Quake II sparse-reader weapon consequence, 0517

One fresh original produced a verified weapon/world consequence: native Enter fire shattered the framed pane in Outer Base. The personally viewed fresh PNG shows the missing pane, flying fragments and a blaster projectile, with HP100. Guard combat, an operated door and passage through the frame are not established. The last forward movement ended in a close wall view at HP92; damage cause and passage remain UNKNOWN. No second original, retry or later gameplay followed.

Evidence is scratch/runs/20261109-quake2-sparse-native-combat-0517/ (R). The decisive image is R/review/panel-fired.png, 383,572 bytes, SHA256 e118b483a5daab7dcde1272ac1e52ffeb54fa5b2377dee19f6a830665a642a2f. The preceding panel-aim.png and early panel-fired-held.png show the pane still present. The final beyond-shattered-panel.png label describes the intended movement, not proof of traversal. All twelve retained fresh PNGs were personally inspected. PNG chunk CRC, IDAT decompression and capture SHA checks pass. Each dependent input followed review of its current PNG; there was no blind route, rebinding, guest write or source-lifetime qualification.

Actual sparse bindings and input

Read CLAUDE/AGENTS, the 0505 sparse-reader and 0450 persistent-caller notes, and the 0435/0425/0415/0350/0325 notes. Personally viewed the prior armor, outdoor HP46 and death images; those remain historical evidence. In particular, 0505's authored tests do not establish the historical0450 string contents.

The fresh startup read, at 05:21:44.407Z, actually decodes:

Engine key Pointer Fresh actual binding
Enter, 13 0x08ff1a10 +attack
Control, 133 0x08ff1f30 +movedown
Mouse1, 200 0x08ff1630 +attack

All eleven capture binding samples and all retained held samples are SNAPSHOT; 82 samples retain the same owning memory identity. Canonical seven EXE windows, process/endpoint/Worker/module/renderer/canvas/memory checks stay mandatory. Worker and renderer module hashes both match CPUa320. The passive reader uses the exact owning shared buffer and sparse PTEs; it makes no RPC/export call, guest write, rebind or mapping change. Existing strict startup observation and trace setup are unchanged and separate from this passive reader.

Eleven numbered commands comprise nine native holds, one loading-only capture and the final no-input quit. The holds include one forward/Control combination. All nine have balanced physical UP acknowledgments and successful replies. The Enter fire hold lasted 1,501ms physically, with eleven wholly-held passive samples retaining [13,0,234314,0,1]; the post-release sample is [0,0,234314,1474,4]. These are latch point samples, not event-consumption, usercmd or weapon proof. The separately inspected projectile and destroyed-pane PNGs establish the visible weapon/world effect. No guard damage or kill is claimed. Native source lifetime remains REVOKED; all-writers and edict ABA remain UNKNOWN.

Preparation and fixed identities

Git source base is c0498e0d7a8eecea56bd2272583424787b313d50 on main. Reader and focused test bytes match published 6e18a9865db6764b10e6f0becc1b9b951a4dac79; composition and caller/shared caller match 1473c5952a52d5d2b4546b93a0d462b64fe5c6cd. Custody is staged directly from published 4341b87723a7d56bc98223215cca4cdc878fcae2, without editing shared custody or Warcraft/Super providers. CPU remains 2,185,596 bytes, SHA256 a320d4c85156644fd12b6134d66b11784f3100b9b135756206c087cdacd43295. Both all-caps flags, strict lineage, pre-CPU/pre-DllMain zero-entry barriers, and work/ACK/close/hold/freshness caps are unchanged. Actual startup receipts retain zero execution entries and trace [1,0x400000,0x500000] before release.

Before allocation, verified all461 retained0505 evidence rows by size/SHA, including46 gzip CRC checks. Nineteen prior test/build log references were SHA/size checked for reuse, including exact0505 focused, oldFAIL/newPASS, eleven composed Node scenarios and locked final build. The new722-provider actual compiled production preflight passed without Chrome or an original; the sole changed existing provider versus0450 is the published sparse reader. Local/remote manifest barriers match SHA256 ba38b6c45392f779efe6863b902af6365fce1260a00b802dd64a9fe14d6deff3.

The actual prepared controller also passed fresh offline open-command/ACK/EOF and EOF-pending lifecycles through the production caller and prepared launch/request branches, using explicit local interpreter fixtures. It starts concurrent drains, awaits OWNED_INPUT_READY, flushes explicit start on one retained PIPE, and uses that same writer for commands. Actual waits and each stream EOF are retained. Launch-started proves transport spawn only; separate actual strict startup/ready evidence establishes this run's launch.

Recovery uses the Git base plus closure-recovery.patch.gz (84,579 uncompressed bytes) and a small own-input/note patch. Media/modules are referenced by existing path, size and SHA, without source-tree archives or duplicate media/modules. No production code changed. No agents/workers, local browser, peer-box action, RAM gate or public deployment was used. Allocation, runtime and initial final validation were above the1.5GB free-/home floor. After archive, shared filesystem growth briefly took free space below that floor. Before further work, losslessly compressed79 own-run JSON files, verifying original SHA/size, round-trip bytes and gzip CRC before removing their raw copies; this recovered1,994,960 bytes. R/lossless-storage-index.json maps retained receipt paths to the gzip storage. No prior run, session or unique evidence was deleted.

First failure, evidence and lifecycle

The first actual runtime failure is the final quit at 05:26:18.839Z, command10: abnormal browser shutdown, UNKNOWN, no replay. Browser.close ACK needed SIGTERM; actual browser exit was0 with both stream EOFs after13,493ms under the unchanged19,875ms nested cap, normal=false. Quit has no successful numbered reply. No later input or capture occurred. This remains a shutdown failure; its historical causes are UNKNOWN. The local caller actually waited1 with distinct stdout/stderr EOFs. Its existing revocation hook cancelled the owned supervisor without gameplay or another capture.

Kernel custody has320 owned identities and320 actual reaps, verified against 3,229 journal events and before-reap witnesses. Drain is complete, command EOF true; remaining/unreaped/ptrace/lineage/resume unresolved sets are empty. Remote supervisor and caller-parent receipts separately retain actual wait1 and both stream EOFs, still recording remoteCustody UNKNOWN. Tracked identities are absent; ports39561/40803, input writer, profile FD writers and runtime lock are drained. The unused profile directory remained before archive, with no writers. Local lifecycle/watchdog actual waits are0 with separate EOFs; boardwatch actually waited-15 with both EOFs. The lifecycle's outer tool return was143; that tool result does not independently prove stream EOF or replace the explicit inner wait receipts. Missing or historical proof is not upgraded.

Exact WC native-current-objective-0512 ARCHIVED PHYSICAL_DRAINED release at 05:19:58Z preceded allocation; WC itself waited for Super0512. An independently observable immutable25-minute watchdog was prearmed before the finite-TTL1495 box. UUID 1b38d1f7-aa3d-4bad-a17f-280dfc235618, box bx_6tbumabf, initial expiry 2026-11-09T05:45:12.457Z. All177 regular remote artifact rows, 6,034,558 row bytes (107 newly stored unique contents), were retrieved and independently verified by SHA/size and GNU gzip CRC/decompression before delegated normal archive. Final state ARCHIVED, stopped05:27:14.480Z; receipt stop_cba893c72832a8f4, completed05:27:14.511Z, forced=false, error=null. Final archiveAfter=null; expiresAt was not returned. PHYSICAL_DRAINED was posted promptly at05:27:38Z.

Next would be a separately assigned fresh guard/door progression attempt. This single task stops after authenticated note publication on main.