Skip to content

[Intel Mac P11/packages] Consolidate model-specific Cirrus audio packages - #557

Draft
andrew-boyd wants to merge 4 commits into
omacom:masterfrom
andrew-boyd:intel-mac/p11-cirrus-audio
Draft

andrew-boyd wants to merge 4 commits into
omacom:masterfrom
andrew-boyd:intel-mac/p11-cirrus-audio

Conversation

@andrew-boyd

@andrew-boyd andrew-boyd commented Sep 20, 2026 •

Copy link
Copy Markdown

[Intel Mac P11/packages] Consolidate model-specific Cirrus audio packages

Consolidate the existing CS8409 and 12-inch MacBook audio recipes as separate, conflicting hardware tracks. The earlier VM-observed 12-inch header-dependency repair is retained; this validation adds no new audio-driver implementation.

Draft consolidation for source review and hardware-owner follow-up. Known defects remain; no hardware or merge-readiness certification is claimed. Part of the 14-feature Intel Mac series with three package companions.

Contribute: open follow-up PRs in andrew-boyd/omarchy-pkgs with base intel-mac/p11-cirrus-audio. Use [Intel Mac P11/packages] in the title and a contributor branch such as intel-mac/p11/describe-change. Link this feature in the series index. Merging a follow-up into this branch updates its roll-up.

This existing draft: #557.

Full Intel Mac scope and current revision

The project covers all 64-bit Intel Macs: pre-T1, T1 and T2. MacBookPro13,3 is only the first available physical test machine. Each feature has its own hardware boundary; one model's success does not validate the rest.

Candidate head: 4cf695f8c4864ce40b064e691d85b8c2c6ae1b82; reconciled upstream: e40f5a5da918e8b923115ebb4c335b6c4828fefe. Published contribution commits and author/co-author metadata remain in the history.

Ready for attended physical testing; Mac hardware remains untested by this effort. The reconciled combined ISO passed fresh unencrypted/encrypted VM installations, desktop/password-lock checks, strict Neovim startup/theme checks, synthetic Apple EFI preservation through a disk wipe, and affected package/DKMS/migration checks. ISO SHA-256: 9e6288ef5cc1eadce519cdd7384c8030286eb938a6288eb218f025a61b433252.

Exact sources, results, retained failures and reproduction. These results cover the pinned combined image; standalone-versus-combined differences remain explicit. Full inventory and retained work tracks.

Scope and dependencies

Preserve separate CS8409, MacBook CS4208 and iMac CS4208 tracks. The two DKMS package tracks conflict on one machine. Source closures do not erase iMac coverage or the MacBookPro13,2 speaker report.

The inventory now records 188 source/context PRs: 101 have dispositions in feature tables and 87 remain only in separate work tracks. This is not 188 implemented PRs or 188-to-14 compression. The 14 feature PRs and three package companions remain the published coordination targets; retained backlog is not claimed complete.

Every affected source PR and recommended action

Current open/closed/merged state is administrative history, not proof that two patches preserve identical behavior. Included/partial identifies selected contributions; alternatives, deferred work and context are not automatically ISO content.

Source and author State Treatment and rationale Recommended action
leifliddy/macbook12-audio-driver#60 — @doctor merged context: Merged driver preparation/DKMS changes underpin package #153. Its preparation files match, while the older pinned Makefiles differ; the package does not contain every change in this driver PR. Already merged; retain evidence and credit. No action requested.
omacom/omarchy-pkgs#153 — @doctor open included: Selected MacBook CS4208 package foundation with the existing recorded matching-kernel preparation follow-up. It conflicts with the separate CS8409 package on one system; do not install both model tracks together. Consider closure only after maintainers accept preserved behavior and hardware coverage.
omacom/omarchy-pkgs#155 — @shawnyeager open superseded: CS8409 package predecessor credited through #249; preserved independently, not an additional companion PR. Maintainers may assess the named successor before closure.
omacom/omarchy-pkgs#182 — @gjinali open alternative: Camera/audio package bundle for omacom/omarchy#7732 with offline source preparation and post-transaction DKMS verification. Compare its delivery guarantees with the selected package companions; the whole bundle is not incorporated merely because package names overlap. Retain for comparison; do not stack competing implementations automatically.
omacom/omarchy-pkgs#249 — @j7j7 open included: Selected CS8409 package foundation, with the previously documented offline preparation/build compatibility work. Review the exact recipe and built archive receipts rather than assuming the source PR is byte-identical to the integrated package. Consider closure only after maintainers accept preserved behavior and hardware coverage.
omacom/omarchy#10458 — @inspiretelapps open included: Selected iMac CS4208 in-tree configuration, with later preservation of administrator settings and audio routes. No new speaker/mixer tuning was inferred without hardware. Consider closure only after maintainers accept preserved behavior and hardware coverage.
omacom/omarchy#11064 — @ozdadirri open alternative: Exact Touch Bar-model/AUR CS8409 installer with a reported MacBookPro13,2 speaker test; upstream closed omacom/omarchy#11625 in its favor. Keep that evidence and model-gating rationale, while retaining the selected pinned package and wider iMac track. The claim that every 13,1/14,1 has a different codec is not adopted as a blanket exclusion. Retain for comparison; do not stack competing implementations automatically.
omacom/omarchy#11623 — @andyholst closed alternative: Combined NVRAM/audio proposal. Its calibration applies a shared file by PCI ID and can retain the embedded default MAC when no interface identity is available; audio uses live git clone/pull and stock linux-headers. Retain the signal/audio reports, but prefer the selected model/subsystem calibration paths and pinned packaged audio delivery. The bundle is not fully superseded by either single roll-up. Already closed; retain evidence and credit. No action requested.
omacom/omarchy#11625 — @andyholst closed alternative: Unpinned direct-clone CS8409 delivery with differing installer/migration gates. Already closed; retain evidence and credit. No action requested.
omacom/omarchy#12286 — @teomurgi open alternative: Complete conflicting MacBook proposal includes automatic EFI writes and freeze sleep policy; preserve separately. Retain for comparison; do not stack competing implementations automatically.
omacom/omarchy#6921 — @sunoxen-lab open included: MacBook CS4208 baseline combined with the existing linked mixer/current-kernel upgrade contribution; uses package #153. Consider closure only after maintainers accept preserved behavior and hardware coverage.
omacom/omarchy#7140 — @shawnyeager closed superseded: CS8409 predecessor credited through omacom/omarchy#9516; narrower model coverage retained as comparison. Already closed; retain evidence and credit. No action requested.
omacom/omarchy#7732 — @gjinali open alternative: Complete MacBookPro13,1 camera/audio/SPI transition bundle with reported hardware validation, dependent on package #182. Preserve as a broader model-specific alternative; no one feature roll-up supersedes all of its coverage. Retain for comparison; do not stack competing implementations automatically.
omacom/omarchy#8285 — @jack-lech open alternative: Different CS8409 -git package name and installer/cache ownership; not stacked. Retain for comparison; do not stack competing implementations automatically.
omacom/omarchy#9516 — @j7j7 closed included: Selected CS8409 package-backed installer including its distinct iMac track. Later exact-model/header checks preserve custom configuration. Upstream closure in favor of omacom/omarchy#8285 does not erase the wider model evidence or source credit. Already closed; retain evidence and credit. No action requested.
sunoxen-lab/omarchy#1 — @doctor open partial: Existing mixer/service/current-kernel upgrade work included. Its separate s2idle proposal is retained as an unselected alternative. Keep uncovered work open; this feature alone does not fully supersede it.

Validation and attribution

Required fresh-image checks are satisfied, with qualified reruns. The full source suite passed 273 files with two disclosed skips, but repeated NetworkManager Polkit failures during that suite temporarily locked its test account. The following password/desktop/recovery checks failed there and passed separately in a clean overlay of the same ISO; the original aggregate remains failed. The first build attempt also failed one libfprint USB-emulation test; unchanged-source isolated repeats, two complete suites and normal package checks passed afterward. That intermittent build-test failure remains documented, and its cause is not established. VM fixtures and DKMS builds do not establish physical device behavior. Physical Wi-Fi, audio, camera, Touch Bar, fingerprint, graphics and suspend behavior remain unverified by this effort.

Original source histories remain linked above. Source-authored commits already included in the roll-ups remain ancestors of this revision; additional copied contributions receive native attribution. Original attribution mapping. No original PR is edited or closed by this effort.

…ages

Consolidate existing source contributions for draft review.
Original implementation authors are retained as author/co-authors; the committer assembled and reconciled the candidate. No contributor approval or sign-off is implied.
Known defects and excluded alternatives remain documented in the PR body.

Source-PR: omacom#249
Source-PR: omacom#153
Source-Commit: omacom#2499eb62fe
Source-Commit: omacom#2495816153
Source-Commit: omacom#249a346e20
Source-Commit: omacom#153d9715a7

Co-authored-by: J7 <31825378+j7j7@users.noreply.github.com>
Co-authored-by: Jake Pillai <jakepillai@me.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
@macintog

Copy link
Copy Markdown

Reviewing dd91732c and the pinned driver preparation scripts, I found two delivery prerequisites that the package recipes do not yet cover:

  • Kernel-source preparation at DKMS time: both packages invoke preparation scripts through PRE_BUILD that can download kernel source when it is not already cached. Having the packages in an offline ISO is therefore insufficient by itself. The MacBook CS4208 script checks the archive against a SHA-256 value from sha256sums.asc, but does not verify that manifest's signature. The pinned CS8409 script downloads and extracts the archive over HTTPS without a digest or signature check. Could the offline delivery path explicitly prepare the required source cache, with authenticated verification of those inputs before extraction?
  • Matching kernel headers: pkgbuilds/macbook12-audio-driver-dkms/PKGBUILD hard-depends on linux-headers. On another selected kernel, that installs stock headers without satisfying the selected kernel's build prerequisite. Could the package and companion installer agree on a matching-header requirement rather than assuming the stock linux package?

The existing cache paths mean an offline build can work when the correct sources have already been prepared; the gap is establishing that prerequisite for installation and subsequent kernel updates. I inspected the recipes and preparation scripts, but have not run a fresh package or DKMS build for this revision.

@gjinali

gjinali commented Oct 18, 2026

Copy link
Copy Markdown

Hi @andrew-boyd. I have a physical MacBookPro13,1 (A1708, two-port, no T1), and I maintain #182, which packages CS8409 audio and FaceTime HD for that model. #182 and this series overlap on two packages, so here is how they differ. The maintainer can then pick one, or we can merge the ideas.

snd-hda-macbookpro-dkms

  • Offline DKMS source. This PR's recipe depends on wget and patch, because the upstream DKMS build downloads kernel source while building. Add MacBookPro13,1 camera and audio packages #182 instead prepares the HDA source at package-build time, from a pinned driver commit and pinned kernel.org archives. Installing and rebuilding then need only the target kernel's headers: no network, no $HOME, no uname.
  • Prepared source per kernel series. Add MacBookPro13,1 camera and audio packages #182 carries source for each active kernel series (currently 7.1 and 7.2), and DKMS selects it from the target kernel.
  • Post-DKMS verifier. A pacman hook verifies the module after kernel and driver transactions, for linux-omarchy and the fallback linux. Hold the reboot back when a kernel lacks a required driver omarchy#15905 uses the same verifier to hold back the reboot when a DKMS build failed.
  • Checks against linux-omarchy. The prepared headers are vanilla, and linux-omarchy patches some private HDA headers. Add MacBookPro13,1 camera and audio packages #182's offline DKMS test therefore compares the built module's BTF type layouts with each target kernel's own HDA modules. Its package check fails when a linux-omarchy patch changes a private header the module compiles. 87 shared types match on 7.2.8-9-omarchy and 7.2.11-arch1-1.

facetimehd-dkms
Both PRs build the same upstream 0.7.2. #182 pins the release archive by checksum instead of a git tag, and adds the same post-DKMS verifier.

Testing offer. The series says Mac hardware is untested so far. I can run attended tests of P08, P11 and P06 on this MacBookPro13,1 and report the results on each PR. That covers only this one model. If you'd rather absorb #182's packaging into the P08 and P11 package branches than keep two tracks, I'm happy to open follow-up PRs there.

@ozdadirri

Copy link
Copy Markdown

I can test the CS8409 package on a physical MacBookPro13,2 alongside @gjinali's 13,1. I agree with @macintog that a DKMS build that downloads kernel source won't work offline. From a user's point of view, #182's approach of preparing the source when the package is built (offline rebuilds, plus a check after DKMS runs) looks like the safer one to ship. I'm happy to test whichever recipe the maintainers choose.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

6 participants