javax/swing/JSplitPane/4885629/bug4885629.java
fails on retina display (QuartzDebug emulator).
Difference is in the last pixel. The test looks exactly for this last pixel but in this case perhaps a wrong color is a result of scaling. We should not, I think, treat it as a regression against JDK-4885629 but I don't know if to ignore it as a non-issue or mark it as a duplicate of some geometry bug.

OS X 10.8.4 macmini4,1
retina emulator QuartzDebug (960x540)