[satchlj]

GPT-6.1 Sol on @satchlj's behalf

Follow up the legacy handoff matching coverage from #1517, which has merged. Extract the existing always-on predicate without changing its result, then execute that production function with modeled OF nodes: exact T8122 PPT compatible/offset/label, wrong offset, wrong or extended label, generic and other SoC compatibles, and explicit apple,always-on on the PPT or an unrelated node. Property-bearing nodes must bypass the legacy compatible check. The probe's firmware-owned/inherited-on exclusions stay in place.

Register the fixture in tools/asahi/run-host-tests.py. No secondary-core throttle programming is added.

Validation on alba and arm64 neo-build: the registered PMGR host selection passes 2/2, including the new production-predicate fixtures compiled with GCC 13 and UBSan. The full host suite on integration base 7cb581e passes 224/224, including all eight formerly failing DCP/T6030/split/tunnel cases. The existing integration fixture repair already resolves those failures, so this PR adds no duplicate repair.

The local runner uses Python 3.12 / Rust 1.99 / GCC 13 / Clang 18; CI has different Python/Rust pins and runs on x86_64. The refactored PMGR driver object also compiled successfully in an arm64 kernel build with Clang 18.

No new hardware tests were performed. This is offline fixture coverage and a predicate refactor; the original PR's hardware evidence is unchanged.

[DjDeveloperr]

Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.

The extraction preserves the existing always-on predicate and its evaluation order exactly; the surrounding firmware-owned/inherited-on exclusions remain intact. The fixture executes that production function, is registered in the host runner, and covers exact versus extended labels, offsets, other SoCs and explicit-property precedence.

Independent local validation: both PMGR fixtures pass with GCC; the new PPT fixture also passes with Clang. Four additional source mutations (offset bypass, prefix label match, generic-SoC overmatch and dropping explicit-property precedence) compile and fail runtime assertions as intended. Inlining the helper reconstructs the parent production source ignoring whitespace. An optional Clang rerun of the existing bus fixture was stopped during local I/O contention; it is not counted as passed. I did not repeat the full suite, arm64 driver-object build or hardware tests.

Source approval for this follow-up. The current hosted host-tests check is still queued, so this review does not claim green CI. The prior #1517 exact-base eight-failure disposition is unchanged; this PR makes no new hardware qualification claim.

[DjDeveloperr]

Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.

Merged current integration f9f14be into this source-approved branch as7ba54f6c, preserving contributor history. The sole conflict was the first host-runner insertion: current main adds the merged #1010 MTP frame fixture and its old-transport control, while this PR adds the PMGR PPT fixture. Both are retained exactly once. Removing just the new PMGR entry reproduces the complete current-main runner byte-for-byte; no other runner entries were lost or weakened.

The PMGR driver and new fixture are byte-identical to the reviewed3ec2354b. The actual merged-source PMGR fixture passes with GCC/UBSan; earlier four compiled runtime mutation controls and the exact predicate reconstruction remain applicable to those unchanged bytes. The final net diff remains the same three source/fixture/registry paths and diff-check passes. No kernel build or hardware experiment was performed.

Existing source approval stands. Fresh host run43088401124 is in progress on7ba54f6c; the old head’s successful CI is not a new-head pass. Merge waits that result. #1517’s earlier hardware and exact-base test qualifications retain their original scope.

[DjDeveloperr]

Codex (GPT-6.1 Sol), reviewing on @DjDeveloperr’s behalf.

Approved current head 7ba54f6. The conflict notification is stale/transient: GitHub reports mergeable and an independent three-way registry merge against current main254c9bd3 is clean. Removing the single PMGR registration from the merge result reconstructs main’s runner byte-for-byte, including its newer MTP/floor cases and existing negative controls.

The PMGR production predicate and fixture remain identical to the previously reviewed/tested source; main has not changed either file. Explicit apple,always-on precedence and firmware-owned/inherited-on exclusions remain intact. Prior GCC/UBSan and four runtime mutation controls retain their exact source scope; no redundant test or kernel build was needed.

Current host workflow43088401124/job130758697503 passed, including execution of the host-test step. This is predicate/fixture coverage, not new hardware qualification or secondary-core throttle programming.