]> 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
Simplify the mparse_open() interface.
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.213 2016/01/08 02:53:13 schwarze Exp $ */
2 /*
3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011-2016 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 AUTHORS DISCLAIM ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS 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 #include "config.h"
19
20 #include <sys/types.h>
21 #include <sys/stat.h>
22 #include <sys/wait.h>
23
24 #include <assert.h>
25 #include <ctype.h>
26 #if HAVE_ERR
27 #include <err.h>
28 #endif
29 #include <errno.h>
30 #include <fcntl.h>
31 #if HAVE_FTS
32 #include <fts.h>
33 #else
34 #include "compat_fts.h"
35 #endif
36 #include <getopt.h>
37 #include <limits.h>
38 #include <stddef.h>
39 #include <stdio.h>
40 #include <stdint.h>
41 #include <stdlib.h>
42 #include <string.h>
43 #include <unistd.h>
44
45 #include <sqlite3.h>
46
47 #include "mandoc_aux.h"
48 #include "mandoc_ohash.h"
49 #include "mandoc.h"
50 #include "roff.h"
51 #include "mdoc.h"
52 #include "man.h"
53 #include "manconf.h"
54 #include "mansearch.h"
55
56 extern int mansearch_keymax;
57 extern const char *const mansearch_keynames[];
58
59 #define SQL_EXEC(_v) \
60 if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
61 say("", "%s: %s", (_v), sqlite3_errmsg(db))
62 #define SQL_BIND_TEXT(_s, _i, _v) \
63 if (SQLITE_OK != sqlite3_bind_text \
64 ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
65 say(mlink->file, "%s", sqlite3_errmsg(db))
66 #define SQL_BIND_INT(_s, _i, _v) \
67 if (SQLITE_OK != sqlite3_bind_int \
68 ((_s), (_i)++, (_v))) \
69 say(mlink->file, "%s", sqlite3_errmsg(db))
70 #define SQL_BIND_INT64(_s, _i, _v) \
71 if (SQLITE_OK != sqlite3_bind_int64 \
72 ((_s), (_i)++, (_v))) \
73 say(mlink->file, "%s", sqlite3_errmsg(db))
74 #define SQL_STEP(_s) \
75 if (SQLITE_DONE != sqlite3_step((_s))) \
76 say(mlink->file, "%s", sqlite3_errmsg(db))
77
78 enum op {
79 OP_DEFAULT = 0, /* new dbs from dir list or default config */
80 OP_CONFFILE, /* new databases from custom config file */
81 OP_UPDATE, /* delete/add entries in existing database */
82 OP_DELETE, /* delete entries from existing database */
83 OP_TEST /* change no databases, report potential problems */
84 };
85
86 struct str {
87 const struct mpage *mpage; /* if set, the owning parse */
88 uint64_t mask; /* bitmask in sequence */
89 char key[]; /* rendered text */
90 };
91
92 struct inodev {
93 ino_t st_ino;
94 dev_t st_dev;
95 };
96
97 struct mpage {
98 struct inodev inodev; /* used for hashing routine */
99 int64_t pageid; /* pageid in mpages SQL table */
100 char *sec; /* section from file content */
101 char *arch; /* architecture from file content */
102 char *title; /* title from file content */
103 char *desc; /* description from file content */
104 struct mlink *mlinks; /* singly linked list */
105 int form; /* format from file content */
106 int name_head_done;
107 };
108
109 struct mlink {
110 char file[PATH_MAX]; /* filename rel. to manpath */
111 char *dsec; /* section from directory */
112 char *arch; /* architecture from directory */
113 char *name; /* name from file name (not empty) */
114 char *fsec; /* section from file name suffix */
115 struct mlink *next; /* singly linked list */
116 struct mpage *mpage; /* parent */
117 int dform; /* format from directory */
118 int fform; /* format from file name suffix */
119 int gzip; /* filename has a .gz suffix */
120 };
121
122 enum stmt {
123 STMT_DELETE_PAGE = 0, /* delete mpage */
124 STMT_INSERT_PAGE, /* insert mpage */
125 STMT_INSERT_LINK, /* insert mlink */
126 STMT_INSERT_NAME, /* insert name */
127 STMT_SELECT_NAME, /* retrieve existing name flags */
128 STMT_INSERT_KEY, /* insert parsed key */
129 STMT__MAX
130 };
131
132 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
133 const struct roff_node *);
134
135 struct mdoc_handler {
136 mdoc_fp fp; /* optional handler */
137 uint64_t mask; /* set unless handler returns 0 */
138 };
139
140 static void dbclose(int);
141 static void dbadd(struct mpage *);
142 static void dbadd_mlink(const struct mlink *mlink);
143 static void dbadd_mlink_name(const struct mlink *mlink);
144 static int dbopen(int);
145 static void dbprune(void);
146 static void filescan(const char *);
147 static void mlink_add(struct mlink *, const struct stat *);
148 static void mlink_check(struct mpage *, struct mlink *);
149 static void mlink_free(struct mlink *);
150 static void mlinks_undupe(struct mpage *);
151 static void mpages_free(void);
152 static void mpages_merge(struct mparse *);
153 static void names_check(void);
154 static void parse_cat(struct mpage *, int);
155 static void parse_man(struct mpage *, const struct roff_meta *,
156 const struct roff_node *);
157 static void parse_mdoc(struct mpage *, const struct roff_meta *,
158 const struct roff_node *);
159 static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
160 const struct roff_node *);
161 static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
162 const struct roff_node *);
163 static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
164 static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
165 const struct roff_node *);
166 static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
167 const struct roff_node *);
168 static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
169 const struct roff_node *);
170 static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
171 const struct roff_node *);
172 static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
173 const struct roff_node *);
174 static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
175 const struct roff_node *);
176 static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
177 const struct roff_node *);
178 static void putkey(const struct mpage *, char *, uint64_t);
179 static void putkeys(const struct mpage *, char *, size_t, uint64_t);
180 static void putmdockey(const struct mpage *,
181 const struct roff_node *, uint64_t);
182 static int render_string(char **, size_t *);
183 static void say(const char *, const char *, ...);
184 static int set_basedir(const char *, int);
185 static int treescan(void);
186 static size_t utf8(unsigned int, char [7]);
187
188 static char tempfilename[32];
189 static int nodb; /* no database changes */
190 static int mparse_options; /* abort the parse early */
191 static int use_all; /* use all found files */
192 static int debug; /* print what we're doing */
193 static int warnings; /* warn about crap */
194 static int write_utf8; /* write UTF-8 output; else ASCII */
195 static int exitcode; /* to be returned by main */
196 static enum op op; /* operational mode */
197 static char basedir[PATH_MAX]; /* current base directory */
198 static struct ohash mpages; /* table of distinct manual pages */
199 static struct ohash mlinks; /* table of directory entries */
200 static struct ohash names; /* table of all names */
201 static struct ohash strings; /* table of all strings */
202 static sqlite3 *db = NULL; /* current database */
203 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
204 static uint64_t name_mask;
205
206 static const struct mdoc_handler mdocs[MDOC_MAX] = {
207 { NULL, 0 }, /* Ap */
208 { NULL, 0 }, /* Dd */
209 { NULL, 0 }, /* Dt */
210 { NULL, 0 }, /* Os */
211 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
212 { parse_mdoc_head, TYPE_Ss }, /* Ss */
213 { NULL, 0 }, /* Pp */
214 { NULL, 0 }, /* D1 */
215 { NULL, 0 }, /* Dl */
216 { NULL, 0 }, /* Bd */
217 { NULL, 0 }, /* Ed */
218 { NULL, 0 }, /* Bl */
219 { NULL, 0 }, /* El */
220 { NULL, 0 }, /* It */
221 { NULL, 0 }, /* Ad */
222 { NULL, TYPE_An }, /* An */
223 { NULL, TYPE_Ar }, /* Ar */
224 { NULL, TYPE_Cd }, /* Cd */
225 { NULL, TYPE_Cm }, /* Cm */
226 { NULL, TYPE_Dv }, /* Dv */
227 { NULL, TYPE_Er }, /* Er */
228 { NULL, TYPE_Ev }, /* Ev */
229 { NULL, 0 }, /* Ex */
230 { NULL, TYPE_Fa }, /* Fa */
231 { parse_mdoc_Fd, 0 }, /* Fd */
232 { NULL, TYPE_Fl }, /* Fl */
233 { parse_mdoc_Fn, 0 }, /* Fn */
234 { NULL, TYPE_Ft }, /* Ft */
235 { NULL, TYPE_Ic }, /* Ic */
236 { NULL, TYPE_In }, /* In */
237 { NULL, TYPE_Li }, /* Li */
238 { parse_mdoc_Nd, 0 }, /* Nd */
239 { parse_mdoc_Nm, 0 }, /* Nm */
240 { NULL, 0 }, /* Op */
241 { NULL, 0 }, /* Ot */
242 { NULL, TYPE_Pa }, /* Pa */
243 { NULL, 0 }, /* Rv */
244 { NULL, TYPE_St }, /* St */
245 { parse_mdoc_Va, TYPE_Va }, /* Va */
246 { parse_mdoc_Va, TYPE_Vt }, /* Vt */
247 { parse_mdoc_Xr, 0 }, /* Xr */
248 { NULL, 0 }, /* %A */
249 { NULL, 0 }, /* %B */
250 { NULL, 0 }, /* %D */
251 { NULL, 0 }, /* %I */
252 { NULL, 0 }, /* %J */
253 { NULL, 0 }, /* %N */
254 { NULL, 0 }, /* %O */
255 { NULL, 0 }, /* %P */
256 { NULL, 0 }, /* %R */
257 { NULL, 0 }, /* %T */
258 { NULL, 0 }, /* %V */
259 { NULL, 0 }, /* Ac */
260 { NULL, 0 }, /* Ao */
261 { NULL, 0 }, /* Aq */
262 { NULL, TYPE_At }, /* At */
263 { NULL, 0 }, /* Bc */
264 { NULL, 0 }, /* Bf */
265 { NULL, 0 }, /* Bo */
266 { NULL, 0 }, /* Bq */
267 { NULL, TYPE_Bsx }, /* Bsx */
268 { NULL, TYPE_Bx }, /* Bx */
269 { NULL, 0 }, /* Db */
270 { NULL, 0 }, /* Dc */
271 { NULL, 0 }, /* Do */
272 { NULL, 0 }, /* Dq */
273 { NULL, 0 }, /* Ec */
274 { NULL, 0 }, /* Ef */
275 { NULL, TYPE_Em }, /* Em */
276 { NULL, 0 }, /* Eo */
277 { NULL, TYPE_Fx }, /* Fx */
278 { NULL, TYPE_Ms }, /* Ms */
279 { NULL, 0 }, /* No */
280 { NULL, 0 }, /* Ns */
281 { NULL, TYPE_Nx }, /* Nx */
282 { NULL, TYPE_Ox }, /* Ox */
283 { NULL, 0 }, /* Pc */
284 { NULL, 0 }, /* Pf */
285 { NULL, 0 }, /* Po */
286 { NULL, 0 }, /* Pq */
287 { NULL, 0 }, /* Qc */
288 { NULL, 0 }, /* Ql */
289 { NULL, 0 }, /* Qo */
290 { NULL, 0 }, /* Qq */
291 { NULL, 0 }, /* Re */
292 { NULL, 0 }, /* Rs */
293 { NULL, 0 }, /* Sc */
294 { NULL, 0 }, /* So */
295 { NULL, 0 }, /* Sq */
296 { NULL, 0 }, /* Sm */
297 { NULL, 0 }, /* Sx */
298 { NULL, TYPE_Sy }, /* Sy */
299 { NULL, TYPE_Tn }, /* Tn */
300 { NULL, 0 }, /* Ux */
301 { NULL, 0 }, /* Xc */
302 { NULL, 0 }, /* Xo */
303 { parse_mdoc_Fo, 0 }, /* Fo */
304 { NULL, 0 }, /* Fc */
305 { NULL, 0 }, /* Oo */
306 { NULL, 0 }, /* Oc */
307 { NULL, 0 }, /* Bk */
308 { NULL, 0 }, /* Ek */
309 { NULL, 0 }, /* Bt */
310 { NULL, 0 }, /* Hf */
311 { NULL, 0 }, /* Fr */
312 { NULL, 0 }, /* Ud */
313 { NULL, TYPE_Lb }, /* Lb */
314 { NULL, 0 }, /* Lp */
315 { NULL, TYPE_Lk }, /* Lk */
316 { NULL, TYPE_Mt }, /* Mt */
317 { NULL, 0 }, /* Brq */
318 { NULL, 0 }, /* Bro */
319 { NULL, 0 }, /* Brc */
320 { NULL, 0 }, /* %C */
321 { NULL, 0 }, /* Es */
322 { NULL, 0 }, /* En */
323 { NULL, TYPE_Dx }, /* Dx */
324 { NULL, 0 }, /* %Q */
325 { NULL, 0 }, /* br */
326 { NULL, 0 }, /* sp */
327 { NULL, 0 }, /* %U */
328 { NULL, 0 }, /* Ta */
329 { NULL, 0 }, /* ll */
330 };
331
332
333 int
334 mandocdb(int argc, char *argv[])
335 {
336 struct manconf conf;
337 struct mparse *mp;
338 const char *path_arg, *progname;
339 size_t j, sz;
340 int ch, i;
341
342 #if HAVE_PLEDGE
343 if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) {
344 perror("pledge");
345 return (int)MANDOCLEVEL_SYSERR;
346 }
347 #endif
348
349 memset(&conf, 0, sizeof(conf));
350 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
351
352 /*
353 * We accept a few different invocations.
354 * The CHECKOP macro makes sure that invocation styles don't
355 * clobber each other.
356 */
357 #define CHECKOP(_op, _ch) do \
358 if (OP_DEFAULT != (_op)) { \
359 warnx("-%c: Conflicting option", (_ch)); \
360 goto usage; \
361 } while (/*CONSTCOND*/0)
362
363 path_arg = NULL;
364 op = OP_DEFAULT;
365
366 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
367 switch (ch) {
368 case 'a':
369 use_all = 1;
370 break;
371 case 'C':
372 CHECKOP(op, ch);
373 path_arg = optarg;
374 op = OP_CONFFILE;
375 break;
376 case 'D':
377 debug++;
378 break;
379 case 'd':
380 CHECKOP(op, ch);
381 path_arg = optarg;
382 op = OP_UPDATE;
383 break;
384 case 'n':
385 nodb = 1;
386 break;
387 case 'p':
388 warnings = 1;
389 break;
390 case 'Q':
391 mparse_options |= MPARSE_QUICK;
392 break;
393 case 'T':
394 if (strcmp(optarg, "utf8")) {
395 warnx("-T%s: Unsupported output format",
396 optarg);
397 goto usage;
398 }
399 write_utf8 = 1;
400 break;
401 case 't':
402 CHECKOP(op, ch);
403 dup2(STDOUT_FILENO, STDERR_FILENO);
404 op = OP_TEST;
405 nodb = warnings = 1;
406 break;
407 case 'u':
408 CHECKOP(op, ch);
409 path_arg = optarg;
410 op = OP_DELETE;
411 break;
412 case 'v':
413 /* Compatibility with espie@'s makewhatis. */
414 break;
415 default:
416 goto usage;
417 }
418
419 argc -= optind;
420 argv += optind;
421
422 #if HAVE_PLEDGE
423 if (nodb) {
424 if (pledge("stdio rpath", NULL) == -1) {
425 perror("pledge");
426 return (int)MANDOCLEVEL_SYSERR;
427 }
428 }
429 #endif
430
431 if (OP_CONFFILE == op && argc > 0) {
432 warnx("-C: Too many arguments");
433 goto usage;
434 }
435
436 exitcode = (int)MANDOCLEVEL_OK;
437 mchars_alloc();
438 mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
439 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
440 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
441
442 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
443
444 /*
445 * Most of these deal with a specific directory.
446 * Jump into that directory first.
447 */
448 if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
449 goto out;
450
451 if (dbopen(1)) {
452 /*
453 * The existing database is usable. Process
454 * all files specified on the command-line.
455 */
456 #if HAVE_PLEDGE
457 if (!nodb) {
458 if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) {
459 perror("pledge");
460 exitcode = (int)MANDOCLEVEL_SYSERR;
461 goto out;
462 }
463 }
464 #endif
465 use_all = 1;
466 for (i = 0; i < argc; i++)
467 filescan(argv[i]);
468 if (OP_TEST != op)
469 dbprune();
470 } else {
471 /*
472 * Database missing or corrupt.
473 * Recreate from scratch.
474 */
475 exitcode = (int)MANDOCLEVEL_OK;
476 op = OP_DEFAULT;
477 if (0 == treescan())
478 goto out;
479 if (0 == dbopen(0))
480 goto out;
481 }
482 if (OP_DELETE != op)
483 mpages_merge(mp);
484 dbclose(OP_DEFAULT == op ? 0 : 1);
485 } else {
486 /*
487 * If we have arguments, use them as our manpaths.
488 * If we don't, grok from manpath(1) or however else
489 * manconf_parse() wants to do it.
490 */
491 if (argc > 0) {
492 conf.manpath.paths = mandoc_reallocarray(NULL,
493 argc, sizeof(char *));
494 conf.manpath.sz = (size_t)argc;
495 for (i = 0; i < argc; i++)
496 conf.manpath.paths[i] = mandoc_strdup(argv[i]);
497 } else
498 manconf_parse(&conf, path_arg, NULL, NULL);
499
500 if (conf.manpath.sz == 0) {
501 exitcode = (int)MANDOCLEVEL_BADARG;
502 say("", "Empty manpath");
503 }
504
505 /*
506 * First scan the tree rooted at a base directory, then
507 * build a new database and finally move it into place.
508 * Ignore zero-length directories and strip trailing
509 * slashes.
510 */
511 for (j = 0; j < conf.manpath.sz; j++) {
512 sz = strlen(conf.manpath.paths[j]);
513 if (sz && conf.manpath.paths[j][sz - 1] == '/')
514 conf.manpath.paths[j][--sz] = '\0';
515 if (0 == sz)
516 continue;
517
518 if (j) {
519 mandoc_ohash_init(&mpages, 6,
520 offsetof(struct mpage, inodev));
521 mandoc_ohash_init(&mlinks, 6,
522 offsetof(struct mlink, file));
523 }
524
525 if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
526 continue;
527 if (0 == treescan())
528 continue;
529 if (0 == dbopen(0))
530 continue;
531
532 mpages_merge(mp);
533 if (warnings && !nodb &&
534 ! (MPARSE_QUICK & mparse_options))
535 names_check();
536 dbclose(0);
537
538 if (j + 1 < conf.manpath.sz) {
539 mpages_free();
540 ohash_delete(&mpages);
541 ohash_delete(&mlinks);
542 }
543 }
544 }
545 out:
546 manconf_free(&conf);
547 mparse_free(mp);
548 mchars_free();
549 mpages_free();
550 ohash_delete(&mpages);
551 ohash_delete(&mlinks);
552 return exitcode;
553 usage:
554 progname = getprogname();
555 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
556 " %s [-aDnpQ] [-Tutf8] dir ...\n"
557 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
558 " %s [-Dnp] -u dir [file ...]\n"
559 " %s [-Q] -t file ...\n",
560 progname, progname, progname, progname, progname);
561
562 return (int)MANDOCLEVEL_BADARG;
563 }
564
565 /*
566 * Scan a directory tree rooted at "basedir" for manpages.
567 * We use fts(), scanning directory parts along the way for clues to our
568 * section and architecture.
569 *
570 * If use_all has been specified, grok all files.
571 * If not, sanitise paths to the following:
572 *
573 * [./]man*[/<arch>]/<name>.<section>
574 * or
575 * [./]cat<section>[/<arch>]/<name>.0
576 *
577 * TODO: accomodate for multi-language directories.
578 */
579 static int
580 treescan(void)
581 {
582 char buf[PATH_MAX];
583 FTS *f;
584 FTSENT *ff;
585 struct mlink *mlink;
586 int dform, gzip;
587 char *dsec, *arch, *fsec, *cp;
588 const char *path;
589 const char *argv[2];
590
591 argv[0] = ".";
592 argv[1] = (char *)NULL;
593
594 f = fts_open((char * const *)argv,
595 FTS_PHYSICAL | FTS_NOCHDIR, NULL);
596 if (f == NULL) {
597 exitcode = (int)MANDOCLEVEL_SYSERR;
598 say("", "&fts_open");
599 return 0;
600 }
601
602 dsec = arch = NULL;
603 dform = FORM_NONE;
604
605 while ((ff = fts_read(f)) != NULL) {
606 path = ff->fts_path + 2;
607 switch (ff->fts_info) {
608
609 /*
610 * Symbolic links require various sanity checks,
611 * then get handled just like regular files.
612 */
613 case FTS_SL:
614 if (realpath(path, buf) == NULL) {
615 if (warnings)
616 say(path, "&realpath");
617 continue;
618 }
619 if (strstr(buf, basedir) != buf
620 #ifdef HOMEBREWDIR
621 && strstr(buf, HOMEBREWDIR) != buf
622 #endif
623 ) {
624 if (warnings) say("",
625 "%s: outside base directory", buf);
626 continue;
627 }
628 /* Use logical inode to avoid mpages dupe. */
629 if (stat(path, ff->fts_statp) == -1) {
630 if (warnings)
631 say(path, "&stat");
632 continue;
633 }
634 /* FALLTHROUGH */
635
636 /*
637 * If we're a regular file, add an mlink by using the
638 * stored directory data and handling the filename.
639 */
640 case FTS_F:
641 if ( ! strcmp(path, MANDOC_DB))
642 continue;
643 if ( ! use_all && ff->fts_level < 2) {
644 if (warnings)
645 say(path, "Extraneous file");
646 continue;
647 }
648 gzip = 0;
649 fsec = NULL;
650 while (fsec == NULL) {
651 fsec = strrchr(ff->fts_name, '.');
652 if (fsec == NULL || strcmp(fsec+1, "gz"))
653 break;
654 gzip = 1;
655 *fsec = '\0';
656 fsec = NULL;
657 }
658 if (fsec == NULL) {
659 if ( ! use_all) {
660 if (warnings)
661 say(path,
662 "No filename suffix");
663 continue;
664 }
665 } else if ( ! strcmp(++fsec, "html")) {
666 if (warnings)
667 say(path, "Skip html");
668 continue;
669 } else if ( ! strcmp(fsec, "ps")) {
670 if (warnings)
671 say(path, "Skip ps");
672 continue;
673 } else if ( ! strcmp(fsec, "pdf")) {
674 if (warnings)
675 say(path, "Skip pdf");
676 continue;
677 } else if ( ! use_all &&
678 ((dform == FORM_SRC &&
679 strncmp(fsec, dsec, strlen(dsec))) ||
680 (dform == FORM_CAT && strcmp(fsec, "0")))) {
681 if (warnings)
682 say(path, "Wrong filename suffix");
683 continue;
684 } else
685 fsec[-1] = '\0';
686
687 mlink = mandoc_calloc(1, sizeof(struct mlink));
688 if (strlcpy(mlink->file, path,
689 sizeof(mlink->file)) >=
690 sizeof(mlink->file)) {
691 say(path, "Filename too long");
692 free(mlink);
693 continue;
694 }
695 mlink->dform = dform;
696 mlink->dsec = dsec;
697 mlink->arch = arch;
698 mlink->name = ff->fts_name;
699 mlink->fsec = fsec;
700 mlink->gzip = gzip;
701 mlink_add(mlink, ff->fts_statp);
702 continue;
703
704 case FTS_D:
705 case FTS_DP:
706 break;
707
708 default:
709 if (warnings)
710 say(path, "Not a regular file");
711 continue;
712 }
713
714 switch (ff->fts_level) {
715 case 0:
716 /* Ignore the root directory. */
717 break;
718 case 1:
719 /*
720 * This might contain manX/ or catX/.
721 * Try to infer this from the name.
722 * If we're not in use_all, enforce it.
723 */
724 cp = ff->fts_name;
725 if (ff->fts_info == FTS_DP) {
726 dform = FORM_NONE;
727 dsec = NULL;
728 break;
729 }
730
731 if ( ! strncmp(cp, "man", 3)) {
732 dform = FORM_SRC;
733 dsec = cp + 3;
734 } else if ( ! strncmp(cp, "cat", 3)) {
735 dform = FORM_CAT;
736 dsec = cp + 3;
737 } else {
738 dform = FORM_NONE;
739 dsec = NULL;
740 }
741
742 if (dsec != NULL || use_all)
743 break;
744
745 if (warnings)
746 say(path, "Unknown directory part");
747 fts_set(f, ff, FTS_SKIP);
748 break;
749 case 2:
750 /*
751 * Possibly our architecture.
752 * If we're descending, keep tabs on it.
753 */
754 if (ff->fts_info != FTS_DP && dsec != NULL)
755 arch = ff->fts_name;
756 else
757 arch = NULL;
758 break;
759 default:
760 if (ff->fts_info == FTS_DP || use_all)
761 break;
762 if (warnings)
763 say(path, "Extraneous directory part");
764 fts_set(f, ff, FTS_SKIP);
765 break;
766 }
767 }
768
769 fts_close(f);
770 return 1;
771 }
772
773 /*
774 * Add a file to the mlinks table.
775 * Do not verify that it's a "valid" looking manpage (we'll do that
776 * later).
777 *
778 * Try to infer the manual section, architecture, and page name from the
779 * path, assuming it looks like
780 *
781 * [./]man*[/<arch>]/<name>.<section>
782 * or
783 * [./]cat<section>[/<arch>]/<name>.0
784 *
785 * See treescan() for the fts(3) version of this.
786 */
787 static void
788 filescan(const char *file)
789 {
790 char buf[PATH_MAX];
791 struct stat st;
792 struct mlink *mlink;
793 char *p, *start;
794
795 assert(use_all);
796
797 if (0 == strncmp(file, "./", 2))
798 file += 2;
799
800 /*
801 * We have to do lstat(2) before realpath(3) loses
802 * the information whether this is a symbolic link.
803 * We need to know that because for symbolic links,
804 * we want to use the orginal file name, while for
805 * regular files, we want to use the real path.
806 */
807 if (-1 == lstat(file, &st)) {
808 exitcode = (int)MANDOCLEVEL_BADARG;
809 say(file, "&lstat");
810 return;
811 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
812 exitcode = (int)MANDOCLEVEL_BADARG;
813 say(file, "Not a regular file");
814 return;
815 }
816
817 /*
818 * We have to resolve the file name to the real path
819 * in any case for the base directory check.
820 */
821 if (NULL == realpath(file, buf)) {
822 exitcode = (int)MANDOCLEVEL_BADARG;
823 say(file, "&realpath");
824 return;
825 }
826
827 if (OP_TEST == op)
828 start = buf;
829 else if (strstr(buf, basedir) == buf)
830 start = buf + strlen(basedir);
831 #ifdef HOMEBREWDIR
832 else if (strstr(buf, HOMEBREWDIR) == buf)
833 start = buf;
834 #endif
835 else {
836 exitcode = (int)MANDOCLEVEL_BADARG;
837 say("", "%s: outside base directory", buf);
838 return;
839 }
840
841 /*
842 * Now we are sure the file is inside our tree.
843 * If it is a symbolic link, ignore the real path
844 * and use the original name.
845 * This implies passing stuff like "cat1/../man1/foo.1"
846 * on the command line won't work. So don't do that.
847 * Note the stat(2) can still fail if the link target
848 * doesn't exist.
849 */
850 if (S_IFLNK & st.st_mode) {
851 if (-1 == stat(buf, &st)) {
852 exitcode = (int)MANDOCLEVEL_BADARG;
853 say(file, "&stat");
854 return;
855 }
856 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
857 say(file, "Filename too long");
858 return;
859 }
860 start = buf;
861 if (OP_TEST != op && strstr(buf, basedir) == buf)
862 start += strlen(basedir);
863 }
864
865 mlink = mandoc_calloc(1, sizeof(struct mlink));
866 mlink->dform = FORM_NONE;
867 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
868 sizeof(mlink->file)) {
869 say(start, "Filename too long");
870 free(mlink);
871 return;
872 }
873
874 /*
875 * First try to guess our directory structure.
876 * If we find a separator, try to look for man* or cat*.
877 * If we find one of these and what's underneath is a directory,
878 * assume it's an architecture.
879 */
880 if (NULL != (p = strchr(start, '/'))) {
881 *p++ = '\0';
882 if (0 == strncmp(start, "man", 3)) {
883 mlink->dform = FORM_SRC;
884 mlink->dsec = start + 3;
885 } else if (0 == strncmp(start, "cat", 3)) {
886 mlink->dform = FORM_CAT;
887 mlink->dsec = start + 3;
888 }
889
890 start = p;
891 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
892 *p++ = '\0';
893 mlink->arch = start;
894 start = p;
895 }
896 }
897
898 /*
899 * Now check the file suffix.
900 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
901 */
902 p = strrchr(start, '\0');
903 while (p-- > start && '/' != *p && '.' != *p)
904 /* Loop. */ ;
905
906 if ('.' == *p) {
907 *p++ = '\0';
908 mlink->fsec = p;
909 }
910
911 /*
912 * Now try to parse the name.
913 * Use the filename portion of the path.
914 */
915 mlink->name = start;
916 if (NULL != (p = strrchr(start, '/'))) {
917 mlink->name = p + 1;
918 *p = '\0';
919 }
920 mlink_add(mlink, &st);
921 }
922
923 static void
924 mlink_add(struct mlink *mlink, const struct stat *st)
925 {
926 struct inodev inodev;
927 struct mpage *mpage;
928 unsigned int slot;
929
930 assert(NULL != mlink->file);
931
932 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
933 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
934 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
935 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
936
937 if ('0' == *mlink->fsec) {
938 free(mlink->fsec);
939 mlink->fsec = mandoc_strdup(mlink->dsec);
940 mlink->fform = FORM_CAT;
941 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
942 mlink->fform = FORM_SRC;
943 else
944 mlink->fform = FORM_NONE;
945
946 slot = ohash_qlookup(&mlinks, mlink->file);
947 assert(NULL == ohash_find(&mlinks, slot));
948 ohash_insert(&mlinks, slot, mlink);
949
950 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
951 inodev.st_ino = st->st_ino;
952 inodev.st_dev = st->st_dev;
953 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
954 sizeof(struct inodev), inodev.st_ino);
955 mpage = ohash_find(&mpages, slot);
956 if (NULL == mpage) {
957 mpage = mandoc_calloc(1, sizeof(struct mpage));
958 mpage->inodev.st_ino = inodev.st_ino;
959 mpage->inodev.st_dev = inodev.st_dev;
960 ohash_insert(&mpages, slot, mpage);
961 } else
962 mlink->next = mpage->mlinks;
963 mpage->mlinks = mlink;
964 mlink->mpage = mpage;
965 }
966
967 static void
968 mlink_free(struct mlink *mlink)
969 {
970
971 free(mlink->dsec);
972 free(mlink->arch);
973 free(mlink->name);
974 free(mlink->fsec);
975 free(mlink);
976 }
977
978 static void
979 mpages_free(void)
980 {
981 struct mpage *mpage;
982 struct mlink *mlink;
983 unsigned int slot;
984
985 mpage = ohash_first(&mpages, &slot);
986 while (NULL != mpage) {
987 while (NULL != (mlink = mpage->mlinks)) {
988 mpage->mlinks = mlink->next;
989 mlink_free(mlink);
990 }
991 free(mpage->sec);
992 free(mpage->arch);
993 free(mpage->title);
994 free(mpage->desc);
995 free(mpage);
996 mpage = ohash_next(&mpages, &slot);
997 }
998 }
999
1000 /*
1001 * For each mlink to the mpage, check whether the path looks like
1002 * it is formatted, and if it does, check whether a source manual
1003 * exists by the same name, ignoring the suffix.
1004 * If both conditions hold, drop the mlink.
1005 */
1006 static void
1007 mlinks_undupe(struct mpage *mpage)
1008 {
1009 char buf[PATH_MAX];
1010 struct mlink **prev;
1011 struct mlink *mlink;
1012 char *bufp;
1013
1014 mpage->form = FORM_CAT;
1015 prev = &mpage->mlinks;
1016 while (NULL != (mlink = *prev)) {
1017 if (FORM_CAT != mlink->dform) {
1018 mpage->form = FORM_NONE;
1019 goto nextlink;
1020 }
1021 (void)strlcpy(buf, mlink->file, sizeof(buf));
1022 bufp = strstr(buf, "cat");
1023 assert(NULL != bufp);
1024 memcpy(bufp, "man", 3);
1025 if (NULL != (bufp = strrchr(buf, '.')))
1026 *++bufp = '\0';
1027 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1028 if (NULL == ohash_find(&mlinks,
1029 ohash_qlookup(&mlinks, buf)))
1030 goto nextlink;
1031 if (warnings)
1032 say(mlink->file, "Man source exists: %s", buf);
1033 if (use_all)
1034 goto nextlink;
1035 *prev = mlink->next;
1036 mlink_free(mlink);
1037 continue;
1038 nextlink:
1039 prev = &(*prev)->next;
1040 }
1041 }
1042
1043 static void
1044 mlink_check(struct mpage *mpage, struct mlink *mlink)
1045 {
1046 struct str *str;
1047 unsigned int slot;
1048
1049 /*
1050 * Check whether the manual section given in a file
1051 * agrees with the directory where the file is located.
1052 * Some manuals have suffixes like (3p) on their
1053 * section number either inside the file or in the
1054 * directory name, some are linked into more than one
1055 * section, like encrypt(1) = makekey(8).
1056 */
1057
1058 if (FORM_SRC == mpage->form &&
1059 strcasecmp(mpage->sec, mlink->dsec))
1060 say(mlink->file, "Section \"%s\" manual in %s directory",
1061 mpage->sec, mlink->dsec);
1062
1063 /*
1064 * Manual page directories exist for each kernel
1065 * architecture as returned by machine(1).
1066 * However, many manuals only depend on the
1067 * application architecture as returned by arch(1).
1068 * For example, some (2/ARM) manuals are shared
1069 * across the "armish" and "zaurus" kernel
1070 * architectures.
1071 * A few manuals are even shared across completely
1072 * different architectures, for example fdformat(1)
1073 * on amd64, i386, sparc, and sparc64.
1074 */
1075
1076 if (strcasecmp(mpage->arch, mlink->arch))
1077 say(mlink->file, "Architecture \"%s\" manual in "
1078 "\"%s\" directory", mpage->arch, mlink->arch);
1079
1080 /*
1081 * XXX
1082 * parse_cat() doesn't set NAME_TITLE yet.
1083 */
1084
1085 if (FORM_CAT == mpage->form)
1086 return;
1087
1088 /*
1089 * Check whether this mlink
1090 * appears as a name in the NAME section.
1091 */
1092
1093 slot = ohash_qlookup(&names, mlink->name);
1094 str = ohash_find(&names, slot);
1095 assert(NULL != str);
1096 if ( ! (NAME_TITLE & str->mask))
1097 say(mlink->file, "Name missing in NAME section");
1098 }
1099
1100 /*
1101 * Run through the files in the global vector "mpages"
1102 * and add them to the database specified in "basedir".
1103 *
1104 * This handles the parsing scheme itself, using the cues of directory
1105 * and filename to determine whether the file is parsable or not.
1106 */
1107 static void
1108 mpages_merge(struct mparse *mp)
1109 {
1110 char any[] = "any";
1111 struct mpage *mpage, *mpage_dest;
1112 struct mlink *mlink, *mlink_dest;
1113 struct roff_man *man;
1114 char *sodest;
1115 char *cp;
1116 int fd;
1117 unsigned int pslot;
1118
1119 if ( ! nodb)
1120 SQL_EXEC("BEGIN TRANSACTION");
1121
1122 mpage = ohash_first(&mpages, &pslot);
1123 while (mpage != NULL) {
1124 mlinks_undupe(mpage);
1125 if ((mlink = mpage->mlinks) == NULL) {
1126 mpage = ohash_next(&mpages, &pslot);
1127 continue;
1128 }
1129
1130 name_mask = NAME_MASK;
1131 mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1132 mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1133 mparse_reset(mp);
1134 man = NULL;
1135 sodest = NULL;
1136
1137 if ((fd = mparse_open(mp, mlink->file)) == -1) {
1138 say(mlink->file, "&open");
1139 goto nextpage;
1140 }
1141
1142 /*
1143 * Interpret the file as mdoc(7) or man(7) source
1144 * code, unless it is known to be formatted.
1145 */
1146 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1147 mparse_readfd(mp, fd, mlink->file);
1148 close(fd);
1149 mparse_result(mp, &man, &sodest);
1150 }
1151
1152 if (sodest != NULL) {
1153 mlink_dest = ohash_find(&mlinks,
1154 ohash_qlookup(&mlinks, sodest));
1155 if (mlink_dest == NULL) {
1156 mandoc_asprintf(&cp, "%s.gz", sodest);
1157 mlink_dest = ohash_find(&mlinks,
1158 ohash_qlookup(&mlinks, cp));
1159 free(cp);
1160 }
1161 if (mlink_dest != NULL) {
1162
1163 /* The .so target exists. */
1164
1165 mpage_dest = mlink_dest->mpage;
1166 while (1) {
1167 mlink->mpage = mpage_dest;
1168
1169 /*
1170 * If the target was already
1171 * processed, add the links
1172 * to the database now.
1173 * Otherwise, this will
1174 * happen when we come
1175 * to the target.
1176 */
1177
1178 if (mpage_dest->pageid)
1179 dbadd_mlink_name(mlink);
1180
1181 if (mlink->next == NULL)
1182 break;
1183 mlink = mlink->next;
1184 }
1185
1186 /* Move all links to the target. */
1187
1188 mlink->next = mlink_dest->next;
1189 mlink_dest->next = mpage->mlinks;
1190 mpage->mlinks = NULL;
1191 }
1192 goto nextpage;
1193 } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1194 mdoc_validate(man);
1195 mpage->form = FORM_SRC;
1196 mpage->sec = man->meta.msec;
1197 mpage->sec = mandoc_strdup(
1198 mpage->sec == NULL ? "" : mpage->sec);
1199 mpage->arch = man->meta.arch;
1200 mpage->arch = mandoc_strdup(
1201 mpage->arch == NULL ? "" : mpage->arch);
1202 mpage->title = mandoc_strdup(man->meta.title);
1203 } else if (man != NULL && man->macroset == MACROSET_MAN) {
1204 man_validate(man);
1205 mpage->form = FORM_SRC;
1206 mpage->sec = mandoc_strdup(man->meta.msec);
1207 mpage->arch = mandoc_strdup(mlink->arch);
1208 mpage->title = mandoc_strdup(man->meta.title);
1209 } else {
1210 mpage->form = FORM_CAT;
1211 mpage->sec = mandoc_strdup(mlink->dsec);
1212 mpage->arch = mandoc_strdup(mlink->arch);
1213 mpage->title = mandoc_strdup(mlink->name);
1214 }
1215 putkey(mpage, mpage->sec, TYPE_sec);
1216 if (*mpage->arch != '\0')
1217 putkey(mpage, mpage->arch, TYPE_arch);
1218
1219 for ( ; mlink != NULL; mlink = mlink->next) {
1220 if ('\0' != *mlink->dsec)
1221 putkey(mpage, mlink->dsec, TYPE_sec);
1222 if ('\0' != *mlink->fsec)
1223 putkey(mpage, mlink->fsec, TYPE_sec);
1224 putkey(mpage, '\0' == *mlink->arch ?
1225 any : mlink->arch, TYPE_arch);
1226 putkey(mpage, mlink->name, NAME_FILE);
1227 }
1228
1229 assert(mpage->desc == NULL);
1230 if (man != NULL && man->macroset == MACROSET_MDOC)
1231 parse_mdoc(mpage, &man->meta, man->first);
1232 else if (man != NULL)
1233 parse_man(mpage, &man->meta, man->first);
1234 else
1235 parse_cat(mpage, fd);
1236 if (mpage->desc == NULL)
1237 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1238
1239 if (warnings && !use_all)
1240 for (mlink = mpage->mlinks; mlink;
1241 mlink = mlink->next)
1242 mlink_check(mpage, mlink);
1243
1244 dbadd(mpage);
1245 mlink = mpage->mlinks;
1246
1247 nextpage:
1248 ohash_delete(&strings);
1249 ohash_delete(&names);
1250 mpage = ohash_next(&mpages, &pslot);
1251 }
1252
1253 if (0 == nodb)
1254 SQL_EXEC("END TRANSACTION");
1255 }
1256
1257 static void
1258 names_check(void)
1259 {
1260 sqlite3_stmt *stmt;
1261 const char *name, *sec, *arch, *key;
1262
1263 sqlite3_prepare_v2(db,
1264 "SELECT name, sec, arch, key FROM ("
1265 "SELECT name AS key, pageid FROM names "
1266 "WHERE bits & ? AND NOT EXISTS ("
1267 "SELECT pageid FROM mlinks "
1268 "WHERE mlinks.pageid == names.pageid "
1269 "AND mlinks.name == names.name"
1270 ")"
1271 ") JOIN ("
1272 "SELECT sec, arch, name, pageid FROM mlinks "
1273 "GROUP BY pageid"
1274 ") USING (pageid);",
1275 -1, &stmt, NULL);
1276
1277 if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK)
1278 say("", "%s", sqlite3_errmsg(db));
1279
1280 while (sqlite3_step(stmt) == SQLITE_ROW) {
1281 name = (const char *)sqlite3_column_text(stmt, 0);
1282 sec = (const char *)sqlite3_column_text(stmt, 1);
1283 arch = (const char *)sqlite3_column_text(stmt, 2);
1284 key = (const char *)sqlite3_column_text(stmt, 3);
1285 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1286 '\0' == *arch ? "" : "/",
1287 '\0' == *arch ? "" : arch, key);
1288 }
1289 sqlite3_finalize(stmt);
1290 }
1291
1292 static void
1293 parse_cat(struct mpage *mpage, int fd)
1294 {
1295 FILE *stream;
1296 char *line, *p, *title;
1297 size_t linesz, plen, titlesz;
1298 ssize_t len;
1299 int offs;
1300
1301 stream = (-1 == fd) ?
1302 fopen(mpage->mlinks->file, "r") :
1303 fdopen(fd, "r");
1304 if (NULL == stream) {
1305 if (-1 != fd)
1306 close(fd);
1307 if (warnings)
1308 say(mpage->mlinks->file, "&fopen");
1309 return;
1310 }
1311
1312 line = NULL;
1313 linesz = 0;
1314
1315 /* Skip to first blank line. */
1316
1317 while (getline(&line, &linesz, stream) != -1)
1318 if (*line == '\n')
1319 break;
1320
1321 /*
1322 * Assume the first line that is not indented
1323 * is the first section header. Skip to it.
1324 */
1325
1326 while (getline(&line, &linesz, stream) != -1)
1327 if (*line != '\n' && *line != ' ')
1328 break;
1329
1330 /*
1331 * Read up until the next section into a buffer.
1332 * Strip the leading and trailing newline from each read line,
1333 * appending a trailing space.
1334 * Ignore empty (whitespace-only) lines.
1335 */
1336
1337 titlesz = 0;
1338 title = NULL;
1339
1340 while ((len = getline(&line, &linesz, stream)) != -1) {
1341 if (*line != ' ')
1342 break;
1343 offs = 0;
1344 while (isspace((unsigned char)line[offs]))
1345 offs++;
1346 if (line[offs] == '\0')
1347 continue;
1348 title = mandoc_realloc(title, titlesz + len - offs);
1349 memcpy(title + titlesz, line + offs, len - offs);
1350 titlesz += len - offs;
1351 title[titlesz - 1] = ' ';
1352 }
1353 free(line);
1354
1355 /*
1356 * If no page content can be found, or the input line
1357 * is already the next section header, or there is no
1358 * trailing newline, reuse the page title as the page
1359 * description.
1360 */
1361
1362 if (NULL == title || '\0' == *title) {
1363 if (warnings)
1364 say(mpage->mlinks->file,
1365 "Cannot find NAME section");
1366 fclose(stream);
1367 free(title);
1368 return;
1369 }
1370
1371 title[titlesz - 1] = '\0';
1372
1373 /*
1374 * Skip to the first dash.
1375 * Use the remaining line as the description (no more than 70
1376 * bytes).
1377 */
1378
1379 if (NULL != (p = strstr(title, "- "))) {
1380 for (p += 2; ' ' == *p || '\b' == *p; p++)
1381 /* Skip to next word. */ ;
1382 } else {
1383 if (warnings)
1384 say(mpage->mlinks->file,
1385 "No dash in title line");
1386 p = title;
1387 }
1388
1389 plen = strlen(p);
1390
1391 /* Strip backspace-encoding from line. */
1392
1393 while (NULL != (line = memchr(p, '\b', plen))) {
1394 len = line - p;
1395 if (0 == len) {
1396 memmove(line, line + 1, plen--);
1397 continue;
1398 }
1399 memmove(line - 1, line + 1, plen - len);
1400 plen -= 2;
1401 }
1402
1403 mpage->desc = mandoc_strdup(p);
1404 fclose(stream);
1405 free(title);
1406 }
1407
1408 /*
1409 * Put a type/word pair into the word database for this particular file.
1410 */
1411 static void
1412 putkey(const struct mpage *mpage, char *value, uint64_t type)
1413 {
1414 char *cp;
1415
1416 assert(NULL != value);
1417 if (TYPE_arch == type)
1418 for (cp = value; *cp; cp++)
1419 if (isupper((unsigned char)*cp))
1420 *cp = _tolower((unsigned char)*cp);
1421 putkeys(mpage, value, strlen(value), type);
1422 }
1423
1424 /*
1425 * Grok all nodes at or below a certain mdoc node into putkey().
1426 */
1427 static void
1428 putmdockey(const struct mpage *mpage,
1429 const struct roff_node *n, uint64_t m)
1430 {
1431
1432 for ( ; NULL != n; n = n->next) {
1433 if (NULL != n->child)
1434 putmdockey(mpage, n->child, m);
1435 if (n->type == ROFFT_TEXT)
1436 putkey(mpage, n->string, m);
1437 }
1438 }
1439
1440 static void
1441 parse_man(struct mpage *mpage, const struct roff_meta *meta,
1442 const struct roff_node *n)
1443 {
1444 const struct roff_node *head, *body;
1445 char *start, *title;
1446 char byte;
1447 size_t sz;
1448
1449 if (NULL == n)
1450 return;
1451
1452 /*
1453 * We're only searching for one thing: the first text child in
1454 * the BODY of a NAME section. Since we don't keep track of
1455 * sections in -man, run some hoops to find out whether we're in
1456 * the correct section or not.
1457 */
1458
1459 if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1460 body = n;
1461 assert(body->parent);
1462 if (NULL != (head = body->parent->head) &&
1463 1 == head->nchild &&
1464 NULL != (head = (head->child)) &&
1465 head->type == ROFFT_TEXT &&
1466 0 == strcmp(head->string, "NAME") &&
1467 NULL != body->child) {
1468
1469 /*
1470 * Suck the entire NAME section into memory.
1471 * Yes, we might run away.
1472 * But too many manuals have big, spread-out
1473 * NAME sections over many lines.
1474 */
1475
1476 title = NULL;
1477 deroff(&title, body);
1478 if (NULL == title)
1479 return;
1480
1481 /*
1482 * Go through a special heuristic dance here.
1483 * Conventionally, one or more manual names are
1484 * comma-specified prior to a whitespace, then a
1485 * dash, then a description. Try to puzzle out
1486 * the name parts here.
1487 */
1488
1489 start = title;
1490 for ( ;; ) {
1491 sz = strcspn(start, " ,");
1492 if ('\0' == start[sz])
1493 break;
1494
1495 byte = start[sz];
1496 start[sz] = '\0';
1497
1498 /*
1499 * Assume a stray trailing comma in the
1500 * name list if a name begins with a dash.
1501 */
1502
1503 if ('-' == start[0] ||
1504 ('\\' == start[0] && '-' == start[1]))
1505 break;
1506
1507 putkey(mpage, start, NAME_TITLE);
1508 if ( ! (mpage->name_head_done ||
1509 strcasecmp(start, meta->title))) {
1510 putkey(mpage, start, NAME_HEAD);
1511 mpage->name_head_done = 1;
1512 }
1513
1514 if (' ' == byte) {
1515 start += sz + 1;
1516 break;
1517 }
1518
1519 assert(',' == byte);
1520 start += sz + 1;
1521 while (' ' == *start)
1522 start++;
1523 }
1524
1525 if (start == title) {
1526 putkey(mpage, start, NAME_TITLE);
1527 if ( ! (mpage->name_head_done ||
1528 strcasecmp(start, meta->title))) {
1529 putkey(mpage, start, NAME_HEAD);
1530 mpage->name_head_done = 1;
1531 }
1532 free(title);
1533 return;
1534 }
1535
1536 while (isspace((unsigned char)*start))
1537 start++;
1538
1539 if (0 == strncmp(start, "-", 1))
1540 start += 1;
1541 else if (0 == strncmp(start, "\\-\\-", 4))
1542 start += 4;
1543 else if (0 == strncmp(start, "\\-", 2))
1544 start += 2;
1545 else if (0 == strncmp(start, "\\(en", 4))
1546 start += 4;
1547 else if (0 == strncmp(start, "\\(em", 4))
1548 start += 4;
1549
1550 while (' ' == *start)
1551 start++;
1552
1553 mpage->desc = mandoc_strdup(start);
1554 free(title);
1555 return;
1556 }
1557 }
1558
1559 for (n = n->child; n; n = n->next) {
1560 if (NULL != mpage->desc)
1561 break;
1562 parse_man(mpage, meta, n);
1563 }
1564 }
1565
1566 static void
1567 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1568 const struct roff_node *n)
1569 {
1570
1571 assert(NULL != n);
1572 for (n = n->child; NULL != n; n = n->next) {
1573 switch (n->type) {
1574 case ROFFT_ELEM:
1575 case ROFFT_BLOCK:
1576 case ROFFT_HEAD:
1577 case ROFFT_BODY:
1578 case ROFFT_TAIL:
1579 if (NULL != mdocs[n->tok].fp)
1580 if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1581 break;
1582 if (mdocs[n->tok].mask)
1583 putmdockey(mpage, n->child,
1584 mdocs[n->tok].mask);
1585 break;
1586 default:
1587 assert(n->type != ROFFT_ROOT);
1588 continue;
1589 }
1590 if (NULL != n->child)
1591 parse_mdoc(mpage, meta, n);
1592 }
1593 }
1594
1595 static int
1596 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1597 const struct roff_node *n)
1598 {
1599 char *start, *end;
1600 size_t sz;
1601
1602 if (SEC_SYNOPSIS != n->sec ||
1603 NULL == (n = n->child) ||
1604 n->type != ROFFT_TEXT)
1605 return 0;
1606
1607 /*
1608 * Only consider those `Fd' macro fields that begin with an
1609 * "inclusion" token (versus, e.g., #define).
1610 */
1611
1612 if (strcmp("#include", n->string))
1613 return 0;
1614
1615 if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1616 return 0;
1617
1618 /*
1619 * Strip away the enclosing angle brackets and make sure we're
1620 * not zero-length.
1621 */
1622
1623 start = n->string;
1624 if ('<' == *start || '"' == *start)
1625 start++;
1626
1627 if (0 == (sz = strlen(start)))
1628 return 0;
1629
1630 end = &start[(int)sz - 1];
1631 if ('>' == *end || '"' == *end)
1632 end--;
1633
1634 if (end > start)
1635 putkeys(mpage, start, end - start + 1, TYPE_In);
1636 return 0;
1637 }
1638
1639 static void
1640 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1641 {
1642 char *cp;
1643 size_t sz;
1644
1645 if (n->type != ROFFT_TEXT)
1646 return;
1647
1648 /* Skip function pointer punctuation. */
1649
1650 cp = n->string;
1651 while (*cp == '(' || *cp == '*')
1652 cp++;
1653 sz = strcspn(cp, "()");
1654
1655 putkeys(mpage, cp, sz, TYPE_Fn);
1656 if (n->sec == SEC_SYNOPSIS)
1657 putkeys(mpage, cp, sz, NAME_SYN);
1658 }
1659
1660 static int
1661 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1662 const struct roff_node *n)
1663 {
1664
1665 if (n->child == NULL)
1666 return 0;
1667
1668 parse_mdoc_fname(mpage, n->child);
1669
1670 for (n = n->child->next; n != NULL; n = n->next)
1671 if (n->type == ROFFT_TEXT)
1672 putkey(mpage, n->string, TYPE_Fa);
1673
1674 return 0;
1675 }
1676
1677 static int
1678 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1679 const struct roff_node *n)
1680 {
1681
1682 if (n->type != ROFFT_HEAD)
1683 return 1;
1684
1685 if (n->child != NULL)
1686 parse_mdoc_fname(mpage, n->child);
1687
1688 return 0;
1689 }
1690
1691 static int
1692 parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1693 const struct roff_node *n)
1694 {
1695 char *cp;
1696
1697 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1698 return 0;
1699
1700 if (n->nchild == 1 && n->child->type == ROFFT_TEXT)
1701 return 1;
1702
1703 cp = NULL;
1704 deroff(&cp, n);
1705 if (cp != NULL) {
1706 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1707 n->type == ROFFT_BODY ? TYPE_Va : 0));
1708 free(cp);
1709 }
1710
1711 return 0;
1712 }
1713
1714 static int
1715 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1716 const struct roff_node *n)
1717 {
1718 char *cp;
1719
1720 if (NULL == (n = n->child))
1721 return 0;
1722
1723 if (NULL == n->next) {
1724 putkey(mpage, n->string, TYPE_Xr);
1725 return 0;
1726 }
1727
1728 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1729 putkey(mpage, cp, TYPE_Xr);
1730 free(cp);
1731 return 0;
1732 }
1733
1734 static int
1735 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1736 const struct roff_node *n)
1737 {
1738
1739 if (n->type == ROFFT_BODY)
1740 deroff(&mpage->desc, n);
1741 return 0;
1742 }
1743
1744 static int
1745 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1746 const struct roff_node *n)
1747 {
1748
1749 if (SEC_NAME == n->sec)
1750 putmdockey(mpage, n->child, NAME_TITLE);
1751 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1752 if (n->child == NULL)
1753 putkey(mpage, meta->name, NAME_SYN);
1754 else
1755 putmdockey(mpage, n->child, NAME_SYN);
1756 }
1757 if ( ! (mpage->name_head_done ||
1758 n->child == NULL || n->child->string == NULL ||
1759 strcasecmp(n->child->string, meta->title))) {
1760 putkey(mpage, n->child->string, ROFFT_HEAD);
1761 mpage->name_head_done = 1;
1762 }
1763 return 0;
1764 }
1765
1766 static int
1767 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1768 const struct roff_node *n)
1769 {
1770
1771 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1772 }
1773
1774 static int
1775 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1776 const struct roff_node *n)
1777 {
1778
1779 return n->type == ROFFT_HEAD;
1780 }
1781
1782 /*
1783 * Add a string to the hash table for the current manual.
1784 * Each string has a bitmask telling which macros it belongs to.
1785 * When we finish the manual, we'll dump the table.
1786 */
1787 static void
1788 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1789 {
1790 struct ohash *htab;
1791 struct str *s;
1792 const char *end;
1793 unsigned int slot;
1794 int i, mustfree;
1795
1796 if (0 == sz)
1797 return;
1798
1799 mustfree = render_string(&cp, &sz);
1800
1801 if (TYPE_Nm & v) {
1802 htab = &names;
1803 v &= name_mask;
1804 if (v & NAME_FIRST)
1805 name_mask &= ~NAME_FIRST;
1806 if (debug > 1)
1807 say(mpage->mlinks->file,
1808 "Adding name %*s, bits=%d", sz, cp, v);
1809 } else {
1810 htab = &strings;
1811 if (debug > 1)
1812 for (i = 0; i < mansearch_keymax; i++)
1813 if ((uint64_t)1 << i & v)
1814 say(mpage->mlinks->file,
1815 "Adding key %s=%*s",
1816 mansearch_keynames[i], sz, cp);
1817 }
1818
1819 end = cp + sz;
1820 slot = ohash_qlookupi(htab, cp, &end);
1821 s = ohash_find(htab, slot);
1822
1823 if (NULL != s && mpage == s->mpage) {
1824 s->mask |= v;
1825 return;
1826 } else if (NULL == s) {
1827 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1828 memcpy(s->key, cp, sz);
1829 ohash_insert(htab, slot, s);
1830 }
1831 s->mpage = mpage;
1832 s->mask = v;
1833
1834 if (mustfree)
1835 free(cp);
1836 }
1837
1838 /*
1839 * Take a Unicode codepoint and produce its UTF-8 encoding.
1840 * This isn't the best way to do this, but it works.
1841 * The magic numbers are from the UTF-8 packaging.
1842 * They're not as scary as they seem: read the UTF-8 spec for details.
1843 */
1844 static size_t
1845 utf8(unsigned int cp, char out[7])
1846 {
1847 size_t rc;
1848
1849 rc = 0;
1850 if (cp <= 0x0000007F) {
1851 rc = 1;
1852 out[0] = (char)cp;
1853 } else if (cp <= 0x000007FF) {
1854 rc = 2;
1855 out[0] = (cp >> 6 & 31) | 192;
1856 out[1] = (cp & 63) | 128;
1857 } else if (cp <= 0x0000FFFF) {
1858 rc = 3;
1859 out[0] = (cp >> 12 & 15) | 224;
1860 out[1] = (cp >> 6 & 63) | 128;
1861 out[2] = (cp & 63) | 128;
1862 } else if (cp <= 0x001FFFFF) {
1863 rc = 4;
1864 out[0] = (cp >> 18 & 7) | 240;
1865 out[1] = (cp >> 12 & 63) | 128;
1866 out[2] = (cp >> 6 & 63) | 128;
1867 out[3] = (cp & 63) | 128;
1868 } else if (cp <= 0x03FFFFFF) {
1869 rc = 5;
1870 out[0] = (cp >> 24 & 3) | 248;
1871 out[1] = (cp >> 18 & 63) | 128;
1872 out[2] = (cp >> 12 & 63) | 128;
1873 out[3] = (cp >> 6 & 63) | 128;
1874 out[4] = (cp & 63) | 128;
1875 } else if (cp <= 0x7FFFFFFF) {
1876 rc = 6;
1877 out[0] = (cp >> 30 & 1) | 252;
1878 out[1] = (cp >> 24 & 63) | 128;
1879 out[2] = (cp >> 18 & 63) | 128;
1880 out[3] = (cp >> 12 & 63) | 128;
1881 out[4] = (cp >> 6 & 63) | 128;
1882 out[5] = (cp & 63) | 128;
1883 } else
1884 return 0;
1885
1886 out[rc] = '\0';
1887 return rc;
1888 }
1889
1890 /*
1891 * If the string contains escape sequences,
1892 * replace it with an allocated rendering and return 1,
1893 * such that the caller can free it after use.
1894 * Otherwise, do nothing and return 0.
1895 */
1896 static int
1897 render_string(char **public, size_t *psz)
1898 {
1899 const char *src, *scp, *addcp, *seq;
1900 char *dst;
1901 size_t ssz, dsz, addsz;
1902 char utfbuf[7], res[6];
1903 int seqlen, unicode;
1904
1905 res[0] = '\\';
1906 res[1] = '\t';
1907 res[2] = ASCII_NBRSP;
1908 res[3] = ASCII_HYPH;
1909 res[4] = ASCII_BREAK;
1910 res[5] = '\0';
1911
1912 src = scp = *public;
1913 ssz = *psz;
1914 dst = NULL;
1915 dsz = 0;
1916
1917 while (scp < src + *psz) {
1918
1919 /* Leave normal characters unchanged. */
1920
1921 if (strchr(res, *scp) == NULL) {
1922 if (dst != NULL)
1923 dst[dsz++] = *scp;
1924 scp++;
1925 continue;
1926 }
1927
1928 /*
1929 * Found something that requires replacing,
1930 * make sure we have a destination buffer.
1931 */
1932
1933 if (dst == NULL) {
1934 dst = mandoc_malloc(ssz + 1);
1935 dsz = scp - src;
1936 memcpy(dst, src, dsz);
1937 }
1938
1939 /* Handle single-char special characters. */
1940
1941 switch (*scp) {
1942 case '\\':
1943 break;
1944 case '\t':
1945 case ASCII_NBRSP:
1946 dst[dsz++] = ' ';
1947 scp++;
1948 continue;
1949 case ASCII_HYPH:
1950 dst[dsz++] = '-';
1951 /* FALLTHROUGH */
1952 case ASCII_BREAK:
1953 scp++;
1954 continue;
1955 default:
1956 abort();
1957 }
1958
1959 /*
1960 * Found an escape sequence.
1961 * Read past the slash, then parse it.
1962 * Ignore everything except characters.
1963 */
1964
1965 scp++;
1966 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1967 continue;
1968
1969 /*
1970 * Render the special character
1971 * as either UTF-8 or ASCII.
1972 */
1973
1974 if (write_utf8) {
1975 unicode = mchars_spec2cp(seq, seqlen);
1976 if (unicode <= 0)
1977 continue;
1978 addsz = utf8(unicode, utfbuf);
1979 if (addsz == 0)
1980 continue;
1981 addcp = utfbuf;
1982 } else {
1983 addcp = mchars_spec2str(seq, seqlen, &addsz);
1984 if (addcp == NULL)
1985 continue;
1986 if (*addcp == ASCII_NBRSP) {
1987 addcp = " ";
1988 addsz = 1;
1989 }
1990 }
1991
1992 /* Copy the rendered glyph into the stream. */
1993
1994 ssz += addsz;
1995 dst = mandoc_realloc(dst, ssz + 1);
1996 memcpy(dst + dsz, addcp, addsz);
1997 dsz += addsz;
1998 }
1999 if (dst != NULL) {
2000 *public = dst;
2001 *psz = dsz;
2002 }
2003
2004 /* Trim trailing whitespace and NUL-terminate. */
2005
2006 while (*psz > 0 && (*public)[*psz - 1] == ' ')
2007 --*psz;
2008 if (dst != NULL) {
2009 (*public)[*psz] = '\0';
2010 return 1;
2011 } else
2012 return 0;
2013 }
2014
2015 static void
2016 dbadd_mlink(const struct mlink *mlink)
2017 {
2018 size_t i;
2019
2020 i = 1;
2021 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
2022 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
2023 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
2024 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
2025 SQL_STEP(stmts[STMT_INSERT_LINK]);
2026 sqlite3_reset(stmts[STMT_INSERT_LINK]);
2027 }
2028
2029 static void
2030 dbadd_mlink_name(const struct mlink *mlink)
2031 {
2032 uint64_t bits;
2033 size_t i;
2034
2035 dbadd_mlink(mlink);
2036
2037 i = 1;
2038 SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
2039 bits = NAME_FILE & NAME_MASK;
2040 if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
2041 bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
2042 sqlite3_reset(stmts[STMT_SELECT_NAME]);
2043 }
2044
2045 i = 1;
2046 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
2047 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2048 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2049 SQL_STEP(stmts[STMT_INSERT_NAME]);
2050 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2051 }
2052
2053 /*
2054 * Flush the current page's terms (and their bits) into the database.
2055 * Wrap the entire set of additions in a transaction to make sqlite be a
2056 * little faster.
2057 * Also, handle escape sequences at the last possible moment.
2058 */
2059 static void
2060 dbadd(struct mpage *mpage)
2061 {
2062 struct mlink *mlink;
2063 struct str *key;
2064 char *cp;
2065 size_t i;
2066 unsigned int slot;
2067 int mustfree;
2068
2069 mlink = mpage->mlinks;
2070
2071 if (nodb) {
2072 for (key = ohash_first(&names, &slot); NULL != key;
2073 key = ohash_next(&names, &slot))
2074 free(key);
2075 for (key = ohash_first(&strings, &slot); NULL != key;
2076 key = ohash_next(&strings, &slot))
2077 free(key);
2078 if (0 == debug)
2079 return;
2080 while (NULL != mlink) {
2081 fputs(mlink->name, stdout);
2082 if (NULL == mlink->next ||
2083 strcmp(mlink->dsec, mlink->next->dsec) ||
2084 strcmp(mlink->fsec, mlink->next->fsec) ||
2085 strcmp(mlink->arch, mlink->next->arch)) {
2086 putchar('(');
2087 if ('\0' == *mlink->dsec)
2088 fputs(mlink->fsec, stdout);
2089 else
2090 fputs(mlink->dsec, stdout);
2091 if ('\0' != *mlink->arch)
2092 printf("/%s", mlink->arch);
2093 putchar(')');
2094 }
2095 mlink = mlink->next;
2096 if (NULL != mlink)
2097 fputs(", ", stdout);
2098 }
2099 printf(" - %s\n", mpage->desc);
2100 return;
2101 }
2102
2103 if (debug)
2104 say(mlink->file, "Adding to database");
2105
2106 cp = mpage->desc;
2107 i = strlen(cp);
2108 mustfree = render_string(&cp, &i);
2109 i = 1;
2110 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
2111 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
2112 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2113 mpage->pageid = sqlite3_last_insert_rowid(db);
2114 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2115 if (mustfree)
2116 free(cp);
2117
2118 while (NULL != mlink) {
2119 dbadd_mlink(mlink);
2120 mlink = mlink->next;
2121 }
2122 mlink = mpage->mlinks;
2123
2124 for (key = ohash_first(&names, &slot); NULL != key;
2125 key = ohash_next(&names, &slot)) {
2126 assert(key->mpage == mpage);
2127 i = 1;
2128 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2129 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
2130 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2131 SQL_STEP(stmts[STMT_INSERT_NAME]);
2132 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2133 free(key);
2134 }
2135 for (key = ohash_first(&strings, &slot); NULL != key;
2136 key = ohash_next(&strings, &slot)) {
2137 assert(key->mpage == mpage);
2138 i = 1;
2139 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2140 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
2141 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2142 SQL_STEP(stmts[STMT_INSERT_KEY]);
2143 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2144 free(key);
2145 }
2146 }
2147
2148 static void
2149 dbprune(void)
2150 {
2151 struct mpage *mpage;
2152 struct mlink *mlink;
2153 size_t i;
2154 unsigned int slot;
2155
2156 if (0 == nodb)
2157 SQL_EXEC("BEGIN TRANSACTION");
2158
2159 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2160 mpage = ohash_next(&mpages, &slot)) {
2161 mlink = mpage->mlinks;
2162 if (debug)
2163 say(mlink->file, "Deleting from database");
2164 if (nodb)
2165 continue;
2166 for ( ; NULL != mlink; mlink = mlink->next) {
2167 i = 1;
2168 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2169 i, mlink->dsec);
2170 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2171 i, mlink->arch);
2172 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2173 i, mlink->name);
2174 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2175 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2176 }
2177 }
2178
2179 if (0 == nodb)
2180 SQL_EXEC("END TRANSACTION");
2181 }
2182
2183 /*
2184 * Close an existing database and its prepared statements.
2185 * If "real" is not set, rename the temporary file into the real one.
2186 */
2187 static void
2188 dbclose(int real)
2189 {
2190 size_t i;
2191 int status;
2192 pid_t child;
2193
2194 if (nodb)
2195 return;
2196
2197 for (i = 0; i < STMT__MAX; i++) {
2198 sqlite3_finalize(stmts[i]);
2199 stmts[i] = NULL;
2200 }
2201
2202 sqlite3_close(db);
2203 db = NULL;
2204
2205 if (real)
2206 return;
2207
2208 if ('\0' == *tempfilename) {
2209 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2210 exitcode = (int)MANDOCLEVEL_SYSERR;
2211 say(MANDOC_DB, "&rename");
2212 }
2213 return;
2214 }
2215
2216 switch (child = fork()) {
2217 case -1:
2218 exitcode = (int)MANDOCLEVEL_SYSERR;
2219 say("", "&fork cmp");
2220 return;
2221 case 0:
2222 execlp("cmp", "cmp", "-s",
2223 tempfilename, MANDOC_DB, (char *)NULL);
2224 say("", "&exec cmp");
2225 exit(0);
2226 default:
2227 break;
2228 }
2229 if (-1 == waitpid(child, &status, 0)) {
2230 exitcode = (int)MANDOCLEVEL_SYSERR;
2231 say("", "&wait cmp");
2232 } else if (WIFSIGNALED(status)) {
2233 exitcode = (int)MANDOCLEVEL_SYSERR;
2234 say("", "cmp died from signal %d", WTERMSIG(status));
2235 } else if (WEXITSTATUS(status)) {
2236 exitcode = (int)MANDOCLEVEL_SYSERR;
2237 say(MANDOC_DB,
2238 "Data changed, but cannot replace database");
2239 }
2240
2241 *strrchr(tempfilename, '/') = '\0';
2242 switch (child = fork()) {
2243 case -1:
2244 exitcode = (int)MANDOCLEVEL_SYSERR;
2245 say("", "&fork rm");
2246 return;
2247 case 0:
2248 execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
2249 say("", "&exec rm");
2250 exit((int)MANDOCLEVEL_SYSERR);
2251 default:
2252 break;
2253 }
2254 if (-1 == waitpid(child, &status, 0)) {
2255 exitcode = (int)MANDOCLEVEL_SYSERR;
2256 say("", "&wait rm");
2257 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2258 exitcode = (int)MANDOCLEVEL_SYSERR;
2259 say("", "%s: Cannot remove temporary directory",
2260 tempfilename);
2261 }
2262 }
2263
2264 /*
2265 * This is straightforward stuff.
2266 * Open a database connection to a "temporary" database, then open a set
2267 * of prepared statements we'll use over and over again.
2268 * If "real" is set, we use the existing database; if not, we truncate a
2269 * temporary one.
2270 * Must be matched by dbclose().
2271 */
2272 static int
2273 dbopen(int real)
2274 {
2275 const char *sql;
2276 int rc, ofl;
2277
2278 if (nodb)
2279 return 1;
2280
2281 *tempfilename = '\0';
2282 ofl = SQLITE_OPEN_READWRITE;
2283
2284 if (real) {
2285 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2286 if (SQLITE_OK != rc) {
2287 exitcode = (int)MANDOCLEVEL_SYSERR;
2288 if (SQLITE_CANTOPEN != rc)
2289 say(MANDOC_DB, "%s", sqlite3_errstr(rc));
2290 return 0;
2291 }
2292 goto prepare_statements;
2293 }
2294
2295 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2296
2297 remove(MANDOC_DB "~");
2298 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2299 if (SQLITE_OK == rc)
2300 goto create_tables;
2301 if (MPARSE_QUICK & mparse_options) {
2302 exitcode = (int)MANDOCLEVEL_SYSERR;
2303 say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
2304 return 0;
2305 }
2306
2307 (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2308 sizeof(tempfilename));
2309 if (NULL == mkdtemp(tempfilename)) {
2310 exitcode = (int)MANDOCLEVEL_SYSERR;
2311 say("", "&%s", tempfilename);
2312 return 0;
2313 }
2314 (void)strlcat(tempfilename, "/" MANDOC_DB,
2315 sizeof(tempfilename));
2316 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2317 if (SQLITE_OK != rc) {
2318 exitcode = (int)MANDOCLEVEL_SYSERR;
2319 say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
2320 return 0;
2321 }
2322
2323 create_tables:
2324 sql = "CREATE TABLE \"mpages\" (\n"
2325 " \"desc\" TEXT NOT NULL,\n"
2326 " \"form\" INTEGER NOT NULL,\n"
2327 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2328 ");\n"
2329 "\n"
2330 "CREATE TABLE \"mlinks\" (\n"
2331 " \"sec\" TEXT NOT NULL,\n"
2332 " \"arch\" TEXT NOT NULL,\n"
2333 " \"name\" TEXT NOT NULL,\n"
2334 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2335 "ON DELETE CASCADE\n"
2336 ");\n"
2337 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2338 "\n"
2339 "CREATE TABLE \"names\" (\n"
2340 " \"bits\" INTEGER NOT NULL,\n"
2341 " \"name\" TEXT NOT NULL,\n"
2342 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2343 "ON DELETE CASCADE,\n"
2344 " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
2345 ");\n"
2346 "\n"
2347 "CREATE TABLE \"keys\" (\n"
2348 " \"bits\" INTEGER NOT NULL,\n"
2349 " \"key\" TEXT NOT NULL,\n"
2350 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2351 "ON DELETE CASCADE\n"
2352 ");\n"
2353 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2354
2355 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2356 exitcode = (int)MANDOCLEVEL_SYSERR;
2357 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2358 sqlite3_close(db);
2359 return 0;
2360 }
2361
2362 prepare_statements:
2363 if (SQLITE_OK != sqlite3_exec(db,
2364 "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2365 exitcode = (int)MANDOCLEVEL_SYSERR;
2366 say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2367 sqlite3_errmsg(db));
2368 sqlite3_close(db);
2369 return 0;
2370 }
2371
2372 sql = "DELETE FROM mpages WHERE pageid IN "
2373 "(SELECT pageid FROM mlinks WHERE "
2374 "sec=? AND arch=? AND name=?)";
2375 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2376 sql = "INSERT INTO mpages "
2377 "(desc,form) VALUES (?,?)";
2378 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2379 sql = "INSERT INTO mlinks "
2380 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2381 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2382 sql = "SELECT bits FROM names where pageid = ?";
2383 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
2384 sql = "INSERT INTO names "
2385 "(bits,name,pageid) VALUES (?,?,?)";
2386 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2387 sql = "INSERT INTO keys "
2388 "(bits,key,pageid) VALUES (?,?,?)";
2389 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2390
2391 #ifndef __APPLE__
2392 /*
2393 * When opening a new database, we can turn off
2394 * synchronous mode for much better performance.
2395 */
2396
2397 if (real && SQLITE_OK != sqlite3_exec(db,
2398 "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2399 exitcode = (int)MANDOCLEVEL_SYSERR;
2400 say(MANDOC_DB, "PRAGMA synchronous: %s",
2401 sqlite3_errmsg(db));
2402 sqlite3_close(db);
2403 return 0;
2404 }
2405 #endif
2406
2407 return 1;
2408 }
2409
2410 static int
2411 set_basedir(const char *targetdir, int report_baddir)
2412 {
2413 static char startdir[PATH_MAX];
2414 static int getcwd_status; /* 1 = ok, 2 = failure */
2415 static int chdir_status; /* 1 = changed directory */
2416 char *cp;
2417
2418 /*
2419 * Remember the original working directory, if possible.
2420 * This will be needed if the second or a later directory
2421 * on the command line is given as a relative path.
2422 * Do not error out if the current directory is not
2423 * searchable: Maybe it won't be needed after all.
2424 */
2425 if (0 == getcwd_status) {
2426 if (NULL == getcwd(startdir, sizeof(startdir))) {
2427 getcwd_status = 2;
2428 (void)strlcpy(startdir, strerror(errno),
2429 sizeof(startdir));
2430 } else
2431 getcwd_status = 1;
2432 }
2433
2434 /*
2435 * We are leaving the old base directory.
2436 * Do not use it any longer, not even for messages.
2437 */
2438 *basedir = '\0';
2439
2440 /*
2441 * If and only if the directory was changed earlier and
2442 * the next directory to process is given as a relative path,
2443 * first go back, or bail out if that is impossible.
2444 */
2445 if (chdir_status && '/' != *targetdir) {
2446 if (2 == getcwd_status) {
2447 exitcode = (int)MANDOCLEVEL_SYSERR;
2448 say("", "getcwd: %s", startdir);
2449 return 0;
2450 }
2451 if (-1 == chdir(startdir)) {
2452 exitcode = (int)MANDOCLEVEL_SYSERR;
2453 say("", "&chdir %s", startdir);
2454 return 0;
2455 }
2456 }
2457
2458 /*
2459 * Always resolve basedir to the canonicalized absolute
2460 * pathname and append a trailing slash, such that
2461 * we can reliably check whether files are inside.
2462 */
2463 if (NULL == realpath(targetdir, basedir)) {
2464 if (report_baddir || errno != ENOENT) {
2465 exitcode = (int)MANDOCLEVEL_BADARG;
2466 say("", "&%s: realpath", targetdir);
2467 }
2468 return 0;
2469 } else if (-1 == chdir(basedir)) {
2470 if (report_baddir || errno != ENOENT) {
2471 exitcode = (int)MANDOCLEVEL_BADARG;
2472 say("", "&chdir");
2473 }
2474 return 0;
2475 }
2476 chdir_status = 1;
2477 cp = strchr(basedir, '\0');
2478 if ('/' != cp[-1]) {
2479 if (cp - basedir >= PATH_MAX - 1) {
2480 exitcode = (int)MANDOCLEVEL_SYSERR;
2481 say("", "Filename too long");
2482 return 0;
2483 }
2484 *cp++ = '/';
2485 *cp = '\0';
2486 }
2487 return 1;
2488 }
2489
2490 static void
2491 say(const char *file, const char *format, ...)
2492 {
2493 va_list ap;
2494 int use_errno;
2495
2496 if ('\0' != *basedir)
2497 fprintf(stderr, "%s", basedir);
2498 if ('\0' != *basedir && '\0' != *file)
2499 fputc('/', stderr);
2500 if ('\0' != *file)
2501 fprintf(stderr, "%s", file);
2502
2503 use_errno = 1;
2504 if (NULL != format) {
2505 switch (*format) {
2506 case '&':
2507 format++;
2508 break;
2509 case '\0':
2510 format = NULL;
2511 break;
2512 default:
2513 use_errno = 0;
2514 break;
2515 }
2516 }
2517 if (NULL != format) {
2518 if ('\0' != *basedir || '\0' != *file)
2519 fputs(": ", stderr);
2520 va_start(ap, format);
2521 vfprintf(stderr, format, ap);
2522 va_end(ap);
2523 }
2524 if (use_errno) {
2525 if ('\0' != *basedir || '\0' != *file || NULL != format)
2526 fputs(": ", stderr);
2527 perror(NULL);
2528 } else
2529 fputc('\n', stderr);
2530 }
A git conversion mirror of mandoc.bsd.lv