The interesting part to me was the variations in the emitted instructions, based upon the source. IOW, it would be easy to mistakenly think that you weren't going to get jumps if you only used a few case branches to test things out.