The shapes of `raw_colors` and `raw_image` are different.

[manipopopo]

For the Fuji X-Pro 1 raw example, the shape of raw_colors is (3288, 5118), and the shape of raw_image is (3288, 5120).

Maybe the raw_colors could be created by

np.pad(pattern, ((0, max(height - n, 0)), (0, max(width - n, 0))), mode='wrap')

instead of https://github.com/letmaik/rawpy/blob/ae822649e72ad112d9aa86a8001fa5cbe708d7a4/rawpy/_rawpy.pyx#L615

[letmaik]

This is the output from exiftool:

Raw Image Full Size : 5120x3288 Raw Image Crop Top Left : 12 28 Raw Image Cropped Size : 4896x3264 Fuji Layout : 12 12 12 12 X-Trans Layout : GRBGBR BGGRGG RGGBGG GBRGRB RGGBGG BGGRGG

And raw_pattern from rawpy:

>>> r.raw_pattern
array([[1, 1, 0, 1, 1, 2],
       [1, 1, 2, 1, 1, 0],
       [2, 0, 1, 0, 2, 1],
       [1, 1, 2, 1, 1, 0],
       [1, 1, 0, 1, 1, 2],
       [0, 2, 1, 2, 0, 1]], dtype=uint8)

We could of course just pad to the correct size as you suggest but I'm wondering what the underlying issue is and whether that's even the correct solution.

[letmaik]

I opened #144. It's probably all fine. I guess there's no reason why the raw size would have to be a multiple of the Bayer pattern size.