Implement CoCo 3 Quadrapassel and deliver build 01

Add gameplay, keyboard/joystick controls, disk-backed high scores and settings,
standard CMOC effects, and the approved C graphics renderer. Include the clearer
attributed font, title/credits updates, source-verified Modern and Classic (40.2)
scoring, meaningful regression fixtures, and recorded guest validation.

Deliver QUADPSL01.DSK with fresh high scores. Retain source, licenses and selected
guest evidence; exclude personal disks, private connections and build scratch.

Host rules/input/persistence/video checks and memory guards pass. Official
512 KB playtests pass the recorded checks; actual 128 KB/physical joystick
qualification and the normal-window buzzing retest remain pending.
This commit is contained in:
Brian Holman committed 2026-11-11 03:14:14 -07:00
1 parent 8a4b44bc73
commit d412e35c4e
79 files changed
+4187 -16

No files matched your search

+123
View File
@@ -0,0 +1,123 @@
# Audio routing correction — November 10, 2026
Brian reported constant noise during keyboard play, including while idle. That
is outside the specification: sound should consist of brief event effects with
silence between them. His earlier pacing feedback remains keyboard evidence;
physical joystick testing is still pending.
## Correction
The old code initialized CardGame audio once at startup, then continuously called
`readJoystickPositions`. The public CMOC documentation explicitly states that
analog sound routing shares hardware with joystick measurements. The permitted
CardGame example's initSound.c selects the DAC and enables sound; beep.c does not
close that path on return. This leaves audio routing unmanaged between events.
The game now calls the documented `Player_disableSound()` helper before joystick
sampling and at startup. Each enabled effect calls `initSound()` immediately
before `beep()` and `Player_disableSound()` immediately afterward. This keeps the
speaker path closed while idle and while the joystick selector is being changed.
The implementation uses standard CardGame and bsound APIs only. It installs no
Player IRQ, supplies no custom interrupt/sound driver, and adds no direct hardware
access. The quiet movement volume and existing event pitches/durations are kept.
The half-frequency constants are the same values the previous code obtained by
integer division; gameplay timing, scoring and save format are unchanged.
Public references read through the installed MCP manual:
- `programming-in-c/api.html#api-player-disablesound`: analog-path enable/disable
shares hardware with joystick measurements; link bsound.
- `programming-in-c/api.html#api-beep` and public cardgame.h: initialization,
positive period/duration and bounded volume.
- Public Player.h: standalone enable/disable helpers; Player_init/shutdown are
associated with the optional player/IRQ service, which this game does not use.
- Pierre Sarrazin's local CardGame 0.1.15 initSound.c/beep.c (public-domain
examples) were read for reference. No assembly or register code was copied.
## Actual MCP captures
Official VCC Plus 3.5.5 (48), disposable CoCo 3 / 512 KB workers, stock fast mode.
Each WAV captures 120 emulated fields directly, at 44,100 Hz, PCM16 stereo,
88,651 or 88,652 sample frames. Both channels were checked.
| Context | Result |
| --- | --- |
| MC6809 title idle | Every sample zero |
| MC6809 gameplay idle | Every sample zero in the captured no-event window |
| MC6809 right-move action window | Short effect, peak 1,792 in each channel; 10 changing edges; last 500 ms exactly zero |
| MC6809 idle after movement | Every sample zero |
| MC6809 hard-drop/lock window | Short effect, peak 5,120 in each channel; 10 changing edges; last 500 ms exactly zero |
| MC6809 pause menu idle | Every sample zero |
| MC6809 Sound OFF, following a left-key action | Every sample zero; OFF was also verified on the guest settings screen |
| MC6809 Sound OFF idle | Every sample zero |
| HD6309 gameplay idle, reset emulation mode | Every sample zero |
| HD6309 right-move action window | Short effect, peak 1,792 in each channel; 10 changing edges; last 500 ms exactly zero |
The movement peak is lower than the landing peak, preserving relative volume.
Positive effect captures establish that silent windows are not simply a globally
muted capture endpoint. Captures start after bounded input, so a silent post-key
window alone does not establish whether an earlier very short effect played.
The unchanged baseline disk was also captured. Its sampled idle and post-movement
windows were zero. **The reported continuous noise was not reproduced in the
isolated worker.** The routing defect was corrected from the public contracts
and source usage; confirmation that Brian's actual-app idle noise is gone still
requires retesting that environment. These are direct guest PCM checks, not a
claim of human listening, Mac output-device qualification or a 128 KB pass.
`tools/check_audio.py` checks all PCM channels for exact digital silence, or for
a nonconstant effect with a silent final 500 ms. Example:
```sh
python3 tools/check_audio.py test-output/audio/after-idle-game.wav --expect silent
python3 tools/check_audio.py test-output/audio/after-lock-effect.wav --expect effect
```
Raw WAVs and full receipts remain in ignored test-output/audio/. The MCP baseline
used a freshly built disposable disk. The pre-update local writable image also
contained QPSAVE0.DAT/QPSAVE1.DAT; that original is preserved under ignored
local-media-backups/. A second private copy there replaces only QUAD.BIN with
the corrected executable. Every other file and the DOS boot track were verified
byte-identical. These personal save files are excluded from Git and MCP inputs.
The personal emulator session was not operated or reset.
## Build identity
- Project revision 16, Release, build `CFC0A2F2-9247-4B1C-9CF8-8CC1AFE896B6`.
- Artifact `261716EE-81C8-487D-A7BF-E5C9F51BC044`.
- Binary: 26,370 bytes, SHA-256
`5c9f3be05fde79e09a5a583740e676f2fb06eddc136dd9d419f62f12da6bebfe`.
- Pristine Quadrapassel.DSK: 161,280 bytes, SHA-256
`d23bc2a1aa60970becaf7ff8be064efc8ea47a5043adf9661d9a6c22bcedfd9b`.
- Code/constants end $7AE8; static data end $1318; existing memory checks passed.
- MC6809 run `0B67B031-7CEE-4CD7-98CA-649B295FECB3`.
- HD6309 run `356B3452-D3AA-4EBA-8111-1D41EC9F9277`.
- All owned runs are stopped. No implementation commit or push was requested.
## November 11 — normal-window buzz still reported
Brian confirmed the font candidate buzzes on its title screen before Start,
including after quitting/reopening VCC Plus. The real 512 KB CoCo 3 does not
buzz. During the authorized desktop check, pausing VCC Plus stopped the noise;
Run was then restored without resetting the game or replacing mounted media.
This establishes a normal-window playback symptom, not its cause.
The visible normal setup is MC6809/512 KB at 1x, with an FD-502/floppy controller,
Speech/Sound Cartridge, Orchestra-90 and a stopped Hello World cassette. App
sound is enabled at 50%. Those peripherals differ from the supported disposable
worker; their presence alone does not establish that any cartridge causes the
buzz. The public MCP audio tool captures only the disposable worker, not the
normal window or microphone. No claim of hearing the actual speaker is made.
The new standard candidate reasserts Player_disableSound after joystick button/
position sampling as well as before it. This is a defensive public-library
change, not a demonstrated cure. It adds no custom audio code or hardware access.
Official build 51, project revision 2, run
`0174149F-48E6-4EB6-8D96-BC1BF0BF4C87`: title and gameplay idle captures are
zero in both channels; movement peak 1792 in both channels, nine changing edges,
last 500 ms zero. Each direct stereo capture covers 120 fields; the effect
capture contains 88,651 frames, idle captures 88,652 at 44,100 Hz. Checks pass.
Owned runs are stopped. The normal-window buzz remains unconfirmed fixed and
needs the new candidate retested in that same environment. Full final identity
and preserved-score disk are in TITLE_AND_SCORING_UPDATE.md.
+35
View File
@@ -0,0 +1,35 @@
# Numbered game disks
Brian requested this delivery convention on November 11, 2026:
`QUADPSLxx.DSK`, with a two-digit build number and fresh high scores for each
new delivered build. Begin at 01. The next changed delivery is 02. This number
identifies the game disk, independently of the VCC Plus application build.
| Build | Date | Filename | Contents / validation |
| --- | --- | --- | --- |
| 01 | 2026-11-11 | QUADPSL01.DSK | Current title spacing, credits, scoring labels and extra standard-library mute; byte-identical to the verified clean official artifact described in TITLE_AND_SCORING_UPDATE.md |
Build 01 is 161,280 bytes, SHA-256:
`4ac171e87c51dfd99b302b9e035a91a59c030144e5ae0661a6d2dfc84ba3a63f`.
It contains only QUAD.BIN, WORDS.DAT and QPTEXT.DAT, plus the existing DOS boot
track. Neither QPSAVE file is present, so the game starts with default preferences
and an empty scoreboard, creating save data through the normal save path.
The official build recipe continues producing the clean `Quadrapassel.DSK`
staging artifact. After a successful qualified build, inspect its directory for
unexpected files and personal QPSAVE data, then copy it without overwriting to
the next `QUADPSLxx.DSK`. Verify the copy hash equals the official artifact hash
and record the delivery here. Rebuilding unchanged content does not consume a
new number; a changed delivered candidate does. Do not reuse an old numbered
filename, migrate personal scores automatically, or overwrite/remount a user's
disk. Keep numbered deliveries distinct from writable playtest copies.
The name change alone required no guest code rebuild or new gameplay test:
build 01's bytes match the already-tested artifact. It does not establish that
the reported normal-window buzzing is fixed. Existing personal score disks and
backups were left unchanged.
Brian authorized committing and pushing the implementation and build 01 on
November 11. The Git commit adding QUADPSL01.DSK records its source snapshot;
use `git log -- QUADPSL01.DSK` to identify that commit. The game bytes were
qualified before committing and are unchanged from the official artifact.
+41
View File
@@ -0,0 +1,41 @@
# Reviewed decisions
## November 10, 2026 — C graphics exception
Brian's explicit decision, relayed by the coordinating chat **Deploy VCC Plus
Gitea** (`01a1c3a8-8bde-77f0-b8af-d0d29b874c7c`): **“Approve the C graphics
exception.”** The approved wording was:
> C code for screen-memory mapping and drawing changed regions, using CMOC/BGraph
> where practical, with no assembly or emulator changes and explicit 128 KB validation.
Allowed scope: the narrowly proposed CoCo 3 C framebuffer/MMU renderer in
`src/platform/coco3_video.c/.h` and associated memory/build documentation. Prefer
standard CMOC/BGraph primitives where practical. Bound dirty-region/text drawing,
preserve code, stack, BASIC graphics/disk workspace and interrupt behavior during
temporary mapping, and restore the state needed by standard APIs and disk calls.
Retain the public-API fallback and compare measured drawing and visible input.
This approval does **not** authorize hand-written assembly, custom sound/input/
interrupt drivers, emulator source access or changes, custom app builds, or a
replacement MCP integration. Official VCC Plus 3.5.5 (48) and reference boundaries
remain required. Further exceptions need review. Actual 128 KB testing remains
separate from the 512 KB MCP worker; Brian offered to perform that manual check.
### Evidence prompting review
Official build 48, Release `-O2`, isolated MC6809, 512 KB, NTSC-rate worker:
- One 8×8 block erase/redraw through `hset`: worst 4 BASIC ticks at ordinary speed.
- T-piece one-cell shift: only four changed 8×8 cells (256 plotted pixels), worst
12 ticks at ordinary approximately 0.895 MHz speed.
- Same 60-shift test using standard `setHighSpeed(TRUE)` at approximately
1.79 MHz: p95 and worst both 6 ticks, approximately 100 ms.
Build `B928140E-6147-4DB5-BEF0-729678C498B0`, MCP revision 2, artifact
`12931271-4662-4B9F-B138-03913F23E58B`, disk SHA-256
`87f9d22c0afe7313f08031b9b4d88c2e0b6eeb4e81050739d108d82fe1fe76ad`.
These measure foreground drawing cost, not full-game input-to-display latency.
Smaller blocks, sparse outlines and staged `hset` drawing remain possible API-only
alternatives, with board-area, appearance or latency compromises. No direct screen
code was implemented before this approval was received and recorded.
+79
View File
@@ -0,0 +1,79 @@
# Font and block-outline update — November 11, 2026
Brian approved the installed official VCC Plus 3.6.0 (51), project reconnection
and isolated build/playtest. Base source revision remains
`8a4b44bc7358b1a999296d14d8c5272e57b65992`, main, plus uncommitted implementation.
Nothing was committed or pushed.
## Result
Menu/HUD text now uses unchanged public-domain CardGame 0.1.15 glyph row data
by Pierre Sarrazin, also used by Color8. Original files, notices and hashes are
retained in `third_party/cardgame-font/`; the bounded subset is in
`src/platform/pierre_font.h`. The renderer uses six visible columns and seven
rows in the existing 8x8 cells. Lowercase input maps to uppercase glyphs.
No other game's renderer or assembly was copied.
Modern tile outlines now use palette slot 10, RGB(1,1,1); the black background,
seven piece colors and Plain/Clean styles are preserved. The title credit asset
is exactly `(c) 2026 Brian Holman`, displayed in the uppercase character set.
Pierre's font contribution is also credited on the credits screen.
## Validation and identity
- `python3 tools/test_video.py`: passed ASan/UBSan. All supported glyphs match
the original font pixel-for-pixel, including parentheses and lowercase mapping.
Last-cell boundaries, all eleven palette colors, all seven Modern block colors,
black background and Plain/Clean styles pass.
- Official MCP Release disk build: project
`84772EFC-B651-4A98-B09E-932247575EA7`, revision 1,
build `FAD8FBEA-F07D-4FE8-AE45-E569EEA5E5CC`,
artifact `4566B606-6576-482A-BA3E-BCC673EC729B`, zero diagnostics.
- QUAD.bin: 26,589 bytes, SHA-256
`9eec745d50544884d74211492f0118360d79212f05864d9e81fa43a1a56ed041`.
- Clean Quadrapassel.DSK: 161,280 bytes, SHA-256
`12f0ee508d5b48bbbe44d1f2b701eb31ae5bd5d3cbb863e1005540524d5dbfb3`.
Local deliverables match the official artifact hashes.
- Enforced map: code/constants end $7AFE, static data end $13DD. Existing
$7B00/$1400 bounds and 1,024-byte stack reserve retained. Alphabet, punctuation
codes and packing table use 197 bytes in the existing low data region and
are only read by the renderer. No new graphics mode or 512 KB requirement.
- Approved C renderer's compiler output still reads the MMU bank via the separate
read function, performs the bounded copy and restores the prior mapping.
- Owned run `1EBBD56F-EB22-4E17-BF49-8C943FB25F04`: MC6809, 512 KB, current
artifact verified. DOS startup, title, settings, normal settings save/return,
gameplay, Hold/drop, title credit and gray outlines inspected. Run stopped.
- Direct 120-field stereo WAV capture: idle peak 0 in both channels; movement
peak 1792 in each channel, nine changing edges, final 500 ms silent.
`tools/check_audio.py` passed both checks. Raw receipts/WAVs remain ignored.
Actual framebuffer evidence (320x192 HSCREEN 2):
![Title](screenshots/font-title-512k.png)
![Settings](screenshots/font-settings-512k.png)
![Gameplay](screenshots/font-gameplay-512k.png)
These are supported paused guest framebuffer captures, not host-test renderings.
The host pixel preview is supplemental only. The larger title lettering remains
code-drawn from its existing compact shape data.
## Existing saves
A new ignored copy, `local-media-backups/2026-11-11-font-update/Quadrapassel-Font-Update-With-Saves.DSK`,
contains the new QUAD.BIN and QPTEXT.DAT. QPSAVE0.DAT, QPSAVE1.DAT, WORDS.DAT and
DOS track 34 were verified unchanged. The source disk and a before-update copy
are retained. Verification hashes are in that folder's private VERIFICATION.json.
Updated saved-disk SHA-256:
`f5ec52d909eb6adc58c7feb5c8497ca04b25a6f7a49247a9793554787970cb60`.
The clean top-level disk contains no personal saves. The normal running emulator
and its mounted media were not replaced or reset.
## Limits
This update was tested in the official 512 KB worker. Actual 128 KB testing,
physical display appearance and physical joystick checks remain pending. Earlier
user-reported physical 512 KB gameplay and prior earned-score cold-boot evidence
are recorded separately in IMPLEMENTATION_STATUS.md. The font update changes
neither persistence codec nor rules. It does not establish a new 6309 pass or a
full timing/stack qualification. The memory budget has only two code bytes and
35 static bytes spare, so further feature work needs compaction first.
+164
View File
@@ -0,0 +1,164 @@
# CoCo 3 implementation status — November 11, 2026
The current disk contains a playable CoCo 3 game with standard CMOC sound,
settings and disk-backed high scores. Actual 128 KB qualification remains
pending; CoCo 2 adaptation is deferred at Brian's direction.
## Current numbered delivery
Brian requested `QUADPSLxx.DSK` with a two-digit build number and fresh high
scores. The first is **QUADPSL01.DSK**, byte-identical to the clean official
artifact below. No QPSAVE files are included; old personal score disks remain
untouched. The next changed delivery is 02. See [BUILDS.md](BUILDS.md).
## Current title, scoring-label and audio follow-up
The standard Release candidate now has an eight-pixel gap before the title
copyright, the credits entry `Brian Holman - Mac & Coco`, and explicit
**Scoring Rules: Modern / Classic (40.2)**. Keep high scores is the default;
no personal records were erased. Inspection of Brian's supplied Mac source
confirmed the existing scoring formulas, including the Classic 40.2 bottom-row
bonus. All 25 cases also pass through actual piece drops.
Official build 51, project revision 2, build
`DFDE8C57-3165-44B5-9027-E8B2F16DDF28`, artifact
`BB277DE8-07A2-437B-9998-FFD1FB2C36C5`, zero diagnostics.
Clean disk SHA-256:
`4ac171e87c51dfd99b302b9e035a91a59c030144e5ae0661a6d2dfc84ba3a63f`.
Code/constants end $7AF4; static data end $13DE. The normal-window buzz persists
in Brian's prior candidate after restart and stops when paused; it remains
unconfirmed fixed. An extra standard-library mute after input sampling passes
worker silence/effect checks. Details, captures and score-preservation proof:
[TITLE_AND_SCORING_UPDATE.md](TITLE_AND_SCORING_UPDATE.md).
## Previous font candidate
The November 11 font/border/copyright candidate passed an official VCC Plus
3.6.0 (51) Release build and MC6809 / 512 KB disposable playtest. The font uses
Pierre Sarrazin's public-domain CardGame data, with attribution. Modern outlines
are dark gray while the background stays black. Title text is
`(c) 2026 Brian Holman`, displayed in the font's uppercase character set.
- Project `84772EFC-B651-4A98-B09E-932247575EA7`, revision 1.
- Build `FAD8FBEA-F07D-4FE8-AE45-E569EEA5E5CC`, artifact
`4566B606-6576-482A-BA3E-BCC673EC729B`; no compiler diagnostics.
- Binary 26,589 bytes, SHA-256
`9eec745d50544884d74211492f0118360d79212f05864d9e81fa43a1a56ed041`.
- Pristine DSK SHA-256
`12f0ee508d5b48bbbe44d1f2b701eb31ae5bd5d3cbb863e1005540524d5dbfb3`.
- Code/constants end $7AFE; static data end $13DD. Existing limits and
1,024-byte stack reserve retained. Further growth needs compaction.
- Pixel tests compare against the unmodified upstream font; palette, all seven
Modern tile colors, Plain/Clean behavior and edge-cell clipping pass under
AddressSanitizer/UndefinedBehaviorSanitizer.
- Actual title, settings and gameplay framebuffer captures inspected. Hold/drop
exercised; a normal settings save wrote the disposable disk successfully.
- Captured idle audio is zero in both channels; movement has a brief effect and
silent final 500 ms. Owned run stopped.
- New private disk copy preserves both existing high-score files byte-for-byte.
Details, screenshots, saved-disk identity and remaining limits:
[FONT_UPDATE.md](FONT_UPDATE.md). Persistence code and gameplay rules are unchanged;
prior cold-boot score evidence below remains applicable to that unchanged code.
Actual 128 KB and physical joystick qualification remain pending.
## Previous audio candidate — November 10 build identity
- Base Git revision: `8a4b44bc7358b1a999296d14d8c5272e57b65992`, main, with
uncommitted implementation. No implementation commit or push was requested.
- Official application: VCC Plus 3.5.5 (48), supported C-project MCP connection.
- Previous code build: project revision 16, Release,
build `CFC0A2F2-9247-4B1C-9CF8-8CC1AFE896B6`,
artifact `261716EE-81C8-487D-A7BF-E5C9F51BC044`.
- Binary: 26,370 bytes,
SHA-256 `5c9f3be05fde79e09a5a583740e676f2fb06eddc136dd9d419f62f12da6bebfe`.
- Previous pristine `Quadrapassel.DSK`: 161,280 bytes,
SHA-256 `d23bc2a1aa60970becaf7ff8be064efc8ea47a5043adf9661d9a6c22bcedfd9b`.
This matched that official MCP artifact, not a saved playtest copy.
- Linked code/constants end at $7AE8 (24 bytes before the enforced $7B00 limit).
Writable static data ends at $1318 (232 bytes before $1400). The stack reserve
is 1,024 bytes; detailed contract and qualification limits are in MEMORY_LAYOUT.md.
- The pristine disk has 50 free granules and no save files. Two 35-sector save
files need four granules each. DOS loader tracks are reserved by the template.
## Implemented
- All 28 piece states, safe above-board behavior, specified kicks, gravity,
hard/soft drop, persistent Hold swap, shadow and simultaneous compaction.
- Modern/Classic 32-bit saturating scoring, level transitions and Relaxed mode.
- Version-1 xorshift32 seeded stream, inverse-square balancing, prefill and
Difficult Pieces metric/selection. Difficult search is currently synchronous.
- Keyboard controls and a one-button joystick input state machine, settings,
seed editor, pause/actions/confirmation screens, score browser and initials.
- Code-drawn title and cached tiles; menu focus updates without full redraws.
Scene text is a bounded, checksummed disk asset sharing the display cache.
- Standard CardGame `initSound`/`beep` effects, with the standard bsound mute
helper closing audio around joystick sampling and between effects. No custom driver.
- Versioned QPSAVE0.DAT/QPSAVE1.DAT store for all 34 categories, ten records each,
initials and scoring-rule markers. CRC/range/length validation, generation
selection, older-slot recovery, header-last commits and readback verification.
Only those two named files are written; errors are visible and play can continue.
## Evidence and limits
| Check | Result |
| --- | --- |
| Host rule fixtures, sanitizer build | Passed: 28 states, all kick fixtures, bounds/top-out, compaction, 25 scoring cases, gravity, Hold, RNG/prefill and Difficult metrics |
| Host input fixtures, sanitizer build | Passed: repeat/conflicts/edges, fire timing, hysteresis, diagonal and disconnected-stick keyboard behavior |
| Host persistence fixtures, sanitizer build | Passed: 34 categories, stable ties, rule markers, 32-bit values, CRC, read-only reset, all 36 logical torn-write boundaries and recovery |
| Official Release build and memory bounds | Passed, final build above, no compiler diagnostics |
| MC6809 / 512 KB | Passed: DOS title, gameplay, movement/rotation/Hold/hard drop, settings, cached settings redraw and fast menu focus |
| Settings save and cold-boot reload | Passed on revision 12: Sound OFF and Custom Seed ON reloaded from the same disposable disk |
| Earned high score save and cold-boot reload | Passed on revision 12: seed zero, 12 authored placements, one line/40 points, initials BRH, Modern marker in difficulty-0 Normal category; survived another cold boot |
| Clean BASIC return and manual fallback | Passed on revision 13: Quit returned OK; PCLEAR1, LOADM"QUAD.BIN", EXEC reached the title |
| HD6309 / 512 KB | Revision 16 idle/effect audio checks and prior title/gameplay framebuffer smoke passed in reset emulation mode; not a native-mode qualification |
| Physical CoCo 3 / 512 KB | Brian reported successful gameplay on November 11, 2026 ("It worked great"); RAM confirmed as 512 KB. Exact disk hash, CPU upgrade status and individual audio/save/joystick checks were not specified |
| CoCo 3 / 128 KB | Pending actual application/manual test; MCP workers remain fixed at 512 KB |
| Physical joystick ports and joystick-only usability | Pending; logical fixtures/neutral samples are not physical-port passes |
| Guest read-only/corrupt media and interrupted FAT/directory writes | Pending; logical host fault tests do not establish these physical disk cases |
| Direct guest audio, idle/mute and bounded effects | Passed captured windows on 6809 and 6309; actual-app reported noise retest pending; see AUDIO_VALIDATION.md |
| Current input/render latency, sound-on worst case, clears and Difficult search | Not fully qualified; earlier renderer probe timing is not a complete-game timing pass |
| PAL/50 Hz, physical Intel, CoCo 2 | Not qualified by these runs |
| Debug/Probe configurations with the final split layout | Not qualified; use Release for this disk |
The earned-score run was `9AFA5124-4FDE-4E3D-95EE-228FCE945951`.
Its retained writable disk SHA-256 is
`44ca3987b8d3a1e02ee41f1a0de8bd587be5a6835f8bdc5186b22138e6be0355`.
That disk is separate from the pristine deliverable. All owned runs are stopped.
Final menu/cache and 6309 changes did not alter the validated persistence codec.
## Capture interpretation
Presented-frame PNGs sometimes omitted unchanged screen content during these
runs. The supported paused `capture_playtest_graphics` preview showed the title,
HUD/frame and cached settings intact in the current physical HSCREEN 2 buffer.
Visual verification therefore uses these direct guest-framebuffer previews in
addition to ordinary screenshots. No emulator internals were inspected and no
emulator change is required for the current disk.
Selected evidence is under docs/screenshots/, including
high-score-saved-512k.png, high-score-reloaded-512k.png,
title-clean-512k.png, gameplay-framebuffer-512k.png,
settings-cached-framebuffer-512k.png and
gameplay-6309-framebuffer-512k.png. Reference art is not runtime evidence.
Next qualification: follow TESTING_128K.md, measure complete-game latency and
Difficult search separately, then physical joystick/read-only/fault cases.
Further feature growth needs code compaction before exceeding the enforced limit.
## Latest user feedback and sound correction
Brian reported good keyboard pacing, then constant audio noise even while idle.
The standard-library audio routing correction and direct MCP stereo WAV checks
are recorded in [AUDIO_VALIDATION.md](AUDIO_VALIDATION.md). The isolated worker
did not reproduce the original constant noise; the actual application needs a
retest. The save codec and gameplay rules were not changed.
## November 11 — physical CoCo 3 playtest
Brian reported playing Quadrapassel successfully on a real Color Computer 3
and described it as working great. He confirmed the RAM size as **512 KB**.
This is user-reported physical-hardware gameplay evidence, separate from the
isolated emulator runs. It does not establish a 128 KB pass or separate passes
for the audio correction, persistence, joystick ports or every game option.
The exact tested disk/build and detailed machine configuration were not supplied.
+61
View File
@@ -0,0 +1,61 @@
# MCP kickoff and connection inventory
Verified November 10, 2026 against installed official VCC Plus **3.5.5 (48)**.
Keep private grant files and tokens outside Git. This is a connection record,
not permission to inspect emulator implementation or another game's project.
| Connection | Live tools used | Verification |
| --- | --- | --- |
| Quadrapassel C project | `get_capabilities`, `read_project_inputs`, `refresh_project`, `build_project`, `get_build`, owned-playtest tools | Project `A1AFAA7E-C427-4437-8D1C-EB68D8C225EF`; this repository's root Makefile, files and hashes |
| Installed manuals/API (through the C connection) | `search_help`, `read_topic` | Read HSCREEN/HSET, keyboard, joystick, timer, SOUND, beep and PCLEAR topics |
| Archive Library (separate connection) | `search_archive`, `get_archive_citation` | Both succeeded; found Tandy's Color Computer 3 Extended Basic, PDF pages 152–153 |
At this chat's start no VCC Plus tools appeared in its tool catalog. Connections
were exported through the installed app's normal native interface. Its public
manual, `vibe-coding-with-chatgpt/01-first-connection.html`, explicitly permits
ordinary STDIO MCP through `/Applications/VCC Plus.app/Contents/Helpers/vcc-mcp`
with the two arguments `--grant` and the private grant filename. The unmodified
installed helper is used by a small MCP client; no alternative server, private
bridge or integration endpoint is used.
Archive Library has its own link button and **Save Archive Search Connection…**
flow. Brian explicitly approved this connection on November 10. Its advertised
build-48 tools return citations/locations, rather than full page text. Use the
Archive Library native reader for the original page when needed. Do not infer
that Archive MCP is unavailable from the C connection's tool inventory.
For future kickoff checks, distinguish **configured**, **tools discovered**, and
**successful request**. Inventory each required connection separately before
implementation or web fallback. Confirm the intended project and current revision
before builds. Connection identity and advertised tools may change after a new
export; discover them again, rather than inventing tool names or endpoints.
Example verified Archive reference:
- Title: *Color Computer 3 Extended Basic (Tandy)*.
- Citation: `sha256:26d4ca51f3477263f3e85269091104079ae88ad83c842cef0e3bc531324e7f5e:pdf_page:152`.
- Read in Archive Library: printed page 143, “Adding Pages (The PCLEAR Command)”.
It describes the default four PMODE pages and reserving fewer pages.
Initial hardware-reference web lookups occurred before discovering the separate
Archive connection. That was a discovery error, not a demonstrated Archive gap.
Brian asked that the kickoff feedback be sent to **Deploy VCC Plus Gitea**; the
approved concise feedback was delivered there. No emulator update is requested.
## November 11 reconnection
Brian approved installed official VCC Plus 3.6.0 (51) and an idle desktop window.
The same repository was opened through C Development and its private project
connection renewed outside Git. A successful `get_capabilities` returned project
`84772EFC-B651-4A98-B09E-932247575EA7`, session
`185C8449-CB9B-4A38-81C8-D43CDA4A3ABE`, app build 51, approved Make recipe,
and the expected Quadrapassel files. This connection still advertises fixed
512 KB disposable workers, MC6809/HD6309, graphics and audio captures.
Archive access is separate and was not needed for this local font adaptation.
After Brian's November 11 restart, the old private grant expired. The same
project connection was renewed through the supported interface. Current project
`3D6A502C-E06D-46BC-9D19-191D741A16B9`, revision 2, built and ran successfully
on official build 51. get_machine_status read normal-session configuration;
no writable disks or peripherals were removed to satisfy native-run blockers.
The new candidate used only disposable playtests.
+49
View File
@@ -0,0 +1,49 @@
# CoCo 3 memory contract
Current candidate uses stock Disk BASIC and `PCLEAR1`, then
`LOADM"QUAD.BIN":EXEC`. `PCLEAR0` was tested in an isolated official build-48
MC6809/512 KB machine and returned `?FC ERROR`; it is not used.
The standard CMOC linker places code/constant data at $1400 and writable static
data at $0E00. The one reserved legacy PMODE page ($0E00–$13FF) is deliberately
used for the game's static variables. This is a CoCo 3-only program: it never
calls PMODE or PCLS. HSCREEN 2 uses its separate physical graphics banks $30–$33.
Graphics is initialized once before preferences are created. Scene changes clear
only the framebuffer through the approved renderer, keeping ROM initialization
outside live game state. No BASIC user program or prior
low-resolution picture is preserved by launching
this game. BASIC system RAM, text RAM and disk buffers below $0E00 remain intact.
The November 11 layout/scoring-label update ends code/constants at $7AF4 and static data at
$13DE (1,502 bytes including initialized variables). The limits remain unchanged.
The font uses 336 row bytes plus 12 punctuation codes; 198 bytes of font/packing
data are stored in the low data region and only read by the renderer. $7B00–$7FFF is kept free for
the foreground/ROM stack, with CMOC configured for 1,024 bytes. The split layout has passed settings and earned-score save/reload checks on
MC6809/512 KB in the official app. Full stack qualification and actual 128 KB
testing remain pending. This is a checked
address budget, not a claim that 128 KB has passed.
`tools/check_memory.pl` verifies every nonempty link-map section independently.
This is necessary because CMOC's `--limit` checks `program_end`, which is the low
static-data end when `--data` is used. The build fails if code/constants extend
past $7B00 or static data extends past $1400. No data or stack may occupy the
renderer’s temporary $8000–$9FFF MMU window.
The approved C renderer temporarily maps task-0 slot 4 to graphics banks $30–$33,
copies bounded chunks and restores the prior mapping before standard APIs run.
CMOC does not implement volatile: a separate read function is used and emitted
reads/stores must be audited in each delivered candidate. No authored assembly,
custom interrupt handler or custom sound driver is used.
Menu text is loaded through standard decbfile APIs from twelve independently
checksummed 512-byte pages in QPTEXT.DAT. The menu buffer shares storage with the
288-byte tile cache and 200-byte board cache; gameplay rebuilds those caches on
resume. Original title lettering uses 84 bytes of shape data and a scanline
scratch buffer, rather than a full-screen bitmap.
The score store keeps one 256-byte page and one 80-byte top-ten table in RAM.
QPSAVE0.DAT/QPSAVE1.DAT each hold 35 sectors, with the commit header written last.
Files longer than the schema allows are rejected and reported; the program does
not truncate an unexpectedly oversized file. Missing/short inactive files can be
created/extended using the documented file API. Only the two named save files
are written. QPTEXT.DAT is read-only game data.
+36
View File
@@ -71,3 +71,39 @@ of pieces above the board. They must not be described as byte-for-byte Mac parit
Hardware feasibility, timing, physical palette appearance, API performance and
disk layout are still to be demonstrated by the implementation. The arithmetic
memory estimates and reference tests do not establish a working CoCo build.
## CMOC example references consulted during implementation
Pierre Sarrazin's local CardGame 0.1.15 `beep-demo.c` (public domain) was consulted
for `initSound`, `getTicksForFrequency` and `beep` usage. CoCoLetters 0.1.0
`file.c`/`file.h` (Copyright 2026 Pierre Sarrazin, GPL-3.0-or-later) were consulted
for `decb_init`/`decb_registerDrive` initialization. The disk adapter follows
that initialization pattern, adds error handling, and uses the installed public
API contracts for writes, lengths and close/flush behavior. No standalone disk
driver, assembly sound routine or other game's rendering engine was copied.
The existing DOS loader is Pierre Sarrazin's public-domain autostart.c 0.1.2,
retained from the official disk-project template with the launch command adapted.
The audio correction also uses the documented bsound Player_disableSound API.
Only its standard mute helper is used; no Player IRQ or custom sound driver is
installed. The CardGame public-domain example implementation was consulted to
understand setup/teardown, without copying its assembly/register code.
## November 11 font update
The menu/HUD font is an unchanged subset of Pierre Sarrazin's public-domain
CardGame 0.1.15 font, also used by Color8. The original two font files, exact
SHA-256 values, source URL and public-domain notice are retained in
`third_party/cardgame-font/`. `src/platform/pierre_font.h` contains 26 letters,
ten digits and twelve punctuation glyphs. No drawing engine or assembly was
imported. Existing lowercase input maps to uppercase glyphs. The new C renderer
uses the original six visible columns/seven rows in each eight-pixel cell.
Brian explicitly permitted inspecting the supplied Mac scoring source on
November 11. The read-only checkout at revision
`be8769a403e506edb930372f963a9c4c83a0881e` confirms the existing Modern and
Classic (40.2) formulas. `scoring_through_drop` in tests/rules_test.c adapts the
GPL-3.0-or-later setup in its tests/test-scoring.vala to exercise actual piece
drops, bottom-row detection and level changes in the CoCo rules engine. No Mac
application code was changed or bundled.
+35
View File
@@ -0,0 +1,35 @@
# Manual CoCo 3 / 128 KB check
Use the official installed VCC Plus application. The MCP worker is fixed at
512 KB; its results below do not establish a 128 KB pass.
1. Preserve your current setup and media. In a separate disposable test setup,
select CoCo 3, 128 KB, MC6809 and ordinary stock speed. Use RGB/NTSC initially.
2. Mount a **writable copy** of the pristine top-level `Quadrapassel.DSK` in drive
0. Do not overwrite a disk containing your own programs or scores. Type `DOS`.
3. Confirm the title, menus and seven piece colors appear. Start a game; test
movement, rotation, soft/hard drop, Hold and pause/resume. Check that the board
frame and score/level/line labels remain visible while pieces move.
4. Change Sound and Custom Seed in Settings, use Back to save, and wait for the
save message to finish. Reboot the same writable disk and verify the settings.
5. Earn a positive score, end the game, enter initials and choose Save to Disk.
Reboot again and verify initials, score, category and Modern/Classic marker.
6. Repeat with sound on. The audio-routing correction should keep the title,
paused game and idle intervals quiet between event sounds. Check menu focus
moves promptly, tones are audible and
distinct, and ordinary actions are responsive. Try Difficult Pieces separately
and report any long stalls; its worst-case timing is not yet qualified.
7. On a separate read-only copy, try saving a setting. Expect an error and
continued play, without a claim that the setting was saved.
8. If available, check both physical joystick ports, neutral calibration,
one-button rotation/actions, and joystick-only menu/game use. Keyboard tests
are not joystick qualification.
The fallback launch sequence is `PCLEAR1`, then `LOADM"QUAD.BIN"`, then `EXEC`.
Do not use PCLEAR0: official build-48 Disk BASIC rejected it with `?FC ERROR`.
The fallback was separately verified on the clean MC6809/512 KB build; actual
128 KB qualification remains pending.
Please record the disk hash, machine/RAM/CPU, video standard, results and a
representative guest screenshot. The game targets 128 KB by address layout; a
manual pass is needed before claiming actual compatibility.
+85
View File
@@ -0,0 +1,85 @@
# Title, credits and scoring labels — November 11, 2026
## Changes
The title's lower decorative blocks move up eight native pixels, leaving a blank
vertical gap before `(c) 2026 Brian Holman`. The credits entry is exactly
`Brian Holman - Mac & Coco`, rendered in the uppercase font. Pierre's unchanged
public-domain ampersand was added to the subset and pixel comparison fixture.
The redundant pixel wrapper was removed; fallback drawing calls standard hset
directly. No graphics mode or memory requirement changed.
Settings now names **Scoring Rules** and displays **Modern** or **Classic (40.2)**.
The confirmation says the new rule starts next game and offers **Keep high
scores**, **Erase all high scores**, **Cancel**. Keep is the initial selection;
the explanation says erasing clears every difficulty. No personal scores were
reset or recalculated during this work.
Brian supplied and authorized reading the Mac scoring source at
`/Users/bholman/Development/Personal/quadrapassel/`, revision
`be8769a403e506edb930372f963a9c4c83a0881e`. The existing CoCo arithmetic matches
src/game.vala, tests/test-scoring.vala and the settings schema in that checkout.
The Mac checkout remains unchanged. Scoring rules were not modified.
| Rows cleared | Regular points × level | Modern bottom-row bonus × level | Classic (40.2) bottom-row bonus × level |
| --- | --- | --- | --- |
| 1 | 40 | 100 | 10000 |
| 2 | 100 | 200 | 10000 |
| 3 | 300 | 300 | 10000 |
| 4 | 1200 | 400 | 10000 |
The bonus applies only when the cleared rows include the bottom row. Both modes
use the newly reached level when crossing ten total lines; Relaxed stays level 1.
A Modern single line above the bottom earns 40 at level 1, while a bottom-row
single earns 140. Classic's bottom-row single earns 10040. Existing records are
preserved even when switching rules and carry their saved M/C marker.
## Verification
- Official VCC Plus 3.6.0 (51), Release, project
`3D6A502C-E06D-46BC-9D19-191D741A16B9`, revision 2.
- Build `DFDE8C57-3165-44B5-9027-E8B2F16DDF28`, artifact
`BB277DE8-07A2-437B-9998-FFD1FB2C36C5`, zero diagnostics.
- QUAD.bin: 26,580 bytes, SHA-256
`4fbd37aeec7f4edb39037a65f9573fc6b6886136ed87a6b3979e26e4c3020eed`.
- Clean Quadrapassel.DSK: 161,280 bytes, SHA-256
`4ac171e87c51dfd99b302b9e035a91a59c030144e5ae0661a6d2dfc84ba3a63f`.
The local standard build matches those official artifact hashes.
- Link guard passes: code/constants end $7AF4; static data end $13DE; existing
$7B00/$1400 limits and 1,024-byte stack reserve retained.
- Host ASan/UBSan rules fixtures pass, including all 25 score cases through both
direct arithmetic and actual vertical-I drops adapted from the supplied Mac
test. Bottom-row detection, compaction and level transitions are exercised.
- Font pixel tests pass, now including the unchanged upstream ampersand.
- Official MC6809/512 KB run `0174149F-48E6-4EB6-8D96-BC1BF0BF4C87`: title gap,
credits, both scoring labels and Keep-default confirmation inspected. Choosing
Keep and saving settings worked. All owned runs are stopped.
- A standard-library mute call now follows analog input sampling too. Direct
title/gameplay idle and positive movement audio checks pass, but Brian's
normal-window buzzing has not been reproduced in this worker or confirmed
fixed. See AUDIO_VALIDATION.md for the reported restart/pause evidence.
Guest framebuffer evidence:
![Title spacing](screenshots/title-spacing-512k.png)
![Credits](screenshots/credits-mac-coco-512k.png)
![Classic setting](screenshots/scoring-classic-512k.png)
![Keep default](screenshots/scoring-keep-default-512k.png)
## Preserved high scores and delivery
The standard clean top-level disk contains no personal save files. An updated
standard-named private copy is at
`local-media-backups/2026-11-11-title-spacing-final/Quadrapassel.DSK`. Its source
was Brian's mounted font-update disk; the source and a before-update backup are
retained. Only QUAD.BIN and QPTEXT.DAT were replaced in the new copy. Both
QPSAVE files, WORDS.DAT and DOS track 34 were verified byte-identical, and the
source disk is unchanged. The private folder contains VERIFICATION.json.
Updated saved-disk SHA-256:
`2ca5f1f95896f660af998d2f913477205e23814b9fede7e7dbe2110e24143d88`.
No normal-session disk was replaced or remounted.
Actual 128 KB, physical joystick and final-candidate physical-machine tests
remain pending. No new HD6309 qualification was performed. Base source remains
main at `8a4b44bc7358b1a999296d14d8c5272e57b65992` plus uncommitted work; no
commit or push was requested.
+13 -3
View File
@@ -2,9 +2,10 @@
## Required installed app
Use `/Applications/VCC Plus.app`, **version 3.5.5, build 48**. These values were
read from the installed app's `Info.plist` on November 10, 2026 after Brian
confirmed installation of official build 48. Verify again at implementation start.
Use `/Applications/VCC Plus.app`, **version 3.6.0, build 51**, explicitly approved by Brian
on November 11, 2026 for reconnection and isolated build/playtest. The installed
Info.plist was verified. Earlier build-48 evidence remains historical.
Verify again at implementation start.
If the app differs, report it and wait for Brian to approve a different version.
Do not downgrade, overwrite, or build a replacement app.
@@ -139,3 +140,12 @@ For XRoar, give Brian a pristine DSK copy, exact machine/RAM/clock/ROM requireme
loading steps, controls, expected results and an evidence checklist. Do not
redistribute private ROMs. Separate user-confirmed CoCo 2 results from VCC Plus
CoCo 3 compatibility runs. Keep screenshots limited to the guest display.
## Numbered game delivery
Brian's latest November 11 convention is `QUADPSLxx.DSK`, beginning with 01,
with empty high scores/default preferences for each new delivery. Keep the
official recipe and MCP artifact at its existing staging name, Quadrapassel.DSK.
Verify no QPSAVE data is present, copy to a new numbered filename without
overwriting, compare hashes, and record it in BUILDS.md. Old personal disks
remain untouched; automatic score migration is no longer part of delivery.
+13
View File
@@ -0,0 +1,13 @@
# CoCoMan start-screen reference
Brian explicitly authorized sharing this start-screen image with the Quadrapassel implementation chat on November 10, 2026.
![CoCoMan start screen](cocoman-start-screen.png)
Unchanged screenshot supplied from CoCoMan, originally named `convoy38-disk-title.png`. The image retains its visible Brian K. Holman credit.
Use it as a visual reference for a polished CoCo title screen: strong title lettering, a composed illustration, a limited bright palette against black, and a clear keyboard/joystick start prompt. Quadrapassel should use its own game identity and falling-block imagery.
This image-specific authorization does not expand access to CoCoMan source, documentation, other assets, chat history, or its MCP connection. It supplements the existing Quadrapassel title reference.
SHA-256: `2e141915f02743a54db5b9bcdc93d36d74fe080fb7dfd4da945d290b6941cdfc`
Binary file not shown.

After

Width:  |  Height:  |  Size: 18 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 9.6 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 10 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 9.5 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 5.6 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 3.6 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 6.7 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 4.7 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 3.2 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 8.7 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 3.2 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 9.3 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 9.3 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 6.9 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 4.0 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 6.8 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 5.4 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 9.8 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 4.7 KiB