Entry 282

Open questions after the triage: the disc, the formats, the engine core and the tools

discformatdecompenginetoolingbuild

This entry answers the question left open by The retail executable was shipped unstripped, An R5900 disassembler checked against llvm-objdump and the relocations, The DATA.BIN index lives in the executable, Reading the DWARF 1 debug information, Model chunks, the frame section, and a clean walk over every archive, How the game is put together: boot, modes, frames and tasks, Kite's stumbles: a round DualSense stick against a DualShock 2's threshold, eemu in Rust: piney-eemu, identical to eemu.py instruction for instruction and 20 times faster, Story start points, the way out of the dungeon to the desktop, and the board's announcements, "One core for the four volumes: detection, the first generated tables, and a mail from Helba", "The port's functions audited across the four volumes: voldiff.py ported", "The later volumes' New Game to the desktop: the carry's pointers, each volume's mail, the streams' archives, the bound constants", The table generator in Rust, The carry in Rust, The symbol transfer in Rust, Harnesses on Mutation's own code, The pad's vibration, The console at the window's size, A shot run as an animated WebP, Overlays and DATA.BIN, their last unknowns, One game from the four discs with piney-build, and a launcher, The event scripts move into the build, Five table groups move into the build, Every table group moves into the build, The placement groups move into the build, The generator reads the discs, The tree cleaned of the games' dialogue, Mutation's str7100, str8800 and str0300 are Infection's code, Mutation's whole story under the autopilot, 101 to 116.

FilesUNKNOWNS.md

This entry closes the open questions of 29 entries about the disc, the file formats, the executable, the engine's modes and tasks, the pad, the build and the project's own tools. The triage of 2026-09-28 (UNKNOWNS.md) split every entry's open paragraph into single questions and checked each one against the later log, the docs and the code. The answered ones are listed below with their evidence. What is still open is restated at the end, with the entry that asked it, the play questions first. Open questions of these entries that belong to another subsystem are restated in that subsystem's entry: 286, 288, 289 and 291.

Answered

  • 2: relocation type 123 — answered by 243 (the micro address >> 3 in the 15-bit immediate) / docs/engine/executable.md
  • 2: what the mc_* symbols resolve to — answered by 243 (VU entry points, 21 referenced)
  • 2: how much of the DWARF survives — answered by 6 (all of it; tools/dwarf1.py)
  • 2: what the MWo3 header's overlay id is used for — answered by 242 (nothing reads it) / docs/formats/prg.md
  • 4: the VU1 microcode is not disassembled — answered by 17 (tools/vu.py, tools/test_vu.py)
  • 5: what fills directCCSTbl and which files use category 19 — answered by 242 (ccAddFileList; no volume adds one) / docs/formats/data-bin.md "Category 19"
  • 6: the contents of the .mwcats sections — answered by 243 (per-function size records) / docs/engine/executable.md
  • 10: the rigid ITOF12, the ST scale and culling inferred from the EE side — answered by 17 (the microcode settles them) and 244 (ST scale)
  • 10: the 15 extra words per mmat in the size of mtype & 4 models — answered by 244 (an overstated size nothing reads) / docs/formats/ccs-model.md
  • 10: the model flag bits above the blend type, and zoffs — answered by 244 (bit 3 dropped; zoffs never read)
  • 10: Anime sub-kinds other than 0x0102 — answered by 32, 244 (all eleven kinds) / docs/formats/ccs.md "Anime sub-kinds"
  • 10: the texture and CLUT flag bits and the CLUT's unknown_1 — answered by 244 (unknown_1 is cpsm) / docs/formats/ccs.md "Texture and CLUT flags"
  • 15: ccEvent::Execute, the event script interpreter — answered by 18, 40 / docs/engine/event-vm.md
  • 82: WORLD_MAN::SetEventData — answered by 87
  • 82: Mac Anu's gate-address window empty — answered by 90 (FieldUi::announce_lines)
  • 82: streams 4-6's subtitles — answered by 85
  • 82: event 10's sound 8 — answered by 176 (the sound instruction's eleven cases)
  • 181: how the later volumes pick their event voice tables — answered by 226 (each disc's own ccEvVoiceRequest, decoded by groups)
  • 204: OUT's and QUA's ccParticle::Setup switch — answered by 205 (decoded by register)
  • 211: whether the carry's Rust port gives the same cache byte for byte — answered by 212 (piney-gen carry --check)
  • 211: the six trade and shop failures — answered by 214
  • 212: whether the symbol transfer (xfer.py transfer, .syms) ports exactly — answered by 213
  • 213: the five older Python generators remain — answered by 256, 257, 258 (every table group in the build; the Python generators are gone from tools/)
  • 219: Carmina Gade's DrawMap and Mutation's changed town maps — answered by 225
  • 256: how many of the 337 shared statics' users change when their groups move — answered by 257 (every table group moved into the build)
  • 257: the generator still reads its inputs from work/ — answered by 259 (the generator reads the discs)
  • 258: the generator reads the executables, overlays, symbols and DWARF from work/ — answered by 259
Still unknown
  • (play) [[259]]: a later volume's image played without Infection's disc falls back to work/, and crate::die ends the process — a player who owns only a later disc. - [[4]]: the EE encodings the game never uses (the debug and performance-counter moves, rsqrt.s, max.s/min.s, vrnext/vrget, the pmfhl variants) are decoded from the manuals only. - [[5]]: whether any requested name can miss its category table (the walk would pass the nameless terminator); the game relies on never asking. - [[5]]: the two s16 fields (+0x24, +0x26) of a sceneFileList entry. - [[6]]: the four-byte member missing from ccChunkIndex, and the 18 other classes' unexplained gaps. - [[242]]: what category 19 (a loose scene file outside the archive) was for; nothing in the code says. - [[15]]: what modes 1 and 0x1000 do (docs/engine/overview.md Unknown). - [[15]]: the full list of tasks each mode starts and their priorities; the port orders each ported mode's tasks itself (docs/engine/overview.md Unknown). - [[15]]: ccSystem::Ctrl's other branches (screen-mode changes, the flags at +0xbd4/+0xbd5); only the vibration countdown was read ([[232]]). - [[69]]: the load before F0 (ccLoadResourceFL) takes no frames in the port (docs/engine/event-vm.md Unknown). - [[82]]: the load time between the mode change and the desktop's set-up (docs/engine/event-vm.md Unknown). - [[62]]: a DualShock 2's real values at full diagonals: recalled, not measured. - [[62]]: half-way diagonals read slightly stronger, so the analogue walk may feel different; not measured. - [[232]]: the rumble has not been felt on a real pad (gilrs needs the pad's force-feedback node). - [[232]]: the pad's connection states (ccPad::Ctrl's other states, scePadInfoAct) are not modelled; a connected pad is a ready DualShock. - [[63]]: piney-eemu: subclass overrides of load, store and set do not reach the interpreter; exec overrides run slowly. - [[63]]: piney-eemu's RAM cannot grow past 32 MB; its registers are live views. - [[63]]: iopemu's machine stays on eemu.py. - [[82]]: a replayed story start lacks the earlier experience and items; its mails and posts are as ccEventFlagSet leaves them. - [[82]]: story:12 starts on the desktop where the game's first block would run in the top page's set-up. - [[148]]: Kite starts event 29's dungeon session as a ghost (dead 4); possibly the story start's doing. - [[149]]: a mid-dungeon story start leaves the room undrawn in its picture. - [[86]]: the unported-default counts say which defaults ran, not how often. - [[176]]: the host-call check plays blocks in isolation; a method reached only through earlier state may not show. - [[125]]: stream_shot has no party members (the field's StreamMenu draws them). - [[161]]: decoding sd9 a second time after sd4 makes the interpreter's DecodeSetup read a bad pointer (test order; docs/engine/dungeon.md Unknown). - [[181]]: whether xfer's ref and content passes misname anything beyond the tables they were built for. - [[182]]: a reader that swallows its error (.ok()?) would not show in smoke runs. - [[198]]: voldiff's "recompiled" rule is a heuristic: a changed field or a swapped comparison passes. - [[198]]: some OUT and QUA "changed" menus call nothing Infection's do, which points at misplaced carried names; to check in xfer.py. - [[202]], [[204]], [[206]]: finders for the remaining carry misses: Mutation 5, Outbreak 23, Quarantine 28 after [[206]]. - [[204]]: readers that take an Infection constant through a shape the scan does not see; not surveyed. - [[219]]: the other harnesses still name some Infection addresses of their own, moved to va() as each runs on a later volume. - [[271]]: why piney-gen syms did not carry Func_str7100, Func_str8800 and Func_str0300 to Mutation; a guess: jump-table relocations. - [[233]]: the game's log lines (stderr) do not reach the in-game console, and the console has no horizontal scroll. - [[241]]: nothing records from the window, and there is no key to record live play. - [[247]]: the window's switch from the launcher to a part is checked headless only. - [[255]]: how long an image's first start takes on a slow drive. - [[258]]: the default build at ~/.local/share/piney/game, at data version 3, has to be remade. - [[259]]: whether root() (the repository path, compiled in) is reached by a build run on another machine. - [[262]]: the dialogue scan misses lines under four words, and text inside the games' textures. - [[262]]: the history's blobs are not scanned; the history will not be published as it is (plans/release.md). - [[263]]: how the pilot should find a Data Bug's protect break for Data Drain. - [[276]]: how the autopilot's run compares in time with a player's.