Codetank_GALAGADIAG.rom — TMS9918 sprite-pipeline bisection probe
=================================================================
Purpose: find out why Galaga's player ship (hardware sprite) is missing on
the real P-LAB Apple-1 while the starfield (name table) renders fine — and
while POM1 renders everything correctly under every condition it can model.

Same 28c256 / 4000R format as the other CodeTank carts. The probe is written
into BOTH 16 KB halves, so the board jumper does NOT matter — burn it, set
either jumper, and boot with:  4000R

It replays Galaga's EXACT video path in 4 steps. The white blocks in the
top-left corner tell you which step is on screen. Press ANY KEY to advance.

  STEP 1 (1 block)  Galaga's exact register init (16x16 sprites, SAT $1B00,
                    sprite patterns $3800) + ship pattern upload.
                    EXPECT: 1 block + a row of dots. No ship yet.
  STEP 2 (2 blocks) ONE static VBlank-gated SAT write: ship + terminator.
                    EXPECT: the ship sits motionless, bottom centre.
  STEP 3 (3 blocks) Galaga's per-frame SAT rebuild (VBlank-gated), forever.
                    EXPECT: the ship sweeps smoothly left-right.
  STEP 4 (4 blocks) The same rebuild WITHOUT the VBlank gate.
                    EXPECT: ship still sweeps (slight flicker is OK).
                    Any key -> back to STEP 3 (cycle for filming).

READ THE RESULT — report the FIRST step where the ship is absent:
  * absent from STEP 2 on  -> a single gated SAT write already fails on your
                              silicon (SAT address / gated-write path).
  * OK in 2, absent in 3   -> the per-frame rebuild is the killer.
  * OK in 2+3, absent in 4 -> active-display SAT writes drop harder than the
                              openMSX model predicts — we widen the pads.
  * OK everywhere          -> the video pipeline is innocent; we bisect
                              Galaga's game-logic path next.

A short phone-video of the whole run is perfect. Grazie mille!
