-const char pcrs_rcs[] = "$Id: pcrs.c,v 1.1 2001/05/13 21:57:07 administrator Exp $";
-
/*********************************************************************
*
- * File : $Source: /home/administrator/cvs/ijb/pcrs.c,v $
+ * File : $Source: /cvsroot/ijbswa/current/pcrs.c,v $
*
- * Purpose : This is the pre-pre-alpha realease of libpcrs. It is only
- * published at this (ugly) stage of development, because it is
- * needed for a new feature in JunkBuster.
+ * Purpose : pcrs is a supplement to the pcre library by Philip Hazel
+ * <ph10@cam.ac.uk> and adds Perl-style substitution. That
+ * is, it mimics Perl's 's' operator. See pcrs(3) for details.
*
- * Apart from the code being quite a mess, no inconsistencies,
- * memory leaks or functional bugs **should** be present.
+ * WARNING: This file contains additional functions and bug
+ * fixes that aren't part of the latest official pcrs package
+ * (which apparently is no longer maintained).
*
- * While you ROTFL at the code, you could just as well mail me
- * (andreas@oesterhelt.org) with advice for improvement.
+ * Copyright : Written and Copyright (C) 2000, 2001 by Andreas S. Oesterhelt
+ * <andreas@oesterhelt.org>
*
- * pcrs is a supplement to the brilliant pcre library by Philip
- * Hazel (ph10@cam.ac.uk) and adds Perl-style substitution. That
- * is, it mimics Perl's 's' operator.
+ * Copyright (C) 2006, 2007 Fabian Keil <fk@fabiankeil.de>
*
- * Currently, there's no documentation besides comments and the
- * source itself ;-)
+ * This program is free software; you can redistribute it
+ * and/or modify it under the terms of the GNU General
+ * Public License as published by the Free Software
+ * Foundation; either version 2 of the License, or (at
+ * your option) any later version.
*
- * Copyright : Written and copyright by andreas@oesterhelt.org
+ * This program is distributed in the hope that it will
+ * be useful, but WITHOUT ANY WARRANTY; without even the
+ * implied warranty of MERCHANTABILITY or FITNESS FOR A
+ * PARTICULAR PURPOSE. See the GNU General Public
+ * License for more details.
*
- * Revisions :
- * $Log: pcrs.c,v $
+ * The GNU General Public License should be included with
+ * this file. If not, you can view it at
+ * http://www.gnu.org/copyleft/gpl.html
+ * or write to the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*
*********************************************************************/
-\f
-#include <pcre.h>
+
#include <string.h>
+#include <ctype.h>
+#include <assert.h>
+
+/*
+ * Include project.h just so that the right pcre.h gets
+ * included from there
+ */
+#include "project.h"
+
+/* For snprintf only */
+#include "miscutil.h"
+/* For xtoi */
+#include "encode.h"
+
#include "pcrs.h"
-const char pcrs_h_rcs[] = PCRS_H_VERSION;
+/*
+ * Internal prototypes
+ */
+
+static int pcrs_parse_perl_options(const char *optstring, unsigned int *flags);
+static pcrs_substitute *pcrs_compile_replacement(const char *replacement, int trivialflag,
+ int capturecount, int *errptr);
+static int is_hex_sequence(const char *sequence);
/*********************************************************************
*
- * Function : my_strsep
+ * Function : pcrs_strerror
*
- * Description : Convenience function. It acts like strsep, except that
- * it respects quoting of the delimiter character with the
- * quote character. (And, of course, quoting the quote char
- * with itself.) Called from `pcrs_make_job'.
+ * Description : Return a string describing a given error code.
*
* Parameters :
- * 1 : token = current token
- * 2 : text = string to tokenize
- * 3 : delimiter = single character deliminter
- * 4 : quote_char = character to cause quoting
+ * 1 : error = the error code
*
- * Returns : -1 => failure, else the length of the token found.
- * In the latter case, *text is the token's start.
+ * Returns : char * to the descriptive string
*
*********************************************************************/
-int my_strsep(char *token, char **text, char delimiter, char quote_char)
+const char *pcrs_strerror(const int error)
{
- int i, k=0, limit, quoted = FALSE;
+ static char buf[100];
- limit = strlen(*text);
- if ( 0 == limit )
+ if (error != 0)
{
- return -1;
- }
+ switch (error)
+ {
+ /* Passed-through PCRE error: */
+ case PCREn(ERROR_NOMEMORY): return "(pcre:) No memory";
- token[0] = '\0';
+ /* Shouldn't happen unless PCRE or PCRS bug, or user messed with compiled job: */
+ case PCREn(ERROR_NULL): return "(pcre:) NULL code or subject or ovector";
+ case PCREn(ERROR_BADOPTION): return "(pcre:) Unrecognized option bit";
+ case PCREn(ERROR_BADMAGIC): return "(pcre:) Bad magic number in code";
+#if defined(PCRE_ERROR_UNKNOWN_NODE)
+ case PCRE_ERROR_UNKNOWN_NODE: return "(pcre:) Bad node in pattern";
+#endif
+ /* Can't happen / not passed: */
+ case PCREn(ERROR_NOSUBSTRING): return "(pcre:) Fire in power supply";
+ case PCREn(ERROR_NOMATCH): return "(pcre:) Water in power supply";
- for (i=0; i < limit; i++)
- {
- if (text[0][i] == delimiter && !quoted)
- {
- *text += 1;
- break;
- }
- else if (text[0][i] == quote_char && !quoted && i+1 < limit && text[0][i+1] == delimiter)
- {
- quoted = TRUE;
- continue;
+#ifdef PCRE_ERROR_MATCHLIMIT
+ /*
+ * Only reported by PCRE versions newer than our own.
+ */
+ case PCREn(ERROR_MATCHLIMIT): return "(pcre:) Match limit reached";
+#endif /* def PCRE_ERROR_MATCHLIMIT */
+ /* PCRS errors: */
+ case PCRS_ERR_NOMEM: return "(pcrs:) No memory";
+ case PCRS_ERR_CMDSYNTAX: return "(pcrs:) Syntax error while parsing command";
+ case PCRS_ERR_STUDY: return "(pcrs:) PCRE error while studying the pattern";
+ case PCRS_ERR_BADJOB: return "(pcrs:) Bad job - NULL job, pattern or substitute";
+ case PCRS_WARN_BADREF: return "(pcrs:) Backreference out of range";
+ case PCRS_WARN_TRUNCATION:
+ return "(pcrs:) At least one variable was too big and has been truncated before compilation";
+
+ default:
+#ifdef HAVE_PCRE2
+ pcre2_get_error_message(error, (PCRE2_UCHAR8*)buf, sizeof(buf));
+#else
+ snprintf(buf, sizeof(buf),
+ "Error code %d. For details, check the pcre documentation.",
+ error);
+#endif
+ return buf;
}
- token[k++] = text[0][i];
- quoted = FALSE;
}
- token[k] = '\0';
- *text += i;
- return k;
+ /* error >= 0: No error */
+ return "(pcrs:) Everything's just fine. Thanks for asking.";
}
/*********************************************************************
*
- * Function : pcrs_compile_perl_options
+ * Function : pcrs_parse_perl_options
*
* Description : This function parses a string containing the options to
* Perl's s/// operator. It returns an integer that is the
* pcre equivalent of the symbolic optstring.
- * Since pcre doesn't know about Perl's 'g' (global) option,
- * but pcrs needs it, the globalflag integer is set if 'g'
- * is encountered.
+ * Since pcre doesn't know about Perl's 'g' (global) or pcrs',
+ * 'T' (trivial) options but pcrs needs them, the corresponding
+ * flags are set if 'g'or 'T' is encountered.
+ * Note: The 'T' and 'U' options do not conform to Perl.
*
* Parameters :
* 1 : optstring = string with options in perl syntax
- * 2 : globalflag = see description
+ * 2 : flags = see description
*
- * Returns : option integer suitable for pcre
+ * Returns : option integer suitable for pcre
*
*********************************************************************/
-int pcrs_compile_perl_options(char *optstring, int *globalflag)
+static int pcrs_parse_perl_options(const char *optstring, unsigned int *flags)
{
- int i, rc = 0;
- *globalflag = 0;
- for (i=0; i < strlen(optstring); i++)
+ size_t i;
+ int rc = 0;
+ *flags = 0;
+
+ if (NULL == optstring) return 0;
+
+ for (i = 0; i < strlen(optstring); i++)
{
switch(optstring[i])
{
- case 'e': break;
- case 'g': *globalflag = 1; break;
- case 'i': rc |= PCRE_CASELESS; break;
- case 'm': rc |= PCRE_MULTILINE; break;
+ case 'e': break; /* ToDo ;-) */
+ case 'g': *flags |= PCRS_GLOBAL; break;
+ case 'i': rc |= PCREn(CASELESS); break;
+ case 'm': rc |= PCREn(MULTILINE); break;
case 'o': break;
- case 's': rc |= PCRE_DOTALL; break;
- case 'x': rc |= PCRE_EXTENDED; break;
- default: break;
+ case 's': rc |= PCREn(DOTALL); break;
+ case 'x': rc |= PCREn(EXTENDED); break;
+ case 'D': *flags |= PCRS_DYNAMIC; break;
+ case 'U': rc |= PCREn(UNGREEDY); break;
+ case 'T': *flags |= PCRS_TRIVIAL; break;
+ default: break;
}
}
return rc;
}
+#ifdef FUZZ
+/*********************************************************************
+ *
+ * Function : pcrs_compile_fuzzed_replacement
+ *
+ * Description : Wrapper around pcrs_compile_replacement() for
+ * fuzzing purposes.
+ *
+ * Parameters :
+ * 1 : replacement = replacement part of s/// operator
+ * in perl syntax
+ * 2 : errptr = pointer to an integer in which error
+ * conditions can be returned.
+ *
+ * Returns : pcrs_substitute data structure, or NULL if an
+ * error is encountered. In that case, *errptr has
+ * the reason.
+ *
+ *********************************************************************/
+extern pcrs_substitute *pcrs_compile_fuzzed_replacement(const char *replacement, int *errptr)
+{
+ int capturecount = PCRS_MAX_SUBMATCHES; /* XXX: fuzzworthy? */
+ int trivial_flag = 0; /* We don't want to fuzz strncpy() */
+
+ *errptr = 0; /* XXX: Should pcrs_compile_replacement() do this? */
+
+ return pcrs_compile_replacement(replacement, trivial_flag, capturecount, errptr);
+
+}
+#endif
+
+
/*********************************************************************
*
* Function : pcrs_compile_replacement
* Parameters :
* 1 : replacement = replacement part of s/// operator
* in perl syntax
- * 2 : errptr = pointer to an integer in which error
+ * 2 : trivialflag = Flag that causes backreferences to be
+ * ignored.
+ * 3 : capturecount = Number of capturing subpatterns in
+ * the pattern. Needed for $+ handling.
+ * 4 : errptr = pointer to an integer in which error
* conditions can be returned.
*
* Returns : pcrs_substitute data structure, or NULL if an
* the reason.
*
*********************************************************************/
-pcrs_substitute *pcrs_compile_replacement(char *replacement, int *errptr)
+static pcrs_substitute *pcrs_compile_replacement(const char *replacement, int trivialflag, int capturecount, int *errptr)
{
- int length, i, k = 0, l = 0, quoted = 0, idx;
- char *text, *num_ptr, *numbers = "0123456789";
+ int i, k, l, quoted;
+ char *text;
pcrs_substitute *r;
+#ifndef FUZZ
+ size_t length;
+#else
+ static size_t length;
+#endif
+ i = k = l = quoted = 0;
- r = (pcrs_substitute *)malloc(sizeof(pcrs_substitute));
- if (r == NULL) return NULL;
- memset(r, '\0', sizeof(pcrs_substitute));
+ /*
+ * Sanity check
+ */
+ if (NULL == replacement)
+ {
+ replacement = "";
+ }
- text = strdup(replacement); /* must be free()d by caller */
- if (text == NULL)
+ /*
+ * Get memory or fail
+ */
+ if (NULL == (r = (pcrs_substitute *)malloc(sizeof(pcrs_substitute))))
{
*errptr = PCRS_ERR_NOMEM;
- free(r);
return NULL;
}
+ memset(r, '\0', sizeof(pcrs_substitute));
length = strlen(replacement);
- for (i=0; i < length; i++)
+ if (NULL == (text = (char *)malloc(length + 1)))
{
- /* Backslash treatment */
- if (replacement[i] == '\\')
- {
- if (quoted)
- {
- text[k++] = replacement[i];
- quoted = 0;
- }
- else
- {
- quoted = 1;
- }
- continue;
- }
+ free(r);
+ *errptr = PCRS_ERR_NOMEM;
+ return NULL;
+ }
+ memset(text, '\0', length + 1);
+
- /* Dollar treatment */
- if (replacement[i] == '$' && !quoted && i < length - 1)
+ /*
+ * In trivial mode, just copy the substitute text
+ */
+ if (trivialflag)
+ {
+ strlcpy(text, replacement, length + 1);
+ k = (int)length;
+ }
+
+ /*
+ * Else, parse, cut out and record all backreferences
+ */
+ else
+ {
+ while (i < (int)length)
{
- if (strchr("0123456789&", replacement[i + 1]) == NULL)
+ /* Quoting */
+ if (replacement[i] == '\\')
{
- text[k++] = replacement[i];
+ if (quoted)
+ {
+ text[k++] = replacement[i++];
+ quoted = 0;
+ }
+ else
+ {
+ if (replacement[i+1] && strchr("tnrfae0", replacement[i+1]))
+ {
+ switch (replacement[++i])
+ {
+ case 't':
+ text[k++] = '\t';
+ break;
+ case 'n':
+ text[k++] = '\n';
+ break;
+ case 'r':
+ text[k++] = '\r';
+ break;
+ case 'f':
+ text[k++] = '\f';
+ break;
+ case 'a':
+ text[k++] = 7;
+ break;
+ case 'e':
+ text[k++] = 27;
+ break;
+ case '0':
+ text[k++] = '\0';
+ break;
+ }
+ i++;
+ }
+ else if (is_hex_sequence(&replacement[i]))
+ {
+ /*
+ * Replace a hex sequence with a single
+ * character with the sequence's ascii value.
+ * e.g.: '\x7e' => '~'
+ */
+ const int ascii_value = xtoi(&replacement[i+2]);
+
+ assert(ascii_value >= 0);
+ assert(ascii_value < 256);
+ text[k++] = (char)ascii_value;
+ i += 4;
+ }
+ else
+ {
+ quoted = 1;
+ i++;
+ }
+ }
+ continue;
}
- else
+
+ /* Backreferences */
+ if (replacement[i] == '$' && !quoted && i < (int)(length - 1))
{
- r->block_length[l] = k - r->block_offset[l];
- r->backref[l] = 0;
- if (replacement[i + 1] != '&')
+ char *symbol, symbols[] = "'`+&";
+ if (l >= PCRS_MAX_SUBMATCHES)
{
- while ((num_ptr = strchr(numbers, replacement[++i])) != NULL && i < length)
+ freez(text);
+ freez(r);
+ *errptr = PCRS_WARN_BADREF;
+ return NULL;
+ }
+ r->block_length[l] = (size_t)(k - r->block_offset[l]);
+
+ /* Numerical backreferences */
+ if (isdigit((int)replacement[i + 1]))
+ {
+ while (i < (int)length && isdigit((int)replacement[++i]))
+ {
+ r->backref[l] = r->backref[l] * 10 + replacement[i] - 48;
+ }
+ if (r->backref[l] > capturecount)
{
- idx = num_ptr - numbers;
- r->backref[l] = r->backref[l] * 10 + idx;
+ freez(text);
+ freez(r);
+ *errptr = PCRS_WARN_BADREF;
+ return NULL;
}
- i--;
}
+
+ /* Symbolic backreferences: */
+ else if (NULL != (symbol = strchr(symbols, replacement[i + 1])))
+ {
+
+ if (symbol - symbols == 2) /* $+ */
+ {
+ r->backref[l] = capturecount;
+ }
+ else if (symbol - symbols == 3) /* $& */
+ {
+ r->backref[l] = 0;
+ }
+ else /* $' or $` */
+ {
+ r->backref[l] = (int)(PCRS_MAX_SUBMATCHES + 1 - (symbol - symbols));
+ }
+ i += 2;
+ }
+
+ /* Invalid backref -> plain '$' */
else
- i++;
- if (r->backref[l] < PCRS_MAX_SUBMATCHES)
+ {
+ goto plainchar;
+ }
+
+ assert(r->backref[l] < PCRS_MAX_SUBMATCHES + 2);
+ /* Valid and in range? -> record */
+ if ((0 <= r->backref[l]) &&
+ (r->backref[l] < PCRS_MAX_SUBMATCHES + 2) &&
+ (l < PCRS_MAX_SUBMATCHES - 1))
+ {
r->backref_count[r->backref[l]] += 1;
- l++;
- r->block_offset[l] = k;
+ r->block_offset[++l] = k;
+ }
+ else
+ {
+ freez(text);
+ freez(r);
+ *errptr = PCRS_WARN_BADREF;
+ return NULL;
+ }
+ continue;
}
- continue;
- }
- /* Plain char treatment */
- text[k++] = replacement[i];
- quoted = 0;
- }
+plainchar:
+ /* Plain chars are copied */
+ text[k++] = replacement[i++];
+ quoted = 0;
+ }
+ } /* -END- if (!trivialflag) */
- text[k] = '\0';
+ /*
+ * Finish & return
+ */
r->text = text;
r->backrefs = l;
- r->block_length[l] = k - r->block_offset[l];
+ r->length = (size_t)k;
+ r->block_length[l] = (size_t)(k - r->block_offset[l]);
+
return r;
}
* Function : pcrs_free_job
*
* Description : Frees the memory used by a pcrs_job struct and its
- * dependant structures. Returns a pointer to the next
- * job, if there was any, or NULL otherwise.
+ * dependent structures.
*
* Parameters :
* 1 : job = pointer to the pcrs_job structure to be freed
*
- * Returns : a pointer to the next job, if there was any, or
- * NULL otherwise.
+ * Returns : a pointer to the next job, if there was any, or
+ * NULL otherwise.
*
*********************************************************************/
pcrs_job *pcrs_free_job(pcrs_job *job)
else
{
next = job->next;
- if (job->pattern != NULL) free(job->pattern);
- if (job->hints != NULL) free(job->hints);
+ if (job->pattern != NULL)
+ {
+#ifdef HAVE_PCRE2
+ pcre2_code_free(job->pattern);
+#else
+ free(job->pattern);
+#endif
+ }
+#ifndef HAVE_PCRE2
+ if (job->hints != NULL)
+ {
+#ifdef PCRE_CONFIG_JIT
+ pcre_free_study(job->hints);
+#else
+ free(job->hints);
+#endif
+ }
+#endif
if (job->substitute != NULL)
{
if (job->substitute->text != NULL) free(job->substitute->text);
/*********************************************************************
*
- * Function : pcrs_make_job
+ * Function : pcrs_free_joblist
*
- * Description : Main entry point. Takes a string with a Perl-style
- * s/// command and returns a corresponding pcrs_job,
- * or NULL if compiling the job fails at any stage.
- * Diagnostics could obviously be improved.
+ * Description : Iterates through a chained list of pcrs_job's and
+ * frees them using pcrs_free_job.
+ *
+ * Parameters :
+ * 1 : joblist = pointer to the first pcrs_job structure to
+ * be freed
+ *
+ * Returns : N/A
+ *
+ *********************************************************************/
+void pcrs_free_joblist(pcrs_job *joblist)
+{
+ while (NULL != (joblist = pcrs_free_job(joblist))) {};
+
+ return;
+
+}
+
+
+/*********************************************************************
+ *
+ * Function : pcrs_compile_command
+ *
+ * Description : Parses a string with a Perl-style s/// command,
+ * calls pcrs_compile, and returns a corresponding
+ * pcrs_job, or NULL if parsing or compiling the job
+ * fails.
*
* Parameters :
* 1 : command = string with perl-style s/// command
* has the reason.
*
*********************************************************************/
-pcrs_job *pcrs_make_job(char *command, int *errptr)
+pcrs_job *pcrs_compile_command(const char *command, int *errptr)
{
- char *dummy, *token, delimiter;
- const char *error;
- int i = 0, globalflag;
+ int i, k, l, quoted = FALSE;
+ size_t limit;
+ char delimiter;
+ char *tokens[4];
pcrs_job *newjob;
- /* Get and init memory */
- if ((newjob = (pcrs_job *)malloc(sizeof(pcrs_job))) == NULL)
- {
- *errptr = PCRS_ERR_NOMEM;
- return NULL;
- }
- memset(newjob, '\0', sizeof(pcrs_job));
+ k = l = 0;
- /* Command too short? */
- if (strlen(command) < 4)
+ /*
+ * Tokenize the perl command
+ */
+ limit = strlen(command);
+ if (limit < 4)
{
*errptr = PCRS_ERR_CMDSYNTAX;
- pcrs_free_job(newjob);
return NULL;
}
+ else
+ {
+ delimiter = command[1];
+ }
+
+ tokens[l] = (char *) malloc(limit + 1);
- /* Split command into tokens and handle them */
- delimiter = command[1];
- token = (char *)malloc(strlen(command)); /* current token */
- dummy = (char *)malloc(strlen(command)); /* must store pattern, since we can't */
- /* use it until the options are known */
- while (my_strsep(token, &command, delimiter, '\\') >= 0)
+ for (i = 0; i <= (int)limit; i++)
{
- switch (i)
+
+ if (command[i] == delimiter && !quoted)
{
- /* We don't care about the command and assume 's' */
- case 0:
+ if (l == 3)
+ {
+ l = -1;
break;
+ }
+ tokens[0][k++] = '\0';
+ tokens[++l] = tokens[0] + k;
+ continue;
+ }
- /* The pattern */
- case 1:
- strcpy(dummy, token);
- break;
+ else if (command[i] == '\\' && !quoted)
+ {
+ quoted = TRUE;
+ if (command[i+1] == delimiter) continue;
+ }
+ else
+ {
+ quoted = FALSE;
+ }
+ tokens[0][k++] = command[i];
+ }
- /* The substitute */
- case 2:
- newjob->substitute = pcrs_compile_replacement(token, errptr);
- if (newjob->substitute == NULL)
- {
- pcrs_free_job(newjob);
- return NULL;
- }
- break;
+ /*
+ * Syntax error ?
+ */
+ if (l != 3)
+ {
+ *errptr = PCRS_ERR_CMDSYNTAX;
+ free(tokens[0]);
+ return NULL;
+ }
- /* The options */
- case 3:
- newjob->options = pcrs_compile_perl_options(token, &globalflag);
- newjob->globalflag = globalflag;
- break;
+ newjob = pcrs_compile(tokens[1], tokens[2], tokens[3], errptr);
+ free(tokens[0]);
+ return newjob;
- /* There shouldn't be anything else! */
- default:
- *errptr = PCRS_ERR_CMDSYNTAX;
- pcrs_free_job(newjob);
- return NULL;
- }
- i++;
+}
+
+
+/*********************************************************************
+ *
+ * Function : pcrs_compile
+ *
+ * Description : Takes the three arguments to a perl s/// command
+ * and compiles a pcrs_job structure from them.
+ *
+ * Parameters :
+ * 1 : pattern = string with perl-style pattern
+ * 2 : substitute = string with perl-style substitute
+ * 3 : options = string with perl-style options
+ * 4 : errptr = pointer to an integer in which error
+ * conditions can be returned.
+ *
+ * Returns : a corresponding pcrs_job data structure, or NULL
+ * if an error was encountered. In that case, *errptr
+ * has the reason.
+ *
+ *********************************************************************/
+pcrs_job *pcrs_compile(const char *pattern, const char *substitute, const char *options, int *errptr)
+{
+ pcrs_job *newjob;
+ unsigned int flags;
+ int capturecount;
+#ifdef HAVE_PCRE2
+ int ret;
+#else
+ int pcre_study_options = 0;
+ const char *error;
+#endif
+
+ *errptr = 0;
+
+ /*
+ * Handle NULL arguments
+ */
+ if (pattern == NULL) pattern = "";
+ if (substitute == NULL) substitute = "";
+
+
+ /*
+ * Get and init memory
+ */
+ if (NULL == (newjob = (pcrs_job *)malloc(sizeof(pcrs_job))))
+ {
+ *errptr = PCRS_ERR_NOMEM;
+ return NULL;
}
- free(token);
+ memset(newjob, '\0', sizeof(pcrs_job));
+
- /* Compile the pattern */
- newjob->pattern = pcre_compile(dummy, newjob->options, &error, errptr, NULL);
+ /*
+ * Evaluate the options
+ */
+ newjob->options = pcrs_parse_perl_options(options, &flags);
+ newjob->flags = flags;
+
+
+ /*
+ * Compile the pattern
+ */
+#ifdef HAVE_PCRE2
+ PCRE2_SIZE error_offset;
+ newjob->pattern = pcre2_compile((const unsigned char *)pattern,
+ PCRE2_ZERO_TERMINATED, (unsigned)newjob->options, errptr,
+ &error_offset, NULL);
+#else
+ newjob->pattern = pcre_compile(pattern, newjob->options, &error, errptr, NULL);
+#endif
if (newjob->pattern == NULL)
{
pcrs_free_job(newjob);
return NULL;
}
- free(dummy);
+#if defined(PCRE_STUDY_JIT_COMPILE) || defined(HAVE_PCRE2)
+#ifdef DISABLE_PCRE_JIT_COMPILATION
+#warning PCRE_STUDY_JIT_COMPILE is supported but Privoxy has been configured not to use it
+#else
+ if (!(flags & PCRS_DYNAMIC))
+ {
+#ifdef HAVE_PCRE2
+ /* Try to enable JIT compilation but continue if it's unsupported. */
+ if ((ret = pcre2_jit_compile(newjob->pattern, PCRE2_JIT_COMPLETE)) &&
+ (ret != PCRE2_ERROR_JIT_BADOPTION))
+ {
+ *errptr = ret;
+ pcrs_free_job(newjob);
+ return NULL;
+ }
+#else
+ pcre_study_options = PCRE_STUDY_JIT_COMPILE;
+#endif
+ }
+#endif
+#endif
+
+#ifndef HAVE_PCRE2
/*
* Generate hints. This has little overhead, since the
* hints will be NULL for a boring pattern anyway.
*/
- newjob->hints = pcre_study(newjob->pattern, 0, &error);
+ newjob->hints = pcre_study(newjob->pattern, pcre_study_options, &error);
if (error != NULL)
{
*errptr = PCRS_ERR_STUDY;
pcrs_free_job(newjob);
return NULL;
}
+#endif
+
+ /*
+ * Determine the number of capturing subpatterns.
+ * This is needed for handling $+ in the substitute.
+ */
+#ifdef HAVE_PCRE2
+ if (0 > (*errptr = pcre2_pattern_info(newjob->pattern, PCRE2_INFO_CAPTURECOUNT, &capturecount)))
+#else
+ if (0 > (*errptr = pcre_fullinfo(newjob->pattern, newjob->hints, PCRE_INFO_CAPTURECOUNT, &capturecount)))
+#endif
+ {
+ pcrs_free_job(newjob);
+ return NULL;
+ }
+
+
+ /*
+ * Compile the substitute
+ */
+ if (NULL == (newjob->substitute = pcrs_compile_replacement(substitute, newjob->flags & PCRS_TRIVIAL, capturecount, errptr)))
+ {
+ pcrs_free_job(newjob);
+ return NULL;
+ }
return newjob;
/*********************************************************************
*
- * Function : create_pcrs_job
+ * Function : pcrs_execute_list
+ *
+ * Description : This is a multiple job wrapper for pcrs_execute().
+ * Apply the regular substitutions defined by the jobs in
+ * the joblist to the subject.
+ * The subject itself is left untouched, memory for the result
+ * is malloc()ed and it is the caller's responsibility to free
+ * the result when it's no longer needed.
+ *
+ * Note: For convenient string handling, a null byte is
+ * appended to the result. It does not count towards the
+ * result_length, though.
*
- * Description : Create a job from all its components, if you don't
- * have a Perl command to start from. Rather boring.
*
* Parameters :
- * 1 : pattern = pointer to pcre pattern
- * 2 : hints = pointer to pcre hints
- * 3 : options = options in pcre format
- * 4 : globalflag = flag that indicates if global matching is desired
- * 5 : substitute = pointer to pcrs_substitute data structure
- * 2 : errptr = pointer to an integer in which error
- * conditions can be returned.
+ * 1 : joblist = the chained list of pcrs_jobs to be executed
+ * 2 : subject = the subject string
+ * 3 : subject_length = the subject's length
+ * 4 : result = char** for returning the result
+ * 5 : result_length = size_t* for returning the result's length
*
- * Returns : pcrs_job structure, or NULL if an error was encountered.
- * In that case, *errptr has the reason why.
+ * Returns : On success, the number of substitutions that were made.
+ * May be > 1 if job->flags contained PCRS_GLOBAL
+ * On failure, the (negative) pcre error code describing the
+ * failure, which may be translated to text using pcrs_strerror().
*
*********************************************************************/
-pcrs_job *create_pcrs_job(pcre *pattern, pcre_extra *hints, int options, int globalflag, pcrs_substitute *substitute, int *errptr)
+int pcrs_execute_list(pcrs_job *joblist, char *subject, size_t subject_length, char **result, size_t *result_length)
{
- pcrs_job *newjob;
+ pcrs_job *job;
+ char *old, *new = NULL;
+ int hits, total_hits;
+
+ old = subject;
+ *result_length = subject_length;
+ total_hits = 0;
- if ((newjob = (pcrs_job *)malloc(sizeof(pcrs_job))) == NULL)
+ for (job = joblist; job != NULL; job = job->next)
{
- *errptr = PCRS_ERR_NOMEM;
- return NULL;
- }
- memset(newjob, '\0', sizeof(pcrs_job));
+ hits = pcrs_execute(job, old, *result_length, &new, result_length);
- newjob->pattern = pattern;
- newjob->hints = hints;
- newjob->options = options;
- newjob->globalflag = globalflag;
- newjob->substitute = substitute;
+ if (old != subject) free(old);
+
+ if (hits < 0)
+ {
+ return(hits);
+ }
+ else
+ {
+ total_hits += hits;
+ old = new;
+ }
+ }
- return(newjob);
+ *result = new;
+ return(total_hits);
}
/*********************************************************************
*
- * Function : pcrs_exec_substitution
+ * Function : pcrs_execute
+ *
+ * Description : Apply the regular substitution defined by the job to the
+ * subject.
+ * The subject itself is left untouched, memory for the result
+ * is malloc()ed and it is the caller's responsibility to free
+ * the result when it's no longer needed.
*
- * Description : Modify the subject by executing the regular substitution
- * defined by the job. Since the result may be longer than
- * the subject, its space requirements are precalculated in
- * the matching phase and new memory is allocated accordingly.
- * It is the caller's responsibility to free the result when
- * it's no longer needed.
+ * Note: For convenient string handling, a null byte is
+ * appended to the result. It does not count towards the
+ * result_length, though.
*
* Parameters :
* 1 : job = the pcrs_job to be executed
* 2 : subject = the subject (== original) string
* 3 : subject_length = the subject's length
- * 4 : result = char** for returning the result
- * 5 : result_length = int* for returning the result's length
+ * 4 : result = char** for returning the result (NULL on error)
+ * 5 : result_length = size_t* for returning the result's length
*
- * Returns : the number of substitutions that were made. May be > 1
- * if job->globalflag was set
+ * Returns : On success, the number of substitutions that were made.
+ * May be > 1 if job->flags contained PCRS_GLOBAL
+ * On failure, the (negative) pcre error code describing the
+ * failure, which may be translated to text using pcrs_strerror().
*
*********************************************************************/
-int pcrs_exec_substitution(pcrs_job *job, char *subject, int subject_length, char **result, int *result_length)
+int pcrs_execute(pcrs_job *job, const char *subject, size_t subject_length, char **result, size_t *result_length)
{
- int offsets[3 * PCRS_MAX_SUBMATCHES],
- offset = 0, i=0, k, matches_found, newsize, submatches;
- pcrs_match matches[PCRS_MAX_MATCHES];
+ int offset,
+ i, k,
+ matches_found,
+ submatches,
+ max_matches = PCRS_MAX_MATCH_INIT;
+ size_t newsize;
+#ifdef HAVE_PCRE2
+ pcrs_match *matches, *dummy;
+ pcre2_match_data *pcre2_matches;
+ size_t *offsets;
+#else
+ pcrs_match *matches, *dummy;
+ int offsets[3 * PCRS_MAX_SUBMATCHES];
+#endif
char *result_offset;
- newsize=subject_length;
+ offset = i = 0;
+ *result = NULL;
+
+ /*
+ * Sanity check & memory allocation
+ */
+ if (job == NULL || job->pattern == NULL || job->substitute == NULL || NULL == subject)
+ {
+ return(PCRS_ERR_BADJOB);
+ }
- /* Find.. */
- while ((submatches = pcre_exec(job->pattern, job->hints, subject, subject_length, offset, 0, offsets, 99)) > 0)
+#ifdef HAVE_PCRE2
+ if (NULL == (pcre2_matches = pcre2_match_data_create_from_pattern(job->pattern, NULL)))
{
+ return(PCRS_ERR_NOMEM);
+ }
+ offsets = pcre2_get_ovector_pointer(pcre2_matches);
+#endif
+ if (NULL == (matches = (pcrs_match *)malloc((size_t)max_matches * sizeof(pcrs_match))))
+ {
+ return(PCRS_ERR_NOMEM);
+ }
+ memset(matches, '\0', (size_t)max_matches * sizeof(pcrs_match));
+
+ /*
+ * Find the pattern and calculate the space
+ * requirements for the result
+ */
+ newsize = subject_length;
+
+#ifdef HAVE_PCRE2
+ while ((submatches = pcre2_match(job->pattern, (const unsigned char *)subject,
+ subject_length, (size_t)offset, 0, pcre2_matches, NULL)) > 0)
+#else
+ while ((submatches = pcre_exec(job->pattern, job->hints, subject, (int)subject_length, offset, 0, offsets, 3 * PCRS_MAX_SUBMATCHES)) > 0)
+#endif
+ {
+ job->flags |= PCRS_SUCCESS;
matches[i].submatches = submatches;
- for (k=0; k < submatches; k++)
+
+ for (k = 0; k < submatches; k++)
{
- matches[i].submatch_offset[k] = offsets[2 * k];
- matches[i].submatch_length[k] = offsets[2 * k + 1] - offsets[2 * k]; /* Non-found optional submatches have length -1-(-1)==0 */
- newsize += matches[i].submatch_length[k] * job->substitute->backref_count[k]; /* reserve mem for each submatch as often as it is ref'd */
+ matches[i].submatch_offset[k] = (int)offsets[2 * k];
+
+ /* Note: Non-found optional submatches have length -1-(-1)==0 */
+ matches[i].submatch_length[k] = (size_t)(offsets[2 * k + 1] - offsets[2 * k]);
+
+ /* reserve mem for each submatch as often as it is ref'd */
+ newsize += matches[i].submatch_length[k] * (size_t)job->substitute->backref_count[k];
+ }
+ /* plus replacement text size minus match text size */
+ newsize += job->substitute->length - matches[i].submatch_length[0];
+
+ /* chunk before match */
+ matches[i].submatch_offset[PCRS_MAX_SUBMATCHES] = 0;
+ matches[i].submatch_length[PCRS_MAX_SUBMATCHES] = (size_t)offsets[0];
+ newsize += (size_t)offsets[0] * (size_t)job->substitute->backref_count[PCRS_MAX_SUBMATCHES];
+
+ /* chunk after match */
+ matches[i].submatch_offset[PCRS_MAX_SUBMATCHES + 1] = (int)offsets[1];
+ matches[i].submatch_length[PCRS_MAX_SUBMATCHES + 1] = subject_length - (size_t)offsets[1] - 1;
+ newsize += (subject_length - (size_t)offsets[1]) * (size_t)job->substitute->backref_count[PCRS_MAX_SUBMATCHES + 1];
+
+ /* Storage for matches exhausted? -> Extend! */
+ if (++i >= max_matches)
+ {
+ max_matches = (int)(max_matches * PCRS_MAX_MATCH_GROW);
+ if (NULL == (dummy = (pcrs_match *)realloc(matches, (size_t)max_matches * sizeof(pcrs_match))))
+ {
+ free(matches);
+ return(PCRS_ERR_NOMEM);
+ }
+ matches = dummy;
}
- newsize += strlen(job->substitute->text) - matches[i].submatch_length[0]; /* plus replacement text size minus match text size */
/* Non-global search or limit reached? */
- if (++i >= PCRS_MAX_MATCHES || !job->globalflag ) break;
+ if (!(job->flags & PCRS_GLOBAL)) break;
/* Don't loop on empty matches */
if (offsets[1] == offset)
- if (offset < subject_length)
+ if ((size_t)offset < subject_length)
offset++;
else
break;
/* Go find the next one */
else
- offset = offsets[1];
+ offset = (int)offsets[1];
+ }
+ /* Pass pcre error through if (bad) failure */
+#ifdef HAVE_PCRE2
+ if (submatches < PCRE2_ERROR_NOMATCH)
+#else
+ if (submatches < PCRE_ERROR_NOMATCH)
+#endif
+ {
+ free(matches);
+#ifdef HAVE_PCRE2
+ pcre2_match_data_free(pcre2_matches);
+#endif
+ return submatches;
}
- if (submatches < -1) return submatches; /* Pass pcre error through */
matches_found = i;
- /* ..get memory ..*/
- if ((*result = (char *)malloc(newsize)) == NULL) /* must be free()d by caller */
+
+ /*
+ * Get memory for the result (must be freed by caller!)
+ * and append terminating null byte.
+ */
+ if ((*result = (char *)malloc(newsize + 1
+#ifdef HAVE_PCRE2
+ /*
+ * Work around to prevent invalid reads in the jit code.
+ */
+ + 16
+#endif
+ )) == NULL)
{
+ free(matches);
+#ifdef HAVE_PCRE2
+ pcre2_match_data_free(pcre2_matches);
+#endif
return PCRS_ERR_NOMEM;
}
+ else
+ {
+ (*result)[newsize] = '\0';
+ }
+
- /* ..and replace */
+ /*
+ * Replace
+ */
offset = 0;
result_offset = *result;
- for (i=0; i < matches_found; i++)
+ for (i = 0; i < matches_found; i++)
{
- memcpy(result_offset, subject + offset, matches[i].submatch_offset[0] - offset); /* copy the chunk preceding the match */
+ /* copy the chunk preceding the match */
+ memcpy(result_offset, subject + offset, (size_t)(matches[i].submatch_offset[0] - offset));
result_offset += matches[i].submatch_offset[0] - offset;
/* For every segment of the substitute.. */
- for (k=0; k <= job->substitute->backrefs; k++)
+ for (k = 0; k <= job->substitute->backrefs; k++)
{
/* ...copy its text.. */
memcpy(result_offset, job->substitute->text + job->substitute->block_offset[k], job->substitute->block_length[k]);
result_offset += job->substitute->block_length[k];
- /* ..plus, if it's not the last chunk (i.e.: There IS a backref).. */
+ /* ..plus, if it's not the last chunk, i.e.: There *is* a backref.. */
if (k != job->substitute->backrefs
- /* ..and a nonempty match.. */
- && matches[i].submatch_length[job->substitute->backref[k]] > 0
- /* ..and in legal range, ... */
- && job->substitute->backref[k] <= PCRS_MAX_SUBMATCHES)
+ /* ..in legal range.. */
+ && job->substitute->backref[k] < PCRS_MAX_SUBMATCHES + 2
+ /* ..and referencing a real submatch.. */
+ && job->substitute->backref[k] < matches[i].submatches
+ /* ..that is nonempty.. */
+ && matches[i].submatch_length[job->substitute->backref[k]] > 0)
{
- /* copy the submatch that is ref'd. */
+ /* ..copy the submatch that is ref'd. */
memcpy(
result_offset,
subject + matches[i].submatch_offset[job->substitute->backref[k]],
result_offset += matches[i].submatch_length[job->substitute->backref[k]];
}
}
- offset = matches[i].submatch_offset[0] + matches[i].submatch_length[0];
+ offset = matches[i].submatch_offset[0] + (int)matches[i].submatch_length[0];
}
/* Copy the rest. */
- memcpy(result_offset, subject + offset, subject_length - offset);
+ memcpy(result_offset, subject + offset, subject_length - (size_t)offset);
*result_length = newsize;
+#ifdef HAVE_PCRE2
+ pcre2_match_data_free(pcre2_matches);
+#endif
+ free(matches);
return matches_found;
}
+#define is_hex_digit(x) ((x) && strchr("0123456789ABCDEF", toupper(x)))
+
+/*********************************************************************
+ *
+ * Function : is_hex_sequence
+ *
+ * Description : Checks the first four characters of a string
+ * and decides if they are a valid hex sequence
+ * (like '\x40').
+ *
+ * Parameters :
+ * 1 : sequence = The string to check
+ *
+ * Returns : Non-zero if it's valid sequence, or
+ * Zero if it isn't.
+ *
+ *********************************************************************/
+static int is_hex_sequence(const char *sequence)
+{
+ return (sequence[0] == '\\' &&
+ sequence[1] == 'x' &&
+ is_hex_digit(sequence[2]) &&
+ is_hex_digit(sequence[3]));
+}
+
+
+/*
+ * Functions below this line are only part of the pcrs version
+ * included in Privoxy. If you use any of them you should not
+ * try to dynamically link against external pcrs versions.
+ */
+
+/*********************************************************************
+ *
+ * Function : pcrs_job_is_dynamic
+ *
+ * Description : Checks if a job has the "D" (dynamic) option set.
+ *
+ * Parameters :
+ * 1 : job = The job to check
+ *
+ * Returns : TRUE if the job is indeed dynamic, otherwise
+ * FALSE
+ *
+ *********************************************************************/
+int pcrs_job_is_dynamic(char *job)
+{
+ const char delimiter = job[1];
+ const size_t length = strlen(job);
+ char *option;
+
+ if (length < 5)
+ {
+ /*
+ * The shortest valid (but useless)
+ * dynamic pattern is "s@@@D"
+ */
+ return FALSE;
+ }
+
+ /*
+ * Everything between the last character
+ * and the last delimiter is an option ...
+ */
+ for (option = job + length; *option != delimiter; option--)
+ {
+ if (*option == 'D')
+ {
+ /*
+ * ... and if said option is 'D' the job is dynamic.
+ */
+ return TRUE;
+ }
+ }
+ return FALSE;
+
+}
+
+
+/*********************************************************************
+ *
+ * Function : pcrs_get_delimiter
+ *
+ * Description : Tries to find a character that is safe to
+ * be used as a pcrs delimiter for a certain string.
+ *
+ * Parameters :
+ * 1 : string = The string to search in
+ *
+ * Returns : A safe delimiter if one was found, otherwise '\0'.
+ *
+ *********************************************************************/
+char pcrs_get_delimiter(const char *string)
+{
+ /*
+ * Some characters that are unlikely to
+ * be part of pcrs replacement strings.
+ */
+ static const char delimiters[] = "><#+*~%^-:;!@";
+ const char *d = delimiters;
+
+ /* Take the first delimiter that isn't part of the string */
+ while (*d && NULL != strchr(string, *d))
+ {
+ d++;
+ }
+ return *d;
+
+}
+
+
+/*********************************************************************
+ *
+ * Function : pcrs_execute_single_command
+ *
+ * Description : Apply single pcrs command to the subject.
+ * The subject itself is left untouched, memory for the result
+ * is malloc()ed and it is the caller's responsibility to free
+ * the result when it's no longer needed.
+ *
+ * Parameters :
+ * 1 : subject = the subject (== original) string
+ * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
+ * 3 : hits = int* for returning the number of modifications
+ *
+ * Returns : NULL in case of errors, otherwise the
+ * result of the pcrs command.
+ *
+ *********************************************************************/
+char *pcrs_execute_single_command(const char *subject, const char *pcrs_command, int *hits)
+{
+ size_t buffer_size, new_size;
+ char *result = NULL;
+ pcrs_job *job;
+
+ assert(subject);
+ assert(pcrs_command);
+
+ *hits = 0;
+ buffer_size = strlen(subject);
+
+ job = pcrs_compile_command(pcrs_command, hits);
+ if (NULL != job)
+ {
+ *hits = pcrs_execute(job, subject, buffer_size, &result, &new_size);
+ if (*hits < 0)
+ {
+ freez(result);
+ }
+ pcrs_free_job(job);
+ }
+ return result;
+
+}
+
+
+/*********************************************************************
+ *
+ * Function : pcrs_compile_dynamic_command
+ *
+ * Description : Takes a dynamic pcrs command, fills in the
+ * values of the variables and compiles it.
+ *
+ * Parameters :
+ * 1 : pcrs_command = The dynamic pcrs command to compile
+ * 2 : v = NULL terminated array of variables and their values.
+ * 3 : error = pcrs error code
+ *
+ * Returns : NULL in case of hard errors, otherwise the
+ * compiled pcrs job.
+ *
+ *********************************************************************/
+pcrs_job *pcrs_compile_dynamic_command(char *pcrs_command, const struct pcrs_variable v[], int *error)
+{
+ char buf[PCRS_BUFFER_SIZE];
+ const char *original_pcrs_command = pcrs_command;
+ char *pcrs_command_tmp = NULL;
+ pcrs_job *job = NULL;
+ int truncation = 0;
+ char d;
+ int ret;
+
+ while ((NULL != v->name) && (NULL != pcrs_command))
+ {
+ assert(NULL != v->value);
+
+ if (NULL == strstr(pcrs_command, v->name))
+ {
+ /*
+ * Skip the substitution if the variable
+ * name isn't part of the pattern.
+ */
+ v++;
+ continue;
+ }
+
+ /* Use pcrs to replace the variable with its value. */
+ d = pcrs_get_delimiter(v->value);
+ if ('\0' == d)
+ {
+ /* No proper delimiter found */
+ *error = PCRS_ERR_CMDSYNTAX;
+ freez(pcrs_command_tmp);
+ return NULL;
+ }
+
+ /*
+ * Variable names are supposed to contain alpha
+ * numerical characters plus '_' only.
+ */
+ assert(NULL == strchr(v->name, d));
+
+ ret = snprintf(buf, sizeof(buf), "s%c\\$%s%c%s%cDgT", d, v->name, d, v->value, d);
+ assert(ret >= 0);
+ if (ret >= sizeof(buf))
+ {
+ /*
+ * Value didn't completely fit into buffer,
+ * overwrite the end of the substitution text
+ * with a truncation message and close the pattern
+ * properly.
+ */
+ static const char warning[] = "... [too long, truncated]";
+ const size_t trailer_size = sizeof(warning) + 4; /* 4 for d + "DgT" */
+ char *trailer_start = buf + sizeof(buf) - trailer_size;
+
+ ret = snprintf(trailer_start, trailer_size, "%s%cDgT", warning, d);
+ assert(ret == trailer_size - 1);
+ assert(sizeof(buf) == strlen(buf) + 1);
+ truncation = 1;
+ }
+
+ pcrs_command_tmp = pcrs_execute_single_command(pcrs_command, buf, error);
+ if (NULL == pcrs_command_tmp)
+ {
+ return NULL;
+ }
+
+ if (pcrs_command != original_pcrs_command)
+ {
+ freez(pcrs_command);
+ }
+ pcrs_command = pcrs_command_tmp;
+
+ v++;
+ }
+
+ job = pcrs_compile_command(pcrs_command, error);
+ if (pcrs_command != original_pcrs_command)
+ {
+ freez(pcrs_command);
+ }
+
+ if (truncation)
+ {
+ *error = PCRS_WARN_TRUNCATION;
+ }
+
+ return job;
+
+}
+
+
/*
Local Variables:
tab-width: 3