https://bugs.kde.org/show_bug.cgi?id=515044

--- Comment #4 from Rodrigo <[email protected]> ---
Hi everyone,

I would like to provide additional debugging data from Fedora 44 (KDE Plasma 6,
Qt 6.10, FFmpeg backend 8.1.2) running on an AMD Ryzen 5 3500U with Radeon Vega
8 Graphics (Raven Ridge). This hardware also lacks silicon-level AV1 decoding
support.

My report (bug #523042) was marked as a duplicate of this one, and I believe
our testing on Fedora can help pin down the root cause of this Qt Multimedia
pipeline failure.

We performed exhaustive testing trying to force a software decoding fallback
(using dav1d/libdav1d, which works flawlessly on the same machine under
GStreamer via 'gst-play-1.0'). Here are the results of the workarounds we
attempted:

1. Setting Harald's suggested variable:
   QT_FFMPEG_DECODING_HW_DEVICE_TYPES=, dragon output.mp4
   (Result: Playback still fails. The pipeline fails to fall back to CPU
decoding and throws "Failed to get pixel format" / "Get current frame error"
once hardware paths are stripped).

2. Forcing Qt6 software mode:
   QT_MULTIMEDIA_FORCE_SOFTWARE=1 dragon output.mp4
   (Result: Ignored. The backend still probes VA-API and fails: "[av1] Failed
setup for format vaapi: hwaccel initialisation returned error.")

3. Disabling VA-API driver completely:
   LIBVA_DRIVER_NAME=none dragon output.mp4
   (Result: The pipeline falls back to Vulkan, but then crashes immediately due
to lack of hardware support: "[av1] Device does not support the
VK_KHR_video_decode_queue extension! Failed setup for format vulkan")

4. Trying other Qt AVPlay wrappers:
   QT_AVPLAY_HWACCEL=none dragon output.mp4
   (Result: Ignored. Still attempts VA-API and fails).

Conclusion:
This confirms Harald's suspicion in Comment 1 and 2. The bug lies deep within
the Qt6 Multimedia FFmpeg backend. The pipeline is tightly bound to hardware
acceleration formats for AV1 streams. It completely lacks a functional,
automatic "software decoding fallback" path. Once all hardware acceleration
probes (VA-API, Vulkan) fail, the pipeline terminates abruptly rather than
falling back to CPU rendering via libdav1d.

Hopefully, this extra testing helps when escalating this issue upstream to the
Qt Multimedia/QtProject trackers.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to