docs: record final managed RC field acceptance and remaining gates
Record supplied fd25cb8 managed-image watch, representative post-warmup media transfers and degraded recovery continuation without claiming avatar or durable family qualification. Require explicit Dockerfile.target and positive entrypoint/marker checks before image promotion; preserve the user README badge.
Co-authored-by: GPT-6 <gpt-6@openai.com>
This commit is contained in:
1 parent
fd25cb8939
commit
9984f21466
3 files changed
+111
-24
No files matched your search
+21
-2
@@ -4,7 +4,26 @@
|
||||
|
||||
## Текущий статус
|
||||
|
||||
### Peer TLS rejection classification fix (2026-10-12)
|
||||
### Final managed RC field acceptance (2026-10-12): DEGRADED, not READY
|
||||
|
||||
- Supplied final field results, not network checks rerun by this docs update: source `fd25cb89398300bb9849db4b5511af75bf86eccf`, correct managed image `sha256:73e955e7fdc6efaef88c62ebe4f7afeab71c29d29ba3c88b2a2b7547bb647da5`, active container `a4e845a6d7…`, official managed PID 1, isolated SOCKS `127.0.0.1:11082`. The earlier `37823…` prototype image is **not active** and is not this acceptance image. Private sanitized proof: `/home/agent/zapret2-autotune-rc-fd25cb8/proof/final.safe.json`. The later docs commit is not the tested image source.
|
||||
- Watch started `2026-10-12T20:15:07.686Z`; check at `2026-10-12T20:30:54.826Z` completed over 15 minutes with DockerRestartCount0 and monitor alive. A fresh orchestrator check after the report returned SOCKS greeting `0500` and WWW HTTP200 with SSLVerifyResult0. Managed `socks5h` probes respect native SS DNS `-d 8.8.8.8`; no local-resolution bypass or certificate bypass is claimed.
|
||||
- At this observation, experimental journal generation9, IDs [1,3,9], hash prefix `0e34…` and covered scope were retained unchanged. Saved candidate3 remains experimental, without qualification/evidence. These are snapshot values, not universal IDs/generation. Runtime recovery classification accepted validated peer-TLS/payload rejections as nonpass, with writable-store/ownership guards and the current job-bound transaction hooks intact. Genuine no-qualified-strategy recovery advanced its cooldown **naturally** while the monitor stayed alive; no cooldown reset or fabricated ERROR/source result was used. This validates degraded recovery continuation, not successful avatar repair or family promotion.
|
||||
- Fresh actual Range media results, **post-warmup**, not cold qualification:
|
||||
|
||||
| Sampled hostname | TLS1.2 | TLS1.3 |
|
||||
| --- | --- | --- |
|
||||
| `rr4---sn-axq7sn7z.googlevideo.com` | 3/3 | 3/3 |
|
||||
| `rr5---sn-n8v7kn76.googlevideo.com` | 3/3 | 3/3 |
|
||||
| `rr18---sn-axq7sn7z.googlevideo.com` | 3/3 | 3/3 |
|
||||
|
||||
All **18** transfers verified HTTP206, 65536 bytes, MP4 `ftyp` and valid Range `0..65535`. Separately, initial TLS1.2 warmup had three timeouts; initial TLS1.3 warmup had three passes. Three hosts across two roots provide representative field transfer evidence only; no cold/all-host or durable family qualification was created. The older rr5 `sn-ixh` counterexample was **not tested** here and is not resolved by the new rr5 `sn-n8v7kn76` sample.
|
||||
- Thumbnail `i.ytimg.com`: JPEG, 2888 bytes, SSLVerifyResult0, decoded 3/3 on each TLS protocol. Fresh watch metadata discovered an allowed exact `yt3.ggpht.com` avatar URL, but actual avatar transfers returned curl28 on both TLS protocols, zero samples. No saved avatar evidence or qualified avatar profile exists. Overall live health remains **DEGRADED / NOT READY**; URL discovery and successful media do not close the avatar gate.
|
||||
- Other production/test targets on 1080/11080/11081 retained their IDs/start times; Xray was untouched. No further remote operations accompany this docs update. Build explicitly with `-f Dockerfile.target` and assert the managed entrypoint plus target marker **before promotion**: a tag alone previously mistook root `Dockerfile`'s prototype for the managed image.
|
||||
|
||||
Remaining work: actual avatar transfer/strategy qualification, then durable representative multi-host family qualification and mode acceptance. Automatic family qualification is not implemented; no experimental candidate was promoted. Restart/recreation observations from older RCs remain separate from this final 15-minute watch. The historical open-media/fallback/deployment labels below are superseded only by the specific final results above, not by blanket production approval.
|
||||
|
||||
### Earlier peer TLS rejection classification fix (2026-10-12)
|
||||
|
||||
- Follow-up independent production repro exposed a separate decoder rejection bug: a bounded/origin-verified truncated image degraded live health but failed owned verification as infrastructure before remaining seeds. Image decoding now emits typed payload rejections (invalid magic/dimensions, truncated PNG/JPEG/GIF, body budget and content-type mismatch); both live health and candidate verification share that class, without accepting any payload or fabricating evidence. Missing/corrupt normalized evidence, malformed transfer metrics/Player JSON, arbitrary same-text errors, explicit guards and mixed cancellation remain fatal; image rejection cannot become media absence. Permanent production monitor tests cover joined workers, seed exhaustion/two-hour pause and a later seed's fresh both-TLS 3/3 evidence with preserved frontend/GGC profiles. The independent overlay repro, affected Go tests and race tests, vet, CLI build and diff checks passed locally; no remote deployment was performed.
|
||||
|
||||
@@ -25,7 +44,7 @@
|
||||
- Avatar discovery now conditionally falls back from a homepage with absent metadata/no allowed avatar to one fresh watch page for the validated configured video ID, on the same trusted HTTPS WWW origin and proxy. Malformed JSON, request failures, cancellation and safety failures do not authorize fallback. Reported watch metadata contained an allowed avatar URL where the homepage did not; the updated fallback has not been field-deployed or image-qualified. URL discovery alone does not qualify the avatar route/profile or close a family gate.
|
||||
- Local source verification for the watch fallback: fresh `go test ./... -count=1 -timeout=180s`, focused avatar inventory/parser/classifier, experimental production-probe and private-inventory regression tests under `-race`, and `git diff --check` passed. These are unit/fixture checks only, not new field evidence.
|
||||
|
||||
Remaining acceptance gates (not asserted defects):
|
||||
Historical acceptance gates at that RC (updated by final field results above):
|
||||
|
||||
1. Representative fresh signed media evidence: at least three unique hosts across at least two roots/prefixes, post-warmup 3/3 for both TLS protocols plus actual media transfer. Do not impose uniform static/cold behavior on the accepted learning experiment or label it normally qualified without evidence.
|
||||
2. Qualify the avatar route/profile: only the thumbnail image profile is currently active; browser-visible images do not close this gate.
|
||||
|
||||
Reference in new issue
Block a user