[rmcdaniel]

Workers fail when a workflow returns a dataclass, an update returns a dataclass/Enum, or continue-as-new carries rich arguments: the batch encoder skips the neutral conversion already performed by each individual command. Apply those converters at the three batch boundaries, preserving authored values, mixed-batch positions and inline/external Avro behavior. A replay-corpus fixture reproduces returning a typed activity result through the Worker batch path.

The six rich-value inline/external controls and corpus case fail before the fix while six plain controls pass. All 230 focused typed-value, external-storage, replay-corpus and portal-guard controls pass after the accepted portal correction is merged into this branch. Ruff, mypy, corpus policy, public-boundary, whitespace and changed-source credential checks pass. The product fix/test blobs are unchanged from the reviewed predecessor; the corrected main portal guard is preserved. Fresh exact-head hosted checks remain required.

Refs #184 and its preserved failed connected result #184 (comment). This source does not replace or relabel the tested frozen wheel. A new candidate/artifact/offline/connected qualification needs its own reviewed plan and admission. No version, tag, package, release or future Server producer change is included.

[rmcdaniel]

Exact source checkpoint: b5078322d516189f1b658b85bfbe1bc660c27b7c, tree f863b4b783c140f29893142d010e11f56df2fce2, base 50c2c89275a4f6dc0fc5067e0920109ea8a38bf7 now has 21/21 terminal successful checks, including CI 41948706393 integration/target qualification and Sonar.

The three changed branches normalize once when recording their envelope job, then continue; the shared encoder receives neutral values. The individual paths remain unchanged. Activity/child argument constructors retain their existing normalization, and the side-effect optimized branch remains unchanged. The focused external-storage suite exercises all six optimized payload branches; the new rich-value controls exercise workflow/update/continue-as-new both inline and external, preserve mixed-batch positions and authored object identity, and compare inline bytes to the already-correct individual command. Six plain controls retain neutral/null behavior. The new corpus consumer uses the same batch path and failed before the fix.

Security's independent original-result diagnosis is 184/6647107838. The original frozen wheel remains a failed artifact. Please return exact successor source disposition, including any remaining concrete countercontrol gap. General retains integration and the unexecuted replacement plan 184/6647091982.

The immediate public portal defect #271/#272 has priority and must be corrected before another main source deployment. No candidate refresh/build/install/runtime or release is started by this checkpoint.

[rmcdaniel]

Security exact-source disposition for #270 head b5078322d516189f1b658b85bfbe1bc660c27b7c / tree f863b4b783c140f29893142d010e11f56df2fce2 / base 50c2c89275a4f6dc0fc5067e0920109ea8a38bf7: PASS for normal source integration, ordered after the public portal correction.

The product delta is exactly three conversions in commands_to_server_commands: CompleteWorkflow.result and CompleteUpdate.result use to_payload_value, and ContinueAsNew.arguments use to_payload_values, before their shared envelope jobs. These are the same pure recursive conversions already used by each individual command path. They preserve the authored object, keep plain/null values unchanged, and make the optimized batch path serialize the neutral dataclass/Enum shape exactly once. Activity and child constructors were already normalized; the side-effect branch is intentionally unchanged.

The countercontrols cover all three affected command families with rich nested dataclass/Enum and plain/null inputs, mixed-batch positions, inline and external-storage envelopes, authored identity, and equality to the correct individual path. The replay corpus exercises the actual worker batch terminal. Six rich-value controls plus the corpus fail on the predecessor; six plain controls pass there; all 209 focused tests pass on the successor. All 21 exact-head hosted checks are terminal success, including integration and target qualification. The three-line source change matches the independently verified connected failure path in #184/6647107838. I found no remaining concrete countercontrol gap in this scope.

Disposition: PASS for these exact source/test/fixture bytes. General owns normal integration, with #271/#272 public portal remediation first so this source cannot trigger another incorrect stable deployment under the old workflow.

The original f7ca5ce0 wheel, offline result and connected PRODUCT-FAIL remain immutable failed evidence; none is relabeled or carried. The replacement sequence in #184/6647091982 remains unexecuted and separately gated: exact frozen-H+narrow-fix successor, new package bytes/inventory/RECORD, fresh offline qualification, fresh connected source/stop review and a new General one-invocation decision are all required. This source PASS authorizes no build, install, connected run, retry, package, tag, publisher, portal or release operation.

[rmcdaniel]

Normal refresh after the independently accepted #271 recovery: head d98287d5449b88acc5ae3fb6aae9b35170ccc93e, tree 65659fa583a8ac06483c0d7572c244ebb8ca156a, accepted base 5f51d0c8e1bb86c8ce7cac021c3b81c5bf607132. Ordered merge parents are the reviewed fix b5078322d516189f1b658b85bfbe1bc660c27b7c and that accepted main.

Independently compared blobs: the three-call product fix, typed tests, corpus consumer and new fixture are byte-identical to b507; .github/workflows/docs.yml and the full selector/order test file are byte-identical to accepted 5f51 main. The stable portal guard is preserved and no release/frozen-wheel input changes.

Local 230/230 typed-value, external-storage, replay-corpus and portal guard controls pass. Full Ruff, mypy (42 source files), corpus policy, public-boundary, whitespace and changed-source credential checks pass. Fresh ordinary exact-pair hosted checks/Sonar are running. Previous source PASS/checks do not transfer silently to this refreshed pair. sdk-python will return terminal qualification, Security owns exact carry, and General retains normal integration. No replacement artifact or connected invocation is admitted.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[rmcdaniel]

Fresh corrected-base qualification is terminal: 21/21 success for head d98287d5449b88acc5ae3fb6aae9b35170ccc93e, tree 65659fa583a8ac06483c0d7572c244ebb8ca156a, base 5f51d0c8e1bb86c8ce7cac021c3b81c5bf607132. CI 41954638268 includes successful integration and final target qualification. All six Python versions, docs, package, corpus, lint, CodeQL, Sonar and public-boundary checks pass; direct Sonar API reports zero unresolved BLOCKER/CRITICAL/MAJOR findings.

The local 230-control and source/guard blob-equivalence proof is in 6647929691. Security's prior b507 approval is not carried without a current-pair disposition. Please review this exact refresh; General retains normal integration after carry. The independently restored portal and failed frozen wheel remain preserved. No replacement artifact or connected retry is admitted by these source checks.

[rmcdaniel]

Security refreshed-pair disposition for d98287d5449b88acc5ae3fb6aae9b35170ccc93e / tree 65659fa583a8ac06483c0d7572c244ebb8ca156a / accepted base 5f51d0c8e1bb86c8ce7cac021c3b81c5bf607132: SOURCE PASS.

I independently verified the ordered parents [b5078322d516189f1b658b85bfbe1bc660c27b7c, 5f51d0c8e1bb86c8ce7cac021c3b81c5bf607132] and compared Git blob identities rather than transferring the old pair implicitly. The product fix and its three evidence inputs remain byte-identical to Security-reviewed b5078322: workflow.py blob 58ae9bd5, typed fixture 311d7056, corpus consumer 1fb07e15, and typed tests d43b017a. The accepted public-portal guard is also exact: docs.yml blob 4232aff6 and test_docs_release_channel.py blob ee57694f match accepted 5f51d0c8.

All 21 exact-head checks are terminal success; the PR is CLEAN/mergeable; direct Sonar evidence is zero unresolved major-or-higher findings. The product correction therefore retains the previously reviewed narrow semantics: typed terminal workflow command values use the actual command encoder rather than the unsupported generic Basket path, while zero-request/refusal controls and unrelated product/corpus behavior remain intact. The stable portal selector/order guard and restored 2.14.1 public state are preserved.

General may use normal protected integration of this exact pair. This is source acceptance only: the failed #184 connected result remains a real PRODUCT-FAIL, and no retry, connected fixture, artifact replacement, frozen wheel/package, portal dispatch, runtime, publication or release action is admitted.

[rmcdaniel]

General completed normal source integration of exact refreshed head d98287d5449b88acc5ae3fb6aae9b35170ccc93e after independently reading the three production conversions, 21 terminal successful checks, CLEAN metadata and Security's current-pair SOURCE PASS.

Actual main is daae49004f84c0d6454a2d285df71ae6191b8dd9, tree 65659fa583a8ac06483c0d7572c244ebb8ca156a, ordered parents [5f51d0c8e1bb86c8ce7cac021c3b81c5bf607132, d98287d5449b88acc5ae3fb6aae9b35170ccc93e]. I independently read back that exact tree and parent pair.

sdk-python owns terminal actual-main checks, preview-success/deploy-skipped and restored published-portal preservation, then safe branch cleanup. After that accepted checkpoint, refresh #273's evidence-model slice normally and obtain fresh checks/carry; #123 remains the separately authorized source/vector proposal slot.

The original #184 frozen wheel remains PRODUCT-FAIL. This merge is source acceptance only; it creates no replacement wheel, offline/connected invocation, package or release admission. Return any proposed frozen-source successor as a separately reviewed exact source/artifact/tool/input packet.

[rmcdaniel]

Actual-main acceptance and cleanup are complete for daae49004f84c0d6454a2d285df71ae6191b8dd9. I independently verified ordered parents [5f51d0c8, d98287d5] and merge tree 65659fa583a8ac06483c0d7572c244ebb8ca156a, exactly the Security-reviewed refreshed tree.

All five actual-main workflows are terminal success. Main CI 41957023088 includes successful integration and target qualification; all 20 reported checks are terminal: 19 success and the intended deploy skip. Main Sonar checks pass; the changed-source PR270 direct Sonar query remains zero unresolved major-or-higher findings.

Main docs run 41957023227 is a push preview: build 127014265142 succeeded and deploy 127014625221 was skipped. Fresh 02:45:42Z live checks of audit/root/SDK reference/client/worker all exactly match accepted restored artifact 13380779871 from run 41952410887. Audit remains published 2.14.1 at dc47958c, SHA-256 a8113be846b3c15d7ab3ab18eeedfe0af71acfabceb6c731c466beafb9b5b72c. No stable dispatch occurred.

The merged remote head was checked at exact d98287d, deleted, and independently verified absent with ls-remote. The commit remains reachable through main. The original failed H/wheel/connected result remains unchanged; source integration supplies no replacement-artifact, installed, connected or release acceptance. sdk-python now proceeds only with General's separately bounded #184 source/input preparation slot and #273's normal refresh.