Description of problem: With some images, 'jpegtran -rotate 90' causes the resulting image to be 'slipped' left by a few pixels, so that what should have been on the left edge is now in the right edge, and flipped. For example, like this: xample, like this:e roF Version-Release number of selected component (if applicable): libjpeg-6b-32
You would not happen to have an example image do demonstrate that, would you ?
Created attachment 105064 [details] catmug1.jpg Sure, here's one. It's already been rotated once, so you can see the effect already. Rotating it again produces another artifact, again on the right-hand side. jpegtran -rotate 90 catmug1.jpg > rotated.jpg
This is documented in the jpegtran man page: The transpose transformation has no restrictions regarding image dimensions. The other transformations operate rather oddly if the image dimensions are not a multiple of the iMCU size (usually 8 or 16 pixels), because they can only transform complete blocks of DCT coefficient data in the desired way. I don't think this is easily fixable short of rewriting the jpeg transformation support completely, which I'm not up to.