Java's lack of unsigned types is a real pain a library like this. Lots of widening takes place. Thankfully .NET has a much saner type system. Converting this library to .NET meant throwing out quite a bit of ugly code.

Some Java APIs use int when reading a byte from stream, where -1 means EOF. That makes comparisons like this one work.

BTW some IDEs (IntelliJ for example) will automatically flag errors such as this, as it can prove that the condition is never met.

@drewnoakes FindBugs found it for me, I simply must have forgotten to run it though FindBugs at the time. It does the same as e.g IntelliJ and point out that the condition can never be met.

Not having unsigned types is one thing (although bad enough), but actually naming the type byte while being signed is just wrong as I see it. I wonder how many bugs that decision has caused during Java's lifetime. If it was called anything with "int", (small, short, tiny etc) you'd expect it, but to me a byte is a collection of bits where the concept of sign "doesn't exist".