[WildRackoon]

Somes changes related to the dithering conversion code:

• Fix RGB cache error values at the end of a row, not set to the right values (Blue values was used for R,G and B ?)
  I believe this is a mistake, this affects only the last column of the image, so it is likely very minimal, but I still really would like a second opinion on that.

• Improved error diffusion readability, now the Floyd–Steinberg weights are readable through the code.

• We should add more comments for posterity IMO, most of this dates back to the original PIL fork, the implementation is nicely optimized, but that makes the code slightly more confusing.

I mean look at this 😅:

Pillow/src/libImaging/Convert.c

Line 1284 in 8494b06

/* propagate errors (don't ask ;-) */

Future improvement includes adding more dithering kernels in the same optimized manner.

If anyone needs more details on those changes, I published a gist for that here: https://gist.github.com/WildRackoon/0030fc3d4db3f4af73e2cdac19bbdf6d

[radarhere]

but I still have to make some more tests with regards to that.

Are you saying this PR is not yet ready for review?

[WildRackoon]

but I still have to make some more tests with regards to that.

Are you saying this PR is not yet ready for review?

I edited my comment, It builds fine and works as intented: the result is the same as before with a bunch of test images I tried.

But it is not like we can test that with a ground truth and improve the test suite to that end, so someone's second opinion would be nice.

[radarhere]

You can see that our test suite is failing with your changes. I've created WildRackoon#1 to update our expected results.

[WildRackoon]

Many thanks @radarhere, haven't had time to figure out the related tests yet, just cherry-picked your commit now.

[radarhere]

What do you think of #9829? That PR purports to include this.

[WildRackoon]

LGTM as respects the ideas behind #9338
Nice refactors too