is a bit overkill, but I'm going to get rid of them altogether soon, so we can leave that as is.

At this point, I would only wait for ros2/rmw#294 to be merged.

[asorbini]

I created #15 to get rid of all "feature flags" in master. If you want, I think you could rebase your changes on asorbini/rolling-no-ff (or rebase them from master, once that PR is merged).

Apologies for the back and forth, but the recent streak of PRs made me realize getting rid of these flags had more priority than I thought.

[anamud]

I created #15 to get rid of all "feature flags" in master. If you want, I think you could rebase your changes on asorbini/rolling-no-ff (or rebase them from master, once that PR is merged).

Apologies for the back and forth, but the recent streak of PRs made me realize getting rid of these flags had more priority than I thought.

@asorbini: Thanks for the review. I understand the rationale to get rid of the flags. I prefer to wait until the PR is merged before rebasing.

[asorbini]

#15 was merged, so you can rebase from master now and remove the feature flags

[anamud]

#15 was merged, so you can rebase from master now and remove the feature flags

Thanks, @asorbini for the notice. I have now rebased to master and updated the PR. I removed conditional compilation altogether assuming a rolling release model.

[asorbini]

LGTM.

It would be good to run CI with the other changed repositories, but I'm assuming that will happen as part of merging of (one of the) other PRs (let me know and I can also start some jobs).

In general, the PR can be merged once ros2/rmw#294 is merged.

[anamud]

@asorbini @wjwwood I removed a conflicting linkage error reported by the CI system. Please confirm that that change is correct.

I am unable to test this PR out properly due to lack of a licensed connextdds system at my side. Can you please suggest an alternative test method?

Since the PRs related to unique network flows are now approved, I would appreciate it if you can start a CI job at your side. Here is an updated ros2 rolling repository list that I use for testing.

[asorbini]

Hi @anamud, the change looks good to me, I think it's fine to have these function have C++ linkage, apologies for not spotting that in my earlier reviews.

About the licensing issue: I'm not a legal expert and this is not legal advice, but maybe this ROS 2 development work counts towards "research purposes" covered by the "non-commercial license" included in Debian package rti-connext-dds-5.3.1. If that is the case, than you could also submit a request for a permanent "reserach and non-commercial use" license from this RTI webpage and have access to Connext for other platforms too. This would also give you access to newer versions than 5.3.1.

Here's some CI jobs to validate the changes using your list:

• Linux:
• macOS:
• Windows: