streamer(vulkan): prioritize capable dGPUs and fix AppImage GLib clash

[quadsteel]

Summary

Resolves Vulkan hardware initialization failure, Wayland syncobj teardown crash, and AppImage driver loading conflicts on Linux hybrid GPU (Intel/NVIDIA) systems:

1. Intelligent Vulkan Device Selection & Scoring (native/opennow-streamer-platform-linux):

   • Replaces unconditional device index 0 binding with capability scoring.
   • Prioritizes discrete GPUs (DISCRETE_GPU) over integrated GPUs (INTEGRATED_GPU).
   • Evaluates PRIME render offload variables (__NV_PRIME_RENDER_OFFLOAD=1, DRI_PRIME=1) and user overrides (OPENNOW_VK_DEVICE_INDEX, OPENNOW_VK_VENDOR_ID).
   • Verifies queue isolation (GRAPHICS, VIDEO_DECODE_KHR, COMPUTE, TRANSFER) and iterates over candidate devices with graceful fallback if a candidate fails initialization.

2. Clean Teardown for Wayland Surfaces (LinuxVulkanGraphics.cpp, vulkan_device.rs):

   • Calls vkDeviceWaitIdle() before instance destruction and buffer release, ensuring sync objects and timeline semaphores (wp_linux_drm_syncobj) are drained cleanly before teardown.

3. AppImage GLib/GObject Host Exclusion (qt-build.yml, qt-release-candidate.yml):

   • Excludes libglib-2.0.so.0, libgobject-2.0.so.0, libgio-2.0.so.0, and libgmodule-2.0.so.0 from linuxdeploy bundling to prevent symbol collisions (undefined symbol: g_string_copy) when loading host VA-API drivers.

Verification

• Live Hybrid GPU Host (Intel UHD 630 + NVIDIA RTX 2060):
  ./build/opennow-qt/opennow-qt --smoke-test --allow-multiple-instances --reduced-motion
  • Correctly auto-selected NVIDIA GeForce RTX 2060 (vendor 0x10de, device 0x1f15).
  • Codecs and queue isolation initialized successfully (h264, h265).
  • Clean shutdown with exit code 0 without syncobj assertion or SIGSEGV.
• Unit & Integration Tests:
  • cargo test --manifest-path native/opennow-streamer/Cargo.toml --workspace: 183 passed, 0 failed.
  • cargo test --manifest-path native/opennow-core/Cargo.toml: 248 passed, 0 failed.
  • ctest --test-dir build/opennow-qt -R opennow-embedded-orchestration-tests -V: 98 passed, 0 failed.
  • python3 -m unittest opennow-qt/tests/test_ci_workflow.py opennow-qt/tests/test_bundled_deb.py: 21 passed.
  • npm run locales:check: passed.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 497816a4-f119-42c4-81cc-f4008f0ce0c0

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[capy-ai]

Thanks for this — the hybrid-GPU scoring, vkDeviceWaitIdle teardown, and AppImage GLib exclusion all check out, and I verified the approach locally.

I couldn't push to your fork from this environment, so I opened #992 carrying your commit verbatim (authorship preserved) plus a small fix-up commit on top:

• The two new scoring tests mutated process-wide env vars, which flaked when cargo ran them on parallel threads — scoring now takes an explicit preferences value with a thin from_env wrapper, and the tests exercise the pure function.
• cargo fmt for the touched package (CI's formatting check was failing on Linux/macOS), a clippy::unnecessary_sort_by cleanup, and two small Rust nits (redundant unsafe, CString::unwrap → graceful fallback).

Once #992 goes green I'll merge it and close this one as superseded. Your commit stays yours in history.

[capy-ai]

Superseded by #992, which carries your commit verbatim plus the CI fixes — merged into dev. Thanks again for the fix!