static bool
InflateUTF8StringToBuffer(JSContext *cx, const UTF8Chars src, jschar *dst, size_t *dstlenp,
bool *isAsciip)
{
*isAsciip = true;
// First, count how many jschars need to be in the inflated string.
// |i| is the index into |src|, and |j| is the the index into |dst|.
size_t srclen = src.length();
uint32_t j = 0;
for (uint32_t i = 0; i < srclen; i++, j++) {
uint32_t v = uint32_t(src[i]);
if (!(v & 0x80)) {
// ASCII code unit. Simple copy.
if (action == Copy)
dst[j] = jschar(v);
} else {
// Non-ASCII code unit. Determine its length in bytes (n).
*isAsciip = false;
uint32_t n = 1;
while (v & (0x80 >> n))
n++;
#define INVALID(report, arg, n2) \
do { \
if (action == CountAndReportInvalids) { \
report(cx, arg); \
return false; \
} else { \
if (action == Copy) \
dst[j] = jschar(REPLACE_UTF8); \
else \
JS_ASSERT(action == CountAndIgnoreInvalids); \
n = n2; \
goto invalidMultiByteCodeUnit; \
} \
} while (0)
// Check the leading byte.
if (n < 2 || n > 4)
INVALID(ReportInvalidCharacter, i, 1);
// Check that |src| is large enough to hold an n-byte code unit.
if (i + n > srclen)
INVALID(ReportBufferTooSmall, /* dummy = */ 0, 1);
// Check the second byte. From Unicode Standard v6.2, Table 3-7
// Well-Formed UTF-8 Byte Sequences.
if ((v == 0xE0 && ((uint8_t)src[i + 1] & 0xE0) != 0xA0) || // E0 A0~BF
(v == 0xED && ((uint8_t)src[i + 1] & 0xE0) != 0x80) || // ED 80~9F
(v == 0xF0 && ((uint8_t)src[i + 1] & 0xF0) == 0x80) || // F0 90~BF
(v == 0xF4 && ((uint8_t)src[i + 1] & 0xF0) != 0x80)) // F4 80~8F
{
INVALID(ReportInvalidCharacter, i, 1);
}
// Check the continuation bytes.
for (uint32_t m = 1; m < n; m++)
if ((src[i + m] & 0xC0) != 0x80)
INVALID(ReportInvalidCharacter, i, m);
// Determine the code unit's length in jschars and act accordingly.
v = Utf8ToOneUcs4Char((uint8_t *)&src[i], n);
if (v < 0x10000) {
// The n-byte UTF8 code unit will fit in a single jschar.
if (action == Copy)
dst[j] = jschar(v);
} else {
v -= 0x10000;
if (v <= 0xFFFFF) {
// The n-byte UTF8 code unit will fit in two jschars.
if (action == Copy)
dst[j] = jschar((v >> 10) + 0xD800);
j++;
if (action == Copy)
dst[j] = jschar((v & 0x3FF) + 0xDC00);
} else {
// The n-byte UTF8 code unit won't fit in two jschars.
INVALID(ReportTooBigCharacter, v, 1);
}
}
invalidMultiByteCodeUnit:
// Move i to the last byte of the multi-byte code unit; the loop
// header will do the final i++ to move to the start of the next
// code unit.
i += n - 1;
}
}
static bool
InflateUTF8StringToBuffer(ContextT* cx, const UTF8Chars src, CharT* dst, size_t* dstlenp,
JS::SmallestEncoding *smallestEncoding)
{
if (Action != AssertNoInvalids)
*smallestEncoding = JS::SmallestEncoding::ASCII;
auto RequireLatin1 = [&smallestEncoding]{
*smallestEncoding = std::max(JS::SmallestEncoding::Latin1, *smallestEncoding);
};
auto RequireUTF16 = [&smallestEncoding]{
*smallestEncoding = JS::SmallestEncoding::UTF16;
};
// Count how many code units need to be in the inflated string.
// |i| is the index into |src|, and |j| is the the index into |dst|.
size_t srclen = src.length();
uint32_t j = 0;
for (uint32_t i = 0; i < srclen; i++, j++) {
uint32_t v = uint32_t(src[i]);
if (!(v & 0x80)) {
// ASCII code unit. Simple copy.
if (Action == Copy)
dst[j] = CharT(v);
} else {
// Non-ASCII code unit. Determine its length in bytes (n).
uint32_t n = 1;
while (v & (0x80 >> n))
n++;
#define INVALID(report, arg, n2) \
do { \
if (Action == CountAndReportInvalids) { \
report(cx, arg); \
return false; \
} else if (Action == AssertNoInvalids) { \
MOZ_CRASH("invalid UTF-8 string: " # report); \
} else { \
if (Action == Copy) { \
if (std::is_same<decltype(dst[0]), Latin1Char>::value) \
dst[j] = CharT(REPLACE_UTF8_LATIN1); \
else \
dst[j] = CharT(REPLACE_UTF8); \
} else { \
MOZ_ASSERT(Action == CountAndIgnoreInvalids || \
Action == FindEncoding); \
} \
n = n2; \
goto invalidMultiByteCodeUnit; \
} \
} while (0)
// Check the leading byte.
if (n < 2 || n > 4)
INVALID(ReportInvalidCharacter, i, 1);
// Check that |src| is large enough to hold an n-byte code unit.
if (i + n > srclen)
INVALID(ReportBufferTooSmall, /* dummy = */ 0, 1);
// Check the second byte. From Unicode Standard v6.2, Table 3-7
// Well-Formed UTF-8 Byte Sequences.
if ((v == 0xE0 && ((uint8_t)src[i + 1] & 0xE0) != 0xA0) || // E0 A0~BF
(v == 0xED && ((uint8_t)src[i + 1] & 0xE0) != 0x80) || // ED 80~9F
(v == 0xF0 && ((uint8_t)src[i + 1] & 0xF0) == 0x80) || // F0 90~BF
(v == 0xF4 && ((uint8_t)src[i + 1] & 0xF0) != 0x80)) // F4 80~8F
{
INVALID(ReportInvalidCharacter, i, 1);
}
// Check the continuation bytes.
for (uint32_t m = 1; m < n; m++) {
if ((src[i + m] & 0xC0) != 0x80)
INVALID(ReportInvalidCharacter, i, m);
}
// Determine the code unit's length in CharT and act accordingly.
v = JS::Utf8ToOneUcs4Char((uint8_t*)&src[i], n);
if (Action != AssertNoInvalids) {
if (v > 0xff) {
RequireUTF16();
if (Action == FindEncoding) {
MOZ_ASSERT(dst == nullptr);
return true;
}
} else {
RequireLatin1();
}
}
if (v < 0x10000) {
// The n-byte UTF8 code unit will fit in a single CharT.
if (Action == Copy)
dst[j] = CharT(v);
} else {
v -= 0x10000;
if (v <= 0xFFFFF) {
// The n-byte UTF8 code unit will fit in two CharT units.
if (Action == Copy)
dst[j] = CharT((v >> 10) + 0xD800);
j++;
if (Action == Copy)
dst[j] = CharT((v & 0x3FF) + 0xDC00);
} else {
// The n-byte UTF8 code unit won't fit in two CharT units.
INVALID(ReportTooBigCharacter, v, 1);
}
}
invalidMultiByteCodeUnit:
// Move i to the last byte of the multi-byte code unit; the loop
// header will do the final i++ to move to the start of the next
// code unit.
i += n - 1;
if (Action != AssertNoInvalids)
RequireUTF16();
}
}
