I see what you're getting at, but that's not the mistake that was made here. People working on it knew what it should do. It's in the spec so Java agrees with y

I see what you're getting at, but that's not the mistake that was made here. People working on it knew what it should do. It's in the spec so Java agrees with you. This bug is a little narrower than this makes it out to be, it ONLY fails when the array on the left is a String[ ]. The way String operations get compiled down to bytecode changed a lot around those releases so the bug is more like an unforeseen interaction between optimization passes than a naive MST/Macro implementation. The "funny" thing here to me is the surprising gap in the regression testing given the quality of most of the JDK. go figure