TODO

Add command to switch banks (for FPGA based)

nkinfile_open_mem: the buffer pointer should be const

Project:
  Move each target to its own repository?
  Make libnklabs a pure library..

STM32:
  * Use Nucleos as-is (no jumpers)

  How do we handle STM32F746 flash memory? Sector sizes are not uniform..
  Missing flash size for info

  low power modes?  We need lpuart..

  * get openocd to work for all stm32s

  * make command to just copy the .hex to the fake usb flash device

Atmel:
  * i2c not working well on atsame70 (crash during i2c scan), but ok on others
  use bitbang for atsame70..

  low power modes?  which is the best clock to use?

I2C:
  * allow use of bitbang i2c, or fpga i2c..

missing documentation:
  target descriptions
  * nkpin
  nkcrclib
  nkmcuflash?
  nkdisplay when we do it

bootloader:
  * We should include an example standard bootloader to show off the ymodem/zmodem firmware udpate

nkymodem:
  * document options
  * ymodem send should send back the file we downloaded..
  use nkoutfile for yrecv?
     it would need nk_outfile_write
  add dbase upload/download using ymodem

  nkblock_t abstraction?
    nk_block_read
    nk_block_write

event driven:
  * multi-line gets for dbase

global:
  * probably only app should print error messages

displays:
  link multiple displays together as one
  nkoutfile_t for them? maybe vt100 emul..
  displays: share CLI code: maybe include display size in struct for this

very small systems:
  f030: try to reduce size, don't use float for any driver..
  maybe have a reduced nucleo.mx that doesn't include everything

tab complete sub-commands
  Now it just prints help, but completion would be better
    Redesign command table so that sub-commands are in it? (Makes table large..)
    Complex algorithm to parse sub-commands from help?

mcuflash:

  nkmcuflash rd and wr do 32-bit works, but on stm32 min. size is 8 bytes- add read / modify / write?

nkchecked:
  better error return for nk_checked_read- maybe return ssize_t. [done]
    but this means nkinfile needs to expect ssize_t from block_read.

nkinfile:
  How do we handle read errors for nkinfile operations?  Maybe use ssize_t
  for block_read..



sdio driver + fat filesystem

arduino functionality:
  digitalRead, digitalWrite, pinMode
  attachInterrupt, detachInterrupt
  analogRead, analogReference,
  analogWrite
  pwm out, pulse in

Add more MCU peripheral drivers:

  CAN, ADC, DAC, etc..

Add more targets:

  PICO-PI
  Add RISCV
  Add Xilinx
  Add TI
  PIC32
  Some example in an RTOS

GPIO:

  Pinmode to alternate functions
  A way to read back the current mode
  Include i2c pin expander gpio pins

nkdbase:

  more tests..
  try all input number formats, including negatives..

  make sure skipping works..

  should error output be an argument instead of nkstderr?  Better for threads

  Finish schema compiler

Nkscan:

  iscan ignores case and squashes case on word and identifier.  Maybe word/identifier should be individually controllable?
  maybe case ignore should be changable in the format string?

  need tests for iscan

  * Expression parser for nkscan?  Allow 0x before decimal?

  Some kind of options parser where we don't care about order- we can build this externally
    while tryopt1 || tryopt2 || tryopt3 || tryopt4

  Always allow whitespace before end of string?  Always allow whitespace before start of conversion?
  Do not ignore whitespace before doubles? (but strtod does this to match C-library standard).

  Versions of nkscan and friends which take a C-string, like sscanf?

  Also vs. sscanf:
    sscanf has character classes
    sscanf automatically skips whitespace before conversions (except for a few special cases)
    sscanf '*' modifier scans but doesn't save result

Get rid of global arrays:

  instead have a wrapper function with the array on the stack which calls the underlying function

Naming:

  nkinfile_t vs. nk_infile_t  type name
  nkdbase vs. nk_dbase        file name

Notes:
  output freezes on ST- fixed by reboot linux

  It seems like a nice idea to use non-standard format sequences in nk_printf-
  but we can not if we also want gcc's built-in format string checking to work.  Therefore we
  have other printing functions:

	nk_indent
	nk_byte_hex_dump
	nk_dbase_fprint

        nkprintf / nkscan:

          Print not just a C-string, but also from an nkinfile

          But this then means that print depends on nkinfile...
          Likewise: Save output to an nkoutfile, not just to a C-string

  sched bug:
            in some implementations, we only update current_time at
		timeouts, but time still passes...  This means if we
            schedule something in the futre we are using the wrong time to
            do it.  Ideally, current_time is based on the actual timer
            value. [this is now required, too many problems with nkymodem
            without it].

            Even so, system works not too bad as it is: interrupts usually
            schedule things for now, and that works.  It's only when we
            schedule things for the future that it breaks.

            this comes up when we try to do a tickless implementation on low
	    power CPUs.  Also the zynq one is probably wrong..