The g_rc_function[RC_Channel_aux::k_flap_auto] ptr came out as NULL
during one HIL run on a desktop CPU, which led to ArduPlane
crashing. I am not yet sure if this can happen in real flight, but I
think the NULL check is worthwhile to be sure.
read_airspeed() is only available when in a non-HIL build. The
optimiser normally removes this call, but when debugging in a desktop
build, it is nice to avoid using the optimiser
Add code to disable throttle if we are on the ground and in FBW_B or higher. We believe we are on the ground if speed < 5 and alt < 5. Also check that we are not trying to perform a takeoff.
These files are not generated by cmake. They are necessary for
the make based build to work. To build using this method
type make in the directoy such as ArudPlane/ArduCopter. Make
sure that you do not call cmake in the ArduPlane etc directory
as this will overwrite this Makefile with the autogenerated cmake
version. Cmake should be called from withing a build directory.
this moves the mavlink send logic into GCS_Mavlink.pde, and also
ensures we only ever send parameters and waypoints when there is
sufficient space in the serial send buffer
this avoids a varient of the gcc excessive stack usage bug, by
wrapping the send calls in NOINLINE functions. This saves us a lot of
stack space, and strangely enough produces slightly smaller code!
I am concerned that some of the mavlink send routines may impact on
sensor calibration, so its safer to just send the minimum information
to keep the GCS happy
Fix an issue with initialization of channels that on-line changed their function.
For the curious people that the code size overhead of having any aux channel do any function (with this code) is 44 bytes.
To see wich code I'm talking about, do a git diff 05057ac2d455..this_commit (replace this commit with the commit hash of this commit)
Amilcar Lucas
Andrew Tridgell
Doug Weibel
James Goppert
Janne M
Jason Short
Michael Oborne
unknown