this fixes a race condition between the DMA completion handler
updating registers in px4io and the mixer used for handling the
override state. The register set code could set r_page_servos[]
between the time when pwm_limit_calc() is called and the servos are
actually output.
* Loosen thresholds for gyro consistency check until temperature compensated units are the norm
* Cut down string lengths so they make it through the MAVLink transport as a whole
This change makes the mixer load and reset operation closer to thread-safe. It was guarded one-way before and in only one location. This change ensures that its being locked out from both directions. The accesses to the locking variables still need work because they are non-atomic.
This change makes the operation more robust as it flags the whole group invalid in the first step. This should not be confused with being thread-safe - to be thread-safe, all accesses to _first and the following linked list need to be guarded by a mutex. This should be done outside of the mixer in the driver though, as the method depends on the board architecture.
NuttX had the CRTSCTS define incorrectly set for only output flow control, which broke our flow control logic. This commit patches NuttX and puts in addition a guard in place to prevent any future issue with the non-POSIX define being incorrect.
This has been debugged and identified by @ecmnet, which was the main contribution for this patch.
This patch fixes two issues:
* It sends the message on the first call, making sure that the first update gets sent out.
* It improves the rate scheduling. In an experiment with 0.5, 50 and 250 Hz all rates were correct within 0.3% of the intended rate.
The rotor count was incorrect which meant that control surfaces like elevons were scaled incorrectly. This was the main reason for really bad SITL performance
in particular:
- SW release version (in addition to the git hash)
- OS version (tag + git hash if exists)
- mcu version & revision & UUID
- toolchain version
The uuid can be disabled via parameter, it's enabled by default.
_global_pos_sub->update(&_global_pos_time, &global_pos); could return
false and in that case global_pos was not set but still accessed.
This is prevented by checking if timestamp == 0.
boards define BOARD_NAME, so board_name() is not necessary. HW_ARCH was
just a wrapper around board_name().
This patch simplifies to having only one common method px4_board_name().
