[P3] Stop double-running compatibility tests in full CI

[alexeygrigorev]

Problem

Push CI has two separate owners for compatibility verification, but the full Django job currently invokes the local aggregate target unconditionally:

• At the current origin/main snapshot 0577393b4c709c4e3653d3a5971b2a7af31b57cf, make test depends on test-compatibility.
• The push workflow full Django branch calls make test regardless of the plan disposition.
• The quality job separately builds and runs the compatibility contract only when compatibility_mode == 'rerun'.

Therefore, a full push with compatibility_mode=skip still runs compatibility once through Django, while rerun runs the same contract once in Django and once in quality. This is a CI scheduling and evidence-contract defect that adds avoidable minutes; it does not change product behavior.

Authority and current evidence

• CI smart test-selection audit, punch-list item 9
• Selective CI ownership, plan, and evidence rules
• Verification strategy: infrastructure and deployment tests
• Versioned quality-contract target list

The current scheduled full-regression job intentionally calls local make test, so it is a separate full-regression contract and must remain explicit.

Scope

1. Add a CI-only test-django-full Make target, or an equivalent explicit conditional target, that runs the existing full Django test recipe without the test-compatibility prerequisite. The preferred target keeps the local make test contract unchanged.
2. Change the push workflow full Django branch to use that CI-only command. Keep the quality job as the sole compatibility owner when the verification plan says compatibility_mode=rerun; do not weaken the plan disposition.
3. Update the versioned verification command and evidence contracts so the full Django command, ownership metadata where applicable, evidence allowlist, and recorded envelope all agree on the exact command. The current dynamic command mapping and command allowlist are here and here.
4. Add or update CI contract tests and the selective-CI documentation to pin the target prerequisite, workflow mode conditions, plan command, evidence command, and local/scheduled distinction.

Non-goals

• Do not change compatibility tests, compatibility fixtures, URL/link/SEO behavior, or the compatibility contract itself.
• Do not remove compatibility from local make test; the scheduled full-regression job may continue to call it and must execute compatibility once.
• Do not change the quality-contract-v1 primitive target list or its aggregate semantics.
• Do not redesign classifier/profile selection, ownership closures, evidence reuse, or the four-hour scheduled backstop.
• Do not change Django application code, public routes, templates, browser behavior, deployment behavior, or dependencies.

Dependencies and status

• No blocking issue dependency is identified. The implementation depends on the current selective-CI plan/evidence contract and the existing compatibility_mode output consumed by push CI.
• The scheduled full-regression behavior is an explicit compatibility boundary: its Django job calls make test and must stay separate from push CI.
• Ready for engineering. If an equivalent conditional prerequisite is chosen instead of test-django-full, it must still expose an unambiguous command in the plan and evidence envelope.

Acceptance criteria

• The CI-only full Django command runs the same full Django suite currently reached by make test, but does not depend on or execute test-compatibility. Local make test still depends on test-compatibility.
• The push workflow full Django branch invokes the CI-only command and no longer invokes local make test.
• With compatibility_mode=skip, the push Django and quality jobs execute no compatibility contract command, and the plan records the compatibility component as skipped.
• With compatibility_mode=rerun, the push quality job executes the compatibility contract exactly once, the push Django job executes it zero times, and the aggregate evidence report contains one compatibility execution.
• The scheduled full-regression path continues to run local make test and executes compatibility exactly once; existing local make test semantics are preserved.
• The full-profile Django plan command, ownership metadata if applicable, evidence allowlist, recorded envelope, and workflow invocation agree on the exact CI-only command. The quality-contract primitive target list remains unchanged.
• Contract tests pin the Make target prerequisite behavior and push workflow behavior for skip, rerun, and the full Django profile, including a regression assertion that the scheduled job remains on local make test.
• make test-ci, the relevant verification plan/evidence/report checks, and the existing local/scheduled workflow contract checks pass with no product-page changes.

Implementation slices

1. Make target: add the CI-only full Django target and preserve make test as the local compatibility-inclusive target.
2. Plan and evidence contract: update the full Django command mapping, component command allowlist, ownership metadata if required by the selected design, and envelope/fixture expectations.
3. Push workflow: switch only the full Django branch; retain quality compatibility gating on rerun and preserve all artifact/result recording.
4. Tests and documentation: add focused Make/verification/workflow contract assertions for skip, rerun, full, local, and scheduled paths; document the two intentional commands and run the repository CI checks.

Verification and browser applicability

• CI/Django/integration: verify target prerequisites and dry-run command expansion; generate and validate a current verification plan; exercise the workflow/evidence contracts; run make test-ci and the focused contract tests.
• Repository/operations: inspect both push and scheduled workflow YAML, confirm the compatibility component is classified exactly once, and confirm no evidence command mismatch or duplicate execution.
• Browser and screenshots: not_applicable. This is an internal CI/process change with no page-render, route, accessibility, or responsive behavior change, so no web.dtcdev.click browser session or screenshot is required.

[alexeygrigorev]

Grooming Complete

PM verdict: READY FOR ENGINEERING

I read AGENTS.md, _docs/PROCESS.md, the product-manager role guidance, the linked CI audit, the selective-CI design, the verification specification, and the current workflow/Makefile contracts at origin/main 0577393b4c709c4e3653d3a5971b2a7af31b57cf.

Exact current-state evidence:

• Makefile:test depends on test-compatibility.
• The push full Django branch calls make test without consulting the compatibility disposition.
• The quality compatibility step runs on compatibility_mode == 'rerun'.
• The scheduled full-regression Django job intentionally calls local make test.

The rewritten issue now specifies the CI-only full-Django target, exact skip/rerun/full-profile behavior, plan/evidence command-contract updates, explicit non-goals, local/scheduled preservation, implementation slices, and focused contract verification. It records browser and screenshots as not_applicable because there is no page-render impact.

No blocking issue dependency was found. Labels are now P3, infra, testing, and bug; no needs grooming label was present. No repository files were edited and no commit, push, merge, or PR was made.

[alexeygrigorev]

Engineer handoff — tester-ready

Status: ready for independent tester verification. No dependency blocker.

• Worktree: /home/alexey/git/dtc-website/.tmp/backlog-213-engineer
• Base/head: 0577393b4c709c4e3653d3a5971b2a7af31b57cf / 0577393b4c709c4e3653d3a5971b2a7af31b57cf
• Worktree remains uncommitted and detached. No commit, push, merge, or PR was made.
• Final plan: .tmp/verification-213/verification-plan.json
• Selection artifact: .tmp/verification-213/ci-selection.json
• Plan SHA-256: b6385e10967e4259038b373bb5e1758773693cccc1546b46dfd8cdf082bb68cd
• Graph SHA-256: e0850384988f84551c0091e4fd5dd62e823038fa36f9d4d6fc0ea0ec6d207c2e (policy 5)
• Source-manifest SHA-256: 609f6b571c04e523aba15facbc2768cbdf466d0aab601b2a8e144394a70bcf40

Implemented only the duplicate-compatibility CI contract:

• Added test-django-full; local make test remains test-compatibility plus the same full Django suite.
• Push full-profile Django runs make test-django-full; quality remains the sole compatibility owner when the plan disposition is rerun.
• The scheduled full-regression planner explicitly records make test and the scheduled job still runs that compatibility-inclusive aggregate exactly once.
• Updated ownership metadata, runner command mapping, evidence allowlist, workflow guards, focused tests, and selective-CI documentation.

Final plan buckets:

• rerun: selector, container, django (make test-django-full), evidence_validation, playwright, quality
• reused: none
• skipped: none
• not_applicable: compatibility, content_invariants, screenshots

For this CI-only candidate, compatibility is not_applicable because it is not required by the impact closure; this is the plan schema’s non-required skip representation. The scheduled override plan was separately validated with Django command make test and compatibility not_applicable.

Verification completed:

• Focused contracts: 95 passed in 11.90s
• make test-ci: 592 passed in 45.09s
• Ruff check and format check: passed
• make -n test and make -n test-django-full: confirmed compatibility is only in the local aggregate
• Final plan validation and worktree-manifest match: passed
• Browser and screenshots: not_applicable for this internal CI/process change

No full Django/Playwright runtime evidence envelopes were generated in the focused engineer pass. Tester should independently recompute the plan with consumer=tester, run the required reruns, validate the evidence/report contract, and freeze the tester-final report from this unchanged worktree.

[alexeygrigorev]

QA Review — FAIL

Independent tester verification is complete for issue #213. The candidate is not ready for PM acceptance because the required versioned quality contract fails on the changed push-workflow command.

Identity and provenance

• Tester worktree: /home/alexey/git/dtc-website/.tmp/backlog-213-tester
• Exact base/head: 0577393b4c709c4e3653d3a5971b2a7af31b57cf / 0577393b4c709c4e3653d3a5971b2a7af31b57cf
• Observed tester HEAD: 0577393b4c709c4e3653d3a5971b2a7af31b57cf; detached and uncommitted; no commit, push, merge, or PR.
• Candidate tracked-diff SHA-256: b6eb0f89060af1d6658c996a6772034b2d2e72a74063114f9b97429c39b52ada; reproduced byte-for-byte from the engineer worktree.
• Candidate scope: 12 modified tracked files plus tests_ci/test_django_full_target.py; no product code, dependency, Dockerfile, quality-contract primitive, or portability-guard file changed.
• Engineer handoff canonical plan SHA-256: b6385e10967e4259038b373bb5e1758773693cccc1546b46dfd8cdf082bb68cd.
• Tester recomputed canonical plan SHA-256: dcecc4126707c7cfaed0d2bb349a73e76b211fa8bec7112758c82226e36583e9. The only plan drift is the expected recomputation timestamp; base/head, selection bytes, graph, source manifest, render impact, commands, and dispositions are identical.
• Graph SHA-256: e0850384988f84551c0091e4fd5dd62e823038fa36f9d4d6fc0ea0ec6d207c2e, policy 5.
• Source-manifest SHA-256: 609f6b571c04e523aba15facbc2768cbdf466d0aab601b2a8e144394a70bcf40.
• Selection SHA-256: 5b3a4e10f25635f902e967036a3275265124236c73027bdc6c59ee63e30164c7.

Four-bucket plan disposition

• rerun: selector (make verification-plan), container (make verification-container), Django (make test-django-full), evidence validation (make test-ci), Playwright (make test-playwright-core), quality (quality-contract-v1).
• reused: none.
• skipped: none.
• not_applicable: compatibility (outside_impact_closure), content invariants (outside_large_content_closure), screenshots (render_inputs_unchanged_without_reusable_visual). Screenshots are explicitly not_applicable for this internal CI/process change.
• Plan render impact: false; no browser page or screenshot gate is required.
• Report verdict: failure.

Plan, selection, report, logs, and evidence are retained at:

• Plan: /home/alexey/git/dtc-website/.tmp/backlog-213-tester/.tmp/verification-213-tester-run1/verification-plan.json
• Selection: /home/alexey/git/dtc-website/.tmp/backlog-213-tester/.tmp/verification-213-tester-run1/ci-selection.json
• Report: /home/alexey/git/dtc-website/.tmp/backlog-213-tester/.tmp/verification-213-tester-run1/verification-report.json
• Evidence directory: /home/alexey/git/dtc-website/.tmp/backlog-213-tester/.tmp/verification-213-tester-run1/evidence/
• Report SHA-256: 1afefe493afb823da9e8f84dbfa82e56dc6240deff42ed860afd4f177ec0266a

Required verification results

Component	Result and counts	Evidence
Focused issue contracts	PASS — 95 passed in 14.34s	logs/focused-contracts.log
Evidence validation / CI contracts	PASS — make test-ci, 592 passed in 49.23s	envelope 3faf80eb53f38a081a17c9cba49e29b677ae787cc77788266326354ea7a43591; output evidence_validation-output.log@sha256:d99353ce8ba01772c8bb3f0e5d313993d184b13e6f744e68ddfe53d0be91ae01
Quality contract	FAIL — quality-contract-v1 exits 2 at database-portability-check	envelope 63910023e72dc9cfabc813d3fda4c9bbea570fce68e54b38b8e1e26ad676d01c; output quality-output.log@sha256:2ccd320b954e8b59ec2a33c984ccd2e96025a13e3009eb728c29ddf13dea948f
Full Django	PASS — make test-django-full; 2304 total, 2295 passed, 9 skipped	envelope 7eaa2c6a059ee33d7708e69378188478cf88598d3242d049f78e96f71ba80190; output django-output.log@sha256:ac0451fb00d9ac6758ebf60cb6ce17e939f3622d53d60b3005f4e5fe5f3bb71f
Playwright core	PASS — 203 passed, 28 deselected; flake policy attempted/passed/failed 203/203/0; 1261.08s	envelope b8f204155404de113af7d6a153554acddd0a62657ad6b88dc0fcf9b9c82ad5a7; output playwright-output.log@sha256:aac113cdaf31be9ba1081e768288b524396af4c147fdacd507a109c8e33e3a89
Container	PASS — 13 assertions; container-check.json status pass	envelope 53f61be4f396c00c140fa88d88e6516746f93e06f70d079b96f36629a7ddab00; container-check.json@sha256:6d100f25b01e1d9f7e7cd0b81d5c37a8705a0e69b2c830a7e918f536300b3f00
Selector	PASS — one validated plan assertion	envelope c1d6c35d5588863824b9c0e39b8917799eedef7703fbfa670cc02816a48e609c; plan output SHA 8f5f5deb25fe6c11ee8e9bdb1632fc0a74e0e43a859eeffc7593425d76511703

The bounded aggregate command was:

VERIFY_PLAN=.tmp/verification-213-tester-run1/verification-plan.json VERIFY_EVIDENCE_DIR=.tmp/verification-213-tester-run1/evidence VERIFY_REPORT=.tmp/verification-213-tester-run1/verification-report.json VERIFY_ISSUE=213 VERIFY_WORKTREE=backlog-213-tester VERIFY_CONSUMER=tester VERIFY_PHASE=tester VERIFY_PRODUCER_ROLE=tester make verification-run

It reached terminal completion with exit 2 because the quality component failed; all other required rerun components completed and their evidence directories/report validated.

Compatibility, local, and scheduled contracts

• Makefile:168-174: local make test remains test-compatibility test-django-full; test-django-full has no compatibility prerequisite.
• Actual make test: compatibility ran once (pytest compatibility/tests -q, 305 passed in 70.86s), then Django ran once (2304 total, 2295 passed, 9 skipped).
• make -n test: exactly one compatibility command and one Django command.
• make -n test-django-full: zero compatibility tokens and one Django command.
• Push full Django branch .github/workflows/ci.yml:684-687 invokes make test-django-full) and no exact make test`.
• Push quality compatibility step remains guarded by compatibility_mode == 'rerun'; the focused workflow contracts pass.
• Scheduled planner .github/workflows/scheduled-full-regression.yml:54-58 records --full-django-command "make test"; scheduled Django execution and evidence recording remain on make test at lines 220-232.
• Scheduled override plan validated with Django command make test; invalid --full-django-command make test-unknown exits 2.
• Planner/runner/evidence fail-closed checks pass: invalid planner command is rejected, full runner maps to make test-django-full, the Django evidence allowlist contains only make test, make test-django-full, and make test-ci-focused, and the wrong-command evidence test rejects `make test) for a push full plan.
• validate-plan, validate-evidence-directory --consumer tester, and validate-report all pass structurally.
• Scope audit passed with no unrelated CI behavior changes in the candidate diff.

Blocking finding

1. Candidate-caused quality-contract failure. The candidate changes push full Django to make test-django-full at .github/workflows/ci.yml:687, but the unchanged database portability guard still requires the exact command make test at scripts/check_database_portability.py:1254. The tester quality envelope therefore reports:

ordinary Django CI does not run the owned full SQLite harness

Reproduction:

• Clean base 0577393…: uv run --frozen python scripts/check_database_portability.py passes; log /home/alexey/git/dtc-website/.tmp/verification-213-base-probe/database-portability.log.
• Candidate: the same command exits 1; log /home/alexey/git/dtc-website/.tmp/backlog-213-tester/.tmp/verification-213-tester-run1/logs/candidate-database-portability.log.

The portability guard (and its contract tests, if applicable) must be updated to recognize the compatibility-free push target while preserving the scheduled make test boundary, then the quality envelope must be rerun. Until then, criterion “make test-ci, relevant verification plan/evidence/report checks, and existing local/scheduled workflow contract checks pass” is unchecked.

Tester-final: FAIL.

[alexeygrigorev]

Engineer rework handoff — tester-ready

Status: ready for independent tester verification after the prior #213 QA failure. No dependency blocker.

Identity and scope

• Worktree: /home/alexey/git/dtc-website/.tmp/backlog-213-rework
• Base/head: 0577393b4c709c4e3653d3a5971b2a7af31b57cf / 0577393b4c709c4e3653d3a5971b2a7af31b57cf
• Detached and uncommitted. No commit, push, merge, or PR was made.
• The accepted candidate was reproduced byte-for-byte; the rework adds only the portability guard correction and its focused contract test:
  • scripts/check_database_portability.py:39,1255-1269
  • ci/tests/test_database_portability.py:22-33

Diagnosis and correction

The candidate changed push full-profile Django execution to make test-django-full at .github/workflows/ci.yml:687, but the unchanged portability guard required the exact line make test. The guard therefore failed with:

ordinary Django CI does not run the owned full SQLite harness

Reproduction before the correction is retained at:

/home/alexey/git/dtc-website/.tmp/backlog-213-rework/.tmp/verification-213-rework/portability-before.log

The guard now requires make test-factories and make test-migrations, then accepts exactly one full-Django command from the explicit legacy/CI allowlist: make test or make test-django-full. It still enforces their order. Focused regression coverage exercises both command forms.

Preserved contracts:

• Local make test remains test-compatibility test-django-full (Makefile:168); make -n test has one compatibility command and one Django command.
• make -n test-django-full has zero compatibility commands.
• Push CI uses make test-django-full; scheduled full regression still runs make test and records --full-django-command "make test" (.github/workflows/scheduled-full-regression.yml:57,208,222).
• The quality-contract-v1 primitive target list, evidence envelope behavior, and A failed component's artifact collection must not mask its failure envelope #202/[P1] Establish an explicit Playwright flake policy #208 contracts are unchanged.

Plan and report

• Plan: /home/alexey/git/dtc-website/.tmp/backlog-213-rework/.tmp/verification-213-rework/verification-plan.json
• Selection: /home/alexey/git/dtc-website/.tmp/backlog-213-rework/.tmp/verification-213-rework/ci-selection.json
• Evidence directory: /home/alexey/git/dtc-website/.tmp/backlog-213-rework/.tmp/verification-213-rework/evidence/
• Report: /home/alexey/git/dtc-website/.tmp/backlog-213-rework/.tmp/verification-213-rework/verification-report.json
• Scheduled override plan: /home/alexey/git/dtc-website/.tmp/backlog-213-rework/.tmp/verification-213-rework/scheduled-verification-plan.json
• Focused contract log: /home/alexey/git/dtc-website/.tmp/backlog-213-rework/.tmp/verification-213-rework/focused-contracts.log
• Plan file SHA-256: 92d133b2daf64ce2b495f01d2620ef2984afc1bc280e6ae4f4143eeba497340c
• Canonical plan SHA-256: 1aab678f241f76f991a38e080cc335f8098bf380f90a240633b9ba0157a1a6d7
• Selection SHA-256: 5b3a4e10f25635f902e967036a3275265124236c73027bdc6c59ee63e30164c7
• Report file SHA-256: ef08f9cad427c29d3eadcc5cafdf8d866d5b0c81f4ad932439f7df283cc218fa
• Graph SHA-256: e0850384988f84551c0091e4fd5dd62e823038fa36f9d4d6fc0ea0ec6d207c2e (policy 5)
• Worktree source-manifest SHA-256: d2383b5cf2123a138e1877f9178775e199c352138be2d91f1a6555e034677e78
• Plan profile/reason: full / dependency_toolchain
• Render impact: false; screenshots: not_applicable
• Four buckets: rerun = selector, container, Django, evidence validation, Playwright, quality; reused = none; skipped = none; not applicable = compatibility, content invariants, screenshots.

Terminal-success evidence

Component	Result	Evidence
Focused issue contracts	PASS — 370 passed	focused-contracts.log
Selector	PASS — one validated plan assertion	envelope f53a2d3c7734fd379b473abd69eeb5d6125d7528e3fd6cdb0175a1ea0e20390e
Quality	PASS — 594 passed; portability check passed	evidence/quality-output.log; envelope b53e0f84cc118504483d8d9938766d127bf6526710be2e899f8387c02030f46d
Evidence validation	PASS — make test-ci, 594 passed	evidence/evidence_validation-output.log; envelope 202f96b67d125d63d9a024f1b6dd585045dd5342883791a4b17095b183d4b41c
Full Django	PASS — make test-django-full, 2304 total, 2295 passed, 9 skipped	evidence/django-output.log; envelope e150a6a44e7083661a48033c9ad3a51336479877fabdd72a81486a60fdcca70c
Playwright core	PASS — 203 passed, 28 deselected; flake policy attempted/passed/failed 203/203/0	evidence/playwright-output.log; envelope 30c358e3bdc45736706bfe160acdd54b08a7f85ca7a824955f5bd7a7c79cacc1
Container	PASS — 13 assertions; status pass	evidence/container-check.json; envelope aa13077878a0199140f6cf30b16cbf825080b84b63c0df23bf43c0e120197ebd
Aggregate report	PASS — verification-report.json verdict success; no invalid evidence	report path above

Additional targeted regression command for #202/#208 artifact, flake-policy, and marker-registry behavior passed: 52 tests.

The bounded command was:

VERIFY_PLAN=.tmp/verification-213-rework/verification-plan.json VERIFY_EVIDENCE_DIR=.tmp/verification-213-rework/evidence VERIFY_REPORT=.tmp/verification-213-rework/verification-report.json VERIFY_ISSUE=213 VERIFY_WORKTREE=backlog-213-rework VERIFY_CONSUMER=engineer VERIFY_PHASE=engineer VERIFY_PRODUCER_ROLE=engineer make verification-run

The worktree is frozen at this uncommitted state for the independent tester.

[alexeygrigorev]

QA Review — PASS

Tester-final: PASS. Independent recheck of the corrected candidate is complete.
All relative artifact paths below resolve under the exact tester worktree above.

• Exact worktree: /home/alexey/git/dtc-website/.tmp/backlog-213-rework-tester
• Base/head: 0577393b4c709c4e3653d3a5971b2a7af31b57cf / 0577393b4c709c4e3653d3a5971b2a7af31b57cf
• Observed HEAD: 0577393b4c709c4e3653d3a5971b2a7af31b57cf
• Detached, uncommitted; no commit, push, merge, or PR.
• Candidate scope is exactly 15 changed paths: 14 modified tracked paths plus one untracked added test.

Changed paths

.github/workflows/ci.yml
.github/workflows/scheduled-full-regression.yml
Makefile
_docs/ci/change-selective-ci.md
ci/evidence.py
ci/ownership.json
ci/runner.py
ci/tests/test_database_portability.py
ci/tests/test_workflows.py
ci/verification.py
scripts/check_database_portability.py
tests_ci/test_django_full_target.py
tests_ci/test_evidence.py
tests_ci/test_runner.py
tests_ci/test_verification.py

Independent plan

• Command: VERIFY_BASE_SHA=0577393b4c709c4e3653d3a5971b2a7af31b57cf VERIFY_HEAD_SHA=0577393b4c709c4e3653d3a5971b2a7af31b57cf VERIFY_OUTPUT_DIR=.tmp/verification-213-rework-tester VERIFY_CONSUMER=tester make verification-plan
• Plan: .tmp/verification-213-rework-tester/verification-plan.json; SHA-256 1823a28680e474a5d899f8142d8e72e3ccfce74a77d70b3bc32b92427e7f8694
• Selection: .tmp/verification-213-rework-tester/ci-selection.json; SHA-256 5b3a4e10f25635f902e967036a3275265124236c73027bdc6c59ee63e30164c7
• Graph SHA-256 e0850384988f84551c0091e4fd5dd62e823038fa36f9d4d6fc0ea0ec6d207c2e; policy 5.
• Source-manifest SHA-256 d2383b5cf2123a138e1877f9178775e199c352138be2d91f1a6555e034677e78.
• Profile/reason: full / dependency_toolchain; browser profile core; render impact false.
• Semantic comparison with the rework handoff plan was byte-identical after removing only each plan's generated created_at timestamp.
• Dispositions: rerun = selector, container, django, evidence_validation, playwright, quality; reused = none; skipped = none; not applicable = compatibility, content_invariants, screenshots.

Fresh plan-driven evidence

Command:

VERIFY_PLAN=.tmp/verification-213-rework-tester/verification-plan.json VERIFY_EVIDENCE_DIR=.tmp/verification-213-rework-tester/evidence VERIFY_REPORT=.tmp/verification-213-rework-tester/verification-report.json VERIFY_ISSUE=213 VERIFY_WORKTREE=backlog-213-rework-tester VERIFY_CONSUMER=tester VERIFY_PHASE=tester VERIFY_PRODUCER_ROLE=tester make verification-run

• Report: .tmp/verification-213-rework-tester/verification-report.json; SHA-256 cdce52e414d772c09829b6722ca520e9c9625c1182a712ef0735dc82e172dd8f; report plan_sha256 9def28da96109dc7a7abef1a52851ddf962699be1448726465816fdfca48b4f6.
• Evidence directory: .tmp/verification-213-rework-tester/evidence/.
• Aggregate: verdict=success, invalid_evidence=[], no pending entries.
• All six envelopes were independently validated with ci.verification validate-evidence; verification-evidence-check and verification-report-check passed.

Component	Command	Result/count	Envelope	Key artifact
selector	make verification-plan	success; 1 assertion, 15 changed paths, 5 direct nodes, 0 downstream	a7477bcb52572c61c97a9a2e99b311907eb2644edb2ed609b49e29625dacad85	selector-plan.json 1823a28680e474a5d899f8142d8e72e3ccfce74a77d70b3bc32b92427e7f8694
quality	quality-contract-v1	success; 594 passed, 0 failed, 0 skipped; Ruff format 593 files; mypy 438 files	e6281b2b01ca4f068243166e8f2faf4ee6e6d457674743ca6484ca65b708ffd0	quality-output.log ba9301d7244f32986bc809cf299e1adb24f1f733fe1da22d2bd114ca2b83c0a6
django	make test-django-full	success; 2,304 run, 2,295 passed, 9 skipped, 0 failed	7fd06011a3f5a2b322e0debd54c58fa25a132e8fe8d4d6228b1d93967a8ef5a7	django-output.log 312b3e1cba83ce3d7ad0aa0500db77e822125a629b94d6fb533906b9aa704941
playwright	make test-playwright-core	success; 203/203 selected passed, 28 deselected; attempted/passed/failed/rerun/quarantined 203/203/0/0/0, complete 1	9df67683242b530d1b0254d5283e68d71dfdf20dd18d81ba697ef51b4ac67d69	playwright-output.log 5b03c8f64d9860ceed451098ade7ced80b7a267369186c8e66108e5aba6b0649
container	make verification-container	success; 13 assertions, 0 failed	6959633d15d09e234632cdd430e1fb18bee4fabc64ba9c427841d4b568b4a2ff	container-check.json 6d100f25b01e1d9f7e7cd0b81d5c37a8705a0e69b2c830a7e918f536300b3f00
evidence validation	make test-ci	success; 594 passed, 0 failed, 0 skipped	37ec76bf372f602e85cb2ebd54818894612fd28661f7abb4503605479fa7e8ee	evidence_validation-output.log 40ee2cf53e98c86e54fff749c217d15b72d1725feded1d06e2fdcc270377022a

All envelope files and environment/result artifacts are retained below the evidence directory. The six envelope file SHA-256 values are: selector 1c0eab532192f11bf8a2c3e64b34043b7151cb44ffffb7b6ec2228180d588cdc, quality b80604f9bf92af00d728232037455e7ed59ef47ebf7fc75e232d4e4255bd52b3, Django 326a6c9f1c72a763fb6de70fa3b996f3935f282e595e55b869db24f584b0a7fd, Playwright e5991ebe432759faeb19d3bea14bd79aabbf1972291168bea5a05ed9dff68024, container 5299020ab9d6b060ab2cde10106b4f2fc877e57a9bee882b242e92a42ea9622a, and evidence validation 8f5b300f88b521f97c94dd44ce462777af64ad9653fc7c671b52cedeb8b59105.

Contract and regression checks

• Focused issue contract command: uv run --frozen pytest ci/tests/test_database_portability.py ci/tests/test_workflows.py tests_ci/test_django_full_target.py tests_ci/test_evidence.py tests_ci/test_runner.py tests_ci/test_verification.py -q — 370 passed; log .tmp/verification-213-rework-tester/focused-contracts.log, SHA-256 38f019997692429f841ebb69fa33e9df1ef71db9a16a8f590b42f4b8ff69dfb3.
• A failed component's artifact collection must not mask its failure envelope #202 failed-artifact and [P1] Establish an explicit Playwright flake policy #208 flake-policy regressions: uv run --frozen pytest tests_ci/test_failed_artifact_recording.py tests_ci/test_flake_policy.py -q — 12 passed; log .tmp/verification-213-rework-tester/issue-202-208-regressions.log, SHA-256 7963693828a6e51352da2bc1b8c670ec59311ae09f30e9e548489a57557b256a.
• [P1] Establish an explicit Playwright flake policy #208 marker-registry/factory contract: make test-factories — 139 passed; terminal success. Quarantine monitor: make test-playwright-quarantined — 231 deselected, 0 selected, policy complete; log .tmp/verification-213-rework-tester/playwright-quarantine-output.log, SHA-256 06ba356c975fc2d0637c22bf9f82b5e324d81ac5dedd8079d385797313165bc4.
• Workflow prechecks: make test-migrations — 16 passed; make -n test shows exactly one compatibility command followed by one Django command; make -n test-django-full shows zero compatibility commands. Dry-run log .tmp/verification-213-rework-tester/make-contract-dry-run.log, SHA-256 a004e292a443b49952a549ef1f469449297e73f6b78c8dcec273734a6fb2110b.
• Local aggregate: make test — compatibility 305 passed exactly once, then Django 2,304 run, 9 skipped, 0 failed.
• Portability guard probe: .tmp/verification-213-rework-tester/portability-guard-probe.log, SHA-256 ecd9dd70bcc52cd851673ad2954154d65d06c780ef839f0bc39baba251260077; allowlist exactly ('make test', 'make test-django-full'); 2 accepted cases (one each), 3 fail-closed cases (zero, duplicate, unknown full command).
• Invalid planner command make test-unknown exited 2 with argparse rejection. Existing focused evidence tests also reject a mismatched make test envelope for a push full plan.

Scheduled boundary

• Recomputed scheduled plan: .tmp/verification-213-rework-tester/scheduled/verification-plan.json; SHA-256 2924961ec8c7bba8c3c8fb52551c69913cb137520d869bbca5362c6819b03bee.
• Scheduled selection: .tmp/verification-213-rework-tester/scheduled/ci-selection.json; SHA-256 5b3a4e10f25635f902e967036a3275265124236c73027bdc6c59ee63e30164c7.
• Scheduled plan is source_mode=commit, profile=full, reason=diff_empty, and records Django command make test; it validates successfully. Workflow inspection and contract tests confirm scheduled Django executes make test, while push full Django executes make test-django-full.
• The unchanged quality primitive target list remains quality-contract-v1's 10 targets; no ci/quality_contract.py or product/dependency/Docker path changed.

Screenshots and disposition

This is an internal CI/process-only issue with no product-page render impact. Plan render impact is false, required captures are empty, and screenshots are explicitly not_applicable; no browser screenshot is required or pending.

Final tester disposition: PASS.

[alexeygrigorev]

Product Acceptance

ACCEPT

Issue #213 is accepted from the product and process perspective.

• Push full CI now invokes the compatibility-free make test-django-full target. The quality job owns compatibility only when compatibility_mode = rerun: skip runs it zero times, while rerun runs it exactly once in quality and zero times in Django.
• Local make test remains compatibility-inclusive, and the scheduled full-regression planner and Django job explicitly retain make test, so the local/scheduled contract executes compatibility exactly once.
• The portability guard accepts exactly one full-Django command from the supported allowlist (make test or make test-django-full) and fails closed for zero, duplicate, or unknown commands. The corrected probe and focused regression tests pass.
• Planner, runner, ownership, workflow, and evidence contracts agree on the push and scheduled commands; invalid planner input and mismatched evidence are rejected. The quality-contract-v1 target list is unchanged.
• The independent corrected tester-final is PASS: make test-ci passed 594 tests, full Django passed 2,304 tests (2,295 passed, 9 skipped), Playwright passed 203/203, quality passed 594/594, and all six evidence envelopes validate. The report has verdict=success, invalid_evidence=[], no reused/skipped/pending components, and screenshots are correctly not_applicable for this CI/process-only change.

The earlier tester failure was concretely corrected in the portability guard and its focused regression coverage.

Ready for a focused commit with Closes #213, followed by the prescribed local merge.

[alexeygrigorev]

Software Engineer Rebase Handoff — #213

Status: ready for independent tester verification after reconciliation onto the current main. The independent tester and PM must both re-run their gates after this rebase; prior #213 evidence is not being reused.

Identity and scope

• Worktree: /home/alexey/git/dtc-website/.tmp/backlog-213-rebase-engineer
• Exact base: 52f943d21997bf37926f4aee89ea27f5a39d431a (origin/main, issue [P0] Root-cause and fix the Playwright harness deadlock #206 merge)
• Exact head: 52f943d21997bf37926f4aee89ea27f5a39d431a (detached, uncommitted candidate changes in the worktree)
• Approved source ported: 5095ae3d8353370bc7abc07c2b28fb8719a8dd26, parent 0577393b4c709c4e3653d3a5971b2a7af31b57cf
• No commit, push, merge, or pull request was made.

The approved parent-to-commit patch was applied with a three-way patch application. It applied cleanly, produced no conflict markers or reject files, and its binary patch digest matches the approved patch: 41264862dc4e4d53cfec9375ee1ac8fae838e8ed3849d11f49718a728dc6b853.

The current #206 delta is confined to conftest.py and test_support/tests/test_playwright_harness.py; neither was modified by the port. The #202 failed-artifact envelope implementation and the #208 marker/flake policy implementation were not weakened: the only evidence-contract addition is the make test-django-full Django command allowlist, while scheduled/local make test, artifact recording, marker registration, and flake policy boundaries remain explicit. Targeted regressions for all three issues passed.

Changed paths — exactly 15

.github/workflows/ci.yml
.github/workflows/scheduled-full-regression.yml
Makefile
_docs/ci/change-selective-ci.md
ci/evidence.py
ci/ownership.json
ci/runner.py
ci/tests/test_database_portability.py
ci/tests/test_workflows.py
ci/verification.py
scripts/check_database_portability.py
tests_ci/test_django_full_target.py
tests_ci/test_evidence.py
tests_ci/test_runner.py
tests_ci/test_verification.py

Verification plan and report

• Plan: .tmp/verification-213-rebase-engineer/verification-plan.json
  • file SHA-256: c1de10aa4accbb3e711381f1d83d6dc9ed0532cac8fba78483eb709f362ba7b9
  • canonical plan SHA-256: adc5ddfaba75f6c8348925910c31d4f42bfd88270b7ad242d646a46a2b6b42de
• Selection: .tmp/verification-213-rebase-engineer/ci-selection.json
  • file/canonical SHA-256: 141ab0abd85ea8ea9e30c4b2f508e9611848f71613f7cf41defe36e9e52899ee
• Graph SHA-256: e0850384988f84551c0091e4fd5dd62e823038fa36f9d4d6fc0ea0ec6d207c2e (policy 5)
• Source-manifest SHA-256: 24bc89aef9210e6623e1872c04981b9221ac2d3ce5c26db279be1b1fcff5d4af
• Profile/reason: full / dependency_toolchain; direct nodes: 5; downstream nodes: 0; render impact: false
• Dispositions: rerun = selector, quality, django, playwright, container, evidence_validation; reused = none; skipped = none; not_applicable = compatibility (outside_impact_closure), content invariants (outside_large_content_closure), screenshots (render_inputs_unchanged_without_reusable_visual)
• Report: .tmp/verification-213-rebase-engineer/verification-report.json; verdict success; file SHA-256 d995391fef02363f728c8772fb502220ad4743c0ab96063231c37bd2e9a67f49; report plan SHA adc5ddfaba75f6c8348925910c31d4f42bfd88270b7ad242d646a46a2b6b42de; invalid evidence []; verification-state SHA 2f361e381dbe87470849f66fa5ef158b7921e14367e29bcf7b2b1dccba24a9e6

Plan-driven terminal-success envelope

All artifacts are below .tmp/verification-213-rebase-engineer/evidence/.

Component	Command and result	Envelope	Key output artifact
selector	make verification-plan; 1 assertion, 15 changed paths	39f05fced784577b8326ffe42221187e708df3d8c2d8e2e543a708af7cd321ee	selector-plan.json@sha256:c1de10aa4accbb3e711381f1d83d6dc9ed0532cac8fba78483eb709f362ba7b9
quality	quality-contract-v1; 594 passed, 0 failed, 0 skipped	818d04e7fd2224b499313062c82f2949f63ec0cf302ea2c0dee37e4807c7293f	quality-output.log@sha256:4264d23e4854bca262f877357dde9c1c578dd42b16f6f671b97f61a11529297c
django	make test-django-full; 2,304 run, 2,295 passed, 9 skipped, 0 failed	2bda88a830c96fc2b4c4cc4dea31d6a1099c465e6222c2a050853924e29119a5	django-output.log@sha256:4903ef333e7838e29adb24e63b9c7304456caaf19f1b6b198726c82fee9bedf2
playwright	make test-playwright-core; 203 passed, 28 deselected; flake attempted/passed/failed 203/203/0	ee1d98955278c0dd07fe58dd93569e95b49505f576a32f9aa60392eaf3a263c2	playwright-output.log@sha256:f72c998ac7469050a83899ceabc08973ce80ef67ffa346e005446e71e9643e51
container	make verification-container; 13 assertions, 0 failed	24b4725ccb1b698110d8a9a68cc195bb677ed36c74a65f4cd0d79f762b08bccb	container-check.json@sha256:0ba63c264de0042cd584c91dc98fb6d370e72f005e24642c3f352f8e6b54bdba
evidence validation	make test-ci; 594 passed, 0 failed, 0 skipped	e414f7805a162d27ea87bf57b36651ec42324398cd9c41d24a7509eacfadb0f4	evidence_validation-output.log@sha256:66a313b9d32b36d13c3d1e7ac9b41836d50a43f9232937ce17f1f98f8119f135

The bounded command was:

VERIFY_PLAN=.tmp/verification-213-rebase-engineer/verification-plan.json VERIFY_EVIDENCE_DIR=.tmp/verification-213-rebase-engineer/evidence VERIFY_REPORT=.tmp/verification-213-rebase-engineer/verification-report.json VERIFY_ISSUE=213 VERIFY_WORKTREE=backlog-213-rebase-engineer VERIFY_CONSUMER=engineer VERIFY_PHASE=engineer VERIFY_PRODUCER_ROLE=engineer make verification-run

Additional contract evidence

• Focused [P3] Stop double-running compatibility tests in full CI #213 contracts: uv run --frozen pytest ci/tests/test_database_portability.py ci/tests/test_workflows.py tests_ci/test_django_full_target.py tests_ci/test_evidence.py tests_ci/test_runner.py tests_ci/test_verification.py -q — 370 passed; .tmp/verification-213-rebase-engineer/focused-contracts.log SHA-256 cee52f253221772faf478b8d6f85b149f2e623cd839f2ae14c05f6e988d6e25b.
• [P0] Root-cause and fix the Playwright harness deadlock #206 lifecycle, A failed component's artifact collection must not mask its failure envelope #202 failed-artifact envelope, and [P1] Establish an explicit Playwright flake policy #208 flake/marker regressions: uv run --frozen pytest test_support/tests/test_playwright_harness.py tests_ci/test_failed_artifact_recording.py tests_ci/test_flake_policy.py test_support/tests/test_marker_registry.py -q — 55 passed; .tmp/verification-213-rebase-engineer/issue-202-206-208-regressions.log SHA-256 2c6dc877e6b491add0549bf72bd7293d94b336b1f258e524f5109cdeb1b97220.
• [P1] Establish an explicit Playwright flake policy #208 factory/runtime/safety/marker target: make test-factories — 139 passed; .tmp/verification-213-rebase-engineer/factories-marker-policy.log SHA-256 6301bd3f441915429c34f29a664eb4196d62b9cccf5a85315d35493af1828cbe.
• Scheduled quarantine policy: make test-playwright-quarantined — 231 deselected, 0 selected, policy complete=1, exit 0; .tmp/verification-213-rebase-engineer/playwright-quarantine-policy.log SHA-256 7020f2a3827a494f4c830c2c471a532c54282eec2cd5378d438abe503f4358c6.
• Local aggregate: make test — compatibility 305 passed; Django 2,304 run, 2,295 passed, 9 skipped; .tmp/verification-213-rebase-engineer/local-make-test.log SHA-256 71fe8f8f4d195039e5e849bab0914b632accf93bc9cc144a51554d7326bdb67c.
• make -n test confirmed one compatibility command plus one Django command; make -n test-django-full confirmed zero compatibility commands.
• Scheduled override plan .tmp/verification-213-rebase-engineer/scheduled-verification-plan.json validated with --full-django-command 'make test'; file SHA-256 94debd87ef90d2e9e3e8f64d18bda6b871e1b6a00ea799ff2aecc041cdad0b1a.
• Final validate-plan, validate-evidence-directory, and validate-report checks passed. git diff --check passed.

Browser screenshots are not_applicable for this internal CI/process change. The worktree is frozen, uncommitted, and ready for the independent tester to recompute the plan from the exact base/head and for PM acceptance after tester pass.

[alexeygrigorev]

QA Review — PASS

Tester-final: PASS. Independent recheck after the current-main rebase is complete. Prior #213 tester/PM evidence was reviewed but not reused.

Identity and exact scope

• Tester worktree: /home/alexey/git/dtc-website/.tmp/backlog-213-rebase-tester
• Exact base/head: 52f943d / 52f943d
• Observed HEAD: 52f943d
• Detached and uncommitted. No commit, push, merge, or pull request.
• Candidate reproduced byte-for-byte from /home/alexey/git/dtc-website/.tmp/backlog-213-rebase-engineer: tracked binary diff SHA-256 89fbfe0df0bdf1bff224e8c50a5a14c4eb4f1f81c8a26c7708122cc46ba0da90; untracked tests_ci/test_django_full_target.py SHA-256 6104c8114f3b3dff4ccc792125441416bc3d5a41d3c6b4f76188e4ac64b6af11.
• Candidate scope is exactly 15 paths: 14 modified tracked paths plus tests_ci/test_django_full_target.py.
• Engineer frozen plan file SHA-256 c1de10aa4accbb3e711381f1d83d6dc9ed0532cac8fba78483eb709f362ba7b9; selection SHA-256 141ab0abd85ea8ea9e30c4b2f508e9611848f71613f7cf41defe36e9e52899ee.
• Tester plan: .tmp/verification-213-rebase-tester/verification-plan.json, file SHA-256 d1ef3d6aed5f0a7e1dd21aaaaf343e037661d6c67bd9b730c9b0a2ab07bae25a; canonical plan SHA-256 3e39eccc41ab7225906b8c85bfef3db8221c2c851cdd852d0fe52ab841166885; report plan SHA-256 matches it. Semantic plan comparison is identical after removing only each generated created_at.
• Graph SHA-256 e0850384988f84551c0091e4fd5dd62e823038fa36f9d4d6fc0ea0ec6d207c2e, policy 5; source-manifest SHA-256 24bc89aef9210e6623e1872c04981b9221ac2d3ce5c26db279be1b1fcff5d4af (current manifest remained unchanged after testing); 5 direct nodes, 0 downstream nodes; profile full, reason dependency_toolchain; render impact false.

Four-bucket tester plan

• rerun: selector, quality, django, playwright, container, evidence_validation
• reused: none
• skipped: none
• not_applicable: compatibility (outside_impact_closure), content invariants (outside_large_content_closure), screenshots (render_inputs_unchanged_without_reusable_visual)

Fresh plan-driven envelope

Command:

VERIFY_PLAN=.tmp/verification-213-rebase-tester/verification-plan.json VERIFY_EVIDENCE_DIR=.tmp/verification-213-rebase-tester/evidence VERIFY_REPORT=.tmp/verification-213-rebase-tester/verification-report.json VERIFY_ISSUE=213 VERIFY_WORKTREE=backlog-213-rebase-tester VERIFY_CONSUMER=tester VERIFY_PHASE=tester VERIFY_PRODUCER_ROLE=tester make verification-run

Terminal result: exit 0; report .tmp/verification-213-rebase-tester/verification-report.json, SHA-256 a9ca6613dfae25043fe3d4f8c77a3f12f36e2ab56f5ac29f6c77c7404c2b82a3; verdict=success, invalid_evidence=[], six rerun results successful, no pending entries.

Component	Result/counts	Evidence ID	Key artifact SHA-256
selector	success; 1 assertion, 15 changed paths, 5 direct, 0 downstream	c88f6d9f0ec5e151f8396f25a3d8d716b81a44a1bed1cd22a63e2d53f905b2a9	selector plan d1ef3d6aed5f0a7e1dd21aaaaf343e037661d6c67bd9b730c9b0a2ab07bae25a
quality	success; 594 passed, 0 failed, 0 skipped	e93e543e420af825c6d0b45724a6d0efb10ca445892793372783b6e6a1572010	output 84bdadb59d6b08af60891ac1d80cee03b10882fa59882e26a6e7e6053fae1c2f
django	success; 2,304 run, 2,295 passed, 9 skipped, 0 failed	21ab7b7830d60f3a806e4520f3fb2305e005e88fcc29331c873d3e5ef7a4c785	output 5032b8183e5db1a7b8a7ba62dc8535b447b079ffd7fdf8702e50f66cfc7ed188
playwright	success; 203 passed, 28 deselected; attempted/passed/failed/rerun/quarantined 203/203/0/0/0	336bf55309b02dd4c096113142e6de044a7046553c4d3fd1784658ac2ed80136	output c036e5b3bb2dc1cf8f87a16ede6cbb39dde160adba5b2558c31cb65a305b4c53
container	success; 13 assertions, 0 failed	35922a046342f1a4b30dae00299dbc091d493d336380ec3c1cf2a4977d06153d	container-check.json 0ba63c264de0042cd584c91dc98fb6d370e72f005e24642c3f352f8e6b54bdba
evidence validation	success; make test-ci, 594 passed, 0 failed, 0 skipped	e58080cca19f8be54725d45a0e8b97c2fa2ba003f2dd370a1b0c028efe61506f	output 0853aa44712922da36ee6a0c9c3aa8687cc154cc415a39b06ef6d173b1e8ff7c

All six envelopes independently passed validate-evidence-directory and validate-report.

Compatibility, planner, runner, evidence, and regression checks

• Focused [P3] Stop double-running compatibility tests in full CI #213 contracts: uv run --frozen pytest ci/tests/test_database_portability.py ci/tests/test_workflows.py tests_ci/test_django_full_target.py tests_ci/test_evidence.py tests_ci/test_runner.py tests_ci/test_verification.py -q — 370 passed; log SHA-256 3b046d5e2d9f0b0abfccf93a85c52c74e363b090f50d2c2d53976de8ae70a2c5.
• [P0] Root-cause and fix the Playwright harness deadlock #206 lifecycle, A failed component's artifact collection must not mask its failure envelope #202 failed-artifact envelopes, and [P1] Establish an explicit Playwright flake policy #208 flake/marker policy: uv run --frozen pytest test_support/tests/test_playwright_harness.py tests_ci/test_failed_artifact_recording.py tests_ci/test_flake_policy.py test_support/tests/test_marker_registry.py -q — 55 passed; log SHA-256 e4bfd4d3302f7d10f3cb8c99a483bc13ec7c488f3e9981d91c877f89e4a5e347.
• [P1] Establish an explicit Playwright flake policy #208 factory/runtime/safety/marker target: make test-factories — 139 passed; quarantine policy: 231 deselected, 0 selected, complete=1. Logs SHA-256 0adf6a9e472763c6b937b70e5d7931534f030997659b222c86f758138d1c5ca2 and 60b3f4853a9b1e40e493414354a9fc985174bea71bbeca8a052b19bad1aef8f4.
• Migration guard: make test-migrations — 16 passed; log SHA-256 c200e726933631f38bcbcc8a3810dc5e2205a35c5daea511577afa3be5e6943c.
• Portability guard exact-one-of probe: allowlist exactly make test / make test-django-full; both accepted once; zero, duplicate, and unknown full-Django command cases all rejected. Log SHA-256 66041808fe64c85b24ea31afa480359564c3077a94876e0d0d170023c7549efc.
• Invalid planner command make test-unknown rejected with exit 2; log SHA-256 172e0443a6b5566186d6cbc4bd7aafaeda3bf4f1d6fcd31835472043b1316d42.
• Planner/runner/evidence facts: default full command and full runner command are make test-django-full; Django evidence allowlist is exactly make test, make test-django-full, make test-ci-focused; the versioned quality target list remains unchanged.
• Push full workflow uses make test-django-full; quality compatibility steps are conditional on compatibility_mode == rerun, and the push quality/Django artifacts contain zero compatibility command executions. Scheduled planner override and Django job both retain make test; scheduled plan .tmp/verification-213-rebase-tester/scheduled/verification-plan.json validates with SHA-256 836bdddd9281592e0c4352977a371e9f768e0eed236b9b268fbf6bf079374156.
• Actual local aggregate DTC_TEST_RUN_ID=issue-213-rebase-tester-local make test passed compatibility 305 tests exactly once, then Django 2,304 run / 2,295 passed / 9 skipped; log SHA-256 93c2cbbecb36bd0354f0aad6644d05844562ad5a366950efe8c76747cc9c13b1.
• Dry-run contract confirms local make test has one compatibility plus one Django command, while make test-django-full has zero compatibility plus one Django command; log SHA-256 c38dea6148441d996ffb7961a3ed4e1fbc4e5067a243530dd227505e7672995f.

Screenshots

Not applicable. This is an internal CI/process-only change with no product-page render impact; the plan has render.impact=false and no required captures. No screenshot is pending.

All issue acceptance criteria and the rebase handoff contracts pass. Tester disposition: PASS. PM re-acceptance remains the next lifecycle gate; the worktree stays uncommitted and no merge/push/PR was performed.

[alexeygrigorev]

Product Acceptance

ACCEPT

I completed the PM gate for the rebased candidate using the engineer handoff and independent tester PASS.

• Base and scope: remote origin/main currently resolves to 52f943d21997bf37926f4aee89ea27f5a39d431a; candidate HEAD equals that SHA and remains detached and uncommitted. The candidate tracked diff SHA-256 is 89fbfe0df0bdf1bff224e8c50a5a14c4eb4f1f81c8a26c7708122cc46ba0da90, matching the tester reproduction. The worktree has exactly 14 modified tracked paths plus the documented added tests_ci/test_django_full_target.py path: exactly 15 paths, with no other non-ignored files.
• Issue contract: make test-django-full runs the existing full Django recipe without compatibility; local make test remains compatibility-inclusive; push full Django uses only the new target; scheduled full regression explicitly retains make test; planner, runner, ownership, evidence-allowlist, portability-guard, and workflow contracts agree; and the quality-contract-v1 target list is unchanged.
• Verification evidence: the tester report is verdict=success with invalid_evidence=[]. make test-ci passed 594 tests; quality passed 594; Django ran 2,304 tests with 2,295 passed and 9 skipped; Playwright passed 203 with 28 deselected and flake policy 203/203/0; container verification passed 13 assertions; and selector verification passed. All six evidence envelopes and the report validation passed, with no reused, required-skipped, or pending components. Compatibility is recorded using the current schema’s non-required not_applicable disposition (outside_impact_closure); the workflow runs compatibility only when compatibility_mode == rerun, so push Django does not execute it.
• Product impact: this is an internal CI/process change only; render impact is false and screenshots are correctly not_applicable.

All issue acceptance criteria are satisfied. The candidate is accepted for the next lifecycle step. No commit, push, merge, or issue close was performed.