Skip to content

fix(ci): move mirror and scorecard off standards 75bd5994 (Mirror startup_failure since 2026-11-09) - #688

Merged
hyperpolymath merged 1 commit into
mainfrom
claude/mirror-standards-pin
Nov 17, 2026
Merged

hyperpolymath merged 1 commit into
mainfrom
claude/mirror-standards-pin

Conversation

@hyperpolymath

Copy link
Copy Markdown
Owner

Why

Mirror to Git Forges has ended in startup_failure on every push to main since 2026-11-09 (first at 1c22722, 34 runs), while mirror.yml has been unchanged since 2026-11-01. It is the only red workflow on main.

Diagnosis. The cause is supported by evidence below, but the failure message itself could not be read from this session.

  • At 75bd5994, standards' own actions.lock records the callee's edge as dtolnay/rust-toolchain@e2a55d2f with ref: 'v1'.
  • Upstream's v1 tag now resolves to 683b588c (git ls-remote https://github.com/dtolnay/rust-toolchain).
  • GitHub validates the callee lock at the called SHA (recorded in scripts/lib/stale-reusable-pins.tsv, Window B).
  • Of this repo's three 75bd5994 callers, mirror-reusable is the only one whose closure includes dtolnay. That explains why hypatia-scan and scorecard still start at the same pin.
  • Standards HEAD has already re-locked to 683b588c.

The startup_failure annotation text is not exposed by the APIs available to this session. Confirmation is the first Mirror run on main after merge, because Mirror runs only on push and dispatch, not on pull_request.

What

File Change
.github/workflows/mirror.yml, scorecard.yml 75bd5994 → b49a33e9 (standards main, 2026-11-17)
scripts/lib/pinned-reusable-sources.json Re-attested both sources: commit, blob, content SHA-256, direct uses. Added a standards@b49a33e9 parent record. Added dtolnay/rust-toolchain@683b588c (v1) and ossf/scorecard-action@fa7117db (v2.5.0) records, taken from standards' lock and checked against upstream tags. Shrank the 75bd5994 parent to its remaining three sources. Removed the two records this change orphans.
scripts/lib/lock-closure-receipts.tsv Both groups re-emitted with audit-lock-closure.sh --emit-receipt, not hand-edited. Provenance comment updated.

Interface checks against 75bd5994:

  • mirror-reusable adds only three optional R2 secrets, an R2 job gated on vars.R2_BUNDLE_MIRROR_ENABLED, and a visibility-guard job with permissions: {}. That guard is a privacy improvement: forge pushes now also require a public repo or an explicit *_PRIVATE_CONFIRMED.
  • scorecard-reusable keeps the same inputs, secrets and per-job permissions.

Deliberately not in this PR: hypatia-scan stays at 75bd5994. Its newer reusable changes the scanner pins (Hypatia, panic-attack, pons) and the reporting, so it changes what is found and needs its own review.

Validation (local replay of the Pinned source closure CI job, with PINNED_SOURCE_CACHE set)

  • pinned-reusable-sources-test.sh: 55 passed, 0 failed
  • audit-pinned-reusable-sources.sh --fetch-sources: findings=0, not-run=0, verified-roots=6 (also on the historical-scanner tree)
  • scanner-caller-contract-test.sh: passed
  • lock-closure-test.sh: 47 passed, 0 failed
  • audit-lock-closure.sh: fatal=0 review=0 accepted=3 not-run=0 source-attested-records=14
  • pinned-source-entrypoints-test.sh: 9 passed, 0 failed
  • check-lock-sync.sh: rc 0

Follow-up, estate-wide (not done here)

About 20 or more hyperpolymath repos call mirror-reusable; code search finds e.g. modshells, echidna, verisimdb and panic-attack. Any still pinned at 75bd5994 would hit the same failure. Once this PR's first main run confirms the diagnosis, a stale-reusable-pins.tsv row (mirror-reusable@75bd5994 → b49a33e9) lets fix-stale-reusable-pin.sh and Hypatia PIN-003 repair them. I have not added that row yet, because it would act on a diagnosis that is still unconfirmed.


Generated by Claude Code

Mirror to Git Forges has ended in startup_failure on every push to main
since 2026-11-09 (1c22722), with mirror.yml unchanged since 2026-11-01.

At 75bd5994, standards' own actions.lock records the callee's
dtolnay/rust-toolchain edge as `ref: 'v1'` -> e2a55d2f. Upstream's v1 tag
now resolves to 683b588c (git ls-remote), and GitHub validates the
callee lock at the called SHA (scripts/lib/stale-reusable-pins.tsv,
Window B). mirror-reusable is the only one of this repo's three 75bd5994
callers that pulls in dtolnay, which is why hypatia-scan and scorecard
still start at the same pin. Standards HEAD re-locked to 683b588c.

Move mirror and scorecard to standards b49a33e9 and re-attest the
pinned-source closure with the repo's tools:
- pinned-reusable-sources.json: new commit, blob, content SHA-256 and
  direct uses for both sources; a parent record for b49a33e9; records
  for dtolnay/rust-toolchain@683b588c (v1) and
  ossf/scorecard-action@fa7117db (v2.5.0), from standards' lock and
  checked against upstream; the 75bd5994 parent record shrinks to its
  remaining sources; the two records this change orphans are removed.
- lock-closure-receipts.tsv: both groups re-emitted with
  audit-lock-closure.sh --emit-receipt.

Interface checks: mirror-reusable only adds three optional R2 secrets
and a permissions:{} visibility-guard job; scorecard-reusable's inputs,
secrets and per-job permissions are unchanged.

hypatia-scan stays at 75bd5994 on purpose: its newer reusable changes
the scanner pins and what gets reported, so it gets its own change.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01J4QvDhFeexmokLqLRfZNce
@sonarqubecloud

Copy link
Copy Markdown

@hyperpolymath
hyperpolymath marked this pull request as ready for review November 17, 2026 04:35
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 17, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ❌ Failed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-17T04:37:24.727125Z cd53ea4 Draft marked ready
🔒 Security Review ❌ Failed 2026-11-17T04:35:11.108840Z
We couldn’t complete this review. Comment @codex security review to try again.
cd53ea4 Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

Repository owner deleted a comment from coderabbitai Bot Nov 17, 2026
Repository owner deleted a comment from chatgpt-codex-connector Bot Nov 17, 2026
@hyperpolymath
hyperpolymath merged commit af2bd27 into main Nov 17, 2026
73 checks passed
@hyperpolymath
hyperpolymath deleted the claude/mirror-standards-pin branch November 17, 2026 04:48
hyperpolymath added a commit that referenced this pull request Nov 17, 2026
…fixes red Pinned source closure) (#698)

## Why

`main` (dec1e7c) is red on "Pinned source closure":
`audit-pinned-reusable-sources.sh --fetch-sources` reports `FATAL
SOURCE-ACTIVE-MISSING`. The probable cause in the task (#688 moving
mirror/scorecard) was **not** the cause — mirror and scorecard were
already attested at b49a33e. The actual drift is `hypatia-scan.yml` job
`scanner-credential`, bumped by #686 to
`hyperpolymath/hypatia/.github/workflows/credential-read-check.yml@29a9e4a8…`
while the manifest still binds `@b93f0615…`. Both commits carry the same
blob (`15c597e…`).

## What

Fallback only: attestation for the **current** pins, **no standards
bump**.
- `scripts/lib/pinned-reusable-sources.json`: commit binding b93f0615 →
29a9e4a8 (blob and content SHA-256 unchanged, re-verified from fetched
objects).
- `scripts/lib/lock-closure-receipts.tsv`: receipt row and comment moved
to 29a9e4a8.
- No workflow, `actions.lock`, SPDX or `.a2ml` change.

## Evidence (baseline origin/main → after; gawk installed locally, fresh
cache each)

| Check | Baseline | After |
|---|---|---|
| `audit-pinned-reusable-sources.sh --fetch-sources` | findings=1
(SOURCE-ACTIVE-MISSING) | findings=0 not-run=0 verified-roots=6 |
| `audit-lock-closure.sh` | fatal=1 review=5 accepted=2 not-run=7 |
fatal=0 review=0 accepted=3 not-run=0 source-attested-records=13 |
| `check-lock-sync.sh` | rc=0 | rc=0 |
| pinned-reusable-sources-test | 55/0 | 55/0 |
| lock-closure-test | 45 pass / 2 fail | 47/0 |
| pinned-source-entrypoints-test | 7 pass / 2 fail | 9/0 (after the CI
historical-source fetch step) |
| prepared-migration `preflight.sh --quiet` | — | rc=0 |
| `STANDARDS_DIR=<standards @3b57589d> tests/fleet-report-back-test.sh`
| — | 69 passed, 0 failed |

## Not done

- **The four-pin bump to standards 3b57589 is blocked.**
`governance-reusable.yml` at 3b57589 contains job-level nested reusable
calls (`deed-conformance.yml@bfa851f…` line 436,
`k9-conformance.yml@bfa851f…` line 445). The verifier deliberately
returns NOT-RUN `SOURCE-LOCAL` for nested reusables, so the governance
pin cannot be attested without extending the verifier's reviewed profile
— out of scope. `hypatia-scan-reusable` at 3b57589 also adds
`actions/download-artifact@9000827…`, absent from the current
receipts/lock closure, so it needs a lock regeneration (tool-owned,
never hand-edited) plus review. Needs an owner decision on the verifier
profile first.
- **actionlint and mikefarah yq v4 were not run**: installation (go
install, release binaries) was denied by this session's permission
classifier. No workflow file is touched here, but actionlint is
unverified. The "Regenerate vendor lock" CI step (`gh actions-lock`) was
not run locally either.
- Compat-branch hypatia pin (75bd599) left as is.

---
_Generated by [Claude
Code](https://claude.ai/code/session_01J4QvDjMjpyixLqzirvwV1s)_

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

<!-- coderabbit-ai-generated:v1 -->

## Summary by CodeRabbit

* **Chores**
  * Updated the pinned workflow source to a newer revision.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Co-authored-by: Claude <noreply@anthropic.com>
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.

3 participants