Another pull from my growing-towards-beta CGI library: sgcgi_url_unescape().

Note the use strcpy, which is faster than the equivalent memmove()ing. To ensure 64-bit safety, I plan to rename this function and then conditionally compile it to point to either strcpy or a 64-bit-safe memmove() implementation of strcpy.

However, even though copy order isn't guaranteed for strcpy, on 16-bit and 32-bit systems, all known implementations copy byte-by-byte from lower addresses to higher addresses. Some 64-bit optimized compilers may copy 8-byte chunks, making the assumption of full linearity unstable at best.

I know it sounds like I'm justifying use of nonstandard code for convenience . . . *blush* . . . it's just something that putting in a -DPEDANTIC type of preprocessor flag could fix if broken, and its SO much faster!