* with the POSIX routines of the same names.
*/
+#ifdef IN_XSUB_RE
+/* We *really* need to overwrite these symbols: */
+# define Perl_pregcomp my_regcomp
+# define Perl_regdump my_regdump
+# define Perl_regprop my_regprop
+/* *These* symbols are masked to allow static link. */
+# define Perl_pregfree my_regfree
+# define Perl_regnext my_regnext
+#endif
+
/*SUPPRESS 112*/
/*
* pregcomp and pregexec -- regsub and regerror are not used in perl
*/
#include "EXTERN.h"
#include "perl.h"
-#include "INTERN.h"
+
+#ifndef IN_XSUB_RE
+# include "INTERN.h"
+#endif
#define REG_COMP_C
#include "regcomp.h"
* Flags to be passed up and down.
*/
#define WORST 0 /* Worst case. */
-#define HASWIDTH 0x1 /* Known never to match null string. */
+#define HASWIDTH 0x1 /* Known to match non-null strings. */
#define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
#define SPSTART 0x4 /* Starts with * or +. */
#define TRYAGAIN 0x8 /* Weeded out a declaration. */
* Forward declarations for pregcomp()'s friends.
*/
-static char* regwhite _((char *, char *));
#ifndef PERL_OBJECT
static regnode *reg _((I32, I32 *));
static regnode *reganode _((U8, U32));
static regnode *regpiece _((I32 *));
static void reginsert _((U8, regnode *));
static void regoptail _((regnode *, regnode *));
-static void regset _((char *, I32));
static void regtail _((regnode *, regnode *));
+static char* regwhite _((char *, char *));
static char* nextchar _((void));
-
static void re_croak2 _((const char* pat1,const char* pat2,...)) __attribute__((noreturn));
#endif
#define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
#define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
-#define SF_FIX_SHIFT_EOL (+2)
-#define SF_FL_SHIFT_EOL (+4)
+#ifdef NO_UNARY_PLUS
+# define SF_FIX_SHIFT_EOL (0+2)
+# define SF_FL_SHIFT_EOL (0+4)
+#else
+# define SF_FIX_SHIFT_EOL (+2)
+# define SF_FL_SHIFT_EOL (+4)
+#endif
#define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
#define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
/* deltap: Write maxlen-minlen here. */
/* last: Stop before this one. */
{
+ dTHR;
I32 min = 0, pars = 0, code;
regnode *scan = *scanp, *next;
I32 delta = 0;
/* Allow dumping */
n = scan + (*OPERAND(scan) + 2 - 1)/sizeof(regnode) + 2;
while (n <= stop) {
+ /* Purify reports a benign UMR here sometimes, because we
+ * don't initialize the OP() slot of a node when that node
+ * is occupied by just the trailing null of the string in
+ * an EXACT node */
if (regkind[(U8)OP(n)] != NOTHING || OP(n) == NOTHING) {
OP(n) = OPTIMIZED;
NEXT_OFF(n) = 0;
if (!scan) /* It was not CURLYX, but CURLY. */
scan = next;
if (dowarn && (minnext + deltanext == 0)
- && !(data->flags & (SF_HAS_PAR|SF_IN_PAR)))
+ && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
+ && maxcount <= 10000) /* Complement check for big count */
warn("Strange *+?{} on zero-length expression");
min += minnext * mincount;
is_inf |= (maxcount == REG_INFTY && (minnext + deltanext) > 0
delta += (minnext + deltanext) * maxcount - minnext * mincount;
/* Try powerful optimization CURLYX => CURLYN. */
-#ifdef REGALIGN_STRUCT
if ( OP(oscan) == CURLYX && data
&& data->flags & SF_IN_PAR
&& !(data->flags & SF_HAS_EVAL)
NEXT_OFF(nxt+ 1) = 0; /* just for consistancy. */
#endif
}
-#endif
nogo:
/* Try optimization CURLYX => CURLYM. */
if ( OP(oscan) == CURLYX && data
-#ifdef REGALIGN_STRUCT
&& !(data->flags & SF_HAS_PAR)
-#else
- && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
-#endif
&& !(data->flags & SF_HAS_EVAL)
&& !deltanext ) {
/* XXXX How to optimize if data == 0? */
&& (OP(nxt2) != WHILEM))
nxt = nxt2;
OP(nxt2) = SUCCEED; /* Whas WHILEM */
-#ifdef REGALIGN_STRUCT
/* Need to optimize away parenths. */
if (data->flags & SF_IN_PAR) {
/* Set the parenth number. */
study_chunk(&nxt1, &deltanext, nxt, NULL, 0);
} else
oscan->flags = 0;
-#endif
}
if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
pars++;
if (data && (fl & SF_HAS_EVAL))
data->flags |= SF_HAS_EVAL;
optimize_curly_tail:
-#ifdef REGALIGN
if (OP(oscan) != CURLYX) {
while (regkind[(U8)OP(next = regnext(oscan))] == NOTHING
&& NEXT_OFF(next))
NEXT_OFF(oscan) += NEXT_OFF(next);
}
-#endif
continue;
default: /* REF only? */
if (flags & SCF_DO_SUBSTR) {
} else if (OP(scan) == OPEN) {
pars++;
} else if (OP(scan) == CLOSE && ARG(scan) == is_par) {
-#ifdef REGALIGN_STRUCT
next = regnext(scan);
if ( next && (OP(next) != WHILEM) && next < last)
-#endif
is_par = 0; /* Disable optimization */
} else if (OP(scan) == EVAL) {
if (data)
STATIC I32
add_data(I32 n, char *s)
{
- if (rx->data) {
- Renewc(rx->data,
- sizeof(*rx->data) + sizeof(void*) * (rx->data->count + n - 1),
+ dTHR;
+ if (regcomp_rx->data) {
+ Renewc(regcomp_rx->data,
+ sizeof(*regcomp_rx->data) + sizeof(void*) * (regcomp_rx->data->count + n - 1),
char, struct reg_data);
- Renew(rx->data->what, rx->data->count + n, U8);
- rx->data->count += n;
+ Renew(regcomp_rx->data->what, regcomp_rx->data->count + n, U8);
+ regcomp_rx->data->count += n;
} else {
- Newc(1207, rx->data, sizeof(*rx->data) + sizeof(void*) * (n - 1),
+ Newc(1207, regcomp_rx->data, sizeof(*regcomp_rx->data) + sizeof(void*) * (n - 1),
char, struct reg_data);
- New(1208, rx->data->what, n, U8);
- rx->data->count = n;
+ New(1208, regcomp_rx->data->what, n, U8);
+ regcomp_rx->data->count = n;
}
- Copy(s, rx->data->what + rx->data->count - n, n, U8);
- return rx->data->count - n;
+ Copy(s, regcomp_rx->data->what + regcomp_rx->data->count - n, n, U8);
+ return regcomp_rx->data->count - n;
}
/*
regexp *
pregcomp(char *exp, char *xend, PMOP *pm)
{
+ dTHR;
register regexp *r;
regnode *scan;
SV **longest;
regseen = 0;
seen_zerolen = *exp == '^' ? -1 : 0;
+ seen_evals = 0;
extralen = 0;
/* First pass: determine size, legality. */
- regparse = exp;
+ regcomp_parse = exp;
regxend = xend;
regnaughty = 0;
regnpar = 1;
/* Small enough for pointer-storage convention?
If extralen==0, this means that we will not need long jumps. */
-#ifndef REGALIGN_STRUCT
- if (regsize >= 0x10000L && extralen)
- FAIL("regexp too big");
-#else
if (regsize >= 0x10000L && extralen)
regsize += extralen;
else
extralen = 0;
-#endif
/* Allocate space and initialize. */
Newc(1001, r, sizeof(regexp) + (unsigned)regsize * sizeof(regnode),
r->prelen = xend - exp;
r->precomp = regprecomp;
r->subbeg = r->subbase = NULL;
- rx = r;
+ r->nparens = regnpar - 1; /* set early to validate backrefs */
+ regcomp_rx = r;
/* Second pass: emit code. */
- regparse = exp;
+ regcomp_parse = exp;
regxend = xend;
regnaughty = 0;
regnpar = 1;
regcode = r->program;
+ /* Store the count of eval-groups for security checks: */
+ regcode->next_off = ((seen_evals > U16_MAX) ? U16_MAX : seen_evals);
regc((U8)MAGIC, (char*) regcode++);
r->data = 0;
if (reg(0, &flags) == NULL)
return(NULL);
/* Dig out information for optimizations. */
+ r->reganch = pm->op_pmflags & PMf_COMPILETIME;
pm->op_pmflags = regflags;
- r->reganch = 0;
r->regstclass = NULL;
r->naughty = regnaughty >= 10; /* Probably an expensive pattern. */
scan = r->program + 1; /* First BRANCH. */
r->check_substr = r->anchored_substr = r->float_substr = Nullsv;
}
- r->nparens = regnpar - 1;
r->minlen = minlen;
if (regseen & REG_SEEN_GPOS)
r->reganch |= ROPT_GPOS_SEEN;
if (regseen & REG_SEEN_LOOKBEHIND)
r->reganch |= ROPT_LOOKBEHIND_SEEN;
+ if (regseen & REG_SEEN_EVAL)
+ r->reganch |= ROPT_EVAL_SEEN;
Newz(1002, r->startp, regnpar, char*);
Newz(1002, r->endp, regnpar, char*);
DEBUG_r(regdump(r));
reg(I32 paren, I32 *flagp)
/* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
{
+ dTHR;
register regnode *ret; /* Will be the head of the group. */
register regnode *br;
register regnode *lastbr;
I32 flags, oregflags = regflags, have_branch = 0, open = 0;
char c;
- *flagp = HASWIDTH; /* Tentatively. */
+ *flagp = 0; /* Tentatively. */
/* Make an OPEN node, if parenthesized. */
if (paren) {
- if (*regparse == '?') {
- regparse++;
- paren = *regparse++;
+ if (*regcomp_parse == '?') {
+ U16 posflags = 0, negflags = 0;
+ U16 *flagsp = &posflags;
+
+ regcomp_parse++;
+ paren = *regcomp_parse++;
ret = NULL; /* For look-ahead/behind. */
switch (paren) {
case '<':
-#ifndef REGALIGN_STRUCT
- FAIL("lookbehind non-implemented without REGALIGN_STRUCT");
-#endif
regseen |= REG_SEEN_LOOKBEHIND;
- if (*regparse == '!')
+ if (*regcomp_parse == '!')
paren = ',';
- if (*regparse != '=' && *regparse != '!')
+ if (*regcomp_parse != '=' && *regcomp_parse != '!')
goto unknown;
- regparse++;
+ regcomp_parse++;
case '=':
case '!':
seen_zerolen++;
FAIL2("Sequence (?%c...) not implemented", (int)paren);
break;
case '#':
- while (*regparse && *regparse != ')')
- regparse++;
- if (*regparse != ')')
+ while (*regcomp_parse && *regcomp_parse != ')')
+ regcomp_parse++;
+ if (*regcomp_parse != ')')
FAIL("Sequence (?#... not terminated");
nextchar();
*flagp = TRYAGAIN;
dTHR;
I32 count = 1, n = 0;
char c;
- char *s = regparse;
+ char *s = regcomp_parse;
SV *sv;
OP_4tree *sop, *rop;
seen_zerolen++;
- while (count && (c = *regparse)) {
- if (c == '\\' && regparse[1])
- regparse++;
+ regseen |= REG_SEEN_EVAL;
+ while (count && (c = *regcomp_parse)) {
+ if (c == '\\' && regcomp_parse[1])
+ regcomp_parse++;
else if (c == '{')
count++;
else if (c == '}')
count--;
- regparse++;
+ regcomp_parse++;
}
- if (*regparse != ')')
+ if (*regcomp_parse != ')')
FAIL("Sequence (?{...}) not terminated or not {}-balanced");
if (!SIZE_ONLY) {
AV *av;
- if (regparse - 1 - s)
- sv = newSVpv(s, regparse - 1 - s);
+ if (regcomp_parse - 1 - s)
+ sv = newSVpv(s, regcomp_parse - 1 - s);
else
sv = newSVpv("", 0);
rop = sv_compile_2op(sv, &sop, "re", &av);
n = add_data(3, "nso");
- rx->data->data[n] = (void*)rop;
- rx->data->data[n+1] = (void*)av;
- rx->data->data[n+2] = (void*)sop;
+ regcomp_rx->data->data[n] = (void*)rop;
+ regcomp_rx->data->data[n+1] = (void*)av;
+ regcomp_rx->data->data[n+2] = (void*)sop;
SvREFCNT_dec(sv);
} else { /* First pass */
+ if (reginterp_cnt < ++seen_evals && curcop != &compiling)
+ /* No compiled RE interpolated, has runtime
+ components ===> unsafe. */
+ FAIL("Eval-group not allowed at runtime, use re 'eval'");
if (tainted)
FAIL("Eval-group in insecure regular expression");
}
}
case '(':
{
- if (regparse[0] == '?') {
- if (regparse[1] == '=' || regparse[1] == '!'
- || regparse[1] == '<'
- || regparse[1] == '{') { /* Lookahead or eval. */
+ if (regcomp_parse[0] == '?') {
+ if (regcomp_parse[1] == '=' || regcomp_parse[1] == '!'
+ || regcomp_parse[1] == '<'
+ || regcomp_parse[1] == '{') { /* Lookahead or eval. */
I32 flag;
ret = reg_node(LOGICAL);
regtail(ret, reg(1, &flag));
goto insert_if;
}
- } else if (regparse[0] >= '1' && regparse[0] <= '9' ) {
- parno = atoi(regparse++);
+ } else if (regcomp_parse[0] >= '1' && regcomp_parse[0] <= '9' ) {
+ parno = atoi(regcomp_parse++);
- while (isDIGIT(*regparse))
- regparse++;
+ while (isDIGIT(*regcomp_parse))
+ regcomp_parse++;
ret = reganode(GROUPP, parno);
if ((c = *nextchar()) != ')')
FAIL2("Switch (?(number%c not recognized", c);
regtail(ret, ender);
return ret;
} else {
- FAIL2("Unknown condition for (?(%.2s", regparse);
+ FAIL2("Unknown condition for (?(%.2s", regcomp_parse);
}
}
case 0:
FAIL("Sequence (? incomplete");
break;
default:
- --regparse;
- while (*regparse && strchr("iogcmsx", *regparse))
- pmflag(®flags, *regparse++);
+ --regcomp_parse;
+ parse_flags:
+ while (*regcomp_parse && strchr("iogcmsx", *regcomp_parse)) {
+ if (*regcomp_parse != 'o')
+ pmflag(flagsp, *regcomp_parse);
+ ++regcomp_parse;
+ }
+ if (*regcomp_parse == '-') {
+ flagsp = &negflags;
+ ++regcomp_parse;
+ goto parse_flags;
+ }
+ regflags |= posflags;
+ regflags &= ~negflags;
+ if (*regcomp_parse == ':') {
+ regcomp_parse++;
+ paren = ':';
+ break;
+ }
unknown:
- if (*regparse != ')')
- FAIL2("Sequence (?%c...) not recognized", *regparse);
+ if (*regcomp_parse != ')')
+ FAIL2("Sequence (?%c...) not recognized", *regcomp_parse);
nextchar();
*flagp = TRYAGAIN;
return NULL;
br = regbranch(&flags, 1);
if (br == NULL)
return(NULL);
- if (*regparse == '|') {
+ if (*regcomp_parse == '|') {
if (!SIZE_ONLY && extralen) {
reginsert(BRANCHJ, br);
} else
regtail(ret, br); /* OPEN -> first. */
} else if (paren != '?') /* Not Conditional */
ret = br;
- if (!(flags&HASWIDTH))
- *flagp &= ~HASWIDTH;
+ if (flags&HASWIDTH)
+ *flagp |= HASWIDTH;
*flagp |= flags&SPSTART;
lastbr = br;
- while (*regparse == '|') {
+ while (*regcomp_parse == '|') {
if (!SIZE_ONLY && extralen) {
ender = reganode(LONGJMP,0);
regtail(NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */
return(NULL);
regtail(lastbr, br); /* BRANCH -> BRANCH. */
lastbr = br;
- if (!(flags&HASWIDTH))
- *flagp &= ~HASWIDTH;
+ if (flags&HASWIDTH)
+ *flagp |= HASWIDTH;
*flagp |= flags&SPSTART;
}
ender = reganode(CLOSE, parno);
break;
case '<':
- case '>':
case ',':
case '=':
case '!':
- ender = reg_node(SUCCEED);
*flagp &= ~HASWIDTH;
+ /* FALL THROUGH */
+ case '>':
+ ender = reg_node(SUCCEED);
break;
case 0:
ender = reg_node(END);
if (paren == '>')
node = SUSPEND, flag = 0;
reginsert(node,ret);
-#ifdef REGALIGN_STRUCT
ret->flags = flag;
-#endif
regtail(ret, reg_node(TAIL));
}
}
/* Check for proper termination. */
- if (paren && (regparse >= regxend || *nextchar() != ')')) {
+ if (paren && (regcomp_parse >= regxend || *nextchar() != ')')) {
FAIL("unmatched () in regexp");
- } else if (!paren && regparse < regxend) {
- if (*regparse == ')') {
+ } else if (!paren && regcomp_parse < regxend) {
+ if (*regcomp_parse == ')') {
FAIL("unmatched () in regexp");
} else
FAIL("junk on end of regexp"); /* "Can't happen". */
STATIC regnode *
regbranch(I32 *flagp, I32 first)
{
+ dTHR;
register regnode *ret;
register regnode *chain = NULL;
register regnode *latest;
*flagp = WORST; /* Tentatively. */
- regparse--;
+ regcomp_parse--;
nextchar();
- while (regparse < regxend && *regparse != '|' && *regparse != ')') {
+ while (regcomp_parse < regxend && *regcomp_parse != '|' && *regcomp_parse != ')') {
flags &= ~TRYAGAIN;
latest = regpiece(&flags);
if (latest == NULL) {
STATIC regnode *
regpiece(I32 *flagp)
{
+ dTHR;
register regnode *ret;
register char op;
register char *next;
I32 flags;
- char *origparse = regparse;
+ char *origparse = regcomp_parse;
char *maxpos;
I32 min;
I32 max = REG_INFTY;
return(NULL);
}
- op = *regparse;
+ op = *regcomp_parse;
- if (op == '{' && regcurly(regparse)) {
- next = regparse + 1;
+ if (op == '{' && regcurly(regcomp_parse)) {
+ next = regcomp_parse + 1;
maxpos = Nullch;
while (isDIGIT(*next) || *next == ',') {
if (*next == ',') {
if (*next == '}') { /* got one */
if (!maxpos)
maxpos = next;
- regparse++;
- min = atoi(regparse);
+ regcomp_parse++;
+ min = atoi(regcomp_parse);
if (*maxpos == ',')
maxpos++;
else
- maxpos = regparse;
+ maxpos = regcomp_parse;
max = atoi(maxpos);
if (!max && *maxpos != '0')
max = REG_INFTY; /* meaning "infinity" */
else if (max >= REG_INFTY)
FAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
- regparse = next;
+ regcomp_parse = next;
nextchar();
do_curly:
if (SIZE_ONLY)
extralen += 3;
}
-#ifdef REGALIGN_STRUCT
ret->flags = 0;
-#endif
if (min > 0)
- *flagp = (WORST|HASWIDTH);
+ *flagp = WORST;
+ if (max > 0)
+ *flagp |= HASWIDTH;
if (max && max < min)
FAIL("Can't do {n,m} with n > m");
if (!SIZE_ONLY) {
nextchar();
- *flagp = (op != '+') ? (WORST|SPSTART) : (WORST|HASWIDTH);
+ *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
if (op == '*' && (flags&SIMPLE)) {
reginsert(STAR, ret);
-#ifdef REGALIGN_STRUCT
ret->flags = 0;
-#endif
regnaughty += 4;
}
else if (op == '*') {
goto do_curly;
} else if (op == '+' && (flags&SIMPLE)) {
reginsert(PLUS, ret);
-#ifdef REGALIGN_STRUCT
ret->flags = 0;
-#endif
regnaughty += 3;
}
else if (op == '+') {
nest_check:
if (dowarn && !SIZE_ONLY && !(flags&HASWIDTH) && max > 10000) {
warn("%.*s matches null string many times",
- regparse - origparse, origparse);
+ regcomp_parse - origparse, origparse);
}
- if (*regparse == '?') {
+ if (*regcomp_parse == '?') {
nextchar();
reginsert(MINMOD, ret);
-#ifdef REGALIGN
regtail(ret, ret + NODE_STEP_REGNODE);
-#else
- regtail(ret, ret + 3);
-#endif
}
- if (ISMULT2(regparse))
+ if (ISMULT2(regcomp_parse))
FAIL("nested *?+ in regexp");
return(ret);
STATIC regnode *
regatom(I32 *flagp)
{
+ dTHR;
register regnode *ret = 0;
I32 flags;
*flagp = WORST; /* Tentatively. */
tryagain:
- switch (*regparse) {
+ switch (*regcomp_parse) {
case '^':
seen_zerolen++;
nextchar();
ret = reg_node(BOL);
break;
case '$':
- if (regparse[1])
+ if (regcomp_parse[1])
seen_zerolen++;
nextchar();
if (regflags & PMf_MULTILINE)
*flagp |= HASWIDTH|SIMPLE;
break;
case '[':
- regparse++;
+ regcomp_parse++;
ret = regclass();
*flagp |= HASWIDTH|SIMPLE;
break;
*flagp |= TRYAGAIN;
return NULL;
}
- FAIL2("internal urp in regexp at /%s/", regparse);
+ FAIL2("internal urp in regexp at /%s/", regcomp_parse);
/* Supposed to be caught earlier. */
break;
case '{':
- if (!regcurly(regparse)) {
- regparse++;
+ if (!regcurly(regcomp_parse)) {
+ regcomp_parse++;
goto defchar;
}
/* FALL THROUGH */
FAIL("?+*{} follows nothing in regexp");
break;
case '\\':
- switch (*++regparse) {
+ switch (*++regcomp_parse) {
case 'A':
seen_zerolen++;
ret = reg_node(SBOL);
*flagp |= SIMPLE;
nextchar();
break;
+ case 'z':
+ ret = reg_node(EOS);
+ *flagp |= SIMPLE;
+ seen_zerolen++; /* Do not optimize RE away */
+ nextchar();
+ break;
case 'w':
ret = reg_node((regflags & PMf_LOCALE) ? ALNUML : ALNUM);
*flagp |= HASWIDTH|SIMPLE;
case '1': case '2': case '3': case '4':
case '5': case '6': case '7': case '8': case '9':
{
- I32 num = atoi(regparse);
+ I32 num = atoi(regcomp_parse);
if (num > 9 && num >= regnpar)
goto defchar;
else {
+ if (!SIZE_ONLY && num > regcomp_rx->nparens)
+ FAIL("reference to nonexistent group");
regsawback = 1;
ret = reganode((regflags & PMf_FOLD)
? ((regflags & PMf_LOCALE) ? REFFL : REFF)
: REF, num);
*flagp |= HASWIDTH;
- while (isDIGIT(*regparse))
- regparse++;
- regparse--;
+ while (isDIGIT(*regcomp_parse))
+ regcomp_parse++;
+ regcomp_parse--;
nextchar();
}
}
break;
case '\0':
- if (regparse >= regxend)
+ if (regcomp_parse >= regxend)
FAIL("trailing \\ in regexp");
/* FALL THROUGH */
default:
case '#':
if (regflags & PMf_EXTENDED) {
- while (regparse < regxend && *regparse != '\n') regparse++;
- if (regparse < regxend)
+ while (regcomp_parse < regxend && *regcomp_parse != '\n') regcomp_parse++;
+ if (regcomp_parse < regxend)
goto tryagain;
}
/* FALL THROUGH */
char *oldp, *s;
I32 numlen;
- regparse++;
+ regcomp_parse++;
defchar:
ret = reg_node((regflags & PMf_FOLD)
: EXACT);
s = (char *) OPERAND(ret);
regc(0, s++); /* save spot for len */
- for (len = 0, p = regparse - 1;
+ for (len = 0, p = regcomp_parse - 1;
len < 127 && p < regxend;
len++)
{
case 'A':
case 'G':
case 'Z':
+ case 'z':
case 'w':
case 'W':
case 'b':
regc(ender, s++);
}
loopdone:
- regparse = p - 1;
+ regcomp_parse = p - 1;
nextchar();
if (len < 0)
FAIL("internal disaster in regexp");
*OPERAND(ret) = len;
regc('\0', s++);
if (SIZE_ONLY) {
-#ifdef REGALIGN_STRUCT
regsize += (len + 2 + sizeof(regnode) - 1) / sizeof(regnode);
-#endif
} else {
regcode += (len + 2 + sizeof(regnode) - 1) / sizeof(regnode);
}
return(ret);
}
-static char *
+STATIC char *
regwhite(char *p, char *e)
{
while (p < e) {
return p;
}
-STATIC void
-regset(char *opnd, register I32 c)
-{
- if (SIZE_ONLY)
- return;
- c &= 0xFF;
- opnd[1 + (c >> 3)] |= (1 << (c & 7));
-}
-
STATIC regnode *
regclass(void)
{
+ dTHR;
register char *opnd, *s;
register I32 Class;
register I32 lastclass = 1234;
ret = reg_node(ANYOF);
for (Class = 0; Class < 33; Class++)
regc(0, s++);
- if (*regparse == '^') { /* Complement of range. */
+ if (*regcomp_parse == '^') { /* Complement of range. */
regnaughty++;
- regparse++;
+ regcomp_parse++;
if (!SIZE_ONLY)
*opnd |= ANYOF_INVERT;
}
} else {
regsize += ANY_SKIP;
}
- if (*regparse == ']' || *regparse == '-')
+ if (*regcomp_parse == ']' || *regcomp_parse == '-')
goto skipcond; /* allow 1st char to be ] or - */
- while (regparse < regxend && *regparse != ']') {
+ while (regcomp_parse < regxend && *regcomp_parse != ']') {
skipcond:
- Class = UCHARAT(regparse++);
- if (Class == '[' && regparse + 1 < regxend &&
+ Class = UCHARAT(regcomp_parse++);
+ if (Class == '[' && regcomp_parse + 1 < regxend &&
/* I smell either [: or [= or [. -- POSIX has been here, right? */
- (*regparse == ':' || *regparse == '=' || *regparse == '.')) {
- char posixccc = *regparse;
- char* posixccs = regparse++;
+ (*regcomp_parse == ':' || *regcomp_parse == '=' || *regcomp_parse == '.')) {
+ char posixccc = *regcomp_parse;
+ char* posixccs = regcomp_parse++;
- while (regparse < regxend && *regparse != posixccc)
- regparse++;
- if (regparse == regxend)
+ while (regcomp_parse < regxend && *regcomp_parse != posixccc)
+ regcomp_parse++;
+ if (regcomp_parse == regxend)
/* Grandfather lone [:, [=, [. */
- regparse = posixccs;
+ regcomp_parse = posixccs;
else {
- regparse++; /* skip over the posixccc */
- if (*regparse == ']') {
+ regcomp_parse++; /* skip over the posixccc */
+ if (*regcomp_parse == ']') {
/* Not Implemented Yet.
* (POSIX Extended Character Classes, that is)
* The text between e.g. [: and :] would start
- * at posixccs + 1 and stop at regparse - 2. */
+ * at posixccs + 1 and stop at regcomp_parse - 2. */
if (dowarn && !SIZE_ONLY)
warn("Character class syntax [%c %c] is reserved for future extensions", posixccc, posixccc);
- regparse++; /* skip over the ending ] */
+ regcomp_parse++; /* skip over the ending ] */
}
}
}
if (Class == '\\') {
- Class = UCHARAT(regparse++);
+ Class = UCHARAT(regcomp_parse++);
switch (Class) {
case 'w':
- if (regflags & PMf_LOCALE) {
- if (!SIZE_ONLY)
+ if (!SIZE_ONLY) {
+ if (regflags & PMf_LOCALE)
*opnd |= ANYOF_ALNUML;
- }
- else {
- for (Class = 0; Class < 256; Class++)
- if (isALNUM(Class))
- regset(opnd, Class);
+ else {
+ for (Class = 0; Class < 256; Class++)
+ if (isALNUM(Class))
+ ANYOF_SET(opnd, Class);
+ }
}
lastclass = 1234;
continue;
case 'W':
- if (regflags & PMf_LOCALE) {
- if (!SIZE_ONLY)
+ if (!SIZE_ONLY) {
+ if (regflags & PMf_LOCALE)
*opnd |= ANYOF_NALNUML;
- }
- else {
- for (Class = 0; Class < 256; Class++)
- if (!isALNUM(Class))
- regset(opnd, Class);
+ else {
+ for (Class = 0; Class < 256; Class++)
+ if (!isALNUM(Class))
+ ANYOF_SET(opnd, Class);
+ }
}
lastclass = 1234;
continue;
case 's':
- if (regflags & PMf_LOCALE) {
- if (!SIZE_ONLY)
+ if (!SIZE_ONLY) {
+ if (regflags & PMf_LOCALE)
*opnd |= ANYOF_SPACEL;
- }
- else {
- for (Class = 0; Class < 256; Class++)
- if (isSPACE(Class))
- regset(opnd, Class);
+ else {
+ for (Class = 0; Class < 256; Class++)
+ if (isSPACE(Class))
+ ANYOF_SET(opnd, Class);
+ }
}
lastclass = 1234;
continue;
case 'S':
- if (regflags & PMf_LOCALE) {
- if (!SIZE_ONLY)
+ if (!SIZE_ONLY) {
+ if (regflags & PMf_LOCALE)
*opnd |= ANYOF_NSPACEL;
- }
- else {
- for (Class = 0; Class < 256; Class++)
- if (!isSPACE(Class))
- regset(opnd, Class);
+ else {
+ for (Class = 0; Class < 256; Class++)
+ if (!isSPACE(Class))
+ ANYOF_SET(opnd, Class);
+ }
}
lastclass = 1234;
continue;
case 'd':
- for (Class = '0'; Class <= '9'; Class++)
- regset(opnd, Class);
+ if (!SIZE_ONLY) {
+ for (Class = '0'; Class <= '9'; Class++)
+ ANYOF_SET(opnd, Class);
+ }
lastclass = 1234;
continue;
case 'D':
- for (Class = 0; Class < '0'; Class++)
- regset(opnd, Class);
- for (Class = '9' + 1; Class < 256; Class++)
- regset(opnd, Class);
+ if (!SIZE_ONLY) {
+ for (Class = 0; Class < '0'; Class++)
+ ANYOF_SET(opnd, Class);
+ for (Class = '9' + 1; Class < 256; Class++)
+ ANYOF_SET(opnd, Class);
+ }
lastclass = 1234;
continue;
case 'n':
Class = '\007';
break;
case 'x':
- Class = scan_hex(regparse, 2, &numlen);
- regparse += numlen;
+ Class = scan_hex(regcomp_parse, 2, &numlen);
+ regcomp_parse += numlen;
break;
case 'c':
- Class = UCHARAT(regparse++);
+ Class = UCHARAT(regcomp_parse++);
Class = toCTRL(Class);
break;
case '0': case '1': case '2': case '3': case '4':
case '5': case '6': case '7': case '8': case '9':
- Class = scan_oct(--regparse, 3, &numlen);
- regparse += numlen;
+ Class = scan_oct(--regcomp_parse, 3, &numlen);
+ regcomp_parse += numlen;
break;
}
}
}
else {
lastclass = Class;
- if (*regparse == '-' && regparse+1 < regxend &&
- regparse[1] != ']') {
- regparse++;
+ if (*regcomp_parse == '-' && regcomp_parse+1 < regxend &&
+ regcomp_parse[1] != ']') {
+ regcomp_parse++;
range = 1;
continue; /* do it next time */
}
}
- for ( ; lastclass <= Class; lastclass++)
- regset(opnd, lastclass);
+ if (!SIZE_ONLY) {
+ for ( ; lastclass <= Class; lastclass++)
+ ANYOF_SET(opnd, lastclass);
+ }
lastclass = Class;
}
- if (*regparse != ']')
+ if (*regcomp_parse != ']')
FAIL("unmatched [] in regexp");
nextchar();
+ /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */
+ if (!SIZE_ONLY && (*opnd & (0xFF ^ ANYOF_INVERT)) == ANYOF_FOLD) {
+ for (Class = 0; Class < 256; ++Class) {
+ if (ANYOF_TEST(opnd, Class)) {
+ I32 cf = fold[Class];
+ ANYOF_SET(opnd, cf);
+ }
+ }
+ *opnd &= ~ANYOF_FOLD;
+ }
+ /* optimize inverted simple patterns (e.g. [^a-z]) */
+ if (!SIZE_ONLY && (*opnd & 0xFF) == ANYOF_INVERT) {
+ for (Class = 0; Class < 32; ++Class)
+ opnd[1 + Class] ^= 0xFF;
+ *opnd = 0;
+ }
return ret;
}
STATIC char*
nextchar(void)
{
- char* retval = regparse++;
+ dTHR;
+ char* retval = regcomp_parse++;
for (;;) {
- if (*regparse == '(' && regparse[1] == '?' &&
- regparse[2] == '#') {
- while (*regparse && *regparse != ')')
- regparse++;
- regparse++;
+ if (*regcomp_parse == '(' && regcomp_parse[1] == '?' &&
+ regcomp_parse[2] == '#') {
+ while (*regcomp_parse && *regcomp_parse != ')')
+ regcomp_parse++;
+ regcomp_parse++;
continue;
}
if (regflags & PMf_EXTENDED) {
- if (isSPACE(*regparse)) {
- regparse++;
+ if (isSPACE(*regcomp_parse)) {
+ regcomp_parse++;
continue;
}
- else if (*regparse == '#') {
- while (*regparse && *regparse != '\n')
- regparse++;
- regparse++;
+ else if (*regcomp_parse == '#') {
+ while (*regcomp_parse && *regcomp_parse != '\n')
+ regcomp_parse++;
+ regcomp_parse++;
continue;
}
}
STATIC regnode * /* Location. */
reg_node(U8 op)
{
+ dTHR;
register regnode *ret;
register regnode *ptr;
ret = regcode;
if (SIZE_ONLY) {
SIZE_ALIGN(regsize);
-#ifdef REGALIGN_STRUCT
regsize += 1;
-#else
- regsize += 3;
-#endif
return(ret);
}
STATIC regnode * /* Location. */
reganode(U8 op, U32 arg)
{
+ dTHR;
register regnode *ret;
register regnode *ptr;
ret = regcode;
if (SIZE_ONLY) {
SIZE_ALIGN(regsize);
-#ifdef REGALIGN
regsize += 2;
-#else
- regsize += 5;
-#endif
return(ret);
}
STATIC void
regc(U8 b, char* s)
{
+ dTHR;
if (!SIZE_ONLY)
*s = b;
}
STATIC void
reginsert(U8 op, regnode *opnd)
{
+ dTHR;
register regnode *src;
register regnode *dst;
register regnode *place;
src = NEXTOPER(place);
FILL_ADVANCE_NODE(place, op);
Zero(src, offset, regnode);
-#if defined(REGALIGN) && !defined(REGALIGN_STRUCT)
- src[offset + 1] = '\177';
-#endif
}
/*
STATIC void
regtail(regnode *p, regnode *val)
{
+ dTHR;
register regnode *scan;
register regnode *temp;
register I32 offset;
scan = temp;
}
-#ifdef REGALIGN
-# ifdef REGALIGN_STRUCT
if (reg_off_by_arg[OP(scan)]) {
ARG_SET(scan, val - scan);
} else {
NEXT_OFF(scan) = val - scan;
}
-# else
- offset = val - scan;
-# ifndef lint
- *(short*)(scan+1) = offset;
-# endif
-#endif
-#else
- if (OP(scan) == BACK)
- offset = scan - val;
- else
- offset = val - scan;
- *(scan+1) = (offset>>8)&0377;
- *(scan+2) = offset&0377;
-#endif
}
/*
STATIC void
regoptail(regnode *p, regnode *val)
{
+ dTHR;
/* "Operandless" and "op != BRANCH" are synonymous in practice. */
if (p == NULL || SIZE_ONLY)
return;
return TRUE;
}
-#ifdef DEBUGGING
STATIC regnode *
dumpuntil(regnode *start, regnode *node, regnode *last, SV* sv, I32 l)
{
+#ifdef DEBUGGING
register char op = EXACT; /* Arbitrary non-END op. */
register regnode *next, *onode;
if (OP(node) == OPTIMIZED)
goto after_print;
regprop(sv, node);
- PerlIO_printf(Perl_debug_log, "%4d%*s%s", node - start,
+ PerlIO_printf(Perl_debug_log, "%4d:%*s%s", node - start,
2*l + 1, "", SvPVX(sv));
if (next == NULL) /* Next ptr. */
PerlIO_printf(Perl_debug_log, "(0)");
else if (op == WHILEM)
l--;
}
+#endif /* DEBUGGING */
return node;
}
void
regdump(regexp *r)
{
+#ifdef DEBUGGING
+ dTHR;
SV *sv = sv_newmortal();
(void)dumpuntil(r->program, r->program + 1, NULL, sv, 0);
if (r->reganch & ROPT_IMPLICIT)
PerlIO_printf(Perl_debug_log, "implicit ");
PerlIO_printf(Perl_debug_log, "minlen %ld ", (long) r->minlen);
+ if (r->reganch & ROPT_EVAL_SEEN)
+ PerlIO_printf(Perl_debug_log, "with eval ");
PerlIO_printf(Perl_debug_log, "\n");
+#endif /* DEBUGGING */
}
/*
void
regprop(SV *sv, regnode *o)
{
+#ifdef DEBUGGING
+ dTHR;
register char *p = 0;
- sv_setpv(sv, ":");
+ sv_setpvn(sv, "", 0);
switch (OP(o)) {
case BOL:
p = "BOL";
case EOL:
p = "EOL";
break;
+ case EOS:
+ p = "EOS";
+ break;
case MEOL:
p = "MEOL";
break;
sv_catpvf(sv, "CURLY {%d,%d}", ARG1(o), ARG2(o));
break;
case CURLYM:
-#ifdef REGALIGN_STRUCT
sv_catpvf(sv, "CURLYM[%d] {%d,%d}", o->flags, ARG1(o), ARG2(o));
-#else
- sv_catpvf(sv, "CURLYM {%d,%d}", ARG1(o), ARG2(o));
-#endif
break;
case CURLYN:
-#ifdef REGALIGN_STRUCT
sv_catpvf(sv, "CURLYN[%d] {%d,%d}", o->flags, ARG1(o), ARG2(o));
-#else
- sv_catpvf(sv, "CURLYN {%d,%d}", ARG1(o), ARG2(o));
-#endif
break;
case CURLYX:
sv_catpvf(sv, "CURLYX {%d,%d}", ARG1(o), ARG2(o));
p = "GPOS";
break;
case UNLESSM:
-#ifdef REGALIGN_STRUCT
sv_catpvf(sv, "UNLESSM[-%d]", o->flags);
-#else
- p = "UNLESSM";
-#endif
break;
case IFMATCH:
-#ifdef REGALIGN_STRUCT
sv_catpvf(sv, "IFMATCH[-%d]", o->flags);
-#else
- p = "IFMATCH";
-#endif
break;
case SUCCEED:
p = "SUCCEED";
}
if (p)
sv_catpv(sv, p);
+#endif /* DEBUGGING */
}
-#endif /* DEBUGGING */
void
pregfree(struct regexp *r)
{
+ dTHR;
if (!r || (--r->refcnt > 0))
return;
if (r->precomp)
Safefree(r->precomp);
if (r->subbase)
Safefree(r->subbase);
- if (r->substrs)
+ if (r->substrs) {
+ if (r->anchored_substr)
+ SvREFCNT_dec(r->anchored_substr);
+ if (r->float_substr)
+ SvREFCNT_dec(r->float_substr);
Safefree(r->substrs);
- if (r->anchored_substr)
- SvREFCNT_dec(r->anchored_substr);
- if (r->float_substr)
- SvREFCNT_dec(r->float_substr);
+ }
if (r->data) {
int n = r->data->count;
while (--n >= 0) {
regnode *
regnext(register regnode *p)
{
+ dTHR;
register I32 offset;
if (p == ®dummy)
if (offset == 0)
return(NULL);
-#ifdef REGALIGN
return(p+offset);
-#else
- if (OP(p) == BACK)
- return(p-offset);
- else
- return(p+offset);
-#endif
}
-#ifdef I_STDARG
-static void
+STATIC void
re_croak2(const char* pat1,const char* pat2,...)
-#else
-/*VARARGS0*/
-static void
-re_croak2(const char* pat1,const char* pat2, va_alist)
- const char* pat1;
- const char* pat2;
- va_dcl
-#endif
{
va_list args;
STRLEN l1 = strlen(pat1);
l2 = 510 - l1;
Copy(pat1, buf, l1 , char);
Copy(pat2, buf + l1, l2 , char);
- buf[l1 + l2 + 1] = '\n';
- buf[l1 + l2 + 2] = '\0';
-#ifdef I_STDARG
+ buf[l1 + l2] = '\n';
+ buf[l1 + l2 + 1] = '\0';
va_start(args, pat2);
-#else
- va_start(args);
-#endif
message = mess(buf, &args);
va_end(args);
l1 = strlen(message);
https://perl5.git.perl.org / perl5.git / blobdiff