The plain int type is only guaranteed to be at least 16 bits, and
yet the flags variable was expecting at least 22 bits. This turns
out to be true for x86 and x86-64 systems, but to ensure compatibility,
the variable has been changed to an unsigned long int. Also added
the "L" suffix to flags that were more than 16 bits.
Most of these were defining various integers as char types, probably
in the naive belief that this is necessarily good because it uses less
RAM. There were also several unnecessary unsigned ints, though.
These have all been changed to just "int", so the compiler can decide
exactly what type to use.