Add a fuzzer for ImageCms.buildTransform

[jvoisin]

No description provided.

[radarhere]

Suggested change

	t = ImageCms.buildTransform(p1, p2, trans1, trans2)
	ImageCms.buildTransform(p1, p2, trans1, trans2)

Lint fix.

[radarhere]

Hi. A few questions.

1. Could you tell us what inspired you to add this? What about ImageCms.buildTransform() or ImageCms.createProfile() made you think that additional testing was necessary?

2. If I understand your code correctly, you're picking a random mode from a list for each of four arguments for fuzz_cms(). The first argument goes to ImageCms.createProfile(). Looking at

   Pillow/src/PIL/ImageCms.py

   Lines 767 to 769 in 6782a07

   	:param colorSpace: String, the color space of the profile you wish to
   	create.
   	Currently only "LAB", "XYZ", and "sRGB" are supported.

   
   only "LAB", "XYZ" and "sRGB" are supported though, and of those, only "LAB" is in your list of modes. The only time your code passes is when all four arguments are "LAB" . Otherwise, your code fails with Color space not supported for on-the-fly profile creation, right?

3. Our other fuzzing scenarios are for when a user is opening an image or a font, scenarios were the input is not trusted and may have come from anywhere. What made you think this was a scenario for fuzzing, and not something that could be exhaustively tested in our normal test suite?