Stronghold original after the ReadFile retry repair
TG150-STRONGHOLD-CRUSADER-20261103 uses published
500de599fa9e1839d090404f713f908fbe9b98ab for one bounded ordinary original
demo observation. This continues original progression. Setup, world and visible
hostile engagement are not established by the preparation checks below.
Evidence is scratch/runs/20261105-stronghold-readfile-repaired-gameplay/.
Source-only preparation began against newer main 39554ba54; shared ownership
began against a239eaece3f109b2322bcd32b3274bc72681ec16 only after formats1430
FINAL 15:11:01Z, ReturnFire1435 FINAL 15:27:01Z and Lords1450 FINAL 15:35:30Z,
each releasing runtime and files with ALLBUILDSFINISHED. The independent gate
verified 179 recorded PIDs absent, 15 profiles absent, 24 refused ports, four
free locks, no heavy runtime and over 9.3 GB free disk. The first scan checked
12 profiles; adding predecessor release-gate records supplied the full 15.
Lords1450's actual Chrome/probe exit status remains unknown; absence supplies
no normal-exit evidence.
Only tools/stronghold-local-runtime.cjs and its existing composed-entry test
change: named assignment, ordered predecessor releases, task cutoff and exact
published dependency/module bindings. Native coverage, first-failure retention,
image/thread/epoch identity, input/reply/review and presentation guards remain
unchanged. The browser still has a twelve-minute finite lifetime and a
sixty-second bootstrap limit. Historical accepted pins stay rejected.
The published ReadFile modules are unchanged, SHA-256
10462bd9693cf4f51d382d33767674b736b84e74346f328306d59e35c5e60031
and compatibility
eb59d32227cc18233f93c7b757e0700f9558f0a31af7e3d38090669d4fc15158.
Inspection of 299 source files found only the two published renderer changes
from ddc9e0e51136132e8c00725756ad91d45f6ad6ab since the ReadFile repair.
Because host-imports imports the shader module, the 52 ReadFile and 14
first-failure cases were rerun once on current dependencies: all passed.
Unchanged causal baseline and census controls are referenced by path/hash.
The fresh actual composed-entry check passed 2429 filesystem and HTTP rows,
48 negative controls, actual input-command/transport checks and server closure,
with zero Chrome or original launches. Source/module/media mutation controls
remain enabled. flock /tmp/wine-build.lock bash tools/build.sh passed
at 15:37:48Z; both output module hashes remain the exact published repair.
The single original outcome is recorded below.
Canonical media is checked in place, without copies:
test/binaries/candidates/stronghold-crusader-demo/installed/App_Executables/Stronghold Crusader multiplayer demo.exe,
SHA-256 9024c533095f2227b0105b7e8a9e4cf28ec4fb206b165629787b2c97cf99f13f.
Git SHA and relevant binary diff/new-file patches provide source recovery;
there are no source snapshots or media archives. No leased box or subworker
is used; box ID/state/expiry are N/A and lifecycle remains with the coordinator.
No public deployment is authorized.
The runtime binding and focused test were published on main as
7b383873a3bb9ebc8420e98c1bb1458d52b171d0; strict pinned-SSH pull/push passed.
The one original attempt began at 15:38:39.643Z against that exact commit.
Personally inspected fresh images show startup/loading art, the title screen,
main menu, Choose Type and Custom Game hover. Every requested live review was
answered before its deadline. Unrequested fresh boundary images were also
inspected; the final boundary-5 image received its explicit viewed receipt
after terminal failure. No Custom Game DOWN was delivered: the final physical
input was the hover move at 15:43:28.119Z. Setup, world and ordinary visible
hostile engagement remain unestablished.
Command 11 failed closed at 15:44:13.347Z with
Stronghold continuous profile/image/epoch coverage unavailable. The owning
worker had been instrumented before instantiation (thread 1, slot 0, endpoint 1,
profile 2, actual image 0x400000, epoch 965192883). The immutable first
failure occurred earlier, during startup at 15:39:30.060Z, before block
observation was attached. Reports 0, 4 and 6 retain the identical complete
record; zero native blocks were recorded. Later host annotations are not guest
execution evidence.
The first record has reason 5, predicate site 0, depth 1, previous reason 0,
callback entries/returns 6/5, EIP 0x4aa829, ESP 0x074ffe58, process epoch 1
and invocation root 2. Its top frame is kind 1, target 0x4a9810, saved return
0x07500590, entry ESP 0x074ffebc, parent EIP 0x4672a7, parent ESP
0x074ffee0, with zero tokens. Current thunk EIP is 0x07500290.
first-failure.json retains all 36 words and the actual native caller offsets;
first-failure-retention.json preserves equality and final physical input.
This differs from the prior generated false-kind-1 ReadFile park signature.
It does not prove a causal emulator mismatch or authorize guessing a repair.
An offline source-index decoder failed its binary-export check (both initial
and import-hoisted order); combined WAT is not the compiler input. Neither
unvalidated mapping is used, and no diagnostic rebuild or new original ran.
The first concrete command failure triggered normal browser closure; Chrome actually exited code 0, signal null at 15:44:13.570Z, and the finite caller exited code 0. That caller status describes orderly teardown, not gameplay success. No second original, forced readiness, epoch change, frame pop or fallback ran. Post-observation hashing verifies all 2429 sealed source/module/media rows unchanged. Cleanup independently verifies seven recorded process identities absent, the owned profile removed, HTTP 34287 and CDP 41249 refused, four locks free, no heavy runtime, no owned watcher, and over 9.2 GB free disk. No box was allocated or changed; ID/state/expiry remain N/A.
The concrete blocker is the newly retained startup coverage gap above. Next work should explain its native caller and callback lifecycle from these retained bytes before any separately authorized original observation. No emulator change is made on this evidence. Existing gameplay/coverage guards remain intact.