Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
50 changes: 50 additions & 0 deletions .github/workflows/cbz-publisher-cas.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,50 @@
{
"name": "CBZ publisher atomic receipt controls",
"on": {
"pull_request": {
"paths": [
"tools/current_site_publication.py",
"tools/current_site_data.py",
"tests/test_current_site_publication.py",
"tools/cbz_site_publisher_cas_mutations.py",
"tools/cbz/reference_crawler_mutation_check.py",
".github/workflows/cbz-publisher-cas.yml"
]
}
},
"permissions": {
"contents": "read"
},
"jobs": {
"controls": {
"runs-on": "ubuntu-24.04",
"timeout-minutes": 15,
"steps": [
{
"uses": "actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1",
"with": {
"ref": "${{ github.sha }}",
"persist-credentials": false
}
},
{
"name": "Run publication atomic-head controls",
"shell": "bash",
"run": "set -euo pipefail\nmkdir -p \"$RUNNER_TEMP/cbz-publisher-cas\"\ntest \"$(git rev-parse HEAD)\" = \"$GITHUB_SHA\"\npython3 --version > \"$RUNNER_TEMP/cbz-publisher-cas/python.txt\"\nfor mode in normal optimized; do\n log=\"$RUNNER_TEMP/cbz-publisher-cas/unit-$mode.log\"\n if [ \"$mode\" = optimized ]; then\n python3 -B -O -S -m unittest -v tests.test_current_site_publication > \"$log\" 2>&1 || { cat \"$log\"; exit 1; }\n else\n python3 -B -S -m unittest -v tests.test_current_site_publication > \"$log\" 2>&1 || { cat \"$log\"; exit 1; }\n fi\n cat \"$log\"\n grep -Eq '^Ran 40 tests in ' \"$log\"\n tail -n 1 \"$log\" | grep -Fxq 'OK'\n if grep -Fq ' ... skipped ' \"$log\"; then echo 'skipped publication control' >&2; exit 1; fi\n if [ \"$mode\" = optimized ]; then\n python3 -B -O -S tools/cbz_site_publisher_cas_mutations.py --output \"$RUNNER_TEMP/cbz-publisher-cas/mutations-$mode\" | tee \"$RUNNER_TEMP/cbz-publisher-cas/mutations-$mode.log\"\n else\n python3 -B -S tools/cbz_site_publisher_cas_mutations.py --output \"$RUNNER_TEMP/cbz-publisher-cas/mutations-$mode\" | tee \"$RUNNER_TEMP/cbz-publisher-cas/mutations-$mode.log\"\n fi\ndone\n"
},
{
"name": "Retain publication control evidence",
"if": "${{ always() }}",
"uses": "actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9",
"with": {
"name": "cbz-publisher-cas-${{ github.run_id }}-${{ github.run_attempt }}",
"path": "${{ runner.temp }}/cbz-publisher-cas",
"if-no-files-found": "warn",
"include-hidden-files": true,
"retention-days": 30
}
}
]
}
}
}
2 changes: 1 addition & 1 deletion .github/workflows/cbz-site-recovery.yml
Original file line number Diff line number Diff line change
Expand Up @@ -52,7 +52,7 @@
{
"name": "Run recovery controls",
"shell": "bash",
"run": "set -euo pipefail\nmkdir -p \"$RUNNER_TEMP/cbz-site-recovery-controls\"\ntest \"$(git rev-parse HEAD)\" = \"$GITHUB_SHA\"\npython3 --version > \"$RUNNER_TEMP/cbz-site-recovery-controls/python.txt\"\nfor mode in normal optimized; do\n log=\"$RUNNER_TEMP/cbz-site-recovery-controls/unit-$mode.log\"\n if [ \"$mode\" = optimized ]; then\n python3 -B -O -S -m unittest -v tests.test_cbz_site_recovery.SiteRecoveryTests tests.test_cbz_site_recovery_workflow.RecoveryWorkflowTests > \"$log\" 2>&1\n else\n python3 -B -S -m unittest -v tests.test_cbz_site_recovery.SiteRecoveryTests tests.test_cbz_site_recovery_workflow.RecoveryWorkflowTests > \"$log\" 2>&1\n fi\n grep -Eq '^Ran 34 tests in ' \"$log\"\n tail -n 1 \"$log\" | grep -Fxq 'OK'\n if grep -Fq ' ... skipped ' \"$log\"; then echo 'skipped recovery control' >&2; exit 1; fi\n if [ \"$mode\" = optimized ]; then\n python3 -B -O -S tools/cbz_site_recovery_mutations.py --output \"$RUNNER_TEMP/cbz-site-recovery-controls/mutations-$mode\" > \"$RUNNER_TEMP/cbz-site-recovery-controls/mutations-$mode.log\" 2>&1\n else\n python3 -B -S tools/cbz_site_recovery_mutations.py --output \"$RUNNER_TEMP/cbz-site-recovery-controls/mutations-$mode\" > \"$RUNNER_TEMP/cbz-site-recovery-controls/mutations-$mode.log\" 2>&1\n fi\ndone\n"
"run": "set -euo pipefail\nmkdir -p \"$RUNNER_TEMP/cbz-site-recovery-controls\"\ntest \"$(git rev-parse HEAD)\" = \"$GITHUB_SHA\"\npython3 --version > \"$RUNNER_TEMP/cbz-site-recovery-controls/python.txt\"\nfor mode in normal optimized; do\n log=\"$RUNNER_TEMP/cbz-site-recovery-controls/unit-$mode.log\"\n if [ \"$mode\" = optimized ]; then\n python3 -B -O -S -m unittest -v tests.test_cbz_site_recovery.SiteRecoveryTests tests.test_cbz_site_recovery_workflow.RecoveryWorkflowTests > \"$log\" 2>&1 || { cat \"$log\"; exit 1; }\n else\n python3 -B -S -m unittest -v tests.test_cbz_site_recovery.SiteRecoveryTests tests.test_cbz_site_recovery_workflow.RecoveryWorkflowTests > \"$log\" 2>&1 || { cat \"$log\"; exit 1; }\n fi\n grep -Eq '^Ran 34 tests in ' \"$log\"\n tail -n 1 \"$log\" | grep -Fxq 'OK'\n if grep -Fq ' ... skipped ' \"$log\"; then echo 'skipped recovery control' >&2; exit 1; fi\n if [ \"$mode\" = optimized ]; then\n python3 -B -O -S tools/cbz_site_recovery_mutations.py --output \"$RUNNER_TEMP/cbz-site-recovery-controls/mutations-$mode\" > \"$RUNNER_TEMP/cbz-site-recovery-controls/mutations-$mode.log\" 2>&1 || { cat \"$RUNNER_TEMP/cbz-site-recovery-controls/mutations-$mode.log\"; exit 1; }\n else\n python3 -B -S tools/cbz_site_recovery_mutations.py --output \"$RUNNER_TEMP/cbz-site-recovery-controls/mutations-$mode\" > \"$RUNNER_TEMP/cbz-site-recovery-controls/mutations-$mode.log\" 2>&1 || { cat \"$RUNNER_TEMP/cbz-site-recovery-controls/mutations-$mode.log\"; exit 1; }\n fi\ndone\n"
},
{
"name": "Retain recovery control evidence",
Expand Down
8 changes: 8 additions & 0 deletions BASELINE.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,13 @@
# CBZ baseline ledger

## Normal Pages publisher atomic-head candidate — 10 November 2026

The original publisher's ancestor-rollback test failed in local and hosted
runs. A stacked source candidate uses an atomic expected-head non-force update
and passed the scoped hosted controls described in STATUS. It is not yet a
published recovery, authenticated platform result, mathematical proof or final
baseline. The actual red and intermediate failures remain preserved.

## Pages failure and guarded recovery boundary — 10 November 2026

The two attempts of site run `42737966287` failed during deployment and did
Expand Down
44 changes: 44 additions & 0 deletions DECISIONS.md
Original file line number Diff line number Diff line change
Expand Up @@ -755,3 +755,47 @@ Author-family correction1 moves only the two runner.temp bindings into prepare-s
the recovery tool/test blobs unchanged. Workspace #69 still pins older
Main `31bc4c1f`; actual Pages recovery and fresh current checks remain
open. No mathematical or scientific claim is changed.

## 2026-11-10 — Guard normal publication with an atomic expected head

- Preserve the actual local normal/optimized rollback failures, the hosted
229-test run with one failure, the first candidate's synthetic-token error
and its stacked recovery-control failure. Their exact sources and hosted
logs are retained under `cbz/evidence/site-publication-cas-20261110/`.
- Replace the normal publisher's REST ref patch in this candidate with fixed
repository/ref `updateRefs`, exact `beforeOid` and `force: false`.
Retain immutable object verification, bounded no-retry mutation and ref
readback. Unknown write outcomes must be resolved from the actual ref.
Adapt the recovery wrapper to the same interface.
- At tested merge `d243e1a2c833e5039b9cd0d1ceb9aa5682c78e35`, hosted
controls ran 40 tests without skips in each mode, 11 clean-baseline and
designated-oracle mutations per mode, and refused three unauthorized
assertion origins and one runtime error per mode. Artifact digest and
readback limits are recorded in STATUS. Authenticated ref mutation and site
recovery were not executed by this pull-request control.
- This candidate changes no mathematical statements, proofs, quarantines or
scientific status. Normal required-check integration and fresh source/ref
observations remain necessary.

## 2026-11-10 — Compose publisher candidate after Main #799

- Retain the complete recomposed recovery decisions prefix, including the
entire Main #799 ledger, then append the exact publisher decision suffix.
The scoped hosted controls predate this composition; rerun required checks
and compare exact source and merge trees before integration.
- Workspace #69's Main pin is stale at this observation. No receipt write,
public deployment, final gate or scientific acceptance follows.

## 2026-11-10 — Compose publisher after Main #803 and recovery

- Preserve the complete recovery/#803 decisions prefix and the original
publisher suffix; retain the #803 workflow, dependency identities and plan.
All publisher and recovery source/test blobs retain the prior tested
identities. Fresh required checks must qualify the recomposed tree.
- Scoped hosted readback job `129618998455` checked publisher artifact
`13734195859` and recovery artifact `13733836323`: exact ZIP digests,
unique members and CRCs, tested-source hashes, unit outcomes and
selected mutation origins in both modes. The initial readback failures
`129618510111`, `129618705245` and `129618849837` remain preserved
on the selected-text custody candidate. These observations do not perform
an authenticated receipt write or establish a final gate.
15 changes: 13 additions & 2 deletions GOVERNANCE.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,16 @@
# CBZ-0 governance

## Normal Pages publisher atomic-head boundary — 10 November 2026

A candidate replaces the normal publisher's REST ref patch with an atomic
expected-head, non-force update of the fixed receipt ref. Immutable Git
object verification and readback remain required, and an unknown write outcome
requires inspection of the actual ref. Its scoped commands, hosted observations
and limits are in [STATUS](STATUS.md#normal-pages-publisher-atomic-head-candidate--observed-10-november-2026)
and [TOOLING](TOOLING.md#normal-pages-publisher-atomic-head-controls--10-november-2026).
Authenticated platform behavior and publication success are not inferred.
This operation has no mathematical or scientific effect.

## Guarded Pages receipt recovery — 10 November 2026

The recovery path is limited to a canceled prepared Pages candidate with a
Expand All @@ -11,8 +22,8 @@ recovery observer. A platform failure or disagreement stops the operation;
if a write's readback fails, inspect the actual ref before inferring its state.
Workflow concurrency coordinates
the normal publisher and recovery, not external actors or atomic public reads.
The normal publisher's separate REST ref write does not yet have this atomic
expected-head guarantee.
At Main `31bc4c1f`, the normal publisher's separate REST ref write lacked
this guarantee; the later candidate above addresses it subject to integration.
The [status](STATUS.md#guarded-pages-receipt-recovery--observed-10-november-2026)
and [tool registry](TOOLING.md#guarded-pages-receipt-recovery--10-november-2026)
give exact commands, observations and limits. This is neither mathematical nor
Expand Down
7 changes: 7 additions & 0 deletions PROOF_INDEX.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,12 @@
# CBZ proof and quarantine index

## Publication atomic-head candidate — 10 November 2026

Receipt ref safety and hosted control results are operational provenance, not
mathematical proof or quarantine disposition. The original publisher failure
and scoped correction are recorded in STATUS and TOOLING. This entry closes no
proof and claims no scientific acceptance.

## Pages receipt recovery boundary — 10 November 2026

Pages publication receipts and their guarded recovery are operational
Expand Down
61 changes: 58 additions & 3 deletions STATUS.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,60 @@
# CBZ-0 execution status

## Normal Pages publisher atomic-head candidate — observed 10 November 2026

The normal publisher at Main `31bc4c1f5b66e92bfbf327366956944ea691289c`
had an ancestor-rollback race: after its ref read, a non-force REST update
could still accept a descendant if the ref had moved back. The new adverse
test failed in both local Python modes, and hosted current-site job
`129610695154` ran 229 tests with one failure; its optimized step was not
reached. The original test tree and full job log are retained under
`cbz/evidence/site-publication-cas-20261110/`. The first correction attempt
also produced one hosted fixture error and a separate recovery-control
failure; their logs remain in that evidence directory.

The subsequent stacked candidate replaces the normal publisher's REST ref
patch with GitHub's atomic `updateRefs`, binding the exact `beforeOid`,
the new commit, the fixed receipt ref and `force: false`. It retains immutable
object checks and post-write readback; a failed response can have an unknown
remote outcome, so inspect the actual ref. Hosted controls job `129613200124`
at tested merge `d243e1a2c833e5039b9cd0d1ceb9aa5682c78e35`, tree
`dafca516877750f5e06603729e862ca9e7420b7a`, ran 40 tests without skips per
Python mode. It also ran 11 clean-baseline/designated-oracle mutations and
rejected three unauthorized assertion origins plus one runtime error per mode.
Its artifact `13733117677` was reported with SHA-256
`8c84b2b1fb2a9ced407ac72e7994a973d54313bbcbb6509e171587c5780822b3`;
the full hosted job log is retained, but the artifact was not independently
unpacked in this observation. Stacked recovery controls job `129613199744`
also passed; its recovery-dispatch job was skipped on the pull request.
These are scoped checks, not actual authenticated ref mutation or full G0.

The later recomposed tested merge `cd913c55ea6ff6187fedb6f5102ce0356aa38e24`
had passing publisher and recovery component jobs. Independent readback job
`129618998455` checked the actual ZIP digests and CRCs, tested-source hashes,
unit logs and selected mutation-origin records from artifacts `13734195859`
and `13733836323`. It observed 40 publisher tests and 11 mutation pairs,
and 34 recovery tests and 37 mutation pairs, in each Python mode, with the
recorded adverse-origin rejections. Three earlier inspection failures remain
preserved. These results apply to that tested merge and do not qualify this
new whole composition or any live Pages write.

From Main, after the candidate is present, rerun:

```sh
python3 -B -S -m unittest -v tests.test_current_site_publication
python3 -B -O -S -m unittest -v tests.test_current_site_publication
CBZ_PUBLISH_NORMAL=$(mktemp -d)/mutations
CBZ_PUBLISH_OPTIMIZED=$(mktemp -d)/mutations
python3 -B -S tools/cbz_site_publisher_cas_mutations.py --output "$CBZ_PUBLISH_NORMAL"
python3 -B -O -S tools/cbz_site_publisher_cas_mutations.py --output "$CBZ_PUBLISH_OPTIMIZED"
```

The source and hosted result are a candidate until normal guarded integration.
Authenticated GraphQL and Pages status, actual recovery and current publication
remain unobserved. The Main `31bc4c1f` site build failed its prepared-receipt
readback and skipped deploy; that job's original log is retained. No final
G0–G11, no-skip federation regression or CBZ-0 is inferred.

## Main pytest component merge — observed 10 November 2026

Main #803 merged as `97663b478822848db5262b2fffcbacd37d6231b4`, with
Expand Down Expand Up @@ -58,9 +113,9 @@ recovery is not evidence of success; inspect the actual ref if write readback
fails because the remote write may already have taken effect. The normal publisher
and recovery dispatch share a workflow concurrency group, but external actors
and public reads are not atomic. Platform denial or an unavailable status is
not bypassed. The existing normal publisher's REST non-force ref update does
not provide the same atomic expected-head guarantee; its correction remains
open. The source-bound local checks below ran in both modes; hosted
not bypassed. At Main `31bc4c1f`, the normal publisher's REST ref update did not provide
this atomic expected-head guarantee. The later candidate above addresses it
subject to its own checks and integration. The source-bound local checks below ran in both modes; hosted
checks and an actual recovery must be observed separately.

From Main, run the focused tests and designated mutation campaign:
Expand Down
37 changes: 35 additions & 2 deletions TOOLING.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,38 @@
# CBZ mission-tool registry

## Normal Pages publisher atomic-head controls — 10 November 2026

`tools/current_site_publication.py` verifies immutable receipt objects and,
in this candidate, advances the fixed receipt ref through GitHub GraphQL
`updateRefs` with the exact expected old commit and `force: false`. It
rejects direct REST ref patches and retains bounded readback and unknown-outcome
handling. The recovery wrapper uses the same publication interface while
maintaining its own checks. This scope does not authenticate future GitHub
responses or establish public deployment.

From Main, run:

```sh
python3 -B -S -m unittest -v tests.test_current_site_publication
python3 -B -O -S -m unittest -v tests.test_current_site_publication
CBZ_PUBLISH_NORMAL=$(mktemp -d)/mutations
CBZ_PUBLISH_OPTIMIZED=$(mktemp -d)/mutations
python3 -B -S tools/cbz_site_publisher_cas_mutations.py --output "$CBZ_PUBLISH_NORMAL"
python3 -B -O -S tools/cbz_site_publisher_cas_mutations.py --output "$CBZ_PUBLISH_OPTIMIZED"
```

Hosted tested merge `d243e1a2c833e5039b9cd0d1ceb9aa5682c78e35` ran 40
tests without skips per mode, 11 designated mutation pairs per mode, and
rejected three unauthorized assertion origins and one runtime error per mode.
The tracer is scoped to trusted synchronous unittest. The corresponding
artifact's reported digest and the full job log are identified in STATUS;
its ZIP was not independently unpacked here. The later recomposed artifact
`13734195859` was independently read back in job `129618998455`: digest,
CRCs, tested-source hashes, unit results and selected mutation origins were
checked in both modes. The accompanying recovery archive `13733836323` was
also checked. These are scoped component observations, not final G0.
Earlier real red and candidate failure logs are retained. This is not a full auditor, final G0 or proof.

## Guarded Pages receipt recovery — 10 November 2026

`tools/cbz_site_receipt_recovery.py` narrowly reconciles a canceled prepared
Expand Down Expand Up @@ -33,8 +66,8 @@ failures per mode, and three subject/helper/lifecycle-origin refusals per mode.
Its existing traceback recorder is scoped to trusted synchronous unittest.
The original 28-test/28-mutation campaign missed a rollback race; its adverse
test and the earlier failed fixture and mutation runs remain in contribution evidence.
The normal publisher's separate REST ref update still lacks the atomic
expected-head guarantee and is not qualified by these recovery checks.
The normal publisher's REST ref update at Main `31bc4c1f` lacked this
atomic expected-head guarantee; the later candidate above has separate checks.
These checks do not prove an authenticated Pages status is available or that a
real dispatch succeeded; reads and external actor state are not atomic. They
are not a full tool audit, final G0 or mathematical certification.
Expand Down
Loading
Loading