OrganicArt startup: retained evidence cannot identify the cause
Run scratch/runs/20261118-oasaver-no-present-startup-cause-repair/ audits
the two closed scopes from 20261118-fallingl-oasaver-paired-moving-scenes.
The concrete limitation is missing causal startup observations. No applicable
startup remedy is established, and no fresh original was run. OrganicArt
software/WebGL motion remains open. Accepted historical Falling Leaves pairs
remain accepted and were not rerun.
What the actual retained evidence establishes
Both ordinary registered OASAVER.SCR /s launches used published source
5883f89a504a0bf176083098460facc2c690f017 and supplied module
eb2395e1c115980dd78dbb646a26f755a4119a2a1e086942e715240e07c34bc6.
The offline audit rehashes the retained module and every supplied registered
media/DLL entry against each arm's before/after pins. HTTP delivery of that
module and real Microsoft D3DXOF is verified. This does not recover actual
instantiated module bytes: both deferred inventories are empty.
| Requested route | Binding duration | Last successful DLL request | Retained result |
|---|---|---|---|
| software, guest Worker | 15,004 ms | binaries/dlls/d3dxof.dll, 200 |
UNKNOWN_CLOSED |
| WebGL, guest Worker | 15,006 ms | binaries/dlls/d3dxof.dll, 200 |
UNKNOWN_CLOSED |
Both logs show native D3DRM import binding and initial DLL entry returns, followed by creation/showing of the Screen Saver window. Neither records a guest trap or page exception. The same-directory D3DXOF 404 is followed by a successful DLL-directory fetch. Optional shell/type-library/UI requests remain in the logs; none is shown to cause the admission timeout. Fetch success alone does not prove completion of guest asset decoding or scene construction.
The recorded state has one live visible 640×480 Screen Saver window, guest endpoint 1, bridge endpoint 1/epoch 1 and fixed 1280×960 viewport. Ordinary startup changes the screen backing/presentation geometry. No baseline, secondary inspector attachment, whole PNG or completed executor trace exists.
Why “no PRESENT” is not an independently measured cause
The exact released helper uses InspectionScope.bind(). Its readiness value
conjoins one visible window, presenter.presents > 0 and current composed
display membership. The retained geometry state omits both the presenter
counter and composed display membership. Consequently ready:false cannot
separate zero completed presentations from absent composition. The two retained
rows are change-only deduplicated observations, not evidence that only two
polls executed or that the browser event loop stalled.
Executor observations and Debugger byte inventory occur after successful binding and capture. Closure prevents both, correctly. Runtime-only inspection was enabled during startup; early Debugger was false on both arms. There is no recorded guest EIP, bounded public API/loading trace, asset decoder completion, native modal owner/content or separately sampled PRESENT count. Neither an initial native modal nor unfinished loading nor presentation starvation can be selected as the actual cause from these artifacts. Historical CLI scene-clock, device-rebuild, alpha and missing-logo findings do not supply this missing measurement for these threaded browser scopes.
The already published Device2/deferred-Debugger controls were read as distinct authored evidence. They are not an OrganicArt native-RM startup reproduction. No measured causal target exists for an old-fail/new-pass native control or behavior repair here. No helper/runtime change, native remedy, locked-build acceptance or repaired-original claim is made. The released helper's focused deferred inventory success/error and late-capture closure tests pass again.
Stop condition and recovery
The two conditional fresh-original authorizations require an applicable native-proved remedy first; that prerequisite is unmet. No original retry, closed endpoint request, input, guard change, deadline extension, forced viewport, rebaseline, asset suppression or fake paint was performed. Resolving the remaining cause requires a separately authorized diagnostic original scope that records the missing startup facts before its existing finite closure. No Windows behavior was assumed and no Wine reference was used.
audit.cjs records named retained evidence hashes and rechecks both original
media lists without copying them. audit.json distinguishes delivery identity
from unknown actual execution identity. Git plus current-relevant.patch
records this task's source context; the original source pins, observer suffixes,
module and recovery recipes remain in the predecessor run. No source archive,
media copy, screenshot series or capture bundle was created. Artifact metadata
and cleanup precede final result.json.
No runtime or native-supervisor scope was started, so new native/watchdog waits
are null. The predecessor's actual waits/EOFs, no-leak receipts and cleanup
remain separately referenced and are not relabeled as new execution. The board
watcher is stopped and awaited. Free home disk remains above 1.5 GB; no RAM gate.
Box ID null, state NOT_CREATED, expiry null, stop receipt null. Ownership is
released after docs-only publication. Full27, parity, FPS and audio remain open.