Yes, I developed it offline until it passed the visible test cases and then uploaded it. I only ran through the random test cases once. I'll take another look some time.

Actually, there are a few test cases that come up in the random input testing that your solution doesn't account for. Sometimes the given epsilon is greater than pi. Also the %.11v formatting option tends to also remove trailing 0s which some of the random input solutions expect to find. I hadn't appreciated how random the random testing was as it wouldn't always fail for me when I submitted my solution for testing. I'm sure I ran through the debug/test cycle more often than you!