**         does not understand UTF-16, it may falsely consider UTF-16 text
**         to be binary.
**
** (-1) -- The content appears to consist entirely of text, with lines
**         delimited by carriage-return, line-feed pairs; however, the
**         encoding may not be UTF-8.
**
************************************ WARNING **********************************
**
** This function does not validate that the blob content is properly formed
** UTF-8.  It assumes that all code points are the same size.  It does not
** validate any code points.  It makes no attempt to detect if any [invalid]
** switches between UTF-8 and other encodings occur.
**
** The only code points that this function cares about are the NUL character,
** carriage-return, and line-feed.
**
************************************ WARNING **********************************
*/
int looks_like_utf8(const Blob *pContent){
  const char *z = blob_buffer(pContent);
  unsigned int n = blob_size(pContent);
  int j, c;
  int result = 1;  /* Assume UTF-8 text with no CR/NL */

  if( j>LENGTH_MASK ){
    return 0;  /* Very long line -> binary */
  }
  return result;  /* No problems seen -> not binary */
}

/*
** Define the type needed to represent a Unicode (UTF-16) character.
*/
#ifndef WCHAR_T
#  ifdef _WIN32
#    define WCHAR_T wchar_t
#  else
#    define WCHAR_T unsigned short
#  endif
#endif

/*
** Maximum length of a line in a text file, in UTF-16 characters.  (4096)
** The number of bytes represented by this value cannot exceed LENGTH_MASK

** bytes, because that is the line buffer size used by the diff engine.
*/
#define UTF16_LENGTH_MASK_SZ  (LENGTH_MASK_SZ-(sizeof(WCHAR_T)-sizeof(char)))
#define UTF16_LENGTH_MASK     ((1<<UTF16_LENGTH_MASK_SZ)-1)

/*
** The carriage-return / line-feed characters in the UTF-16be and UTF-16le
** encodings.
*/
#define UTF16BE_CR  ((WCHAR_T)'\r')
#define UTF16BE_LF  ((WCHAR_T)'\n')
#define UTF16LE_CR  (((WCHAR_T)'\r')<<(sizeof(char)<<3))
#define UTF16LE_LF  (((WCHAR_T)'\n')<<(sizeof(char)<<3))


/*
** This function attempts to scan each logical line within the blob to
** determine the type of content it appears to contain.  Possible return
** values are:
**
**  (1) -- The content appears to consist entirely of text, with lines
**         delimited by line-feed characters; however, the encoding may
**         not be UTF-16.
**
**  (0) -- The content appears to be binary because it contains embedded
**         NUL characters or an extremely long line.  Since this function
**         does not understand UTF-8, it may falsely consider UTF-8 text
**         to be binary.
**
** (-1) -- The content appears to consist entirely of text, with lines
**         delimited by carriage-return, line-feed pairs; however, the
**         encoding may not be UTF-16.
**
************************************ WARNING **********************************
**
** This function does not validate that the blob content is properly formed
** UTF-16.  It assumes that all code points are the same size.  It does not
** validate any code points.  It makes no attempt to detect if any [invalid]
** switches between the UTF-16be and UTF-16le encodings occur.
**
** The only code points that this function cares about are the NUL character,
** carriage-return, and line-feed.
**
************************************ WARNING **********************************
*/
int looks_like_utf16(const Blob *pContent){
  const WCHAR_T *z = (WCHAR_T *)blob_buffer(pContent);
  unsigned int n = blob_size(pContent);
  int j, c;
  int result = 1;  /* Assume UTF-16 text with no CR/NL */

  /* Check individual lines.
  */
  if( n==0 ) return result;  /* Empty file -> text */
  if( n%2 ) return 0;  /* Odd number of bytes -> binary (or UTF-8) */
  c = *z;
  if( c==0 ) return 0;  /* NUL character in a file -> binary */
  j = ((c!=UTF16BE_LF) && (c!=UTF16LE_LF));
  while( (n-=2)>0 ){
    c = *++z; ++j;
    if( c==0 ) return 0;  /* NUL character in a file -> binary */
    if( c==UTF16BE_LF || c==UTF16LE_LF ){
      int c2 = z[-1];
      if( c2==UTF16BE_CR || c2==UTF16LE_CR ){
        result = -1;  /* Contains CR/NL, continue */
      }
      if( j>UTF16_LENGTH_MASK ){
        return 0;  /* Very long line -> binary */