pxt-microbit
pxt-microbit copied to clipboard
microsoft
Document `input.runningTimeMicros()` overflows after 18 minutes
This caught me by surprised and it's a MakeCode specific limitation, as CODAL returns a 32 bit unsigned int and MakeCode drops the 2 most significant bits: https://github.com/microsoft/pxt-microbit/blob/975c890f2b4b1cf6534ee893b91c8e740739824f/libs/core/control.cpp#L239-L245 I guess the signed bit makes sense, but don't know why the second bit is dropped as well.
It was done in commit https://github.com/microsoft/pxt-microbit/commit/15b4b47e6dbeb40026e6211a3f8e5706c456791e as part of PR https://github.com/microsoft/pxt-microbit/pull/2572 (512b4dd2af68171f8843d4c602c061d333b62971).
This should probably be documented, as 18 minutes is a relatively short time that can be easily reached.
Docs: https://github.com/microsoft/pxt-microbit/blob/975c890f2b4b1cf6534ee893b91c8e740739824f/docs/reference/input/running-time-micros.md#L1-L16
@pelikhan out of curiosity, do you know why the 2 most significant bits from CODAL/DAL system_timer_current_time_us are masked out in MakeCode? (it was done in https://github.com/microsoft/pxt-microbit/commit/512b4dd2af68171f8843d4c602c061d333b62971)
https://github.com/microsoft/pxt-microbit/blob/975c890f2b4b1cf6534ee893b91c8e740739824f/libs/core/control.cpp#L239-L245
The signed bit makes sense, but not sure why bit 30 is removed as well, was there a technical reason?
Integers up to 0x3fff_ffff are unboxed. Anything larger is boxed as a double. We could change this function to return a double (I think the codal call returns uint64_t).
Ah, that's interesting to know, thanks!
And yes, both DAL and CODAL return uint64_t (codal-microbit-v2 set CODAL_TIMESTAMP in target.json).
Not sure if changing the size range of input.runningTimeMicros() would be considered a breaking change (probably no, as it's still a number in JS/TS), so up to you, although if it's left as it is, it's worth noting the overflow in the docs.
