PNG differences between OS after upgrading from Pillow 11.0.0

[stefan6419846]

What did you do?

Upgrade from Pillow 11.0.0 to 11.3.0.

What did you expect to happen?

The output for one Pillow version is equivalent on different operating systems.

What actually happened?

The output of Pillow 11.3.0 is different for PNG files when running on Ubuntu and Windows.

What are your OS, Python and Pillow versions?

• OS: Microsoft Windows Server 2022 10.0.20348 and Ubuntu 22.04.5
• Python: 3.13.5
• Pillow: 11.3.0

Corresponding CI runs: https://github.com/py-pdf/pypdf/actions/runs/16620592573

These are the images for two of the tests:

• test_flate_decode__not_rectangular, actual on Windows: https://github.com/user-attachments/assets/c6ae29d2-7967-45bd-abee-619fe758a5db
• test_flate_decode__not_rectangular, expected and Ubuntu: https://github.com/user-attachments/assets/ae23c82f-db14-4f0a-9fc4-a19ff72d8b1f
• test_pa_image_extraction, actual on Windows: https://github.com/user-attachments/assets/73f6c746-9fa3-49e3-ac1a-6686ff2882cc
• test_pa_image_extraction, expected and Ubuntu: https://github.com/user-attachments/assets/ffebdd80-8bda-4ec1-a0da-830d36d619e6

It is unexpected for me that the Windows builds suddenly break without a major version change, although there are no real visible changes.

[radarhere]

Hi. This would be a duplicate of #8749. Here are the key points.

#8749 (comment)

you're experiencing the effects of #8500, where we switched from zlib to zlib-ng in our wheels. See also https://pillow.readthedocs.io/en/stable/releasenotes/11.1.0.html#zlib-ng-in-wheels. As noted in the PR, this changes the output, but improves performance.

#8749 (comment)

I don’t think that file bytes comparison is a valid test for PNG files. There are several different ways that a PNG can be encoded losslessly into different bytes — there are different filter methods that can be chosen arbitrarily by the encoder to improve compression on a scan line by scan line basis. On top of that, the underlying compression can be done that are bit different, but still decompress to the same stream of bits.

So, all in all, I think that bit differences in png files is an invalid failure for a test.

#8749 (comment)

I'm not sure what the purpose of the test is, but I would suggest you compare the pixels in the images, rather than the bytes.

Let us know if you have any further questions.

[stefan6419846]

Thanks for the hint - I have been searching for Windows-specific issues and did not consider including macOS as well. We will switch to proper pixel comparison to mitigate this on our side.