I’m developing a media foundation-based h264 player.
I open my URL.
The framework creates and initializes my media source, which in turn creates and initializes my 2 streams, audio and video.
Then it asks for video samples until the end of the file is reached. The log is filled with CMFTransformDetours::ProcessOutput failed hr=0xC00D6D72 MF_E_TRANSFORM_NEED_MORE_INPUT
Then after my video stream sends MEEndOfStream, the framework asks for a few more audio samples, finally transitions the state to playing, and starts to play audio only.
What does the framework tries to find in my video stream that isn’t there?
The same file plays OK by the same player code if opened by the built-in stream source. mftrace.exe says when the built-in stream source plays the file, the first video sample is 38 bytes longer then when my stream source plays it (all other samples are exactly the same length).
38 bytes is exactly the size of my MF_MT_MPEG_SEQUENCE_HEADER value for the video stream (i.e. 00 00 01 + SPS + 00 00 01 + PPS). I’ve tried to prepend MF_MT_MPEG_SEQUENCE_HEADER value to my first sample, didn’t help.
System-provided stream source sets some undocumented attribute on video samples, GUID = {19124E7C-AD4B-465F-BB18-20186287B6AF}, values are 8-bytes binary values like “09 00 00 00 29 0d 00 00”, different on each frame — what’s that and can it be the reason?
What else can I try?
Is there any documentation on what exactly does the MF h264 decoder wants on input?
Thanks in advance.