1 /* $OpenBSD: pat_rep.c,v 1.30 2005/08/05 08:30:10 djm Exp $ */
2 /* $NetBSD: pat_rep.c,v 1.4 1995/03/21 09:07:33 cgd Exp $ */
5 * Copyright (c) 1992 Keith Muller.
6 * Copyright (c) 1992, 1993
7 * The Regents of the University of California. All rights reserved.
9 * This code is derived from software contributed to Berkeley by
10 * Keith Muller of the University of California, San Diego.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 static const char sccsid[] = "@(#)pat_rep.c 8.2 (Berkeley) 4/18/94";
41 static const char rcsid[] = "$OpenBSD: pat_rep.c,v 1.30 2005/08/05 08:30:10 djm Exp $";
45 #include <sys/types.h>
48 #include <sys/param.h>
60 * routines to handle pattern matching, name modification (regular expression
61 * substitution and interactive renames), and destination name modification for
62 * copy (-rw). Both file name and link names are adjusted as required in these
66 #define MAXSUBEXP 10 /* max subexpressions, DO NOT CHANGE */
67 static PATTERN *pathead = NULL; /* file pattern match list head */
68 static PATTERN *pattail = NULL; /* file pattern match list tail */
69 static REPLACE *rephead = NULL; /* replacement string list head */
70 static REPLACE *reptail = NULL; /* replacement string list tail */
72 static int rep_name(char *, size_t, int *, int);
73 static int tty_rename(ARCHD *);
74 static int fix_path(char *, int *, char *, int);
75 static int fn_match(char *, char *, char **);
76 static char * range_match(char *, int);
77 static int resub(regex_t *, regmatch_t *, char *, char *, char *, char *);
81 * parses the -s replacement string; compiles the regular expression
82 * and stores the compiled value and it's replacement string together in
83 * replacement string list. Input to this function is of the form:
85 * The first char in the string specifies the delimiter used by this
86 * replacement string. "Old" is a regular expression in "ed" format which
87 * is compiled by regcomp() and is applied to filenames. "new" is the
88 * substitution string; p and g are options flags for printing and global
89 * replacement (over the single filename)
91 * 0 if a proper replacement string and regular expression was added to
92 * the list of replacement patterns; -1 otherwise.
105 * throw out the bad parameters
107 if ((str == NULL) || (*str == '\0')) {
108 paxwarn(1, "Empty replacement string");
113 * first character in the string specifies what the delimiter is for
116 for (pt1 = str+1; *pt1; pt1++) {
125 paxwarn(1, "Invalid replacement string %s", str);
130 * allocate space for the node that handles this replacement pattern
131 * and split out the regular expression and try to compile it
133 if ((rep = (REPLACE *)malloc(sizeof(REPLACE))) == NULL) {
134 paxwarn(1, "Unable to allocate memory for replacement string");
139 if ((res = regcomp(&(rep->rcmp), str+1, 0)) != 0) {
140 regerror(res, &(rep->rcmp), rebuf, sizeof(rebuf));
141 paxwarn(1, "%s while compiling regular expression %s", rebuf, str);
142 (void)free((char *)rep);
147 * put the delimiter back in case we need an error message and
148 * locate the delimiter at the end of the replacement string
149 * we then point the node at the new substitution string
152 for (pt2 = pt1; *pt2; pt2++) {
161 regfree(&(rep->rcmp));
162 (void)free((char *)rep);
163 paxwarn(1, "Invalid replacement string %s", str);
173 * set the options if any
175 while (*pt2 != '\0') {
186 regfree(&(rep->rcmp));
187 (void)free((char *)rep);
189 paxwarn(1, "Invalid replacement string option %s", str);
196 * all done, link it in at the end
199 if (rephead == NULL) {
200 reptail = rephead = rep;
210 * add a pattern match to the pattern match list. Pattern matches are used
211 * to select which archive members are extracted. (They appear as
212 * arguments to pax in the list and read modes). If no patterns are
213 * supplied to pax, all members in the archive will be selected (and the
214 * pattern match list is empty).
216 * 0 if the pattern was added to the list, -1 otherwise
220 pat_add(char *str, char *chdname)
227 if ((str == NULL) || (*str == '\0')) {
228 paxwarn(1, "Empty pattern string");
233 * allocate space for the pattern and store the pattern. the pattern is
234 * part of argv so do not bother to copy it, just point at it. Add the
235 * node to the end of the pattern list
237 if ((pt = (PATTERN *)malloc(sizeof(PATTERN))) == NULL) {
238 paxwarn(1, "Unable to allocate memory for pattern string");
244 pt->plen = strlen(str);
247 pt->chdname = chdname;
249 if (pathead == NULL) {
250 pattail = pathead = pt;
260 * complain if any the user supplied pattern did not result in a match to
261 * a selected archive member.
271 * walk down the list checking the flags to make sure MTCH was set,
274 for (pt = pathead; pt != NULL; pt = pt->fow) {
278 paxwarn(1, "WARNING! These patterns were not matched:");
281 (void)fprintf(stderr, "%s\n", pt->pstr);
287 * the archive member which matches a pattern was selected. Mark the
288 * pattern as having selected an archive member. arcn->pat points at the
289 * pattern that was matched. arcn->pat is set in pat_match()
291 * NOTE: When the -c option is used, we are called when there was no match
292 * by pat_match() (that means we did match before the inverted sense of
293 * the logic). Now this seems really strange at first, but with -c we
294 * need to keep track of those patterns that cause an archive member to NOT
295 * be selected (it found an archive member with a specified pattern)
297 * 0 if the pattern pointed at by arcn->pat was tagged as creating a
298 * match, -1 otherwise.
309 * if no patterns just return
311 if ((pathead == NULL) || ((pt = arcn->pat) == NULL))
315 * when we are NOT limited to a single match per pattern mark the
324 * we reach this point only when we allow a single selected match per
325 * pattern, if the pattern matches a directory and we do not have -d
326 * (dflag) we are done with this pattern. We may also be handed a file
327 * in the subtree of a directory. in that case when we are operating
328 * with -d, this pattern was already selected and we are done
330 if (pt->flgs & DIR_MTCH)
333 if (!dflag && ((pt->pend != NULL) || (arcn->type == PAX_DIR))) {
335 * ok we matched a directory and we are allowing
336 * subtree matches but because of the -n only its children will
337 * match. This is tagged as a DIR_MTCH type.
338 * WATCH IT, the code assumes that pt->pend points
339 * into arcn->name and arcn->name has not been modified.
340 * If not we will have a big mess. Yup this is another kludge
344 * if this was a prefix match, remove trailing part of path
345 * so we can copy it. Future matches will be exact prefix match
347 if (pt->pend != NULL)
350 if ((pt->pstr = strdup(arcn->name)) == NULL) {
351 paxwarn(1, "Pattern select out of memory");
352 if (pt->pend != NULL)
359 * put the trailing / back in the source string
361 if (pt->pend != NULL) {
365 pt->plen = strlen(pt->pstr);
368 * strip off any trailing /, this should really never happen
371 if (*(pt->pstr + len) == '/') {
372 *(pt->pstr + len) = '\0';
375 pt->flgs = DIR_MTCH | MTCH;
381 * we are then done with this pattern, so we delete it from the list
382 * because it can never be used for another match.
383 * Seems kind of strange to do for a -c, but the pax spec is really
384 * vague on the interaction of -c, -n and -d. We assume that when -c
385 * and the pattern rejects a member (i.e. it matched it) it is done.
386 * In effect we place the order of the flags as having -c last.
390 while ((pt != NULL) && (pt != arcn->pat)) {
397 * should never happen....
399 paxwarn(1, "Pattern list inconsistent");
403 (void)free((char *)pt);
410 * see if this archive member matches any supplied pattern, if a match
411 * is found, arcn->pat is set to point at the potential pattern. Later if
412 * this archive member is "selected" we process and mark the pattern as
413 * one which matched a selected archive member (see pat_sel())
415 * 0 if this archive member should be processed, 1 if it should be
416 * skipped and -1 if we are done with all patterns (and pax should quit
417 * looking for more members)
421 pat_match(ARCHD *arcn)
428 * if there are no more patterns and we have -n (and not -c) we are
429 * done. otherwise with no patterns to match, matches all
431 if (pathead == NULL) {
438 * have to search down the list one at a time looking for a match.
443 * check for a file name match unless we have DIR_MTCH set in
444 * this pattern then we want a prefix match
446 if (pt->flgs & DIR_MTCH) {
448 * this pattern was matched before to a directory
449 * as we must have -n set for this (but not -d). We can
450 * only match CHILDREN of that directory so we must use
451 * an exact prefix match (no wildcards).
453 if ((arcn->name[pt->plen] == '/') &&
454 (strncmp(pt->pstr, arcn->name, pt->plen) == 0))
456 } else if (fn_match(pt->pstr, arcn->name, &pt->pend) == 0)
462 * return the result, remember that cflag (-c) inverts the sense of a
466 return(cflag ? 0 : 1);
469 * we had a match, now when we invert the sense (-c) we reject this
470 * member. However we have to tag the pattern a being successful, (in a
471 * match, not in selecting a archive member) so we call pat_sel() here.
477 if (pat_sel(arcn) < 0)
486 * 0 if this archive member should be processed, 1 if it should be
487 * skipped and -1 if we are done with all patterns (and pax should quit
488 * looking for more members)
489 * Note: *pend may be changed to show where the prefix ends.
493 fn_match(char *pattern, char *string, char **pend)
500 switch (c = *pattern++) {
503 * Ok we found an exact match
509 * Check if it is a prefix match
511 if ((dflag == 1) || (*string != '/'))
515 * It is a prefix match, remember where the trailing
521 if ((test = *string++) == '\0')
527 * Collapse multiple *'s.
533 * Optimized hack for pattern with a * at the end
539 * General case, use recursion.
541 while ((test = *string) != '\0') {
542 if (!fn_match(pattern, string, pend))
551 if (((test = *string++) == '\0') ||
552 ((pattern = range_match(pattern, test)) == NULL))
566 range_match(char *pattern, int test)
573 if ((negate = (*pattern == '!')) != 0)
576 while ((c = *pattern++) != ']') {
583 if ((*pattern == '-') && ((c2 = pattern[1]) != '\0') &&
585 if ((c <= test) && (test <= c2))
588 } else if (c == test)
591 return (ok == negate ? NULL : pattern);
596 * modify a selected file name. first attempt to apply replacement string
597 * expressions, then apply interactive file rename. We apply replacement
598 * string expressions to both filenames and file links (if we didn't the
599 * links would point to the wrong place, and we could never be able to
600 * move an archive that has a file link in it). When we rename files
601 * interactively, we store that mapping (old name to user input name) so
602 * if we spot any file links to the old file name in the future, we will
603 * know exactly how to fix the file link.
605 * 0 continue to process file, 1 skip this file, -1 pax is finished
609 mod_name(ARCHD *arcn)
614 * Strip off leading '/' if appropriate.
615 * Currently, this option is only set for the tar format.
617 while (rmleadslash && arcn->name[0] == '/') {
618 if (arcn->name[1] == '\0') {
621 (void)memmove(arcn->name, &arcn->name[1],
625 if (rmleadslash < 2) {
627 paxwarn(0, "Removing leading / from absolute path names in the archive");
630 while (rmleadslash && arcn->ln_name[0] == '/' &&
631 (arcn->type == PAX_HLK || arcn->type == PAX_HRG)) {
632 if (arcn->ln_name[1] == '\0') {
633 arcn->ln_name[0] = '.';
635 (void)memmove(arcn->ln_name, &arcn->ln_name[1],
636 strlen(arcn->ln_name));
639 if (rmleadslash < 2) {
641 paxwarn(0, "Removing leading / from absolute path names in the archive");
646 * IMPORTANT: We have a problem. what do we do with symlinks?
647 * Modifying a hard link name makes sense, as we know the file it
648 * points at should have been seen already in the archive (and if it
649 * wasn't seen because of a read error or a bad archive, we lose
650 * anyway). But there are no such requirements for symlinks. On one
651 * hand the symlink that refers to a file in the archive will have to
652 * be modified to so it will still work at its new location in the
653 * file system. On the other hand a symlink that points elsewhere (and
654 * should continue to do so) should not be modified. There is clearly
655 * no perfect solution here. So we handle them like hardlinks. Clearly
656 * a replacement made by the interactive rename mapping is very likely
657 * to be correct since it applies to a single file and is an exact
658 * match. The regular expression replacements are a little harder to
659 * justify though. We claim that the symlink name is only likely
660 * to be replaced when it points within the file tree being moved and
661 * in that case it should be modified. what we really need to do is to
662 * call an oracle here. :)
664 if (rephead != NULL) {
666 * we have replacement strings, modify the name and the link
669 if ((res = rep_name(arcn->name, sizeof(arcn->name), &(arcn->nlen), 1)) != 0)
672 if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
673 (arcn->type == PAX_HRG)) &&
674 ((res = rep_name(arcn->ln_name, sizeof(arcn->ln_name), &(arcn->ln_nlen), 0)) != 0))
680 * perform interactive file rename, then map the link if any
682 if ((res = tty_rename(arcn)) != 0)
684 if ((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
685 (arcn->type == PAX_HRG))
686 sub_name(arcn->ln_name, &(arcn->ln_nlen), sizeof(arcn->ln_name));
693 * Prompt the user for a replacement file name. A "." keeps the old name,
694 * a empty line skips the file, and an EOF on reading the tty, will cause
695 * pax to stop processing and exit. Otherwise the file name input, replaces
698 * 0 process this file, 1 skip this file, -1 we need to exit pax
702 tty_rename(ARCHD *arcn)
704 char tmpname[PAXPATHLEN+2];
708 * prompt user for the replacement name for a file, keep trying until
709 * we get some reasonable input. Archives may have more than one file
710 * on them with the same name (from updates etc). We print verbose info
711 * on the file so the user knows what is up.
713 tty_prnt("\nATTENTION: %s interactive file rename operation.\n", argv0);
717 tty_prnt("Input new name, or a \".\" to keep the old name, ");
718 tty_prnt("or a \"return\" to skip this file.\n");
719 tty_prnt("Input > ");
720 if (tty_read(tmpname, sizeof(tmpname)) < 0)
722 if (strcmp(tmpname, "..") == 0) {
723 tty_prnt("Try again, illegal file name: ..\n");
726 if (strlen(tmpname) > PAXPATHLEN) {
727 tty_prnt("Try again, file name too long\n");
734 * empty file name, skips this file. a "." leaves it alone
736 if (tmpname[0] == '\0') {
737 tty_prnt("Skipping file.\n");
740 if ((tmpname[0] == '.') && (tmpname[1] == '\0')) {
741 tty_prnt("Processing continues, name unchanged.\n");
746 * ok the name changed. We may run into links that point at this
747 * file later. we have to remember where the user sent the file
748 * in order to repair any links.
750 tty_prnt("Processing continues, name changed to: %s\n", tmpname);
751 res = add_name(arcn->name, arcn->nlen, tmpname);
752 arcn->nlen = strlcpy(arcn->name, tmpname, sizeof(arcn->name));
753 if (arcn->nlen >= sizeof(arcn->name))
754 arcn->nlen = sizeof(arcn->name) - 1; /* XXX truncate? */
762 * fix up the file name and the link name (if any) so this file will land
763 * in the destination directory (used during copy() -rw).
765 * 0 if ok, -1 if failure (name too long)
769 set_dest(ARCHD *arcn, char *dest_dir, int dir_len)
771 if (fix_path(arcn->name, &(arcn->nlen), dest_dir, dir_len) < 0)
775 * It is really hard to deal with symlinks here, we cannot be sure
776 * if the name they point was moved (or will be moved). It is best to
779 if ((arcn->type != PAX_HLK) && (arcn->type != PAX_HRG))
782 if (fix_path(arcn->ln_name, &(arcn->ln_nlen), dest_dir, dir_len) < 0)
789 * concatenate dir_name and or_name and store the result in or_name (if
790 * it fits). This is one ugly function.
792 * 0 if ok, -1 if the final name is too long
796 fix_path(char *or_name, int *or_len, char *dir_name, int dir_len)
804 * we shift the or_name to the right enough to tack in the dir_name
805 * at the front. We make sure we have enough space for it all before
806 * we start. since dest always ends in a slash, we skip of or_name
807 * if it also starts with one.
810 src = start + *or_len;
811 dest = src + dir_len;
816 if ((len = dest - or_name) > PAXPATHLEN) {
817 paxwarn(1, "File name %s/%s, too long", dir_name, start);
823 * enough space, shift
827 src = dir_name + dir_len - 1;
830 * splice in the destination directory name
832 while (src >= dir_name)
835 *(or_name + len) = '\0';
841 * walk down the list of replacement strings applying each one in order.
842 * when we find one with a successful substitution, we modify the name
843 * as specified. if required, we print the results. if the resulting name
844 * is empty, we will skip this archive member. We use the regexp(3)
845 * routines (regexp() ought to win a prize as having the most cryptic
846 * library function manual page).
848 * name is the file name we are going to apply the regular expressions to
849 * (and may be modified)
850 * nsize is the size of the name buffer.
851 * nlen is the length of this name (and is modified to hold the length of
853 * prnt is a flag that says whether to print the final result.
855 * 0 if substitution was successful, 1 if we are to skip the file (the name
860 rep_name(char *name, size_t nsize, int *nlen, int prnt)
869 regmatch_t pm[MAXSUBEXP];
870 char nname[PAXPATHLEN+1]; /* final result of all replacements */
871 char buf1[PAXPATHLEN+1]; /* where we work on the name */
874 * copy the name into buf1, where we will work on it. We need to keep
875 * the orig string around so we can print out the result of the final
876 * replacement. We build up the final result in nname. inpt points at
877 * the string we apply the regular expression to. prnt is used to
878 * suppress printing when we handle replacements on the link field
879 * (the user already saw that substitution go by)
882 (void)strlcpy(buf1, name, sizeof(buf1));
885 endpt = outpt + PAXPATHLEN;
888 * try each replacement string in order
894 * check for a successful substitution, if not go to
895 * the next pattern, or cleanup if we were global
897 if (regexec(&(pt->rcmp), inpt, MAXSUBEXP, pm, 0) != 0)
901 * ok we found one. We have three parts, the prefix
902 * which did not match, the section that did and the
903 * tail (that also did not match). Copy the prefix to
904 * the final output buffer (watching to make sure we
905 * do not create a string too long).
908 rpt = inpt + pm[0].rm_so;
910 while ((inpt < rpt) && (outpt < endpt))
916 * for the second part (which matched the regular
917 * expression) apply the substitution using the
918 * replacement string and place it the prefix in the
919 * final output. If we have problems, skip it.
921 if ((res = resub(&(pt->rcmp),pm,pt->nstr,oinpt,outpt,endpt))
924 paxwarn(1, "Replacement name error %s",
931 * we set up to look again starting at the first
932 * character in the tail (of the input string right
933 * after the last character matched by the regular
934 * expression (inpt always points at the first char in
935 * the string to process). If we are not doing a global
936 * substitution, we will use inpt to copy the tail to
937 * the final result. Make sure we do not overrun the
940 inpt += pm[0].rm_eo - pm[0].rm_so;
942 if ((outpt == endpt) || (*inpt == '\0'))
946 * if the user wants global we keep trying to
947 * substitute until it fails, then we are done.
949 } while (pt->flgs & GLOB);
955 * a successful substitution did NOT occur, try the next one
962 * we had a substitution, copy the last tail piece (if there is
963 * room) to the final result
965 while ((outpt < endpt) && (*inpt != '\0'))
969 if ((outpt == endpt) && (*inpt != '\0')) {
971 paxwarn(1,"Replacement name too long %s >> %s",
977 * inform the user of the result if wanted
979 if (prnt && (pt->flgs & PRNT)) {
981 (void)fprintf(stderr,"%s >> <empty string>\n",
984 (void)fprintf(stderr,"%s >> %s\n", name, nname);
988 * if empty inform the caller this file is to be skipped
989 * otherwise copy the new name over the orig name and return
993 *nlen = strlcpy(name, nname, nsize);
1000 * apply the replacement to the matched expression. expand out the old
1001 * style ed(1) subexpression expansion.
1003 * -1 if error, or the number of characters added to the destination.
1007 resub(regex_t *rp, regmatch_t *pm, char *src, char *inpt, char *dest,
1019 subexcnt = rp->re_nsub;
1020 while ((dpt < destend) && ((c = *spt++) != '\0')) {
1022 * see if we just have an ordinary replacement character
1023 * or we refer to a subexpression.
1027 } else if ((c == '\\') && (*spt >= '0') && (*spt <= '9')) {
1029 * make sure there is a subexpression as specified
1031 if ((len = *spt++ - '0') > subexcnt)
1036 * Ordinary character, just copy it
1038 if ((c == '\\') && (*spt != '\0'))
1045 * continue if the subexpression is bogus
1047 if ((pmpt->rm_so < 0) || (pmpt->rm_eo < 0) ||
1048 ((len = pmpt->rm_eo - pmpt->rm_so) <= 0))
1052 * copy the subexpression to the destination.
1053 * fail if we run out of space or the match string is damaged
1055 if (len > (destend - dpt))
1057 strncpy(dpt, inpt + pmpt->rm_so, len);