FFmpeg's swscale does not properly copy all planes when the pixel format is NV12, so instead we will handle cases where nothing actually changes and we just have to copy memory.
There is a small but important difference in the way OBS Studio and FFmpeg handle the FPS time base. This change fixes the problem and makes all encoders go back to sanity.
This prevents the uninformed from using an encoder that they really shouldn't be using, thus reducing the amount of issues that come from showing too many encoders.
Some encoders don't have all options available with the automatically detected color format, so allowing the user to override the detected color format opens up more options.
Instead of instantly dropping frames on EAGAIN, try first recieving a packet (if not already done so) and then resubmit the same frame. If it still doesn't work, drop the frame as we have no way to deal with that situation yet.
* Improved encoding loop can now deal with early and late receive requests.
* Lots of warnings fixed.
* Removed all threading options (they were broken anyway).
* Added some profiler calls.
This implements video encoding for the generic encoder by using frame queue and a set lag amount (in frames). The lag amount determines for how many frames no repeated attempts at submitting or retrieving a frame should be made, and after that many frames have been submitted the generic encoder will try to retrieve packets if at all possible. If not possible, it will delay for as long as possible to show an Encoder overloaded message.
Michael Fabian 'Xaymar' Dirks