marfrit/marfrit-packages
0
0
Fork 1
Code Issues 2 Pull Requests Actions Packages Projects Releases Wiki Activity

fresnel-fourier stack: bump libva to iter38b + fix firefox HW-decode defaults (#8) #9

Merged
marfrit merged 2 commits from claude-noether/marfrit-packages:noether/libva-iter38b-and-firefox-prefs into main 2026-05-15 06:13:20 +00:00
Conversation 0 Commits 2 Files Changed 3
+48 -12
claude-noether commented 2026-05-15 06:10:53 +00:00
Contributor
Copy Link
Copy Source

Two PKGBUILD-level changes for the fresnel-fourier production stack.

1. libva-v4l2-request-fourier: bump pin iter8 → iter38b (7ac934e)

Carries the fork from libva-multiplanar's iter8 close (2026-05-06) to fresnel-fourier's iter38b close (2026-05-14):

  • iter25 — rkvdec image_fmt pre-seed (HEVC frame-1 fix)
  • iter31 — VAAPI st_rps_bits → V4L2 slice_params (HEVC frame-2+ fix)
  • iter33 — VP8 uncompressed-header prepend for hantro first_part_offset
  • iter35 — MPEG-2 verified libva-correct
  • iter38 — multi-device probe so a single libva session serves rkvdec + hantro
  • iter38b — bounds check uses MAX_PROFILES (11), not MAX_CONFIG_ATTRIBUTES (10)

Required by the fresnel-fourier README which currently says the package isn't built (the README is wrong about that — the package exists, just on the iter8 pin). vainfo lists all 10 profiles in one session, byte-exact bench verified across 5 codecs.

68 commits between old pin (65969da3) and new (7ac934e0). pkgver() recomputes the rev count at build time; placeholder bumped from r280 to r348.

2. firefox-fourier: ship RK3399 HW-decode default prefs (closes #8)

Issue #8 documented that the patch series (0001..0004) builds the routing path but doesn't actually engage on a fresh profile because three other prefs are false upstream:

  • widget.dmabuf.force-enabled (panfrost EGL probe doesn't fire)
  • media.hardware-video-decoding.force-enabled (Intel-tuned cost heuristic picks SW)
  • media.ffvpx-hw.enabled (FFvpx PDM HW-capable variant gating for VP8/VP9)

Fixed by shipping a vendor-default file at /usr/lib/firefox-fourier/browser/defaults/preferences/rockchip-fourier-defaults.js. Lower precedence than user.js / about:config — power users can still disable for debugging without touching the package.

pkgrel 1 → 2.

CI impact

After merge, two builds will trigger sequentially per the workflow's needs: chain:

  1. libva-v4l2-request-fourier-aarch64
  2. firefox-fourier-aarch64 (slow — Firefox build, hours)

The mpv-fourier iter2 build (run #92) currently in progress is unaffected.

Test plan

  • CI green for libva-v4l2-request-fourier-aarch64
  • CI green for firefox-fourier-aarch64
  • pacman -Syu on fresnel pulls the new libva (iter38b) and firefox (pkgrel=2)
  • Fresh Firefox profile on fresnel — empty user.js, just media.autoplay.default=0 — autoplay H.264 → MOZ_LOG shows Requesting pixel format VAAPI_VLD + Got VA-API DMABufSurface
  • Same sweep across HEVC / VP8 / VP9 (per #8's repro matrix)

Bot: claude-noether

Two PKGBUILD-level changes for the fresnel-fourier production stack. ## 1. `libva-v4l2-request-fourier`: bump pin iter8 → iter38b (`7ac934e`) Carries the fork from libva-multiplanar's iter8 close (2026-05-06) to fresnel-fourier's iter38b close (2026-05-14): - iter25 — rkvdec `image_fmt` pre-seed (HEVC frame-1 fix) - iter31 — VAAPI `st_rps_bits` → V4L2 slice_params (HEVC frame-2+ fix) - iter33 — VP8 uncompressed-header prepend for hantro `first_part_offset` - iter35 — MPEG-2 verified libva-correct - **iter38** — multi-device probe so a single libva session serves rkvdec + hantro - **iter38b** — bounds check uses `MAX_PROFILES` (11), not `MAX_CONFIG_ATTRIBUTES` (10) Required by the [`fresnel-fourier` README](https://git.reauktion.de/marfrit/fresnel-fourier/src/branch/main/README.md) which currently says the package isn't built (the README is wrong about that — the package exists, just on the iter8 pin). vainfo lists all 10 profiles in one session, byte-exact bench verified across 5 codecs. 68 commits between old pin (`65969da3`) and new (`7ac934e0`). `pkgver()` recomputes the rev count at build time; placeholder bumped from `r280` to `r348`. ## 2. `firefox-fourier`: ship RK3399 HW-decode default prefs (closes #8) Issue #8 documented that the patch series (0001..0004) builds the routing path but doesn't actually engage on a fresh profile because three other prefs are `false` upstream: - `widget.dmabuf.force-enabled` (panfrost EGL probe doesn't fire) - `media.hardware-video-decoding.force-enabled` (Intel-tuned cost heuristic picks SW) - `media.ffvpx-hw.enabled` (FFvpx PDM HW-capable variant gating for VP8/VP9) Fixed by shipping a vendor-default file at `/usr/lib/firefox-fourier/browser/defaults/preferences/rockchip-fourier-defaults.js`. Lower precedence than `user.js` / `about:config` — power users can still disable for debugging without touching the package. `pkgrel 1 → 2`. ## CI impact After merge, two builds will trigger sequentially per the workflow's `needs:` chain: 1. `libva-v4l2-request-fourier-aarch64` 2. `firefox-fourier-aarch64` (slow — Firefox build, hours) The mpv-fourier iter2 build (run #92) currently in progress is unaffected. ## Test plan - [ ] CI green for libva-v4l2-request-fourier-aarch64 - [ ] CI green for firefox-fourier-aarch64 - [ ] `pacman -Syu` on fresnel pulls the new libva (iter38b) and firefox (pkgrel=2) - [ ] Fresh Firefox profile on fresnel — empty `user.js`, just `media.autoplay.default=0` — autoplay H.264 → MOZ_LOG shows `Requesting pixel format VAAPI_VLD` + `Got VA-API DMABufSurface` - [ ] Same sweep across HEVC / VP8 / VP9 (per #8's repro matrix) Bot: claude-noether
claude-noether added 2 commits 2026-05-15 06:10:54 +00:00
libva-v4l2-request-fourier: bump pin iter8 -> iter38b (7ac934e) 81cc050bf3 
Carries the fork from libva-multiplanar iter8 close (2026-05-06) to
fresnel-fourier iter38b (2026-05-14): multi-device probe so a single
libva session serves rkvdec (H.264 + HEVC + VP9) + hantro (MPEG-2 +
VP8) without LIBVA_V4L2_REQUEST_VIDEO_PATH overrides, plus the
MAX_PROFILES bounds-check fix from iter38b.

Required by the fresnel-fourier README's HW-decode stack (vainfo
exposes all 10 profiles in one session, byte-exact bench verified
across 5 codecs).

68 commits between the old pin (65969da3) and the new (7ac934e0).
pkgver() recomputes the rev count at build time; the static placeholder
is bumped from r280 to r348 to reflect the new pin.
firefox-fourier: ship RK3399 HW-decode default prefs (closes #8) 62a594ab59 
Three prefs that gate the patched VAAPI / V4L2-request code path are
'false' upstream and need to be 'true' for the libva-v4l2-request
backend to actually engage on RK3399 / panfrost EGL:

  widget.dmabuf.force-enabled
  media.hardware-video-decoding.force-enabled
  media.ffvpx-hw.enabled

Without them, Gecko's DMA-BUF probe silently fails (panfrost EGL
doesn't trip the right detection), the (Intel-tuned) HW-vs-SW cost
heuristic picks SW, and the FFvpx PDM never selects its HW-capable
variant for VP8/VP9. Net result: fresh profiles SW-decode despite the
0001..0004 unlock patches being applied.

Shipped as a vendor-default file at:
  /usr/lib/firefox-fourier/browser/defaults/preferences/rockchip-fourier-defaults.js

Lower precedence than user.js / about:config — power users debugging
HW decode can still flip them off without touching the package.

pkgrel 1 -> 2.

Closes: #8
marfrit merged commit 2f78136479 into main 2026-05-15 06:13:20 +00:00
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
No items
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
marfrit (Markus Fritsche)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: marfrit/marfrit-packages#9
Delete Branch "claude-noether/marfrit-packages:noether/libva-iter38b-and-firefox-prefs"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?