Saving TIFF with compression throws Segmentation Fault 11 in 5.4.1, but not 5.3.0

[jeremydmoore]

What did you do?

Opened a bitonal TIFF created with Adobe Photoshop CC (version 20.0.2) then saved it with Group4 compression using Pillow.

TIFF Image should be downloadable from Github repository here: https://github.com/photosbyjeremy/images/blob/master/test_bitonal_text.tif.zip

What did you expect to happen?

I expected the image to save with compression.

What actually happened?

In Pillow 5.4.1 there was a Segmentation Fault 11, but in Pillow 5.3.0 it works just fine.

What are your OS, Python and Pillow versions?

• OS: macOS 10.14.3
• Python: 3.6.8 (and I actually tried 3.7.1 first, but didn't screenshot those attempts)
• Pillow: 5.4.1 and 5.3.0

Please include code that reproduces the issue and whenever possible, an image that demonstrates the issue. Please upload images to GitHub, not to third-party file hosting sites. If necessary, add the image to a zip or tar archive.

The best reproductions are self-contained scripts with minimal dependencies. If you are using a framework such as plone, Django, or buildout, try to replicate the issue just using Pillow.

from PIL import Image
image = Image.open('test_bitonal_text.tif')
image.save('test_pillow_save.tif', compression='group4')

[jeremydmoore]

[hugovk]

Git bisect points to 6ead422 from PR #3513.

$ python3 3677.py
Python(31119,0x7fffb3bf1380) malloc: *** mach_vm_map(size=18446744072426508288) failed (error code=3)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug
test_pillow_save.tif: Failed to allocate memory for custom tag binary object (-1283043808 elements of 1 bytes each).
Traceback (most recent call last):
  File "3677.py", line 3, in <module>
    image.save('test_pillow_save.tif', compression='group4')
  File "/Users/hugo/github/Pillow/src/PIL/Image.py", line 1988, in save
    save_handler(self, fp, filename)
  File "/Users/hugo/github/Pillow/src/PIL/TiffImagePlugin.py", line 1552, in _save
    e = Image._getencoder(im.mode, 'libtiff', a, im.encoderconfig)
  File "/Users/hugo/github/Pillow/src/PIL/Image.py", line 477, in _getencoder
    return encoder(mode, *args + extra)
RuntimeError: Error setting from dictionary

• OS: macOS High Sierra 10.13.6
• Python: 3.7.2

[jeremydmoore]

If I replace the code in PIL.TiffImagePlugin.py changed by 6ead422 with that from 5.3.0 the save succeeds in 5.4.1

LEFT: TiffImagePlugin.py from 5.4.1 that works RIGHT: TiffImagePlugin.py from 5.3.0 that works

[radarhere]

I have found the problematic tags. The following code runs without error.

from PIL import Image
image = Image.open('test_bitonal_text.tif')
del image.tag_v2[34377]
del image.tag_v2[33723]
del image.tag_v2[700]
image.save('test_pillow_save.tif', compression='group4')

The following code is all that is needed to reproduce the segmentation fault.

from PIL import Image, TiffImagePlugin
TiffImagePlugin.WRITE_LIBTIFF = True
image = Image.new("RGB", (100, 100))
image.save('out.tif', tiffinfo={700: b"a"})

The test suite is able to save byte data for other tag numbers, so the problem is specific to these numbers. The seg fault is occuring at

Pillow/src/libImaging/TiffDecode.c

Line 432 in d167f9e

status = TIFFVSetField(clientstate->tiff, tag, ap);

[radarhere]

While an error still occurs for this issue, the segfault has been fixed in #4033

[radarhere]

The error now received here is

from PIL import Image
image = Image.open('test_bitonal_text.tif')
image.save('test_pillow_save.tif', compression='group4')

TypeError: an integer is required (got type bytes)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "code.py", line 3, in <module>
    image.save('test_pillow_save.tif', compression='group4')
  File "PIL/Image.py", line 2054, in save
  File "PIL/TiffImagePlugin.py", line 1598, in _save
  File "PIL/Image.py", line 428, in _getencoder
SystemError: <built-in function libtiff_encoder> returned a result with an error set

[zoj613]

Also getting the SystemError: <built-in function libtiff_encoder> returned a result with an error set when daving a tif file through tesserocr. Anyone with a solution yet?

[radarhere]

I'm actually working on a solution at the moment. To be sure that your problem is solved though, in case your error is triggered by a different case than the one in this issue, could we have a simple self-contained example of your situation?

[zoj613]

I'm actually working on a solution at the moment. To be sure that your problem is solved though, in case your error is triggered by a different case than the one in this issue, could we have a simple self-contained example of your situation?

it gets triggered when performing OCR on a tiff PIL image via the tesserocr library. SOmething like:

import tesserocr
from tesserocr import PSM
tesserocr.image_to_text(scanned_tiff_img, psm=PSM.SPARSE_TEXT_OSD)

Then I get SystemError: <built-in function libtiff_encoder> returned a result with an error set. From what I remember from looking at the stacktrace, it jammed when the library tried to save the tiff.
My version of PIL is 7.1.2

[hugovk]

@zoj613 Please could you post the full stacktrace?