version 8.12

• improved C128 support
• removed broken floating point support from c128 target (its FP routines are currently incompatible with prog8)
• cx16 api change: renamed cx16.push/pop_vera_context() to save/restore_vera_context()
• cx16: added cx16.poweroff_system() and cx16.set_leds_brightness()
• cx16: fix golden ram area
• various bugfixes

version 8.11

• A,X,Y now are valid identifiers
• floating point literals such as .123 (without leading zero) are now also valid
• added divmod() and divmodw() builtin functions
• added txt.width() and txt.height() for the 'virtual' target
• added a memory map diagram to the 'targetsystem' chapter in the documentation.
• CX16 system init code no longer resets the Vera display chip
• CX16 fixed wrong ubyte to float conversion
• CX16 got a few more Vera registers and some new audio ROM routines
• CX16 graphics.plot() colors fix
• CX16 FB_set_palette() fix
• added cx16 bubbleuniverse example
• removed 2 problematic ZP locations from the free list on the C64 target
• fixed the cx16/bdmusic and sincos examples
• got rid of too greedy asm optimizer that caused issues when writing to I/O registers
• added another optimization for simple for loops
• fix occasional crash in pokew() code gen
• documentation menu fix on readthedocs
• when now allowed to use simple 1,2,3 etc value literals in its cases for a word variable, without needing a cast to word
• improved some error messages
• fixed some problems with array intialization values
• fixed some issues in the IR codegen
• fixed some bugs and compiler crashes
• updated 3rd party library versions
• new IR return value mechanism
• experimental '-newexpr' option to select new expression codegen that doesn't use the software eval stack.

version 8.10

• uninitialized variables are now put into a 'bss' section instead of in the resulting PRG itself, this can shrink the PRG size by hundreds or even thousands of bytes
• added varshigh compiler option to move the bss section out of regular program memory space altogether, to allow PRGs to be larger. On the c64 it is moved to $c000+ and on the cx16 to bank 1 of banked ram at $a000+
• the internal expression evaluation stack as been reduced to 1 page of memory (was 2 pages)
• on the cx16, the eval stack was moved from bottom of golden ram to top of golden ram to make it easier to load other things in golden ram ($0400-$06ff is free now).
• the experimental noreinit compiler option has been removed for now
• removed cx16.callrom() , just use cx16.callfar()
• cx16 programs now turn off the mouse cursor at startup
• added streaming audio player example to cx16/examples/pcmaudio
• internal change: code generators uses unified interface and no longer uses internal compiler Ast
• vm improvements
• bugfixes
• documentation improvements

version 8.9

• removed workaround for black cursor bug in older X16 kernal
• cx16.numbanks() now returns a word because the result can be >255
• fixed cx16/keyboardhandler example
• sys.wait() behaves better
• no longer allow ~ (bitwise invert) on booleans
• some boolean logic optimizations
• fixed zsound example
• fix some casting type errors
• upgrade to kotlin 1.8.0

version 8.8.1

Patches a few nasty issues found since the last release that couldn't wait for the next full release.

• fix compiler crash on hoisting certain variable declarations from inner scopes to subroutine scope.
• don't remove non-trivial initialization assignments for unused variables
• don't remove consecutive assigns to certain memory locations or memorymapped vars (I/O space)
• now able to compile "not xx in array" expression (for 6502 targets)
• "xx not in array" is not synonymous for "not xx in array"
• as a bonus, includes an updated "christmas season" snow example

version 8.8

• fix several FP rom routine addresses on cx16. Some floating point operations didn't work before.
• it's now possible to use symbols that are the same name as 6502 instructions
• removed unused command line option keepIR
• API change: diskio.list_files doesn't have an internal buffer anymore, you now have to supply a buffer + size yourself. Function renamed to list_filenames
• added cx16diskio.f_seek() to seek to a position in an opened file
• diskio file lister routine now also remember the file type (prg, seq, dir)
• fixed endless loop in optimizer on certain expressions
• fixed gfx2 text color and more robust screen mode reset on real hardware (retains display mode)
• added cx16 snow example for the season
• added cx16 diskspeed example
• added more $03xx vector definitions to C64/C128/CX16 syslib
• various other bugfixes and optimizations
• various changes in IR

version 8.7

• rnd and rndw are no longer builtin functions, but are now regular routines in the math module.
• rnd and rndf routines can now be seeded with rndseed and rndseedf routines. You can get a reproducible sequence of random numbers this way.
• new integer rnd routine that is smaller and faster than the previous one.
• added cx16diskio.vload_raw to load headerless file into VRAM.
• added gfx2.pget to get a pixel color value.
• bit shifts of 8 and more no longer silently result in 0 but give a warning.
• removed silent promotion of byte -> word bit shift that would occur if it was a simple assignment.
• fixed indexing pointer variable with word index.
• fixed various compiler problems with negating array values.
• documentation improvements and spelling corrections.
• other bug fixes.
• the IR and VM have been rewritten to use a graph based code chunk representation. A dead code remover has been added that can remove all unused/unreachable code lines. NOTE: this codegen backend still only works in the vm target.

version 8.6.2

• fix compiler issue on windows (line endings...)
• fix for the documentation build error on readthedocs.io (should be up to date again)
• ".p8virt" virtual machine source file format removed - vm now directly reads and executes the ".p8ir" IR file.
• code cleanups in IR and VM.

version 8.6.1

• added string.lowerchar() and string.upperchar()
• added a few more complex vm examples
• fixed a lot of issues in the IR and VM that I forgot to fix in the 8.6 release..... finally the IR and VM are feature complete and can run complex prog8 programs such as textelite.

version 8.5

• removed the 16 bits sin/cos routines from math library (sin16, sin16r etc) - just create your own specialized tables if you need
• added '-keepIR' option to save the IR file if it's generated.
• command line '-vm' option now also reads .p8ir files
• fixed operator precedence: bitwise must come before comparisons
• added cx16.kbdbuf_clear() routine
• several documentation improvements
• added c64 starfield example
• lots of VM fixes and improvements.
• (internal) massive rewrite of VM code generator, now has Intermediate Representation step. Intended to also be used for future code generators.
• several bug fixes