2 * Copyright (c) 1992 Keith Muller.
3 * Copyright (c) 1992, 1993
4 * The Regents of the University of California. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
7 * Keith Muller of the University of California, San Diego.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 static char sccsid[] = "@(#)pat_rep.c 8.2 (Berkeley) 4/18/94";
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
42 #include <sys/types.h>
57 * routines to handle pattern matching, name modification (regular expression
58 * substitution and interactive renames), and destination name modification for
59 * copy (-rw). Both file name and link names are adjusted as required in these
63 #define MAXSUBEXP 10 /* max subexpressions, DO NOT CHANGE */
64 static PATTERN *pathead = NULL; /* file pattern match list head */
65 static PATTERN *pattail = NULL; /* file pattern match list tail */
66 static REPLACE *rephead = NULL; /* replacement string list head */
67 static REPLACE *reptail = NULL; /* replacement string list tail */
69 static int rep_name(char *, int *, int);
70 static int tty_rename(ARCHD *);
71 static int fix_path(char *, int *, char *, int);
72 static int fn_match(char *, char *, char **);
73 static char * range_match(char *, int);
75 static int resub(regexp *, char *, char *, char *);
77 static int resub(regex_t *, regmatch_t *, char *, char *, char *, char *);
82 * parses the -s replacement string; compiles the regular expression
83 * and stores the compiled value and it's replacement string together in
84 * replacement string list. Input to this function is of the form:
86 * The first char in the string specifies the delimiter used by this
87 * replacement string. "Old" is a regular expression in "ed" format which
88 * is compiled by regcomp() and is applied to filenames. "new" is the
89 * substitution string; p and g are options flags for printing and global
90 * replacement (over the single filename)
92 * 0 if a proper replacement string and regular expression was added to
93 * the list of replacement patterns; -1 otherwise.
108 * throw out the bad parameters
110 if ((str == NULL) || (*str == '\0')) {
111 paxwarn(1, "Empty replacement string");
116 * first character in the string specifies what the delimiter is for
119 if ((pt1 = strchr(str+1, *str)) == NULL) {
120 paxwarn(1, "Invalid replacement string %s", str);
125 * allocate space for the node that handles this replacement pattern
126 * and split out the regular expression and try to compile it
128 if ((rep = (REPLACE *)malloc(sizeof(REPLACE))) == NULL) {
129 paxwarn(1, "Unable to allocate memory for replacement string");
135 if ((rep->rcmp = regcomp(str+1)) == NULL) {
137 if ((res = regcomp(&(rep->rcmp), str+1, 0)) != 0) {
138 regerror(res, &(rep->rcmp), rebuf, sizeof(rebuf));
139 paxwarn(1, "%s while compiling regular expression %s", rebuf, str);
146 * put the delimiter back in case we need an error message and
147 * locate the delimiter at the end of the replacement string
148 * we then point the node at the new substitution string
151 if ((pt2 = strchr(pt1, *str)) == NULL) {
158 paxwarn(1, "Invalid replacement string %s", str);
168 * set the options if any
170 while (*pt2 != '\0') {
188 paxwarn(1, "Invalid replacement string option %s", str);
195 * all done, link it in at the end
198 if (rephead == NULL) {
199 reptail = rephead = rep;
209 * add a pattern match to the pattern match list. Pattern matches are used
210 * to select which archive members are extracted. (They appear as
211 * arguments to pax in the list and read modes). If no patterns are
212 * supplied to pax, all members in the archive will be selected (and the
213 * pattern match list is empty).
215 * 0 if the pattern was added to the list, -1 otherwise
219 pat_add(char *str, char *chdnam)
226 if ((str == NULL) || (*str == '\0')) {
227 paxwarn(1, "Empty pattern string");
232 * allocate space for the pattern and store the pattern. the pattern is
233 * part of argv so do not bother to copy it, just point at it. Add the
234 * node to the end of the pattern list
236 if ((pt = (PATTERN *)malloc(sizeof(PATTERN))) == NULL) {
237 paxwarn(1, "Unable to allocate memory for pattern string");
243 pt->plen = strlen(str);
246 pt->chdname = chdnam;
248 if (pathead == NULL) {
249 pattail = pathead = pt;
259 * complain if any the user supplied pattern did not result in a match to
260 * a selected archive member.
270 * walk down the list checking the flags to make sure MTCH was set,
273 for (pt = pathead; pt != NULL; pt = pt->fow) {
277 paxwarn(1, "WARNING! These patterns were not matched:");
280 (void)fprintf(stderr, "%s\n", pt->pstr);
286 * the archive member which matches a pattern was selected. Mark the
287 * pattern as having selected an archive member. arcn->pat points at the
288 * pattern that was matched. arcn->pat is set in pat_match()
290 * NOTE: When the -c option is used, we are called when there was no match
291 * by pat_match() (that means we did match before the inverted sense of
292 * the logic). Now this seems really strange at first, but with -c we
293 * need to keep track of those patterns that cause an archive member to NOT
294 * be selected (it found an archive member with a specified pattern)
296 * 0 if the pattern pointed at by arcn->pat was tagged as creating a
297 * match, -1 otherwise.
308 * if no patterns just return
310 if ((pathead == NULL) || ((pt = arcn->pat) == NULL))
314 * when we are NOT limited to a single match per pattern mark the
323 * we reach this point only when we allow a single selected match per
324 * pattern, if the pattern matches a directory and we do not have -d
325 * (dflag) we are done with this pattern. We may also be handed a file
326 * in the subtree of a directory. in that case when we are operating
327 * with -d, this pattern was already selected and we are done
329 if (pt->flgs & DIR_MTCH)
332 if (!dflag && ((pt->pend != NULL) || (arcn->type == PAX_DIR))) {
334 * ok we matched a directory and we are allowing
335 * subtree matches but because of the -n only its children will
336 * match. This is tagged as a DIR_MTCH type.
337 * WATCH IT, the code assumes that pt->pend points
338 * into arcn->name and arcn->name has not been modified.
339 * If not we will have a big mess. Yup this is another kludge
343 * if this was a prefix match, remove trailing part of path
344 * so we can copy it. Future matches will be exact prefix match
346 if (pt->pend != NULL)
349 if ((pt->pstr = strdup(arcn->name)) == NULL) {
350 paxwarn(1, "Pattern select out of memory");
351 if (pt->pend != NULL)
358 * put the trailing / back in the source string
360 if (pt->pend != NULL) {
364 pt->plen = strlen(pt->pstr);
367 * strip off any trailing /, this should really never happen
370 if (*(pt->pstr + len) == '/') {
371 *(pt->pstr + len) = '\0';
374 pt->flgs = DIR_MTCH | MTCH;
380 * we are then done with this pattern, so we delete it from the list
381 * because it can never be used for another match.
382 * Seems kind of strange to do for a -c, but the pax spec is really
383 * vague on the interaction of -c -n and -d. We assume that when -c
384 * and the pattern rejects a member (i.e. it matched it) it is done.
385 * In effect we place the order of the flags as having -c last.
389 while ((pt != NULL) && (pt != arcn->pat)) {
396 * should never happen....
398 paxwarn(1, "Pattern list inconsistent");
409 * see if this archive member matches any supplied pattern, if a match
410 * is found, arcn->pat is set to point at the potential pattern. Later if
411 * this archive member is "selected" we process and mark the pattern as
412 * one which matched a selected archive member (see pat_sel())
414 * 0 if this archive member should be processed, 1 if it should be
415 * skipped and -1 if we are done with all patterns (and pax should quit
416 * looking for more members)
420 pat_match(ARCHD *arcn)
427 * if there are no more patterns and we have -n (and not -c) we are
428 * done. otherwise with no patterns to match, matches all
430 if (pathead == NULL) {
437 * have to search down the list one at a time looking for a match.
442 * check for a file name match unless we have DIR_MTCH set in
443 * this pattern then we want a prefix match
445 if (pt->flgs & DIR_MTCH) {
447 * this pattern was matched before to a directory
448 * as we must have -n set for this (but not -d). We can
449 * only match CHILDREN of that directory so we must use
450 * an exact prefix match (no wildcards).
452 if ((arcn->name[pt->plen] == '/') &&
453 (strncmp(pt->pstr, arcn->name, pt->plen) == 0))
455 } else if (fn_match(pt->pstr, arcn->name, &pt->pend) == 0)
461 * return the result, remember that cflag (-c) inverts the sense of a
465 return(cflag ? 0 : 1);
468 * We had a match, now when we invert the sense (-c) we reject this
469 * member. However we have to tag the pattern a being successful, (in a
470 * match, not in selecting an archive member) so we call pat_sel() here.
476 if (pat_sel(arcn) < 0)
485 * 0 if this archive member should be processed, 1 if it should be
486 * skipped and -1 if we are done with all patterns (and pax should quit
487 * looking for more members)
488 * Note: *pend may be changed to show where the prefix ends.
492 fn_match(char *pattern, char *string, char **pend)
499 switch (c = *pattern++) {
502 * Ok we found an exact match
508 * Check if it is a prefix match
510 if ((dflag == 1) || (*string != '/'))
514 * It is a prefix match, remember where the trailing
520 if ((test = *string++) == '\0')
526 * Collapse multiple *'s.
532 * Optimized hack for pattern with a * at the end
538 * General case, use recursion.
540 while ((test = *string) != '\0') {
541 if (!fn_match(pattern, string, pend))
550 if (((test = *string++) == '\0') ||
551 ((pattern = range_match(pattern, test)) == NULL))
565 range_match(char *pattern, int test)
572 if ((negate = (*pattern == '!')) != 0)
575 while ((c = *pattern++) != ']') {
582 if ((*pattern == '-') && ((c2 = pattern[1]) != '\0') &&
584 if ((c <= test) && (test <= c2))
587 } else if (c == test)
590 return (ok == negate ? NULL : pattern);
595 * modify a selected file name. first attempt to apply replacement string
596 * expressions, then apply interactive file rename. We apply replacement
597 * string expressions to both filenames and file links (if we didn't the
598 * links would point to the wrong place, and we could never be able to
599 * move an archive that has a file link in it). When we rename files
600 * interactively, we store that mapping (old name to user input name) so
601 * if we spot any file links to the old file name in the future, we will
602 * know exactly how to fix the file link.
604 * 0 continue to process file, 1 skip this file, -1 pax is finished
608 mod_name(ARCHD *arcn)
613 * Strip off leading '/' if appropriate.
614 * Currently, this option is only set for the tar format.
616 if (rmleadslash && arcn->name[0] == '/') {
617 if (arcn->name[1] == '\0') {
620 (void)memmove(arcn->name, &arcn->name[1],
624 if (rmleadslash < 2) {
626 paxwarn(0, "Removing leading / from absolute path names in the archive");
629 if (rmleadslash && arcn->ln_name[0] == '/' &&
630 (arcn->type == PAX_HLK || arcn->type == PAX_HRG)) {
631 if (arcn->ln_name[1] == '\0') {
632 arcn->ln_name[0] = '.';
634 (void)memmove(arcn->ln_name, &arcn->ln_name[1],
635 strlen(arcn->ln_name));
638 if (rmleadslash < 2) {
640 paxwarn(0, "Removing leading / from absolute path names in the archive");
645 * IMPORTANT: We have a problem. what do we do with symlinks?
646 * Modifying a hard link name makes sense, as we know the file it
647 * points at should have been seen already in the archive (and if it
648 * wasn't seen because of a read error or a bad archive, we lose
649 * anyway). But there are no such requirements for symlinks. On one
650 * hand the symlink that refers to a file in the archive will have to
651 * be modified to so it will still work at its new location in the
652 * file system. On the other hand a symlink that points elsewhere (and
653 * should continue to do so) should not be modified. There is clearly
654 * no perfect solution here. So we handle them like hardlinks. Clearly
655 * a replacement made by the interactive rename mapping is very likely
656 * to be correct since it applies to a single file and is an exact
657 * match. The regular expression replacements are a little harder to
658 * justify though. We claim that the symlink name is only likely
659 * to be replaced when it points within the file tree being moved and
660 * in that case it should be modified. what we really need to do is to
661 * call an oracle here. :)
663 if (rephead != NULL) {
665 * we have replacement strings, modify the name and the link
668 if ((res = rep_name(arcn->name, &(arcn->nlen), 1)) != 0)
671 if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
672 (arcn->type == PAX_HRG)) &&
673 ((res = rep_name(arcn->ln_name, &(arcn->ln_nlen), 0)) != 0))
679 * perform interactive file rename, then map the link if any
681 if ((res = tty_rename(arcn)) != 0)
683 if ((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
684 (arcn->type == PAX_HRG))
685 sub_name(arcn->ln_name, &(arcn->ln_nlen), sizeof(arcn->ln_name));
692 * Prompt the user for a replacement file name. A "." keeps the old name,
693 * a empty line skips the file, and an EOF on reading the tty, will cause
694 * pax to stop processing and exit. Otherwise the file name input, replaces
697 * 0 process this file, 1 skip this file, -1 we need to exit pax
701 tty_rename(ARCHD *arcn)
703 char tmpname[PAXPATHLEN+2];
707 * prompt user for the replacement name for a file, keep trying until
708 * we get some reasonable input. Archives may have more than one file
709 * on them with the same name (from updates etc). We print verbose info
710 * on the file so the user knows what is up.
712 tty_prnt("\nATTENTION: %s interactive file rename operation.\n", argv0);
716 tty_prnt("Input new name, or a \".\" to keep the old name, ");
717 tty_prnt("or a \"return\" to skip this file.\n");
718 tty_prnt("Input > ");
719 if (tty_read(tmpname, sizeof(tmpname)) < 0)
721 if (strcmp(tmpname, "..") == 0) {
722 tty_prnt("Try again, illegal file name: ..\n");
725 if (strlen(tmpname) > PAXPATHLEN) {
726 tty_prnt("Try again, file name too long\n");
733 * empty file name, skips this file. a "." leaves it alone
735 if (tmpname[0] == '\0') {
736 tty_prnt("Skipping file.\n");
739 if ((tmpname[0] == '.') && (tmpname[1] == '\0')) {
740 tty_prnt("Processing continues, name unchanged.\n");
745 * ok the name changed. We may run into links that point at this
746 * file later. we have to remember where the user sent the file
747 * in order to repair any links.
749 tty_prnt("Processing continues, name changed to: %s\n", tmpname);
750 res = add_name(arcn->name, arcn->nlen, tmpname);
751 arcn->nlen = l_strncpy(arcn->name, tmpname, sizeof(arcn->name) - 1);
752 arcn->name[arcn->nlen] = '\0';
760 * fix up the file name and the link name (if any) so this file will land
761 * in the destination directory (used during copy() -rw).
763 * 0 if ok, -1 if failure (name too long)
767 set_dest(ARCHD *arcn, char *dest_dir, int dir_len)
769 if (fix_path(arcn->name, &(arcn->nlen), dest_dir, dir_len) < 0)
773 * It is really hard to deal with symlinks here, we cannot be sure
774 * if the name they point was moved (or will be moved). It is best to
777 if ((arcn->type != PAX_HLK) && (arcn->type != PAX_HRG))
780 if (fix_path(arcn->ln_name, &(arcn->ln_nlen), dest_dir, dir_len) < 0)
787 * concatenate dir_name and or_name and store the result in or_name (if
788 * it fits). This is one ugly function.
790 * 0 if ok, -1 if the final name is too long
794 fix_path( char *or_name, int *or_len, char *dir_name, int dir_len)
802 * we shift the or_name to the right enough to tack in the dir_name
803 * at the front. We make sure we have enough space for it all before
804 * we start. since dest always ends in a slash, we skip of or_name
805 * if it also starts with one.
808 src = start + *or_len;
809 dest = src + dir_len;
814 if ((len = dest - or_name) > PAXPATHLEN) {
815 paxwarn(1, "File name %s/%s, too long", dir_name, start);
821 * enough space, shift
825 src = dir_name + dir_len - 1;
828 * splice in the destination directory name
830 while (src >= dir_name)
833 *(or_name + len) = '\0';
839 * walk down the list of replacement strings applying each one in order.
840 * when we find one with a successful substitution, we modify the name
841 * as specified. if required, we print the results. if the resulting name
842 * is empty, we will skip this archive member. We use the regexp(3)
843 * routines (regexp() ought to win a prize as having the most cryptic
844 * library function manual page).
846 * name is the file name we are going to apply the regular expressions to
847 * (and may be modified)
848 * nlen is the length of this name (and is modified to hold the length of
850 * prnt is a flag that says whether to print the final result.
852 * 0 if substitution was successful, 1 if we are to skip the file (the name
857 rep_name(char *name, int *nlen, int prnt)
867 regmatch_t pm[MAXSUBEXP];
869 char nname[PAXPATHLEN+1]; /* final result of all replacements */
870 char buf1[PAXPATHLEN+1]; /* where we work on the name */
873 * copy the name into buf1, where we will work on it. We need to keep
874 * the orig string around so we can print out the result of the final
875 * replacement. We build up the final result in nname. inpt points at
876 * the string we apply the regular expression to. prnt is used to
877 * suppress printing when we handle replacements on the link field
878 * (the user already saw that substitution go by)
881 (void)strcpy(buf1, name);
884 endpt = outpt + PAXPATHLEN;
887 * try each replacement string in order
892 * check for a successful substitution, if not go to
893 * the next pattern, or cleanup if we were global
896 if (regexec(pt->rcmp, inpt) == 0)
898 if (regexec(&(pt->rcmp), inpt, MAXSUBEXP, pm, 0) != 0)
903 * ok we found one. We have three parts, the prefix
904 * which did not match, the section that did and the
905 * tail (that also did not match). Copy the prefix to
906 * the final output buffer (watching to make sure we
907 * do not create a string too long).
911 rpt = pt->rcmp->startp[0];
913 rpt = inpt + pm[0].rm_so;
916 while ((inpt < rpt) && (outpt < endpt))
922 * for the second part (which matched the regular
923 * expression) apply the substitution using the
924 * replacement string and place it the prefix in the
925 * final output. If we have problems, skip it.
928 if ((res = resub(pt->rcmp,pt->nstr,outpt,endpt)) < 0) {
930 if ((res = resub(&(pt->rcmp),pm,inpt,pt->nstr,outpt,endpt))
934 paxwarn(1, "Replacement name error %s",
941 * we set up to look again starting at the first
942 * character in the tail (of the input string right
943 * after the last character matched by the regular
944 * expression (inpt always points at the first char in
945 * the string to process). If we are not doing a global
946 * substitution, we will use inpt to copy the tail to
947 * the final result. Make sure we do not overrun the
951 inpt = pt->rcmp->endp[0];
953 inpt += pm[0].rm_eo - pm[0].rm_so;
956 if ((outpt == endpt) || (*inpt == '\0'))
960 * if the user wants global we keep trying to
961 * substitute until it fails, then we are done.
963 } while (pt->flgs & GLOB);
969 * a successful substitution did NOT occur, try the next one
976 * we had a substitution, copy the last tail piece (if there is
977 * room) to the final result
979 while ((outpt < endpt) && (*inpt != '\0'))
983 if ((outpt == endpt) && (*inpt != '\0')) {
985 paxwarn(1,"Replacement name too long %s >> %s",
991 * inform the user of the result if wanted
993 if (prnt && (pt->flgs & PRNT)) {
995 (void)fprintf(stderr,"%s >> <empty string>\n",
998 (void)fprintf(stderr,"%s >> %s\n", name, nname);
1002 * if empty inform the caller this file is to be skipped
1003 * otherwise copy the new name over the orig name and return
1007 *nlen = l_strncpy(name, nname, PAXPATHLEN + 1);
1008 name[PAXPATHLEN] = '\0';
1016 * apply the replacement to the matched expression. expand out the old
1017 * style ed(1) subexpression expansion.
1019 * -1 if error, or the number of characters added to the destination.
1023 resub(regexp *prog, char *src, char *dest, char *destend)
1033 while ((dpt < destend) && ((c = *spt++) != '\0')) {
1036 else if ((c == '\\') && (*spt >= '0') && (*spt <= '9'))
1039 if ((c == '\\') && ((*spt == '\\') || (*spt == '&')))
1044 if ((prog->startp[no] == NULL) || (prog->endp[no] == NULL) ||
1045 ((len = prog->endp[no] - prog->startp[no]) <= 0))
1049 * copy the subexpression to the destination.
1050 * fail if we run out of space or the match string is damaged
1052 if (len > (destend - dpt))
1053 len = destend - dpt;
1054 if (l_strncpy(dpt, prog->startp[no], len) != len)
1065 * apply the replacement to the matched expression. expand out the old
1066 * style ed(1) subexpression expansion.
1068 * -1 if error, or the number of characters added to the destination.
1072 resub(regex_t *rp, regmatch_t *pm, char *orig, char *src, char *dest,
1084 subexcnt = rp->re_nsub;
1085 while ((dpt < destend) && ((c = *spt++) != '\0')) {
1087 * see if we just have an ordinary replacement character
1088 * or we refer to a subexpression.
1092 } else if ((c == '\\') && (*spt >= '0') && (*spt <= '9')) {
1094 * make sure there is a subexpression as specified
1096 if ((len = *spt++ - '0') > subexcnt)
1101 * Ordinary character, just copy it
1103 if ((c == '\\') && ((*spt == '\\') || (*spt == '&')))
1110 * continue if the subexpression is bogus
1112 if ((pmpt->rm_so < 0) || (pmpt->rm_eo < 0) ||
1113 ((len = pmpt->rm_eo - pmpt->rm_so) <= 0))
1117 * copy the subexpression to the destination.
1118 * fail if we run out of space or the match string is damaged
1120 if (len > (destend - dpt))
1121 len = destend - dpt;
1122 if (l_strncpy(dpt, orig + pmpt->rm_so, len) != len)