Make the types explicit in quaternion_helper.hpp. #625
Conversation
Windows started complaining that there was possible loss of precision. That came about because we were using the 'double' versions of cmath functions (sqrt, pow, etc). However, it turns out that Ogre only deals with floats by default anyway (and we don't change the default when we vendor it), so switch all calculations to floats. Signed-off-by: Chris Lalancette <clalancette@openrobotics.org>
|
CI:
|
|
Just a question for my own info... What does "we don't change the default when we vendor it" mean? What does it mean "to vendor" something? |
Vendoring in general means that we are building a library/utility that isn't a core part of ROS 2. We vendor a number of libraries in ROS 2: Ogre, spdlog, libyaml, etc. The reasons for doing that vary, but often revolve around the library shipped by default in Linux being too old (spdlog is one example of that), or the library not being available in a convenient way in Windows or macOS. In this case, we are vendoring Ogre here in the RViz package: https://github.com/ros2/rviz/tree/ros2/rviz_ogre_vendor . By default, Ogre uses |
|
@clalancette Thanks for the explanation. |
|
All right, CI is green all around, and I have approvals. Merging this with the hope that we get closer to green on Windows overnight. |
Windows started complaining that there was possible loss of precision. That came about because we were using the 'double' versions of cmath functions (sqrt, pow, etc). However, it turns out that Ogre only deals with floats by default anyway (and we don't change the default when we vendor it), so switch all calculations to floats. Signed-off-by: Chris Lalancette <clalancette@openrobotics.org>
Windows started complaining that there was possible loss of precision. That came about because we were using the 'double' versions of cmath functions (sqrt, pow, etc). However, it turns out that Ogre only deals with floats by default anyway (and we don't change the default when we vendor it), so switch all calculations to floats. Signed-off-by: Chris Lalancette <clalancette@openrobotics.org> Co-authored-by: Chris Lalancette <clalancette@openrobotics.org>
Windows started complaining that there was possible loss of
precision. That came about because we were using the 'double'
versions of cmath functions (sqrt, pow, etc). However, it
turns out that Ogre only deals with floats by default anyway
(and we don't change the default when we vendor it), so switch
all calculations to floats.
Signed-off-by: Chris Lalancette clalancette@openrobotics.org
This should fix the last warning we are currently getting from Windows, as seen in https://ci.ros2.org/view/nightly/job/nightly_win_rel/1750/ (as one example).