Fixed Python errors when saving a (0, 0) TIFF image

[radarhere]

At the moment, saving a zero by zero image to TIFF gives a ZeroDivisionError.

>>> from PIL import Image
>>> Image.new("RGB", (0, 0)).save("out.tiff")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "PIL/Image.py", line 2235, in save
    save_handler(self, fp, filename)
  File "PIL/TiffImagePlugin.py", line 1586, in _save
    strips_per_image = (im.size[1] + rows_per_strip - 1) // rows_per_strip
ZeroDivisionError: integer division or modulo by zero

This PR ensures that rows_per_strip, and then later strip_byte_counts, are not zero, allowing the code to progress forward to a SystemError instead.

>>> Image.new("RGB", (0, 0)).save("out.tiff")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "PIL/Image.py", line 2240, in save
  File "PIL/TiffImagePlugin.py", line 1759, in _save
  File "PIL/ImageFile.py", line 531, in _save
SystemError: tile cannot extend outside image

[radarhere]

cc @kmilos

[kmilos]

Why not just handle the problematic division instead, is it not enough?

[radarhere]

I set rows_per_strip to 1 since I presume that setting ROWSPERSTRIP to zero is not ideal.

I set strip_byte_counts to 1 since it is used in range later, and can't be zero.

[kmilos]

This was actually raising two different kinds of error in 8.1 and before: SystemError indeed if uncompressed (no libtiff), but a different

RuntimeError: Error setting from dictionary


_TIFFVSetField: out.tiff: Bad value 0 for "RowsPerStrip" tag.

when using libtiff...

So what's the behaviour you want, both raising the same SystemError now I guess?

[radarhere]

I created this PR more just out of the thought that we shouldn't raise a ZeroDivisionError.

[kmilos]

As a side effect I think it is better both Pillow and libtiff writers now throw the same SystemError for this corner case, so please feel free to merge as far as I'm concerned.