a) Anchored substring;
b) Fixed substring;
c) Whether we are anchored (beginning-of-line or \G);
- d) First node (of those at offset 0) which may distingush positions;
+ d) First node (of those at offset 0) which may distinguish positions;
We use a)b)d) and multiline-part of c), and try to find a position in the
string which does not contradict any of them.
*/
prog->float_substr = prog->float_utf8 = NULL; /* clear */
check = NULL; /* abort */
s = strpos;
- /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevent flag
+ /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag
see http://bugs.activestate.com/show_bug.cgi?id=87173 */
if (prog->intflags & PREGf_IMPLICIT)
prog->extflags &= ~RXf_ANCH_MBOL;
}
if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
goto fail;
- /* Check is floating subtring. */
+ /* Check is floating substring. */
retry_floating_check:
t = check_at - start_shift;
DEBUG_EXECUTE_r( what = "floating" );
/* We know what class it must start with. */
switch (OP(c)) {
+ case ANYOFV:
case ANYOF:
- if (utf8_target) {
+ if (utf8_target || OP(c) == ANYOFV) {
REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_NONBITMAP) ||
!UTF8_IS_INVARIANT((U8)s[0]) ?
reginclass(prog, c, (U8*)s, 0, utf8_target) :
goto phooey;
} else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK))
{
- /* the warning about reginfo.ganch being used without intialization
+ /* the warning about reginfo.ganch being used without initialization
is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN
and we only enter this block when the same bit is set. */
char *tmp_s = reginfo.ganch - prog->gofs;
(?>A), which pop all the backtrack states associated with A before
continuing.
-Bascktrack state structs are allocated in slabs of about 4K in size.
+Backtrack state structs are allocated in slabs of about 4K in size.
PL_regmatch_state and st always point to the currently active state,
and PL_regmatch_slab points to the slab currently containing
PL_regmatch_state. The first time regmatch() is called, the first slab is
SV *popmark = NULL; /* are we looking for a mark? */
SV *sv_commit = NULL; /* last mark name seen in failure */
SV *sv_yes_mark = NULL; /* last mark name we have seen
- during a successfull match */
+ during a successful match */
U32 lastopen = 0; /* last open we saw */
bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0;
SV* const oreplsv = GvSV(PL_replgv);
ST.nextword,
tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
PL_colors[0], PL_colors[1],
- (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
+ (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII
)
: "not compiled under -Dr",
PL_colors[5] );
OP(scan) == BOUNDL))
sayNO;
break;
+ case ANYOFV:
case ANYOF:
- if (utf8_target) {
+ if (utf8_target || state_num == ANYOFV) {
STRLEN inclasslen = PL_regeol - locinput;
if (locinput >= PL_regeol)
sayNO;
if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target))
- goto anyof_fail;
+ sayNO;
locinput += inclasslen;
nextchr = UCHARAT(locinput);
break;
if (!nextchr && locinput >= PL_regeol)
sayNO;
if (!REGINCLASS(rex, scan, (U8*)locinput))
- goto anyof_fail;
+ sayNO;
nextchr = UCHARAT(++locinput);
break;
}
- anyof_fail:
- /* If we might have the case of the German sharp s
- * in a casefolding Unicode character class. */
-
- if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
- locinput += SHARP_S_SKIP;
- nextchr = UCHARAT(locinput);
- }
- else
- sayNO;
break;
/* Special char classes - The defines start on line 129 or so */
CCC_TRY_AFF_U( ALNUM, ALNUML, perl_word, "a", isALNUM_LC_utf8, isWORDCHAR_L1, isALNUM_LC);
re_fold_t folder;
const U8 *fold_array;
- folder = NULL; /* NULL assumes will be NREF, REF: no
- folding */
- fold_array = NULL;
-
PL_reg_flags |= RF_tainted;
folder = foldEQ_locale;
fold_array = PL_fold_locale;
char *tmpeol = loceol;
while (hardcount < max
&& foldEQ_utf8(scan, &tmpeol, 0, utf8_target,
- STRING(p), NULL, 1, UTF_PATTERN))
+ STRING(p), NULL, 1, cBOOL(UTF_PATTERN)))
{
scan = tmpeol;
tmpeol = loceol;
}
}
break;
+ case ANYOFV:
case ANYOF:
if (utf8_target) {
loceol = PL_regeol;
bytes in p were matched. If there was no match, the value is undefined,
possibly changed from the input.
+ Note that this can be a synthetic start class, a combination of various
+ nodes, so things you think might be mutually exclusive, such as locale,
+ aren't. It can match both locale and non-locale
+
*/
STATIC bool
if (c < 256) {
if (ANYOF_BITMAP_TEST(n, c))
match = TRUE;
- else if (flags & ANYOF_FOLD) {
- U8 f;
-
- if (flags & ANYOF_LOCALE) {
- PL_reg_flags |= RF_tainted;
- f = PL_fold_locale[c];
- }
- else
- f = PL_fold[c];
- if (f != c && ANYOF_BITMAP_TEST(n, f))
- match = TRUE;
+ else if (flags & ANYOF_NON_UTF8_LATIN1_ALL
+ && ! utf8_target
+ && ! isASCII(c))
+ {
+ match = TRUE;
}
-
- if (!match && ANYOF_CLASS_TEST_ANY_SET(n)) {
- PL_reg_flags |= RF_tainted; /* CLASS implies LOCALE */
- if (
- (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
- ) /* How's that for a conditional? */
+
+ else if (flags & ANYOF_LOCALE) {
+ PL_reg_flags |= RF_tainted;
+
+ if ((flags & ANYOF_LOC_NONBITMAP_FOLD)
+ && ANYOF_BITMAP_TEST(n, PL_fold_locale[c]))
{
match = TRUE;
}
+ else if (ANYOF_CLASS_TEST_ANY_SET(n) &&
+ ((ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
+ ) /* How's that for a conditional? */
+ ) {
+ match = TRUE;
+ }
}
}
if (!match) {
if (utf8_target && (flags & ANYOF_UNICODE_ALL)) {
if (c >= 256
- || ((flags & ANYOF_FOLD) /* Latin1 1 that has a non-Latin1 fold
- should match */
+ || ((flags & ANYOF_LOC_NONBITMAP_FOLD) /* Latin1 1 that has a
+ non-Latin1 fold
+ should match */
&& _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c)))
{
match = TRUE;
if (utf8_target) {
utf8_p = (U8 *) p;
} else {
- STRLEN len = 1;
+
+ /* Not utf8. Convert as much of the string as available up
+ * to the limit of how far the (single) character in the
+ * pattern can possibly match (no need to go further). If
+ * the node is a straight ANYOF or not folding, it can't
+ * match more than one. Otherwise, It can match up to how
+ * far a single char can fold to. Since not utf8, each
+ * character is a single byte, so the max it can be in
+ * bytes is the same as the max it can be in characters */
+ STRLEN len = (OP(n) == ANYOF
+ || ! (flags & ANYOF_LOC_NONBITMAP_FOLD))
+ ? 1
+ : (maxlen < UTF8_MAX_FOLD_CHAR_EXPAND)
+ ? maxlen
+ : UTF8_MAX_FOLD_CHAR_EXPAND;
utf8_p = bytes_to_utf8(p, &len);
}
- if (swash_fetch(sw, utf8_p, 1))
+
+ if (swash_fetch(sw, utf8_p, 1)) /* See if in the swash */
match = TRUE;
- else if (flags & ANYOF_FOLD) {
- if (!match && lenp && av) {
+ else if (flags & ANYOF_LOC_NONBITMAP_FOLD) {
+
+ /* Here, we need to test if the fold of the target string
+ * matches. In the case of a multi-char fold that is
+ * caught by regcomp.c, it has stored all such folds into
+ * 'av'; we linearly check to see if any match the target
+ * string (folded). We know that the originals were each
+ * one character, but we don't currently know how many
+ * characters/bytes each folded to, except we do know that
+ * there are small limits imposed by Unicode. XXX A
+ * performance enhancement would be to have regcomp.c store
+ * the max number of chars/bytes that are in an av entry,
+ * as, say the 0th element. Even better would be to have a
+ * hash of the few characters that can start a multi-char
+ * fold to the max number of chars of those folds.
+ *
+ * Further down, if there isn't a
+ * match in the av, we will check if there is another
+ * fold-type match. For that, we also need the fold, but
+ * only the first character. No sense in folding it twice,
+ * so we do it here, even if there isn't any multi-char
+ * fold, so we always fold at least the first character.
+ * If the node is a straight ANYOF node, or there is only
+ * one character available in the string, or if there isn't
+ * any av, that's all we have to fold. In the case of a
+ * multi-char fold, we do have guarantees in Unicode that
+ * it can only expand up to so many characters and so many
+ * bytes. We keep track so don't exceed either.
+ *
+ * If there is a match, we will need to advance (if lenp is
+ * specified) the match pointer in the target string. But
+ * what we are comparing here isn't that string directly,
+ * but its fold, whose length may differ from the original.
+ * As we go along in constructing the fold, therefore, we
+ * create a map so that we know how many bytes in the
+ * source to advance given that we have matched a certain
+ * number of bytes in the fold. This map is stored in
+ * 'map_fold_len_back'. The first character in the fold
+ * has array element 1 contain the number of bytes in the
+ * source that folded to it; the 2nd is the cumulative
+ * number to match it; ... */
+ U8 map_fold_len_back[UTF8_MAX_FOLD_CHAR_EXPAND] = { 0 };
+ U8 folded[UTF8_MAXBYTES_CASE+1];
+ STRLEN foldlen = 0; /* num bytes in fold of 1st char */
+ STRLEN foldlen_for_av; /* num bytes in fold of all chars */
+
+ if (OP(n) == ANYOF || maxlen == 1 || ! lenp || ! av) {
+
+ /* Here, only need to fold the first char of the target
+ * string */
+ to_utf8_fold(utf8_p, folded, &foldlen);
+ foldlen_for_av = foldlen;
+ map_fold_len_back[1] = UTF8SKIP(utf8_p);
+ }
+ else {
+
+ /* Here, need to fold more than the first char. Do so
+ * up to the limits */
+ UV which_char = 0;
+ U8* source_ptr = utf8_p; /* The source for the fold
+ is the regex target
+ string */
+ U8* folded_ptr = folded;
+ U8* e = utf8_p + maxlen; /* Can't go beyond last
+ available byte in the
+ target string */
+ while (which_char < UTF8_MAX_FOLD_CHAR_EXPAND
+ && source_ptr < e)
+ {
+
+ /* Fold the next character */
+ U8 this_char_folded[UTF8_MAXBYTES_CASE+1];
+ STRLEN this_char_foldlen;
+ to_utf8_fold(source_ptr,
+ this_char_folded,
+ &this_char_foldlen);
+
+ /* Bail if it would exceed the byte limit for
+ * folding a single char. */
+ if (this_char_foldlen + folded_ptr - folded >
+ UTF8_MAXBYTES_CASE)
+ {
+ break;
+ }
+
+ /* Save the first character's folded length, in
+ * case we have to use it later */
+ if (! foldlen) {
+ foldlen = this_char_foldlen;
+ }
+
+ /* Here, add the fold of this character */
+ Copy(this_char_folded,
+ folded_ptr,
+ this_char_foldlen,
+ U8);
+ which_char++;
+ map_fold_len_back[which_char] =
+ map_fold_len_back[which_char - 1]
+ + UTF8SKIP(source_ptr);
+ folded_ptr += this_char_foldlen;
+ source_ptr += UTF8SKIP(source_ptr);
+ }
+ *folded_ptr = '\0';
+ foldlen_for_av = folded_ptr - folded;
+ }
+
+
+ /* Do the linear search to see if the fold is in the list
+ * of multi-char folds. (Useless to look if won't be able
+ * to store that it is a multi-char fold in *lenp) */
+ if (lenp && av) {
I32 i;
for (i = 0; i <= av_len(av); i++) {
SV* const sv = *av_fetch(av, i, FALSE);
STRLEN len;
const char * const s = SvPV_const(sv, len);
- if (len <= maxlen && memEQ(s, (char*)utf8_p, len)) {
- *lenp = len;
+ if (len <= foldlen_for_av && memEQ(s,
+ (char*)folded,
+ len))
+ {
+
+ /* Advance the target string ptr to account for
+ * this fold, but have to translate from the
+ * folded length to the corresponding source
+ * length. The array is indexed by how many
+ * characters in the match */
+ *lenp = map_fold_len_back[
+ utf8_length(folded, folded + len)];
match = TRUE;
break;
}
}
}
- if (!match) {
- U8 folded[UTF8_MAXBYTES_CASE+1];
-
- /* See if the folded version matches */
- STRLEN foldlen;
- to_utf8_fold(utf8_p, folded, &foldlen);
- if (swash_fetch(sw, folded, 1)) { /* 1 => is utf8 */
- match = TRUE;
- }
- else {
- /* The fold in a few cases of an above Latin1 char
- * is in the Latin1 range, and hence may be in the
- * bitmap */
- if (UTF8_IS_INVARIANT(*folded)
- && ANYOF_BITMAP_TEST(n, UNI_TO_NATIVE(*folded)))
- {
- match = TRUE;
+ if (!match) { /* See if the folded version matches */
+ SV** listp;
+
+ /* Consider "k" =~ /[K]/i. The line above would have
+ * just folded the 'k' to itself, and that isn't going
+ * to match 'K'. So we look through the closure of
+ * everything that folds to 'k'. That will find the
+ * 'K'. Initialize the list, if necessary */
+ if (! PL_utf8_foldclosures) {
+
+ /* If the folds haven't been read in, call a fold
+ * function to force that */
+ if (! PL_utf8_tofold) {
+ U8 dummy[UTF8_MAXBYTES+1];
+ STRLEN dummy_len;
+ to_utf8_fold((U8*) "A", dummy, &dummy_len);
}
- else if (UTF8_IS_DOWNGRADEABLE_START(*folded)
- && ANYOF_BITMAP_TEST(n,
- UNI_TO_NATIVE(
- TWO_BYTE_UTF8_TO_UNI(folded[0],
- folded[1]))))
- { /* Since the fold comes from internally
- * generated data, we can safely assume it is
- * valid utf8 in the test above */
+ PL_utf8_foldclosures =
+ _swash_inversion_hash(PL_utf8_tofold);
+ }
- match = TRUE;
- }
- if (! match) {
- SV** listp;
-
- /* Consider "k" =~ /[K]/i. The line above
- * would have just folded the 'k' to itself,
- * and that isn't going to match 'K'. So we
- * look through the closure of everything that
- * folds to 'k'. That will find the 'K'.
- * Initialize the list, if necessary */
- if (! PL_utf8_foldclosures) {
-
- /* If the folds haven't been read in, call a
- * fold function to force that */
- if (! PL_utf8_tofold) {
- U8 dummy[UTF8_MAXBYTES+1];
- STRLEN dummy_len;
- to_utf8_fold((U8*) "A",
- dummy, &dummy_len);
- }
- PL_utf8_foldclosures =
- _swash_inversion_hash(PL_utf8_tofold);
+ /* The data structure is a hash with the keys every
+ * character that is folded to, like 'k', and the
+ * values each an array of everything that folds to its
+ * key. e.g. [ 'k', 'K', KELVIN_SIGN ] */
+ if ((listp = hv_fetch(PL_utf8_foldclosures,
+ (char *) folded, foldlen, FALSE)))
+ {
+ AV* list = (AV*) *listp;
+ IV i;
+ for (i = 0; i <= av_len(list); i++) {
+ SV** try_p = av_fetch(list, i, FALSE);
+ char* try_c;
+ if (try_p == NULL) {
+ Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure");
}
-
- /* The data structure is a hash with the keys
- * every character that is folded to, like 'k',
- * and the values each an array of everything
- * that folds to its key. e.g. [ 'k', 'K',
- * KELVIN_SIGN ] */
- if ((listp = hv_fetch(PL_utf8_foldclosures,
- (char *) folded, foldlen, FALSE)))
+ /* Don't have to worry about embedded nulls
+ * since NULL isn't folded or foldable */
+ try_c = SvPVX(*try_p);
+
+ /* The fold in a few cases of an above Latin1
+ * char is in the Latin1 range, and hence may
+ * be in the bitmap */
+ if (UTF8_IS_INVARIANT(*try_c)
+ && ANYOF_BITMAP_TEST(n,
+ UNI_TO_NATIVE(*try_c)))
{
- AV* list = (AV*) *listp;
- IV i;
- for (i = 0; i <= av_len(list); i++) {
- SV** try_p = av_fetch(list, i, FALSE);
- char* try_c;
- if (try_p == NULL) {
- Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure");
- }
- /* Don't have to worry about embeded
- * nulls since NULL isn't folded or
- * foldable */
- try_c = SvPVX(*try_p);
- if (UTF8_IS_INVARIANT(*try_c)
- && ANYOF_BITMAP_TEST(n,
- UNI_TO_NATIVE(*try_c)))
- {
- match = TRUE;
- break;
- }
- else if
- (UTF8_IS_DOWNGRADEABLE_START(*try_c)
- && ANYOF_BITMAP_TEST(n,
- UNI_TO_NATIVE(
+ match = TRUE;
+ break;
+ }
+ else if
+ (UTF8_IS_DOWNGRADEABLE_START(*try_c)
+ && ANYOF_BITMAP_TEST(n, UNI_TO_NATIVE(
TWO_BYTE_UTF8_TO_UNI(try_c[0],
- try_c[1]))))
- {
- match = TRUE;
- break;
- } else if (swash_fetch(sw,
- (U8*) try_c, 1))
- {
- match = TRUE;
- break;
- }
- }
+ try_c[1]))))
+ {
+ /* Since the fold comes from internally
+ * generated data, we can safely assume it
+ * is valid utf8 in the test above */
+ match = TRUE;
+ break;
+ } else if (swash_fetch(sw, (U8*) try_c, 1)) {
+ match = TRUE;
+ break;
}
}
- }
+ }
}
}