ffmpeg.SWS_FAST_BILINEAR missing after updating to 8.0.0

[vpenades]

After updating to 8.0.0, I've got compile errors related to ffmpeg.SWS_FAST_BILINEAR not existing anymore.

I've checked the source code and it's there, so I don't know what might be the issue.

[zgabi]

It was added to an enum in the ffmpeg header:

typedef enum SwsFlags {
    /**
     * Scaler selection options. Only one may be active at a time.
     */
    SWS_FAST_BILINEAR = 1 <<  0, ///< fast bilinear filtering
    SWS_BILINEAR      = 1 <<  1, ///< bilinear filtering
    SWS_BICUBIC       = 1 <<  2, ///< 2-tap cubic B-spline
    SWS_X             = 1 <<  3, ///< experimental
    SWS_POINT         = 1 <<  4, ///< nearest neighbor
    SWS_AREA          = 1 <<  5, ///< area averaging
...
}

So it was also added to an enum in the C# wrapper:

public enum SwsFlags : int
{
    /// <summary>fast bilinear filtering</summary>
    @SWS_FAST_BILINEAR = 1,
    /// <summary>bilinear filtering</summary>
    @SWS_BILINEAR = 2,
    /// <summary>2-tap cubic B-spline</summary>
    @SWS_BICUBIC = 4,
...
}

It is a little bit strange that in the sws_getContext function it accepts an integer, to you have to cast the enum value to int:

            var rgbToYuvCtx = sws_getContext(_c->width, _c->height, AVPixelFormat.AV_PIX_FMT_GRAY8, _c->width, _c->height,
                AVPixelFormat.AV_PIX_FMT_YUV420P, (int)SwsFlags.SWS_BICUBIC, null, null, null);

But this comes from the native ffmpeg...

[vpenades]

It is a little bit strange that in the sws_getContext function it accepts an integer

I guess it's one of those C++ things that simply let enums to be passed as values.

Anyway, first thing I did was to check your examples, and saw them unchanged, that's why I opened the issue.

I guess you'll have to update the examples to the latest version

[zgabi]

Where is that example?
Btw, I'm just a user of the ffmpeg.AutoGen. I think your previous reply was meant for @Ruslan-B :)

[vpenades]

Where is that example?

it's here

Where is that example? Btw, I'm just a user of the ffmpeg.AutoGen. I think your previous reply was meant for @Ruslan-B :)

Indeed 😃, I guess the examples need to be updated accordingly

[zgabi]

Check the example in the 8.0 branch:
https://github.com/Ruslan-B/FFmpeg.AutoGen/blob/8.0/FFmpeg.AutoGen.Example/MediaDecoder.cs

It seems that the 8.0 changes were not merged to the master branch. (Not just the sample, nothing)

[Ruslan-B]

I am planning to throw out cpp sharp any way, none AI magic ) science. However will take time, working on floks population, still deep in semantics. Don't blame me for not merging github full of AI rules nowadays, they change rules, bot stuck in code. It is a AI mess =^_^=

[Ruslan-B]

Up side, positive, it will give use normal interface for semi managed c like c#. However looking on trends, most-probaly ffmpeg will fall as well - why do you needed any way just read and extract media? Right do it one recode id for web and that it.
Philosophical question, why we are using ffmpeg?

[vpenades]

On a personal note: I am using ffmpeg.(autogen) for unit testing. yeah

We work on AI vision sensors, so we have lots of unit tests that do this:

• load a pre-recorded Mp4
• extract individual frames
• process individual frames, usually with some onnx vision models
• annotate the frames with data (usually drawing rectangles over the frames)
• gather the modified frames back into an mp4, so it can be reviewed

A recurring issue I've been having with this approach is that I've never been able to replay the output mp4 at the same framerate as the input mp4, in our case it's not a great deal because we use it for testing.