Open jabbenseth opened 6 years ago
ping
Any chance to fix the two failing builds from my side? IMHO they are in no way related to my changes but were existing before (#103) but I may be wrong here
@ipa-jba Sorry for the long delay on this. I'll try to get around reviewing pending PRs in the next few days.
the failing tests seem to be unrelated to this PR, it can be reproduced on master and is tracked at https://github.com/ros/dynamic_reconfigure/issues/103
great to hear. However the impact of this pr might be quite substantial and require work on "unknown buggy" packages/config files
great to hear. However the impact of this pr might be quite substantial and require work on "unknown buggy" packages/config files
Correct, we will need to build every known ROS package using dynamic_reconfigure
to see how many do not comply. If this proves to require many changes in downstream packages we will need to hold off that change for the next ROS release (ROS Noetic) and make sure to file PR on every package that requires a change
@ros-pull-request-builder retest this please
@ipa-jba I finally found the hours needed to test this on Kinetic (tested all packages depending directly on dynamic reconfigure).
There are several packages failing, some are warranted failures but for some the check seems a bit too restrictive. For example the comparison is made regardless of the parameter type, for string for example it may not make sense to do a comparison (arguably it may not make much sense to give a range to a string parameter in the first place..).
Some packages seem to set out of range values purposefully, I guess it's to reflect the "uninitialized" state. Though to confirm I guess we should reach out to the maintainers of these packages to see if it's a valid assumption.
List of packages failing with this change (10 out of 159):
RE:
For example the comparison is made regardless of the parameter type, for string for example it may not make sense to do a comparison (arguably it may not make much sense to give a range to a string parameter in the first place..).
This was actually a broken chekc in dynamic reconfigure and should have failed regardless of this PR (addressed in https://github.com/ros/dynamic_reconfigure/pull/122)
@mjcarroll do you think this can get merged for noetic ? this would be a great improvement for debugging / improving robustness of systems
a) the range for a parameter is impossible to meet b) the default is out of range
Having those errors in a cfg file can be tedious work to find/debug