Skip to content

TODO.md

Latest commit

 

History

History
37 lines (32 loc) · 2.47 KB

TODO.md

File metadata and controls

37 lines (32 loc) · 2.47 KB

TODO

This is just a random scratchpad of things that could possibly be improved, in no particular order.

  • Align realtimer getValue() with comparator getValue() so PulseCounter and FrequencyCounter can use either.

  • reading DS18B20 hangs if debug logging is blocked on a full output FIFO..... and shouldn't.

  • Do what yalla does, in order to remain compilable on avr-gcc 6+, which no longer accepts avr-libc's "reinterpret_cast_ inside constexpr (see https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63715).

  • make a timeout on the RFM12 failing to initialize (INT pin not going high)

  • re-investigate if we can have write(myStruct) instead of write(&myStruct) when decltype(struct)::Protocol exists

  • Have ChunkedFifo maintain its own Fifo of fixed size internally, to workaround the -Os "bug" (or maybe data is a temp that's copied and gone??)

  • log::debug should always invoke toString(T), which defaults to dec() for decimals, but also defines Microsecond<> as "...ms", etc.

  • Make RFM12 OOK support optional, so you can create the RFM12 driver without using a comparator with suitable timer prescaler

  • Scan for chunked-fifo like API, drop whole chunk if not matched. Or, if scan can work within a chunk, make up a different name, e.g. match()

  • Create a TypedFifo, templated on sizeof(T), without the generic write and read methods;

    • only one T at a time
    • maybe on(T, lambda)

  • find out why pulseCounter.minimumLength is somehow applied x2

  • Rewrite SerialConfig to be a static template class, and remove (for now) ability to change serial configs at runtime. That'll create much faster software serial, and removes the need to juggle pointers in the fifo.

  • maintain "last sent" pin value in pulse counter, comparing with pin->isHigh() after each change, inserting dummies on mismatch . That way we don't need the double memory anymore after all.

  • consider pin factory methods to imply default state as input / output

  • make namespace HAL::Atmel consistent (everything public goes there, everything private in sub-namespaces)

  • Change fifo.abortedWrites into a flag "overflow", and add two other flags "full" and "hasData", so fastread() and fastwrite() don't have to invoke complex calculations.

  • create template variants of the PinChange handlers that don't allow runtime setting of onRising / onFalling but rather do it in template, to save cycles.