Michel Hidalgo

> when a thirdparty contributor releases a new message package or when an existing message is updated, geometry_msgs_java won't be automatically updated Correct. There wouldn't be any special logic for...

Alright, https://github.com/ros2/design/pull/310 is up. Get there and bash it till its rough edges are gone!

Discussion at https://github.com/ros2/design/pull/310 is still on-going, but I think we all agree about the need for a unified code generation CLI. I'll open up a ticket to track the progress...

> In order to support length checking, a new string class seems to be necessary. Indeed, a `bounded_basic_string` with checks upon construction and assignment would do. We should make it...

IIRC this has not been implemented. This issue is about changing typesupport packages to provide API that can help `rcl` interact with action-specific fields in type erased action goals, results,...

This should be straightforward to implement. @xander-m2k contributions are most welcome.

Hmm, this is indeed peculiar. It's not immediately obvious to me why [`sizeof(...)`](https://github.com/ros2/rosidl/blob/05ef0b5c43450784c0e3fe0ebc97e8ce2cf1d881/rosidl_typesupport_introspection_c/resource/msg__type_support.c.em#L247) isn't doing what it is supposed to.

Indeed. Switching to type erased function pointers should do the trick.

As @clalancette, `inject_on_return` works everywhere but on RHEL/CentOS. For reasons that remain obscure to me, unlike other Linux distributions, binaries in CentOS yield `.plt` trampoline addresses in lieu of that...

Thank you for the investigation @y-okumura-isp. I'll try to come back to this later on. If you're right, perhaps we can export these flags from `mimick` targets directly.