temperature_compensation: use correct type for params (int -> int32_t)

This commit is contained in:
Beat Küng
2017-10-27 10:43:15 +02:00
committed by Lorenz Meier
parent 96616d831e
commit c3b47b71de
@@ -118,7 +118,7 @@ private:
*/
struct SensorCalData1D {
int ID;
int32_t ID;
float x5;
float x4;
float x3;
@@ -162,7 +162,7 @@ private:
*/
struct SensorCalData3D {
int ID; /**< sensor device ID*/
int32_t ID; /**< sensor device ID*/
float x3[3]; /**< x^3 term of polynomial */
float x2[3]; /**< x^2 term of polynomial */
float x1[3]; /**< x^1 term of polynomial */
@@ -187,11 +187,11 @@ private:
// create a struct containing all thermal calibration parameters
struct Parameters {
int gyro_tc_enable;
int32_t gyro_tc_enable;
SensorCalData3D gyro_cal_data[GYRO_COUNT_MAX];
int accel_tc_enable;
int32_t accel_tc_enable;
SensorCalData3D accel_cal_data[ACCEL_COUNT_MAX];
int baro_tc_enable;
int32_t baro_tc_enable;
SensorCalData1D baro_cal_data[BARO_COUNT_MAX];
};