diff --git a/docs/compatibility.md b/docs/compatibility.md index de73356..5093c1f 100644 --- a/docs/compatibility.md +++ b/docs/compatibility.md @@ -1,5 +1,26 @@ # Compatibility and validation +## Phase 1 wave 2 bounded observer evidence + +Full authorized wave2 `scripts/test.sh` exits0:462 Python+2 HUD checks, +all existing managed projects and36 GPU/render checks pass with zero Vulkan +validation errors (GTX1070); SDL stream/lifecycle checks and539 original hashes +pass. The initial sandbox run retained8 denied-loopback errors; no tests were +skipped, and the authorized rerun passed. Viewer27 named Node checks and the +observer23 managed checks retain their separate passing receipts. Logs are +local under `artifacts/fidelity/wave2-test*.log`. + +Plan01-04 passes23 managed checks,7 port-process checks and protected groups +Movement28/Director23/Checkpoint4/BodyOwnership3. The real LV1B→VR04→LV1B +fixture retains one observer clock. Native Linux SDL/Vulkan under private Xvfb +on/off passes with4 identical GPU readbacks and zero validation errors, using +actual dirty-source/diff and DLL identities recorded in its local comparison +report. No physical input/audio, full route, original-DxWnd or native Windows +acceptance follows. Debug and isolated Linux self-contained publish pass; +NU1900 is the existing unavailable vulnerability-metadata warning. All539 +originals match. Exact receipts/limitations: +`../.planning/phases/01-reproducible-fidelity-baseline/01-04-SUMMARY.md`. + ## Phase 1 resumed — capability ledger complete User resumed from the usage pause. Plan01-03 is complete:25 capabilities cover diff --git a/docs/decisions.md b/docs/decisions.md index 9251328..60ca468 100644 --- a/docs/decisions.md +++ b/docs/decisions.md @@ -1,5 +1,17 @@ # Decisions +## Passive evidence observation — phase 1 wave 2 + +Keep one optional application-owned clock across Run continuations. Record raw, +requested and actual gate-accepted input separately after existing owner calls; +copy state without new gameplay ownership or callback reads. Script activity, +explicit skip actions and transitions carry actual observed provenance. Observer +failures disable recording with incomplete diagnostics and do not reset/retry +simulation. Output is caller-selected for standalone packages, bounded and +exclusive; fixtures/headless/debug-injected runs are labeled port_fixture. +Observer overhead and incomplete coverage stay explicit. Existing algorithms, +TSK retained/fresh ordering and collision admission gates are preserved. + ## Phase 1 resumed — capability ledger complete User resumed from the usage pause. Plan01-03 is complete:25 capabilities cover diff --git a/docs/progress.md b/docs/progress.md index ff7011b..4eff6a2 100644 --- a/docs/progress.md +++ b/docs/progress.md @@ -1,5 +1,27 @@ # Progress — 2026-11-07 +## Phase 1 wave 2 — review, ledger and passive observation + +Full authorized wave2 `scripts/test.sh` exits0:462 Python+2 HUD checks, +all existing managed projects and36 GPU/render checks pass with zero Vulkan +validation errors (GTX1070); SDL stream/lifecycle checks and539 original hashes +pass. The initial sandbox run retained8 denied-loopback errors; no tests were +skipped, and the authorized rerun passed. Viewer27 named Node checks and the +observer23 managed checks retain their separate passing receipts. Logs are +local under `artifacts/fidelity/wave2-test*.log`. + +Plans01-02/03/04 are implemented: paired actual-speed local playback with exclusive +audio/optional overlays;25 capabilities covering33 requirements and33 checkpoint +rows; optional bounded input/control/continuation observations. Observer checks: +23 managed,7 actual subprocesses, protected Movement28/Director23/Checkpoint4/ +BodyOwnership3. Private-Xvfb native SDL/Vulkan on/off has4 identical readbacks +and zero validation errors; exact source-diff/binary identities and limits are in +`../.planning/phases/01-reproducible-fidelity-baseline/01-04-SUMMARY.md`. +Linux self-contained publish used an isolated artifact directory. All539 original +hashes match. This is bounded tooling/fixture evidence, not physical input/audio, +whole-original, full-route or native Windows acceptance; actual obligations stay +OPEN0/120. Next: Windows checkpoint bundle (plan01-05). + ## Phase 1 resumed — capability ledger complete User resumed from the usage pause. Plan01-03 is complete:25 capabilities cover diff --git a/docs/tasks.md b/docs/tasks.md index c37ce9e..328a594 100644 --- a/docs/tasks.md +++ b/docs/tasks.md @@ -1,5 +1,17 @@ # Tasks +## Phase 1 execution — wave 2 + +- [x]01-02 paired local review:21 Python/27 Node checks and real Chromium + synthetic-media smoke; actual audible original/native review stays open. +- [x]01-03 full ledger:25 capabilities/33 requirements/33 checkpoint rows; + 31 ledger tests and103 combined fidelity Python tests pass. +- [x]01-04 passive observer:23 managed/7 process tests and protected suites; + native private-Xvfb on/off4 readbacks identical, zero validation errors. +- [ ]01-05 prepare complete Windows/DxWnd local checkpoint bundle. +- [ ]01-06 native Linux captures;01-07 combined hands-on session and strict audit. +- [ ]Actual original/native runtime acceptance OPEN0/120. + ## Phase 1 resumed — capability ledger complete User resumed from the usage pause. Plan01-03 is complete:25 capabilities cover