]> git.cameronkatri.com Git - mandoc.git/blob - mandocdb.c
git.ckatri.com / mandoc.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
The files mandoc.c and mandoc.h contained both specialised low-level
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.119 2014/03/23 11:25:26 schwarze Exp $ */
2 /*
3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011, 2012, 2013, 2014 Ingo Schwarze <schwarze@openbsd.org>
5 *
6 * Permission to use, copy, modify, and distribute this software for any
7 * purpose with or without fee is hereby granted, provided that the above
8 * copyright notice and this permission notice appear in all copies.
9 *
10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17 */
18 #ifdef HAVE_CONFIG_H
19 #include "config.h"
20 #endif
21
22 #include <sys/stat.h>
23 #include <sys/wait.h>
24
25 #include <assert.h>
26 #include <ctype.h>
27 #include <errno.h>
28 #include <fcntl.h>
29 #include <fts.h>
30 #include <getopt.h>
31 #include <limits.h>
32 #include <stddef.h>
33 #include <stdio.h>
34 #include <stdint.h>
35 #include <stdlib.h>
36 #include <string.h>
37 #include <unistd.h>
38
39 #ifdef HAVE_OHASH
40 #include <ohash.h>
41 #else
42 #include "compat_ohash.h"
43 #endif
44 #include <sqlite3.h>
45
46 #include "mdoc.h"
47 #include "man.h"
48 #include "mandoc.h"
49 #include "mandoc_aux.h"
50 #include "manpath.h"
51 #include "mansearch.h"
52
53 extern int mansearch_keymax;
54 extern const char *const mansearch_keynames[];
55
56 #define SQL_EXEC(_v) \
57 if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
58 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
59 #define SQL_BIND_TEXT(_s, _i, _v) \
60 if (SQLITE_OK != sqlite3_bind_text \
61 ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
62 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
63 #define SQL_BIND_INT(_s, _i, _v) \
64 if (SQLITE_OK != sqlite3_bind_int \
65 ((_s), (_i)++, (_v))) \
66 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
67 #define SQL_BIND_INT64(_s, _i, _v) \
68 if (SQLITE_OK != sqlite3_bind_int64 \
69 ((_s), (_i)++, (_v))) \
70 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
71 #define SQL_STEP(_s) \
72 if (SQLITE_DONE != sqlite3_step((_s))) \
73 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
74
75 enum op {
76 OP_DEFAULT = 0, /* new dbs from dir list or default config */
77 OP_CONFFILE, /* new databases from custom config file */
78 OP_UPDATE, /* delete/add entries in existing database */
79 OP_DELETE, /* delete entries from existing database */
80 OP_TEST /* change no databases, report potential problems */
81 };
82
83 enum form {
84 FORM_NONE, /* format is unknown */
85 FORM_SRC, /* format is -man or -mdoc */
86 FORM_CAT /* format is cat */
87 };
88
89 struct str {
90 char *rendered; /* key in UTF-8 or ASCII form */
91 const struct mpage *mpage; /* if set, the owning parse */
92 uint64_t mask; /* bitmask in sequence */
93 char key[]; /* may contain escape sequences */
94 };
95
96 struct inodev {
97 ino_t st_ino;
98 dev_t st_dev;
99 };
100
101 struct mpage {
102 struct inodev inodev; /* used for hashing routine */
103 int64_t recno; /* id in mpages SQL table */
104 enum form form; /* format from file content */
105 char *sec; /* section from file content */
106 char *arch; /* architecture from file content */
107 char *title; /* title from file content */
108 char *desc; /* description from file content */
109 struct mlink *mlinks; /* singly linked list */
110 };
111
112 struct mlink {
113 char file[PATH_MAX]; /* filename rel. to manpath */
114 enum form dform; /* format from directory */
115 enum form fform; /* format from file name suffix */
116 char *dsec; /* section from directory */
117 char *arch; /* architecture from directory */
118 char *name; /* name from file name (not empty) */
119 char *fsec; /* section from file name suffix */
120 struct mlink *next; /* singly linked list */
121 struct mpage *mpage; /* parent */
122 };
123
124 enum stmt {
125 STMT_DELETE_PAGE = 0, /* delete mpage */
126 STMT_INSERT_PAGE, /* insert mpage */
127 STMT_INSERT_LINK, /* insert mlink */
128 STMT_INSERT_KEY, /* insert parsed key */
129 STMT__MAX
130 };
131
132 typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
133
134 struct mdoc_handler {
135 mdoc_fp fp; /* optional handler */
136 uint64_t mask; /* set unless handler returns 0 */
137 };
138
139 static void dbclose(int);
140 static void dbadd(struct mpage *, struct mchars *);
141 static void dbadd_mlink(const struct mlink *mlink);
142 static int dbopen(int);
143 static void dbprune(void);
144 static void filescan(const char *);
145 static void *hash_alloc(size_t, void *);
146 static void hash_free(void *, size_t, void *);
147 static void *hash_halloc(size_t, void *);
148 static void mlink_add(struct mlink *, const struct stat *);
149 static int mlink_check(struct mpage *, struct mlink *);
150 static void mlink_free(struct mlink *);
151 static void mlinks_undupe(struct mpage *);
152 static void mpages_free(void);
153 static void mpages_merge(struct mchars *, struct mparse *);
154 static void parse_cat(struct mpage *);
155 static void parse_man(struct mpage *, const struct man_node *);
156 static void parse_mdoc(struct mpage *, const struct mdoc_node *);
157 static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
158 static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
159 static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
160 static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
161 static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
162 static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
163 static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
164 static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
165 static void putkey(const struct mpage *, char *, uint64_t);
166 static void putkeys(const struct mpage *,
167 const char *, size_t, uint64_t);
168 static void putmdockey(const struct mpage *,
169 const struct mdoc_node *, uint64_t);
170 static void render_key(struct mchars *, struct str *);
171 static void say(const char *, const char *, ...);
172 static int set_basedir(const char *);
173 static int treescan(void);
174 static size_t utf8(unsigned int, char [7]);
175
176 static char tempfilename[32];
177 static char *progname;
178 static int nodb; /* no database changes */
179 static int mparse_options; /* abort the parse early */
180 static int use_all; /* use all found files */
181 static int verb; /* print what we're doing */
182 static int warnings; /* warn about crap */
183 static int write_utf8; /* write UTF-8 output; else ASCII */
184 static int exitcode; /* to be returned by main */
185 static enum op op; /* operational mode */
186 static char basedir[PATH_MAX]; /* current base directory */
187 static struct ohash mpages; /* table of distinct manual pages */
188 static struct ohash mlinks; /* table of directory entries */
189 static struct ohash strings; /* table of all strings */
190 static sqlite3 *db = NULL; /* current database */
191 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
192
193 static const struct mdoc_handler mdocs[MDOC_MAX] = {
194 { NULL, 0 }, /* Ap */
195 { NULL, 0 }, /* Dd */
196 { NULL, 0 }, /* Dt */
197 { NULL, 0 }, /* Os */
198 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
199 { parse_mdoc_head, TYPE_Ss }, /* Ss */
200 { NULL, 0 }, /* Pp */
201 { NULL, 0 }, /* D1 */
202 { NULL, 0 }, /* Dl */
203 { NULL, 0 }, /* Bd */
204 { NULL, 0 }, /* Ed */
205 { NULL, 0 }, /* Bl */
206 { NULL, 0 }, /* El */
207 { NULL, 0 }, /* It */
208 { NULL, 0 }, /* Ad */
209 { NULL, TYPE_An }, /* An */
210 { NULL, TYPE_Ar }, /* Ar */
211 { NULL, TYPE_Cd }, /* Cd */
212 { NULL, TYPE_Cm }, /* Cm */
213 { NULL, TYPE_Dv }, /* Dv */
214 { NULL, TYPE_Er }, /* Er */
215 { NULL, TYPE_Ev }, /* Ev */
216 { NULL, 0 }, /* Ex */
217 { NULL, TYPE_Fa }, /* Fa */
218 { parse_mdoc_Fd, 0 }, /* Fd */
219 { NULL, TYPE_Fl }, /* Fl */
220 { parse_mdoc_Fn, 0 }, /* Fn */
221 { NULL, TYPE_Ft }, /* Ft */
222 { NULL, TYPE_Ic }, /* Ic */
223 { NULL, TYPE_In }, /* In */
224 { NULL, TYPE_Li }, /* Li */
225 { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
226 { parse_mdoc_Nm, TYPE_Nm }, /* Nm */
227 { NULL, 0 }, /* Op */
228 { NULL, 0 }, /* Ot */
229 { NULL, TYPE_Pa }, /* Pa */
230 { NULL, 0 }, /* Rv */
231 { NULL, TYPE_St }, /* St */
232 { NULL, TYPE_Va }, /* Va */
233 { parse_mdoc_body, TYPE_Va }, /* Vt */
234 { parse_mdoc_Xr, 0 }, /* Xr */
235 { NULL, 0 }, /* %A */
236 { NULL, 0 }, /* %B */
237 { NULL, 0 }, /* %D */
238 { NULL, 0 }, /* %I */
239 { NULL, 0 }, /* %J */
240 { NULL, 0 }, /* %N */
241 { NULL, 0 }, /* %O */
242 { NULL, 0 }, /* %P */
243 { NULL, 0 }, /* %R */
244 { NULL, 0 }, /* %T */
245 { NULL, 0 }, /* %V */
246 { NULL, 0 }, /* Ac */
247 { NULL, 0 }, /* Ao */
248 { NULL, 0 }, /* Aq */
249 { NULL, TYPE_At }, /* At */
250 { NULL, 0 }, /* Bc */
251 { NULL, 0 }, /* Bf */
252 { NULL, 0 }, /* Bo */
253 { NULL, 0 }, /* Bq */
254 { NULL, TYPE_Bsx }, /* Bsx */
255 { NULL, TYPE_Bx }, /* Bx */
256 { NULL, 0 }, /* Db */
257 { NULL, 0 }, /* Dc */
258 { NULL, 0 }, /* Do */
259 { NULL, 0 }, /* Dq */
260 { NULL, 0 }, /* Ec */
261 { NULL, 0 }, /* Ef */
262 { NULL, TYPE_Em }, /* Em */
263 { NULL, 0 }, /* Eo */
264 { NULL, TYPE_Fx }, /* Fx */
265 { NULL, TYPE_Ms }, /* Ms */
266 { NULL, 0 }, /* No */
267 { NULL, 0 }, /* Ns */
268 { NULL, TYPE_Nx }, /* Nx */
269 { NULL, TYPE_Ox }, /* Ox */
270 { NULL, 0 }, /* Pc */
271 { NULL, 0 }, /* Pf */
272 { NULL, 0 }, /* Po */
273 { NULL, 0 }, /* Pq */
274 { NULL, 0 }, /* Qc */
275 { NULL, 0 }, /* Ql */
276 { NULL, 0 }, /* Qo */
277 { NULL, 0 }, /* Qq */
278 { NULL, 0 }, /* Re */
279 { NULL, 0 }, /* Rs */
280 { NULL, 0 }, /* Sc */
281 { NULL, 0 }, /* So */
282 { NULL, 0 }, /* Sq */
283 { NULL, 0 }, /* Sm */
284 { NULL, 0 }, /* Sx */
285 { NULL, TYPE_Sy }, /* Sy */
286 { NULL, TYPE_Tn }, /* Tn */
287 { NULL, 0 }, /* Ux */
288 { NULL, 0 }, /* Xc */
289 { NULL, 0 }, /* Xo */
290 { parse_mdoc_head, 0 }, /* Fo */
291 { NULL, 0 }, /* Fc */
292 { NULL, 0 }, /* Oo */
293 { NULL, 0 }, /* Oc */
294 { NULL, 0 }, /* Bk */
295 { NULL, 0 }, /* Ek */
296 { NULL, 0 }, /* Bt */
297 { NULL, 0 }, /* Hf */
298 { NULL, 0 }, /* Fr */
299 { NULL, 0 }, /* Ud */
300 { NULL, TYPE_Lb }, /* Lb */
301 { NULL, 0 }, /* Lp */
302 { NULL, TYPE_Lk }, /* Lk */
303 { NULL, TYPE_Mt }, /* Mt */
304 { NULL, 0 }, /* Brq */
305 { NULL, 0 }, /* Bro */
306 { NULL, 0 }, /* Brc */
307 { NULL, 0 }, /* %C */
308 { NULL, 0 }, /* Es */
309 { NULL, 0 }, /* En */
310 { NULL, TYPE_Dx }, /* Dx */
311 { NULL, 0 }, /* %Q */
312 { NULL, 0 }, /* br */
313 { NULL, 0 }, /* sp */
314 { NULL, 0 }, /* %U */
315 { NULL, 0 }, /* Ta */
316 };
317
318 int
319 main(int argc, char *argv[])
320 {
321 int ch, i;
322 size_t j, sz;
323 const char *path_arg;
324 struct mchars *mc;
325 struct manpaths dirs;
326 struct mparse *mp;
327 struct ohash_info mpages_info, mlinks_info;
328
329 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
330 memset(&dirs, 0, sizeof(struct manpaths));
331
332 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
333 mpages_info.halloc = mlinks_info.halloc = hash_halloc;
334 mpages_info.hfree = mlinks_info.hfree = hash_free;
335
336 mpages_info.key_offset = offsetof(struct mpage, inodev);
337 mlinks_info.key_offset = offsetof(struct mlink, file);
338
339 progname = strrchr(argv[0], '/');
340 if (progname == NULL)
341 progname = argv[0];
342 else
343 ++progname;
344
345 /*
346 * We accept a few different invocations.
347 * The CHECKOP macro makes sure that invocation styles don't
348 * clobber each other.
349 */
350 #define CHECKOP(_op, _ch) do \
351 if (OP_DEFAULT != (_op)) { \
352 fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
353 goto usage; \
354 } while (/*CONSTCOND*/0)
355
356 path_arg = NULL;
357 op = OP_DEFAULT;
358
359 while (-1 != (ch = getopt(argc, argv, "aC:d:nQT:tu:vW")))
360 switch (ch) {
361 case ('a'):
362 use_all = 1;
363 break;
364 case ('C'):
365 CHECKOP(op, ch);
366 path_arg = optarg;
367 op = OP_CONFFILE;
368 break;
369 case ('d'):
370 CHECKOP(op, ch);
371 path_arg = optarg;
372 op = OP_UPDATE;
373 break;
374 case ('n'):
375 nodb = 1;
376 break;
377 case ('Q'):
378 mparse_options |= MPARSE_QUICK;
379 break;
380 case ('T'):
381 if (strcmp(optarg, "utf8")) {
382 fprintf(stderr, "-T%s: Unsupported "
383 "output format\n", optarg);
384 goto usage;
385 }
386 write_utf8 = 1;
387 break;
388 case ('t'):
389 CHECKOP(op, ch);
390 dup2(STDOUT_FILENO, STDERR_FILENO);
391 op = OP_TEST;
392 nodb = warnings = 1;
393 break;
394 case ('u'):
395 CHECKOP(op, ch);
396 path_arg = optarg;
397 op = OP_DELETE;
398 break;
399 case ('v'):
400 verb++;
401 break;
402 case ('W'):
403 warnings = 1;
404 break;
405 default:
406 goto usage;
407 }
408
409 argc -= optind;
410 argv += optind;
411
412 if (OP_CONFFILE == op && argc > 0) {
413 fprintf(stderr, "-C: Too many arguments\n");
414 goto usage;
415 }
416
417 exitcode = (int)MANDOCLEVEL_OK;
418 mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL, NULL);
419 mc = mchars_alloc();
420
421 ohash_init(&mpages, 6, &mpages_info);
422 ohash_init(&mlinks, 6, &mlinks_info);
423
424 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
425 /*
426 * Force processing all files.
427 */
428 use_all = 1;
429
430 /*
431 * All of these deal with a specific directory.
432 * Jump into that directory then collect files specified
433 * on the command-line.
434 */
435 if (0 == set_basedir(path_arg))
436 goto out;
437 for (i = 0; i < argc; i++)
438 filescan(argv[i]);
439 if (0 == dbopen(1))
440 goto out;
441 if (OP_TEST != op)
442 dbprune();
443 if (OP_DELETE != op)
444 mpages_merge(mc, mp);
445 dbclose(1);
446 } else {
447 /*
448 * If we have arguments, use them as our manpaths.
449 * If we don't, grok from manpath(1) or however else
450 * manpath_parse() wants to do it.
451 */
452 if (argc > 0) {
453 dirs.paths = mandoc_calloc
454 (argc, sizeof(char *));
455 dirs.sz = (size_t)argc;
456 for (i = 0; i < argc; i++)
457 dirs.paths[i] = mandoc_strdup(argv[i]);
458 } else
459 manpath_parse(&dirs, path_arg, NULL, NULL);
460
461 /*
462 * First scan the tree rooted at a base directory, then
463 * build a new database and finally move it into place.
464 * Ignore zero-length directories and strip trailing
465 * slashes.
466 */
467 for (j = 0; j < dirs.sz; j++) {
468 sz = strlen(dirs.paths[j]);
469 if (sz && '/' == dirs.paths[j][sz - 1])
470 dirs.paths[j][--sz] = '\0';
471 if (0 == sz)
472 continue;
473
474 if (j) {
475 ohash_init(&mpages, 6, &mpages_info);
476 ohash_init(&mlinks, 6, &mlinks_info);
477 }
478
479 if (0 == set_basedir(dirs.paths[j]))
480 goto out;
481 if (0 == treescan())
482 goto out;
483 if (0 == set_basedir(dirs.paths[j]))
484 goto out;
485 if (0 == dbopen(0))
486 goto out;
487
488 mpages_merge(mc, mp);
489 dbclose(0);
490
491 if (j + 1 < dirs.sz) {
492 mpages_free();
493 ohash_delete(&mpages);
494 ohash_delete(&mlinks);
495 }
496 }
497 }
498 out:
499 set_basedir(NULL);
500 manpath_free(&dirs);
501 mchars_free(mc);
502 mparse_free(mp);
503 mpages_free();
504 ohash_delete(&mpages);
505 ohash_delete(&mlinks);
506 return(exitcode);
507 usage:
508 fprintf(stderr, "usage: %s [-anQvW] [-C file] [-Tutf8]\n"
509 " %s [-anQvW] [-Tutf8] dir ...\n"
510 " %s [-nQvW] [-Tutf8] -d dir [file ...]\n"
511 " %s [-nvW] -u dir [file ...]\n"
512 " %s [-Q] -t file ...\n",
513 progname, progname, progname,
514 progname, progname);
515
516 return((int)MANDOCLEVEL_BADARG);
517 }
518
519 /*
520 * Scan a directory tree rooted at "basedir" for manpages.
521 * We use fts(), scanning directory parts along the way for clues to our
522 * section and architecture.
523 *
524 * If use_all has been specified, grok all files.
525 * If not, sanitise paths to the following:
526 *
527 * [./]man*[/<arch>]/<name>.<section>
528 * or
529 * [./]cat<section>[/<arch>]/<name>.0
530 *
531 * TODO: accomodate for multi-language directories.
532 */
533 static int
534 treescan(void)
535 {
536 FTS *f;
537 FTSENT *ff;
538 struct mlink *mlink;
539 int dform;
540 char *dsec, *arch, *fsec, *cp;
541 const char *path;
542 const char *argv[2];
543
544 argv[0] = ".";
545 argv[1] = (char *)NULL;
546
547 /*
548 * Walk through all components under the directory, using the
549 * logical descent of files.
550 */
551 f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
552 if (NULL == f) {
553 exitcode = (int)MANDOCLEVEL_SYSERR;
554 say("", NULL);
555 return(0);
556 }
557
558 dsec = arch = NULL;
559 dform = FORM_NONE;
560
561 while (NULL != (ff = fts_read(f))) {
562 path = ff->fts_path + 2;
563 /*
564 * If we're a regular file, add an mlink by using the
565 * stored directory data and handling the filename.
566 */
567 if (FTS_F == ff->fts_info) {
568 if (0 == strcmp(path, MANDOC_DB))
569 continue;
570 if ( ! use_all && ff->fts_level < 2) {
571 if (warnings)
572 say(path, "Extraneous file");
573 continue;
574 } else if (NULL == (fsec =
575 strrchr(ff->fts_name, '.'))) {
576 if ( ! use_all) {
577 if (warnings)
578 say(path,
579 "No filename suffix");
580 continue;
581 }
582 } else if (0 == strcmp(++fsec, "html")) {
583 if (warnings)
584 say(path, "Skip html");
585 continue;
586 } else if (0 == strcmp(fsec, "gz")) {
587 if (warnings)
588 say(path, "Skip gz");
589 continue;
590 } else if (0 == strcmp(fsec, "ps")) {
591 if (warnings)
592 say(path, "Skip ps");
593 continue;
594 } else if (0 == strcmp(fsec, "pdf")) {
595 if (warnings)
596 say(path, "Skip pdf");
597 continue;
598 } else if ( ! use_all &&
599 ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
600 (FORM_CAT == dform && strcmp(fsec, "0")))) {
601 if (warnings)
602 say(path, "Wrong filename suffix");
603 continue;
604 } else
605 fsec[-1] = '\0';
606
607 mlink = mandoc_calloc(1, sizeof(struct mlink));
608 strlcpy(mlink->file, path, sizeof(mlink->file));
609 mlink->dform = dform;
610 mlink->dsec = dsec;
611 mlink->arch = arch;
612 mlink->name = ff->fts_name;
613 mlink->fsec = fsec;
614 mlink_add(mlink, ff->fts_statp);
615 continue;
616 } else if (FTS_D != ff->fts_info &&
617 FTS_DP != ff->fts_info) {
618 if (warnings)
619 say(path, "Not a regular file");
620 continue;
621 }
622
623 switch (ff->fts_level) {
624 case (0):
625 /* Ignore the root directory. */
626 break;
627 case (1):
628 /*
629 * This might contain manX/ or catX/.
630 * Try to infer this from the name.
631 * If we're not in use_all, enforce it.
632 */
633 cp = ff->fts_name;
634 if (FTS_DP == ff->fts_info)
635 break;
636
637 if (0 == strncmp(cp, "man", 3)) {
638 dform = FORM_SRC;
639 dsec = cp + 3;
640 } else if (0 == strncmp(cp, "cat", 3)) {
641 dform = FORM_CAT;
642 dsec = cp + 3;
643 } else {
644 dform = FORM_NONE;
645 dsec = NULL;
646 }
647
648 if (NULL != dsec || use_all)
649 break;
650
651 if (warnings)
652 say(path, "Unknown directory part");
653 fts_set(f, ff, FTS_SKIP);
654 break;
655 case (2):
656 /*
657 * Possibly our architecture.
658 * If we're descending, keep tabs on it.
659 */
660 if (FTS_DP != ff->fts_info && NULL != dsec)
661 arch = ff->fts_name;
662 else
663 arch = NULL;
664 break;
665 default:
666 if (FTS_DP == ff->fts_info || use_all)
667 break;
668 if (warnings)
669 say(path, "Extraneous directory part");
670 fts_set(f, ff, FTS_SKIP);
671 break;
672 }
673 }
674
675 fts_close(f);
676 return(1);
677 }
678
679 /*
680 * Add a file to the mlinks table.
681 * Do not verify that it's a "valid" looking manpage (we'll do that
682 * later).
683 *
684 * Try to infer the manual section, architecture, and page name from the
685 * path, assuming it looks like
686 *
687 * [./]man*[/<arch>]/<name>.<section>
688 * or
689 * [./]cat<section>[/<arch>]/<name>.0
690 *
691 * See treescan() for the fts(3) version of this.
692 */
693 static void
694 filescan(const char *file)
695 {
696 char buf[PATH_MAX];
697 struct stat st;
698 struct mlink *mlink;
699 char *p, *start;
700
701 assert(use_all);
702
703 if (0 == strncmp(file, "./", 2))
704 file += 2;
705
706 if (NULL == realpath(file, buf)) {
707 exitcode = (int)MANDOCLEVEL_BADARG;
708 say(file, NULL);
709 return;
710 }
711
712 if (strstr(buf, basedir) == buf)
713 start = buf + strlen(basedir) + 1;
714 else if (OP_TEST == op)
715 start = buf;
716 else {
717 exitcode = (int)MANDOCLEVEL_BADARG;
718 say("", "%s: outside base directory", buf);
719 return;
720 }
721
722 if (-1 == stat(buf, &st)) {
723 exitcode = (int)MANDOCLEVEL_BADARG;
724 say(file, NULL);
725 return;
726 } else if ( ! (S_IFREG & st.st_mode)) {
727 exitcode = (int)MANDOCLEVEL_BADARG;
728 say(file, "Not a regular file");
729 return;
730 }
731
732 mlink = mandoc_calloc(1, sizeof(struct mlink));
733 strlcpy(mlink->file, start, sizeof(mlink->file));
734
735 /*
736 * First try to guess our directory structure.
737 * If we find a separator, try to look for man* or cat*.
738 * If we find one of these and what's underneath is a directory,
739 * assume it's an architecture.
740 */
741 if (NULL != (p = strchr(start, '/'))) {
742 *p++ = '\0';
743 if (0 == strncmp(start, "man", 3)) {
744 mlink->dform = FORM_SRC;
745 mlink->dsec = start + 3;
746 } else if (0 == strncmp(start, "cat", 3)) {
747 mlink->dform = FORM_CAT;
748 mlink->dsec = start + 3;
749 }
750
751 start = p;
752 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
753 *p++ = '\0';
754 mlink->arch = start;
755 start = p;
756 }
757 }
758
759 /*
760 * Now check the file suffix.
761 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
762 */
763 p = strrchr(start, '\0');
764 while (p-- > start && '/' != *p && '.' != *p)
765 /* Loop. */ ;
766
767 if ('.' == *p) {
768 *p++ = '\0';
769 mlink->fsec = p;
770 }
771
772 /*
773 * Now try to parse the name.
774 * Use the filename portion of the path.
775 */
776 mlink->name = start;
777 if (NULL != (p = strrchr(start, '/'))) {
778 mlink->name = p + 1;
779 *p = '\0';
780 }
781 mlink_add(mlink, &st);
782 }
783
784 static void
785 mlink_add(struct mlink *mlink, const struct stat *st)
786 {
787 struct inodev inodev;
788 struct mpage *mpage;
789 unsigned int slot;
790
791 assert(NULL != mlink->file);
792
793 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
794 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
795 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
796 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
797
798 if ('0' == *mlink->fsec) {
799 free(mlink->fsec);
800 mlink->fsec = mandoc_strdup(mlink->dsec);
801 mlink->fform = FORM_CAT;
802 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
803 mlink->fform = FORM_SRC;
804 else
805 mlink->fform = FORM_NONE;
806
807 slot = ohash_qlookup(&mlinks, mlink->file);
808 assert(NULL == ohash_find(&mlinks, slot));
809 ohash_insert(&mlinks, slot, mlink);
810
811 inodev.st_ino = st->st_ino;
812 inodev.st_dev = st->st_dev;
813 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
814 sizeof(struct inodev), inodev.st_ino);
815 mpage = ohash_find(&mpages, slot);
816 if (NULL == mpage) {
817 mpage = mandoc_calloc(1, sizeof(struct mpage));
818 mpage->inodev.st_ino = inodev.st_ino;
819 mpage->inodev.st_dev = inodev.st_dev;
820 ohash_insert(&mpages, slot, mpage);
821 } else
822 mlink->next = mpage->mlinks;
823 mpage->mlinks = mlink;
824 mlink->mpage = mpage;
825 }
826
827 static void
828 mlink_free(struct mlink *mlink)
829 {
830
831 free(mlink->dsec);
832 free(mlink->arch);
833 free(mlink->name);
834 free(mlink->fsec);
835 free(mlink);
836 }
837
838 static void
839 mpages_free(void)
840 {
841 struct mpage *mpage;
842 struct mlink *mlink;
843 unsigned int slot;
844
845 mpage = ohash_first(&mpages, &slot);
846 while (NULL != mpage) {
847 while (NULL != (mlink = mpage->mlinks)) {
848 mpage->mlinks = mlink->next;
849 mlink_free(mlink);
850 }
851 free(mpage->sec);
852 free(mpage->arch);
853 free(mpage->title);
854 free(mpage->desc);
855 free(mpage);
856 mpage = ohash_next(&mpages, &slot);
857 }
858 }
859
860 /*
861 * For each mlink to the mpage, check whether the path looks like
862 * it is formatted, and if it does, check whether a source manual
863 * exists by the same name, ignoring the suffix.
864 * If both conditions hold, drop the mlink.
865 */
866 static void
867 mlinks_undupe(struct mpage *mpage)
868 {
869 char buf[PATH_MAX];
870 struct mlink **prev;
871 struct mlink *mlink;
872 char *bufp;
873
874 mpage->form = FORM_CAT;
875 prev = &mpage->mlinks;
876 while (NULL != (mlink = *prev)) {
877 if (FORM_CAT != mlink->dform) {
878 mpage->form = FORM_NONE;
879 goto nextlink;
880 }
881 if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) {
882 if (warnings)
883 say(mlink->file, "Filename too long");
884 goto nextlink;
885 }
886 bufp = strstr(buf, "cat");
887 assert(NULL != bufp);
888 memcpy(bufp, "man", 3);
889 if (NULL != (bufp = strrchr(buf, '.')))
890 *++bufp = '\0';
891 strlcat(buf, mlink->dsec, PATH_MAX);
892 if (NULL == ohash_find(&mlinks,
893 ohash_qlookup(&mlinks, buf)))
894 goto nextlink;
895 if (warnings)
896 say(mlink->file, "Man source exists: %s", buf);
897 if (use_all)
898 goto nextlink;
899 *prev = mlink->next;
900 mlink_free(mlink);
901 continue;
902 nextlink:
903 prev = &(*prev)->next;
904 }
905 }
906
907 static int
908 mlink_check(struct mpage *mpage, struct mlink *mlink)
909 {
910 int match;
911
912 match = 1;
913
914 /*
915 * Check whether the manual section given in a file
916 * agrees with the directory where the file is located.
917 * Some manuals have suffixes like (3p) on their
918 * section number either inside the file or in the
919 * directory name, some are linked into more than one
920 * section, like encrypt(1) = makekey(8).
921 */
922
923 if (FORM_SRC == mpage->form &&
924 strcasecmp(mpage->sec, mlink->dsec)) {
925 match = 0;
926 say(mlink->file, "Section \"%s\" manual in %s directory",
927 mpage->sec, mlink->dsec);
928 }
929
930 /*
931 * Manual page directories exist for each kernel
932 * architecture as returned by machine(1).
933 * However, many manuals only depend on the
934 * application architecture as returned by arch(1).
935 * For example, some (2/ARM) manuals are shared
936 * across the "armish" and "zaurus" kernel
937 * architectures.
938 * A few manuals are even shared across completely
939 * different architectures, for example fdformat(1)
940 * on amd64, i386, sparc, and sparc64.
941 */
942
943 if (strcasecmp(mpage->arch, mlink->arch)) {
944 match = 0;
945 say(mlink->file, "Architecture \"%s\" manual in "
946 "\"%s\" directory", mpage->arch, mlink->arch);
947 }
948
949 if (strcasecmp(mpage->title, mlink->name))
950 match = 0;
951
952 return(match);
953 }
954
955 /*
956 * Run through the files in the global vector "mpages"
957 * and add them to the database specified in "basedir".
958 *
959 * This handles the parsing scheme itself, using the cues of directory
960 * and filename to determine whether the file is parsable or not.
961 */
962 static void
963 mpages_merge(struct mchars *mc, struct mparse *mp)
964 {
965 char any[] = "any";
966 struct ohash_info str_info;
967 struct mpage *mpage, *mpage_dest;
968 struct mlink *mlink, *mlink_dest;
969 struct mdoc *mdoc;
970 struct man *man;
971 char *sodest;
972 char *cp;
973 int match;
974 unsigned int pslot;
975 enum mandoclevel lvl;
976
977 str_info.alloc = hash_alloc;
978 str_info.halloc = hash_halloc;
979 str_info.hfree = hash_free;
980 str_info.key_offset = offsetof(struct str, key);
981
982 if (0 == nodb)
983 SQL_EXEC("BEGIN TRANSACTION");
984
985 mpage = ohash_first(&mpages, &pslot);
986 while (NULL != mpage) {
987 mlinks_undupe(mpage);
988 if (NULL == mpage->mlinks) {
989 mpage = ohash_next(&mpages, &pslot);
990 continue;
991 }
992
993 ohash_init(&strings, 6, &str_info);
994 mparse_reset(mp);
995 mdoc = NULL;
996 man = NULL;
997
998 /*
999 * Try interpreting the file as mdoc(7) or man(7)
1000 * source code, unless it is already known to be
1001 * formatted. Fall back to formatted mode.
1002 */
1003 if (FORM_CAT != mpage->mlinks->dform ||
1004 FORM_CAT != mpage->mlinks->fform) {
1005 lvl = mparse_readfd(mp, -1, mpage->mlinks->file);
1006 if (lvl < MANDOCLEVEL_FATAL)
1007 mparse_result(mp, &mdoc, &man, &sodest);
1008 }
1009
1010 if (NULL != sodest) {
1011 mlink_dest = ohash_find(&mlinks,
1012 ohash_qlookup(&mlinks, sodest));
1013 if (NULL != mlink_dest) {
1014
1015 /* The .so target exists. */
1016
1017 mpage_dest = mlink_dest->mpage;
1018 mlink = mpage->mlinks;
1019 while (1) {
1020 mlink->mpage = mpage_dest;
1021
1022 /*
1023 * If the target was already
1024 * processed, add the links
1025 * to the database now.
1026 * Otherwise, this will
1027 * happen when we come
1028 * to the target.
1029 */
1030
1031 if (mpage_dest->recno)
1032 dbadd_mlink(mlink);
1033
1034 if (NULL == mlink->next)
1035 break;
1036 mlink = mlink->next;
1037 }
1038
1039 /* Move all links to the target. */
1040
1041 mlink->next = mlink_dest->next;
1042 mlink_dest->next = mpage->mlinks;
1043 mpage->mlinks = NULL;
1044 }
1045 ohash_delete(&strings);
1046 mpage = ohash_next(&mpages, &pslot);
1047 continue;
1048 } else if (NULL != mdoc) {
1049 mpage->form = FORM_SRC;
1050 mpage->sec =
1051 mandoc_strdup(mdoc_meta(mdoc)->msec);
1052 mpage->arch = mdoc_meta(mdoc)->arch;
1053 mpage->arch = mandoc_strdup(
1054 NULL == mpage->arch ? "" : mpage->arch);
1055 mpage->title =
1056 mandoc_strdup(mdoc_meta(mdoc)->title);
1057 } else if (NULL != man) {
1058 mpage->form = FORM_SRC;
1059 mpage->sec =
1060 mandoc_strdup(man_meta(man)->msec);
1061 mpage->arch =
1062 mandoc_strdup(mpage->mlinks->arch);
1063 mpage->title =
1064 mandoc_strdup(man_meta(man)->title);
1065 } else {
1066 mpage->form = FORM_CAT;
1067 mpage->sec =
1068 mandoc_strdup(mpage->mlinks->dsec);
1069 mpage->arch =
1070 mandoc_strdup(mpage->mlinks->arch);
1071 mpage->title =
1072 mandoc_strdup(mpage->mlinks->name);
1073 }
1074 putkey(mpage, mpage->sec, TYPE_sec);
1075 putkey(mpage, '\0' == *mpage->arch ?
1076 any : mpage->arch, TYPE_arch);
1077
1078 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1079 if ('\0' != *mlink->dsec)
1080 putkey(mpage, mlink->dsec, TYPE_sec);
1081 if ('\0' != *mlink->fsec)
1082 putkey(mpage, mlink->fsec, TYPE_sec);
1083 putkey(mpage, '\0' == *mlink->arch ?
1084 any : mlink->arch, TYPE_arch);
1085 putkey(mpage, mlink->name, TYPE_Nm);
1086 }
1087
1088 if (warnings && !use_all) {
1089 match = 0;
1090 for (mlink = mpage->mlinks; mlink;
1091 mlink = mlink->next)
1092 if (mlink_check(mpage, mlink))
1093 match = 1;
1094 } else
1095 match = 1;
1096
1097 if (NULL != mdoc) {
1098 if (NULL != (cp = mdoc_meta(mdoc)->name))
1099 putkey(mpage, cp, TYPE_Nm);
1100 assert(NULL == mpage->desc);
1101 parse_mdoc(mpage, mdoc_node(mdoc));
1102 putkey(mpage, NULL != mpage->desc ?
1103 mpage->desc : mpage->mlinks->name, TYPE_Nd);
1104 } else if (NULL != man)
1105 parse_man(mpage, man_node(man));
1106 else
1107 parse_cat(mpage);
1108
1109 dbadd(mpage, mc);
1110 ohash_delete(&strings);
1111 mpage = ohash_next(&mpages, &pslot);
1112 }
1113
1114 if (0 == nodb)
1115 SQL_EXEC("END TRANSACTION");
1116 }
1117
1118 static void
1119 parse_cat(struct mpage *mpage)
1120 {
1121 FILE *stream;
1122 char *line, *p, *title;
1123 size_t len, plen, titlesz;
1124
1125 if (NULL == (stream = fopen(mpage->mlinks->file, "r"))) {
1126 if (warnings)
1127 say(mpage->mlinks->file, NULL);
1128 return;
1129 }
1130
1131 /* Skip to first blank line. */
1132
1133 while (NULL != (line = fgetln(stream, &len)))
1134 if ('\n' == *line)
1135 break;
1136
1137 /*
1138 * Assume the first line that is not indented
1139 * is the first section header. Skip to it.
1140 */
1141
1142 while (NULL != (line = fgetln(stream, &len)))
1143 if ('\n' != *line && ' ' != *line)
1144 break;
1145
1146 /*
1147 * Read up until the next section into a buffer.
1148 * Strip the leading and trailing newline from each read line,
1149 * appending a trailing space.
1150 * Ignore empty (whitespace-only) lines.
1151 */
1152
1153 titlesz = 0;
1154 title = NULL;
1155
1156 while (NULL != (line = fgetln(stream, &len))) {
1157 if (' ' != *line || '\n' != line[len - 1])
1158 break;
1159 while (len > 0 && isspace((unsigned char)*line)) {
1160 line++;
1161 len--;
1162 }
1163 if (1 == len)
1164 continue;
1165 title = mandoc_realloc(title, titlesz + len);
1166 memcpy(title + titlesz, line, len);
1167 titlesz += len;
1168 title[titlesz - 1] = ' ';
1169 }
1170
1171 /*
1172 * If no page content can be found, or the input line
1173 * is already the next section header, or there is no
1174 * trailing newline, reuse the page title as the page
1175 * description.
1176 */
1177
1178 if (NULL == title || '\0' == *title) {
1179 if (warnings)
1180 say(mpage->mlinks->file,
1181 "Cannot find NAME section");
1182 assert(NULL == mpage->desc);
1183 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1184 putkey(mpage, mpage->mlinks->name, TYPE_Nd);
1185 fclose(stream);
1186 free(title);
1187 return;
1188 }
1189
1190 title = mandoc_realloc(title, titlesz + 1);
1191 title[titlesz] = '\0';
1192
1193 /*
1194 * Skip to the first dash.
1195 * Use the remaining line as the description (no more than 70
1196 * bytes).
1197 */
1198
1199 if (NULL != (p = strstr(title, "- "))) {
1200 for (p += 2; ' ' == *p || '\b' == *p; p++)
1201 /* Skip to next word. */ ;
1202 } else {
1203 if (warnings)
1204 say(mpage->mlinks->file,
1205 "No dash in title line");
1206 p = title;
1207 }
1208
1209 plen = strlen(p);
1210
1211 /* Strip backspace-encoding from line. */
1212
1213 while (NULL != (line = memchr(p, '\b', plen))) {
1214 len = line - p;
1215 if (0 == len) {
1216 memmove(line, line + 1, plen--);
1217 continue;
1218 }
1219 memmove(line - 1, line + 1, plen - len);
1220 plen -= 2;
1221 }
1222
1223 assert(NULL == mpage->desc);
1224 mpage->desc = mandoc_strdup(p);
1225 putkey(mpage, mpage->desc, TYPE_Nd);
1226 fclose(stream);
1227 free(title);
1228 }
1229
1230 /*
1231 * Put a type/word pair into the word database for this particular file.
1232 */
1233 static void
1234 putkey(const struct mpage *mpage, char *value, uint64_t type)
1235 {
1236 char *cp;
1237
1238 assert(NULL != value);
1239 if (TYPE_arch == type)
1240 for (cp = value; *cp; cp++)
1241 if (isupper((unsigned char)*cp))
1242 *cp = _tolower((unsigned char)*cp);
1243 putkeys(mpage, value, strlen(value), type);
1244 }
1245
1246 /*
1247 * Grok all nodes at or below a certain mdoc node into putkey().
1248 */
1249 static void
1250 putmdockey(const struct mpage *mpage,
1251 const struct mdoc_node *n, uint64_t m)
1252 {
1253
1254 for ( ; NULL != n; n = n->next) {
1255 if (NULL != n->child)
1256 putmdockey(mpage, n->child, m);
1257 if (MDOC_TEXT == n->type)
1258 putkey(mpage, n->string, m);
1259 }
1260 }
1261
1262 static void
1263 parse_man(struct mpage *mpage, const struct man_node *n)
1264 {
1265 const struct man_node *head, *body;
1266 char *start, *sv, *title;
1267 char byte;
1268 size_t sz, titlesz;
1269
1270 if (NULL == n)
1271 return;
1272
1273 /*
1274 * We're only searching for one thing: the first text child in
1275 * the BODY of a NAME section. Since we don't keep track of
1276 * sections in -man, run some hoops to find out whether we're in
1277 * the correct section or not.
1278 */
1279
1280 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1281 body = n;
1282 assert(body->parent);
1283 if (NULL != (head = body->parent->head) &&
1284 1 == head->nchild &&
1285 NULL != (head = (head->child)) &&
1286 MAN_TEXT == head->type &&
1287 0 == strcmp(head->string, "NAME") &&
1288 NULL != (body = body->child) &&
1289 MAN_TEXT == body->type) {
1290
1291 title = NULL;
1292 titlesz = 0;
1293
1294 /*
1295 * Suck the entire NAME section into memory.
1296 * Yes, we might run away.
1297 * But too many manuals have big, spread-out
1298 * NAME sections over many lines.
1299 */
1300
1301 for ( ; NULL != body; body = body->next) {
1302 if (MAN_TEXT != body->type)
1303 break;
1304 if (0 == (sz = strlen(body->string)))
1305 continue;
1306 title = mandoc_realloc
1307 (title, titlesz + sz + 1);
1308 memcpy(title + titlesz, body->string, sz);
1309 titlesz += sz + 1;
1310 title[titlesz - 1] = ' ';
1311 }
1312 if (NULL == title)
1313 return;
1314
1315 title = mandoc_realloc(title, titlesz + 1);
1316 title[titlesz] = '\0';
1317
1318 /* Skip leading space. */
1319
1320 sv = title;
1321 while (isspace((unsigned char)*sv))
1322 sv++;
1323
1324 if (0 == (sz = strlen(sv))) {
1325 free(title);
1326 return;
1327 }
1328
1329 /* Erase trailing space. */
1330
1331 start = &sv[sz - 1];
1332 while (start > sv && isspace((unsigned char)*start))
1333 *start-- = '\0';
1334
1335 if (start == sv) {
1336 free(title);
1337 return;
1338 }
1339
1340 start = sv;
1341
1342 /*
1343 * Go through a special heuristic dance here.
1344 * Conventionally, one or more manual names are
1345 * comma-specified prior to a whitespace, then a
1346 * dash, then a description. Try to puzzle out
1347 * the name parts here.
1348 */
1349
1350 for ( ;; ) {
1351 sz = strcspn(start, " ,");
1352 if ('\0' == start[sz])
1353 break;
1354
1355 byte = start[sz];
1356 start[sz] = '\0';
1357
1358 /*
1359 * Assume a stray trailing comma in the
1360 * name list if a name begins with a dash.
1361 */
1362
1363 if ('-' == start[0] ||
1364 ('\\' == start[0] && '-' == start[1]))
1365 break;
1366
1367 putkey(mpage, start, TYPE_Nm);
1368
1369 if (' ' == byte) {
1370 start += sz + 1;
1371 break;
1372 }
1373
1374 assert(',' == byte);
1375 start += sz + 1;
1376 while (' ' == *start)
1377 start++;
1378 }
1379
1380 if (sv == start) {
1381 putkey(mpage, start, TYPE_Nm);
1382 free(title);
1383 return;
1384 }
1385
1386 while (isspace((unsigned char)*start))
1387 start++;
1388
1389 if (0 == strncmp(start, "-", 1))
1390 start += 1;
1391 else if (0 == strncmp(start, "\\-\\-", 4))
1392 start += 4;
1393 else if (0 == strncmp(start, "\\-", 2))
1394 start += 2;
1395 else if (0 == strncmp(start, "\\(en", 4))
1396 start += 4;
1397 else if (0 == strncmp(start, "\\(em", 4))
1398 start += 4;
1399
1400 while (' ' == *start)
1401 start++;
1402
1403 assert(NULL == mpage->desc);
1404 mpage->desc = mandoc_strdup(start);
1405 putkey(mpage, mpage->desc, TYPE_Nd);
1406 free(title);
1407 return;
1408 }
1409 }
1410
1411 for (n = n->child; n; n = n->next) {
1412 if (NULL != mpage->desc)
1413 break;
1414 parse_man(mpage, n);
1415 }
1416 }
1417
1418 static void
1419 parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1420 {
1421
1422 assert(NULL != n);
1423 for (n = n->child; NULL != n; n = n->next) {
1424 switch (n->type) {
1425 case (MDOC_ELEM):
1426 /* FALLTHROUGH */
1427 case (MDOC_BLOCK):
1428 /* FALLTHROUGH */
1429 case (MDOC_HEAD):
1430 /* FALLTHROUGH */
1431 case (MDOC_BODY):
1432 /* FALLTHROUGH */
1433 case (MDOC_TAIL):
1434 if (NULL != mdocs[n->tok].fp)
1435 if (0 == (*mdocs[n->tok].fp)(mpage, n))
1436 break;
1437 if (mdocs[n->tok].mask)
1438 putmdockey(mpage, n->child,
1439 mdocs[n->tok].mask);
1440 break;
1441 default:
1442 assert(MDOC_ROOT != n->type);
1443 continue;
1444 }
1445 if (NULL != n->child)
1446 parse_mdoc(mpage, n);
1447 }
1448 }
1449
1450 static int
1451 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1452 {
1453 const char *start, *end;
1454 size_t sz;
1455
1456 if (SEC_SYNOPSIS != n->sec ||
1457 NULL == (n = n->child) ||
1458 MDOC_TEXT != n->type)
1459 return(0);
1460
1461 /*
1462 * Only consider those `Fd' macro fields that begin with an
1463 * "inclusion" token (versus, e.g., #define).
1464 */
1465
1466 if (strcmp("#include", n->string))
1467 return(0);
1468
1469 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1470 return(0);
1471
1472 /*
1473 * Strip away the enclosing angle brackets and make sure we're
1474 * not zero-length.
1475 */
1476
1477 start = n->string;
1478 if ('<' == *start || '"' == *start)
1479 start++;
1480
1481 if (0 == (sz = strlen(start)))
1482 return(0);
1483
1484 end = &start[(int)sz - 1];
1485 if ('>' == *end || '"' == *end)
1486 end--;
1487
1488 if (end > start)
1489 putkeys(mpage, start, end - start + 1, TYPE_In);
1490 return(0);
1491 }
1492
1493 static int
1494 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1495 {
1496 char *cp;
1497
1498 if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1499 return(0);
1500
1501 /*
1502 * Parse: .Fn "struct type *name" "char *arg".
1503 * First strip away pointer symbol.
1504 * Then store the function name, then type.
1505 * Finally, store the arguments.
1506 */
1507
1508 if (NULL == (cp = strrchr(n->string, ' ')))
1509 cp = n->string;
1510
1511 while ('*' == *cp)
1512 cp++;
1513
1514 putkey(mpage, cp, TYPE_Fn);
1515
1516 if (n->string < cp)
1517 putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1518
1519 for (n = n->next; NULL != n; n = n->next)
1520 if (MDOC_TEXT == n->type)
1521 putkey(mpage, n->string, TYPE_Fa);
1522
1523 return(0);
1524 }
1525
1526 static int
1527 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1528 {
1529 char *cp;
1530
1531 if (NULL == (n = n->child))
1532 return(0);
1533
1534 if (NULL == n->next) {
1535 putkey(mpage, n->string, TYPE_Xr);
1536 return(0);
1537 }
1538
1539 if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) {
1540 perror(NULL);
1541 exit((int)MANDOCLEVEL_SYSERR);
1542 }
1543 putkey(mpage, cp, TYPE_Xr);
1544 free(cp);
1545 return(0);
1546 }
1547
1548 static int
1549 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1550 {
1551 size_t sz;
1552
1553 if (MDOC_BODY != n->type)
1554 return(0);
1555
1556 /*
1557 * Special-case the `Nd' because we need to put the description
1558 * into the document table.
1559 */
1560
1561 for (n = n->child; NULL != n; n = n->next) {
1562 if (MDOC_TEXT == n->type) {
1563 if (NULL != mpage->desc) {
1564 sz = strlen(mpage->desc) +
1565 strlen(n->string) + 2;
1566 mpage->desc = mandoc_realloc(
1567 mpage->desc, sz);
1568 strlcat(mpage->desc, " ", sz);
1569 strlcat(mpage->desc, n->string, sz);
1570 } else
1571 mpage->desc = mandoc_strdup(n->string);
1572 }
1573 if (NULL != n->child)
1574 parse_mdoc_Nd(mpage, n);
1575 }
1576 return(1);
1577 }
1578
1579 static int
1580 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1581 {
1582
1583 return(SEC_NAME == n->sec ||
1584 (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type));
1585 }
1586
1587 static int
1588 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1589 {
1590
1591 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1592 }
1593
1594 static int
1595 parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1596 {
1597
1598 return(MDOC_HEAD == n->type);
1599 }
1600
1601 static int
1602 parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1603 {
1604
1605 return(MDOC_BODY == n->type);
1606 }
1607
1608 /*
1609 * Add a string to the hash table for the current manual.
1610 * Each string has a bitmask telling which macros it belongs to.
1611 * When we finish the manual, we'll dump the table.
1612 */
1613 static void
1614 putkeys(const struct mpage *mpage,
1615 const char *cp, size_t sz, uint64_t v)
1616 {
1617 struct str *s;
1618 const char *end;
1619 uint64_t mask;
1620 unsigned int slot;
1621 int i;
1622
1623 if (0 == sz)
1624 return;
1625
1626 if (verb > 1) {
1627 for (i = 0, mask = 1;
1628 i < mansearch_keymax;
1629 i++, mask <<= 1)
1630 if (mask & v)
1631 break;
1632 say(mpage->mlinks->file, "Adding key %s=%*s",
1633 mansearch_keynames[i], sz, cp);
1634 }
1635
1636 end = cp + sz;
1637 slot = ohash_qlookupi(&strings, cp, &end);
1638 s = ohash_find(&strings, slot);
1639
1640 if (NULL != s && mpage == s->mpage) {
1641 s->mask |= v;
1642 return;
1643 } else if (NULL == s) {
1644 s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1645 memcpy(s->key, cp, sz);
1646 ohash_insert(&strings, slot, s);
1647 }
1648 s->mpage = mpage;
1649 s->mask = v;
1650 }
1651
1652 /*
1653 * Take a Unicode codepoint and produce its UTF-8 encoding.
1654 * This isn't the best way to do this, but it works.
1655 * The magic numbers are from the UTF-8 packaging.
1656 * They're not as scary as they seem: read the UTF-8 spec for details.
1657 */
1658 static size_t
1659 utf8(unsigned int cp, char out[7])
1660 {
1661 size_t rc;
1662
1663 rc = 0;
1664 if (cp <= 0x0000007F) {
1665 rc = 1;
1666 out[0] = (char)cp;
1667 } else if (cp <= 0x000007FF) {
1668 rc = 2;
1669 out[0] = (cp >> 6 & 31) | 192;
1670 out[1] = (cp & 63) | 128;
1671 } else if (cp <= 0x0000FFFF) {
1672 rc = 3;
1673 out[0] = (cp >> 12 & 15) | 224;
1674 out[1] = (cp >> 6 & 63) | 128;
1675 out[2] = (cp & 63) | 128;
1676 } else if (cp <= 0x001FFFFF) {
1677 rc = 4;
1678 out[0] = (cp >> 18 & 7) | 240;
1679 out[1] = (cp >> 12 & 63) | 128;
1680 out[2] = (cp >> 6 & 63) | 128;
1681 out[3] = (cp & 63) | 128;
1682 } else if (cp <= 0x03FFFFFF) {
1683 rc = 5;
1684 out[0] = (cp >> 24 & 3) | 248;
1685 out[1] = (cp >> 18 & 63) | 128;
1686 out[2] = (cp >> 12 & 63) | 128;
1687 out[3] = (cp >> 6 & 63) | 128;
1688 out[4] = (cp & 63) | 128;
1689 } else if (cp <= 0x7FFFFFFF) {
1690 rc = 6;
1691 out[0] = (cp >> 30 & 1) | 252;
1692 out[1] = (cp >> 24 & 63) | 128;
1693 out[2] = (cp >> 18 & 63) | 128;
1694 out[3] = (cp >> 12 & 63) | 128;
1695 out[4] = (cp >> 6 & 63) | 128;
1696 out[5] = (cp & 63) | 128;
1697 } else
1698 return(0);
1699
1700 out[rc] = '\0';
1701 return(rc);
1702 }
1703
1704 /*
1705 * Store the rendered version of a key, or alias the pointer
1706 * if the key contains no escape sequences.
1707 */
1708 static void
1709 render_key(struct mchars *mc, struct str *key)
1710 {
1711 size_t sz, bsz, pos;
1712 char utfbuf[7], res[6];
1713 char *buf;
1714 const char *seq, *cpp, *val;
1715 int len, u;
1716 enum mandoc_esc esc;
1717
1718 assert(NULL == key->rendered);
1719
1720 res[0] = '\\';
1721 res[1] = '\t';
1722 res[2] = ASCII_NBRSP;
1723 res[3] = ASCII_HYPH;
1724 res[4] = ASCII_BREAK;
1725 res[5] = '\0';
1726
1727 val = key->key;
1728 bsz = strlen(val);
1729
1730 /*
1731 * Pre-check: if we have no stop-characters, then set the
1732 * pointer as ourselvse and get out of here.
1733 */
1734 if (strcspn(val, res) == bsz) {
1735 key->rendered = key->key;
1736 return;
1737 }
1738
1739 /* Pre-allocate by the length of the input */
1740
1741 buf = mandoc_malloc(++bsz);
1742 pos = 0;
1743
1744 while ('\0' != *val) {
1745 /*
1746 * Halt on the first escape sequence.
1747 * This also halts on the end of string, in which case
1748 * we just copy, fallthrough, and exit the loop.
1749 */
1750 if ((sz = strcspn(val, res)) > 0) {
1751 memcpy(&buf[pos], val, sz);
1752 pos += sz;
1753 val += sz;
1754 }
1755
1756 switch (*val) {
1757 case (ASCII_HYPH):
1758 buf[pos++] = '-';
1759 val++;
1760 continue;
1761 case ('\t'):
1762 /* FALLTHROUGH */
1763 case (ASCII_NBRSP):
1764 buf[pos++] = ' ';
1765 val++;
1766 /* FALLTHROUGH */
1767 case (ASCII_BREAK):
1768 continue;
1769 default:
1770 break;
1771 }
1772 if ('\\' != *val)
1773 break;
1774
1775 /* Read past the slash. */
1776
1777 val++;
1778
1779 /*
1780 * Parse the escape sequence and see if it's a
1781 * predefined character or special character.
1782 */
1783
1784 esc = mandoc_escape
1785 ((const char **)&val, &seq, &len);
1786 if (ESCAPE_ERROR == esc)
1787 break;
1788 if (ESCAPE_SPECIAL != esc)
1789 continue;
1790
1791 /*
1792 * Render the special character
1793 * as either UTF-8 or ASCII.
1794 */
1795
1796 if (write_utf8) {
1797 if (0 == (u = mchars_spec2cp(mc, seq, len)))
1798 continue;
1799 cpp = utfbuf;
1800 if (0 == (sz = utf8(u, utfbuf)))
1801 continue;
1802 sz = strlen(cpp);
1803 } else {
1804 cpp = mchars_spec2str(mc, seq, len, &sz);
1805 if (NULL == cpp)
1806 continue;
1807 if (ASCII_NBRSP == *cpp) {
1808 cpp = " ";
1809 sz = 1;
1810 }
1811 }
1812
1813 /* Copy the rendered glyph into the stream. */
1814
1815 bsz += sz;
1816 buf = mandoc_realloc(buf, bsz);
1817 memcpy(&buf[pos], cpp, sz);
1818 pos += sz;
1819 }
1820
1821 buf[pos] = '\0';
1822 key->rendered = buf;
1823 }
1824
1825 static void
1826 dbadd_mlink(const struct mlink *mlink)
1827 {
1828 size_t i;
1829
1830 i = 1;
1831 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1832 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1833 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1834 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->recno);
1835 SQL_STEP(stmts[STMT_INSERT_LINK]);
1836 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1837 }
1838
1839 /*
1840 * Flush the current page's terms (and their bits) into the database.
1841 * Wrap the entire set of additions in a transaction to make sqlite be a
1842 * little faster.
1843 * Also, handle escape sequences at the last possible moment.
1844 */
1845 static void
1846 dbadd(struct mpage *mpage, struct mchars *mc)
1847 {
1848 struct mlink *mlink;
1849 struct str *key;
1850 size_t i;
1851 unsigned int slot;
1852
1853 if (verb)
1854 say(mpage->mlinks->file, "Adding to database");
1855
1856 if (nodb)
1857 return;
1858
1859 i = 1;
1860 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
1861 SQL_STEP(stmts[STMT_INSERT_PAGE]);
1862 mpage->recno = sqlite3_last_insert_rowid(db);
1863 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1864
1865 for (mlink = mpage->mlinks; mlink; mlink = mlink->next)
1866 dbadd_mlink(mlink);
1867
1868 for (key = ohash_first(&strings, &slot); NULL != key;
1869 key = ohash_next(&strings, &slot)) {
1870 assert(key->mpage == mpage);
1871 if (NULL == key->rendered)
1872 render_key(mc, key);
1873 i = 1;
1874 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1875 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
1876 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->recno);
1877 SQL_STEP(stmts[STMT_INSERT_KEY]);
1878 sqlite3_reset(stmts[STMT_INSERT_KEY]);
1879 if (key->rendered != key->key)
1880 free(key->rendered);
1881 free(key);
1882 }
1883 }
1884
1885 static void
1886 dbprune(void)
1887 {
1888 struct mpage *mpage;
1889 struct mlink *mlink;
1890 size_t i;
1891 unsigned int slot;
1892
1893 if (0 == nodb)
1894 SQL_EXEC("BEGIN TRANSACTION");
1895
1896 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
1897 mpage = ohash_next(&mpages, &slot)) {
1898 mlink = mpage->mlinks;
1899 if (verb)
1900 say(mlink->file, "Deleting from database");
1901 if (nodb)
1902 continue;
1903 for ( ; NULL != mlink; mlink = mlink->next) {
1904 i = 1;
1905 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1906 i, mlink->dsec);
1907 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1908 i, mlink->arch);
1909 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1910 i, mlink->name);
1911 SQL_STEP(stmts[STMT_DELETE_PAGE]);
1912 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
1913 }
1914 }
1915
1916 if (0 == nodb)
1917 SQL_EXEC("END TRANSACTION");
1918 }
1919
1920 /*
1921 * Close an existing database and its prepared statements.
1922 * If "real" is not set, rename the temporary file into the real one.
1923 */
1924 static void
1925 dbclose(int real)
1926 {
1927 size_t i;
1928 int status;
1929 pid_t child;
1930
1931 if (nodb)
1932 return;
1933
1934 for (i = 0; i < STMT__MAX; i++) {
1935 sqlite3_finalize(stmts[i]);
1936 stmts[i] = NULL;
1937 }
1938
1939 sqlite3_close(db);
1940 db = NULL;
1941
1942 if (real)
1943 return;
1944
1945 if ('\0' == *tempfilename) {
1946 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
1947 exitcode = (int)MANDOCLEVEL_SYSERR;
1948 say(MANDOC_DB, "%s", strerror(errno));
1949 }
1950 return;
1951 }
1952
1953 switch (child = fork()) {
1954 case (-1):
1955 exitcode = (int)MANDOCLEVEL_SYSERR;
1956 say("fork cmp", "%s", strerror(errno));
1957 return;
1958 case (0):
1959 execlp("cmp", "cmp", "-s",
1960 tempfilename, MANDOC_DB, NULL);
1961 say("exec cmp", "%s", strerror(errno));
1962 exit(0);
1963 default:
1964 break;
1965 }
1966 if (-1 == waitpid(child, &status, 0)) {
1967 exitcode = (int)MANDOCLEVEL_SYSERR;
1968 say("wait cmp", "%s", strerror(errno));
1969 } else if (WIFSIGNALED(status)) {
1970 exitcode = (int)MANDOCLEVEL_SYSERR;
1971 say("cmp", "Died from a signal");
1972 } else if (WEXITSTATUS(status)) {
1973 exitcode = (int)MANDOCLEVEL_SYSERR;
1974 say(MANDOC_DB,
1975 "Data changed, but cannot replace database");
1976 }
1977
1978 *strrchr(tempfilename, '/') = '\0';
1979 switch (child = fork()) {
1980 case (-1):
1981 exitcode = (int)MANDOCLEVEL_SYSERR;
1982 say("fork rm", "%s", strerror(errno));
1983 return;
1984 case (0):
1985 execlp("rm", "rm", "-rf", tempfilename, NULL);
1986 say("exec rm", "%s", strerror(errno));
1987 exit((int)MANDOCLEVEL_SYSERR);
1988 default:
1989 break;
1990 }
1991 if (-1 == waitpid(child, &status, 0)) {
1992 exitcode = (int)MANDOCLEVEL_SYSERR;
1993 say("wait rm", "%s", strerror(errno));
1994 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
1995 exitcode = (int)MANDOCLEVEL_SYSERR;
1996 say(tempfilename,
1997 "Cannot remove temporary directory");
1998 }
1999 }
2000
2001 /*
2002 * This is straightforward stuff.
2003 * Open a database connection to a "temporary" database, then open a set
2004 * of prepared statements we'll use over and over again.
2005 * If "real" is set, we use the existing database; if not, we truncate a
2006 * temporary one.
2007 * Must be matched by dbclose().
2008 */
2009 static int
2010 dbopen(int real)
2011 {
2012 const char *sql;
2013 int rc, ofl;
2014
2015 if (nodb)
2016 return(1);
2017
2018 *tempfilename = '\0';
2019 ofl = SQLITE_OPEN_READWRITE;
2020
2021 if (real) {
2022 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2023 if (SQLITE_OK != rc) {
2024 exitcode = (int)MANDOCLEVEL_SYSERR;
2025 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2026 return(0);
2027 }
2028 goto prepare_statements;
2029 }
2030
2031 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2032
2033 remove(MANDOC_DB "~");
2034 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2035 if (SQLITE_OK == rc)
2036 goto create_tables;
2037 if (MPARSE_QUICK & mparse_options) {
2038 exitcode = (int)MANDOCLEVEL_SYSERR;
2039 say(MANDOC_DB "~", "%s", sqlite3_errmsg(db));
2040 return(0);
2041 }
2042
2043 if (strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2044 sizeof(tempfilename)) >= sizeof(tempfilename)) {
2045 exitcode = (int)MANDOCLEVEL_SYSERR;
2046 say("/tmp/mandocdb.XXXXXX", "Filename too long");
2047 return(0);
2048 }
2049 if (NULL == mkdtemp(tempfilename)) {
2050 exitcode = (int)MANDOCLEVEL_SYSERR;
2051 say(tempfilename, "%s", strerror(errno));
2052 return(0);
2053 }
2054 if (strlcat(tempfilename, "/" MANDOC_DB,
2055 sizeof(tempfilename)) >= sizeof(tempfilename)) {
2056 exitcode = (int)MANDOCLEVEL_SYSERR;
2057 say(tempfilename, "Filename too long");
2058 return(0);
2059 }
2060 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2061 if (SQLITE_OK != rc) {
2062 exitcode = (int)MANDOCLEVEL_SYSERR;
2063 say(tempfilename, "%s", sqlite3_errmsg(db));
2064 return(0);
2065 }
2066
2067 create_tables:
2068 sql = "CREATE TABLE \"mpages\" (\n"
2069 " \"form\" INTEGER NOT NULL,\n"
2070 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2071 ");\n"
2072 "\n"
2073 "CREATE TABLE \"mlinks\" (\n"
2074 " \"sec\" TEXT NOT NULL,\n"
2075 " \"arch\" TEXT NOT NULL,\n"
2076 " \"name\" TEXT NOT NULL,\n"
2077 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
2078 "ON DELETE CASCADE\n"
2079 ");\n"
2080 "\n"
2081 "CREATE TABLE \"keys\" (\n"
2082 " \"bits\" INTEGER NOT NULL,\n"
2083 " \"key\" TEXT NOT NULL,\n"
2084 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
2085 "ON DELETE CASCADE\n"
2086 ");\n";
2087
2088 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2089 exitcode = (int)MANDOCLEVEL_SYSERR;
2090 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2091 return(0);
2092 }
2093
2094 prepare_statements:
2095 SQL_EXEC("PRAGMA foreign_keys = ON");
2096 sql = "DELETE FROM mpages WHERE id IN "
2097 "(SELECT pageid FROM mlinks WHERE "
2098 "sec=? AND arch=? AND name=?)";
2099 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2100 sql = "INSERT INTO mpages "
2101 "(form) VALUES (?)";
2102 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2103 sql = "INSERT INTO mlinks "
2104 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2105 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2106 sql = "INSERT INTO keys "
2107 "(bits,key,pageid) VALUES (?,?,?)";
2108 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2109
2110 #ifndef __APPLE__
2111 /*
2112 * When opening a new database, we can turn off
2113 * synchronous mode for much better performance.
2114 */
2115
2116 if (real)
2117 SQL_EXEC("PRAGMA synchronous = OFF");
2118 #endif
2119
2120 return(1);
2121 }
2122
2123 static void *
2124 hash_halloc(size_t sz, void *arg)
2125 {
2126
2127 return(mandoc_calloc(sz, 1));
2128 }
2129
2130 static void *
2131 hash_alloc(size_t sz, void *arg)
2132 {
2133
2134 return(mandoc_malloc(sz));
2135 }
2136
2137 static void
2138 hash_free(void *p, size_t sz, void *arg)
2139 {
2140
2141 free(p);
2142 }
2143
2144 static int
2145 set_basedir(const char *targetdir)
2146 {
2147 static char startdir[PATH_MAX];
2148 static int fd;
2149
2150 /*
2151 * Remember where we started by keeping a fd open to the origin
2152 * path component: throughout this utility, we chdir() a lot to
2153 * handle relative paths, and by doing this, we can return to
2154 * the starting point.
2155 */
2156 if ('\0' == *startdir) {
2157 if (NULL == getcwd(startdir, PATH_MAX)) {
2158 exitcode = (int)MANDOCLEVEL_SYSERR;
2159 if (NULL != targetdir)
2160 say(".", NULL);
2161 return(0);
2162 }
2163 if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
2164 exitcode = (int)MANDOCLEVEL_SYSERR;
2165 say(startdir, NULL);
2166 return(0);
2167 }
2168 if (NULL == targetdir)
2169 targetdir = startdir;
2170 } else {
2171 if (-1 == fd)
2172 return(0);
2173 if (-1 == fchdir(fd)) {
2174 close(fd);
2175 basedir[0] = '\0';
2176 exitcode = (int)MANDOCLEVEL_SYSERR;
2177 say(startdir, NULL);
2178 return(0);
2179 }
2180 if (NULL == targetdir) {
2181 close(fd);
2182 return(1);
2183 }
2184 }
2185 if (NULL == realpath(targetdir, basedir)) {
2186 basedir[0] = '\0';
2187 exitcode = (int)MANDOCLEVEL_BADARG;
2188 say(targetdir, NULL);
2189 return(0);
2190 } else if (-1 == chdir(basedir)) {
2191 exitcode = (int)MANDOCLEVEL_BADARG;
2192 say("", NULL);
2193 return(0);
2194 }
2195 return(1);
2196 }
2197
2198 static void
2199 say(const char *file, const char *format, ...)
2200 {
2201 va_list ap;
2202
2203 if ('\0' != *basedir)
2204 fprintf(stderr, "%s", basedir);
2205 if ('\0' != *basedir && '\0' != *file)
2206 fputs("//", stderr);
2207 if ('\0' != *file)
2208 fprintf(stderr, "%s", file);
2209 fputs(": ", stderr);
2210
2211 if (NULL == format) {
2212 perror(NULL);
2213 return;
2214 }
2215
2216 va_start(ap, format);
2217 vfprintf(stderr, format, ap);
2218 va_end(ap);
2219
2220 fputc('\n', stderr);
2221 }
A git conversion mirror of mandoc.bsd.lv