Run37074392471
completed with failure. The lead independently reconciled six native ZIP archives
against authenticated GitHub source/run metadata, byte lengths and SHA256 digests,
then checked ten report identities, unique cases and native summaries. Normal
units pass1021/1022 Linux/macOS and1016/1022 Windows; scalar execution is skipped.
Recovery passes136/136 Linux/macOS and135/136 Windows. The one Windows projection
case ProjectionCrashRecoversEffectsOutboxReceiptOutcomeAndCheckpointTogether
at MutationApplied3 reports TaskCanceledException; retained evidence does not
establish its cause. Each OS still has1000 unique successful seeded atomic
process-trial rows; those rows do not erase the separate failed projection case.
All118 analyzer cases pass in each OS report and in the standalone analyzer job.
The exact receipt (report removed from repository) retains all archives,
jobs, report hashes and failed case identities.
Both image jobs again fail at native context-inspect before client/runtime/load execution. Their source has the same absent Docker format argument documented below; native stderr bytes were not captured, so no more specific runtime cause is invented. RF3, comparisons and measured profiles are skipped. The source repairs, new binary WAL, isolated Linux cell pipeline and published TimeSeries consumer update each require qualification at a new delivered source. Historical three-OS reports remain unchanged by the owner's future Linux-only direction.
Run37073331174 completed with failure. Root independently verified all four native test archives against authenticated GitHub byte lengths, SHA-256 digests and source/run metadata, then checked ten reports' source revisions, unique case identities and summaries. Normal units pass1021/1022 on Linux/macOS and1016/1022 on Windows. All136 recovery and118 analyzer cases pass on each OS; each OS retains1000 unique process-trial rows within four concurrent storage trials. No native report contains skipped, cancelled, timed-out or flaky cases. Scalar invocation was skipped after the normal unit failure. Exact jobs, reports and archive receipts (report removed from repository) preserve these boundaries.
RF3 and comparison preparation both fail before restore/build/client/workload
execution: native context-show exits0, then context-inspect exits1. The exact
source reads dockerArgument.format but never defines it. The
byte-verified image receipt preserves
both command sequences. Add the named --format value without weakening Engine,
image, cleanup or ownership checks. The Linux/macOS metadata regression reads
configId from a parser that returns configImageId; its source correction
keeps manifest receipt assertions on configId. Five additional Windows Node
contract failures have only generic retained errors; their cause remains open.
Neither repair is qualified until it is included in a new pushed SHA and its GitHub gates pass. Both measured profiles were skipped; dry-run benchmark JSON inside unit artifacts is not real comparison evidence. Process-kill trials do not qualify power-loss durability. The owner-directed Linux-only isolated native one/two/three-node matrix is newer source and remains unqualified by this run.
Run37060131271 completes with failure. All12 native reports are independently source/hash/count verified:3543 outcomes,3521passed and22errors, with zero native skipped/cancelled/ timed-out/flaky/in-progress cases. Full solution build, formatter, governance and 118analyzer cases pass on each OS; unit suites pass 871/871 each OS. Recovery passes 136/136 Linux/macOS and fails 116/136 Windows. Genuine Docker/Aspire RF3 passes 46/46, including additive series SDK/MCP and Chrome cases. Exact jobs, native byte hashes, case identities and process receipts retain this source's evidence; later candidate code inherits no qualification.
Windows19 seeded batches (0..18) and one subscription MutationApplied3 case cancel at marker waits, receipt writes or command-file read. These errors do not identify a ZoneTree atomicity defect. The preserving four-slot storage-trial candidate retains1000trials, every fault/assertion and original15s/20s deadlines; its own exact-SHA full recovery proof is pending.
Comparison is 2/4, and both measured profiles are skipped. Raw smoke has six targets/96cases: KeyLoad has zero samples because its caller supplied no three RF3 endpoints; Rabbit lacks its authenticated management client; PostgreSQL event readback reaches the interface default. Other-engine samples are present but do not qualify the failed harness. The separate series raw report has48samples, 20attempts/checks and only one timed attempt per operation; RF3 KeyLoad and one Timescale server have distinct acknowledgement/topology contracts. No throughput, p99, statistical performance winner or server-resource claim follows. The byte-verified diagnostic receipt (report removed from repository) records these observations and their limits.
The exact comparisons parent remains Waiting with no terminal exit observation; the pinned Aspire13.6.0 DCP Executable watch times out after60seconds. This matches the upstream watch repair as a supported causal inference, not proof of child exit. No published servicing package carrying that fix was available in the observed feed. Native terminal/exit-zero/report gates remain mandatory.
Main scheduling is now actually observed under AC-CQ-017: the started f25 dispatch survived the later c486 push, ended at20:31:14Z and c486 jobs began at20:31:16Z or later. Pending-run replacement and PR replacement semantics remain unchanged; the two source revisions have independent qualification results.
Owner clarified SIMD on2026-10-02: .NET intrinsics first, Rust only after profiling. Five existing public vector validation/golden/real-store cases pass on each OS atc486. Validation blocks and the full hardware-disabled CI suite are later source pending their own exact-SHA tests; scoring formulas/accumulation remain unchanged. Coverage, actual activation movement, full nine-engine Docker/six-profile graph, server cost, retention/rollups, endurance and power-loss gates remain open.
Run37057708780 retains all12 independently source/hash/count-checked native reports. Full solution build, formatter, governance and118 analyzer cases pass on all three OSes. Units are 868/870 on each OS, process recovery passes 136/136 on each OS, and native Docker/Aspire RF3 is 44/46, including the real Chrome administration flow. Comparison remains 2/4 with both measured profiles skipped. Exact report hashes, case failures and six terminal job receipts retain3540 outcomes:3530passed,10errors and zero native skipped/cancelled/timed-out/ flaky/in-progress cases. Successful step conclusions are preserved; absent literal process exit integers remain null.
The new valid aggregate fixture accidentally reused sample ID5 with changed content; the real deduplication gate correctly rejected it. Two catalog constants still expected50 after the three accepted series tools raised the exact catalog to53. The RF3 cancellation fixture expected a CLR exception while the existing SDK returns a failed Cancelled Result. The R9 source repair (report removed from repository) adds a changed-content Conflict/unchanged-state/healthy-append regression, repairs the valid corpus, updates only the frozen catalog counts, and asserts all three SDK cancellation results plus a successful following read. These source repairs pass an all25-project development build and canonical formatter; their own delivered-SHA GitHub tests remain pending.
The d186/run37056851814 workflow was cancelled after retaining seven native reports, including these real failures and one Chrome cancellation. The following 2fc0/run37057526531 was cancelled before jobs/artifacts existed. Neither is a full qualification. The partial d186 receipt preserves its actual source and scope. The later f627 Chrome pass does not prove the cause of the earlier cancellation.
Main CI now preserves an already running main qualification when another main push/dispatch arrives; PR replacement cancellation remains enabled. The one-line source change still needs a real two-submission GitHub scheduler receipt under REQ-CQ-008/AC-CQ-017. GitHub's pending-run replacement behavior is not changed. The full workflow, numeric coverage, actual activation movement, per-operation server costs, retention/rollups, endurance and power-loss gates remain unfinished.
Run37049469093 retains all12 source-checked native reports. Full solution build, formatter, governance and118 analyzer cases pass on all three OSes. Units pass 809/809 on each OS, process recovery passes 136/136 on each OS, and native Docker/Aspire RF3 passes 41/41 through the real .NET/MCP clients. The complete workflow still fails: comparison is 2/4, both native runner-completion predicates remain Waiting, and both measurement profiles are skipped. Exact native hashes, case transitions, jobs and artifact metadata retain3352 outcomes:3350passed,2errors and zero native skipped/cancelled/timed-out/ flaky/in-progress cases. Successful GitHub step conclusions are retained; their literal process exit integers were not logged. Comparison records literal exit2.
All11 prior macOS replica-readiness setup errors and both Windows missing-required elapsed errors now pass with their original assertions and bounds. The four W5 startup diagnostic methods pass on each OS, including the real output-drain streams. That does not execute every environmental startup/restart failure branch or establish its cause. ZIP digests are retained as GitHub metadata, not independently verified downloads. No local runtime tests ran. Numeric coverage, actual activation movement, server-resource/performance, endurance and power-loss gates remain open.
ADR-052 and TimeSeries map AC-SERIES-008..012 to new latest, half-open raw aggregates and dense UTC windows using published ManagedCode.TimeSeries10.0.0. The shared native reverse storage visitor preserves forward scans; typed SDK/HTTP/MCP operations retain persisted authorization, one read budget, request actors and RF3 storage ownership. New unit and genuine SDK/MCP follower-restart cases are authored. Their own delivered-SHA qualification is pending; e8d1a9eb1 does not qualify this later source.
Run37044499074 retains all12 source-checked native reports. Full build, formatter, governance and118 analyzer cases pass on all three OSes; units are 805/805 on each OS. Native Docker/Aspire RF3 passes41/41, including real .NET/MCP clients, Chrome and five new safe-diagnostic regressions. Recovery is 136/136 Linux,125/136 macOS and134/136 Windows. Exact native hashes, failures, previous-case transitions, job URLs and artifact metadata (report removed from repository) retain the complete scope.
Prior checkpoint stages7/8 and Windows target-WAL reopening now pass. All11 new macOS replica-readiness cases fail during fixture construction at ReplicaSnapshotFiles.RejectLinks; two Windows missing-required-file cases exceed the preserved6s observation cap. These remain real failed gates, with fixture path and scheduling review required before a preserving repair. The four prior Windows startup timeouts pass without a demonstrated source-cause repair; the additive startup diagnostic still requires its own delivered-SHA CI.
Both comparison completion predicates still fail with the runner in Waiting; retained smoke/TimeSeries reports are unqualified and measured profiles skipped. Native RF3 success leaves the environmental restart failure branch unexecuted. Numeric coverage, actual activation movement, server-resource costs, endurance and power-loss gates remain open. No local runtime qualification ran.
Source join (report removed from repository) retains all six source hashes, accepted feature/task ownership, native red baseline and authored source chronology. It reuses the existing physical-temp resolver for the macOS readiness fixture, isolates only two Windows intrinsic missing-file timing cases, and adds closed bounded startup exit/drain diagnostics. All25 projects pass an enabled Release development build with0warnings/0errors; canonical formatter and static governance pass. No local runtime tests ran. The subsequent exact e8d1a9eb1 run above qualifies all12 new startup outcomes and the full recovery suites on three OSes. Comparison remains failed and unexecuted environmental branches remain open; the full goal stays in progress.
Run37042082081 retains all12 source-checked native reports. Full build, formatter, governance and118 analyzer cases pass on each OS. Units are805/805 on Linux/macOS and801/805 on Windows: four real comparison-host startup cases reach their unchanged20-second deadline. Recovery remains119/121 Linux/macOS and118/121 Windows. Native Docker/Aspire RF3 passes36/36, including real .NET/MCP SDK calls and Chrome administration. Comparison remains2/4 with measured profiles skipped. Exact reports, failure identities, job URLs and artifact hashes (report removed from repository) are retained.
W4 source repairs optional-metadata absence and expected source/target file readiness, and adds safe restart failure receipts. The source join (report removed from repository) records source hashes, preserved bounds, an all25-project development build with0warnings/0errors, full canonical formatter exit0 and static governance pass; its new136 recovery and41 RF3 case inventory is unexecuted until the delivered-SHA CI. Four Windows startup failures still need diagnostics to distinguish process exit wait from redirected-reader drain; no cause, timeout increase or configuration change is inferred. Coverage, actual activation movement, server-resource costs, endurance and power-loss gates remain open.
Run37038422388 completes with failed recovery, native restart and comparison gates. All three OSes pass full build, formatter, governance,118/118 analyzer cases and805/805 units. Recovery is119/121 Linux/macOS and118/121 Windows. RF3 is35/36: all EventStreams SDK/MCP data assertions and the real Chrome admin browsing case pass. The native node2 leader-loss/minority restart fails again. Comparison remains2/4, with measured profiles skipped. Complete native hashes, failures, job URLs and archive metadata (report removed from repository) retain the exact source.
Recovery still fails absent checkpoint metadata stages7/8 on every OS and a target metadata-WAL reopen after snapshot publication on Windows. All20 parameterized50-trial seeded crash cases and all6 W3 readiness cases pass on every OS. The prior Windows source-store and elapsed-bound cases pass without a source repair; this does not establish stability or remove their W4 regression/fix obligations. No skipped native suite counts as passing.
Root W4 handles sanctioned metadata absence, all expected source/target ownership under one existing deadline, two isolated intrinsic timing measurements and additive safe failed-restart receipts. These source refinements require their own delivered-SHA CI. Both comparison predicates still fail with the runner in Waiting; retained TimeSeries/smoke reports remain unqualified. No local tests, recovery, AppHost or benchmarks ran.
Run37036628601 completes with failed recovery, RF3 and comparison gates. The full solution build, formatter, governance, 118/118 analyzer cases and 805/805 unit cases pass on each of Linux, macOS and Windows. Recovery is119/121 on Linux/macOS and116/121 on Windows; RF3 is32/36; comparison is2/4. All native suites ran with zero skipped cases. Exact native report hashes, failures, job URLs and artifact receipts (report removed from repository) retain the complete source scope.
The prior admin unit, catalog cursor, discovery schema and native logging cases pass. RF3 still fails three EventStreams expected-data assertions and the real Chrome admin browse case; aed338c carries a later independent admin-owned source correction pending its own CI. Native leader-loss/minority passes this run, without a demonstrated root-cause repair or stability claim.
Recovery still fails both absent checkpoint metadata paths on every OS. Windows also fails source-store metadata/ownership reopening and the existing permanent-lock elapsed observation bound (6.9586565s versus the preserved6s bound). All twenty parameterized50-trial process-crash cases pass on every OS; five of six W3 readiness cases pass on Windows, and all six pass on Linux/macOS. Root W4 probes and real-file regressions remain source work pending exact GitHub qualification. Do not describe this as a green recovery gate.
Both native comparison completion predicates still fail with the runner in Waiting; retained smoke/TimeSeries reports remain unqualified and measured profiles were skipped. No local tests, recovery, AppHost or benchmarks ran.
Run37032546228 completes with failed runtime gates. Full solution build, formatter, governance and118/118 analyzer cases pass on Ubuntu, macOS and Windows. Unit suites are800/804 on each OS; recovery is119/121 on Linux/macOS and118/121 on Windows; Docker/Aspire RF3 is27/36 and comparison2/4, with no skipped native cases. The exact native reports, case identities, SHA-256 hashes, jobs and artifact receipts (report removed from repository) preserve the complete scope.
The W3 search-byte and exact raw-read-budget cases pass across all three OSes. Official MCP initialization succeeds and thirteen of the prior seventeen failing MCP cases now pass; the other four reach existing catalog or event-byte assertions. Six real-file readiness cases and all twenty parameterized seeded process-crash cases pass on every OS. The complete recovery gate still fails: checkpoint stages7/8 have no live tree/0.meta.wal, and a separate Windows replica-snapshot caller still reopens before the WAL becomes available. Native diagnostic capture still misses ToolsCall; the two new admin unit cases and one catalog cursor argument also fail on every OS. Each failure remains in the task plan; none is reclassified as passing.
The RF3 failures are one discovery-schema assertion, three event-byte assertions, three admin authorization/shell cases, the admin browser case and one native node2 restart. Concurrent admin work owns its source and associated catalog/assertion corrections. Root investigates recovery and native restart separately. The two comparison completion predicates still fail with the runner in Waiting; preserved smoke/TimeSeries JSON remains unqualified and measured profiles were skipped. No local tests, recovery, AppHost or benchmarks ran.
Run37029344985 completed failed at the matrix/RF3/comparison build gates after the all-source main checkpoint. The standalone analyzer suite passes; unit, process-recovery, RF3 and comparison suites were not executed. Native reports, compiler diagnostics and artifact receipts (report removed from repository) retain the exact source and job scope. This does not supersede the prior full runtime evidence below.
The joined correction preserves the admin dashboard clock/comparison/async policy and adds the real metadata-WAL readiness probe under REQ/AC-STORAGE-012. Six real-file cases cover all three holder paths, cancellation, pre-cancellation and the unchanged five-second permanent-lock bound. Every original recovery caller,50 trials and15-second deadline remains. The tests-first source receipt (report removed from repository) records the hashes and source join. All25 projects compile with0warnings/0errors in a development build; full corrected-source GitHub qualification remains pending.
Run37021991878 completed with required failures. Full solution build/format/governance and118/118 analyzer regressions pass on all three OSes. Unit suites are787/788 each; all four prior W2 failures now pass. Ubuntu search stored bytes, macOS exact topic-read bytes and Windows concurrent native logging capture fail independently. Complete native reports, hashes, jobs and archive receipts (report removed from repository) retain precise cases. The historical53-case ledger now has52 exact cases passing on all OSes and one corrected fixture method with all current cases passing; the retired argument is not claimed executed.
Recovery executes after every successful build even if unit tests fail:115/115 on Linux/macOS and104/115 on Windows. Eleven Windows errors involve metadata-WAL sharing or per-trial cancellation. The readiness probe omits that WAL; the holder is unknown. Preserve all seeds, deadlines and assertions, and do not claim a product/dependency cause without evidence.
RF3 is12/29: nine of26 caller cases and three separate real-file receipt cases pass. The retained-replica scenario passes this run without a demonstrated root-cause repair and still needs stability proof. Seventeen official MCP cases receive discovery -32603 then fallback Initialize HTTP400 at ProtocolRevisionCount. Pinned SDK/.NET source uses unknown-length content; maximum-capacity buffering projects at least240,947,200 bytes into the134,217,728-byte default control pool. Bounded actual-capacity framing repairs are source joined, pending renewed official RF3 proof. Guards, auth, quotas, protocol and pool defaults remain exact.
Comparison is2/4. Existing report bytes now survive cleanup: unqualified schema3 smoke and separate TimeSeries reports are retained. TimeSeries records20 successful attempts and20 checks total across its three targets and48 samples. Native terminal/zero-exit gates fail and foreign-schema assertion is not reached. Report existence does not qualify performance or refresh the published website; Aspire13.6 native watch remains an external blocker.
The new main source packet repairs recorded-time fixture coordination, same-store raw-byte boundaries, native logging scheduling and unknown-length MCP buffers. Its exact-SHA CI remains pending. No local tests, recovery or benchmarks ran.
Previous completed canonical workflow: run37015193756, source ad594642b4f1a05ac5df0fff0a33b562f4aebf87, branch codex/runtime-qualification-20261002. It completed with failure. Subsequent W2 repairs are source-reviewed and development-built, awaiting the next exact-SHA GitHub run. Shared checkout HEAD/index and independent website changes remain preserved.
| Gate | Result | Exact job |
|---|---|---|
| Standalone analyzer regressions | 88 passed,0 failed,0 skipped | analyzer |
| Ubuntu full build/format/governance/analyzers | Passed; unit778 passed/3 failed/0 skipped, total781 | Ubuntu |
| macOS full build/format/governance/analyzers | Passed; unit777 passed/4 failed/0 skipped, total781 | macOS |
| Windows full build/format/governance/analyzers | Passed; unit779 passed/2 failed/0 skipped, total781 | Windows |
| Process recovery | Not run after unit failure on all OSes | Same matrix jobs |
| Docker/Aspire RF3 real SDKs | 8 passed/18 failed/0 skipped, total26 | RF3 |
| Comparison suite | 2 passed/2 failed/0 skipped; later measured profiles not run | comparison |
Counts qualify only the recorded source and scope. Five vector finite/golden/real-store tests pass on every native OS run; no software-fallback invocation or new SIMD validation optimization is qualified. All12 artifacts were downloaded; actual report hashes and GitHub archive metadata are in the complete receipt (report removed from repository). Extracted-report hashes are verified; raw downloaded ZIP digests were not verified. No local tests, recovery, AppHost or load qualification ran.
| Test / gate | Actual failure at ad594642 | Accepted next source / evidence |
|---|---|---|
| MissingAndMalformedStoredBodiesStillRejectThenRecover, all OSes | Atomic missing-body Corruption propagates; fixture expected a result | Exact exception and unchanged index/body/counters/position/apply assertions |
| SingleMessageReceiveStopsAtFirstReadyRecordAndKeepsFifo, all OSes | One delivery scans257 index entries | VisitRange stops after enough deliveries or256 examined records; no mutation during callback; expiry/FIFO/cap regressions |
| CancellationWhileJsonIsGrowingStopsBeforeRawCsvIsPublished, Ubuntu/macOS | Observer misses early cutoff or whole-write cancellation | Arm actual dedicated-thread file observer before writing; preserve20k×4096 corpus,10s,OCE,<quarter cutoff,no MD/CSV; observe all cleanup tasks |
| MidBodyCancellationMapsReadFailureAndClientCanSendNextRequest, macOS | Actual server RequestAborted wait times out | One bounded second incomplete chunk after cancelled caller result; same5s and actual abort/client reuse; closed stage evidence |
| Seventeen official MCP cases | Discovery/initialize still fails HTTP400; recorded legacy initialize is a secondary fallback | Keep current protocol/authority, add closed stage/method diagnostics to distinguish first discovery rejection |
| Retained replica restart scenario | Write while one replica is stopped maps an unexpected exception to RecoveryRequired | Keep first bounded RF3 failure file before later failures overwrite it; root cause still open |
| Two native comparison completion gates | Runner state remains Waiting with unavailable creation/start/exit timestamps | Pinned Aspire13.6 executable watch terminates after a Polly watch-response timeout; unpublished upstream fix remains external qualification blocker |
TimeSeries partial report contains20 successful attempts and20 matching correctness checks across TimescaleDB2.30.2-pg18, RF3 KeyLoad and published ManagedCode.TimeSeries10.0.0 in-memory buckets. They exercise48 samples, range/boundary/offset/empty/invalid handling, aggregation and cleanup. The whole TimeSeries test failed its native completion gate and did not reach the separate foreign-schema assertion; report existence does not make that test pass. No matched durability or performance winner is claimed.
Official Aspire source at13.6.0 awaits the watch response under a one-minute timeout. Upstream fix f48a7b1 retries that timeout; no compatible published patch was available in the feed read. Keep native terminal/zero-exit checks. Preserve existing report bytes after app stop, and run recovery independently after a successful build without ignoring unit failure.
Run37005805424 at6949fa0 previously failed51/53/50 of770 unit cases,20 of26 RF3 cases and2 of4 comparison cases. At ad594642,48 of those exact unit cases pass on all OSes and one corrected catalog fixture method passes all current cases. Four distinct cases remain; the retired catalog argument is not reported as executed. The full53-case attribution and new per-OS proof (report removed from repository) retain the original symptoms.
Tombstone/null-versus-empty regressions, strict frame decoding, malformed bounded HTTP problems, retained quota corruption and matching stream command IDs now have actual multi-OS unit proof. Per-request fixture isolation and leader-loss recovery pass in RF3; official MCP, retained replica recovery and native comparison completion still fail. Root TestResults retention works, including complete analyzer/RF3/comparison reports. Progressive report byte/schema tests pass; its real early-cancellation gate remains open. W2 source packets require renewed full CI.
flowchart LR
A[Exact source checkpoint] --> B[Canonical GitHub CI]
B --> C[Passed build and static gates]
B --> D[Failed runtime suites]
D --> E[Scoped source and regression repairs]
E --> F[Next exact source CI pending]
The complete run receipt (report removed from repository) retains native summaries, the four-case union, every exact job, all12 GitHub archive receipts and downloaded report hashes. The historical repair ledger (report removed from repository) maps each prior case to its actual new proof or remaining failure. Production readiness, coverage, software fallback, server-resource budgets, activation movement, fault/endurance and power-loss gates remain open.