9 |
|
|
10 |
Written by: Philip Hazel <ph10@cam.ac.uk> |
Written by: Philip Hazel <ph10@cam.ac.uk> |
11 |
|
|
12 |
Copyright (c) 1997-1999 University of Cambridge |
Copyright (c) 1997-2000 University of Cambridge |
13 |
|
|
14 |
----------------------------------------------------------------------------- |
----------------------------------------------------------------------------- |
15 |
Permission is granted to anyone to use this software for any purpose on any |
Permission is granted to anyone to use this software for any purpose on any |
82 |
"*", "*?", "+", "+?", "?", "??", "{", "{", "{", |
"*", "*?", "+", "+?", "?", "??", "{", "{", "{", |
83 |
"*", "*?", "+", "+?", "?", "??", "{", "{", "{", |
"*", "*?", "+", "+?", "?", "??", "{", "{", "{", |
84 |
"*", "*?", "+", "+?", "?", "??", "{", "{", |
"*", "*?", "+", "+?", "?", "??", "{", "{", |
85 |
"class", "Ref", |
"class", "Ref", "Recurse", |
86 |
"Alt", "Ket", "KetRmax", "KetRmin", "Assert", "Assert not", |
"Alt", "Ket", "KetRmax", "KetRmin", "Assert", "Assert not", |
87 |
"AssertB", "AssertB not", "Reverse", "Once", "Cond", "Cref", |
"AssertB", "AssertB not", "Reverse", "Once", "Cond", "Cref", |
88 |
"Brazero", "Braminzero", "Bra" |
"Brazero", "Braminzero", "Bra" |
107 |
0, 0, -ESC_z /* x - z */ |
0, 0, -ESC_z /* x - z */ |
108 |
}; |
}; |
109 |
|
|
110 |
|
/* Tables of names of POSIX character classes and their lengths. The list is |
111 |
|
terminated by a zero length entry. The first three must be alpha, upper, lower, |
112 |
|
as this is assumed for handling case independence. */ |
113 |
|
|
114 |
|
static const char *posix_names[] = { |
115 |
|
"alpha", "lower", "upper", |
116 |
|
"alnum", "ascii", "cntrl", "digit", "graph", |
117 |
|
"print", "punct", "space", "word", "xdigit" }; |
118 |
|
|
119 |
|
static const uschar posix_name_lengths[] = { |
120 |
|
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 }; |
121 |
|
|
122 |
|
/* Table of class bit maps for each POSIX class; up to three may be combined |
123 |
|
to form the class. */ |
124 |
|
|
125 |
|
static const int posix_class_maps[] = { |
126 |
|
cbit_lower, cbit_upper, -1, /* alpha */ |
127 |
|
cbit_lower, -1, -1, /* lower */ |
128 |
|
cbit_upper, -1, -1, /* upper */ |
129 |
|
cbit_digit, cbit_lower, cbit_upper, /* alnum */ |
130 |
|
cbit_print, cbit_cntrl, -1, /* ascii */ |
131 |
|
cbit_cntrl, -1, -1, /* cntrl */ |
132 |
|
cbit_digit, -1, -1, /* digit */ |
133 |
|
cbit_graph, -1, -1, /* graph */ |
134 |
|
cbit_print, -1, -1, /* print */ |
135 |
|
cbit_punct, -1, -1, /* punct */ |
136 |
|
cbit_space, -1, -1, /* space */ |
137 |
|
cbit_word, -1, -1, /* word */ |
138 |
|
cbit_xdigit,-1, -1 /* xdigit */ |
139 |
|
}; |
140 |
|
|
141 |
|
|
142 |
/* Definition to allow mutual recursion */ |
/* Definition to allow mutual recursion */ |
143 |
|
|
144 |
static BOOL |
static BOOL |
193 |
|
|
194 |
|
|
195 |
/************************************************* |
/************************************************* |
196 |
* Return info about a compiled pattern * |
* (Obsolete) Return info about compiled pattern * |
197 |
*************************************************/ |
*************************************************/ |
198 |
|
|
199 |
/* This function picks potentially useful data out of the private |
/* This is the original "info" function. It picks potentially useful data out |
200 |
structure. The public options are passed back in an int - though the |
of the private structure, but its interface was too rigid. It remains for |
201 |
re->options field has been expanded to a long int, all the public options |
backwards compatibility. The public options are passed back in an int - though |
202 |
|
the re->options field has been expanded to a long int, all the public options |
203 |
at the low end of it, and so even on 16-bit systems this will still be OK. |
at the low end of it, and so even on 16-bit systems this will still be OK. |
204 |
Therefore, I haven't changed the API for pcre_info(). |
Therefore, I haven't changed the API for pcre_info(). |
205 |
|
|
210 |
or -1 if multiline and all branches start ^, |
or -1 if multiline and all branches start ^, |
211 |
or -2 otherwise |
or -2 otherwise |
212 |
|
|
213 |
Returns: number of identifying extraction brackets |
Returns: number of capturing subpatterns |
214 |
or negative values on error |
or negative values on error |
215 |
*/ |
*/ |
216 |
|
|
229 |
|
|
230 |
|
|
231 |
|
|
232 |
|
/************************************************* |
233 |
|
* Return info about compiled pattern * |
234 |
|
*************************************************/ |
235 |
|
|
236 |
|
/* This is a newer "info" function which has an extensible interface so |
237 |
|
that additional items can be added compatibly. |
238 |
|
|
239 |
|
Arguments: |
240 |
|
external_re points to compiled code |
241 |
|
external_study points to study data, or NULL |
242 |
|
what what information is required |
243 |
|
where where to put the information |
244 |
|
|
245 |
|
Returns: 0 if data returned, negative on error |
246 |
|
*/ |
247 |
|
|
248 |
|
int |
249 |
|
pcre_fullinfo(const pcre *external_re, const pcre_extra *study_data, int what, |
250 |
|
void *where) |
251 |
|
{ |
252 |
|
const real_pcre *re = (const real_pcre *)external_re; |
253 |
|
const real_pcre_extra *study = (const real_pcre_extra *)study_data; |
254 |
|
|
255 |
|
if (re == NULL || where == NULL) return PCRE_ERROR_NULL; |
256 |
|
if (re->magic_number != MAGIC_NUMBER) return PCRE_ERROR_BADMAGIC; |
257 |
|
|
258 |
|
switch (what) |
259 |
|
{ |
260 |
|
case PCRE_INFO_OPTIONS: |
261 |
|
*((unsigned long int *)where) = re->options & PUBLIC_OPTIONS; |
262 |
|
break; |
263 |
|
|
264 |
|
case PCRE_INFO_SIZE: |
265 |
|
*((size_t *)where) = re->size; |
266 |
|
break; |
267 |
|
|
268 |
|
case PCRE_INFO_CAPTURECOUNT: |
269 |
|
*((int *)where) = re->top_bracket; |
270 |
|
break; |
271 |
|
|
272 |
|
case PCRE_INFO_BACKREFMAX: |
273 |
|
*((int *)where) = re->top_backref; |
274 |
|
break; |
275 |
|
|
276 |
|
case PCRE_INFO_FIRSTCHAR: |
277 |
|
*((int *)where) = |
278 |
|
((re->options & PCRE_FIRSTSET) != 0)? re->first_char : |
279 |
|
((re->options & PCRE_STARTLINE) != 0)? -1 : -2; |
280 |
|
break; |
281 |
|
|
282 |
|
case PCRE_INFO_FIRSTTABLE: |
283 |
|
*((const uschar **)where) = |
284 |
|
(study != NULL && (study->options & PCRE_STUDY_MAPPED) != 0)? |
285 |
|
study->start_bits : NULL; |
286 |
|
break; |
287 |
|
|
288 |
|
case PCRE_INFO_LASTLITERAL: |
289 |
|
*((int *)where) = |
290 |
|
((re->options & PCRE_REQCHSET) != 0)? re->req_char : -1; |
291 |
|
break; |
292 |
|
|
293 |
|
default: return PCRE_ERROR_BADOPTION; |
294 |
|
} |
295 |
|
|
296 |
|
return 0; |
297 |
|
} |
298 |
|
|
299 |
|
|
300 |
|
|
301 |
#ifdef DEBUG |
#ifdef DEBUG |
302 |
/************************************************* |
/************************************************* |
356 |
int options, BOOL isclass, compile_data *cd) |
int options, BOOL isclass, compile_data *cd) |
357 |
{ |
{ |
358 |
const uschar *ptr = *ptrptr; |
const uschar *ptr = *ptrptr; |
359 |
int c = *(++ptr) & 255; /* Ensure > 0 on signed-char systems */ |
int c, i; |
|
int i; |
|
360 |
|
|
361 |
|
c = *(++ptr) & 255; /* Ensure > 0 on signed-char systems */ |
362 |
if (c == 0) *errorptr = ERR1; |
if (c == 0) *errorptr = ERR1; |
363 |
|
|
364 |
/* Digits or letters may have special meaning; all others are literals. */ |
/* Digits or letters may have special meaning; all others are literals. */ |
723 |
|
|
724 |
|
|
725 |
/************************************************* |
/************************************************* |
726 |
|
* Check for POSIX class syntax * |
727 |
|
*************************************************/ |
728 |
|
|
729 |
|
/* This function is called when the sequence "[:" or "[." or "[=" is |
730 |
|
encountered in a character class. It checks whether this is followed by an |
731 |
|
optional ^ and then a sequence of letters, terminated by a matching ":]" or |
732 |
|
".]" or "=]". |
733 |
|
|
734 |
|
Argument: |
735 |
|
ptr pointer to the initial [ |
736 |
|
endptr where to return the end pointer |
737 |
|
cd pointer to compile data |
738 |
|
|
739 |
|
Returns: TRUE or FALSE |
740 |
|
*/ |
741 |
|
|
742 |
|
static BOOL |
743 |
|
check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd) |
744 |
|
{ |
745 |
|
int terminator; /* Don't combine these lines; the Solaris cc */ |
746 |
|
terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */ |
747 |
|
if (*(++ptr) == '^') ptr++; |
748 |
|
while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++; |
749 |
|
if (*ptr == terminator && ptr[1] == ']') |
750 |
|
{ |
751 |
|
*endptr = ptr; |
752 |
|
return TRUE; |
753 |
|
} |
754 |
|
return FALSE; |
755 |
|
} |
756 |
|
|
757 |
|
|
758 |
|
|
759 |
|
|
760 |
|
/************************************************* |
761 |
|
* Check POSIX class name * |
762 |
|
*************************************************/ |
763 |
|
|
764 |
|
/* This function is called to check the name given in a POSIX-style class entry |
765 |
|
such as [:alnum:]. |
766 |
|
|
767 |
|
Arguments: |
768 |
|
ptr points to the first letter |
769 |
|
len the length of the name |
770 |
|
|
771 |
|
Returns: a value representing the name, or -1 if unknown |
772 |
|
*/ |
773 |
|
|
774 |
|
static int |
775 |
|
check_posix_name(const uschar *ptr, int len) |
776 |
|
{ |
777 |
|
register int yield = 0; |
778 |
|
while (posix_name_lengths[yield] != 0) |
779 |
|
{ |
780 |
|
if (len == posix_name_lengths[yield] && |
781 |
|
strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield; |
782 |
|
yield++; |
783 |
|
} |
784 |
|
return -1; |
785 |
|
} |
786 |
|
|
787 |
|
|
788 |
|
|
789 |
|
|
790 |
|
/************************************************* |
791 |
* Compile one branch * |
* Compile one branch * |
792 |
*************************************************/ |
*************************************************/ |
793 |
|
|
930 |
goto FAILED; |
goto FAILED; |
931 |
} |
} |
932 |
|
|
933 |
|
/* Handle POSIX class names. Perl allows a negation extension of the |
934 |
|
form [:^name]. A square bracket that doesn't match the syntax is |
935 |
|
treated as a literal. We also recognize the POSIX constructions |
936 |
|
[.ch.] and [=ch=] ("collating elements") and fault them, as Perl |
937 |
|
5.6 does. */ |
938 |
|
|
939 |
|
if (c == '[' && |
940 |
|
(ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') && |
941 |
|
check_posix_syntax(ptr, &tempptr, cd)) |
942 |
|
{ |
943 |
|
BOOL local_negate = FALSE; |
944 |
|
int posix_class, i; |
945 |
|
register const uschar *cbits = cd->cbits; |
946 |
|
|
947 |
|
if (ptr[1] != ':') |
948 |
|
{ |
949 |
|
*errorptr = ERR31; |
950 |
|
goto FAILED; |
951 |
|
} |
952 |
|
|
953 |
|
ptr += 2; |
954 |
|
if (*ptr == '^') |
955 |
|
{ |
956 |
|
local_negate = TRUE; |
957 |
|
ptr++; |
958 |
|
} |
959 |
|
|
960 |
|
posix_class = check_posix_name(ptr, tempptr - ptr); |
961 |
|
if (posix_class < 0) |
962 |
|
{ |
963 |
|
*errorptr = ERR30; |
964 |
|
goto FAILED; |
965 |
|
} |
966 |
|
|
967 |
|
/* If matching is caseless, upper and lower are converted to |
968 |
|
alpha. This relies on the fact that the class table starts with |
969 |
|
alpha, lower, upper as the first 3 entries. */ |
970 |
|
|
971 |
|
if ((options & PCRE_CASELESS) != 0 && posix_class <= 2) |
972 |
|
posix_class = 0; |
973 |
|
|
974 |
|
/* Or into the map we are building up to 3 of the static class |
975 |
|
tables, or their negations. */ |
976 |
|
|
977 |
|
posix_class *= 3; |
978 |
|
for (i = 0; i < 3; i++) |
979 |
|
{ |
980 |
|
int taboffset = posix_class_maps[posix_class + i]; |
981 |
|
if (taboffset < 0) break; |
982 |
|
if (local_negate) |
983 |
|
for (c = 0; c < 32; c++) class[c] |= ~cbits[c+taboffset]; |
984 |
|
else |
985 |
|
for (c = 0; c < 32; c++) class[c] |= cbits[c+taboffset]; |
986 |
|
} |
987 |
|
|
988 |
|
ptr = tempptr + 1; |
989 |
|
class_charcount = 10; /* Set > 1; assumes more than 1 per class */ |
990 |
|
continue; |
991 |
|
} |
992 |
|
|
993 |
/* Backslash may introduce a single character, or it may introduce one |
/* Backslash may introduce a single character, or it may introduce one |
994 |
of the specials, which just set a flag. Escaped items are checked for |
of the specials, which just set a flag. Escaped items are checked for |
995 |
validity in the pre-compiling pass. The sequence \b is a special case. |
validity in the pre-compiling pass. The sequence \b is a special case. |
1017 |
continue; |
continue; |
1018 |
|
|
1019 |
case ESC_w: |
case ESC_w: |
1020 |
for (c = 0; c < 32; c++) |
for (c = 0; c < 32; c++) class[c] |= cbits[c+cbit_word]; |
|
class[c] |= (cbits[c+cbit_digit] | cbits[c+cbit_word]); |
|
1021 |
continue; |
continue; |
1022 |
|
|
1023 |
case ESC_W: |
case ESC_W: |
1024 |
for (c = 0; c < 32; c++) |
for (c = 0; c < 32; c++) class[c] |= ~cbits[c+cbit_word]; |
|
class[c] |= ~(cbits[c+cbit_digit] | cbits[c+cbit_word]); |
|
1025 |
continue; |
continue; |
1026 |
|
|
1027 |
case ESC_s: |
case ESC_s: |
1584 |
ptr++; |
ptr++; |
1585 |
break; |
break; |
1586 |
|
|
1587 |
|
case 'R': /* Pattern recursion */ |
1588 |
|
*code++ = OP_RECURSE; |
1589 |
|
ptr++; |
1590 |
|
continue; |
1591 |
|
|
1592 |
default: /* Option setting */ |
default: /* Option setting */ |
1593 |
set = unset = 0; |
set = unset = 0; |
1594 |
optset = &set; |
optset = &set; |
2244 |
real_pcre *re; |
real_pcre *re; |
2245 |
int length = 3; /* For initial BRA plus length */ |
int length = 3; /* For initial BRA plus length */ |
2246 |
int runlength; |
int runlength; |
2247 |
int c, size, reqchar, countlits; |
int c, reqchar, countlits; |
2248 |
int bracount = 0; |
int bracount = 0; |
2249 |
int top_backref = 0; |
int top_backref = 0; |
2250 |
int branch_extra = 0; |
int branch_extra = 0; |
2251 |
int branch_newextra; |
int branch_newextra; |
2252 |
unsigned int brastackptr = 0; |
unsigned int brastackptr = 0; |
2253 |
|
size_t size; |
2254 |
uschar *code; |
uschar *code; |
2255 |
const uschar *ptr; |
const uschar *ptr; |
2256 |
compile_data compile_block; |
compile_data compile_block; |
2478 |
ptr += 2; |
ptr += 2; |
2479 |
break; |
break; |
2480 |
|
|
2481 |
|
/* A recursive call to the regex is an extension, to provide the |
2482 |
|
facility which can be obtained by $(?p{perl-code}) in Perl 5.6. */ |
2483 |
|
|
2484 |
|
case 'R': |
2485 |
|
if (ptr[3] != ')') |
2486 |
|
{ |
2487 |
|
*errorptr = ERR29; |
2488 |
|
goto PCRE_ERROR_RETURN; |
2489 |
|
} |
2490 |
|
ptr += 3; |
2491 |
|
length += 1; |
2492 |
|
break; |
2493 |
|
|
2494 |
/* Lookbehinds are in Perl from version 5.005 */ |
/* Lookbehinds are in Perl from version 5.005 */ |
2495 |
|
|
2496 |
case '<': |
case '<': |
2793 |
return NULL; |
return NULL; |
2794 |
} |
} |
2795 |
|
|
2796 |
/* Put in the magic number and the options. */ |
/* Put in the magic number, and save the size, options, and table pointer */ |
2797 |
|
|
2798 |
re->magic_number = MAGIC_NUMBER; |
re->magic_number = MAGIC_NUMBER; |
2799 |
|
re->size = size; |
2800 |
re->options = options; |
re->options = options; |
2801 |
re->tables = tables; |
re->tables = tables; |
2802 |
|
|
3391 |
ecode += 3; |
ecode += 3; |
3392 |
break; |
break; |
3393 |
|
|
3394 |
|
/* Recursion matches the current regex, nested. If there are any capturing |
3395 |
|
brackets started but not finished, we have to save their starting points |
3396 |
|
and reinstate them after the recursion. However, we don't know how many |
3397 |
|
such there are (offset_top records the completed total) so we just have |
3398 |
|
to save all the potential data. There may be up to 99 such values, which |
3399 |
|
is a bit large to put on the stack, but using malloc for small numbers |
3400 |
|
seems expensive. As a compromise, the stack is used when there are fewer |
3401 |
|
than 16 values to store; otherwise malloc is used. A problem is what to do |
3402 |
|
if the malloc fails ... there is no way of returning to the top level with |
3403 |
|
an error. Save the top 15 values on the stack, and accept that the rest |
3404 |
|
may be wrong. */ |
3405 |
|
|
3406 |
|
case OP_RECURSE: |
3407 |
|
{ |
3408 |
|
BOOL rc; |
3409 |
|
int *save; |
3410 |
|
int stacksave[15]; |
3411 |
|
|
3412 |
|
c = md->offset_max; |
3413 |
|
|
3414 |
|
if (c < 16) save = stacksave; else |
3415 |
|
{ |
3416 |
|
save = (int *)(pcre_malloc)((c+1) * sizeof(int)); |
3417 |
|
if (save == NULL) |
3418 |
|
{ |
3419 |
|
save = stacksave; |
3420 |
|
c = 15; |
3421 |
|
} |
3422 |
|
} |
3423 |
|
|
3424 |
|
for (i = 1; i <= c; i++) |
3425 |
|
save[i] = md->offset_vector[md->offset_end - i]; |
3426 |
|
rc = match(eptr, md->start_pattern, offset_top, md, ims, FALSE, eptrb); |
3427 |
|
for (i = 1; i <= c; i++) |
3428 |
|
md->offset_vector[md->offset_end - i] = save[i]; |
3429 |
|
if (save != stacksave) (pcre_free)(save); |
3430 |
|
if (!rc) return FALSE; |
3431 |
|
|
3432 |
|
/* In case the recursion has set more capturing values, save the final |
3433 |
|
number, then move along the subject till after the recursive match, |
3434 |
|
and advance one byte in the pattern code. */ |
3435 |
|
|
3436 |
|
offset_top = md->end_offset_top; |
3437 |
|
eptr = md->end_match_ptr; |
3438 |
|
ecode++; |
3439 |
|
} |
3440 |
|
break; |
3441 |
|
|
3442 |
/* "Once" brackets are like assertion brackets except that after a match, |
/* "Once" brackets are like assertion brackets except that after a match, |
3443 |
the point in the subject string is not moved back. Thus there can never be |
the point in the subject string is not moved back. Thus there can never be |
4507 |
(offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL; |
(offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL; |
4508 |
if (re->magic_number != MAGIC_NUMBER) return PCRE_ERROR_BADMAGIC; |
if (re->magic_number != MAGIC_NUMBER) return PCRE_ERROR_BADMAGIC; |
4509 |
|
|
4510 |
|
match_block.start_pattern = re->code; |
4511 |
match_block.start_subject = (const uschar *)subject; |
match_block.start_subject = (const uschar *)subject; |
4512 |
match_block.end_subject = match_block.start_subject + length; |
match_block.end_subject = match_block.start_subject + length; |
4513 |
end_subject = match_block.end_subject; |
end_subject = match_block.end_subject; |