Browse Source

Bug: 'utf8.codes' accepts spurious continuation bytes

pull/28/head
Roberto Ierusalimschy 2 years ago
parent
commit
a1089b415a
  1. 27
      lutf8lib.c
  2. 12
      testes/utf8.lua

27
lutf8lib.c

@ -25,6 +25,9 @@
#define MAXUTF 0x7FFFFFFFu
#define MSGInvalid "invalid UTF-8 code"
/*
** Integer type for decoded UTF-8 values; MAXUTF needs 31 bits.
*/
@ -35,7 +38,8 @@ typedef unsigned long utfint;
#endif
#define iscont(p) ((*(p) & 0xC0) == 0x80)
#define iscont(c) (((c) & 0xC0) == 0x80)
#define iscontp(p) iscont(*(p))
/* from strlib */
@ -65,7 +69,7 @@ static const char *utf8_decode (const char *s, utfint *val, int strict) {
int count = 0; /* to count number of continuation bytes */
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? */
if (!iscont(cc)) /* not a continuation byte? */
return NULL; /* invalid byte sequence */
res = (res << 6) | (cc & 0x3F); /* add lower 6 bits from cont. byte */
}
@ -140,7 +144,7 @@ static int codepoint (lua_State *L) {
utfint code;
s = utf8_decode(s, &code, !lax);
if (s == NULL)
return luaL_error(L, "invalid UTF-8 code");
return luaL_error(L, MSGInvalid);
lua_pushinteger(L, code);
n++;
}
@ -190,16 +194,16 @@ static int byteoffset (lua_State *L) {
"position out of bounds");
if (n == 0) {
/* find beginning of current byte sequence */
while (posi > 0 && iscont(s + posi)) posi--;
while (posi > 0 && iscontp(s + posi)) posi--;
}
else {
if (iscont(s + posi))
if (iscontp(s + posi))
return luaL_error(L, "initial position is a continuation byte");
if (n < 0) {
while (n < 0 && posi > 0) { /* move back */
do { /* find beginning of previous character */
posi--;
} while (posi > 0 && iscont(s + posi));
} while (posi > 0 && iscontp(s + posi));
n++;
}
}
@ -208,7 +212,7 @@ static int byteoffset (lua_State *L) {
while (n > 0 && posi < (lua_Integer)len) {
do { /* find beginning of next character */
posi++;
} while (iscont(s + posi)); /* (cannot pass final '\0') */
} while (iscontp(s + posi)); /* (cannot pass final '\0') */
n--;
}
}
@ -226,15 +230,15 @@ static int iter_aux (lua_State *L, int strict) {
const char *s = luaL_checklstring(L, 1, &len);
lua_Unsigned n = (lua_Unsigned)lua_tointeger(L, 2);
if (n < len) {
while (iscont(s + n)) n++; /* skip continuation bytes */
while (iscontp(s + n)) n++; /* go to next character */
}
if (n >= len) /* (also handles original 'n' being negative) */
return 0; /* no more codepoints */
else {
utfint code;
const char *next = utf8_decode(s + n, &code, strict);
if (next == NULL)
return luaL_error(L, "invalid UTF-8 code");
if (next == NULL || iscontp(next))
return luaL_error(L, MSGInvalid);
lua_pushinteger(L, n + 1);
lua_pushinteger(L, code);
return 2;
@ -253,7 +257,8 @@ static int iter_auxlax (lua_State *L) {
static int iter_codes (lua_State *L) {
int lax = lua_toboolean(L, 2);
luaL_checkstring(L, 1);
const char *s = luaL_checkstring(L, 1);
luaL_argcheck(L, !iscontp(s), 1, MSGInvalid);
lua_pushcfunction(L, lax ? iter_auxlax : iter_auxstrict);
lua_pushvalue(L, 1);
lua_pushinteger(L, 0);

12
testes/utf8.lua

@ -97,9 +97,15 @@ do -- error indication in utf8.len
assert(not a and b == p)
end
check("abc\xE3def", 4)
check("汉字\x80", #("汉字") + 1)
check("\xF4\x9F\xBF", 1)
check("\xF4\x9F\xBF\xBF", 1)
-- spurious continuation bytes
check("汉字\x80", #("汉字") + 1)
check("\x80hello", 1)
check("hel\x80lo", 4)
check("汉字\xBF", #("汉字") + 1)
check("\xBFhello", 1)
check("hel\xBFlo", 4)
end
-- errors in utf8.codes
@ -112,12 +118,16 @@ do
end
errorcodes("ab\xff")
errorcodes("\u{110000}")
errorcodes("in\x80valid")
errorcodes("\xbfinvalid")
errorcodes("αλφ\xBFα")
-- calling interation function with invalid arguments
local f = utf8.codes("")
assert(f("", 2) == nil)
assert(f("", -1) == nil)
assert(f("", math.mininteger) == nil)
end
-- error in initial position for offset

Loading…
Cancel
Save