From 4aca87515a5083ae0e31ce3177189fd43b6d05ac Mon Sep 17 00:00:00 2001 From: Andreas Baumann Date: Sat, 3 Jan 2015 13:58:15 +0100 Subject: patch to Vanilla Tomato 1.28 --- release/src/router/busybox/editors/sed.c | 1878 ++++++++++++++++-------------- 1 file changed, 1023 insertions(+), 855 deletions(-) (limited to 'release/src/router/busybox/editors/sed.c') diff --git a/release/src/router/busybox/editors/sed.c b/release/src/router/busybox/editors/sed.c index 6beba066..eb31f7d2 100644 --- a/release/src/router/busybox/editors/sed.c +++ b/release/src/router/busybox/editors/sed.c @@ -1,27 +1,39 @@ +/* vi: set sw=4 ts=4: */ /* * sed.c - very minimalist version of sed * * Copyright (C) 1999,2000,2001 by Lineo, inc. and Mark Whitley * Copyright (C) 1999,2000,2001 by Mark Whitley * Copyright (C) 2002 Matt Kraai - * Copyright (C) 2003 by Glenn McGrath + * Copyright (C) 2003 by Glenn McGrath + * Copyright (C) 2003,2004 by Rob Landley * - * 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. - * - * 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. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * MAINTAINER: Rob Landley * + * Licensed under GPL version 2, see file LICENSE in this tarball for details. */ +/* Code overview. + + Files are laid out to avoid unnecessary function declarations. So for + example, every function add_cmd calls occurs before add_cmd in this file. + + add_cmd() is called on each line of sed command text (from a file or from + the command line). It calls get_address() and parse_cmd_args(). The + resulting sed_cmd_t structures are appended to a linked list + (G.sed_cmd_head/G.sed_cmd_tail). + + add_input_file() adds a FILE* to the list of input files. We need to + know all input sources ahead of time to find the last line for the $ match. + + process_files() does actual sedding, reading data lines from each input FILE * + (which could be stdin) and applying the sed command list (sed_cmd_head) to + each of the resulting lines. + + sed_main() is where external code calls into this, with a command line. +*/ + + /* Supported features and commands in this version of sed: @@ -30,114 +42,104 @@ - commands: (p)rint, (d)elete, (s)ubstitue (with g & I flags) - edit commands: (a)ppend, (i)nsert, (c)hange - file commands: (r)ead - - backreferences in substitution expressions (\1, \2...\9) + - backreferences in substitution expressions (\0, \1, \2...\9) - grouped commands: {cmd1;cmd2} - transliteration (y/source-chars/dest-chars/) - pattern space hold space storing / swapping (g, h, x) - - labels / branching (: label, b, t) + - labels / branching (: label, b, t, T) (Note: Specifying an address (range) to match is *optional*; commands default to the whole pattern space if no specific address match was requested.) - Unsupported features: - - - GNU extensions - - and lots, lots more. - - Bugs: - - - Cant subst globally using ^ or $ in regex, eg. "aah" | sed 's/^a/b/g' + Todo: + - Create a wrapper around regex to make libc's regex conform with sed Reference http://www.opengroup.org/onlinepubs/007904975/utilities/sed.html */ -#include -#include /* for getopt() */ -#include -#include /* for strdup() */ -#include -#include /* for isspace() */ -#include -#include "busybox.h" +#include "libbb.h" +#include "xregex.h" +/* Each sed command turns into one of these structures. */ typedef struct sed_cmd_s { - /* Order by alignment requirements */ + /* Ordered by alignment requirements: currently 36 bytes on x86 */ + struct sed_cmd_s *next; /* Next command (linked list, NULL terminated) */ /* address storage */ - regex_t *beg_match; /* sed -e '/match/cmd' */ - regex_t *end_match; /* sed -e '/match/,/end_match/cmd' */ - - int beg_line; /* 'sed 1p' 0 == no begining line, apply commands to all lines */ - int end_line; /* 'sed 1,3p' 0 == no end line, use only beginning. -1 == $ */ - - /* inversion flag */ - int invert; /* the '!' after the address */ -// int block_cmd; /* This command is part of a group that has a command address */ - - /* SUBSTITUTION COMMAND SPECIFIC FIELDS */ - - /* sed -e 's/sub_match/replace/' */ - regex_t *sub_match; - char *replace; + regex_t *beg_match; /* sed -e '/match/cmd' */ + regex_t *end_match; /* sed -e '/match/,/end_match/cmd' */ + regex_t *sub_match; /* For 's/sub_match/string/' */ + int beg_line; /* 'sed 1p' 0 == apply commands to all lines */ + int end_line; /* 'sed 1,3p' 0 == one line only. -1 = last line ($) */ - /* EDIT COMMAND (a,i,c) SPECIFIC FIELDS */ - char *editline; + FILE *sw_file; /* File (sw) command writes to, -1 for none. */ + char *string; /* Data string for (saicytb) commands. */ - /* FILE COMMAND (r) SPECIFIC FIELDS */ - char *filename; + unsigned which_match; /* (s) Which match to replace (0 for all) */ - /* SUBSTITUTION COMMAND SPECIFIC FIELDS */ + /* Bitfields (gcc won't group them if we don't) */ + unsigned invert:1; /* the '!' after the address */ + unsigned in_match:1; /* Next line also included in match? */ + unsigned sub_p:1; /* (s) print option */ - unsigned int num_backrefs:4; /* how many back references (\1..\9) */ - /* Note: GNU/POSIX sed does not save more than nine backrefs, so - * we only use 4 bits to hold the number */ - unsigned int sub_g:1; /* sed -e 's/foo/bar/g' (global) */ - unsigned int sub_p:1; /* sed -e 's/foo/bar/p' (print substitution) */ - - /* TRANSLATE COMMAND */ - char *translate; + char sw_last_char; /* Last line written by (sw) had no '\n' */ /* GENERAL FIELDS */ - /* the command */ - char cmd; /* p,d,s (add more at your leisure :-) */ + char cmd; /* The command char: abcdDgGhHilnNpPqrstwxy:={} */ +} sed_cmd_t; - /* Branch commands */ - char *label; +static const char semicolon_whitespace[] ALIGN1 = "; \n\r\t\v"; - /* next command in list (sequential list of specified commands) */ - struct sed_cmd_s *next; +struct globals { + /* options */ + int be_quiet, regex_type; + FILE *nonstdout; + char *outname, *hold_space; -} sed_cmd_t; + /* List of input files */ + int input_file_count, current_input_file; + FILE **input_file_list; + regmatch_t regmatch[10]; + regex_t *previous_regex_ptr; -/* externs */ -extern void xregcomp(regex_t * preg, const char *regex, int cflags); -extern int optind; /* in unistd.h */ -extern char *optarg; /* ditto */ + /* linked list of sed commands */ + sed_cmd_t sed_cmd_head, *sed_cmd_tail; -/* globals */ -/* options */ -static int be_quiet = 0; -static const char bad_format_in_subst[] = - "bad format in substitution expression"; + /* Linked list of append lines */ + llist_t *append_head; -/* linked list of sed commands */ -static sed_cmd_t sed_cmd_head; -static sed_cmd_t *sed_cmd_tail = &sed_cmd_head; + char *add_cmd_line; -const char *const semicolon_whitespace = "; \n\r\t\v\0"; -static regex_t *previous_regex_ptr = NULL; + struct pipeline { + char *buf; /* Space to hold string */ + int idx; /* Space used */ + int len; /* Space allocated */ + } pipeline; +}; +#define G (*(struct globals*)&bb_common_bufsiz1) +void BUG_sed_globals_too_big(void); +#define INIT_G() do { \ + if (sizeof(struct globals) > COMMON_BUFSIZE) \ + BUG_sed_globals_too_big(); \ + G.sed_cmd_tail = &G.sed_cmd_head; \ +} while (0) -#ifdef CONFIG_FEATURE_CLEAN_UP -static void destroy_cmd_strs(void) +#if ENABLE_FEATURE_CLEAN_UP +static void sed_free_and_close_stuff(void) { - sed_cmd_t *sed_cmd = sed_cmd_head.next; + sed_cmd_t *sed_cmd = G.sed_cmd_head.next; + + llist_free(G.append_head, free); while (sed_cmd) { sed_cmd_t *sed_cmd_next = sed_cmd->next; + if (sed_cmd->sw_file) + xprint_and_close_file(sed_cmd->sw_file); + if (sed_cmd->beg_match) { regfree(sed_cmd->beg_match); free(sed_cmd->beg_match); @@ -150,75 +152,119 @@ static void destroy_cmd_strs(void) regfree(sed_cmd->sub_match); free(sed_cmd->sub_match); } - free(sed_cmd->replace); + free(sed_cmd->string); free(sed_cmd); sed_cmd = sed_cmd_next; } + + free(G.hold_space); + + while (G.current_input_file < G.input_file_count) + fclose(G.input_file_list[G.current_input_file++]); } +#else +void sed_free_and_close_stuff(void); #endif +/* If something bad happens during -i operation, delete temp file */ + +static void cleanup_outname(void) +{ + if (G.outname) unlink(G.outname); +} + +/* strcpy, replacing "\from" with 'to'. If to is NUL, replacing "\any" with 'any' */ + +static void parse_escapes(char *dest, const char *string, int len, char from, char to) +{ + int i = 0; + + while (i < len) { + if (string[i] == '\\') { + if (!to || string[i+1] == from) { + *dest++ = to ? to : string[i+1]; + i += 2; + continue; + } + *dest++ = string[i++]; + } + /* TODO: is it safe wrt a string with trailing '\\' ? */ + *dest++ = string[i++]; + } + *dest = '\0'; +} + +static char *copy_parsing_escapes(const char *string, int len) +{ + char *dest = xmalloc(len + 1); + + parse_escapes(dest, string, len, 'n', '\n'); + /* GNU sed also recognizes \t */ + parse_escapes(dest, dest, strlen(dest), 't', '\t'); + return dest; +} + + /* * index_of_next_unescaped_regexp_delim - walks left to right through a string * beginning at a specified index and returns the index of the next regular - * expression delimiter (typically a forward * slash ('/')) not preceeded by - * a backslash ('\'). + * expression delimiter (typically a forward slash ('/')) not preceded by + * a backslash ('\'). A negative delimiter disables square bracket checking. */ -static int index_of_next_unescaped_regexp_delim(const char delimiter, - const char *str) +static int index_of_next_unescaped_regexp_delim(int delimiter, const char *str) { int bracket = -1; int escaped = 0; int idx = 0; char ch; + if (delimiter < 0) { + bracket--; + delimiter = -delimiter; + } + for (; (ch = str[idx]); idx++) { - if (bracket != -1) { + if (bracket >= 0) { if (ch == ']' && !(bracket == idx - 1 || (bracket == idx - 2 - && str[idx - 1] == '^'))) + && str[idx - 1] == '^'))) bracket = -1; } else if (escaped) escaped = 0; else if (ch == '\\') escaped = 1; - else if (ch == '[') + else if (bracket == -1 && ch == '[') bracket = idx; else if (ch == delimiter) return idx; } /* if we make it to here, we've hit the end of the string */ - return -1; + bb_error_msg_and_die("unmatched '%c'", delimiter); } +/* + * Returns the index of the third delimiter + */ static int parse_regex_delim(const char *cmdstr, char **match, char **replace) { const char *cmdstr_ptr = cmdstr; char delimiter; int idx = 0; - /* verify that the 's' is followed by something. That something + /* verify that the 's' or 'y' is followed by something. That something * (typically a 'slash') is now our regexp delimiter... */ if (*cmdstr == '\0') - bb_error_msg_and_die(bad_format_in_subst); - else - delimiter = *cmdstr_ptr; - - cmdstr_ptr++; + bb_error_msg_and_die("bad format in substitution expression"); + delimiter = *cmdstr_ptr++; /* save the match string */ idx = index_of_next_unescaped_regexp_delim(delimiter, cmdstr_ptr); - if (idx == -1) { - bb_error_msg_and_die(bad_format_in_subst); - } - *match = bb_xstrndup(cmdstr_ptr, idx); + *match = copy_parsing_escapes(cmdstr_ptr, idx); /* save the replacement string */ cmdstr_ptr += idx + 1; - idx = index_of_next_unescaped_regexp_delim(delimiter, cmdstr_ptr); - if (idx == -1) { - bb_error_msg_and_die(bad_format_in_subst); - } - *replace = bb_xstrndup(cmdstr_ptr, idx); + idx = index_of_next_unescaped_regexp_delim(-delimiter, cmdstr_ptr); + *replace = copy_parsing_escapes(cmdstr_ptr, idx); return ((cmdstr_ptr - cmdstr) + idx); } @@ -226,99 +272,133 @@ static int parse_regex_delim(const char *cmdstr, char **match, char **replace) /* * returns the index in the string just past where the address ends. */ -static int get_address(char *my_str, int *linenum, regex_t ** regex) +static int get_address(const char *my_str, int *linenum, regex_t ** regex) { - int idx = 0; - - if (isdigit(my_str[idx])) { - char *endstr; + const char *pos = my_str; - *linenum = strtol(my_str, &endstr, 10); + if (isdigit(*my_str)) { + *linenum = strtol(my_str, (char**)&pos, 10); /* endstr shouldnt ever equal NULL */ - idx = endstr - my_str; - } else if (my_str[idx] == '$') { + } else if (*my_str == '$') { *linenum = -1; - idx++; - } else if (my_str[idx] == '/' || my_str[idx] == '\\') { - int idx_start = 1; + pos++; + } else if (*my_str == '/' || *my_str == '\\') { + int next; char delimiter; + char *temp; delimiter = '/'; - if (my_str[idx] == '\\') { - idx_start++; - delimiter = my_str[++idx]; - } + if (*my_str == '\\') delimiter = *++pos; + next = index_of_next_unescaped_regexp_delim(delimiter, ++pos); + temp = copy_parsing_escapes(pos, next); + *regex = xmalloc(sizeof(regex_t)); + xregcomp(*regex, temp, G.regex_type|REG_NEWLINE); + free(temp); + /* Move position to next character after last delimiter */ + pos += (next+1); + } + return pos - my_str; +} + +/* Grab a filename. Whitespace at start is skipped, then goes to EOL. */ +static int parse_file_cmd(/*sed_cmd_t *sed_cmd,*/ const char *filecmdstr, char **retval) +{ + int start = 0, idx, hack = 0; + + /* Skip whitespace, then grab filename to end of line */ + while (isspace(filecmdstr[start])) + start++; + idx = start; + while (filecmdstr[idx] && filecmdstr[idx] != '\n') idx++; - idx += index_of_next_unescaped_regexp_delim(delimiter, my_str + idx); - if (idx == -1) { - bb_error_msg_and_die("unterminated match expression"); - } - my_str[idx] = '\0'; - *regex = (regex_t *) xmalloc(sizeof(regex_t)); - xregcomp(*regex, my_str + idx_start, REG_NEWLINE); - idx++; /* so it points to the next character after the last '/' */ - } + /* If lines glued together, put backslash back. */ + if (filecmdstr[idx] == '\n') + hack = 1; + if (idx == start) + bb_error_msg_and_die("empty filename"); + *retval = xstrndup(filecmdstr+start, idx-start+hack+1); + if (hack) + (*retval)[idx] = '\\'; + return idx; } -static int parse_subst_cmd(sed_cmd_t * const sed_cmd, const char *substr) +static int parse_subst_cmd(sed_cmd_t *sed_cmd, const char *substr) { - int cflags = 0; + int cflags = G.regex_type; char *match; - int idx = 0; - int j; + int idx; /* - * the string that gets passed to this function should look like this: - * s/match/replace/gIp + * A substitution command should look something like this: + * s/match/replace/ #gIpw * || | ||| * mandatory optional - * - * (all three of the '/' slashes are mandatory) */ - idx = parse_regex_delim(substr, &match, &sed_cmd->replace); + idx = parse_regex_delim(substr, &match, &sed_cmd->string); /* determine the number of back references in the match string */ /* Note: we compute this here rather than in the do_subst_command() * function to save processor time, at the expense of a little more memory * (4 bits) per sed_cmd */ - /* sed_cmd->num_backrefs = 0; *//* XXX: not needed? --apparently not */ - for (j = 0; match[j]; j++) { - /* GNU/POSIX sed does not save more than nine backrefs */ - if (match[j] == '\\' && match[j + 1] == '(' - && sed_cmd->num_backrefs <= 9) - sed_cmd->num_backrefs++; - } - /* process the flags */ + + sed_cmd->which_match = 1; while (substr[++idx]) { + /* Parse match number */ + if (isdigit(substr[idx])) { + if (match[0] != '^') { + /* Match 0 treated as all, multiple matches we take the last one. */ + const char *pos = substr + idx; +/* FIXME: error check? */ + sed_cmd->which_match = (unsigned)strtol(substr+idx, (char**) &pos, 10); + idx = pos - substr; + } + continue; + } + /* Skip spaces */ + if (isspace(substr[idx])) continue; + switch (substr[idx]) { + /* Replace all occurrences */ case 'g': - sed_cmd->sub_g = 1; - break; - /* Hmm, i dont see the I option mentioned in the standard */ - case 'I': - cflags |= REG_ICASE; + if (match[0] != '^') + sed_cmd->which_match = 0; break; + /* Print pattern space */ case 'p': sed_cmd->sub_p = 1; break; + /* Write to file */ + case 'w': + { + char *temp; + idx += parse_file_cmd(/*sed_cmd,*/ substr+idx, &temp); + break; + } + /* Ignore case (gnu exension) */ + case 'I': + cflags |= REG_ICASE; + break; + /* Comment */ + case '#': + while (substr[++idx]) /*skip all*/; + /* Fall through */ + /* End of command */ + case ';': + case '}': + goto out; default: - /* any whitespace or semicolon trailing after a s/// is ok */ - if (strchr(semicolon_whitespace, substr[idx])) - goto out; - /* else */ bb_error_msg_and_die("bad option in substitution expression"); } } - - out: +out: /* compile the match string into a regex */ if (*match != '\0') { /* If match is empty, we use last regex used at runtime */ - sed_cmd->sub_match = (regex_t *) xmalloc(sizeof(regex_t)); + sed_cmd->sub_match = xmalloc(sizeof(regex_t)); xregcomp(sed_cmd->sub_match, match, cflags); } free(match); @@ -326,855 +406,943 @@ static int parse_subst_cmd(sed_cmd_t * const sed_cmd, const char *substr) return idx; } -static void replace_slash_n(char *string) -{ - int i; - int remaining = strlen(string); - - for (i = 0; string[i]; i++) { - if ((string[i] == '\\') && (string[i + 1] == 'n')) { - string[i] = '\n'; - memmove(string + i + 1, string + i + 1, remaining - 1); - } else { - remaining--; - } - } -} - -static int parse_translate_cmd(sed_cmd_t * const sed_cmd, const char *cmdstr) -{ - char *match; - char *replace; - int idx; - int i; - - idx = parse_regex_delim(cmdstr, &match, &replace); - replace_slash_n(match); - replace_slash_n(replace); - sed_cmd->translate = xcalloc(1, (strlen(match) + 1) * 2); - for (i = 0; (match[i] != 0) && (replace[i] != 0); i++) { - sed_cmd->translate[i * 2] = match[i]; - sed_cmd->translate[(i * 2) + 1] = replace[i]; - } - return (idx + 1); -} - -static int parse_edit_cmd(sed_cmd_t * sed_cmd, const char *editstr) -{ - int i, j; - - /* - * the string that gets passed to this function should look like this: - * - * need one of these - * | - * | this backslash (immediately following the edit command) is mandatory - * | | - * [aic]\ - * TEXT1\ - * TEXT2\ - * TEXTN - * - * as soon as we hit a TEXT line that has no trailing '\', we're done. - * this means a command like: - * - * i\ - * INSERTME - * - * is a-ok. - * - */ - if ((*editstr != '\\') || ((editstr[1] != '\n') && (editstr[1] != '\r'))) { - bb_error_msg_and_die("bad format in edit expression"); - } - - /* store the edit line text */ - sed_cmd->editline = xmalloc(strlen(&editstr[2]) + 2); - for (i = 2, j = 0; - editstr[i] != '\0' && strchr("\r\n", editstr[i]) == NULL; i++, j++) { - if ((editstr[i] == '\\') && strchr("\n\r", editstr[i + 1]) != NULL) { - sed_cmd->editline[j] = '\n'; - i++; - } else - sed_cmd->editline[j] = editstr[i]; - } - - /* figure out if we need to add a newline */ - if (sed_cmd->editline[j - 1] != '\n') - sed_cmd->editline[j++] = '\n'; - - /* terminate string */ - sed_cmd->editline[j] = '\0'; - - return i; -} - - -static int parse_file_cmd(sed_cmd_t * sed_cmd, const char *filecmdstr) -{ - int idx = 0; - int filenamelen = 0; - - /* - * the string that gets passed to this function should look like this: - * '[ ]filename' - * | | - * | a filename - * | - * optional whitespace - - * re: the file to be read, the GNU manual says the following: "Note that - * if filename cannot be read, it is treated as if it were an empty file, - * without any error indication." Thus, all of the following commands are - * perfectly leagal: - * - * sed -e '1r noexist' - * sed -e '1r ;' - * sed -e '1r' - */ - - /* the file command may be followed by whitespace; move past it. */ - while (isspace(filecmdstr[++idx])) {; - } - - /* the first non-whitespace we get is a filename. the filename ends when we - * hit a normal sed command terminator or end of string */ - filenamelen = strcspn(&filecmdstr[idx], semicolon_whitespace); - sed_cmd->filename = xmalloc(filenamelen + 1); - safe_strncpy(sed_cmd->filename, &filecmdstr[idx], filenamelen + 1); - return idx + filenamelen; -} - /* * Process the commands arguments */ -static char *parse_cmd_str(sed_cmd_t * sed_cmd, char *cmdstr) +static const char *parse_cmd_args(sed_cmd_t *sed_cmd, const char *cmdstr) { /* handle (s)ubstitution command */ - if (sed_cmd->cmd == 's') { + if (sed_cmd->cmd == 's') cmdstr += parse_subst_cmd(sed_cmd, cmdstr); - } /* handle edit cmds: (a)ppend, (i)nsert, and (c)hange */ else if (strchr("aic", sed_cmd->cmd)) { if ((sed_cmd->end_line || sed_cmd->end_match) && sed_cmd->cmd != 'c') bb_error_msg_and_die ("only a beginning address can be specified for edit commands"); - cmdstr += parse_edit_cmd(sed_cmd, cmdstr); - } + for (;;) { + if (*cmdstr == '\n' || *cmdstr == '\\') { + cmdstr++; + break; + } else if (isspace(*cmdstr)) + cmdstr++; + else + break; + } + sed_cmd->string = xstrdup(cmdstr); + /* "\anychar" -> "anychar" */ + parse_escapes(sed_cmd->string, sed_cmd->string, strlen(cmdstr), '\0', '\0'); + cmdstr += strlen(cmdstr); /* handle file cmds: (r)ead */ - else if (sed_cmd->cmd == 'r') { + } else if (strchr("rw", sed_cmd->cmd)) { if (sed_cmd->end_line || sed_cmd->end_match) - bb_error_msg_and_die("Command only uses one address"); - cmdstr += parse_file_cmd(sed_cmd, cmdstr); - } + bb_error_msg_and_die("command only uses one address"); + cmdstr += parse_file_cmd(/*sed_cmd,*/ cmdstr, &sed_cmd->string); + if (sed_cmd->cmd == 'w') { + sed_cmd->sw_file = xfopen_for_write(sed_cmd->string); + sed_cmd->sw_last_char = '\n'; + } /* handle branch commands */ - else if (strchr(":bt", sed_cmd->cmd)) { + } else if (strchr(":btT", sed_cmd->cmd)) { int length; - cmdstr += strspn(cmdstr, " "); - length = strcspn(cmdstr, "; \n"); - sed_cmd->label = strndup(cmdstr, length); - cmdstr += length; + cmdstr = skip_whitespace(cmdstr); + length = strcspn(cmdstr, semicolon_whitespace); + if (length) { + sed_cmd->string = xstrndup(cmdstr, length); + cmdstr += length; + } } /* translation command */ else if (sed_cmd->cmd == 'y') { - cmdstr += parse_translate_cmd(sed_cmd, cmdstr); + char *match, *replace; + int i = cmdstr[0]; + + cmdstr += parse_regex_delim(cmdstr, &match, &replace)+1; + /* \n already parsed, but \delimiter needs unescaping. */ + parse_escapes(match, match, strlen(match), i, i); + parse_escapes(replace, replace, strlen(replace), i, i); + + sed_cmd->string = xzalloc((strlen(match) + 1) * 2); + for (i = 0; match[i] && replace[i]; i++) { + sed_cmd->string[i*2] = match[i]; + sed_cmd->string[i*2+1] = replace[i]; + } + free(match); + free(replace); } /* if it wasnt a single-letter command that takes no arguments * then it must be an invalid command. */ - else if (strchr("dgGhHnNpPqx={}", sed_cmd->cmd) == 0) { - bb_error_msg_and_die("Unsupported command %c", sed_cmd->cmd); + else if (strchr("dDgGhHlnNpPqx={}", sed_cmd->cmd) == 0) { + bb_error_msg_and_die("unsupported command %c", sed_cmd->cmd); } /* give back whatever's left over */ - return (cmdstr); + return cmdstr; } -static char *add_cmd(sed_cmd_t * sed_cmd, char *cmdstr) + +/* Parse address+command sets, skipping comment lines. */ + +static void add_cmd(const char *cmdstr) { - /* Skip over leading whitespace and semicolons */ - cmdstr += strspn(cmdstr, semicolon_whitespace); + sed_cmd_t *sed_cmd; + int temp; - /* if we ate the whole thing, that means there was just trailing - * whitespace or a final / no-op semicolon. either way, get out */ - if (*cmdstr == '\0') { - return (NULL); + /* Append this line to any unfinished line from last time. */ + if (G.add_cmd_line) { + char *tp = xasprintf("%s\n%s", G.add_cmd_line, cmdstr); + free(G.add_cmd_line); + cmdstr = G.add_cmd_line = tp; } - /* if this is a comment, jump past it and keep going */ - if (*cmdstr == '#') { - /* "#n" is the same as using -n on the command line */ - if (cmdstr[1] == 'n') { - be_quiet++; - } - return (strpbrk(cmdstr, "\n\r")); + /* If this line ends with backslash, request next line. */ + temp = strlen(cmdstr); + if (temp && cmdstr[--temp] == '\\') { + if (!G.add_cmd_line) + G.add_cmd_line = xstrdup(cmdstr); + G.add_cmd_line[temp] = '\0'; + return; } - /* parse the command - * format is: [addr][,addr]cmd - * |----||-----||-| - * part1 part2 part3 - */ + /* Loop parsing all commands in this line. */ + while (*cmdstr) { + /* Skip leading whitespace and semicolons */ + cmdstr += strspn(cmdstr, semicolon_whitespace); + + /* If no more commands, exit. */ + if (!*cmdstr) break; + + /* if this is a comment, jump past it and keep going */ + if (*cmdstr == '#') { + /* "#n" is the same as using -n on the command line */ + if (cmdstr[1] == 'n') + G.be_quiet++; + cmdstr = strpbrk(cmdstr, "\n\r"); + if (!cmdstr) break; + continue; + } - /* first part (if present) is an address: either a '$', a number or a /regex/ */ - cmdstr += get_address(cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match); + /* parse the command + * format is: [addr][,addr][!]cmd + * |----||-----||-| + * part1 part2 part3 + */ - /* second part (if present) will begin with a comma */ - if (*cmdstr == ',') { - int idx; + sed_cmd = xzalloc(sizeof(sed_cmd_t)); - cmdstr++; - idx = get_address(cmdstr, &sed_cmd->end_line, &sed_cmd->end_match); - if (idx == 0) { - bb_error_msg_and_die("get_address: no address found in string\n" - "\t(you probably didn't check the string you passed me)"); - } - cmdstr += idx; - } + /* first part (if present) is an address: either a '$', a number or a /regex/ */ + cmdstr += get_address(cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match); - /* skip whitespace before the command */ - while (isspace(*cmdstr)) { - cmdstr++; - } - - /* there my be the inversion flag between part2 and part3 */ - if (*cmdstr == '!') { - sed_cmd->invert = 1; - cmdstr++; + /* second part (if present) will begin with a comma */ + if (*cmdstr == ',') { + int idx; -#ifdef SED_FEATURE_STRICT_CHECKING - /* According to the spec - * It is unspecified whether s can follow a '!' character, - * and conforming applications shall not follow a '!' character - * with s. - */ - if (isblank(cmdstr[idx]) { - bb_error_msg_and_die("blank follows '!'");} -#else - /* skip whitespace before the command */ - while (isspace(*cmdstr)) { cmdstr++; + idx = get_address(cmdstr, &sed_cmd->end_line, &sed_cmd->end_match); + if (!idx) + bb_error_msg_and_die("no address after comma"); + cmdstr += idx; } -#endif - } - /* last part (mandatory) will be a command */ - if (*cmdstr == '\0') - bb_error_msg_and_die("missing command"); - - sed_cmd->cmd = *cmdstr; - cmdstr++; + /* skip whitespace before the command */ + cmdstr = skip_whitespace(cmdstr); - cmdstr = parse_cmd_str(sed_cmd, cmdstr); + /* Check for inversion flag */ + if (*cmdstr == '!') { + sed_cmd->invert = 1; + cmdstr++; - /* Add the command to the command array */ - sed_cmd_tail->next = sed_cmd; - sed_cmd_tail = sed_cmd_tail->next; + /* skip whitespace before the command */ + cmdstr = skip_whitespace(cmdstr); + } - return (cmdstr); -} + /* last part (mandatory) will be a command */ + if (!*cmdstr) + bb_error_msg_and_die("missing command"); + sed_cmd->cmd = *(cmdstr++); + cmdstr = parse_cmd_args(sed_cmd, cmdstr); -static void add_cmd_str(char *cmdstr) -{ -#ifdef CONFIG_FEATURE_SED_EMBEDED_NEWLINE - char *cmdstr_ptr = cmdstr; - - /* HACK: convert "\n" to match tranlated '\n' string */ - while ((cmdstr_ptr = strstr(cmdstr_ptr, "\\n")) != NULL) { - cmdstr = xrealloc(cmdstr, strlen(cmdstr) + 2); - cmdstr_ptr = strstr(cmdstr, "\\n"); - memmove(cmdstr_ptr + 1, cmdstr_ptr, strlen(cmdstr_ptr) + 1); - cmdstr_ptr[0] = '\\'; - cmdstr_ptr += 3; + /* Add the command to the command array */ + G.sed_cmd_tail->next = sed_cmd; + G.sed_cmd_tail = G.sed_cmd_tail->next; } -#endif - do { - sed_cmd_t *sed_cmd; - - sed_cmd = xcalloc(1, sizeof(sed_cmd_t)); - cmdstr = add_cmd(sed_cmd, cmdstr); - } while (cmdstr && strlen(cmdstr)); -} - -static void load_cmd_file(char *filename) -{ - FILE *cmdfile; - char *line; - char *nextline; - char *e; - - cmdfile = bb_xfopen(filename, "r"); - - while ((line = bb_get_line_from_file(cmdfile)) != NULL) { - /* if a line ends with '\' it needs the next line appended to it */ - while (((e = last_char_is(line, '\n')) != NULL) - && (e > line) && (e[-1] == '\\') - && ((nextline = bb_get_line_from_file(cmdfile)) != NULL)) { - line = xrealloc(line, (e - line) + 1 + strlen(nextline) + 1); - strcat(line, nextline); - free(nextline); - } - /* eat trailing newline (if any) --if I don't do this, edit commands - * (aic) will print an extra newline */ - chomp(line); - add_cmd_str(line); - free(line); - } + /* If we glued multiple lines together, free the memory. */ + free(G.add_cmd_line); + G.add_cmd_line = NULL; } -struct pipeline { - char *buf; - int idx; - int len; -}; +/* Append to a string, reallocating memory as necessary. */ -#define PIPE_MAGIC 0x7f #define PIPE_GROW 64 -void pipe_putc(struct pipeline *const pipeline, char c) +static void pipe_putc(char c) { - if (pipeline->buf[pipeline->idx] == PIPE_MAGIC) { - pipeline->buf = xrealloc(pipeline->buf, pipeline->len + PIPE_GROW); - memset(pipeline->buf + pipeline->len, 0, PIPE_GROW); - pipeline->len += PIPE_GROW; - pipeline->buf[pipeline->len - 1] = PIPE_MAGIC; + if (G.pipeline.idx == G.pipeline.len) { + G.pipeline.buf = xrealloc(G.pipeline.buf, + G.pipeline.len + PIPE_GROW); + G.pipeline.len += PIPE_GROW; } - pipeline->buf[pipeline->idx++] = (c); + G.pipeline.buf[G.pipeline.idx++] = c; } -#define pipeputc(c) pipe_putc(pipeline, c) - -#if 0 -{ - if (pipeline[pipeline_idx] == PIPE_MAGIC) { - pipeline = xrealloc(pipeline, pipeline_len + PIPE_GROW); - memset(pipeline + pipeline_len, 0, PIPE_GROW); - pipeline_len += PIPE_GROW; - pipeline[pipeline_len - 1] = PIPE_MAGIC; - } - pipeline[pipeline_idx++] = (c); -} -#endif - -static void print_subst_w_backrefs(const char *line, const char *replace, - regmatch_t * regmatch, struct pipeline *const pipeline, int matches) +static void do_subst_w_backrefs(char *line, char *replace) { - int i; + int i,j; /* go through the replacement string */ for (i = 0; replace[i]; i++) { /* if we find a backreference (\1, \2, etc.) print the backref'ed * text */ - if (replace[i] == '\\' && isdigit(replace[i + 1])) { - int j; - char tmpstr[2]; - int backref; - - ++i; /* i now indexes the backref number, instead of the leading slash */ - tmpstr[0] = replace[i]; - tmpstr[1] = 0; - backref = atoi(tmpstr); - /* print out the text held in regmatch[backref] */ - if (backref <= matches && regmatch[backref].rm_so != -1) - for (j = regmatch[backref].rm_so; j < regmatch[backref].rm_eo; - j++) - pipeputc(line[j]); - } - - /* if we find a backslash escaped character, print the character */ - else if (replace[i] == '\\') { - ++i; - pipeputc(replace[i]); + if (replace[i] == '\\') { + unsigned backref = replace[++i] - '0'; + if (backref <= 9) { + /* print out the text held in G.regmatch[backref] */ + if (G.regmatch[backref].rm_so != -1) { + j = G.regmatch[backref].rm_so; + while (j < G.regmatch[backref].rm_eo) + pipe_putc(line[j++]); + } + continue; + } + /* I _think_ it is impossible to get '\' to be + * the last char in replace string. Thus we dont check + * for replace[i] == NUL. (counterexample anyone?) */ + /* if we find a backslash escaped character, print the character */ + pipe_putc(replace[i]); + continue; } - - /* if we find an unescaped '&' print out the whole matched text. - * fortunately, regmatch[0] contains the indicies to the whole matched - * expression (kinda seems like it was designed for just such a - * purpose...) */ - else if (replace[i] == '&') { - int j; - - for (j = regmatch[0].rm_so; j < regmatch[0].rm_eo; j++) - pipeputc(line[j]); + /* if we find an unescaped '&' print out the whole matched text. */ + if (replace[i] == '&') { + j = G.regmatch[0].rm_so; + while (j < G.regmatch[0].rm_eo) + pipe_putc(line[j++]); + continue; } - /* nothing special, just print this char of the replacement string to stdout */ - else - pipeputc(replace[i]); + /* Otherwise just output the character. */ + pipe_putc(replace[i]); } } -static int do_subst_command(sed_cmd_t * sed_cmd, char **line) +static int do_subst_command(sed_cmd_t *sed_cmd, char **line) { - char *hackline = *line; - struct pipeline thepipe = { NULL, 0, 0 }; - struct pipeline *const pipeline = &thepipe; + char *oldline = *line; int altered = 0; - int result; - regmatch_t *regmatch = NULL; + unsigned match_count = 0; regex_t *current_regex; + /* Handle empty regex. */ if (sed_cmd->sub_match == NULL) { - current_regex = previous_regex_ptr; - } else { - previous_regex_ptr = current_regex = sed_cmd->sub_match; - } - result = regexec(current_regex, hackline, 0, NULL, 0); - - /* we only proceed if the substitution 'search' expression matches */ - if (result == REG_NOMATCH) { + current_regex = G.previous_regex_ptr; + if (!current_regex) + bb_error_msg_and_die("no previous regexp"); + } else + G.previous_regex_ptr = current_regex = sed_cmd->sub_match; + + /* Find the first match */ + if (REG_NOMATCH == regexec(current_regex, oldline, 10, G.regmatch, 0)) return 0; - } - /* whaddaya know, it matched. get the number of back references */ - regmatch = xmalloc(sizeof(regmatch_t) * (sed_cmd->num_backrefs + 1)); - - /* allocate more PIPE_GROW bytes - if replaced string is larger than original */ - thepipe.len = strlen(hackline) + PIPE_GROW; - thepipe.buf = xcalloc(1, thepipe.len); - /* buffer magic */ - thepipe.buf[thepipe.len - 1] = PIPE_MAGIC; - - /* and now, as long as we've got a line to try matching and if we can match - * the search string, we make substitutions */ - while ((*hackline || !altered) - && (regexec(current_regex, hackline, sed_cmd->num_backrefs + 1, - regmatch, 0) != REG_NOMATCH)) { + /* Initialize temporary output buffer. */ + G.pipeline.buf = xmalloc(PIPE_GROW); + G.pipeline.len = PIPE_GROW; + G.pipeline.idx = 0; + + /* Now loop through, substituting for matches */ + do { int i; + /* Work around bug in glibc regexec, demonstrated by: + echo " a.b" | busybox sed 's [^ .]* x g' + The match_count check is so not to break + echo "hi" | busybox sed 's/^/!/g' */ + if (!G.regmatch[0].rm_so && !G.regmatch[0].rm_eo && match_count) { + pipe_putc(*oldline++); + continue; + } + + match_count++; + + /* If we aren't interested in this match, output old line to + end of match and continue */ + if (sed_cmd->which_match + && (sed_cmd->which_match != match_count) + ) { + for (i = 0; i < G.regmatch[0].rm_eo; i++) + pipe_putc(*oldline++); + continue; + } + /* print everything before the match */ - for (i = 0; i < regmatch[0].rm_so; i++) - pipeputc(hackline[i]); + for (i = 0; i < G.regmatch[0].rm_so; i++) + pipe_putc(oldline[i]); /* then print the substitution string */ - print_subst_w_backrefs(hackline, sed_cmd->replace, regmatch, pipeline, - sed_cmd->num_backrefs); + do_subst_w_backrefs(oldline, sed_cmd->string); /* advance past the match */ - hackline += regmatch[0].rm_eo; + oldline += G.regmatch[0].rm_eo; /* flag that something has changed */ altered++; /* if we're not doing this globally, get out now */ - if (!sed_cmd->sub_g) { + if (sed_cmd->which_match) break; - } - } - for (; *hackline; hackline++) - pipeputc(*hackline); - if (thepipe.buf[thepipe.idx] == PIPE_MAGIC) - thepipe.buf[thepipe.idx] = 0; + } while (*oldline && (regexec(current_regex, oldline, 10, G.regmatch, 0) != REG_NOMATCH)); + + /* Copy rest of string into output pipeline */ - /* cleanup */ - free(regmatch); + while (*oldline) + pipe_putc(*oldline++); + pipe_putc(0); free(*line); - *line = thepipe.buf; + *line = G.pipeline.buf; return altered; } -static sed_cmd_t *branch_to(const char *label) +/* Set command pointer to point to this label. (Does not handle null label.) */ +static sed_cmd_t *branch_to(char *label) { sed_cmd_t *sed_cmd; - for (sed_cmd = sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) { - if ((sed_cmd->label) && (strcmp(sed_cmd->label, label) == 0)) { - break; + for (sed_cmd = G.sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) { + if (sed_cmd->cmd == ':' && sed_cmd->string && !strcmp(sed_cmd->string, label)) { + return sed_cmd; } } + bb_error_msg_and_die("can't find label for jump to '%s'", label); +} - /* If no match returns last command */ - return (sed_cmd); +static void append(char *s) +{ + llist_add_to_end(&G.append_head, xstrdup(s)); } -static void process_file(FILE * file) +static void flush_append(void) { - char *pattern_space; /* Posix requires it be able to hold at least 8192 bytes */ - char *hold_space = NULL; /* Posix requires it be able to hold at least 8192 bytes */ - static int linenum = 0; /* GNU sed does not restart counting lines at EOF */ - unsigned int still_in_range = 0; - int altered; - int force_print; - - pattern_space = bb_get_chomped_line_from_file(file); - if (pattern_space == NULL) { - return; - } - - /* go through every line in the file */ - do { - char *next_line; - sed_cmd_t *sed_cmd; - int substituted = 0; - /* This enables whole blocks of commands to be mask'ed out if the lead address doesnt match */ - int block_mask = 1; - - /* Read one line in advance so we can act on the last line, the '$' address */ - next_line = bb_get_chomped_line_from_file(file); - - linenum++; - altered = 0; - force_print = 0; - - /* for every line, go through all the commands */ - for (sed_cmd = sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) { - int deleted = 0; - - /* - * entry point into sedding... - */ - int matched = ( - /* no range necessary */ - (sed_cmd->beg_line == 0 && sed_cmd->end_line == 0 - && sed_cmd->beg_match == NULL - && sed_cmd->end_match == NULL) || - /* this line number is the first address we're looking for */ - (sed_cmd->beg_line > 0 && (sed_cmd->beg_line == linenum)) || - /* this line matches our first address regex */ - (sed_cmd->beg_match - && (regexec(sed_cmd->beg_match, pattern_space, 0, NULL, - 0) == 0)) || - /* we are currently within the beginning & ending address range */ - still_in_range || ((sed_cmd->beg_line == -1) - && (next_line == NULL)) - ); - - if (sed_cmd->cmd == '{') { - block_mask = block_mask & matched; - } -// matched &= block_mask; + char *data; - if (sed_cmd->invert ^ (matched & block_mask)) { - /* Update last used regex incase a blank substitute BRE is found */ - if (sed_cmd->beg_match) { - previous_regex_ptr = sed_cmd->beg_match; - } + /* Output appended lines. */ + while ((data = (char *)llist_pop(&G.append_head))) { + fprintf(G.nonstdout, "%s\n", data); + free(data); + } +} - /* - * actual sedding - */ - switch (sed_cmd->cmd) { - case '=': - printf("%d\n", linenum); - break; - case 'P':{ - /* Write the current pattern space upto the first newline */ - char *tmp = strchr(pattern_space, '\n'); +static void add_input_file(FILE *file) +{ + G.input_file_list = xrealloc_vector(G.input_file_list, 2, G.input_file_count); + G.input_file_list[G.input_file_count++] = file; +} - if (tmp) { - *tmp = '\0'; - } +/* Get next line of input from G.input_file_list, flushing append buffer and + * noting if we ran out of files without a newline on the last line we read. + */ +enum { + NO_EOL_CHAR = 1, + LAST_IS_NUL = 2, +}; +static char *get_next_line(char *gets_char) +{ + char *temp = NULL; + int len; + char gc; + + flush_append(); + + /* will be returned if last line in the file + * doesn't end with either '\n' or '\0' */ + gc = NO_EOL_CHAR; + while (G.current_input_file < G.input_file_count) { + FILE *fp = G.input_file_list[G.current_input_file]; + /* Read line up to a newline or NUL byte, inclusive, + * return malloc'ed char[]. length of the chunk read + * is stored in len. NULL if EOF/error */ + temp = bb_get_chunk_from_file(fp, &len); + if (temp) { + /* len > 0 here, it's ok to do temp[len-1] */ + char c = temp[len-1]; + if (c == '\n' || c == '\0') { + temp[len-1] = '\0'; + gc = c; + if (c == '\0') { + int ch = fgetc(fp); + if (ch != EOF) + ungetc(ch, fp); + else + gc = LAST_IS_NUL; } - case 'p': /* Write the current pattern space to output */ - puts(pattern_space); - break; - case 'd': - altered++; - deleted = 1; - break; + } + /* else we put NO_EOL_CHAR into *gets_char */ + break; - case 's': - - /* - * Some special cases for 's' printing to make it compliant with - * GNU sed printing behavior (aka "The -n | s///p Matrix"): - * - * -n ONLY = never print anything regardless of any successful - * substitution - * - * s///p ONLY = always print successful substitutions, even if - * the pattern_space is going to be printed anyway (pattern_space - * will be printed twice). - * - * -n AND s///p = print ONLY a successful substitution ONE TIME; - * no other lines are printed - this is the reason why the 'p' - * flag exists in the first place. - */ - -#ifdef CONFIG_FEATURE_SED_EMBEDED_NEWLINE - /* HACK: escape newlines twice so regex can match them */ - { - int offset = 0; - - while (strchr(pattern_space + offset, '\n') != NULL) { - char *tmp; - - pattern_space = - xrealloc(pattern_space, - strlen(pattern_space) + 2); - tmp = strchr(pattern_space + offset, '\n'); - memmove(tmp + 1, tmp, strlen(tmp) + 1); - tmp[0] = '\\'; - tmp[1] = 'n'; - offset = tmp - pattern_space + 2; - } - } -#endif - /* we print the pattern_space once, unless we were told to be quiet */ - substituted = do_subst_command(sed_cmd, &pattern_space); + /* NB: I had the idea of peeking next file(s) and returning + * NO_EOL_CHAR only if it is the *last* non-empty + * input file. But there is a case where this won't work: + * file1: "a woo\nb woo" + * file2: "c no\nd no" + * sed -ne 's/woo/bang/p' input1 input2 => "a bang\nb bang" + * (note: *no* newline after "b bang"!) */ + } + /* Close this file and advance to next one */ + fclose(fp); + G.current_input_file++; + } + *gets_char = gc; + return temp; +} -#ifdef CONFIG_FEATURE_SED_EMBEDED_NEWLINE - /* undo HACK: escape newlines twice so regex can match them */ - { - char *tmp = pattern_space; +/* Output line of text. */ +/* Note: + * The tricks with NO_EOL_CHAR and last_puts_char are there to emulate gnu sed. + * Without them, we had this: + * echo -n thingy >z1 + * echo -n again >z2 + * >znull + * sed "s/i/z/" z1 z2 znull | hexdump -vC + * output: + * gnu sed 4.1.5: + * 00000000 74 68 7a 6e 67 79 0a 61 67 61 7a 6e |thzngy.agazn| + * bbox: + * 00000000 74 68 7a 6e 67 79 61 67 61 7a 6e |thzngyagazn| + */ +static void puts_maybe_newline(char *s, FILE *file, char *last_puts_char, char last_gets_char) +{ + char lpc = *last_puts_char; - while ((tmp = strstr(tmp, "\\n")) != NULL) { - memmove(tmp, tmp + 1, strlen(tmp + 1) + 1); - tmp[0] = '\n'; - } - } -#endif - altered |= substituted; - if (!be_quiet && altered && ((sed_cmd->next == NULL) - || (sed_cmd->next->cmd != 's'))) { - force_print = 1; - } + /* Need to insert a '\n' between two files because first file's + * last line wasn't terminated? */ + if (lpc != '\n' && lpc != '\0') { + fputc('\n', file); + lpc = '\n'; + } + fputs(s, file); + + /* 'x' - just something which is not '\n', '\0' or NO_EOL_CHAR */ + if (s[0]) + lpc = 'x'; + + /* had trailing '\0' and it was last char of file? */ + if (last_gets_char == LAST_IS_NUL) { + fputc('\0', file); + lpc = 'x'; /* */ + } else + /* had trailing '\n' or '\0'? */ + if (last_gets_char != NO_EOL_CHAR) { + fputc(last_gets_char, file); + lpc = last_gets_char; + } - /* we also print the line if we were given the 'p' flag - * (this is quite possibly the second printing) */ - if ((sed_cmd->sub_p) && altered) { - puts(pattern_space); - } - break; - case 'a': - puts(pattern_space); - fputs(sed_cmd->editline, stdout); - altered++; - break; + if (ferror(file)) { + xfunc_error_retval = 4; /* It's what gnu sed exits with... */ + bb_error_msg_and_die(bb_msg_write_error); + } + *last_puts_char = lpc; +} - case 'i': - fputs(sed_cmd->editline, stdout); - break; +#define sed_puts(s, n) (puts_maybe_newline(s, G.nonstdout, &last_puts_char, n)) - case 'c': - /* single-address case */ - if ((sed_cmd->end_match == NULL && sed_cmd->end_line == 0) - /* multi-address case */ - /* - matching text */ - || (sed_cmd->end_match - && (regexec(sed_cmd->end_match, pattern_space, 0, - NULL, 0) == 0)) - /* - matching line numbers */ - || (sed_cmd->end_line > 0 - && sed_cmd->end_line == linenum)) { - fputs(sed_cmd->editline, stdout); - } - altered++; +static int beg_match(sed_cmd_t *sed_cmd, const char *pattern_space) +{ + int retval = sed_cmd->beg_match && !regexec(sed_cmd->beg_match, pattern_space, 0, NULL, 0); + if (retval) + G.previous_regex_ptr = sed_cmd->beg_match; + return retval; +} - break; +/* Process all the lines in all the files */ - case 'r':{ - FILE *outfile; - - outfile = fopen(sed_cmd->filename, "r"); - if (outfile) { - char *line; - - while ((line = - bb_get_chomped_line_from_file(outfile)) != - NULL) { - pattern_space = - xrealloc(pattern_space, - strlen(line) + strlen(pattern_space) + 2); - strcat(pattern_space, "\n"); - strcat(pattern_space, line); - } - bb_xprint_and_close_file(outfile); - } +static void process_files(void) +{ + char *pattern_space, *next_line; + int linenum = 0; + char last_puts_char = '\n'; + char last_gets_char, next_gets_char; + sed_cmd_t *sed_cmd; + int substituted; + + /* Prime the pump */ + next_line = get_next_line(&next_gets_char); + + /* go through every line in each file */ + again: + substituted = 0; + + /* Advance to next line. Stop if out of lines. */ + pattern_space = next_line; + if (!pattern_space) return; + last_gets_char = next_gets_char; + + /* Read one line in advance so we can act on the last line, + * the '$' address */ + next_line = get_next_line(&next_gets_char); + linenum++; + restart: + /* for every line, go through all the commands */ + for (sed_cmd = G.sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) { + int old_matched, matched; + + old_matched = sed_cmd->in_match; + + /* Determine if this command matches this line: */ + + /* Are we continuing a previous multi-line match? */ + sed_cmd->in_match = sed_cmd->in_match + /* Or is no range necessary? */ + || (!sed_cmd->beg_line && !sed_cmd->end_line + && !sed_cmd->beg_match && !sed_cmd->end_match) + /* Or did we match the start of a numerical range? */ + || (sed_cmd->beg_line > 0 && (sed_cmd->beg_line == linenum)) + /* Or does this line match our begin address regex? */ + || (beg_match(sed_cmd, pattern_space)) + /* Or did we match last line of input? */ + || (sed_cmd->beg_line == -1 && next_line == NULL); + + /* Snapshot the value */ + + matched = sed_cmd->in_match; + + /* Is this line the end of the current match? */ + + if (matched) { + sed_cmd->in_match = !( + /* has the ending line come, or is this a single address command? */ + (sed_cmd->end_line ? + sed_cmd->end_line == -1 ? + !next_line + : (sed_cmd->end_line <= linenum) + : !sed_cmd->end_match + ) + /* or does this line matches our last address regex */ + || (sed_cmd->end_match && old_matched + && (regexec(sed_cmd->end_match, + pattern_space, 0, NULL, 0) == 0)) + ); + } + /* Skip blocks of commands we didn't match. */ + if (sed_cmd->cmd == '{') { + if (sed_cmd->invert ? matched : !matched) { + while (sed_cmd->cmd != '}') { + sed_cmd = sed_cmd->next; + if (!sed_cmd) + bb_error_msg_and_die("unterminated {"); } - break; - case 'q': /* Branch to end of script and quit */ - deleted = 1; - /* Exit the outer while loop */ - free(next_line); - next_line = NULL; - break; - case 'n': /* Read next line from input */ - free(pattern_space); - pattern_space = next_line; - next_line = bb_get_chomped_line_from_file(file); - linenum++; - break; - case 'N': /* Append the next line to the current line */ - if (next_line) { - pattern_space = - realloc(pattern_space, - strlen(pattern_space) + strlen(next_line) + 2); - strcat(pattern_space, "\n"); - strcat(pattern_space, next_line); - next_line = bb_get_chomped_line_from_file(file); - linenum++; - } - break; - case 'b': - sed_cmd = branch_to(sed_cmd->label); - break; - case 't': - if (substituted) { - sed_cmd = branch_to(sed_cmd->label); - } - break; - case 'y':{ - int i; + } + continue; + } + + /* Okay, so did this line match? */ + if (sed_cmd->invert ? !matched : matched) { + /* Update last used regex in case a blank substitute BRE is found */ + if (sed_cmd->beg_match) { + G.previous_regex_ptr = sed_cmd->beg_match; + } - for (i = 0; pattern_space[i] != 0; i++) { - int j; + /* actual sedding */ + switch (sed_cmd->cmd) { - for (j = 0; sed_cmd->translate[j]; j += 2) { - if (pattern_space[i] == sed_cmd->translate[j]) { - pattern_space[i] = sed_cmd->translate[j + 1]; - } - } - } - } + /* Print line number */ + case '=': + fprintf(G.nonstdout, "%d\n", linenum); + break; + + /* Write the current pattern space up to the first newline */ + case 'P': + { + char *tmp = strchr(pattern_space, '\n'); + + if (tmp) { + *tmp = '\0'; + /* TODO: explain why '\n' below */ + sed_puts(pattern_space, '\n'); + *tmp = '\n'; break; - case 'g': /* Replace pattern space with hold space */ + } + /* Fall Through */ + } + + /* Write the current pattern space to output */ + case 'p': + /* NB: we print this _before_ the last line + * (of current file) is printed. Even if + * that line is nonterminated, we print + * '\n' here (gnu sed does the same) */ + sed_puts(pattern_space, '\n'); + break; + /* Delete up through first newline */ + case 'D': + { + char *tmp = strchr(pattern_space, '\n'); + + if (tmp) { + tmp = xstrdup(tmp+1); free(pattern_space); - pattern_space = strdup(hold_space); - break; - case 'G': { /* Append newline and hold space to pattern space */ - int pattern_space_size = 0; - if (pattern_space) { - pattern_space_size = strlen(pattern_space); - } - pattern_space = xrealloc(pattern_space, pattern_space_size + strlen(hold_space) + 2); - strcat(pattern_space, "\n"); - strcat(pattern_space, hold_space); - break; + pattern_space = tmp; + goto restart; } - case 'h': /* Replace hold space with pattern space */ - free(hold_space); - hold_space = strdup(pattern_space); - break; - case 'H': { /* Append newline and pattern space to hold space */ - int hold_space_size = 0; - if (hold_space) { - hold_space_size = strlen(hold_space); - } - hold_space = xrealloc(hold_space, hold_space_size + strlen(pattern_space) + 2); - strcat(hold_space, "\n"); - strcat(hold_space, pattern_space); + } + /* discard this line. */ + case 'd': + goto discard_line; + + /* Substitute with regex */ + case 's': + if (!do_subst_command(sed_cmd, &pattern_space)) break; + substituted |= 1; + + /* handle p option */ + if (sed_cmd->sub_p) + sed_puts(pattern_space, last_gets_char); + /* handle w option */ + if (sed_cmd->sw_file) + puts_maybe_newline( + pattern_space, sed_cmd->sw_file, + &sed_cmd->sw_last_char, last_gets_char); + break; + + /* Append line to linked list to be printed later */ + case 'a': + append(sed_cmd->string); + break; + + /* Insert text before this line */ + case 'i': + sed_puts(sed_cmd->string, '\n'); + break; + + /* Cut and paste text (replace) */ + case 'c': + /* Only triggers on last line of a matching range. */ + if (!sed_cmd->in_match) + sed_puts(sed_cmd->string, NO_EOL_CHAR); + goto discard_line; + + /* Read file, append contents to output */ + case 'r': + { + FILE *rfile; + + rfile = fopen_for_read(sed_cmd->string); + if (rfile) { + char *line; + + while ((line = xmalloc_fgetline(rfile)) + != NULL) + append(line); + xprint_and_close_file(rfile); } - case 'x':{ - /* Swap hold and pattern space */ - char *tmp = pattern_space; - pattern_space = hold_space; - hold_space = tmp; + + break; + } + + /* Write pattern space to file. */ + case 'w': + puts_maybe_newline( + pattern_space, sed_cmd->sw_file, + &sed_cmd->sw_last_char, last_gets_char); + break; + + /* Read next line from input */ + case 'n': + if (!G.be_quiet) + sed_puts(pattern_space, last_gets_char); + if (next_line) { + free(pattern_space); + pattern_space = next_line; + last_gets_char = next_gets_char; + next_line = get_next_line(&next_gets_char); + substituted = 0; + linenum++; break; } + /* fall through */ + + /* Quit. End of script, end of input. */ + case 'q': + /* Exit the outer while loop */ + free(next_line); + next_line = NULL; + goto discard_commands; + + /* Append the next line to the current line */ + case 'N': + { + int len; + /* If no next line, jump to end of script and exit. */ + if (next_line == NULL) { + /* Jump to end of script and exit */ + free(next_line); + next_line = NULL; + goto discard_line; + /* append next_line, read new next_line. */ } + len = strlen(pattern_space); + pattern_space = realloc(pattern_space, len + strlen(next_line) + 2); + pattern_space[len] = '\n'; + strcpy(pattern_space + len+1, next_line); + last_gets_char = next_gets_char; + next_line = get_next_line(&next_gets_char); + linenum++; + break; } - /* - * exit point from sedding... - */ - if (matched) { - if ( - /* this is a single-address command or... */ - (sed_cmd->end_line == 0 && sed_cmd->end_match == NULL) - /* If only one address */ - /* we were in the middle of our address range (this - * isn't the first time through) and.. */ - || ((still_in_range == 1) - /* this line number is the last address we're looking for or... */ - && ((sed_cmd->end_line > 0 - && (sed_cmd->end_line == linenum)) - /* this line matches our last address regex */ - || (sed_cmd->end_match - && (regexec(sed_cmd->end_match, pattern_space, - 0, NULL, 0) == 0))))) { - /* we're out of our address range */ - still_in_range = 0; - } else { - /* didn't hit the exit? then we're still in the middle of an address range */ - still_in_range = 1; + /* Test/branch if substitution occurred */ + case 't': + if (!substituted) break; + substituted = 0; + /* Fall through */ + /* Test/branch if substitution didn't occur */ + case 'T': + if (substituted) break; + /* Fall through */ + /* Branch to label */ + case 'b': + if (!sed_cmd->string) goto discard_commands; + else sed_cmd = branch_to(sed_cmd->string); + break; + /* Transliterate characters */ + case 'y': + { + int i, j; + + for (i = 0; pattern_space[i]; i++) { + for (j = 0; sed_cmd->string[j]; j += 2) { + if (pattern_space[i] == sed_cmd->string[j]) { + pattern_space[i] = sed_cmd->string[j + 1]; + break; + } + } } + + break; } + case 'g': /* Replace pattern space with hold space */ + free(pattern_space); + pattern_space = xstrdup(G.hold_space ? G.hold_space : ""); + break; + case 'G': /* Append newline and hold space to pattern space */ + { + int pattern_space_size = 2; + int hold_space_size = 0; + + if (pattern_space) + pattern_space_size += strlen(pattern_space); + if (G.hold_space) + hold_space_size = strlen(G.hold_space); + pattern_space = xrealloc(pattern_space, + pattern_space_size + hold_space_size); + if (pattern_space_size == 2) + pattern_space[0] = 0; + strcat(pattern_space, "\n"); + if (G.hold_space) + strcat(pattern_space, G.hold_space); + last_gets_char = '\n'; - if (sed_cmd->cmd == '}') { - block_mask = 1; + break; } + case 'h': /* Replace hold space with pattern space */ + free(G.hold_space); + G.hold_space = xstrdup(pattern_space); + break; + case 'H': /* Append newline and pattern space to hold space */ + { + int hold_space_size = 2; + int pattern_space_size = 0; + + if (G.hold_space) + hold_space_size += strlen(G.hold_space); + if (pattern_space) + pattern_space_size = strlen(pattern_space); + G.hold_space = xrealloc(G.hold_space, + hold_space_size + pattern_space_size); + + if (hold_space_size == 2) + *G.hold_space = 0; + strcat(G.hold_space, "\n"); + if (pattern_space) + strcat(G.hold_space, pattern_space); - if (deleted) break; + } + case 'x': /* Exchange hold and pattern space */ + { + char *tmp = pattern_space; + pattern_space = G.hold_space ? : xzalloc(1); + last_gets_char = '\n'; + G.hold_space = tmp; + break; + } + } } + } - /* we will print the line unless we were told to be quiet or if the - * line was altered (via a 'd'elete or 's'ubstitution), in which case - * the altered line was already printed */ - if ((!be_quiet && !altered) || force_print) { - puts(pattern_space); + /* + * exit point from sedding... + */ + discard_commands: + /* we will print the line unless we were told to be quiet ('-n') + or if the line was suppressed (ala 'd'elete) */ + if (!G.be_quiet) + sed_puts(pattern_space, last_gets_char); + + /* Delete and such jump here. */ + discard_line: + flush_append(); + free(pattern_space); + + goto again; +} + +/* It is possible to have a command line argument with embedded + * newlines. This counts as multiple command lines. + * However, newline can be escaped: 's/e/z\z/' + * We check for this. + */ + +static void add_cmd_block(char *cmdstr) +{ + char *sv, *eol; + + cmdstr = sv = xstrdup(cmdstr); + do { + eol = strchr(cmdstr, '\n'); + next: + if (eol) { + /* Count preceding slashes */ + int slashes = 0; + char *sl = eol; + + while (sl != cmdstr && *--sl == '\\') + slashes++; + /* Odd number of preceding slashes - newline is escaped */ + if (slashes & 1) { + overlapping_strcpy(eol - 1, eol); + eol = strchr(eol, '\n'); + goto next; + } + *eol = '\0'; } - free(pattern_space); - pattern_space = next_line; - } while (pattern_space); + add_cmd(cmdstr); + cmdstr = eol + 1; + } while (eol); + free(sv); } -extern int sed_main(int argc, char **argv) +int sed_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE; +int sed_main(int argc UNUSED_PARAM, char **argv) { - int opt, status = EXIT_SUCCESS; + enum { + OPT_in_place = 1 << 0, + }; + unsigned opt; + llist_t *opt_e, *opt_f; + int status = EXIT_SUCCESS; + + INIT_G(); -#ifdef CONFIG_FEATURE_CLEAN_UP /* destroy command strings on exit */ - if (atexit(destroy_cmd_strs) == -1) - bb_perror_msg_and_die("atexit"); -#endif + if (ENABLE_FEATURE_CLEAN_UP) atexit(sed_free_and_close_stuff); - /* do normal option parsing */ - while ((opt = getopt(argc, argv, "ne:f:")) > 0) { - switch (opt) { - case 'n': - be_quiet++; - break; - case 'e':{ - char *str_cmd = strdup(optarg); + /* Lie to autoconf when it starts asking stupid questions. */ + if (argv[1] && !strcmp(argv[1], "--version")) { + puts("This is not GNU sed version 4.0"); + return 0; + } - add_cmd_str(str_cmd); - free(str_cmd); - break; - } - case 'f': - load_cmd_file(optarg); - break; - default: - bb_show_usage(); + /* do normal option parsing */ + opt_e = opt_f = NULL; + opt_complementary = "e::f::" /* can occur multiple times */ + "nn"; /* count -n */ + opt = getopt32(argv, "irne:f:", &opt_e, &opt_f, + &G.be_quiet); /* counter for -n */ + //argc -= optind; + argv += optind; + if (opt & OPT_in_place) { // -i + atexit(cleanup_outname); + } + if (opt & 0x2) G.regex_type |= REG_EXTENDED; // -r + //if (opt & 0x4) G.be_quiet++; // -n + while (opt_e) { // -e + add_cmd_block(llist_pop(&opt_e)); + } + while (opt_f) { // -f + char *line; + FILE *cmdfile; + cmdfile = xfopen_for_read(llist_pop(&opt_f)); + while ((line = xmalloc_fgetline(cmdfile)) != NULL) { + add_cmd(line); + free(line); } + fclose(cmdfile); } - - /* if we didn't get a pattern from a -e and no command file was specified, - * argv[optind] should be the pattern. no pattern, no worky */ - if (sed_cmd_head.next == NULL) { - if (argv[optind] == NULL) + /* if we didn't get a pattern from -e or -f, use argv[0] */ + if (!(opt & 0x18)) { + if (!*argv) bb_show_usage(); - else { - char *str_cmd = strdup(argv[optind]); - - add_cmd_str(strdup(str_cmd)); - free(str_cmd); - optind++; - } + add_cmd_block(*argv++); } + /* Flush any unfinished commands. */ + add_cmd(""); + + /* By default, we write to stdout */ + G.nonstdout = stdout; - /* argv[(optind)..(argc-1)] should be names of file to process. If no + /* argv[0..(argc-1)] should be names of file to process. If no * files were specified or '-' was specified, take input from stdin. * Otherwise, we process all the files specified. */ - if (argv[optind] == NULL || (strcmp(argv[optind], "-") == 0)) { - process_file(stdin); + if (argv[0] == NULL) { + if (opt & OPT_in_place) + bb_error_msg_and_die(bb_msg_requires_arg, "-i"); + add_input_file(stdin); + process_files(); } else { int i; FILE *file; - for (i = optind; i < argc; i++) { - file = bb_wfopen(argv[i], "r"); - if (file) { - process_file(file); - fclose(file); - } else + for (i = 0; argv[i]; i++) { + struct stat statbuf; + int nonstdoutfd; + + if (LONE_DASH(argv[i]) && !(opt & OPT_in_place)) { + add_input_file(stdin); + process_files(); + continue; + } + file = fopen_or_warn(argv[i], "r"); + if (!file) { status = EXIT_FAILURE; + continue; + } + if (!(opt & OPT_in_place)) { + add_input_file(file); + continue; + } + + G.outname = xasprintf("%sXXXXXX", argv[i]); + nonstdoutfd = mkstemp(G.outname); + if (-1 == nonstdoutfd) + bb_perror_msg_and_die("cannot create temp file %s", G.outname); + G.nonstdout = fdopen(nonstdoutfd, "w"); + + /* Set permissions of output file */ + + fstat(fileno(file), &statbuf); + fchmod(nonstdoutfd, statbuf.st_mode); + add_input_file(file); + process_files(); + fclose(G.nonstdout); + + G.nonstdout = stdout; + /* unlink(argv[i]); */ + xrename(G.outname, argv[i]); + free(G.outname); + G.outname = NULL; } + if (G.input_file_count > G.current_input_file) + process_files(); } return status; -- cgit v1.2.3-54-g00ecf