[ooaklee]

CI-only guardrail — held draft, no hardware-topic merge

Enable source-hygiene checks for the three exact dependency bases used by
the held 7.2.5 topic stack:

• x/sync-jg-7.2.5-20260928 ([QUALIFICATION - DO NOT MERGE] Refresh SP11 beta onto rebased 7.2.5 #51): camera [HELD 7.2.5] media: qcom: port guarded SP11 camera parity #56, TX [HELD 7.2.5] ASoC: preserve Denali TX DMIC capture #57, battery [HELD 7.2.5] platform/surface: preserve SP11 battery providers #59 and GPI [HELD 7.2.5] dmaengine: qcom: preserve GPI event-ring ordering #61.
• x/sp11-tx-7.2.5-20260928 ([HELD 7.2.5] ASoC: preserve Denali TX DMIC capture #57): Golden [HELD 7.2.5] ASoC: preserve guarded Golden protected playback #58.
• x/sp11-battery-7.2.5-20260928 ([HELD 7.2.5] platform/surface: preserve SP11 battery providers #59): profile [HELD 7.2.5] platform/surface: preserve SP11 profile and fan policy #60.

The workflow currently selects only beta-based PRs. GitHub's
PR branch filters match the target branch,
so the stacked ranges have local validation but no hosted checks.

Exact scope

One commit, 42c8a5d765b4e978e9c308dc603b09719e9cd28a, on the pinned refresh
bdae86ee3b4ef317df96b0e2e1b3a77c15b9a221. Only
.github/workflows/sp11-integration-checks.yml changes: three literal PR-base
filters and two comments. No wildcard or privileged pull_request_target.

The beta-only push filter, read-only permissions, pinned checkout,
disabled credential persistence, sparse/bounded history and existing
fail-closed aggregate/per-commit validator are unchanged. No kernel,
packaging, userspace, contributor metadata, DCO exception or source-style
allowlist is modified. The tested v4 and all held topic heads stay pinned.

Validation

• actionlint 1.7.12 passes the changed workflow; external shellcheck and
  pyflakes integrations were disabled. The validator script is unchanged.
• Structural comparison permits only the three exact added base names;
  push/permission/checkout/validator and all other YAML structure match.
• Four allowed and eight excluded literal PR-target cases pass.
• Seven deliberate policy mutations are rejected: wildcard/duplicate target,
  widened push, write permission, persisted credentials, validator bypass
  and privileged PR event.
• Exact one-commit review range passes the existing integration validator.
• A local two-parent synthetic PR merge passes automatic base/head inference
  and checks exactly one topic commit. Half-specified ranges and an ordinary
  one-parent checkout with pull_request context fail closed.
• The new staging-base PR's actual hosted workflow
  run 36892320186
  passes. The log checks synthetic merge
  0f6fd42167c6c2eae650f18fdab8582a22e902ba, with the pinned refresh as
  first parent and this policy head as second parent, and reports exactly
  one topic commit passing per-commit checks. The fetched merge tree matches
  the policy head. This verifies the new refresh-base PR trigger only.

OpenCode's Zai-Codex/GLM 5.1 review agrees with this narrow policy and held
rollout. Its review is not hardware acceptance or a second device inspection.

Rollout boundary

This dedicated draft targets the refresh dependency base, not beta, and does
not merge the refresh or any hardware topic. Its hosted check, if successful,
proves this new policy-PR path only; it does not retroactively establish hosted
coverage for #56–61. Their applicable refs must receive the reviewed policy
through an explicitly agreed rollout and a triggering event, then their exact
base/head check results must be inspected. No retargeting, source-head rewrite,
unqualified merge or false manual-HEAD success is performed here.

The v4 audio/input functional results are recorded separately in #55.
Cold-power-on, DSP/protection/runtime state, suspend, remaining feature groups,
non-Denali behavior, contributor attestation and explicit acceptance of deviations
remain open. CI success must not bypass those gates or discard integration behavior.

October 4 qualification cross-reference

The v8 qualification checkpoint closes the bounded
camera/quiet-playback Power Saver check, not this CI policy or its rollout.
#62/#64/#66 remain held pending complete reviewed rollout with workflow-write
access, including the prepared fail-closed Git-read correction. Earlier sparse
hosted checks remain limited; local complete-input results are separate.
These workflows, scripts and policy are fork-only and must not be exported
upstream. No source, policy, branch ref or draft state is changed here.

[ooaklee]

Closed as an inactive/consolidated review vehicle during the 20 October kernel queue cleanup—not merged, newly hardware-qualified or accepted for upstream by this closure.

Previously recorded v8 acceptance evidence stands unchanged; this closure only retires the review vehicle.

The original source head and contributor history remain retained. Where a closed topic branch is removed, its verified archive/pr-62-20261020 tag preserves the exact head for recovery. Released/recovery/dependency branches remain untouched.

The separate tested-source staging route and all outstanding upstream reconstruction, provenance, fork-CI and deviation/promotion gates are tracked in OE #33. Experimental reverse-engineering PRs remain separate and unmerged. Beta/default branches and installed kernels are unchanged.

Retained exact source: 42c8a5d765b4e978e9c308dc603b09719e9cd28a.
Dated consolidation record: ooaklee/linux-surface-pro-11-oe#33 (comment)