Repository navigation
M9: host acceptance, endurance evidence and race hardening - #14
Open
JordanD120 wants to merge 6 commits into
Open
JordanD120 wants to merge 6 commits into
JordanD120 wants to merge 6 commits into
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Tasks 126/133/127/134/135 add host acceptance traceability, accelerated endurance and demonstrated reliability hardening above M8. Records the user's M8 verdict at
e1511c1fa39a155801e5c1855872747c9b78c8f8. This separate branch targetscodex/m8-offline-usability; preserve #5 → #6 → #7 → #8 → #9 → #10 → #11 → #12 → #13 → M9, all unmerged.The acceptance matrix traces all six official objectives and 40 SRs to implementation, automated scope, manual source claims and unresolved gates. Exact wording/source statuses and provisional release identity remain unchanged. The deferred operator checklist covers separately authorized installation, electrical/polarity validation, calibration, camera/model/crop approval, physical STOP/interlocks, usability and supervised real 72-hour endurance.
Demonstrated reliability defects are corrected: an older running state response could overwrite a newer paused DOM; unconfirmed Pause unnecessarily sent an optional stale revision and received 409 during concurrent worker activity; healthy queued logging could be mistaken for a fatal worker failure; and status reads unnecessarily applied writer setup under SQLite contention. The client discards superseded polls and omits only the optional revision for unconfirmed Pause. Status queries now use read-only/query-only connections. Bounded per-path host writer coordination shares the existing busy budget, preserving concurrent reads/backups and explicit over-budget failure. Bounded timestamp-normalization reuse fixes the demonstrated 100,000-row deadline regression without changing UTC ordering or deadlines. M1 sources/permissions/confirmed-start checks remain unchanged. Fresh per-tick persistence evidence distinguishes pending from failure; actuator scheduling skips that queued tick, while unknown false results and real IO errors still degrade/revoke.
The synthetic endurance runner advances 72 hours with one-second synthetic input/status heartbeats, hourly fixture pump scheduling, lighting edges, 72 captures/inferences and 864 batched five-minute synthetic records. It checks gateway dispatch context, durable media receipts, integrity/FKs, strict CSV/image exports, backup hashes and isolated verified restores, bounded to 64 MiB/1000 files without pruning. Actual elapsed runtime is reported separately; this is not 72 real unattended physical hours or agronomic acceptance.
New tests cover matrix/audit fidelity, queued logging races, stale Pause/Start boundaries, actual corruption of disposable SQLite, ENOSPC/quota refusal, sensor/camera/model loss, clock rollback and forcible termination of a real disposable process with retained partial stage/interrupted receipt and no permission resumption. The repeated Chromium test exercises all seven screens, forced response/preparation ordering, start/pause, focus/keyboard/touch, transport loss, recovery and stale images. All existing tests/assertions/timeouts are retained. All 264 tracked Python checks, actual-sketch C++ and all seven Chromium workflows passed source-head PR CI and push CI. Local Windows results retain two original symlink-privilege skips; POSIX CI runs those checks. Nine preserved root-app checks passed locally. Three final-source M9 browser campaigns cover 84 complete screen visits and nine repeated start/pause cycles, plus ordering/transport/recovery challenges. The 72-hour accelerated local campaign took 240.932 real seconds and retained 27837074 logical bytes in its temporary workspace. Exact logs, hashes, traces, screenshots, endurance and CI evidence are committed. Tested implementation:
8c45f0d6736a5294067c9441b29c25e9e5a2532d; review head:8027b9f5b424aa5fc226ea4c3897ee3e2fbfdab3.Implementation/readiness, acceptance matrix, reliability findings, security review, operator checklist—not executed.
Historical successful-response M4 timeout and a separate pre-dispatch schedule-save response wait remain causally unproven; corrected demonstrated races do not establish every past cause. HTTP read/download identity authorization, production roles/Host/Origin policy, HTTP concurrency limits, trusted Python/filesystem races, target endurance/performance, real sensor/camera/model/crop/electrical settings, requirements release, retention and human usability remain open. HTML responses now use no-store and all responses nosniff; no production access is enabled.
Host-only review. No Pi deployment, physical interfaces, firmware flashing, real camera/model downloads, live actuator operation, production POST enablement/new exposure, historical deletion/retention automation, Gantt changes or merges. Original 15 hashes, untracked work and accepted contracts remain preserved. Return for M9 review; physical/production readiness is not established.
Final review-head complete campaigns: PR CI and push CI, both successful.
The source-range secret scan is clean. The full evidence-range scan flags two API-named SHA-256 entries; both were independently recomputed against HMI_API_CHECKLIST.json and api_requests.json and verified as content-hash false positives.