Yesterday I spent the whole day debugging a relatively elaborate program that produced wrong results compared to a very similar program that gave me the expected results with the same input. At the last moment I realized the problem was a floating point parameter obtained from @ARGV using the package Getopt::Long. The problem was that the parameter was 1.0e-4 in one program and 1e-4 in the other. The first was read as 0.0001; the second was misread as 1. I was using version 2.57. The error has been corrected in version 2.58.