asorbini
commented
3 years ago Two unit tests failed with this change:
1: [ RUN ] TestMessageTypeSupportDispatch.get_handle_function
1: /tmp/ws/src/rosidl_typesupport/rosidl_typesupport_c/test/test_message_type_support_dispatch.cpp:91: Failure
1: Value of: rcutils_error_is_set()
1: Actual: false
1: Expected: true
1:
1: >>> [rcutils|error_handling.c:108] rcutils_set_error_state()
1: This error state is being overwritten:
1:
1: 'Failed to get symbol 'test_message_type_support' in library: error not set, at /tmp/ws/src/rosidl_typesupport/rosidl_typesupport_c/src/type_support_dispatch.hpp:102'
1:
1: with this new error message:
1:
1: 'Failed to find symbol 'test_message_type_support2' in library, at /tmp/ws/src/rosidl_typesupport/rosidl_typesupport_c/src/type_support_dispatch.hpp:96'
1:
1: rcutils_reset_error() should be called after error handling to avoid this.
1: <<<
1: [ FAILED ] TestMessageTypeSupportDispatch.get_handle_function (641 ms)2: [ RUN ] TestServiceTypeSupportDispatch.get_handle_function
2: /tmp/ws/src/rosidl_typesupport/rosidl_typesupport_cpp/test/test_service_type_support_dispatch.cpp:88: Failure
2: Value of: rcutils_error_is_set()
2: Actual: false
2: Expected: true
2: [ FAILED ] TestServiceTypeSupportDispatch.get_handle_function (642 ms)I'll wait for input from the maintainers before taking any further action.
This PR changes the severity level of two messages logged by the
get_typesupport_handle_function()functions inrosidl_typesupport_candrosidl_typesupport_cpp, from error to debug.As far as I can gather from inspecting and debugging the code, these functions are used to look up a specific type support by name, and the current "error condition" is actually just a benign case of an available type support not matching the sought after
identifier.I'm still unsure of how the condition is actually triggered, but these error messages show up at times in applications running on
rmw_connextddsand can become quite intrusive (see ros2/rmw_connextdds#21 for an example).After debugging
rmw_connextdds, I've confirmed that these errors are printed when the RMW tries to look up one of the type supports it needs (e.g. see here). The error is always printed only forrosidl_typesupport_cpp, and I haven't been able to determine why it doesn't also appear at times for other type supports, or why it doesn't happen all the time.Any input on uncovering these answers is of course welcome, but since the condition seems to be benign (even if only from empirical observation), I would like to downgrade the errors to reduce their intrusiveness.