Nikolai Morin

Results 124 comments of Nikolai Morin

This issue hinges on #126.

One thing I realized is that if we don't use a global context, component nodes will have to communicate across participants, which might be more expensive, not sure.

Backtrace: ``` #0 0x00007ff7b5e56b86 in eprosima::fastrtps::types::TypeIdentifier::operator=(eprosima::fastrtps::types::TypeIdentifier const&) () from /opt/ros/rolling/lib/libfastrtps.so.2.6 #1 0x00007ff7b5e8569b in ?? () from /opt/ros/rolling/lib/libfastrtps.so.2.6 #2 0x00007ff7b5e7f377 in register_builtin_annotations_types(eprosima::fastrtps::types::TypeObjectFactory*) () from /opt/ros/rolling/lib/libfastrtps.so.2.6 #3 0x00007ff7b5e74e63 in eprosima::fastrtps::types::TypeObjectFactory::get_instance() () from...

@ros2-rust/dev We should see if we can find an upstream ticket for this and then close this.

Potentially related to https://github.com/eProsima/Fast-DDS/issues/2635

Let's exclude `rclrs_examples` for now too.

Forgot to close this!

@Nizerlak Ok, did you want to present an idea of your own or should I expand on my comment above?

The log crate mentioned above already defines logging macros. We'd just have to implement the backend that logs to rosout.

Yes, `rcl` should definitely be used. Good point about the `FATAL` severity. That probably means `log` is out. We'll need to think about what we want our API to look...