Replace fabsF macro with std::abs call to avoid having float / double mismatch warnings #28642
Conversation
…oat / double mismatch warnings AP_Math: Replace fabsF macro with std::abs call in templates to avoid having float / double mismatch warnings
|
Perhaps fabsF should be removed everywhere and replaced by std::abs? |
| @@ -733,7 +733,7 @@ void ModeGuided::pos_control_run() | |||
|
|
|||
| float pos_offset_z_buffer = 0.0; // Vertical buffer size in m | |||
| if (guided_pos_terrain_alt) { | |||
| pos_offset_z_buffer = MIN(copter.wp_nav->get_terrain_margin() * 100.0, 0.5 * fabsF(guided_pos_target_cm.z)); | |||
| pos_offset_z_buffer = MIN(copter.wp_nav->get_terrain_margin() * 100.0, 0.5 * std::abs(guided_pos_target_cm.z)); | |||
There was a problem hiding this comment.
if the variable is a float, will it promote to double by making this change? That would slow things a lot on older boards
|
cppreference shows only integer versions? |
|
we could add a float cast to the float version of fabsF() |
|
The float versions are here: https://en.cppreference.com/w/cpp/numeric/math/fabs |
|
But if you want to avoid using std library then let me try with the float cast |
I thought this was the point - it's a feature not a bug. We want the right version that has been configured, using abs removes the type-safety |
|
Closing this PR. Going with #28897 instead with the cast method instead of the switch to std::abs proposed in this PR. |
The fabsF macro is defined as either the float version or double version in ftype.h depending on HAL_WITH_EKF_DOUBLE. If it is defined as the float version and a double is used as the argument then we get a compiler warning in the Qurt build. If those are replaced with std:abs calls then that clears up the warnings since it is overloaded for both floats and doubles.