GP2X Reading Pixels - Every Other Line Gets Skipped

[GernotFrisch]

I have some code to read from the back buffer, which works perfectly on windows debug version. However on the GP2X it seems that it skips every other line and reads a rectangle that's twice as big as the one I specified, but scaled by 1/2 in vertical.

Here's the code:

	unsigned long scanline = Screen.getWidth();
	Buffer = (unsigned short*)gl_malloc(Width*Height*sizeof(unsigned short)); //memset(Buffer, 0, Width*Height*2);
	register unsigned short* pScreen = (unsigned short*)Screen.getColorBuffer();
	register unsigned short* pBuf = Buffer;
	for(y = ClipY; y < ClipHeight; y++)
	{
		for(x = ClipX; x < ClipWidth; x++)
		{
			*pBuf++ = *pScreen++;
		}
		pBuf-=ClipWidth;
		pBuf+=Width;
		pScreen-=ClipWidth;
		pScreen+=scanline;
	}

gl_malloc returns a mmap'ed memory from the upper 32 MB (mmuhack). The Screen.getColorBuffer is also in this memory space.
The graphic I am grabbing from looks correct. If I load an image into my sprite class, it works, too. But grabbed images are scaled by 1/2 in vertical, but the do have the correct size. I mean: I grab 32x32, and it returns me a grab of 32x64 scaled to 32x32.

Any ideas what this might be?

 

[rixed]

Any ideas what this might be?

Different address for odd and even lines in your GP2X video mode ?
I don't know what the default video mode is, but it may be using this feature.

 

[Squidge]

Your 'for' goes from ClipX to ClipWidth, shouldn't this be taken for in your routines after the for loops, rather than just subtracting ClipWidth?

 

[Exophase]

You should check what Screen.getWidth is actually returning, I don't know the exact interface you're using there but if it were in units of bytes instead of shorts then it'd be causing the problem you're describing.

 

[GernotFrisch]

SOLVED.
I don't know why/what caused it, but it's gone now. So strange...