]> git.cameronkatri.com Git - mandoc.git/blob - compat_fts.c
Improve HTML formatting of .Bl -tag.
[mandoc.git] / compat_fts.c
1 #include "config.h"
2
3 #if HAVE_FTS
4
5 int dummy;
6
7 #else
8
9 /* $Id: compat_fts.c,v 1.12 2016/10/18 23:58:12 schwarze Exp $ */
10 /* $OpenBSD: fts.c,v 1.56 2016/09/21 04:38:56 guenther Exp $ */
11
12 /*-
13 * Copyright (c) 1990, 1993, 1994
14 * The Regents of the University of California. All rights reserved.
15 *
16 * Redistribution and use in source and binary forms, with or without
17 * modification, are permitted provided that the following conditions
18 * are met:
19 * 1. Redistributions of source code must retain the above copyright
20 * notice, this list of conditions and the following disclaimer.
21 * 2. Redistributions in binary form must reproduce the above copyright
22 * notice, this list of conditions and the following disclaimer in the
23 * documentation and/or other materials provided with the distribution.
24 * 3. Neither the name of the University nor the names of its contributors
25 * may be used to endorse or promote products derived from this software
26 * without specific prior written permission.
27 *
28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * SUCH DAMAGE.
39 */
40
41 #include <sys/stat.h>
42 #include <sys/types.h>
43
44 #include <dirent.h>
45 #include <errno.h>
46 #include <fcntl.h>
47 #include <limits.h>
48 #include <stdlib.h>
49 #include <string.h>
50 #include <unistd.h>
51 #include "compat_fts.h"
52
53 #define MAXIMUM(a, b) (((a) > (b)) ? (a) : (b))
54
55 static FTSENT *fts_alloc(FTS *, const char *, size_t);
56 static FTSENT *fts_build(FTS *);
57 static void fts_lfree(FTSENT *);
58 static void fts_load(FTS *, FTSENT *);
59 static size_t fts_maxarglen(char * const *);
60 static void fts_padjust(FTS *, FTSENT *);
61 static int fts_palloc(FTS *, size_t);
62 static FTSENT *fts_sort(FTS *, FTSENT *, int);
63 static unsigned short fts_stat(FTS *, FTSENT *);
64
65 #define ISDOT(a) (a[0] == '.' && (!a[1] || (a[1] == '.' && !a[2])))
66 #ifndef O_DIRECTORY
67 #define O_DIRECTORY 0
68 #endif
69 #ifndef O_CLOEXEC
70 #define O_CLOEXEC 0
71 #endif
72 #ifndef PATH_MAX
73 #define PATH_MAX 4096
74 #endif
75
76 #define CLR(opt) (sp->fts_options &= ~(opt))
77 #define ISSET(opt) (sp->fts_options & (opt))
78 #define SET(opt) (sp->fts_options |= (opt))
79
80 FTS *
81 fts_open(char * const *argv, int options,
82 int (*compar)(const FTSENT **, const FTSENT **))
83 {
84 FTS *sp;
85 FTSENT *p, *root;
86 int nitems;
87 FTSENT *parent, *tmp;
88
89 /* Options check. */
90 if (options & ~FTS_OPTIONMASK) {
91 errno = EINVAL;
92 return (NULL);
93 }
94
95 /* At least one path must be specified. */
96 if (*argv == NULL) {
97 errno = EINVAL;
98 return (NULL);
99 }
100
101 /* Allocate/initialize the stream */
102 if ((sp = calloc(1, sizeof(FTS))) == NULL)
103 return (NULL);
104 sp->fts_compar = compar;
105 sp->fts_options = options;
106
107 /*
108 * Start out with 1K of path space, and enough, in any case,
109 * to hold the user's paths.
110 */
111 if (fts_palloc(sp, MAXIMUM(fts_maxarglen(argv), PATH_MAX)))
112 goto mem1;
113
114 /* Allocate/initialize root's parent. */
115 if ((parent = fts_alloc(sp, "", 0)) == NULL)
116 goto mem2;
117 parent->fts_level = FTS_ROOTPARENTLEVEL;
118
119 /* Allocate/initialize root(s). */
120 for (root = NULL, nitems = 0; *argv; ++argv, ++nitems) {
121 if ((p = fts_alloc(sp, *argv, strlen(*argv))) == NULL)
122 goto mem3;
123 p->fts_level = FTS_ROOTLEVEL;
124 p->fts_parent = parent;
125 p->fts_accpath = p->fts_name;
126 p->fts_info = fts_stat(sp, p);
127
128 /* Command-line "." and ".." are real directories. */
129 if (p->fts_info == FTS_DOT)
130 p->fts_info = FTS_D;
131
132 /*
133 * If comparison routine supplied, traverse in sorted
134 * order; otherwise traverse in the order specified.
135 */
136 if (compar) {
137 p->fts_link = root;
138 root = p;
139 } else {
140 p->fts_link = NULL;
141 if (root == NULL)
142 tmp = root = p;
143 else {
144 tmp->fts_link = p;
145 tmp = p;
146 }
147 }
148 }
149 if (compar && nitems > 1)
150 root = fts_sort(sp, root, nitems);
151
152 /*
153 * Allocate a dummy pointer and make fts_read think that we've just
154 * finished the node before the root(s); set p->fts_info to FTS_INIT
155 * so that everything about the "current" node is ignored.
156 */
157 if ((sp->fts_cur = fts_alloc(sp, "", 0)) == NULL)
158 goto mem3;
159 sp->fts_cur->fts_link = root;
160 sp->fts_cur->fts_info = FTS_INIT;
161
162 if (nitems == 0)
163 free(parent);
164
165 return (sp);
166
167 mem3: fts_lfree(root);
168 free(parent);
169 mem2: free(sp->fts_path);
170 mem1: free(sp);
171 return (NULL);
172 }
173
174 static void
175 fts_load(FTS *sp, FTSENT *p)
176 {
177 size_t len;
178 char *cp;
179
180 /*
181 * Load the stream structure for the next traversal. Since we don't
182 * actually enter the directory until after the preorder visit, set
183 * the fts_accpath field specially so the chdir gets done to the right
184 * place and the user can access the first node. From fts_open it's
185 * known that the path will fit.
186 */
187 len = p->fts_pathlen = p->fts_namelen;
188 memmove(sp->fts_path, p->fts_name, len + 1);
189 if ((cp = strrchr(p->fts_name, '/')) && (cp != p->fts_name || cp[1])) {
190 len = strlen(++cp);
191 memmove(p->fts_name, cp, len + 1);
192 p->fts_namelen = len;
193 }
194 p->fts_accpath = p->fts_path = sp->fts_path;
195 sp->fts_dev = p->fts_dev;
196 }
197
198 int
199 fts_close(FTS *sp)
200 {
201 FTSENT *freep, *p;
202
203 /*
204 * This still works if we haven't read anything -- the dummy structure
205 * points to the root list, so we step through to the end of the root
206 * list which has a valid parent pointer.
207 */
208 if (sp->fts_cur) {
209 for (p = sp->fts_cur; p->fts_level >= FTS_ROOTLEVEL;) {
210 freep = p;
211 p = p->fts_link ? p->fts_link : p->fts_parent;
212 free(freep);
213 }
214 free(p);
215 }
216
217 /* Free up child linked list, sort array, path buffer, stream ptr.*/
218 if (sp->fts_child)
219 fts_lfree(sp->fts_child);
220 free(sp->fts_array);
221 free(sp->fts_path);
222 free(sp);
223
224 return (0);
225 }
226
227 /*
228 * Special case of "/" at the end of the path so that slashes aren't
229 * appended which would cause paths to be written as "....//foo".
230 */
231 #define NAPPEND(p) \
232 (p->fts_path[p->fts_pathlen - 1] == '/' \
233 ? p->fts_pathlen - 1 : p->fts_pathlen)
234
235 FTSENT *
236 fts_read(FTS *sp)
237 {
238 FTSENT *p, *tmp;
239 int instr;
240 char *t;
241
242 /* If finished or unrecoverable error, return NULL. */
243 if (sp->fts_cur == NULL || ISSET(FTS_STOP))
244 return (NULL);
245
246 /* Set current node pointer. */
247 p = sp->fts_cur;
248
249 /* Save and zero out user instructions. */
250 instr = p->fts_instr;
251 p->fts_instr = FTS_NOINSTR;
252
253 /* Directory in pre-order. */
254 if (p->fts_info == FTS_D) {
255 /* If skipped or crossed mount point, do post-order visit. */
256 if (instr == FTS_SKIP ||
257 (ISSET(FTS_XDEV) && p->fts_dev != sp->fts_dev)) {
258 if (sp->fts_child) {
259 fts_lfree(sp->fts_child);
260 sp->fts_child = NULL;
261 }
262 p->fts_info = FTS_DP;
263 return (p);
264 }
265
266 /*
267 * If haven't read do so. If the read fails, fts_build sets
268 * FTS_STOP or the fts_info field of the node.
269 */
270 if (sp->fts_child) {
271 /* nothing */
272 } else if ((sp->fts_child = fts_build(sp)) == NULL) {
273 if (ISSET(FTS_STOP))
274 return (NULL);
275 return (p);
276 }
277 p = sp->fts_child;
278 sp->fts_child = NULL;
279 goto name;
280 }
281
282 /* Move to the next node on this level. */
283 next: tmp = p;
284 if ((p = p->fts_link)) {
285 free(tmp);
286
287 /*
288 * If reached the top, return to the original directory (or
289 * the root of the tree), and load the paths for the next root.
290 */
291 if (p->fts_level == FTS_ROOTLEVEL) {
292 fts_load(sp, p);
293 return (sp->fts_cur = p);
294 }
295
296 /*
297 * User may have called fts_set on the node. If skipped,
298 * ignore. If followed, get a file descriptor so we can
299 * get back if necessary.
300 */
301 if (p->fts_instr == FTS_SKIP)
302 goto next;
303
304 name: t = sp->fts_path + NAPPEND(p->fts_parent);
305 *t++ = '/';
306 memmove(t, p->fts_name, p->fts_namelen + 1);
307 return (sp->fts_cur = p);
308 }
309
310 /* Move up to the parent node. */
311 p = tmp->fts_parent;
312 free(tmp);
313
314 if (p->fts_level == FTS_ROOTPARENTLEVEL) {
315 /*
316 * Done; free everything up and set errno to 0 so the user
317 * can distinguish between error and EOF.
318 */
319 free(p);
320 errno = 0;
321 return (sp->fts_cur = NULL);
322 }
323
324 /* NUL terminate the pathname. */
325 sp->fts_path[p->fts_pathlen] = '\0';
326
327 p->fts_info = p->fts_errno ? FTS_ERR : FTS_DP;
328 return (sp->fts_cur = p);
329 }
330
331 /*
332 * Fts_set takes the stream as an argument although it's not used in this
333 * implementation; it would be necessary if anyone wanted to add global
334 * semantics to fts using fts_set. An error return is allowed for similar
335 * reasons.
336 */
337 int
338 fts_set(FTS *sp, FTSENT *p, int instr)
339 {
340 if (instr && instr != FTS_NOINSTR && instr != FTS_SKIP) {
341 errno = EINVAL;
342 return (1);
343 }
344 p->fts_instr = instr;
345 return (0);
346 }
347
348 /*
349 * This is the tricky part -- do not casually change *anything* in here. The
350 * idea is to build the linked list of entries that are used by fts_children
351 * and fts_read. There are lots of special cases.
352 *
353 * The real slowdown in walking the tree is the stat calls. If FTS_NOSTAT is
354 * set and it's a physical walk (so that symbolic links can't be directories),
355 * we can do things quickly. First, if it's a 4.4BSD file system, the type
356 * of the file is in the directory entry. Otherwise, we assume that the number
357 * of subdirectories in a node is equal to the number of links to the parent.
358 * The former skips all stat calls. The latter skips stat calls in any leaf
359 * directories and for any files after the subdirectories in the directory have
360 * been found, cutting the stat calls by about 2/3.
361 */
362 static FTSENT *
363 fts_build(FTS *sp)
364 {
365 struct dirent *dp;
366 FTSENT *p, *head;
367 FTSENT *cur, *tail;
368 DIR *dirp;
369 void *oldaddr;
370 size_t dlen, len, maxlen;
371 int nitems, level, doadjust;
372 int saved_errno;
373 char *cp;
374
375 /* Set current node pointer. */
376 cur = sp->fts_cur;
377
378 /*
379 * Open the directory for reading. If this fails, we're done.
380 * If being called from fts_read, set the fts_info field.
381 */
382 if ((dirp = opendir(cur->fts_accpath)) == NULL) {
383 cur->fts_info = FTS_DNR;
384 cur->fts_errno = errno;
385 return (NULL);
386 }
387
388 /*
389 * Figure out the max file name length that can be stored in the
390 * current path -- the inner loop allocates more path as necessary.
391 * We really wouldn't have to do the maxlen calculations here, we
392 * could do them in fts_read before returning the path, but it's a
393 * lot easier here since the length is part of the dirent structure.
394 *
395 * If not changing directories set a pointer so that can just append
396 * each new name into the path.
397 */
398 len = NAPPEND(cur);
399 cp = sp->fts_path + len;
400 *cp++ = '/';
401 len++;
402 maxlen = sp->fts_pathlen - len;
403
404 /*
405 * fts_level is signed so we must prevent it from wrapping
406 * around to FTS_ROOTLEVEL and FTS_ROOTPARENTLEVEL.
407 */
408 level = cur->fts_level;
409 if (level < FTS_MAXLEVEL)
410 level++;
411
412 /* Read the directory, attaching each entry to the `link' pointer. */
413 doadjust = 0;
414 for (head = tail = NULL, nitems = 0; dirp && (dp = readdir(dirp));) {
415 if (ISDOT(dp->d_name))
416 continue;
417
418 #if HAVE_DIRENT_NAMLEN
419 dlen = dp->d_namlen;
420 #else
421 dlen = strlen(dp->d_name);
422 #endif
423
424 if (!(p = fts_alloc(sp, dp->d_name, dlen)))
425 goto mem1;
426 if (dlen >= maxlen) { /* include space for NUL */
427 oldaddr = sp->fts_path;
428 if (fts_palloc(sp, dlen + len + 1)) {
429 /*
430 * No more memory for path or structures. Save
431 * errno, free up the current structure and the
432 * structures already allocated.
433 */
434 mem1: saved_errno = errno;
435 free(p);
436 fts_lfree(head);
437 (void)closedir(dirp);
438 cur->fts_info = FTS_ERR;
439 SET(FTS_STOP);
440 errno = saved_errno;
441 return (NULL);
442 }
443 /* Did realloc() change the pointer? */
444 if (oldaddr != sp->fts_path) {
445 doadjust = 1;
446 cp = sp->fts_path + len;
447 }
448 maxlen = sp->fts_pathlen - len;
449 }
450
451 p->fts_level = level;
452 p->fts_parent = sp->fts_cur;
453 p->fts_pathlen = len + dlen;
454 if (p->fts_pathlen < len) {
455 /*
456 * If we wrap, free up the current structure and
457 * the structures already allocated, then error
458 * out with ENAMETOOLONG.
459 */
460 free(p);
461 fts_lfree(head);
462 (void)closedir(dirp);
463 cur->fts_info = FTS_ERR;
464 SET(FTS_STOP);
465 errno = ENAMETOOLONG;
466 return (NULL);
467 }
468
469 /* Build a file name for fts_stat to stat. */
470 p->fts_accpath = p->fts_path;
471 memmove(cp, p->fts_name, p->fts_namelen + 1);
472 /* Stat it. */
473 p->fts_info = fts_stat(sp, p);
474
475 /* We walk in directory order so "ls -f" doesn't get upset. */
476 p->fts_link = NULL;
477 if (head == NULL)
478 head = tail = p;
479 else {
480 tail->fts_link = p;
481 tail = p;
482 }
483 ++nitems;
484 }
485 if (dirp)
486 (void)closedir(dirp);
487
488 /*
489 * If realloc() changed the address of the path, adjust the
490 * addresses for the rest of the tree and the dir list.
491 */
492 if (doadjust)
493 fts_padjust(sp, head);
494
495 /*
496 * If not changing directories, reset the path back to original
497 * state.
498 */
499 if (len == sp->fts_pathlen || nitems == 0)
500 --cp;
501 *cp = '\0';
502
503 /* If didn't find anything, return NULL. */
504 if (!nitems) {
505 cur->fts_info = FTS_DP;
506 return (NULL);
507 }
508
509 /* Sort the entries. */
510 if (sp->fts_compar && nitems > 1)
511 head = fts_sort(sp, head, nitems);
512 return (head);
513 }
514
515 static unsigned short
516 fts_stat(FTS *sp, FTSENT *p)
517 {
518 FTSENT *t;
519 dev_t dev;
520 ino_t ino;
521 struct stat *sbp;
522
523 /* If user needs stat info, stat buffer already allocated. */
524 sbp = p->fts_statp;
525
526 if (lstat(p->fts_accpath, sbp)) {
527 p->fts_errno = errno;
528 memset(sbp, 0, sizeof(struct stat));
529 return (FTS_NS);
530 }
531
532 if (S_ISDIR(sbp->st_mode)) {
533 /*
534 * Set the device/inode. Used to find cycles and check for
535 * crossing mount points. Also remember the link count, used
536 * in fts_build to limit the number of stat calls. It is
537 * understood that these fields are only referenced if fts_info
538 * is set to FTS_D.
539 */
540 dev = p->fts_dev = sbp->st_dev;
541 ino = p->fts_ino = sbp->st_ino;
542 p->fts_nlink = sbp->st_nlink;
543
544 if (ISDOT(p->fts_name))
545 return (FTS_DOT);
546
547 /*
548 * Cycle detection is done by brute force when the directory
549 * is first encountered. If the tree gets deep enough or the
550 * number of symbolic links to directories is high enough,
551 * something faster might be worthwhile.
552 */
553 for (t = p->fts_parent;
554 t->fts_level >= FTS_ROOTLEVEL; t = t->fts_parent)
555 if (ino == t->fts_ino && dev == t->fts_dev) {
556 p->fts_cycle = t;
557 return (FTS_DC);
558 }
559 return (FTS_D);
560 }
561 if (S_ISLNK(sbp->st_mode))
562 return (FTS_SL);
563 if (S_ISREG(sbp->st_mode))
564 return (FTS_F);
565 return (FTS_DEFAULT);
566 }
567
568 static FTSENT *
569 fts_sort(FTS *sp, FTSENT *head, int nitems)
570 {
571 FTSENT **ap, *p;
572
573 /*
574 * Construct an array of pointers to the structures and call qsort(3).
575 * Reassemble the array in the order returned by qsort. If unable to
576 * sort for memory reasons, return the directory entries in their
577 * current order. Allocate enough space for the current needs plus
578 * 40 so don't realloc one entry at a time.
579 */
580 if (nitems > sp->fts_nitems) {
581 struct _ftsent **a;
582
583 sp->fts_nitems = nitems + 40;
584 if ((a = reallocarray(sp->fts_array,
585 sp->fts_nitems, sizeof(FTSENT *))) == NULL) {
586 free(sp->fts_array);
587 sp->fts_array = NULL;
588 sp->fts_nitems = 0;
589 return (head);
590 }
591 sp->fts_array = a;
592 }
593 for (ap = sp->fts_array, p = head; p; p = p->fts_link)
594 *ap++ = p;
595 qsort(sp->fts_array, nitems, sizeof(FTSENT *), sp->fts_compar);
596 for (head = *(ap = sp->fts_array); --nitems; ++ap)
597 ap[0]->fts_link = ap[1];
598 ap[0]->fts_link = NULL;
599 return (head);
600 }
601
602 static FTSENT *
603 fts_alloc(FTS *sp, const char *name, size_t namelen)
604 {
605 FTSENT *p;
606 size_t len;
607
608 len = sizeof(FTSENT) + namelen;
609 if ((p = calloc(1, len)) == NULL)
610 return (NULL);
611
612 p->fts_path = sp->fts_path;
613 p->fts_namelen = namelen;
614 p->fts_instr = FTS_NOINSTR;
615 p->fts_statp = malloc(sizeof(struct stat));
616 if (p->fts_statp == NULL) {
617 free(p);
618 return (NULL);
619 }
620 memcpy(p->fts_name, name, namelen);
621
622 return (p);
623 }
624
625 static void
626 fts_lfree(FTSENT *head)
627 {
628 FTSENT *p;
629
630 /* Free a linked list of structures. */
631 while ((p = head)) {
632 head = head->fts_link;
633 free(p);
634 }
635 }
636
637 /*
638 * Allow essentially unlimited paths; find, rm, ls should all work on any tree.
639 * Most systems will allow creation of paths much longer than PATH_MAX, even
640 * though the kernel won't resolve them. Add the size (not just what's needed)
641 * plus 256 bytes so don't realloc the path 2 bytes at a time.
642 */
643 static int
644 fts_palloc(FTS *sp, size_t more)
645 {
646 char *p;
647
648 /*
649 * Check for possible wraparound.
650 */
651 more += 256;
652 if (sp->fts_pathlen + more < sp->fts_pathlen) {
653 free(sp->fts_path);
654 sp->fts_path = NULL;
655 errno = ENAMETOOLONG;
656 return (1);
657 }
658 sp->fts_pathlen += more;
659 p = realloc(sp->fts_path, sp->fts_pathlen);
660 if (p == NULL) {
661 free(sp->fts_path);
662 sp->fts_path = NULL;
663 return (1);
664 }
665 sp->fts_path = p;
666 return (0);
667 }
668
669 /*
670 * When the path is realloc'd, have to fix all of the pointers in structures
671 * already returned.
672 */
673 static void
674 fts_padjust(FTS *sp, FTSENT *head)
675 {
676 FTSENT *p;
677 char *addr = sp->fts_path;
678
679 #define ADJUST(p) { \
680 if ((p)->fts_accpath != (p)->fts_name) { \
681 (p)->fts_accpath = \
682 (char *)addr + ((p)->fts_accpath - (p)->fts_path); \
683 } \
684 (p)->fts_path = addr; \
685 }
686 /* Adjust the current set of children. */
687 for (p = sp->fts_child; p; p = p->fts_link)
688 ADJUST(p);
689
690 /* Adjust the rest of the tree, including the current level. */
691 for (p = head; p->fts_level >= FTS_ROOTLEVEL;) {
692 ADJUST(p);
693 p = p->fts_link ? p->fts_link : p->fts_parent;
694 }
695 }
696
697 static size_t
698 fts_maxarglen(char * const *argv)
699 {
700 size_t len, max;
701
702 for (max = 0; *argv; ++argv)
703 if ((len = strlen(*argv)) > max)
704 max = len;
705 return (max + 1);
706 }
707
708 #endif