On Linux (and in most of posix-like systems) the baudrate set in struct
termios via cfset[io]speed is actually the index of the standard
baudrates, it's not the baudrate per se (hence the use of B* macros).
On Linux it's possible to set the baudrate by passing the baudrate
itself to TCSETS2 ioctl. Unfortunately it's not possible to include
asm/termbits.h and termios.h in the same compilation unit, so we need to
split it. Here we split into a Linux and a NuttX implementation.
The following is a Barometer sensor driver for the LPS25H Barometer that
is integrated in the 96Boards STM32 Sensor mezzanine board.
the update includes the .cpp and .h files of the driver as well as the
updates required in AP_Baro.cpp.
All of the init strings that were sent in the NMEA driver are for GPS protocols
which have binary drivers, which provide far more features and are more robust.
It also appears that due to driver changes the config strings for SIRF/UBLOX
were no longer correct anyways).
The mavlink message field SYSTEM_TIME.time_unix_usec works fine with GPS_AUTO_SWITCH == 0 (no switch) or ==1 (usebest)
But when GPS_AUTO_SWITCH == 2 (blend) then state[GPS_BLENDED_INSTANCE].last_gps_time_ms gets initialized with 0 and never rewritten.
The consequence: SYSTEM_TIME.time_unix_usec gets stuck at zero.
The solution: Do not reset state[GPS_BLENDED_INSTANCE].last_gps_time_ms because it would overwrite the correct value already set on line 1149
Commit b9877e0d38
(GCS_MAVLink: make per channel perf counter non-static) made the
perf counters to be available per instance but missed the fact that
the perf infra doesn't copy the string.
Fix this by maintaining a the string inside the object.
It's not being sold, there are just a few (different) engineering
samples built and there are no plans for this to go forward for people
that were pushing it.