issue #195 lexer / utf8 improvements from Lua 5.4

src/llex.c

@@ -1,5 +1,5 @@
 /*
-** $Id: llex.c,v 2.96.1.1 2017/04/19 17:20:42 roberto Exp $
+** $Id: llex.c $
 ** Lexical Analyzer
 ** See Copyright Notice in lua.h
 */
@@ -29,7 +29,7 @@
 
 
 
-#define next(ls) (ls->current = zgetc(ls->z))
+#define next(ls)	(ls->current = zgetc(ls->z))
 
 
 
@@ -88,7 +88,10 @@ void luaX_init (lua_State *L) {
 const char *luaX_token2str (LexState *ls, int token) {
   if (token < FIRST_RESERVED) {  /* single-byte symbols? */
     lua_assert(token == cast_uchar(token));
-    return luaO_pushfstring(ls->L, "'%c'", token);
+    if (lisprint(token))
+      return luaO_pushfstring(ls->L, "'%c'", token);
+    else  /* control character */
+      return luaO_pushfstring(ls->L, "'<\\%d>'", token);
   }
   else {
     const char *s = luaX_tokens[token - FIRST_RESERVED];
@@ -224,8 +227,16 @@ static int check_next2 (LexState *ls, const char *set) {
 
 /* LUA_NUMBER */
 /*
-** this function is quite liberal in what it accepts, as 'luaO_str2num'
+** This function is quite liberal in what it accepts, as 'luaO_str2num'
-** will reject ill-formed numerals.
+** will reject ill-formed numerals. Roughly, it accepts the following
+** pattern:
+**
+**   %d(%x|%.|([Ee][+-]?))* | 0[Xx](%x|%.|([Pp][+-]?))*
+**
+** The only tricky part is to accept [+-] only after a valid exponent
+** mark, to avoid reading '3-4' or '0xe+1' as a single number.
+**
+** The caller might have already read an initial dot.
 */
 static int read_numeral (LexState *ls, SemInfo *seminfo) {
   TValue obj;
@@ -236,14 +247,14 @@ static int read_numeral (LexState *ls, SemInfo *seminfo) {
   if (first == '0' && check_next2(ls, "xX"))  /* hexadecimal? */
     expo = "Pp";
   for (;;) {
-    if (check_next2(ls, expo))  /* exponent part? */
+    if (check_next2(ls, expo))  /* exponent mark? */
       check_next2(ls, "-+");  /* optional exponent sign */
-    if (lisxdigit(ls->current))
+    else if (lisxdigit(ls->current) || ls->current == '.')  /* '%x|%.' */
-      save_and_next(ls);
-    else if (ls->current == '.')
       save_and_next(ls);
     else break;
   }
+  if (lislalpha(ls->current))  /* is numeral touching a letter? */
+    save_and_next(ls);  /* force an error */
   save(ls, '\0');
   if (luaO_str2num(luaZ_buffer(ls->buff), &obj) == 0)  /* format error? */
     lexerror(ls, "malformed number", TK_FLT);
@@ -260,12 +271,12 @@ static int read_numeral (LexState *ls, SemInfo *seminfo) {
 
 
 /*
-** skip a sequence '[=*[' or ']=*]'; if sequence is well formed, return
+** reads a sequence '[=*[' or ']=*]', leaving the last bracket.
-** its number of '='s; otherwise, return a negative number (-1 iff there
+** If sequence is well formed, return its number of '='s + 2; otherwise,
-** are no '='s after initial bracket)
+** return 1 if there is no '='s or 0 otherwise (an unfinished '[==...').
 */
-static int skip_sep (LexState *ls) {
+static size_t skip_sep (LexState *ls) {
-  int count = 0;
+  size_t count = 0;
   int s = ls->current;
   lua_assert(s == '[' || s == ']');
   save_and_next(ls);
@@ -273,11 +284,13 @@ static int skip_sep (LexState *ls) {
     save_and_next(ls);
     count++;
   }
-  return (ls->current == s) ? count : (-count) - 1;
+  return (ls->current == s) ? count + 2
+         : (count == 0) ? 1
+         : 0;
 }
 
 
-static void read_long_string (LexState *ls, SemInfo *seminfo, int sep) {
+static void read_long_string (LexState *ls, SemInfo *seminfo, size_t sep) {
   int line = ls->linenumber;  /* initial line (for error message) */
   save_and_next(ls);  /* skip 2nd '[' */
   if (currIsNewline(ls))  /* string starts with a newline? */
@@ -311,8 +324,8 @@ static void read_long_string (LexState *ls, SemInfo *seminfo, int sep) {
     }
   } endloop:
   if (seminfo)
-    seminfo->ts = luaX_newstring(ls, luaZ_buffer(ls->buff) + (2 + sep),
+    seminfo->ts = luaX_newstring(ls, luaZ_buffer(ls->buff) + sep,
-                                     luaZ_bufflen(ls->buff) - 2*(2 + sep));
+                                     luaZ_bufflen(ls->buff) - 2 * sep);
 }
 
 
@@ -348,8 +361,8 @@ static unsigned long readutf8esc (LexState *ls) {
   r = gethexa(ls);  /* must have at least one digit */
   while ((save_and_next(ls), lisxdigit(ls->current))) {
     i++;
+    esccheck(ls, r <= (0x7FFFFFFFu >> 4), "UTF-8 value too large");
     r = (r << 4) + luaO_hexavalue(ls->current);
-    esccheck(ls, r <= 0x10FFFF, "UTF-8 value too large");
   }
   esccheck(ls, ls->current == '}', "missing '}'");
   next(ls);  /* skip '}' */
@@ -494,9 +507,9 @@ static int llex (LexState *ls, SemInfo *seminfo) {
         /* else is a comment */
         next(ls);
         if (ls->current == '[') {  /* long comment? */
-          int sep = skip_sep(ls);
+          size_t sep = skip_sep(ls);
           luaZ_resetbuffer(ls->buff);  /* 'skip_sep' may dirty the buffer */
-          if (sep >= 0) {
+          if (sep >= 2) {
             read_long_string(ls, NULL, sep);  /* skip long comment */
             luaZ_resetbuffer(ls->buff);  /* previous call may dirty the buff. */
             break;
@@ -508,12 +521,12 @@ static int llex (LexState *ls, SemInfo *seminfo) {
         break;
       }
       case '[': {  /* long string or simply '[' */
-        int sep = skip_sep(ls);
+        size_t sep = skip_sep(ls);
-        if (sep >= 0) {
+        if (sep >= 2) {
           read_long_string(ls, seminfo, sep);
           return TK_STRING;
         }
-        else if (sep != -1)  /* '[=...' missing second bracket */
+        else if (sep == 0)  /* '[=...' missing second bracket? */
           lexerror(ls, "invalid long string delimiter", TK_STRING);
         return '[';
       }

src/lobject.c

@@ -1,5 +1,5 @@
 /*
-** $Id: lobject.c,v 2.113.1.1 2017/04/19 17:29:57 roberto Exp $
+** $Id: lobject.c $
 ** Some generic functions over Lua objects
 ** See Copyright Notice in lua.h
 */
@@ -345,7 +345,7 @@ size_t luaO_str2num (const char *s, TValue *o) {
 
 int luaO_utf8esc (char *buff, unsigned long x) {
   int n = 1;  /* number of bytes put in buffer (backwards) */
-  lua_assert(x <= 0x10FFFF);
+  lua_assert(x <= 0x7FFFFFFFu);
   if (x < 0x80)  /* ascii? */
     buff[UTF8BUFFSZ - 1] = cast(char, x);
   else {  /* need continuation bytes */
@@ -442,7 +442,7 @@ const char *luaO_pushvfstring (lua_State *L, const char *fmt, va_list argp) {
       }
       case 'U': {  /* an 'int' as a UTF-8 sequence */
         char buff[UTF8BUFFSZ];
-        int l = luaO_utf8esc(buff, cast(long, va_arg(argp, long)));
+        int l = luaO_utf8esc(buff, va_arg(argp, long));
         pushstr(L, buff + UTF8BUFFSZ - l, l);
         break;
       }

src/lutf8lib.c

@@ -20,7 +20,20 @@
 #include "lauxlib.h"
 #include "lualib.h"
 
-#define MAXUNICODE	0x10FFFF
+
+#define MAXUNICODE	0x10FFFFu
+
+#define MAXUTF		0x7FFFFFFFu
+
+/*
+** Integer type for decoded UTF-8 values; MAXUTF needs 31 bits.
+*/
+#if (UINT_MAX >> 30) >= 1
+typedef unsigned int utfint;
+#else
+typedef unsigned long utfint;
+#endif
+
 
 #define iscont(p)	((*(p) & 0xC0) == 0x80)
 
@@ -35,53 +48,62 @@ static lua_Integer u_posrelat (lua_Integer pos, size_t len) {
 
 
 /*
-** Decode one UTF-8 sequence, returning NULL if byte sequence is invalid.
+** Decode one UTF-8 sequence, returning NULL if byte sequence is
+** invalid.  The array 'limits' stores the minimum value for each
+** sequence length, to check for overlong representations. Its first
+** entry forces an error for non-ascii bytes with no continuation
+** bytes (count == 0).
 */
-static const char *utf8_decode (const char *o, int *val) {
+static const char *utf8_decode (const char *s, utfint *val, int strict) {
-  static const unsigned int limits[] = {0xFF, 0x7F, 0x7FF, 0xFFFF};
+  static const utfint limits[] =
-  const unsigned char *s = (const unsigned char *)o;
+        {~(utfint)0, 0x80, 0x800, 0x10000u, 0x200000u, 0x4000000u};
-  unsigned int c = s[0];
+  unsigned int c = (unsigned char)s[0];
-  unsigned int res = 0;  /* final result */
+  utfint res = 0;  /* final result */
   if (c < 0x80)  /* ascii? */
     res = c;
   else {
     int count = 0;  /* to count number of continuation bytes */
-    while (c & 0x40) {  /* still have continuation bytes? */
+    for (; c & 0x40; c <<= 1) {  /* while it needs continuation bytes... */
-      int cc = s[++count];  /* read next byte */
+      unsigned int cc = (unsigned char)s[++count];  /* read next byte */
       if ((cc & 0xC0) != 0x80)  /* not a continuation byte? */
         return NULL;  /* invalid byte sequence */
       res = (res << 6) | (cc & 0x3F);  /* add lower 6 bits from cont. byte */
-      c <<= 1;  /* to test next bit */
     }
-    res |= ((c & 0x7F) << (count * 5));  /* add first byte */
+    res |= ((utfint)(c & 0x7F) << (count * 5));  /* add first byte */
-    if (count > 3 || res > MAXUNICODE || res <= limits[count])
+    if (count > 5 || res > MAXUTF || res < limits[count])
       return NULL;  /* invalid byte sequence */
     s += count;  /* skip continuation bytes read */
   }
+  if (strict) {
+    /* check for invalid code points; too large or surrogates */
+    if (res > MAXUNICODE || (0xD800u <= res && res <= 0xDFFFu))
+      return NULL;
+  }
   if (val) *val = res;
-  return (const char *)s + 1;  /* +1 to include first byte */
+  return s + 1;  /* +1 to include first byte */
 }
 
 
 /*
-** utf8len(s [, i [, j]]) --> number of characters that start in the
+** utf8len(s [, i [, j [, lax]]]) --> number of characters that
-** range [i,j], or nil + current position if 's' is not well formed in
+** start in the range [i,j], or nil + current position if 's' is not
-** that interval
+** well formed in that interval
 */
 static int utflen (lua_State *L) {
-  int n = 0;
+  lua_Integer n = 0;  /* counter for the number of characters */
-  size_t len;
+  size_t len;  /* string length in bytes */
   const char *s = luaL_checklstring(L, 1, &len);
   lua_Integer posi = u_posrelat(luaL_optinteger(L, 2, 1), len);
   lua_Integer posj = u_posrelat(luaL_optinteger(L, 3, -1), len);
+  int lax = lua_toboolean(L, 4);
   luaL_argcheck(L, 1 <= posi && --posi <= (lua_Integer)len, 2,
-                   "initial position out of string");
+                   "initial position out of bounds");
   luaL_argcheck(L, --posj < (lua_Integer)len, 3,
-                   "final position out of string");
+                   "final position out of bounds");
   while (posi <= posj) {
-    const char *s1 = utf8_decode(s + posi, NULL);
+    const char *s1 = utf8_decode(s + posi, NULL, !lax);
     if (s1 == NULL) {  /* conversion error? */
-      lua_pushnil(L);  /* return nil ... */
+      luaL_pushfail(L);  /* return fail ... */
       lua_pushinteger(L, posi + 1);  /* ... and current position */
       return 2;
     }
@@ -94,28 +116,29 @@ static int utflen (lua_State *L) {
 
 
 /*
-** codepoint(s, [i, [j]])  -> returns codepoints for all characters
+** codepoint(s, [i, [j [, lax]]]) -> returns codepoints for all
-** that start in the range [i,j]
+** characters that start in the range [i,j]
 */
 static int codepoint (lua_State *L) {
   size_t len;
   const char *s = luaL_checklstring(L, 1, &len);
   lua_Integer posi = u_posrelat(luaL_optinteger(L, 2, 1), len);
   lua_Integer pose = u_posrelat(luaL_optinteger(L, 3, posi), len);
+  int lax = lua_toboolean(L, 4);
   int n;
   const char *se;
-  luaL_argcheck(L, posi >= 1, 2, "out of range");
+  luaL_argcheck(L, posi >= 1, 2, "out of bounds");
-  luaL_argcheck(L, pose <= (lua_Integer)len, 3, "out of range");
+  luaL_argcheck(L, pose <= (lua_Integer)len, 3, "out of bounds");
   if (posi > pose) return 0;  /* empty interval; return no values */
   if (pose - posi >= INT_MAX)  /* (lua_Integer -> int) overflow? */
     return luaL_error(L, "string slice too long");
-  n = (int)(pose -  posi) + 1;
+  n = (int)(pose -  posi) + 1;  /* upper bound for number of returns */
   luaL_checkstack(L, n, "string slice too long");
-  n = 0;
+  n = 0;  /* count the number of returns */
-  se = s + pose;
+  se = s + pose;  /* string end */
   for (s += posi - 1; s < se;) {
-    int code;
+    utfint code;
-    s = utf8_decode(s, &code);
+    s = utf8_decode(s, &code, !lax);
     if (s == NULL)
       return luaL_error(L, "invalid UTF-8 code");
     lua_pushinteger(L, code);
@@ -126,8 +149,8 @@ static int codepoint (lua_State *L) {
 
 
 static void pushutfchar (lua_State *L, int arg) {
-  lua_Integer code = luaL_checkinteger(L, arg);
+  lua_Unsigned code = (lua_Unsigned)luaL_checkinteger(L, arg);
-  luaL_argcheck(L, 0 <= code && code <= MAXUNICODE, arg, "value out of range");
+  luaL_argcheck(L, code <= MAXUTF, arg, "value out of range");
   lua_pushfstring(L, "%U", (long)code);
 }
 
@@ -164,7 +187,7 @@ static int byteoffset (lua_State *L) {
   lua_Integer posi = (n >= 0) ? 1 : len + 1;
   posi = u_posrelat(luaL_optinteger(L, 3, posi), len);
   luaL_argcheck(L, 1 <= posi && --posi <= (lua_Integer)len, 3,
-                   "position out of range");
+                   "position out of bounds");
   if (n == 0) {
     /* find beginning of current byte sequence */
     while (posi > 0 && iscont(s + posi)) posi--;
@@ -193,12 +216,12 @@ static int byteoffset (lua_State *L) {
   if (n == 0)  /* did it find given character? */
     lua_pushinteger(L, posi + 1);
   else  /* no such character */
-    lua_pushnil(L);
+    luaL_pushfail(L);
   return 1;
 }
 
 
-static int iter_aux (lua_State *L) {
+static int iter_aux (lua_State *L, int strict) {
   size_t len;
   const char *s = luaL_checklstring(L, 1, &len);
   lua_Integer n = lua_tointeger(L, 2) - 1;
@@ -211,9 +234,9 @@ static int iter_aux (lua_State *L) {
   if (n >= (lua_Integer)len)
     return 0;  /* no more codepoints */
   else {
-    int code;
+    utfint code;
-    const char *next = utf8_decode(s + n, &code);
+    const char *next = utf8_decode(s + n, &code, strict);
-    if (next == NULL || iscont(next))
+    if (next == NULL)
       return luaL_error(L, "invalid UTF-8 code");
     lua_pushinteger(L, n + 1);
     lua_pushinteger(L, code);
@@ -222,9 +245,19 @@ static int iter_aux (lua_State *L) {
 }
 
 
+static int iter_auxstrict (lua_State *L) {
+  return iter_aux(L, 1);
+}
+
+static int iter_auxlax (lua_State *L) {
+  return iter_aux(L, 0);
+}
+
+
 static int iter_codes (lua_State *L) {
+  int lax = lua_toboolean(L, 2);
   luaL_checkstring(L, 1);
-  lua_pushcfunction(L, iter_aux);
+  lua_pushcfunction(L, lax ? iter_auxlax : iter_auxstrict);
   lua_pushvalue(L, 1);
   lua_pushinteger(L, 0);
   return 3;
@@ -232,7 +265,7 @@ static int iter_codes (lua_State *L) {
 
 
 /* pattern to match a single UTF-8 character */
-#define UTF8PATT	"[\0-\x7F\xC2-\xF4][\x80-\xBF]*"
+#define UTF8PATT	"[\0-\x7F\xC2-\xFD][\x80-\xBF]*"
 
 
 static const luaL_Reg funcs[] = {

tests/lua53/literals.lua

@@ -56,16 +56,23 @@ assert("abc\z
 assert("\u{0}\u{00000000}\x00\0" == string.char(0, 0, 0, 0))
 
 -- limits for 1-byte sequences
-assert("\u{0}\u{7F}" == "\x00\z\x7F")
+assert("\u{0}\u{7F}" == "\x00\x7F")
 
 -- limits for 2-byte sequences
-assert("\u{80}\u{7FF}" == "\xC2\x80\z\xDF\xBF")
+assert("\u{80}\u{7FF}" == "\xC2\x80\xDF\xBF")
 
 -- limits for 3-byte sequences
-assert("\u{800}\u{FFFF}" ==   "\xE0\xA0\x80\z\xEF\xBF\xBF")
+assert("\u{800}\u{FFFF}" ==   "\xE0\xA0\x80\xEF\xBF\xBF")
 
 -- limits for 4-byte sequences
-assert("\u{10000}\u{10FFFF}" == "\xF0\x90\x80\x80\z\xF4\x8F\xBF\xBF")
+assert("\u{10000}\u{1FFFFF}" == "\xF0\x90\x80\x80\xF7\xBF\xBF\xBF")
+
+-- limits for 5-byte sequences
+assert("\u{200000}\u{3FFFFFF}" == "\xF8\x88\x80\x80\x80\xFB\xBF\xBF\xBF\xBF")
+
+-- limits for 6-byte sequences
+assert("\u{4000000}\u{7FFFFFFF}" ==
+       "\xFC\x84\x80\x80\x80\x80\xFD\xBF\xBF\xBF\xBF\xBF")
 
 
 -- Error in escape sequences
@@ -94,7 +101,7 @@ lexerror([["xyz\300"]], [[\300"]])
 lexerror([["   \256"]], [[\256"]])
 
 -- errors in UTF-8 sequences
-lexerror([["abc\u{110000}"]], [[abc\u{110000]])   -- too large
+lexerror([["abc\u{100000000}"]], [[abc\u{100000000]])   -- too large
 lexerror([["abc\u11r"]], [[abc\u1]])    -- missing '{'
 lexerror([["abc\u"]], [[abc\u"]])    -- missing '{'
 lexerror([["abc\u{11r"]], [[abc\u{11r]])    -- missing '}'

tests/lua53/utf8.lua

@@ -21,62 +21,59 @@ local justone = "^" .. utf8.charpattern .. "$"
 
 -- 't' is the list of codepoints of 's'
 local function checksyntax (s, t)
+  -- creates a string "return '\u{t[1]}...\u{t[n]}'"
   local ts = {"return '"}
   for i = 1, #t do ts[i + 1] = string.format("\\u{%x}", t[i]) end
   ts[#t + 2] = "'"
   ts = table.concat(ts)
+  -- its execution should result in 's'
   assert(assert(load(ts))() == s)
 end
 
-assert(utf8.offset("alo", 5) == nil)
+assert(not utf8.offset("alo", 5))
-assert(utf8.offset("alo", -4) == nil)
+assert(not utf8.offset("alo", -4))
 
--- 't' is the list of codepoints of 's'
+-- 'check' makes several tests over the validity of string 's'.
-local function check (s, t)
+-- 't' is the list of codepoints of 's'.
-  local l = utf8.len(s) 
+local function check (s, t, nonstrict)
+  local l = utf8.len(s, 1, -1, nonstrict)
   assert(#t == l and len(s) == l)
-  assert(utf8.char(table.unpack(t)) == s)
+  assert(utf8.char(table.unpack(t)) == s)   -- 't' and 's' are equivalent
 
   assert(utf8.offset(s, 0) == 1)
 
   checksyntax(s, t)
 
-  local t1 = {utf8.codepoint(s, 1, -1)}
+  -- creates new table with all codepoints of 's'
+  local t1 = {utf8.codepoint(s, 1, -1, nonstrict)}
   assert(#t == #t1)
-  for i = 1, #t do assert(t[i] == t1[i]) end
+  for i = 1, #t do assert(t[i] == t1[i]) end   -- 't' is equal to 't1'
 
-  for i = 1, l do
+  for i = 1, l do   -- for all codepoints
     local pi = utf8.offset(s, i)        -- position of i-th char
     local pi1 = utf8.offset(s, 2, pi)   -- position of next char
     assert(string.find(string.sub(s, pi, pi1 - 1), justone))
     assert(utf8.offset(s, -1, pi1) == pi)
     assert(utf8.offset(s, i - l - 1) == pi)
-    assert(pi1 - pi == #utf8.char(utf8.codepoint(s, pi)))
+    assert(pi1 - pi == #utf8.char(utf8.codepoint(s, pi, pi, nonstrict)))
-    for j = pi, pi1 - 1 do 
+    for j = pi, pi1 - 1 do
       assert(utf8.offset(s, 0, j) == pi)
     end
     for j = pi + 1, pi1 - 1 do
       assert(not utf8.len(s, j))
     end
-   assert(utf8.len(s, pi, pi) == 1)
+   assert(utf8.len(s, pi, pi, nonstrict) == 1)
-   assert(utf8.len(s, pi, pi1 - 1) == 1)
+   assert(utf8.len(s, pi, pi1 - 1, nonstrict) == 1)
-   assert(utf8.len(s, pi) == l - i + 1)
+   assert(utf8.len(s, pi, -1, nonstrict) == l - i + 1)
-   assert(utf8.len(s, pi1) == l - i)
+   assert(utf8.len(s, pi1, -1, nonstrict) == l - i)
-   assert(utf8.len(s, 1, pi) == i)
+   assert(utf8.len(s, 1, pi, nonstrict) == i)
   end
 
   local i = 0
-  for p, c in utf8.codes(s) do
+  for p, c in utf8.codes(s, nonstrict) do
     i = i + 1
     assert(c == t[i] and p == utf8.offset(s, i))
-    assert(utf8.codepoint(s, p) == c)
+    assert(utf8.codepoint(s, p, p, nonstrict) == c)
-  end
-  assert(i == #t)
-
-  i = 0
-  for p, c in utf8.codes(s) do
-    i = i + 1
-    assert(c == t[i] and p == utf8.offset(s, i)) 
   end
   assert(i == #t)
 
@@ -105,23 +102,30 @@ do    -- error indication in utf8.len
   check("\xF4\x9F\xBF\xBF", 1)
 end
 
--- error in utf8.codes
+-- errors in utf8.codes
-checkerror("invalid UTF%-8 code",
+do
-  function ()
+  local function errorcodes (s)
-    local s = "ab\xff"
+    checkerror("invalid UTF%-8 code",
-    for c in utf8.codes(s) do assert(c) end
+      function ()
-  end)
+        for c in utf8.codes(s) do assert(c) end
-
+      end)
+  end
+  errorcodes("ab\xff")
+  errorcodes("\u{110000}")
+end
 
 -- error in initial position for offset
-checkerror("position out of range", utf8.offset, "abc", 1, 5)
+checkerror("position out of bounds", utf8.offset, "abc", 1, 5)
-checkerror("position out of range", utf8.offset, "abc", 1, -4)
+checkerror("position out of bounds", utf8.offset, "abc", 1, -4)
-checkerror("position out of range", utf8.offset, "", 1, 2)
+checkerror("position out of bounds", utf8.offset, "", 1, 2)
-checkerror("position out of range", utf8.offset, "", 1, -1)
+checkerror("position out of bounds", utf8.offset, "", 1, -1)
 checkerror("continuation byte", utf8.offset, "𦧺", 1, 2)
 checkerror("continuation byte", utf8.offset, "𦧺", 1, 2)
 checkerror("continuation byte", utf8.offset, "\x80", 1)
 
+-- error in indices for len
+checkerror("out of bounds", utf8.len, "abc", 0, 2)
+checkerror("out of bounds", utf8.len, "abc", 1, 4)
 
 
 local s = "hello World"
@@ -136,19 +140,27 @@ do
   local t = {utf8.codepoint(s,1,#s - 1)}
   assert(#t == 3 and t[1] == 225 and t[2] == 233 and t[3] == 237)
   checkerror("invalid UTF%-8 code", utf8.codepoint, s, 1, #s)
-  checkerror("out of range", utf8.codepoint, s, #s + 1)
+  checkerror("out of bounds", utf8.codepoint, s, #s + 1)
   t = {utf8.codepoint(s, 4, 3)}
   assert(#t == 0)
-  checkerror("out of range", utf8.codepoint, s, -(#s + 1), 1)
+  checkerror("out of bounds", utf8.codepoint, s, -(#s + 1), 1)
-  checkerror("out of range", utf8.codepoint, s, 1, #s + 1)
+  checkerror("out of bounds", utf8.codepoint, s, 1, #s + 1)
+  -- surrogates
+  assert(utf8.codepoint("\u{D7FF}") == 0xD800 - 1)
+  assert(utf8.codepoint("\u{E000}") == 0xDFFF + 1)
+  assert(utf8.codepoint("\u{D800}", 1, 1, true) == 0xD800)
+  assert(utf8.codepoint("\u{DFFF}", 1, 1, true) == 0xDFFF)
+  assert(utf8.codepoint("\u{7FFFFFFF}", 1, 1, true) == 0x7FFFFFFF)
 end
 
 assert(utf8.char() == "")
-assert(utf8.char(97, 98, 99) == "abc")
+assert(utf8.char(0, 97, 98, 99, 1) == "\0abc\1")
 
 assert(utf8.codepoint(utf8.char(0x10FFFF)) == 0x10FFFF)
+assert(utf8.codepoint(utf8.char(0x7FFFFFFF), 1, 1, true) == (1<<31) - 1)
 
-checkerror("value out of range", utf8.char, 0x10FFFF + 1)
+checkerror("value out of range", utf8.char, 0x7FFFFFFF + 1)
+checkerror("value out of range", utf8.char, -1)
 
 local function invalid (s)
   checkerror("invalid UTF%-8 code", utf8.codepoint, s)
@@ -158,6 +170,10 @@ end
 -- UTF-8 representation for 0x11ffff (value out of valid range)
 invalid("\xF4\x9F\xBF\xBF")
 
+-- surrogates
+invalid("\u{D800}")
+invalid("\u{DFFF}")
+
 -- overlong sequences
 invalid("\xC0\x80")          -- zero
 invalid("\xC1\xBF")          -- 0x7F (should be coded in 1 byte)
@@ -183,6 +199,21 @@ s = "\0 \x7F\z
 s = string.gsub(s, " ", "")
 check(s, {0,0x7F, 0x80,0x7FF, 0x800,0xFFFF, 0x10000,0x10FFFF})
 
+do
+  -- original UTF-8 values
+  local s = "\u{4000000}\u{7FFFFFFF}"
+  assert(#s == 12)
+  check(s, {0x4000000, 0x7FFFFFFF}, true)
+
+  s = "\u{200000}\u{3FFFFFF}"
+  assert(#s == 10)
+  check(s, {0x200000, 0x3FFFFFF}, true)
+
+  s = "\u{10000}\u{1fffff}"
+  assert(#s == 8)
+  check(s, {0x10000, 0x1FFFFF}, true)
+end
+
 x = "日本語a-4\0éó"
 check(x, {26085, 26412, 35486, 97, 45, 52, 0, 233, 243})