Timestamps seed to system clock

[ThadHouse]

Currently the timestamps on systemcore seed to system clock. We probably don't want this, and should instead just use monotonic clock directly without any seeding.

[AustinSchuh]

Consider switching to clock_gettime(CLOCK_MONOTONIC, ...) too. That'll provide nanosecond accurate time.

FYI, a lot of the kernel syscalls really want to use CLOCK_MONOTONIC not CLOCK_MONOTONIC_RAW. I've gone down that path in the past, ended up backing out and just going with CLOCK_MONOTONIC.

(Feel free to steal anything you like from https://github.com/RealtimeRoboticsGroup/aos/blob/main/aos/time/time.cc )

[ThadHouse]

Is C++ steady_clock generally driven by CLOCK_MONOTONIC? That way teams could use either? And you'd just get NS precision if you use the C function?

[AustinSchuh]

https://github.com/llvm-mirror/libcxx/blob/master/src/chrono.cpp#L217

https://github.com/gcc-mirror/gcc/blob/master/libstdc++-v3/src/c++11/chrono.cc#L87

Looks like it.

[KangarooKoala]

How much overlap is there between this and #7906?

[ThadHouse]

How much overlap is there between this and #7906?

None. Different parts of the code