Changes in the validation of UTF-8

All UTF-8 encoding functionality (including the escape
sequence '\u') accepts all values from the original UTF-8
specification (with sequences of up to six bytes).

By default, the decoding functions in the UTF-8 library do not
accept invalid Unicode code points, such as surrogates. A new
parameter 'nonstrict' makes them accept all code points up to
(2^31)-1, as in the original UTF-8 specification.

llex.c

@@ -335,7 +335,7 @@ static unsigned long readutf8esc (LexState *ls) {
   while ((save_and_next(ls), lisxdigit(ls->current))) {
     i++;
     r = (r << 4) + luaO_hexavalue(ls->current);
-    esccheck(ls, r <= 0x10FFFF, "UTF-8 value too large");
+    esccheck(ls, r <= 0x7FFFFFFFu, "UTF-8 value too large");
   }
   esccheck(ls, ls->current == '}', "missing '}'");
   next(ls);  /* skip '}' */

lobject.c

@@ -343,7 +343,7 @@ 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 */
@@ -435,9 +435,9 @@ const char *luaO_pushvfstring (lua_State *L, const char *fmt, va_list argp) {
         pushstr(L, buff, l);
         break;
       }
-      case 'U': {  /* an 'int' as a UTF-8 sequence */
+      case 'U': {  /* a 'long' 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;
       }

lutf8lib.c

@@ -21,12 +21,14 @@
 #include "lualib.h"
 
 
-#define MAXUNICODE	0x10FFFF
+#define MAXUNICODE	0x10FFFFu
+
+#define MAXUTF		0x7FFFFFFFu
 
 /*
-** Integer type for decoded UTF-8 values; MAXUNICODE needs 21 bits.
+** Integer type for decoded UTF-8 values; MAXUTF needs 31 bits.
 */
-#if LUAI_BITSINT >= 21
+#if LUAI_BITSINT >= 31
 typedef	unsigned int utfint;
 #else
 typedef unsigned long utfint;
@@ -46,38 +48,46 @@ 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, utfint *val) {
-  static const unsigned int limits[] = {0xFF, 0x7F, 0x7FF, 0xFFFF};
-  const unsigned char *s = (const unsigned char *)o;
-  unsigned int c = s[0];
+static const char *utf8_decode (const char *s, utfint *val, int strict) {
+  static const utfint limits[] =
+        {~(utfint)0, 0x80, 0x800, 0x10000u, 0x200000u, 0x4000000u};
+  unsigned int c = (unsigned char)s[0];
   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? */
-      int cc = s[++count];  /* read next byte */
+    for (; c & 0x40; c <<= 1) {  /* while it needs continuation bytes... */
+      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 |= ((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
-** range [i,j], or nil + current position if 's' is not well formed in
-** that interval
+** utf8len(s [, i [, j [, nonstrict]]]) --> number of characters that
+** start in the range [i,j], or nil + current position if 's' is not
+** well formed in that interval
 */
 static int utflen (lua_State *L) {
   lua_Integer n = 0;  /* counter for the number of characters */
@@ -85,12 +95,13 @@ static int utflen (lua_State *L) {
   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 nonstrict = lua_toboolean(L, 4);
   luaL_argcheck(L, 1 <= posi && --posi <= (lua_Integer)len, 2,
                    "initial position out of string");
   luaL_argcheck(L, --posj < (lua_Integer)len, 3,
                    "final position out of string");
   while (posi <= posj) {
-    const char *s1 = utf8_decode(s + posi, NULL);
+    const char *s1 = utf8_decode(s + posi, NULL, !nonstrict);
     if (s1 == NULL) {  /* conversion error? */
       lua_pushnil(L);  /* return nil ... */
       lua_pushinteger(L, posi + 1);  /* ... and current position */
@@ -105,14 +116,15 @@ static int utflen (lua_State *L) {
 
 
 /*
-** codepoint(s, [i, [j]])  -> returns codepoints for all characters
-** that start in the range [i,j]
+** codepoint(s, [i, [j [, nonstrict]]]) -> returns codepoints for all
+** 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 nonstrict = lua_toboolean(L, 4);
   int n;
   const char *se;
   luaL_argcheck(L, posi >= 1, 2, "out of range");
@@ -126,7 +138,7 @@ static int codepoint (lua_State *L) {
   se = s + pose;  /* string end */
   for (s += posi - 1; s < se;) {
     utfint code;
-    s = utf8_decode(s, &code);
+    s = utf8_decode(s, &code, !nonstrict);
     if (s == NULL)
       return luaL_error(L, "invalid UTF-8 code");
     lua_pushinteger(L, code);
@@ -137,8 +149,8 @@ static int codepoint (lua_State *L) {
 
 
 static void pushutfchar (lua_State *L, int arg) {
-  lua_Integer code = luaL_checkinteger(L, arg);
-  luaL_argcheck(L, 0 <= code && code <= MAXUNICODE, arg, "value out of range");
+  lua_Unsigned code = (lua_Unsigned)luaL_checkinteger(L, arg);
+  luaL_argcheck(L, code <= MAXUTF, arg, "value out of range");
   lua_pushfstring(L, "%U", (long)code);
 }
 
@@ -209,7 +221,7 @@ static int byteoffset (lua_State *L) {
 }
 
 
-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;
@@ -223,8 +235,8 @@ static int iter_aux (lua_State *L) {
     return 0;  /* no more codepoints */
   else {
     utfint code;
-    const char *next = utf8_decode(s + n, &code);
-    if (next == NULL || iscont(next))
+    const char *next = utf8_decode(s + n, &code, strict);
+    if (next == NULL)
       return luaL_error(L, "invalid UTF-8 code");
     lua_pushinteger(L, n + 1);
     lua_pushinteger(L, code);
@@ -233,9 +245,19 @@ static int iter_aux (lua_State *L) {
 }
 
 
+static int iter_auxstrict (lua_State *L) {
+  return iter_aux(L, 1);
+}
+
+static int iter_auxnostrict (lua_State *L) {
+  return iter_aux(L, 0);
+}
+
+
 static int iter_codes (lua_State *L) {
+  int nonstrict = lua_toboolean(L, 2);
   luaL_checkstring(L, 1);
-  lua_pushcfunction(L, iter_aux);
+  lua_pushcfunction(L, nonstrict ? iter_auxnostrict : iter_auxstrict);
   lua_pushvalue(L, 1);
   lua_pushinteger(L, 0);
   return 3;
@@ -243,7 +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[] = {

manual/manual.of

@@ -1004,6 +1004,8 @@ the escape sequence @T{\u{@rep{XXX}}}
 (note the mandatory enclosing brackets),
 where @rep{XXX} is a sequence of one or more hexadecimal digits
 representing the character code point.
+This code point can be any value smaller than @M{2@sp{31}}.
+(Lua uses the original UTF-8 specification here.)
 
 Literal strings can also be defined using a long format
 enclosed by @def{long brackets}.
@@ -6899,6 +6901,7 @@ x = string.gsub("$name-$version.tar.gz", "%$(%w+)", t)
 }
 
 @LibEntry{string.len (s)|
+
 Receives a string and returns its length.
 The empty string @T{""} has length 0.
 Embedded zeros are counted,
@@ -6907,6 +6910,7 @@ so @T{"a\000bc\000"} has length 5.
 }
 
 @LibEntry{string.lower (s)|
+
 Receives a string and returns a copy of this string with all
 uppercase letters changed to lowercase.
 All other characters are left unchanged.
@@ -6915,6 +6919,7 @@ The definition of what an uppercase letter is depends on the current locale.
 }
 
 @LibEntry{string.match (s, pattern [, init])|
+
 Looks for the first @emph{match} of
 @id{pattern} @see{pm} in the string @id{s}.
 If it finds one, then @id{match} returns
@@ -6946,6 +6951,7 @@ The format string cannot have the variable-length options
 }
 
 @LibEntry{string.rep (s, n [, sep])|
+
 Returns a string that is the concatenation of @id{n} copies of
 the string @id{s} separated by the string @id{sep}.
 The default value for @id{sep} is the empty string
@@ -6958,11 +6964,13 @@ with a single call to this function.)
 }
 
 @LibEntry{string.reverse (s)|
+
 Returns a string that is the string @id{s} reversed.
 
 }
 
 @LibEntry{string.sub (s, i [, j])|
+
 Returns the substring of @id{s} that
 starts at @id{i}  and continues until @id{j};
 @id{i} and @id{j} can be negative.
@@ -6998,6 +7006,7 @@ this function also returns the index of the first unread byte in @id{s}.
 }
 
 @LibEntry{string.upper (s)|
+
 Receives a string and returns a copy of this string with all
 lowercase letters changed to uppercase.
 All other characters are left unchanged.
@@ -7318,8 +7327,24 @@ or one plus the length of the subject string.
 As in the string library,
 negative indices count from the end of the string.
 
+Functions that create byte sequences
+accept all values up to @T{0x7FFFFFFF},
+as defined in the original UTF-8 specification;
+that implies byte sequences of up to six bytes.
+
+Functions that interpret byte sequences only accept
+valid sequences (well formed and not overlong).
+By default, they only accept byte sequences
+that result in valid Unicode code points,
+rejecting values larger than @T{10FFFF} and surrogates.
+A boolean argument @id{nonstrict}, when available,
+lifts these checks,
+so that all values up to @T{0x7FFFFFFF} are accepted.
+(Not well formed and overlong sequences are still rejected.)
+
 
 @LibEntry{utf8.char (@Cdots)|
+
 Receives zero or more integers,
 converts each one to its corresponding UTF-8 byte sequence
 and returns a string with the concatenation of all these sequences.
@@ -7327,14 +7352,15 @@ and returns a string with the concatenation of all these sequences.
 }
 
 @LibEntry{utf8.charpattern|
-The pattern (a string, not a function) @St{[\0-\x7F\xC2-\xF4][\x80-\xBF]*}
+
+The pattern (a string, not a function) @St{[\0-\x7F\xC2-\xFD][\x80-\xBF]*}
 @see{pm},
 which matches exactly one UTF-8 byte sequence,
 assuming that the subject is a valid UTF-8 string.
 
 }
 
-@LibEntry{utf8.codes (s)|
+@LibEntry{utf8.codes (s [, nonstrict])|
 
 Returns values so that the construction
 @verbatim{
@@ -7347,7 +7373,8 @@ It raises an error if it meets any invalid byte sequence.
 
 }
 
-@LibEntry{utf8.codepoint (s [, i [, j]])|
+@LibEntry{utf8.codepoint (s [, i [, j [, nonstrict]]])|
+
 Returns the codepoints (as integers) from all characters in @id{s}
 that start between byte position @id{i} and @id{j} (both included).
 The default for @id{i} is 1 and for @id{j} is @id{i}.
@@ -7355,7 +7382,8 @@ It raises an error if it meets any invalid byte sequence.
 
 }
 
-@LibEntry{utf8.len (s [, i [, j]])|
+@LibEntry{utf8.len (s [, i [, j [, nonstrict]]])|
+
 Returns the number of UTF-8 characters in string @id{s}
 that start between positions @id{i} and @id{j} (both inclusive).
 The default for @id{i} is @num{1} and for @id{j} is @num{-1}.
@@ -7365,6 +7393,7 @@ returns a false value plus the position of the first invalid byte.
 }
 
 @LibEntry{utf8.offset (s, n [, i])|
+
 Returns the position (in bytes) where the encoding of the
 @id{n}-th character of @id{s}
 (counting from position @id{i}) starts.
@@ -8755,6 +8784,12 @@ You can enclose the call in parentheses if you need to
 discard these extra results.
 }
 
+@item{
+By default, the decoding functions in the @Lid{utf8} library
+do not accept surrogates as valid code points.
+An extra parameter in these functions makes them more permissive.
+}
+
 }
 
 }

testes/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 '}'

testes/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(utf8.offset("alo", -4) == nil)
 
--- 't' is the list of codepoints of 's'
-local function check (s, t)
-  local l = utf8.len(s) 
+-- 'check' makes several tests over the validity of string 's'.
+-- 't' is the list of codepoints of '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)))
-    for j = pi, pi1 - 1 do 
+    assert(pi1 - pi == #utf8.char(utf8.codepoint(s, pi, pi, nonstrict)))
+    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, pi1 - 1) == 1)
-   assert(utf8.len(s, pi) == l - i + 1)
-   assert(utf8.len(s, pi1) == l - i)
-   assert(utf8.len(s, 1, pi) == i)
+   assert(utf8.len(s, pi, pi, nonstrict) == 1)
+   assert(utf8.len(s, pi, pi1 - 1, nonstrict) == 1)
+   assert(utf8.len(s, pi, -1, nonstrict) == l - i + 1)
+   assert(utf8.len(s, pi1, -1, nonstrict) == l - i)
+   assert(utf8.len(s, 1, pi, -1, 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)
-  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)) 
+    assert(utf8.codepoint(s, p, p, nonstrict) == c)
   end
   assert(i == #t)
 
@@ -105,13 +102,17 @@ do    -- error indication in utf8.len
   check("\xF4\x9F\xBF\xBF", 1)
 end
 
--- error in utf8.codes
-checkerror("invalid UTF%-8 code",
-  function ()
-    local s = "ab\xff"
-    for c in utf8.codes(s) do assert(c) end
-  end)
-
+-- errors in utf8.codes
+do
+  local function errorcodes (s)
+    checkerror("invalid UTF%-8 code",
+      function ()
+        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)
@@ -141,14 +142,22 @@ do
   assert(#t == 0)
   checkerror("out of range", utf8.codepoint, s, -(#s + 1), 1)
   checkerror("out of range", 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 +167,10 @@ 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 +196,21 @@ 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})