It's possible to use the internal clock in PCA96895 if we account for
the drift it contains. This is a bit different from solutions in other
projects like the Adafruit library and the PX4 firmware: instead of
applying a correction to the final frequency we apply the correction to
the clock since this is the source of the error.
With this fix we arrived to much better results across different lots of
sensors.
The Navio board continues to use the external clock and should have no
difference behavior.
This commit changes the way libraries headers are included in source files:
- If the header is in the same directory the source belongs to, so the
notation '#include ""' is used with the path relative to the directory
containing the source.
- If the header is outside the directory containing the source, then we use
the notation '#include <>' with the path relative to libraries folder.
Some of the advantages of such approach:
- Only one search path for libraries headers.
- OSs like Windows may have a better lookup time.
this makes it possible to bootup ardupilot before the desired network
interface is available. This is very useful for when using 3G dongles
in aircraft
The current implementation doesn't throw an error on a malformed path string.
i.e. udp:192.168.1.1.14550 instead of udp:192.168.1.1:14550 may result in a memory leak or whatsoever.
The commit fixes the issue and outputs a nice error message if anything's wrong.
DMA is getting stopped in the separate method now. This is the best we
can get at the current time. It does yield slightly better experience
and works in the majority of cases.
The patch is a no bulletproof solution, though.
There's a possibility of corruption in case of e.g. a SIGKILL. There's
no signal framework at the time and the commit doesn't add one. That's
why all signals are handled in the same erroneous way. This is not a
good nor a final solution to the issue.
For the issue at hand a better fix might be porting the code to kernel
space but it's a rather tediuos task that we cannot undertake in the
couple of weeks.
The issue has already come up. There's no deinitialization mechanisms at the moment. As APM is rather software than firmware on Linux, there're some clean-up work that needs to be done. This commit triggers deinitialization of RCInput on a panic.
