shekh

Maybe it depends on how to encode sample? I used vfw codec feeding rgba32. There is some dithering/noise but after some averaging I get 0xFF00, not 0xFFF0. If I understand...

It seems SAMPLE_HEADER.input_format can be used as a clue.

It is not uncommon approach to scale rgb values when changing bitdepth. Some examples: Convert 8-bit image in Photoshop to 16-bit. 0xFF is scaled to 0xFFFF. r210 decoder in FFmpeg...

I am trying to verify something. I made white rectangle and saved it as rgba using rgba32 source format. On decoding to B64A it is seen as: r,g,b=0xFF00, a=0xFFFF. But...

Good to know. I planned to collect some feedback and then have second attempt. What do you think about replacing "C:\\Program Files\\Common Files\\" with SHGetSpecialFolderPathW(CSIDL_PROGRAM_FILES_COMMON)? Leaving it as is leads...

Ok, atlbase can be removed with no loss by adding this fragment to settings.h ``` #include class CComBSTR { public: char* str; CComBSTR() { str = NULL; } CComBSTR(LPCSTR pSrc)...

Great. Now it looks like it does not need the registry key to exist (SOFTWARE\\CineForm\\ColorProcessing) Is it still good idea to think: if there is any AM db the key...

96 * 64 does not make decoder crash, but it does not decode correctly with any method. input: [s96.zip](https://github.com/gopro/cineform-sdk/files/1556453/s96.zip) output: [s96-vfw.zip](https://github.com/gopro/cineform-sdk/files/1556479/s96-vfw.zip)  Similar results when encoding either through vfw v210,...

I wanted smallest resolution to isolate issue #24. Somehow rejected idea to start with 8 * 8

My understanding of safe abort (it seems to work ok in VirtualDub2): 1) stop trying to send new frames for compression 2) consume all compressed frames 3) release objects Hope...