X-Git-Url: http://www.privoxy.org/gitweb/?p=privoxy.git;a=blobdiff_plain;f=pcrs.c;h=83100983b6475d0b7d75cbae6280db8595227d28;hp=4baaf72b8119a37b8ef649593d3aac9042994f21;hb=56f3964a2bd367ca422feef5c72f537aa43c17e8;hpb=9577b50e7aaeedaa8e97df6f9ef78122217de524 diff --git a/pcrs.c b/pcrs.c index 4baaf72b..83100983 100644 --- a/pcrs.c +++ b/pcrs.c @@ -1,29 +1,21 @@ -const char pcrs_rcs[] = "$Id: pcrs.c,v 1.6 2001/06/03 11:03:48 oes Exp $"; - /********************************************************************* * * File : $Source: /cvsroot/ijbswa/current/pcrs.c,v $ * - * Purpose : This is the alpha release of libpcrs. It is only published - * at this early stage of development, because it is - * needed for a new feature in JunkBuster. - * - * While no inconsistencies, memory leaks or functional bugs - * are known at this time, there *could* be plenty ;-). Also, - * Many pcre-specific options are not yet supported, and - * error handling needs improvement. + * Purpose : pcrs is a supplement to the pcre library by Philip Hazel + * and adds Perl-style substitution. That + * is, it mimics Perl's 's' operator. See pcrs(3) for details. * - * 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. + * 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additional functions and bug + * fixes that aren't part of the latest official pcrs package + * (which apparently is no longer maintained). * - * Currently, there's no documentation besides comments and the - * source itself ;-) - * - * Copyright : Written and Copyright (C) 2000 by Andreas Oesterhelt + * Copyright : Written and Copyright (C) 2000, 2001 by Andreas S. Oesterhelt * * - * This program is free software; you can redistribute it + * Copyright (C) 2006, 2007 Fabian Keil + * + * 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 @@ -41,237 +33,145 @@ const char pcrs_rcs[] = "$Id: pcrs.c,v 1.6 2001/06/03 11:03:48 oes Exp $"; * or write to the Free Software Foundation, Inc., 59 * Temple Place - Suite 330, Boston, MA 02111-1307, USA. * - * Revisions : - * $Log: pcrs.c,v $ - * Revision 1.6 2001/06/03 11:03:48 oes - * Makefile/in - * - * introduced cgi.c - * - * actions.c: - * - * adapted to new enlist_unique arg format - * - * conf loadcfg.c - * - * introduced confdir option - * - * filters.c filtrers.h - * - * extracted-CGI relevant stuff - * - * jbsockets.c - * - * filled comment - * - * jcc.c - * - * support for new cgi mechansim - * - * list.c list.h - * - * functions for new list type: "map" - * extended enlist_unique - * - * miscutil.c .h - * introduced bindup() - * - * parsers.c parsers.h - * - * deleted const struct interceptors - * - * pcrs.c - * added FIXME - * - * project.h - * - * added struct map - * added struct http_response - * changes struct interceptors to struct cgi_dispatcher - * moved HTML stuff to cgi.h - * - * re_filterfile: - * - * changed - * - * showargs.c - * NO TIME LEFT - * - * Revision 1.5 2001/05/29 09:50:24 jongfoster - * Unified blocklist/imagelist/permissionslist. - * File format is still under discussion, but the internal changes - * are (mostly) done. - * - * Also modified interceptor behaviour: - * - We now intercept all URLs beginning with one of the following - * prefixes (and *only* these prefixes): - * * http://i.j.b/ - * * http://ijbswa.sf.net/config/ - * * http://ijbswa.sourceforge.net/config/ - * - New interceptors "home page" - go to http://i.j.b/ to see it. - * - Internal changes so that intercepted and fast redirect pages - * are not replaced with an image. - * - Interceptors now have the option to send a binary page direct - * to the client. (i.e. ijb-send-banner uses this) - * - Implemented show-url-info interceptor. (Which is why I needed - * the above interceptors changes - a typical URL is - * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif". - * The previous mechanism would not have intercepted that, and - * if it had been intercepted then it then it would have replaced - * it with an image.) - * - * Revision 1.4 2001/05/25 14:12:40 oes - * Fixed bug: Empty substitutes now detected - * - * Revision 1.3 2001/05/25 11:03:55 oes - * Added sanity check for NULL jobs to pcrs_exec_substitution - * - * Revision 1.2 2001/05/22 18:46:04 oes - * - * - Enabled filtering banners by size rather than URL - * by adding patterns that replace all standard banner - * sizes with the "Junkbuster" gif to the re_filterfile - * - * - Enabled filtering WebBugs by providing a pattern - * which kills all 1x1 images - * - * - Added support for PCRE_UNGREEDY behaviour to pcrs, - * which is selected by the (nonstandard and therefore - * capital) letter 'U' in the option string. - * It causes the quantifiers to be ungreedy by default. - * Appending a ? turns back to greedy (!). - * - * - Added a new interceptor ijb-send-banner, which - * sends back the "Junkbuster" gif. Without imagelist or - * MSIE detection support, or if tinygif = 1, or the - * URL isn't recognized as an imageurl, a lame HTML - * explanation is sent instead. - * - * - Added new feature, which permits blocking remote - * script redirects and firing back a local redirect - * to the browser. - * The feature is conditionally compiled, i.e. it - * can be disabled with --disable-fast-redirects, - * plus it must be activated by a "fast-redirects" - * line in the config file, has its own log level - * and of course wants to be displayed by show-proxy-args - * Note: Boy, all the #ifdefs in 1001 locations and - * all the fumbling with configure.in and acconfig.h - * were *way* more work than the feature itself :-( - * - * - Because a generic redirect template was needed for - * this, tinygif = 3 now uses the same. - * - * - Moved GIFs, and other static HTTP response templates - * to project.h - * - * - Some minor fixes - * - * - Removed some >400 CRs again (Jon, you really worked - * a lot! ;-) - * - * Revision 1.1.1.1 2001/05/15 13:59:02 oes - * Initial import of version 2.9.3 source tree - * - * *********************************************************************/ - -#include + #include +#include +#include + +/* + * 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, 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; - - limit = strlen(*text); - if ( 0 == limit ) - { - return -1; - } - - token[0] = '\0'; + static char buf[100]; - for (i=0; i < limit; i++) + if (error != 0) { - if (text[0][i] == delimiter && !quoted) + switch (error) { - *text += 1; - break; - } - else if (text[0][i] == quote_char && !quoted && i+1 < limit && text[0][i+1] == delimiter) - { - quoted = TRUE; - continue; + /* Passed-through PCRE error: */ + case PCRE_ERROR_NOMEMORY: return "(pcre:) No memory"; + + /* Shouldn't happen unless PCRE or PCRS bug, or user messed with compiled job: */ + case PCRE_ERROR_NULL: return "(pcre:) NULL code or subject or ovector"; + case PCRE_ERROR_BADOPTION: return "(pcre:) Unrecognized option bit"; + case PCRE_ERROR_BADMAGIC: return "(pcre:) Bad magic number in code"; + case PCRE_ERROR_UNKNOWN_NODE: return "(pcre:) Bad node in pattern"; + + /* Can't happen / not passed: */ + case PCRE_ERROR_NOSUBSTRING: return "(pcre:) Fire in power supply"; + case PCRE_ERROR_NOMATCH: return "(pcre:) Water in power supply"; + +#ifdef PCRE_ERROR_MATCHLIMIT + /* + * Only reported by PCRE versions newer than our own. + */ + case PCRE_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"; + + /* + * XXX: With the exception of PCRE_ERROR_MATCHLIMIT we + * only catch PCRE errors that can happen with our internal + * version. If Privoxy is linked against a newer + * PCRE version all bets are off ... + */ + default: + snprintf(buf, sizeof(buf), + "Error code %d. For details, check the pcre documentation.", + error); + 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, int *flags) { size_t i; int rc = 0; - *globalflag = 0; - for (i=0; i < strlen(optstring); i++) + *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 'e': break; /* ToDo ;-) */ + case 'g': *flags |= PCRS_GLOBAL; break; case 'i': rc |= PCRE_CASELESS; break; case 'm': rc |= PCRE_MULTILINE; break; case 'o': break; case 's': rc |= PCRE_DOTALL; break; case 'x': rc |= PCRE_EXTENDED; break; + case 'D': *flags |= PCRS_DYNAMIC; break; case 'U': rc |= PCRE_UNGREEDY; break; - default: break; + case 'T': *flags |= PCRS_TRIVIAL; break; + default: break; } } return rc; @@ -279,6 +179,38 @@ int pcrs_compile_perl_options(char *optstring, int *globalflag) } +#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 @@ -291,7 +223,11 @@ int pcrs_compile_perl_options(char *optstring, int *globalflag) * 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 $+ handling. + * 4 : errptr = pointer to an integer in which error * conditions can be returned. * * Returns : pcrs_substitute data structure, or NULL if an @@ -299,82 +235,212 @@ int pcrs_compile_perl_options(char *optstring, int *globalflag) * the reason. * *********************************************************************/ -pcrs_substitute *pcrs_compile_replacement(char *replacement, int *errptr) +static pcrs_substitute *pcrs_compile_replacement(const char *replacement, int trivialflag, int capturecount, int *errptr) { - int length, i, k = 0, l = 0, quoted = 0, idx; - char *text, *num_ptr, *numbers = "0123456789"; + int i, k, l, quoted; + char *text; pcrs_substitute *r; +#ifndef FUZZ + size_t length; +#else + static size_t length; +#endif + i = k = l = quoted = 0; - r = (pcrs_substitute *)malloc(sizeof(pcrs_substitute)); - if (r == NULL) return NULL; - memset(r, '\0', sizeof(pcrs_substitute)); + /* + * Sanity check + */ + if (NULL == replacement) + { + replacement = ""; + } - text = strdup(replacement); /* must be free()d by caller */ - if (text == NULL) + /* + * Get memory or fail + */ + if (NULL == (r = (pcrs_substitute *)malloc(sizeof(pcrs_substitute)))) { *errptr = PCRS_ERR_NOMEM; - free(r); return NULL; } + memset(r, '\0', sizeof(pcrs_substitute)); length = strlen(replacement); - for (i=0; i < length; i++) + if (NULL == (text = (char *)malloc(length + 1))) { - /* Backslash treatment */ - if (replacement[i] == '\\') - { - if (quoted) - { - text[k++] = replacement[i]; - quoted = 0; - } - else - { - quoted = 1; - } - continue; - } + free(r); + *errptr = PCRS_ERR_NOMEM; + return NULL; + } + memset(text, '\0', length + 1); + + + /* + * In trivial mode, just copy the substitute text + */ + if (trivialflag) + { + strlcpy(text, replacement, length + 1); + k = (int)length; + } - /* Dollar treatment */ - if (replacement[i] == '$' && !quoted && i < length - 1) + /* + * 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) + { + 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 ((num_ptr = strchr(numbers, replacement[++i])) != NULL && i < length) + 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; } @@ -385,14 +451,13 @@ pcrs_substitute *pcrs_compile_replacement(char *replacement, int *errptr) * 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) @@ -407,7 +472,14 @@ pcrs_job *pcrs_free_job(pcrs_job *job) { next = job->next; if (job->pattern != NULL) free(job->pattern); - if (job->hints != NULL) free(job->hints); + if (job->hints != NULL) + { +#ifdef PCRE_CONFIG_JIT + pcre_free_study(job->hints); +#else + free(job->hints); +#endif + } if (job->substitute != NULL) { if (job->substitute->text != NULL) free(job->substitute->text); @@ -422,12 +494,35 @@ pcrs_job *pcrs_free_job(pcrs_job *job) /********************************************************************* * - * 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 @@ -439,97 +534,153 @@ pcrs_job *pcrs_free_job(pcrs_job *job) * 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; } - - /* 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) + else { - switch (i) - { - /* We don't care about the command and assume 's' */ - case 0: - break; + delimiter = command[1]; + } - /* The pattern */ - case 1: - strcpy(dummy, token); - break; + tokens[l] = (char *) malloc(limit + 1); - /* The substitute */ - case 2: - if ((newjob->substitute = pcrs_compile_replacement(token, errptr)) == NULL) - { - pcrs_free_job(newjob); - return NULL; - } - else - { - break; - } + for (i = 0; i <= (int)limit; i++) + { - /* The options */ - case 3: - newjob->options = pcrs_compile_perl_options(token, &globalflag); - newjob->globalflag = globalflag; + if (command[i] == delimiter && !quoted) + { + if (l == 3) + { + l = -1; break; + } + tokens[0][k++] = '\0'; + tokens[++l] = tokens[0] + k; + continue; + } - /* There shouldn't be anything else! */ - default: - *errptr = PCRS_ERR_CMDSYNTAX; - pcrs_free_job(newjob); - return NULL; + else if (command[i] == '\\' && !quoted) + { + quoted = TRUE; + if (command[i+1] == delimiter) continue; } - i++; + else + { + quoted = FALSE; + } + tokens[0][k++] = command[i]; } - free(token); - /* We have a valid substitute? */ - if (newjob->substitute == NULL) + /* + * Syntax error ? + */ + if (l != 3) { *errptr = PCRS_ERR_CMDSYNTAX; - pcrs_free_job(newjob); + free(tokens[0]); return NULL; } - /* Compile the pattern */ - newjob->pattern = pcre_compile(dummy, newjob->options, &error, errptr, NULL); + newjob = pcrs_compile(tokens[1], tokens[2], tokens[3], errptr); + free(tokens[0]); + return newjob; + +} + + +/********************************************************************* + * + * 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; + int flags; + int capturecount; + const char *error; + int pcre_study_options = 0; + + *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; + } + memset(newjob, '\0', sizeof(pcrs_job)); + + + /* + * Evaluate the options + */ + newjob->options = pcrs_parse_perl_options(options, &flags); + newjob->flags = flags; + + + /* + * Compile the pattern + */ + newjob->pattern = pcre_compile(pattern, newjob->options, &error, errptr, NULL); if (newjob->pattern == NULL) { pcrs_free_job(newjob); return NULL; } - free(dummy); + + +#ifdef PCRE_STUDY_JIT_COMPILE + if (!(flags & PCRS_DYNAMIC)) + { + pcre_study_options = PCRE_STUDY_JIT_COMPILE; + } +#endif /* * 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; @@ -537,6 +688,27 @@ pcrs_job *pcrs_make_job(char *command, int *errptr) return NULL; } + + /* + * Determine the number of capturing subpatterns. + * This is needed for handling $+ in the substitute. + */ + if (0 > (*errptr = pcre_fullinfo(newjob->pattern, newjob->hints, PCRE_INFO_CAPTURECOUNT, &capturecount))) + { + 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; } @@ -544,107 +716,175 @@ pcrs_job *pcrs_make_job(char *command, int *errptr) /********************************************************************* * - * 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); - return(newjob); + if (hits < 0) + { + return(hits); + } + else + { + total_hits += hits; + old = new; + } + } + + *result = new; + return(total_hits); } /********************************************************************* * - * Function : pcrs_exec_substitution + * Function : pcrs_execute * - * 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. + * 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. * - * FIXME: MUST HANDLE SUBJECTS THAT ARE LONGER THAN subject_length - CORRECTLY! --oes + * 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]; + offset, + i, k, + matches_found, + submatches, + max_matches = PCRS_MAX_MATCH_INIT; + size_t newsize; + pcrs_match *matches, *dummy; char *result_offset; + offset = i = 0; + *result = NULL; - /* Sanity first */ - if (job == NULL || job->pattern == NULL || job->substitute == NULL) + /* + * Sanity check & memory allocation + */ + if (job == NULL || job->pattern == NULL || job->substitute == NULL || NULL == subject) { - *result = NULL; return(PCRS_ERR_BADJOB); } - newsize=subject_length; + 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; - /* Find.. */ - while ((submatches = pcre_exec(job->pattern, job->hints, subject, subject_length, offset, 0, offsets, 99)) > 0) + while ((submatches = pcre_exec(job->pattern, job->hints, subject, (int)subject_length, offset, 0, offsets, 3 * PCRS_MAX_SUBMATCHES)) > 0) { + 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 */ + + /* 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] = 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; @@ -652,41 +892,59 @@ int pcrs_exec_substitution(pcrs_job *job, char *subject, int subject_length, cha else offset = offsets[1]; } - if (submatches < -1) return submatches; /* Pass pcre error through */ + /* Pass pcre error through if (bad) failure */ + if (submatches < PCRE_ERROR_NOMATCH) + { + free(matches); + return submatches; + } 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)) == NULL) { + free(matches); 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]], @@ -695,18 +953,282 @@ int pcrs_exec_substitution(pcrs_job *job, char *subject, int subject_length, cha 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; + 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 size; + char *result = NULL; + pcrs_job *job; + + assert(subject); + assert(pcrs_command); + + *hits = 0; + size = strlen(subject); + + job = pcrs_compile_command(pcrs_command, hits); + if (NULL != job) + { + *hits = pcrs_execute(job, subject, size, &result, &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