I32 seen_zerolen;
regnode_offset *open_parens; /* offsets to open parens */
regnode_offset *close_parens; /* offsets to close parens */
+ I32 parens_buf_size; /* #slots malloced open/close_parens */
regnode *end_op; /* END node in program */
I32 utf8; /* whether the pattern is utf8 or not */
I32 orig_utf8; /* whether the pattern was originally in utf8 */
#define RExC_maxlen (pRExC_state->maxlen)
#define RExC_npar (pRExC_state->npar)
#define RExC_total_parens (pRExC_state->total_par)
+#define RExC_parens_buf_size (pRExC_state->parens_buf_size)
#define RExC_nestroot (pRExC_state->nestroot)
#define RExC_seen_zerolen (pRExC_state->seen_zerolen)
#define RExC_utf8 (pRExC_state->utf8)
} \
} STMT_END
-#define BRANCH_MAX_OFFSET U16_MAX
#define REQUIRE_BRANCHJ(flagp, restart_retval) \
STMT_START { \
RExC_use_BRANCHJ = 1; \
/* Used to point after bad bytes for an error message, but avoid skipping
* past a nul byte. */
-#define SKIP_IF_CHAR(s) (!*(s) ? 0 : UTF ? UTF8SKIP(s) : 1)
+#define SKIP_IF_CHAR(s, e) (!*(s) ? 0 : UTF ? UTF8_SAFE_SKIP(s, e) : 1)
/* Set up to clean up after our imminent demise */
#define PREPARE_TO_DIE \
* another SSC or a regular ANYOF class. Can create false positives. */
SV* anded_cp_list;
+ U8 and_with_flags = (OP(and_with) == ANYOFH) ? 0 : ANYOF_FLAGS(and_with);
U8 anded_flags;
PERL_ARGS_ASSERT_SSC_AND;
* the code point inversion list and just the relevant flags */
if (is_ANYOF_SYNTHETIC(and_with)) {
anded_cp_list = ((regnode_ssc *)and_with)->invlist;
- anded_flags = ANYOF_FLAGS(and_with);
+ anded_flags = and_with_flags;
/* XXX This is a kludge around what appears to be deficiencies in the
* optimizer. If we make S_ssc_anything() add in the WARN_SUPER flag,
else {
anded_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, and_with);
if (OP(and_with) == ANYOFD) {
- anded_flags = ANYOF_FLAGS(and_with) & ANYOF_COMMON_FLAGS;
+ anded_flags = and_with_flags & ANYOF_COMMON_FLAGS;
}
else {
- anded_flags = ANYOF_FLAGS(and_with)
+ anded_flags = and_with_flags
&( ANYOF_COMMON_FLAGS
|ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
|ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP);
- if (ANYOFL_UTF8_LOCALE_REQD(ANYOF_FLAGS(and_with))) {
+ if (ANYOFL_UTF8_LOCALE_REQD(and_with_flags)) {
anded_flags &=
ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
}
* <= (C1 & ~C2) | (P1 & ~P2)
* */
- if ((ANYOF_FLAGS(and_with) & ANYOF_INVERT)
+ if ((and_with_flags & ANYOF_INVERT)
&& ! is_ANYOF_SYNTHETIC(and_with))
{
unsigned int i;
/* If either P1 or P2 is empty, the intersection will be also; can skip
* the loop */
- if (! (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL)) {
+ if (! (and_with_flags & ANYOF_MATCHES_POSIXL)) {
ANYOF_POSIXL_ZERO(ssc);
}
else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
else {
ssc->invlist = anded_cp_list;
ANYOF_POSIXL_ZERO(ssc);
- if (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL) {
+ if (and_with_flags & ANYOF_MATCHES_POSIXL) {
ANYOF_POSIXL_OR((regnode_charclass_posixl*) and_with, ssc);
}
}
}
else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)
- || (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL))
+ || (and_with_flags & ANYOF_MATCHES_POSIXL))
{
/* One or the other of P1, P2 is non-empty. */
- if (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL) {
+ if (and_with_flags & ANYOF_MATCHES_POSIXL) {
ANYOF_POSIXL_AND((regnode_charclass_posixl*) and_with, ssc);
}
ssc_union(ssc, anded_cp_list, FALSE);
SV* ored_cp_list;
U8 ored_flags;
+ U8 or_with_flags = (OP(or_with) == ANYOFH) ? 0 : ANYOF_FLAGS(or_with);
PERL_ARGS_ASSERT_SSC_OR;
* the code point inversion list and just the relevant flags */
if (is_ANYOF_SYNTHETIC(or_with)) {
ored_cp_list = ((regnode_ssc*) or_with)->invlist;
- ored_flags = ANYOF_FLAGS(or_with);
+ ored_flags = or_with_flags;
}
else {
ored_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, or_with);
- ored_flags = ANYOF_FLAGS(or_with) & ANYOF_COMMON_FLAGS;
+ ored_flags = or_with_flags & ANYOF_COMMON_FLAGS;
if (OP(or_with) != ANYOFD) {
ored_flags
- |= ANYOF_FLAGS(or_with)
+ |= or_with_flags
& ( ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
|ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP);
- if (ANYOFL_UTF8_LOCALE_REQD(ANYOF_FLAGS(or_with))) {
+ if (ANYOFL_UTF8_LOCALE_REQD(or_with_flags)) {
ored_flags |=
ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
}
* (which results in actually simpler code than the non-inverted case)
* */
- if ((ANYOF_FLAGS(or_with) & ANYOF_INVERT)
+ if ((or_with_flags & ANYOF_INVERT)
&& ! is_ANYOF_SYNTHETIC(or_with))
{
/* We ignore P2, leaving P1 going forward */
} /* else Not inverted */
- else if (ANYOF_FLAGS(or_with) & ANYOF_MATCHES_POSIXL) {
+ else if (or_with_flags & ANYOF_MATCHES_POSIXL) {
ANYOF_POSIXL_OR((regnode_charclass_posixl*)or_with, ssc);
if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
unsigned int i;
trie_words = newAV();
});
- re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
+ re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, GV_ADD);
assert(re_trie_maxbuff);
if (!SvIOK(re_trie_maxbuff)) {
sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
STRLEN l;
const char * const s = SvPV_const(data->last_found, l);
SSize_t old = b - data->last_start_min;
+ assert(old >= 0);
if (UTF)
- old = utf8_hop((U8*)s, old) - (U8*)s;
+ old = utf8_hop_forward((U8*)s, old,
+ (U8 *) SvEND(data->last_found))
+ - (U8*)s;
l -= old;
/* Get the added string: */
last_str = newSVpvn_utf8(s + old, l, UTF);
last, &data_fake, stopparen,
recursed_depth, NULL, f, depth+1);
if (scan->flags) {
- if (deltanext) {
- FAIL("Variable length lookbehind not implemented");
- }
- else if (minnext > (I32)U8_MAX) {
+ if ( deltanext < 0
+ || deltanext > (I32) U8_MAX
+ || minnext > (I32)U8_MAX
+ || minnext + deltanext > (I32)U8_MAX)
+ {
FAIL2("Lookbehind longer than %" UVuf " not implemented",
(UV)U8_MAX);
}
- scan->flags = (U8)minnext;
+
+ /* The 'next_off' field has been repurposed to count the
+ * additional starting positions to try beyond the initial
+ * one. (This leaves it at 0 for non-variable length
+ * matches to avoid breakage for those not using this
+ * extension) */
+ if (deltanext) {
+ scan->next_off = deltanext;
+ ckWARNexperimental(RExC_parse,
+ WARN_EXPERIMENTAL__VLB,
+ "Variable length lookbehind is experimental");
+ }
+ scan->flags = (U8)minnext + deltanext;
}
if (data) {
if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
stopparen, recursed_depth, NULL,
f, depth+1);
if (scan->flags) {
- if (deltanext) {
- FAIL("Variable length lookbehind not implemented");
- }
- else if (*minnextp > (I32)U8_MAX) {
+ assert(0); /* This code has never been tested since this
+ is normally not compiled */
+ if ( deltanext < 0
+ || deltanext > (I32) U8_MAX
+ || *minnextp > (I32)U8_MAX
+ || *minnextp + deltanext > (I32)U8_MAX)
+ {
FAIL2("Lookbehind longer than %" UVuf " not implemented",
(UV)U8_MAX);
}
- scan->flags = (U8)*minnextp;
+
+ if (deltanext) {
+ scan->next_off = deltanext;
+ }
+ scan->flags = (U8)*minnextp + deltanext;
}
*minnextp += min;
RExC_naughty = 0;
RExC_npar = 1;
+ RExC_parens_buf_size = 0;
RExC_emit_start = RExC_rxi->program;
pRExC_state->code_index = 0;
return;
default:
fail_modifiers:
- RExC_parse += SKIP_IF_CHAR(RExC_parse);
+ RExC_parse += SKIP_IF_CHAR(RExC_parse, RExC_end);
/* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
vFAIL2utf8f("Sequence (%" UTF8f "...) not recognized",
UTF8fARG(UTF, RExC_parse-seqstart, seqstart));
I32 freeze_paren = 0;
I32 after_freeze = 0;
I32 num; /* numeric backreferences */
+ SV * max_open; /* Max number of unclosed parens */
char * parse_start = RExC_parse; /* MJD */
char * const oregcomp_parse = RExC_parse;
PERL_ARGS_ASSERT_REG;
DEBUG_PARSE("reg ");
+
+ max_open = get_sv(RE_COMPILE_RECURSION_LIMIT, GV_ADD);
+ assert(max_open);
+ if (!SvIOK(max_open)) {
+ sv_setiv(max_open, RE_COMPILE_RECURSION_INIT);
+ }
+ if (depth > 4 * SvIV(max_open)) { /* We increase depth by 4 for each open
+ paren */
+ vFAIL("Too many nested open parens");
+ }
+
*flagp = 0; /* Tentatively. */
/* Having this true makes it feasible to have a lot fewer tests for the
} /* End of switch */
if ( ! op ) {
- RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
+ RExC_parse += UTF
+ ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
+ : 1;
if (has_upper || verb_len == 0) {
vFAIL2utf8f(
"Unknown verb pattern '%" UTF8f "'",
return handle_named_backref(pRExC_state, flagp,
parse_start, ')');
}
- RExC_parse += SKIP_IF_CHAR(RExC_parse);
+ RExC_parse += SKIP_IF_CHAR(RExC_parse, RExC_end);
/* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
vFAIL3("Sequence (%.*s...) not recognized",
RExC_parse-seqstart, seqstart);
case '?': /* (??...) */
is_logical = 1;
if (*RExC_parse != '{') {
- RExC_parse += SKIP_IF_CHAR(RExC_parse);
+ RExC_parse += SKIP_IF_CHAR(RExC_parse, RExC_end);
/* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
vFAIL2utf8f(
"Sequence (%" UTF8f "...) not recognized",
insert_if_check_paren:
if (UCHARAT(RExC_parse) != ')') {
- RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
+ RExC_parse += UTF
+ ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
+ : 1;
vFAIL("Switch condition not recognized");
}
nextchar(pRExC_state);
#endif
return ret;
}
- RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
+ RExC_parse += UTF
+ ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
+ : 1;
vFAIL("Unknown switch condition (?(...))");
}
case '[': /* (?[ ... ]) */
RExC_parse--; /* for vFAIL to print correctly */
vFAIL("Sequence (? incomplete");
break;
+
+ case ')':
+ if (RExC_strict) { /* [perl #132851] */
+ ckWARNreg(RExC_parse, "Empty (?) without any modifiers");
+ }
+ /* FALLTHROUGH */
default: /* e.g., (?i) */
RExC_parse = (char *) seqstart + 1;
parse_flags:
if (! ALL_PARENS_COUNTED) {
/* If we are in our first pass through (and maybe only pass),
* we need to allocate memory for the capturing parentheses
- * data structures. Since we start at npar=1, when it reaches
- * 2, for the first time it has something to put in it. Above
- * 2 means we extend what we already have */
- if (RExC_npar == 2) {
+ * data structures.
+ */
+
+ if (!RExC_parens_buf_size) {
+ /* first guess at number of parens we might encounter */
+ RExC_parens_buf_size = 10;
+
/* setup RExC_open_parens, which holds the address of each
* OPEN tag, and to make things simpler for the 0 index the
* start of the program - this is used later for offsets */
- Newxz(RExC_open_parens, RExC_npar, regnode_offset);
+ Newxz(RExC_open_parens, RExC_parens_buf_size,
+ regnode_offset);
RExC_open_parens[0] = 1; /* +1 for REG_MAGIC */
/* setup RExC_close_parens, which holds the address of each
* CLOSE tag, and to make things simpler for the 0 index
* the end of the program - this is used later for offsets
* */
- Newxz(RExC_close_parens, RExC_npar, regnode_offset);
+ Newxz(RExC_close_parens, RExC_parens_buf_size,
+ regnode_offset);
/* we dont know where end op starts yet, so we dont need to
* set RExC_close_parens[0] like we do RExC_open_parens[0]
* above */
}
- else {
- Renew(RExC_open_parens, RExC_npar, regnode_offset);
- Zero(RExC_open_parens + RExC_npar - 1, 1, regnode_offset);
+ else if (RExC_npar > RExC_parens_buf_size) {
+ I32 old_size = RExC_parens_buf_size;
+
+ RExC_parens_buf_size *= 2;
+
+ Renew(RExC_open_parens, RExC_parens_buf_size,
+ regnode_offset);
+ Zero(RExC_open_parens + old_size,
+ RExC_parens_buf_size - old_size, regnode_offset);
- Renew(RExC_close_parens, RExC_npar, regnode_offset);
- Zero(RExC_close_parens + RExC_npar - 1, 1, regnode_offset);
+ Renew(RExC_close_parens, RExC_parens_buf_size,
+ regnode_offset);
+ Zero(RExC_close_parens + old_size,
+ RExC_parens_buf_size - old_size, regnode_offset);
}
}
RETURN_FAIL_ON_RESTART(flags, flagp);
FAIL2("panic: regbranch returned failure, flags=%#" UVxf, (UV) flags);
}
- REGTAIL(pRExC_state, lastbr, br); /* BRANCH -> BRANCH. */
+ if (! REGTAIL(pRExC_state, lastbr, br)) { /* BRANCH -> BRANCH. */
+ REQUIRE_BRANCHJ(flagp, 0);
+ }
lastbr = br;
*flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
}
(IV)(ender - lastbr)
);
);
- REGTAIL(pRExC_state, lastbr, ender);
+ if (! REGTAIL(pRExC_state, lastbr, ender)) {
+ REQUIRE_BRANCHJ(flagp, 0);
+ }
if (have_branch) {
char is_nothing= 1;
for (br = REGNODE_p(ret); br; br = regnext(br)) {
const U8 op = PL_regkind[OP(br)];
if (op == BRANCH) {
- REGTAIL_STUDY(pRExC_state,
- REGNODE_OFFSET(NEXTOPER(br)),
- ender);
+ if (! REGTAIL_STUDY(pRExC_state,
+ REGNODE_OFFSET(NEXTOPER(br)),
+ ender))
+ {
+ REQUIRE_BRANCHJ(flagp, 0);
+ }
if ( OP(NEXTOPER(br)) != NOTHING
|| regnext(NEXTOPER(br)) != REGNODE_p(ender))
is_nothing= 0;
Set_Node_Cur_Length(REGNODE_p(ret), parse_start);
Set_Node_Offset(REGNODE_p(ret), parse_start + 1);
FLAGS(REGNODE_p(ret)) = flag;
- REGTAIL_STUDY(pRExC_state, ret, reg_node(pRExC_state, TAIL));
+ if (! REGTAIL_STUDY(pRExC_state, ret, reg_node(pRExC_state, TAIL)))
+ {
+ REQUIRE_BRANCHJ(flagp, 0);
+ }
}
}
/* FIXME adding one for every branch after the first is probably
* excessive now we have TRIE support. (hv) */
MARK_NAUGHTY(1);
- REGTAIL(pRExC_state, chain, latest);
+ if (! REGTAIL(pRExC_state, chain, latest)) {
+ /* XXX We could just redo this branch, but figuring out what
+ * bookkeeping needs to be reset is a pain, and it's likely
+ * that other branches that goto END will also be too large */
+ REQUIRE_BRANCHJ(flagp, 0);
+ }
}
chain = latest;
- if ( chain > (SSize_t) BRANCH_MAX_OFFSET
- && ! RExC_use_BRANCHJ)
- {
- /* XXX We could just redo this branch, but figuring out what
- * bookkeeping needs to be reset is a pain */
- REQUIRE_BRANCHJ(flagp, 0);
- }
c++;
}
if (chain == 0) { /* Loop ran zero times. */
else {
/* Point to the first byte of the final character */
- s = (char *) utf8_hop((U8 *) s, -1);
+ s = (char *) utf8_hop_back((U8 *) s, -1, (U8 *) s0);
while (s >= s0) { /* Search backwards until find
a non-problematic char */
RExC_parse = RExC_end;
}
else if (RExC_parse != save_parse) {
- RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
+ RExC_parse += (UTF)
+ ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
+ : 1;
}
vFAIL("Expecting '(?flags:(?[...'");
}
} /* The \p isn't immediately followed by a '{' */
else if (! isALPHA(*RExC_parse)) {
- RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
+ RExC_parse += (UTF)
+ ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
+ : 1;
vFAIL2("Character following \\%c must be '{' or a "
"single-character Unicode property name",
(U8) value);
SV * prop_definition = parse_uniprop_string(
name, n, UTF, FOLD,
FALSE, /* This is compile-time */
+
+ /* We can't defer this defn when
+ * the full result is required in
+ * this call */
+ ! cBOOL(ret_invlist),
+
&user_defined,
msg,
0 /* Base level */
RExC_parse += numlen;
if (numlen != 3) {
if (strict) {
- RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
+ RExC_parse += (UTF)
+ ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
+ : 1;
vFAIL("Need exactly 3 octal digits");
}
else if ( numlen < 3 /* like \08, \178 */
* bitmap, optimize to indicate that */
if ( start[0] >= NUM_ANYOF_CODE_POINTS
&& ! LOC
- && ! upper_latin1_only_utf8_matches)
+ && ! upper_latin1_only_utf8_matches
+ && anyof_flags == 0)
{
+ UV highest_cp = invlist_highest(cp_list);
+
+ /* If the lowest and highest code point in the class have the same
+ * UTF-8 first byte, then all do, and we can store that byte for
+ * regexec.c to use so that it can more quickly scan the target
+ * string for potential matches for this class. We co-opt the the
+ * flags field for this. Zero means, they don't have the same
+ * first byte. We do accept here very large code points (for
+ * future use), but don't bother with this optimization for them,
+ * as it would cause other complications */
+ if (highest_cp > IV_MAX) {
+ anyof_flags = 0;
+ }
+ else {
+ U8 low_utf8[UTF8_MAXBYTES+1];
+ U8 high_utf8[UTF8_MAXBYTES+1];
+
+ (void) uvchr_to_utf8(low_utf8, start[0]);
+ (void) uvchr_to_utf8(high_utf8, invlist_highest(cp_list));
+
+ anyof_flags = (low_utf8[0] == high_utf8[0])
+ ? low_utf8[0]
+ : 0;
+ }
+
op = ANYOFH;
}
} /* End of seeing if can optimize it into a different node */
stored here for just
this occasion */
TRUE, /* run time */
+ FALSE, /* This call must find the defn */
si, /* The property definition */
&user_defined,
msg,
|| UTF8_IS_INVARIANT(*RExC_parse)
|| UTF8_IS_START(*RExC_parse));
- RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
+ RExC_parse += (UTF)
+ ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
+ : 1;
skip_to_be_ignored_text(pRExC_state, &RExC_parse,
FALSE /* Don't force /x */ );
}
/*
-- regtail - set the next-pointer at the end of a node chain of p to val.
+- regtail - set the next-pointer at the end of a node chain of p to val. If
+ that value won't fit in the space available, instead returns FALSE.
+ (Except asserts if we can't fit in the largest space the regex
+ engine is designed for.)
- SEE ALSO: regtail_study
*/
-STATIC void
+STATIC bool
S_regtail(pTHX_ RExC_state_t * pRExC_state,
const regnode_offset p,
const regnode_offset val,
}
if (reg_off_by_arg[OP(REGNODE_p(scan))]) {
+ assert(val - scan <= U32_MAX);
ARG_SET(REGNODE_p(scan), val - scan);
}
else {
+ if (val - scan > U16_MAX) {
+ /* Since not all callers check the return value, populate this with
+ * something that won't loop and will likely lead to a crash if
+ * execution continues */
+ NEXT_OFF(REGNODE_p(scan)) = U16_MAX;
+ return FALSE;
+ }
NEXT_OFF(REGNODE_p(scan)) = val - scan;
}
+
+ return TRUE;
}
#ifdef DEBUGGING
Currently only used when in DEBUG mode. The macro REGTAIL_STUDY() is used
to control which is which.
+This used to return a value that was ignored. It was a problem that it is
+#ifdef'd to be another function that didn't return a value. khw has changed it
+so both currently return a pass/fail return.
+
*/
/* TODO: All four parms should be const */
-STATIC U8
+STATIC bool
S_regtail_study(pTHX_ RExC_state_t *pRExC_state, regnode_offset p,
const regnode_offset val, U32 depth)
{
bool unfolded_multi_char; /* Unexamined in this routine */
if (join_exact(pRExC_state, scan, &min,
&unfolded_multi_char, 1, REGNODE_p(val), depth+1))
- return EXACT;
+ return TRUE; /* Was return EXACT */
}
#endif
if ( exact ) {
);
});
if (reg_off_by_arg[OP(REGNODE_p(scan))]) {
+ assert(val - scan <= U32_MAX);
ARG_SET(REGNODE_p(scan), val - scan);
}
else {
+ if (val - scan > U16_MAX) {
+ NEXT_OFF(REGNODE_p(scan)) = U16_MAX;
+ return FALSE;
+ }
NEXT_OFF(REGNODE_p(scan)) = val - scan;
}
- return exact;
+ return TRUE; /* Was 'return exact' */
}
#endif
/* 2: embedded, otherwise 1 */
Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags);
else if (k == ANYOF) {
- const U8 flags = ANYOF_FLAGS(o);
+ const U8 flags = (OP(o) == ANYOFH) ? 0 : ANYOF_FLAGS(o);
bool do_sep = FALSE; /* Do we need to separate various components of
the output? */
/* Set if there is still an unresolved user-defined property */
/* And finally the matching, closing ']' */
Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
+ if (OP(o) == ANYOFH && FLAGS(o) != 0) {
+ Perl_sv_catpvf(aTHX_ sv, " (First UTF-8 byte=\\x%02x)", FLAGS(o));
+ }
+
+
SvREFCNT_dec(unresolved);
}
else if (k == ANYOFM) {
assert(FLAGS(o) < C_ARRAY_LENGTH(bounds));
sv_catpv(sv, bounds[FLAGS(o)]);
}
- else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
- Perl_sv_catpvf(aTHX_ sv, "[%d]", -(o->flags));
+ else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH)) {
+ Perl_sv_catpvf(aTHX_ sv, "[%d", -(o->flags));
+ if (o->next_off) {
+ Perl_sv_catpvf(aTHX_ sv, "..-%d", o->flags - o->next_off);
+ }
+ Perl_sv_catpvf(aTHX_ sv, "]");
+ }
else if (OP(o) == SBOL)
Perl_sv_catpvf(aTHX_ sv, " /%s/", o->flags ? "\\A" : "^");
if (!dsv)
dsv = (REGEXP*) newSV_type(SVt_REGEXP);
else {
+ assert(SvTYPE(dsv) == SVt_REGEXP || (SvTYPE(dsv) == SVt_PVLV));
+
+ /* our only valid caller, sv_setsv_flags(), should have done
+ * a SV_CHECK_THINKFIRST_COW_DROP() by now */
+ assert(!SvOOK(dsv));
+ assert(!SvIsCOW(dsv));
+ assert(!SvROK(dsv));
+
+ if (SvPVX_const(dsv)) {
+ if (SvLEN(dsv))
+ Safefree(SvPVX(dsv));
+ SvPVX(dsv) = NULL;
+ }
+ SvLEN_set(dsv, 0);
+ SvCUR_set(dsv, 0);
SvOK_off((SV *)dsv);
+
if (islv) {
/* For PVLVs, the head (sv_any) points to an XPVLV, while
* the LV's xpvlenu_rx will point to a regexp body, which
2: something we no longer hold a reference on
so we need to copy it locally. */
RX_WRAPPED(dstr) = SAVEPVN(RX_WRAPPED_const(sstr), SvCUR(sstr)+1);
+ /* set malloced length to a non-zero value so it will be freed
+ * (otherwise in combination with SVf_FAKE it looks like an alien
+ * buffer). It doesn't have to be the actual malloced size, since it
+ * should never be grown */
+ SvLEN_set(dstr, SvCUR(sstr)+1);
ret->mother_re = NULL;
}
#endif /* PERL_IN_XSUB_RE */
const bool is_utf8, /* ? Is 'name' encoded in UTF-8 */
const bool to_fold, /* ? Is this under /i */
const bool runtime, /* ? Are we in compile- or run-time */
+ const bool deferrable, /* Is it ok for this property's full definition
+ to be deferred until later? */
SV* contents, /* The property's definition */
bool *user_defined_ptr, /* This will be set TRUE as we wouldn't be
getting called unless this is thought to be
Perl_sv_catpvf(aTHX_ msg, "%" UTF8f,
UTF8fARG(is_contents_utf8, s - s0, s0));
sv_catpvs(msg, "\"");
- goto return_msg;
+ goto return_failure;
}
/* Accumulate this digit into the value */
Perl_sv_catpvf(aTHX_ msg, "%" UTF8f,
UTF8fARG(is_contents_utf8, s - s0, s0));
sv_catpvs(msg, "\"");
- goto return_msg;
+ goto return_failure;
}
max = (max << 4) + READ_XDIGIT(s);
Perl_sv_catpvf(aTHX_ msg, "%" UTF8f,
UTF8fARG(is_contents_utf8, s - s0, s0));
sv_catpvs(msg, "\"");
- goto return_msg;
+ goto return_failure;
}
#if 0 /* See explanation at definition above of get_extended_utf8_msg() */
this_definition = parse_uniprop_string(s0, s - s0,
is_utf8, to_fold, runtime,
+ deferrable,
user_defined_ptr, msg,
(name_len == 0)
? level /* Don't increase level
: level + 1
);
if (this_definition == NULL) {
- goto return_msg; /* 'msg' should have had the reason appended to
- it by the above call */
+ goto return_failure; /* 'msg' should have had the reason
+ appended to it by the above call */
}
if (! is_invlist(this_definition)) { /* Unknown at this time */
}
/* Otherwise, add some explanatory text, but we will return success */
+ goto return_msg;
+
+ return_failure:
+ running_definition = NULL;
return_msg:
const bool is_utf8, /* ? Is 'name' encoded in UTF-8 */
const bool to_fold, /* ? Is this under /i */
const bool runtime, /* TRUE if this is being called at run time */
+ const bool deferrable, /* TRUE if it's ok for the definition to not be
+ known at this call */
bool *user_defined_ptr, /* Upon return from this function it will be
set to TRUE if any component is a
user-defined property */
is_utf8,
to_fold,
runtime,
+ deferrable,
user_defined_ptr,
msg,
level + 1);
* handle it */
prop_definition = handle_user_defined_property(name, name_len,
is_utf8, to_fold, runtime,
+ deferrable,
POPs, user_defined_ptr,
msg,
level);
}
- /* Here, we have the results of the expansion. Replace the
- * placeholder with them. We need exclusive access to the hash,
- * and we can't let anyone else in, between when we delete the
- * placeholder and add the permanent entry */
+ /* Here, we have the results of the expansion. Delete the
+ * placeholder, and if the definition is now known, replace it with
+ * that definition. We need exclusive access to the hash, and we
+ * can't let anyone else in, between when we delete the placeholder
+ * and add the permanent entry */
USER_PROP_MUTEX_LOCK;
S_delete_recursion_entry(aTHX_ SvPVX(fq_name));
* compile time, it might just be that the subroutine for that
* property hasn't been encountered yet, but at runtime, it's
* an error to try to use an undefined one */
- if (runtime) {
+ if (! deferrable) {
if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
sv_catpvs(msg, "Unknown user-defined property name");
goto append_name_to_msg;
0, /* Not UTF-8 */
0, /* Not folded */
runtime,
+ deferrable,
pu_definition,
&dummy,
msg,
https://perl5.git.perl.org / perl5.git / blobdiff