Thumbnailing a single gif: 'IOError: codec configuration error when reading image file'

[Lavr18]

What did you do?

I was trying to thumbnail a gif image and save it

from PIL import Image

im = Image.open("Pictures/image.gif")
im.thumbnail((240,160))
im.save("Thumbnails/thumbnail.gif")

What did you expect to happen?

I expeted to get a gif thumbnail in the "/Thumbnails/" directory.
This same code worked for jpg, png, bmp formats.

What actually happened?

This is the log:

Traceback (most recent call last):
  File "testpy.py", line 4, in <module>
    im.thumbnail((240,180))
  File "/usr/lib/python2.7/dist-packages/PIL/Image.py", line 1843, in thumbnail
    im = self.resize(size, resample)
  File "/usr/lib/python2.7/dist-packages/PIL/Image.py", line 1541, in resize
    self.load()
  File "/usr/lib/python2.7/dist-packages/PIL/ImageFile.py", line 244, in load
    raise_ioerror(err_code)
  File "/usr/lib/python2.7/dist-packages/PIL/ImageFile.py", line 59, in raise_ioerror
    raise IOError(message + " when reading image file")
IOError: codec configuration error when reading image file

The issue doesn't occur though with ImageFile.LOAD_TRUNCATED_IMAGES = True, but in this case the thumbnail is completely black with the size of 213x160 instead of 240x160.

What versions of Pillow and Python are you using?

I am using Pillow 4.3.0, Python 2.7.13

[hugovk]

Does it work with other gif files? Please can you share the problematic gif?

[Lavr18]

This works with gif images saved from the internet. For example, with this:

The error occurs when working with gif images made by my raspistill camera. For example, with this:

Edit:
This error happens when I specifically select encoding mode as gif in raspistill camera.
If I don't select encoding mode at all, it works well.
But if not selecting encoding, then all the images are treated as jpg, which is not what I want.

[hugovk]

Slightly smaller reproduction:

from PIL import Image
im = Image.open("2811.gif")
im.load()  # IOError

https://github.com/python-pillow/Pillow/blob/master/PIL/ImageFile.py#L236 returns -1, -8 in n, err_code = decoder.decode(b).

Adding this beforehand allows it to be loaded, but it's a blank, black image:

from PIL import Image, ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True

[wiredfool]

It's kind of a funky gif, it's setting a bit depth of 11:

>>> im.tile
[('gif', (0, 0, 100, 100), 792, (11, False))]

That gets set in GifImagePlugin:

                # image data
                bits = i8(self.fp.read(1))
                self.__offset = self.fp.tell()
                self.tile = [("gif",
                             (x0, y0, x1, y1),
                             self.__offset,
                             (bits, interlace))]

Which is then failing in the GIF decoder:

	/* Initialise state */
	if (context->bits < 0 || context->bits > 8) {
	    state->errcode = IMAGING_CODEC_CONFIG;
	    return -1;
	}

Looking at the GIF, it does have a setting of 0B for the number of bits, which is not correct, or at least, not allowed by the decoder.

[wiredfool]

According to the spec: https://www.w3.org/Graphics/GIF/spec-gif89a.txt (appendix f) it appears that the legal range is 1-12 bits. There doesn't appear to be anything in the decoder that's relying on that bit length to be < 8, so it appears to be a safe change.

[wiredfool]

@Lavr18 Can we use that image in the test suite?

[Lavr18]

@wiredfool Yes, sure