1 /* Id: mandocdb.c,v 1.101 2014/01/05 04:48:40 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
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_NONE, /* format is unknown */
80 FORM_SRC, /* format is -man or -mdoc */
81 FORM_CAT /* format is cat */
82 };
83
84 struct str {
85 char *rendered; /* key in UTF-8 or ASCII form */
86 const struct mpage *mpage; /* if set, the owning parse */
87 uint64_t mask; /* bitmask in sequence */
88 char key[]; /* may contain escape sequences */
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 fform; /* 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 struct mlink *next; /* singly linked list */
115 };
116
117 enum stmt {
118 STMT_DELETE_PAGE = 0, /* delete mpage */
119 STMT_INSERT_PAGE, /* insert mpage */
120 STMT_INSERT_LINK, /* insert mlink */
121 STMT_INSERT_KEY, /* insert parsed key */
122 STMT__MAX
123 };
124
125 typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
126
127 struct mdoc_handler {
128 mdoc_fp fp; /* optional handler */
129 uint64_t mask; /* set unless handler returns 0 */
130 };
131
132 static void dbclose(int);
133 static void dbindex(const struct mpage *, struct mchars *);
134 static int dbopen(int);
135 static void dbprune(void);
136 static void filescan(const char *);
137 static void *hash_alloc(size_t, void *);
138 static void hash_free(void *, size_t, void *);
139 static void *hash_halloc(size_t, void *);
140 static void mlink_add(struct mlink *, const struct stat *);
141 static int mlink_check(struct mpage *, struct mlink *);
142 static void mlink_free(struct mlink *);
143 static void mlinks_undupe(struct mpage *);
144 static void mpages_free(void);
145 static void mpages_merge(struct mchars *, struct mparse *);
146 static void parse_cat(struct mpage *);
147 static void parse_man(struct mpage *, const struct man_node *);
148 static void parse_mdoc(struct mpage *, const struct mdoc_node *);
149 static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
150 static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
151 static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
152 static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
153 static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
154 static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
155 static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
156 static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
157 static void putkey(const struct mpage *,
158 const char *, uint64_t);
159 static void putkeys(const struct mpage *,
160 const char *, size_t, uint64_t);
161 static void putmdockey(const struct mpage *,
162 const struct mdoc_node *, uint64_t);
163 static void render_key(struct mchars *, struct str *);
164 static void say(const char *, const char *, ...);
165 static int set_basedir(const char *);
166 static int treescan(void);
167 static size_t utf8(unsigned int, char [7]);
168
169 static char *progname;
170 static int use_all; /* use all found files */
171 static int nodb; /* no database changes */
172 static int verb; /* print what we're doing */
173 static int warnings; /* warn about crap */
174 static int write_utf8; /* write UTF-8 output; else ASCII */
175 static int exitcode; /* to be returned by main */
176 static enum op op; /* operational mode */
177 static char basedir[PATH_MAX]; /* current base directory */
178 static struct ohash mpages; /* table of distinct manual pages */
179 static struct ohash mlinks; /* table of directory entries */
180 static struct ohash strings; /* table of all strings */
181 static sqlite3 *db = NULL; /* current database */
182 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
183
184 static const struct mdoc_handler mdocs[MDOC_MAX] = {
185 { NULL, 0 }, /* Ap */
186 { NULL, 0 }, /* Dd */
187 { NULL, 0 }, /* Dt */
188 { NULL, 0 }, /* Os */
189 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
190 { parse_mdoc_head, TYPE_Ss }, /* Ss */
191 { NULL, 0 }, /* Pp */
192 { NULL, 0 }, /* D1 */
193 { NULL, 0 }, /* Dl */
194 { NULL, 0 }, /* Bd */
195 { NULL, 0 }, /* Ed */
196 { NULL, 0 }, /* Bl */
197 { NULL, 0 }, /* El */
198 { NULL, 0 }, /* It */
199 { NULL, 0 }, /* Ad */
200 { NULL, TYPE_An }, /* An */
201 { NULL, TYPE_Ar }, /* Ar */
202 { NULL, TYPE_Cd }, /* Cd */
203 { NULL, TYPE_Cm }, /* Cm */
204 { NULL, TYPE_Dv }, /* Dv */
205 { NULL, TYPE_Er }, /* Er */
206 { NULL, TYPE_Ev }, /* Ev */
207 { NULL, 0 }, /* Ex */
208 { NULL, TYPE_Fa }, /* Fa */
209 { parse_mdoc_Fd, 0 }, /* Fd */
210 { NULL, TYPE_Fl }, /* Fl */
211 { parse_mdoc_Fn, 0 }, /* Fn */
212 { NULL, TYPE_Ft }, /* Ft */
213 { NULL, TYPE_Ic }, /* Ic */
214 { NULL, TYPE_In }, /* In */
215 { NULL, TYPE_Li }, /* Li */
216 { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
217 { parse_mdoc_Nm, TYPE_Nm }, /* Nm */
218 { NULL, 0 }, /* Op */
219 { NULL, 0 }, /* Ot */
220 { NULL, TYPE_Pa }, /* Pa */
221 { NULL, 0 }, /* Rv */
222 { NULL, TYPE_St }, /* St */
223 { NULL, TYPE_Va }, /* Va */
224 { parse_mdoc_body, TYPE_Va }, /* Vt */
225 { parse_mdoc_Xr, 0 }, /* Xr */
226 { NULL, 0 }, /* %A */
227 { NULL, 0 }, /* %B */
228 { NULL, 0 }, /* %D */
229 { NULL, 0 }, /* %I */
230 { NULL, 0 }, /* %J */
231 { NULL, 0 }, /* %N */
232 { NULL, 0 }, /* %O */
233 { NULL, 0 }, /* %P */
234 { NULL, 0 }, /* %R */
235 { NULL, 0 }, /* %T */
236 { NULL, 0 }, /* %V */
237 { NULL, 0 }, /* Ac */
238 { NULL, 0 }, /* Ao */
239 { NULL, 0 }, /* Aq */
240 { NULL, TYPE_At }, /* At */
241 { NULL, 0 }, /* Bc */
242 { NULL, 0 }, /* Bf */
243 { NULL, 0 }, /* Bo */
244 { NULL, 0 }, /* Bq */
245 { NULL, TYPE_Bsx }, /* Bsx */
246 { NULL, TYPE_Bx }, /* Bx */
247 { NULL, 0 }, /* Db */
248 { NULL, 0 }, /* Dc */
249 { NULL, 0 }, /* Do */
250 { NULL, 0 }, /* Dq */
251 { NULL, 0 }, /* Ec */
252 { NULL, 0 }, /* Ef */
253 { NULL, TYPE_Em }, /* Em */
254 { NULL, 0 }, /* Eo */
255 { NULL, TYPE_Fx }, /* Fx */
256 { NULL, TYPE_Ms }, /* Ms */
257 { NULL, 0 }, /* No */
258 { NULL, 0 }, /* Ns */
259 { NULL, TYPE_Nx }, /* Nx */
260 { NULL, TYPE_Ox }, /* Ox */
261 { NULL, 0 }, /* Pc */
262 { NULL, 0 }, /* Pf */
263 { NULL, 0 }, /* Po */
264 { NULL, 0 }, /* Pq */
265 { NULL, 0 }, /* Qc */
266 { NULL, 0 }, /* Ql */
267 { NULL, 0 }, /* Qo */
268 { NULL, 0 }, /* Qq */
269 { NULL, 0 }, /* Re */
270 { NULL, 0 }, /* Rs */
271 { NULL, 0 }, /* Sc */
272 { NULL, 0 }, /* So */
273 { NULL, 0 }, /* Sq */
274 { NULL, 0 }, /* Sm */
275 { NULL, 0 }, /* Sx */
276 { NULL, TYPE_Sy }, /* Sy */
277 { NULL, TYPE_Tn }, /* Tn */
278 { NULL, 0 }, /* Ux */
279 { NULL, 0 }, /* Xc */
280 { NULL, 0 }, /* Xo */
281 { parse_mdoc_head, 0 }, /* Fo */
282 { NULL, 0 }, /* Fc */
283 { NULL, 0 }, /* Oo */
284 { NULL, 0 }, /* Oc */
285 { NULL, 0 }, /* Bk */
286 { NULL, 0 }, /* Ek */
287 { NULL, 0 }, /* Bt */
288 { NULL, 0 }, /* Hf */
289 { NULL, 0 }, /* Fr */
290 { NULL, 0 }, /* Ud */
291 { NULL, TYPE_Lb }, /* Lb */
292 { NULL, 0 }, /* Lp */
293 { NULL, TYPE_Lk }, /* Lk */
294 { NULL, TYPE_Mt }, /* Mt */
295 { NULL, 0 }, /* Brq */
296 { NULL, 0 }, /* Bro */
297 { NULL, 0 }, /* Brc */
298 { NULL, 0 }, /* %C */
299 { NULL, 0 }, /* Es */
300 { NULL, 0 }, /* En */
301 { NULL, TYPE_Dx }, /* Dx */
302 { NULL, 0 }, /* %Q */
303 { NULL, 0 }, /* br */
304 { NULL, 0 }, /* sp */
305 { NULL, 0 }, /* %U */
306 { NULL, 0 }, /* Ta */
307 };
308
309 int
main(int argc,char * argv[])310 main(int argc, char *argv[])
311 {
312 int ch, i;
313 size_t j, sz;
314 const char *path_arg;
315 struct mchars *mc;
316 struct manpaths dirs;
317 struct mparse *mp;
318 struct ohash_info mpages_info, mlinks_info;
319
320 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
321 memset(&dirs, 0, sizeof(struct manpaths));
322
323 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
324 mpages_info.halloc = mlinks_info.halloc = hash_halloc;
325 mpages_info.hfree = mlinks_info.hfree = hash_free;
326
327 mpages_info.key_offset = offsetof(struct mpage, inodev);
328 mlinks_info.key_offset = offsetof(struct mlink, file);
329
330 progname = strrchr(argv[0], '/');
331 if (progname == NULL)
332 progname = argv[0];
333 else
334 ++progname;
335
336 /*
337 * We accept a few different invocations.
338 * The CHECKOP macro makes sure that invocation styles don't
339 * clobber each other.
340 */
341 #define CHECKOP(_op, _ch) do \
342 if (OP_DEFAULT != (_op)) { \
343 fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
344 goto usage; \
345 } while (/*CONSTCOND*/0)
346
347 path_arg = NULL;
348 op = OP_DEFAULT;
349
350 while (-1 != (ch = getopt(argc, argv, "aC:d:nT:tu:vW")))
351 switch (ch) {
352 case ('a'):
353 use_all = 1;
354 break;
355 case ('C'):
356 CHECKOP(op, ch);
357 path_arg = optarg;
358 op = OP_CONFFILE;
359 break;
360 case ('d'):
361 CHECKOP(op, ch);
362 path_arg = optarg;
363 op = OP_UPDATE;
364 break;
365 case ('n'):
366 nodb = 1;
367 break;
368 case ('T'):
369 if (strcmp(optarg, "utf8")) {
370 fprintf(stderr, "-T%s: Unsupported "
371 "output format\n", optarg);
372 goto usage;
373 }
374 write_utf8 = 1;
375 break;
376 case ('t'):
377 CHECKOP(op, ch);
378 dup2(STDOUT_FILENO, STDERR_FILENO);
379 op = OP_TEST;
380 nodb = warnings = 1;
381 break;
382 case ('u'):
383 CHECKOP(op, ch);
384 path_arg = optarg;
385 op = OP_DELETE;
386 break;
387 case ('v'):
388 verb++;
389 break;
390 case ('W'):
391 warnings = 1;
392 break;
393 default:
394 goto usage;
395 }
396
397 argc -= optind;
398 argv += optind;
399
400 if (OP_CONFFILE == op && argc > 0) {
401 fprintf(stderr, "-C: Too many arguments\n");
402 goto usage;
403 }
404
405 exitcode = (int)MANDOCLEVEL_OK;
406 mp = mparse_alloc(MPARSE_AUTO,
407 MANDOCLEVEL_FATAL, NULL, NULL, NULL);
408 mc = mchars_alloc();
409
410 ohash_init(&mpages, 6, &mpages_info);
411 ohash_init(&mlinks, 6, &mlinks_info);
412
413 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
414 /*
415 * Force processing all files.
416 */
417 use_all = 1;
418
419 /*
420 * All of these deal with a specific directory.
421 * Jump into that directory then collect files specified
422 * on the command-line.
423 */
424 if (0 == set_basedir(path_arg))
425 goto out;
426 for (i = 0; i < argc; i++)
427 filescan(argv[i]);
428 if (0 == dbopen(1))
429 goto out;
430 if (OP_TEST != op)
431 dbprune();
432 if (OP_DELETE != op)
433 mpages_merge(mc, mp);
434 dbclose(1);
435 } else {
436 /*
437 * If we have arguments, use them as our manpaths.
438 * If we don't, grok from manpath(1) or however else
439 * manpath_parse() wants to do it.
440 */
441 if (argc > 0) {
442 dirs.paths = mandoc_calloc
443 (argc, sizeof(char *));
444 dirs.sz = (size_t)argc;
445 for (i = 0; i < argc; i++)
446 dirs.paths[i] = mandoc_strdup(argv[i]);
447 } else
448 manpath_parse(&dirs, path_arg, NULL, NULL);
449
450 /*
451 * First scan the tree rooted at a base directory, then
452 * build a new database and finally move it into place.
453 * Ignore zero-length directories and strip trailing
454 * slashes.
455 */
456 for (j = 0; j < dirs.sz; j++) {
457 sz = strlen(dirs.paths[j]);
458 if (sz && '/' == dirs.paths[j][sz - 1])
459 dirs.paths[j][--sz] = '\0';
460 if (0 == sz)
461 continue;
462
463 if (j) {
464 ohash_init(&mpages, 6, &mpages_info);
465 ohash_init(&mlinks, 6, &mlinks_info);
466 }
467
468 if (0 == set_basedir(dirs.paths[j]))
469 goto out;
470 if (0 == treescan())
471 goto out;
472 if (0 == set_basedir(dirs.paths[j]))
473 goto out;
474 if (0 == dbopen(0))
475 goto out;
476
477 mpages_merge(mc, mp);
478 dbclose(0);
479
480 if (j + 1 < dirs.sz) {
481 mpages_free();
482 ohash_delete(&mpages);
483 ohash_delete(&mlinks);
484 }
485 }
486 }
487 out:
488 set_basedir(NULL);
489 manpath_free(&dirs);
490 mchars_free(mc);
491 mparse_free(mp);
492 mpages_free();
493 ohash_delete(&mpages);
494 ohash_delete(&mlinks);
495 return(exitcode);
496 usage:
497 fprintf(stderr, "usage: %s [-anvW] [-C file] [-Tutf8]\n"
498 " %s [-anvW] [-Tutf8] dir ...\n"
499 " %s [-nvW] [-Tutf8] -d dir [file ...]\n"
500 " %s [-nvW] -u dir [file ...]\n"
501 " %s -t file ...\n",
502 progname, progname, progname,
503 progname, progname);
504
505 return((int)MANDOCLEVEL_BADARG);
506 }
507
508 /*
509 * Scan a directory tree rooted at "basedir" for manpages.
510 * We use fts(), scanning directory parts along the way for clues to our
511 * section and architecture.
512 *
513 * If use_all has been specified, grok all files.
514 * If not, sanitise paths to the following:
515 *
516 * [./]man*[/<arch>]/<name>.<section>
517 * or
518 * [./]cat<section>[/<arch>]/<name>.0
519 *
520 * TODO: accomodate for multi-language directories.
521 */
522 static int
treescan(void)523 treescan(void)
524 {
525 FTS *f;
526 FTSENT *ff;
527 struct mlink *mlink;
528 int dform;
529 char *dsec, *arch, *fsec, *cp;
530 const char *path;
531 const char *argv[2];
532
533 argv[0] = ".";
534 argv[1] = (char *)NULL;
535
536 /*
537 * Walk through all components under the directory, using the
538 * logical descent of files.
539 */
540 f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
541 if (NULL == f) {
542 exitcode = (int)MANDOCLEVEL_SYSERR;
543 say("", NULL);
544 return(0);
545 }
546
547 dsec = arch = NULL;
548 dform = FORM_NONE;
549
550 while (NULL != (ff = fts_read(f))) {
551 path = ff->fts_path + 2;
552 /*
553 * If we're a regular file, add an mlink by using the
554 * stored directory data and handling the filename.
555 */
556 if (FTS_F == ff->fts_info) {
557 if (0 == strcmp(path, MANDOC_DB))
558 continue;
559 if ( ! use_all && ff->fts_level < 2) {
560 if (warnings)
561 say(path, "Extraneous file");
562 continue;
563 } else if (NULL == (fsec =
564 strrchr(ff->fts_name, '.'))) {
565 if ( ! use_all) {
566 if (warnings)
567 say(path,
568 "No filename suffix");
569 continue;
570 }
571 } else if (0 == strcmp(++fsec, "html")) {
572 if (warnings)
573 say(path, "Skip html");
574 continue;
575 } else if (0 == strcmp(fsec, "gz")) {
576 if (warnings)
577 say(path, "Skip gz");
578 continue;
579 } else if (0 == strcmp(fsec, "ps")) {
580 if (warnings)
581 say(path, "Skip ps");
582 continue;
583 } else if (0 == strcmp(fsec, "pdf")) {
584 if (warnings)
585 say(path, "Skip pdf");
586 continue;
587 } else if ( ! use_all &&
588 ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
589 (FORM_CAT == dform && strcmp(fsec, "0")))) {
590 if (warnings)
591 say(path, "Wrong filename suffix");
592 continue;
593 } else
594 fsec[-1] = '\0';
595
596 mlink = mandoc_calloc(1, sizeof(struct mlink));
597 strlcpy(mlink->file, path, sizeof(mlink->file));
598 mlink->dform = dform;
599 mlink->dsec = dsec;
600 mlink->arch = arch;
601 mlink->name = ff->fts_name;
602 mlink->fsec = fsec;
603 mlink_add(mlink, ff->fts_statp);
604 continue;
605 } else if (FTS_D != ff->fts_info &&
606 FTS_DP != ff->fts_info) {
607 if (warnings)
608 say(path, "Not a regular file");
609 continue;
610 }
611
612 switch (ff->fts_level) {
613 case (0):
614 /* Ignore the root directory. */
615 break;
616 case (1):
617 /*
618 * This might contain manX/ or catX/.
619 * Try to infer this from the name.
620 * If we're not in use_all, enforce it.
621 */
622 cp = ff->fts_name;
623 if (FTS_DP == ff->fts_info)
624 break;
625
626 if (0 == strncmp(cp, "man", 3)) {
627 dform = FORM_SRC;
628 dsec = cp + 3;
629 } else if (0 == strncmp(cp, "cat", 3)) {
630 dform = FORM_CAT;
631 dsec = cp + 3;
632 } else {
633 dform = FORM_NONE;
634 dsec = NULL;
635 }
636
637 if (NULL != dsec || use_all)
638 break;
639
640 if (warnings)
641 say(path, "Unknown directory part");
642 fts_set(f, ff, FTS_SKIP);
643 break;
644 case (2):
645 /*
646 * Possibly our architecture.
647 * If we're descending, keep tabs on it.
648 */
649 if (FTS_DP != ff->fts_info && NULL != dsec)
650 arch = ff->fts_name;
651 else
652 arch = NULL;
653 break;
654 default:
655 if (FTS_DP == ff->fts_info || use_all)
656 break;
657 if (warnings)
658 say(path, "Extraneous directory part");
659 fts_set(f, ff, FTS_SKIP);
660 break;
661 }
662 }
663
664 fts_close(f);
665 return(1);
666 }
667
668 /*
669 * Add a file to the mlinks table.
670 * Do not verify that it's a "valid" looking manpage (we'll do that
671 * later).
672 *
673 * Try to infer the manual section, architecture, and page name from the
674 * path, assuming it looks like
675 *
676 * [./]man*[/<arch>]/<name>.<section>
677 * or
678 * [./]cat<section>[/<arch>]/<name>.0
679 *
680 * See treescan() for the fts(3) version of this.
681 */
682 static void
filescan(const char * file)683 filescan(const char *file)
684 {
685 char buf[PATH_MAX];
686 struct stat st;
687 struct mlink *mlink;
688 char *p, *start;
689
690 assert(use_all);
691
692 if (0 == strncmp(file, "./", 2))
693 file += 2;
694
695 if (NULL == realpath(file, buf)) {
696 exitcode = (int)MANDOCLEVEL_BADARG;
697 say(file, NULL);
698 return;
699 } else if (OP_TEST != op && strstr(buf, basedir) != buf) {
700 exitcode = (int)MANDOCLEVEL_BADARG;
701 say("", "%s: outside base directory", buf);
702 return;
703 } else if (-1 == stat(buf, &st)) {
704 exitcode = (int)MANDOCLEVEL_BADARG;
705 say(file, NULL);
706 return;
707 } else if ( ! (S_IFREG & st.st_mode)) {
708 exitcode = (int)MANDOCLEVEL_BADARG;
709 say(file, "Not a regular file");
710 return;
711 }
712 start = buf + strlen(basedir);
713 mlink = mandoc_calloc(1, sizeof(struct mlink));
714 strlcpy(mlink->file, start, sizeof(mlink->file));
715
716 /*
717 * First try to guess our directory structure.
718 * If we find a separator, try to look for man* or cat*.
719 * If we find one of these and what's underneath is a directory,
720 * assume it's an architecture.
721 */
722 if (NULL != (p = strchr(start, '/'))) {
723 *p++ = '\0';
724 if (0 == strncmp(start, "man", 3)) {
725 mlink->dform = FORM_SRC;
726 mlink->dsec = start + 3;
727 } else if (0 == strncmp(start, "cat", 3)) {
728 mlink->dform = FORM_CAT;
729 mlink->dsec = start + 3;
730 }
731
732 start = p;
733 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
734 *p++ = '\0';
735 mlink->arch = start;
736 start = p;
737 }
738 }
739
740 /*
741 * Now check the file suffix.
742 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
743 */
744 p = strrchr(start, '\0');
745 while (p-- > start && '/' != *p && '.' != *p)
746 /* Loop. */ ;
747
748 if ('.' == *p) {
749 *p++ = '\0';
750 mlink->fsec = p;
751 }
752
753 /*
754 * Now try to parse the name.
755 * Use the filename portion of the path.
756 */
757 mlink->name = start;
758 if (NULL != (p = strrchr(start, '/'))) {
759 mlink->name = p + 1;
760 *p = '\0';
761 }
762 mlink_add(mlink, &st);
763 }
764
765 static void
mlink_add(struct mlink * mlink,const struct stat * st)766 mlink_add(struct mlink *mlink, const struct stat *st)
767 {
768 struct inodev inodev;
769 struct mpage *mpage;
770 unsigned int slot;
771
772 assert(NULL != mlink->file);
773
774 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
775 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
776 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
777 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
778
779 if ('0' == *mlink->fsec) {
780 free(mlink->fsec);
781 mlink->fsec = mandoc_strdup(mlink->dsec);
782 mlink->fform = FORM_CAT;
783 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
784 mlink->fform = FORM_SRC;
785 else
786 mlink->fform = FORM_NONE;
787
788 slot = ohash_qlookup(&mlinks, mlink->file);
789 assert(NULL == ohash_find(&mlinks, slot));
790 ohash_insert(&mlinks, slot, mlink);
791
792 inodev.st_ino = st->st_ino;
793 inodev.st_dev = st->st_dev;
794 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
795 sizeof(struct inodev), inodev.st_ino);
796 mpage = ohash_find(&mpages, slot);
797 if (NULL == mpage) {
798 mpage = mandoc_calloc(1, sizeof(struct mpage));
799 mpage->inodev.st_ino = inodev.st_ino;
800 mpage->inodev.st_dev = inodev.st_dev;
801 ohash_insert(&mpages, slot, mpage);
802 } else
803 mlink->next = mpage->mlinks;
804 mpage->mlinks = mlink;
805 }
806
807 static void
mlink_free(struct mlink * mlink)808 mlink_free(struct mlink *mlink)
809 {
810
811 free(mlink->dsec);
812 free(mlink->arch);
813 free(mlink->name);
814 free(mlink->fsec);
815 free(mlink);
816 }
817
818 static void
mpages_free(void)819 mpages_free(void)
820 {
821 struct mpage *mpage;
822 struct mlink *mlink;
823 unsigned int slot;
824
825 mpage = ohash_first(&mpages, &slot);
826 while (NULL != mpage) {
827 while (NULL != (mlink = mpage->mlinks)) {
828 mpage->mlinks = mlink->next;
829 mlink_free(mlink);
830 }
831 free(mpage->sec);
832 free(mpage->arch);
833 free(mpage->title);
834 free(mpage->desc);
835 free(mpage);
836 mpage = ohash_next(&mpages, &slot);
837 }
838 }
839
840 /*
841 * For each mlink to the mpage, check whether the path looks like
842 * it is formatted, and if it does, check whether a source manual
843 * exists by the same name, ignoring the suffix.
844 * If both conditions hold, drop the mlink.
845 */
846 static void
mlinks_undupe(struct mpage * mpage)847 mlinks_undupe(struct mpage *mpage)
848 {
849 char buf[PATH_MAX];
850 struct mlink **prev;
851 struct mlink *mlink;
852 char *bufp;
853
854 mpage->form = FORM_CAT;
855 prev = &mpage->mlinks;
856 while (NULL != (mlink = *prev)) {
857 if (FORM_CAT != mlink->dform) {
858 mpage->form = FORM_NONE;
859 goto nextlink;
860 }
861 if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) {
862 if (warnings)
863 say(mlink->file, "Filename too long");
864 goto nextlink;
865 }
866 bufp = strstr(buf, "cat");
867 assert(NULL != bufp);
868 memcpy(bufp, "man", 3);
869 if (NULL != (bufp = strrchr(buf, '.')))
870 *++bufp = '\0';
871 strlcat(buf, mlink->dsec, PATH_MAX);
872 if (NULL == ohash_find(&mlinks,
873 ohash_qlookup(&mlinks, buf)))
874 goto nextlink;
875 if (warnings)
876 say(mlink->file, "Man source exists: %s", buf);
877 if (use_all)
878 goto nextlink;
879 *prev = mlink->next;
880 mlink_free(mlink);
881 continue;
882 nextlink:
883 prev = &(*prev)->next;
884 }
885 }
886
887 static int
mlink_check(struct mpage * mpage,struct mlink * mlink)888 mlink_check(struct mpage *mpage, struct mlink *mlink)
889 {
890 int match;
891
892 match = 1;
893
894 /*
895 * Check whether the manual section given in a file
896 * agrees with the directory where the file is located.
897 * Some manuals have suffixes like (3p) on their
898 * section number either inside the file or in the
899 * directory name, some are linked into more than one
900 * section, like encrypt(1) = makekey(8).
901 */
902
903 if (FORM_SRC == mpage->form &&
904 strcasecmp(mpage->sec, mlink->dsec)) {
905 match = 0;
906 say(mlink->file, "Section \"%s\" manual in %s directory",
907 mpage->sec, mlink->dsec);
908 }
909
910 /*
911 * Manual page directories exist for each kernel
912 * architecture as returned by machine(1).
913 * However, many manuals only depend on the
914 * application architecture as returned by arch(1).
915 * For example, some (2/ARM) manuals are shared
916 * across the "armish" and "zaurus" kernel
917 * architectures.
918 * A few manuals are even shared across completely
919 * different architectures, for example fdformat(1)
920 * on amd64, i386, sparc, and sparc64.
921 */
922
923 if (strcasecmp(mpage->arch, mlink->arch)) {
924 match = 0;
925 say(mlink->file, "Architecture \"%s\" manual in "
926 "\"%s\" directory", mpage->arch, mlink->arch);
927 }
928
929 if (strcasecmp(mpage->title, mlink->name))
930 match = 0;
931
932 return(match);
933 }
934
935 /*
936 * Run through the files in the global vector "mpages"
937 * and add them to the database specified in "basedir".
938 *
939 * This handles the parsing scheme itself, using the cues of directory
940 * and filename to determine whether the file is parsable or not.
941 */
942 static void
mpages_merge(struct mchars * mc,struct mparse * mp)943 mpages_merge(struct mchars *mc, struct mparse *mp)
944 {
945 struct ohash_info str_info;
946 struct mpage *mpage;
947 struct mlink *mlink;
948 struct mdoc *mdoc;
949 struct man *man;
950 const char *cp;
951 int match;
952 unsigned int pslot;
953 enum mandoclevel lvl;
954
955 str_info.alloc = hash_alloc;
956 str_info.halloc = hash_halloc;
957 str_info.hfree = hash_free;
958 str_info.key_offset = offsetof(struct str, key);
959
960 mpage = ohash_first(&mpages, &pslot);
961 while (NULL != mpage) {
962 mlinks_undupe(mpage);
963 if (NULL == mpage->mlinks) {
964 mpage = ohash_next(&mpages, &pslot);
965 continue;
966 }
967
968 ohash_init(&strings, 6, &str_info);
969 mparse_reset(mp);
970 mdoc = NULL;
971 man = NULL;
972
973 /*
974 * Try interpreting the file as mdoc(7) or man(7)
975 * source code, unless it is already known to be
976 * formatted. Fall back to formatted mode.
977 */
978 if (FORM_CAT != mpage->mlinks->dform ||
979 FORM_CAT != mpage->mlinks->fform) {
980 lvl = mparse_readfd(mp, -1, mpage->mlinks->file);
981 if (lvl < MANDOCLEVEL_FATAL)
982 mparse_result(mp, &mdoc, &man);
983 }
984
985 if (NULL != mdoc) {
986 mpage->form = FORM_SRC;
987 mpage->sec =
988 mandoc_strdup(mdoc_meta(mdoc)->msec);
989 mpage->arch = mdoc_meta(mdoc)->arch;
990 mpage->arch = mandoc_strdup(
991 NULL == mpage->arch ? "" : mpage->arch);
992 mpage->title =
993 mandoc_strdup(mdoc_meta(mdoc)->title);
994 } else if (NULL != man) {
995 mpage->form = FORM_SRC;
996 mpage->sec =
997 mandoc_strdup(man_meta(man)->msec);
998 mpage->arch =
999 mandoc_strdup(mpage->mlinks->arch);
1000 mpage->title =
1001 mandoc_strdup(man_meta(man)->title);
1002 } else {
1003 mpage->form = FORM_CAT;
1004 mpage->sec =
1005 mandoc_strdup(mpage->mlinks->dsec);
1006 mpage->arch =
1007 mandoc_strdup(mpage->mlinks->arch);
1008 mpage->title =
1009 mandoc_strdup(mpage->mlinks->name);
1010 }
1011 putkey(mpage, mpage->sec, TYPE_sec);
1012 putkey(mpage, '\0' == *mpage->arch ?
1013 "any" : mpage->arch, TYPE_arch);
1014
1015 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1016 if ('\0' != *mlink->dsec)
1017 putkey(mpage, mlink->dsec, TYPE_sec);
1018 if ('\0' != *mlink->fsec)
1019 putkey(mpage, mlink->fsec, TYPE_sec);
1020 putkey(mpage, '\0' == *mlink->arch ?
1021 "any" : mlink->arch, TYPE_arch);
1022 putkey(mpage, mlink->name, TYPE_Nm);
1023 }
1024
1025 if (warnings && !use_all) {
1026 match = 0;
1027 for (mlink = mpage->mlinks; mlink;
1028 mlink = mlink->next)
1029 if (mlink_check(mpage, mlink))
1030 match = 1;
1031 } else
1032 match = 1;
1033
1034 if (NULL != mdoc) {
1035 if (NULL != (cp = mdoc_meta(mdoc)->name))
1036 putkey(mpage, cp, TYPE_Nm);
1037 assert(NULL == mpage->desc);
1038 parse_mdoc(mpage, mdoc_node(mdoc));
1039 putkey(mpage, NULL != mpage->desc ?
1040 mpage->desc : mpage->mlinks->name, TYPE_Nd);
1041 } else if (NULL != man)
1042 parse_man(mpage, man_node(man));
1043 else
1044 parse_cat(mpage);
1045
1046 dbindex(mpage, mc);
1047 ohash_delete(&strings);
1048 mpage = ohash_next(&mpages, &pslot);
1049 }
1050 }
1051
1052 static void
parse_cat(struct mpage * mpage)1053 parse_cat(struct mpage *mpage)
1054 {
1055 FILE *stream;
1056 char *line, *p, *title;
1057 size_t len, plen, titlesz;
1058
1059 if (NULL == (stream = fopen(mpage->mlinks->file, "r"))) {
1060 if (warnings)
1061 say(mpage->mlinks->file, NULL);
1062 return;
1063 }
1064
1065 /* Skip to first blank line. */
1066
1067 while (NULL != (line = fgetln(stream, &len)))
1068 if ('\n' == *line)
1069 break;
1070
1071 /*
1072 * Assume the first line that is not indented
1073 * is the first section header. Skip to it.
1074 */
1075
1076 while (NULL != (line = fgetln(stream, &len)))
1077 if ('\n' != *line && ' ' != *line)
1078 break;
1079
1080 /*
1081 * Read up until the next section into a buffer.
1082 * Strip the leading and trailing newline from each read line,
1083 * appending a trailing space.
1084 * Ignore empty (whitespace-only) lines.
1085 */
1086
1087 titlesz = 0;
1088 title = NULL;
1089
1090 while (NULL != (line = fgetln(stream, &len))) {
1091 if (' ' != *line || '\n' != line[len - 1])
1092 break;
1093 while (len > 0 && isspace((unsigned char)*line)) {
1094 line++;
1095 len--;
1096 }
1097 if (1 == len)
1098 continue;
1099 title = mandoc_realloc(title, titlesz + len);
1100 memcpy(title + titlesz, line, len);
1101 titlesz += len;
1102 title[titlesz - 1] = ' ';
1103 }
1104
1105 /*
1106 * If no page content can be found, or the input line
1107 * is already the next section header, or there is no
1108 * trailing newline, reuse the page title as the page
1109 * description.
1110 */
1111
1112 if (NULL == title || '\0' == *title) {
1113 if (warnings)
1114 say(mpage->mlinks->file,
1115 "Cannot find NAME section");
1116 assert(NULL == mpage->desc);
1117 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1118 putkey(mpage, mpage->mlinks->name, TYPE_Nd);
1119 fclose(stream);
1120 free(title);
1121 return;
1122 }
1123
1124 title = mandoc_realloc(title, titlesz + 1);
1125 title[titlesz] = '\0';
1126
1127 /*
1128 * Skip to the first dash.
1129 * Use the remaining line as the description (no more than 70
1130 * bytes).
1131 */
1132
1133 if (NULL != (p = strstr(title, "- "))) {
1134 for (p += 2; ' ' == *p || '\b' == *p; p++)
1135 /* Skip to next word. */ ;
1136 } else {
1137 if (warnings)
1138 say(mpage->mlinks->file,
1139 "No dash in title line");
1140 p = title;
1141 }
1142
1143 plen = strlen(p);
1144
1145 /* Strip backspace-encoding from line. */
1146
1147 while (NULL != (line = memchr(p, '\b', plen))) {
1148 len = line - p;
1149 if (0 == len) {
1150 memmove(line, line + 1, plen--);
1151 continue;
1152 }
1153 memmove(line - 1, line + 1, plen - len);
1154 plen -= 2;
1155 }
1156
1157 assert(NULL == mpage->desc);
1158 mpage->desc = mandoc_strdup(p);
1159 putkey(mpage, mpage->desc, TYPE_Nd);
1160 fclose(stream);
1161 free(title);
1162 }
1163
1164 /*
1165 * Put a type/word pair into the word database for this particular file.
1166 */
1167 static void
putkey(const struct mpage * mpage,const char * value,uint64_t type)1168 putkey(const struct mpage *mpage, const char *value, uint64_t type)
1169 {
1170
1171 assert(NULL != value);
1172 putkeys(mpage, value, strlen(value), type);
1173 }
1174
1175 /*
1176 * Grok all nodes at or below a certain mdoc node into putkey().
1177 */
1178 static void
putmdockey(const struct mpage * mpage,const struct mdoc_node * n,uint64_t m)1179 putmdockey(const struct mpage *mpage,
1180 const struct mdoc_node *n, uint64_t m)
1181 {
1182
1183 for ( ; NULL != n; n = n->next) {
1184 if (NULL != n->child)
1185 putmdockey(mpage, n->child, m);
1186 if (MDOC_TEXT == n->type)
1187 putkey(mpage, n->string, m);
1188 }
1189 }
1190
1191 static void
parse_man(struct mpage * mpage,const struct man_node * n)1192 parse_man(struct mpage *mpage, const struct man_node *n)
1193 {
1194 const struct man_node *head, *body;
1195 char *start, *sv, *title;
1196 char byte;
1197 size_t sz, titlesz;
1198
1199 if (NULL == n)
1200 return;
1201
1202 /*
1203 * We're only searching for one thing: the first text child in
1204 * the BODY of a NAME section. Since we don't keep track of
1205 * sections in -man, run some hoops to find out whether we're in
1206 * the correct section or not.
1207 */
1208
1209 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1210 body = n;
1211 assert(body->parent);
1212 if (NULL != (head = body->parent->head) &&
1213 1 == head->nchild &&
1214 NULL != (head = (head->child)) &&
1215 MAN_TEXT == head->type &&
1216 0 == strcmp(head->string, "NAME") &&
1217 NULL != (body = body->child) &&
1218 MAN_TEXT == body->type) {
1219
1220 title = NULL;
1221 titlesz = 0;
1222
1223 /*
1224 * Suck the entire NAME section into memory.
1225 * Yes, we might run away.
1226 * But too many manuals have big, spread-out
1227 * NAME sections over many lines.
1228 */
1229
1230 for ( ; NULL != body; body = body->next) {
1231 if (MAN_TEXT != body->type)
1232 break;
1233 if (0 == (sz = strlen(body->string)))
1234 continue;
1235 title = mandoc_realloc
1236 (title, titlesz + sz + 1);
1237 memcpy(title + titlesz, body->string, sz);
1238 titlesz += sz + 1;
1239 title[titlesz - 1] = ' ';
1240 }
1241 if (NULL == title)
1242 return;
1243
1244 title = mandoc_realloc(title, titlesz + 1);
1245 title[titlesz] = '\0';
1246
1247 /* Skip leading space. */
1248
1249 sv = title;
1250 while (isspace((unsigned char)*sv))
1251 sv++;
1252
1253 if (0 == (sz = strlen(sv))) {
1254 free(title);
1255 return;
1256 }
1257
1258 /* Erase trailing space. */
1259
1260 start = &sv[sz - 1];
1261 while (start > sv && isspace((unsigned char)*start))
1262 *start-- = '\0';
1263
1264 if (start == sv) {
1265 free(title);
1266 return;
1267 }
1268
1269 start = sv;
1270
1271 /*
1272 * Go through a special heuristic dance here.
1273 * Conventionally, one or more manual names are
1274 * comma-specified prior to a whitespace, then a
1275 * dash, then a description. Try to puzzle out
1276 * the name parts here.
1277 */
1278
1279 for ( ;; ) {
1280 sz = strcspn(start, " ,");
1281 if ('\0' == start[sz])
1282 break;
1283
1284 byte = start[sz];
1285 start[sz] = '\0';
1286
1287 putkey(mpage, start, TYPE_Nm);
1288
1289 if (' ' == byte) {
1290 start += sz + 1;
1291 break;
1292 }
1293
1294 assert(',' == byte);
1295 start += sz + 1;
1296 while (' ' == *start)
1297 start++;
1298 }
1299
1300 if (sv == start) {
1301 putkey(mpage, start, TYPE_Nm);
1302 free(title);
1303 return;
1304 }
1305
1306 while (isspace((unsigned char)*start))
1307 start++;
1308
1309 if (0 == strncmp(start, "-", 1))
1310 start += 1;
1311 else if (0 == strncmp(start, "\\-\\-", 4))
1312 start += 4;
1313 else if (0 == strncmp(start, "\\-", 2))
1314 start += 2;
1315 else if (0 == strncmp(start, "\\(en", 4))
1316 start += 4;
1317 else if (0 == strncmp(start, "\\(em", 4))
1318 start += 4;
1319
1320 while (' ' == *start)
1321 start++;
1322
1323 assert(NULL == mpage->desc);
1324 mpage->desc = mandoc_strdup(start);
1325 putkey(mpage, mpage->desc, TYPE_Nd);
1326 free(title);
1327 return;
1328 }
1329 }
1330
1331 for (n = n->child; n; n = n->next) {
1332 if (NULL != mpage->desc)
1333 break;
1334 parse_man(mpage, n);
1335 }
1336 }
1337
1338 static void
parse_mdoc(struct mpage * mpage,const struct mdoc_node * n)1339 parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1340 {
1341
1342 assert(NULL != n);
1343 for (n = n->child; NULL != n; n = n->next) {
1344 switch (n->type) {
1345 case (MDOC_ELEM):
1346 /* FALLTHROUGH */
1347 case (MDOC_BLOCK):
1348 /* FALLTHROUGH */
1349 case (MDOC_HEAD):
1350 /* FALLTHROUGH */
1351 case (MDOC_BODY):
1352 /* FALLTHROUGH */
1353 case (MDOC_TAIL):
1354 if (NULL != mdocs[n->tok].fp)
1355 if (0 == (*mdocs[n->tok].fp)(mpage, n))
1356 break;
1357 if (mdocs[n->tok].mask)
1358 putmdockey(mpage, n->child,
1359 mdocs[n->tok].mask);
1360 break;
1361 default:
1362 assert(MDOC_ROOT != n->type);
1363 continue;
1364 }
1365 if (NULL != n->child)
1366 parse_mdoc(mpage, n);
1367 }
1368 }
1369
1370 static int
parse_mdoc_Fd(struct mpage * mpage,const struct mdoc_node * n)1371 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1372 {
1373 const char *start, *end;
1374 size_t sz;
1375
1376 if (SEC_SYNOPSIS != n->sec ||
1377 NULL == (n = n->child) ||
1378 MDOC_TEXT != n->type)
1379 return(0);
1380
1381 /*
1382 * Only consider those `Fd' macro fields that begin with an
1383 * "inclusion" token (versus, e.g., #define).
1384 */
1385
1386 if (strcmp("#include", n->string))
1387 return(0);
1388
1389 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1390 return(0);
1391
1392 /*
1393 * Strip away the enclosing angle brackets and make sure we're
1394 * not zero-length.
1395 */
1396
1397 start = n->string;
1398 if ('<' == *start || '"' == *start)
1399 start++;
1400
1401 if (0 == (sz = strlen(start)))
1402 return(0);
1403
1404 end = &start[(int)sz - 1];
1405 if ('>' == *end || '"' == *end)
1406 end--;
1407
1408 if (end > start)
1409 putkeys(mpage, start, end - start + 1, TYPE_In);
1410 return(0);
1411 }
1412
1413 static int
parse_mdoc_Fn(struct mpage * mpage,const struct mdoc_node * n)1414 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1415 {
1416 const char *cp;
1417
1418 if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1419 return(0);
1420
1421 /*
1422 * Parse: .Fn "struct type *name" "char *arg".
1423 * First strip away pointer symbol.
1424 * Then store the function name, then type.
1425 * Finally, store the arguments.
1426 */
1427
1428 if (NULL == (cp = strrchr(n->string, ' ')))
1429 cp = n->string;
1430
1431 while ('*' == *cp)
1432 cp++;
1433
1434 putkey(mpage, cp, TYPE_Fn);
1435
1436 if (n->string < cp)
1437 putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1438
1439 for (n = n->next; NULL != n; n = n->next)
1440 if (MDOC_TEXT == n->type)
1441 putkey(mpage, n->string, TYPE_Fa);
1442
1443 return(0);
1444 }
1445
1446 static int
parse_mdoc_Xr(struct mpage * mpage,const struct mdoc_node * n)1447 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1448 {
1449 char *cp;
1450
1451 if (NULL == (n = n->child))
1452 return(0);
1453
1454 if (NULL == n->next) {
1455 putkey(mpage, n->string, TYPE_Xr);
1456 return(0);
1457 }
1458
1459 if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) {
1460 perror(NULL);
1461 exit((int)MANDOCLEVEL_SYSERR);
1462 }
1463 putkey(mpage, cp, TYPE_Xr);
1464 free(cp);
1465 return(0);
1466 }
1467
1468 static int
parse_mdoc_Nd(struct mpage * mpage,const struct mdoc_node * n)1469 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1470 {
1471 size_t sz;
1472
1473 if (MDOC_BODY != n->type)
1474 return(0);
1475
1476 /*
1477 * Special-case the `Nd' because we need to put the description
1478 * into the document table.
1479 */
1480
1481 for (n = n->child; NULL != n; n = n->next) {
1482 if (MDOC_TEXT == n->type) {
1483 if (NULL != mpage->desc) {
1484 sz = strlen(mpage->desc) +
1485 strlen(n->string) + 2;
1486 mpage->desc = mandoc_realloc(
1487 mpage->desc, sz);
1488 strlcat(mpage->desc, " ", sz);
1489 strlcat(mpage->desc, n->string, sz);
1490 } else
1491 mpage->desc = mandoc_strdup(n->string);
1492 }
1493 if (NULL != n->child)
1494 parse_mdoc_Nd(mpage, n);
1495 }
1496 return(1);
1497 }
1498
1499 static int
parse_mdoc_Nm(struct mpage * mpage,const struct mdoc_node * n)1500 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1501 {
1502
1503 return(SEC_NAME == n->sec ||
1504 (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type));
1505 }
1506
1507 static int
parse_mdoc_Sh(struct mpage * mpage,const struct mdoc_node * n)1508 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1509 {
1510
1511 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1512 }
1513
1514 static int
parse_mdoc_head(struct mpage * mpage,const struct mdoc_node * n)1515 parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1516 {
1517
1518 return(MDOC_HEAD == n->type);
1519 }
1520
1521 static int
parse_mdoc_body(struct mpage * mpage,const struct mdoc_node * n)1522 parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1523 {
1524
1525 return(MDOC_BODY == n->type);
1526 }
1527
1528 /*
1529 * Add a string to the hash table for the current manual.
1530 * Each string has a bitmask telling which macros it belongs to.
1531 * When we finish the manual, we'll dump the table.
1532 */
1533 static void
putkeys(const struct mpage * mpage,const char * cp,size_t sz,uint64_t v)1534 putkeys(const struct mpage *mpage,
1535 const char *cp, size_t sz, uint64_t v)
1536 {
1537 struct str *s;
1538 unsigned int slot;
1539 const char *end;
1540
1541 if (0 == sz)
1542 return;
1543
1544 end = cp + sz;
1545 slot = ohash_qlookupi(&strings, cp, &end);
1546 s = ohash_find(&strings, slot);
1547
1548 if (NULL != s && mpage == s->mpage) {
1549 s->mask |= v;
1550 return;
1551 } else if (NULL == s) {
1552 s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1553 memcpy(s->key, cp, sz);
1554 ohash_insert(&strings, slot, s);
1555 }
1556 s->mpage = mpage;
1557 s->mask = v;
1558 }
1559
1560 /*
1561 * Take a Unicode codepoint and produce its UTF-8 encoding.
1562 * This isn't the best way to do this, but it works.
1563 * The magic numbers are from the UTF-8 packaging.
1564 * They're not as scary as they seem: read the UTF-8 spec for details.
1565 */
1566 static size_t
utf8(unsigned int cp,char out[7])1567 utf8(unsigned int cp, char out[7])
1568 {
1569 size_t rc;
1570
1571 rc = 0;
1572 if (cp <= 0x0000007F) {
1573 rc = 1;
1574 out[0] = (char)cp;
1575 } else if (cp <= 0x000007FF) {
1576 rc = 2;
1577 out[0] = (cp >> 6 & 31) | 192;
1578 out[1] = (cp & 63) | 128;
1579 } else if (cp <= 0x0000FFFF) {
1580 rc = 3;
1581 out[0] = (cp >> 12 & 15) | 224;
1582 out[1] = (cp >> 6 & 63) | 128;
1583 out[2] = (cp & 63) | 128;
1584 } else if (cp <= 0x001FFFFF) {
1585 rc = 4;
1586 out[0] = (cp >> 18 & 7) | 240;
1587 out[1] = (cp >> 12 & 63) | 128;
1588 out[2] = (cp >> 6 & 63) | 128;
1589 out[3] = (cp & 63) | 128;
1590 } else if (cp <= 0x03FFFFFF) {
1591 rc = 5;
1592 out[0] = (cp >> 24 & 3) | 248;
1593 out[1] = (cp >> 18 & 63) | 128;
1594 out[2] = (cp >> 12 & 63) | 128;
1595 out[3] = (cp >> 6 & 63) | 128;
1596 out[4] = (cp & 63) | 128;
1597 } else if (cp <= 0x7FFFFFFF) {
1598 rc = 6;
1599 out[0] = (cp >> 30 & 1) | 252;
1600 out[1] = (cp >> 24 & 63) | 128;
1601 out[2] = (cp >> 18 & 63) | 128;
1602 out[3] = (cp >> 12 & 63) | 128;
1603 out[4] = (cp >> 6 & 63) | 128;
1604 out[5] = (cp & 63) | 128;
1605 } else
1606 return(0);
1607
1608 out[rc] = '\0';
1609 return(rc);
1610 }
1611
1612 /*
1613 * Store the rendered version of a key, or alias the pointer
1614 * if the key contains no escape sequences.
1615 */
1616 static void
render_key(struct mchars * mc,struct str * key)1617 render_key(struct mchars *mc, struct str *key)
1618 {
1619 size_t sz, bsz, pos;
1620 char utfbuf[7], res[5];
1621 char *buf;
1622 const char *seq, *cpp, *val;
1623 int len, u;
1624 enum mandoc_esc esc;
1625
1626 assert(NULL == key->rendered);
1627
1628 res[0] = '\\';
1629 res[1] = '\t';
1630 res[2] = ASCII_NBRSP;
1631 res[3] = ASCII_HYPH;
1632 res[4] = '\0';
1633
1634 val = key->key;
1635 bsz = strlen(val);
1636
1637 /*
1638 * Pre-check: if we have no stop-characters, then set the
1639 * pointer as ourselvse and get out of here.
1640 */
1641 if (strcspn(val, res) == bsz) {
1642 key->rendered = key->key;
1643 return;
1644 }
1645
1646 /* Pre-allocate by the length of the input */
1647
1648 buf = mandoc_malloc(++bsz);
1649 pos = 0;
1650
1651 while ('\0' != *val) {
1652 /*
1653 * Halt on the first escape sequence.
1654 * This also halts on the end of string, in which case
1655 * we just copy, fallthrough, and exit the loop.
1656 */
1657 if ((sz = strcspn(val, res)) > 0) {
1658 memcpy(&buf[pos], val, sz);
1659 pos += sz;
1660 val += sz;
1661 }
1662
1663 if (ASCII_HYPH == *val) {
1664 buf[pos++] = '-';
1665 val++;
1666 continue;
1667 } else if ('\t' == *val || ASCII_NBRSP == *val) {
1668 buf[pos++] = ' ';
1669 val++;
1670 continue;
1671 } else if ('\\' != *val)
1672 break;
1673
1674 /* Read past the slash. */
1675
1676 val++;
1677
1678 /*
1679 * Parse the escape sequence and see if it's a
1680 * predefined character or special character.
1681 */
1682
1683 esc = mandoc_escape
1684 ((const char **)&val, &seq, &len);
1685 if (ESCAPE_ERROR == esc)
1686 break;
1687 if (ESCAPE_SPECIAL != esc)
1688 continue;
1689
1690 /*
1691 * Render the special character
1692 * as either UTF-8 or ASCII.
1693 */
1694
1695 if (write_utf8) {
1696 if (0 == (u = mchars_spec2cp(mc, seq, len)))
1697 continue;
1698 cpp = utfbuf;
1699 if (0 == (sz = utf8(u, utfbuf)))
1700 continue;
1701 sz = strlen(cpp);
1702 } else {
1703 cpp = mchars_spec2str(mc, seq, len, &sz);
1704 if (NULL == cpp)
1705 continue;
1706 if (ASCII_NBRSP == *cpp) {
1707 cpp = " ";
1708 sz = 1;
1709 }
1710 }
1711
1712 /* Copy the rendered glyph into the stream. */
1713
1714 bsz += sz;
1715 buf = mandoc_realloc(buf, bsz);
1716 memcpy(&buf[pos], cpp, sz);
1717 pos += sz;
1718 }
1719
1720 buf[pos] = '\0';
1721 key->rendered = buf;
1722 }
1723
1724 /*
1725 * Flush the current page's terms (and their bits) into the database.
1726 * Wrap the entire set of additions in a transaction to make sqlite be a
1727 * little faster.
1728 * Also, handle escape sequences at the last possible moment.
1729 */
1730 static void
dbindex(const struct mpage * mpage,struct mchars * mc)1731 dbindex(const struct mpage *mpage, struct mchars *mc)
1732 {
1733 struct mlink *mlink;
1734 struct str *key;
1735 const char *desc;
1736 int64_t recno;
1737 size_t i;
1738 unsigned int slot;
1739
1740 if (verb)
1741 say(mpage->mlinks->file, "Adding to index");
1742
1743 if (nodb)
1744 return;
1745
1746 desc = "";
1747 if (NULL != mpage->desc && '\0' != *mpage->desc) {
1748 key = ohash_find(&strings,
1749 ohash_qlookup(&strings, mpage->desc));
1750 assert(NULL != key);
1751 if (NULL == key->rendered)
1752 render_key(mc, key);
1753 desc = key->rendered;
1754 }
1755
1756 SQL_EXEC("BEGIN TRANSACTION");
1757
1758 i = 1;
1759 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, desc);
1760 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
1761 SQL_STEP(stmts[STMT_INSERT_PAGE]);
1762 recno = sqlite3_last_insert_rowid(db);
1763 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1764
1765 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1766 i = 1;
1767 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->file);
1768 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1769 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1770 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1771 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, recno);
1772 SQL_STEP(stmts[STMT_INSERT_LINK]);
1773 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1774 }
1775
1776 for (key = ohash_first(&strings, &slot); NULL != key;
1777 key = ohash_next(&strings, &slot)) {
1778 assert(key->mpage == mpage);
1779 if (NULL == key->rendered)
1780 render_key(mc, key);
1781 i = 1;
1782 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1783 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
1784 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
1785 SQL_STEP(stmts[STMT_INSERT_KEY]);
1786 sqlite3_reset(stmts[STMT_INSERT_KEY]);
1787 if (key->rendered != key->key)
1788 free(key->rendered);
1789 free(key);
1790 }
1791
1792 SQL_EXEC("END TRANSACTION");
1793 }
1794
1795 static void
dbprune(void)1796 dbprune(void)
1797 {
1798 struct mpage *mpage;
1799 struct mlink *mlink;
1800 size_t i;
1801 unsigned int slot;
1802
1803 if (nodb)
1804 return;
1805
1806 mpage = ohash_first(&mpages, &slot);
1807 while (NULL != mpage) {
1808 mlink = mpage->mlinks;
1809 i = 1;
1810 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE], i, mlink->file);
1811 SQL_STEP(stmts[STMT_DELETE_PAGE]);
1812 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
1813 if (verb)
1814 say(mlink->file, "Deleted from index");
1815 mpage = ohash_next(&mpages, &slot);
1816 }
1817 }
1818
1819 /*
1820 * Close an existing database and its prepared statements.
1821 * If "real" is not set, rename the temporary file into the real one.
1822 */
1823 static void
dbclose(int real)1824 dbclose(int real)
1825 {
1826 size_t i;
1827
1828 if (nodb)
1829 return;
1830
1831 for (i = 0; i < STMT__MAX; i++) {
1832 sqlite3_finalize(stmts[i]);
1833 stmts[i] = NULL;
1834 }
1835
1836 sqlite3_close(db);
1837 db = NULL;
1838
1839 if (real)
1840 return;
1841
1842 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
1843 exitcode = (int)MANDOCLEVEL_SYSERR;
1844 say(MANDOC_DB, NULL);
1845 }
1846 }
1847
1848 /*
1849 * This is straightforward stuff.
1850 * Open a database connection to a "temporary" database, then open a set
1851 * of prepared statements we'll use over and over again.
1852 * If "real" is set, we use the existing database; if not, we truncate a
1853 * temporary one.
1854 * Must be matched by dbclose().
1855 */
1856 static int
dbopen(int real)1857 dbopen(int real)
1858 {
1859 const char *file, *sql;
1860 int rc, ofl;
1861
1862 if (nodb)
1863 return(1);
1864
1865 ofl = SQLITE_OPEN_READWRITE;
1866 if (0 == real) {
1867 file = MANDOC_DB "~";
1868 if (-1 == remove(file) && ENOENT != errno) {
1869 exitcode = (int)MANDOCLEVEL_SYSERR;
1870 say(file, NULL);
1871 return(0);
1872 }
1873 ofl |= SQLITE_OPEN_EXCLUSIVE;
1874 } else
1875 file = MANDOC_DB;
1876
1877 rc = sqlite3_open_v2(file, &db, ofl, NULL);
1878 if (SQLITE_OK == rc)
1879 goto prepare_statements;
1880 if (SQLITE_CANTOPEN != rc) {
1881 exitcode = (int)MANDOCLEVEL_SYSERR;
1882 say(file, NULL);
1883 return(0);
1884 }
1885
1886 sqlite3_close(db);
1887 db = NULL;
1888
1889 if (SQLITE_OK != (rc = sqlite3_open(file, &db))) {
1890 exitcode = (int)MANDOCLEVEL_SYSERR;
1891 say(file, NULL);
1892 return(0);
1893 }
1894
1895 sql = "CREATE TABLE \"mpages\" (\n"
1896 " \"desc\" TEXT NOT NULL,\n"
1897 " \"form\" INTEGER NOT NULL,\n"
1898 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1899 ");\n"
1900 "\n"
1901 "CREATE TABLE \"mlinks\" (\n"
1902 " \"file\" TEXT NOT NULL,\n"
1903 " \"sec\" TEXT NOT NULL,\n"
1904 " \"arch\" TEXT NOT NULL,\n"
1905 " \"name\" TEXT NOT NULL,\n"
1906 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1907 "ON DELETE CASCADE,\n"
1908 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1909 ");\n"
1910 "\n"
1911 "CREATE TABLE \"keys\" (\n"
1912 " \"bits\" INTEGER NOT NULL,\n"
1913 " \"key\" TEXT NOT NULL,\n"
1914 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1915 "ON DELETE CASCADE,\n"
1916 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1917 ");\n"
1918 "\n"
1919 "CREATE INDEX \"key_index\" ON keys (key);\n";
1920
1921 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1922 exitcode = (int)MANDOCLEVEL_SYSERR;
1923 say(file, "%s", sqlite3_errmsg(db));
1924 return(0);
1925 }
1926
1927 prepare_statements:
1928 SQL_EXEC("PRAGMA foreign_keys = ON");
1929 sql = "DELETE FROM mpages where file=?";
1930 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
1931 sql = "INSERT INTO mpages "
1932 "(desc,form) VALUES (?,?)";
1933 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
1934 sql = "INSERT INTO mlinks "
1935 "(file,sec,arch,name,pageid) VALUES (?,?,?,?,?)";
1936 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1937 sql = "INSERT INTO keys "
1938 "(bits,key,pageid) VALUES (?,?,?)";
1939 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1940
1941 #ifndef __APPLE__
1942 /*
1943 * When opening a new database, we can turn off
1944 * synchronous mode for much better performance.
1945 */
1946
1947 if (real)
1948 SQL_EXEC("PRAGMA synchronous = OFF");
1949 #endif
1950
1951 return(1);
1952 }
1953
1954 static void *
hash_halloc(size_t sz,void * arg)1955 hash_halloc(size_t sz, void *arg)
1956 {
1957
1958 return(mandoc_calloc(sz, 1));
1959 }
1960
1961 static void *
hash_alloc(size_t sz,void * arg)1962 hash_alloc(size_t sz, void *arg)
1963 {
1964
1965 return(mandoc_malloc(sz));
1966 }
1967
1968 static void
hash_free(void * p,size_t sz,void * arg)1969 hash_free(void *p, size_t sz, void *arg)
1970 {
1971
1972 free(p);
1973 }
1974
1975 static int
set_basedir(const char * targetdir)1976 set_basedir(const char *targetdir)
1977 {
1978 static char startdir[PATH_MAX];
1979 static int fd;
1980
1981 /*
1982 * Remember where we started by keeping a fd open to the origin
1983 * path component: throughout this utility, we chdir() a lot to
1984 * handle relative paths, and by doing this, we can return to
1985 * the starting point.
1986 */
1987 if ('\0' == *startdir) {
1988 if (NULL == getcwd(startdir, PATH_MAX)) {
1989 exitcode = (int)MANDOCLEVEL_SYSERR;
1990 if (NULL != targetdir)
1991 say(".", NULL);
1992 return(0);
1993 }
1994 if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
1995 exitcode = (int)MANDOCLEVEL_SYSERR;
1996 say(startdir, NULL);
1997 return(0);
1998 }
1999 if (NULL == targetdir)
2000 targetdir = startdir;
2001 } else {
2002 if (-1 == fd)
2003 return(0);
2004 if (-1 == fchdir(fd)) {
2005 close(fd);
2006 basedir[0] = '\0';
2007 exitcode = (int)MANDOCLEVEL_SYSERR;
2008 say(startdir, NULL);
2009 return(0);
2010 }
2011 if (NULL == targetdir) {
2012 close(fd);
2013 return(1);
2014 }
2015 }
2016 if (NULL == realpath(targetdir, basedir)) {
2017 basedir[0] = '\0';
2018 exitcode = (int)MANDOCLEVEL_BADARG;
2019 say(targetdir, NULL);
2020 return(0);
2021 } else if (-1 == chdir(basedir)) {
2022 exitcode = (int)MANDOCLEVEL_BADARG;
2023 say("", NULL);
2024 return(0);
2025 }
2026 return(1);
2027 }
2028
2029 static void
say(const char * file,const char * format,...)2030 say(const char *file, const char *format, ...)
2031 {
2032 va_list ap;
2033
2034 if ('\0' != *basedir)
2035 fprintf(stderr, "%s", basedir);
2036 if ('\0' != *basedir && '\0' != *file)
2037 fputs("//", stderr);
2038 if ('\0' != *file)
2039 fprintf(stderr, "%s", file);
2040 fputs(": ", stderr);
2041
2042 if (NULL == format) {
2043 perror(NULL);
2044 return;
2045 }
2046
2047 va_start(ap, format);
2048 vfprintf(stderr, format, ap);
2049 va_end(ap);
2050
2051 fputc('\n', stderr);
2052 }
2053