This repository does not ship the game: it ships the way to build it back from your own copy and to check that what comes out is exactly the cartridge.
Your own dump of the cartridge, in the root and with this name:
| file | hyperolympic2.rom |
| size | 16,384 bytes |
| sha256 | f254764f4bd1634f50abcdabe5813d133d719197216071772845983ffabad22d |
Plus pasmo, z80dasm and Python 3. For the emulator measurements, openMSX.
make comprueba # is this the right dump?
make # trace, generate the listing, reassemble it and run the testsmake finishes green or it does not finish. The line that matters is this one:
ensamblado : 16384 bytes f254764f...abad22d
original : 16384 bytes f254764f...abad22d
OK: reproducible byte a bytemake verify # the test that decides: reassembling has to give the ROM back
make sanity # what reassembling CANNOT catch
make densidad # how much is commented, routine by routine
make imagenes # draw blocks of the ROM so you can look at them
make test # the listing tests
make web # rebuild this websiteverify is not enoughReassembling and getting the same bytes proves the listing is faithful, not that it is correct. If some graphics were being read as instructions the bytes would come out identical all the same: the only thing lying would be the listing.
That is why make sanity runs three more checks, and those three are the ones that really earn their keep:
No annotation may cite an address from the sibling cartridge, or name its events.
This cartridge and Hyper Olympic 1 are nearly the same program: 83.4 % of that one's instructions turn up here. Making use of that is the sensible thing to do, and the annotations for the shared code are brought over with tools/porta_notas.py, which lines both ROMs up instruction by instruction with the sixteen-bit operands zeroed. But that puts each comment at the right address without changing what it says: a comment that is true over there can be a lie here.
tools/repasa_el_porte.py is the guard. The rule is exact and has no heuristic: every address cited inside a comment —or inside a page of this website— has to be the start of an instruction or fall in a data range of this cartridge, and no annotation may name the other one's events. It forgives three things, each with its reason written into the code: the zero slots of the tables indexed from one, sentences that are deliberately comparing with the sibling, and the two comparison pages, whose subject is precisely the other binary.
It is not a decorative guard. It caught nine cited addresses that do not exist here —six off by one byte and three pointing at a different routine— and a block heading that still titled the flight with the name of the sibling's event, the hammer, which this cartridge does not have.
And there is a second guard, tools/cifras_portadas.py, which watches what the first one cannot see: if a comment is identical to one in the sibling, the instruction it annotates has to be the same too, once the sixteen-bit operands are zeroed. It caught the two comments at 0x4389, which said "32 sprites" when the ld bc,01afeh here loads 26.
src/hyperolympic2.asm | the listing; generated, not hand-edited |
src/hyperolympic2.notes | the annotations, anchored to addresses |
src/hyperolympic2.entries | the entry points, each one justified |
src/hyperolympic2.nocode | the areas the tracer is forbidden to enter |
medidas/ | what was measured in openMSX, with the raw tables |
tools/ | the tracer, the listing generator and the data walkers |
docs/ | this website |
The comments live apart from the listing, anchored to the address they describe. That way they survive a re-trace: if tomorrow the tracer splits the binary differently, the comments still land where they belong.
The cartridge image. Neither this one nor any other: it belongs to Konami and is not distributed. See LEGAL-NOTICE.md.