static const char malformed_text[] = "Malformed UTF-8 character";
static const char unees[] =
"Malformed UTF-8 character (unexpected end of string)";
+
+/* Be sure to synchronize this message with the similar one in regcomp.c */
static const char cp_above_legal_max[] =
- "Use of code point 0x%" UVXf " is deprecated; the permissible max is 0x%" UVXf;
+ "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
PERL_ARGS_ASSERT__FORCE_OUT_MALFORMED_UTF8_MESSAGE;
ENTER;
- SAVESPTR(PL_dowarn);
+ SAVEI8(PL_dowarn);
SAVESPTR(PL_curcop);
PL_dowarn = G_WARN_ALL_ON|G_WARN_ON;
}
}
+STATIC HV *
+S_new_msg_hv(pTHX_ const char * const message, /* The message text */
+ U32 categories, /* Packed warning categories */
+ U32 flag) /* Flag associated with this message */
+{
+ /* Creates, populates, and returns an HV* that describes an error message
+ * for the translators between UTF8 and code point */
+
+ SV* msg_sv = newSVpv(message, 0);
+ SV* category_sv = newSVuv(categories);
+ SV* flag_bit_sv = newSVuv(flag);
+
+ HV* msg_hv = newHV();
+
+ PERL_ARGS_ASSERT_NEW_MSG_HV;
+
+ (void) hv_stores(msg_hv, "text", msg_sv);
+ (void) hv_stores(msg_hv, "warn_categories", category_sv);
+ (void) hv_stores(msg_hv, "flag_bit", flag_bit_sv);
+
+ return msg_hv;
+}
+
/*
=for apidoc uvoffuni_to_utf8_flags
=cut
*/
-#define HANDLE_UNICODE_SURROGATE(uv, flags) \
+U8 *
+Perl_uvoffuni_to_utf8_flags(pTHX_ U8 *d, UV uv, const UV flags)
+{
+ PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS;
+
+ return uvoffuni_to_utf8_flags_msgs(d, uv, flags, NULL);
+}
+
+/* All these formats take a single UV code point argument */
+const char surrogate_cp_format[] = "UTF-16 surrogate U+%04" UVXf;
+const char nonchar_cp_format[] = "Unicode non-character U+%04" UVXf
+ " is not recommended for open interchange";
+const char super_cp_format[] = "Code point 0x%" UVXf " is not Unicode,"
+ " may not be 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, msgs) \
STMT_START { \
if (flags & UNICODE_WARN_SURROGATE) { \
- Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE), \
- "UTF-16 surrogate U+%04" UVXf, uv); \
+ U32 category = packWARN(WARN_SURROGATE); \
+ const char * format = surrogate_cp_format; \
+ if (msgs) { \
+ *msgs = new_msg_hv(Perl_form(aTHX_ format, uv), \
+ category, \
+ UNICODE_GOT_SURROGATE); \
+ } \
+ else { \
+ Perl_ck_warner_d(aTHX_ category, format, uv); \
+ } \
} \
if (flags & UNICODE_DISALLOW_SURROGATE) { \
return NULL; \
} \
} STMT_END;
-#define HANDLE_UNICODE_NONCHAR(uv, flags) \
+#define HANDLE_UNICODE_NONCHAR(uv, flags, msgs) \
STMT_START { \
if (flags & UNICODE_WARN_NONCHAR) { \
- Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR), \
- "Unicode non-character U+%04" UVXf " is not " \
- "recommended for open interchange", uv); \
+ U32 category = packWARN(WARN_NONCHAR); \
+ const char * format = nonchar_cp_format; \
+ if (msgs) { \
+ *msgs = new_msg_hv(Perl_form(aTHX_ format, uv), \
+ category, \
+ UNICODE_GOT_NONCHAR); \
+ } \
+ else { \
+ Perl_ck_warner_d(aTHX_ category, format, uv); \
+ } \
} \
if (flags & UNICODE_DISALLOW_NONCHAR) { \
return NULL; \
#define MARK UTF_CONTINUATION_MARK
#define MASK UTF_CONTINUATION_MASK
+/*
+=for apidoc uvchr_to_utf8_flags_msgs
+
+THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES.
+
+Most code should use C<L</uvchr_to_utf8_flags>()> rather than call this directly.
+
+This function is for code that wants any warning and/or error messages to be
+returned to the caller rather than be displayed. All messages that would have
+been displayed if all lexcial warnings are enabled will be returned.
+
+It is just like C<L</uvchr_to_utf8_flags>> but it takes an extra parameter
+placed after all the others, C<msgs>. If this parameter is 0, this function
+behaves identically to C<L</uvchr_to_utf8_flags>>. Otherwise, C<msgs> should
+be a pointer to an C<HV *> variable, in which this function creates a new HV to
+contain any appropriate messages. The hash has three key-value pairs, as
+follows:
+
+=over 4
+
+=item C<text>
+
+The text of the message as a C<SVpv>.
+
+=item C<warn_categories>
+
+The warning category (or categories) packed into a C<SVuv>.
+
+=item C<flag>
+
+A single flag bit associated with this message, in a C<SVuv>.
+The bit corresponds to some bit in the C<*errors> return value,
+such as C<UNICODE_GOT_SURROGATE>.
+
+=back
+
+It's important to note that specifying this parameter as non-null will cause
+any warnings this function would otherwise generate to be suppressed, and
+instead be placed in C<*msgs>. The caller can check the lexical warnings state
+(or not) when choosing what to do with the returned messages.
+
+The caller, of course, is responsible for freeing any returned HV.
+
+=cut
+*/
+
+/* Undocumented; we don't want people using this. Instead they should use
+ * uvchr_to_utf8_flags_msgs() */
U8 *
-Perl_uvoffuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
+Perl_uvoffuni_to_utf8_flags_msgs(pTHX_ U8 *d, UV uv, const UV flags, HV** msgs)
{
- PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS;
+ PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS_MSGS;
+
+ if (msgs) {
+ *msgs = NULL;
+ }
if (OFFUNI_IS_INVARIANT(uv)) {
*d++ = LATIN1_TO_NATIVE(uv);
if (UNLIKELY( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv)
|| UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
{
- HANDLE_UNICODE_NONCHAR(uv, flags);
+ HANDLE_UNICODE_NONCHAR(uv, flags, msgs);
}
else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) {
- HANDLE_UNICODE_SURROGATE(uv, flags);
+ HANDLE_UNICODE_SURROGATE(uv, flags, msgs);
}
}
#endif
* 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_ABOVE_31_BIT(uv)
- && (flags & UNICODE_WARN_ABOVE_31_BIT)))
+ if ( (flags & UNICODE_WARN_SUPER)
+ || ( (flags & UNICODE_WARN_PERL_EXTENDED)
+ && UNICODE_IS_PERL_EXTENDED(uv)))
{
- Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
+ const char * format = super_cp_format;
+ U32 category = packWARN(WARN_NON_UNICODE);
+ U32 flag = UNICODE_GOT_SUPER;
+
+ /* Choose the more dire applicable warning */
+ if (UNICODE_IS_PERL_EXTENDED(uv)) {
+ format = perl_extended_cp_format;
+ if (flags & (UNICODE_WARN_PERL_EXTENDED
+ |UNICODE_DISALLOW_PERL_EXTENDED))
+ {
+ flag = UNICODE_GOT_PERL_EXTENDED;
+ }
+ }
- /* Choose the more dire applicable warning */
- (UNICODE_IS_ABOVE_31_BIT(uv))
- ? "Code point 0x%" UVXf " is not Unicode, and not portable"
- : "Code point 0x%" UVXf " is not Unicode, may not be portable",
- uv);
+ if (msgs) {
+ *msgs = new_msg_hv(Perl_form(aTHX_ format, uv),
+ category, flag);
+ }
+ else {
+ Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE), format, uv);
+ }
}
- if (flags & UNICODE_DISALLOW_SUPER
- || ( UNICODE_IS_ABOVE_31_BIT(uv)
- && (flags & UNICODE_DISALLOW_ABOVE_31_BIT)))
+ if ( (flags & UNICODE_DISALLOW_SUPER)
+ || ( (flags & UNICODE_DISALLOW_PERL_EXTENDED)
+ && UNICODE_IS_PERL_EXTENDED(uv)))
{
return NULL;
}
}
else if (UNLIKELY(UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) {
- HANDLE_UNICODE_NONCHAR(uv, flags);
+ HANDLE_UNICODE_NONCHAR(uv, flags, msgs);
}
/* Test for and handle 4-byte result. In the test immediately below, the
characters. The end-plane non-characters for EBCDIC were
handled just above */
if (UNLIKELY(UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv))) {
- HANDLE_UNICODE_NONCHAR(uv, flags);
+ HANDLE_UNICODE_NONCHAR(uv, flags, msgs);
}
else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) {
- HANDLE_UNICODE_SURROGATE(uv, flags);
+ HANDLE_UNICODE_SURROGATE(uv, flags, msgs);
}
#endif
*(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 on EBCDIC
+platforms,these flags can apply to code points that actually do fit in 31 bits.
+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.
+ * 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.
*
- * 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:
- *
- * 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_ARGS_ASSERT_IS_UTF8_CP_ABOVE_31_BITS;
- PERL_UNUSED_ARG(e);
+ assert(! UTF8_IS_INVARIANT(*s) && e > s);
-#endif
-
- 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 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 if 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));
+ /* 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;
+ }
-#endif
+ /* 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;
+ }
-}
+ /* 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 */
+ /* Here, ASCII and EBCDIC rejoin:
+ * On ASCII: We have an overlong sequence starting with FF
+ * On EBCDIC: We have a sequence starting with FE. */
- PERL_ARGS_ASSERT_DOES_UTF8_OVERFLOW;
- assert(s <= e && s + UTF8SKIP(s) >= e);
+ { /* For C89, use a block so the declaration can be close to its use */
-#if ! defined(UV_IS_QUAD) && ! defined(EBCDIC)
+#ifdef EBCDIC
- /* On 32 bit ASCII machines, many overlongs that start with FF don't
- * overflow */
+ /* 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";
- 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));
- }
+#else
-#endif
+ /* 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";
- for (x = s; x < e; x++, y++) {
- /* If this byte is larger than the corresponding highest UTF-8 byte, it
- * overflows */
- if (UNLIKELY(NATIVE_UTF8_TO_I8(*x) > *y)) {
- return TRUE;
+#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));
}
- /* If not the same as this byte, it must be smaller, doesn't overflow */
- if (LIKELY(NATIVE_UTF8_TO_I8(*x) != *y)) {
- 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;
}
-
- /* 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;
}
-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;
+ }
- return len >= sizeof(FF_OVERLONG_PREFIX) - 1
- && UNLIKELY(memEQ(s, FF_OVERLONG_PREFIX,
- sizeof(FF_OVERLONG_PREFIX) - 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
+
+}
+
+#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
* return will be larger than 'e - s'.
*
* This function assumes that the code point represented is UTF-8 variant.
- * The caller should have excluded this possibility before calling this
- * function.
+ * The caller should have excluded the possibility of it being invariant
+ * before calling this function.
*
* 'flags' can be 0, or any combination of the UTF8_DISALLOW_foo flags
* accepted by L</utf8n_to_uvchr>. If non-zero, this function will return
PERL_ARGS_ASSERT__IS_UTF8_CHAR_HELPER;
assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE
- |UTF8_DISALLOW_ABOVE_31_BIT)));
+ |UTF8_DISALLOW_PERL_EXTENDED)));
assert(! UTF8_IS_INVARIANT(*s));
/* A variant char must begin with a start byte */
if (flags && isUTF8_POSSIBLY_PROBLEMATIC(*s)) {
const U8 s0 = NATIVE_UTF8_TO_I8(s[0]);
- /* The code below is derived from this table. Keep in mind that legal
- * continuation bytes range between \x80..\xBF for UTF-8, and
- * \xA0..\xBF for I8. Anything above those aren't continuation bytes.
- * Hence, we don't have to test the upper edge because if any of those
- * are encountered, the sequence is malformed, and will fail elsewhere
- * in this function.
+ /* Here, we are disallowing some set of largish code points, and the
+ * first byte indicates the sequence is for a code point that could be
+ * in the excluded set. We generally don't have to look beyond this or
+ * the second byte to see if the sequence is actually for one of the
+ * excluded classes. The code below is derived from this table:
+ *
* UTF-8 UTF-EBCDIC I8
* U+D800: \xED\xA0\x80 \xF1\xB6\xA0\xA0 First surrogate
* U+DFFF: \xED\xBF\xBF \xF1\xB7\xBF\xBF Final surrogate
* U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0 First above Unicode
*
+ * Keep in mind that legal continuation bytes range between \x80..\xBF
+ * for UTF-8, and \xA0..\xBF for I8. Anything above those aren't
+ * continuation bytes. Hence, we don't have to test the upper edge
+ * because if any of those is encountered, the sequence is malformed,
+ * and would fail elsewhere in this function.
+ *
+ * The code here likewise assumes that there aren't other
+ * malformations; again the function should fail elsewhere because of
+ * these. For example, an overlong beginning with FC doesn't actually
+ * have to be a super; it could actually represent a small code point,
+ * even U+0000. But, since overlongs (and other malformations) are
+ * illegal, the function should return FALSE in either case.
*/
#ifdef EBCDIC /* On EBCDIC, these are actually I8 bytes */
# 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)
return 0; /* Above Unicode */
}
- if ( (flags & UTF8_DISALLOW_ABOVE_31_BIT)
- && UNLIKELY(is_utf8_cp_above_31_bits(s, e)))
+ if ( (flags & UTF8_DISALLOW_PERL_EXTENDED)
+ && 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;
}
return UTF8SKIP(s);
}
-STATIC char *
-S__byte_dump_string(pTHX_ const U8 * s, const STRLEN len)
+char *
+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', each in a \xXY format. */
+ * 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);
- *d++ = '\\';
- *d++ = 'x';
+ if (format) {
+ if (s > start) {
+ *d++ = ' ';
+ }
+ }
+ else {
+ *d++ = '\\';
+ *d++ = 'x';
+ }
if (high_nibble < 10) {
*d++ = high_nibble + '0';
return Perl_form(aTHX_ "%s: %s (unexpected non-continuation byte 0x%02x,"
" %s after start byte 0x%02x; need %d bytes, got %d)",
malformed_text,
- _byte_dump_string(s, print_len),
+ _byte_dump_string(s, print_len, 0),
*(s + non_cont_byte_pos),
where,
*s,
character. Also, if UTF-8 warnings haven't been lexically disabled, a warning
is raised. Some UTF-8 input sequences may contain multiple malformations.
This function tries to find every possible one in each call, so multiple
-warnings can be raised for each sequence.
+warnings can be raised for the same sequence.
Various ALLOW flags can be set in C<flags> to allow (and not warn on)
individual types of malformations, such as the sequence being overlong (that
(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.
-
-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.)
+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.
-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
return utf8n_to_uvchr_error(s, curlen, retlen, flags, NULL);
}
+/* The tables below come from http://bjoern.hoehrmann.de/utf-8/decoder/dfa/,
+ * which requires this copyright notice */
+
+/* Copyright (c) 2008-2009 Bjoern Hoehrmann <bjoern@hoehrmann.de>
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal in
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+of the Software, and to permit persons to whom the Software is furnished to do
+so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+*/
+
+#if 0
+static U8 utf8d_C9[] = {
+ /* The first part of the table maps bytes to character classes that
+ * to reduce the size of the transition table and create bitmasks. */
+ 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-1F*/
+ 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-3F*/
+ 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-5F*/
+ 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-7F*/
+ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, /*-9F*/
+ 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, /*-BF*/
+ 8,8,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /*-DF*/
+ 10,3,3,3,3,3,3,3,3,3,3,3,3,4,3,3, 11,6,6,6,5,8,8,8,8,8,8,8,8,8,8,8, /*-FF*/
+
+ /* The second part is a transition table that maps a combination
+ * of a state of the automaton and a character class to a state. */
+ 0,12,24,36,60,96,84,12,12,12,48,72, 12,12,12,12,12,12,12,12,12,12,12,12,
+ 12, 0,12,12,12,12,12, 0,12, 0,12,12, 12,24,12,12,12,12,12,24,12,24,12,12,
+ 12,12,12,12,12,12,12,24,12,12,12,12, 12,24,12,12,12,12,12,12,12,24,12,12,
+ 12,12,12,12,12,12,12,36,12,36,12,12, 12,36,12,12,12,12,12,36,12,36,12,12,
+ 12,36,12,12,12,12,12,12,12,12,12,12
+};
+
+#endif
+
+#ifndef EBCDIC
+
+/* This is a version of the above table customized for Perl that doesn't
+ * exclude surrogates and accepts start bytes up through F7 (representing
+ * 2**21 - 1). */
+static U8 dfa_tab_for_perl[] = {
+ /* The first part of the table maps bytes to character classes to reduce
+ * the size of the transition table and create bitmasks. */
+ 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-1F*/
+ 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-3F*/
+ 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-5F*/
+ 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-7F*/
+ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, /*-9F*/
+ 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, /*-BF*/
+ 8,8,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /*-DF*/
+ 10,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, 11,4,4,4,4,4,4,4,8,8,8,8,8,8,8,8, /*-FF*/
+
+ /* The second part is a transition table that maps a combination
+ * of a state of the automaton and a character class to a state. */
+ 0,12,24,36,96,12,12,12,12,12,48,72, 12,12,12,12,12,12,12,12,12,12,12,12,/*23*/
+ 12, 0,12,12,12,12,12, 0,12, 0,12,12, 12,24,12,12,12,12,12,24,12,24,12,12,/*47*/
+ 12,12,12,12,12,12,12,24,12,12,12,12, 12,24,12,12,12,12,12,12,12,24,12,12,/*71*/
+ 12,12,12,12,12,12,12,36,12,36,12,12, 12,36,12,12,12,12,12,36,12,36,12,12,/*95*/
+ 12,36,12,12,12,12,12,36,12,36,12,12 /* 96- 107 */
+
+ /* The customization was to repurpose the surrogates type '4' to instead be
+ * for start bytes F1-F7. Types 5 and 6 are now unused, and their entries in
+ * the transition part of the table are set to 12, so are illegal.
+ *
+ * To do higher code points would require expansion and some rearrangement of
+ * the table. The type '1' entries for continuation bytes 80-8f would have to
+ * be split into several types, because they aren't treated uniformly for
+ * higher start bytes, since overlongs for F8 are 80-87; FC: 80-83; and FE:
+ * 80-81. We start needing to worry about overflow if FE is included.
+ * Ignoring, FE and FF, we could use type 5 for F9-FB, and 6 for FD (remember
+ * from the web site that these are used to right shift). FE would
+ * necessarily be type 7; and FF, type 8. And new states would have to be
+ * created for F8 and FC (and FE and FF if used), so quite a bit of work would
+ * be involved.
+ *
+ * XXX Better would be to customize the table so that the noncharacters are
+ * excluded. This again is non trivial, but doing so would simplify the code
+ * that uses this, and might make it small enough to make it inlinable */
+};
+
+#endif
+
/*
=for apidoc utf8n_to_uvchr_error
Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
This function is for code that needs to know what the precise malformation(s)
-are when an error is found.
+are when an error is found. If you also need to know the generated warning
+messages, use L</utf8n_to_uvchr_msgs>() instead.
It is like C<L</utf8n_to_uvchr>> but it takes an extra parameter placed after
all the others, C<errors>. If this parameter is 0, this function behaves
=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>
The input sequence was malformed in that there is some other sequence that
evaluates to the same code point, but that sequence is shorter than this one.
+Until Unicode 3.1, it was legal for programs to accept this malformation, but
+it was discovered that this created security issues.
+
=item C<UTF8_GOT_NONCHAR>
The code point represented by the input UTF-8 sequence is for a Unicode
=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>
flag to suppress any warnings, and then examine the C<*errors> return.
=cut
+
+Also implemented as a macro in utf8.h
*/
UV
Perl_utf8n_to_uvchr_error(pTHX_ const U8 *s,
- STRLEN curlen,
- STRLEN *retlen,
- const U32 flags,
- U32 * errors)
+ STRLEN curlen,
+ STRLEN *retlen,
+ const U32 flags,
+ U32 * errors)
+{
+ PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_ERROR;
+
+ return utf8n_to_uvchr_msgs(s, curlen, retlen, flags, errors, NULL);
+}
+
+/*
+
+=for apidoc utf8n_to_uvchr_msgs
+
+THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES.
+Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
+
+This function is for code that needs to know what the precise malformation(s)
+are when an error is found, and wants the corresponding warning and/or error
+messages to be returned to the caller rather than be displayed. All messages
+that would have been displayed if all lexcial warnings are enabled will be
+returned.
+
+It is just like C<L</utf8n_to_uvchr_error>> but it takes an extra parameter
+placed after all the others, C<msgs>. If this parameter is 0, this function
+behaves identically to C<L</utf8n_to_uvchr_error>>. Otherwise, C<msgs> should
+be a pointer to an C<AV *> variable, in which this function creates a new AV to
+contain any appropriate messages. The elements of the array are ordered so
+that the first message that would have been displayed is in the 0th element,
+and so on. Each element is a hash with three key-value pairs, as follows:
+
+=over 4
+
+=item C<text>
+
+The text of the message as a C<SVpv>.
+
+=item C<warn_categories>
+
+The warning category (or categories) packed into a C<SVuv>.
+
+=item C<flag>
+
+A single flag bit associated with this message, in a C<SVuv>.
+The bit corresponds to some bit in the C<*errors> return value,
+such as C<UTF8_GOT_LONG>.
+
+=back
+
+It's important to note that specifying this parameter as non-null will cause
+any warnings this function would otherwise generate to be suppressed, and
+instead be placed in C<*msgs>. The caller can check the lexical warnings state
+(or not) when choosing what to do with the returned messages.
+
+If the flag C<UTF8_CHECK_ONLY> is passed, no warnings are generated, and hence
+no AV is created.
+
+The caller, of course, is responsible for freeing any returned AV.
+
+=cut
+*/
+
+UV
+Perl_utf8n_to_uvchr_msgs(pTHX_ const U8 *s,
+ STRLEN curlen,
+ STRLEN *retlen,
+ const U32 flags,
+ U32 * errors,
+ AV ** msgs)
{
const U8 * const s0 = s;
- U8 * send = NULL; /* (initialized to silence compilers' wrong
- warning) */
+ const U8 * send = s0 + curlen;
U32 possible_problems = 0; /* A bit is set here for each potential problem
found as we go along */
- UV uv = *s;
+ UV uv = (UV) -1;
STRLEN expectlen = 0; /* How long should this sequence be?
(initialized to silence compilers' wrong
warning) */
* 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) */
- PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_ERROR;
+ UV state = 0;
+
+ PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_MSGS;
if (errors) {
*errors = 0;
}
/* An invariant is trivially well-formed */
- if (UTF8_IS_INVARIANT(uv)) {
- return uv;
+ if (UTF8_IS_INVARIANT(*s0)) {
+ return *s0;
+ }
+
+#ifndef EBCDIC
+
+ /* Measurements show that this dfa is somewhat faster than the regular code
+ * below, so use it first, dropping down for the non-normal cases. */
+
+# define PERL_UTF8_DECODE_REJECT 12
+
+ while (s < send && LIKELY(state != PERL_UTF8_DECODE_REJECT)) {
+ UV type = dfa_tab_for_perl[*s];
+
+ if (state != 0) {
+ uv = (*s & 0x3fu) | (uv << UTF_ACCUMULATION_SHIFT);
+ state = dfa_tab_for_perl[256 + state + type];
+ }
+ else {
+ uv = (0xff >> type) & (*s);
+ state = dfa_tab_for_perl[256 + type];
+ }
+
+ if (state == 0) {
+
+ /* If this could be a code point that the flags don't allow (the first
+ * surrogate is the first such possible one), delve further, but we already
+ * have calculated 'uv' */
+ if ( (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE
+ |UTF8_WARN_ILLEGAL_INTERCHANGE))
+ && uv >= UNICODE_SURROGATE_FIRST)
+ {
+ curlen = s + 1 - s0;
+ goto got_uv;
+ }
+
+ return uv;
+ }
+
+ s++;
}
+ /* Here, is some sort of failure. Use the full mechanism */
+
+ uv = *s0;
+
+#endif
+
/* A continuation character can't start a valid sequence */
if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) {
possible_problems |= UTF8_GOT_CONTINUATION;
/* Setup the loop end point, making sure to not look past the end of the
* input string, and flag it as too short if the size isn't big enough. */
- send = (U8*) s0;
if (UNLIKELY(curlen < expectlen)) {
possible_problems |= UTF8_GOT_SHORT;
avail_len = curlen;
- send += curlen;
}
else {
- send += expectlen;
+ send = (U8*) s0 + expectlen;
}
- adjusted_send = send;
/* Now, loop through the remaining bytes in the character's sequence,
* accumulating each into the working value as we go. */
* if allowing this malformation. */
possible_problems |= UTF8_GOT_NON_CONTINUATION;
break;
- } /* End of loop through the character's bytes */
-
- /* Save how many bytes were actually in the character */
- curlen = s - s0;
-
- /* A convenience macro that matches either of the too-short conditions. */
-# define UTF8_GOT_TOO_SHORT (UTF8_GOT_SHORT|UTF8_GOT_NON_CONTINUATION)
-
- if (UNLIKELY(possible_problems & UTF8_GOT_TOO_SHORT)) {
- uv_so_far = uv;
- uv = UNICODE_REPLACEMENT;
- }
+ } /* End of loop through the character's bytes */
+
+ /* Save how many bytes were actually in the character */
+ curlen = s - s0;
/* Note that there are two types of too-short malformation. One is when
* there is actual wrong data before the normal termination of the
* This means that we were passed data for a partial character, but it is
* valid as far as we saw. The other is definitely invalid. This
* distinction could be important to a caller, so the two types are kept
- * separate. */
+ * separate.
+ *
+ * A convenience macro that matches either of the too-short conditions. */
+# define UTF8_GOT_TOO_SHORT (UTF8_GOT_SHORT|UTF8_GOT_NON_CONTINUATION)
+
+ if (UNLIKELY(possible_problems & UTF8_GOT_TOO_SHORT)) {
+ uv_so_far = uv;
+ 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;
}
* overlong */
if ( ( LIKELY(! possible_problems)
&& UNLIKELY(expectlen > (STRLEN) OFFUNISKIP(uv)))
- || ( UNLIKELY( possible_problems)
+ || ( 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;
- if (UNLIKELY(possible_problems & UTF8_GOT_TOO_SHORT)) {
+ if ( UNLIKELY( possible_problems & UTF8_GOT_TOO_SHORT)
+
+ /* The calculation in the 'true' branch of this 'if'
+ * below won't work if overflows, and isn't needed
+ * anyway. Further below we handle all overflow
+ * cases */
+ && LIKELY(! (possible_problems & UTF8_GOT_OVERFLOW)))
+ {
UV min_uv = uv_so_far;
STRLEN i;
* bytes. There is no single code point it could be for, but there
* may be enough information present to determine if what we have
* so far is for an unallowed code point, such as for a surrogate.
- * The code below has the intelligence to determine this, but just
- * for non-overlong UTF-8 sequences. What we do here is calculate
- * the smallest code point the input could represent if there were
- * no too short malformation. Then we compute and save the UTF-8
- * for that, which is what the code below looks at instead of the
- * raw input. It turns out that the smallest such code point is
- * all we need. */
+ * The code further below has the intelligence to determine this,
+ * but just for non-overlong UTF-8 sequences. What we do here is
+ * calculate the smallest code point the input could represent if
+ * there were no too short malformation. Then we compute and save
+ * the UTF-8 for that, which is what the code below looks at
+ * instead of the raw input. It turns out that the smallest such
+ * code point is all we need. */
for (i = curlen; i < expectlen; i++) {
min_uv = UTF8_ACCUMULATE(min_uv,
I8_TO_NATIVE_UTF8(UTF_CONTINUATION_MARK));
}
- Newx(adjusted_s0, OFFUNISKIP(min_uv) + 1, U8);
- SAVEFREEPV((U8 *) adjusted_s0); /* Needed because we may not get
- to free it ourselves if
- warnings are made fatal */
- adjusted_send = uvoffuni_to_utf8_flags(adjusted_s0, min_uv, 0);
+ adjusted_s0 = temp_char_buf;
+ (void) uvoffuni_to_utf8_flags(adjusted_s0, min_uv, 0);
}
}
- /* Now check that the input isn't for a problematic code point not allowed
- * by the input parameters. */
- /* isn't problematic if < this */
- if ( ( ( LIKELY(! possible_problems) && uv >= UNICODE_SURROGATE_FIRST)
+ got_uv:
+
+ /* Here, we have found all the possible problems, except for when the input
+ * is for a problematic code point not allowed by the input parameters. */
+
+ /* uv is valid for overlongs */
+ if ( ( ( LIKELY(! (possible_problems & ~UTF8_GOT_LONG))
+
+ /* isn't problematic if < this */
+ && uv >= UNICODE_SURROGATE_FIRST)
|| ( UNLIKELY(possible_problems)
/* if overflow, we know without looking further
* 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_DISALLOW_ABOVE_31_BIT
+ |UTF8_DISALLOW_PERL_EXTENDED
|UTF8_WARN_NONCHAR
|UTF8_WARN_SURROGATE
|UTF8_WARN_SUPER
- |UTF8_WARN_ABOVE_31_BIT))
- /* 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 */
* some subsitute value, typically the REPLACEMENT
* CHARACTER.
* s0 points to the first byte of the character
- * send points to just after where that (potentially
- * partial) character ends
- * 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
+ * s points to just after were we left off processing
+ * the character
+ * send points to just after where that character should
+ * end, based on how many bytes the start byte tells
+ * us should be in it, but no further than s0 +
+ * avail_len
*/
if (UNLIKELY(possible_problems)) {
bool disallowed = FALSE;
const U32 orig_problems = possible_problems;
+ if (msgs) {
+ *msgs = NULL;
+ }
+
while (possible_problems) { /* Handle each possible problem */
UV pack_warn = 0;
char * message = NULL;
+ U32 this_flag_bit = 0;
/* Each 'if' clause handles one problem. They are ordered so that
* the first ones' messages will be displayed before the later
- * ones; this is kinda in decreasing severity order */
+ * ones; this is kinda in decreasing severity order. But the
+ * overlong must come last, as it changes 'uv' looked at by the
+ * others */
if (possible_problems & UTF8_GOT_OVERFLOW) {
- /* Overflow means also got a super and above 31 bits, but we
- * handle all three cases here */
+ /* Overflow means also got a super and are using Perl's
+ * extended UTF-8, but we handle all three cases here */
possible_problems
- &= ~(UTF8_GOT_OVERFLOW|UTF8_GOT_SUPER|UTF8_GOT_ABOVE_31_BIT);
+ &= ~(UTF8_GOT_OVERFLOW|UTF8_GOT_SUPER|UTF8_GOT_PERL_EXTENDED);
*errors |= UTF8_GOT_OVERFLOW;
/* But the API says we flag all errors found */
*errors |= UTF8_GOT_SUPER;
}
if (flags
- & (UTF8_WARN_ABOVE_31_BIT|UTF8_DISALLOW_ABOVE_31_BIT))
+ & (UTF8_WARN_PERL_EXTENDED|UTF8_DISALLOW_PERL_EXTENDED))
{
- *errors |= UTF8_GOT_ABOVE_31_BIT;
+ *errors |= UTF8_GOT_PERL_EXTENDED;
}
/* Disallow if any of the three categories say to */
- if ( ! (flags & UTF8_ALLOW_OVERFLOW)
+ if ( ! (flags & UTF8_ALLOW_OVERFLOW)
|| (flags & ( UTF8_DISALLOW_SUPER
- |UTF8_DISALLOW_ABOVE_31_BIT)))
+ |UTF8_DISALLOW_PERL_EXTENDED)))
{
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_ABOVE_31_BIT)))
+ /* 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
* parent-child relationship. Even if it works now to
* raise the warning if either is enabled, it wouldn't
* necessarily do so in the future. We output (only) the
- * most dire warning*/
+ * most dire warning */
if (! (flags & UTF8_CHECK_ONLY)) {
- if (ckWARN_d(WARN_UTF8)) {
+ if (msgs || ckWARN_d(WARN_UTF8)) {
pack_warn = packWARN(WARN_UTF8);
}
- else if (ckWARN_d(WARN_NON_UNICODE)) {
+ else if (msgs || ckWARN_d(WARN_NON_UNICODE)) {
pack_warn = packWARN(WARN_NON_UNICODE);
}
if (pack_warn) {
message = Perl_form(aTHX_ "%s: %s (overflows)",
malformed_text,
- _byte_dump_string(s0, send - s0));
+ _byte_dump_string(s0, curlen, 0));
+ this_flag_bit = UTF8_GOT_OVERFLOW;
}
}
}
assert(0);
disallowed = TRUE;
- if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) {
+ if ( (msgs
+ || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY))
+ {
pack_warn = packWARN(WARN_UTF8);
message = Perl_form(aTHX_ "%s (empty string)",
malformed_text);
+ this_flag_bit = UTF8_GOT_EMPTY;
}
}
}
if (! (flags & UTF8_ALLOW_CONTINUATION)) {
disallowed = TRUE;
- if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) {
+ if (( msgs
+ || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY))
+ {
pack_warn = packWARN(WARN_UTF8);
message = Perl_form(aTHX_
"%s: %s (unexpected continuation byte 0x%02x,"
" with no preceding start byte)",
malformed_text,
- _byte_dump_string(s0, 1), *s0);
+ _byte_dump_string(s0, 1, 0), *s0);
+ this_flag_bit = UTF8_GOT_CONTINUATION;
}
}
}
if (! (flags & UTF8_ALLOW_SHORT)) {
disallowed = TRUE;
- if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) {
+ if (( msgs
+ || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY))
+ {
pack_warn = packWARN(WARN_UTF8);
message = Perl_form(aTHX_
- "%s: %s (too short; %d byte%s available, need %d)",
- malformed_text,
- _byte_dump_string(s0, send - s0),
- (int)avail_len,
- avail_len == 1 ? "" : "s",
- (int)expectlen);
+ "%s: %s (too short; %d byte%s available, need %d)",
+ malformed_text,
+ _byte_dump_string(s0, send - s0, 0),
+ (int)avail_len,
+ avail_len == 1 ? "" : "s",
+ (int)expectlen);
+ this_flag_bit = UTF8_GOT_SHORT;
}
}
if (! (flags & UTF8_ALLOW_NON_CONTINUATION)) {
disallowed = TRUE;
- if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) {
+ if (( msgs
+ || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY))
+ {
/* If we don't know for sure that the input length is
* valid, avoid as much as possible reading past the
printlen,
s - s0,
(int) expectlen));
- }
- }
- }
- else if (possible_problems & UTF8_GOT_LONG) {
- possible_problems &= ~UTF8_GOT_LONG;
- *errors |= UTF8_GOT_LONG;
-
- if (flags & UTF8_ALLOW_LONG) {
-
- /* We don't allow the actual overlong value, unless the
- * special extra bit is also set */
- if (! (flags & ( UTF8_ALLOW_LONG_AND_ITS_VALUE
- & ~UTF8_ALLOW_LONG)))
- {
- uv = UNICODE_REPLACEMENT;
- }
- }
- else {
- disallowed = TRUE;
-
- if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) {
- pack_warn = packWARN(WARN_UTF8);
-
- /* These error types cause 'uv' to be something that
- * isn't what was intended, so can't use it in the
- * message. The other error types either can't
- * generate an overlong, or else the 'uv' is valid */
- if (orig_problems &
- (UTF8_GOT_TOO_SHORT|UTF8_GOT_OVERFLOW))
- {
- message = Perl_form(aTHX_
- "%s: %s (any UTF-8 sequence that starts"
- " with \"%s\" is overlong which can and"
- " should be represented with a"
- " different, shorter sequence)",
- malformed_text,
- _byte_dump_string(s0, send - s0),
- _byte_dump_string(s0, curlen));
- }
- else {
- U8 tmpbuf[UTF8_MAXBYTES+1];
- const U8 * const e = uvoffuni_to_utf8_flags(tmpbuf,
- uv, 0);
- message = Perl_form(aTHX_
- "%s: %s (overlong; instead use %s to represent"
- " U+%0*" UVXf ")",
- malformed_text,
- _byte_dump_string(s0, send - s0),
- _byte_dump_string(tmpbuf, e - tmpbuf),
- ((uv < 256) ? 2 : 4), /* Field width of 2 for
- small code points */
- uv);
- }
+ this_flag_bit = UTF8_GOT_NON_CONTINUATION;
}
}
}
*errors |= UTF8_GOT_SURROGATE;
if ( ! (flags & UTF8_CHECK_ONLY)
- && ckWARN_d(WARN_SURROGATE))
+ && (msgs || ckWARN_d(WARN_SURROGATE)))
{
pack_warn = packWARN(WARN_SURROGATE);
message = Perl_form(aTHX_
"UTF-16 surrogate (any UTF-8 sequence that"
" starts with \"%s\" is for a surrogate)",
- _byte_dump_string(s0, curlen));
+ _byte_dump_string(s0, curlen, 0));
}
else {
- message = Perl_form(aTHX_
- "UTF-16 surrogate U+%04" UVXf, uv);
+ message = Perl_form(aTHX_ surrogate_cp_format, uv);
}
+ this_flag_bit = UTF8_GOT_SURROGATE;
}
}
*errors |= UTF8_GOT_SUPER;
if ( ! (flags & UTF8_CHECK_ONLY)
- && ckWARN_d(WARN_NON_UNICODE))
+ && (msgs || ckWARN_d(WARN_NON_UNICODE)))
{
pack_warn = packWARN(WARN_NON_UNICODE);
"Any UTF-8 sequence that starts with"
" \"%s\" is for a non-Unicode code point,"
" may not be portable",
- _byte_dump_string(s0, curlen));
+ _byte_dump_string(s0, curlen, 0));
}
else {
- message = Perl_form(aTHX_
- "Code point 0x%04" UVXf " is not"
- " Unicode, may not be portable",
- uv);
+ message = Perl_form(aTHX_ super_cp_format, uv);
}
+ this_flag_bit = UTF8_GOT_SUPER;
}
}
- /* 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_ABOVE_31_BIT
- |UTF8_WARN_SUPER
- |UTF8_DISALLOW_ABOVE_31_BIT))
- && ( ( 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_ABOVE_31_BIT(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_ABOVE_31_BIT|UTF8_WARN_SUPER))
- && ckWARN_d(WARN_UTF8))
+ && (flags & (UTF8_WARN_PERL_EXTENDED|UTF8_WARN_SUPER))
+ && (msgs || ckWARN_d(WARN_NON_UNICODE)))
{
- pack_warn = packWARN(WARN_UTF8);
+ 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));
+ perl_extended_cp_format, uv);
}
else {
message = Perl_form(aTHX_
- "Code point 0x%" UVXf " is not Unicode,"
- " and not portable",
- uv);
+ "Any UTF-8 sequence that starts with"
+ " \"%s\" is a Perl extension, and"
+ " so is not portable",
+ _byte_dump_string(s0, curlen, 0));
}
+ this_flag_bit = UTF8_GOT_PERL_EXTENDED;
}
- if (flags & ( UTF8_WARN_ABOVE_31_BIT
- |UTF8_DISALLOW_ABOVE_31_BIT))
+ if (flags & ( UTF8_WARN_PERL_EXTENDED
+ |UTF8_DISALLOW_PERL_EXTENDED))
{
- *errors |= UTF8_GOT_ABOVE_31_BIT;
+ *errors |= UTF8_GOT_PERL_EXTENDED;
- if (flags & UTF8_DISALLOW_ABOVE_31_BIT) {
+ if (flags & UTF8_DISALLOW_PERL_EXTENDED) {
disallowed = TRUE;
}
}
*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;
*errors |= UTF8_GOT_NONCHAR;
if ( ! (flags & UTF8_CHECK_ONLY)
- && ckWARN_d(WARN_NONCHAR))
+ && (msgs || ckWARN_d(WARN_NONCHAR)))
{
/* The code above should have guaranteed that we don't
* get here with errors other than overlong */
& ~(UTF8_GOT_LONG|UTF8_GOT_NONCHAR)));
pack_warn = packWARN(WARN_NONCHAR);
- message = Perl_form(aTHX_ "Unicode non-character"
- " U+%04" UVXf " is not recommended"
- " for open interchange", uv);
+ message = Perl_form(aTHX_ nonchar_cp_format, uv);
+ this_flag_bit = UTF8_GOT_NONCHAR;
}
}
disallowed = TRUE;
*errors |= UTF8_GOT_NONCHAR;
}
+ }
+ else if (possible_problems & UTF8_GOT_LONG) {
+ possible_problems &= ~UTF8_GOT_LONG;
+ *errors |= UTF8_GOT_LONG;
+
+ if (flags & UTF8_ALLOW_LONG) {
+
+ /* We don't allow the actual overlong value, unless the
+ * special extra bit is also set */
+ if (! (flags & ( UTF8_ALLOW_LONG_AND_ITS_VALUE
+ & ~UTF8_ALLOW_LONG)))
+ {
+ uv = UNICODE_REPLACEMENT;
+ }
+ }
+ else {
+ disallowed = TRUE;
+
+ if (( msgs
+ || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY))
+ {
+ pack_warn = packWARN(WARN_UTF8);
+
+ /* These error types cause 'uv' to be something that
+ * isn't what was intended, so can't use it in the
+ * message. The other error types either can't
+ * generate an overlong, or else the 'uv' is valid */
+ if (orig_problems &
+ (UTF8_GOT_TOO_SHORT|UTF8_GOT_OVERFLOW))
+ {
+ message = Perl_form(aTHX_
+ "%s: %s (any UTF-8 sequence that starts"
+ " with \"%s\" is overlong which can and"
+ " should be represented with a"
+ " different, shorter sequence)",
+ malformed_text,
+ _byte_dump_string(s0, send - s0, 0),
+ _byte_dump_string(s0, curlen, 0));
+ }
+ else {
+ U8 tmpbuf[UTF8_MAXBYTES+1];
+ const U8 * const e = uvoffuni_to_utf8_flags(tmpbuf,
+ uv, 0);
+ /* 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 ")",
+ malformed_text,
+ _byte_dump_string(s0, send - s0, 0),
+ _byte_dump_string(tmpbuf, e - tmpbuf, 0),
+ preface,
+ ((uv < 256) ? 2 : 4), /* Field width of 2 for
+ small code points */
+ UNI_TO_NATIVE(uv));
+ }
+ this_flag_bit = UTF8_GOT_LONG;
+ }
+ }
} /* End of looking through the possible flags */
/* Display the message (if any) for the problem being handled in
* this iteration of the loop */
if (message) {
- if (PL_op)
+ if (msgs) {
+ assert(this_flag_bit);
+
+ if (*msgs == NULL) {
+ *msgs = newAV();
+ }
+
+ av_push(*msgs, newRV_noinc((SV*) new_msg_hv(message,
+ pack_warn,
+ this_flag_bit)));
+ }
+ else if (PL_op)
Perl_warner(aTHX_ pack_warn, "%s in %s", message,
OP_DESC(PL_op));
else
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
UV
Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
{
+ PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF;
+
assert(s < send);
return utf8n_to_uvchr(s, send - s, retlen,
Only in very rare circumstances should code need to be dealing in Unicode
(as opposed to native) code points. In those few cases, use
-C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> instead.
+C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> instead. If you
+are not absolutely sure this is one of those cases, then assume it isn't and
+use plain C<utf8_to_uvchr_buf> instead.
Returns the Unicode (not-native) code point of the first character in the
string C<s> which
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
*/
} else {
/* diag_listed_as: Malformed UTF-8 character%s */
Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8),
- "%s %s%s",
- unexpected_non_continuation_text(u - 1, 2, 1, 2),
- PL_op ? " in " : "",
- PL_op ? OP_DESC(PL_op) : "");
+ "%s %s%s",
+ unexpected_non_continuation_text(u - 2, 2, 1, 2),
+ PL_op ? " in " : "",
+ PL_op ? OP_DESC(PL_op) : "");
return -2;
}
} else {
/*
=for apidoc utf8_to_bytes
-Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
+Converts a string C<"s"> of length C<*lenp> from UTF-8 into native byte encoding.
Unlike L</bytes_to_utf8>, this over-writes the original string, and
-updates C<len> to contain the new length.
-Returns zero on failure, setting C<len> to -1.
+updates C<*lenp> to contain the new length.
+Returns zero on failure (leaving C<"s"> unchanged) setting C<*lenp> to -1.
+
+Upon successful return, the number of variants in the string can be computed by
+having saved the value of C<*lenp> before the call, and subtracting the
+after-call value of C<*lenp> from it.
If you need a copy of the string, see L</bytes_from_utf8>.
*/
U8 *
-Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len)
+Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *lenp)
{
- U8 * const save = s;
- U8 * const send = s + *len;
- U8 *d;
+ U8 * first_variant;
PERL_ARGS_ASSERT_UTF8_TO_BYTES;
PERL_UNUSED_CONTEXT;
- /* ensure valid UTF-8 and chars < 256 before updating string */
- while (s < send) {
- if (! UTF8_IS_INVARIANT(*s)) {
- if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) {
- *len = ((STRLEN) -1);
- return 0;
+ /* This is a no-op if no variants at all in the input */
+ if (is_utf8_invariant_string_loc(s, *lenp, (const U8 **) &first_variant)) {
+ return s;
+ }
+
+ {
+ U8 * const save = s;
+ U8 * const send = s + *lenp;
+ U8 * d;
+
+ /* Nothing before the first variant needs to be changed, so start the real
+ * work there */
+ s = first_variant;
+ while (s < send) {
+ if (! UTF8_IS_INVARIANT(*s)) {
+ if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) {
+ *lenp = ((STRLEN) -1);
+ return 0;
+ }
+ s++;
}
s++;
}
- s++;
- }
- d = s = save;
- while (s < send) {
- U8 c = *s++;
- if (! UTF8_IS_INVARIANT(c)) {
- /* Then it is two-byte encoded */
- c = EIGHT_BIT_UTF8_TO_NATIVE(c, *s);
- s++;
- }
- *d++ = c;
+ /* Is downgradable, so do it */
+ d = s = first_variant;
+ while (s < send) {
+ U8 c = *s++;
+ if (! UVCHR_IS_INVARIANT(c)) {
+ /* Then it is two-byte encoded */
+ c = EIGHT_BIT_UTF8_TO_NATIVE(c, *s);
+ s++;
+ }
+ *d++ = c;
+ }
+ *d = '\0';
+ *lenp = d - save;
+
+ return save;
}
- *d = '\0';
- *len = d - save;
- return save;
}
/*
=for apidoc bytes_from_utf8
-Converts a string C<s> of length C<len> from UTF-8 into native byte encoding.
-Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to
-the newly-created string, and updates C<len> to contain the new
-length. Returns the original string if no conversion occurs, C<len>
-is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to
-0 if C<s> is converted or consisted entirely of characters that are invariant
-in UTF-8 (i.e., US-ASCII on non-EBCDIC machines).
+Converts a potentially UTF-8 encoded string C<s> of length C<*lenp> into native
+byte encoding. On input, the boolean C<*is_utf8p> gives whether or not C<s> is
+actually encoded in UTF-8.
+
+Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, this is non-destructive of
+the input string.
+
+Do nothing if C<*is_utf8p> is 0, or if there are code points in the string
+not expressible in native byte encoding. In these cases, C<*is_utf8p> and
+C<*lenp> are unchanged, and the return value is the original C<s>.
+
+Otherwise, C<*is_utf8p> is set to 0, and the return value is a pointer to a
+newly created string containing a downgraded copy of C<s>, and whose length is
+returned in C<*lenp>, updated. The new string is C<NUL>-terminated. The
+caller is responsible for arranging for the memory used by this string to get
+freed.
+
+Upon successful return, the number of variants in the string can be computed by
+having saved the value of C<*lenp> before the call, and subtracting the
+after-call value of C<*lenp> from it.
=cut
+
+There is a macro that avoids this function call, but this is retained for
+anyone who calls it with the Perl_ prefix */
+
+U8 *
+Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *lenp, bool *is_utf8p)
+{
+ PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
+ PERL_UNUSED_CONTEXT;
+
+ return bytes_from_utf8_loc(s, lenp, is_utf8p, NULL);
+}
+
+/*
+No = here because currently externally undocumented
+for apidoc bytes_from_utf8_loc
+
+Like C<L</bytes_from_utf8>()>, but takes an extra parameter, a pointer to where
+to store the location of the first character in C<"s"> that cannot be
+converted to non-UTF8.
+
+If that parameter is C<NULL>, this function behaves identically to
+C<bytes_from_utf8>.
+
+Otherwise if C<*is_utf8p> is 0 on input, the function behaves identically to
+C<bytes_from_utf8>, except it also sets C<*first_non_downgradable> to C<NULL>.
+
+Otherwise, the function returns a newly created C<NUL>-terminated string
+containing the non-UTF8 equivalent of the convertible first portion of
+C<"s">. C<*lenp> is set to its length, not including the terminating C<NUL>.
+If the entire input string was converted, C<*is_utf8p> is set to a FALSE value,
+and C<*first_non_downgradable> is set to C<NULL>.
+
+Otherwise, C<*first_non_downgradable> set to point to the first byte of the
+first character in the original string that wasn't converted. C<*is_utf8p> is
+unchanged. Note that the new string may have length 0.
+
+Another way to look at it is, if C<*first_non_downgradable> is non-C<NULL> and
+C<*is_utf8p> is TRUE, this function starts at the beginning of C<"s"> and
+converts as many characters in it as possible stopping at the first one it
+finds that can't be converted to non-UTF-8. C<*first_non_downgradable> is
+set to point to that. The function returns the portion that could be converted
+in a newly created C<NUL>-terminated string, and C<*lenp> is set to its length,
+not including the terminating C<NUL>. If the very first character in the
+original could not be converted, C<*lenp> will be 0, and the new string will
+contain just a single C<NUL>. If the entire input string was converted,
+C<*is_utf8p> is set to FALSE and C<*first_non_downgradable> is set to C<NULL>.
+
+Upon successful return, the number of variants in the converted portion of the
+string can be computed by having saved the value of C<*lenp> before the call,
+and subtracting the after-call value of C<*lenp> from it.
+
+=cut
+
+
*/
U8 *
-Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8)
+Perl_bytes_from_utf8_loc(const U8 *s, STRLEN *lenp, bool *is_utf8p, const U8** first_unconverted)
{
U8 *d;
- const U8 *start = s;
- const U8 *send;
- I32 count = 0;
+ const U8 *original = s;
+ U8 *converted_start;
+ const U8 *send = s + *lenp;
- PERL_ARGS_ASSERT_BYTES_FROM_UTF8;
- PERL_UNUSED_CONTEXT;
- if (!*is_utf8)
- return (U8 *)start;
-
- /* ensure valid UTF-8 and chars < 256 before converting string */
- for (send = s + *len; s < send;) {
- if (! UTF8_IS_INVARIANT(*s)) {
- if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) {
- return (U8 *)start;
- }
- count++;
- s++;
- }
- s++;
+ PERL_ARGS_ASSERT_BYTES_FROM_UTF8_LOC;
+
+ if (! *is_utf8p) {
+ if (first_unconverted) {
+ *first_unconverted = NULL;
+ }
+
+ return (U8 *) original;
}
- *is_utf8 = FALSE;
+ Newx(d, (*lenp) + 1, U8);
- Newx(d, (*len) - count + 1, U8);
- s = start; start = d;
+ converted_start = d;
while (s < send) {
- U8 c = *s++;
- if (! UTF8_IS_INVARIANT(c)) {
- /* Then it is two-byte encoded */
- c = EIGHT_BIT_UTF8_TO_NATIVE(c, *s);
+ U8 c = *s++;
+ if (! UTF8_IS_INVARIANT(c)) {
+
+ /* Then it is multi-byte encoded. If the code point is above 0xFF,
+ * have to stop now */
+ if (UNLIKELY (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s - 1, send))) {
+ if (first_unconverted) {
+ *first_unconverted = s - 1;
+ goto finish_and_return;
+ }
+ else {
+ Safefree(converted_start);
+ return (U8 *) original;
+ }
+ }
+
+ c = EIGHT_BIT_UTF8_TO_NATIVE(c, *s);
s++;
- }
- *d++ = c;
+ }
+ *d++ = c;
+ }
+
+ /* Here, converted the whole of the input */
+ *is_utf8p = FALSE;
+ if (first_unconverted) {
+ *first_unconverted = NULL;
}
+
+ finish_and_return:
*d = '\0';
- *len = d - start;
- return (U8 *)start;
+ *lenp = d - converted_start;
+
+ /* Trim unused space */
+ Renew(converted_start, *lenp + 1, U8);
+
+ return converted_start;
}
/*
=for apidoc bytes_to_utf8
-Converts a string C<s> of length C<len> bytes from the native encoding into
+Converts a string C<s> of length C<*lenp> bytes from the native encoding into
UTF-8.
-Returns a pointer to the newly-created string, and sets C<len> to
-reflect the new length in bytes.
+Returns a pointer to the newly-created string, and sets C<*lenp> to
+reflect the new length in bytes. The caller is responsible for arranging for
+the memory used by this string to get freed.
+
+Upon successful return, the number of variants in the string can be computed by
+having saved the value of C<*lenp> before the call, and subtracting it from the
+after-call value of C<*lenp>.
A C<NUL> character will be written after the end of the string.
=cut
*/
-/* This logic is duplicated in sv_catpvn_flags, so any bug fixes will
- likewise need duplication. */
-
U8*
-Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len)
+Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *lenp)
{
- const U8 * const send = s + (*len);
+ const U8 * const send = s + (*lenp);
U8 *d;
U8 *dst;
PERL_ARGS_ASSERT_BYTES_TO_UTF8;
PERL_UNUSED_CONTEXT;
- Newx(d, (*len) * 2 + 1, U8);
+ Newx(d, (*lenp) * 2 + 1, U8);
dst = d;
while (s < send) {
append_utf8_from_native_byte(*s, &d);
s++;
}
+
*d = '\0';
- *len = d-dst;
+ *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().
*
- * Destination must be pre-extended to 3/2 source. Do not use in-place.
- * We optimize for native, for obvious reasons. */
+ * 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.
+ *
+ * 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)
PERL_ARGS_ASSERT_UTF16_TO_UTF8;
if (bytelen & 1)
- Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %" UVuf, (UV)bytelen);
+ Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %" UVuf,
+ (UV)bytelen);
pend = p + bytelen;
*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);
}
UV
-Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s)
+Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp,
+ const char S_or_s)
{
/* We have the latin1-range values compiled into the core, so just use
* those, converting the result to UTF-8. The only difference between upper
return 'S';
#endif
default:
- Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
+ Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect"
+ " '%c' to map to '%c'",
+ c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS);
NOT_REACHED; /* NOTREACHED */
}
}
* needs to have space for UTF8_MAXBYTES_CASE+1 bytes
* LENP will be set to the length in bytes of the string of changed characters
*
- * The functions return the ordinal of the first character in the string of OUTP */
-#define CALL_UPPER_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_toupper, "ToUc", "")
-#define CALL_TITLE_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_totitle, "ToTc", "")
-#define CALL_LOWER_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_tolower, "ToLc", "")
+ * The functions return the ordinal of the first character in the string of
+ * OUTP */
+#define CALL_UPPER_CASE(uv, s, d, lenp) \
+ _to_utf8_case(uv, s, d, lenp, &PL_utf8_toupper, "ToUc", "")
+#define CALL_TITLE_CASE(uv, s, d, lenp) \
+ _to_utf8_case(uv, s, d, lenp, &PL_utf8_totitle, "ToTc", "")
+#define CALL_LOWER_CASE(uv, s, d, lenp) \
+ _to_utf8_case(uv, s, d, lenp, &PL_utf8_tolower, "ToLc", "")
/* This additionally has the input parameter 'specials', which if non-zero will
* cause this to use the specials hash for folding (meaning get full case
* folding); otherwise, when zero, this implies a simple case fold */
-#define CALL_FOLD_CASE(uv, s, d, lenp, specials) _to_utf8_case(uv, s, d, lenp, &PL_utf8_tofold, "ToCf", (specials) ? "" : NULL)
+#define CALL_FOLD_CASE(uv, s, d, lenp, specials) \
+_to_utf8_case(uv, s, d, lenp, &PL_utf8_tofold, "ToCf", (specials) ? "" : NULL)
UV
Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
}
UV
-Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const unsigned int flags)
+Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp,
+ const unsigned int flags)
{
/* Corresponds to to_lower_latin1(); <flags> bits meanings:
* FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited
PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
if (flags & FOLD_FLAGS_LOCALE) {
- /* Treat a UTF-8 locale as not being in locale at all */
+ /* Treat a UTF-8 locale as not being in locale at all, except for
+ * potentially warning */
+ _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
if (IN_UTF8_CTYPE_LOCALE) {
flags &= ~FOLD_FLAGS_LOCALE;
}
else {
- _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
goto needs_full_generality;
}
}
needs_full_generality:
uvchr_to_utf8(utf8_c, c);
- return _toFOLD_utf8_flags(utf8_c, p, lenp, flags);
+ return _toFOLD_utf8_flags(utf8_c, utf8_c + sizeof(utf8_c),
+ p, lenp, flags);
}
}
}
PERL_STATIC_INLINE bool
-S_is_utf8_common_with_len(pTHX_ const U8 *const p, const U8 * const e, SV **swash,
- const char *const swashname, SV* const invlist)
+S_is_utf8_common_with_len(pTHX_ const U8 *const p, const U8 * const e,
+ SV **swash, const char *const swashname,
+ SV* const invlist)
{
/* returns a boolean giving whether or not the UTF8-encoded character that
* starts at <p>, and extending no further than <e - 1> is in the swash
Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
}
+ if (instr(file, "mathoms.c")) {
+ Perl_warner(aTHX_ WARN_DEPRECATED,
+ "In %s, line %d, starting in Perl v5.30, %s()"
+ " will be removed. Avoid this message by"
+ " converting to use %s().\n",
+ file, line, name, alternative);
+ }
+ else {
Perl_warner(aTHX_ WARN_DEPRECATED,
"In %s, line %d, starting in Perl v5.30, %s() will"
" require an additional parameter. Avoid this"
" message by converting to use %s().\n",
file, line, name, alternative);
+ }
}
}
}
bool
Perl__is_utf8_FOO(pTHX_ U8 classnum,
- const U8 *p,
+ const U8 * const p,
const char * const name,
const char * const alternative,
const bool use_utf8,
warn_on_first_deprecated_use(name, alternative, use_locale, file, line);
if (use_utf8 && UTF8_IS_ABOVE_LATIN1(*p)) {
- SV * invlist;
switch (classnum) {
case _CC_WORDCHAR:
return is_VERTWS_high(p);
case _CC_IDFIRST:
if (! PL_utf8_perl_idstart) {
- invlist = _new_invlist_C_array(_Perl_IDStart_invlist);
+ PL_utf8_perl_idstart
+ = _new_invlist_C_array(_Perl_IDStart_invlist);
}
- return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart", invlist);
+ return is_utf8_common(p, &PL_utf8_perl_idstart,
+ "_Perl_IDStart", NULL);
case _CC_IDCONT:
if (! PL_utf8_perl_idcont) {
- invlist = _new_invlist_C_array(_Perl_IDCont_invlist);
+ PL_utf8_perl_idcont
+ = _new_invlist_C_array(_Perl_IDCont_invlist);
}
- return is_utf8_common(p, &PL_utf8_perl_idcont, "_Perl_IDCont", invlist);
+ return is_utf8_common(p, &PL_utf8_perl_idcont,
+ "_Perl_IDCont", NULL);
}
}
return is_utf8_common(p, &PL_utf8_mark, "IsM", NULL);
}
-/*
-=for apidoc to_utf8_case
-
-Instead use the appropriate one of L</toUPPER_utf8>,
-L</toTITLE_utf8>,
-L</toLOWER_utf8>,
-or L</toFOLD_utf8>.
-
-C<p> contains the pointer to the UTF-8 string encoding
-the character that is being converted. This routine assumes that the character
-at C<p> is well-formed.
-
-C<ustrp> is a pointer to the character buffer to put the
-conversion result to. C<lenp> is a pointer to the length
-of the result.
-
-C<swashp> is a pointer to the swash to use.
-
-Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>,
-and loaded by C<SWASHNEW>, using F<lib/utf8_heavy.pl>. C<special> (usually,
-but not always, a multicharacter mapping), is tried first.
-
-C<special> is a string, normally C<NULL> or C<"">. C<NULL> means to not use
-any special mappings; C<""> means to use the special mappings. Values other
-than these two are treated as the name of the hash containing the special
-mappings, like C<"utf8::ToSpecLower">.
-
-C<normal> is a string like C<"ToLower"> which means the swash
-C<%utf8::ToLower>.
-
-Code points above the platform's C<IV_MAX> will raise a deprecation warning,
-unless those are turned off.
-
-=cut */
-
-UV
-Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
- SV **swashp, const char *normal, const char *special)
-{
- PERL_ARGS_ASSERT_TO_UTF8_CASE;
-
- return _to_utf8_case(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp, swashp, normal, special);
-}
-
/* change namve uv1 to 'from' */
STATIC UV
S__to_utf8_case(pTHX_ const UV uv1, const U8 *p, U8* ustrp, STRLEN *lenp,
if (ckWARN_d(WARN_SURROGATE)) {
const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
- "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04" UVXf, desc, uv1);
+ "Operation \"%s\" returns its argument for"
+ " UTF-16 surrogate U+%04" UVXf, desc, uv1);
}
goto cases_to_self;
}
}
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;
Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
- "Operation \"%s\" returns its argument for non-Unicode code point 0x%04" UVXf, desc, uv1);
+ "Operation \"%s\" returns its argument for"
+ " non-Unicode code point 0x%04" UVXf, desc, uv1);
}
goto cases_to_self;
}
> HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C))
{
- /* As of this writing, this means we avoid swash creation
- * for anything beyond low Plane 1 */
+ /* As of Unicode 10.0, this means we avoid swash creation
+ * for anything beyond high Plane 1 (below emojis) */
goto cases_to_self;
}
#endif
}
if (!*swashp) /* load on-demand */
- *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL);
+ *swashp = _core_swash_init("utf8", normal, &PL_sv_undef,
+ 4, 0, NULL, NULL);
if (special) {
/* It might be "special" (sometimes, but not always,
}
STATIC UV
-S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp)
+S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result,
+ U8* const ustrp, STRLEN *lenp)
{
/* This is called when changing the case of a UTF-8-encoded character above
* the Latin1 range, and the operation is in a non-UTF-8 locale. If the
* p points to the original string whose case was changed; assumed
* by this routine to be well-formed
* result the code point of the first character in the changed-case string
- * ustrp points to the changed-case string (<result> represents its first char)
+ * ustrp points to the changed-case string (<result> represents its
+ * first char)
* lenp points to the length of <ustrp> */
UV original; /* To store the first code point of <p> */
/* 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 %s(\"\\x{%" UVXf "}\") on non-UTF-8 locale; "
- "resolved to \"\\x{%" UVXf "}\".",
+ "Can't do %s(\"\\x{%" UVXf "}\") on non-UTF-8"
+ " locale; resolved to \"\\x{%" UVXf "}\".",
OP_DESC(PL_op),
original,
original);
return original;
}
+STATIC U32
+S_check_and_deprecate(pTHX_ const U8 *p,
+ const U8 **e,
+ const unsigned int type, /* See below */
+ const bool use_locale, /* Is this a 'LC_'
+ macro call? */
+ const char * const file,
+ const unsigned line)
+{
+ /* This is a temporary function to deprecate the unsafe calls to the case
+ * changing macros and functions. It keeps all the special stuff in just
+ * one place.
+ *
+ * It updates *e with the pointer to the end of the input string. If using
+ * the old-style macros, *e is NULL on input, and so this function assumes
+ * the input string is long enough to hold the entire UTF-8 sequence, and
+ * sets *e accordingly, but it then returns a flag to pass the
+ * utf8n_to_uvchr(), to tell it that this size is a guess, and to avoid
+ * using the full length if possible.
+ *
+ * It also does the assert that *e > p when *e is not NULL. This should be
+ * migrated to the callers when this function gets deleted.
+ *
+ * The 'type' parameter is used for the caller to specify which case
+ * changing function this is called from: */
+
+# define DEPRECATE_TO_UPPER 0
+# define DEPRECATE_TO_TITLE 1
+# define DEPRECATE_TO_LOWER 2
+# define DEPRECATE_TO_FOLD 3
+
+ U32 utf8n_flags = 0;
+ const char * name;
+ const char * alternative;
+
+ PERL_ARGS_ASSERT_CHECK_AND_DEPRECATE;
+
+ if (*e == NULL) {
+ utf8n_flags = _UTF8_NO_CONFIDENCE_IN_CURLEN;
+ *e = p + UTF8SKIP(p);
+
+ /* For mathoms.c calls, we use the function name we know is stored
+ * there. It could be part of a larger path */
+ if (type == DEPRECATE_TO_UPPER) {
+ name = instr(file, "mathoms.c")
+ ? "to_utf8_upper"
+ : "toUPPER_utf8";
+ alternative = "toUPPER_utf8_safe";
+ }
+ else if (type == DEPRECATE_TO_TITLE) {
+ name = instr(file, "mathoms.c")
+ ? "to_utf8_title"
+ : "toTITLE_utf8";
+ alternative = "toTITLE_utf8_safe";
+ }
+ else if (type == DEPRECATE_TO_LOWER) {
+ name = instr(file, "mathoms.c")
+ ? "to_utf8_lower"
+ : "toLOWER_utf8";
+ alternative = "toLOWER_utf8_safe";
+ }
+ else if (type == DEPRECATE_TO_FOLD) {
+ name = instr(file, "mathoms.c")
+ ? "to_utf8_fold"
+ : "toFOLD_utf8";
+ alternative = "toFOLD_utf8_safe";
+ }
+ else Perl_croak(aTHX_ "panic: Unexpected case change type");
+
+ warn_on_first_deprecated_use(name, alternative, use_locale, file, line);
+ }
+ else {
+ assert (p < *e);
+ }
+
+ return utf8n_flags;
+}
+
/* The process for changing the case is essentially the same for the four case
* change types, except there are complications for folding. Otherwise the
* difference is only which case to change to. To make sure that they all do
* going on. */
#define CASE_CHANGE_BODY_START(locale_flags, LC_L1_change_macro, L1_func, \
L1_func_extra_param) \
+ \
if (flags & (locale_flags)) { \
+ _CHECK_AND_WARN_PROBLEMATIC_LOCALE; \
/* Treat a UTF-8 locale as not being in locale at all */ \
if (IN_UTF8_CTYPE_LOCALE) { \
flags &= ~(locale_flags); \
} \
- else { \
- _CHECK_AND_WARN_PROBLEMATIC_LOCALE; \
- } \
} \
\
if (UTF8_IS_INVARIANT(*p)) { \
return L1_func(*p, ustrp, lenp, L1_func_extra_param); \
} \
} \
- else if UTF8_IS_DOWNGRADEABLE_START(*p) { \
+ else if UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e) { \
if (flags & (locale_flags)) { \
result = LC_L1_change_macro(EIGHT_BIT_UTF8_TO_NATIVE(*p, \
*(p+1))); \
ustrp, lenp, L1_func_extra_param); \
} \
} \
- else { /* malformed UTF-8 */ \
- result = valid_utf8_to_uvchr(p, NULL); \
+ else { /* malformed UTF-8 or ord above 255 */ \
+ STRLEN len_result; \
+ result = utf8n_to_uvchr(p, e - p, &len_result, UTF8_CHECK_ONLY); \
+ if (len_result == (STRLEN) -1) { \
+ _force_out_malformed_utf8_message(p, e, utf8n_flags, \
+ 1 /* Die */ ); \
+ }
#define CASE_CHANGE_BODY_END(locale_flags, change_macro) \
result = change_macro(result, p, ustrp, lenp); \
/*
=for apidoc to_utf8_upper
-Instead use L</toUPPER_utf8>.
+Instead use L</toUPPER_utf8_safe>.
=cut */
* be used. */
UV
-Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, bool flags)
+Perl__to_utf8_upper_flags(pTHX_ const U8 *p,
+ const U8 *e,
+ U8* ustrp,
+ STRLEN *lenp,
+ bool flags,
+ const char * const file,
+ const int line)
{
UV result;
+ const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_UPPER,
+ cBOOL(flags), file, line);
PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS;
/*
=for apidoc to_utf8_title
-Instead use L</toTITLE_utf8>.
+Instead use L</toTITLE_utf8_safe>.
=cut */
*/
UV
-Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, bool flags)
+Perl__to_utf8_title_flags(pTHX_ const U8 *p,
+ const U8 *e,
+ U8* ustrp,
+ STRLEN *lenp,
+ bool flags,
+ const char * const file,
+ const int line)
{
UV result;
+ const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_TITLE,
+ cBOOL(flags), file, line);
PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS;
/*
=for apidoc to_utf8_lower
-Instead use L</toLOWER_utf8>.
+Instead use L</toLOWER_utf8_safe>.
=cut */
*/
UV
-Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, bool flags)
+Perl__to_utf8_lower_flags(pTHX_ const U8 *p,
+ const U8 *e,
+ U8* ustrp,
+ STRLEN *lenp,
+ bool flags,
+ const char * const file,
+ const int line)
{
UV result;
+ const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_LOWER,
+ cBOOL(flags), file, line);
PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS;
/*
=for apidoc to_utf8_fold
-Instead use L</toFOLD_utf8>.
+Instead use L</toFOLD_utf8_safe>.
=cut */
*/
UV
-Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags)
+Perl__to_utf8_fold_flags(pTHX_ const U8 *p,
+ const U8 *e,
+ U8* ustrp,
+ STRLEN *lenp,
+ U8 flags,
+ const char * const file,
+ const int line)
{
UV result;
+ const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_FOLD,
+ cBOOL(flags), file, line);
PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS;
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; "
*/
SV*
-Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none)
+Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv,
+ I32 minbits, I32 none)
{
PERL_ARGS_ASSERT_SWASH_INIT;
* public interface, and returning a copy prevents others from doing
* mischief on the original */
- return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL));
+ return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none,
+ NULL, NULL));
}
SV*
-Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p)
+Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv,
+ I32 minbits, I32 none, SV* invlist,
+ U8* const flags_p)
{
/*NOTE NOTE NOTE - If you want to use "return" in this routine you MUST
assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
assert(! invlist || minbits == 1);
- PL_curpm= NULL; /* reset PL_curpm so that we dont get confused between the regex
- that triggered the swash init and the swash init perl logic itself.
- See perl #122747 */
+ PL_curpm= NULL; /* reset PL_curpm so that we dont get confused between the
+ regex that triggered the swash init and the swash init
+ perl logic itself. See perl #122747 */
/* If data was passed in to go out to utf8_heavy to find the swash of, do
* so */
/* Must have at least 8 bits to get the mappings */
if (bits != 8 && bits != 16 && bits != 32) {
- Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %" UVuf,
- (UV)bits);
+ Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"
+ UVuf, (UV)bits);
}
if (specials_p) { /* It might be "special" (sometimes, but not always, a
while ((from_list = (AV *) hv_iternextsv(specials_inverse,
&char_to, &to_len)))
{
- if (av_tindex_nomg(from_list) > 0) {
+ if (av_tindex_skip_len_mg(from_list) > 0) {
SSize_t i;
/* We iterate over all combinations of i,j to place each code
* point on each list */
- for (i = 0; i <= av_tindex_nomg(from_list); i++) {
+ for (i = 0; i <= av_tindex_skip_len_mg(from_list); i++) {
SSize_t j;
AV* i_list = newAV();
SV** entryp = av_fetch(from_list, i, FALSE);
if (entryp == NULL) {
- Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
+ Perl_croak(aTHX_ "panic: av_fetch() unexpectedly"
+ " failed");
}
if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) {
- Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp));
+ Perl_croak(aTHX_ "panic: unexpected entry for %s",
+ SvPVX(*entryp));
}
if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp),
(SV*) i_list, FALSE))
}
/* For DEBUG_U: UV u = valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/
- for (j = 0; j <= av_tindex_nomg(from_list); j++) {
+ for (j = 0; j <= av_tindex_skip_len_mg(from_list); j++) {
entryp = av_fetch(from_list, j, FALSE);
if (entryp == NULL) {
Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
/* Look through list to see if this inverse mapping already is
* listed, or if there is a mapping to itself already */
- for (i = 0; i <= av_tindex_nomg(list); i++) {
+ for (i = 0; i <= av_tindex_skip_len_mg(list); i++) {
SV** entryp = av_fetch(list, i, FALSE);
SV* entry;
UV uv;
/* The first number is a count of the rest */
l++;
if (!grok_atoUV((const char *)l, &elements, &after_atou)) {
- Perl_croak(aTHX_ "panic: Expecting a valid count of elements at start of inversion list");
+ Perl_croak(aTHX_ "panic: Expecting a valid count of elements"
+ " at start of inversion list");
}
if (elements == 0) {
invlist = _new_invlist(0);
}
else {
- while (isSPACE(*l)) l++;
l = (U8 *) after_atou;
- /* Get the 0th element, which is needed to setup the inversion list */
+ /* Get the 0th element, which is needed to setup the inversion list
+ * */
while (isSPACE(*l)) l++;
if (!grok_atoUV((const char *)l, &element0, &after_atou)) {
- Perl_croak(aTHX_ "panic: Expecting a valid 0th element for inversion list");
+ Perl_croak(aTHX_ "panic: Expecting a valid 0th element for"
+ " inversion list");
}
l = (U8 *) after_atou;
- invlist = _setup_canned_invlist(elements, element0, &other_elements_ptr);
+ invlist = _setup_canned_invlist(elements, element0,
+ &other_elements_ptr);
elements--;
/* Then just populate the rest of the input */
while (elements-- > 0) {
if (l > lend) {
- Perl_croak(aTHX_ "panic: Expecting %" UVuf " more elements than available", elements);
+ Perl_croak(aTHX_ "panic: Expecting %" UVuf " more"
+ " elements than available", elements);
}
while (isSPACE(*l)) l++;
- if (!grok_atoUV((const char *)l, other_elements_ptr++, &after_atou)) {
- Perl_croak(aTHX_ "panic: Expecting a valid element in inversion list");
+ if (!grok_atoUV((const char *)l, other_elements_ptr++,
+ &after_atou))
+ {
+ Perl_croak(aTHX_ "panic: Expecting a valid element"
+ " in inversion list");
}
l = (U8 *) after_atou;
}
* 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++;
}
Perl_check_utf8_print(pTHX_ const U8* s, const STRLEN len)
{
/* 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. */
+ * 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 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;
* do for the non-chars and above-unicodes */
UV uv = utf8_to_uvchr_buf(s, e, NULL);
Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
- "Unicode surrogate U+%04" UVXf " is illegal in UTF-8", uv);
+ "Unicode surrogate U+%04" UVXf " is illegal in UTF-8",
+ uv);
ok = FALSE;
}
}
- else if (UNLIKELY(UTF8_IS_NONCHAR(s, e)) && (ckWARN_d(WARN_NONCHAR))) {
+ else if ( UNLIKELY(UTF8_IS_NONCHAR(s, e))
+ && (ckWARN_d(WARN_NONCHAR)))
+ {
/* A side effect of this function will be to warn */
(void) utf8n_to_uvchr(s, e - s, NULL, UTF8_WARN_NONCHAR);
ok = FALSE;
=cut */
char *
-Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags)
+Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim,
+ UV flags)
{
int truncated = 0;
const char *s, *e;
/*
=for apidoc foldEQ_utf8
-Returns true if the leading portions of the strings C<s1> and C<s2> (either or both
-of which may be in UTF-8) are the same case-insensitively; false otherwise.
-How far into the strings to compare is determined by other input parameters.
+Returns true if the leading portions of the strings C<s1> and C<s2> (either or
+both of which may be in UTF-8) are the same case-insensitively; false
+otherwise. How far into the strings to compare is determined by other input
+parameters.
If C<u1> is true, the string C<s1> is assumed to be in UTF-8-encoded Unicode;
-otherwise it is assumed to be in native 8-bit encoding. Correspondingly for C<u2>
-with respect to C<s2>.
-
-If the byte length C<l1> is non-zero, it says how far into C<s1> to check for fold
-equality. In other words, C<s1>+C<l1> will be used as a goal to reach. The
-scan will not be considered to be a match unless the goal is reached, and
-scanning won't continue past that goal. Correspondingly for C<l2> with respect to
-C<s2>.
-
-If C<pe1> is non-C<NULL> and the pointer it points to is not C<NULL>, that pointer is
-considered an end pointer to the position 1 byte past the maximum point
-in C<s1> beyond which scanning will not continue under any circumstances.
+otherwise it is assumed to be in native 8-bit encoding. Correspondingly for
+C<u2> with respect to C<s2>.
+
+If the byte length C<l1> is non-zero, it says how far into C<s1> to check for
+fold equality. In other words, C<s1>+C<l1> will be used as a goal to reach.
+The scan will not be considered to be a match unless the goal is reached, and
+scanning won't continue past that goal. Correspondingly for C<l2> with respect
+to C<s2>.
+
+If C<pe1> is non-C<NULL> and the pointer it points to is not C<NULL>, that
+pointer is considered an end pointer to the position 1 byte past the maximum
+point in C<s1> beyond which scanning will not continue under any circumstances.
(This routine assumes that UTF-8 encoded input strings are not malformed;
-malformed input can cause it to read past C<pe1>).
-This means that if both C<l1> and C<pe1> are specified, and C<pe1>
-is less than C<s1>+C<l1>, the match will never be successful because it can
-never
+malformed input can cause it to read past C<pe1>). This means that if both
+C<l1> and C<pe1> are specified, and C<pe1> is less than C<s1>+C<l1>, the match
+will never be successful because it can never
get as far as its goal (and in fact is asserted against). Correspondingly for
C<pe2> with respect to C<s2>.
* FOLDEQ_S2_FOLDS_SANE
*/
I32
-Perl_foldEQ_utf8_flags(pTHX_ const char *s1, char **pe1, UV l1, bool u1, const char *s2, char **pe2, UV l2, bool u2, U32 flags)
+Perl_foldEQ_utf8_flags(pTHX_ const char *s1, char **pe1, UV l1, bool u1,
+ const char *s2, char **pe2, UV l2, bool u2,
+ U32 flags)
{
const U8 *p1 = (const U8*)s1; /* Point to current char */
const U8 *p2 = (const U8*)s2;
*foldbuf1 = toFOLD(*p1);
}
else if (u1) {
- _toFOLD_utf8_flags(p1, foldbuf1, &n1, flags_for_folder);
+ _toFOLD_utf8_flags(p1, e1, foldbuf1, &n1, flags_for_folder);
}
else { /* Not UTF-8, get UTF-8 fold */
_to_uni_fold_flags(*p1, foldbuf1, &n1, flags_for_folder);
*foldbuf2 = toFOLD(*p2);
}
else if (u2) {
- _toFOLD_utf8_flags(p2, foldbuf2, &n2, flags_for_folder);
+ _toFOLD_utf8_flags(p2, e2, foldbuf2, &n2, flags_for_folder);
}
else {
_to_uni_fold_flags(*p2, foldbuf2, &n2, flags_for_folder);
{
PERL_ARGS_ASSERT_UVUNI_TO_UTF8;
- return Perl_uvoffuni_to_utf8_flags(aTHX_ d, uv, 0);
+ return uvoffuni_to_utf8_flags(d, uv, 0);
}
/*
https://perl5.git.perl.org / perl5.git / blobdiff