Frame data

[eddymhawejj]

Hi there,

Thank for sharing this easy implementation of giflib.
I have a question regarding the Frame class.

Assuming that pixelBuffer represent the data buffer in every frame, is it somehow possible to get the size of that buffer so I could load a particular frame from a gif as a normal image?

[Chlumsky]

I'm not sure I understand the question. You can load a frame as an image by copying its pixels from EasyGifReader::Frame::pixels(). You need to copy exactly 4*width()*height() bytes (4 bytes per pixel), which should be the size of the buffer you asked about?

[eddymhawejj]

Thank you for the answer!
I believe that you did answer my question but unfortunately it isn't working for my case.

What i'm trying to do is to create a directx texture (D3DX9) using the first frame from the gif.
I'm using the function D3DXCreateTextureFromFileInMemoryEx as follow

Here's a sample of the code

		try
		{
			EasyGifReader gifReader = EasyGifReader::openMemory(p_DataBuf, dataSize);

			int frameCount = gifReader.frameCount();
			int width = gifReader.width();
			int height = gifReader.height();
			bool loop = gifReader.repeatInfinitely();

			fprintf("Frame count %d width %d height %d loop %d", frameCount, width, height, loop);


			const std::uint32_t* framePixels = gifReader.begin().pixels(); // first pixel of the gif
			
			
			D3DXIMAGE_INFO imgInfo;
			HRESULT hr = D3DXCreateTextureFromFileInMemoryEx(m_3dDevice,
				framePixels,
				(4*width*height), // Your answer implemented here
				D3DX_DEFAULT_NONPOW2,
				D3DX_DEFAULT_NONPOW2,
				1,
				0,
				D3DFMT_A8R8G8B8,
				D3DPOOL_MANAGED,
				D3DX_FILTER_NONE,
				D3DX_FILTER_NONE,
				0xffff00ff,
				&imgInfo,
				NULL,
				&m_lpd3dTexture);
			if (FAILED(hr)) {
				fprintf("%s: failed to create texture with error: %s %s",
					m_stFileName.c_str(), DXGetErrorString(hr), DXGetErrorDescription(hr));
				return false;
			}
		}

Where the create function returns failed with the following output

3.gif: failed to create texture with error: D3DXERR_INVALIDDATA Invalid data

[Chlumsky]

That's probably right. I would suggest reordering the bytes as needed during a copy operation (instead of just memcpy).

[eddymhawejj]

I ended up reordering it in nextFrame() which i found a bit less expensive.

1 more question, would you kindly share which giflib version you used while developing?
There seem to be some memory issues with reading the buffer from memory as that issue doesn't happen at all when building my application in debug mode.

Edit: downgraded from 5.2.1 to 5.1.2 and everything seem to work normally now! thanks for the support!

Edit2: seems like it was just a coincidence that it worked several times in a row, the problem still exist. Still interested in hearing which version you used.

[Chlumsky]

I have honestly no idea and I'm doubtful that this issue comes from Giflib, which is over 30 years old and used by millions of people. I think it's much more likely the bug is in your project, or even mine. Are you perhaps freeing the memory right after opening the GIF? I'm afraid you will just have to debug a bit more.

[eddymhawejj]

I agree on this point however it could be that the optimization flags I have in my vs project triggered this behavior.
And I'm pretty sure that the issue is not in my project as all other formats except gifs works properly through the dx api by reading from memory.

Soo long story short, I had to dig a little bit into the library itself after not finding anything in your code.
Then I found out that the exception thrown is always in the same place in dgif_lib.c

    /* this can only happen on a severely malformed GIF */
    if (BitsPerPixel > 8 || Private->RunningBits > 32) {
	GifFile->Error = D_GIF_ERR_READ_FAILED;	/* somewhat bogus error code */
	return GIF_ERROR;    /* Failed to read Code size. */
    }

Apparently RunningBits is sometimes larger than 32 which doesn't make much sense as RunningBits is assigned for the first time in the same function just 3 lines below the check.

So I went back to DGifOpen and initialized Private->RunningBits and Private->RunningCode to 0 and now everything works ...
And that explains why it worked in debug mode as all variables are assigned to 0.

Not sure if I should raise a bug in giflib tho, doesn't seem that maintained.

[Chlumsky]

Interesting, I think you should, actually.