Neue Testversion 19.40b10

[Herbert123]

Bug time!

Open the attached file. Activate the Curves adjustment layer at the top of the layer stack. Looks good.

Now deactivate that adjustment layer, and activate the same curve adjustment, but in this case the child curve adjustment layer. Banding galore!

Something goes quite wrong with the internal processing when the adjustment curve is applied as a child. This also happens with other adjustment layers, such as the histogram, hue editor, and others.

[photoken]

Banding galore!

Maybe it has to do with how you created the image? I noticed that the bottom background layer is 16-bit, while the other layers are 32-bit. Changing that bottom layer to 32-bit did not cure the banding, so I created a new image as 32-bit right from the get-go and I don't notice banding.

Here's my test:

darktest ken 02.pld

(Oooops! My first attachment didn't use 32-bit adjustment layers. This "02" image does.)

[Hoogo]

It helps when you delete the sRGB profile from the document.
I guess that curves work only on pixels, and the vector is turned into 8Bit RGB on the fly using that profile? Just a guess...

[Herbert123]

Banding galore!

Maybe it has to do with how you created the image? I noticed that the bottom background layer is 16-bit, while the other layers are 32-bit. Changing that bottom layer to 32-bit did not cure the banding, so I created a new image as 32-bit right from the get-go and I don't notice banding.

Here's my test:

darktest ken 02.pld

(Oooops! My first attachment didn't use 32-bit adjustment layers. This "02" image does.)

The 32bpc adjustment layers were just a test to see if that would make a difference. It does not. There is something going on: when I turn off the profile, it looks fine in both cases. With a colour profile, the child adjustment layer causes banding. But it is inconsistent: why would a child adjustment layer cause banding, and the same adjustment applied at the top of the layer stack does not?

I also tried this with a 32bpc document, and the result is the same.

[bkh]

Now deactivate that adjustment layer, and activate the same curve adjustment, but in this case the child curve adjustment layer. Banding galore!

Something goes quite wrong with the internal processing when the adjustment curve is applied as a child. This also happens with other adjustment layers, such as the histogram, hue editor, and others.

Looks as if this has already been fixed in 19.40b11. Didn't expect the Hubers to be so fast, phantastic.

Cheers

Burkhard.

[evren]

Hubers amazing...
PS's coders army vs PL's 2 german rambo :)