[ooaklee]

This is a test-only carrier; do not merge it. It packages the isolated GPI event-pointer ordering candidate on the 7.2.5 refresh from #51 so touchscreen and pen behavior can be qualified before further feature groups advance.

The candidate is now 7.2.5-jg-0sp11v1, restarting the SP11 counter for the new upstream patch series. The branch name retains the original v24 label to preserve this PR and its history; it is not the current package version. The old v24 bundle is retained as historical, uninstalled evidence and must not be used for the next trial.

Relative to #51, only drivers/dma/qcom/gpi.c and debian.qcom-x1e/changelog change. There is no Golden audio, TX DMIC, camera repair, battery-provider or platform-profile topic in this candidate. Existing base features remain present.

Exact current source

• Upstream: e57ec2987d4540fa89280d349f79c4d9bcd7cf27 (7.2.5-jg-0).
• Refresh dependency: [QUALIFICATION - DO NOT MERGE] Refresh SP11 beta onto rebased 7.2.5 #51 at bdae86ee3b4ef317df96b0e2e1b3a77c15b9a221.
• GPI port: f9c68b9c62f651652bfe6cbe7de0f1c18effd63e, identical stable patch ID to dmaengine: qcom: gpi: order event ring pointer reads #45's bd9f5e29b2a8.
• Current candidate: e89ea031c208450ca56f342346d74bff0eb809b1.
• Current tree: 031f1497ee56ca03c86caa8e3a9c82339824d21a.
• Debian version: 7.2.5-jg-0sp11v1.
• ABI derived from the actual packaging rules: 7.2.5-jg-0sp11v1-qcom-x1e.

The last commit changes the changelog only. A real rebuild is in progress in separate work/output directories; existing package files are not being renamed. The v1 bundle has not yet completed artifact qualification and is not ready to install.

Version and tooling checks

• Linux-container evaluation of the repository's packaging rules confirms the version, revision, ABI, do_dtbs=true and do_stubble=false.
• Actual Debian comparisons confirm 7.2.0-jg-0sp11v23 < 7.2.5-jg-0sp11v1 < 7.2.5-jg-0sp11v24. Neither 7.2.5 candidate is installed on the qualification machine; its shared boot-support package is still the v23 version. Renumbering an already-installed same-series v24 would require separate downgrade handling.
• The clean producer remains Lexr v0.5.0-rc.4, revision fceefddeef16f5cff04aba7e72950692bd89a20a.
• Lexr's status, installer, boot-support and release-preparation unit tests pass. Its build/installer ABI parsers accept the proposed identity.
• OpenCode/GLM 5.1 supports this scoped renumber and rebuild, not hardware qualification. Current Lexr userspace/hardware diagnostics have no 7.2.5 compatibility policy. Its camera capture command also has a patch-line-blind minimum-generation gate; the FullIO release-pairing range accepts neither v24 nor v1. These limitations remain explicit; no gates were weakened and no userspace components were replaced.

Historical v24 artifact evidence — not a v1 build result

The earlier source 0f8b4e536e9cec37a308405eb844266285c5dbe7, tree 476c0974945478b54fdaf50afb9251a5ccdc4d1a, completed a full Lexr build and local inspection at version 7.2.5-jg-0sp11v24:

• Five checksummed packages: image, modules, ABI headers, common headers and lexr-kernel-boot-support.
• All 7,816 modules had the v24 ABI; required input/audio modules and touch/pen configuration were present.
• Source mode resolved to external-required, with zero embedded DTBs. Both Denali DTBs matched the independent focused builds: OLED 5aa8a21becb67f2f119960be271b86f6f471d37b71a8d45a8d97e7f712ac0325; LCD 2f621e90ee735762cea50bab5762118f7a990093d5f93c2387c5be5cd3075d2a.
• Variant-specific boot-support paths/digests/selectors, executable lifecycle hooks and exact-ABI image/support triggers passed structural checks. This did not prove installed-system convergence.
• Relocated and runtime-only bundle inspection passed; runtime selection retained boot support.
• Historical v24 boot image SHA-256: 1caddb591a3f5b3ed4df665baa538d81bb690e5063a219ee0ee757339bac9cd0. Do not reuse this digest as the v1 image identity.

Nonfatal upstream/toolchain warnings were retained. These artifacts were never a public kernel release or a hardware pass. Version-sensitive package/module/boot-support checks must be repeated on the completed v1 bundle.

Physical gate

1. Finish the v1 build and verify exact source provenance, package checksums and identities, every module ABI, configuration, packaged DTBs and boot-support lifecycle. Pin the collector to the actual new image digest; repeat integrity checks after transfer.
2. Independently verify the known-good v23 fallback, package health, boot entries and recovery availability. Use a persistent remote session. Run current Lexr preflight and install dry-run, review the plan, then verify the actual installation receipt and boot doctor before any reboot.
3. Before a supervised candidate boot, establish candidate-only speaker isolation without changing the fallback: desktop mute did not prevent earlier static. This is an input-only trial, not an audio test. Boot without screen contact and capture exact kernel/DTB identity, input driver state and persistent kernel logs.
4. Under supervised testing, check first contact, sustained touch, idle/recovery, pen interaction and repeated suspend/resume. Stop on an Oops, new taint, unexpected sound, input failure or freeze.
5. Reconfirm the known-good fallback after the trial. Keep older v10/[QUALIFICATION - DO NOT MERGE] SP11 v10 jg-1 touch test #50 evidence separate: it concerns the 7.2.2 base. X1P/LCD and non-SP11 GPI-client testing remain separate gates.

The refresh and GPI candidate remain unqualified on hardware. v23's working 7.2.0 result is the control; the upstream SPI-core fix alone is not proof that the historical touchscreen freeze is resolved. Close this packaging carrier without merging once its evidence has served topic qualification.

[ooaklee]

September 29 qualification result — hold remains

The installed and booted candidate was 7.2.5-jg-0sp11v1-qcom-x1e from this PR at e89ea031c208450ca56f342346d74bff0eb809b1, despite the historical v24 branch name. Installation succeeded with the fix now merged as ooaklee/lexr.sh#72. The user reported sustained touch/pen success, but suspend hard-failed and required a forced power cycle. The prior boot log recorded PM: suspend entry (deep), then filesystem sync completion and no resume completion; a blocked probe worker was present before suspend and recurred after reboot. Causation remains unresolved.

ALSA reported no sound cards. This input-only build does not contain the held TX/Golden audio stack, so it is not a full audio-parity test or a hardware failure verdict on that untested port. The user returned to working v23, which remains the verified default and recovery kernel.

The six refreshed topic refs and dependency mapping are now published in the refresh checkpoint. All existing PR heads are unchanged. Do not repeat failed suspend or audio stimuli unattended, merge this packaging carrier, or treat the positive input result as complete hardware qualification.

[ooaklee]

Closing this superseded review/qualification vehicle without merging. Current review work is tracked in ooaklee/linux_ms_dev_kit-sp11#51, ooaklee/linux_ms_dev_kit-sp11#61, ooaklee/linux_ms_dev_kit-sp11#68; ooaklee/linux_ms_dev_kit-sp11#68 is the single combined 7.2.5 qualification candidate. This does not discard the original feature intent or close any outstanding acceptance gate.

The current combined source is 8ba2b75713feb5c72bbc11dd8cce4048e23cdeea, ABI 7.2.5-jg-0sp11v8-qcom-x1e; see the experimental v8 release. Ordinary-use results on one X1E/OLED unit are recorded, but the release remains experimental and the detailed hardware, lifecycle, deviation, CI and provenance gates in ooaklee/linux_ms_dev_kit-sp11#68 remain open.

This PR's branch and exact head e89ea031c208, authored history, discussion and test evidence are retained unchanged. Closing is administrative supersession, not a merge, an acceptance decision, or a claim that the old and current trees are identical. Beta and the repository default are unchanged.

The historical branch label contains v24, but this PR's later package identity was 7.2.5 v1. Its failed deep-suspend result remains recorded here. The old v1 head is not an ancestor of v8; the GPI source was replayed and separately checked. Source-port correspondence must not be confused with identical commit ancestry.

The two additional deferred GPI corrections are explicitly deferred, not removed from an accepted fix list. V8's existing working GPI/touch implementation remains intact; the known residual risks and separate width-cast disposition are not waived.