LBreakout original pause control and terminal visible nonresponse, 2026-11-10
The original executable supports ordinary local-game pause/resume with p.
The new generated native workflow is published and passes offline consumer
controls. Its single authorized original experiment stopped at the first
unconfirmed visible pause. Score remained 0, the ball stayed attached, and no
life loss, launch, score increase or first-level clear was observed.
Evidence: scratch/runs/20261110-lbreakout-native-pause-score-0057/.
Code commit 001103a7c was committed on main after focused tests and locked
build, then immediately pulled/pushed using strict SSH. The run used that exact
commit, module SHA-256
3e491c0840bb7771328738a0d160b4bbaa3d7ab72074cbd38a7b41e5d7e8ec18,
and 824 SHA/size pins. Exact Git SHA and relevant binary diffs, including small
run-specific helper patches, are retained. No source/media archive was made.
Accepted unchanged fixture evidence from 4fc62b7d7 was hash-verified and
reused without another fixture. Bound-owner selection from 420e12095 remains
unchanged. Nox, Q3, providers and the passive lane observer were not edited.
The authentic EXE hash is
cc1fd5c3d229cc1d4f910e658f5abefa80ef8ebf4b5e47801648f0be0e1ca244.
Its embedded help explicitly says p pauses the game. Original disassembly
shows _handle_default_key at 0x41e474 comparing symbol 112, branching to
0x41e75c. For a local game with an active player, state 8 becomes state 20
and calls _display_text("Pause"); a subsequent p in state 20 restores
state 8. This is an actual original handler, not a guessed pause key.
pause_workflow_source() extends the actual generated browser caller with
ordinary p and bounded play commands. A play command requires an attestation
that the exact previously reviewed image visibly showed pause. It admits that
image through the existing bound-owner chain, sends balanced native p to
resume, a short supported paddle/fire pulse, and balanced native p to pause
before capture. Each UP remains independently scheduled. A failure prevents
later play/pause/input/capture. No guest routine is invoked artificially, guest
memory written, emulator frozen, clock changed or deadline extended.
The actual generated entry point ran offline with process/CDP boundary adapters,
three guest Workers, module/owner binding, pause/resume/play/re-pause, independent
release, wrong command/image binding negatives and sticky first visual failure.
The existing generated caller's unknown-input and refusal controls also passed.
A separate full original launcher completed ordinary close, actual parent wait
0 and independent stdout/stderr EOFs. These are offline controls, not visual
pause acceptance. Initial offline setup errors (an out-of-range test deadline
and a changed helper hash) were fixed here before passing. Seven permanent
inspection/handoff tests, original SDL/handler controls, three native-key tests,
and flock /tmp/wine-build.lock bash tools/build.sh passed before the code
commit. There was no local Chrome or nested worker.
The one original reached Loading, the main menu, Local Game submenu and the
first Late At Night board. All nine returned images were individually reviewed
before dependent input. At the ready board the paddle/attached ball were
centered, the bricks and life indicators intact, and score 0. The next command
armed the accepted passive observer and sent ordinary native p for 80 ms,
then waited 200 ms before capture.
Passive records show:
- Trusted native
KeyPDOWN at 01:03:22.085Z and UP at 01:03:22.164Z. - Independently scheduled UP at its original 80 ms deadline; both ACKs passed.
- Original SDL DX5 consumption of scan 25/data 128 at 01:03:22.157Z and
SDL_PrivateKeyboardentry with symbol 112, down 1. - The observer stopped normally at its existing 200001-boundary limit. It did not record the original pause handler invocation, pause state or P release consumption. The saved page input records show the released P still queued.
The first post-P image at 01:03:22.475Z was byte-identical to level-ready.png
(SHA-256 f6b1be794506bd3d4beda9022cd9665af03f31e80b5fe59200de741ba2f0a7c7).
It showed no Pause overlay. That does not prove the original ignored the key:
SDL consumed its correct symbol. The last saved guest PC was SDL_MapRGB+0xa,
consistent with ongoing surface drawing; it does not establish pause-state or
display completion. The specific terminal condition is that visible pause was
unconfirmed at this first bounded capture. No later capture was used to wait
for an overlay, and no resume/play or Space command was sent.
Immediately upon that image review, owning native cancellation sent SIGTERM to Chrome 6313 and Node 6304 at 01:03:36Z. Node's actual parent wait was -15, with separate stdout/stderr EOF records. No independent Chrome reap is claimed. There was no later game protocol, input, capture, retry, quit or extension.
All 121 remote artifact references, 103 unique contents, were independently
retrieved and SHA/size verified; all 824 remote pins remained unchanged. All
nine PNGs also passed full pngjs decode/CRC verification before normal stop.
Fresh box bx_qyqc2khs had prearmed independently observable immutable watchdog
UUID 7d8ebab5-f789-4b39-bfee-fcd72796a284, original API expiry
2026-11-10T01:25:55.247Z and watchdog deadline 01:25:55.952Z. The sole delegated
normal stop, stop_8cf53214e1f68593, completed unforced at 01:04:32.581Z.
Independent final API state is archived, expiry null. The watchdog's actual
wait returned 0 with both EOFs, separately from the original launcher records.
Inherited boxes, including bx_nmaechnr, were untouched. Home stayed above
1.5 GB free.
The remaining issue is visible completion of the original pause transition; a 200 ms post-pulse capture did not confirm it. Any further original experiment requires separate authorization. This worker releases its files and stops.