]> 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
Another step on the way to clear naming, this time regarding mlinks:
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.83 2013/12/27 14:29:28 schwarze Exp $ */
2 /*
3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011, 2012, 2013 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
24 #include <assert.h>
25 #include <ctype.h>
26 #include <errno.h>
27 #include <fcntl.h>
28 #include <fts.h>
29 #include <getopt.h>
30 #include <limits.h>
31 #include <stddef.h>
32 #include <stdio.h>
33 #include <stdint.h>
34 #include <stdlib.h>
35 #include <string.h>
36 #include <unistd.h>
37
38 #ifdef HAVE_OHASH
39 #include <ohash.h>
40 #else
41 #include "compat_ohash.h"
42 #endif
43 #include <sqlite3.h>
44
45 #include "mdoc.h"
46 #include "man.h"
47 #include "mandoc.h"
48 #include "manpath.h"
49 #include "mansearch.h"
50
51 #define SQL_EXEC(_v) \
52 if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
53 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
54 #define SQL_BIND_TEXT(_s, _i, _v) \
55 if (SQLITE_OK != sqlite3_bind_text \
56 ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
57 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
58 #define SQL_BIND_INT(_s, _i, _v) \
59 if (SQLITE_OK != sqlite3_bind_int \
60 ((_s), (_i)++, (_v))) \
61 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
62 #define SQL_BIND_INT64(_s, _i, _v) \
63 if (SQLITE_OK != sqlite3_bind_int64 \
64 ((_s), (_i)++, (_v))) \
65 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
66 #define SQL_STEP(_s) \
67 if (SQLITE_DONE != sqlite3_step((_s))) \
68 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
69
70 enum op {
71 OP_DEFAULT = 0, /* new dbs from dir list or default config */
72 OP_CONFFILE, /* new databases from custom config file */
73 OP_UPDATE, /* delete/add entries in existing database */
74 OP_DELETE, /* delete entries from existing database */
75 OP_TEST /* change no databases, report potential problems */
76 };
77
78 enum form {
79 FORM_SRC, /* format is -man or -mdoc */
80 FORM_CAT, /* format is cat */
81 FORM_NONE /* format is unknown */
82 };
83
84 struct str {
85 char *utf8; /* key in UTF-8 form */
86 const struct mpage *mpage; /* if set, the owning parse */
87 uint64_t mask; /* bitmask in sequence */
88 char key[]; /* the string itself */
89 };
90
91 struct inodev {
92 ino_t st_ino;
93 dev_t st_dev;
94 };
95
96 struct mpage {
97 struct inodev inodev; /* used for hashing routine */
98 enum form form; /* format from file content */
99 char *sec; /* section from file content */
100 char *arch; /* architecture from file content */
101 char *title; /* title from file content */
102 char *desc; /* description from file content */
103 struct mlink *mlinks; /* singly linked list */
104 };
105
106 struct mlink {
107 char file[PATH_MAX]; /* filename rel. to manpath */
108 enum form dform; /* format from directory */
109 enum form sform; /* format from file name suffix */
110 char *dsec; /* section from directory */
111 char *arch; /* architecture from directory */
112 char *name; /* name from file name (not empty) */
113 char *fsec; /* section from file name suffix */
114 };
115
116 struct title {
117 char *title; /* name(sec/arch) given inside the file */
118 char *file; /* file name in case of mismatch */
119 };
120
121 enum stmt {
122 STMT_DELETE_PAGE = 0, /* delete mpage */
123 STMT_INSERT_PAGE, /* insert mpage */
124 STMT_INSERT_LINK, /* insert mlink */
125 STMT_INSERT_KEY, /* insert parsed key */
126 STMT__MAX
127 };
128
129 typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
130
131 struct mdoc_handler {
132 mdoc_fp fp; /* optional handler */
133 uint64_t mask; /* set unless handler returns 0 */
134 };
135
136 static void dbclose(int);
137 static void dbindex(const struct mpage *, struct mchars *);
138 static int dbopen(int);
139 static void dbprune(void);
140 static void filescan(const char *);
141 static void *hash_alloc(size_t, void *);
142 static void hash_free(void *, size_t, void *);
143 static void *hash_halloc(size_t, void *);
144 static int inocheck(const struct stat *);
145 static void mlink_add(int, const char *, const char *, const char *,
146 const char *, const char *, const struct stat *);
147 static void mlink_free(struct mlink *);
148 static void mpages_free(void);
149 static void mpages_merge(struct mchars *, struct mparse *, int);
150 static void parse_cat(struct mpage *);
151 static void parse_man(struct mpage *, const struct man_node *);
152 static void parse_mdoc(struct mpage *, const struct mdoc_node *);
153 static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
154 static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
155 static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
156 static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
157 static int parse_mdoc_In(struct mpage *, const struct mdoc_node *);
158 static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
159 static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
160 static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
161 static int parse_mdoc_St(struct mpage *, const struct mdoc_node *);
162 static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
163 static void putkey(const struct mpage *,
164 const char *, uint64_t);
165 static void putkeys(const struct mpage *,
166 const char *, size_t, uint64_t);
167 static void putmdockey(const struct mpage *,
168 const struct mdoc_node *, uint64_t);
169 static void say(const char *, const char *, ...);
170 static int set_basedir(const char *);
171 static int treescan(void);
172 static size_t utf8(unsigned int, char [7]);
173 static void utf8key(struct mchars *, struct str *);
174
175 static char *progname;
176 static int use_all; /* use all found files */
177 static int nodb; /* no database changes */
178 static int verb; /* print what we're doing */
179 static int warnings; /* warn about crap */
180 static int exitcode; /* to be returned by main */
181 static enum op op; /* operational mode */
182 static char basedir[PATH_MAX]; /* current base directory */
183 static struct ohash mpages; /* table of distinct manual pages */
184 static struct ohash mlinks; /* table of directory entries */
185 static struct ohash strings; /* table of all strings */
186 static sqlite3 *db = NULL; /* current database */
187 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
188
189 static const struct mdoc_handler mdocs[MDOC_MAX] = {
190 { NULL, 0 }, /* Ap */
191 { NULL, 0 }, /* Dd */
192 { NULL, 0 }, /* Dt */
193 { NULL, 0 }, /* Os */
194 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
195 { parse_mdoc_head, TYPE_Ss }, /* Ss */
196 { NULL, 0 }, /* Pp */
197 { NULL, 0 }, /* D1 */
198 { NULL, 0 }, /* Dl */
199 { NULL, 0 }, /* Bd */
200 { NULL, 0 }, /* Ed */
201 { NULL, 0 }, /* Bl */
202 { NULL, 0 }, /* El */
203 { NULL, 0 }, /* It */
204 { NULL, 0 }, /* Ad */
205 { NULL, TYPE_An }, /* An */
206 { NULL, TYPE_Ar }, /* Ar */
207 { NULL, TYPE_Cd }, /* Cd */
208 { NULL, TYPE_Cm }, /* Cm */
209 { NULL, TYPE_Dv }, /* Dv */
210 { NULL, TYPE_Er }, /* Er */
211 { NULL, TYPE_Ev }, /* Ev */
212 { NULL, 0 }, /* Ex */
213 { NULL, TYPE_Fa }, /* Fa */
214 { parse_mdoc_Fd, 0 }, /* Fd */
215 { NULL, TYPE_Fl }, /* Fl */
216 { parse_mdoc_Fn, 0 }, /* Fn */
217 { NULL, TYPE_Ft }, /* Ft */
218 { NULL, TYPE_Ic }, /* Ic */
219 { parse_mdoc_In, TYPE_In }, /* In */
220 { NULL, TYPE_Li }, /* Li */
221 { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
222 { parse_mdoc_Nm, TYPE_Nm }, /* Nm */
223 { NULL, 0 }, /* Op */
224 { NULL, 0 }, /* Ot */
225 { NULL, TYPE_Pa }, /* Pa */
226 { NULL, 0 }, /* Rv */
227 { parse_mdoc_St, 0 }, /* St */
228 { NULL, TYPE_Va }, /* Va */
229 { parse_mdoc_body, TYPE_Va }, /* Vt */
230 { parse_mdoc_Xr, 0 }, /* Xr */
231 { NULL, 0 }, /* %A */
232 { NULL, 0 }, /* %B */
233 { NULL, 0 }, /* %D */
234 { NULL, 0 }, /* %I */
235 { NULL, 0 }, /* %J */
236 { NULL, 0 }, /* %N */
237 { NULL, 0 }, /* %O */
238 { NULL, 0 }, /* %P */
239 { NULL, 0 }, /* %R */
240 { NULL, 0 }, /* %T */
241 { NULL, 0 }, /* %V */
242 { NULL, 0 }, /* Ac */
243 { NULL, 0 }, /* Ao */
244 { NULL, 0 }, /* Aq */
245 { NULL, TYPE_At }, /* At */
246 { NULL, 0 }, /* Bc */
247 { NULL, 0 }, /* Bf */
248 { NULL, 0 }, /* Bo */
249 { NULL, 0 }, /* Bq */
250 { NULL, TYPE_Bsx }, /* Bsx */
251 { NULL, TYPE_Bx }, /* Bx */
252 { NULL, 0 }, /* Db */
253 { NULL, 0 }, /* Dc */
254 { NULL, 0 }, /* Do */
255 { NULL, 0 }, /* Dq */
256 { NULL, 0 }, /* Ec */
257 { NULL, 0 }, /* Ef */
258 { NULL, TYPE_Em }, /* Em */
259 { NULL, 0 }, /* Eo */
260 { NULL, TYPE_Fx }, /* Fx */
261 { NULL, TYPE_Ms }, /* Ms */
262 { NULL, 0 }, /* No */
263 { NULL, 0 }, /* Ns */
264 { NULL, TYPE_Nx }, /* Nx */
265 { NULL, TYPE_Ox }, /* Ox */
266 { NULL, 0 }, /* Pc */
267 { NULL, 0 }, /* Pf */
268 { NULL, 0 }, /* Po */
269 { NULL, 0 }, /* Pq */
270 { NULL, 0 }, /* Qc */
271 { NULL, 0 }, /* Ql */
272 { NULL, 0 }, /* Qo */
273 { NULL, 0 }, /* Qq */
274 { NULL, 0 }, /* Re */
275 { NULL, 0 }, /* Rs */
276 { NULL, 0 }, /* Sc */
277 { NULL, 0 }, /* So */
278 { NULL, 0 }, /* Sq */
279 { NULL, 0 }, /* Sm */
280 { NULL, 0 }, /* Sx */
281 { NULL, TYPE_Sy }, /* Sy */
282 { NULL, TYPE_Tn }, /* Tn */
283 { NULL, 0 }, /* Ux */
284 { NULL, 0 }, /* Xc */
285 { NULL, 0 }, /* Xo */
286 { parse_mdoc_head, 0 }, /* Fo */
287 { NULL, 0 }, /* Fc */
288 { NULL, 0 }, /* Oo */
289 { NULL, 0 }, /* Oc */
290 { NULL, 0 }, /* Bk */
291 { NULL, 0 }, /* Ek */
292 { NULL, 0 }, /* Bt */
293 { NULL, 0 }, /* Hf */
294 { NULL, 0 }, /* Fr */
295 { NULL, 0 }, /* Ud */
296 { NULL, TYPE_Lb }, /* Lb */
297 { NULL, 0 }, /* Lp */
298 { NULL, TYPE_Lk }, /* Lk */
299 { NULL, TYPE_Mt }, /* Mt */
300 { NULL, 0 }, /* Brq */
301 { NULL, 0 }, /* Bro */
302 { NULL, 0 }, /* Brc */
303 { NULL, 0 }, /* %C */
304 { NULL, 0 }, /* Es */
305 { NULL, 0 }, /* En */
306 { NULL, TYPE_Dx }, /* Dx */
307 { NULL, 0 }, /* %Q */
308 { NULL, 0 }, /* br */
309 { NULL, 0 }, /* sp */
310 { NULL, 0 }, /* %U */
311 { NULL, 0 }, /* Ta */
312 };
313
314 int
315 main(int argc, char *argv[])
316 {
317 int ch, i;
318 size_t j, sz;
319 const char *path_arg;
320 struct mchars *mc;
321 struct manpaths dirs;
322 struct mparse *mp;
323 struct ohash_info mpages_info, mlinks_info;
324
325 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
326 memset(&dirs, 0, sizeof(struct manpaths));
327
328 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
329 mpages_info.halloc = mlinks_info.halloc = hash_halloc;
330 mpages_info.hfree = mlinks_info.hfree = hash_free;
331
332 mpages_info.key_offset = offsetof(struct mpage, inodev);
333 mlinks_info.key_offset = offsetof(struct mlink, file);
334
335 progname = strrchr(argv[0], '/');
336 if (progname == NULL)
337 progname = argv[0];
338 else
339 ++progname;
340
341 /*
342 * We accept a few different invocations.
343 * The CHECKOP macro makes sure that invocation styles don't
344 * clobber each other.
345 */
346 #define CHECKOP(_op, _ch) do \
347 if (OP_DEFAULT != (_op)) { \
348 fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
349 goto usage; \
350 } while (/*CONSTCOND*/0)
351
352 path_arg = NULL;
353 op = OP_DEFAULT;
354
355 while (-1 != (ch = getopt(argc, argv, "aC:d:ntu:vW")))
356 switch (ch) {
357 case ('a'):
358 use_all = 1;
359 break;
360 case ('C'):
361 CHECKOP(op, ch);
362 path_arg = optarg;
363 op = OP_CONFFILE;
364 break;
365 case ('d'):
366 CHECKOP(op, ch);
367 path_arg = optarg;
368 op = OP_UPDATE;
369 break;
370 case ('n'):
371 nodb = 1;
372 break;
373 case ('t'):
374 CHECKOP(op, ch);
375 dup2(STDOUT_FILENO, STDERR_FILENO);
376 op = OP_TEST;
377 nodb = warnings = 1;
378 break;
379 case ('u'):
380 CHECKOP(op, ch);
381 path_arg = optarg;
382 op = OP_DELETE;
383 break;
384 case ('v'):
385 verb++;
386 break;
387 case ('W'):
388 warnings = 1;
389 break;
390 default:
391 goto usage;
392 }
393
394 argc -= optind;
395 argv += optind;
396
397 if (OP_CONFFILE == op && argc > 0) {
398 fprintf(stderr, "-C: Too many arguments\n");
399 goto usage;
400 }
401
402 exitcode = (int)MANDOCLEVEL_OK;
403 mp = mparse_alloc(MPARSE_AUTO,
404 MANDOCLEVEL_FATAL, NULL, NULL, NULL);
405 mc = mchars_alloc();
406
407 ohash_init(&mpages, 6, &mpages_info);
408 ohash_init(&mlinks, 6, &mlinks_info);
409
410 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
411 /*
412 * Force processing all files.
413 */
414 use_all = 1;
415
416 /*
417 * All of these deal with a specific directory.
418 * Jump into that directory then collect files specified
419 * on the command-line.
420 */
421 if (0 == set_basedir(path_arg))
422 goto out;
423 for (i = 0; i < argc; i++)
424 filescan(argv[i]);
425 if (0 == dbopen(1))
426 goto out;
427 if (OP_TEST != op)
428 dbprune();
429 if (OP_DELETE != op)
430 mpages_merge(mc, mp, 0);
431 dbclose(1);
432 } else {
433 /*
434 * If we have arguments, use them as our manpaths.
435 * If we don't, grok from manpath(1) or however else
436 * manpath_parse() wants to do it.
437 */
438 if (argc > 0) {
439 dirs.paths = mandoc_calloc
440 (argc, sizeof(char *));
441 dirs.sz = (size_t)argc;
442 for (i = 0; i < argc; i++)
443 dirs.paths[i] = mandoc_strdup(argv[i]);
444 } else
445 manpath_parse(&dirs, path_arg, NULL, NULL);
446
447 /*
448 * First scan the tree rooted at a base directory, then
449 * build a new database and finally move it into place.
450 * Ignore zero-length directories and strip trailing
451 * slashes.
452 */
453 for (j = 0; j < dirs.sz; j++) {
454 sz = strlen(dirs.paths[j]);
455 if (sz && '/' == dirs.paths[j][sz - 1])
456 dirs.paths[j][--sz] = '\0';
457 if (0 == sz)
458 continue;
459
460 if (j) {
461 ohash_init(&mpages, 6, &mpages_info);
462 ohash_init(&mlinks, 6, &mlinks_info);
463 }
464
465 if (0 == set_basedir(dirs.paths[j]))
466 goto out;
467 if (0 == treescan())
468 goto out;
469 if (0 == set_basedir(dirs.paths[j]))
470 goto out;
471 if (0 == dbopen(0))
472 goto out;
473
474 mpages_merge(mc, mp, warnings && !use_all);
475 dbclose(0);
476
477 if (j + 1 < dirs.sz) {
478 mpages_free();
479 ohash_delete(&mpages);
480 ohash_delete(&mlinks);
481 }
482 }
483 }
484 out:
485 set_basedir(NULL);
486 manpath_free(&dirs);
487 mchars_free(mc);
488 mparse_free(mp);
489 mpages_free();
490 ohash_delete(&mpages);
491 ohash_delete(&mlinks);
492 return(exitcode);
493 usage:
494 fprintf(stderr, "usage: %s [-anvW] [-C file]\n"
495 " %s [-anvW] dir ...\n"
496 " %s [-nvW] -d dir [file ...]\n"
497 " %s [-nvW] -u dir [file ...]\n"
498 " %s -t file ...\n",
499 progname, progname, progname,
500 progname, progname);
501
502 return((int)MANDOCLEVEL_BADARG);
503 }
504
505 /*
506 * Scan a directory tree rooted at "basedir" for manpages.
507 * We use fts(), scanning directory parts along the way for clues to our
508 * section and architecture.
509 *
510 * If use_all has been specified, grok all files.
511 * If not, sanitise paths to the following:
512 *
513 * [./]man*[/<arch>]/<name>.<section>
514 * or
515 * [./]cat<section>[/<arch>]/<name>.0
516 *
517 * TODO: accomodate for multi-language directories.
518 */
519 static int
520 treescan(void)
521 {
522 FTS *f;
523 FTSENT *ff;
524 int dform;
525 char *sec;
526 const char *dsec, *arch, *cp, *path;
527 const char *argv[2];
528
529 argv[0] = ".";
530 argv[1] = (char *)NULL;
531
532 /*
533 * Walk through all components under the directory, using the
534 * logical descent of files.
535 */
536 f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
537 if (NULL == f) {
538 exitcode = (int)MANDOCLEVEL_SYSERR;
539 say("", NULL);
540 return(0);
541 }
542
543 dsec = arch = NULL;
544 dform = FORM_NONE;
545
546 while (NULL != (ff = fts_read(f))) {
547 path = ff->fts_path + 2;
548 /*
549 * If we're a regular file, add an mlink by using the
550 * stored directory data and handling the filename.
551 * Disallow duplicate (hard-linked) files.
552 */
553 if (FTS_F == ff->fts_info) {
554 if (0 == strcmp(path, MANDOC_DB))
555 continue;
556 if ( ! use_all && ff->fts_level < 2) {
557 if (warnings)
558 say(path, "Extraneous file");
559 continue;
560 } else if (inocheck(ff->fts_statp)) {
561 if (warnings)
562 say(path, "Duplicate file");
563 continue;
564 } else if (NULL == (sec =
565 strrchr(ff->fts_name, '.'))) {
566 if ( ! use_all) {
567 if (warnings)
568 say(path,
569 "No filename suffix");
570 continue;
571 }
572 } else if (0 == strcmp(++sec, "html")) {
573 if (warnings)
574 say(path, "Skip html");
575 continue;
576 } else if (0 == strcmp(sec, "gz")) {
577 if (warnings)
578 say(path, "Skip gz");
579 continue;
580 } else if (0 == strcmp(sec, "ps")) {
581 if (warnings)
582 say(path, "Skip ps");
583 continue;
584 } else if (0 == strcmp(sec, "pdf")) {
585 if (warnings)
586 say(path, "Skip pdf");
587 continue;
588 } else if ( ! use_all &&
589 ((FORM_SRC == dform && strcmp(sec, dsec)) ||
590 (FORM_CAT == dform && strcmp(sec, "0")))) {
591 if (warnings)
592 say(path, "Wrong filename suffix");
593 continue;
594 } else
595 sec[-1] = '\0';
596 mlink_add(dform, path, ff->fts_name, dsec, sec,
597 arch, ff->fts_statp);
598 continue;
599 } else if (FTS_D != ff->fts_info &&
600 FTS_DP != ff->fts_info) {
601 if (warnings)
602 say(path, "Not a regular file");
603 continue;
604 }
605
606 switch (ff->fts_level) {
607 case (0):
608 /* Ignore the root directory. */
609 break;
610 case (1):
611 /*
612 * This might contain manX/ or catX/.
613 * Try to infer this from the name.
614 * If we're not in use_all, enforce it.
615 */
616 dsec = NULL;
617 dform = FORM_NONE;
618 cp = ff->fts_name;
619 if (FTS_DP == ff->fts_info)
620 break;
621
622 if (0 == strncmp(cp, "man", 3)) {
623 dform = FORM_SRC;
624 dsec = cp + 3;
625 } else if (0 == strncmp(cp, "cat", 3)) {
626 dform = FORM_CAT;
627 dsec = cp + 3;
628 }
629
630 if (NULL != dsec || use_all)
631 break;
632
633 if (warnings)
634 say(path, "Unknown directory part");
635 fts_set(f, ff, FTS_SKIP);
636 break;
637 case (2):
638 /*
639 * Possibly our architecture.
640 * If we're descending, keep tabs on it.
641 */
642 arch = NULL;
643 if (FTS_DP != ff->fts_info && NULL != dsec)
644 arch = ff->fts_name;
645 break;
646 default:
647 if (FTS_DP == ff->fts_info || use_all)
648 break;
649 if (warnings)
650 say(path, "Extraneous directory part");
651 fts_set(f, ff, FTS_SKIP);
652 break;
653 }
654 }
655
656 fts_close(f);
657 return(1);
658 }
659
660 /*
661 * Add a file to the file vector.
662 * Do not verify that it's a "valid" looking manpage (we'll do that
663 * later).
664 *
665 * Try to infer the manual section, architecture, and page name from the
666 * path, assuming it looks like
667 *
668 * [./]man*[/<arch>]/<name>.<section>
669 * or
670 * [./]cat<section>[/<arch>]/<name>.0
671 *
672 * Stuff this information directly into the mlink vector.
673 * See treescan() for the fts(3) version of this.
674 */
675 static void
676 filescan(const char *file)
677 {
678 char buf[PATH_MAX];
679 const char *sec, *arch, *name, *dsec;
680 char *p, *start;
681 int dform;
682 struct stat st;
683
684 assert(use_all);
685
686 if (0 == strncmp(file, "./", 2))
687 file += 2;
688
689 if (NULL == realpath(file, buf)) {
690 exitcode = (int)MANDOCLEVEL_BADARG;
691 say(file, NULL);
692 return;
693 } else if (OP_TEST != op && strstr(buf, basedir) != buf) {
694 exitcode = (int)MANDOCLEVEL_BADARG;
695 say("", "%s: outside base directory", buf);
696 return;
697 } else if (-1 == stat(buf, &st)) {
698 exitcode = (int)MANDOCLEVEL_BADARG;
699 say(file, NULL);
700 return;
701 } else if ( ! (S_IFREG & st.st_mode)) {
702 exitcode = (int)MANDOCLEVEL_BADARG;
703 say(file, "Not a regular file");
704 return;
705 } else if (inocheck(&st)) {
706 if (warnings)
707 say(file, "Duplicate file");
708 return;
709 }
710 start = buf + strlen(basedir);
711 sec = arch = name = dsec = NULL;
712 dform = FORM_NONE;
713
714 /*
715 * First try to guess our directory structure.
716 * If we find a separator, try to look for man* or cat*.
717 * If we find one of these and what's underneath is a directory,
718 * assume it's an architecture.
719 */
720 if (NULL != (p = strchr(start, '/'))) {
721 *p++ = '\0';
722 if (0 == strncmp(start, "man", 3)) {
723 dform = FORM_SRC;
724 dsec = start + 3;
725 } else if (0 == strncmp(start, "cat", 3)) {
726 dform = FORM_CAT;
727 dsec = start + 3;
728 }
729
730 start = p;
731 if (NULL != dsec && NULL != (p = strchr(start, '/'))) {
732 *p++ = '\0';
733 arch = start;
734 start = p;
735 }
736 }
737
738 /*
739 * Now check the file suffix.
740 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
741 */
742 p = strrchr(start, '\0');
743 while (p-- > start && '/' != *p && '.' != *p)
744 /* Loop. */ ;
745
746 if ('.' == *p) {
747 *p++ = '\0';
748 sec = p;
749 }
750
751 /*
752 * Now try to parse the name.
753 * Use the filename portion of the path.
754 */
755 name = start;
756 if (NULL != (p = strrchr(start, '/'))) {
757 name = p + 1;
758 *p = '\0';
759 }
760
761 mlink_add(dform, file, name, dsec, sec, arch, &st);
762 }
763
764 static int
765 inocheck(const struct stat *st)
766 {
767 struct inodev inodev;
768 uint32_t hash;
769
770 memset(&inodev, 0, sizeof(inodev));
771 inodev.st_ino = hash = st->st_ino;
772 inodev.st_dev = st->st_dev;
773
774 return(NULL != ohash_find(&mpages, ohash_lookup_memory(
775 &mpages, (char *)&inodev, sizeof(inodev), hash)));
776 }
777
778 static void
779 mlink_add(int dform, const char *file, const char *name, const char *dsec,
780 const char *sec, const char *arch, const struct stat *st)
781 {
782 struct inodev inodev;
783 struct mpage *mpage;
784 struct mlink *mlink;
785 int sform;
786 unsigned int slot;
787
788 assert(NULL != file);
789
790 if (NULL == name)
791 name = "";
792 if (NULL == sec)
793 sec = "";
794 if (NULL == dsec)
795 dsec = "";
796 if (NULL == arch)
797 arch = "";
798
799 if ('0' == *sec) {
800 sec = dsec;
801 sform = FORM_CAT;
802 } else if ('1' <= *sec && '9' >= *sec)
803 sform = FORM_SRC;
804 else
805 sform = FORM_NONE;
806
807 mlink = mandoc_calloc(1, sizeof(struct mlink));
808 strlcpy(mlink->file, file, PATH_MAX);
809 mlink->dform = dform;
810 mlink->sform = sform;
811 mlink->dsec = mandoc_strdup(dsec);
812 mlink->arch = mandoc_strdup(arch);
813 mlink->name = mandoc_strdup(name);
814 mlink->fsec = mandoc_strdup(sec);
815
816 slot = ohash_qlookup(&mlinks, mlink->file);
817 assert(NULL == ohash_find(&mlinks, slot));
818 ohash_insert(&mlinks, slot, mlink);
819
820 inodev.st_ino = st->st_ino;
821 inodev.st_dev = st->st_dev;
822 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
823 sizeof(struct inodev), inodev.st_ino);
824 mpage = ohash_find(&mpages, slot);
825 if (NULL == mpage) {
826 mpage = mandoc_calloc(1, sizeof(struct mpage));
827 mpage->inodev.st_ino = inodev.st_ino;
828 mpage->inodev.st_dev = inodev.st_dev;
829 ohash_insert(&mpages, slot, mpage);
830 } else
831 abort();
832 mpage->mlinks = mlink;
833 }
834
835 static void
836 mlink_free(struct mlink *mlink)
837 {
838
839 free(mlink->dsec);
840 free(mlink->arch);
841 free(mlink->name);
842 free(mlink->fsec);
843 free(mlink);
844 }
845
846 static void
847 mpages_free(void)
848 {
849 struct mpage *mpage;
850 struct mlink *mlink;
851 unsigned int slot;
852
853 mpage = ohash_first(&mpages, &slot);
854 while (NULL != mpage) {
855 while (NULL != (mlink = mpage->mlinks)) {
856 mpage->mlinks = NULL;
857 mlink_free(mlink);
858 }
859 free(mpage->sec);
860 free(mpage->arch);
861 free(mpage->title);
862 free(mpage->desc);
863 free(mpage);
864 mpage = ohash_next(&mpages, &slot);
865 }
866 }
867
868 /*
869 * Run through the files in the global vector "mpages"
870 * and add them to the database specified in "basedir".
871 *
872 * This handles the parsing scheme itself, using the cues of directory
873 * and filename to determine whether the file is parsable or not.
874 */
875 static void
876 mpages_merge(struct mchars *mc, struct mparse *mp, int check_reachable)
877 {
878 struct ohash title_table;
879 struct ohash_info title_info, str_info;
880 char buf[PATH_MAX];
881 struct mpage *mpage;
882 struct mdoc *mdoc;
883 struct man *man;
884 struct title *title_entry;
885 char *bufp, *title_str;
886 const char *cp;
887 size_t sz;
888 int match;
889 unsigned int pslot, tslot;
890 enum mandoclevel lvl;
891
892 str_info.alloc = hash_alloc;
893 str_info.halloc = hash_halloc;
894 str_info.hfree = hash_free;
895 str_info.key_offset = offsetof(struct str, key);
896
897 if (check_reachable) {
898 title_info.alloc = hash_alloc;
899 title_info.halloc = hash_halloc;
900 title_info.hfree = hash_free;
901 title_info.key_offset = offsetof(struct title, title);
902 ohash_init(&title_table, 6, &title_info);
903 }
904
905 mpage = ohash_first(&mpages, &pslot);
906 while (NULL != mpage) {
907 /*
908 * If we're a catpage (as defined by our path), then see
909 * if a manpage exists by the same name (ignoring the
910 * suffix).
911 * If it does, then we want to use it instead of our
912 * own.
913 */
914 if ( ! use_all && FORM_CAT == mpage->mlinks->dform) {
915 sz = strlcpy(buf, mpage->mlinks->file, PATH_MAX);
916 if (sz >= PATH_MAX) {
917 if (warnings)
918 say(mpage->mlinks->file,
919 "Filename too long");
920 mpage = ohash_next(&mpages, &pslot);
921 continue;
922 }
923 bufp = strstr(buf, "cat");
924 assert(NULL != bufp);
925 memcpy(bufp, "man", 3);
926 if (NULL != (bufp = strrchr(buf, '.')))
927 *++bufp = '\0';
928 strlcat(buf, mpage->mlinks->dsec, PATH_MAX);
929 if (NULL != ohash_find(&mlinks,
930 ohash_qlookup(&mlinks, buf))) {
931 if (warnings)
932 say(mpage->mlinks->file, "Man "
933 "source exists: %s", buf);
934 mpage = ohash_next(&mpages, &pslot);
935 continue;
936 }
937 }
938
939 ohash_init(&strings, 6, &str_info);
940 mparse_reset(mp);
941 mdoc = NULL;
942 man = NULL;
943 match = 1;
944
945 /*
946 * Try interpreting the file as mdoc(7) or man(7)
947 * source code, unless it is already known to be
948 * formatted. Fall back to formatted mode.
949 */
950 if (FORM_CAT != mpage->mlinks->dform ||
951 FORM_CAT != mpage->mlinks->sform) {
952 lvl = mparse_readfd(mp, -1, mpage->mlinks->file);
953 if (lvl < MANDOCLEVEL_FATAL)
954 mparse_result(mp, &mdoc, &man);
955 }
956
957 if (NULL != mdoc) {
958 mpage->form = FORM_SRC;
959 mpage->sec =
960 mandoc_strdup(mdoc_meta(mdoc)->msec);
961 mpage->arch = mdoc_meta(mdoc)->arch;
962 if (NULL == mpage->arch)
963 mpage->arch = "";
964 mpage->arch = mandoc_strdup(mpage->arch);
965 mpage->title =
966 mandoc_strdup(mdoc_meta(mdoc)->title);
967 } else if (NULL != man) {
968 mpage->form = FORM_SRC;
969 mpage->sec =
970 mandoc_strdup(man_meta(man)->msec);
971 mpage->arch =
972 mandoc_strdup(mpage->mlinks->arch);
973 mpage->title =
974 mandoc_strdup(man_meta(man)->title);
975 } else {
976 mpage->form = FORM_CAT;
977 mpage->sec =
978 mandoc_strdup(mpage->mlinks->dsec);
979 mpage->arch =
980 mandoc_strdup(mpage->mlinks->arch);
981 mpage->title =
982 mandoc_strdup(mpage->mlinks->name);
983 }
984
985 /*
986 * Check whether the manual section given in a file
987 * agrees with the directory where the file is located.
988 * Some manuals have suffixes like (3p) on their
989 * section number either inside the file or in the
990 * directory name, some are linked into more than one
991 * section, like encrypt(1) = makekey(8). Do not skip
992 * manuals for such reasons.
993 */
994 if (warnings && !use_all && FORM_SRC == mpage->form &&
995 strcasecmp(mpage->sec, mpage->mlinks->dsec)) {
996 match = 0;
997 say(mpage->mlinks->file, "Section \"%s\" "
998 "manual in %s directory",
999 mpage->sec, mpage->mlinks->dsec);
1000 }
1001
1002 /*
1003 * Manual page directories exist for each kernel
1004 * architecture as returned by machine(1).
1005 * However, many manuals only depend on the
1006 * application architecture as returned by arch(1).
1007 * For example, some (2/ARM) manuals are shared
1008 * across the "armish" and "zaurus" kernel
1009 * architectures.
1010 * A few manuals are even shared across completely
1011 * different architectures, for example fdformat(1)
1012 * on amd64, i386, sparc, and sparc64.
1013 * Thus, warn about architecture mismatches,
1014 * but don't skip manuals for this reason.
1015 */
1016 if (warnings && !use_all &&
1017 strcasecmp(mpage->arch, mpage->mlinks->arch)) {
1018 match = 0;
1019 say(mpage->mlinks->file, "Architecture \"%s\" "
1020 "manual in \"%s\" directory",
1021 mpage->arch, mpage->mlinks->arch);
1022 }
1023 if (warnings && !use_all &&
1024 strcasecmp(mpage->title, mpage->mlinks->name))
1025 match = 0;
1026
1027 putkey(mpage, mpage->mlinks->name, TYPE_Nm);
1028
1029 if (NULL != mdoc) {
1030 if (NULL != (cp = mdoc_meta(mdoc)->name))
1031 putkey(mpage, cp, TYPE_Nm);
1032 assert(NULL == mpage->desc);
1033 parse_mdoc(mpage, mdoc_node(mdoc));
1034 putkey(mpage, NULL != mpage->desc ?
1035 mpage->desc : mpage->mlinks->name, TYPE_Nd);
1036 } else if (NULL != man)
1037 parse_man(mpage, man_node(man));
1038 else
1039 parse_cat(mpage);
1040
1041 /*
1042 * Build a title string for the file. If it matches
1043 * the location of the file, remember the title as
1044 * found; else, remember it as missing.
1045 */
1046
1047 if (check_reachable) {
1048 if (-1 == asprintf(&title_str, "%s(%s%s%s)",
1049 mpage->title, mpage->sec,
1050 '\0' == *mpage->arch ? "" : "/",
1051 mpage->arch)) {
1052 perror(NULL);
1053 exit((int)MANDOCLEVEL_SYSERR);
1054 }
1055 tslot = ohash_qlookup(&title_table, title_str);
1056 title_entry = ohash_find(&title_table, tslot);
1057 if (NULL == title_entry) {
1058 title_entry = mandoc_malloc(
1059 sizeof(struct title));
1060 title_entry->title = title_str;
1061 title_entry->file = mandoc_strdup(
1062 match ? "" : mpage->mlinks->file);
1063 ohash_insert(&title_table, tslot,
1064 title_entry);
1065 } else {
1066 if (match)
1067 *title_entry->file = '\0';
1068 free(title_str);
1069 }
1070 }
1071
1072 dbindex(mpage, mc);
1073 ohash_delete(&strings);
1074 mpage = ohash_next(&mpages, &pslot);
1075 }
1076
1077 if (check_reachable) {
1078 title_entry = ohash_first(&title_table, &tslot);
1079 while (NULL != title_entry) {
1080 if ('\0' != *title_entry->file)
1081 say(title_entry->file,
1082 "Probably unreachable, title is %s",
1083 title_entry->title);
1084 free(title_entry->title);
1085 free(title_entry->file);
1086 free(title_entry);
1087 title_entry = ohash_next(&title_table, &tslot);
1088 }
1089 ohash_delete(&title_table);
1090 }
1091 }
1092
1093 static void
1094 parse_cat(struct mpage *mpage)
1095 {
1096 FILE *stream;
1097 char *line, *p, *title;
1098 size_t len, plen, titlesz;
1099
1100 if (NULL == (stream = fopen(mpage->mlinks->file, "r"))) {
1101 if (warnings)
1102 say(mpage->mlinks->file, NULL);
1103 return;
1104 }
1105
1106 /* Skip to first blank line. */
1107
1108 while (NULL != (line = fgetln(stream, &len)))
1109 if ('\n' == *line)
1110 break;
1111
1112 /*
1113 * Assume the first line that is not indented
1114 * is the first section header. Skip to it.
1115 */
1116
1117 while (NULL != (line = fgetln(stream, &len)))
1118 if ('\n' != *line && ' ' != *line)
1119 break;
1120
1121 /*
1122 * Read up until the next section into a buffer.
1123 * Strip the leading and trailing newline from each read line,
1124 * appending a trailing space.
1125 * Ignore empty (whitespace-only) lines.
1126 */
1127
1128 titlesz = 0;
1129 title = NULL;
1130
1131 while (NULL != (line = fgetln(stream, &len))) {
1132 if (' ' != *line || '\n' != line[len - 1])
1133 break;
1134 while (len > 0 && isspace((unsigned char)*line)) {
1135 line++;
1136 len--;
1137 }
1138 if (1 == len)
1139 continue;
1140 title = mandoc_realloc(title, titlesz + len);
1141 memcpy(title + titlesz, line, len);
1142 titlesz += len;
1143 title[titlesz - 1] = ' ';
1144 }
1145
1146 /*
1147 * If no page content can be found, or the input line
1148 * is already the next section header, or there is no
1149 * trailing newline, reuse the page title as the page
1150 * description.
1151 */
1152
1153 if (NULL == title || '\0' == *title) {
1154 if (warnings)
1155 say(mpage->mlinks->file,
1156 "Cannot find NAME section");
1157 assert(NULL == mpage->desc);
1158 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1159 putkey(mpage, mpage->mlinks->name, TYPE_Nd);
1160 fclose(stream);
1161 free(title);
1162 return;
1163 }
1164
1165 title = mandoc_realloc(title, titlesz + 1);
1166 title[titlesz] = '\0';
1167
1168 /*
1169 * Skip to the first dash.
1170 * Use the remaining line as the description (no more than 70
1171 * bytes).
1172 */
1173
1174 if (NULL != (p = strstr(title, "- "))) {
1175 for (p += 2; ' ' == *p || '\b' == *p; p++)
1176 /* Skip to next word. */ ;
1177 } else {
1178 if (warnings)
1179 say(mpage->mlinks->file,
1180 "No dash in title line");
1181 p = title;
1182 }
1183
1184 plen = strlen(p);
1185
1186 /* Strip backspace-encoding from line. */
1187
1188 while (NULL != (line = memchr(p, '\b', plen))) {
1189 len = line - p;
1190 if (0 == len) {
1191 memmove(line, line + 1, plen--);
1192 continue;
1193 }
1194 memmove(line - 1, line + 1, plen - len);
1195 plen -= 2;
1196 }
1197
1198 assert(NULL == mpage->desc);
1199 mpage->desc = mandoc_strdup(p);
1200 putkey(mpage, mpage->desc, TYPE_Nd);
1201 fclose(stream);
1202 free(title);
1203 }
1204
1205 /*
1206 * Put a type/word pair into the word database for this particular file.
1207 */
1208 static void
1209 putkey(const struct mpage *mpage, const char *value, uint64_t type)
1210 {
1211
1212 assert(NULL != value);
1213 putkeys(mpage, value, strlen(value), type);
1214 }
1215
1216 /*
1217 * Grok all nodes at or below a certain mdoc node into putkey().
1218 */
1219 static void
1220 putmdockey(const struct mpage *mpage,
1221 const struct mdoc_node *n, uint64_t m)
1222 {
1223
1224 for ( ; NULL != n; n = n->next) {
1225 if (NULL != n->child)
1226 putmdockey(mpage, n->child, m);
1227 if (MDOC_TEXT == n->type)
1228 putkey(mpage, n->string, m);
1229 }
1230 }
1231
1232 static void
1233 parse_man(struct mpage *mpage, const struct man_node *n)
1234 {
1235 const struct man_node *head, *body;
1236 char *start, *sv, *title;
1237 char byte;
1238 size_t sz, titlesz;
1239
1240 if (NULL == n)
1241 return;
1242
1243 /*
1244 * We're only searching for one thing: the first text child in
1245 * the BODY of a NAME section. Since we don't keep track of
1246 * sections in -man, run some hoops to find out whether we're in
1247 * the correct section or not.
1248 */
1249
1250 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1251 body = n;
1252 assert(body->parent);
1253 if (NULL != (head = body->parent->head) &&
1254 1 == head->nchild &&
1255 NULL != (head = (head->child)) &&
1256 MAN_TEXT == head->type &&
1257 0 == strcmp(head->string, "NAME") &&
1258 NULL != (body = body->child) &&
1259 MAN_TEXT == body->type) {
1260
1261 title = NULL;
1262 titlesz = 0;
1263
1264 /*
1265 * Suck the entire NAME section into memory.
1266 * Yes, we might run away.
1267 * But too many manuals have big, spread-out
1268 * NAME sections over many lines.
1269 */
1270
1271 for ( ; NULL != body; body = body->next) {
1272 if (MAN_TEXT != body->type)
1273 break;
1274 if (0 == (sz = strlen(body->string)))
1275 continue;
1276 title = mandoc_realloc
1277 (title, titlesz + sz + 1);
1278 memcpy(title + titlesz, body->string, sz);
1279 titlesz += sz + 1;
1280 title[titlesz - 1] = ' ';
1281 }
1282 if (NULL == title)
1283 return;
1284
1285 title = mandoc_realloc(title, titlesz + 1);
1286 title[titlesz] = '\0';
1287
1288 /* Skip leading space. */
1289
1290 sv = title;
1291 while (isspace((unsigned char)*sv))
1292 sv++;
1293
1294 if (0 == (sz = strlen(sv))) {
1295 free(title);
1296 return;
1297 }
1298
1299 /* Erase trailing space. */
1300
1301 start = &sv[sz - 1];
1302 while (start > sv && isspace((unsigned char)*start))
1303 *start-- = '\0';
1304
1305 if (start == sv) {
1306 free(title);
1307 return;
1308 }
1309
1310 start = sv;
1311
1312 /*
1313 * Go through a special heuristic dance here.
1314 * Conventionally, one or more manual names are
1315 * comma-specified prior to a whitespace, then a
1316 * dash, then a description. Try to puzzle out
1317 * the name parts here.
1318 */
1319
1320 for ( ;; ) {
1321 sz = strcspn(start, " ,");
1322 if ('\0' == start[sz])
1323 break;
1324
1325 byte = start[sz];
1326 start[sz] = '\0';
1327
1328 putkey(mpage, start, TYPE_Nm);
1329
1330 if (' ' == byte) {
1331 start += sz + 1;
1332 break;
1333 }
1334
1335 assert(',' == byte);
1336 start += sz + 1;
1337 while (' ' == *start)
1338 start++;
1339 }
1340
1341 if (sv == start) {
1342 putkey(mpage, start, TYPE_Nm);
1343 free(title);
1344 return;
1345 }
1346
1347 while (isspace((unsigned char)*start))
1348 start++;
1349
1350 if (0 == strncmp(start, "-", 1))
1351 start += 1;
1352 else if (0 == strncmp(start, "\\-\\-", 4))
1353 start += 4;
1354 else if (0 == strncmp(start, "\\-", 2))
1355 start += 2;
1356 else if (0 == strncmp(start, "\\(en", 4))
1357 start += 4;
1358 else if (0 == strncmp(start, "\\(em", 4))
1359 start += 4;
1360
1361 while (' ' == *start)
1362 start++;
1363
1364 assert(NULL == mpage->desc);
1365 mpage->desc = mandoc_strdup(start);
1366 putkey(mpage, mpage->desc, TYPE_Nd);
1367 free(title);
1368 return;
1369 }
1370 }
1371
1372 for (n = n->child; n; n = n->next) {
1373 if (NULL != mpage->desc)
1374 break;
1375 parse_man(mpage, n);
1376 }
1377 }
1378
1379 static void
1380 parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1381 {
1382
1383 assert(NULL != n);
1384 for (n = n->child; NULL != n; n = n->next) {
1385 switch (n->type) {
1386 case (MDOC_ELEM):
1387 /* FALLTHROUGH */
1388 case (MDOC_BLOCK):
1389 /* FALLTHROUGH */
1390 case (MDOC_HEAD):
1391 /* FALLTHROUGH */
1392 case (MDOC_BODY):
1393 /* FALLTHROUGH */
1394 case (MDOC_TAIL):
1395 if (NULL != mdocs[n->tok].fp)
1396 if (0 == (*mdocs[n->tok].fp)(mpage, n))
1397 break;
1398 if (mdocs[n->tok].mask)
1399 putmdockey(mpage, n->child,
1400 mdocs[n->tok].mask);
1401 break;
1402 default:
1403 assert(MDOC_ROOT != n->type);
1404 continue;
1405 }
1406 if (NULL != n->child)
1407 parse_mdoc(mpage, n);
1408 }
1409 }
1410
1411 static int
1412 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1413 {
1414 const char *start, *end;
1415 size_t sz;
1416
1417 if (SEC_SYNOPSIS != n->sec ||
1418 NULL == (n = n->child) ||
1419 MDOC_TEXT != n->type)
1420 return(0);
1421
1422 /*
1423 * Only consider those `Fd' macro fields that begin with an
1424 * "inclusion" token (versus, e.g., #define).
1425 */
1426
1427 if (strcmp("#include", n->string))
1428 return(0);
1429
1430 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1431 return(0);
1432
1433 /*
1434 * Strip away the enclosing angle brackets and make sure we're
1435 * not zero-length.
1436 */
1437
1438 start = n->string;
1439 if ('<' == *start || '"' == *start)
1440 start++;
1441
1442 if (0 == (sz = strlen(start)))
1443 return(0);
1444
1445 end = &start[(int)sz - 1];
1446 if ('>' == *end || '"' == *end)
1447 end--;
1448
1449 if (end > start)
1450 putkeys(mpage, start, end - start + 1, TYPE_In);
1451 return(1);
1452 }
1453
1454 static int
1455 parse_mdoc_In(struct mpage *mpage, const struct mdoc_node *n)
1456 {
1457
1458 if (NULL != n->child && MDOC_TEXT == n->child->type)
1459 return(0);
1460
1461 putkey(mpage, n->child->string, TYPE_In);
1462 return(1);
1463 }
1464
1465 static int
1466 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1467 {
1468 const char *cp;
1469
1470 if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1471 return(0);
1472
1473 /*
1474 * Parse: .Fn "struct type *name" "char *arg".
1475 * First strip away pointer symbol.
1476 * Then store the function name, then type.
1477 * Finally, store the arguments.
1478 */
1479
1480 if (NULL == (cp = strrchr(n->string, ' ')))
1481 cp = n->string;
1482
1483 while ('*' == *cp)
1484 cp++;
1485
1486 putkey(mpage, cp, TYPE_Fn);
1487
1488 if (n->string < cp)
1489 putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1490
1491 for (n = n->next; NULL != n; n = n->next)
1492 if (MDOC_TEXT == n->type)
1493 putkey(mpage, n->string, TYPE_Fa);
1494
1495 return(0);
1496 }
1497
1498 static int
1499 parse_mdoc_St(struct mpage *mpage, const struct mdoc_node *n)
1500 {
1501
1502 if (NULL == n->child || MDOC_TEXT != n->child->type)
1503 return(0);
1504
1505 putkey(mpage, n->child->string, TYPE_St);
1506 return(1);
1507 }
1508
1509 static int
1510 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1511 {
1512 char *cp;
1513
1514 if (NULL == (n = n->child))
1515 return(0);
1516
1517 if (NULL == n->next) {
1518 putkey(mpage, n->string, TYPE_Xr);
1519 return(0);
1520 }
1521
1522 if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) {
1523 perror(NULL);
1524 exit((int)MANDOCLEVEL_SYSERR);
1525 }
1526 putkey(mpage, cp, TYPE_Xr);
1527 free(cp);
1528 return(0);
1529 }
1530
1531 static int
1532 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1533 {
1534 size_t sz;
1535
1536 if (MDOC_BODY != n->type)
1537 return(0);
1538
1539 /*
1540 * Special-case the `Nd' because we need to put the description
1541 * into the document table.
1542 */
1543
1544 for (n = n->child; NULL != n; n = n->next) {
1545 if (MDOC_TEXT == n->type) {
1546 if (NULL != mpage->desc) {
1547 sz = strlen(mpage->desc) +
1548 strlen(n->string) + 2;
1549 mpage->desc = mandoc_realloc(
1550 mpage->desc, sz);
1551 strlcat(mpage->desc, " ", sz);
1552 strlcat(mpage->desc, n->string, sz);
1553 } else
1554 mpage->desc = mandoc_strdup(n->string);
1555 }
1556 if (NULL != n->child)
1557 parse_mdoc_Nd(mpage, n);
1558 }
1559 return(1);
1560 }
1561
1562 static int
1563 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1564 {
1565
1566 if (SEC_NAME == n->sec)
1567 return(1);
1568 else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
1569 return(0);
1570
1571 return(1);
1572 }
1573
1574 static int
1575 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1576 {
1577
1578 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1579 }
1580
1581 static int
1582 parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1583 {
1584
1585 return(MDOC_HEAD == n->type);
1586 }
1587
1588 static int
1589 parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1590 {
1591
1592 return(MDOC_BODY == n->type);
1593 }
1594
1595 /*
1596 * Add a string to the hash table for the current manual.
1597 * Each string has a bitmask telling which macros it belongs to.
1598 * When we finish the manual, we'll dump the table.
1599 */
1600 static void
1601 putkeys(const struct mpage *mpage,
1602 const char *cp, size_t sz, uint64_t v)
1603 {
1604 struct str *s;
1605 unsigned int slot;
1606 const char *end;
1607
1608 if (0 == sz)
1609 return;
1610
1611 end = cp + sz;
1612 slot = ohash_qlookupi(&strings, cp, &end);
1613 s = ohash_find(&strings, slot);
1614
1615 if (NULL != s && mpage == s->mpage) {
1616 s->mask |= v;
1617 return;
1618 } else if (NULL == s) {
1619 s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1620 memcpy(s->key, cp, sz);
1621 ohash_insert(&strings, slot, s);
1622 }
1623 s->mpage = mpage;
1624 s->mask = v;
1625 }
1626
1627 /*
1628 * Take a Unicode codepoint and produce its UTF-8 encoding.
1629 * This isn't the best way to do this, but it works.
1630 * The magic numbers are from the UTF-8 packaging.
1631 * They're not as scary as they seem: read the UTF-8 spec for details.
1632 */
1633 static size_t
1634 utf8(unsigned int cp, char out[7])
1635 {
1636 size_t rc;
1637
1638 rc = 0;
1639 if (cp <= 0x0000007F) {
1640 rc = 1;
1641 out[0] = (char)cp;
1642 } else if (cp <= 0x000007FF) {
1643 rc = 2;
1644 out[0] = (cp >> 6 & 31) | 192;
1645 out[1] = (cp & 63) | 128;
1646 } else if (cp <= 0x0000FFFF) {
1647 rc = 3;
1648 out[0] = (cp >> 12 & 15) | 224;
1649 out[1] = (cp >> 6 & 63) | 128;
1650 out[2] = (cp & 63) | 128;
1651 } else if (cp <= 0x001FFFFF) {
1652 rc = 4;
1653 out[0] = (cp >> 18 & 7) | 240;
1654 out[1] = (cp >> 12 & 63) | 128;
1655 out[2] = (cp >> 6 & 63) | 128;
1656 out[3] = (cp & 63) | 128;
1657 } else if (cp <= 0x03FFFFFF) {
1658 rc = 5;
1659 out[0] = (cp >> 24 & 3) | 248;
1660 out[1] = (cp >> 18 & 63) | 128;
1661 out[2] = (cp >> 12 & 63) | 128;
1662 out[3] = (cp >> 6 & 63) | 128;
1663 out[4] = (cp & 63) | 128;
1664 } else if (cp <= 0x7FFFFFFF) {
1665 rc = 6;
1666 out[0] = (cp >> 30 & 1) | 252;
1667 out[1] = (cp >> 24 & 63) | 128;
1668 out[2] = (cp >> 18 & 63) | 128;
1669 out[3] = (cp >> 12 & 63) | 128;
1670 out[4] = (cp >> 6 & 63) | 128;
1671 out[5] = (cp & 63) | 128;
1672 } else
1673 return(0);
1674
1675 out[rc] = '\0';
1676 return(rc);
1677 }
1678
1679 /*
1680 * Store the UTF-8 version of a key, or alias the pointer if the key has
1681 * no UTF-8 transcription marks in it.
1682 */
1683 static void
1684 utf8key(struct mchars *mc, struct str *key)
1685 {
1686 size_t sz, bsz, pos;
1687 char utfbuf[7], res[5];
1688 char *buf;
1689 const char *seq, *cpp, *val;
1690 int len, u;
1691 enum mandoc_esc esc;
1692
1693 assert(NULL == key->utf8);
1694
1695 res[0] = '\\';
1696 res[1] = '\t';
1697 res[2] = ASCII_NBRSP;
1698 res[3] = ASCII_HYPH;
1699 res[4] = '\0';
1700
1701 val = key->key;
1702 bsz = strlen(val);
1703
1704 /*
1705 * Pre-check: if we have no stop-characters, then set the
1706 * pointer as ourselvse and get out of here.
1707 */
1708 if (strcspn(val, res) == bsz) {
1709 key->utf8 = key->key;
1710 return;
1711 }
1712
1713 /* Pre-allocate by the length of the input */
1714
1715 buf = mandoc_malloc(++bsz);
1716 pos = 0;
1717
1718 while ('\0' != *val) {
1719 /*
1720 * Halt on the first escape sequence.
1721 * This also halts on the end of string, in which case
1722 * we just copy, fallthrough, and exit the loop.
1723 */
1724 if ((sz = strcspn(val, res)) > 0) {
1725 memcpy(&buf[pos], val, sz);
1726 pos += sz;
1727 val += sz;
1728 }
1729
1730 if (ASCII_HYPH == *val) {
1731 buf[pos++] = '-';
1732 val++;
1733 continue;
1734 } else if ('\t' == *val || ASCII_NBRSP == *val) {
1735 buf[pos++] = ' ';
1736 val++;
1737 continue;
1738 } else if ('\\' != *val)
1739 break;
1740
1741 /* Read past the slash. */
1742
1743 val++;
1744 u = 0;
1745
1746 /*
1747 * Parse the escape sequence and see if it's a
1748 * predefined character or special character.
1749 */
1750 esc = mandoc_escape
1751 ((const char **)&val, &seq, &len);
1752 if (ESCAPE_ERROR == esc)
1753 break;
1754
1755 if (ESCAPE_SPECIAL != esc)
1756 continue;
1757 if (0 == (u = mchars_spec2cp(mc, seq, len)))
1758 continue;
1759
1760 /*
1761 * If we have a Unicode codepoint, try to convert that
1762 * to a UTF-8 byte string.
1763 */
1764 cpp = utfbuf;
1765 if (0 == (sz = utf8(u, utfbuf)))
1766 continue;
1767
1768 /* Copy the rendered glyph into the stream. */
1769
1770 sz = strlen(cpp);
1771 bsz += sz;
1772
1773 buf = mandoc_realloc(buf, bsz);
1774
1775 memcpy(&buf[pos], cpp, sz);
1776 pos += sz;
1777 }
1778
1779 buf[pos] = '\0';
1780 key->utf8 = buf;
1781 }
1782
1783 /*
1784 * Flush the current page's terms (and their bits) into the database.
1785 * Wrap the entire set of additions in a transaction to make sqlite be a
1786 * little faster.
1787 * Also, UTF-8-encode the description at the last possible moment.
1788 */
1789 static void
1790 dbindex(const struct mpage *mpage, struct mchars *mc)
1791 {
1792 struct str *key;
1793 const char *desc;
1794 int64_t recno;
1795 size_t i;
1796 unsigned int slot;
1797
1798 if (verb)
1799 say(mpage->mlinks->file, "Adding to index");
1800
1801 if (nodb)
1802 return;
1803
1804 desc = "";
1805 if (NULL != mpage->desc && '\0' != *mpage->desc) {
1806 key = ohash_find(&strings,
1807 ohash_qlookup(&strings, mpage->desc));
1808 assert(NULL != key);
1809 if (NULL == key->utf8)
1810 utf8key(mc, key);
1811 desc = key->utf8;
1812 }
1813
1814 SQL_EXEC("BEGIN TRANSACTION");
1815
1816 i = 1;
1817 /*
1818 * XXX The following three lines are obsolete
1819 * and only kept for backward compatibility
1820 * until apropos(1) and friends have caught up.
1821 */
1822 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->file);
1823 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->dsec);
1824 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->arch);
1825 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, desc);
1826 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
1827 SQL_STEP(stmts[STMT_INSERT_PAGE]);
1828 recno = sqlite3_last_insert_rowid(db);
1829 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1830
1831 i = 1;
1832 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mpage->mlinks->dsec);
1833 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mpage->mlinks->arch);
1834 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mpage->mlinks->file);
1835 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, recno);
1836 SQL_STEP(stmts[STMT_INSERT_LINK]);
1837 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1838
1839 for (key = ohash_first(&strings, &slot); NULL != key;
1840 key = ohash_next(&strings, &slot)) {
1841 assert(key->mpage == mpage);
1842 if (NULL == key->utf8)
1843 utf8key(mc, key);
1844 i = 1;
1845 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1846 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->utf8);
1847 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
1848 SQL_STEP(stmts[STMT_INSERT_KEY]);
1849 sqlite3_reset(stmts[STMT_INSERT_KEY]);
1850 if (key->utf8 != key->key)
1851 free(key->utf8);
1852 free(key);
1853 }
1854
1855 SQL_EXEC("END TRANSACTION");
1856 }
1857
1858 static void
1859 dbprune(void)
1860 {
1861 struct mpage *mpage;
1862 struct mlink *mlink;
1863 size_t i;
1864 unsigned int slot;
1865
1866 if (nodb)
1867 return;
1868
1869 mpage = ohash_first(&mpages, &slot);
1870 while (NULL != mpage) {
1871 mlink = mpage->mlinks;
1872 i = 1;
1873 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE], i, mlink->file);
1874 SQL_STEP(stmts[STMT_DELETE_PAGE]);
1875 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
1876 if (verb)
1877 say(mlink->file, "Deleted from index");
1878 mpage = ohash_next(&mpages, &slot);
1879 }
1880 }
1881
1882 /*
1883 * Close an existing database and its prepared statements.
1884 * If "real" is not set, rename the temporary file into the real one.
1885 */
1886 static void
1887 dbclose(int real)
1888 {
1889 size_t i;
1890
1891 if (nodb)
1892 return;
1893
1894 for (i = 0; i < STMT__MAX; i++) {
1895 sqlite3_finalize(stmts[i]);
1896 stmts[i] = NULL;
1897 }
1898
1899 sqlite3_close(db);
1900 db = NULL;
1901
1902 if (real)
1903 return;
1904
1905 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
1906 exitcode = (int)MANDOCLEVEL_SYSERR;
1907 say(MANDOC_DB, NULL);
1908 }
1909 }
1910
1911 /*
1912 * This is straightforward stuff.
1913 * Open a database connection to a "temporary" database, then open a set
1914 * of prepared statements we'll use over and over again.
1915 * If "real" is set, we use the existing database; if not, we truncate a
1916 * temporary one.
1917 * Must be matched by dbclose().
1918 */
1919 static int
1920 dbopen(int real)
1921 {
1922 const char *file, *sql;
1923 int rc, ofl;
1924
1925 if (nodb)
1926 return(1);
1927
1928 ofl = SQLITE_OPEN_READWRITE;
1929 if (0 == real) {
1930 file = MANDOC_DB "~";
1931 if (-1 == remove(file) && ENOENT != errno) {
1932 exitcode = (int)MANDOCLEVEL_SYSERR;
1933 say(file, NULL);
1934 return(0);
1935 }
1936 ofl |= SQLITE_OPEN_EXCLUSIVE;
1937 } else
1938 file = MANDOC_DB;
1939
1940 rc = sqlite3_open_v2(file, &db, ofl, NULL);
1941 if (SQLITE_OK == rc)
1942 goto prepare_statements;
1943 if (SQLITE_CANTOPEN != rc) {
1944 exitcode = (int)MANDOCLEVEL_SYSERR;
1945 say(file, NULL);
1946 return(0);
1947 }
1948
1949 sqlite3_close(db);
1950 db = NULL;
1951
1952 if (SQLITE_OK != (rc = sqlite3_open(file, &db))) {
1953 exitcode = (int)MANDOCLEVEL_SYSERR;
1954 say(file, NULL);
1955 return(0);
1956 }
1957
1958 /*
1959 * XXX The first three columns in table mpages are obsolete
1960 * and only kept for backward compatibility
1961 * until apropos(1) and friends have caught up.
1962 */
1963 sql = "CREATE TABLE \"mpages\" (\n"
1964 " \"file\" TEXT NOT NULL,\n"
1965 " \"sec\" TEXT NOT NULL,\n"
1966 " \"arch\" TEXT NOT NULL,\n"
1967 " \"desc\" TEXT NOT NULL,\n"
1968 " \"form\" INTEGER NOT NULL,\n"
1969 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1970 ");\n"
1971 "\n"
1972 "CREATE TABLE \"mlinks\" (\n"
1973 " \"sec\" TEXT NOT NULL,\n"
1974 " \"arch\" TEXT NOT NULL,\n"
1975 " \"name\" TEXT NOT NULL,\n"
1976 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1977 "ON DELETE CASCADE,\n"
1978 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1979 ");\n"
1980 "\n"
1981 "CREATE TABLE \"keys\" (\n"
1982 " \"bits\" INTEGER NOT NULL,\n"
1983 " \"key\" TEXT NOT NULL,\n"
1984 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1985 "ON DELETE CASCADE,\n"
1986 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1987 ");\n"
1988 "\n"
1989 "CREATE INDEX \"key_index\" ON keys (key);\n";
1990
1991 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1992 exitcode = (int)MANDOCLEVEL_SYSERR;
1993 say(file, "%s", sqlite3_errmsg(db));
1994 return(0);
1995 }
1996
1997 prepare_statements:
1998 SQL_EXEC("PRAGMA foreign_keys = ON");
1999 sql = "DELETE FROM mpages where file=?";
2000 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2001 sql = "INSERT INTO mpages "
2002 "(file,sec,arch,desc,form) VALUES (?,?,?,?,?)";
2003 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2004 sql = "INSERT INTO mlinks "
2005 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2006 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2007 sql = "INSERT INTO keys "
2008 "(bits,key,pageid) VALUES (?,?,?)";
2009 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2010
2011 #ifndef __APPLE__
2012 /*
2013 * When opening a new database, we can turn off
2014 * synchronous mode for much better performance.
2015 */
2016
2017 if (real)
2018 SQL_EXEC("PRAGMA synchronous = OFF");
2019 #endif
2020
2021 return(1);
2022 }
2023
2024 static void *
2025 hash_halloc(size_t sz, void *arg)
2026 {
2027
2028 return(mandoc_calloc(sz, 1));
2029 }
2030
2031 static void *
2032 hash_alloc(size_t sz, void *arg)
2033 {
2034
2035 return(mandoc_malloc(sz));
2036 }
2037
2038 static void
2039 hash_free(void *p, size_t sz, void *arg)
2040 {
2041
2042 free(p);
2043 }
2044
2045 static int
2046 set_basedir(const char *targetdir)
2047 {
2048 static char startdir[PATH_MAX];
2049 static int fd;
2050
2051 /*
2052 * Remember where we started by keeping a fd open to the origin
2053 * path component: throughout this utility, we chdir() a lot to
2054 * handle relative paths, and by doing this, we can return to
2055 * the starting point.
2056 */
2057 if ('\0' == *startdir) {
2058 if (NULL == getcwd(startdir, PATH_MAX)) {
2059 exitcode = (int)MANDOCLEVEL_SYSERR;
2060 if (NULL != targetdir)
2061 say(".", NULL);
2062 return(0);
2063 }
2064 if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
2065 exitcode = (int)MANDOCLEVEL_SYSERR;
2066 say(startdir, NULL);
2067 return(0);
2068 }
2069 if (NULL == targetdir)
2070 targetdir = startdir;
2071 } else {
2072 if (-1 == fd)
2073 return(0);
2074 if (-1 == fchdir(fd)) {
2075 close(fd);
2076 basedir[0] = '\0';
2077 exitcode = (int)MANDOCLEVEL_SYSERR;
2078 say(startdir, NULL);
2079 return(0);
2080 }
2081 if (NULL == targetdir) {
2082 close(fd);
2083 return(1);
2084 }
2085 }
2086 if (NULL == realpath(targetdir, basedir)) {
2087 basedir[0] = '\0';
2088 exitcode = (int)MANDOCLEVEL_BADARG;
2089 say(targetdir, NULL);
2090 return(0);
2091 } else if (-1 == chdir(basedir)) {
2092 exitcode = (int)MANDOCLEVEL_BADARG;
2093 say("", NULL);
2094 return(0);
2095 }
2096 return(1);
2097 }
2098
2099 static void
2100 say(const char *file, const char *format, ...)
2101 {
2102 va_list ap;
2103
2104 if ('\0' != *basedir)
2105 fprintf(stderr, "%s", basedir);
2106 if ('\0' != *basedir && '\0' != *file)
2107 fputs("//", stderr);
2108 if ('\0' != *file)
2109 fprintf(stderr, "%s", file);
2110 fputs(": ", stderr);
2111
2112 if (NULL == format) {
2113 perror(NULL);
2114 return;
2115 }
2116
2117 va_start(ap, format);
2118 vfprintf(stderr, format, ap);
2119 va_end(ap);
2120
2121 fputc('\n', stderr);
2122 }
A git conversion mirror of mandoc.bsd.lv