Experiment formal console source provenance results 01 - kmosoti/FabricO11y GitHub Wiki
Console product-source provenance: finite engineering results
Recorded 2026-10-18. Alpha acceptance remains owner-paused. These are bounded engineering controls and registered measurements, not browser acceptance, release qualification or an optimality certificate. The final local fast profile, both receipt validators and documentation checks subsequently passed at 1f6a33215429b9ee3d7a9479f9abb2ea3462fc59; current-head required hosted checks remain pending (see final integration below). No product package was rebuilt or refrozen; independent Python delivery, query and rate oracles are unchanged.
Problem and retained source authorities
Chrome baseline-02 at harness e08caaaa5d48106030b03bb08b6ec2d03f585f5d stopped before browser launch: the console stager compared a frozen product receipt against the newer harness checkout's Cargo.lock. The receipt pin is 5fd251faa4c5145a4813bdb8e3c8a56812d5e52887c8e040101440aa891d0c13; product checkout e078ba23b543126011a2136dd1f7e98ca65370eb retains Cargo.lock 1bec3939e821e84d60dd4a5b532e6f2acabaee309ee46aee68c5f16532c2496e. Harness Cargo.lock was cf5160aab5c004c5b98c1f9bd8a7d28699ad01b89790fb7c3f2ebe830a61efcd. The failed browser attempt and its stop/failure evidence remain intact; no browser result is credited.
The prospective protocol preceded implementation and registers the controls and measurement boundaries. The Rust primitive takes independently pinned product root, exact commit, build root and receipt digest. It retains bounded full-byte hashing, descriptor-relative no-follow file admission, tracked-source checks, bounded Git subprocesses and before/after source identities. The receipt's source field is not authority. Default development staging still checks its current checkout and refuses stale builds.
The legacy adapter requires a caller-pinned native helper snapshot and owned effects callback for explicit packaged mode. It retains asset inventory, PNG and 8MiB budget checks, bounds the receipt before parsing/helper invocation and each asset before verification, and writes the retained verified bytes without reopening input assets. The browser uses its existing owned process boundary; no new Python process supervisor was introduced.
Observed progression and failures retained
All paths below are relative to the mounted DATA root /run/media/kmosoti/data/FabricO11y. Commands, raw logs, per-run resource receipts, source/driver hashes and actual stop states are archived under results/console-source-controls-01. Passing subsets retain their own revision attribution.
| Record | Source | Actual outcome | Final whole-stop UTC |
|---|---|---|---|
| X310 existing stager RED | dae9270001613cdb883be8c6413db4d590bed341 |
exit1; exact different source: Cargo.lock; destination absent | 2026-10-18T05:59:36.092072+00:00 |
| X311 primitive controls | b30430e5436614de55e5685a75baab5236294b39 |
exit0;9 passed | 2026-10-18T06:02:30.117221+00:00 |
| X312 skipped-source-hash mutant | ab7fef00237eb9a1cacc8a20e2c14ebcc599b922 |
exit101; exactly1 named corruption assertion failed,8 filtered | 2026-10-18T06:03:42.750820+00:00 |
| X313 diagnostic debug measurement | 607e041de2b4b93ebeae3dcad72ee413375084c7 |
exit0;35 observations0; not optimized comparison | 2026-10-18T06:05:53.270989+00:00 |
| X314 optimized repeat | 709fdaf5713108cf6f19648db1f202170e35c370 |
exit0;35 observations0 | 2026-10-18T06:07:42.368669+00:00 |
| X315 adapter fixture failures | 87a60a7d6650f423a4ba9bf9fdc0c41e7555348b |
exit101; primitive9 passed; adapter3 passed2 failed; subsequent gates not run | 2026-10-18T06:13:52.355020+00:00 |
| X316 focused subset | babd209245980b9bd41016cb5958ec2a4c2b804b |
all focused commands0;9 primitive+6 adapter+1 health;18 packaging+18 bridge+74 UI | 2026-10-18T06:18:08.615763+00:00 |
| X317 two defect regressions | d27005875005e107829899307c2a1620a8eb9f51 |
exit101; exactly6 passed2 named failures | 2026-10-18T06:21:40.056650+00:00 |
| X318 corrected focused controls | 76d9f2bcbaec6d0248ed74b58fb1641eb435a55f |
all focused commands0;9 primitive+8 adapter+1 health;18 packaging+18 bridge+74 UI;Clippy0 | 2026-10-18T06:24:44.592400+00:00 |
X312 changes exactly one source-hash read to a no-op. The existing control commits corrupt product source and pins that new clean commit, so the clean-source guard cannot mask the omitted receipt digest comparison. Compilation succeeded; committed_source_corruption_is_rejected_by_the_receipt_hash_control failed at assert!(!rejected.status.success()), detecting acceptance of corrupted source. Its isolated worktree, private target and failure fixture scratch/console-source-979603-0 remain retained. An ordinary failed exit101 remains a failed test run; only this named outcome supports defect-detection credit.
X315 preserved two fixture problems: missing browser-module import path and an obsolete expected diagnostic string. They were corrected in a separate test-only commit; no rejection outcome was weakened. The original failed fixture and log remain archived.
X317 then independently exposed two production defects: actual browser identity AST raised KeyError: console_source_helper; actual staging copied bytes changed after verification rather than the verified body. Both named regressions failed, other6 passed, and compilation succeeded. Fixtures scratch/console-source-adapter-988884-12 and -13 remain retained. Two small production corrections changed final identity roles to the recorded immutable snapshot map and staged only retained verified asset bytes. The unchanged regressions pass in X318.
The executed adapter tests require the named wrong root/commit/helper and receipt pin, forged helper-report identity, receipt race, oversized receipt, missing/unlisted/modified/symlinked asset, PNG metadata, boolean size, declared aggregate budget and sparse oversized-asset refusals. They also verify default stale-source refusal and successful explicitly pinned staging. Child diagnostic assertions execute inside the Rust tests; successful child outputs are not individually published as independent acceptance receipts. The callback process-ownership control is source/AST coverage only; fixture adapters execute an injected callback and actual native CLI. No actual packaged browser or real owned callback lifecycle was exercised by these controls.
Registered measurement
X313's debug result is preserved as diagnostic: median Python/Rust wall milliseconds were25.410781/16.283152 for17 sources and32.801081/72.524082 for512 sources. It is not adoption performance evidence against optimized hashlib.
Before X314, a separate prospective protocol amendment required optimized Rust, changing only build profile and fresh output. Source implementation, corpus construction, host, seven alternating pairs and measurement method stayed fixed. git diff confirms the measured Rust verifier, shared file reader and measurement example bytes at709fdaf equal those at current76d9f2b; later adapter fixes are outside this hash-loop comparison.
| Warm corpus | Files / source bytes | Python median wall ms | Release Rust median wall ms | Python / Rust maximum RSS KiB |
|---|---|---|---|---|
| Canonical product | 17 /268836 | 23.466608 | 3.247566 | 15980 /3528 |
| Deterministic synthetic | 512 /1048576 | 32.792373 | 20.129424 | 16164 /4720 |
Seven additional complete native canonical CLI observations include Git/provenance work: median19.855039ms, maximum7492KiB RSS. These are reported separately, not compared to a Python hash-only loop. GNU time median CPU seconds were canonical Python0.01/Rust0.00, synthetic0.02/0.01, complete CLI0.01; its coarse resolution cannot establish zero CPU cost. All35 subprocess observations exited0 with matching counts/digests. Process startup is included; corpus construction, compilation and driver aggregation are excluded from paired durations. Filesystem metadata is warm and these are small fixtures; Rust also applies stricter file-admission semantics. No whole-browser, full-profile, GPU or universal performance conclusion follows. No further provider/backend/parameter tuning was run.
Hardware/tool/version observations are archived in results/console-source-controls-01/x313-machine-input-pins-01.json. Both debug and release ELFs were copied to immutable mode0500 archives with SHA-256 manifests before any overwrite. X314 records all17 actual product file hashes, which matched the same original receipt before and after execution. Raw35 argv/stdout/stderr/GNU-time/exit observations are in the measurement JSON; each pair alternates Python/Rust order.
Actual commands and custody
Common contained source-driver invocations preserve exact inner argv in their JSON. Every project command ran under python3 tools/resource_group.py -- ..., default16GiB high/20GiB max, no swap,1800s deadline, with strict whole-DATA namespace census and full reservations. Formatter1GiB; targeted tests/build/Clippy2GiB; measurement1GiB. Rust used the retained1.99 toolchain explicitly and private inner DATA targets; mutant and production targets are separate. No uncapped or shared mutant-target fallback occurred.
Actual command families:
python3 tools/resource_group.py -- python3 -B tools/packaging/stage_console.py DATA/results/console-replacement-tls-01-a DATA/results/console-source-controls-01/red-console-stage
cargo test -p xtask --test console_source -- --test-threads=1
cargo test -p xtask --test console_source committed_source_corruption_is_rejected_by_the_receipt_hash_control -- --exact --test-threads=1
cargo build --offline --release -p xtask --bin xtask --example console_source_measure
TARGET/release/examples/console_source_measure PRODUCT_ROOT BUILD_RECEIPT e078ba23b543126011a2136dd1f7e98ca65370eb TARGET/release/xtask NEW_OUTPUT
cargo test --offline -p xtask --test console_source_adapter -- --test-threads=1
cargo test --offline -p xtask --test console_source --test console_source_adapter --test live_browser_health -- --test-threads=1
python3 -B tools/packaging/test_packaging.py
env PYTHONPATH=tools/qualification:tools/ci:tools/ui python3 -B -m unittest -v test_console_bridge test_console_bridge_smoke test_release_runtime
env PYTHONPATH=tools/ui:tools/qualification:tools python3 -B -m unittest discover -v -s tools/ui -p test_*.py
cargo clippy --offline -p xtask --all-targets --all-features -- -D warnings
DATA expands to the mounted root above; TARGET is DATA/scratch/console-source-measure-target-01, PRODUCT_ROOT is DATA/scratch/candidate-build-replacement-tls-01-a, BUILD_RECEIPT is DATA/results/console-replacement-tls-01-a/build.json, and optimized NEW_OUTPUT is DATA/results/console-source-measurements-01/run-release-01. Exact absolute argv and environment pins are in each plan/outer record, including the debug-profile command. Format mutations were separately recorded, reviewed and committed before clean-source tests; they are not test passes. The first X311 source guard refused an earlier source pin before workload after a protocol commit crossed setup; that refusal is retained without workload credit.
Successful fixture scratch was removed by existing cleanup paths. Failed fixtures, failed outer evidence, original source trees and isolated mutant target remain intentionally preserved. Every launched run above has a confirmed whole-service stop; no project unit remains active. Full fast, current-head required hosted checks and alpha acceptance remain pending. This page closes only this finite reporting checkpoint.
Primary aggregate hashes (DATA-local evidence archive):
| Record | SHA-256 |
|---|---|
results/console-source-controls-01/red-stager-actual-01.json |
c450363b289e0448bdba332245feb3f4b46a70ee760f9cde14db566c65ba504a |
results/console-source-controls-01/x311-actual-results-01.json |
b734403cd6ac2c05bebb20d8fb8677ca9160e4f157f34629d4ea3a4489682edf |
results/console-source-controls-01/x312-actual-results-01.json |
8702ca23daa4c59e89889629734d9a71f79236563470a193fa9e94ea6fe5e975 |
results/console-source-controls-01/x313-actual-results-01.json |
fe07f998305bc118900454af5856168105e7d005f7b84d2977fff4db38f33298 |
results/console-source-controls-01/x314-actual-results-01.json |
62be7d654c009847e0b2937cbb29ddabde342458f9e3bc0490330228c5d4fe6b |
results/console-source-controls-01/x315-actual-results-01.json |
dcc4009ce6b1b707694f7a5545502cb4744b94b9bcbbfe575f7d91c376a852f0 |
results/console-source-controls-01/x316-actual-results-01.json |
71bfc2beb4ad758ab63eb22802b7ec356888bb90397dcdaead66a5b66586f239 |
results/console-source-controls-01/x317-actual-results-01.json |
68ab05d7ec544d5857d1cbcf551be2672488e2f3938d104284a6615a900c30b3 |
results/console-source-controls-01/x318-actual-results-01.json |
59990d4b0b4146fc69d8c55daaca1b266eb3a12cc9444410c030fdc55fad0e45 |
results/console-source-measurements-01/run-01/measurement.json |
82653559a9975398fb03a5a97c5a53cbd09198c167bdf462d7af5304a75381c2 |
results/console-source-measurements-01/run-release-01/measurement.json |
ab08591396d5946e1f92181e768bbee3929a7b266a9820b87c3f2f96ee1ee2bc |
Final local integration: interruption, custody and bounded build
The initial X319 full-fast attempt at c5295a65a020195b4eb867747a98f938aac81e30 was interrupted2, not passed. Its strict periodic census reached96,394,987,761 bytes during WASM compilation after18 stdout-reported passing gates. The original12GiB reservation had been admitted from81,445,731,031 bytes; it did not guarantee the actual footprint. The existing guard stopped its exact owned service, whole-stop confirmed at2026-10-18T06:36:48.824534Z. The incomplete canonical marker and79 staged raw files (218328 bytes) remain archived and readback-hashed. Validators/docs were not run on that interrupted prefix, and no prefix was credited as a complete profile.
Root reviewed exact cache-only cleanup plans before deletion. Nine completed private targets were removed only after86 unique linked ELF bodies had been archived and each decompressed blob readback-hashed. Then the interrupted private production target's232 linked paths/164 unique executable/shared-object ELF bodies were archived and readback-verified; checker and all console regression binaries are included. The checked c529 source archive matched Git objects on readback. Raw receipts/logs, meaningful failed fixtures, all frozen0500 artifacts, frozen product/package sources, shared DATA/cargo and acceptance evidence were excluded. Relocatable object files, rlib/rmeta and incremental compiler state were disposable cache. Phase1 cleanup exited0/whole-stop2026-10-18T06:45:13.156027Z; phase2 exited0/whole-stop2026-10-18T06:51:12.929665Z. The post-cleanup strict namespace census was79,627,378,033 bytes, UID/GID0; no storage relocation or threshold relaxation occurred.
A separate prospective policy commit preceded the X320 retry. A fresh private DATA/scratch/console-source-bounded-target-01 used local dev/test line-table debug information and no incremental storage, explicitly retaining codegen256, opt0, debug assertions and overflow checks. Release/product settings and every registry command/CI configuration remained unchanged. The purpose was a bounded resource correction, not another algorithm/backend experiment or a speedup claim. Cargo's documented line-table mode keeps filename/line backtraces but omits variable/parameter debug info. Raw compiler invocations show codegen-units=256, debuginfo=line-tables-only and no incremental flag; Cargo omits rustc flags for retained default opt0/assertions-on/overflow-on, while the explicit profile environment is archived. Build-script/dependency overrides remain separately visible in the raw logs.
The exact new environment, inside the launcher, was:
RUSTUP_TOOLCHAIN=1.99.0
CARGO_TARGET_DIR=DATA/scratch/console-source-bounded-target-01
CARGO_PROFILE_DEV_DEBUG=line-tables-only
CARGO_PROFILE_TEST_DEBUG=line-tables-only
CARGO_INCREMENTAL=0
CARGO_PROFILE_DEV_CODEGEN_UNITS=256
CARGO_PROFILE_TEST_CODEGEN_UNITS=256
CARGO_PROFILE_DEV_OPT_LEVEL=0
CARGO_PROFILE_TEST_OPT_LEVEL=0
CARGO_PROFILE_DEV_DEBUG_ASSERTIONS=true
CARGO_PROFILE_TEST_DEBUG_ASSERTIONS=true
CARGO_PROFILE_DEV_OVERFLOW_CHECKS=true
CARGO_PROFILE_TEST_OVERFLOW_CHECKS=true
CARGO_TERM_VERBOSE=true
cargo xtask checks --profile fast --light-workers 2 --receipts DATA/results/console-source-full-gates-01/fast-receipts-02
Retained direct Rust1.99 PATH,2 build jobs, default resource limits/noSwap/1800s, full12GiB logical reserve and strict95GB periodic census remained enforced. X320 admitted79,627,402,908+12,884,901,888 bytes. All20 original fast checks actually ran and passed at clean 1f6a33215429b9ee3d7a9479f9abb2ea3462fc59; light cohort maximum2 workers, heavy checks serial. Maximum scheduled fast DATA was86,404,172,134 bytes. The private target was observed at5,835,373,093 apparent bytes after fast; this is an operational observation, not a controlled cache-footprint or runtime speed comparison with the differently accumulated X319 target.
| Current local gate | Actual exit | Whole-stop UTC |
|---|---|---|
console-source-bounded-fast-01 |
0 | 2026-10-18T07:00:49.028778+00:00 |
console-source-bounded-native-verify-01 |
0 | 2026-10-18T07:00:58.974045+00:00 |
console-source-bounded-python-verify-01 |
0 | 2026-10-18T07:01:03.978673+00:00 |
console-source-bounded-docs-deps-01 |
0 | 2026-10-18T07:01:08.953722+00:00 |
console-source-bounded-docs-links-01 |
0 | 2026-10-18T07:01:14.866009+00:00 |
console-source-bounded-docs-checker-01 |
0 | 2026-10-18T07:01:38.741181+00:00 |
console-source-bounded-docs-hooks-01 |
0 | 2026-10-18T07:01:44.758083+00:00 |
After fast0, the native command was cargo run --locked -p xtask --features verification-sha-openssl -- verify-receipts --profile fast --receipts DATA/results/console-source-full-gates-01/fast-receipts-02 under the same local build environment. The unchanged independent Python command was python3 -B tools/ci/verify_receipts.py --profile fast --receipts DATA/results/console-source-full-gates-01/fast-receipts-02. Both completed0 on the same20 current-source receipt/log identities. No running marker remained.
Documentation then ran the frozen Bun install with DATA cache followed by bun tools/docs/check.mjs, bun tools/docs/check.test.mjs (55/55) and python3 -B tools/docs/test_hooks.py (10/10); all0/stopped. Existing wrappers set the pinned Bun path/cache; exact absolute argv appear in the local plan/outer records. Source remained clean before and after every gate; all temporary launcher directories were removed and no project unit remained active at local closeout.
The initial PR82 head was1f6a332. Its later hosted browser failure and the subsequent engineering controls are recorded below; that historical local pass is not a pass at a newer revision. These local results are not browser acceptance or full-soak qualification; alpha remains explicitly owner-paused with no automatic restart after CI/merge. No further optional tuning or 200-program rewrite is a prerequisite.
Additional immutable DATA-local records:
| Record | SHA-256 |
|---|---|
results/console-source-full-gates-01/x319-interrupted-actual-01.json |
d1c3e193ea2349ad7d73d55ca4ae92e145bf798d31e3dbef6d7ea46d6e401079 |
results/console-source-full-gates-01/interrupted-stage-archive-manifest-01.json |
2693cdcd6797eea4fe43b90f9187e6158f8cf5b7c2c7b1327bccc1d5f20015d1 |
results/console-source-full-gates-01/owned-cache-cleanup-plan-01.json |
1f4dbee7a94659df513416e0b4faa368ec8dc40491c51d9294a8554b90981549 |
results/console-source-full-gates-01/owned-cache-custody-01/completed-private-targets-custody-manifest-01.json |
f636521613a060a418e416baace57c475f5490946d89695ac154d4a66fc6c2e7 |
results/console-source-full-gates-01/owned-cache-custody-01/x319-private-target-custody-manifest-01.json |
4c6c6535ed08d38d2b91da2e3443502504700c611dbd6418186a5671e20f2200 |
results/console-source-full-gates-01/phase1-cleanup-outer-01.json |
11b434c1dbb8c8ddf3dd468e82e897a818754405c861704c47c269e4ae47a765 |
results/console-source-full-gates-01/phase2-cleanup-outer-01.json |
f911ee189add3e1d60bcb19a6e3c5ba57e3d002adc38e6a097131f477a4d9c4e |
results/console-source-full-gates-01/phase2-post-strict-census-01.json |
4896b1186104101397dedab14e594649b1664808848815f26db1c95d866c00fa |
results/console-source-full-gates-01/plan-bounded-01.json |
d21da37f187640b91d118b18808fd440105c066e450a331dcd7b6d40c1cca70d |
results/console-source-full-gates-01/effective-compiler-samples-02.json |
af9d02bdbd8097aac1d86ca1bebfe9ebbb06eab1491063297bfa66f6ca156c7e |
results/console-source-full-gates-01/x320-actual-terminal-01.json |
e35264eb71162b46acfda5e2f32f13accfef86ef723f60ef4603a618def7c814 |
Hosted browser failure and native process handoff follow-up
These are engineering observations, not alpha acceptance. The product source/package freeze and independent delivery/query/rate oracles are unchanged. The final source at this reporting checkpoint is 352cafb; its local full profile/docs and current required hosted checks remain pending. Alpha remains owner-paused. No release, endurance qualification, GPU or universal performance claim follows from these results.
X321 archived203 unique linked ELF bodies and the checked source/settings before the root-reviewed removal of the completed private X320 target and its ordinary docs dependency installation. The ELF archive was995,849,506 bytes, SHA b9436ab311b90920720d3b7a0d3820261ccec3af40f12bd9653b4878d5fcfe25; every body was readback-hashed. The post-cleanup strict namespace census was80,632,291,216 bytes. Shared caches, source, all failed fixtures, original results and frozen0500 executables were excluded. This was a historical storage observation, not a guarantee of later admission. The strict census counts hardlinks per path; GNUdu without --count-links can deduplicate those paths and must not be substituted for that guard.
The initial hosted fast job at1f6a332 failed in the existing response-header consumer: it indexed cache_control in a request row with no HTTP response. Of97 raw rows,96 had HTTP responses; one completed POST /v1/console/query had statusnull and no rejection kind recorded. The original failure is not retrospectively classified as a cancellation. Its670,127-byte downloaded artifact ZIP matched GitHub's SHA fac114e0781f08c7d3bf43bd33952dbcca1f9a0d34e96a3e965e173c23173696. That revision's security, both extended shards and final extended aggregate subsequently succeeded, but fast failed, so the revision was not mergeable.
An instrumentation-only packaged Chrome probe atb7c4e47 then reproduced the failure with explicit evidence: index62 was a finished fetch-stage AbortError, same request signal aborted at rejection, statusnull; the other96 rows were HTTP responses. All five owned process groups stopped, but the unchanged header consumer still exited1. This discriminated an actual non-HTTP terminal state from a missing security header. The correction retains every raw row, requires real HTTP rows with valid integer statuses and all original no-store/CSP/COOP/CORP headers, separately records only narrowly witnessed query cancellations, and refuses other rejection/pending/metadata states. The bounded ledger limit remains512. The registered adversarial control uses actual production consumer/fixture-wrapper source; it does not manufacture a new independent product oracle.
| Run | Revision | Actual outcome and preserved limit |
|---|---|---|
| X322 | 3691a809 | Compiled targeted Rust/legacy-AST regression exited101: witnessed cancellation raised the original KeyError; malformed terminal rows and missing diagnostic guards were also exposed. One test failed;8 filtered. This is a named RED, not a setup failure. |
| X323 | 7f0bba85 | Targeted ledger1, primitive9, adapter9, browser-health1, packaging18, cleanup/bridge18 and UI74 tests passed; all-target/all-feature Clippy exited0. Six outer cells stopped by07:31:59.577835Z. |
| X324 | 7f0bba85 | Probe02 exited1 before browser launch because the retained prior fixed scratch directory existed. No JS result or browser credit. |
| X325 | 7f0bba85 | Probe03 exited0/stopped07:36:42.232540Z with363 browser checks,97 HTTP responses and zero cancellation rows. It did not exercise the actual JS cancellation branch. |
| X326 | d82efb61 | The unchanged targeted ledger control passed. Probe04 then exited1 before browser launch: the fast source-verifier process finished before live process binding; inner cleanup also rejected an unregistered child. Only the outer cgroup's confirmed stop is credited. |
| X327 | 416e3cf0โ9d7a8776 | Contained rustfmt exited0; only three Rust files were mechanically formatted and committed separately. Formatting is a mutation, not validation. |
| X328 | 9d7a8776 | Three native unit tests passed. The delegated real Linux fixture exited101 because a completed positive launch left browser-0, which the next attempted launch reused. The append-after-release mutant had not executed; no mutant detection credit. |
| X329 | c7bdebe5 | Monotonic per-domain attempted-launch names removed that collision. The complete delegated Linux fixture, targeted ledger control and Clippy all exited0/stopped08:06:28.214186Z. Its JSON was asserted but not emitted; raw timing metrics are not claimed for this run. |
| X330 | 7508a989 | A test-only reporting replay of the same Linux fixture exited0/stopped08:08:09.914659Z and emitted the complete control report. Native ELF/test/source bodies were archived and readback-verified. |
| X331 | 7508a989 | Packaged engineering Chrome probe05 exited0/stopped08:12:22.337238Z:367 checks,97 HTTP responses plus two separately witnessed native-fetch cancellations, final identity checks and all nine owned groups cleaned. No acceptance credit. |
| X332โX333 | 4e52cbe7โ1caae9d9 | The new CI contract failed exactly for the old workflow's absent delegated native fixture/pinned entry operands (23 other tests passed), then all24 tests passed after the separate workflow integration. |
| X334 | 1caae9d9โ352cafb0 | Hosted fast/remaining/security failed the unformatted metrics print line. A contained rustfmt-only commit changed that line's layout; local format check and24 CI contracts then exited0/stopped08:25:56.754151Z. Earlier runtime observations remain revision-bound. |
Every row's full literal argv, source, storage reservation, stdout/stderr hash, actual exit and resource receipt is in its DATA-local plan/outer record. Project commands used the existing resource launcher; default outer16/20GiB high/max, noSwap and1800s remained enforced. Rust controls used retained Rust1.99, explicit private production-only targets and the already registered line-tables-only/nonincremental dev/test profile. The real Linux fixture separately applied and verified its original stricter3.5/4GB high/max, zeroSwap,200% CPU and512-task limits. No shared target was used across mutant/worktree variants.
Representative literal commands after the launcher/environment prefix were:
cargo test -p xtask --test console_source_adapter browser_fetch_ledger_consumer_separates_http_responses_from_witnessed_query_cancellation -- --exact --nocapture
cargo test -p xtask browser_enter::tests -- --nocapture
cargo test -p xtask --test browser_entry actual_parent_seam_binds_before_releasing_native_short_child -- --ignored --exact --nocapture
cargo clippy -p xtask --all-targets --all-features -- -D warnings
python3 -B tools/ci/test_ci_contracts.py
The ignored real Linux fixture command was explicitly launched with resource_group --delegate. Its observed positive /usr/bin/true exited0. The same release-time guard checked a still-live PID/startticks/pidfd and real subgroup membership before sending RELEASE; the copied append-after-release mutant was rejected by that same guard. Wrong placement, RELEASE EOF/timeout/framing, cleanup fallback and a foreign child were refused, all attempted subgroup paths were unique, and the process domain ended empty. This tests the real parent seam; it is not VM or installation acceptance.
X330's finite report recorded positive interval36,517,989ns and whole fixture260,386,336ns. The262,144-byte cgroup peak is post-entry charged memory, not whole helper RSS. RUSAGE_CHILDREN maximum50,512KiB spans all fixture subprocesses, not an isolated helper. These are one-run observations, not speed/memory comparisons. The current native entry ELF is archived0500 with SHA 7220128d28c8edb609e31d13c0a2b6effa5953e427c78e8a713dd9cef8b1c5dd; the optimized console-source verifier remains the independent existing 3f629a8aโฆ artifact. Raw source archives and actual Cargo compilation invocations bind their provenance; no product package was rebuilt.
Probe05 preserved99 raw fetch rows. Its native Request.signal pre-aborted control observed started43.5ms, abort43.59999996423721ms and finished43.89999997615814ms. Its in-flight init.signal control observed started593.5ms, abort594ms and finished594ms. Both produced fetch-stage AbortError with the same signal aborted at rejection, and both controls were accepted; all97 actual HTTP responses still met the security-header oracle. The nine process bindings recorded pinned native helper identity, PID/startticks, release state and entry interval; all nine final group records were kernel-empty and children-reaped. The final immutable executable map covered server, spindle, console-source verifier and native entry helper. Product-source verification matched17 original files. All successful probe fixtures were removed; no project unit remained active.
The failed01 and04 fixed scratch trees were not deleted: root-reviewed same-filesystem atomic renames moved them intact into their original outer failure evidence directories, preserving device/inode/mode and54,517,445/26,942,016 bytes respectively. Probe02's setup failure and X328's failed native fixture remain preserved. The old1c hosted failures remain failures; any superseded/cancelled job is not a pass on the new head. Current local fullfast/docs, current required hosted results and tested-checkout custody still gate merge. A separate conservative cache-custody/cleanup plan is pending and provides no acceptance credit.
Additional primary records (DATA/results/console-source-full-gates-01/):
| Record | SHA-256 |
|---|---|
pr82-fast-failure-actual-01.json |
8189806fb90e711da18189ff0af936114b9a369b45a7eaa13a9c30f82080e8f5 |
chrome-fetch-diagnostic-actual-01.json |
542e3253ab8a51d86768d5b589a10d6efde733f214f5b71e4591f6fbea3c0bf7 |
fetch-green-terminal-01.json |
26db9ce89d0b8a1fdb0f34c2b95daa1af23bc5e653fac8800fa852a1cf6c0a49 |
chrome-fetch-diagnostic-actual-03.json |
21f86caa3bec75a5aec2da617c9a9baae4b3ff7125771d69d88f27c91ca1bddd |
fetch-composition-actual-01.json |
9589ee6b28149caaf99ad50f2160aea429d9f4dda49e8d6991dc34017895eeb9 |
browser-entry-controls-actual-01.json |
a317a32035fca23ea4416eaedd142484d99a9aa155a133cd08aad1eaf48d9069 |
browser-entry-controls-actual-02.json |
0e7caf4cf1d5ca5036478c7cf0da0812d77fcd4d0031d4af3be2d3528d0f7833 |
browser-entry-real-report-01.json |
96a9365d3cf14ea7c51da0c526f9635f6895599b2813b7ec2b4ae2b00b843bef |
native-entry-artifacts-02/manifest.json |
56eef075b694da758c0d20f3460ce12cff2c30741356e9028432d33c3b638d88 |
native-browser-composition-actual-05.json |
6017c0f764fc6060cf9d24dd7824a86abe17fde5c3cae49da67c3e9671d15cae |
fetch-failure-fixture-relocation-01.json |
f8043c9afe9e452c03c9b903729efc4dbd7b0814aa372b363e6058149d58497a |
fetch-failure-fixture-relocation-04.json |
b635385a3d0d2a752966f3a6532ceba40c50beddcc09cb0892067fcfe9763fe8 |
Final local verification and retained-failure compatibility (2026-10-18)
X335 removed only the 29 explicitly approved completed private targets and per-worktree documentation dependency directories after complete path/body/source custody and archive readback. Shared caches, frozen product builds and executables, source fixtures, all failure evidence and results were excluded. The complete archive is SHA 4718851ddfcaba1a8c69251b8d93f956de328726960600a151c4afc6ad5e88b1; custody manifest 2372eb9cf639a2b59fd415c8b3908be248994c0d78ba571f2e0acb9cb07132d7. Cleanup exited0/stopped08:34:11.571989Z. The subsequent strict namespace census measured77,817,770,532 bytes; no later workload fit was inferred from this earlier observation.
X336 at clean 352cafb09d0b8319fa7793892e3a40db4d014deb admitted a fresh private target with17,000,000,000-byte reservation against77,817,787,667 bytes. All20 original fast checks completed0/stopped08:43:51.165357Z. Native OpenSSL complete receipt verification and the unchanged Python complete validator each exited0, followed by frozen dependencies, documentation check,55 probes and10 hooks. The final stop was08:44:41.371337Z. The same registered line-tables-only/nonincremental dev/test settings, explicit256 codegen units/opt-level0/debug assertions/overflow checks and default containment remained in force; the8GiB+1 sparse refusal fixture was unchanged. This is engineering verification, not browser or soak acceptance.
The hosted fast run at head352 nevertheless failed its API-cache negative detector. Downloaded artifact12130882499 matched the GitHub digest d29ac2afbdfbbb548b1da94bb3e9fb2fff0fb34c5ef415e4f70430f48806d9a3. Its actual tested synthetic checkout fd09ea5ade7e0f8dbdb4b87cf34fc13a65ddd7e8 has parents main 8a8b53421ce76e92eec65698fbce694035b767ed and head352. The negative browser receipt contained the exact intended error, both required browser/passkey witnesses and all four unique kernel-empty/reaped process identities with integer exits, no cleanup error and no remaining owned children. Its only unmet predicate was the detector's exact removed-fixture prose: failed fixtures were intentionally retained. The hosted run remains failed; this diagnosis does not retroactively turn it into a pass.
After a separate prospective protocol correction, X337 at test-only40fba699 exited1/stopped08:47:16.876806Z: exactly test_observed_hosted_defect_receipts(disposition='retained') failed among four tests. X338 at e1769872790be9aa19586c13e6059aee0fd01a3b accepted exactly the removed-success and retained-failure dispositions while preserving every process identity, kernel-empty, reap, exit/error and exact-defect predicate. The same four-test suite,24 CI contracts, detector applied to the unchanged actual hosted negative receipt, documentation check,55 probes and10 hooks all exited0/stopped. Corruptions and unrelated errors remained refusals for both dispositions. No browser or Cargo runtime was changed or rerun for this checker-only delta.
The352โe176 diff consists only of the protocol, checker, checker test and two minimized fixture/origin files. Earlier full compiled/runtime and native/Python complete-validator results remain attributed352; they are not relabeled as executions at e176. Current-head hosted checks and tested-checkout custody still gate merge. Alpha acceptance remains owner-paused.
Primary records in DATA/results/console-source-full-gates-01/:
| Record | SHA-256 |
|---|---|
final-local-terminal-01.json |
7fcd8499f181e25e4fde6f0ce67f219b125174f63fb981d332bd77a6d71c44ee |
pr82-third-fast-cleanup-mismatch-actual-01.json |
af588c020b39a38611a7eadc6680480adc11f838bd86fc4c8411abd689911598 |
pr82-third-fast-negative-browser-01.json |
b12ff444cac0d4f083f0d9d1b411485cbba781d38930b97fd391dacf7e392e37 |
cleanup-disposition-terminal-01.json |
2749d2a5082553c967967f842fa71bff001cb50cdea1add7f6109c8d09196d2c |
The last focused local stop was 2026-10-18T08:49:08.326043+00:00. No end-to-end speedup, qualification or acceptance-completion claim follows from these finite results.
Engineering integration completed; alpha resumed separately
The current PR head e1769872790be9aa19586c13e6059aee0fd01a3b completed all five required hosted checks: fast13m27s, security1m00s, semantic shard25m43s, remaining shard26m25s and aggregate48s. The fast artifacts contain20 successful receipts with matching output hashes at tested synthetic checkout 5d490494acb3f21aa2daf7512f8d116b19b2dfa9. Its native handoff fixture completed, actual Chrome positive recorded354 checks and clean process shutdown, and the API-cache negative recorded87 preceding checks before the exact intended failure and successful detector grade. Both browser records retain four kernel-empty/reaped groups and no owned children. These CI observations are not frozen-package acceptance.
The extended aggregate retained exactly the ten original producer receipt/log byte pairs (one semantic and nine remaining), all from that same synthetic checkout. Every producer's original contained outer exited0 and confirmed stop. The Rust collector and unchanged Python complete-profile validator also exited0/stopped. Artifact ZIP hashes were checked against GitHub metadata before inspection. Ordinary match-head merge of PR82 produced main 1ff56e6b79a03dc6ca7ef50498af988db95dd5e3 at09:22:16Z, with the tested base/head parents and exactly the tested tree b9288625925339dd1b54a69588465b647caa2c2a. No release or tag was created.
After full cache-body/source/settings/receipt custody and byte/type/mode/link readback, X340 removed only the two approved completed local target/dependency directories. Actual exit0/stop08:57:35.134643Z and strict all-UID DATA census79,184,952,556 bytes were recorded. The complete preserved archive is1,345,783,203 bytes, SHA 05e4d58f97d38f5d6caa92b6ad6bbe22c9099f2e3d23b0981da6d8a4233ae813; custody manifest 0020fe54cab93d24c16909be2f810f977abdd968e0c7d14849c3beed0c8ce7f1. Shared caches, source, product builds and all failed fixtures/results/frozen executables were excluded. Future reservations still require fresh actual censuses.
This completes the bounded engineering prerequisite, without a universal optimization claim or a requirement to rewrite every inventoried program. The owner/root then explicitly resumed alpha acceptance. The separate original70-cell browser union was rebound mechanically to clean main1ff, with seven explicit pinned operand pairs on68 live vectors, unchanged stage/extract vectors and a fresh Chrome baseline-03 output. Stage0 is reused; all engineering probes remain outside acceptance. The first new acceptance cell was running at this publication checkpoint, not yet graded. Its original gates, controlled-clock limits, real expiry durations, process cleanup and independent oracles remain mandatory. #43/#46 and the actual7,200-second FULL soak remain incomplete; no acceptance-completion claim follows from this engineering report.
Additional primary records in DATA/results/console-source-full-gates-01/:
| Record | SHA-256 |
|---|---|
post-final-cache-custody-01/custody-manifest-01.json |
0020fe54cab93d24c16909be2f810f977abdd968e0c7d14849c3beed0c8ce7f1 |
post-final-cache-cleanup-terminal-01.json |
92ff48f7af39912527291080e3265ee226f6ad7821f49ecf29d88a727c5b38c6 |
pr82-fourth-fast-custody-actual-01.json |
d3a288a26ac0fa59a268c5dcda1b25901779b8d1ca27649df6b4035455ae8f76 |
pr82-fourth-producer-shards-custody-01.json |
e3db72da02ca853df23e7d3a3178006ac8ed575a2b0843134518ebf7f93b54cd |
pr82-merge-ready-custody-01.json |
a78a7e862e1effabc18472fbcf2cf943ddc1821f7bd089799f5958098ff76fd5 |
pr82-merged-tree-and-resume-review-01.json |
ff6973eec6e4d57b945b2ed5c4bc565664d14dbc598834458f7e35b02327debd |
Native BiDi viewport and confirmation controls
The separate exact-package Chrome baseline at main 1ff56e6 completed 367 checks. The next Firefox baseline failed its existing injected pagination-overflow assertion: the CSS mutation applied, but Firefox 157.0.1 returned an actual 500-CSS-pixel viewport after a classic 390-pixel window request. The mutated button's right edge was approximately 489.47 pixels, so the unchanged overflow predicate correctly remained false. This is a failed harness stimulus, not evidence that the product violated the viewport contract. The original stopped receipt and 130,098,017-byte failed fixture remain preserved.
A prospectively registered calibration used Bun's native WebSocket implementation as an independent experimental client, without product server or authentication. Classic outer390ร844 produced outer500ร844/inner500ร758; BiDi390ร844 produced Classic JavaScript inner390ร844 at DPR1. CSS zoom200% and font18px were retained. viewport:null restored inner500, and classic desktop1440ร1100 restored inner1440ร1014. Window-handle/root-context equality was observed, not assumed.
The bounded Rust xtask browser-bidi-viewport --ws-url LOOPBACK_URL --context CLASSIC_HANDLE --mode narrow|restore primitive uses pinned Tungstenite0.30.0, one absolute five-second deadline, 64KiB frame/message bounds, a cumulative2MiB input budget and strict context/JSON checks. Its ACK explicitly does not verify geometry. Twelve focused controls and scoped Clippy exited0 at ba323f0fcb96b14c981ae04427d48117659a3eb3; earlier E0639 compilation and test-only Clippy failures remain failed, with no tests credited to compilation failure. The archived executable is SHA 4335ad0166025dc48fce82bc244075b312dfdcaa02d2d0e74f6ca8c3ae61d38a (mode0500). Its actual ignored Linux entry fixture also exited0 at 12e36cc, using that exact executable and the original stricter delegated memory/CPU/task/no-swap limits.
The independent Bun/Rust comparison completed0/stopped10:13:52.391664Z. Bun measured390 and null restoration, then closed its socket before Rust connected. Classic WebDriver JavaScript independently measured Rust390 and restoration to500 with unchanged DPR and zoom/font settings. A valid-shaped no-op ACK stub exited0 but left500 and was rejected by the actual390 geometry oracle. All three child exits and two owned groups were recorded and cleaned. No transport or timing performance ranking follows from this comparison.
The thin caller keeps the existing pagination font, zoom and overflow predicate. It also records the console-source verifier in the existing process-exit inventory: the original Chrome receipt had nine stopped groups but only eight listed exits. This producer correction leaves the strict negative grader unchanged. Early structural fixtures failed because of incomplete AST traversal, missing injected namespace and missing closure scope; those failures remain recorded. Source-derived controls include the missing-append mutant, repeated restoration with unique launch labels and preservation of an earlier cleanup error.
Two actual packaged positive attempts then found distinct harness defects. At f936097, missing nonlocal viewport_websocket_url left the outer endpoint empty; exit1/stopped10:30:07.186996Z retained the sticky restoration error and five cleaned groups. At 5b8ce3c, viewport390 and the unchanged injected overflow rejection succeeded, but the existing explicit POST /alert/accept returned404 after reader revocation; exit1/stopped10:37:51.863297Z retained seven cleaned groups. Neither attempt is credited as a successful packaged trial.
A separate blank-page causal probe completed0/stopped10:48:17.991361Z: three sessions, five real confirmations and four native viewport calls. Classic-default preserved the exact confirmation for explicit acceptance. BiDi-default dismissed it both before any WebSocket connection and after the native pair closed; unhandledPromptBehavior: {"confirm":"ignore"} preserved it for the same explicit accept before and after. No event subscription, auto-accept or fallback was used. The Firefox-only capability correction and its omitted/auto-accept mutants are recorded in source31e27b9. Raw capabilities/session URLs remain private DATA evidence.
At clean31e27b9, two adapter controls, scoped Clippy, 18 bridge/cleanup controls and74 UI controls all exited0. The final two packaged Firefox engineering diagnostics used the unchanged frozen product source e078ba23b543126011a2136dd1f7e98ca65370eb, console receipt 5fd251faa4c5145a4813bdb8e3c8a56812d5e52887c8e040101440aa891d0c13, provenance helper3f629a8a and native helper4335ad01. The original baseline retained --recovery only; the negative retained --inject-api-cache-defect.
| Actual command/cell | Exit and whole-stop | Observation |
|---|---|---|
live_browser_check.py --browser firefox ... --recovery positive03 |
0;10:53:43.030009Z | 344 checks;390ร844 at zoom200%/18px; original injected overflow detected; offline recovery reached; eight groups/eight integer exit0 records; fixture removed |
live_browser_check.py --browser firefox ... --inject-api-cache-defect negative03 |
1;10:54:52.937993Z | 78 preceding checks; exact worker caches only public shell paths error; seven groups/seven integer exit0 records; failed fixture retained |
unchanged check-browser-defect.py 1 negative-03/browser.json |
0;10:55:07.826789Z | Original exact-cause, browser/passkey and complete identity/reap/exit predicates accepted the intended negative |
All groups were kernel-empty and reaped, with no cleanup errors or owned children remaining. The expected-negative133,119,667-byte fixture was then moved atomically, without deletion or byte changes, into its exact stopped outer evidence directory; exit0/stopped10:56:32.302003Z. Earlier failed fixtures likewise remain preserved. Every workload used the resource launcher, strict all-UID95GB DATA census, fresh reservations and default16/20GiB/no-swap/1800s outer containment. Browser domains retained approximately3.5/4GB memory, CPU200% and512-task limits. Focused Rust steps reserved2GiB and these browser/admin steps1GiB; no full-profile admission is inferred from those smaller reservations.
These are finite engineering controls on the observed Linux/Firefox configuration. They grant no #45 acceptance, elapsed endurance, FULL-soak or product qualification credit. Independent product oracles and frozen product packages are unchanged. Final full20 local checks, documentation checks and current-head hosted required gates remain pending at this publication checkpoint; acceptance remains held for this internal harness repair.
Primary raw records below are under the private DATA root; hashes bind the exact archived JSON, whose commands, source/artifact pins, log/resource hashes, exits, stops and cleanup remain available:
| DATA-relative record | SHA-256 |
|---|---|
results/browser-viewport-primitive-01/source-controls-terminal-01.json |
eab701573624247f639765afcb036774d9cfbbf736674855580d9de1b7212cb5 |
results/firefox-viewport-calibration-01/actual-terminal-01.json |
a0738f4c191ed298d2ff9f7a9e36e9ab7f21b967913ab821bfa02472b28292f3 |
results/firefox-viewport-native-comparison-01/actual-terminal-03.json |
fa7b481afd5035150dd1d61cd4b7d87b8682a65a197e49f4a5ae54fc427f2936 |
results/browser-viewport-primitive-01/current-native-entry-terminal-01.json |
1142e0ef5ba81e9af07a036e751d5eccc210e4f88d99ba1444487ddbbebbb1a6 |
results/browser-viewport-primitive-01/adapter-source-terminal-01.json |
8acc199367181e870ad8f049ed058910070b07fade3b42cea8f5b150ba80c276 |
results/firefox-viewport-packaged-diagnostics-01/positive-failed-terminal-01.json |
a45bed51e7ba45549d4d742d5867eb3791f38d11e9c2e3b4a8c11dd8b60bed2a |
results/firefox-viewport-packaged-diagnostics-01/positive-failed-terminal-02.json |
92ee83de0963c20efda0dd8b2c5601eb7ef4ec9d1347d72fa32a08712f9bf0d5 |
results/firefox-prompt-lifecycle-diagnostic-01/actual-terminal-01.json |
400a62aec3cdf3e5e36d0b2cad2830763e0fb3e96ff6665d03752431d2abfee0 |
results/firefox-viewport-packaged-diagnostics-01/actual-terminal-03.json |
73756a7aae41a64afda2ee3e1298c19b2e5a87a18b6a766c5a0e63d5d59d1b39 |
results/browser-viewport-primitive-01/packaged-negative-fixture-relocation-03.json |
49f18267ebf52d685bd796c589bd0c36186b2ddfe984c8ba3a69b163d8c61180 |
Final viewport integration checks
At clean 1c2a624a7697bd93ad63a6ff97920dfe51f0b5e0, all20 original local fast commands completed0 under the registered two-light-worker dispatcher with heavy work serial. The native OpenSSL complete-profile verifier and unchanged Python complete-profile validator each completed0 on those same receipts. Frozen Bun dependencies, documentation check,55 probes and10 hooks also completed0, with final whole-stop 2026-10-18T11:18:31.727125Z. The seven contained cell records retain source-clean before/after states, exact arguments, output hashes, enforced resources and confirmed stops. The full terminal index is DATA/results/browser-viewport-primitive-01/final-local-terminal-01.json, SHA 34221056114d854fb4e42c850ed43b17764391203ed4a9a191c4471d57521ecb.
The fresh worktree-private build used Rust1.99.0 and the previously registered dev/test settings: line-tables-only debug information, incremental disabled,256 codegen units, optimization0, debug assertions and overflow checks enabled. The original8GiB+1 sparse refusal fixture and every fast command remained unchanged. An actual strict all-UID census of77,870,386,407 bytes admitted the full17,000,000,000-byte reservation below95,000,000,000. The highest30-second sampled DATA observation was85,033,665,896 bytes; samples are not a peak-storage proof. Fast outer runtime was521.246s: workspace tests343.263s, workspace check46.279s, Wasm check45.596s, kernel properties25.170s and fuzz corpus24.208s were the largest recorded checks. These are run observations, not a scheduling speedup comparison.
Before that admission, the18 explicitly approved obsolete owned cache roots were fully archived and byte/type/mode/link/identity checked, then removed under separate root approval. The restorable archive is410,078,886 bytes, SHA 51d48211c28a31731cadbbaf678baf080465e6e536fa8e8275ec278c7482bf69; complete custody totals501,055,519 bytes. Manifest SHA 91695e6355fd30ad5e2f6e227c27cfabf22d939b470d9ccb0c38c65d5d93dc2a retains the exact roots, regular-body hashes, path metadata and source/settings/receipt joins. Cleanup completed0/stopped11:08:34.224198Z. Current required compilers, shared caches, frozen packages/executables, source, results and all meaningful failed fixtures were excluded. The now-completed final private target requires its own reviewed custody before any later removal.
The original main1ff Chrome03 receipt remains a finite367-check observation: nine stopped groups but eight listed process exits do not satisfy the required complete exit-map invariant. A fresh Chrome acceptance baseline is therefore required after the corrected harness is merged and explicitly resumed; its raw receipt will not be retrofitted. The engineering Firefox03 positive/negative likewise remain outside acceptance. PR83 is open at1c2a624; hosted fast/security succeeded, while current extended shards, final aggregate and tested-checkout artifact custody remain pending at this checkpoint. No merge or acceptance completion is claimed.
Viewport engineering merged
All five current-head hosted gates completed successfully: fast12m30s, security1m17s, semantic shard22m12s, remaining shard30m17s and required aggregate42s. The artifacts bind head 1c2a624a7697bd93ad63a6ff97920dfe51f0b5e0 to actual tested synthetic checkout b9b18d40ee20548be3fa30156c417ff415fbfe00, with base1ff56e6 and tree 57b056fe53933bc839284706db272aaacb4e8e0f. All20 fast receipt/log hashes and the native Linux handoff report were preserved, together with actual CI Chrome positive354 checks and exact cache-negative87 preceding checks, each with four complete stopped-group/exit records. These CI trials are not frozen-package acceptance.
The extended aggregate contains exactly the ten original producer receipt/log byte pairs from that same checkout, with the original1+9 individual contained producers each exiting0 and confirming stop. The Rust byte collector and unchanged Python extended complete-profile validator exited0/stopped. The native verify-receipts CLI result is recorded separately for local and hosted fast checks; it is not relabeled as an extended-aggregate invocation. Complete artifact custody is DATA/results/browser-viewport-primitive-01/pr83-merge-ready-custody-01.json, SHA db29a491e39420ed1600f9660e888ded818a68fc8d7418762bf8bcd2b245b53e.
Ordinary match-head merge of PR83 at11:37:23Z produced main 73cccefc208ebc5ee5f59d1e256f26107ef2cd8c, with exact parents1ff56e6/1c2a624 and that same tested tree. Merge/fresh-board proof is pr83-merged-tree-and-board-review-01.json, SHA 7d63f824e5c12ca6511074e69bee8cb9727b272cf783a48e4e36151b8f0069a8. Local results remain attributed1c2 and hosted executionb9b; no main-revision rerun is claimed.
The completed final private target and documentation installation were fully archived and read back before separately approved exact-root cleanup. Archive1,378,329,852 bytes has SHA 35e3cf066229bc3b3ce21b315c142d5eca626a3e09f4ef210421d8f473d64b5b; manifest SHA 74ed1e9a61720eb3812d7348397d30050d8cfae48e7e91c141be9faa4632025d. Cleanup exited0/stopped11:27:10.504673Z; strict all-UID DATA census79,267,653,599 bytes was below the future owner reservation threshold at that observation. Later jobs still require fresh admission; no future fit is inferred. Results, failed fixtures, shared caches, source and frozen executables/packages remain preserved.
A GitHub CLI raw-log download initially exited1 because the client refused terminal escape bytes. That capture refusal is preserved; a distinct bounded download with explicit raw-output permission completed0/stopped and kept the bytes private, without rendering them or changing any hosted result. The reporting-only allowance read at11:34:46Z returned54% used in a10,080-minute window; there is no weekly stopping boundary or billing/reset change.
This completes the bounded viewport engineering prerequisites. Root explicitly resumed the original alpha continuation after merge and fresh coordination review, but the next mechanical70-case binding and fresh Chrome04/Firefox02 acceptance runs were not yet admitted at this publication checkpoint. Frozen product artifacts remain unchanged; engineering Firefox diagnostics retain no acceptance credit. No release or tag was created.
Explicit helper adoption and security capture counterexample
The resumed original matrix reached 23/70 completed records: 22 fresh cells plus the valid original stage reuse. The immutable membership snapshot is DATA/results/browser-replacement-continuation-04/completed-membership-snapshot-23-01.json, SHA 6e91191d5200c2ed648225ec6d6458b7a0a1930c037ec8990a66b047235a90b4. Fresh Chrome04 completed367 checks with nine stopped groups and nine integer process exits; fresh Firefox02 completed344 checks with eight groups and eight exit0 records. Both real cursor-expiry cells retained their approximately903-second waits. Expected negatives retained their original executable grading: standalone cache/nonresident graders where registered, and the existing in-run exact semantic-rejection predicates for the role/rejection cells. Failed fixtures remain evidence; no receipt was retrofitted.
The next Chrome security-positive cell unexpectedly exited1 with timeout: actual WebAuthn owner registration, so the matrix is held, not completed. Its primary SHA is 7830301f7c709a17d85199327c711df9b4dd4d945f4a70c612d34fc2b772efbe. All three groups were empty and reaped, the complete process-exit map was preserved, and the outer stop was confirmed. The entire59,083,774-byte failed fixture was atomically relocated into its exact stopped outer evidence directory without deletion. Custody receipt browser-replacement-continuation-04/custody-chrome-security-positive-failure-receipt-01.json has SHA 42cdd5f9b96d2ed6f8e64e96b5f404788d6bd47a4365e6a4c070951e2a9724dd. There are47 outstanding records:46 live cells, including this failed positive, and the unchanged migration extraction.
A source-derived no-network Bun differential reproduced the legacy security wrapper's API mismatch: URL plus init dispatched and captured the body, while the actual Request-only form threw before dispatch. Its terminal index SHA is 66c1a19f5f86d234ef1b6bb61e47d960816ffaccaf8fe84b641f1331b8c16231. A first proposed repair passed the explicit pinned-Bun effect test but failed the independent actual Chrome03 experiment: constructing a new Request from the original transferred its body before forwarding, yielding bodyUsed=true and a TypeError. This counterexample is retained, not overwritten by the Bun result.
The minimal correction clones a Request before constructing the inspection request. Chrome04 then refused mismatched source operands before a browser session; Chrome05 failed on an incorrectly truncated experimental extraction. Both are setup-only failures, with no candidate verdict. Chrome06 used the actual unique AST-selected source literal and completed all six original cases and16 predicates: original Request-only failure, legacy URL/init control, candidate Request-only, candidate URL/init, effective init-body override, and malformed-JSON refusal. Original inputs remained unread at dispatch and the forwarded effective bodies were checked. DELETE returned200; both owned groups were empty and reaped, children were empty, cleanup had no error, and exits were client0 plus intentional driver-15. Failed experimental profiles from03โ05 remain preserved. These are finite harness experiments, with no acceptance or performance-ranking credit.
| DATA-relative terminal record | SHA-256 |
|---|---|
results/chrome-fetch-browser-effect-01/terminal-03.json |
a258f1a2a696ddc52bddf0e4153205e724511b7a5a96d9f0353c60d4efd2de3c |
results/chrome-fetch-browser-effect-01/terminal-04.json |
9008f40d69d2eb49776c459a6b98b57c394e35e9ad60aad6f0a39e4b4887045b |
results/chrome-fetch-browser-effect-01/terminal-05.json |
77c07e00e092bfc950192f934337b9d704cd2a710bab9d54cf292530342b3fbb |
results/chrome-fetch-browser-effect-01/terminal-06.json |
7a6341b31d3f26634000a2783ff01aea3b26748e1ee9440a5eb6e3fa14bc283c |
Current-source qualification callers also needed explicit native-helper adoption after the shared launch API changed. The old frozen5220 API was internally consistent; it is not relabeled as a failing historical run. The combined change propagates the explicit helper pair across R3, release/query/stress, both ProductionAccess modes, finite bridge/trust/native-cleanup fixtures and owner walkthrough. CandidateFixture owns a verified mode0500 copy, includes its bytes in immutable input hashes, and checks the recorded snapshot binding at closeout. Caller replacement leaves the owned bytes stable; snapshot byte or recorded-binding mutation is refused. The pure current owner argv guard admits the explicit pair without performing filesystem effects, while the independent admission pair is checked before system effects. Frozen owner/pressure runtimes are not silently rebound.
Local evidence remains revision scoped. The Rust qualification wrapper at f5b63d23 executed the existing11 Python modules,89 tests, exit0; census and source-adoption controls also ran there. A generated namespace error in another target was preserved and corrected in test-only24300b9. Release adoption, the pinned ignored Bun effect test, Clippy and formatter then completed0; a separate CI fixture NameError was preserved and corrected atdca7eba9. Native Linux handoff and affected documentation gates completed0 there. Exact compiled-input joins carry those unchanged results to72c3f897; no historical result is relabeled as a new run. New archived helper SHA 4c6cfe3ecf85177492cd8e612da0bcbcf1705cdf7486cd6400c0df7bb9d8069c is bound to the actual ignored Linux fixture and manifest SHA 071592fcc43bf5a5f461929771eebe0c77f1a300b81bcf01e884f5c1c279c347. It is not yet admitted into the acceptance binding.
The first combined CI head exposed an additional verification-policy coverage gap: its positive command's security flag caused an early return after96 security checks, omitting the original normal-positive scope. That result is valid only as security evidence; merge was held. A separate policy commit restored the original normal positive and added a distinct security positive, preserving the API-cache negative. Contract tests require all three commands and reject omission or misplaced flags. At final clean head 72c3f897a9c26b3f1eeb82a4057006e5ddd46cf9, the24 CI contracts, formatter, frozen documentation dependencies/check,55 documentation probes and10 hooks all exited0 with confirmed stops. Terminal index results/release-browser-entry-01/policy-local-terminal-01.json has SHA 3ab4d1037081ba2fcc93beb7ac869a600ca8ba5b33f1b1cbc0d3d7e9ca32a2fd.
PR85 contains this complete combined repair; PR84 remains unmerged. Current72 fast and security jobs have succeeded, while both extended shards and final aggregate remain pending at this checkpoint. Current fast artifact custody must verify three distinct normal/security/negative raw receipts and all original20 receipt/log joins before merge review. The old dca extended run was canceled by the normal new push, not passed. All local project work retains strict95GB logical DATA admission, default containment, private bounded build profiles and preserved failure evidence. The 5GiB sparse Python control and original independent oracles remain unchanged. No new browser-continuity certificate, S1, full soak, #45 completion, release or tag is claimed.
Combined repair merged and original continuation resumed
All five required checks completed successfully at final head 72c3f897a9c26b3f1eeb82a4057006e5ddd46cf9: fast10m23s, security1m14s, semantic22m53s, remaining26m00s and aggregate35s. Exact artifact custody binds the actual tested merge dc9c233aaff29647236a0a293c9222ceb20485f5, parents73cc/72c3, to tree 29627167336fc59786a9d3b2650b423a9cf742c4. All20 original fast receipt/log pairs and10 original extended producer pairs were preserved and matched through aggregation. Actual native/Python fast complete validators, the extended Rust byte collector and unchanged Python complete validator all exited0 with confirmed stops; original extended producers retain their1+9 serial job receipts. Native Linux release/binding controls all ran successfully.
The current fast artifact contains three distinct records from that same checkout: normal Chrome354 checks with four stopped groups and four integer exits, separate security96 with two groups/two exits, and API-cache negative87 preceding checks with the exact registered error and four groups/four exits. All group identities, kernel emptiness, reaping, owned-child emptiness and fixture dispositions passed the unchanged predicates. Intentional ChromeDriver-15 exits remain recorded. The earlier dca96-only artifact remains historical security evidence and is not mixed into this current result. Full joined custody is DATA/results/release-browser-entry-01/pr85-72-merge-ready-custody-01.json, SHA 2bf96fab9e127b2d9900870e7e23e89295559283c99d8449e71fec4a5f1439b5; its contained join exited0/stopped15:11:40.899077Z.
Ordinary match-head merge of PR85 at15:13:56Z produced main 3e97661cc9e0ba70b2c9b6830362f25019752dd0, with exact parents73cc/72c3 and the same tested tree. Fresh main/epic/coordination-board proof is pr85-merged-tree-and-board-review-01.json, SHA 93f9ec155c63ea99e68c3ca034c748b97bc7e9d7e9345bc9d05422ef63919a1f. No product package, source freeze, release or tag changed.
Before resuming, the runtime05 guard changed only the repository and frozen-revision literals. All31 registered source paths, eight critical artifact pins,20 console build/stage asset bodies and21 native compiled-input hashes were rechecked. The archived4c6cfe3e helper is byte-joined to the actual merged Rust/manifests/lock/registry inputs and its actual native fixture; it is not claimed byte-equal to the older4335 helper. Binding05 SHA 8ccca9306fe22a1f8b70a2a531ac0fe2a1bbe9080a769bcdbf222a0158ea5c60 preserves the completed23 records and changes only pending guard/native-pair operands plus the failed security-positive cell's unique retry label/output. All47 pending outputs were absent; original expected outcomes, elapsed scopes, reservations, negative graders and unchanged extraction remain. Two administrative comparison assertions failed before any workload: an invented old output prefix, then an incorrect46-entry assumption for the reviewed47-entry order. Their records are retained; the admitted executable vectors did not change.
Root explicitly resumed that exact original scope. The first security-positive retry exited0/stopped15:25:09.683085Z with97 checks, all eight original security probes, the frozen e078/5fd/3f source/build join and entry4c6 snapshot join. Its three stopped groups were empty/reaped/error-free, the three integer exits retained verifier0/server0/driver-15, and owned children were empty with the fixture removed. Terminal SHA 6cc254f1991be2f40b8f79b6ed308b296158652987481ba47ccfeab662122411 adds one completed original cell; the earlier unexpected failure remains preserved. The remaining original cells continue serially under fresh per-cell strict census and containment. This is not #45 closure or a new continuity/S1/full-soak certificate.
A narrowly reviewed storage step preserved a fixed2529-file,3,459,714,615-byte subset of obsolete non-ELF incremental metadata before removal. The first gzip archive hit its exact cap+1 sentinel and exited2; its partial bytes and failure remain. The separate pinned single-thread xz attempt completed0, full archive/body/path/identity readback and stop: archive1,006,439,092 bytes, SHA 7b612a8b4fea8153ed460fa07840ee6e0f08127f50ffe35c173790d49944f515, manifest SHA ad5b5a250ad1c82cecca828b56d37bf32a9d9bae28e0865352fddeb7be44ee21. Separately approved exact-file cleanup used held no-follow directory descriptors and preserved every directory, ELF, object, library, lock, other file, source, failed fixture and both archives. Actual0/stopped15:12:54.354058Z yielded strict all-UID/GID0 DATA79,613,379,658 bytes. Adding the original owner reservation13,958,643,712 gave93,572,023,370 at that observation; later admission still requires a fresh census and no future fit is inferred.
Visible query evidence counts: product counterexample and engineering regression
The resumed old-candidate matrix reached 32 completed records and then stopped on an unexpected positive failure: UI response property rejected: gap. This was a failed registered acceptance diagnostic, primary SHA 742c3be811392c5fcfb12e507d4577d6ce1f4cd2849d04059c92b9690da5db82; it is neither a passed gate nor a server-generated gap claim. The original response experiment injected the controlled response. Actual UI text showed 200 observations and incomplete coverage, but no visible reported-gap count. Source inspection found that QueryResults exposed the raw gap array only inside collapsed details; the visible gap summary was present only in Overview. The registered console-access contract requires a visible distinction between gaps and partial answers. All five groups and process exits were preserved with confirmed cleanup, and the entire failed fixture remains in its exact stopped-unit evidence directory. The receipt-path deviation and failed metadata correction attempt are explicitly retained in response-positive-gap-custody-path-reconciliation-01.json, SHA 4fbc87420207619315a13117261edd9c9ab153b09875225ce791c5924f606c62.
The minimal product repair at cccf6bb8ab313433c1dfae8327a25d02d4832876 appends the gap and unavailable-source array lengths to the existing QueryResults status. No browser oracle changed: the response grader remains SHA c02b2dd6107bd50f327630dcc503e04c4dd747c1d6284fcca14e31fdda513375. PR86 contains that repair plus its deterministic regression fixture and documentation. The private console build and source-bound staging completed0. This engineering preview explicitly paired old e078 native executables with new ccc UI assets: 120 selected non-UI source hashes match, but the whole Cargo.lock does not. It is not a new native build, package extraction, candidate freeze or acceptance result.
The unchanged --ui-responses Chrome preview completed0 with 173 checks and all ten original response cases. Its gap row visibly showed 1 reported gaps ยท 0 unavailable sources; the original timeout case observed actual abort after 16.156593227 seconds. The targeted --inject-ui-response-defect gap run then completed with expected exit1, 51 preceding checks and the exact original error. Each run has five kernel-empty, reaped groups, no stop errors, five integer process exits (including intentional ChromeDriver-15), owned-child emptiness and confirmed outer stop. Positive primary SHA is 602e6d6b0e49756e4dfee64c6b6c58a8c6e85f3ff5114b2f9f8940d52d4a9b1d; negative primary SHA is fcdd7db10fbf3289760b376921838c353a77697ca7b64e157e76e169109159a5.
Negative-fixture custody initially failed before moving because a metadata assertion read a nonexistent primary label; that actual1 remains preserved. An intermediate command-prefix correction was caught before execution and remains NOTRUN. Corrected custody03 joined the actual primary command to the outer command after its Python prefix, checked the actual outer label and all original hash/cleanup/identity guards, and completed0. Exactly one same-device move preserved inode13141748 and 59,712,151 bytes; source scratch is absent. Custody receipt SHA is 1348806631c39b233a853d681279b1e8ebce109c64015654c1a6dc85585dd251.
At the same clean ccc source, formatter, frozen documentation dependencies, link/document check, 55 checker probes and ten hooks completed0 with confirmed stops. The complete revision-scoped preview and local index is DATA/results/query-gap-summary-01/preview-and-local-terminal-01.json, SHA 2b05ae882b24433aa356e373c0ecacd0db3a80484af034c5daa46f664578ac69. Every project job used the existing default containment, exclusive outputs and fresh strict95GB DATA admission. Current PR86 hosted fast and extended checks remain pending at this checkpoint; the focused-security workflow does not trigger for this UI/docs-only diff. No nonexistent current security job is claimed.
The changed packaged UI requires a fresh product candidate, native/console/package provenance and the original affected installation, service, browser, owner and soak gates. All old e078 32 observations remain historical; none closes a new-candidate gate. The full original browser union remains 70, with unchanged expected outcomes, windows, reserves and negative controls. No release or tag was created.
Query evidence repair merged; new candidate still pending
All four current PR86 required checks succeeded: fast10m30s, semantic23m04s, remaining26m12s and aggregate47s. The security workflow's actual path filters exclude this UI/docs-only diff; no new focused-security execution is claimed. Exact comparison of all filtered inputs from72 toccc is empty; the separate ccc formatter ran0. Historical focused-security receipts remain revision attributed and may be used only where the registered freeze policy permits.
The final contained artifact join completed0/stopped16:23:46.800168Z. It checks all20 original fast receipt/log pairs, ten original extended pairs and their actual stopped producers, exact aggregate byte joins, native handoff flags, native/Python complete verifiers and collector results. Three current browser records remain distinct: normal0/354, security0/96 and API-cache negative expected1/87 with its unchanged exact cause and registered grader. Group/process maps, reaping, kernel emptiness, owned-child emptiness and original cleanup predicates are retained. Complete custody is DATA/results/query-gap-summary-01/pr86-merge-ready-custody-01.json, SHA 24d2c440e86e42be8a12c92390b59f53bc853e196ac9813692afb34013c84d2b.
Ordinary match-head merge at16:24:07Z produced b1e5874fbf9c6feb611d98ed40386ac44a703fe1, parents3e976/ccc, with tree 0f493f6bc680dac316c535f29c19647cf791f9db exactly equal to the actual tested synthetic merge44719ed7. Source epoch is1792340646. Fresh main and every coordination-board comment page were read; proof02 SHA is 5198b0c878b2e942b7e80b868cc1c4587533fb6c8483d7d4e547eacb7a7fccfd. The initial first-page-only freshness summary is preserved and explicitly superseded by the six-page observation; no external instruction delta was found.
Two fresh detached clean source worktrees and a mechanically derived original18-cell build sequence are prepared for the new source. They preserve independent console/native package builds, reproducibility and mutation checks, inspection, resource limits and prospective storage reservations. The new candidate freeze/admission and actual package builds remain pending. The prior e07832 observations and mixed-native engineering preview are not new-candidate acceptance credit. No release or tag exists.