This caught my attention:

static wchar_t
charref(char *x, char *y)
{
    wchar_t wc;
    size_t ret;

    if (!(patglobflags & GF_MULTIBYTE) || !(STOUC(*x) & 0x80))
        return (wchar_t) STOUC(*x);

well, this is definitely not valid for arbitrary multibyte character set. I am 
just curious if it is possible to consistently assume that UTF-8 is in use? 
That can definitely simplify things.