GitLab

Combined with the previous change to rule_stack_t, this makes sure that
the number of bytes allocated for the parser's rule stack is unchanged for
the initial allocation, and for subsequent resizes of the stack.

What does change is that now the rules stack can hold more entries for the
same number of bytes.  The result is that there will be less resizes of the
rule stack.

Takes less memory when parsing.  Restrictions introduced by this change
are:
- maximum parse tree size is 65535 bytes (was 4G)
- maximum number of expressions in a tuple, list, dict, etc is 65535
  (was 4G)

Given the amount of memory on the device is so small, these new limits
should never be reached.

If an exception happens during script compilation there will not be any
traceback information, yet it's still important to display such an error on
the display.

os.uname().release retains the release string without any prefix.

The REPL banner now looks something like this:

    MicroPython v1.9.2-34-gd64154c73 on 2017-09-01; micro:bit 1.0.0-beta.1 with nRF51822
    Type "help()" for more information.
    >>>

The output of os.uname() in unchanged.

So it can be accessed outside the modos.c file.  Also rename VERSION to
MICROBIT_VERSION to match.

So we can have full control over how the version strings appear.

Commit 13bcee29 fixed this once, but then
commit 4c6160dc seems to have accidentally
put it back to PULLUP.

To prevent warnings about casting functions when building with C++.

Now that there are tags in this repo YOTTA_BUILD_VCS_DESCRIPTION includes
the tag in its description and so there is no longer any need to explicitly
put RELEASE in the version string.

Also, the RELEASE string is exposed to the linker in case that is needed.

build/firmware.hex now has UICR info added to it, but otherwise should work
the same as before, in particular with tools/makecombinedhex.py.

The address of the symbol microbit_version_string will give the address of
the start of the version string in the firmware.

The old behavour prioritises code embedded in the known flash
location over a main.py file in the file system.
This change reverses that order.

This updates the micro:bit version string to match more closely with
the upstream MicroPython one, and also be semver compatible.

It's useful for editors to be able to determine some sense of linear
progression of builds for betas, so using 'beta1', 'beta2' etc for the
beta releases is desireable

The latest DAL no longer relies on the newlib code that pulls in big
locale support, so we can have a bigger heap for all builds now.

This patch puts all the neopixel code directly into the neopixel module:
getting, setting and clearing pixels is done directly in the class methods,
and the only external code is the sendNeopixelBuffer() function which does
the low-level output of the data.

This saves code size, makes it a little more efficient, eliminates the use
of malloc to allocate the memory for the pixel data buffer (it now uses the
uPy heap not the mbed heap and so doesn't leak memory), and allows in the
future to support different colour layouts/formats of the pixels by making
the RGB index macros dynamic variables.