I've posted some questions worth discussing here : #348 (comment)

@adityapande-1995 @clalancette I have added a more comprehensive sample code in commit 5aa51d7

@adityapande-1995 @clalancette
Hello and sorry for the comment. Is there anything I can do to help with the review of this pull request?

@osrf-jenkins retest this please

@adityapande-1995 @clalancette
Let me present a thought experiment that might make clearer what I am trying to achieve with this pull request.

How would you test a node that is subscribed to a topic with a single int field, that publishes the same topic where the int was incremented by one?

@osrf-jenkins retest this please

@adityapande-1995 @methylDragon
Friendly reminder about this PR.

Until now, I have been using my fork at work for our project, but it would give us more confidence if we could just use the official repository, or even better if this PR could end up in the released version on https://index.ros.org/packages/.

@osrf-jenkins retest this please

@LastStarDust my apologies, taking a look now

@adityapande-1995 @methylDragon I fixed some linting errors and now the checks are finally passing.

Hello, could we get a maintainer to look at this PR?

@methylDragon is OOO, but I will ping him to look when he is back.

@Ryanf55 Hello, I would like to resume working on this PR. Could you please take a look at the conversations you opened and resolve the ones you think are resolved?

@osrf-jenkins retest this please

@Ryanf55 Hello, I would like to resume working on this PR. Could you please take a look at the conversations you opened and resolve the ones you think are resolved?

Yep, all of the issues we found are still problems, and we never found a good workaround for doing this generically in humble. For now, we disabled our python automated testing as a requirement for PR's because this problem is so difficult to deal with, and have focused testing without the middleware. Once we can move to Jazzy, then the event handlers should make this kind of testing much more reliable because you can do a proper wait on the interfaces to come up.

Sorry, we've just spent an exorbitant amount of time on this trying to get this reliable in humble and could never figure out a way that wasn't extremely coupled to knowing exactly what was running and manual graph introspection hard-coded into our test framework. I'm optimistic that upgrading distros will make it easier, but this way of testing in python is not something we're going to pursue any further unless anyone can give us a reasonable approach that works without hard-coded timeouts/sleeps for discovery that fail on slower runners.

@Ryanf55 Thank you for sharing the situation. I am sorry we could not figure out how to make this work reliably in Humble.
Anyway, can I resolve the discussions opened by you in this PR, since it applies to rolling?

@Ryanf55 Thank you for sharing the situation. I am sorry we could not figure out how to make this work reliably in Humble. Anyway, can I resolve the discussions opened by you in this PR, since it applies to rolling?

Looks like I don't have authority to resolve threads, but consider the comments OBE. I'd be happy for this to be merged as-is, and then a follow up to go in that adds the event support.

@adityapande-1995 @methylDragon @mjcarroll
It took me a while to figure it out, but the unit tests are failing because in the CI environment the ros2 command does not support the topic argument. However, until some months ago the tests were passing (as they are passing on my machine) so something might have changed in the CI environment.

Here is the offending code:

def trigger_callback():
    command = 'ros2 topic pub --once --max-wait-time-secs 10 --keep-alive 1 \
        /input std_msgs/msg/String "data: Hello World"'
    p = Popen(command, shell=True, stdout=PIPE, stderr=PIPE)
    stdout, stderr = p.communicate()
    print("stdout: ", stdout)
    print("stderr: ", stderr)
    print('Callback triggered')

and here is the CI output:

15:06:12 stderr:  b"usage: ros2 [-h] [--use-python-default-buffering]\n            Call `ros2 <command> -h` for more detailed usage. ...\nros2: error: argument Call `ros2 <command> -h` for more detailed usage.: invalid choice: 'topic' (choose from 'daemon', 'extension_points', 'extensions', 'pkg')\n"

Please let me know how you want me to proceed. If the CI environment should be fixed or I should avoid using the ros2 topic command.

It took me a while to figure it out, but the unit tests are failing because in the CI environment the ros2 command does not support the topic argument.

That's because there is no test_depend on ros2topic, so it is not available in the environment.

That said, I don't think we can introduce that dependency. In particular, I'm worried about a circular dependency here, since ros2topic already depends on launch_ros. If you are only trying to publish a message, then I'd say it is easier to use the rclpy APIs directly.

@clalancette
Thank you so much for the suggestion.
@adityapande-1995 @methylDragon @mjcarroll @clalancette
Now all tests are passing. You can resume the review.

@mjcarroll Friendly ping for review.

@mjcarroll Friendly ping for review.