]> 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
Add an option -Q (quick) to mandocdb(8)
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.102 2014/01/05 20:26:36 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 nodb; /* no database changes */
171 static int quick; /* abort the parse early */
172 static int use_all; /* use all found files */
173 static int verb; /* print what we're doing */
174 static int warnings; /* warn about crap */
175 static int write_utf8; /* write UTF-8 output; else ASCII */
176 static int exitcode; /* to be returned by main */
177 static enum op op; /* operational mode */
178 static char basedir[PATH_MAX]; /* current base directory */
179 static struct ohash mpages; /* table of distinct manual pages */
180 static struct ohash mlinks; /* table of directory entries */
181 static struct ohash strings; /* table of all strings */
182 static sqlite3 *db = NULL; /* current database */
183 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
184
185 static const struct mdoc_handler mdocs[MDOC_MAX] = {
186 { NULL, 0 }, /* Ap */
187 { NULL, 0 }, /* Dd */
188 { NULL, 0 }, /* Dt */
189 { NULL, 0 }, /* Os */
190 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
191 { parse_mdoc_head, TYPE_Ss }, /* Ss */
192 { NULL, 0 }, /* Pp */
193 { NULL, 0 }, /* D1 */
194 { NULL, 0 }, /* Dl */
195 { NULL, 0 }, /* Bd */
196 { NULL, 0 }, /* Ed */
197 { NULL, 0 }, /* Bl */
198 { NULL, 0 }, /* El */
199 { NULL, 0 }, /* It */
200 { NULL, 0 }, /* Ad */
201 { NULL, TYPE_An }, /* An */
202 { NULL, TYPE_Ar }, /* Ar */
203 { NULL, TYPE_Cd }, /* Cd */
204 { NULL, TYPE_Cm }, /* Cm */
205 { NULL, TYPE_Dv }, /* Dv */
206 { NULL, TYPE_Er }, /* Er */
207 { NULL, TYPE_Ev }, /* Ev */
208 { NULL, 0 }, /* Ex */
209 { NULL, TYPE_Fa }, /* Fa */
210 { parse_mdoc_Fd, 0 }, /* Fd */
211 { NULL, TYPE_Fl }, /* Fl */
212 { parse_mdoc_Fn, 0 }, /* Fn */
213 { NULL, TYPE_Ft }, /* Ft */
214 { NULL, TYPE_Ic }, /* Ic */
215 { NULL, TYPE_In }, /* In */
216 { NULL, TYPE_Li }, /* Li */
217 { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
218 { parse_mdoc_Nm, TYPE_Nm }, /* Nm */
219 { NULL, 0 }, /* Op */
220 { NULL, 0 }, /* Ot */
221 { NULL, TYPE_Pa }, /* Pa */
222 { NULL, 0 }, /* Rv */
223 { NULL, TYPE_St }, /* St */
224 { NULL, TYPE_Va }, /* Va */
225 { parse_mdoc_body, TYPE_Va }, /* Vt */
226 { parse_mdoc_Xr, 0 }, /* Xr */
227 { NULL, 0 }, /* %A */
228 { NULL, 0 }, /* %B */
229 { NULL, 0 }, /* %D */
230 { NULL, 0 }, /* %I */
231 { NULL, 0 }, /* %J */
232 { NULL, 0 }, /* %N */
233 { NULL, 0 }, /* %O */
234 { NULL, 0 }, /* %P */
235 { NULL, 0 }, /* %R */
236 { NULL, 0 }, /* %T */
237 { NULL, 0 }, /* %V */
238 { NULL, 0 }, /* Ac */
239 { NULL, 0 }, /* Ao */
240 { NULL, 0 }, /* Aq */
241 { NULL, TYPE_At }, /* At */
242 { NULL, 0 }, /* Bc */
243 { NULL, 0 }, /* Bf */
244 { NULL, 0 }, /* Bo */
245 { NULL, 0 }, /* Bq */
246 { NULL, TYPE_Bsx }, /* Bsx */
247 { NULL, TYPE_Bx }, /* Bx */
248 { NULL, 0 }, /* Db */
249 { NULL, 0 }, /* Dc */
250 { NULL, 0 }, /* Do */
251 { NULL, 0 }, /* Dq */
252 { NULL, 0 }, /* Ec */
253 { NULL, 0 }, /* Ef */
254 { NULL, TYPE_Em }, /* Em */
255 { NULL, 0 }, /* Eo */
256 { NULL, TYPE_Fx }, /* Fx */
257 { NULL, TYPE_Ms }, /* Ms */
258 { NULL, 0 }, /* No */
259 { NULL, 0 }, /* Ns */
260 { NULL, TYPE_Nx }, /* Nx */
261 { NULL, TYPE_Ox }, /* Ox */
262 { NULL, 0 }, /* Pc */
263 { NULL, 0 }, /* Pf */
264 { NULL, 0 }, /* Po */
265 { NULL, 0 }, /* Pq */
266 { NULL, 0 }, /* Qc */
267 { NULL, 0 }, /* Ql */
268 { NULL, 0 }, /* Qo */
269 { NULL, 0 }, /* Qq */
270 { NULL, 0 }, /* Re */
271 { NULL, 0 }, /* Rs */
272 { NULL, 0 }, /* Sc */
273 { NULL, 0 }, /* So */
274 { NULL, 0 }, /* Sq */
275 { NULL, 0 }, /* Sm */
276 { NULL, 0 }, /* Sx */
277 { NULL, TYPE_Sy }, /* Sy */
278 { NULL, TYPE_Tn }, /* Tn */
279 { NULL, 0 }, /* Ux */
280 { NULL, 0 }, /* Xc */
281 { NULL, 0 }, /* Xo */
282 { parse_mdoc_head, 0 }, /* Fo */
283 { NULL, 0 }, /* Fc */
284 { NULL, 0 }, /* Oo */
285 { NULL, 0 }, /* Oc */
286 { NULL, 0 }, /* Bk */
287 { NULL, 0 }, /* Ek */
288 { NULL, 0 }, /* Bt */
289 { NULL, 0 }, /* Hf */
290 { NULL, 0 }, /* Fr */
291 { NULL, 0 }, /* Ud */
292 { NULL, TYPE_Lb }, /* Lb */
293 { NULL, 0 }, /* Lp */
294 { NULL, TYPE_Lk }, /* Lk */
295 { NULL, TYPE_Mt }, /* Mt */
296 { NULL, 0 }, /* Brq */
297 { NULL, 0 }, /* Bro */
298 { NULL, 0 }, /* Brc */
299 { NULL, 0 }, /* %C */
300 { NULL, 0 }, /* Es */
301 { NULL, 0 }, /* En */
302 { NULL, TYPE_Dx }, /* Dx */
303 { NULL, 0 }, /* %Q */
304 { NULL, 0 }, /* br */
305 { NULL, 0 }, /* sp */
306 { NULL, 0 }, /* %U */
307 { NULL, 0 }, /* Ta */
308 };
309
310 int
311 main(int argc, char *argv[])
312 {
313 int ch, i;
314 size_t j, sz;
315 const char *path_arg;
316 struct mchars *mc;
317 struct manpaths dirs;
318 struct mparse *mp;
319 struct ohash_info mpages_info, mlinks_info;
320
321 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
322 memset(&dirs, 0, sizeof(struct manpaths));
323
324 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
325 mpages_info.halloc = mlinks_info.halloc = hash_halloc;
326 mpages_info.hfree = mlinks_info.hfree = hash_free;
327
328 mpages_info.key_offset = offsetof(struct mpage, inodev);
329 mlinks_info.key_offset = offsetof(struct mlink, file);
330
331 progname = strrchr(argv[0], '/');
332 if (progname == NULL)
333 progname = argv[0];
334 else
335 ++progname;
336
337 /*
338 * We accept a few different invocations.
339 * The CHECKOP macro makes sure that invocation styles don't
340 * clobber each other.
341 */
342 #define CHECKOP(_op, _ch) do \
343 if (OP_DEFAULT != (_op)) { \
344 fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
345 goto usage; \
346 } while (/*CONSTCOND*/0)
347
348 path_arg = NULL;
349 op = OP_DEFAULT;
350
351 while (-1 != (ch = getopt(argc, argv, "aC:d:nQT:tu:vW")))
352 switch (ch) {
353 case ('a'):
354 use_all = 1;
355 break;
356 case ('C'):
357 CHECKOP(op, ch);
358 path_arg = optarg;
359 op = OP_CONFFILE;
360 break;
361 case ('d'):
362 CHECKOP(op, ch);
363 path_arg = optarg;
364 op = OP_UPDATE;
365 break;
366 case ('n'):
367 nodb = 1;
368 break;
369 case ('Q'):
370 quick = 1;
371 break;
372 case ('T'):
373 if (strcmp(optarg, "utf8")) {
374 fprintf(stderr, "-T%s: Unsupported "
375 "output format\n", optarg);
376 goto usage;
377 }
378 write_utf8 = 1;
379 break;
380 case ('t'):
381 CHECKOP(op, ch);
382 dup2(STDOUT_FILENO, STDERR_FILENO);
383 op = OP_TEST;
384 nodb = warnings = 1;
385 break;
386 case ('u'):
387 CHECKOP(op, ch);
388 path_arg = optarg;
389 op = OP_DELETE;
390 break;
391 case ('v'):
392 verb++;
393 break;
394 case ('W'):
395 warnings = 1;
396 break;
397 default:
398 goto usage;
399 }
400
401 argc -= optind;
402 argv += optind;
403
404 if (OP_CONFFILE == op && argc > 0) {
405 fprintf(stderr, "-C: Too many arguments\n");
406 goto usage;
407 }
408
409 exitcode = (int)MANDOCLEVEL_OK;
410 mp = mparse_alloc(MPARSE_AUTO,
411 MANDOCLEVEL_FATAL, NULL, NULL, quick);
412 mc = mchars_alloc();
413
414 ohash_init(&mpages, 6, &mpages_info);
415 ohash_init(&mlinks, 6, &mlinks_info);
416
417 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
418 /*
419 * Force processing all files.
420 */
421 use_all = 1;
422
423 /*
424 * All of these deal with a specific directory.
425 * Jump into that directory then collect files specified
426 * on the command-line.
427 */
428 if (0 == set_basedir(path_arg))
429 goto out;
430 for (i = 0; i < argc; i++)
431 filescan(argv[i]);
432 if (0 == dbopen(1))
433 goto out;
434 if (OP_TEST != op)
435 dbprune();
436 if (OP_DELETE != op)
437 mpages_merge(mc, mp);
438 dbclose(1);
439 } else {
440 /*
441 * If we have arguments, use them as our manpaths.
442 * If we don't, grok from manpath(1) or however else
443 * manpath_parse() wants to do it.
444 */
445 if (argc > 0) {
446 dirs.paths = mandoc_calloc
447 (argc, sizeof(char *));
448 dirs.sz = (size_t)argc;
449 for (i = 0; i < argc; i++)
450 dirs.paths[i] = mandoc_strdup(argv[i]);
451 } else
452 manpath_parse(&dirs, path_arg, NULL, NULL);
453
454 /*
455 * First scan the tree rooted at a base directory, then
456 * build a new database and finally move it into place.
457 * Ignore zero-length directories and strip trailing
458 * slashes.
459 */
460 for (j = 0; j < dirs.sz; j++) {
461 sz = strlen(dirs.paths[j]);
462 if (sz && '/' == dirs.paths[j][sz - 1])
463 dirs.paths[j][--sz] = '\0';
464 if (0 == sz)
465 continue;
466
467 if (j) {
468 ohash_init(&mpages, 6, &mpages_info);
469 ohash_init(&mlinks, 6, &mlinks_info);
470 }
471
472 if (0 == set_basedir(dirs.paths[j]))
473 goto out;
474 if (0 == treescan())
475 goto out;
476 if (0 == set_basedir(dirs.paths[j]))
477 goto out;
478 if (0 == dbopen(0))
479 goto out;
480
481 mpages_merge(mc, mp);
482 dbclose(0);
483
484 if (j + 1 < dirs.sz) {
485 mpages_free();
486 ohash_delete(&mpages);
487 ohash_delete(&mlinks);
488 }
489 }
490 }
491 out:
492 set_basedir(NULL);
493 manpath_free(&dirs);
494 mchars_free(mc);
495 mparse_free(mp);
496 mpages_free();
497 ohash_delete(&mpages);
498 ohash_delete(&mlinks);
499 return(exitcode);
500 usage:
501 fprintf(stderr, "usage: %s [-anQvW] [-C file] [-Tutf8]\n"
502 " %s [-anQvW] [-Tutf8] dir ...\n"
503 " %s [-nQvW] [-Tutf8] -d dir [file ...]\n"
504 " %s [-nvW] -u dir [file ...]\n"
505 " %s [-Q] -t file ...\n",
506 progname, progname, progname,
507 progname, progname);
508
509 return((int)MANDOCLEVEL_BADARG);
510 }
511
512 /*
513 * Scan a directory tree rooted at "basedir" for manpages.
514 * We use fts(), scanning directory parts along the way for clues to our
515 * section and architecture.
516 *
517 * If use_all has been specified, grok all files.
518 * If not, sanitise paths to the following:
519 *
520 * [./]man*[/<arch>]/<name>.<section>
521 * or
522 * [./]cat<section>[/<arch>]/<name>.0
523 *
524 * TODO: accomodate for multi-language directories.
525 */
526 static int
527 treescan(void)
528 {
529 FTS *f;
530 FTSENT *ff;
531 struct mlink *mlink;
532 int dform;
533 char *dsec, *arch, *fsec, *cp;
534 const char *path;
535 const char *argv[2];
536
537 argv[0] = ".";
538 argv[1] = (char *)NULL;
539
540 /*
541 * Walk through all components under the directory, using the
542 * logical descent of files.
543 */
544 f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
545 if (NULL == f) {
546 exitcode = (int)MANDOCLEVEL_SYSERR;
547 say("", NULL);
548 return(0);
549 }
550
551 dsec = arch = NULL;
552 dform = FORM_NONE;
553
554 while (NULL != (ff = fts_read(f))) {
555 path = ff->fts_path + 2;
556 /*
557 * If we're a regular file, add an mlink by using the
558 * stored directory data and handling the filename.
559 */
560 if (FTS_F == ff->fts_info) {
561 if (0 == strcmp(path, MANDOC_DB))
562 continue;
563 if ( ! use_all && ff->fts_level < 2) {
564 if (warnings)
565 say(path, "Extraneous file");
566 continue;
567 } else if (NULL == (fsec =
568 strrchr(ff->fts_name, '.'))) {
569 if ( ! use_all) {
570 if (warnings)
571 say(path,
572 "No filename suffix");
573 continue;
574 }
575 } else if (0 == strcmp(++fsec, "html")) {
576 if (warnings)
577 say(path, "Skip html");
578 continue;
579 } else if (0 == strcmp(fsec, "gz")) {
580 if (warnings)
581 say(path, "Skip gz");
582 continue;
583 } else if (0 == strcmp(fsec, "ps")) {
584 if (warnings)
585 say(path, "Skip ps");
586 continue;
587 } else if (0 == strcmp(fsec, "pdf")) {
588 if (warnings)
589 say(path, "Skip pdf");
590 continue;
591 } else if ( ! use_all &&
592 ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
593 (FORM_CAT == dform && strcmp(fsec, "0")))) {
594 if (warnings)
595 say(path, "Wrong filename suffix");
596 continue;
597 } else
598 fsec[-1] = '\0';
599
600 mlink = mandoc_calloc(1, sizeof(struct mlink));
601 strlcpy(mlink->file, path, sizeof(mlink->file));
602 mlink->dform = dform;
603 mlink->dsec = dsec;
604 mlink->arch = arch;
605 mlink->name = ff->fts_name;
606 mlink->fsec = fsec;
607 mlink_add(mlink, ff->fts_statp);
608 continue;
609 } else if (FTS_D != ff->fts_info &&
610 FTS_DP != ff->fts_info) {
611 if (warnings)
612 say(path, "Not a regular file");
613 continue;
614 }
615
616 switch (ff->fts_level) {
617 case (0):
618 /* Ignore the root directory. */
619 break;
620 case (1):
621 /*
622 * This might contain manX/ or catX/.
623 * Try to infer this from the name.
624 * If we're not in use_all, enforce it.
625 */
626 cp = ff->fts_name;
627 if (FTS_DP == ff->fts_info)
628 break;
629
630 if (0 == strncmp(cp, "man", 3)) {
631 dform = FORM_SRC;
632 dsec = cp + 3;
633 } else if (0 == strncmp(cp, "cat", 3)) {
634 dform = FORM_CAT;
635 dsec = cp + 3;
636 } else {
637 dform = FORM_NONE;
638 dsec = NULL;
639 }
640
641 if (NULL != dsec || use_all)
642 break;
643
644 if (warnings)
645 say(path, "Unknown directory part");
646 fts_set(f, ff, FTS_SKIP);
647 break;
648 case (2):
649 /*
650 * Possibly our architecture.
651 * If we're descending, keep tabs on it.
652 */
653 if (FTS_DP != ff->fts_info && NULL != dsec)
654 arch = ff->fts_name;
655 else
656 arch = NULL;
657 break;
658 default:
659 if (FTS_DP == ff->fts_info || use_all)
660 break;
661 if (warnings)
662 say(path, "Extraneous directory part");
663 fts_set(f, ff, FTS_SKIP);
664 break;
665 }
666 }
667
668 fts_close(f);
669 return(1);
670 }
671
672 /*
673 * Add a file to the mlinks table.
674 * Do not verify that it's a "valid" looking manpage (we'll do that
675 * later).
676 *
677 * Try to infer the manual section, architecture, and page name from the
678 * path, assuming it looks like
679 *
680 * [./]man*[/<arch>]/<name>.<section>
681 * or
682 * [./]cat<section>[/<arch>]/<name>.0
683 *
684 * See treescan() for the fts(3) version of this.
685 */
686 static void
687 filescan(const char *file)
688 {
689 char buf[PATH_MAX];
690 struct stat st;
691 struct mlink *mlink;
692 char *p, *start;
693
694 assert(use_all);
695
696 if (0 == strncmp(file, "./", 2))
697 file += 2;
698
699 if (NULL == realpath(file, buf)) {
700 exitcode = (int)MANDOCLEVEL_BADARG;
701 say(file, NULL);
702 return;
703 } else if (OP_TEST != op && strstr(buf, basedir) != buf) {
704 exitcode = (int)MANDOCLEVEL_BADARG;
705 say("", "%s: outside base directory", buf);
706 return;
707 } else if (-1 == stat(buf, &st)) {
708 exitcode = (int)MANDOCLEVEL_BADARG;
709 say(file, NULL);
710 return;
711 } else if ( ! (S_IFREG & st.st_mode)) {
712 exitcode = (int)MANDOCLEVEL_BADARG;
713 say(file, "Not a regular file");
714 return;
715 }
716 start = buf + strlen(basedir);
717 mlink = mandoc_calloc(1, sizeof(struct mlink));
718 strlcpy(mlink->file, start, sizeof(mlink->file));
719
720 /*
721 * First try to guess our directory structure.
722 * If we find a separator, try to look for man* or cat*.
723 * If we find one of these and what's underneath is a directory,
724 * assume it's an architecture.
725 */
726 if (NULL != (p = strchr(start, '/'))) {
727 *p++ = '\0';
728 if (0 == strncmp(start, "man", 3)) {
729 mlink->dform = FORM_SRC;
730 mlink->dsec = start + 3;
731 } else if (0 == strncmp(start, "cat", 3)) {
732 mlink->dform = FORM_CAT;
733 mlink->dsec = start + 3;
734 }
735
736 start = p;
737 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
738 *p++ = '\0';
739 mlink->arch = start;
740 start = p;
741 }
742 }
743
744 /*
745 * Now check the file suffix.
746 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
747 */
748 p = strrchr(start, '\0');
749 while (p-- > start && '/' != *p && '.' != *p)
750 /* Loop. */ ;
751
752 if ('.' == *p) {
753 *p++ = '\0';
754 mlink->fsec = p;
755 }
756
757 /*
758 * Now try to parse the name.
759 * Use the filename portion of the path.
760 */
761 mlink->name = start;
762 if (NULL != (p = strrchr(start, '/'))) {
763 mlink->name = p + 1;
764 *p = '\0';
765 }
766 mlink_add(mlink, &st);
767 }
768
769 static void
770 mlink_add(struct mlink *mlink, const struct stat *st)
771 {
772 struct inodev inodev;
773 struct mpage *mpage;
774 unsigned int slot;
775
776 assert(NULL != mlink->file);
777
778 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
779 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
780 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
781 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
782
783 if ('0' == *mlink->fsec) {
784 free(mlink->fsec);
785 mlink->fsec = mandoc_strdup(mlink->dsec);
786 mlink->fform = FORM_CAT;
787 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
788 mlink->fform = FORM_SRC;
789 else
790 mlink->fform = FORM_NONE;
791
792 slot = ohash_qlookup(&mlinks, mlink->file);
793 assert(NULL == ohash_find(&mlinks, slot));
794 ohash_insert(&mlinks, slot, mlink);
795
796 inodev.st_ino = st->st_ino;
797 inodev.st_dev = st->st_dev;
798 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
799 sizeof(struct inodev), inodev.st_ino);
800 mpage = ohash_find(&mpages, slot);
801 if (NULL == mpage) {
802 mpage = mandoc_calloc(1, sizeof(struct mpage));
803 mpage->inodev.st_ino = inodev.st_ino;
804 mpage->inodev.st_dev = inodev.st_dev;
805 ohash_insert(&mpages, slot, mpage);
806 } else
807 mlink->next = mpage->mlinks;
808 mpage->mlinks = mlink;
809 }
810
811 static void
812 mlink_free(struct mlink *mlink)
813 {
814
815 free(mlink->dsec);
816 free(mlink->arch);
817 free(mlink->name);
818 free(mlink->fsec);
819 free(mlink);
820 }
821
822 static void
823 mpages_free(void)
824 {
825 struct mpage *mpage;
826 struct mlink *mlink;
827 unsigned int slot;
828
829 mpage = ohash_first(&mpages, &slot);
830 while (NULL != mpage) {
831 while (NULL != (mlink = mpage->mlinks)) {
832 mpage->mlinks = mlink->next;
833 mlink_free(mlink);
834 }
835 free(mpage->sec);
836 free(mpage->arch);
837 free(mpage->title);
838 free(mpage->desc);
839 free(mpage);
840 mpage = ohash_next(&mpages, &slot);
841 }
842 }
843
844 /*
845 * For each mlink to the mpage, check whether the path looks like
846 * it is formatted, and if it does, check whether a source manual
847 * exists by the same name, ignoring the suffix.
848 * If both conditions hold, drop the mlink.
849 */
850 static void
851 mlinks_undupe(struct mpage *mpage)
852 {
853 char buf[PATH_MAX];
854 struct mlink **prev;
855 struct mlink *mlink;
856 char *bufp;
857
858 mpage->form = FORM_CAT;
859 prev = &mpage->mlinks;
860 while (NULL != (mlink = *prev)) {
861 if (FORM_CAT != mlink->dform) {
862 mpage->form = FORM_NONE;
863 goto nextlink;
864 }
865 if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) {
866 if (warnings)
867 say(mlink->file, "Filename too long");
868 goto nextlink;
869 }
870 bufp = strstr(buf, "cat");
871 assert(NULL != bufp);
872 memcpy(bufp, "man", 3);
873 if (NULL != (bufp = strrchr(buf, '.')))
874 *++bufp = '\0';
875 strlcat(buf, mlink->dsec, PATH_MAX);
876 if (NULL == ohash_find(&mlinks,
877 ohash_qlookup(&mlinks, buf)))
878 goto nextlink;
879 if (warnings)
880 say(mlink->file, "Man source exists: %s", buf);
881 if (use_all)
882 goto nextlink;
883 *prev = mlink->next;
884 mlink_free(mlink);
885 continue;
886 nextlink:
887 prev = &(*prev)->next;
888 }
889 }
890
891 static int
892 mlink_check(struct mpage *mpage, struct mlink *mlink)
893 {
894 int match;
895
896 match = 1;
897
898 /*
899 * Check whether the manual section given in a file
900 * agrees with the directory where the file is located.
901 * Some manuals have suffixes like (3p) on their
902 * section number either inside the file or in the
903 * directory name, some are linked into more than one
904 * section, like encrypt(1) = makekey(8).
905 */
906
907 if (FORM_SRC == mpage->form &&
908 strcasecmp(mpage->sec, mlink->dsec)) {
909 match = 0;
910 say(mlink->file, "Section \"%s\" manual in %s directory",
911 mpage->sec, mlink->dsec);
912 }
913
914 /*
915 * Manual page directories exist for each kernel
916 * architecture as returned by machine(1).
917 * However, many manuals only depend on the
918 * application architecture as returned by arch(1).
919 * For example, some (2/ARM) manuals are shared
920 * across the "armish" and "zaurus" kernel
921 * architectures.
922 * A few manuals are even shared across completely
923 * different architectures, for example fdformat(1)
924 * on amd64, i386, sparc, and sparc64.
925 */
926
927 if (strcasecmp(mpage->arch, mlink->arch)) {
928 match = 0;
929 say(mlink->file, "Architecture \"%s\" manual in "
930 "\"%s\" directory", mpage->arch, mlink->arch);
931 }
932
933 if (strcasecmp(mpage->title, mlink->name))
934 match = 0;
935
936 return(match);
937 }
938
939 /*
940 * Run through the files in the global vector "mpages"
941 * and add them to the database specified in "basedir".
942 *
943 * This handles the parsing scheme itself, using the cues of directory
944 * and filename to determine whether the file is parsable or not.
945 */
946 static void
947 mpages_merge(struct mchars *mc, struct mparse *mp)
948 {
949 struct ohash_info str_info;
950 struct mpage *mpage;
951 struct mlink *mlink;
952 struct mdoc *mdoc;
953 struct man *man;
954 const char *cp;
955 int match;
956 unsigned int pslot;
957 enum mandoclevel lvl;
958
959 str_info.alloc = hash_alloc;
960 str_info.halloc = hash_halloc;
961 str_info.hfree = hash_free;
962 str_info.key_offset = offsetof(struct str, key);
963
964 mpage = ohash_first(&mpages, &pslot);
965 while (NULL != mpage) {
966 mlinks_undupe(mpage);
967 if (NULL == mpage->mlinks) {
968 mpage = ohash_next(&mpages, &pslot);
969 continue;
970 }
971
972 ohash_init(&strings, 6, &str_info);
973 mparse_reset(mp);
974 mdoc = NULL;
975 man = NULL;
976
977 /*
978 * Try interpreting the file as mdoc(7) or man(7)
979 * source code, unless it is already known to be
980 * formatted. Fall back to formatted mode.
981 */
982 if (FORM_CAT != mpage->mlinks->dform ||
983 FORM_CAT != mpage->mlinks->fform) {
984 lvl = mparse_readfd(mp, -1, mpage->mlinks->file);
985 if (lvl < MANDOCLEVEL_FATAL)
986 mparse_result(mp, &mdoc, &man);
987 }
988
989 if (NULL != mdoc) {
990 mpage->form = FORM_SRC;
991 mpage->sec =
992 mandoc_strdup(mdoc_meta(mdoc)->msec);
993 mpage->arch = mdoc_meta(mdoc)->arch;
994 mpage->arch = mandoc_strdup(
995 NULL == mpage->arch ? "" : mpage->arch);
996 mpage->title =
997 mandoc_strdup(mdoc_meta(mdoc)->title);
998 } else if (NULL != man) {
999 mpage->form = FORM_SRC;
1000 mpage->sec =
1001 mandoc_strdup(man_meta(man)->msec);
1002 mpage->arch =
1003 mandoc_strdup(mpage->mlinks->arch);
1004 mpage->title =
1005 mandoc_strdup(man_meta(man)->title);
1006 } else {
1007 mpage->form = FORM_CAT;
1008 mpage->sec =
1009 mandoc_strdup(mpage->mlinks->dsec);
1010 mpage->arch =
1011 mandoc_strdup(mpage->mlinks->arch);
1012 mpage->title =
1013 mandoc_strdup(mpage->mlinks->name);
1014 }
1015 putkey(mpage, mpage->sec, TYPE_sec);
1016 putkey(mpage, '\0' == *mpage->arch ?
1017 "any" : mpage->arch, TYPE_arch);
1018
1019 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1020 if ('\0' != *mlink->dsec)
1021 putkey(mpage, mlink->dsec, TYPE_sec);
1022 if ('\0' != *mlink->fsec)
1023 putkey(mpage, mlink->fsec, TYPE_sec);
1024 putkey(mpage, '\0' == *mlink->arch ?
1025 "any" : mlink->arch, TYPE_arch);
1026 putkey(mpage, mlink->name, TYPE_Nm);
1027 }
1028
1029 if (warnings && !use_all) {
1030 match = 0;
1031 for (mlink = mpage->mlinks; mlink;
1032 mlink = mlink->next)
1033 if (mlink_check(mpage, mlink))
1034 match = 1;
1035 } else
1036 match = 1;
1037
1038 if (NULL != mdoc) {
1039 if (NULL != (cp = mdoc_meta(mdoc)->name))
1040 putkey(mpage, cp, TYPE_Nm);
1041 assert(NULL == mpage->desc);
1042 parse_mdoc(mpage, mdoc_node(mdoc));
1043 putkey(mpage, NULL != mpage->desc ?
1044 mpage->desc : mpage->mlinks->name, TYPE_Nd);
1045 } else if (NULL != man)
1046 parse_man(mpage, man_node(man));
1047 else
1048 parse_cat(mpage);
1049
1050 dbindex(mpage, mc);
1051 ohash_delete(&strings);
1052 mpage = ohash_next(&mpages, &pslot);
1053 }
1054 }
1055
1056 static void
1057 parse_cat(struct mpage *mpage)
1058 {
1059 FILE *stream;
1060 char *line, *p, *title;
1061 size_t len, plen, titlesz;
1062
1063 if (NULL == (stream = fopen(mpage->mlinks->file, "r"))) {
1064 if (warnings)
1065 say(mpage->mlinks->file, NULL);
1066 return;
1067 }
1068
1069 /* Skip to first blank line. */
1070
1071 while (NULL != (line = fgetln(stream, &len)))
1072 if ('\n' == *line)
1073 break;
1074
1075 /*
1076 * Assume the first line that is not indented
1077 * is the first section header. Skip to it.
1078 */
1079
1080 while (NULL != (line = fgetln(stream, &len)))
1081 if ('\n' != *line && ' ' != *line)
1082 break;
1083
1084 /*
1085 * Read up until the next section into a buffer.
1086 * Strip the leading and trailing newline from each read line,
1087 * appending a trailing space.
1088 * Ignore empty (whitespace-only) lines.
1089 */
1090
1091 titlesz = 0;
1092 title = NULL;
1093
1094 while (NULL != (line = fgetln(stream, &len))) {
1095 if (' ' != *line || '\n' != line[len - 1])
1096 break;
1097 while (len > 0 && isspace((unsigned char)*line)) {
1098 line++;
1099 len--;
1100 }
1101 if (1 == len)
1102 continue;
1103 title = mandoc_realloc(title, titlesz + len);
1104 memcpy(title + titlesz, line, len);
1105 titlesz += len;
1106 title[titlesz - 1] = ' ';
1107 }
1108
1109 /*
1110 * If no page content can be found, or the input line
1111 * is already the next section header, or there is no
1112 * trailing newline, reuse the page title as the page
1113 * description.
1114 */
1115
1116 if (NULL == title || '\0' == *title) {
1117 if (warnings)
1118 say(mpage->mlinks->file,
1119 "Cannot find NAME section");
1120 assert(NULL == mpage->desc);
1121 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1122 putkey(mpage, mpage->mlinks->name, TYPE_Nd);
1123 fclose(stream);
1124 free(title);
1125 return;
1126 }
1127
1128 title = mandoc_realloc(title, titlesz + 1);
1129 title[titlesz] = '\0';
1130
1131 /*
1132 * Skip to the first dash.
1133 * Use the remaining line as the description (no more than 70
1134 * bytes).
1135 */
1136
1137 if (NULL != (p = strstr(title, "- "))) {
1138 for (p += 2; ' ' == *p || '\b' == *p; p++)
1139 /* Skip to next word. */ ;
1140 } else {
1141 if (warnings)
1142 say(mpage->mlinks->file,
1143 "No dash in title line");
1144 p = title;
1145 }
1146
1147 plen = strlen(p);
1148
1149 /* Strip backspace-encoding from line. */
1150
1151 while (NULL != (line = memchr(p, '\b', plen))) {
1152 len = line - p;
1153 if (0 == len) {
1154 memmove(line, line + 1, plen--);
1155 continue;
1156 }
1157 memmove(line - 1, line + 1, plen - len);
1158 plen -= 2;
1159 }
1160
1161 assert(NULL == mpage->desc);
1162 mpage->desc = mandoc_strdup(p);
1163 putkey(mpage, mpage->desc, TYPE_Nd);
1164 fclose(stream);
1165 free(title);
1166 }
1167
1168 /*
1169 * Put a type/word pair into the word database for this particular file.
1170 */
1171 static void
1172 putkey(const struct mpage *mpage, const char *value, uint64_t type)
1173 {
1174
1175 assert(NULL != value);
1176 putkeys(mpage, value, strlen(value), type);
1177 }
1178
1179 /*
1180 * Grok all nodes at or below a certain mdoc node into putkey().
1181 */
1182 static void
1183 putmdockey(const struct mpage *mpage,
1184 const struct mdoc_node *n, uint64_t m)
1185 {
1186
1187 for ( ; NULL != n; n = n->next) {
1188 if (NULL != n->child)
1189 putmdockey(mpage, n->child, m);
1190 if (MDOC_TEXT == n->type)
1191 putkey(mpage, n->string, m);
1192 }
1193 }
1194
1195 static void
1196 parse_man(struct mpage *mpage, const struct man_node *n)
1197 {
1198 const struct man_node *head, *body;
1199 char *start, *sv, *title;
1200 char byte;
1201 size_t sz, titlesz;
1202
1203 if (NULL == n)
1204 return;
1205
1206 /*
1207 * We're only searching for one thing: the first text child in
1208 * the BODY of a NAME section. Since we don't keep track of
1209 * sections in -man, run some hoops to find out whether we're in
1210 * the correct section or not.
1211 */
1212
1213 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1214 body = n;
1215 assert(body->parent);
1216 if (NULL != (head = body->parent->head) &&
1217 1 == head->nchild &&
1218 NULL != (head = (head->child)) &&
1219 MAN_TEXT == head->type &&
1220 0 == strcmp(head->string, "NAME") &&
1221 NULL != (body = body->child) &&
1222 MAN_TEXT == body->type) {
1223
1224 title = NULL;
1225 titlesz = 0;
1226
1227 /*
1228 * Suck the entire NAME section into memory.
1229 * Yes, we might run away.
1230 * But too many manuals have big, spread-out
1231 * NAME sections over many lines.
1232 */
1233
1234 for ( ; NULL != body; body = body->next) {
1235 if (MAN_TEXT != body->type)
1236 break;
1237 if (0 == (sz = strlen(body->string)))
1238 continue;
1239 title = mandoc_realloc
1240 (title, titlesz + sz + 1);
1241 memcpy(title + titlesz, body->string, sz);
1242 titlesz += sz + 1;
1243 title[titlesz - 1] = ' ';
1244 }
1245 if (NULL == title)
1246 return;
1247
1248 title = mandoc_realloc(title, titlesz + 1);
1249 title[titlesz] = '\0';
1250
1251 /* Skip leading space. */
1252
1253 sv = title;
1254 while (isspace((unsigned char)*sv))
1255 sv++;
1256
1257 if (0 == (sz = strlen(sv))) {
1258 free(title);
1259 return;
1260 }
1261
1262 /* Erase trailing space. */
1263
1264 start = &sv[sz - 1];
1265 while (start > sv && isspace((unsigned char)*start))
1266 *start-- = '\0';
1267
1268 if (start == sv) {
1269 free(title);
1270 return;
1271 }
1272
1273 start = sv;
1274
1275 /*
1276 * Go through a special heuristic dance here.
1277 * Conventionally, one or more manual names are
1278 * comma-specified prior to a whitespace, then a
1279 * dash, then a description. Try to puzzle out
1280 * the name parts here.
1281 */
1282
1283 for ( ;; ) {
1284 sz = strcspn(start, " ,");
1285 if ('\0' == start[sz])
1286 break;
1287
1288 byte = start[sz];
1289 start[sz] = '\0';
1290
1291 putkey(mpage, start, TYPE_Nm);
1292
1293 if (' ' == byte) {
1294 start += sz + 1;
1295 break;
1296 }
1297
1298 assert(',' == byte);
1299 start += sz + 1;
1300 while (' ' == *start)
1301 start++;
1302 }
1303
1304 if (sv == start) {
1305 putkey(mpage, start, TYPE_Nm);
1306 free(title);
1307 return;
1308 }
1309
1310 while (isspace((unsigned char)*start))
1311 start++;
1312
1313 if (0 == strncmp(start, "-", 1))
1314 start += 1;
1315 else if (0 == strncmp(start, "\\-\\-", 4))
1316 start += 4;
1317 else if (0 == strncmp(start, "\\-", 2))
1318 start += 2;
1319 else if (0 == strncmp(start, "\\(en", 4))
1320 start += 4;
1321 else if (0 == strncmp(start, "\\(em", 4))
1322 start += 4;
1323
1324 while (' ' == *start)
1325 start++;
1326
1327 assert(NULL == mpage->desc);
1328 mpage->desc = mandoc_strdup(start);
1329 putkey(mpage, mpage->desc, TYPE_Nd);
1330 free(title);
1331 return;
1332 }
1333 }
1334
1335 for (n = n->child; n; n = n->next) {
1336 if (NULL != mpage->desc)
1337 break;
1338 parse_man(mpage, n);
1339 }
1340 }
1341
1342 static void
1343 parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1344 {
1345
1346 assert(NULL != n);
1347 for (n = n->child; NULL != n; n = n->next) {
1348 switch (n->type) {
1349 case (MDOC_ELEM):
1350 /* FALLTHROUGH */
1351 case (MDOC_BLOCK):
1352 /* FALLTHROUGH */
1353 case (MDOC_HEAD):
1354 /* FALLTHROUGH */
1355 case (MDOC_BODY):
1356 /* FALLTHROUGH */
1357 case (MDOC_TAIL):
1358 if (NULL != mdocs[n->tok].fp)
1359 if (0 == (*mdocs[n->tok].fp)(mpage, n))
1360 break;
1361 if (mdocs[n->tok].mask)
1362 putmdockey(mpage, n->child,
1363 mdocs[n->tok].mask);
1364 break;
1365 default:
1366 assert(MDOC_ROOT != n->type);
1367 continue;
1368 }
1369 if (NULL != n->child)
1370 parse_mdoc(mpage, n);
1371 }
1372 }
1373
1374 static int
1375 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1376 {
1377 const char *start, *end;
1378 size_t sz;
1379
1380 if (SEC_SYNOPSIS != n->sec ||
1381 NULL == (n = n->child) ||
1382 MDOC_TEXT != n->type)
1383 return(0);
1384
1385 /*
1386 * Only consider those `Fd' macro fields that begin with an
1387 * "inclusion" token (versus, e.g., #define).
1388 */
1389
1390 if (strcmp("#include", n->string))
1391 return(0);
1392
1393 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1394 return(0);
1395
1396 /*
1397 * Strip away the enclosing angle brackets and make sure we're
1398 * not zero-length.
1399 */
1400
1401 start = n->string;
1402 if ('<' == *start || '"' == *start)
1403 start++;
1404
1405 if (0 == (sz = strlen(start)))
1406 return(0);
1407
1408 end = &start[(int)sz - 1];
1409 if ('>' == *end || '"' == *end)
1410 end--;
1411
1412 if (end > start)
1413 putkeys(mpage, start, end - start + 1, TYPE_In);
1414 return(0);
1415 }
1416
1417 static int
1418 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1419 {
1420 const char *cp;
1421
1422 if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1423 return(0);
1424
1425 /*
1426 * Parse: .Fn "struct type *name" "char *arg".
1427 * First strip away pointer symbol.
1428 * Then store the function name, then type.
1429 * Finally, store the arguments.
1430 */
1431
1432 if (NULL == (cp = strrchr(n->string, ' ')))
1433 cp = n->string;
1434
1435 while ('*' == *cp)
1436 cp++;
1437
1438 putkey(mpage, cp, TYPE_Fn);
1439
1440 if (n->string < cp)
1441 putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1442
1443 for (n = n->next; NULL != n; n = n->next)
1444 if (MDOC_TEXT == n->type)
1445 putkey(mpage, n->string, TYPE_Fa);
1446
1447 return(0);
1448 }
1449
1450 static int
1451 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1452 {
1453 char *cp;
1454
1455 if (NULL == (n = n->child))
1456 return(0);
1457
1458 if (NULL == n->next) {
1459 putkey(mpage, n->string, TYPE_Xr);
1460 return(0);
1461 }
1462
1463 if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) {
1464 perror(NULL);
1465 exit((int)MANDOCLEVEL_SYSERR);
1466 }
1467 putkey(mpage, cp, TYPE_Xr);
1468 free(cp);
1469 return(0);
1470 }
1471
1472 static int
1473 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1474 {
1475 size_t sz;
1476
1477 if (MDOC_BODY != n->type)
1478 return(0);
1479
1480 /*
1481 * Special-case the `Nd' because we need to put the description
1482 * into the document table.
1483 */
1484
1485 for (n = n->child; NULL != n; n = n->next) {
1486 if (MDOC_TEXT == n->type) {
1487 if (NULL != mpage->desc) {
1488 sz = strlen(mpage->desc) +
1489 strlen(n->string) + 2;
1490 mpage->desc = mandoc_realloc(
1491 mpage->desc, sz);
1492 strlcat(mpage->desc, " ", sz);
1493 strlcat(mpage->desc, n->string, sz);
1494 } else
1495 mpage->desc = mandoc_strdup(n->string);
1496 }
1497 if (NULL != n->child)
1498 parse_mdoc_Nd(mpage, n);
1499 }
1500 return(1);
1501 }
1502
1503 static int
1504 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1505 {
1506
1507 return(SEC_NAME == n->sec ||
1508 (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type));
1509 }
1510
1511 static int
1512 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1513 {
1514
1515 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1516 }
1517
1518 static int
1519 parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1520 {
1521
1522 return(MDOC_HEAD == n->type);
1523 }
1524
1525 static int
1526 parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1527 {
1528
1529 return(MDOC_BODY == n->type);
1530 }
1531
1532 /*
1533 * Add a string to the hash table for the current manual.
1534 * Each string has a bitmask telling which macros it belongs to.
1535 * When we finish the manual, we'll dump the table.
1536 */
1537 static void
1538 putkeys(const struct mpage *mpage,
1539 const char *cp, size_t sz, uint64_t v)
1540 {
1541 struct str *s;
1542 unsigned int slot;
1543 const char *end;
1544
1545 if (0 == sz)
1546 return;
1547
1548 end = cp + sz;
1549 slot = ohash_qlookupi(&strings, cp, &end);
1550 s = ohash_find(&strings, slot);
1551
1552 if (NULL != s && mpage == s->mpage) {
1553 s->mask |= v;
1554 return;
1555 } else if (NULL == s) {
1556 s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1557 memcpy(s->key, cp, sz);
1558 ohash_insert(&strings, slot, s);
1559 }
1560 s->mpage = mpage;
1561 s->mask = v;
1562 }
1563
1564 /*
1565 * Take a Unicode codepoint and produce its UTF-8 encoding.
1566 * This isn't the best way to do this, but it works.
1567 * The magic numbers are from the UTF-8 packaging.
1568 * They're not as scary as they seem: read the UTF-8 spec for details.
1569 */
1570 static size_t
1571 utf8(unsigned int cp, char out[7])
1572 {
1573 size_t rc;
1574
1575 rc = 0;
1576 if (cp <= 0x0000007F) {
1577 rc = 1;
1578 out[0] = (char)cp;
1579 } else if (cp <= 0x000007FF) {
1580 rc = 2;
1581 out[0] = (cp >> 6 & 31) | 192;
1582 out[1] = (cp & 63) | 128;
1583 } else if (cp <= 0x0000FFFF) {
1584 rc = 3;
1585 out[0] = (cp >> 12 & 15) | 224;
1586 out[1] = (cp >> 6 & 63) | 128;
1587 out[2] = (cp & 63) | 128;
1588 } else if (cp <= 0x001FFFFF) {
1589 rc = 4;
1590 out[0] = (cp >> 18 & 7) | 240;
1591 out[1] = (cp >> 12 & 63) | 128;
1592 out[2] = (cp >> 6 & 63) | 128;
1593 out[3] = (cp & 63) | 128;
1594 } else if (cp <= 0x03FFFFFF) {
1595 rc = 5;
1596 out[0] = (cp >> 24 & 3) | 248;
1597 out[1] = (cp >> 18 & 63) | 128;
1598 out[2] = (cp >> 12 & 63) | 128;
1599 out[3] = (cp >> 6 & 63) | 128;
1600 out[4] = (cp & 63) | 128;
1601 } else if (cp <= 0x7FFFFFFF) {
1602 rc = 6;
1603 out[0] = (cp >> 30 & 1) | 252;
1604 out[1] = (cp >> 24 & 63) | 128;
1605 out[2] = (cp >> 18 & 63) | 128;
1606 out[3] = (cp >> 12 & 63) | 128;
1607 out[4] = (cp >> 6 & 63) | 128;
1608 out[5] = (cp & 63) | 128;
1609 } else
1610 return(0);
1611
1612 out[rc] = '\0';
1613 return(rc);
1614 }
1615
1616 /*
1617 * Store the rendered version of a key, or alias the pointer
1618 * if the key contains no escape sequences.
1619 */
1620 static void
1621 render_key(struct mchars *mc, struct str *key)
1622 {
1623 size_t sz, bsz, pos;
1624 char utfbuf[7], res[5];
1625 char *buf;
1626 const char *seq, *cpp, *val;
1627 int len, u;
1628 enum mandoc_esc esc;
1629
1630 assert(NULL == key->rendered);
1631
1632 res[0] = '\\';
1633 res[1] = '\t';
1634 res[2] = ASCII_NBRSP;
1635 res[3] = ASCII_HYPH;
1636 res[4] = '\0';
1637
1638 val = key->key;
1639 bsz = strlen(val);
1640
1641 /*
1642 * Pre-check: if we have no stop-characters, then set the
1643 * pointer as ourselvse and get out of here.
1644 */
1645 if (strcspn(val, res) == bsz) {
1646 key->rendered = key->key;
1647 return;
1648 }
1649
1650 /* Pre-allocate by the length of the input */
1651
1652 buf = mandoc_malloc(++bsz);
1653 pos = 0;
1654
1655 while ('\0' != *val) {
1656 /*
1657 * Halt on the first escape sequence.
1658 * This also halts on the end of string, in which case
1659 * we just copy, fallthrough, and exit the loop.
1660 */
1661 if ((sz = strcspn(val, res)) > 0) {
1662 memcpy(&buf[pos], val, sz);
1663 pos += sz;
1664 val += sz;
1665 }
1666
1667 if (ASCII_HYPH == *val) {
1668 buf[pos++] = '-';
1669 val++;
1670 continue;
1671 } else if ('\t' == *val || ASCII_NBRSP == *val) {
1672 buf[pos++] = ' ';
1673 val++;
1674 continue;
1675 } else if ('\\' != *val)
1676 break;
1677
1678 /* Read past the slash. */
1679
1680 val++;
1681
1682 /*
1683 * Parse the escape sequence and see if it's a
1684 * predefined character or special character.
1685 */
1686
1687 esc = mandoc_escape
1688 ((const char **)&val, &seq, &len);
1689 if (ESCAPE_ERROR == esc)
1690 break;
1691 if (ESCAPE_SPECIAL != esc)
1692 continue;
1693
1694 /*
1695 * Render the special character
1696 * as either UTF-8 or ASCII.
1697 */
1698
1699 if (write_utf8) {
1700 if (0 == (u = mchars_spec2cp(mc, seq, len)))
1701 continue;
1702 cpp = utfbuf;
1703 if (0 == (sz = utf8(u, utfbuf)))
1704 continue;
1705 sz = strlen(cpp);
1706 } else {
1707 cpp = mchars_spec2str(mc, seq, len, &sz);
1708 if (NULL == cpp)
1709 continue;
1710 if (ASCII_NBRSP == *cpp) {
1711 cpp = " ";
1712 sz = 1;
1713 }
1714 }
1715
1716 /* Copy the rendered glyph into the stream. */
1717
1718 bsz += sz;
1719 buf = mandoc_realloc(buf, bsz);
1720 memcpy(&buf[pos], cpp, sz);
1721 pos += sz;
1722 }
1723
1724 buf[pos] = '\0';
1725 key->rendered = buf;
1726 }
1727
1728 /*
1729 * Flush the current page's terms (and their bits) into the database.
1730 * Wrap the entire set of additions in a transaction to make sqlite be a
1731 * little faster.
1732 * Also, handle escape sequences at the last possible moment.
1733 */
1734 static void
1735 dbindex(const struct mpage *mpage, struct mchars *mc)
1736 {
1737 struct mlink *mlink;
1738 struct str *key;
1739 const char *desc;
1740 int64_t recno;
1741 size_t i;
1742 unsigned int slot;
1743
1744 if (verb)
1745 say(mpage->mlinks->file, "Adding to index");
1746
1747 if (nodb)
1748 return;
1749
1750 desc = "";
1751 if (NULL != mpage->desc && '\0' != *mpage->desc) {
1752 key = ohash_find(&strings,
1753 ohash_qlookup(&strings, mpage->desc));
1754 assert(NULL != key);
1755 if (NULL == key->rendered)
1756 render_key(mc, key);
1757 desc = key->rendered;
1758 }
1759
1760 SQL_EXEC("BEGIN TRANSACTION");
1761
1762 i = 1;
1763 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, desc);
1764 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
1765 SQL_STEP(stmts[STMT_INSERT_PAGE]);
1766 recno = sqlite3_last_insert_rowid(db);
1767 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1768
1769 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1770 i = 1;
1771 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->file);
1772 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1773 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1774 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1775 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, recno);
1776 SQL_STEP(stmts[STMT_INSERT_LINK]);
1777 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1778 }
1779
1780 for (key = ohash_first(&strings, &slot); NULL != key;
1781 key = ohash_next(&strings, &slot)) {
1782 assert(key->mpage == mpage);
1783 if (NULL == key->rendered)
1784 render_key(mc, key);
1785 i = 1;
1786 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1787 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
1788 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
1789 SQL_STEP(stmts[STMT_INSERT_KEY]);
1790 sqlite3_reset(stmts[STMT_INSERT_KEY]);
1791 if (key->rendered != key->key)
1792 free(key->rendered);
1793 free(key);
1794 }
1795
1796 SQL_EXEC("END TRANSACTION");
1797 }
1798
1799 static void
1800 dbprune(void)
1801 {
1802 struct mpage *mpage;
1803 struct mlink *mlink;
1804 size_t i;
1805 unsigned int slot;
1806
1807 if (nodb)
1808 return;
1809
1810 mpage = ohash_first(&mpages, &slot);
1811 while (NULL != mpage) {
1812 mlink = mpage->mlinks;
1813 i = 1;
1814 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE], i, mlink->file);
1815 SQL_STEP(stmts[STMT_DELETE_PAGE]);
1816 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
1817 if (verb)
1818 say(mlink->file, "Deleted from index");
1819 mpage = ohash_next(&mpages, &slot);
1820 }
1821 }
1822
1823 /*
1824 * Close an existing database and its prepared statements.
1825 * If "real" is not set, rename the temporary file into the real one.
1826 */
1827 static void
1828 dbclose(int real)
1829 {
1830 size_t i;
1831
1832 if (nodb)
1833 return;
1834
1835 for (i = 0; i < STMT__MAX; i++) {
1836 sqlite3_finalize(stmts[i]);
1837 stmts[i] = NULL;
1838 }
1839
1840 sqlite3_close(db);
1841 db = NULL;
1842
1843 if (real)
1844 return;
1845
1846 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
1847 exitcode = (int)MANDOCLEVEL_SYSERR;
1848 say(MANDOC_DB, NULL);
1849 }
1850 }
1851
1852 /*
1853 * This is straightforward stuff.
1854 * Open a database connection to a "temporary" database, then open a set
1855 * of prepared statements we'll use over and over again.
1856 * If "real" is set, we use the existing database; if not, we truncate a
1857 * temporary one.
1858 * Must be matched by dbclose().
1859 */
1860 static int
1861 dbopen(int real)
1862 {
1863 const char *file, *sql;
1864 int rc, ofl;
1865
1866 if (nodb)
1867 return(1);
1868
1869 ofl = SQLITE_OPEN_READWRITE;
1870 if (0 == real) {
1871 file = MANDOC_DB "~";
1872 if (-1 == remove(file) && ENOENT != errno) {
1873 exitcode = (int)MANDOCLEVEL_SYSERR;
1874 say(file, NULL);
1875 return(0);
1876 }
1877 ofl |= SQLITE_OPEN_EXCLUSIVE;
1878 } else
1879 file = MANDOC_DB;
1880
1881 rc = sqlite3_open_v2(file, &db, ofl, NULL);
1882 if (SQLITE_OK == rc)
1883 goto prepare_statements;
1884 if (SQLITE_CANTOPEN != rc) {
1885 exitcode = (int)MANDOCLEVEL_SYSERR;
1886 say(file, NULL);
1887 return(0);
1888 }
1889
1890 sqlite3_close(db);
1891 db = NULL;
1892
1893 if (SQLITE_OK != (rc = sqlite3_open(file, &db))) {
1894 exitcode = (int)MANDOCLEVEL_SYSERR;
1895 say(file, NULL);
1896 return(0);
1897 }
1898
1899 sql = "CREATE TABLE \"mpages\" (\n"
1900 " \"desc\" TEXT NOT NULL,\n"
1901 " \"form\" INTEGER NOT NULL,\n"
1902 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1903 ");\n"
1904 "\n"
1905 "CREATE TABLE \"mlinks\" (\n"
1906 " \"file\" TEXT NOT NULL,\n"
1907 " \"sec\" TEXT NOT NULL,\n"
1908 " \"arch\" TEXT NOT NULL,\n"
1909 " \"name\" TEXT NOT NULL,\n"
1910 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1911 "ON DELETE CASCADE,\n"
1912 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1913 ");\n"
1914 "\n"
1915 "CREATE TABLE \"keys\" (\n"
1916 " \"bits\" INTEGER NOT NULL,\n"
1917 " \"key\" TEXT NOT NULL,\n"
1918 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1919 "ON DELETE CASCADE,\n"
1920 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1921 ");\n"
1922 "\n"
1923 "CREATE INDEX \"key_index\" ON keys (key);\n";
1924
1925 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1926 exitcode = (int)MANDOCLEVEL_SYSERR;
1927 say(file, "%s", sqlite3_errmsg(db));
1928 return(0);
1929 }
1930
1931 prepare_statements:
1932 SQL_EXEC("PRAGMA foreign_keys = ON");
1933 sql = "DELETE FROM mpages where file=?";
1934 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
1935 sql = "INSERT INTO mpages "
1936 "(desc,form) VALUES (?,?)";
1937 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
1938 sql = "INSERT INTO mlinks "
1939 "(file,sec,arch,name,pageid) VALUES (?,?,?,?,?)";
1940 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1941 sql = "INSERT INTO keys "
1942 "(bits,key,pageid) VALUES (?,?,?)";
1943 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1944
1945 #ifndef __APPLE__
1946 /*
1947 * When opening a new database, we can turn off
1948 * synchronous mode for much better performance.
1949 */
1950
1951 if (real)
1952 SQL_EXEC("PRAGMA synchronous = OFF");
1953 #endif
1954
1955 return(1);
1956 }
1957
1958 static void *
1959 hash_halloc(size_t sz, void *arg)
1960 {
1961
1962 return(mandoc_calloc(sz, 1));
1963 }
1964
1965 static void *
1966 hash_alloc(size_t sz, void *arg)
1967 {
1968
1969 return(mandoc_malloc(sz));
1970 }
1971
1972 static void
1973 hash_free(void *p, size_t sz, void *arg)
1974 {
1975
1976 free(p);
1977 }
1978
1979 static int
1980 set_basedir(const char *targetdir)
1981 {
1982 static char startdir[PATH_MAX];
1983 static int fd;
1984
1985 /*
1986 * Remember where we started by keeping a fd open to the origin
1987 * path component: throughout this utility, we chdir() a lot to
1988 * handle relative paths, and by doing this, we can return to
1989 * the starting point.
1990 */
1991 if ('\0' == *startdir) {
1992 if (NULL == getcwd(startdir, PATH_MAX)) {
1993 exitcode = (int)MANDOCLEVEL_SYSERR;
1994 if (NULL != targetdir)
1995 say(".", NULL);
1996 return(0);
1997 }
1998 if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
1999 exitcode = (int)MANDOCLEVEL_SYSERR;
2000 say(startdir, NULL);
2001 return(0);
2002 }
2003 if (NULL == targetdir)
2004 targetdir = startdir;
2005 } else {
2006 if (-1 == fd)
2007 return(0);
2008 if (-1 == fchdir(fd)) {
2009 close(fd);
2010 basedir[0] = '\0';
2011 exitcode = (int)MANDOCLEVEL_SYSERR;
2012 say(startdir, NULL);
2013 return(0);
2014 }
2015 if (NULL == targetdir) {
2016 close(fd);
2017 return(1);
2018 }
2019 }
2020 if (NULL == realpath(targetdir, basedir)) {
2021 basedir[0] = '\0';
2022 exitcode = (int)MANDOCLEVEL_BADARG;
2023 say(targetdir, NULL);
2024 return(0);
2025 } else if (-1 == chdir(basedir)) {
2026 exitcode = (int)MANDOCLEVEL_BADARG;
2027 say("", NULL);
2028 return(0);
2029 }
2030 return(1);
2031 }
2032
2033 static void
2034 say(const char *file, const char *format, ...)
2035 {
2036 va_list ap;
2037
2038 if ('\0' != *basedir)
2039 fprintf(stderr, "%s", basedir);
2040 if ('\0' != *basedir && '\0' != *file)
2041 fputs("//", stderr);
2042 if ('\0' != *file)
2043 fprintf(stderr, "%s", file);
2044 fputs(": ", stderr);
2045
2046 if (NULL == format) {
2047 perror(NULL);
2048 return;
2049 }
2050
2051 va_start(ap, format);
2052 vfprintf(stderr, format, ap);
2053 va_end(ap);
2054
2055 fputc('\n', stderr);
2056 }
A git conversion mirror of mandoc.bsd.lv