Skip to content

Prepare Native 3.3.1 and complete installed PHP condition-wait qualification - #1108

Merged
rmcdaniel merged 2 commits into
mainfrom
release/general-native-3.3.1
Nov 9, 2026
Merged

rmcdaniel merged 2 commits into
mainfrom
release/general-native-3.3.1

Conversation

@rmcdaniel

@rmcdaniel rmcdaniel commented Nov 9, 2026 •

Copy link
Copy Markdown
Member

Server 3.3.1 packages the seven transition-query reductions merged in #1099. The published Compose default moves to 3.3.1, with notes for unchanged schema, protocol, capability and rollback boundaries. The paired poll-cache writer stays deferred in #1063. This does not complete the performance backlog or claim a benchmark speedup.

php-replay-conditions extends the existing published lifecycle workflow with the normal installed SDK's C1–C5 qualifier. It retains the unchanged thirteen replay scenarios, separates the CLI's progress output from its final report, captures nine raw Native exports, checks the original C2 activity/signal window and continuation lineage, then verifies the owned MySQL wait/timer/task rows in a read-only transaction. Missing or failed evidence refuses. No product artifact is patched.

The Shared fixture selector excludes only scripts/ci/native-qualification-subject.json, which has separate Native promotion contracts. Executable CI helpers, workflow changes and mixed runtime changes still select the full fixtures. Pull-request and merge-group workflows keep their required terminal result; the heavy selection occurs inside the changes job. This PR selects the full backend matrix.

Validation on final head 887cd0f15e3bcdbc8ddfae341704b54ddd685511: all seven normal source workflows passed, 66 successful checks plus the intended cancellation/prepared-local skip, zero statuses. Earlier local checks: 87 isolated Node tests, 13 isolated Python tests, YAML/whitespace and static Docker isolation.

Installed run 42545606994 passed on this runner, using published Native 3.3.0 and normal SDK 3.11.0 with the pinned tooling image. All thirteen replay cases executed with zero skips; all seven C1–C5 cases/nine exports passed; C2's noise signal is between ActivityStarted and ActivityCompleted; C3/C5 lineage matches; MySQL inventory and all reviewed predicates have zero violations. Real PHP lint and cleanup passed. Raw audit. This completes durable-workflow/sdk-php#269's remaining installed coverage on the already published versions.

Source tree 6b167a0980a9a1708e557298e5b698e6f2183da2, base 6d80fbe6bda493b161288adabcb2fd06a48f0cdd: thirteen files, 64,781-byte binary diff / SHA-256 af88dc5a93cc4fc600ca31f8f3f33097f1f0f67d5db099cae697923021f92991.

After ordinary queue/main custody: tag/build 3.3.1, refresh its publication-bound qualification subject, run published corpus and exact-digest/default Compose plus PHP/Sample App follow-through, then promote aliases and publish the GitHub Release after acceptance. No 3.3.1 tag or image exists yet. See #325. Historical source findings and their fixes remain in the PR comments.

Package the seven accepted transition reductions with upgrade notes and the
published Compose default. Keep the paired cache writer deferred.

Select the installed SDK C1-C5 qualifier in the existing hosted replay job,
retain its nine Native exports, verify the original C2 activity/signal window
and CAN lineage, and check the owned MySQL rows in a read-only transaction.
The original thirteen replay scenarios remain required and unchanged.

Exclude only the publication-subject JSON from unrelated Shared fixture
inputs; retain full selection for executable CI helpers and mixed changes.

Co-Authored-By: Codex <noreply@openai.com>
@rmcdaniel

Copy link
Copy Markdown
Member Author

Source review found a runner defect before any installed-profile dispatch. The normal published SDK qualifier emits ok: ... progress lines followed by its final JSON object. This head redirected that mixed stream directly to a JSON report, so the exporter would refuse to decode it.

A narrow successor is prepared locally: retain the original stdout log, accept only the declared progress-line prefix followed by one complete validated report, then run the unchanged export and SQL bindings. Unexpected output, crashes, trailing data and partial reports refuse. Thirteen isolated Python tests pass, including the causal mixed-output regression and the negative cases. The product packages and thirteen original replay scenarios are unchanged. The first local test invocation had a module-import error before any tests; the corrected isolated invocation ran all thirteen successfully.

The current train will finish normally. No runtime dispatch or successor push has occurred; this head will not be merged or used for the installed profile.

Retain the normal SDK CLI stdout verbatim, then require its progress prefix
and one complete validated report before the export and SQL checks. Reject
unexpected output, crashes, trailing data and incomplete reports. Keep the
published package bytes and original replay scenarios unchanged.

Co-Authored-By: Codex <noreply@openai.com>
@rmcdaniel

Copy link
Copy Markdown
Member Author

Attempt 1 is terminal CI PASS on fe9c88473a71e8e1f76d0ccd4566a336d40441e6: 66 success, 1 intended skip (candidate cancellation/prepared-local source qualification), 0 statuses. All seven original workflows succeeded on attempt 1, including every Shared backend lane, PHP feature, Sonar and CodeQL. No retry or cancellation occurred. This does not erase the separately found report-parsing defect, so that head is source-HOLD and had no installed-profile attempt.

Selecting one successor push for 887cd0f15e3bcdbc8ddfae341704b54ddd685511, tree 6b167a0980a9a1708e557298e5b698e6f2183da2, sole parent the first head. The correction touches three files only: raw installed-CLI stdout is retained separately, then the declared progress prefix and one complete report are validated before the existing exports and SQL checks. Unexpected output and incomplete reports refuse. Product packages and the thirteen original replay scenarios are unchanged.

Successor delta: 4,320 bytes / SHA-256 060e0b25c7f2de64c81a781dcb9414c8ecb23a13ec6e97af5dc27c5509f2a2fb. Full delta from current main 6d80fbe6bda493b161288adabcb2fd06a48f0cdd: 64,781 bytes / SHA-256 af88dc5a93cc4fc600ca31f8f3f33097f1f0f67d5db099cae697923021f92991. Thirteen isolated Python tests, static Docker isolation and whitespace checks pass; unchanged Node coverage remains the earlier 87 passing tests. The new PHP lint and installed/MySQL execution are still owed.

Pre-push guards passed: exact head/tree/parent/diffs, clean worktree, main unchanged, PR draft/open at the old head, all first-head workflows terminal, remote exact old head. One exact-old-head leased push starts the normal successor source train. No ready/queue, tag, publication or runtime dispatch is selected here.

@sonarqubecloud

sonarqubecloud Bot commented Nov 9, 2026

Copy link
Copy Markdown

@rmcdaniel

Copy link
Copy Markdown
Member Author

Source and published-artifact qualification are accepted for this exact head. Source CI: 66 success + 1 intended skip, seven original workflows successful on attempt 1, no rerun/cancel. Installed run 42545606994 and raw audit passed all thirteen unchanged replay scenarios, seven C1–C5 cases/nine raw exports, the original C2 activity/signal trigger, CAN lineage, read-only MySQL inventory/predicates and cleanup=0. durable-workflow/sdk-php#269 is completed on published 3.3.0/SDK 3.11.0. The first head's parsing finding remains attributed.

Selecting one mark-ready and ordinary merge-queue entry for head 887cd0f15e3bcdbc8ddfae341704b54ddd685511, tree 6b167a0980a9a1708e557298e5b698e6f2183da2, against current main 6d80fbe6bda493b161288adabcb2fd06a48f0cdd. Full reviewed delta is 64,781 bytes / SHA-256 af88dc5a93cc4fc600ca31f8f3f33097f1f0f67d5db099cae697923021f92991. Pre-checks passed: clean worktree; exact remote/head/base/merge-base/tree/diff; draft, CLEAN/MERGEABLE, no auto-merge; zero statuses; all three required checks green; 67 success including the separate installed dispatch and one expected skip; queue empty.

Use expectedHeadOid matching this head. The queue ceiling is 45 minutes from actual enqueue. If the exact entry remains live then, do one exact dequeue and stop that attempt; no automatic re-entry. After a merge, verify valid signature, ordered parents [6d80fbe6, 887cd0f1], tree equality, original actual-main check outcomes and safe topic deletion. No tag, native-release dispatch, image, alias, GitHub Release or Cloud effect is selected here.

@rmcdaniel
rmcdaniel marked this pull request as ready for review November 9, 2026 01:35
@rmcdaniel
rmcdaniel added this pull request to the merge queue Nov 9, 2026
Merged via the queue into main with commit 0bcadca Nov 9, 2026
68 checks passed
@rmcdaniel

Copy link
Copy Markdown
Member Author

#1108 merged at 2026-11-09 01:59:27 UTC as 0bcadca1bbab1d262f43bff11b802e31af9124cb, inside the selected queue ceiling. No dequeue or re-entry occurred.

Initial actual-main custody: valid signature; ordered parents [6d80fbe6bda493b161288adabcb2fd06a48f0cdd, 887cd0f15e3bcdbc8ddfae341704b54ddd685511]; tree 6b167a0980a9a1708e557298e5b698e6f2183da2, exactly the reviewed head. Current main is this commit. The four queue-associated workflows succeeded on attempt 1.

The original actual-main workflows are running. Topic removal and the 3.3.1 tag/build remain pending terminal main verification. The queue ceiling is satisfied by the merge, so no later dequeue is due. Published-artifact C1–C5/SQL acceptance remains the audited 3.3.0 / SDK 3.11.0 result; 3.3.1 publication and its own qualification are still owed.

@rmcdaniel
rmcdaniel deleted the release/general-native-3.3.1 branch November 9, 2026 02:21
@rmcdaniel

Copy link
Copy Markdown
Member Author

Actual-main custody is complete for 0bcadca1bbab1d262f43bff11b802e31af9124cb.

Valid signature, ordered parents [6d80fbe6, 887cd0f1], tree 6b167a0980a9a1708e557298e5b698e6f2183da2 equal to the reviewed head. Current main remains this commit. Check runs: 71 success, eight expected skips, zero statuses; no failure/cancel/timeout. All nine direct original workflows succeeded on attempt 1. The tenth workflow is the expected skipped Merge queue Sonar callback after the push/main PHP feature suite; its controller only analyzes successful merge_group completions. No check was rerun or waived.

The first cleanup preflight refused before any effect because it required every workflow, including that guarded callback, to say success. After reading the controller and classifying that exact skip, the final guard passed. The topic release/general-native-3.3.1 was deleted once with the exact old-head lease, after clean-worktree/ancestry/zero-open-PR guards; API404 and empty ls-remote confirm absence. Saved WIP refs and the separate qualification-runner ref are unchanged.

This completes source/integration custody only. The 3.3.1 tag, image build/publication, publication-bound qualification, aliases and GitHub Release follow separately under #325. The queue ceiling was satisfied by the 01:59:27 merge; no dequeue remains due.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant