Paint.net update changed the way the "Dents" distortion filter works, causing unwanted artifacts/glitches. Was this an intended change?

[megatrook]

Forgive me if this post isn't in the correct forum section. I looked around a bit and this seems to be the most appropriate place for it.

 

Recently, I was going to use the Dents filter to add some texture to a gradient, but noticed it wasn't behaving as it used to. Before, it would only grab colors from nearby points in the image, but after updating paint.net, it seems to be wrapping around in some way, causing weird color artifacts. Not only that, but in some cases, the Dents filter stops properly functioning for a substantial part of the canvas.

 

I'll include a couple pictures to illustrate what I mean:

 

In this first picture, I first created a white-to-black gradient and then applied a Dents filter, with default settings. As you can see, around the borders of the image, there are white/gray artifacts around the darker region of the gradient. The same thing happens with any combination of colors, some more noticeably than others. In the past, this wouldn't happen at all, and the general organization of the colors of the image would be preserved. No weird out-of-place splotches of color.

 

In this other example, generated after creating a vertical gradient of yellow to orangey-pink and applying the dents filter with different settings, there are still some unwanted colors around the edge of the screen (the pink splotches to the right and left), but there is also a horizontal line down the middle of the picture, and the lower half of the image failed to properly apply the filter like the upper half. I did not select the lower part of the picture and do something different to it, this is just how it happened.

 

Unsure if these changes were intended. If not, I hope this issue gets fixed. If they were intended, then I'd really like if there was a way to use the old behaviour of the dents plugin instead of this new one!

[Rick Brewster]

This may be from changing the floating-point from double-precision to single-precision. I'll look into it for 4.3.3.

 

[Rick Brewster]

This reproduces for me, should be easy to look at the commit history to figure out what's causing it

 

[Rick Brewster]

This has been fixed for the 4.3.3 release