Michel Hidalgo

@bpwilcox yes! I'll take a look at these tomorrow morning first thing. Sorry for the insane delay, I've been too busy.

I'd say user error.

A few comments here and there: > If the context is shutdown in the middle of the call to rclcpp::ok and the construction of the executor in rclcpp::spin_some, then the...

> if the usage of std::once is correct here Use of `std::call_once` seems correct to me too. > IMO, this may be a case of spooky action at distance Hmm,...

> In CI, it happens frequently that flaky tests are more prone to fail on Windows (and with rmw_connext). For time sensitive tests, yes, but this looks different (considering above...

This needs a rebase and CI to get into Rolling.

@mjcarroll I migrated the `two_node_pipeline` demo, but not the rest. I was unsure to what extent we wanted to migrate. WDYT?

I came across the same thing. @claireyywang try pressing `x`... The thing is that these demos spin up a separate thread to support `stdin`-based user interaction. Because of the way...

This might be a duplicate of #290. If nothing is published on the topic, the viewer becomes unresponsive to user input. @gonzodepedro could you validate this?

@clalancette do these processes behave upon SIGTERM on Windows?