xref: /csrg-svn/bin/pax/pat_rep.c (revision 60676)
157266Smuller /*-
257266Smuller  * Copyright (c) 1992 Keith Muller.
3*60676Sbostic  * Copyright (c) 1992, 1993
4*60676Sbostic  *	The Regents of the University of California.  All rights reserved.
557266Smuller  *
657266Smuller  * This code is derived from software contributed to Berkeley by
757266Smuller  * Keith Muller of the University of California, San Diego.
857266Smuller  *
957266Smuller  * %sccs.include.redist.c%
1057266Smuller  */
1157266Smuller 
1257266Smuller #ifndef lint
13*60676Sbostic static char sccsid[] = "@(#)pat_rep.c	8.1 (Berkeley) 05/31/93";
1457266Smuller #endif /* not lint */
1557266Smuller 
1657266Smuller #include <sys/types.h>
1757266Smuller #include <sys/time.h>
1857266Smuller #include <sys/stat.h>
1957266Smuller #include <sys/param.h>
2057266Smuller #include <stdio.h>
2157266Smuller #include <ctype.h>
2257266Smuller #include <string.h>
2357266Smuller #include <unistd.h>
2457266Smuller #include <stdlib.h>
2557266Smuller #include <fnmatch.h>
2657266Smuller #ifdef NET2_REGEX
2757266Smuller #include <regexp.h>
2857266Smuller #else
2957266Smuller #include <regex.h>
3057266Smuller #endif
3157266Smuller #include "pax.h"
3257266Smuller #include "pat_rep.h"
3357266Smuller #include "extern.h"
3457266Smuller 
3557266Smuller /*
3657266Smuller  * routines to handle pattern matching, name modification (regular expression
3757266Smuller  * substitution and interactive renames), and destination name modification for
3857266Smuller  * copy (-rw). Both file name and link names are adjusted as required in these
3957266Smuller  * routines.
4057266Smuller  */
4157266Smuller 
4257266Smuller #define MAXSUBEXP	10		/* max subexpressions, DO NOT CHANGE */
4357266Smuller static PATTERN *pathead = NULL;		/* file pattern match list head */
4457266Smuller static PATTERN *pattail = NULL;		/* file pattern match list tail */
4557266Smuller static REPLACE *rephead = NULL;		/* replacement string list head */
4657266Smuller static REPLACE *reptail = NULL;		/* replacement string list tail */
4757266Smuller 
4857266Smuller static int rep_name __P((char *, int *, int));
4957266Smuller static int tty_rename __P((register ARCHD *));
5057266Smuller static int fix_path __P((char *, int *, char *, int));
5157266Smuller #ifdef NET2_REGEX
5257266Smuller static int resub __P((regexp *, char *, char *, register char *));
5357266Smuller #else
5457266Smuller static int resub __P((regex_t *, regmatch_t *, char *, char *, char *));
5557266Smuller #endif
5657266Smuller 
5757266Smuller /*
5857266Smuller  * rep_add()
5957266Smuller  *	parses the -s replacement string; compiles the regular expression
6057266Smuller  *	and stores the compiled value and it's replacement string together in
6157266Smuller  *	replacement string list. Input to this function is of the form:
6257266Smuller  *		/old/new/pg
6357266Smuller  *	The first char in the string specifies the delimiter used by this
6457266Smuller  *	replacement string. "Old" is a regular expression in "ed" format which
6557266Smuller  *	is compiled by regcomp() and is applied to filenames. "new" is the
6657266Smuller  *	substitution string; p and g are options flags for printing and global
6757266Smuller  *	replacement (over the single filename)
6857266Smuller  * Return:
6957266Smuller  *	0 if a proper replacement string and regular expression was added to
7057266Smuller  *	the list of replacement patterns; -1 otherwise.
7157266Smuller  */
7257266Smuller 
7357266Smuller #if __STDC__
7457266Smuller int
7557266Smuller rep_add(register char *str)
7657266Smuller #else
7757266Smuller int
7857266Smuller rep_add(str)
7957266Smuller 	register char *str;
8057266Smuller #endif
8157266Smuller {
8257266Smuller 	register char *pt1;
8357266Smuller 	register char *pt2;
8457266Smuller 	register REPLACE *rep;
8557266Smuller #	ifndef NET2_REGEX
8657266Smuller 	register int res;
8757266Smuller 	char rebuf[BUFSIZ];
8857266Smuller #	endif
8957266Smuller 
9057266Smuller 	/*
9157266Smuller 	 * throw out the bad parameters
9257266Smuller 	 */
9357266Smuller 	if ((str == NULL) || (*str == '\0')) {
9457266Smuller 		warn(1, "Empty replacement string");
9557266Smuller 		return(-1);
9657266Smuller 	}
9757266Smuller 
9857266Smuller 	/*
9957266Smuller 	 * first character in the string specifies what the delimiter is for
10057266Smuller 	 * this expression
10157266Smuller 	 */
10257266Smuller 	if ((pt1 = strchr(str+1, *str)) == NULL) {
10357266Smuller 		warn(1, "Invalid replacement string %s", str);
10457266Smuller 		return(-1);
10557266Smuller 	}
10657266Smuller 
10757266Smuller 	/*
10857266Smuller 	 * allocate space for the node that handles this replacement pattern
10957266Smuller 	 * and split out the regular expression and try to compile it
11057266Smuller 	 */
11157266Smuller 	if ((rep = (REPLACE *)malloc(sizeof(REPLACE))) == NULL) {
11257266Smuller 		warn(1, "Unable to allocate memory for replacement string");
11357266Smuller 		return(-1);
11457266Smuller 	}
11557266Smuller 
11657266Smuller 	*pt1 = '\0';
11757266Smuller #	ifdef NET2_REGEX
11857266Smuller 	if ((rep->rcmp = regcomp(str+1)) == NULL) {
11957266Smuller #	else
12057266Smuller 	if ((res = regcomp(&(rep->rcmp), str+1, 0)) != 0) {
12157266Smuller 		regerror(res, &(rep->rcmp), rebuf, sizeof(rebuf));
12257266Smuller 		warn(1, "%s while compiling regular expression %s", rebuf, str);
12357266Smuller #	endif
12457266Smuller 		(void)free((char *)rep);
12557266Smuller 		return(-1);
12657266Smuller 	}
12757266Smuller 
12857266Smuller 	/*
12957266Smuller 	 * put the delimiter back in case we need an error message and
13057266Smuller 	 * locate the delimiter at the end of the replacement string
13157266Smuller 	 * we then point the node at the new substitution string
13257266Smuller 	 */
13357266Smuller 	*pt1++ = *str;
13457266Smuller 	if ((pt2 = strchr(pt1, *str)) == NULL) {
13557266Smuller #		ifdef NET2_REGEX
13657266Smuller 		(void)free((char *)rep->rcmp);
13757266Smuller #		else
13857266Smuller 		regfree(&(rep->rcmp));
13957266Smuller #		endif
14057266Smuller 		(void)free((char *)rep);
14157266Smuller 		warn(1, "Invalid replacement string %s", str);
14257266Smuller 		return(-1);
14357266Smuller 	}
14457266Smuller 
14557266Smuller 	*pt2 = '\0';
14657266Smuller 	rep->nstr = pt1;
14757266Smuller 	pt1 = pt2++;
14857266Smuller 	rep->flgs = 0;
14957266Smuller 
15057266Smuller 	/*
15157266Smuller 	 * set the options if any
15257266Smuller 	 */
15357266Smuller 	while (*pt2 != '\0') {
15457266Smuller 		switch(*pt2) {
15557266Smuller 		case 'g':
15657266Smuller 		case 'G':
15757266Smuller 			rep->flgs  |= GLOB;
15857266Smuller 			break;
15957266Smuller 		case 'p':
16057266Smuller 		case 'P':
16157266Smuller 			rep->flgs  |= PRNT;
16257266Smuller 			break;
16357266Smuller 		default:
16457266Smuller #			ifdef NET2_REGEX
16557266Smuller 			(void)free((char *)rep->rcmp);
16657266Smuller #			else
16757266Smuller 			regfree(&(rep->rcmp));
16857266Smuller #			endif
16957266Smuller 			(void)free((char *)rep);
17057266Smuller 			*pt1 = *str;
17157266Smuller 			warn(1, "Invalid replacement string option %s", str);
17257266Smuller 			return(-1);
17357266Smuller 		}
17457266Smuller 		++pt2;
17557266Smuller 	}
17657266Smuller 
17757266Smuller 	/*
17857266Smuller 	 * all done, link it in at the end
17957266Smuller 	 */
18057266Smuller 	rep->fow = NULL;
18157266Smuller 	if (rephead == NULL) {
18257266Smuller 		reptail = rephead = rep;
18357266Smuller 		return(0);
18457266Smuller 	}
18557266Smuller 	reptail->fow = rep;
18657266Smuller 	reptail = rep;
18757266Smuller 	return(0);
18857266Smuller }
18957266Smuller 
19057266Smuller /*
19157266Smuller  * pat_add()
19257266Smuller  *	add a pattern match to the pattern match list. Pattern matches are used
19357266Smuller  *	to select which archive members are extracted. (They appear as
19457266Smuller  *	arguments to pax in the list and read modes). If no patterns are
19557266Smuller  *	supplied to pax, all members in the archive will be selected (and the
19657266Smuller  *	pattern match list is empty).
19757266Smuller  * Return:
19857266Smuller  *	0 if the pattern was added to the list, -1 otherwise
19957266Smuller  */
20057266Smuller 
20157266Smuller #if __STDC__
20257266Smuller int
20357266Smuller pat_add(char *str)
20457266Smuller #else
20557266Smuller int
20657266Smuller pat_add(str)
20757266Smuller 	char *str;
20857266Smuller #endif
20957266Smuller {
21057266Smuller 	register PATTERN *pt;
21157266Smuller 
21257266Smuller 	/*
21357266Smuller 	 * throw out the junk
21457266Smuller 	 */
21557266Smuller 	if ((str == NULL) || (*str == '\0')) {
21657266Smuller 		warn(1, "Empty pattern string");
21757266Smuller 		return(-1);
21857266Smuller 	}
21957266Smuller 
22057266Smuller 	/*
22157266Smuller 	 * allocate space for the pattern and store the pattern. the pattern is
22257266Smuller 	 * part of argv so do not bother to copy it, just point at it. Add the
22357266Smuller 	 * node to the end of the pattern list
22457266Smuller 	 */
22557266Smuller 	if ((pt = (PATTERN *)malloc(sizeof(PATTERN))) == NULL) {
22657266Smuller 		warn(1, "Unable to allocate memory for pattern string");
22757266Smuller 		return(-1);
22857266Smuller 	}
22957266Smuller 
23057266Smuller 	pt->pstr = str;
23157266Smuller 	pt->plen = strlen(str);
23257266Smuller 	pt->fow = NULL;
23357266Smuller 	pt->flgs = 0;
23457266Smuller 	if (pathead == NULL) {
23557266Smuller 		pattail = pathead = pt;
23657266Smuller 		return(0);
23757266Smuller 	}
23857266Smuller 	pattail->fow = pt;
23957266Smuller 	pattail = pt;
24057266Smuller 	return(0);
24157266Smuller }
24257266Smuller 
24357266Smuller /*
24457266Smuller  * pat_chk()
24557266Smuller  *	complain if any the user supplied pattern did not result in a match to
24657266Smuller  *	a selected archive member.
24757266Smuller  */
24857266Smuller 
24957266Smuller #if __STDC__
25057266Smuller void
25157266Smuller pat_chk(void)
25257266Smuller #else
25357266Smuller void
25457266Smuller pat_chk()
25557266Smuller #endif
25657266Smuller {
25757266Smuller 	register PATTERN *pt;
25857266Smuller 	register int wban = 0;
25957266Smuller 
26057266Smuller 	/*
26157266Smuller 	 * walk down the list checking the flags to make sure MTCH was set,
26257266Smuller 	 * if not complain
26357266Smuller 	 */
26457266Smuller 	for (pt = pathead; pt != NULL; pt = pt->fow) {
26557266Smuller 		if (pt->flgs & MTCH)
26657266Smuller 			continue;
26757266Smuller 		if (!wban) {
26857266Smuller 			warn(1, "WARNING! These patterns were not matched:");
26957266Smuller 			++wban;
27057266Smuller 		}
27157266Smuller 		(void)fprintf(stderr, "%s\n", pt->pstr);
27257266Smuller 	}
27357266Smuller }
27457266Smuller 
27557266Smuller /*
27657266Smuller  * pat_sel()
27757266Smuller  *	the archive member which matches a pattern was selected. Mark the
27857266Smuller  *	pattern as having selected an archive member. arcn->pat points at the
27957266Smuller  *	pattern that was matched. arcn->pat is set in pat_match()
28057266Smuller  *
28157266Smuller  *	NOTE: When the -c option is used, we are called when there was no match
28257266Smuller  *	by pat_match() (that means we did match before the inverted sense of
28357266Smuller  *	the logic). Now this seems really strange at first, but with -c  we
28457266Smuller  *	need to keep track of those patterns that cause a archive member to NOT
28557266Smuller  *	be selected (it found an archive member with a specified pattern)
28657266Smuller  * Return:
28757266Smuller  *	0 if the pattern pointed at by arcn->pat was tagged as creating a
28857266Smuller  *	match, -1 otherwise.
28957266Smuller  */
29057266Smuller 
29157266Smuller #if __STDC__
29257266Smuller int
29357266Smuller pat_sel(register ARCHD *arcn)
29457266Smuller #else
29557266Smuller int
29657266Smuller pat_sel(arcn)
29757266Smuller 	register ARCHD *arcn;
29857266Smuller #endif
29957266Smuller {
30057266Smuller 	register PATTERN *pt;
30157266Smuller 	register PATTERN **ppt;
30257266Smuller 	register int len;
30357266Smuller 
30457266Smuller 	/*
30557266Smuller 	 * if no patterns just return
30657266Smuller 	 */
30757266Smuller 	if ((pathead == NULL) || ((pt = arcn->pat) == NULL))
30857266Smuller 		return(0);
30957266Smuller 
31057266Smuller 	/*
31157266Smuller 	 * when we are NOT limited to a single match per pattern mark the
31257266Smuller 	 * the pattern and return
31357266Smuller 	 */
31457266Smuller 	if (!nflag) {
31557266Smuller 		pt->flgs |= MTCH;
31657266Smuller 		if (dflag || (arcn->type != PAX_DIR))
31757266Smuller 			return(0);
31857266Smuller 		/*
31957266Smuller 		 * ok we matched a directory and we are allowing
32057266Smuller 		 * subtree matches. We add this as a DIR_MTCH pattern
32157266Smuller 		 * so all its children will match. Note we know that
32257266Smuller 		 * when successful, pat_add() puts the pattern at the
32357266Smuller 		 * tail (yup a kludge). In the code below will make
32457266Smuller 		 * a dir match pattern
32557266Smuller 		 */
32657266Smuller 		if ((pat_add(arcn->name) < 0) || ((pt = pattail) == NULL))
32757266Smuller 			return(-1);
32857266Smuller 		arcn->pat = pt;
32957266Smuller 	}
33057266Smuller 
33157266Smuller 	/*
33257266Smuller 	 * we reach this point only when we allow a single selected match per
33357266Smuller 	 * pattern, or we have to add a DIR_MATCH pattern. if the pattern
33457266Smuller 	 * matched a directory and we do not have -d * (dflag) we are done
33557266Smuller 	 * with this pattern. We may also be handed a file in the subtree of a
33657266Smuller 	 * directory. in that case when we are operating with -d, this pattern
33757266Smuller 	 * was already selected and we are done
33857266Smuller 	 */
33957266Smuller 	if (pt->flgs & DIR_MTCH)
34057266Smuller 		return(0);
34157266Smuller 
34257266Smuller 	if (!dflag && (arcn->type == PAX_DIR)) {
34357266Smuller 		/*
34457266Smuller 		 * we are allowing subtree matches at directories, mark the
34557266Smuller 		 * node as a directory match so pat_match() will only match
34657266Smuller 		 * children of this directory (we replace the pattern with the
34757266Smuller 		 * directory name to enforce this subtree only match)
34857266Smuller 		 * pat_match() looks for DIR_MTCH to determine what comparison
34957266Smuller 		 * technique to use when it checks for a pattern match
35057266Smuller 	 	 */
35157266Smuller 		if ((pt->pstr = strdup(arcn->name)) == NULL) {
35257266Smuller 			warn(1, "Pattern select out of memory");
35357266Smuller 			return(-1);
35457266Smuller 		}
35557266Smuller 		pt->plen = strlen(pt->pstr);
35657266Smuller 
35757266Smuller 		/*
35857266Smuller 		 * strip off any trailing /, this should really never happen
35957266Smuller 		 */
36057266Smuller 		len = pt->plen - 1;
36157266Smuller 		if (*(pt->pstr + len) == '/') {
36257266Smuller 			*(pt->pstr + len) = '\0';
36357266Smuller 			pt->plen = len;
36457266Smuller 		}
36557266Smuller 		pt->flgs |= DIR_MTCH | MTCH;
36657266Smuller 		return(0);
36757266Smuller 	}
36857266Smuller 
36957266Smuller 	/*
37057266Smuller 	 * it is not a directory, we are then done with this pattern, so we
37157266Smuller 	 * delete it from the list, as it can never be used for another match
37257266Smuller 	 * Seems kind of strange to do for a -c, but the pax spec is really
37357266Smuller 	 * vague on the interaction of -c -n and -d. We assume that when -c
37457266Smuller 	 * and the pattern rejects a member (i.e. it matched it) it is done.
37557266Smuller 	 * In effect we place the order of the flags as having -c last.
37657266Smuller 	 */
37757266Smuller 	pt = pathead;
37857266Smuller 	ppt = &pathead;
37957266Smuller 	while ((pt != NULL) && (pt != arcn->pat)) {
38057266Smuller 		ppt = &(pt->fow);
38157266Smuller 		pt = pt->fow;
38257266Smuller 	}
38357266Smuller 
38457266Smuller 	if (pt == NULL) {
38557266Smuller 		/*
38657266Smuller 		 * should never happen....
38757266Smuller 		 */
38857266Smuller 		warn(1, "Pattern list inconsistant");
38957266Smuller 		return(-1);
39057266Smuller 	}
39157266Smuller 	*ppt = pt->fow;
39257266Smuller 	(void)free((char *)pt);
39357266Smuller 	arcn->pat = NULL;
39457266Smuller 	return(0);
39557266Smuller }
39657266Smuller 
39757266Smuller /*
39857266Smuller  * pat_match()
39957266Smuller  *	see if this archive member matches any supplied pattern, if a match
40057266Smuller  *	is found, arcn->pat is set to point at the potential pattern. Later if
40157266Smuller  *	this archive member is "selected" we process and mark the pattern as
40257266Smuller  *	one which matched a selected archive member (see pat_sel())
40357266Smuller  * Return:
40457266Smuller  *	0 if this archive member should be processed, 1 if it should be
40557266Smuller  *	skipped and -1 if we are done with all patterns (and pax should quit
40657266Smuller  *	looking for more members)
40757266Smuller  */
40857266Smuller 
40957266Smuller #if __STDC__
41057266Smuller int
41157266Smuller pat_match(register ARCHD *arcn)
41257266Smuller #else
41357266Smuller int
41457266Smuller pat_match(arcn)
41557266Smuller 	register ARCHD *arcn;
41657266Smuller #endif
41757266Smuller {
41857266Smuller 	register PATTERN *pt;
41957266Smuller 
42057266Smuller 	arcn->pat = NULL;
42157266Smuller 
42257266Smuller 	/*
42357266Smuller 	 * if there are no more patterns and we have -n (and not -c) we are
42457266Smuller 	 * done. otherwise with no patterns to match, matches all
42557266Smuller 	 */
42657266Smuller 	if (pathead == NULL) {
42757266Smuller 		if (nflag && !cflag)
42857266Smuller 			return(-1);
42957266Smuller 		return(0);
43057266Smuller 	}
43157266Smuller 
43257266Smuller 	/*
43357266Smuller 	 * have to search down the list one at a time looking for a match.
43457266Smuller 	 */
43557266Smuller 	pt = pathead;
43657266Smuller 	while (pt != NULL) {
43757266Smuller 		/*
43857266Smuller 		 * check for a file name match unless we have DIR_MTCH set in
43957266Smuller 		 * this pattern then we want a prefix match
44057266Smuller 		 */
44157266Smuller 
44257266Smuller 		if (pt->flgs & DIR_MTCH) {
44357266Smuller 			/*
44457266Smuller 			 * this pattern was matched before to a directory
44557266Smuller 			 * as we must have -n set for this (but not -d). We can
44657266Smuller 			 * only match CHILDREN of that directory so we must use
44757266Smuller 			 * an exact prefix match
44857266Smuller 			 */
44957266Smuller 			if ((strncmp(pt->pstr, arcn->name, pt->plen) == 0) &&
45057266Smuller 			    (arcn->name[pt->plen] == '/'))
45157266Smuller 				break;
45257266Smuller 		} else if (fnmatch(pt->pstr, arcn->name, 0) == 0)
45357266Smuller 			break;
45457266Smuller 		pt = pt->fow;
45557266Smuller 	}
45657266Smuller 
45757266Smuller 	/*
45857266Smuller 	 * return the result, remember that cflag (-c) inverts the sense of a
45957266Smuller 	 * match
46057266Smuller 	 */
46157266Smuller 	if (pt == NULL)
46257266Smuller 		return(cflag ? 0 : 1);
46357266Smuller 
46457266Smuller 	/*
46557266Smuller 	 * we had a match, now when we invert the sense (-c) we reject this
46657266Smuller 	 * member. However we have to tag the pattern a being successful, (in a
46757266Smuller 	 * match, not in selecting a archive member) so we call pat_sel() here.
46857266Smuller 	 */
46957266Smuller 	arcn->pat = pt;
47057266Smuller 	if (!cflag)
47157266Smuller 		return(0);
47257266Smuller 
47357266Smuller 	if (pat_sel(arcn) < 0)
47457266Smuller 		return(-1);
47557266Smuller 	arcn->pat = NULL;
47657266Smuller 	return(1);
47757266Smuller }
47857266Smuller 
47957266Smuller /*
48057266Smuller  * mod_name()
48157266Smuller  *	modify a selected file name. first attempt to apply replacement string
48257266Smuller  *	expressions, then apply interactive file rename. We apply replacement
48357266Smuller  *	string expressions to both filenames and file links (if we didn't the
48457266Smuller  *	links would point to the wrong place, and we could never be able to
48557266Smuller  *	move an archive that has a file link in it). When we rename files
48657266Smuller  *	interactively, we store that mapping (old name to user input name) so
48757266Smuller  *	if we spot any file links to the old file name in the future, we will
48857266Smuller  *	know exactly how to fix the file link.
48957266Smuller  * Return:
49057266Smuller  *	0 continue to  process file, 1 skip this file, -1 pax is finished
49157266Smuller  */
49257266Smuller 
49357266Smuller #if __STDC__
49457266Smuller int
49557266Smuller mod_name(register ARCHD *arcn)
49657266Smuller #else
49757266Smuller int
49857266Smuller mod_name(arcn)
49957266Smuller 	register ARCHD *arcn;
50057266Smuller #endif
50157266Smuller {
50257266Smuller 	register int res = 0;
50357266Smuller 
50457266Smuller 	/*
50557266Smuller 	 * IMPORTANT: We have a problem. what do we do with symlinks?
50657266Smuller 	 * Modifying a hard link name makes sense, as we know the file it
50757266Smuller 	 * points at should have been seen already in the archive (and if it
50857266Smuller 	 * wasn't seen because of a read error or a bad archive, we lose
50957266Smuller 	 * anyway). But there are no such requirements for symlinks. On one
51057266Smuller 	 * hand the symlink that refers to a file in the archive will have to
51157266Smuller 	 * be modified to so it will still work at its new location in the
51257266Smuller 	 * file system. On the other hand a symlink that points elsewhere (and
51357266Smuller 	 * should continue to do so) should not be modified. There is clearly
51457266Smuller 	 * no perfect solution here. So we handle them like hardlinks. Clearly
51557266Smuller 	 * a replacement made by the interactive rename mapping is very likely
51657266Smuller 	 * to be correct since it applies to a single file and is an exact
51757266Smuller 	 * match. The regular expression replacements are a little harder to
51857266Smuller 	 * justify though. We claim that the symlink name is only likely
51957266Smuller 	 * to be replaced when it points within the file tree being moved and
52057266Smuller 	 * in that case it should be modified. what we really need to do is to
52157266Smuller 	 * call an oracle here. :)
52257266Smuller 	 */
52357266Smuller 	if (rephead != NULL) {
52457266Smuller 		/*
52557266Smuller 		 * we have replacement strings, modify the name and the link
52657266Smuller 		 * name if any.
52757266Smuller 		 */
52857266Smuller 		if ((res = rep_name(arcn->name, &(arcn->nlen), 1)) != 0)
52957266Smuller 			return(res);
53057266Smuller 
53157266Smuller 		if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
53257266Smuller 		    (arcn->type == PAX_HRG)) &&
53357266Smuller 		    ((res = rep_name(arcn->ln_name, &(arcn->ln_nlen), 0)) != 0))
53457266Smuller 			return(res);
53557266Smuller 	}
53657266Smuller 
53757266Smuller 	if (iflag) {
53857266Smuller 		/*
53957266Smuller 		 * perform interactive file rename, then map the link if any
54057266Smuller 		 */
54157266Smuller 		if ((res = tty_rename(arcn)) != 0)
54257266Smuller 			return(res);
54357266Smuller 		if ((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
54457266Smuller 		    (arcn->type == PAX_HRG))
54557266Smuller 			sub_name(arcn->ln_name, &(arcn->ln_nlen));
54657266Smuller 	}
54757266Smuller 	return(res);
54857266Smuller }
54957266Smuller 
55057266Smuller /*
55157266Smuller  * tty_rename()
55257266Smuller  *	Prompt the user for a replacement file name. A "." keeps the old name,
55357266Smuller  *	a empty line skips the file, and an EOF on reading the tty, will cause
55457266Smuller  *	pax to stop processing and exit. Otherwise the file name input, replaces
55557266Smuller  *	the old one.
55657266Smuller  * Return:
55757266Smuller  *	0 process this file, 1 skip this file, -1 we need to exit pax
55857266Smuller  */
55957266Smuller 
56057266Smuller #if __STDC__
56157266Smuller static int
56257266Smuller tty_rename(register ARCHD *arcn)
56357266Smuller #else
56457266Smuller static int
56557266Smuller tty_rename(arcn)
56657266Smuller 	register ARCHD *arcn;
56757266Smuller #endif
56857266Smuller {
56957266Smuller 	char tmpname[PAXPATHLEN+2];
57057266Smuller 	int res;
57157266Smuller 
57257266Smuller 	/*
57357266Smuller 	 * prompt user for the replacement name for a file, keep trying until
57457266Smuller 	 * we get some reasonable input. Archives may have more than one file
57557266Smuller 	 * on them with the same name (from updates etc). We print verbose info
57657266Smuller 	 * on the file so the user knows what is up.
57757266Smuller 	 */
57857266Smuller 	tty_prnt("\nATTENTION: Pax interactive file rename operation.\n");
57957266Smuller 
58057266Smuller 	for (;;) {
58157266Smuller 		ls_tty(arcn);
58257266Smuller 		tty_prnt("Input new name, or a \".\" to keep the old name, ");
58357266Smuller 		tty_prnt("or a \"return\" to skip this file.\n");
58457266Smuller 		tty_prnt("Input > ");
58557266Smuller 		if (tty_read(tmpname, sizeof(tmpname)) < 0)
58657266Smuller 			return(-1);
58757266Smuller 		if (strcmp(tmpname, "..") == 0) {
58857266Smuller 			tty_prnt("Try again, illegal file name: ..\n");
58957266Smuller 			continue;
59057266Smuller 		}
59157266Smuller 		if (strlen(tmpname) > PAXPATHLEN) {
59257266Smuller 			tty_prnt("Try again, file name too long\n");
59357266Smuller 			continue;
59457266Smuller 		}
59557266Smuller 		break;
59657266Smuller 	}
59757266Smuller 
59857266Smuller 	/*
59957266Smuller 	 * empty file name, skips this file. a "." leaves it alone
60057266Smuller 	 */
60157266Smuller 	if (tmpname[0] == '\0') {
60257266Smuller 		tty_prnt("Skipping file.\n");
60357266Smuller 		return(1);
60457266Smuller 	}
60557266Smuller 	if ((tmpname[0] == '.') && (tmpname[1] == '\0')) {
60657266Smuller 		tty_prnt("Processing continues, name unchanged.\n");
60757266Smuller 		return(0);
60857266Smuller 	}
60957266Smuller 
61057266Smuller 	/*
61157266Smuller 	 * ok the name changed. We may run into links that point at this
61257266Smuller 	 * file later. we have to remember where the user sent the file
61357266Smuller 	 * in order to repair any links.
61457266Smuller 	 */
61557266Smuller 	tty_prnt("Processing continues, name changed to: %s\n", tmpname);
61657266Smuller 	res = add_name(arcn->name, arcn->nlen, tmpname);
61757266Smuller 	arcn->nlen = l_strncpy(arcn->name, tmpname, PAXPATHLEN+1);
61857266Smuller 	if (res < 0)
61957266Smuller 		return(-1);
62057266Smuller 	return(0);
62157266Smuller }
62257266Smuller 
62357266Smuller /*
62457266Smuller  * set_dest()
62557266Smuller  *	fix up the file name and the link name (if any) so this file will land
62657266Smuller  *	in the destination directory (used during copy() -rw).
62757266Smuller  * Return:
62857266Smuller  *	0 if ok, -1 if failure (name too long)
62957266Smuller  */
63057266Smuller 
63157266Smuller #if __STDC__
63257266Smuller int
63357266Smuller set_dest(register ARCHD *arcn, char *dest_dir, int dir_len)
63457266Smuller #else
63557266Smuller int
63657266Smuller set_dest(arcn, dest_dir, dir_len)
63757266Smuller 	register ARCHD *arcn;
63857266Smuller 	char *dest_dir;
63957266Smuller 	int dir_len;
64057266Smuller #endif
64157266Smuller {
64257266Smuller 	if (fix_path(arcn->name, &(arcn->nlen), dest_dir, dir_len) < 0)
64357266Smuller 		return(-1);
64457266Smuller 
64557600Smuller 	/*
64657600Smuller 	 * It is really hard to deal with symlinks here, we cannot be sure
64757600Smuller 	 * if the name they point was moved (or will be moved). It is best to
64857600Smuller 	 * leave them alone.
64957600Smuller 	 */
65057600Smuller 	if ((arcn->type != PAX_HLK) && (arcn->type != PAX_HRG))
65157266Smuller 		return(0);
65257266Smuller 
65357266Smuller 	if (fix_path(arcn->ln_name, &(arcn->ln_nlen), dest_dir, dir_len) < 0)
65457266Smuller 		return(-1);
65557266Smuller 	return(0);
65657266Smuller }
65757266Smuller 
65857266Smuller /*
65957266Smuller  * fix_path
66057266Smuller  *	concatenate dir_name and or_name and store the result in or_name (if
66157266Smuller  *	it fits). This is one ugly function.
66257266Smuller  * Return:
66357266Smuller  *	0 if ok, -1 if the final name is too long
66457266Smuller  */
66557266Smuller 
66657266Smuller #if __STDC__
66757266Smuller static int
66857266Smuller fix_path( char *or_name, int *or_len, char *dir_name, int dir_len)
66957266Smuller #else
67057266Smuller static int
67157266Smuller fix_path(or_name, or_len, dir_name, dir_len)
67257266Smuller 	char *or_name;
67357266Smuller 	int *or_len;
67457266Smuller 	char *dir_name;
67557266Smuller 	int dir_len;
67657266Smuller #endif
67757266Smuller {
67857266Smuller 	register char *src;
67957266Smuller 	register char *dest;
68057266Smuller 	register char *start;
68157266Smuller 	int len;
68257266Smuller 
68357266Smuller 	/*
68457266Smuller 	 * we shift the or_name to the right enough to tack in the dir_name
68557266Smuller 	 * at the front. We make sure we have enough space for it all before
68657266Smuller 	 * we start. since dest always ends in a slash, we skip of or_name
68757266Smuller 	 * if it also starts with one.
68857266Smuller 	 */
68957266Smuller 	start = or_name;
69057266Smuller 	src = start + *or_len;
69157266Smuller 	dest = src + dir_len;
69257266Smuller 	if (*start == '/') {
69357266Smuller 		++start;
69457266Smuller 		--dest;
69557266Smuller 	}
69657266Smuller 	if ((len = dest - or_name) > PAXPATHLEN) {
69757266Smuller 		warn(1, "File name %s/%s, too long", dir_name, start);
69857266Smuller 		return(-1);
69957266Smuller 	}
70057266Smuller 	*or_len = len;
70157266Smuller 
70257266Smuller 	/*
70357266Smuller 	 * enough space, shift
70457266Smuller 	 */
70557266Smuller 	while (src >= start)
70657266Smuller 		*dest-- = *src--;
70757266Smuller 	src = dir_name + dir_len - 1;
70857266Smuller 
70957266Smuller 	/*
71057266Smuller 	 * splice in the destination directory name
71157266Smuller 	 */
71257266Smuller 	while (src >= dir_name)
71357266Smuller 		*dest-- = *src--;
71457266Smuller 
71557266Smuller 	*(or_name + len) = '\0';
71657266Smuller 	return(0);
71757266Smuller }
71857266Smuller 
71957266Smuller /*
72057266Smuller  * rep_name()
72157266Smuller  *	walk down the list of replacement strings applying each one in order.
72257266Smuller  *	when we find one with a successful substitution, we modify the name
72357266Smuller  *	as specified. if required, we print the results. if the resulting name
72457266Smuller  *	is empty, we will skip this archive member. We use the regexp(3)
72557266Smuller  *	routines (regexp() ought to win a prize as having the most cryptic
72657266Smuller  *	library function manual page).
72757266Smuller  *	--Parameters--
72857266Smuller  *	name is the file name we are going to apply the regular expressions to
72957266Smuller  *	(and may be modified)
73057266Smuller  *	nlen is the length of this name (and is modified to hold the length of
73157266Smuller  *	the final string).
73257266Smuller  *	prnt is a flag that says whether to print the final result.
73357266Smuller  * Return:
73457266Smuller  *	0 if substitution was successful, 1 if we are to skip the file (the name
73557266Smuller  *	ended up empty)
73657266Smuller  */
73757266Smuller 
73857266Smuller #if __STDC__
73957266Smuller static int
74057266Smuller rep_name(char *name, int *nlen, int prnt)
74157266Smuller #else
74257266Smuller static int
74357266Smuller rep_name(name, nlen, prnt)
74457266Smuller 	char *name;
74557266Smuller 	int *nlen;
74657266Smuller 	int prnt;
74757266Smuller #endif
74857266Smuller {
74957266Smuller 	register REPLACE *pt;
75057266Smuller 	register char *inpt;
75157266Smuller 	register char *outpt;
75257266Smuller 	register char *endpt;
75357266Smuller 	register char *rpt;
75457266Smuller 	register int found = 0;
75557266Smuller 	register int res;
75657266Smuller #	ifndef NET2_REGEX
75757266Smuller 	regmatch_t pm[MAXSUBEXP];
75857266Smuller #	endif
75957266Smuller 	char nname[PAXPATHLEN+1];	/* final result of all replacements */
76057266Smuller 	char buf1[PAXPATHLEN+1];	/* where we work on the name */
76157266Smuller 
76257266Smuller 	/*
76357266Smuller 	 * copy the name into buf1, where we will work on it. We need to keep
76457266Smuller 	 * the orig string around so we can print out the result of the final
76557266Smuller 	 * replacement. We build up the final result in nname. inpt points at
76657266Smuller 	 * the string we apply the regular expression to. prnt is used to
76757266Smuller 	 * suppress printing when we handle replacements on the link field
76857266Smuller 	 * (the user already saw that substitution go by)
76957266Smuller 	 */
77057266Smuller 	pt = rephead;
77157266Smuller 	(void)strcpy(buf1, name);
77257266Smuller 	inpt = buf1;
77357266Smuller 	outpt = nname;
77457266Smuller 	endpt = outpt + PAXPATHLEN;
77557266Smuller 
77657266Smuller 	/*
77757266Smuller 	 * try each replacement string in order
77857266Smuller 	 */
77957266Smuller 	while (pt != NULL) {
78057266Smuller 		do {
78157266Smuller 			/*
78257266Smuller 			 * check for a successful substitution, if not go to
78357266Smuller 			 * the next pattern, or cleanup if we were global
78457266Smuller 			 */
78557266Smuller #			ifdef NET2_REGEX
78657266Smuller 			if (regexec(pt->rcmp, inpt) == 0)
78757266Smuller #			else
78857266Smuller 			if (regexec(&(pt->rcmp), inpt, MAXSUBEXP, pm, 0) != 0)
78957266Smuller #			endif
79057266Smuller 				break;
79157266Smuller 
79257266Smuller 			/*
79357266Smuller 			 * ok we found one. We have three parts, the prefix
79457266Smuller 			 * which did not match, the section that did and the
79557266Smuller 			 * tail (that also did not match). Copy the prefix to
79657266Smuller 			 * the final output buffer (watching to make sure we
79757266Smuller 			 * do not create a string too long).
79857266Smuller 			 */
79957266Smuller 			found = 1;
80057266Smuller #			ifdef NET2_REGEX
80157266Smuller 			rpt = pt->rcmp->startp[0];
80257266Smuller #			else
80357266Smuller 			rpt = inpt + pm[0].rm_so;
80457266Smuller #			endif
80557266Smuller 
80657266Smuller 			while ((inpt < rpt) && (outpt < endpt))
80757266Smuller 				*outpt++ = *inpt++;
80857266Smuller 			if (outpt == endpt)
80957266Smuller 				break;
81057266Smuller 
81157266Smuller 			/*
81257266Smuller 			 * for the second part (which matched the regular
81357266Smuller 			 * expression) apply the substitution using the
81457266Smuller 			 * replacement string and place it the prefix in the
81557266Smuller 			 * final output. If we have problems, skip it.
81657266Smuller 			 */
81757266Smuller #			ifdef NET2_REGEX
81857266Smuller 			if ((res = resub(pt->rcmp,pt->nstr,outpt,endpt)) < 0) {
81957266Smuller #			else
82057266Smuller 			if ((res = resub(&(pt->rcmp),pm,pt->nstr,outpt,endpt))
82157266Smuller 			    < 0) {
82257266Smuller #			endif
82357266Smuller 				if (prnt)
82457266Smuller 					warn(1, "Replacement name error %s",
82557266Smuller 					    name);
82657266Smuller 				return(1);
82757266Smuller 			}
82857266Smuller 			outpt += res;
82957266Smuller 
83057266Smuller 			/*
83157266Smuller 			 * we set up to look again starting at the first
83257266Smuller 			 * character in the tail (of the input string right
83357266Smuller 			 * after the last character matched by the regular
83457266Smuller 			 * expression (inpt always points at the first char in
83557266Smuller 			 * the string to process). If we are not doing a global
83657266Smuller 			 * substitution, we will use inpt to copy the tail to
83757266Smuller 			 * the final result. Make sure we do not overrun the
83857266Smuller 			 * output buffer
83957266Smuller 			 */
84057266Smuller #			ifdef NET2_REGEX
84157266Smuller 			inpt = pt->rcmp->endp[0];
84257266Smuller #			else
84357266Smuller 			inpt += pm[0].rm_eo;
84457266Smuller #			endif
84557266Smuller 
84657266Smuller 			if ((outpt == endpt) || (*inpt == '\0'))
84757266Smuller 				break;
84857266Smuller 
84957266Smuller 			/*
85057266Smuller 			 * if the user wants global we keep trying to
85157266Smuller 			 * substitute until it fails, then we are done.
85257266Smuller 			 */
85357266Smuller 		} while (pt->flgs & GLOB);
85457266Smuller 
85557266Smuller 		if (found)
85657266Smuller 			break;
85757266Smuller 
85857266Smuller 		/*
85957266Smuller 		 * a successful substitution did NOT occur, try the next one
86057266Smuller 		 */
86157266Smuller 		pt = pt->fow;
86257266Smuller 	}
86357266Smuller 
86457266Smuller 	if (found) {
86557266Smuller 		/*
86657266Smuller 		 * we had a substitution, copy the last tail piece (if there is
86757266Smuller 		 * room) to the final result
86857266Smuller 		 */
86957266Smuller 		while ((outpt < endpt) && (*inpt != '\0'))
87057266Smuller 			*outpt++ = *inpt++;
87157266Smuller 
87257266Smuller 		*outpt = '\0';
87357266Smuller 		if ((outpt == endpt) && (*inpt != '\0')) {
87457266Smuller 			if (prnt)
87557266Smuller 				warn(1,"Replacement name too long %s >> %s",
87657266Smuller 				    name, nname);
87757266Smuller 			return(1);
87857266Smuller 		}
87957266Smuller 
88057266Smuller 		/*
88157266Smuller 		 * inform the user of the result if wanted
88257266Smuller 		 */
88357266Smuller 		if (prnt && (pt->flgs & PRNT)) {
88457266Smuller 			if (*nname == '\0')
88557266Smuller 				(void)fprintf(stderr,"%s >> <empty string>\n",
88657266Smuller 				    name);
88757266Smuller 			else
88857266Smuller 				(void)fprintf(stderr,"%s >> %s\n", name, nname);
88957266Smuller 		}
89057266Smuller 
89157266Smuller 		/*
89257266Smuller 		 * if empty inform the caller this file is to be skipped
89357266Smuller 		 * otherwise copy the new name over the orig name and return
89457266Smuller 		 */
89557266Smuller 		if (*nname == '\0')
89657266Smuller 			return(1);
89757266Smuller 		*nlen = l_strncpy(name, nname, PAXPATHLEN + 1);
89857266Smuller 	}
89957266Smuller 	return(0);
90057266Smuller }
90157266Smuller 
90257266Smuller #ifdef NET2_REGEX
90357266Smuller /*
90457266Smuller  * resub()
90557266Smuller  *	apply the replacement to the matched expression. expand out the old
90657266Smuller  * 	style ed(1) subexpression expansion.
90757266Smuller  * Return:
90857266Smuller  *	-1 if error, or the number of characters added to the destination.
90957266Smuller  */
91057266Smuller 
91157266Smuller #if __STDC__
91257266Smuller static int
91357266Smuller resub(regexp *prog, char *src, char *dest, register char *destend)
91457266Smuller #else
91557266Smuller static int
91657266Smuller resub(prog, src, dest, destend)
91757266Smuller 	regexp *prog;
91857266Smuller 	char *src;
91957266Smuller 	char *dest;
92057266Smuller 	register char *destend;
92157266Smuller #endif
92257266Smuller {
92357266Smuller 	register char *spt;
92457266Smuller 	register char *dpt;
92557266Smuller 	register char c;
92657266Smuller 	register int no;
92757266Smuller 	register int len;
92857266Smuller 
92957266Smuller 	spt = src;
93057266Smuller 	dpt = dest;
93157266Smuller 	while ((dpt < destend) && ((c = *spt++) != '\0')) {
93257266Smuller 		if (c == '&')
93357266Smuller 			no = 0;
93457266Smuller 		else if ((c == '\\') && (*spt >= '0') && (*spt <= '9'))
93557266Smuller 			no = *spt++ - '0';
93657266Smuller 		else {
93757266Smuller  			if ((c == '\\') && ((*spt == '\\') || (*spt == '&')))
93857266Smuller  				c = *spt++;
93957266Smuller  			*dpt++ = c;
94057266Smuller 			continue;
94157266Smuller 		}
94257266Smuller  		if ((prog->startp[no] == NULL) || (prog->endp[no] == NULL) ||
94357266Smuller 		    ((len = prog->endp[no] - prog->startp[no]) <= 0))
94457266Smuller 			continue;
94557266Smuller 
94657266Smuller 		/*
94757266Smuller 		 * copy the subexpression to the destination.
94857266Smuller 		 * fail if we run out of space or the match string is damaged
94957266Smuller 		 */
95057266Smuller 		if (len > (destend - dpt))
95157266Smuller 			len = destend - dpt;
95257266Smuller 		if (l_strncpy(dpt, prog->startp[no], len) != len)
95357266Smuller 			return(-1);
95457266Smuller 		dpt += len;
95557266Smuller 	}
95657266Smuller 	return(dpt - dest);
95757266Smuller }
95857266Smuller 
95957266Smuller #else
96057266Smuller 
96157266Smuller /*
96257266Smuller  * resub()
96357266Smuller  *	apply the replacement to the matched expression. expand out the old
96457266Smuller  * 	style ed(1) subexpression expansion.
96557266Smuller  * Return:
96657266Smuller  *	-1 if error, or the number of characters added to the destination.
96757266Smuller  */
96857266Smuller 
96957266Smuller #if __STDC__
97057266Smuller static int
97157266Smuller resub(regex_t *rp, register regmatch_t *pm, char *src, char *dest,
97257266Smuller 	register char *destend)
97357266Smuller #else
97457266Smuller static int
97557266Smuller resub(rp, pm, src, dest, destend)
97657266Smuller 	regex_t *rp;
97757266Smuller 	register regmatch_t *pm;
97857266Smuller 	char *src;
97957266Smuller 	char *dest;
98057266Smuller 	register char *destend;
98157266Smuller #endif
98257266Smuller {
98357266Smuller 	register char *spt;
98457266Smuller 	register char *dpt;
98557266Smuller 	register char c;
98657266Smuller 	register regmatch_t *pmpt;
98757266Smuller 	register int len;
98857266Smuller 	int subexcnt;
98957266Smuller 
99057266Smuller 	spt =  src;
99157266Smuller 	dpt = dest;
99257266Smuller 	subexcnt = rp->re_nsub;
99357266Smuller 	while ((dpt < destend) && ((c = *spt++) != '\0')) {
99457266Smuller 		/*
99557266Smuller 		 * see if we just have an ordinary replacement character
99657266Smuller 		 * or we refer to a subexpression.
99757266Smuller 		 */
99857266Smuller 		if (c == '&') {
99957266Smuller 			pmpt = pm;
100057266Smuller 		} else if ((c == '\\') && (*spt >= '0') && (*spt <= '9')) {
100157266Smuller 			/*
100257266Smuller 			 * make sure there is a subexpression as specified
100357266Smuller 			 */
100457266Smuller 			if ((len = *spt++ - '0') > subexcnt)
100557266Smuller 				return(-1);
100657266Smuller 			pmpt = pm + len;
100757266Smuller 		} else {
100857266Smuller  			/*
100957266Smuller 			 * Ordinary character, just copy it
101057266Smuller 			 */
101157266Smuller  			if ((c == '\\') && ((*spt == '\\') || (*spt == '&')))
101257266Smuller  				c = *spt++;
101357266Smuller  			*dpt++ = c;
101457266Smuller 			continue;
101557266Smuller 		}
101657266Smuller 
101757266Smuller 		/*
101857266Smuller 		 * continue if the subexpression is bogus
101957266Smuller 		 */
102057266Smuller 		if ((pmpt->rm_so < 0) || (pmpt->rm_eo < 0) ||
102157266Smuller 		    ((len = pmpt->rm_eo - pmpt->rm_so) <= 0))
102257266Smuller 			continue;
102357266Smuller 
102457266Smuller 		/*
102557266Smuller 		 * copy the subexpression to the destination.
102657266Smuller 		 * fail if we run out of space or the match string is damaged
102757266Smuller 		 */
102857266Smuller 		if (len > (destend - dpt))
102957266Smuller 			len = destend - dpt;
103057266Smuller 		if (l_strncpy(dpt, src + pmpt->rm_so, len) != len)
103157266Smuller 			return(-1);
103257266Smuller 		dpt += len;
103357266Smuller 	}
103457266Smuller 	return(dpt - dest);
103557266Smuller }
103657266Smuller #endif
1037