Saving .ico saves more sizes than necessary

[hugovk]

What did you do?

from PIL import Image
im = Image.open('python.ico')
im.save('saved.ico')
im_saved = Image.open('saved.ico')

print im.info['sizes']
print im_saved.info['sizes']

assert im_saved.info['sizes'] == set([(16, 16), (24, 24), (32, 32), (48, 48)])

What did you expect to happen?

No assertion, and only relevant sizes saved:

set([(16, 16), (48, 48), (32, 32)])
set([(16, 16), (24, 24), (48, 48), (32, 32)])

$ identify python.ico
python.ico[0] ICO 16x16 16x16+0+0 32-bit sRGB 15.1KB 0.000u 0:00.002
python.ico[1] ICO 32x32 32x32+0+0 32-bit sRGB 15.1KB 0.000u 0:00.008
python.ico[2] ICO 48x48 48x48+0+0 32-bit sRGB 15.1KB 0.000u 0:00.013
$ identify saved.ico
saved.ico[0] PNG 16x16 16x16+0+0 8-bit sRGB 12.7KB 0.000u 0:00.003
saved.ico[1] PNG 24x24 24x24+0+0 8-bit sRGB 12.7KB 0.000u 0:00.006
saved.ico[2] PNG 32x32 32x32+0+0 8-bit sRGB 12.7KB 0.000u 0:00.007
saved.ico[3] PNG 48x48 48x48+0+0 8-bit sRGB 12.7KB 0.000u 0:00.010

What actually happened?

set([(16, 16), (48, 48), (32, 32)])
set([(64, 64), (16, 16), (256, 256), (24, 24), (32, 32), (128, 128), (48, 48)])
Traceback (most recent call last):
  File "test.py", line 9, in <module>
    assert im_saved.info['sizes'] == set([(16, 16), (24, 24), (32, 32), (48, 48)])
AssertionError

$ identify python.ico
python.ico[0] ICO 16x16 16x16+0+0 32-bit sRGB 15.1KB 0.000u 0:00.002
python.ico[1] ICO 32x32 32x32+0+0 32-bit sRGB 15.1KB 0.000u 0:00.008
python.ico[2] ICO 48x48 48x48+0+0 32-bit sRGB 15.1KB 0.000u 0:00.013
$ identify saved.ico
saved.ico[0] PNG 16x16 16x16+0+0 8-bit sRGB 12.7KB 0.000u 0:00.003
saved.ico[1] PNG 24x24 24x24+0+0 8-bit sRGB 12.7KB 0.000u 0:00.006
saved.ico[2] PNG 32x32 32x32+0+0 8-bit sRGB 12.7KB 0.000u 0:00.007
saved.ico[3] PNG 48x48 48x48+0+0 8-bit sRGB 12.7KB 0.000u 0:00.010
saved.ico[4] PNG 48x48 48x48+0+0 8-bit sRGB 12.7KB 0.000u 0:00.011
saved.ico[5] PNG 48x48 48x48+0+0 8-bit sRGB 12.7KB 0.000u 0:00.013
saved.ico[6] PNG 48x48 48x48+0+0 8-bit sRGB 12.7KB 0.000u 0:00.016

What versions of Pillow and Python are you using?

Python 2.7.11
Pillow 3.4.2