In a regression from the shipping Safari, if an absolutly posioned replaced object (i.e. and image) has top, bottom, margin-top, margin-bottom with all values other than auto (the same is true for the horizontal equivilants), it ends up giving incorrect rendering by making top and bottom go to 0.
Created attachment 8523 [details] test case of the horizontal
Created attachment 8524 [details] test case of the vertical
The patch for this is pretty trivial but it will be much easier once bug 9085 is finished an landed.
Created attachment 8539 [details] patch Scratch the last comment. I fixed the patch for bug 9085 to not interfere and was therefore able to make this patch. It very simply adds an else clause to cover the case when all the values are not auto.
Comment on attachment 8539 [details] patch r=me