https://bugs.freedesktop.org/show_bug.cgi?id=105145
--- Comment #13 from k.philipp@gmail.com --- (In reply to Christian König from comment #12)
Unfortunately yes it is.
That is indeed very unfortunate. It complicates stuff a lot :-( I'd still like to see a solution that allows to support PAFF without copying around already-decoded frames/fields all the time. Do you have any ideas? Could we have the decoder put progressive content into progressive surfaces and interlaced content into interlaced surfaces (under the assumption that we will most likely want to deinterlace)?
That HEVC and VP9 only support progressive layout in the output format is also only logical because those formats don't support interlaced content.
At least HEVC does support interlaced content in theory, but as far as I understood it the (hw) decoder does not need special support for that. How would that work then, would we get top/bottom field as two separate quasi-frames?