V4L2 decoder does not correctly implement VIDIOC_G_FMT and VIDIOC_S_FMT

[EwanRoycroft]

According to the V4L2 documentation, VIDIOC_G_FMT is exclusive to capture devices, and you can use this method with other drivers to get information about the output format; yet yami limits it to V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE.

The documentation also says of VIDIOC_S_FMT that:

"Drivers should not return an error code unless the type field is invalid,"

and that, after setting the format as requested:

"the VIDIOC_S_FMT ioctl returns the current format parameters as VIDIOC_G_FMT does."

I have found that this is not the case with yami. If certain fields are left empty, it will throw an error instead of filling them with the current values.

[xuguangxin]

yes. we are not fully aligned with spec. the spec original designed for capture or output device. some items may not reasonable to decode or encoder.
Are you expect this VIDIOC_S_FMT/VIDIOC_G_FMT behavior in our application?
thanks

[EwanRoycroft]

I have a program that uses the V4L2 driver on the Raspberry Pi to decode video. Assuming other drivers follow the spec, it should work cross-platform; but it fails on yami when making calls to VIDIOC_G_FMT / VIDIOC_S_FMT.