Skip to content

fix: repair selected replay navigation and historical audit CI - #7276

Open
antunishdPursuit wants to merge 3 commits into
CrocSwap:mainfrom
antunishdPursuit:fix/selected-ci-navigation-7000
Open

antunishdPursuit wants to merge 3 commits into
CrocSwap:mainfrom
antunishdPursuit:fix/selected-ci-navigation-7000

Conversation

@antunishdPursuit

@antunishdPursuit antunishdPursuit commented Nov 11, 2026 •

Copy link
Copy Markdown
Contributor

What changes

Repair the shared CI navigation and historical audit after #7000 became selected. The guide points to the current replay and retains #6030 as history. The historical audit uses its inspected commit, preserving every source hash and all original rejection/plan controls.

The shallow planning job now checks that the snapshot controls are registered; the full-history replay job runs all four integration controls. A fresh Git-free fixture verifies that onboarding needs no historical objects. This fixes the three missing-commit errors introduced by running those integration tests during planning.

Credit Clear1nsight's #7257 for the pinned historical checkout, unpaidmoon for the original audit, and the retained replay/planner authors. The navigation repair follows #7161 and ToikY's f2206e6. This PR isolates CI maintenance from their research submissions.

Checks

  • ["python", "-B", "-m", "unittest", "discover", "-s", "tests", "-p", "test_research_historical_snapshot.py"]: passed (exit 0). unittest: 2 tests run; 0 failures; 0 errors; 0 skipped; 0 expected failures; 0 unexpected successes.
  • ["python", "-B", "-m", "unittest", "discover", "-s", "tests", "-p", "test_research_current_status.py"]: passed (exit 0). unittest: 6 tests run; 0 failures; 0 errors; 0 skipped; 0 expected failures; 0 unexpected successes.
  • ["python", "-B", "scripts/research_package.py", "--package", "research/historical-selected-replay-gap"]: passed (exit 0).

Not tested

Updated-head hosted CI is pending. No selected-family scientific cold replay or supplier proof is claimed. At original head a54cd49, the broad native run had 11 failures, 23 errors and 90 skips across 760 tests; seven representative failures also reproduce with unchanged main. Those native fixture/platform issues remain unrepaired, and the full native suite is not rerun for this scoped update. Historical checkouts use retained Git objects, not fresh network acquisition.

Prepared with substantial OpenAI Codex assistance.

Isolate the navigation repair already carried in PR7161 and reused by ToikY in f2206e6. Preserve the historical6030 checkpoint and all scientific inputs and assertions. The affected six-test status module and selected-source metadata check pass. Full native onboarding remains limited by pre-existing fixture/platform failures, disclosed in the PR.
Reuse the authenticated repair from Clear1nsight PR CrocSwap#7257. Materialize the inspected historical tree in a fresh private checkout while preserving all source hashes, 19 rejection cases and the nine-package plan. Add fresh-snapshot and corruption controls. All 10 affected tests and the registered replay pass natively; prior broad native-suite limitations remain disclosed.
@antunishdPursuit antunishdPursuit changed the title fix: align CI navigation with selected #7000 replay fix: repair selected replay navigation and historical audit CI Nov 12, 2026
Move snapshot integration controls out of shallow CI planning and into the mandatory registered replay. Keep onboarding routing checks and a Git-free fixture, preserving all four controls and their timeout.
@antunishdPursuit

Copy link
Copy Markdown
Contributor Author

CI passed at 4403cae: planning completed 762 tests with 24 skips, and the registered historical replay passed the original audit and all four snapshot controls. This supersedes the pending hosted-CI statement in the description. The existing scientific and acquisition limits still apply.

Prepared with substantial OpenAI Codex assistance.

@fluorine

Copy link
Copy Markdown
Contributor

Independent test check of this PR. It is a local check, not hosted CI.

Result: this PR fixes the test that fails CI on main. Confidence 85%.

Setup: main e54523517d0f03e58352e647c403a5af3e525a57 and this PR at head 4403caeb39779f42842b85458b1aa52d0720e62e. The merge base equals current main. CPython 3.13.9 on Windows 11, sparse checkout (tests/, scripts/, docs/, certificates/, the touched package, the workflow files). The command is the one in research-packages.yml:

python -B -m unittest discover -s tests -p 'test_research_*.py' -v

Findings

  • Main: test_research_current_status.CurrentCiNavigationTests.test_current_ci_section_matches_selected_record fails. It expects make frontier7000-verify, but docs/ci-verification.md names the Conditional κ .003125804 (+5.22e-7 over #5907): remove helper 470's frame-593 rung, the twin of #5907's removal #6030 replay. This matches the hosted failure.
  • This PR: the same test passes. All 6 tests of test_research_current_status.py pass.
  • New tests: both tests of test_research_historical_snapshot.py pass when research/historical-selected-replay-gap/ and the workflow files are present.
  • Full run: the failing sets of main and this PR are identical except for the new test file. The sparse checkout causes about 50 failures in both runs. Hosted CI on main shows only the one failure.

Not checked: Python 3.12, the other steps of the plan job, the replay job, and the full-history controls that this PR adds. Without those, I cannot say that hosted CI turns green.

Why it matters: the plan job has failed on every recent PR since 2026-11-14, so the replay job is skipped and no PR gets hosted verification.

Prepared by fluorine with substantial Anthropic Claude (Claude Code) assistance. No independent human review of this check exists yet.

🤖 Generated with Claude Code

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants