Skip to content

Clean up error handling in many rcl{_action,_lifecycle} codepaths#1202

Merged
fujitatomoya merged 22 commits intorollingfrom
clalancette/rcl-action-test-action-server-cleanup
Dec 26, 2024
Merged

Clean up error handling in many rcl{_action,_lifecycle} codepaths#1202
fujitatomoya merged 22 commits intorollingfrom
clalancette/rcl-action-test-action-server-cleanup

Conversation

@clalancette
Copy link
Copy Markdown
Contributor

@clalancette clalancette commented Dec 24, 2024

The original impetus from this PR came from running the rcl tests, and seeing output like the following on the console:

>>> [rcutils|error_handling.c:108] rcutils_set_error_state()
This error state is being overwritten:

  'failed to reallocate memory for new transitions, at /home/ubuntu/rolling_ws/src/ros2/rcl/rcl_lifecycle/src/transition_map.c:152, at /home/ubuntu/rolling_ws/src/ros2/rcl/rcl_lifecycle/src/default_state_machine.c:725'

with this new error message:

  'failed to reallocate memory for new transitions on state, at /home/ubuntu/rolling_ws/src/ros2/rcl/rcl_lifecycle/src/transition_map.c:169'

rcutils_reset_error() should be called after error handling to avoid this.
<<<

There are 2 main reasons why this can happen:

  1. The test itself is forcing an error condition, but then forgetting to call rcl_reset_error() before trying another condition.
  2. An error happens in a "lower-layer" method (like rcutils), which sets an error message, and then the rcl* layer also tries to set the error message.

This PR corrects almost all instances of this occurence in the rcl libraries, with the exception of test_arguments, which has quite a few of these and will require a separate PR. Not only will this quite down many of the tests, it also fixes some bugs in error handling in this code.

@clalancette
Shorten the delay in test_action_server setup.
2260193 
Instead of waiting 250ms between setting up 10 goals
(for at least 2.5 seconds), just wait 100ms which reduces
this to 1 second.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Small style cleanups in test_action_server.cpp
1bf2373 
Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Reset the error in rcl_node_type_cache_register_type().
09d26e9 
That is, if rcutils_hash_map_set() fails, it sets its
own error, so overriding it with our own will cause a
warning to print.  Make sure to clear it before setting
our own.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Only unregister a clock jump callback if we have installed it.
33200b7 
This avoids a warning on cleanup in rcl_timer_init2.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Record the return value from rcl_node_type_cache_register_type.
aa5abfa 
Otherwise, in a failure situation we set the error but we
actually return RCL_RET_OK to the upper layers, which is
odd.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Get rid of completely unnecessary return value translation.
344b435 
This generated code was translating an RCL error to an
RCL error, which doesn't make much sense.  Just remove
the duplicate code.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Use the rcl_timer_init2 functionality to start the timer disabled.
25ede71 
Rather than starting it enabled, and then immediately
canceling it.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Don't overwrite the error from rcl_action_goal_handle_get_info()
3b4be6d 
It already sets the error, so rcl_action_server_goal_exists()
should not set it again.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Reset errors before setting new ones when checking action validity
7d8c0cd 
That way we avoid an ugly warning in the error paths.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Move the copying of the options earlier in rcl_subscription_init.
c1fe643 
That way when we go to cleanup in the "fail" case, the
options actually exist and are valid.  This avoids an
ugly warning during cleanup.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Make sure to set the error on failure of rcl_action_get_##_service_name
52a0c77 
This makes it match the generated code for the action_client.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Reset the errors during RCUTILS_FAULT_INJECTION testing.
6d81f9c 
That way subsequent failures won't print out ugly error
strings.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Make sure to return errors in _rcl_parse_resource_match .
d46b8f4 
That is, if rcl_lexer_lookahead2_expect() returns an error,
we should pass that along to higher layers rather than
just ignoring it.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Don't overwrite error by rcl_validate_enclave_name.
9ed7a91 
It leads to ugly warnings.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Add acomment that rmw_validate_namespace_with_size sets the error
89d072e 
Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Make sure to reset error in rcl_node_type_cache_init.
96ef397 
Otherwise we get a warning about overwriting the error
from rcutils_hash_map_init.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Conditionally set error message in rcl_publisher_is_valid.
4d3fcaa 
Only when rcl_context_is_valid doesn't set the error.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Don't overwrite error from rcl_node_get_logger_name.
19a61c9 
It already sets the error in the failure case.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Make sure to reset errors when testing network flow endpoints.
97124e3 
That's because some of the RMW implementations may not support
this feature, and thus set errors.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Make sure to reset errors in rcl_expand_topic_name.
2d39a3e 
That way we can set more useful errors for the upper
layers.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Cleanup wait.c error handling.
7fb850b 
In particular, make sure to not overwrite errors as we
get into error-handling paths, which should clean up
warnings we get.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Make sure to reset errors in rcl_lifecycle tests.
8c3817f 
That way we won't get ugly "overwritten" warnings on
subsequent tests.

Signed-off-by: Chris Lalancette <clalancette@gmail.com>
@clalancette
Copy link
Copy Markdown
Contributor Author

clalancette commented Dec 24, 2024

Pulls: #1202
Gist: https://gist.githubusercontent.com/clalancette/e3f7e4a8739d5a2fe4825922491b9561/raw/01d769156a1c9b934554eac77efc9aa4fb529cff/ros2.repos
BUILD args:
TEST args:
ROS Distro: rolling
Job: ci_launcher
ci_launcher ran: https://ci.ros2.org/job/ci_launcher/14995

  • Linux Build Status
  • Linux-aarch64 Build Status
  • Linux-rhel Build Status
  • Windows Build Status

fujitatomoya
fujitatomoya reviewed Dec 25, 2024
View reviewed changes
Copy link
Copy Markdown
Collaborator

@fujitatomoya fujitatomoya left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm, but a couple of comments.

@fujitatomoya fujitatomoya merged commit 39a5ba8 into rolling Dec 26, 2024
@clalancette clalancette deleted the clalancette/rcl-action-test-action-server-cleanup branch December 26, 2024 16:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@fujitatomoya fujitatomoya fujitatomoya approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@clalancette @fujitatomoya