static const char unees[] =
"Malformed UTF-8 character (unexpected end of string)";
static const char cp_above_legal_max[] =
- "Use of code point 0x%" UVXf " is deprecated; the permissible max is 0x%" UVXf ". This will be fatal in Perl 5.28";
+ "Use of code point 0x%" UVXf " is not allowed; the"
+ " permissible max is 0x%" UVXf;
-#define MAX_NON_DEPRECATED_CP ((UV) (IV_MAX))
+#define MAX_EXTERNALLY_LEGAL_CP ((UV) (IV_MAX))
/*
=head1 Unicode Support
" is not recommended for open interchange";
const char super_cp_format[] = "Code point 0x%" UVXf " is not Unicode,"
" may not be portable";
-const char above_31_bit_cp_format[] = "Code point 0x%" UVXf " is not"
- " Unicode, and not portable";
+const char perl_extended_cp_format[] = "Code point 0x%" UVXf " is not" \
+ " Unicode, requires a Perl extension," \
+ " and so is not portable";
#define HANDLE_UNICODE_SURROGATE(uv, flags) \
STMT_START { \
* performance hit on these high EBCDIC code points. */
if (UNLIKELY(UNICODE_IS_SUPER(uv))) {
- if ( UNLIKELY(uv > MAX_NON_DEPRECATED_CP)
- && ckWARN_d(WARN_DEPRECATED))
- {
- Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
- cp_above_legal_max, uv, MAX_NON_DEPRECATED_CP);
+ if (UNLIKELY(uv > MAX_EXTERNALLY_LEGAL_CP)) {
+ Perl_croak(aTHX_ cp_above_legal_max, uv, MAX_EXTERNALLY_LEGAL_CP);
}
- if ( (flags & UNICODE_WARN_SUPER)
- || ( UNICODE_IS_PERL_EXTENDED(uv)
- && (flags & UNICODE_WARN_PERL_EXTENDED)))
+ if ( (flags & UNICODE_WARN_SUPER)
+ || ( (flags & UNICODE_WARN_PERL_EXTENDED)
+ && UNICODE_IS_PERL_EXTENDED(uv)))
{
Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
/* Choose the more dire applicable warning */
(UNICODE_IS_PERL_EXTENDED(uv))
- ? above_31_bit_cp_format
+ ? perl_extended_cp_format
: super_cp_format,
uv);
}
if ( (flags & UNICODE_DISALLOW_SUPER)
- || ( UNICODE_IS_PERL_EXTENDED(uv)
- && (flags & UNICODE_DISALLOW_PERL_EXTENDED)))
+ || ( (flags & UNICODE_DISALLOW_PERL_EXTENDED)
+ && UNICODE_IS_PERL_EXTENDED(uv)))
{
return NULL;
}
*(d++) = uv;
-This function accepts any UV as input, but very high code points (above
-C<IV_MAX> on the platform) will raise a deprecation warning. This is
-typically 0x7FFF_FFFF in a 32-bit word.
+This function accepts any code point from 0..C<IV_MAX> as input.
+C<IV_MAX> is typically 0x7FFF_FFFF in a 32-bit word.
It is possible to forbid or warn on non-Unicode code points, or those that may
be problematic by using L</uvchr_to_utf8_flags>.
*(d++) = uv;
-If C<flags> is 0, this function accepts any UV as input, but very high code
-points (above C<IV_MAX> for the platform) will raise a deprecation warning.
-This is typically 0x7FFF_FFFF in a 32-bit word.
+If C<flags> is 0, this function accepts any code point from 0..C<IV_MAX> as
+input. C<IV_MAX> is typically 0x7FFF_FFFF in a 32-bit word.
Specifying C<flags> can further restrict what is allowed and not warned on, as
follows:
L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>.
See L<perlunicode/Noncharacter code points>.
-Code points above 0x7FFF_FFFF (2**31 - 1) were never specified in any standard,
-so using them is more problematic than other above-Unicode code points. Perl
-invented an extension to UTF-8 to represent the ones above 2**36-1, so it is
-likely that non-Perl languages will not be able to read files that contain
-these that written by the perl interpreter; nor would Perl understand files
-written by something that uses a different extension. For these reasons, there
-is a separate set of flags that can warn and/or disallow these extremely high
-code points, even if other above-Unicode ones are accepted. These are the
-C<UNICODE_WARN_ABOVE_31_BIT> and C<UNICODE_DISALLOW_ABOVE_31_BIT> flags. These
-are entirely independent from the deprecation warning for code points above
-C<IV_MAX>. On 32-bit machines, it will eventually be forbidden to have any
-code point that needs more than 31 bits to represent. When that happens,
-effectively the C<UNICODE_DISALLOW_ABOVE_31_BIT> flag will always be set on
-32-bit machines. (Of course C<UNICODE_DISALLOW_SUPER> will treat all
-above-Unicode code points, including these, as malformations; and
-C<UNICODE_WARN_SUPER> warns on these.)
-
-On EBCDIC platforms starting in Perl v5.24, the Perl extension for representing
-extremely high code points kicks in at 0x3FFF_FFFF (2**30 -1), which is lower
-than on ASCII. Prior to that, code points 2**31 and higher were simply
-unrepresentable, and a different, incompatible method was used to represent
-code points between 2**30 and 2**31 - 1. The flags C<UNICODE_WARN_ABOVE_31_BIT>
-and C<UNICODE_DISALLOW_ABOVE_31_BIT> have the same function as on ASCII
-platforms, warning and disallowing 2**31 and higher.
+Extremely high code points were never specified in any standard, and require an
+extension to UTF-8 to express, which Perl does. It is likely that programs
+written in something other than Perl would not be able to read files that
+contain these; nor would Perl understand files written by something that uses a
+different extension. For these reasons, there is a separate set of flags that
+can warn and/or disallow these extremely high code points, even if other
+above-Unicode ones are accepted. They are the C<UNICODE_WARN_PERL_EXTENDED>
+and C<UNICODE_DISALLOW_PERL_EXTENDED> flags. For more information see
+L</C<UTF8_GOT_PERL_EXTENDED>>. Of course C<UNICODE_DISALLOW_SUPER> will
+treat all above-Unicode code points, including these, as malformations. (Note
+that the Unicode standard considers anything above 0x10FFFF to be illegal, but
+there are standards predating it that allow up to 0x7FFF_FFFF (2**31 -1))
+
+A somewhat misleadingly named synonym for C<UNICODE_WARN_PERL_EXTENDED> is
+retained for backward compatibility: C<UNICODE_WARN_ABOVE_31_BIT>. Similarly,
+C<UNICODE_DISALLOW_ABOVE_31_BIT> is usable instead of the more accurately named
+C<UNICODE_DISALLOW_PERL_EXTENDED>. The names are misleading because these
+flags can apply to code points that actually do fit in 31 bits. This happens
+on EBCDIC platforms, and sometimes when the L<overlong
+malformation|/C<UTF8_GOT_LONG>> is also present. The new names accurately
+describe the situation in all cases.
=cut
*/
return uvchr_to_utf8_flags(d, uv, flags);
}
-PERL_STATIC_INLINE bool
-S_is_utf8_cp_above_31_bits(const U8 * const s, const U8 * const e)
+#ifndef UV_IS_QUAD
+
+STATIC int
+S_is_utf8_cp_above_31_bits(const U8 * const s,
+ const U8 * const e,
+ const bool consider_overlongs)
{
/* Returns TRUE if the first code point represented by the Perl-extended-
* UTF-8-encoded string starting at 's', and looking no further than 'e -
* the final ones necessary for the complete representation may be beyond
* 'e - 1'.
*
- * The function assumes that the sequence is well-formed UTF-8 as far as it
- * goes, and is for a UTF-8 variant code point. If the sequence is
- * incomplete, the function returns FALSE if there is any well-formed
- * UTF-8 byte sequence that can complete it in such a way that a code point
- * < 2**31 is produced; otherwise it returns TRUE.
- *
- * Getting this exactly right is slightly tricky, and has to be done in
- * several places in this file, so is centralized here. It is based on the
- * following table:
+ * The function also can handle the case where the input is an overlong
+ * sequence. If 'consider_overlongs' is 0, the function assumes the
+ * input is not overlong, without checking, and will return based on that
+ * assumption. If this parameter is 1, the function will go to the trouble
+ * of figuring out if it actually evaluates to above or below 31 bits.
*
- * U+7FFFFFFF (2 ** 31 - 1)
- * ASCII: \xFD\xBF\xBF\xBF\xBF\xBF
- * IBM-1047: \xFE\x41\x41\x41\x41\x41\x41\x42\x73\x73\x73\x73\x73\x73
- * IBM-037: \xFE\x41\x41\x41\x41\x41\x41\x42\x72\x72\x72\x72\x72\x72
- * POSIX-BC: \xFE\x41\x41\x41\x41\x41\x41\x42\x75\x75\x75\x75\x75\x75
- * I8: \xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA1\xBF\xBF\xBF\xBF\xBF\xBF
- * U+80000000 (2 ** 31):
- * ASCII: \xFE\x82\x80\x80\x80\x80\x80
- * [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] 10 11 12 13
- * IBM-1047: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41
- * IBM-037: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41
- * POSIX-BC: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41
- * I8: \xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA2\xA0\xA0\xA0\xA0\xA0\xA0
+ * The sequence is otherwise assumed to be well-formed, without checking.
*/
-#ifdef EBCDIC
-
- /* [0] is start byte [1] [2] [3] [4] [5] [6] [7] */
- const U8 prefix[] = "\x41\x41\x41\x41\x41\x41\x42";
- const STRLEN prefix_len = sizeof(prefix) - 1;
const STRLEN len = e - s;
- const STRLEN cmp_len = MIN(prefix_len, len - 1);
+ int is_overlong;
-#else
-
- PERL_UNUSED_ARG(e);
+ PERL_ARGS_ASSERT_IS_UTF8_CP_ABOVE_31_BITS;
-#endif
+ assert(! UTF8_IS_INVARIANT(*s) && e > s);
- PERL_ARGS_ASSERT_IS_UTF8_CP_ABOVE_31_BITS;
+#ifdef EBCDIC
- assert(! UTF8_IS_INVARIANT(*s));
+ PERL_UNUSED_ARG(consider_overlongs);
-#ifndef EBCDIC
+ /* On the EBCDIC code pages we handle, only the native start byte 0xFE can
+ * mean a 32-bit or larger code point (0xFF is an invariant). 0xFE can
+ * also be the start byte for a 31-bit code point; we need at least 2
+ * bytes, and maybe up through 8 bytes, to determine that. (It can also be
+ * the start byte for an overlong sequence, but for 30-bit or smaller code
+ * points, so we don't have to worry about overlongs on EBCDIC.) */
+ if (*s != 0xFE) {
+ return 0;
+ }
- /* Technically, a start byte of FE can be for a code point that fits into
- * 31 bytes, but not for well-formed UTF-8: doing that requires an overlong
- * malformation. */
- return (*s >= 0xFE);
+ if (len == 1) {
+ return -1;
+ }
#else
- /* On the EBCDIC code pages we handle, only the native start byte 0xFE can
- * mean a 32-bit or larger code point (0xFF is an invariant). For 0xFE, we
- * need at least 2 bytes, and maybe up through 8 bytes, to be sure that the
- * value is above 31 bits. */
- if (*s != 0xFE || len == 1) {
- return FALSE;
+ /* On ASCII, FE and FF are the only start bytes that can evaluate to
+ * needing more than 31 bits. */
+ if (LIKELY(*s < 0xFE)) {
+ return 0;
}
- /* Note that in UTF-EBCDIC, the two lowest possible continuation bytes are
- * \x41 and \x42. */
- return cBOOL(memGT(s + 1, prefix, cmp_len));
-
-#endif
+ /* What we have left are FE and FF. Both of these require more than 31
+ * bits unless they are for overlongs. */
+ if (! consider_overlongs) {
+ return 1;
+ }
-}
+ /* Here, we have FE or FF. If the input isn't overlong, it evaluates to
+ * above 31 bits. But we need more than one byte to discern this, so if
+ * passed just the start byte, it could be an overlong evaluating to
+ * smaller */
+ if (len == 1) {
+ return -1;
+ }
-/* Anything larger than this will overflow the word if it were converted into a UV */
-#if defined(UV_IS_QUAD)
-# ifdef EBCDIC /* Actually is I8 */
-# define HIGHEST_REPRESENTABLE_UTF8 \
- "\xFF\xAF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF"
-# else
-# define HIGHEST_REPRESENTABLE_UTF8 \
- "\xFF\x80\x8F\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF"
-# endif
-#else /* 32-bit */
-# ifdef EBCDIC
-# define HIGHEST_REPRESENTABLE_UTF8 \
- "\xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA3\xBF\xBF\xBF\xBF\xBF\xBF"
-# else
-# define HIGHEST_REPRESENTABLE_UTF8 "\xFE\x83\xBF\xBF\xBF\xBF\xBF"
-# endif
-#endif
+ /* Having excluded len==1, and knowing that FE and FF are both valid start
+ * bytes, we can call the function below to see if the sequence is
+ * overlong. (We don't need the full generality of the called function,
+ * but for these huge code points, speed shouldn't be a consideration, and
+ * the compiler does have enough information, since it's static to this
+ * file, to optimize to just the needed parts.) */
+ is_overlong = is_utf8_overlong_given_start_byte_ok(s, len);
-PERL_STATIC_INLINE bool
-S_does_utf8_overflow(const U8 * const s, const U8 * e)
-{
- const U8 *x;
- const U8 * y = (const U8 *) HIGHEST_REPRESENTABLE_UTF8;
+ /* If it isn't overlong, more than 31 bits are required. */
+ if (is_overlong == 0) {
+ return 1;
+ }
-#if ! defined(UV_IS_QUAD) && ! defined(EBCDIC)
+ /* If it is indeterminate if it is overlong, return that */
+ if (is_overlong < 0) {
+ return -1;
+ }
- const STRLEN len = e - s;
+ /* Here is overlong. Such a sequence starting with FE is below 31 bits, as
+ * the max it can be is 2**31 - 1 */
+ if (*s == 0xFE) {
+ return 0;
+ }
#endif
- /* Returns a boolean as to if this UTF-8 string would overflow a UV on this
- * platform, that is if it represents a code point larger than the highest
- * representable code point. (For ASCII platforms, we could use memcmp()
- * because we don't have to convert each byte to I8, but it's very rare
- * input indeed that would approach overflow, so the loop below will likely
- * only get executed once.
- *
- * 'e' must not be beyond a full character. If it is less than a full
- * character, the function returns FALSE if there is any input beyond 'e'
- * that could result in a non-overflowing code point */
-
- PERL_ARGS_ASSERT_DOES_UTF8_OVERFLOW;
- assert(s <= e && s + UTF8SKIP(s) >= e);
+ /* Here, ASCII and EBCDIC rejoin:
+ * On ASCII: We have an overlong sequence starting with FF
+ * On EBCDIC: We have a sequence starting with FE. */
-#if ! defined(UV_IS_QUAD) && ! defined(EBCDIC)
+ { /* For C89, use a block so the declaration can be close to its use */
- /* On 32 bit ASCII machines, many overlongs that start with FF don't
- * overflow */
+#ifdef EBCDIC
- if (isFF_OVERLONG(s, len)) {
- const U8 max_32_bit_overlong[] = "\xFF\x80\x80\x80\x80\x80\x80\x84";
- return memGE(s, max_32_bit_overlong,
- MIN(len, sizeof(max_32_bit_overlong) - 1));
- }
+ /* U+7FFFFFFF (2 ** 31 - 1)
+ * [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] 10 11 12 13
+ * IBM-1047: \xFE\x41\x41\x41\x41\x41\x41\x42\x73\x73\x73\x73\x73\x73
+ * IBM-037: \xFE\x41\x41\x41\x41\x41\x41\x42\x72\x72\x72\x72\x72\x72
+ * POSIX-BC: \xFE\x41\x41\x41\x41\x41\x41\x42\x75\x75\x75\x75\x75\x75
+ * I8: \xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA1\xBF\xBF\xBF\xBF\xBF\xBF
+ * U+80000000 (2 ** 31):
+ * IBM-1047: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41
+ * IBM-037: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41
+ * POSIX-BC: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41
+ * I8: \xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA2\xA0\xA0\xA0\xA0\xA0\xA0
+ *
+ * and since we know that *s = \xfe, any continuation sequcence
+ * following it that is gt the below is above 31 bits
+ [0] [1] [2] [3] [4] [5] [6] */
+ const U8 conts_for_highest_30_bit[] = "\x41\x41\x41\x41\x41\x41\x42";
-#endif
+#else
- for (x = s; x < e; x++, y++) {
+ /* FF overlong for U+7FFFFFFF (2 ** 31 - 1)
+ * ASCII: \xFF\x80\x80\x80\x80\x80\x80\x81\xBF\xBF\xBF\xBF\xBF
+ * FF overlong for U+80000000 (2 ** 31):
+ * ASCII: \xFF\x80\x80\x80\x80\x80\x80\x82\x80\x80\x80\x80\x80
+ * and since we know that *s = \xff, any continuation sequcence
+ * following it that is gt the below is above 30 bits
+ [0] [1] [2] [3] [4] [5] [6] */
+ const U8 conts_for_highest_30_bit[] = "\x80\x80\x80\x80\x80\x80\x81";
- /* If this byte is larger than the corresponding highest UTF-8
- * byte, it overflows */
- if (UNLIKELY(NATIVE_UTF8_TO_I8(*x) > *y)) {
- return TRUE;
- }
- /* If not the same as this byte, it must be smaller, doesn't
- * overflow */
- if (LIKELY(NATIVE_UTF8_TO_I8(*x) != *y)) {
- return FALSE;
- }
+#endif
+ const STRLEN conts_len = sizeof(conts_for_highest_30_bit) - 1;
+ const STRLEN cmp_len = MIN(conts_len, len - 1);
+
+ /* Now compare the continuation bytes in s with the ones we have
+ * compiled in that are for the largest 30 bit code point. If we have
+ * enough bytes available to determine the answer, or the bytes we do
+ * have differ from them, we can compare the two to get a definitive
+ * answer (Note that in UTF-EBCDIC, the two lowest possible
+ * continuation bytes are \x41 and \x42.) */
+ if (cmp_len >= conts_len || memNE(s + 1,
+ conts_for_highest_30_bit,
+ cmp_len))
+ {
+ return cBOOL(memGT(s + 1, conts_for_highest_30_bit, cmp_len));
}
- /* Got to the end and all bytes are the same. If the input is a whole
- * character, it doesn't overflow. And if it is a partial character,
- * there's not enough information to tell, so assume doesn't overflow */
- return FALSE;
+ /* Here, all the bytes we have are the same as the highest 30-bit code
+ * point, but we are missing so many bytes that we can't make the
+ * determination */
+ return -1;
+ }
}
-PERL_STATIC_INLINE bool
+#endif
+
+PERL_STATIC_INLINE int
S_is_utf8_overlong_given_start_byte_ok(const U8 * const s, const STRLEN len)
{
- /* Overlongs can occur whenever the number of continuation bytes
- * changes. That means whenever the number of leading 1 bits in a start
- * byte increases from the next lower start byte. That happens for start
- * bytes C0, E0, F0, F8, FC, FE, and FF. On modern perls, the following
- * illegal start bytes have already been excluded, so don't need to be
- * tested here;
+ /* Returns an int indicating whether or not the UTF-8 sequence from 's' to
+ * 's' + 'len' - 1 is an overlong. It returns 1 if it is an overlong; 0 if
+ * it isn't, and -1 if there isn't enough information to tell. This last
+ * return value can happen if the sequence is incomplete, missing some
+ * trailing bytes that would form a complete character. If there are
+ * enough bytes to make a definitive decision, this function does so.
+ * Usually 2 bytes sufficient.
+ *
+ * Overlongs can occur whenever the number of continuation bytes changes.
+ * That means whenever the number of leading 1 bits in a start byte
+ * increases from the next lower start byte. That happens for start bytes
+ * C0, E0, F0, F8, FC, FE, and FF. On modern perls, the following illegal
+ * start bytes have already been excluded, so don't need to be tested here;
* ASCII platforms: C0, C1
* EBCDIC platforms C0, C1, C2, C3, C4, E0
- *
- * At least a second byte is required to determine if other sequences will
- * be an overlong. */
+ */
const U8 s0 = NATIVE_UTF8_TO_I8(s[0]);
const U8 s1 = NATIVE_UTF8_TO_I8(s[1]);
# else
if (s0 == 0xE0 && UNLIKELY(s1 < 0xA0)) {
- return TRUE;
+ return 1;
}
# define F0_ABOVE_OVERLONG 0x90
|| (s0 == 0xFC && UNLIKELY(s1 < FC_ABOVE_OVERLONG))
|| (s0 == 0xFE && UNLIKELY(s1 < FE_ABOVE_OVERLONG)))
{
- return TRUE;
+ return 1;
}
/* Check for the FF overlong */
return isFF_OVERLONG(s, len);
}
-PERL_STATIC_INLINE bool
+PERL_STATIC_INLINE int
S_isFF_OVERLONG(const U8 * const s, const STRLEN len)
{
+ /* Returns an int indicating whether or not the UTF-8 sequence from 's' to
+ * 'e' - 1 is an overlong beginning with \xFF. It returns 1 if it is; 0 if
+ * it isn't, and -1 if there isn't enough information to tell. This last
+ * return value can happen if the sequence is incomplete, missing some
+ * trailing bytes that would form a complete character. If there are
+ * enough bytes to make a definitive decision, this function does so. */
+
PERL_ARGS_ASSERT_ISFF_OVERLONG;
- /* Check for the FF overlong. This happens only if all these bytes match;
- * what comes after them doesn't matter. See tables in utf8.h,
+ /* To be an FF overlong, all the available bytes must match */
+ if (LIKELY(memNE(s, FF_OVERLONG_PREFIX,
+ MIN(len, sizeof(FF_OVERLONG_PREFIX) - 1))))
+ {
+ return 0;
+ }
+
+ /* To be an FF overlong sequence, all the bytes in FF_OVERLONG_PREFIX must
+ * be there; what comes after them doesn't matter. See tables in utf8.h,
* utfebcdic.h. */
+ if (len >= sizeof(FF_OVERLONG_PREFIX) - 1) {
+ return 1;
+ }
+
+ /* The missing bytes could cause the result to go one way or the other, so
+ * the result is indeterminate */
+ return -1;
+}
+
+#if defined(UV_IS_QUAD) /* These assume IV_MAX is 2**63-1 */
+# ifdef EBCDIC /* Actually is I8 */
+# define HIGHEST_REPRESENTABLE_UTF8 \
+ "\xFF\xA7\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF"
+# else
+# define HIGHEST_REPRESENTABLE_UTF8 \
+ "\xFF\x80\x87\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF"
+# endif
+#endif
+
+PERL_STATIC_INLINE int
+S_does_utf8_overflow(const U8 * const s,
+ const U8 * e,
+ const bool consider_overlongs)
+{
+ /* Returns an int indicating whether or not the UTF-8 sequence from 's' to
+ * 'e' - 1 would overflow an IV on this platform; that is if it represents
+ * a code point larger than the highest representable code point. It
+ * returns 1 if it does overflow; 0 if it doesn't, and -1 if there isn't
+ * enough information to tell. This last return value can happen if the
+ * sequence is incomplete, missing some trailing bytes that would form a
+ * complete character. If there are enough bytes to make a definitive
+ * decision, this function does so.
+ *
+ * If 'consider_overlongs' is TRUE, the function checks for the possibility
+ * that the sequence is an overlong that doesn't overflow. Otherwise, it
+ * assumes the sequence is not an overlong. This can give different
+ * results only on ASCII 32-bit platforms.
+ *
+ * (For ASCII platforms, we could use memcmp() because we don't have to
+ * convert each byte to I8, but it's very rare input indeed that would
+ * approach overflow, so the loop below will likely only get executed once.)
+ *
+ * 'e' - 1 must not be beyond a full character. */
+
+
+ PERL_ARGS_ASSERT_DOES_UTF8_OVERFLOW;
+ assert(s <= e && s + UTF8SKIP(s) >= e);
+
+#if ! defined(UV_IS_QUAD)
+
+ return is_utf8_cp_above_31_bits(s, e, consider_overlongs);
+
+#else
+
+ PERL_UNUSED_ARG(consider_overlongs);
+
+ {
+ const STRLEN len = e - s;
+ const U8 *x;
+ const U8 * y = (const U8 *) HIGHEST_REPRESENTABLE_UTF8;
+
+ for (x = s; x < e; x++, y++) {
+
+ if (UNLIKELY(NATIVE_UTF8_TO_I8(*x) == *y)) {
+ continue;
+ }
+
+ /* If this byte is larger than the corresponding highest UTF-8
+ * byte, the sequence overflow; otherwise the byte is less than,
+ * and so the sequence doesn't overflow */
+ return NATIVE_UTF8_TO_I8(*x) > *y;
+
+ }
+
+ /* Got to the end and all bytes are the same. If the input is a whole
+ * character, it doesn't overflow. And if it is a partial character,
+ * there's not enough information to tell */
+ if (len < sizeof(HIGHEST_REPRESENTABLE_UTF8) - 1) {
+ return -1;
+ }
+
+ return 0;
+ }
+
+#endif
- return len >= sizeof(FF_OVERLONG_PREFIX) - 1
- && UNLIKELY(memEQ(s, FF_OVERLONG_PREFIX,
- sizeof(FF_OVERLONG_PREFIX) - 1));
}
+#if 0
+
+/* This is the portions of the above function that deal with UV_MAX instead of
+ * IV_MAX. They are left here in case we want to combine them so that internal
+ * uses can have larger code points. The only logic difference is that the
+ * 32-bit EBCDIC platform is treate like the 64-bit, and the 32-bit ASCII has
+ * different logic.
+ */
+
+/* Anything larger than this will overflow the word if it were converted into a UV */
+#if defined(UV_IS_QUAD)
+# ifdef EBCDIC /* Actually is I8 */
+# define HIGHEST_REPRESENTABLE_UTF8 \
+ "\xFF\xAF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF"
+# else
+# define HIGHEST_REPRESENTABLE_UTF8 \
+ "\xFF\x80\x8F\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF"
+# endif
+#else /* 32-bit */
+# ifdef EBCDIC
+# define HIGHEST_REPRESENTABLE_UTF8 \
+ "\xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA3\xBF\xBF\xBF\xBF\xBF\xBF"
+# else
+# define HIGHEST_REPRESENTABLE_UTF8 "\xFE\x83\xBF\xBF\xBF\xBF\xBF"
+# endif
+#endif
+
+#if ! defined(UV_IS_QUAD) && ! defined(EBCDIC)
+
+ /* On 32 bit ASCII machines, many overlongs that start with FF don't
+ * overflow */
+ if (consider_overlongs && isFF_OVERLONG(s, len) > 0) {
+
+ /* To be such an overlong, the first bytes of 's' must match
+ * FF_OVERLONG_PREFIX, which is "\xff\x80\x80\x80\x80\x80\x80". If we
+ * don't have any additional bytes available, the sequence, when
+ * completed might or might not fit in 32 bits. But if we have that
+ * next byte, we can tell for sure. If it is <= 0x83, then it does
+ * fit. */
+ if (len <= sizeof(FF_OVERLONG_PREFIX) - 1) {
+ return -1;
+ }
+
+ return s[sizeof(FF_OVERLONG_PREFIX) - 1] > 0x83;
+ }
+
+/* Starting with the #else, the rest of the function is identical except
+ * 1. we need to move the 'len' declaration to be global to the function
+ * 2. the endif move to just after the UNUSED_ARG.
+ * An empty endif is given just below to satisfy the preprocessor
+ */
+#endif
+
+#endif
+
#undef F0_ABOVE_OVERLONG
#undef F8_ABOVE_OVERLONG
#undef FC_ABOVE_OVERLONG
# define IS_UTF8_2_BYTE_SURROGATE(s0, s1) ((s0) == 0xF1 \
/* B6 and B7 */ \
&& ((s1) & 0xFE ) == 0xB6)
+# define isUTF8_PERL_EXTENDED(s) (*s == I8_TO_NATIVE_UTF8(0xFF))
#else
# define FIRST_START_BYTE_THAT_IS_DEFINITELY_SUPER 0xF5
# define IS_UTF8_2_BYTE_SUPER(s0, s1) ((s0) == 0xF4 && (s1) >= 0x90)
# define IS_UTF8_2_BYTE_SURROGATE(s0, s1) ((s0) == 0xED && (s1) >= 0xA0)
+# define isUTF8_PERL_EXTENDED(s) (*s >= 0xFE)
#endif
if ( (flags & UTF8_DISALLOW_SUPER)
}
if ( (flags & UTF8_DISALLOW_PERL_EXTENDED)
- && UNLIKELY(is_utf8_cp_above_31_bits(s, e)))
+ && UNLIKELY(isUTF8_PERL_EXTENDED(s)))
{
- return 0; /* Above 31 bits */
+ return 0;
}
if (len > 1) {
/* Here is syntactically valid. Next, make sure this isn't the start of an
* overlong. */
- if (len > 1 && is_utf8_overlong_given_start_byte_ok(s, len)) {
+ if (len > 1 && is_utf8_overlong_given_start_byte_ok(s, len) > 0) {
return 0;
}
/* And finally, that the code point represented fits in a word on this
* platform */
- if (does_utf8_overflow(s, e)) {
+ if (0 < does_utf8_overflow(s, e,
+ 0 /* Don't consider overlongs */
+ ))
+ {
return 0;
}
}
char *
-Perl__byte_dump_string(pTHX_ const U8 * s, const STRLEN len, const bool format)
+Perl__byte_dump_string(pTHX_ const U8 * const start, const STRLEN len, const bool format)
{
/* Returns a mortalized C string that is a displayable copy of the 'len'
- * bytes starting at 's'. 'format' gives how to display each byte.
+ * bytes starting at 'start'. 'format' gives how to display each byte.
* Currently, there are only two formats, so it is currently a bool:
* 0 \xab
* 1 ab (that is a space between two hex digit bytes)
const STRLEN output_len = 4 * len + 1; /* 4 bytes per each input, plus a
trailing NUL */
- const U8 * const e = s + len;
+ const U8 * s = start;
+ const U8 * const e = start + len;
char * output;
char * d;
SAVEFREEPV(output);
d = output;
- for (; s < e; s++) {
+ for (s = start; s < e; s++) {
const unsigned high_nibble = (*s & 0xF0) >> 4;
const unsigned low_nibble = (*s & 0x0F);
if (format) {
- *d++ = ' ';
+ if (s > start) {
+ *d++ = ' ';
+ }
}
else {
*d++ = '\\';
(But note that warnings are not raised if lexically disabled nor if
C<UTF8_CHECK_ONLY> is also specified.)
-It is now deprecated to have very high code points (above C<IV_MAX> on the
-platforms) and this function will raise a deprecation warning for these (unless
-such warnings are turned off). This value is typically 0x7FFF_FFFF (2**31 -1)
-in a 32-bit word.
+Extremely high code points were never specified in any standard, and require an
+extension to UTF-8 to express, which Perl does. It is likely that programs
+written in something other than Perl would not be able to read files that
+contain these; nor would Perl understand files written by something that uses a
+different extension. For these reasons, there is a separate set of flags that
+can warn and/or disallow these extremely high code points, even if other
+above-Unicode ones are accepted. They are the C<UTF8_WARN_PERL_EXTENDED> and
+C<UTF8_DISALLOW_PERL_EXTENDED> flags. For more information see
+L</C<UTF8_GOT_PERL_EXTENDED>>. Of course C<UTF8_DISALLOW_SUPER> will treat all
+above-Unicode code points, including these, as malformations.
+(Note that the Unicode standard considers anything above 0x10FFFF to be
+illegal, but there are standards predating it that allow up to 0x7FFF_FFFF
+(2**31 -1))
+
+A somewhat misleadingly named synonym for C<UTF8_WARN_PERL_EXTENDED> is
+retained for backward compatibility: C<UTF8_WARN_ABOVE_31_BIT>. Similarly,
+C<UTF8_DISALLOW_ABOVE_31_BIT> is usable instead of the more accurately named
+C<UTF8_DISALLOW_PERL_EXTENDED>. The names are misleading because these flags
+can apply to code points that actually do fit in 31 bits. This happens on
+EBCDIC platforms, and sometimes when the L<overlong
+malformation|/C<UTF8_GOT_LONG>> is also present. The new names accurately
+describe the situation in all cases.
-Code points above 0x7FFF_FFFF (2**31 - 1) were never specified in any standard,
-so using them is more problematic than other above-Unicode code points. Perl
-invented an extension to UTF-8 to represent the ones above 2**36-1, so it is
-likely that non-Perl languages will not be able to read files that contain
-these; nor would Perl understand files
-written by something that uses a different extension. For these reasons, there
-is a separate set of flags that can warn and/or disallow these extremely high
-code points, even if other above-Unicode ones are accepted. These are the
-C<UTF8_WARN_ABOVE_31_BIT> and C<UTF8_DISALLOW_ABOVE_31_BIT> flags. These
-are entirely independent from the deprecation warning for code points above
-C<IV_MAX>. On 32-bit machines, it will eventually be forbidden to have any
-code point that needs more than 31 bits to represent. When that happens,
-effectively the C<UTF8_DISALLOW_ABOVE_31_BIT> flag will always be set on
-32-bit machines. (Of course C<UTF8_DISALLOW_SUPER> will treat all
-above-Unicode code points, including these, as malformations; and
-C<UTF8_WARN_SUPER> warns on these.)
-
-On EBCDIC platforms starting in Perl v5.24, the Perl extension for representing
-extremely high code points kicks in at 0x3FFF_FFFF (2**30 -1), which is lower
-than on ASCII. Prior to that, code points 2**31 and higher were simply
-unrepresentable, and a different, incompatible method was used to represent
-code points between 2**30 and 2**31 - 1. The flags C<UTF8_WARN_ABOVE_31_BIT>
-and C<UTF8_DISALLOW_ABOVE_31_BIT> have the same function as on ASCII
-platforms, warning and disallowing 2**31 and higher.
All other code points corresponding to Unicode characters, including private
use and those yet to be assigned, are never considered malformed and never
=over 4
-=item C<UTF8_GOT_ABOVE_31_BIT>
+=item C<UTF8_GOT_PERL_EXTENDED>
-The code point represented by the input UTF-8 sequence occupies more than 31
-bits.
-This bit is set only if the input C<flags> parameter contains either the
-C<UTF8_DISALLOW_ABOVE_31_BIT> or the C<UTF8_WARN_ABOVE_31_BIT> flags.
+The input sequence is not standard UTF-8, but a Perl extension. This bit is
+set only if the input C<flags> parameter contains either the
+C<UTF8_DISALLOW_PERL_EXTENDED> or the C<UTF8_WARN_PERL_EXTENDED> flags.
+
+Code points above 0x7FFF_FFFF (2**31 - 1) were never specified in any standard,
+and so some extension must be used to express them. Perl uses a natural
+extension to UTF-8 to represent the ones up to 2**36-1, and invented a further
+extension to represent even higher ones, so that any code point that fits in a
+64-bit word can be represented. Text using these extensions is not likely to
+be portable to non-Perl code. We lump both of these extensions together and
+refer to them as Perl extended UTF-8. There exist other extensions that people
+have invented, incompatible with Perl's.
+
+On EBCDIC platforms starting in Perl v5.24, the Perl extension for representing
+extremely high code points kicks in at 0x3FFF_FFFF (2**30 -1), which is lower
+than on ASCII. Prior to that, code points 2**31 and higher were simply
+unrepresentable, and a different, incompatible method was used to represent
+code points between 2**30 and 2**31 - 1.
+
+On both platforms, ASCII and EBCDIC, C<UTF8_GOT_PERL_EXTENDED> is set if
+Perl extended UTF-8 is used.
+
+In earlier Perls, this bit was named C<UTF8_GOT_ABOVE_31_BIT>, which you still
+may use for backward compatibility. That name is misleading, as this flag may
+be set when the code point actually does fit in 31 bits. This happens on
+EBCDIC platforms, and sometimes when the L<overlong
+malformation|/C<UTF8_GOT_LONG>> is also present. The new name accurately
+describes the situation in all cases.
=item C<UTF8_GOT_CONTINUATION>
=item C<UTF8_GOT_OVERFLOW>
The input sequence was malformed in that it is for a code point that is not
-representable in the number of bits available in a UV on the current platform.
+representable in the number of bits available in an IV on the current platform.
=item C<UTF8_GOT_SHORT>
* too short one. Otherwise the first two are set to 's0' and 'send', and
* the third not used at all */
U8 * adjusted_s0 = (U8 *) s0;
- U8 * adjusted_send = NULL; /* (Initialized to silence compilers' wrong
- warning) */
U8 temp_char_buf[UTF8_MAXBYTES + 1]; /* Used to avoid a Newx in this
routine; see [perl #130921] */
UV uv_so_far = 0; /* (Initialized to silence compilers' wrong warning) */
else {
send += expectlen;
}
- adjusted_send = send;
/* Now, loop through the remaining bytes in the character's sequence,
* accumulating each into the working value as we go. */
uv = UNICODE_REPLACEMENT;
}
- /* Check for overflow. */
- if (UNLIKELY(does_utf8_overflow(s0, send))) {
+ /* Check for overflow. The algorithm requires us to not look past the end
+ * of the current character, even if partial, so the upper limit is 's' */
+ if (UNLIKELY(0 < does_utf8_overflow(s0, s,
+ 1 /* Do consider overlongs */
+ )))
+ {
possible_problems |= UTF8_GOT_OVERFLOW;
uv = UNICODE_REPLACEMENT;
}
|| ( UNLIKELY(possible_problems)
&& ( UNLIKELY(! UTF8_IS_START(*s0))
|| ( curlen > 1
- && UNLIKELY(is_utf8_overlong_given_start_byte_ok(s0,
- send - s0))))))
+ && UNLIKELY(0 < is_utf8_overlong_given_start_byte_ok(s0,
+ s - s0))))))
{
possible_problems |= UTF8_GOT_LONG;
}
adjusted_s0 = temp_char_buf;
- adjusted_send = uvoffuni_to_utf8_flags(adjusted_s0, min_uv, 0);
+ (void) uvoffuni_to_utf8_flags(adjusted_s0, min_uv, 0);
}
}
* and we deal with those in the overflow handling
* code */
&& LIKELY(! (possible_problems & UTF8_GOT_OVERFLOW))
- && isUTF8_POSSIBLY_PROBLEMATIC(*adjusted_s0)))
+ && ( isUTF8_POSSIBLY_PROBLEMATIC(*adjusted_s0)
+ || UNLIKELY(isUTF8_PERL_EXTENDED(s0)))))
&& ((flags & ( UTF8_DISALLOW_NONCHAR
|UTF8_DISALLOW_SURROGATE
|UTF8_DISALLOW_SUPER
|UTF8_WARN_NONCHAR
|UTF8_WARN_SURROGATE
|UTF8_WARN_SUPER
- |UTF8_WARN_PERL_EXTENDED))
- /* In case of a malformation, 'uv' is not valid, and has
- * been changed to something in the Unicode range.
- * Currently we don't output a deprecation message if there
- * is already a malformation, so we don't have to special
- * case the test immediately below */
- || ( UNLIKELY(uv > MAX_NON_DEPRECATED_CP)
- && ckWARN_d(WARN_DEPRECATED))))
+ |UTF8_WARN_PERL_EXTENDED))))
{
/* If there were no malformations, or the only malformation is an
* overlong, 'uv' is valid */
* end, based on how many bytes the start byte tells
* us should be in it, but no further than s0 +
* avail_len
- * adjusted_s0 normally is the same as s0, but in case of an
- * overlong for which the UTF-8 matters below, it is
- * the first byte of the shortest form representation
- * of the input.
- * adjusted_send normally is the same as 'send', but if adjusted_s0
- * is set to something other than s0, this points one
- * beyond its end
*/
if (UNLIKELY(possible_problems)) {
disallowed = TRUE;
}
- /* Likewise, warn if any say to, plus if deprecation warnings
- * are on, because this code point is above IV_MAX */
- if ( ckWARN_d(WARN_DEPRECATED)
- || ! (flags & UTF8_ALLOW_OVERFLOW)
- || (flags & (UTF8_WARN_SUPER|UTF8_WARN_PERL_EXTENDED)))
+ /* Likewise, warn if any say to */
+ if ( ! (flags & UTF8_ALLOW_OVERFLOW)
+ || (flags & (UTF8_WARN_SUPER|UTF8_WARN_PERL_EXTENDED)))
{
/* The warnings code explicitly says it doesn't handle the
}
}
- /* The maximum code point ever specified by a standard was
- * 2**31 - 1. Anything larger than that is a Perl extension
- * that very well may not be understood by other applications
- * (including earlier perl versions on EBCDIC platforms). We
- * test for these after the regular SUPER ones, and before
- * possibly bailing out, so that the slightly more dire warning
- * will override the regular one. */
- if ( (flags & (UTF8_WARN_PERL_EXTENDED
- |UTF8_WARN_SUPER
- |UTF8_DISALLOW_PERL_EXTENDED))
- && ( ( UNLIKELY(orig_problems & UTF8_GOT_TOO_SHORT)
- && UNLIKELY(is_utf8_cp_above_31_bits(
- adjusted_s0,
- adjusted_send)))
- || ( LIKELY(! (orig_problems & UTF8_GOT_TOO_SHORT))
- && UNLIKELY(UNICODE_IS_PERL_EXTENDED(uv)))))
- {
+ /* Test for Perl's extended UTF-8 after the regular SUPER ones,
+ * and before possibly bailing out, so that the more dire
+ * warning will override the regular one. */
+ if (UNLIKELY(isUTF8_PERL_EXTENDED(s0))) {
if ( ! (flags & UTF8_CHECK_ONLY)
&& (flags & (UTF8_WARN_PERL_EXTENDED|UTF8_WARN_SUPER))
&& ckWARN_d(WARN_NON_UNICODE))
{
pack_warn = packWARN(WARN_NON_UNICODE);
- if (orig_problems & UTF8_GOT_TOO_SHORT) {
+ /* If it is an overlong that evaluates to a code point
+ * that doesn't have to use the Perl extended UTF-8, it
+ * still used it, and so we output a message that
+ * doesn't refer to the code point. The same is true
+ * if there was a SHORT malformation where the code
+ * point is not valid. In that case, 'uv' will have
+ * been set to the REPLACEMENT CHAR, and the message
+ * below without the code point in it will be selected
+ * */
+ if (UNICODE_IS_PERL_EXTENDED(uv)) {
message = Perl_form(aTHX_
- "Any UTF-8 sequence that starts with"
- " \"%s\" is for a non-Unicode code"
- " point, and is not portable",
- _byte_dump_string(s0, curlen, 0));
+ perl_extended_cp_format, uv);
}
else {
message = Perl_form(aTHX_
- above_31_bit_cp_format, uv);
+ "Any UTF-8 sequence that starts with"
+ " \"%s\" is a Perl extension, and"
+ " so is not portable",
+ _byte_dump_string(s0, curlen, 0));
}
}
*errors |= UTF8_GOT_SUPER;
disallowed = TRUE;
}
-
- /* The deprecated warning overrides any non-deprecated one. If
- * there are other problems, a deprecation message is not
- * really helpful, so don't bother to raise it in that case.
- * This also keeps the code from having to handle the case
- * where 'uv' is not valid. */
- if ( ! (orig_problems
- & (UTF8_GOT_TOO_SHORT|UTF8_GOT_OVERFLOW))
- && UNLIKELY(uv > MAX_NON_DEPRECATED_CP)
- && ckWARN_d(WARN_DEPRECATED))
- {
- message = Perl_form(aTHX_ cp_above_legal_max,
- uv, MAX_NON_DEPRECATED_CP);
- pack_warn = packWARN(WARN_DEPRECATED);
- }
}
else if (possible_problems & UTF8_GOT_NONCHAR) {
possible_problems &= ~UTF8_GOT_NONCHAR;
U8 tmpbuf[UTF8_MAXBYTES+1];
const U8 * const e = uvoffuni_to_utf8_flags(tmpbuf,
uv, 0);
- const char * preface = (uv <= PERL_UNICODE_MAX)
- ? "U+"
- : "0x";
+ /* Don't use U+ for non-Unicode code points, which
+ * includes those in the Latin1 range */
+ const char * preface = ( uv > PERL_UNICODE_MAX
+#ifdef EBCDIC
+ || uv <= 0xFF
+#endif
+ )
+ ? "0x"
+ : "U+";
message = Perl_form(aTHX_
"%s: %s (overlong; instead use %s to represent"
" %s%0*" UVXf ")",
preface,
((uv < 256) ? 2 : 4), /* Field width of 2 for
small code points */
- uv);
+ UNI_TO_NATIVE(uv));
}
}
}
See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is
returned.
-Code points above the platform's C<IV_MAX> will raise a deprecation warning,
-unless those are turned off.
-
=cut
Also implemented as a macro in utf8.h
next possible position in C<s> that could begin a non-malformed character.
See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned.
-Code points above the platform's C<IV_MAX> will raise a deprecation warning,
-unless those are turned off.
-
=cut
*/
assert(send > s);
- /* Call the low level routine, asking for checks */
return NATIVE_TO_UNI(utf8_to_uvchr_buf(s, send, retlen));
}
/*
=for apidoc utf8_length
-Return the length of the UTF-8 char encoded string C<s> in characters.
-Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end
-up past C<e>, croaks.
+Returns the number of characters in the sequence of UTF-8-encoded bytes starting
+at C<s> and ending at the byte just before C<e>. If <s> and <e> point to the
+same place, it returns 0 with no warning raised.
+
+If C<e E<lt> s> or if the scan would end up past C<e>, it raises a UTF8 warning
+and returns the number of valid characters.
=cut
*/
}
finish_and_return:
- *d = '\0';
- *lenp = d - converted_start;
+ *d = '\0';
+ *lenp = d - converted_start;
/* Trim unused space */
Renew(converted_start, *lenp + 1, U8);
append_utf8_from_native_byte(*s, &d);
s++;
}
+
*d = '\0';
*lenp = d-dst;
+
+ /* Trim unused space */
+ Renew(dst, *lenp + 1, U8);
+
return dst;
}
/*
- * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8.
+ * Convert native (big-endian) UTF-16 to UTF-8. For reversed (little-endian),
+ * use utf16_to_utf8_reversed().
+ *
+ * UTF-16 requires 2 bytes for every code point below 0x10000; otherwise 4 bytes.
+ * UTF-8 requires 1-3 bytes for every code point below 0x1000; otherwise 4 bytes.
+ * UTF-EBCDIC requires 1-4 bytes for every code point below 0x1000; otherwise 4-5 bytes.
*
- * Destination must be pre-extended to 3/2 source. Do not use in-place.
- * We optimize for native, for obvious reasons. */
+ * These functions don't check for overflow. The worst case is every code
+ * point in the input is 2 bytes, and requires 4 bytes on output. (If the code
+ * is never going to run in EBCDIC, it is 2 bytes requiring 3 on output.) Therefore the
+ * destination must be pre-extended to 2 times the source length.
+ *
+ * Do not use in-place. We optimize for native, for obvious reasons. */
U8*
Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen)
*d++ = UTF8_TWO_BYTE_LO(UNI_TO_NATIVE(uv));
continue;
}
+
#define FIRST_HIGH_SURROGATE UNICODE_SURROGATE_FIRST
#define LAST_HIGH_SURROGATE 0xDBFF
#define FIRST_LOW_SURROGATE 0xDC00
#define LAST_LOW_SURROGATE UNICODE_SURROGATE_LAST
+#define FIRST_IN_PLANE1 0x10000
/* This assumes that most uses will be in the first Unicode plane, not
* needing surrogates */
}
p += 2;
uv = ((uv - FIRST_HIGH_SURROGATE) << 10)
- + (low - FIRST_LOW_SURROGATE) + 0x10000;
+ + (low - FIRST_LOW_SURROGATE) + FIRST_IN_PLANE1;
}
}
#ifdef EBCDIC
d = uvoffuni_to_utf8_flags(d, uv, 0);
#else
- if (uv < 0x10000) {
+ if (uv < FIRST_IN_PLANE1) {
*d++ = (U8)(( uv >> 12) | 0xe0);
*d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
*d++ = (U8)(( uv & 0x3f) | 0x80);
}
if (UNLIKELY(UNICODE_IS_SUPER(uv1))) {
- if ( UNLIKELY(uv1 > MAX_NON_DEPRECATED_CP)
- && ckWARN_d(WARN_DEPRECATED))
- {
- Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
- cp_above_legal_max, uv1, MAX_NON_DEPRECATED_CP);
+ if (UNLIKELY(uv1 > MAX_EXTERNALLY_LEGAL_CP)) {
+ Perl_croak(aTHX_ cp_above_legal_max, uv1,
+ MAX_EXTERNALLY_LEGAL_CP);
}
if (ckWARN_d(WARN_NON_UNICODE)) {
const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
if (flags & FOLD_FLAGS_LOCALE) {
# define LONG_S_T LATIN_SMALL_LIGATURE_LONG_S_T_UTF8
- const unsigned int long_s_t_len = sizeof(LONG_S_T) - 1;
-
# ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
# define CAP_SHARP_S LATIN_CAPITAL_LETTER_SHARP_S_UTF8
- const unsigned int cap_sharp_s_len = sizeof(CAP_SHARP_S) - 1;
-
/* Special case these two characters, as what normally gets
* returned under locale doesn't work */
- if (UTF8SKIP(p) == cap_sharp_s_len
- && memEQ((char *) p, CAP_SHARP_S, cap_sharp_s_len))
+ if (memEQs((char *) p, UTF8SKIP(p), CAP_SHARP_S))
{
/* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */
Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
}
else
#endif
- if (UTF8SKIP(p) == long_s_t_len
- && memEQ((char *) p, LONG_S_T, long_s_t_len))
+ if (memEQs((char *) p, UTF8SKIP(p), LONG_S_T))
{
/* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */
Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
* 255/256 boundary which is forbidden under /l, and so the code
* wouldn't catch that they are equivalent (which they are only in
* this release) */
- else if (UTF8SKIP(p) == sizeof(DOTTED_I) - 1
- && memEQ((char *) p, DOTTED_I, sizeof(DOTTED_I) - 1))
- {
+ else if (memEQs((char *) p, UTF8SKIP(p), DOTTED_I)) {
/* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */
Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
"Can't do fc(\"\\x{0130}\") on non-UTF-8 locale; "
* size based on worst possible case, which is each line in the input
* creates 2 elements in the inversion list: 1) the beginning of a
* range in the list; 2) the beginning of a range not in the list. */
- while ((loc = (strchr(loc, '\n'))) != NULL) {
+ while ((loc = (char *) memchr(loc, '\n', lend - (U8 *) loc)) != NULL) {
elements += 2;
loc++;
}
/* May change: warns if surrogates, non-character code points, or
* non-Unicode code points are in 's' which has length 'len' bytes.
* Returns TRUE if none found; FALSE otherwise. The only other validity
- * check is to make sure that this won't exceed the string's length.
- *
- * Code points above the platform's C<IV_MAX> will raise a deprecation
- * warning, unless those are turned off. */
+ * check is to make sure that this won't exceed the string's length nor
+ * overflow */
const U8* const e = s + len;
bool ok = TRUE;
if (UNLIKELY(isUTF8_POSSIBLY_PROBLEMATIC(*s))) {
if (UNLIKELY(UTF8_IS_SUPER(s, e))) {
if ( ckWARN_d(WARN_NON_UNICODE)
- || ( ckWARN_d(WARN_DEPRECATED)
-#ifndef UV_IS_QUAD
- && UNLIKELY(is_utf8_cp_above_31_bits(s, e))
-#else /* Below is 64-bit words */
- /* 2**63 and up meet these conditions provided we have
- * a 64-bit word. */
-# ifdef EBCDIC
- && *s == 0xFE
- && NATIVE_UTF8_TO_I8(s[1]) >= 0xA8
-# else
- && *s == 0xFF
- /* s[1] being above 0x80 overflows */
- && s[2] >= 0x88
-# endif
-#endif
- )) {
+ || UNLIKELY(0 < does_utf8_overflow(s, s + len,
+ 0 /* Don't consider overlongs */
+ )))
+ {
/* A side effect of this function will be to warn */
(void) utf8n_to_uvchr(s, e - s, NULL, UTF8_WARN_SUPER);
ok = FALSE;
https://perl5.git.perl.org / perl5.git / blobdiff