tmerritt
71ada15de0
Pass 3 named the 10 data regions in prg.s, then the project was promoted
out of _wip/ so `make` now proves header + PRG + CHR against the original
ROM instead of only the PRG bank.
Data regions named:
LevelPtrTable $C000-$C03B (60 B) level header data
ObjectPtrTable $C03C-$C043 (8 B) 4 pointer pairs
LevelDataPtrTable $C044-$C79D (1882 B) level data pointers
SpriteInitPtrTable labelled at $C4B0 see residue note
UnreachedCode_D650 $D650-$D696 (71 B) unreached code
UnreachedCode_E144 $E144-$E165 (34 B) unreached code
MusicDataTable $F55B-$FA47 (1261 B) music/sound data
APUDataTable $FAE6-$FB74 (143 B) APU pulse/noise tables
SfxDataTable $FE59-$FFEF (407 B) sound-effect data
VectorTable $FFF6-$FFFF (10 B) padding + NMI/RESET/IRQ
Promotion (per the AGENTS.md checklist):
- git mv "_wip/Donkey Kong (JU)" "Donkey Kong (JU)"
- add copilot/rom.nes -> ../Donkey Kong (JU).nes symlink
- add header.s / chr.s (one-line .incbin of header.bin / chr.bin)
- linker.cfg gains HEADER ($0000/16 B) and CHR ($0000/8 KB)
- Makefile switched from `cmp reconstructed_prg.bin prg.bin` to
`cmp reconstructed.nes rom.nes`
- no _tools path change (this Makefile never referenced _tools/)
and no `make regen` pipeline exists for it
Doc corrections:
- project OVERVIEW.md said Mirroring Vertical; header byte 6 = $00
means Horizontal (the root index was right) — fixed
- root OVERVIEW.md: row 18 now BYTE-IDENTICAL, tier paragraph and
summary table refreshed (they had drifted around the Lunar Pool and
Urban Champion promotions), new per-project detail entry
- counts corrected to 19 top-level / 17 _wip in OVERVIEW.md,
TODO.md and AGENTS.md
Residue recorded rather than guessed:
- Pass 2 is partial: 919 Label_ definitions remain (214 semantic
labels applied); noted in TODO.md so nobody reads the promotion as
full semantic naming
- SpriteInitPtrTable's label sits at $C4B0: the routine at $C807 reads
$C4A7,x/$C4A8,x, but that table starts mid-row and a label there
would split the canonical 16-byte audit-comment line
Verification:
make clean && make -> "Byte-identical ROM verified."
verify_pass1.py -> 19/19 top-level build, Donkey Kong (JU) OK
(6014 instr / 12505 B), all comments canonical
Doc-drift sweep: build tiers, mirroring, and stale status tails
Reconciled the docs against verified reality after the Lunar Pool / Urban
Champion / Donkey Kong (JU) promotions and the 2026-10-19 audit-comment
standardization. No source behaviour changed except the two build fixes below.
Build tier (whole-ROM vs PRG-only)
- AGENTS.md: the 17 `_wip/` projects are real disassemblies at the PRG-build
tier, not "skeletons"; header.s/chr.s/rom.nes are promotion additions;
verify_pass1.py covers the top-level projects.
- Root OVERVIEW.md: `_wip` prose, audit-comment narrative and summary table
refreshed (37.5% figure, SEEDED vs OK wording, re-trace false alarm).
- Promoted projects whose build blocks still claimed `make` emits
`reconstructed_prg.bin` compared to `prg.bin` (Balloon Fight, Baseball,
CIRCUS, Clu Clu Land, Excitebike, NES PowerPad, Pac-Man (USA), Tennis):
now describe the whole-ROM `reconstructed.nes` vs `rom.nes`.
- Lunar Pool / Urban Champion / CIRCUS also had stale `cd "_wip/..."`
build paths and "Promotion" items; retargeted to their top-level folders.
Mirroring (header byte 6 bit 0: 0 = Horizontal, 1 = Vertical)
- Corrected DK Jr. prose "horizontal mirroring (flag 6 = $01)" -> vertical,
and both `_wip/Pac-Man` variant notes "Both use vertical mirroring" ->
horizontal (both are byte-6 = $00).
- Re-audited every project OVERVIEW: 0 rows and 0 prose claims now disagree.
Stale status tails
- Retitled/restatused the `_wip` OVERVIEWs to their real `prg.s` (Golf,
Ice Climber (U), Lode Runner, Millipede, Pac-Man Tengen, Pac-Man (U),
Popeye 0/1, Raid, Raid a2, Hogan's Alley, Mario Bros); Golf's
skeleton-era contents rows and Next Steps corrected.
- Dropped the skeleton-era "## Next Steps (per `PROMPTS.md`)" heading and
the "(from the skeleton)" wording across all project OVERVIEWs.
- Rewrote Lunar Pool / Urban Champion Next Steps (both already Pass 1-3
complete and promoted; their tails still described pending Pass 2 and
promotion); fixed root row 13 (Wild Gunman) and row 20 (Duck Hunt).
Also lands two code fixes carried with this sweep:
- `_wip/Hogan's Alley (JU)`: $E9F1-$E9FD and $EA13-$EA14 reclassified as
data with canonical audit comments (the seeded build had decoded a data
table base as code, breaking the build since 2ae09f7). `make` ->
BYTE-IDENTICAL again; reconstructed_prg.bin/.lbl refreshed.
- `_wip/Pyramid (U)`: `Label_D72E` -> `NMI:` (byte-identical).
Verification:
make (Hogan's Alley, Pyramid) -> BYTE-IDENTICAL, BYTE-IDENTICAL
_tools/verify_pass1.py -> 19/19 build, all canonical,
"No truncated or unfaithful Pass 1"
mirroring audit (rows + prose) -> 0 mismatches
Donkey Kong (JU): complete Pass 2 — all 919 Label_ definitions named
The last non-zero Label_ residue in the repo is now gone. Every one of the
919 address-derived definitions in copilot/prg.s is replaced:
* 319 routine entry points (jsr/jmp targets):
- 102 named from observed behaviour (LoadSpriteInitPointerToZp00 $C807,
CopyVramDataToBuffer $CBDF, CheckSpriteListCollision $D8AD,
UpdateRandomSeed $F4ED, ReadControllers $F50E, UpdateSoundEngine $FA48, ...)
- 217 left as SubC<addr> — honest residue where the body is too
context-dependent to name without guessing (Golf: 108, Mario Bros: 223)
- the 77 already-named entries untouched
* 600 internal branch targets named <Routine>_Block / <Routine>_Block_n in
address order within their owning routine (DK Jr./Balloon Fight convention).
Ownership is control-flow containment, so a loop that physically precedes
its own entry lands correctly (FlushPpuBuffer_Block $F1EC under $F228).
* Two proposal collisions disambiguated by indexing zero-page slot:
UpdateObjectListZp5D/DispatchObjectStateZp5E ($DA16/$DA4C) vs
UpdateObjectListZpAE/DispatchObjectStateZpAF ($E1E9/$E250).
jsr/jmp operands stay numeric (Wild Gunman style); only branch operands are
symbolic. Applied with _tools/pass2.py apply (919 defs + 2,622 refs rewritten,
comments re-aligned); the ROM still rebuilds byte-identically and
verify_pass1.py reports 19/19 canonical with no truncated split.
Docs synced: project OVERVIEW.md gains a Pass 2 Details section, root
OVERVIEW.md rows + conventions, TODO.md section 4 closed. dk_full_map.tsv now
records the complete applied map (its zp rows remain unapplied proposals).
_wip status correction: five false "Pass 2 complete" claims + map audit
The `_wip` tier's "Pass 2 complete" headers counted their draft maps' rows, not
the names actually present in `prg.s`, and described the count of *neutral*
`ZpXX` slots in `locals.inc` as "zero-page semantic aliases". Verified against
the sources and corrected in the root inventory rows, the project OVERVIEWs and
TODO:
project named/total Label_ left drafted map
Pinball (JU) 180 / 681 501 210 rows, 179 applied
Ice Climber (JE) 76 / 973 897 255 rows, 76 applied
Donkey Kong Jr. Math 64 / 1035 971 243 rows, 62 applied
Duck Hunt (JUE) 31 / 472 441 272 rows, 31 applied
Ice Climber (U) 2 / 973 971 none - Pass 2 not started
No `_wip` project is Pass 2 complete. Ice Climber (U) was the worst case: the
2026-10-19 sweep itself restated it "Pass 2 complete" while describing its 971
`Label_` definitions in the same paragraph - a copy-paste from (JE); (U) has no
map file at all.
Drafted maps audited row by row against the committed split: they are not
applicable. 100/243 (DK Jr. Math), 96/255 (Ice Climber (JE)) and 17/210
(Pinball) rows point at addresses that are not instruction boundaries - they
were built against an older code/data split - while most of the rest are
unlabelled boundaries that no operand references, or (Duck Hunt) competing
candidate names for one label (`Label_C2BF` alone carries 205). Rows that are
both insertable and corroborated by control flow: 0 in every map. Recorded as
stale and kept for provenance; Pass 2 here needs a fresh map derived from the
committed split.
Also corrected while verifying:
- Mario Bros entry counts (docs said 252 / 90 + 165 / 477 sites / 156 neutral
slots; the file holds 262 distinct `jsr`/`jmp` targets = 39 semantic + 223
`SubC<addr>`, 478 sites, 151 slots).
- root `_wip` prose line/label ranges (4,482-17,180 lines, 192-971 `Label_`,
11,126 total) and the Ice Climber code figure (74.6% / 6,112 instr, not
74.7% / 6,113).
- TODO §4 re-scoped to the promoted tier, with the `_wip` residue table added.
Verification: `make` across all 17 `_wip` projects -> BYTE-IDENTICAL (17/17);
`_tools/verify_pass1.py` -> 19/19 build, all canonical, "No truncated or
unfaithful Pass 1 found". No source bytes changed.
Balloon Fight (JU): apply locals.inc/symbols.inc names to prg.s (Pass 3)
Substitutes symbol names for 922 numeric operands: 125 hardware registers
(symbols.inc) and 797 of 813 zero-page/RAM operands (locals.inc). The 15
multi-alias addresses were resolved per routine; 16 operands stay numeric
because $13 is used there as a plain high-byte scratch beside Temp12 and
none of its three aliases fits.
- _tools/apply_symbols.py: the substitution tool (dry run by default;
--sites supplies per-routine choices for multi-alias addresses, --report
lists the residue). Only the operand token is rewritten, so the
`; CPU:$.... PRG:.... XX XX` audit comments survive byte for byte.
- prg.s: operand tokens only. make prints "Byte-identical ROM verified."
- OVERVIEW.md (project): instruction quotes now match prg.s, new
alias-reuse table, and section 10 rewritten from the stale "497 Label_"
prose to the real naming state and the 16-site residue.
- OVERVIEW.md (root), TODO.md, AGENTS.md: status, counts, tool list.
verify_pass1.py: 19/19 OK, all audit comments canonical.
Bomberman (U): apply locals.inc/symbols.inc names to prg.s (Pass 3)
Substitutes 362 zero-page operands in copilot/prg.s with the names already
declared in copilot/locals.inc: 354 from single-name addresses and 8 from the
two multi-alias addresses, resolved per site:
$E0/$E1 IndirectPtrLo/Hi in the RAM-clear loop
(InitializeSystemAndRunMainLoop -> Main_WriteSoundDispatchIndex_C317,
which writes $00 across $0100-$07FF via (ptr),y),
SoundDelay0 (the base of the per-channel delay table, indexed by X)
in the sound engine;
No symbols.inc substitution was needed: copilot/prg.s contains no numeric
hardware operands ($2000-$4017) at all. All 362 operands were already written
with two hex digits, so no ca65 `a:` prefix was required.
make -> Byte-identical ROM verified. Operand token only; mnemonics, spacing and
the trailing `; CPU:$.... PRG:.... XX XX` audit comments are untouched.
Remaining numeric operands (758) address bytes that no .inc file names:
505 zero-page (133 addresses), 171 RAM $0100-$07FF (46 addresses) and 82
table reads in $8000+ (54 addresses).
Bomberman (U): substitute 24 in-file table labels at ROM operand sites
Uses the new `--labels` option of `apply_symbols.py` with the
ld65 `-Ln` output (`reconstructed.lbl`). Only labels that prg.s
actually defines and that live in the PRG (+) are considered,
so CHR/header segment labels cannot leak in.
24 numeric operands (`lda ,x`, `adc ,y`, etc.) now read
`lda SystemPaletteData,x`, `adc GameplayObjectClassTable,y`, etc.
All 24 point at exact label addresses; the remaining 58 ROM-space
numeric operands point at interior bytes or unlabelled tables and
stay numeric.
make -> Byte-identical ROM verified.
Bomberman (U): update docs for Pass 3 symbol application
- Project OVERVIEW.md: corrected source file from main.s to prg.s/header.s/chr.s;
output name from reconstructed_bomberman.nes to reconstructed.nes;
rewritten §10 Zero-Page/RAM Symbols table to match locals.inc (with aliases
and unnamed residue documented);
fixed §4 step 1/4 to use IndirectPtrLo/Hi and SoundDelay0 names;
fixed §11 InesHeader location (now header.s);
added §2 Pass 3 summary table;
added §13 notes on legacy main.s and Main_WriteSoundDispatchIndex_C317 misnomer.
- Root OVERVIEW.md §3: updated source line count (7,553), file list, and added
Pass 3 summary.
Both builds verified byte-identical.
Pass 3 symbol application to all 19 top-level projects (2026-10-19)
Ran _tools/apply_symbols.py --write on all 19 promoted projects to substitute
symbolic names from symbols.inc (hardware registers) and locals.inc (zero-page
RAM aliases, forward references) into prg.s operands. The tool preserves the
canonical audit comments (; CPU:$XXXX PRG:XXXX AA BB) and preserves byte
identity — all projects still rebuild byte-identical to their original ROMs.
Per-project operand substitutions:
Baseball (UE) 121 (85 single + 36 hardware)
Tennis (JU) 18 (18 single)
Pac-Man (USA) (Namco) 524 (524 single; 727 multi-alias residues)
CIRCUS 118 (38 single + 80 hardware; 23 multi-alias residues)
Wild Gunman (U) 76 (11 single + 65 hardware)
Joypad Test Cartridge (U) 47 (47 hardware)
NES PowerPad Test Card (U) 89 (89 hardware)
MUSCLE (U) 220 (91 single + 129 hardware)
Lunar Pool (U) 45 (18 single + 27 hardware)
Donkey Kong (JU) 75 (17 single + 58 hardware)
Urban Champion (JU) 70 (4 single + 66 hardware)
Already had symbols applied: Balloon Fight, Bomberman, Burger Time,
Clu Clu Land, Defender 2, Donkey Kong 3, Donkey Kong Jr.
Skipped: Excitebike (JU) — inline prg.s definitions conflict with locals.inc
values and source operands don't match ROM byte encodings; automatic
substitution breaks byte identity.
Documentation updated:
OVERVIEW.md — added Pass 3 symbol notes to all 19 project entries,
added Urban Champion as section #36
TODO.md — current state now reflects complete symbol application across
the promoted tier
Verification: python3 _tools/verify_pass1.py reports all 19 projects OK/SEEDED,
all canonical comments, no truncated or unfaithful Pass 1.
71 lines
10 KiB
Markdown
71 lines
10 KiB
Markdown
# AGENTS.md
|
||
|
||
## What this repo is
|
||
|
||
Thirty-six AI-generated disassemblies of NES ROMs: 19 promoted to top-level folders and 17 still under `_wip/`. All 36 carry real disassemblies (Pass 1 or beyond); the difference is the build tier — the 19 at top level verify **byte-identical for the whole ROM**, the 17 in `_wip/` verify at the PRG level only. The contract (stated in `PROMPTS.md` and verified by `make` in every `copilot/` folder) is **byte-identical rebuilds**: `reconstructed.nes` must match the original ROM exactly. Not for distribution (`README.md`).
|
||
|
||
## Layout
|
||
|
||
- Each project owns a folder with a `copilot/` build dir (`prg.s`, `linker.cfg`, `symbols.inc`, `locals.inc`, `Makefile`; promoted projects add `header.s`, `chr.s` and a copy/symlink `rom.nes`) and a per-project `OVERVIEW.md` at the folder root (power-on flow analysis). Some older ones use `main.s` instead of `prg.s`, and a few lack `prg.bin` (carve the PRG out of `rom.nes` instead).
|
||
- Root `OVERVIEW.md` is the index: vectors, build fixes, and summary for all 36 projects. **Keep it in sync** when a project changes — its DK Jr. entries drifted badly before being corrected, and its vector column later turned out to have five rows wrong from copy-paste duplication.
|
||
- `TODO.md` is the handoff: open work per project, plus the gotchas that cost real time. Read it before starting anything non-trivial.
|
||
- `PROMPTS.md` is the workflow contract. `default_project.tar.gz` is the empty-project skeleton (prg.s, symbols.inc, linker.cfg, Makefile, locals.inc). `_ROMS/` is the master ROM store.
|
||
- `_tools/`: `nesdis.py` (Pass 1 tracer), `pass2.py` and `bytecode_dis.py` (Pass 2/3 emitters), `apply_symbols.py` (substitutes `symbols.inc`/`locals.inc` names into `prg.s` operands), `trace_stage_payload.py` (py65-based runtime tracer), `cdl_reclass.py` (Mesen-CDL-driven code/data split fix), `verify_pass1.py` (rebuild + re-trace check across the top-level projects), `audit_comment_tool.py` (adds canonical audit comments using tracer as ground truth), `splitnes.py`, `chr2gif.py`.
|
||
|
||
## Commands
|
||
|
||
```bash
|
||
cd "<Project>/copilot" && make # builds reconstructed.nes, then cmp vs rom.nes
|
||
python3 _tools/verify_pass1.py # rebuilds + re-traces the 19 top-level projects
|
||
```
|
||
|
||
`make`'s default target includes the byte-identity `cmp` — a build that doesn't print "Byte-identical ROM verified." is a failure. Requires `ca65`/`ld65` on PATH.
|
||
|
||
**There are two build tiers, and the difference is real.** Top-level projects compare the whole ROM against `copilot/rom.nes`. The 17 `_wip/` projects only run `cmp reconstructed_prg.bin prg.bin`, so their **header and CHR are not proven** — do not describe them as byte-identical to the ROM.
|
||
|
||
## Promoting a project out of `_wip/`
|
||
|
||
1. `git mv "_wip/<name>" "<name>"` — the relative `copilot/rom.nes -> ../<name>.nes` symlink survives untouched.
|
||
2. Add `copilot/rom.nes` (a space-free copy or symlink of the original) and switch the Makefile from `cmp reconstructed_prg.bin prg.bin` to `cmp reconstructed.nes rom.nes`.
|
||
2b. A `_wip/` project has no `header.s`/`chr.s` and its `linker.cfg` declares only the PRG memory area, so add all three (copy the shape from `Burger Time (U)/copilot/`: two one-line `.incbin`s and `HEADER: start = $0000, size = $0010` ahead of PRG) or the whole-ROM compare has nothing to link. `Lunar Pool (U)` is the worked example (2026-10-18), and `Donkey Kong (JU)` (2026-10-19) the second — it also fixed a header-byte mirroring error in its project `OVERVIEW.md` (Vertical → Horizontal from byte 6 = `$00`).
|
||
3. Change `../../../_tools/` to `../../_tools/` in the Makefile — the depth changed by one. (No-op if the Makefile never referenced `_tools/` — `Donkey Kong (JU)` had none, and no `make regen` target either.)
|
||
4. Verify `make`, then `make regen` if a regen pipeline exists.
|
||
5. Add it to the root inventory table and re-number the `_wip` rows.
|
||
|
||
|
||
## Irreversible conventions (do not "clean up" past these)
|
||
|
||
- Every instruction carries a trailing audit comment with its CPU address, PRG offset and exact emitted bytes. **The canonical format is `; CPU:$C123 PRG:0123 AD 21 03` — all 19 top-level projects now use this exact format** (standardized 2026-10-19; Donkey Kong (JU) confirmed canonical at its promotion the same day). This is the byte-identity audit trail — preserve it even while renaming labels or restructuring files. The earlier gap (6 variants, 1,013 uncommented lines in Excitebike) was closed 2026-10-19 via `_tools/audit_comment_tool.py`.
|
||
- Unreached/unresolved bytes stay as data (`Label_/Data/` `.byte` or `.incbin` blocks with ROM provenance), never deleted or reformatted.
|
||
- Routine headers follow the template in `prg.s`: global + local description, parameters, returns, clobbers.
|
||
- Two separate symbol files: `symbols.inc` = NES hardware constants only; `locals.inc` = zero-page and forward references. Unresolved zero-page bytes keep neutral `ZpXX` names rather than speculative names. Proven semantic names may sit *beside* the generated `ZpXX` entries as aliases.
|
||
- Pass discipline: Pass 1 = code/data split (stop for review), Pass 2 = label replacement, Pass 3 = semantic naming of data tables/structures. Do not skip straight to guessy rewrites. When a pass is only partly achievable, record the residue and why rather than inventing names.
|
||
- `make regen` (where it exists, e.g. MUSCLE) must keep reproducing `prg.s` byte-for-byte on the standard library alone. Re-check it after any `nesdis.py` change.
|
||
|
||
## Hard-won quirks (details in root `OVERVIEW.md` per-project "Build fix")
|
||
|
||
- The flat disassembler mislabels `bne` as `bpl`; fixes use `bne *-$XX` to reproduce the original bytes (Clu Clu Land, Defender 2, Excitebike).
|
||
- Zero pages are reused across modes; aliases may be required in `locals.inc` (e.g. `RandomStateAlt = $E4` beside `RandomState0 = $54`).
|
||
- Donkey Kong Jr. is architecturally different: reset entry is $C66E, the reset path idles forever in an LFSR loop, and the entire frame state machine runs inside NMI. Its `DeferredTaskHandlerTable` entry 0 deliberately mirrors the IRQ vector word at $BFFE.
|
||
- IRQ vectors often point at unused/data targets in these games — verify before assuming a handler is broken.
|
||
- Unlicensed titles break static-trace assumptions: Pyramid (U) pokes a mapper unlock at `$4100` (beyond the register space) despite a mapper-0 header, and dispatches through `jmp ($0002)` built from RAM — trace via its two queue tables (`$C3A6`, `$CCD5`), and do NOT fall-through-seed gaps that follow `rts`/`jmp`/`brk` (that misclassifies data as code).
|
||
- ROM filenames contain spaces **and parentheses** (e.g. `M.U.S.C.L.E. (U).nes`); builds `incbin` a space-free `rom.nes` inside `copilot/`. Keep that copy/symlink working and don't rename it. Note that a name with spaces/parens is also what makes FCEUX hang headless — symlink it to a clean name and run under Xvfb.
|
||
- `.gitignore` contains `ROMS/*` but the actual directory is `_ROMS` — the pattern matches nothing, so ROM binaries stay tracked. Check before committing large blobs.
|
||
|
||
## Tracer and toolchain gotchas (2026-10-17)
|
||
|
||
Full detail and reasoning in `TODO.md`. These are the ones that silently corrupt results rather than failing loudly.
|
||
|
||
- **Undocumented opcodes silently truncate a trace.** A tracer that knows only the 151 official opcodes halts on the first one used as ordinary code. Joypad Test Cartridge died at `$C09E` on `DCP $3C20` (`$CF`) and filed **3,256 bytes of real code as `.byte`**. `nesdis.py --undocumented` decodes all 105 (default **off**, so existing `make regen` pipelines stay byte-stable).
|
||
- **ca65 V2.19 implements none of the undocumented opcodes**, so they cannot be written as mnemonics. `nesdis.py` emits them as explicit `.byte` with the disassembly in the comment — that keeps them in the code region, so both the split and the audit trail survive. ca65 is also happy to accept plain `nop`, which is official (`0xEA`).
|
||
- **Vectors are not in the iNES header.** They are the three little-endian words at the *end* of the 16 KB PRG (`$FFFA–$FFFF` in the `$C000` mirror). Header bytes 0–9 carry only sizes, mirroring and mapper. Parsing them from `header.bin` as if they were header fields yields garbage.
|
||
- **RESET and IRQ are usually the same address.** The semantic-name map must not overwrite on collision, or the primary entry gets renamed to `IrqHandler`. Use `setdefault` in RESET/NMI/IRQ priority order.
|
||
- **Control-flow targets must win over fall-through.** A fall-through decode whose *operand* spans an address that another path branches to is the wrong decode — no assembler branches mid-instruction. Mario Bros had a `bne` at `$C4BA` targeting `$C528` swallowed by a `NOP #$A5` at `$C527`.
|
||
- **Branch targets inside a surviving instruction must be pruned** from the target set, or the emitter's linear walk consumes those bytes as an operand, the label is referenced but never defined, and **the file does not assemble at all**.
|
||
- **A plain vector trace reaches only ~15% of a real game.** Almost everything sits behind computed dispatch. Look for `jmp ($xx)`, and for dispatchers that pop the table address with `pla` — that form cannot be followed from the dispatcher because the pointer is only knowable in the caller's frame (Mario Bros `$CD9E`). Probing a suspected data region linearly with `nesdis.py --probe` is cheap and settles whether it is code; `--auto-seed` plus a couple of `--extra-seed` addresses took Mario Bros from 14.6% to 72.2%.
|
||
- **When emulating to trace, NMI must fire once per frame, not per `$2002` read.** The main loop polls `$2002` to sync on vblank, and each read clears bit 7, re-arming NMI within the same scanline. Guarding on the vblank flag alone gave **40,900 NMIs across 3,000 frames** and flattened coverage at 26%.
|
||
- **Controller input goes on joy1 = `$4016`** with a real latch/shift register. The poll routine's `lda $4016,X` with X=1 lands on `$4017`, but only bits 0–1 are compared, so driving `$4017` consumes nothing and changes no coverage at all.
|
||
- **`ca65 -l` listings contain no source line numbers** — they emit `<offset>r <n> <bytes> <source text>`. You cannot key listing entries back to source lines, and the listing prints `rr rr` for relocatable words, so byte values must come from `prg.bin`.
|
||
- **py65 is not a build dependency.** Unpack on demand (`pip download --no-deps py65`, then `unzip` into a lib dir; `PY65LIB` env var points at it). `/tmp` is volatile — a reboot wiped it mid-session. Anything reusable belongs in `_tools/`.
|
||
|