To be able to still infer the direction of the thrust vector we
limit it to a minimal length even if MPC_THR_MIN is set to zero.

Note: This is a hotfix for certain specific applications.
The direction of the thrust vector in this corner case is very
likely to get into the tilt limit which is generally undesired.

A low gyro cutoff is needed for most medium/large size drones as the structural natural and blade-pass frequencies are low.
A higher value is still desirable for small platforms surch as racers
or well isolated autopilots and should be tuned by the user.
Specific values for config files are untouched.
The cutoff filter for the D term is disabled here as the required
cutoff frequency for the default D term of the rate controller is higher
than the gyro cutoff. In that case, enabling the D term cutoff would
just add some undesired phase lag to the derivative.

It looks like the atomic builtins are not available using QuRT and the
Hexagon toolchain, so our best bet is to use `volatile` for the atomics.

This fixes a potential dead-lock when 'uorb status' was used via MAVLink
shell.
The dead-lock chain is: DeviceMaster::lock() -> printf -> output to a pipe,
which blocks until a reader reads the data. In that case it's mavlink.
If mavlink makes a call that requires DeviceMaster::lock() (such as
orb_exists), it dead-locks.

This patch moves all printf's out of the locked state.

- when either terrain data was temporarily not valid (flying at high
distance to the ground) or the vehicle was not close to the ground
(_param_ekf2_gnd_max_hgt) the ekf switched to using the land detector
ground effect flag.

Signed-off-by: Roman <bapstroman@gmail.com>

- takeoff help is used for fixed wings, it increases the altitude setpoint
after a launch. A vtol does not need this as it's already sufficiently
high up in the air.

Signed-off-by: Roman <bapstroman@gmail.com>

This reverts commit 4a71984f.

The noise for airspeed is now applied on the Gazebo side and we can
remove this hack.

the condition to enter the rc mode switch evaluation was neglecting
the first connection of an RC when "no RC switch changed". this means
depending on the actual initialization values of _last_sp_man and the
desired mode preselected on the RC while connecting it would not get
evaluated.

pure refactor using De Morgan's law to make the condition more intuitive
since you think about when should I enter and not when should I skip

pure refactor of the huge condition into smaller const bool junks

There were multiple comments not addressed in pr #11904.
See commit 4bcb37f9bc9ef20b521ebcf33e7a0ed08fa86ccd

The Pixhawk cube as an internal mag on SPI. If there was a bad register
read (e.g. on an IMU failure), _register_wait was non-zero and there was
no transfer but _mag still published mpu_report.mag.

Auto mode traj - limit yaw setpoint rate of change when generated in the flight task instead of clamping the yaw rate in the controller

Move yaw setpoint slew rate from AutoLineSmoothVel to Auto. The slew rate is now applied consistently to all the auto FlightTasks

Add missing underscore prefix to a member variable in mavlink_main.cpp.h and delete unnecessary struct specifiers in mavlink_main.h.

Migrate an additional set of 16 variable initilializations from the Mavlink class constructor list to their respective declarations.

By declaring the params above the navigator modes inheriting from
mission block we make sure the params in navigator are set before
read in the constructor of mission block.

This prevents uninitialized padding from being written to dataman.

This variable was not set and lead to uninitialized memory being written
to dataman.

This should not be any functional change. The only difference are the
variable scopes, and early return versus nested ifs.

on airspeed scaling

Signed-off-by: Roman <bapstroman@gmail.com>

the first position when the oa fails as setpoint to avoid jumps

Inhibiting bias estimation using the EKF2_AID_MASK parameter only applies to the delta velocity bias states.