BUGS
- GDC diagnostic disk fails on 9 of 13 tests (tests 4 and 6 - 13).
Details
a. (Rainbow driver) : interaction between DEC's external hardware and the NEC 7220 isn't fully understood (see page 173 of AA-AE36A)
It is also unclear what port $50 actually does when it 'synchronizes R-M-W cycles' between NEC and DEC hardware.
For now, we provide sane defaults for both vector and bitmap units without disturbing colors, display mode(s) or the NEC 7220.
b. the HBLANK / VBLANK ratio is plainly wrong (QUICK TEST / subtest #6),
c. IRQs are flagged as 'erratic' (QUICK TEST / subtest #12).
d. (7220) : incorrect fifo stati are handed out (GDC reports FIFO_EMPTY instead of _FULL when QUICK TEST #4 floods the queue)
e. (7220) : RDAT with MOD 2 used extensively here, but unimplemented (modes other than 0 weren't documented by NEC or INTEL)
Programs with initialization / redraw / reentrance problems (invocation order after reset matters, at least in emulation):
- CANON (high resolution + vectors), Solitaire (SOLIT.EXE) and GDEMO (from GRPHCS.ARC, interactive graphics interpreter '85),
plus the Monitor Aligment Test (from the GDC test disk).
Graphical games that work well: MMIND (MasterMind, after BMP logo), PACMAN, SCRAM, (G)OTELO.
UNIMPLEMENTED:
// - Rainbow 100 A palette quirks (2 bit palette... applies to certain modes only)
UNKNOWN IMPLEMENTATION DETAILS:
// 1. READBACK (hard copy programs like JOBSDUMP definitely use it. See also GDC diagnostics). VRAM_R ?
// 2. UNVERIFIED DIVIDER (31.188 Mhz / 32) is at least close to 1 Mhz (as on the VT240, which uses a very similar design)
// 3. UPD7220 / CORE oddities:
// 3.1. occasional redraw problems (only when screen 1 runs at 60 Hz and screen 2 at 29.99 Hz interlaced = HIRES ?).
// Quote from Haze: "if you have 2 screens running at different refresh rates one of them won't update properly
// (the partial update system gets very confused because it expects both the screens to end at the same time
// and if that isn't the case large parts of one screen end up not updating at all)
3.2 pixels are stretched out too wide at 384 x 240 (not fixable in Rainbow driver, -keepaspect seems to have no effect)
