TestIdeas.txt

TODO:
 - (EI keeps interrupt disabled for next instruction test) DD+FD+EI blocks test done, needs cleanup + release
 - zmakebas to build +3DOS bas file directly?

Test ideas: "???" means I can't any more even guess what exactly the idea was about

 - clipping in all layers/modes (CSpect seems to had various off by one when sprites/tile is 2x resolution, mostly fixed now?)
 - scrolling pace/offset for all layers/modes (CSpect 80x32 tilemode horizontal seems 2x and truncated)

base:
 - DMA bug check

* Palettes:
 - CSpect seems to render 9b colours $B600/$B601 weird, the blue value seems shifted in either case.
 - ??? exercise all ways to upload/download + autoincrement on/off (layer2 + auto-inc + 9b = bug in .27)
 - exercise ink masks / borders ($42)
 - ??? there's no duality! (check tilemap pallette first/second select duality, if they are readable on both bits $43 vs $6B)

* Layer2:
 . "hidden" SRAM access under Bank 5 and 1/2 Bank 7 (hidden by FPGA BRAM redirection)

* LoRes:
 - scroll (vs ULA, half-pixel precision)
 - LoRes enable overrides ulanext on/off
 - ULAplus Radastan original 4bpp mode (or what it is)

* Timex 8x1 colour

* Timex 512x192 hires

* Sprites:
 - NextReg mirrors ($35..$39, $75..$79)
 - core3 limits (128 sprites, 64 patterns)
 - exhausting pixel throughput on single line
 - clipping in various modes even "invalid" like $15 enabling clipping over border, but not enabling sprites over border

* Copper:
 - writing code on fly, changing index while keeping mode, etc..
 - write to $60 at even address is latched also for $63, check if some emulator breaks on this

("." points are optional, more like examples/documenting current ZXN board, than trying
to preserve those features for future ... "-" points are like serious tests where emulator
is expected to follow the behaviour, if it wants to be accurate)