Tomoya Fujita

@mikeferguson good catch. as @clalancette mentioned, this is one of the corner case, i think. and the problem is technically `cross-distribution` communication, which is not supported or recommended. I am...

with your [SSCCE](http://sscce.org/), i can not find the problem. publisher and subscription works fine. ```bash root@tomoyafujita:~/ros2_ws/colcon_ws# ros2 run prover_rclcpp examples_368_sub [INFO] [1702660421.069453964] [DualThreadedNode]: THREAD 13945239655426654374 => Heard 'sss0' at 1702660421069439658...

@alsora just checking my understanding. So if we just rely on RMW `LoanedMessage` for intra-process communication w/o rclcpp one, this should be no problem? since that can be managed by...

AFAIK, no progress. @carmiac are you going to work on this? i think what we need to do is, - add lifecycle subscriber test in this PR. - add `TODO`...

> If the client is terminated, I would expect a CancelCallback. action server does not have connectivity map, so it does not have any idea which action client is disconnected....

> By terminated I mean it crashes, hard. So the dtor would not be called any more. I do not really understand this means. what leads to the crash?

@jmachowinski thanks for sharing your thoughts, appreciate it. > any programming error, e.g. access of a nullptr followed by a segfault. there will be a signal or exception to catch...

how about memory usage on subscription process, i would check the statistics for the process space. besides, if publisher and subscription are in the same host, it uses `shared memory...

> When the program crashes that is new, so the listener crash and you have core file? it would be easier to see the stack trace where exactly it crashes?...

> I deployed both the listener and the talker inside the same container. that brings me an another question, why shared memory transport is used... do you apply fast-dds configuration...