You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Oct 11, 2026. It is now read-only.
Repository navigation
This repository was archived by the owner on Oct 11, 2026. It is now read-only.
Aurora Linux bug roadmap: fix order, dependencies and acceptance criteria #57
Latest regular release: Aurora 2026.10.10.9. The M3 Pro desktop failure #129 is fixed and closed after OpenGL/Vulkan and desktop checks on J516S. Keep newer board-specific failures separate from that result.
Latest kernel increment:#138 aligns AV1 imported-buffer CDF tails, fixing the observed first inter-frame H9 fault. The temporary decoder module passed 12/12 bit-exact 1080p inter-frame outputs plus AV1 intra/H.264/VP9 controls on J516S (34 frames total), and the original module was restored. #134 fixes the separate malformed Wi-Fi event completion wait; nine ARM64 QEMU controls passed. The fresh kernel/headers and generated installer pass package controls; the new full kernel has not had a fresh M1/M2 boot/dock test. Existing Mesa/m1n1 profiles are retained.
Previous kernel batch:#131 hardens Rust device-coredump ownership/read buffers and ISP ring validation. #132 fixes memory-pressure receive refill #69; #133 preserves the matched installer bindings. The full ARM64 package build, 49 KUnit tests, 88 receive/lifecycle controls and 64 generated-installer checks pass, with independent artifact review. No new physical boot or dock test is claimed for this increment.
Order
Priority
Work
Completion criterion
1
P1
#135: Neo intermittent Wi-Fi association, split from #122
All eight saved timeouts have two starts, one calibration and no final report; successful channel11 connections also exist. Establish the missing firmware/worker transition before replaying calibration. #134 fixes a separate malformed-event wait and is released in 2026.10.10.9. #122 retains native-display hardware acceptance.
Kernel #138 is released in 2026.10.10.9 with targeted J516S module hardware acceptance. Reverify the original 60-frame clip and normal browser playback; retain AV1 intra/H.264/VP9 controls. This is not full AV1/10-bit qualification. Userspace #104 is source-merged and remains outside this kernel package release.
3
P1
#82 / #109: M3 Air board-specific graphics acceptance
Current14 J613 and J615 handoff and legacy packages are already shipped. Qualify J615 renderer/readback and legacy Vulkan; exact26.6.2 native J615 is a separate experimental OpenGL-only profile. J61326.6.2 has reporter readback evidence in #35.
Preserve existing sealed keys across content-changing identity-keybag saves and reboot. Unchanged-content reboot success and fingerprint enrollment do not close this failure.
Establish the board/firmware-specific backlight and night-light interface, with actual panel behavior before enabling it.
6
P2
Early boot input, DTB overrides and UKI verification
Early USB PR31 has package/runtime and actual module-image checks. DTB PR13 needs the tested repeated-occurrence repair incorporated. UKI/rekey PR17 passes current-main integration, with three known minor follow-ups. Upstream review/CI and required cold-boot evidence precede release of these boot packages.
Security boundary:#53 retains P0 installed-package acceptance priority. Its source fix exists; custom Hyprland/Aquamarine packages remain outside kernel releases. #75 is a distinct desktop capture crash with kernel causality unproved.
Neo external displays:m1n1 PR26 is blocked by the matched kernel's explicit DCPEXT refusal and the shipping J700 device tree's missing external display controller/routes. A bootloader-only change cannot provide this feature. USB-C data/charging and external display support are separate paths.
Qualification and later work: preserve Lucas's J493 Touch ID and dock hotplug/suspend results in #111 / #117. M1 Max resume #74, direct-USB-C HPD #38 and Thunderbolt link recovery #24 / #28 remain board-specific. Hibernation #99, Neo AVE #19 and broader device activation follow core failures. Omacom's M1 compute stack remains held for exact-head runtime regression evidence.
Historical roadmap and retained acceptance evidence
The entries below retain earlier reproduction, implementation and hardware evidence. Use the current priorities above for work order and the latest linked issue replies for status.
Development landing branch: omacom/linux:integration/aurora. The refreshed Omacom kernel roadmap includes the shared-clock failure in PR #73 and current hardware results. This tracker and its original reports remain available during migration; Omacom kernel issues must be enabled before individual tickets can be recreated.
Status: 2026-10-09. Latest normal release: 2026.10.09.4. The earlier 12.6 stable retains its recorded M1 Air / MBP13 hardware results; newer M3 increments have source/build acceptance, without an exact-release hardware qualification claim. The earlier 12.6 prerelease remains available with its original optional desktop packages. This roadmap orders fixes by impact; a passing run on another board does not close a different Mac’s bug. Board, firmware and actual renderer matter.
Priority labels: P0 session/security boundary; P1 broken core functionality; P2 recovery, reliability and qualification; P3 capabilities, documentation and performance. Milestones follow that order without assigning release dates. status:merged means integrated in source; release or stated hardware acceptance remains recorded below; status:fix-pending means a linked fix still needs integration; needs-retest means the specified current-version result is missing. Successful hardware reports retain their evidence when closed.
1. Fix the lock boundary first — P0
J293: compositor restart after suspend leaves the desktop unlocked (12.4) #53: Lock intent, authenticated unlock ordering and the actual UWSM restart path are fixed in the matched Omarchy source / package pair. Focused startup/unlock/rollback controls and staged package checks pass. Existing unguarded sessions require logout/reboot. Keep open for installed suspend/crash/restart and next-login acceptance; camera recovery stays separate.
2. Unblock core functionality — P1
Merged fixes remain open for the specified release and hardware retests. The persistent-key failure still needs a persistence fix; source fixes for buffer imports and display lifetime await installed hardware acceptance.
Merged in #59: selected-CRTC routing plus displaced-route rollback. Retest a lone TB display at login, through DPMS, suspend and replug on 12.6 stable. Direct USB-C was fixed in 12.1 and verified in 12.2. #45 now has matching CRTC/-EINVAL evidence and the original J316S reporter confirms 12.6 stable fixes it; closed as resolved. J314S has a distinct shared-clock blackout with a candidate in PR #73.
Legacy Honeykrisp plane offsets fixed in Mesa #8; #9 keeps the production offset control active in package builds. Host controls pass with original-code negatives. Fresh aarch64 package built and passed production offset, loader, macro, TLS and MRT controls; real Vulkan/EGL/Zink pixels remain pending. Native 25 GL-only source is unchanged.
Separate direct-USB-C port-specific HPD wake failure on M2 Max. Delivered-high-IRQ recovery merged in #67; this does not establish the cause of a missing IRQ. Retest port2 versus port1, including IRQ delivery.
PMP lookup and firmware-log ownership fixes #101 (superseding #98), and literal Limine repair #102 are released in 2026.10.09.4 after fresh ARM64 build and independent matched-package/installer admission. Corrected J613 current14 power admission reaches firmware startup, but the later setup ACK stall remains open; no all-Air GPU success claim. J613/J615 current14 handoff, D208 notification and brightness source are already released; exact board/firmware GPU readback and night-light/brightness acceptance remain separate. Native J61325 requires exact26.6.2, matched stage1 and native GL-only profile; Satya’s tested migration recipe/guide is forthcoming.
Stock J293 desktop capture can abort after resume. A separately patched Hyprland/Aquamarine pair passed the targeted retest, but custom desktop packages are excluded from the stable release. Keep open for an upstream-compatible desktop fix; kernel causality is unproved.
M1 Max / Studio Display GPU firmware crash after loaded-GPU resume. Capture the first fault and compare exact board/firmware on the stable kernel; separate from #75.
Reverify the original GPU firmware MMU fault independently of the confirmed plane-offset bug. Normal browser/sandbox settings; capture the first fault if it recurs.
One bounded startup retry merged in #61, preserving this Mac's calibration/address and draining retry work at suspend/unregister. Stock J516s boot/radio retest remains pending on 12.6 stable.
START_EP unsent-message retry and required-caller error handling merged in #64. Historical untyped timeout is not attributed; ALS readings and the earlier firmware crash remain to be checked.
Stored backtrace matches released connector entries in Aquamarine teardown. Two narrow guards and the aarch64 package pass production-function, relocation and library-load controls (package fix). Custom Aquamarine packages are excluded from 12.6 stable; installed logout/greeter exit and next login acceptance remain pending.
Restore RX refill after memory pressure exhausts all descriptors, with bounded delayed retry and safe lifecycle cancellation. Existing limitation remains separate from merged #68.
Per-Mac firmware/calibration provisioning and in-tree ANE integration; a real ANE job is required beyond probe/boot success.
Related review queue: #2 is merged through #65 (DP alt-mode lifetime), with current code and author attribution preserved; physical hotplug/resume retest pending. #22 remains blocked on its companion tile/error-handling changes. #26 is integrated through #70; #37 is integrated, with failed-probe cleanup in #71. Collection/probe success alone does not establish enrollment or a real ANE job. #22 results on M1 Pro do not establish base-M2 support. Put display details in #41, not another umbrella issue.
4. Add capabilities and documentation — P3
tests: synchronize stale SMC sampler fixture to avoid read-error race #103: synchronize the synthetic stale-SMC sampler fixture. The full 686 run had one unchanged reason assertion fail while the alternate read-error path safely stopped load; focused stale/read-error controls pass. Keep distinct refusal and cleanup assertions while removing the fixture’s missing-file window.
Feature request: supported hibernation on J413 M2 Air for long lid-closed idle #99: J413 M2 Air hibernation. Prerequisites merged in #100: Apple platform admission currently refuses hibernation; spin-table CPU park/restart and cold firmware/DMA restore must be supported before encrypted image storage, early resume and repeated power-off/recovery qualification. One successful s2idle cycle does not establish hibernation or measured battery drain.
Successful reports archived with their untested features stated; unresolved reliability/qualification issues remain open.
Previously closed #3, #8, #9 and #23 keep their recorded resolutions. Reopen a canonical bug with a new failing case instead of closing a different platform's failure on the strength of these passes. Source permalinks, reproductions and specific retest requests are in each linked issue's triage comment.
Recorded 12.6 hardware baseline
Reviewed Aurora PRs #58–#68 and #70–#72 are merged. 12.6 stable contains the exact kernel/headers tested on the M1 Air and M1 MBP13, plus matched bootloader, Mesa, libfprint and Touch ID packages. Custom Hyprland, Aquamarine and Omarchy desktop packages are excluded. The earlier prerelease remains unchanged; its optional desktop bundle is not part of the stable release. Exact source/package hashes are attached.
Both Macs passed repeated external-display power cycles, same-port replug and monitor input switching. The Air returned both pictures after suspend. The MBP restored kernel display/dock state but its stock desktop aborted during capture (#75); a separately patched desktop survived a 37-minute retest, without those patches shipping in stable. Full raw jobs retain the documented firmware/menu failures; no all-model or full-suite pass is implied. Hardware details remain in #41.
Alongside the current P1 M3 startup/installer work below, remaining implementation priorities are the shared-clock candidate PR #73, #31 keybag persistence, #69 bounded RX refill recovery and the incomplete #22 dual-stream 5K path. #74 requires idle/loaded and older-kernel comparisons before choosing a GPU resume fix. Successful boot reports #76–#78 are archived with their untested features; they do not close #74. Draft #50 needs a current fixture/layout update before integration. Hardware bugs remain open for the specific checks above; related successes do not establish a different Mac's result. Native J613 25G83 is an explicit experimental OpenGL-only profile, requiring exact 26.6.2 firmware and a matched stage 1; Vulkan25 is unavailable.
Current14 versus native25:--m3-gpu selects only a supported existing profile. System firmware alone does not choose the GPU ABI. Current14 J613/J615 and Pro keep legacy Mesa; native J61325 requires exact26.6.2 and admitted stage1, selector j613-25g83-hal200, and a GL-only native profile. Vulkan25 is unavailable.
Native25 onboarding: Satya reports successful source-built stage1 chainload and a working J61325 setup. The existing source contains config filling and ESP boot-path support, but his exact tested recipe and firmware migration/enrollment guide are still forthcoming. Internal keyboard in U-Boot/Limine remains a stated limitation. Do not infer J61525 support or copy another Mac’s calibration.
Development and inbound reports stay in this fork until Omacom enables Issues and changes its default branch. Custom Hyprland/Aquamarine packages remain excluded. Exact native25 migration instructions will follow the tested incoming deliverable rather than an inferred firmware update procedure.
Hardware qualification found two distinct P1 follow-ups: #75 (J293 Hyprland SHM capture assertion after resume/DPMS) and the newly reported #74 (J316c GPU firmware crash after loaded-GPU suspend). Preserve both separately; neither is an established 12.6 regression or a duplicate of #56.
Next order: establish capture-free versus capture-triggered resume for #75; retain #53 lock-boundary acceptance; compare idle/loaded GPU and older-kernel resume for #74. Keep the intermittent J293 required AOP-endpoint startup timeout as a separate unresolved diagnostic, without adding speculative retries. Display reconnect/DPMS evidence is in #41; full desktop-package installation and remaining model/port coverage are still pending.
Air GPU delivery update: Aurora 2026.10.09 now provides a single --m3-gpu install, reboot, and aurora-m3-gpu-check workflow for the supported current14 13-inch/15-inch Air profiles. #83 supplies the J615 panel and matched activation; m1n1 Aurora15 exports that Mac’s own identity and calibration inputs. The separate J613 25G83 profile remains OpenGL-only.
400 installer tests passed, one existing skip; packages and the live M3 Pro checker passed their checks. Remaining Air work: #80 brightness controls, plus physical confirmation of the delivered #79/#82 changes. The package is available now; physical Air feedback does not hold up delivery. Work remains on this fork while Omacom’s repository administration is pending.
M3 Air GPU priority update: #105/#106 are incorporated in merged #110 (a4d7ff4acdef), fixing J613 native25 GPU intent and RAM-dependent firmware identity. ARM64 Asahi compilation and admission controls passed; not yet released or boot-qualified.
Next delivery dependencies: matched package for the J613 fixes and the source-built stage1/own26.6.2 installation path. J615 native25 #107/#108 needs matching Mesa loader, m1n1/stage1 artifacts and assembly bindings. Current14 startup in #35 remains open.
The current roadmap is now in Omacom #56. Closing this older tracker as migrated; unresolved bugs remain open in their linked Omacom tickets. Development is moving to omacom/linux-aurora; this personal repository will be read-only, with its existing releases and historical reports retained. Please continue new findings in the linked destination.
Current priorities — 2026-10-10
Latest regular release: Aurora 2026.10.10.9. The M3 Pro desktop failure #129 is fixed and closed after OpenGL/Vulkan and desktop checks on J516S. Keep newer board-specific failures separate from that result.
Latest kernel increment: #138 aligns AV1 imported-buffer CDF tails, fixing the observed first inter-frame H9 fault. The temporary decoder module passed 12/12 bit-exact 1080p inter-frame outputs plus AV1 intra/H.264/VP9 controls on J516S (34 frames total), and the original module was restored. #134 fixes the separate malformed Wi-Fi event completion wait; nine ARM64 QEMU controls passed. The fresh kernel/headers and generated installer pass package controls; the new full kernel has not had a fresh M1/M2 boot/dock test. Existing Mesa/m1n1 profiles are retained.
Previous kernel batch: #131 hardens Rust device-coredump ownership/read buffers and ISP ring validation. #132 fixes memory-pressure receive refill #69; #133 preserves the matched installer bindings. The full ARM64 package build, 49 KUnit tests, 88 receive/lifecycle controls and 64 generated-installer checks pass, with independent artifact review. No new physical boot or dock test is claimed for this increment.
Security boundary: #53 retains P0 installed-package acceptance priority. Its source fix exists; custom Hyprland/Aquamarine packages remain outside kernel releases. #75 is a distinct desktop capture crash with kernel causality unproved.
Neo external displays: m1n1 PR26 is blocked by the matched kernel's explicit DCPEXT refusal and the shipping J700 device tree's missing external display controller/routes. A bootloader-only change cannot provide this feature. USB-C data/charging and external display support are separate paths.
Qualification and later work: preserve Lucas's J493 Touch ID and dock hotplug/suspend results in #111 / #117. M1 Max resume #74, direct-USB-C HPD #38 and Thunderbolt link recovery #24 / #28 remain board-specific. Hibernation #99, Neo AVE #19 and broader device activation follow core failures. Omacom's M1 compute stack remains held for exact-head runtime regression evidence.
Historical roadmap and retained acceptance evidence
The entries below retain earlier reproduction, implementation and hardware evidence. Use the current priorities above for work order and the latest linked issue replies for status.
Development landing branch: omacom/linux:integration/aurora. The refreshed Omacom kernel roadmap includes the shared-clock failure in PR #73 and current hardware results. This tracker and its original reports remain available during migration; Omacom kernel issues must be enabled before individual tickets can be recreated.
Status: 2026-10-09. Latest normal release: 2026.10.09.4. The earlier 12.6 stable retains its recorded M1 Air / MBP13 hardware results; newer M3 increments have source/build acceptance, without an exact-release hardware qualification claim. The earlier 12.6 prerelease remains available with its original optional desktop packages. This roadmap orders fixes by impact; a passing run on another board does not close a different Mac’s bug. Board, firmware and actual renderer matter.
Priority labels: P0 session/security boundary; P1 broken core functionality; P2 recovery, reliability and qualification; P3 capabilities, documentation and performance. Milestones follow that order without assigning release dates.
status:mergedmeans integrated in source; release or stated hardware acceptance remains recorded below;status:fix-pendingmeans a linked fix still needs integration;needs-retestmeans the specified current-version result is missing. Successful hardware reports retain their evidence when closed.1. Fix the lock boundary first — P0
2. Unblock core functionality — P1
Merged fixes remain open for the specified release and hardware retests. The persistent-key failure still needs a persistence fix; source fixes for buffer imports and display lifetime await installed hardware acceptance.
-EINVALevidence and the original J316S reporter confirms 12.6 stable fixes it; closed as resolved. J314S has a distinct shared-clock blackout with a candidate in PR #73.3. Recover and qualify — P2
Related review queue: #2 is merged through #65 (DP alt-mode lifetime), with current code and author attribution preserved; physical hotplug/resume retest pending. #22 remains blocked on its companion tile/error-handling changes. #26 is integrated through #70; #37 is integrated, with failed-probe cleanup in #71. Collection/probe success alone does not establish enrollment or a real ANE job. #22 results on M1 Pro do not establish base-M2 support. Put display details in #41, not another umbrella issue.
4. Add capabilities and documentation — P3
Addressed / archived during grooming
EBUSY/reboot requirement; safe warm reattach remains future enhancement work.Previously closed #3, #8, #9 and #23 keep their recorded resolutions. Reopen a canonical bug with a new failing case instead of closing a different platform's failure on the strength of these passes. Source permalinks, reproductions and specific retest requests are in each linked issue's triage comment.
Recorded 12.6 hardware baseline
Reviewed Aurora PRs #58–#68 and #70–#72 are merged. 12.6 stable contains the exact kernel/headers tested on the M1 Air and M1 MBP13, plus matched bootloader, Mesa, libfprint and Touch ID packages. Custom Hyprland, Aquamarine and Omarchy desktop packages are excluded. The earlier prerelease remains unchanged; its optional desktop bundle is not part of the stable release. Exact source/package hashes are attached.
Both Macs passed repeated external-display power cycles, same-port replug and monitor input switching. The Air returned both pictures after suspend. The MBP restored kernel display/dock state but its stock desktop aborted during capture (#75); a separately patched desktop survived a 37-minute retest, without those patches shipping in stable. Full raw jobs retain the documented firmware/menu failures; no all-model or full-suite pass is implied. Hardware details remain in #41.
Alongside the current P1 M3 startup/installer work below, remaining implementation priorities are the shared-clock candidate PR #73, #31 keybag persistence, #69 bounded RX refill recovery and the incomplete #22 dual-stream 5K path. #74 requires idle/loaded and older-kernel comparisons before choosing a GPU resume fix. Successful boot reports #76–#78 are archived with their untested features; they do not close #74. Draft #50 needs a current fixture/layout update before integration. Hardware bugs remain open for the specific checks above; related successes do not establish a different Mac's result. Native J613 25G83 is an explicit experimental OpenGL-only profile, requiring exact 26.6.2 firmware and a matched stage 1; Vulkan25 is unavailable.
M3 onboarding priority, 2026-10-09
--m3-gpuselects only a supported existing profile. System firmware alone does not choose the GPU ABI. Current14 J613/J615 and Pro keep legacy Mesa; native J61325 requires exact26.6.2 and admitted stage1, selectorj613-25g83-hal200, and a GL-only native profile. Vulkan25 is unavailable.Development and inbound reports stay in this fork until Omacom enables Issues and changes its default branch. Custom Hyprland/Aquamarine packages remain excluded. Exact native25 migration instructions will follow the tested incoming deliverable rather than an inferred firmware update procedure.