KnownBugs

The s16 VM:

Functions and return

    Functions called from main work if we end with a
    ret (return) instruction:

        function blah
            ret 1;
        end

    But won't return properly if we omit the ret.

    So long as we put a ret statement at the end of
    every function it's fine, if we don't the program
    terminates when it reaches the end of the function
    code - needs attention.

    -- UPDATE - seem to have got this issue fixed,
    still testing but functions now return to the
    caller whether there's a ret instruction or not.


Endianess

    Maybe I'm misunderstanding something - you can use
    the --byte-order option with the linker to tell
    it what byte-order to write the executable file in.

    I'm using the ArrayBuffer to allocate RAM segments
    and passing a boolean (global.little_endian) which
    is set to true if --byte-order 'little-endian' is
    specified at the command line.

    Now the weird thing is, JavaScript writes ArrayBuffer
    in big-endian by default, we specify little-endian
    by passing the boolean true which seems to working
    just fine.

    I can assemble and link an exe and it will be in
    little-endian byte-ordering, and the s16 VM seems
    happy with this.

    Yet for some reason when I use big-endian - the
    native read/write mode...it doesn't work!

    Needs addressing.


The s16a assembler and linker

    There are a few little bugs, to be honest I'm probably
    going to re-write both of these once the VM is stable,
    for now they just exist so that I can build and test
    executables in s16.

    The assembler doesn't like file names and paths because
    I'm running pre-processor directives through a bunch of
    filters.

    In all fairness this should have probably been done
    later during linking, it'd make more sense.

    There are other small bugs but I'm not too fussed, as
    long as I can assemble and run code I'm happy enough
    for now...will update as things progress.