rkmppdec improperly judging interlaced source

[tc]

Summary of the bug:

​https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/rkmppdec.c#L407

frame->interlaced_frame = ((mode & MPP_FRAME_FLAG_FIELD_ORDER_MASK) == MPP_FRAME_FLAG_DEINTERLACED);

my understanding of frame->interlaced is that it should be non-zero when input frames are interlaced.
mpp_frame_get_mode(mppframe) returns a bitfield covered by mpp_frame.h:

#define MPP_FRAME_FLAG_TOP_FIELD        (0x00000001)
/* bottom field only */
#define MPP_FRAME_FLAG_BOT_FIELD        (0x00000002)
/* paired field */
#define MPP_FRAME_FLAG_PAIRED_FIELD     (MPP_FRAME_FLAG_TOP_FIELD|MPP_FRAME_FLAG_BOT_FIELD)
/* paired field with field order of top first */
#define MPP_FRAME_FLAG_TOP_FIRST        (0x00000004)
/* paired field with field order of bottom first */
#define MPP_FRAME_FLAG_BOT_FIRST        (0x00000008)
/* paired field with unknown field order (MBAFF) */
#define MPP_FRAME_FLAG_DEINTERLACED     (MPP_FRAME_FLAG_TOP_FIRST|MPP_FRAME_FLAG_BOT_FIRST)
#define MPP_FRAME_FLAG_FIELD_ORDER_MASK (0x0000000C)

For a typical tff interlaced source, returned mode is 0x7 (0b111). For bff, it would be 0b1011.

However, MPP_FRAME_FLAG_DEINTERLACED masks both TFF and BFF, which is impossible in interlaced sources.
As per discussion here: ​https://github.com/JeffyCN/FFmpeg/issues/10

the correct logic would be inverse of current, i.e.

frame->interlaced_frame = (mode & MPP_FRAME_FLAG_FIELD_ORDER_MASK);

After making this change, the stream is properly marked as interlaced:

Stream #0:0: Video: h264, drm_prime(bt709, top coded first (swapped)), 1440x1080, q=2-31, 3000 kb/s, 29.97 fps, 90k tbn

[Carl Eugen Hoyos]

Please send your patch - made with git format-patch - to the FFmpeg development mailing list.

[tc]

Sorry, I don't have access to mailing list or a way to make a patch. I've posted the change and asking for comment/feedback whether this is reasonable or not.

It's up to the maintainer(s) to decide what to do with this information.