Merge pull request 'picture: no-op codec_set_controls case for VAProfileAV1Profile0' (#10) from noether/picture-av1-noop into master

Reviewed-on: marfrit/libva-v4l2-request-fourier#10
This commit is contained in:
2026-05-20 19:07:12 +00:00
+24
View File
@@ -318,6 +318,30 @@ static VAStatus codec_set_controls(struct request_data *driver_data,
return VA_STATUS_ERROR_OPERATION_FAILED;
break;
case VAProfileAV1Profile0:
/*
* AV1 has no codec-specific V4L2 control dispatch wired up
* yet on this branch (see config.c VAProfileAV1Profile0
* comment). For the daedalus_v4l2 daemon path that's fine:
* AV1 frames are self-describing per-frame (OBU sequence +
* frame headers carry everything libavcodec needs), so the
* bitstream in the V4L2 OUTPUT buffer is sufficient — no
* V4L2_CID_STATELESS_AV1_* controls have to be populated.
*
* Per-codec dispatch in request_switch_device_for_profile
* has already retargeted (video_fd, media_fd) to
* video_fd_daedalus (or video_fd_vpu981 on RK3588 if
* present) by the time we get here; the OUTPUT buffer will
* be queued via that fd and the kernel forwards bytes to
* the daemon as a regular REQ_DECODE. No-op is the
* correct shape.
*
* When the vpu981-targeted V4L2_CID_STATELESS_AV1_* dispatch
* lands from the av1-iter1 operator branch, replace this
* with av1_set_controls(...).
*/
break;
default:
return VA_STATUS_ERROR_UNSUPPORTED_PROFILE;
}