3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * He still hopefully carried some of his gear in his pack: a small tinder-box,
13 * two small shallow pans, the smaller fitting into the larger; inside them a
14 * wooden spoon, a short two-pronged fork and some skewers were stowed; and
15 * hidden at the bottom of the pack in a flat wooden box a dwindling treasure,
18 * [p.653 of _The Lord of the Rings_, IV/iv: "Of Herbs and Stewed Rabbit"]
21 /* This file contains pp ("push/pop") functions that
22 * execute the opcodes that make up a perl program. A typical pp function
23 * expects to find its arguments on the stack, and usually pushes its
24 * results onto the stack, hence the 'pp' terminology. Each OP structure
25 * contains a pointer to the relevant pp_foo() function.
27 * This particular file just contains pp_pack() and pp_unpack(). See the
28 * other pp*.c files for the rest of the pp_ functions.
32 #define PERL_IN_PP_PACK_C
35 /* Types used by pack/unpack */
37 e_no_len, /* no length */
38 e_number, /* number, [] */
42 typedef struct tempsym {
43 const char* patptr; /* current template char */
44 const char* patend; /* one after last char */
45 const char* grpbeg; /* 1st char of ()-group */
46 const char* grpend; /* end of ()-group */
47 I32 code; /* template code (!<>) */
48 I32 length; /* length/repeat count */
49 howlen_t howlen; /* how length is given */
50 int level; /* () nesting level */
51 U32 flags; /* /=4, comma=2, pack=1 */
52 /* and group modifiers */
53 STRLEN strbeg; /* offset of group start */
54 struct tempsym *previous; /* previous group */
57 #define TEMPSYM_INIT(symptr, p, e, f) \
59 (symptr)->patptr = (p); \
60 (symptr)->patend = (e); \
61 (symptr)->grpbeg = NULL; \
62 (symptr)->grpend = NULL; \
63 (symptr)->grpend = NULL; \
65 (symptr)->length = 0; \
66 (symptr)->howlen = e_no_len; \
67 (symptr)->level = 0; \
68 (symptr)->flags = (f); \
69 (symptr)->strbeg = 0; \
70 (symptr)->previous = NULL; \
78 #if defined(HAS_LONG_DOUBLE) && defined(USE_LONG_DOUBLE)
81 U8 bytes[sizeof(long double)];
88 /* Maximum number of bytes to which a byte can grow due to upgrade */
92 * Offset for integer pack/unpack.
94 * On architectures where I16 and I32 aren't really 16 and 32 bits,
95 * which for now are all Crays, pack and unpack have to play games.
99 * These values are required for portability of pack() output.
100 * If they're not right on your machine, then pack() and unpack()
101 * wouldn't work right anyway; you'll need to apply the Cray hack.
102 * (I'd like to check them with #if, but you can't use sizeof() in
103 * the preprocessor.) --???
106 The appropriate SHORTSIZE, INTSIZE, LONGSIZE, and LONGLONGSIZE
107 defines are now in config.h. --Andy Dougherty April 1998
112 /* CROSSCOMPILE and MULTIARCH are going to affect pp_pack() and pp_unpack().
115 #if U16SIZE > SIZE16 || U32SIZE > SIZE32
116 # if BYTEORDER == 0x1234 || BYTEORDER == 0x12345678 /* little-endian */
117 # define OFF16(p) ((char*)(p))
118 # define OFF32(p) ((char*)(p))
120 # if BYTEORDER == 0x4321 || BYTEORDER == 0x87654321 /* big-endian */
121 # define OFF16(p) ((char*)(p) + (sizeof(U16) - SIZE16))
122 # define OFF32(p) ((char*)(p) + (sizeof(U32) - SIZE32))
124 ++++ bad cray byte order
128 # define OFF16(p) ((char *) (p))
129 # define OFF32(p) ((char *) (p))
132 /* Only to be used inside a loop (see the break) */
133 #define SHIFT16(utf8, s, strend, p, datumtype) STMT_START { \
135 if (!uni_to_bytes(aTHX_ &(s), strend, OFF16(p), SIZE16, datumtype)) break; \
137 Copy(s, OFF16(p), SIZE16, char); \
142 /* Only to be used inside a loop (see the break) */
143 #define SHIFT32(utf8, s, strend, p, datumtype) STMT_START { \
145 if (!uni_to_bytes(aTHX_ &(s), strend, OFF32(p), SIZE32, datumtype)) break; \
147 Copy(s, OFF32(p), SIZE32, char); \
152 #define PUSH16(utf8, cur, p) PUSH_BYTES(utf8, cur, OFF16(p), SIZE16)
153 #define PUSH32(utf8, cur, p) PUSH_BYTES(utf8, cur, OFF32(p), SIZE32)
155 #if BYTEORDER == 0x4321 || BYTEORDER == 0x87654321 /* big-endian */
156 # define NEEDS_SWAP(d) (TYPE_ENDIANNESS(d) == TYPE_IS_LITTLE_ENDIAN)
157 #elif BYTEORDER == 0x1234 || BYTEORDER == 0x12345678 /* little-endian */
158 # define NEEDS_SWAP(d) (TYPE_ENDIANNESS(d) == TYPE_IS_BIG_ENDIAN)
160 # error "Unsupported byteorder"
161 /* Need to add code here to re-instate mixed endian support. */
164 /* Only to be used inside a loop (see the break) */
165 #define SHIFT_BYTES(utf8, s, strend, buf, len, datumtype) \
168 if (!uni_to_bytes(aTHX_ &s, strend, \
169 (char *) (buf), len, datumtype)) break; \
171 Copy(s, (char *) (buf), len, char); \
176 #define SHIFT_VAR(utf8, s, strend, var, datumtype) \
177 SHIFT_BYTES(utf8, s, strend, &(var), sizeof(var), datumtype)
179 #define PUSH_VAR(utf8, aptr, var) \
180 PUSH_BYTES(utf8, aptr, &(var), sizeof(var))
182 /* Avoid stack overflow due to pathological templates. 100 should be plenty. */
183 #define MAX_SUB_TEMPLATE_LEVEL 100
185 /* flags (note that type modifiers can also be used as flags!) */
186 #define FLAG_WAS_UTF8 0x40
187 #define FLAG_PARSE_UTF8 0x20 /* Parse as utf8 */
188 #define FLAG_UNPACK_ONLY_ONE 0x10
189 #define FLAG_DO_UTF8 0x08 /* The underlying string is utf8 */
190 #define FLAG_SLASH 0x04
191 #define FLAG_COMMA 0x02
192 #define FLAG_PACK 0x01
195 S_mul128(pTHX_ SV *sv, U8 m)
198 char *s = SvPV(sv, len);
201 PERL_ARGS_ASSERT_MUL128;
203 if (!strnEQ(s, "0000", 4)) { /* need to grow sv */
204 SV * const tmpNew = newSVpvs("0000000000");
206 sv_catsv(tmpNew, sv);
207 SvREFCNT_dec(sv); /* free old sv */
212 while (!*t) /* trailing '\0'? */
215 const U32 i = ((*t - '0') << 7) + m;
216 *(t--) = '0' + (char)(i % 10);
222 /* Explosives and implosives. */
224 #if 'I' == 73 && 'J' == 74
225 /* On an ASCII/ISO kind of system */
226 #define ISUUCHAR(ch) ((ch) >= ' ' && (ch) < 'a')
229 Some other sort of character set - use memchr() so we don't match
232 #define ISUUCHAR(ch) (memchr(PL_uuemap, (ch), sizeof(PL_uuemap)-1) || (ch) == ' ')
236 #define TYPE_IS_SHRIEKING 0x100
237 #define TYPE_IS_BIG_ENDIAN 0x200
238 #define TYPE_IS_LITTLE_ENDIAN 0x400
239 #define TYPE_IS_PACK 0x800
240 #define TYPE_ENDIANNESS_MASK (TYPE_IS_BIG_ENDIAN|TYPE_IS_LITTLE_ENDIAN)
241 #define TYPE_MODIFIERS(t) ((t) & ~0xFF)
242 #define TYPE_NO_MODIFIERS(t) ((t) & 0xFF)
244 # define TYPE_ENDIANNESS(t) ((t) & TYPE_ENDIANNESS_MASK)
245 # define TYPE_NO_ENDIANNESS(t) ((t) & ~TYPE_ENDIANNESS_MASK)
247 # define ENDIANNESS_ALLOWED_TYPES "sSiIlLqQjJfFdDpP("
249 # define DO_BO_UNPACK(var) \
252 my_swabn(&var, sizeof(var)); \
256 # define DO_BO_PACK(var) \
259 my_swabn(&var, sizeof(var)); \
263 #define PACK_SIZE_CANNOT_CSUM 0x80
264 #define PACK_SIZE_UNPREDICTABLE 0x40 /* Not a fixed size element */
265 #define PACK_SIZE_MASK 0x3F
267 #include "packsizetables.c"
270 uni_to_byte(pTHX_ const char **s, const char *end, I32 datumtype)
273 UV val = utf8n_to_uvchr((U8 *) *s, end-*s, &retlen,
274 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
275 /* We try to process malformed UTF-8 as much as possible (preferably with
276 warnings), but these two mean we make no progress in the string and
277 might enter an infinite loop */
278 if (retlen == (STRLEN) -1 || retlen == 0)
279 Perl_croak(aTHX_ "Malformed UTF-8 string in '%c' format in unpack",
280 (int) TYPE_NO_MODIFIERS(datumtype));
282 Perl_ck_warner(aTHX_ packWARN(WARN_UNPACK),
283 "Character in '%c' format wrapped in unpack",
284 (int) TYPE_NO_MODIFIERS(datumtype));
291 #define SHIFT_BYTE(utf8, s, strend, datumtype) ((utf8) ? \
292 uni_to_byte(aTHX_ &(s), (strend), (datumtype)) : \
296 uni_to_bytes(pTHX_ const char **s, const char *end, const char *buf, int buf_len, I32 datumtype)
300 const char *from = *s;
302 const U32 flags = ckWARN(WARN_UTF8) ?
303 UTF8_CHECK_ONLY : (UTF8_CHECK_ONLY | UTF8_ALLOW_ANY);
304 for (;buf_len > 0; buf_len--) {
305 if (from >= end) return FALSE;
306 val = utf8n_to_uvchr((U8 *) from, end-from, &retlen, flags);
307 if (retlen == (STRLEN) -1 || retlen == 0) {
308 from += UTF8SKIP(from);
310 } else from += retlen;
315 *(U8 *)buf++ = (U8)val;
317 /* We have enough characters for the buffer. Did we have problems ? */
320 /* Rewalk the string fragment while warning */
322 const int flags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
323 for (ptr = *s; ptr < from; ptr += UTF8SKIP(ptr)) {
324 if (ptr >= end) break;
325 utf8n_to_uvuni((U8 *) ptr, end-ptr, &retlen, flags);
327 if (from > end) from = end;
330 Perl_ck_warner(aTHX_ packWARN(datumtype & TYPE_IS_PACK ?
331 WARN_PACK : WARN_UNPACK),
332 "Character(s) in '%c' format wrapped in %s",
333 (int) TYPE_NO_MODIFIERS(datumtype),
334 datumtype & TYPE_IS_PACK ? "pack" : "unpack");
341 next_uni_uu(pTHX_ const char **s, const char *end, I32 *out)
345 const UV val = utf8n_to_uvchr((U8 *) *s, end-*s, &retlen, UTF8_CHECK_ONLY);
346 if (val >= 0x100 || !ISUUCHAR(val) ||
347 retlen == (STRLEN) -1 || retlen == 0) {
351 *out = PL_uudmap[val] & 077;
357 S_bytes_to_uni(const U8 *start, STRLEN len, char *dest) {
358 const U8 * const end = start + len;
360 PERL_ARGS_ASSERT_BYTES_TO_UNI;
362 while (start < end) {
363 const UV uv = NATIVE_TO_ASCII(*start);
364 if (UNI_IS_INVARIANT(uv))
365 *dest++ = (char)(U8)UTF_TO_NATIVE(uv);
367 *dest++ = (char)(U8)UTF8_EIGHT_BIT_HI(uv);
368 *dest++ = (char)(U8)UTF8_EIGHT_BIT_LO(uv);
375 #define PUSH_BYTES(utf8, cur, buf, len) \
378 (cur) = bytes_to_uni((U8 *) buf, len, (cur)); \
380 Copy(buf, cur, len, char); \
385 #define GROWING(utf8, cat, start, cur, in_len) \
387 STRLEN glen = (in_len); \
388 if (utf8) glen *= UTF8_EXPAND; \
389 if ((cur) + glen >= (start) + SvLEN(cat)) { \
390 (start) = sv_exp_grow(cat, glen); \
391 (cur) = (start) + SvCUR(cat); \
395 #define PUSH_GROWING_BYTES(utf8, cat, start, cur, buf, in_len) \
397 const STRLEN glen = (in_len); \
399 if (utf8) gl *= UTF8_EXPAND; \
400 if ((cur) + gl >= (start) + SvLEN(cat)) { \
402 SvCUR_set((cat), (cur) - (start)); \
403 (start) = sv_exp_grow(cat, gl); \
404 (cur) = (start) + SvCUR(cat); \
406 PUSH_BYTES(utf8, cur, buf, glen); \
409 #define PUSH_BYTE(utf8, s, byte) \
412 const U8 au8 = (byte); \
413 (s) = bytes_to_uni(&au8, 1, (s)); \
414 } else *(U8 *)(s)++ = (byte); \
417 /* Only to be used inside a loop (see the break) */
418 #define NEXT_UNI_VAL(val, cur, str, end, utf8_flags) \
421 if (str >= end) break; \
422 val = utf8n_to_uvchr((U8 *) str, end-str, &retlen, utf8_flags); \
423 if (retlen == (STRLEN) -1 || retlen == 0) { \
425 Perl_croak(aTHX_ "Malformed UTF-8 string in pack"); \
430 static const char *_action( const tempsym_t* symptr )
432 return (const char *)(( symptr->flags & FLAG_PACK ) ? "pack" : "unpack");
435 /* Returns the sizeof() struct described by pat */
437 S_measure_struct(pTHX_ tempsym_t* symptr)
441 PERL_ARGS_ASSERT_MEASURE_STRUCT;
443 while (next_symbol(symptr)) {
447 switch (symptr->howlen) {
449 Perl_croak(aTHX_ "Within []-length '*' not allowed in %s",
453 /* e_no_len and e_number */
454 len = symptr->length;
458 size = packprops[TYPE_NO_ENDIANNESS(symptr->code)] & PACK_SIZE_MASK;
461 /* endianness doesn't influence the size of a type */
462 switch(TYPE_NO_ENDIANNESS(symptr->code)) {
464 Perl_croak(aTHX_ "Invalid type '%c' in %s",
465 (int)TYPE_NO_MODIFIERS(symptr->code),
467 case '.' | TYPE_IS_SHRIEKING:
468 case '@' | TYPE_IS_SHRIEKING:
472 case 'U': /* XXXX Is it correct? */
475 Perl_croak(aTHX_ "Within []-length '%c' not allowed in %s",
476 (int) TYPE_NO_MODIFIERS(symptr->code),
483 tempsym_t savsym = *symptr;
484 symptr->patptr = savsym.grpbeg;
485 symptr->patend = savsym.grpend;
486 /* XXXX Theoretically, we need to measure many times at
487 different positions, since the subexpression may contain
488 alignment commands, but be not of aligned length.
489 Need to detect this and croak(). */
490 size = measure_struct(symptr);
494 case 'X' | TYPE_IS_SHRIEKING:
495 /* XXXX Is this useful? Then need to treat MEASURE_BACKWARDS.
497 if (!len) /* Avoid division by 0 */
499 len = total % len; /* Assumed: the start is aligned. */
504 Perl_croak(aTHX_ "'X' outside of string in %s", _action( symptr ) );
506 case 'x' | TYPE_IS_SHRIEKING:
507 if (!len) /* Avoid division by 0 */
509 star = total % len; /* Assumed: the start is aligned. */
510 if (star) /* Other portable ways? */
534 size = sizeof(char*);
544 /* locate matching closing parenthesis or bracket
545 * returns char pointer to char after match, or NULL
548 S_group_end(pTHX_ const char *patptr, const char *patend, char ender)
550 PERL_ARGS_ASSERT_GROUP_END;
552 while (patptr < patend) {
553 const char c = *patptr++;
560 while (patptr < patend && *patptr != '\n')
564 patptr = group_end(patptr, patend, ')') + 1;
566 patptr = group_end(patptr, patend, ']') + 1;
568 Perl_croak(aTHX_ "No group ending character '%c' found in template",
574 /* Convert unsigned decimal number to binary.
575 * Expects a pointer to the first digit and address of length variable
576 * Advances char pointer to 1st non-digit char and returns number
579 S_get_num(pTHX_ const char *patptr, I32 *lenptr )
581 I32 len = *patptr++ - '0';
583 PERL_ARGS_ASSERT_GET_NUM;
585 while (isDIGIT(*patptr)) {
586 if (len >= 0x7FFFFFFF/10)
587 Perl_croak(aTHX_ "pack/unpack repeat count overflow");
588 len = (len * 10) + (*patptr++ - '0');
594 /* The marvellous template parsing routine: Using state stored in *symptr,
595 * locates next template code and count
598 S_next_symbol(pTHX_ tempsym_t* symptr )
600 const char* patptr = symptr->patptr;
601 const char* const patend = symptr->patend;
603 PERL_ARGS_ASSERT_NEXT_SYMBOL;
605 symptr->flags &= ~FLAG_SLASH;
607 while (patptr < patend) {
608 if (isSPACE(*patptr))
610 else if (*patptr == '#') {
612 while (patptr < patend && *patptr != '\n')
617 /* We should have found a template code */
618 I32 code = *patptr++ & 0xFF;
619 U32 inherited_modifiers = 0;
621 if (code == ','){ /* grandfather in commas but with a warning */
622 if (((symptr->flags & FLAG_COMMA) == 0) && ckWARN(WARN_UNPACK)){
623 symptr->flags |= FLAG_COMMA;
624 Perl_warner(aTHX_ packWARN(WARN_UNPACK),
625 "Invalid type ',' in %s", _action( symptr ) );
630 /* for '(', skip to ')' */
632 if( isDIGIT(*patptr) || *patptr == '*' || *patptr == '[' )
633 Perl_croak(aTHX_ "()-group starts with a count in %s",
635 symptr->grpbeg = patptr;
636 patptr = 1 + ( symptr->grpend = group_end(patptr, patend, ')') );
637 if( symptr->level >= MAX_SUB_TEMPLATE_LEVEL )
638 Perl_croak(aTHX_ "Too deeply nested ()-groups in %s",
642 /* look for group modifiers to inherit */
643 if (TYPE_ENDIANNESS(symptr->flags)) {
644 if (strchr(ENDIANNESS_ALLOWED_TYPES, TYPE_NO_MODIFIERS(code)))
645 inherited_modifiers |= TYPE_ENDIANNESS(symptr->flags);
648 /* look for modifiers */
649 while (patptr < patend) {
654 modifier = TYPE_IS_SHRIEKING;
655 allowed = "sSiIlLxXnNvV@.";
658 modifier = TYPE_IS_BIG_ENDIAN;
659 allowed = ENDIANNESS_ALLOWED_TYPES;
662 modifier = TYPE_IS_LITTLE_ENDIAN;
663 allowed = ENDIANNESS_ALLOWED_TYPES;
674 if (!strchr(allowed, TYPE_NO_MODIFIERS(code)))
675 Perl_croak(aTHX_ "'%c' allowed only after types %s in %s", *patptr,
676 allowed, _action( symptr ) );
678 if (TYPE_ENDIANNESS(code | modifier) == TYPE_ENDIANNESS_MASK)
679 Perl_croak(aTHX_ "Can't use both '<' and '>' after type '%c' in %s",
680 (int) TYPE_NO_MODIFIERS(code), _action( symptr ) );
681 else if (TYPE_ENDIANNESS(code | modifier | inherited_modifiers) ==
682 TYPE_ENDIANNESS_MASK)
683 Perl_croak(aTHX_ "Can't use '%c' in a group with different byte-order in %s",
684 *patptr, _action( symptr ) );
686 if ((code & modifier)) {
687 Perl_ck_warner(aTHX_ packWARN(WARN_UNPACK),
688 "Duplicate modifier '%c' after '%c' in %s",
689 *patptr, (int) TYPE_NO_MODIFIERS(code),
697 /* inherit modifiers */
698 code |= inherited_modifiers;
700 /* look for count and/or / */
701 if (patptr < patend) {
702 if (isDIGIT(*patptr)) {
703 patptr = get_num( patptr, &symptr->length );
704 symptr->howlen = e_number;
706 } else if (*patptr == '*') {
708 symptr->howlen = e_star;
710 } else if (*patptr == '[') {
711 const char* lenptr = ++patptr;
712 symptr->howlen = e_number;
713 patptr = group_end( patptr, patend, ']' ) + 1;
714 /* what kind of [] is it? */
715 if (isDIGIT(*lenptr)) {
716 lenptr = get_num( lenptr, &symptr->length );
718 Perl_croak(aTHX_ "Malformed integer in [] in %s",
721 tempsym_t savsym = *symptr;
722 symptr->patend = patptr-1;
723 symptr->patptr = lenptr;
724 savsym.length = measure_struct(symptr);
728 symptr->howlen = e_no_len;
733 while (patptr < patend) {
734 if (isSPACE(*patptr))
736 else if (*patptr == '#') {
738 while (patptr < patend && *patptr != '\n')
743 if (*patptr == '/') {
744 symptr->flags |= FLAG_SLASH;
746 if (patptr < patend &&
747 (isDIGIT(*patptr) || *patptr == '*' || *patptr == '['))
748 Perl_croak(aTHX_ "'/' does not take a repeat count in %s",
755 /* at end - no count, no / */
756 symptr->howlen = e_no_len;
761 symptr->patptr = patptr;
765 symptr->patptr = patptr;
770 There is no way to cleanly handle the case where we should process the
771 string per byte in its upgraded form while it's really in downgraded form
772 (e.g. estimates like strend-s as an upper bound for the number of
773 characters left wouldn't work). So if we foresee the need of this
774 (pattern starts with U or contains U0), we want to work on the encoded
775 version of the string. Users are advised to upgrade their pack string
776 themselves if they need to do a lot of unpacks like this on it
779 need_utf8(const char *pat, const char *patend)
783 PERL_ARGS_ASSERT_NEED_UTF8;
785 while (pat < patend) {
788 pat = (const char *) memchr(pat, '\n', patend-pat);
789 if (!pat) return FALSE;
790 } else if (pat[0] == 'U') {
791 if (first || pat[1] == '0') return TRUE;
792 } else first = FALSE;
799 first_symbol(const char *pat, const char *patend) {
800 PERL_ARGS_ASSERT_FIRST_SYMBOL;
802 while (pat < patend) {
803 if (pat[0] != '#') return pat[0];
805 pat = (const char *) memchr(pat, '\n', patend-pat);
813 =for apidoc unpackstring
815 The engine implementing the unpack() Perl function.
817 Using the template pat..patend, this function unpacks the string
818 s..strend into a number of mortal SVs, which it pushes onto the perl
819 argument (@_) stack (so you will need to issue a C<PUTBACK> before and
820 C<SPAGAIN> after the call to this function). It returns the number of
823 The strend and patend pointers should point to the byte following the last
824 character of each string.
826 Although this function returns its values on the perl argument stack, it
827 doesn't take any parameters from that stack (and thus in particular
828 there's no need to do a PUSHMARK before calling it, unlike L</call_pv> for
834 Perl_unpackstring(pTHX_ const char *pat, const char *patend, const char *s, const char *strend, U32 flags)
838 PERL_ARGS_ASSERT_UNPACKSTRING;
840 if (flags & FLAG_DO_UTF8) flags |= FLAG_WAS_UTF8;
841 else if (need_utf8(pat, patend)) {
842 /* We probably should try to avoid this in case a scalar context call
843 wouldn't get to the "U0" */
844 STRLEN len = strend - s;
845 s = (char *) bytes_to_utf8((U8 *) s, &len);
848 flags |= FLAG_DO_UTF8;
851 if (first_symbol(pat, patend) != 'U' && (flags & FLAG_DO_UTF8))
852 flags |= FLAG_PARSE_UTF8;
854 TEMPSYM_INIT(&sym, pat, patend, flags);
856 return unpack_rec(&sym, s, s, strend, NULL );
860 S_unpack_rec(pTHX_ tempsym_t* symptr, const char *s, const char *strbeg, const char *strend, const char **new_s )
864 const I32 start_sp_offset = SP - PL_stack_base;
869 const int bits_in_uv = CHAR_BIT * sizeof(cuv);
871 bool explicit_length;
872 const bool unpack_only_one = (symptr->flags & FLAG_UNPACK_ONLY_ONE) != 0;
873 bool utf8 = (symptr->flags & FLAG_PARSE_UTF8) ? 1 : 0;
875 PERL_ARGS_ASSERT_UNPACK_REC;
877 symptr->strbeg = s - strbeg;
879 while (next_symbol(symptr)) {
882 I32 datumtype = symptr->code;
884 /* do first one only unless in list context
885 / is implemented by unpacking the count, then popping it from the
886 stack, so must check that we're not in the middle of a / */
888 && (SP - PL_stack_base == start_sp_offset + 1)
889 && (datumtype != '/') ) /* XXX can this be omitted */
892 switch (howlen = symptr->howlen) {
894 len = strend - strbeg; /* long enough */
897 /* e_no_len and e_number */
898 len = symptr->length;
902 explicit_length = TRUE;
904 beyond = s >= strend;
906 props = packprops[TYPE_NO_ENDIANNESS(datumtype)];
908 /* props nonzero means we can process this letter. */
909 const long size = props & PACK_SIZE_MASK;
910 const long howmany = (strend - s) / size;
914 if (!checksum || (props & PACK_SIZE_CANNOT_CSUM)) {
915 if (len && unpack_only_one) len = 1;
921 needs_swap = NEEDS_SWAP(datumtype);
923 switch(TYPE_NO_ENDIANNESS(datumtype)) {
925 Perl_croak(aTHX_ "Invalid type '%c' in unpack", (int)TYPE_NO_MODIFIERS(datumtype) );
928 if (howlen == e_no_len)
929 len = 16; /* len is not specified */
937 tempsym_t savsym = *symptr;
938 const U32 group_modifiers = TYPE_MODIFIERS(datumtype & ~symptr->flags);
939 symptr->flags |= group_modifiers;
940 symptr->patend = savsym.grpend;
941 symptr->previous = &savsym;
944 if (len && unpack_only_one) len = 1;
946 symptr->patptr = savsym.grpbeg;
947 if (utf8) symptr->flags |= FLAG_PARSE_UTF8;
948 else symptr->flags &= ~FLAG_PARSE_UTF8;
949 unpack_rec(symptr, s, strbeg, strend, &s);
950 if (s == strend && savsym.howlen == e_star)
951 break; /* No way to continue */
954 savsym.flags = symptr->flags & ~group_modifiers;
958 case '.' | TYPE_IS_SHRIEKING:
962 const bool u8 = utf8 && !(datumtype & TYPE_IS_SHRIEKING);
963 if (howlen == e_star) from = strbeg;
964 else if (len <= 0) from = s;
966 tempsym_t *group = symptr;
968 while (--len && group) group = group->previous;
969 from = group ? strbeg + group->strbeg : strbeg;
972 newSVuv( u8 ? (UV) utf8_length((const U8*)from, (const U8*)s) : (UV) (s-from)) :
973 newSViv(-(u8 ? (IV) utf8_length((const U8*)s, (const U8*)from) : (IV) (from-s)));
977 case '@' | TYPE_IS_SHRIEKING:
979 s = strbeg + symptr->strbeg;
980 if (utf8 && !(datumtype & TYPE_IS_SHRIEKING))
984 Perl_croak(aTHX_ "'@' outside of string in unpack");
989 Perl_croak(aTHX_ "'@' outside of string with malformed UTF-8 in unpack");
992 Perl_croak(aTHX_ "'@' outside of string in unpack");
996 case 'X' | TYPE_IS_SHRIEKING:
997 if (!len) /* Avoid division by 0 */
1000 const char *hop, *last;
1002 hop = last = strbeg;
1004 hop += UTF8SKIP(hop);
1011 Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
1015 len = (s - strbeg) % len;
1021 Perl_croak(aTHX_ "'X' outside of string in unpack");
1022 while (--s, UTF8_IS_CONTINUATION(*s)) {
1024 Perl_croak(aTHX_ "'X' outside of string in unpack");
1029 if (len > s - strbeg)
1030 Perl_croak(aTHX_ "'X' outside of string in unpack" );
1034 case 'x' | TYPE_IS_SHRIEKING: {
1036 if (!len) /* Avoid division by 0 */
1038 if (utf8) ai32 = utf8_length((U8 *) strbeg, (U8 *) s) % len;
1039 else ai32 = (s - strbeg) % len;
1040 if (ai32 == 0) break;
1048 Perl_croak(aTHX_ "'x' outside of string in unpack");
1053 if (len > strend - s)
1054 Perl_croak(aTHX_ "'x' outside of string in unpack");
1059 Perl_croak(aTHX_ "'/' must follow a numeric type in unpack");
1065 /* Preliminary length estimate is assumed done in 'W' */
1066 if (len > strend - s) len = strend - s;
1072 for (l=len, hop=s; l>0; l--, hop += UTF8SKIP(hop)) {
1073 if (hop >= strend) {
1075 Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
1080 Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
1082 } else if (len > strend - s)
1085 if (datumtype == 'Z') {
1086 /* 'Z' strips stuff after first null */
1087 const char *ptr, *end;
1089 for (ptr = s; ptr < end; ptr++) if (*ptr == 0) break;
1090 sv = newSVpvn(s, ptr-s);
1091 if (howlen == e_star) /* exact for 'Z*' */
1092 len = ptr-s + (ptr != strend ? 1 : 0);
1093 } else if (datumtype == 'A') {
1094 /* 'A' strips both nulls and spaces */
1096 if (utf8 && (symptr->flags & FLAG_WAS_UTF8)) {
1097 for (ptr = s+len-1; ptr >= s; ptr--)
1098 if (*ptr != 0 && !UTF8_IS_CONTINUATION(*ptr) &&
1099 !isSPACE_utf8(ptr)) break;
1100 if (ptr >= s) ptr += UTF8SKIP(ptr);
1103 Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
1105 for (ptr = s+len-1; ptr >= s; ptr--)
1106 if (*ptr != 0 && !isSPACE(*ptr)) break;
1109 sv = newSVpvn(s, ptr-s);
1110 } else sv = newSVpvn(s, len);
1114 /* Undo any upgrade done due to need_utf8() */
1115 if (!(symptr->flags & FLAG_WAS_UTF8))
1116 sv_utf8_downgrade(sv, 0);
1124 if (howlen == e_star || len > (strend - s) * 8)
1125 len = (strend - s) * 8;
1128 while (len >= 8 && s < strend) {
1129 cuv += PL_bitcount[uni_to_byte(aTHX_ &s, strend, datumtype)];
1134 cuv += PL_bitcount[*(U8 *)s++];
1137 if (len && s < strend) {
1139 bits = SHIFT_BYTE(utf8, s, strend, datumtype);
1140 if (datumtype == 'b')
1142 if (bits & 1) cuv++;
1147 if (bits & 0x80) cuv++;
1154 sv = sv_2mortal(newSV(len ? len : 1));
1157 if (datumtype == 'b') {
1159 const I32 ai32 = len;
1160 for (len = 0; len < ai32; len++) {
1161 if (len & 7) bits >>= 1;
1163 if (s >= strend) break;
1164 bits = uni_to_byte(aTHX_ &s, strend, datumtype);
1165 } else bits = *(U8 *) s++;
1166 *str++ = bits & 1 ? '1' : '0';
1170 const I32 ai32 = len;
1171 for (len = 0; len < ai32; len++) {
1172 if (len & 7) bits <<= 1;
1174 if (s >= strend) break;
1175 bits = uni_to_byte(aTHX_ &s, strend, datumtype);
1176 } else bits = *(U8 *) s++;
1177 *str++ = bits & 0x80 ? '1' : '0';
1181 SvCUR_set(sv, str - SvPVX_const(sv));
1188 /* Preliminary length estimate, acceptable for utf8 too */
1189 if (howlen == e_star || len > (strend - s) * 2)
1190 len = (strend - s) * 2;
1192 sv = sv_2mortal(newSV(len ? len : 1));
1196 if (datumtype == 'h') {
1199 for (len = 0; len < ai32; len++) {
1200 if (len & 1) bits >>= 4;
1202 if (s >= strend) break;
1203 bits = uni_to_byte(aTHX_ &s, strend, datumtype);
1204 } else bits = * (U8 *) s++;
1206 *str++ = PL_hexdigit[bits & 15];
1210 const I32 ai32 = len;
1211 for (len = 0; len < ai32; len++) {
1212 if (len & 1) bits <<= 4;
1214 if (s >= strend) break;
1215 bits = uni_to_byte(aTHX_ &s, strend, datumtype);
1216 } else bits = *(U8 *) s++;
1218 *str++ = PL_hexdigit[(bits >> 4) & 15];
1223 SvCUR_set(sv, str - SvPVX_const(sv));
1230 if (explicit_length)
1231 /* Switch to "character" mode */
1232 utf8 = (symptr->flags & FLAG_DO_UTF8) ? 1 : 0;
1237 while (len-- > 0 && s < strend) {
1242 aint = utf8n_to_uvchr((U8 *) s, strend-s, &retlen,
1243 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1244 if (retlen == (STRLEN) -1 || retlen == 0)
1245 Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
1249 aint = *(U8 *)(s)++;
1250 if (aint >= 128 && datumtype != 'C') /* fake up signed chars */
1254 else if (checksum > bits_in_uv)
1255 cdouble += (NV)aint;
1263 while (len-- > 0 && s < strend) {
1265 const UV val = utf8n_to_uvchr((U8 *) s, strend-s, &retlen,
1266 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1267 if (retlen == (STRLEN) -1 || retlen == 0)
1268 Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
1272 else if (checksum > bits_in_uv)
1273 cdouble += (NV) val;
1277 } else if (!checksum)
1279 const U8 ch = *(U8 *) s++;
1282 else if (checksum > bits_in_uv)
1283 while (len-- > 0) cdouble += (NV) *(U8 *) s++;
1285 while (len-- > 0) cuv += *(U8 *) s++;
1289 if (explicit_length && howlen != e_star) {
1290 /* Switch to "bytes in UTF-8" mode */
1291 if (symptr->flags & FLAG_DO_UTF8) utf8 = 0;
1293 /* Should be impossible due to the need_utf8() test */
1294 Perl_croak(aTHX_ "U0 mode on a byte string");
1298 if (len > strend - s) len = strend - s;
1300 if (len && unpack_only_one) len = 1;
1304 while (len-- > 0 && s < strend) {
1308 U8 result[UTF8_MAXLEN];
1309 const char *ptr = s;
1311 /* Bug: warns about bad utf8 even if we are short on bytes
1312 and will break out of the loop */
1313 if (!uni_to_bytes(aTHX_ &ptr, strend, (char *) result, 1,
1316 len = UTF8SKIP(result);
1317 if (!uni_to_bytes(aTHX_ &ptr, strend,
1318 (char *) &result[1], len-1, 'U')) break;
1319 auv = utf8n_to_uvuni(result, len, &retlen, UTF8_ALLOW_DEFAULT);
1322 auv = utf8n_to_uvuni((U8*)s, strend - s, &retlen, UTF8_ALLOW_DEFAULT);
1323 if (retlen == (STRLEN) -1 || retlen == 0)
1324 Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
1329 else if (checksum > bits_in_uv)
1330 cdouble += (NV) auv;
1335 case 's' | TYPE_IS_SHRIEKING:
1336 #if SHORTSIZE != SIZE16
1339 SHIFT_VAR(utf8, s, strend, ashort, datumtype);
1340 DO_BO_UNPACK(ashort);
1343 else if (checksum > bits_in_uv)
1344 cdouble += (NV)ashort;
1356 #if U16SIZE > SIZE16
1359 SHIFT16(utf8, s, strend, &ai16, datumtype);
1361 #if U16SIZE > SIZE16
1367 else if (checksum > bits_in_uv)
1368 cdouble += (NV)ai16;
1373 case 'S' | TYPE_IS_SHRIEKING:
1374 #if SHORTSIZE != SIZE16
1376 unsigned short aushort;
1377 SHIFT_VAR(utf8, s, strend, aushort, datumtype);
1378 DO_BO_UNPACK(aushort);
1381 else if (checksum > bits_in_uv)
1382 cdouble += (NV)aushort;
1395 #if U16SIZE > SIZE16
1398 SHIFT16(utf8, s, strend, &au16, datumtype);
1400 if (datumtype == 'n')
1401 au16 = PerlSock_ntohs(au16);
1402 if (datumtype == 'v')
1406 else if (checksum > bits_in_uv)
1407 cdouble += (NV) au16;
1412 case 'v' | TYPE_IS_SHRIEKING:
1413 case 'n' | TYPE_IS_SHRIEKING:
1416 # if U16SIZE > SIZE16
1419 SHIFT16(utf8, s, strend, &ai16, datumtype);
1420 if (datumtype == ('n' | TYPE_IS_SHRIEKING))
1421 ai16 = (I16) PerlSock_ntohs((U16) ai16);
1422 if (datumtype == ('v' | TYPE_IS_SHRIEKING))
1423 ai16 = (I16) vtohs((U16) ai16);
1426 else if (checksum > bits_in_uv)
1427 cdouble += (NV) ai16;
1433 case 'i' | TYPE_IS_SHRIEKING:
1436 SHIFT_VAR(utf8, s, strend, aint, datumtype);
1440 else if (checksum > bits_in_uv)
1441 cdouble += (NV)aint;
1447 case 'I' | TYPE_IS_SHRIEKING:
1450 SHIFT_VAR(utf8, s, strend, auint, datumtype);
1451 DO_BO_UNPACK(auint);
1454 else if (checksum > bits_in_uv)
1455 cdouble += (NV)auint;
1463 SHIFT_VAR(utf8, s, strend, aiv, datumtype);
1467 else if (checksum > bits_in_uv)
1476 SHIFT_VAR(utf8, s, strend, auv, datumtype);
1480 else if (checksum > bits_in_uv)
1486 case 'l' | TYPE_IS_SHRIEKING:
1487 #if LONGSIZE != SIZE32
1490 SHIFT_VAR(utf8, s, strend, along, datumtype);
1491 DO_BO_UNPACK(along);
1494 else if (checksum > bits_in_uv)
1495 cdouble += (NV)along;
1506 #if U32SIZE > SIZE32
1509 SHIFT32(utf8, s, strend, &ai32, datumtype);
1511 #if U32SIZE > SIZE32
1512 if (ai32 > 2147483647) ai32 -= 4294967296;
1516 else if (checksum > bits_in_uv)
1517 cdouble += (NV)ai32;
1522 case 'L' | TYPE_IS_SHRIEKING:
1523 #if LONGSIZE != SIZE32
1525 unsigned long aulong;
1526 SHIFT_VAR(utf8, s, strend, aulong, datumtype);
1527 DO_BO_UNPACK(aulong);
1530 else if (checksum > bits_in_uv)
1531 cdouble += (NV)aulong;
1544 #if U32SIZE > SIZE32
1547 SHIFT32(utf8, s, strend, &au32, datumtype);
1549 if (datumtype == 'N')
1550 au32 = PerlSock_ntohl(au32);
1551 if (datumtype == 'V')
1555 else if (checksum > bits_in_uv)
1556 cdouble += (NV)au32;
1561 case 'V' | TYPE_IS_SHRIEKING:
1562 case 'N' | TYPE_IS_SHRIEKING:
1565 #if U32SIZE > SIZE32
1568 SHIFT32(utf8, s, strend, &ai32, datumtype);
1569 if (datumtype == ('N' | TYPE_IS_SHRIEKING))
1570 ai32 = (I32)PerlSock_ntohl((U32)ai32);
1571 if (datumtype == ('V' | TYPE_IS_SHRIEKING))
1572 ai32 = (I32)vtohl((U32)ai32);
1575 else if (checksum > bits_in_uv)
1576 cdouble += (NV)ai32;
1584 SHIFT_VAR(utf8, s, strend, aptr, datumtype);
1586 /* newSVpv generates undef if aptr is NULL */
1587 mPUSHs(newSVpv(aptr, 0));
1595 while (len > 0 && s < strend) {
1597 ch = SHIFT_BYTE(utf8, s, strend, datumtype);
1598 auv = (auv << 7) | (ch & 0x7f);
1599 /* UTF8_IS_XXXXX not right here - using constant 0x80 */
1607 if (++bytes >= sizeof(UV)) { /* promote to string */
1610 sv = Perl_newSVpvf(aTHX_ "%.*"UVuf, (int)TYPE_DIGITS(UV), auv);
1611 while (s < strend) {
1612 ch = SHIFT_BYTE(utf8, s, strend, datumtype);
1613 sv = mul128(sv, (U8)(ch & 0x7f));
1619 t = SvPV_nolen_const(sv);
1628 if ((s >= strend) && bytes)
1629 Perl_croak(aTHX_ "Unterminated compressed integer in unpack");
1633 if (symptr->howlen == e_star)
1634 Perl_croak(aTHX_ "'P' must have an explicit size in unpack");
1636 if (s + sizeof(char*) <= strend) {
1638 SHIFT_VAR(utf8, s, strend, aptr, datumtype);
1640 /* newSVpvn generates undef if aptr is NULL */
1641 PUSHs(newSVpvn_flags(aptr, len, SVs_TEMP));
1648 SHIFT_VAR(utf8, s, strend, aquad, datumtype);
1649 DO_BO_UNPACK(aquad);
1651 mPUSHs(aquad >= IV_MIN && aquad <= IV_MAX ?
1652 newSViv((IV)aquad) : newSVnv((NV)aquad));
1653 else if (checksum > bits_in_uv)
1654 cdouble += (NV)aquad;
1662 SHIFT_VAR(utf8, s, strend, auquad, datumtype);
1663 DO_BO_UNPACK(auquad);
1665 mPUSHs(auquad <= UV_MAX ?
1666 newSVuv((UV)auquad) : newSVnv((NV)auquad));
1667 else if (checksum > bits_in_uv)
1668 cdouble += (NV)auquad;
1673 #endif /* HAS_QUAD */
1674 /* float and double added gnb@melba.bby.oz.au 22/11/89 */
1678 SHIFT_VAR(utf8, s, strend, afloat, datumtype);
1679 DO_BO_UNPACK(afloat);
1689 SHIFT_VAR(utf8, s, strend, adouble, datumtype);
1690 DO_BO_UNPACK(adouble);
1700 SHIFT_BYTES(utf8, s, strend, anv.bytes, sizeof(anv.bytes), datumtype);
1701 DO_BO_UNPACK(anv.nv);
1708 #if defined(HAS_LONG_DOUBLE) && defined(USE_LONG_DOUBLE)
1712 SHIFT_BYTES(utf8, s, strend, aldouble.bytes, sizeof(aldouble.bytes), datumtype);
1713 DO_BO_UNPACK(aldouble.ld);
1715 mPUSHn(aldouble.ld);
1717 cdouble += aldouble.ld;
1723 const STRLEN l = (STRLEN) (strend - s) * 3 / 4;
1724 sv = sv_2mortal(newSV(l));
1725 if (l) SvPOK_on(sv);
1728 while (next_uni_uu(aTHX_ &s, strend, &len)) {
1733 next_uni_uu(aTHX_ &s, strend, &a);
1734 next_uni_uu(aTHX_ &s, strend, &b);
1735 next_uni_uu(aTHX_ &s, strend, &c);
1736 next_uni_uu(aTHX_ &s, strend, &d);
1737 hunk[0] = (char)((a << 2) | (b >> 4));
1738 hunk[1] = (char)((b << 4) | (c >> 2));
1739 hunk[2] = (char)((c << 6) | d);
1741 sv_catpvn(sv, hunk, (len > 3) ? 3 : len);
1749 /* possible checksum byte */
1750 const char *skip = s+UTF8SKIP(s);
1751 if (skip < strend && *skip == '\n')
1757 while (s < strend && *s > ' ' && ISUUCHAR(*s)) {
1761 len = PL_uudmap[*(U8*)s++] & 077;
1763 if (s < strend && ISUUCHAR(*s))
1764 a = PL_uudmap[*(U8*)s++] & 077;
1767 if (s < strend && ISUUCHAR(*s))
1768 b = PL_uudmap[*(U8*)s++] & 077;
1771 if (s < strend && ISUUCHAR(*s))
1772 c = PL_uudmap[*(U8*)s++] & 077;
1775 if (s < strend && ISUUCHAR(*s))
1776 d = PL_uudmap[*(U8*)s++] & 077;
1779 hunk[0] = (char)((a << 2) | (b >> 4));
1780 hunk[1] = (char)((b << 4) | (c >> 2));
1781 hunk[2] = (char)((c << 6) | d);
1783 sv_catpvn(sv, hunk, (len > 3) ? 3 : len);
1788 else /* possible checksum byte */
1789 if (s + 1 < strend && s[1] == '\n')
1799 if (strchr("fFdD", TYPE_NO_MODIFIERS(datumtype)) ||
1800 (checksum > bits_in_uv &&
1801 strchr("cCsSiIlLnNUWvVqQjJ", TYPE_NO_MODIFIERS(datumtype))) ) {
1804 anv = (NV) (1 << (checksum & 15));
1805 while (checksum >= 16) {
1809 while (cdouble < 0.0)
1811 cdouble = Perl_modf(cdouble / anv, &trouble) * anv;
1812 sv = newSVnv(cdouble);
1815 if (checksum < bits_in_uv) {
1816 UV mask = ((UV)1 << checksum) - 1;
1825 if (symptr->flags & FLAG_SLASH){
1826 if (SP - PL_stack_base - start_sp_offset <= 0)
1828 if( next_symbol(symptr) ){
1829 if( symptr->howlen == e_number )
1830 Perl_croak(aTHX_ "Count after length/code in unpack" );
1832 /* ...end of char buffer then no decent length available */
1833 Perl_croak(aTHX_ "length/code after end of string in unpack" );
1835 /* take top of stack (hope it's numeric) */
1838 Perl_croak(aTHX_ "Negative '/' count in unpack" );
1841 Perl_croak(aTHX_ "Code missing after '/' in unpack" );
1843 datumtype = symptr->code;
1844 explicit_length = FALSE;
1852 return SP - PL_stack_base - start_sp_offset;
1860 I32 gimme = GIMME_V;
1863 const char *pat = SvPV_const(left, llen);
1864 const char *s = SvPV_const(right, rlen);
1865 const char *strend = s + rlen;
1866 const char *patend = pat + llen;
1870 cnt = unpackstring(pat, patend, s, strend,
1871 ((gimme == G_SCALAR) ? FLAG_UNPACK_ONLY_ONE : 0)
1872 | (DO_UTF8(right) ? FLAG_DO_UTF8 : 0));
1875 if ( !cnt && gimme == G_SCALAR )
1876 PUSHs(&PL_sv_undef);
1881 doencodes(U8 *h, const char *s, I32 len)
1883 *h++ = PL_uuemap[len];
1885 *h++ = PL_uuemap[(077 & (s[0] >> 2))];
1886 *h++ = PL_uuemap[(077 & (((s[0] << 4) & 060) | ((s[1] >> 4) & 017)))];
1887 *h++ = PL_uuemap[(077 & (((s[1] << 2) & 074) | ((s[2] >> 6) & 03)))];
1888 *h++ = PL_uuemap[(077 & (s[2] & 077))];
1893 const char r = (len > 1 ? s[1] : '\0');
1894 *h++ = PL_uuemap[(077 & (s[0] >> 2))];
1895 *h++ = PL_uuemap[(077 & (((s[0] << 4) & 060) | ((r >> 4) & 017)))];
1896 *h++ = PL_uuemap[(077 & ((r << 2) & 074))];
1897 *h++ = PL_uuemap[0];
1904 S_is_an_int(pTHX_ const char *s, STRLEN l)
1906 SV *result = newSVpvn(s, l);
1907 char *const result_c = SvPV_nolen(result); /* convenience */
1908 char *out = result_c;
1912 PERL_ARGS_ASSERT_IS_AN_INT;
1920 SvREFCNT_dec(result);
1943 SvREFCNT_dec(result);
1949 SvCUR_set(result, out - result_c);
1953 /* pnum must be '\0' terminated */
1955 S_div128(pTHX_ SV *pnum, bool *done)
1958 char * const s = SvPV(pnum, len);
1962 PERL_ARGS_ASSERT_DIV128;
1966 const int i = m * 10 + (*t - '0');
1967 const int r = (i >> 7); /* r < 10 */
1975 SvCUR_set(pnum, (STRLEN) (t - s));
1980 =for apidoc packlist
1982 The engine implementing pack() Perl function.
1988 Perl_packlist(pTHX_ SV *cat, const char *pat, const char *patend, SV **beglist, SV **endlist )
1993 PERL_ARGS_ASSERT_PACKLIST;
1995 TEMPSYM_INIT(&sym, pat, patend, FLAG_PACK);
1997 /* We're going to do changes through SvPVX(cat). Make sure it's valid.
1998 Also make sure any UTF8 flag is loaded */
1999 SvPV_force_nolen(cat);
2001 sym.flags |= FLAG_PARSE_UTF8 | FLAG_DO_UTF8;
2003 (void)pack_rec( cat, &sym, beglist, endlist );
2006 /* like sv_utf8_upgrade, but also repoint the group start markers */
2008 marked_upgrade(pTHX_ SV *sv, tempsym_t *sym_ptr) {
2011 const char *from_ptr, *from_start, *from_end, **marks, **m;
2012 char *to_start, *to_ptr;
2014 if (SvUTF8(sv)) return;
2016 from_start = SvPVX_const(sv);
2017 from_end = from_start + SvCUR(sv);
2018 for (from_ptr = from_start; from_ptr < from_end; from_ptr++)
2019 if (!NATIVE_IS_INVARIANT(*from_ptr)) break;
2020 if (from_ptr == from_end) {
2021 /* Simple case: no character needs to be changed */
2026 len = (from_end-from_ptr)*UTF8_EXPAND+(from_ptr-from_start)+1;
2027 Newx(to_start, len, char);
2028 Copy(from_start, to_start, from_ptr-from_start, char);
2029 to_ptr = to_start + (from_ptr-from_start);
2031 Newx(marks, sym_ptr->level+2, const char *);
2032 for (group=sym_ptr; group; group = group->previous)
2033 marks[group->level] = from_start + group->strbeg;
2034 marks[sym_ptr->level+1] = from_end+1;
2035 for (m = marks; *m < from_ptr; m++)
2036 *m = to_start + (*m-from_start);
2038 for (;from_ptr < from_end; from_ptr++) {
2039 while (*m == from_ptr) *m++ = to_ptr;
2040 to_ptr = (char *) uvchr_to_utf8((U8 *) to_ptr, *(U8 *) from_ptr);
2044 while (*m == from_ptr) *m++ = to_ptr;
2045 if (m != marks + sym_ptr->level+1) {
2048 Perl_croak(aTHX_ "panic: marks beyond string end, m=%p, marks=%p, "
2049 "level=%d", m, marks, sym_ptr->level);
2051 for (group=sym_ptr; group; group = group->previous)
2052 group->strbeg = marks[group->level] - to_start;
2057 SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
2058 from_start -= SvIVX(sv);
2061 SvFLAGS(sv) &= ~SVf_OOK;
2064 Safefree(from_start);
2065 SvPV_set(sv, to_start);
2066 SvCUR_set(sv, to_ptr - to_start);
2071 /* Exponential string grower. Makes string extension effectively O(n)
2072 needed says how many extra bytes we need (not counting the final '\0')
2073 Only grows the string if there is an actual lack of space
2076 S_sv_exp_grow(pTHX_ SV *sv, STRLEN needed) {
2077 const STRLEN cur = SvCUR(sv);
2078 const STRLEN len = SvLEN(sv);
2081 PERL_ARGS_ASSERT_SV_EXP_GROW;
2083 if (len - cur > needed) return SvPVX(sv);
2084 extend = needed > len ? needed : len;
2085 return SvGROW(sv, len+extend+1);
2090 S_pack_rec(pTHX_ SV *cat, tempsym_t* symptr, SV **beglist, SV **endlist )
2093 tempsym_t lookahead;
2094 I32 items = endlist - beglist;
2095 bool found = next_symbol(symptr);
2096 bool utf8 = (symptr->flags & FLAG_PARSE_UTF8) ? 1 : 0;
2097 bool warn_utf8 = ckWARN(WARN_UTF8);
2099 PERL_ARGS_ASSERT_PACK_REC;
2101 if (symptr->level == 0 && found && symptr->code == 'U') {
2102 marked_upgrade(aTHX_ cat, symptr);
2103 symptr->flags |= FLAG_DO_UTF8;
2106 symptr->strbeg = SvCUR(cat);
2112 SV *lengthcode = NULL;
2113 I32 datumtype = symptr->code;
2114 howlen_t howlen = symptr->howlen;
2115 char *start = SvPVX(cat);
2116 char *cur = start + SvCUR(cat);
2119 #define NEXTFROM (lengthcode ? lengthcode : items-- > 0 ? *beglist++ : &PL_sv_no)
2123 len = strchr("@Xxu", TYPE_NO_MODIFIERS(datumtype)) ?
2127 /* e_no_len and e_number */
2128 len = symptr->length;
2133 packprops_t props = packprops[TYPE_NO_ENDIANNESS(datumtype)];
2135 if (props && !(props & PACK_SIZE_UNPREDICTABLE)) {
2136 /* We can process this letter. */
2137 STRLEN size = props & PACK_SIZE_MASK;
2138 GROWING(utf8, cat, start, cur, (STRLEN) len * size);
2142 /* Look ahead for next symbol. Do we have code/code? */
2143 lookahead = *symptr;
2144 found = next_symbol(&lookahead);
2145 if (symptr->flags & FLAG_SLASH) {
2147 if (!found) Perl_croak(aTHX_ "Code missing after '/' in pack");
2148 if (strchr("aAZ", lookahead.code)) {
2149 if (lookahead.howlen == e_number) count = lookahead.length;
2152 count = sv_len_utf8(*beglist);
2155 if (lookahead.code == 'Z') count++;
2158 if (lookahead.howlen == e_number && lookahead.length < items)
2159 count = lookahead.length;
2162 lookahead.howlen = e_number;
2163 lookahead.length = count;
2164 lengthcode = sv_2mortal(newSViv(count));
2167 needs_swap = NEEDS_SWAP(datumtype);
2169 /* Code inside the switch must take care to properly update
2170 cat (CUR length and '\0' termination) if it updated *cur and
2171 doesn't simply leave using break */
2172 switch(TYPE_NO_ENDIANNESS(datumtype)) {
2174 Perl_croak(aTHX_ "Invalid type '%c' in pack",
2175 (int) TYPE_NO_MODIFIERS(datumtype));
2177 Perl_croak(aTHX_ "'%%' may not be used in pack");
2180 case '.' | TYPE_IS_SHRIEKING:
2182 if (howlen == e_star) from = start;
2183 else if (len == 0) from = cur;
2185 tempsym_t *group = symptr;
2187 while (--len && group) group = group->previous;
2188 from = group ? start + group->strbeg : start;
2191 len = SvIV(fromstr);
2193 case '@' | TYPE_IS_SHRIEKING:
2195 from = start + symptr->strbeg;
2197 if (utf8 && !(datumtype & TYPE_IS_SHRIEKING))
2199 while (len && from < cur) {
2200 from += UTF8SKIP(from);
2204 Perl_croak(aTHX_ "Malformed UTF-8 string in pack");
2206 /* Here we know from == cur */
2208 GROWING(0, cat, start, cur, len);
2209 Zero(cur, len, char);
2211 } else if (from < cur) {
2214 } else goto no_change;
2222 if (len > 0) goto grow;
2223 if (len == 0) goto no_change;
2230 tempsym_t savsym = *symptr;
2231 U32 group_modifiers = TYPE_MODIFIERS(datumtype & ~symptr->flags);
2232 symptr->flags |= group_modifiers;
2233 symptr->patend = savsym.grpend;
2235 symptr->previous = &lookahead;
2238 if (utf8) symptr->flags |= FLAG_PARSE_UTF8;
2239 else symptr->flags &= ~FLAG_PARSE_UTF8;
2240 was_utf8 = SvUTF8(cat);
2241 symptr->patptr = savsym.grpbeg;
2242 beglist = pack_rec(cat, symptr, beglist, endlist);
2243 if (SvUTF8(cat) != was_utf8)
2244 /* This had better be an upgrade while in utf8==0 mode */
2247 if (savsym.howlen == e_star && beglist == endlist)
2248 break; /* No way to continue */
2250 items = endlist - beglist;
2251 lookahead.flags = symptr->flags & ~group_modifiers;
2254 case 'X' | TYPE_IS_SHRIEKING:
2255 if (!len) /* Avoid division by 0 */
2262 hop += UTF8SKIP(hop);
2269 Perl_croak(aTHX_ "Malformed UTF-8 string in pack");
2273 len = (cur-start) % len;
2277 if (len < 1) goto no_change;
2281 Perl_croak(aTHX_ "'%c' outside of string in pack",
2282 (int) TYPE_NO_MODIFIERS(datumtype));
2283 while (--cur, UTF8_IS_CONTINUATION(*cur)) {
2285 Perl_croak(aTHX_ "'%c' outside of string in pack",
2286 (int) TYPE_NO_MODIFIERS(datumtype));
2292 if (cur - start < len)
2293 Perl_croak(aTHX_ "'%c' outside of string in pack",
2294 (int) TYPE_NO_MODIFIERS(datumtype));
2297 if (cur < start+symptr->strbeg) {
2298 /* Make sure group starts don't point into the void */
2300 const STRLEN length = cur-start;
2301 for (group = symptr;
2302 group && length < group->strbeg;
2303 group = group->previous) group->strbeg = length;
2304 lookahead.strbeg = length;
2307 case 'x' | TYPE_IS_SHRIEKING: {
2309 if (!len) /* Avoid division by 0 */
2311 if (utf8) ai32 = utf8_length((U8 *) start, (U8 *) cur) % len;
2312 else ai32 = (cur - start) % len;
2313 if (ai32 == 0) goto no_change;
2325 aptr = SvPV_const(fromstr, fromlen);
2326 if (DO_UTF8(fromstr)) {
2327 const char *end, *s;
2329 if (!utf8 && !SvUTF8(cat)) {
2330 marked_upgrade(aTHX_ cat, symptr);
2331 lookahead.flags |= FLAG_DO_UTF8;
2332 lookahead.strbeg = symptr->strbeg;
2335 cur = start + SvCUR(cat);
2337 if (howlen == e_star) {
2338 if (utf8) goto string_copy;
2342 end = aptr + fromlen;
2343 fromlen = datumtype == 'Z' ? len-1 : len;
2344 while ((I32) fromlen > 0 && s < end) {
2349 Perl_croak(aTHX_ "Malformed UTF-8 string in pack");
2352 if (datumtype == 'Z') len++;
2358 fromlen = len - fromlen;
2359 if (datumtype == 'Z') fromlen--;
2360 if (howlen == e_star) {
2362 if (datumtype == 'Z') len++;
2364 GROWING(0, cat, start, cur, len);
2365 if (!uni_to_bytes(aTHX_ &aptr, end, cur, fromlen,
2366 datumtype | TYPE_IS_PACK))
2367 Perl_croak(aTHX_ "panic: predicted utf8 length not available, "
2368 "for '%c', aptr=%p end=%p cur=%p, fromlen=%"UVuf,
2369 (int)datumtype, aptr, end, cur, (UV)fromlen);
2373 if (howlen == e_star) {
2375 if (datumtype == 'Z') len++;
2377 if (len <= (I32) fromlen) {
2379 if (datumtype == 'Z' && fromlen > 0) fromlen--;
2381 /* assumes a byte expands to at most UTF8_EXPAND bytes on
2383 expected_length <= from_len*UTF8_EXPAND + (len-from_len) */
2384 GROWING(0, cat, start, cur, fromlen*(UTF8_EXPAND-1)+len);
2386 while (fromlen > 0) {
2387 cur = (char *) uvchr_to_utf8((U8 *) cur, * (U8 *) aptr);
2393 if (howlen == e_star) {
2395 if (datumtype == 'Z') len++;
2397 if (len <= (I32) fromlen) {
2399 if (datumtype == 'Z' && fromlen > 0) fromlen--;
2401 GROWING(0, cat, start, cur, len);
2402 Copy(aptr, cur, fromlen, char);
2406 memset(cur, datumtype == 'A' ? ' ' : '\0', len);
2413 const char *str, *end;
2420 str = SvPV_const(fromstr, fromlen);
2421 end = str + fromlen;
2422 if (DO_UTF8(fromstr)) {
2424 utf8_flags = warn_utf8 ? 0 : UTF8_ALLOW_ANY;
2426 utf8_source = FALSE;
2427 utf8_flags = 0; /* Unused, but keep compilers happy */
2429 if (howlen == e_star) len = fromlen;
2430 field_len = (len+7)/8;
2431 GROWING(utf8, cat, start, cur, field_len);
2432 if (len > (I32)fromlen) len = fromlen;
2435 if (datumtype == 'B')
2439 NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
2441 } else bits |= *str++ & 1;
2442 if (l & 7) bits <<= 1;
2444 PUSH_BYTE(utf8, cur, bits);
2449 /* datumtype == 'b' */
2453 NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
2454 if (val & 1) bits |= 0x80;
2455 } else if (*str++ & 1)
2457 if (l & 7) bits >>= 1;
2459 PUSH_BYTE(utf8, cur, bits);
2465 if (datumtype == 'B')
2466 bits <<= 7 - (l & 7);
2468 bits >>= 7 - (l & 7);
2469 PUSH_BYTE(utf8, cur, bits);
2472 /* Determine how many chars are left in the requested field */
2474 if (howlen == e_star) field_len = 0;
2475 else field_len -= l;
2476 Zero(cur, field_len, char);
2482 const char *str, *end;
2489 str = SvPV_const(fromstr, fromlen);
2490 end = str + fromlen;
2491 if (DO_UTF8(fromstr)) {
2493 utf8_flags = warn_utf8 ? 0 : UTF8_ALLOW_ANY;
2495 utf8_source = FALSE;
2496 utf8_flags = 0; /* Unused, but keep compilers happy */
2498 if (howlen == e_star) len = fromlen;
2499 field_len = (len+1)/2;
2500 GROWING(utf8, cat, start, cur, field_len);
2501 if (!utf8 && len > (I32)fromlen) len = fromlen;
2504 if (datumtype == 'H')
2508 NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
2509 if (val < 256 && isALPHA(val))
2510 bits |= (val + 9) & 0xf;
2513 } else if (isALPHA(*str))
2514 bits |= (*str++ + 9) & 0xf;
2516 bits |= *str++ & 0xf;
2517 if (l & 1) bits <<= 4;
2519 PUSH_BYTE(utf8, cur, bits);
2527 NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
2528 if (val < 256 && isALPHA(val))
2529 bits |= ((val + 9) & 0xf) << 4;
2531 bits |= (val & 0xf) << 4;
2532 } else if (isALPHA(*str))
2533 bits |= ((*str++ + 9) & 0xf) << 4;
2535 bits |= (*str++ & 0xf) << 4;
2536 if (l & 1) bits >>= 4;
2538 PUSH_BYTE(utf8, cur, bits);
2544 PUSH_BYTE(utf8, cur, bits);
2547 /* Determine how many chars are left in the requested field */
2549 if (howlen == e_star) field_len = 0;
2550 else field_len -= l;
2551 Zero(cur, field_len, char);
2559 aiv = SvIV(fromstr);
2560 if ((-128 > aiv || aiv > 127))
2561 Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
2562 "Character in 'c' format wrapped in pack");
2563 PUSH_BYTE(utf8, cur, (U8)(aiv & 0xff));
2568 utf8 = (symptr->flags & FLAG_DO_UTF8) ? 1 : 0;
2574 aiv = SvIV(fromstr);
2575 if ((0 > aiv || aiv > 0xff))
2576 Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
2577 "Character in 'C' format wrapped in pack");
2578 PUSH_BYTE(utf8, cur, (U8)(aiv & 0xff));
2583 U8 in_bytes = (U8)IN_BYTES;
2585 end = start+SvLEN(cat)-1;
2586 if (utf8) end -= UTF8_MAXLEN-1;
2590 auv = SvUV(fromstr);
2591 if (in_bytes) auv = auv % 0x100;
2596 SvCUR_set(cat, cur - start);
2598 GROWING(0, cat, start, cur, len+UTF8_MAXLEN);
2599 end = start+SvLEN(cat)-UTF8_MAXLEN;
2601 cur = (char *) uvuni_to_utf8_flags((U8 *) cur,
2604 0 : UNICODE_ALLOW_ANY);
2609 SvCUR_set(cat, cur - start);
2610 marked_upgrade(aTHX_ cat, symptr);
2611 lookahead.flags |= FLAG_DO_UTF8;
2612 lookahead.strbeg = symptr->strbeg;
2615 cur = start + SvCUR(cat);
2616 end = start+SvLEN(cat)-UTF8_MAXLEN;
2619 Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
2620 "Character in 'W' format wrapped in pack");
2625 SvCUR_set(cat, cur - start);
2626 GROWING(0, cat, start, cur, len+1);
2627 end = start+SvLEN(cat)-1;
2629 *(U8 *) cur++ = (U8)auv;
2638 if (!(symptr->flags & FLAG_DO_UTF8)) {
2639 marked_upgrade(aTHX_ cat, symptr);
2640 lookahead.flags |= FLAG_DO_UTF8;
2641 lookahead.strbeg = symptr->strbeg;
2647 end = start+SvLEN(cat);
2648 if (!utf8) end -= UTF8_MAXLEN;
2652 auv = SvUV(fromstr);
2654 U8 buffer[UTF8_MAXLEN], *endb;
2655 endb = uvuni_to_utf8_flags(buffer, auv,
2657 0 : UNICODE_ALLOW_ANY);
2658 if (cur+(endb-buffer)*UTF8_EXPAND >= end) {
2660 SvCUR_set(cat, cur - start);
2661 GROWING(0, cat, start, cur,
2662 len+(endb-buffer)*UTF8_EXPAND);
2663 end = start+SvLEN(cat);
2665 cur = bytes_to_uni(buffer, endb-buffer, cur);
2669 SvCUR_set(cat, cur - start);
2670 GROWING(0, cat, start, cur, len+UTF8_MAXLEN);
2671 end = start+SvLEN(cat)-UTF8_MAXLEN;
2673 cur = (char *) uvuni_to_utf8_flags((U8 *) cur, auv,
2675 0 : UNICODE_ALLOW_ANY);
2680 /* Float and double added by gnb@melba.bby.oz.au 22/11/89 */
2686 anv = SvNV(fromstr);
2687 # if defined(VMS) && !defined(_IEEE_FP)
2688 /* IEEE fp overflow shenanigans are unavailable on VAX and optional
2689 * on Alpha; fake it if we don't have them.
2693 else if (anv < -FLT_MAX)
2695 else afloat = (float)anv;
2697 afloat = (float)anv;
2700 PUSH_VAR(utf8, cur, afloat);
2708 anv = SvNV(fromstr);
2709 # if defined(VMS) && !defined(_IEEE_FP)
2710 /* IEEE fp overflow shenanigans are unavailable on VAX and optional
2711 * on Alpha; fake it if we don't have them.
2715 else if (anv < -DBL_MAX)
2717 else adouble = (double)anv;
2719 adouble = (double)anv;
2721 DO_BO_PACK(adouble);
2722 PUSH_VAR(utf8, cur, adouble);
2727 Zero(&anv, 1, NV); /* can be long double with unused bits */
2731 /* to work round a gcc/x86 bug; don't use SvNV */
2732 anv.nv = sv_2nv(fromstr);
2734 anv.nv = SvNV(fromstr);
2737 PUSH_BYTES(utf8, cur, anv.bytes, sizeof(anv.bytes));
2741 #if defined(HAS_LONG_DOUBLE) && defined(USE_LONG_DOUBLE)
2744 /* long doubles can have unused bits, which may be nonzero */
2745 Zero(&aldouble, 1, long double);
2749 /* to work round a gcc/x86 bug; don't use SvNV */
2750 aldouble.ld = (long double)sv_2nv(fromstr);
2752 aldouble.ld = (long double)SvNV(fromstr);
2754 DO_BO_PACK(aldouble);
2755 PUSH_BYTES(utf8, cur, aldouble.bytes, sizeof(aldouble.bytes));
2760 case 'n' | TYPE_IS_SHRIEKING:
2765 ai16 = (I16)SvIV(fromstr);
2766 ai16 = PerlSock_htons(ai16);
2767 PUSH16(utf8, cur, &ai16);
2770 case 'v' | TYPE_IS_SHRIEKING:
2775 ai16 = (I16)SvIV(fromstr);
2777 PUSH16(utf8, cur, &ai16);
2780 case 'S' | TYPE_IS_SHRIEKING:
2781 #if SHORTSIZE != SIZE16
2783 unsigned short aushort;
2785 aushort = SvUV(fromstr);
2786 DO_BO_PACK(aushort);
2787 PUSH_VAR(utf8, cur, aushort);
2797 au16 = (U16)SvUV(fromstr);
2799 PUSH16(utf8, cur, &au16);
2802 case 's' | TYPE_IS_SHRIEKING:
2803 #if SHORTSIZE != SIZE16
2807 ashort = SvIV(fromstr);
2809 PUSH_VAR(utf8, cur, ashort);
2819 ai16 = (I16)SvIV(fromstr);
2821 PUSH16(utf8, cur, &ai16);
2825 case 'I' | TYPE_IS_SHRIEKING:
2829 auint = SvUV(fromstr);
2831 PUSH_VAR(utf8, cur, auint);
2838 aiv = SvIV(fromstr);
2840 PUSH_VAR(utf8, cur, aiv);
2847 auv = SvUV(fromstr);
2849 PUSH_VAR(utf8, cur, auv);
2856 anv = SvNV(fromstr);
2860 SvCUR_set(cat, cur - start);
2861 Perl_croak(aTHX_ "Cannot compress negative numbers in pack");
2864 /* 0xFFFFFFFFFFFFFFFF may cast to 18446744073709551616.0,
2865 which is == UV_MAX_P1. IOK is fine (instead of UV_only), as
2866 any negative IVs will have already been got by the croak()
2867 above. IOK is untrue for fractions, so we test them
2868 against UV_MAX_P1. */
2869 if (SvIOK(fromstr) || anv < UV_MAX_P1) {
2870 char buf[(sizeof(UV)*CHAR_BIT)/7+1];
2871 char *in = buf + sizeof(buf);
2872 UV auv = SvUV(fromstr);
2875 *--in = (char)((auv & 0x7f) | 0x80);
2878 buf[sizeof(buf) - 1] &= 0x7f; /* clear continue bit */
2879 PUSH_GROWING_BYTES(utf8, cat, start, cur,
2880 in, (buf + sizeof(buf)) - in);
2881 } else if (SvPOKp(fromstr))
2883 else if (SvNOKp(fromstr)) {
2884 /* 10**NV_MAX_10_EXP is the largest power of 10
2885 so 10**(NV_MAX_10_EXP+1) is definitely unrepresentable
2886 given 10**(NV_MAX_10_EXP+1) == 128 ** x solve for x:
2887 x = (NV_MAX_10_EXP+1) * log (10) / log (128)
2888 And with that many bytes only Inf can overflow.
2889 Some C compilers are strict about integral constant
2890 expressions so we conservatively divide by a slightly
2891 smaller integer instead of multiplying by the exact
2892 floating-point value.
2894 #ifdef NV_MAX_10_EXP
2895 /* char buf[1 + (int)((NV_MAX_10_EXP + 1) * 0.47456)]; -- invalid C */
2896 char buf[1 + (int)((NV_MAX_10_EXP + 1) / 2)]; /* valid C */
2898 /* char buf[1 + (int)((308 + 1) * 0.47456)]; -- invalid C */
2899 char buf[1 + (int)((308 + 1) / 2)]; /* valid C */
2901 char *in = buf + sizeof(buf);
2903 anv = Perl_floor(anv);
2905 const NV next = Perl_floor(anv / 128);
2906 if (in <= buf) /* this cannot happen ;-) */
2907 Perl_croak(aTHX_ "Cannot compress integer in pack");
2908 *--in = (unsigned char)(anv - (next * 128)) | 0x80;
2911 buf[sizeof(buf) - 1] &= 0x7f; /* clear continue bit */
2912 PUSH_GROWING_BYTES(utf8, cat, start, cur,
2913 in, (buf + sizeof(buf)) - in);
2922 /* Copy string and check for compliance */
2923 from = SvPV_const(fromstr, len);
2924 if ((norm = is_an_int(from, len)) == NULL)
2925 Perl_croak(aTHX_ "Can only compress unsigned integers in pack");
2927 Newx(result, len, char);
2930 while (!done) *--in = div128(norm, &done) | 0x80;
2931 result[len - 1] &= 0x7F; /* clear continue bit */
2932 PUSH_GROWING_BYTES(utf8, cat, start, cur,
2933 in, (result + len) - in);
2935 SvREFCNT_dec(norm); /* free norm */
2940 case 'i' | TYPE_IS_SHRIEKING:
2944 aint = SvIV(fromstr);
2946 PUSH_VAR(utf8, cur, aint);
2949 case 'N' | TYPE_IS_SHRIEKING:
2954 au32 = SvUV(fromstr);
2955 au32 = PerlSock_htonl(au32);
2956 PUSH32(utf8, cur, &au32);
2959 case 'V' | TYPE_IS_SHRIEKING:
2964 au32 = SvUV(fromstr);
2966 PUSH32(utf8, cur, &au32);
2969 case 'L' | TYPE_IS_SHRIEKING:
2970 #if LONGSIZE != SIZE32
2972 unsigned long aulong;
2974 aulong = SvUV(fromstr);
2976 PUSH_VAR(utf8, cur, aulong);
2986 au32 = SvUV(fromstr);
2988 PUSH32(utf8, cur, &au32);
2991 case 'l' | TYPE_IS_SHRIEKING:
2992 #if LONGSIZE != SIZE32
2996 along = SvIV(fromstr);
2998 PUSH_VAR(utf8, cur, along);
3008 ai32 = SvIV(fromstr);
3010 PUSH32(utf8, cur, &ai32);
3018 auquad = (Uquad_t) SvUV(fromstr);
3020 PUSH_VAR(utf8, cur, auquad);
3027 aquad = (Quad_t)SvIV(fromstr);
3029 PUSH_VAR(utf8, cur, aquad);
3032 #endif /* HAS_QUAD */
3034 len = 1; /* assume SV is correct length */
3035 GROWING(utf8, cat, start, cur, sizeof(char *));
3042 SvGETMAGIC(fromstr);
3043 if (!SvOK(fromstr)) aptr = NULL;
3045 /* XXX better yet, could spirit away the string to
3046 * a safe spot and hang on to it until the result
3047 * of pack() (and all copies of the result) are
3050 if ((SvTEMP(fromstr) || (SvPADTMP(fromstr) &&
3051 !SvREADONLY(fromstr)))) {
3052 Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
3053 "Attempt to pack pointer to temporary value");
3055 if (SvPOK(fromstr) || SvNIOK(fromstr))
3056 aptr = SvPV_nomg_const_nolen(fromstr);
3058 aptr = SvPV_force_flags_nolen(fromstr, 0);
3061 PUSH_VAR(utf8, cur, aptr);
3065 const char *aptr, *aend;
3069 if (len <= 2) len = 45;
3070 else len = len / 3 * 3;
3072 Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
3073 "Field too wide in 'u' format in pack");
3076 aptr = SvPV_const(fromstr, fromlen);
3077 from_utf8 = DO_UTF8(fromstr);
3079 aend = aptr + fromlen;
3080 fromlen = sv_len_utf8_nomg(fromstr);
3081 } else aend = NULL; /* Unused, but keep compilers happy */
3082 GROWING(utf8, cat, start, cur, (fromlen+2) / 3 * 4 + (fromlen+len-1)/len * 2);
3083 while (fromlen > 0) {
3086 U8 hunk[1+63/3*4+1];
3088 if ((I32)fromlen > len)
3094 if (!uni_to_bytes(aTHX_ &aptr, aend, buffer, todo,
3095 'u' | TYPE_IS_PACK)) {
3097 SvCUR_set(cat, cur - start);
3098 Perl_croak(aTHX_ "panic: string is shorter than advertised, "
3099 "aptr=%p, aend=%p, buffer=%p, todo=%ld",
3100 aptr, aend, buffer, (long) todo);
3102 end = doencodes(hunk, buffer, todo);
3104 end = doencodes(hunk, aptr, todo);
3107 PUSH_BYTES(utf8, cur, hunk, end-hunk);
3114 SvCUR_set(cat, cur - start);
3116 *symptr = lookahead;
3125 dVAR; dSP; dMARK; dORIGMARK; dTARGET;
3128 SV *pat_sv = *++MARK;
3129 const char *pat = SvPV_const(pat_sv, fromlen);
3130 const char *patend = pat + fromlen;
3136 packlist(cat, pat, patend, MARK, SP + 1);
3146 * c-indentation-style: bsd
3148 * indent-tabs-mode: nil
3151 * ex: set ts=8 sts=4 sw=4 et: