Michel Hidalgo

@Crola1702 looks like the `io` component in `gz-common5` was not built/released.

Alright, I just stumbled upon this issue too. It's quite inconvenient, as pushing the service to a thread instead of pushing the tests (as @pbaughman had to do) simplifies things...

Ok, there seems to be a fundamental limitation in `asyncio` when it comes to subprocesses. At least on Unix platforms and as of Python 3.6, all child watchers rely on...

> I get this error: TypeError: eval substitution expects 1 argument That's because the Python expression substitution [expects a single argument](https://github.com/ros2/launch/blob/d648e19590d425e5d38411fe83fdb3b59f56d586/launch/launch/substitutions/python_expression.py#L52-L56). Making the argument a single string and escaping single...

> If this is the designed behavior, this is not compatible with ROS1 and I suggest that it would be documented somewhere. I agree with that statement i.e. to document...

> If we use that as a rule for all substitutions that take one argument, it won't be an special case I thought about it. It works great for `command`...

> I'm not sure if we can make that case more flexible, @hidmic do you think that would be possible? I'd think so, yeah. Though I haven't read that grammar...

@mhl787156 just like in ROS1, `eval` substitutions do not use `launch` context as their scope. You have to nest substitutions like `$(eval '$(val variable) == \'value\'')`. As a side note,...

@clalancette I'm inclined to think this would not be a problem if googling `'package api docs'` did not bring old stuff. Can we improve that somehow?

@clalancette @tfoote as another data point, I can't seem to find ROS 2 Humble documentation for interfaces (ie. messages, services, actions). Another discoverability issue? Or is `rosdoc2` lacking the feature...