#define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
/* these are unrolled below in the CCC_TRY_XXX defined */
-#define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
+#ifdef EBCDIC
+ /* Often 'str' is a hard-coded utf8 string instead of utfebcdic. so just
+ * skip the check on EBCDIC platforms */
+# define LOAD_UTF8_CHARCLASS(class,str) LOAD_UTF8_CHARCLASS_NO_CHECK(class)
+#else
+# define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
if (!CAT2(PL_utf8_,class)) { \
bool ok; \
ENTER; save_re_context(); \
ok=CAT2(is_utf8_,class)((const U8*)str); \
- assert(ok); LEAVE; } } STMT_END
+ assert(ok); assert(CAT2(PL_utf8_,class)); LEAVE; } } STMT_END
+#endif
/* Doesn't do an assert to verify that is correct */
#define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \
if (!CAT2(PL_utf8_,class)) { \
- bool throw_away __attribute__unused__; \
+ bool throw_away PERL_UNUSED_DECL; \
ENTER; save_re_context(); \
throw_away = CAT2(is_utf8_,class)((const U8*)" "); \
LEAVE; } } STMT_END
(IV)prog->check_end_shift);
});
- if (flags & REXEC_SCREAM) {
+ if ((flags & REXEC_SCREAM) && SvSCREAM(sv)) {
I32 p = -1; /* Internal iterator of scream. */
I32 * const pp = data ? data->scream_pos : &p;
+ const MAGIC *mg;
+ bool found = FALSE;
- if (PL_screamfirst[BmRARE(check)] >= 0
+ assert(SvMAGICAL(sv));
+ mg = mg_find(sv, PERL_MAGIC_study);
+ assert(mg);
+
+ if (mg->mg_private == 1) {
+ found = ((U8 *)mg->mg_ptr)[BmRARE(check)] != (U8)~0;
+ } else if (mg->mg_private == 2) {
+ found = ((U16 *)mg->mg_ptr)[BmRARE(check)] != (U16)~0;
+ } else {
+ assert (mg->mg_private == 4);
+ found = ((U32 *)mg->mg_ptr)[BmRARE(check)] != (U32)~0;
+ }
+
+ if (found
|| ( BmRARE(check) == '\n'
&& (BmPREVIOUS(check) == SvCUR(check) - 1)
&& SvTAIL(check) ))
folder = foldEQ_latin1; /* /a, except the sharp s one which */
goto do_exactf_non_utf8; /* isn't dealt with by these */
- case EXACTFU:
- if (UTF_PATTERN || utf8_target) {
- utf8_fold_flags = 0;
- goto do_exactf_utf8;
- }
- fold_array = PL_fold_latin1;
- folder = foldEQ_latin1;
- /* XXX This uses the full utf8 fold because if the pattern contains
- * 'ss' it could match LATIN_SMALL_LETTER SHARP_S in the string.
- * There could be a new node type, say EXACTFU_SS, which is
- * generated by regcomp only if there is an 'ss', and then every
- * other case could goto do_exactf_non_utf8;*/
- goto do_exactf_utf8;
-
case EXACTF:
if (UTF_PATTERN || utf8_target) {
- utf8_fold_flags = 0;
+
+ /* regcomp.c already folded this if pattern is in UTF-8 */
+ utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
goto do_exactf_utf8;
}
fold_array = PL_fold;
}
fold_array = PL_fold_locale;
folder = foldEQ_locale;
+ goto do_exactf_non_utf8;
+
+ case EXACTFU:
+ if (UTF_PATTERN || utf8_target) {
+ utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
+ goto do_exactf_utf8;
+ }
+
+ /* Any 'ss' in the pattern should have been replaced by regcomp,
+ * so we don't have to worry here about this single special case
+ * in the Latin1 range */
+ fold_array = PL_fold_latin1;
+ folder = foldEQ_latin1;
/* FALL THROUGH */
pat_string = STRING(c);
ln = STR_LEN(c); /* length to match in octets/bytes */
+ /* We know that we have to match at least 'ln' bytes (which is the
+ * same as characters, since not utf8). If we have to match 3
+ * characters, and there are only 2 availabe, we know without
+ * trying that it will fail; so don't start a match past the
+ * required minimum number from the far end */
e = HOP3c(strend, -((I32)ln), s);
if (!reginfo && e < s) {
break;
do_exactf_utf8:
+ {
+ unsigned expansion;
+
/* If one of the operands is in utf8, we can't use the simpler
* folding above, due to the fact that many different characters
? utf8_length((U8 *) pat_string, (U8 *) pat_end)
: ln;
+ /* We have 'lnc' characters to match in the pattern, but because of
+ * multi-character folding, each character in the target can match
+ * up to 3 characters (Unicode guarantees it will never exceed
+ * this) if it is utf8-encoded; and up to 2 if not (based on the
+ * fact that the Latin 1 folds are already determined, and the
+ * only multi-char fold in that range is the sharp-s folding to
+ * 'ss'. Thus, a pattern character can match as little as 1/3 of a
+ * string character. Adjust lnc accordingly, rounding up, so that
+ * if we need to match at least 4+1/3 chars, that really is 5. */
+ expansion = (utf8_target) ? UTF8_MAX_FOLD_CHAR_EXPAND : 2;
+ lnc = (lnc + expansion - 1) / expansion;
+
+ /* As in the non-UTF8 case, if we have to match 3 characters, and
+ * only 2 are left, it's guaranteed to fail, so don't start a
+ * match that would require us to go beyond the end of the string
+ */
e = HOP3c(strend, -((I32)lnc), s);
if (!reginfo && e < s) {
e = s; /* Due to minlen logic of intuit() */
}
+ /* XXX Note that we could recalculate e to stop the loop earlier,
+ * as the worst case expansion above will rarely be met, and as we
+ * go along we would usually find that e moves further to the left.
+ * This would happen only after we reached the point in the loop
+ * where if there were no expansion we should fail. Unclear if
+ * worth the expense */
+
while (s <= e) {
char *my_strend= (char *)strend;
if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target,
{
goto got_it;
}
- s += UTF8SKIP(s);
+ s += (utf8_target) ? UTF8SKIP(s) : 1;
}
break;
+ }
case BOUNDL:
PL_reg_flags |= RF_tainted;
FBC_BOUND(isALNUM_LC,
dontbother = end_shift;
strend = HOPc(strend, -dontbother);
while ( (s <= last) &&
- ((flags & REXEC_SCREAM)
+ ((flags & REXEC_SCREAM) && SvSCREAM(sv)
? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg,
end_shift, &scream_pos, 0))
: (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)),
utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
float_real = utf8_target ? prog->float_utf8 : prog->float_substr;
- if (flags & REXEC_SCREAM) {
+ if ((flags & REXEC_SCREAM) && SvSCREAM(sv)) {
last = screaminstr(sv, float_real, s - strbeg,
end_shift, &scream_pos, 1); /* last one */
if (!last)
case EXACTFU:
folder = foldEQ_latin1;
fold_array = PL_fold_latin1;
- fold_utf8_flags = 0;
+ fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0;
goto do_exactf;
case EXACTFA:
case EXACTF:
folder = foldEQ;
fold_array = PL_fold;
- fold_utf8_flags = 0;
+ fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0;
do_exactf:
s = STRING(scan);
case NBOUNDU:
case NBOUNDA:
/* was last char in word? */
- if (utf8_target && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET) {
+ if (utf8_target
+ && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET
+ && FLAGS(scan) != REGEX_ASCII_MORE_RESTRICTED_CHARSET)
+ {
if (locinput == PL_bostr)
ln = '\n';
else {
n = isALNUM(nextchr);
break;
case REGEX_ASCII_RESTRICTED_CHARSET:
+ case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
ln = isWORDCHAR_A(ln);
n = isWORDCHAR_A(nextchr);
break;
L* (L | LVT T* | V V* T* | LV V* T*)
That means that if we have seen any L's at all we can quit
- there, but if the next character is a LVT, a V or and LV we
+ there, but if the next character is an LVT, a V, or an LV we
should keep going.
There is a subtlety with Prepend* which showed up in testing.
Note that the Begin, and only the Begin is required in:
| Prepend* Begin Extend*
- Also, Begin contains '! Control'. A Prepend must be a '!
- Control', which means it must be a Begin. What it comes down to
- is that if we match Prepend* and then find no suitable Begin
- afterwards, that if we backtrack the last Prepend, that one will
- be a suitable Begin.
+ Also, Begin contains '! Control'. A Prepend must be a
+ '! Control', which means it must also be a Begin. What it
+ comes down to is that if we match Prepend* and then find no
+ suitable Begin afterwards, that if we backtrack the last
+ Prepend, that one will be a suitable Begin.
*/
if (locinput >= PL_regeol)
/* Match either CR LF or '.', as all the other possibilities
* require utf8 */
locinput++; /* Match the . or CR */
- if (nextchr == '\r'
+ if (nextchr == '\r' /* And if it was CR, and the next is LF,
+ match the LF */
&& locinput < PL_regeol
&& UCHARAT(locinput) == '\n') locinput++;
}
PL_op = (OP_4tree*)rexi->data->data[n];
DEBUG_STATE_r( PerlIO_printf(Perl_debug_log,
" re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
+ /* wrap the call in two SAVECOMPPADs. This ensures that
+ * when the save stack is eventually unwound, all the
+ * accumulated SAVEt_CLEARSV's will be processed with
+ * interspersed SAVEt_COMPPAD's to ensure that lexicals
+ * are cleared in the right pad */
+ SAVECOMPPAD();
PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]);
PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr;
Copy(&saved_state, &PL_reg_state, 1, struct re_save_state);
PL_op = oop;
+ SAVECOMPPAD();
PAD_RESTORE_LOCAL(old_comppad);
PL_curcop = ocurcop;
PL_regeol = saved_regeol;
#define CASE_CLASS(nAmE) \
case nAmE: \
+ if (locinput >= PL_regeol) \
+ sayNO; \
if ((n=is_##nAmE(locinput,utf8_target))) { \
locinput += n; \
nextchr = UCHARAT(locinput); \
sayNO; \
break; \
case N##nAmE: \
+ if (locinput >= PL_regeol) \
+ sayNO; \
if ((n=is_##nAmE(locinput,utf8_target))) { \
sayNO; \
} else { \
/* Here, the string is utf8, and the pattern char is different
* in utf8 than not, so can't compare them directly. Outside the
- * loop, find find the two utf8 bytes that represent c, and then
+ * loop, find the two utf8 bytes that represent c, and then
* look for those in sequence in the utf8 string */
U8 high = UTF8_TWO_BYTE_HI(c);
U8 low = UTF8_TWO_BYTE_LO(c);
case EXACTF:
case EXACTFU:
- utf8_flags = 0;
+ utf8_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
/* The comments for the EXACT case above apply as well to these fold
* ones */
STATIC U8 *
S_reghop3(U8 *s, I32 off, const U8* lim)
{
+ /* return the position 'off' UTF-8 characters away from 's', forward if
+ * 'off' >= 0, backwards if negative. But don't go outside of position
+ * 'lim', which better be < s if off < 0 */
+
dVAR;
PERL_ARGS_ASSERT_REGHOP3;
prog->substrs->data[i].utf8_substr = sv;
sv_utf8_upgrade(sv);
if (SvVALID(prog->substrs->data[i].substr)) {
- const U8 flags = BmFLAGS(prog->substrs->data[i].substr);
- if (flags & FBMcf_TAIL) {
+ if (SvTAIL(prog->substrs->data[i].substr)) {
/* Trim the trailing \n that fbm_compile added last
time. */
SvCUR_set(sv, SvCUR(sv) - 1);
/* Whilst this makes the SV technically "invalid" (as its
buffer is no longer followed by "\0") when fbm_compile()
adds the "\n" back, a "\0" is restored. */
- }
- fbm_compile(sv, flags);
+ fbm_compile(sv, FBMcf_TAIL);
+ } else
+ fbm_compile(sv, 0);
}
if (prog->substrs->data[i].substr == prog->check_substr)
prog->check_utf8 = sv;
SV* sv = newSVsv(prog->substrs->data[i].utf8_substr);
if (sv_utf8_downgrade(sv, TRUE)) {
if (SvVALID(prog->substrs->data[i].utf8_substr)) {
- const U8 flags
- = BmFLAGS(prog->substrs->data[i].utf8_substr);
- if (flags & FBMcf_TAIL) {
+ if (SvTAIL(prog->substrs->data[i].utf8_substr)) {
/* Trim the trailing \n that fbm_compile added last
time. */
SvCUR_set(sv, SvCUR(sv) - 1);
- }
- fbm_compile(sv, flags);
- }
+ fbm_compile(sv, FBMcf_TAIL);
+ } else
+ fbm_compile(sv, 0);
+ }
} else {
SvREFCNT_dec(sv);
sv = &PL_sv_undef;