What is not done, what is not known, and what it would take to close it.
This page used to open by asking someone to play a full game, because the patch was measured instant by instant in the emulator but nobody had sat down and played it. Araubi did, and sent the recording: eight minutes in, the sheet's labels came out as garbage, and shortly after that the machine hung.
It was the patch's fault: the Ring hook pointed one byte before the start of its routine and slipped in an ld (hl),a that ate the separators of the name list, one per sheet drawn. It is written up in Findings, and fixed with a single byte.
One recorded game was worth more than every automated check put together. What still needs playing to be seen:
And one the recording left open: the hang came after the broken panel and was most likely its consequence —with no separator, copying a name never finds where to stop— but that is not proven. With the corrected tape the game is still alive ten minutes in and does not hang; a genuinely long game would close it.
It has been seen on Gandalf (Brujo) and Frodo (Hobbit), and it fits on both. But the sheet's labels vary in length — "Es muy Decidido ," is a long one — and the numbers go in column 20. The other types are unchecked.
The planting loop deliberately skips slots 0x16 and 0x17, and both are enemy units. Of the ten cells holding enemies, nine get planted.
What is not known: why. The game treats them differently in more places — the loop that colours unit cells (0x6AE3) skips them too, and there is a routine, PON_EL_EJERCITO_16_EN_SU_SITIO (0x922C), dedicated to putting 0x16 back where it belongs — so they look like something special rather than an oversight. Making them visible without knowing what they are is asking for trouble.
Today the number appears in the sheet, which is where the game draws the ring. A counter permanently on screen would mean hooking the play loop (0x7F57) and writing into the ZX screen every frame: new code, more risk, and no way to check it without playing for a while.
The hook is located: 0x733E gives the bearer, 0x8333 the value, and 0x7113 knows how to paint the number.
It was asked for here and it is done: with the map repaint the Eye became dark red over the terrain's cream, and the friendly icon a blue shield. Both used to carry the same 0x38, black on white, and at a glance they looked too much alike. Dark red is one of the twelve colours the Spectrum attribute can actually ask for in this port; bright red is not.
In the repaint, tiles 85 to 88 and 93 to 96 come back empty, and on the cassette they had artwork. They are exactly the four quadrants of drawings 0x16 and 0x18 in the two-by-two table at 0x77B5.
What is not known is whether anything asks for them. Hunting for who uses each index of that table turned up owners for 0x00-0x0F (PINTA_LO_DE_ENCIMA, through the terrain's nibble), 0x11 and 0x15 (PINTA_LA_UNIDAD) and 0x13/0x14 (terrain 4). For 0x10, 0x12, 0x16, 0x17 and 0x18 no caller was found, which is not the same as proving them dead. If one of them ever gets drawn in a long game, a hole would show up there. The fix is to repaint them: eight drawings on the canvas.
The nineteen strings have been read out of the emulator's RAM and seen on screen on the map sign, on a named unit's sheet and on a nameless formation's. What is not checked:
Orthanc, Barad-Dur, Minas Tirith...), which was the ask;Ga. Hierro and Puerta N are abbreviations forced by the signpost's width (10x1 and 8x1). Nothing longer fits without redrawing the sign;Mago, Hombre and Hombres. Elfo, Elfos and the two Mago slots read correctly out of RAM but have not each been caught on screen.The four new adjectives have been caught on screen, dumped from the patched cassette: Firme, Virtuoso, Valiente and Fuerte on Gandalf's sheet and on Frodo's, with Aliado a la Comunidad closing both.
The cartridge leaves RAM, VRAM, the VDP and the PSG just as the tape does — compared byte for byte on four machines — and that proves the game starts under the same conditions. It does not prove a game can be finished. The menu and the map have been seen; nothing else.
What is needed is the usual thing: someone playing. If something turned up, it would be a loader that gets RAM right and something else wrong, and the place to look would be what the comparison does not cover: ports other than the VDP and the PSG, and the state of the slots after the jump.
The game is silent on tape and stays silent on cartridge. But a cartridge has room: the ROM can grow to 128 KB without touching the loader, and the per-frame hook the game leaves empty is identified. What is missing is the player and the music, and a decision about where its work area lives: RAM free in every phase comes to about 2.9 KB, plus 13,824 reclaimable bytes at 0x094F-0x3F4E (the two ZX end screens, read exactly once).
That is an extension, not a doubt: there is nothing to find out, it just has to be done.