Skip to content

Install Intel video acceleration on GPUs named only Intel Graphics - #14252

Closed
massisalva wants to merge 2 commits into
omacom:quattrofrom
massisalva:fix/intel-graphics-vaapi
Closed

massisalva wants to merge 2 commits into
omacom:quattrofrom
massisalva:fix/intel-graphics-vaapi

Conversation

@massisalva

@massisalva massisalva commented Oct 4, 2026 •

Copy link
Copy Markdown

On a ThinkPad T14s Gen 6, lspci names the Arrow Lake-U GPU only as Intel Corporation Arrow Lake-U [Intel Graphics] (8086:7d41). None of the names in the detection regex in install/hardware/intel/video-acceleration.sh match it, so the install never adds intel-media-driver, libvpl or vpl-gpu-rt. With no VA-API driver, browsers and players decode video on the CPU.

Add intel graphics to the regex. The older GPUs that should keep libva-intel-driver report names such as 4th Gen Core Processor Integrated Graphics Controller or Mobile 4 Series Chipset Integrated Graphics Controller, and those still don't match.

Hardware setup only runs at install time, so existing installs on these GPUs stay without the driver. A migration installs the same packages there. omarchy-pkg-add skips packages that are already installed, so running it again does nothing.

After installing the packages on the affected machine, vainfo reports the iHD driver with H.264, HEVC, VP9 and AV1 decode.

Fixes #11958
Fixes #13987

🤖 Generated with Claude Code

Arrow Lake-U reports itself in lspci as "Intel Graphics", which the
detection regex missed, leaving video decoding on the CPU. Match that name
and add a migration for existing installs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@greptile-apps

greptile-apps Bot commented Oct 4, 2026 •

Copy link
Copy Markdown
Contributor

RetriggerView in GreptileConfidence Score: 5/5

[Medium impact] The PR appears safe to merge.

Summary

Adds support for Intel GPUs whose names contain only “Intel Graphics.”

  • Intel video setup installs drivers for GPUs named only Intel Graphics.

The previous migration finding is fully addressed. No new actionable issues were found.

Reviews (2) · Last reviewed commit: "Avoid SIGPIPE in Intel Graphics migratio..." · Reviewed by Greptile

Comment thread migrations/1791129708.sh Resolved

@chessing234 chessing234 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewed head 53f516319f6c11c3b873f808a63c473d0c8751c8.

Reviewed the added plain Intel Graphics match and the existing-user migration condition. Both paths select the same acceleration packages for that new device label. No actionable defect found in the changed matching. I did not install packages or verify VA-API on Intel hardware.

@jandrusk

jandrusk commented Oct 6, 2026

Copy link
Copy Markdown

Thanks for the clear writeup on the Arrow Lake-U naming. Heads-up that #9824 also targets this: it switches detection to "any Intel display device" and has an Arrow Lake-U [Intel Graphics] fixture. It doesn't include a migration for existing installs, though, and yours does. It might be worth coordinating so the detection change and the migration land together rather than as two competing regex fixes. #8514 (hybrid GPUs by PCI generation) is in the same area too.

Small nit on the migration: migrations run under bash -euo pipefail, and grep -q exits on the first match. If the upstream grep is still writing, it can take a SIGPIPE, the pipeline then reports failure, and the migration gets marked done without installing anything. Capturing first avoids that:

gpus=$(lspci | grep -iE 'vga|3d|display' || true)
if grep -qi 'intel graphics' <<<"$gpus"; then
  omarchy-pkg-add intel-media-driver libvpl vpl-gpu-rt
fi

@omarchybot

Copy link
Copy Markdown
Collaborator

Thanks for the clear write-up. I read the diff at 53f5163 against quattro, but haven't classified or tested it yet. Each pull request gets a classification from two models before it's tested, and the second one (Codex Medium) couldn't give one today: it is answering, but the daily limit on second-opinion reviews was already used up. So I haven't added a label or run anything on a worker. The pull request goes back in the queue to be classified and reviewed properly.

Two older pull requests change the same detection in install/hardware/intel/video-acceleration.sh, and both cover Arrow Lake-U: #8514 picks the driver by PCI device ID (its test includes 8086:7d41) and adds a migration for existing installs, and #9824 installs intel-media-driver for any Intel display device that isn't GMA. Which of the three should land is a decision for the maintainer.

Nothing is needed from you for now.

@paulogeyer

Copy link
Copy Markdown
Contributor

Compared the PRs for video-acceleration.sh skipping intel-media-driver on GPUs that lspci names only "[Intel Graphics]" (#11958, #13987). Tested on my laptop with a stub harness (fake lspci, recorded omarchy-pkg-add); stock quattro installs nothing for Meteor Lake and Wildcat Lake.

PR Meteor Lake / Wildcat Lake UHD 620 / GMA / NVIDIA-only unchanged Test
#14252 pass / pass pass no
#15094 (closed as duplicate of this one) pass / pass pass yes
#9824 pass / pass pass yes
#8514 not comparable (moves detection to a new helper) not comparable yes

#14252 is the one to merge now that #15094 was closed in its favour: the same one-word change to the name regex, and it passes every case. It has no test; the stub test from #15094 could be carried over. #9824 (class-based detection) and #8514 (PCI-generation helper) are larger rewrites of the same fix.

Suggestion: add Fixes #11958 and Fixes #13987 to #14252.

Capture the lspci output before matching so grep -q can't make the
pipeline fail under pipefail. Add a stub-based test for the install
script and the migration, adapted from omacom#15094.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@massisalva

Copy link
Copy Markdown
Author

Thanks for the reviews. Pushed two changes:

Also linked #11958 and #13987, which this fixes. Happy to coordinate with #9824 or #8514 if the maintainers prefer one of those approaches. The migration would apply on top of either.

@omarchybot omarchybot added bug Something isn't working compatibility Make Omarchy work better on everything! labels Nov 2, 2026
@omarchybot

Copy link
Copy Markdown
Collaborator

Thanks for the fix, @massisalva, and for picking up the SIGPIPE nit and the test from #15094.

Since your last push, #17150 was merged to quattro as cec1d26. It makes the same fix this pull request does: install/hardware/intel/video-acceleration.sh now matches [intel graphics, and migrations/1793452544.sh installs intel-media-driver libvpl vpl-gpu-rt on existing machines with those GPUs. I read both diffs against current quattro. Every GPU name in your test, Arrow Lake-U included, gets the same packages there at install and from the migration, and GMA and non-Intel machines are handled the same. That leaves this pull request with nothing to change for a real device. It also conflicts with quattro now, and its migration would run again on every machine that already ran 1793452544 without doing anything. Nothing was run on a worker, since the fix is already on quattro.

Two models reviewed it, Codex Medium and muse-wonder (via pi, thinking high), and both concluded the same. Neither found a real device that your looser intel graphics match catches and quattro's bracketed match misses. Neither thought the SIGPIPE in quattro's migration can happen here, since the filtered lspci output is a few hundred bytes written in one go. Codex also found that installed_for in the test ignores the script's exit status, so a script that installed the right packages and then failed would still pass. The one thing left here is the wider set of GPU names in the test. quattro's agents/skills/migrations.md says not to test migrations, so if those names are kept, they belong as install-script rows in test/shell.d/hardware-quirks-test.sh.

I've recommended to the maintainer that this be closed as fixed by #17150. #11958 and #13987 are still open, but #17150 fixes them too. Nothing is needed from you.

@massisalva

Copy link
Copy Markdown
Author

Closing as fixed by #17150, which landed the same fix on quattro. Thanks for the thorough review!

@massisalva massisalva closed this Nov 14, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bug Something isn't working compatibility Make Omarchy work better on everything!

Projects

None yet

5 participants