- With gs 8.53, the black shadows arount the big text ("Free Sync") are missing. - With gs 8.57, the shadows are here, but the background of the text is completely wrong (white instead of orange). - With HEAD the background is REALLY strange (blocks of patterns of gray)
Created attachment 3196 [details] in.pdf
Created attachment 3197 [details] r8127.png PNG created with -r36 from HEAD as of rev 8127
Created attachment 3198 [details] gs857.png PNG at -r36 from release 8.57
This is probably a graphics library issue since it is not affected by disabling PDF transparency (also it paints as it is being rendered to the display device on windows or to the x11 device on linux). Adjusting priority to customer level.
Created attachment 3199 [details] simplified sample file This is a JPXDecode issue. This simplified sample PDF file contains a single image in /DeviceGray color space. The image looks mostly white on AR8 (only a few pixels are black) but it is rather dark on Ghostscript.
The image in the simplified sample file has wrong decoding array. Please change [0 255] to [0 1]. Even with this change Ghostscript and AR8 render the file differently. AR8 renders the original sample file with the white text background - the same as gs 8.57. However, gs 8.57 and HEAD render the simplified sample identically. Perhaps, there's more than one problem.
Passing the bug to Alex who handfles the PDF interpreter. As I understood, the image should look better if replace Decoding. Alex, can you code an automatic replacement based on checking the color space ? Maybe not general, - just for simple cases. Then please pass the bug to the owner of JPXDecode.
The original file is correct. I made a mistake when I created the simplified sample file. There's no need to change the PDF interpreter.
This was actually a duplicate of Bug 688869. The issue reappeared in r7958, and is now fixed properly in r8294.