Rust #325: activity cancellation acknowledgement like Server 2.11.1 by rmcdaniel · Pull Request #745 · durable-workflow/server

rmcdaniel · 2026-10-24T13:37:10Z

Part of #325. Security class: R2 (worker protocol surface). Stacked on #744 (base rust-325-train-oct24-c), so it also takes its route off the unserved list; it merges after #744.

Ports POST /api/worker/activity-tasks/{task}/acknowledge-cancellation, the last activity-task route native did not serve, with its cooperative delivery path. Fixture first.

Fixture (recorded from published 2.11.1)

activity-cancellation-acknowledgement follows the cooperative path over raw worker HTTP:

request cancellation;
deliver it to the activity from the workflow task;
acknowledge the stop from the activity worker.

It records every ownership, protocol and validation refusal, the receipt on the status route, the duplicate, a later outcome, a claim polled on 1.19, an immediate cancel, a retried attempt and a run terminated after delivery. The contract pins the published bodies and the ActivityCancellationAcknowledged event. Identities and times come from the observation, and their relations are exact. A unit test runs the contract on the recorded 2.11.1 observation, on the corrected retry, and on 16 mutations it must refuse.

Reference boundary (Workflow #923). Published 2.11.1 delivers a retried activity's cancellation to the execution's first task, which has already completed, instead of the live attempt's task. The live task stays leased, and its acknowledgement is refused as activity_cancellation_not_fenced. The contract accepts exactly that defect's signature, or the correct acknowledgement, and nothing else. Native gives the correct outcome, and a native execution test requires it.

Native

deliver-cancellation accepts call_kind: activity under the default try_cancel policy. It cancels the open execution, its current attempt and that attempt's own task. It records ActivityCancelled under the request, with PHP's attempt and task snapshots.
request-cancellation no longer refuses a run with open activities.
Activity polls bind the immutable claim, as workflow polls do. A claim polled on 1.19 cannot acknowledge.
The heartbeat reports a delivered stop as PHP does (200, cancel_requested, attempt_closed). The status route carries the cancellation_acknowledgement receipt.
The acknowledgement follows PHP's order: protocol, validation, ownership guard, claim, fence, request, recorded cancellation, snapshot, context, duplicate. It records one event with the stop's evidence and the cancelled task's snapshot, and never changes the attempt.
server_capabilities.activity_cancellation_acknowledgement is advertised. Named namespaces admit deliver-cancellation and the new route.

Smaller parity fixes the fixture exposed:

poll leases, the status route and task snapshots format times with microseconds;
stale activity outcome receipts carry the worker protocol fields.

The execution tests that pinned native's older receipt now expect PHP's.

Not ported. Explicit wait_cancellation_completed and abandon activity policies. Native does not record an activity's policy, so every delivery is try_cancel. This is logged in the port log as a remaining gap.

Evidence

Native full suites passed on SQLite, PostgreSQL and MySQL before the stack rebase. After the rebase, activity_cancellation_ack and route_coverage pass on SQLite.
The fixture passes against native locally; published 2.11.1 passes it.
This PR's hosted runs are the qualification. The R2 handoff with the exact SHA follows when they are terminal.

PHP's repair pass raises the task's repair count and projects the run in one transaction, then stamps the dispatch and projects the run again in a separate step. Operator metrics reads runs, tasks and backlog without a shared snapshot, so one response can straddle those steps. Model the four PHP states and accept reads that never go backwards in read order; native must still report one consistent state.

Appends, diagnostics, delivery as the runtime-reserved signal, worker cursor acknowledgements and waits on workflow task completions, and the hand-over of the cursor and unconsumed messages to a continued run, with the message-streams parity fixture recorded from published PHP 2.11.1. A signal consumed by a condition wait now records no SignalApplied unless it is declared, the instance cursor event carries its task snapshot, and a completion reports the next message task, as PHP does. Retention holds runs whose instance has inbound streams until stream cleanup is ported.

Published Workflow aca7928484 reads each operator metrics value with its own query, so one section can straddle a repair. The PHP-only model now matches eight state-dependent reads in their real order, including the timestamps read before their counts, and reads each health warning from its own field. Native stays strict.

Published PHP treats an open_condition_wait's condition_key and condition_definition_fingerprint as optional strings of any form, trimmed and left out when empty. Native required both and a sha256 fingerprint. The wait projection and run summary now read a keyless wait as PHP does.

Published PHP's WorkflowHistoryEvent::record adds the task's snapshot to each event it records with the task model. Native now adds it centrally for those event types, uses the full snapshot on TimerScheduled and the activity events, and leases a timer task to itself while it fires, as RunTimerTask does, so TimerFired carries the leased timer task.

PHP's run description names the run's workflow class on every command and says whether the command's payload preview is omitted. Native left both out. Only a stored external reference is read to decide the flag.

…ients worker.php already waits, inside the original 30 s budget, for the Worker's acknowledgement of its query result before it stops: the independent caller can see the committed result first. The Python and Rust clients stopped immediately, so a fast caller could end the Worker before its acknowledgement arrived (#708 attempt 1, MySQL 8.0, PHP target, timely). Both now wait the same bounded way. The submission and acknowledgement assertions are unchanged.

…nto rust-325-task-snapshots # Conflicts: # rust/tests/execution/worker_history_presentation.rs

In late-handler-sync mode the client stopped as soon as the handler returned, but the installed SDK logs the lost authority after that, so the expired-authority assertion could see none (#713 attempt 2, PHP target). worker.php waits for its Worker's refusal; every non-timely Python mode now also waits, inside the 30 s original budget, for the expired-authority evidence before stopping. Assertions are unchanged.

…fix; run-bound task snapshots Security review of #718: - oversized() reads native's external_payload envelope as well as PHP's external_storage, so a stored reference over the payload limit omits the preview (and the target name, as PHP does). Run input and output previews use the same check. - The described-command query reads a payload only when its first 30 characters equal native's stored prefix, compared case-sensitively on each backend (COLLATE BINARY, binary cast, PostgreSQL text). - record() reads a task snapshot only from a task of the event's own run. Tests: a real stored reference at the 2 MiB limit and one byte above, upper-case/v2/malformed/large-inline counterfactuals, the query itself on SQLite, and a mismatched-run event with no snapshot.

…s maintenance loop (#721) On SQLite the published PHP API server and its maintenance loop share one database file. A contended poll gets PHP's typed backend_lock_pressure refusal, and the published Python run_until helper ends its Worker on it (sdk-python #218). The cohort tests patch replacement, not writer contention: stop the loop for the PHP record, restart it right after, and record and assert its state. The observer still never retries a poll refusal; MySQL and PostgreSQL jobs are unchanged.

…t until a sustained restart (#721) Security review of #722: the restore trap was installed only after a successful stop and disarmed before the restart was checked. Move the window into published-sdk-patch/maintenance.sh: arm restoration before the stop, restart with up to three attempts that each require the loop to stay running across three checks, and disarm only after that. A failed restart records the observed state and fails with restoration still armed. tests/Unit/PatchCohortMaintenanceTest.mjs injects a stop that exits nonzero after taking effect, a failed PHP record, and a start that exits at once, with a fake docker. The previous ordering fails the first and the last.

Security review of #718 at 65772ab: native keeps a signal's stored input reference in workflow_signal_records.arguments; workflow_commands.payload holds the encoded command. Read the exact native sentinel from s.arguments for signals and from c.payload for starts only, case-sensitively per backend. The describe test now proves the real signal row is the source, mutates that row around the limit with the same counterfactuals, shows an oversized sentinel in a signal's command row does not count, and covers the start command's row.

…721) The Docker workflow isolation policy requires a stopped container's name to resolve to github.run_id, run_attempt and job in the same source. Name the SQLite job's maintenance container from GITHUB_RUN_ID, GITHUB_RUN_ATTEMPT and GITHUB_JOB, refuse to act unless it equals the job's $RESOURCE_SCOPE-php-maintenance, and add a control proving a container outside the job scope is never touched.

…st-325-train-oct24-b

docs/rust-server-port.md lists the published Server 2.11.1 routes native does not serve yet, and that prose drifted from the code (it said 125 of 174). An execution test now sends every method of every published API route, from the embedded PHP route table, through the native router, and requires the routes that answer 501 rust_capability_not_implemented to equal rust/tests/fixtures/unserved-php-routes.v2.11.1.txt. A port that serves a route removes it from the list in the same change. At this tree native serves 126 of 175 route methods; 49 remain. The test fails, naming the route, when one is removed from the list.

…d from published 2.11.1 Fixture first for POST /api/worker/activity-tasks/{task}/acknowledge- cancellation, the last activity-task route native does not serve. - The probe follows the cooperative path over raw worker HTTP: request cancellation, deliver it to the activity from the workflow task, then acknowledge from the activity worker. It records every ownership, protocol and validation refusal, the receipt on the status route, the duplicate, a later outcome, a claim polled on 1.19, an immediate cancel, a retried attempt and a run terminated after delivery. - The contract pins the published 2.11.1 bodies and event. Identities and times come from the observation; their relations are exact. - One documented reference boundary: published 2.11.1 delivers a retried activity's cancellation to the execution's first task (Workflow #923), so that attempt's acknowledgement is refused as not fenced. The contract accepts exactly that defect's signature or the correct acknowledgement, nothing else. - The unit test runs the contract on the recorded 2.11.1 observation, on the corrected retry, and on 16 mutations it must refuse. Published 2.11.1 passes 3/3 locally with its maintenance loop running. Native still answers 501 here; this branch adds the native route next.

…livery to activities Native now serves POST /api/worker/activity-tasks/{task}/acknowledge- cancellation like published Server 2.11.1, and the cooperative path that leads to it: - deliver-cancellation accepts call_kind activity with the default try_cancel policy. It cancels the open execution, its current attempt and that attempt's own task, and records ActivityCancelled under the request with PHP's attempt and task snapshots. Published 2.11.1 picks the execution's first task instead, which breaks a retried activity (Workflow #923); native cancels the live task. - request-cancellation no longer refuses a run with open activities. - Activity polls bind the immutable claim, as workflow polls do, so a claim polled on protocol 1.19 cannot acknowledge. - The heartbeat reports a delivered stop as PHP does (200, cancel_requested, attempt_closed), and the status route carries the cancellation_acknowledgement receipt. - The acknowledgement follows PHP's order: protocol, validation, ownership guard, claim, then the fence, request, recorded cancellation, snapshot, context and duplicate checks. It records one ActivityCancellationAcknowledged event with the stop's evidence and the cancelled task's snapshot, and never changes the attempt. - server_capabilities advertises activity_cancellation_acknowledgement, and named namespaces admit deliver-cancellation and the new route. Smaller parity fixes found by the fixture: the status route and task snapshots format times with microseconds, and stale activity outcome receipts carry the worker protocol fields. Explicit wait_cancellation_completed and abandon activity policies stay unported: native does not record an activity's policy, so every delivery is try_cancel. The activity-cancellation-acknowledgement fixture passes 3/3 against native on SQLite locally, taking the correct retried-attempt branch.

…cknowledgement test - The workflow and activity polls report lease_expires_at with microseconds, as PHP does; the status route already does. - Native's stale outcome receipts now carry the worker protocol fields, so the execution tests that pinned the older receipt expect PHP's. - A native execution test requires that a retried activity's cooperative cancellation cancels the live attempt's task, keeps the earlier task completed, fences the earlier attempt and acknowledges the live one once. The shared fixture accepts published 2.11.1's defect (Workflow #923) here; native is held to the correct outcome. Full native SQLite suite: 257 library and 457 execution tests pass. The acknowledgement, lease renewal, one-activity and cancel-leased- activity fixtures pass against native locally.

…into rust-325-train-oct24-c

…st-325-train-oct24-c

…into rust-325-train-oct24-c

…ust-325-train-oct24-c # Conflicts: # .github/workflows/server-parity.yml # rust/src/runtime/store.rs # rust/tests/execution.rs

#709 serves the three message-stream routes, which the route coverage test reported as now served. Native serves 129 of 175 route methods at this tree; 46 remain.

heartbeat_renews_only_progress_deadline_and_cannot_resurrect_elapsed_authority compared two lease_expires_at strings. Native times are whole milliseconds and serialize without a fraction on a whole second, so a claim lease of ...:26Z sorted after a renewal of ...:26.045Z and the assertion failed about one run in a thousand (train #729, PostgreSQL 17 job 120852521096). The test now parses both and reports the values.

…activity or timer Security's R2 finding on #745: native recorded a cooperative cancellation delivery for call_kind activity even when no open activity was at the sequence. That pinned the wrong delivery sequence, and the duplicate path then refused the correct delivery. Native now runs published Workflow 2.11.1's validateCallBoundary for activity and timer deliveries before cancelling anything: - Every typed step at the sequence must have the call's shape, and an authored sequence must have typed history. Either fault is cancellation_delivery_shape_mismatch. A wait's timeout timer belongs to the wait, not to a timer call. - A call that resolved before the request is cancellation_delivery_not_eligible, including a member of a cancelled selection operation. One that resolved after the request still takes the delivery as it is. - The next sequence may be a call not yet authored. - Each validation refusal answers 409 with the bridge's undelivered body in the worker envelope, as Server 2.11.1 does, and records nothing. The activity acknowledgement probe adds a run whose activity completed before the request, with a timer at sequence 2. The contract pins both refusals, the unchanged history, the still-undelivered request and the timer's delivery. Published 2.11.1 passes it 4/4 and the pr1 recording replaces the expected observation. Native passes 3/3. The previous head fails it with a 200 delivery at the completed activity. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…t stack Conflicts were in two workflows. The fixtures workflow drops its old pull_request path list, as main does; the push path list it now classifies pull requests with already carries this stack's two contract tests. Its comparator step runs main's list plus those two tests. The PHPUnit workflow keeps the stack's patch cohort step beside main's merge queue and fixtures result steps. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

rmcdaniel · 2026-10-24T22:02:39Z

Rust Server: fixed in 9c450ca56 on this branch (Security R2 finding 6385591535). The head is now 9c39ac093, which also merges current main (#746, #733). This PR now targets main as the A unit, containing #743 and #744, so hosted CI runs on what would land.

Fix. Native runs published Workflow 2.11.1's validateCallBoundary for activity and timer deliveries before cancelling anything:

Every typed step at the sequence must have the call's shape (assertCompatible). An authored sequence (sequence < next) must have typed history (assertTypedHistoryRecorded). Either fault is cancellation_delivery_shape_mismatch. A wait's timeout timer belongs to the wait, not to a timer call (isPureTimerEvent).
A call resolved before the request is cancellation_delivery_not_eligible (resolvedBeforeRequest, including a member covered by a cancelled selection operation). One resolved after the request still takes the delivery as it is.
sequence == next may still be a call not yet authored.
Each validation refusal answers 409 with the bridge's undelivered body inside the worker envelope. I checked that against Server 2.11.1's CooperativeCancellationController::deliver and DefaultWorkflowTaskBridge::deliverCancellation. Nothing is recorded.

Evidence (all local).

Fixture first: the probe adds a run whose activity completed before the request, with a timer at sequence 2. The contract pins both refusals, unchanged history, the still-undelivered request (duplicate request shows delivery_sequence: null), then the timer's delivery (TimerCancelled, CooperativeCancellationDelivered).
- Published 2.11.1 (ghcr.io/durable-workflow/server@sha256:bf8bed13…, SQLite, with its maintenance loop running) passes it 4 of 4 times. The pr1 recording replaces the expected observation.
- Native at the fix passes 3 of 3, and the runner's compare of the PHP and native pr1 records passes.
Counterfactual: native built from the previous head efc35b461 fails the contract with a 200 delivered: true at the completed activity, exactly as you described.
Contract unit test: 25 of 25. Six new refusals, including that old head's signature.
Native execution suite at 9c450ca56, full --test execution: SQLite 466 passed, PostgreSQL 466 passed, MySQL 8.0 466 passed (each 0 failed, 1 ignored). The new native tests:
- both refusals leave the delivery fields, history, pending timer and completed activity unchanged;
- the request is still deliverable at its real boundary;
- a next-sequence activity delivery succeeds;
- a timed wait's timeout timer is a shape mismatch;
- a timer that fired after the request still takes the delivery.
The merge with main changes CI files and one native test (activity_deadlines.rs). The workflow structure tests (135) and PhpunitFeatureWorkflowContractTest pass on the merged tree.

Ready for your R2 recheck on 9c39ac093 once hosted CI is green.

rmcdaniel · 2026-10-24T22:38:03Z

Rust Server: hosted CI is terminal and green on the exact head 9c39ac0934e2fabd37383ecfeb550c3cc691b035, which targets main. GitHub reports it mergeable and clean.

Shared Server fixtures run 40136659746: success, 22:00:13 to 22:36:38 UTC (36m25s end to end). Its result job is success.
PHPUnit Feature 40136659771: success. SonarQube 40136659738: success.
Across all check runs on the head, 46 succeeded and 2 were path-skipped.
The only non-success entries on the head are from the superseded fixtures run 40136591831: one failed result job and six cancelled jobs. That cycle was started before this PR was retargeted to main and was cancelled when the reopen superseded it. Nothing in it reflects a test result.

Ready for Security's R2 recheck. After acceptance I'll retarget #744 to main so it's marked merged with #743, merge this unit, and delete the merged train branches.

…cel-ack-boundary

rmcdaniel · 2026-10-25T00:03:35Z

Rust Server: main moved after the green cycle (#751, PHPUnit superseded-run cancellation; #738, PHP update-validation registration fence). Branch protection requires an up-to-date head, so this PR now merges current main at 62be4bbe17bd03c4f5663da4f1f31a1413edba39.

The merge is clean. It adds only main's PHP and CI files; native code and this unit's fix (9c450ca56) are unchanged. Locally the workflow structure tests (51) and PhpunitFeatureWorkflowContractTest pass. A fresh hosted cycle is running.

Security's R2 recheck can target 9c450ca56 for the fix itself. The merge-ready head is 62be4bbe1 once its cycle is green.

rmcdaniel · 2026-10-25T01:10:46Z

Rust Server: hosted CI is terminal and green on the exact head 62be4bbe1 (main merged, with #751 and #738), which targets main. GitHub reports it mergeable with a clean state.

Shared Server fixtures, run 40145192948:
- Every job passed on the first attempt except execution-and-codec (corpus).
- That job failed once in legacy-marker-profile.sh, on the published PHP 2.11.1 target. The profile records PHP first and stops there, so no native recording ran. The message was legacy-two-markers-pending: same original poll recovers, and PHP was answering SQLite lock contention with 503 at the time.
- The same check passed on this branch's previous head with the same native code. The merge in between brought only CI and PHP-application changes, which the published image does not run.
- Rerunning only that job (attempt 2) passed, and so did the aggregate result. I'm recording the flaky check separately.
Every other required check is green on this head.

This unit is ready to merge on Security's R2 acceptance of this exact head. Then #744 retargets to main, the train branches are deleted, and the B unit (#739, containing #712) merges main for its own cycle.

rmcdaniel · 2026-10-25T02:43:45Z

Second-round Security review accepts the fix for the prior call-boundary finding at exact current head 62be4bbe17bd03c4f5663da4f1f31a1413edba39.

The fix commit 9c450ca5621b1e551d06c1396e41f02a25d9c737 directly descends from the reviewed defective head and remains byte-unchanged in the relevant runtime, HTTP, execution-test, probe, contract and mutation-test files at the current head. validate_call_boundary now runs before activity/timer cancellation or delivery recording. It requires authored typed history to match the requested call shape, refuses resolutions that predate the cancellation request, preserves the allowed not-yet-authored next boundary, and wraps validation refusals in the published bridge response.

Coverage specifically proves that a completed activity is ineligible, an activity delivery at a timer is a shape mismatch, a timer delivery at an activity is a shape mismatch, a timed-wait timeout is not a pure timer boundary, refusals change no history or consumer state, the real pending timer remains deliverable afterward, and a timer that resolves after the request may still take delivery. The recorded probe and contract now carry those observations, and mutation cases reject the original incorrect acceptance and partial-state outcomes. Current hosted checks on 62be4bbe1 are green across the listed database/cohort/native matrices, CodeQL, Sonar, public-boundary and repository checks.

This accepts the exact fix for the cited finding, not an unreviewed future merge result. The branch is behind current main; merge current main, run the required full cycle, and preserve this behavior through any conflict resolution before final merge.

…cel-ack-boundary

sonarqubecloud · 2026-10-25T03:00:23Z

Quality Gate passed

Issues
12 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
1.2% Duplication on New Code

See analysis details on SonarQube Cloud

The MySQL and PostgreSQL corpus steps start native-a and native-b together, wait up to 30 seconds for native-a, then check native-b once. On run 40156433545 (#745, mysql-execution mariadb-10.11 corpus), native-b was still binding its port: that single check exited 7 ("Failed to connect to parity-native-b port 8080"), and the step failed before recording anything. native-b now gets the same bounded readiness loop. The survivor check after native-a is killed stays a single check, since native-b has been serving for the whole corpus by then. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

rmcdaniel · 2026-10-25T03:47:38Z

Security final recheck accepts the previously reviewed cooperative-cancellation fix at exact current head 6df4e7340b6bca955bb16967011bc0fbe7cc2918.

The delta from accepted head 62be4bbe17bd03c4f5663da4f1f31a1413edba39 contains only current-main PHP/docs/CI changes from #752/#753 and their merge; none of the reviewed Rust runtime, HTTP, execution-test, probe, contract, or mutation-test files changed. GitHub reports the PR clean and mergeable. Shared Server fixtures run 40156433545 is terminal success on this exact head; both rerun jobs are green, the aggregate result is green, and the remaining CodeQL, Sonar, public-boundary, repository, PHPUnit, database, cohort, native, corpus, and SDK-direction checks shown on the head are complete/success except the intentional path-skipped check.

The Security hold on #745 is released for this exact head. GitHub cannot record a formal approval from the shared maintainer identity on its own PR, so this comment is the durable acceptance receipt. Any future head requires revalidation.

rmcdaniel · 2026-10-25T03:47:55Z

Rust Server: exact-head evidence for Security's re-acknowledgement of 6df4e7340b6bca955bb16967011bc0fbe7cc2918. Hosted CI is terminal and green.

Hosted CI on this head

Shared Server fixtures run 40156433545. Attempt 1 failed two jobs, both classified in server-325 and accepted as bounded classifications:
- mysql-execution (mariadb-10.11, corpus), job 121059713123. A native-b readiness race before any fixture was recorded: one connect attempt (curl: (7) Failed to connect to parity-native-b port 8080 after 0 ms), after which native-b logged listening. The workflow fix is 7723ae5, delivered separately.
- mysql-execution (mysql-8.0, sdk-directions), job 121059713069. The native recording of query-sdk-python-rust got HTTP status 0 instead of 200. The same case failed the same way on CI: run the published-SDK direction lanes as their own fixture part #752 (CI-only, no rust/src changes), so it's tracked on Server parity: legacy-marker 503 recovery check fails intermittently on published PHP (SQLite) #754, where comment 6392974785 records the recurrence. The diagnostics fix that names the request is 4ff1f2c, also delivered separately.
Attempt 2 reran only those two jobs: jobs 121064643324 (mysql-8.0 sdk-directions) and 121064643330 (mariadb-10.11 corpus) both succeeded, and the result job 121067825146 succeeded. The run is terminal success with all 34 jobs green.
Other workflows on this head: PHPUnit Feature 40156433483, SonarQube 40156433496, Public Boundary 40156433478 and 40156428759, and PR Rust #325: activity cancellation acknowledgement like Server 2.11.1 #745 40156429665, all terminal success.

Merge structure

Parents: 62be4bbe17bd03c4f5663da4f1f31a1413edba39, the head Security accepted in 6392589855, and main 8c1c07ea0844e4269701b2c55fe5f81197826698 (CI: run the published-SDK direction lanes as their own fixture part #752, Release PHP long-poll wait capacity when its process dies #753).
Tree: de025a9faf20d4fc777a8938a5d811fd5d0aa979.
Merge base: 4290969ee.
The merge was automatic, with no conflicts.

The merge adds only main's changes

All 56 files that only the accepted head changed are blob-identical to 62be4bbe1.
All 17 files that only main changed are blob-identical to 8c1c07ea0.
One file changed on both sides, .github/workflows/server-parity.yml. Its 140 changed lines between 62be4bbe1 and 6df4e7340 are exactly main's changed lines for that file (base to 8c1c07ea0).
The fix files are blob-identical to the accepted head: rust/src/runtime/cooperative.rs and rust/src/runtime/activity_cancellation_ack.rs.

Local qualification on this exact head: the full native shared fixture suite on SQLite, and the SQLite all-targets suite (264 unit, 466 execution).

Holding the final merge for Security's re-acknowledgement of this head.

…ust-325-worker-sessions Conflicts resolved as unions. PHP guards worker-session commands before validating message streams, so native does too. Capabilities advertise both worker sessions and message streams. The activity heartbeat view keeps the session and adds the cancellation receipt. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The unserved-route list arrived with #745 and still named the five worker-session routes this branch serves, so route_coverage failed after the merge. Native now serves 138 of the 175 published route methods. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…-activity Conflicts with #745 resolved as unions. Capabilities advertise both the local-activity manifest and message streams. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…workflow-import Both sides described omitted command previews. Main reads native stored references (a start's payload and a signal's arguments) through per-backend exact-prefix comparisons. This branch also recognised PHP's own external references, which an imported history keeps. The merged query keeps main's comparisons and adds two columns in the same style: php_payload_reference (exact 23-byte PHP prefix) and reason_payload for commandReason(), which never reads either kind of reference. A preview is omitted when either reference is oversized. The query test now pins both new comparisons and still refuses LIKE. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…cancellation-scopes The port log's route count is corrected to this branch's state: 133 of 175 served, 42 unserved as the route list records. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The unserved-route list arrived with #745 and still named it, so route_coverage failed after the merge. Native serves 131 of the 175 published route methods on this branch. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

rmcdaniel and others added 30 commits October 23, 2026 18:25

Merge branch 'main' into rust-325-opmetrics-race

0317e12

Bring the Rust port log's current state up to October 23

3fa5cd5

Describe each command's workflow class and preview flag like PHP

05a6110

PHP's run description names the run's workflow class on every command and says whether the command's payload preview is omitted. Native left both out. Only a stored external reference is read to decide the flag.

Merge remote-tracking branch 'origin/main' into rust-325-opmetrics-race

ca0d23c

Merge remote-tracking branch 'origin/rust-325-history-presentation' i…

37cc705

…nto rust-325-task-snapshots # Conflicts: # rust/tests/execution/worker_history_presentation.rs

Merge remote-tracking branch 'origin/rust-325-opmetrics-race' into ru…

ce14b38

…st-325-train-oct24-b

Merge remote-tracking branch 'origin/rust-325-query-ack-wait' into ru…

0ff26be

…st-325-train-oct24-b

Merge remote-tracking branch 'origin/rust-325-condition-fingerprint' …

a4f3057

…into rust-325-train-oct24-c

Merge remote-tracking branch 'origin/rust-325-task-snapshots' into ru…

c4d65f6

…st-325-train-oct24-c

Merge remote-tracking branch 'origin/rust-325-patch-cohort-schedule' …

1e7cbf3

…into rust-325-train-oct24-c

Merge remote-tracking branch 'origin/rust-325-message-streams' into r…

c8b65b3

…ust-325-train-oct24-c # Conflicts: # .github/workflows/server-parity.yml # rust/src/runtime/store.rs # rust/tests/execution.rs

Message streams are served: drop them from the unserved route list

6bde945

#709 serves the three message-stream routes, which the route coverage test reported as now served. Native serves 129 of 175 route methods at this tree; 46 remain.

Merge remote-tracking branch 'origin/main' into rust-325-train-oct24-b

d24a001

rmcdaniel mentioned this pull request Oct 24, 2026

Rust #325: worker sessions like Server 2.11.1 (#736) #740

Merged

rmcdaniel changed the base branch from rust-325-train-oct24-c to main October 24, 2026 21:59

rmcdaniel closed this Oct 24, 2026

rmcdaniel reopened this Oct 24, 2026

This was referenced Oct 24, 2026

Cancel superseded PHPUnit CI cycles for each pull request #750

Closed

Cancel superseded PHPUnit runs per pull request #751

Merged

rmcdaniel mentioned this pull request Oct 25, 2026

Rust #325: fence held polls against replaced worker registrations (#734) #739

Merged

Merge remote-tracking branch 'origin/main' into rust-325-activity-can…

62be4bb

…cel-ack-boundary

This was referenced Oct 25, 2026

Server parity: legacy-marker 503 recovery check fails intermittently on published PHP (SQLite) #754

Open

CI: run the published-SDK direction lanes as their own fixture part #752

Merged

Merge remote-tracking branch 'origin/main' into rust-325-activity-can…

6df4e73

…cel-ack-boundary

rmcdaniel merged commit 38a96f0 into main Oct 25, 2026
79 of 82 checks passed

rmcdaniel deleted the rust-325-activity-cancel-ack branch October 25, 2026 03:50

Navigation Menu

Repository navigation

Rust #325: activity cancellation acknowledgement like Server 2.11.1 - #745

Rust #325: activity cancellation acknowledgement like Server 2.11.1#745
rmcdaniel merged 38 commits into
mainfrom
rust-325-activity-cancel-ack

rmcdaniel commented Oct 24, 2026

Uh oh!

rmcdaniel commented Oct 24, 2026 •

edited

Loading

Uh oh!

rmcdaniel commented Oct 24, 2026

Uh oh!

rmcdaniel commented Oct 25, 2026

Uh oh!

rmcdaniel commented Oct 25, 2026

Uh oh!

rmcdaniel commented Oct 25, 2026

Uh oh!

sonarqubecloud Bot commented Oct 25, 2026

Uh oh!

rmcdaniel commented Oct 25, 2026

Uh oh!

rmcdaniel commented Oct 25, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

rmcdaniel commented Oct 24, 2026

Fixture (recorded from published 2.11.1)

Native

Evidence

Uh oh!

rmcdaniel commented Oct 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rmcdaniel commented Oct 24, 2026

Uh oh!

rmcdaniel commented Oct 25, 2026

Uh oh!

rmcdaniel commented Oct 25, 2026

Uh oh!

rmcdaniel commented Oct 25, 2026

Uh oh!

sonarqubecloud Bot commented Oct 25, 2026

Quality Gate passed

Uh oh!

rmcdaniel commented Oct 25, 2026

Uh oh!

rmcdaniel commented Oct 25, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

rmcdaniel commented Oct 24, 2026 •

edited

Loading