Truncated/corrupted jpeg2000 image

[Pazaak]

Hi, I have a batch of images like this and I want them converted to a numpy array:
https://drive.google.com/open?id=1hx_s1LZV48ID89CFaTzE3vjUTXN1IcGl

The images seem to be jpeg2000 and truncated. Trying to load them without LOAD_TRUNCATED_IMAGES gives me the usual exception:

from PIL import Image
a = Image.open('broken.jpg').convert("RGB")

OSError: broken data stream when reading image file

Using LOAD_TRUNCATED_IMAGES = True

from PIL import Image, ImageFile
import numpy as np
ImageFile.LOAD_TRUNCATED_IMAGES = True
a = np.asarray(Image.open('broken.jpg').convert("RGB"))
print(np.sum(a))

Gives a zero sum of the array, indicating that the image is not properly loaded. I was expecting an array containing the RGB information of the image.

• OS: Ubuntu 16.04
• Python: 3.68
• Pillow: 6.1.0

Thank you!

[radarhere]

It is registering as truncated because the JPEG2000 decoder is returning an error - different from the more common truncated image errors that are raised directly from ImageFile.

Pillow/src/libImaging/Jpeg2KDecode.c

Lines 622 to 626 in 575a616

	for (n = 1; n < image->numcomps; ++n) {
	if (image->comps[n].dx != 1 || image->comps[n].dy != 1) {
	state->errcode = IMAGING_CODEC_BROKEN;
	state->state = J2K_STATE_FAILED;
	goto quick_exit;

According to comments from openjpeg, dx and dy refer to the 'separation of a sample of ith component with respect to the reference grid' of one of the image components.

[radarhere]

Testing, I find this is resolved by #4996