In some point in past it may have been ttyS2, but right now on kernel
4.16+ the UART is on ttyS5. We could do like is done for I2C and search
on sysfs, but it requires additional changes.
Apparently this code came in part from libuavcan that defines this
struct Control. They also had the same issue detailed on
https://github.com/UAVCAN/libuavcan/issues/116.
The solution here is much simpler though: stick to the design of cmsg()
even if it's C. As per cmsg(3), use a union together with CMSG_SPACE().
Use a const variable instead of a define so we don't polute the
environment with a define specific to this function.
Also remove tabs and replace with proper coding style.
This adds a counter here to only loop to a maximum number of iterations.
Now we avoid situations in which we would be stuck processing packets here.
This also adds some other meaningful defines
Removes compile-time selection of RCInput driver for ocpoc_zynq.
PPM and SBUS input are now colocated on the ocpoc board, and it
only needs to run RCInput_ZYNQ. Pulse input is also inverted
to accommodate SBUS input, which has no effect on PPM input.
On kernels 4.7+ duty_cycle should always be less than period.
Otherways, we'll get a EINVAL.
It makes sense to set duty_cycle to 0, as
duty_cycle doesn't really make sense without a proper period.
A proper way to handle these errors might be to call pwm_adjust_config
in every pwmchip backend but it's unrealistic to hope that all vendors
will do it quickly.
If we are the controlling terminal for a tty device we will receive a
SIGHUP when the device disappears. Currently what happens is that we
simply stop the whole process. We don't want to fall off the sky due
to a bad device. This can happen for any reason, but it's more likely
if the UART is behind a USB connection.
These messages are only valid in a small subset of Linux boards (those
based on BBB and variants). The ToneAlarm class should actually be
split allowing different implementations, but for now let's just disable
the message.