]> 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
move man(7) validation into the dedicated validation phase, too
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.204 2015/10/20 02:01:31 schwarze Exp $ */
2 /*
3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011-2015 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 #include <err.h>
27 #include <errno.h>
28 #include <fcntl.h>
29 #if HAVE_FTS
30 #include <fts.h>
31 #else
32 #include "compat_fts.h"
33 #endif
34 #include <getopt.h>
35 #include <limits.h>
36 #include <stddef.h>
37 #include <stdio.h>
38 #include <stdint.h>
39 #include <stdlib.h>
40 #include <string.h>
41 #include <unistd.h>
42
43 #include <sqlite3.h>
44
45 #include "mandoc_aux.h"
46 #include "mandoc_ohash.h"
47 #include "mandoc.h"
48 #include "roff.h"
49 #include "mdoc.h"
50 #include "man.h"
51 #include "manconf.h"
52 #include "mansearch.h"
53
54 extern int mansearch_keymax;
55 extern const char *const mansearch_keynames[];
56
57 #define SQL_EXEC(_v) \
58 if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
59 say("", "%s: %s", (_v), sqlite3_errmsg(db))
60 #define SQL_BIND_TEXT(_s, _i, _v) \
61 if (SQLITE_OK != sqlite3_bind_text \
62 ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
63 say(mlink->file, "%s", sqlite3_errmsg(db))
64 #define SQL_BIND_INT(_s, _i, _v) \
65 if (SQLITE_OK != sqlite3_bind_int \
66 ((_s), (_i)++, (_v))) \
67 say(mlink->file, "%s", sqlite3_errmsg(db))
68 #define SQL_BIND_INT64(_s, _i, _v) \
69 if (SQLITE_OK != sqlite3_bind_int64 \
70 ((_s), (_i)++, (_v))) \
71 say(mlink->file, "%s", sqlite3_errmsg(db))
72 #define SQL_STEP(_s) \
73 if (SQLITE_DONE != sqlite3_step((_s))) \
74 say(mlink->file, "%s", sqlite3_errmsg(db))
75
76 enum op {
77 OP_DEFAULT = 0, /* new dbs from dir list or default config */
78 OP_CONFFILE, /* new databases from custom config file */
79 OP_UPDATE, /* delete/add entries in existing database */
80 OP_DELETE, /* delete entries from existing database */
81 OP_TEST /* change no databases, report potential problems */
82 };
83
84 struct str {
85 const struct mpage *mpage; /* if set, the owning parse */
86 uint64_t mask; /* bitmask in sequence */
87 char key[]; /* rendered text */
88 };
89
90 struct inodev {
91 ino_t st_ino;
92 dev_t st_dev;
93 };
94
95 struct mpage {
96 struct inodev inodev; /* used for hashing routine */
97 int64_t pageid; /* pageid in mpages SQL table */
98 char *sec; /* section from file content */
99 char *arch; /* architecture from file content */
100 char *title; /* title from file content */
101 char *desc; /* description from file content */
102 struct mlink *mlinks; /* singly linked list */
103 int form; /* format from file content */
104 int name_head_done;
105 };
106
107 struct mlink {
108 char file[PATH_MAX]; /* filename rel. to manpath */
109 char *dsec; /* section from directory */
110 char *arch; /* architecture from directory */
111 char *name; /* name from file name (not empty) */
112 char *fsec; /* section from file name suffix */
113 struct mlink *next; /* singly linked list */
114 struct mpage *mpage; /* parent */
115 int dform; /* format from directory */
116 int fform; /* format from file name suffix */
117 int gzip; /* filename has a .gz suffix */
118 };
119
120 enum stmt {
121 STMT_DELETE_PAGE = 0, /* delete mpage */
122 STMT_INSERT_PAGE, /* insert mpage */
123 STMT_INSERT_LINK, /* insert mlink */
124 STMT_INSERT_NAME, /* insert name */
125 STMT_SELECT_NAME, /* retrieve existing name flags */
126 STMT_INSERT_KEY, /* insert parsed key */
127 STMT__MAX
128 };
129
130 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
131 const struct roff_node *);
132
133 struct mdoc_handler {
134 mdoc_fp fp; /* optional handler */
135 uint64_t mask; /* set unless handler returns 0 */
136 };
137
138 static void dbclose(int);
139 static void dbadd(struct mpage *);
140 static void dbadd_mlink(const struct mlink *mlink);
141 static void dbadd_mlink_name(const struct mlink *mlink);
142 static int dbopen(int);
143 static void dbprune(void);
144 static void filescan(const char *);
145 static void mlink_add(struct mlink *, const struct stat *);
146 static void mlink_check(struct mpage *, struct mlink *);
147 static void mlink_free(struct mlink *);
148 static void mlinks_undupe(struct mpage *);
149 static void mpages_free(void);
150 static void mpages_merge(struct mparse *);
151 static void names_check(void);
152 static void parse_cat(struct mpage *, int);
153 static void parse_man(struct mpage *, const struct roff_meta *,
154 const struct roff_node *);
155 static void parse_mdoc(struct mpage *, const struct roff_meta *,
156 const struct roff_node *);
157 static int parse_mdoc_body(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_Xr(struct mpage *, const struct roff_meta *,
175 const struct roff_node *);
176 static void putkey(const struct mpage *, char *, uint64_t);
177 static void putkeys(const struct mpage *, char *, size_t, uint64_t);
178 static void putmdockey(const struct mpage *,
179 const struct roff_node *, uint64_t);
180 static int render_string(char **, size_t *);
181 static void say(const char *, const char *, ...);
182 static int set_basedir(const char *, int);
183 static int treescan(void);
184 static size_t utf8(unsigned int, char [7]);
185
186 extern char *__progname;
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 { NULL, TYPE_Va }, /* Va */
246 { parse_mdoc_body, TYPE_Va }, /* 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;
339 size_t j, sz;
340 int ch, i;
341
342 memset(&conf, 0, sizeof(conf));
343 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
344
345 /*
346 * We accept a few different invocations.
347 * The CHECKOP macro makes sure that invocation styles don't
348 * clobber each other.
349 */
350 #define CHECKOP(_op, _ch) do \
351 if (OP_DEFAULT != (_op)) { \
352 warnx("-%c: Conflicting option", (_ch)); \
353 goto usage; \
354 } while (/*CONSTCOND*/0)
355
356 path_arg = NULL;
357 op = OP_DEFAULT;
358
359 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
360 switch (ch) {
361 case 'a':
362 use_all = 1;
363 break;
364 case 'C':
365 CHECKOP(op, ch);
366 path_arg = optarg;
367 op = OP_CONFFILE;
368 break;
369 case 'D':
370 debug++;
371 break;
372 case 'd':
373 CHECKOP(op, ch);
374 path_arg = optarg;
375 op = OP_UPDATE;
376 break;
377 case 'n':
378 nodb = 1;
379 break;
380 case 'p':
381 warnings = 1;
382 break;
383 case 'Q':
384 mparse_options |= MPARSE_QUICK;
385 break;
386 case 'T':
387 if (strcmp(optarg, "utf8")) {
388 warnx("-T%s: Unsupported output format",
389 optarg);
390 goto usage;
391 }
392 write_utf8 = 1;
393 break;
394 case 't':
395 CHECKOP(op, ch);
396 dup2(STDOUT_FILENO, STDERR_FILENO);
397 op = OP_TEST;
398 nodb = warnings = 1;
399 break;
400 case 'u':
401 CHECKOP(op, ch);
402 path_arg = optarg;
403 op = OP_DELETE;
404 break;
405 case 'v':
406 /* Compatibility with espie@'s makewhatis. */
407 break;
408 default:
409 goto usage;
410 }
411
412 argc -= optind;
413 argv += optind;
414
415 if (OP_CONFFILE == op && argc > 0) {
416 warnx("-C: Too many arguments");
417 goto usage;
418 }
419
420 exitcode = (int)MANDOCLEVEL_OK;
421 mchars_alloc();
422 mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
423 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
424 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
425
426 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
427
428 /*
429 * Most of these deal with a specific directory.
430 * Jump into that directory first.
431 */
432 if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
433 goto out;
434
435 if (dbopen(1)) {
436 /*
437 * The existing database is usable. Process
438 * all files specified on the command-line.
439 */
440 use_all = 1;
441 for (i = 0; i < argc; i++)
442 filescan(argv[i]);
443 if (OP_TEST != op)
444 dbprune();
445 } else {
446 /*
447 * Database missing or corrupt.
448 * Recreate from scratch.
449 */
450 exitcode = (int)MANDOCLEVEL_OK;
451 op = OP_DEFAULT;
452 if (0 == treescan())
453 goto out;
454 if (0 == dbopen(0))
455 goto out;
456 }
457 if (OP_DELETE != op)
458 mpages_merge(mp);
459 dbclose(OP_DEFAULT == op ? 0 : 1);
460 } else {
461 /*
462 * If we have arguments, use them as our manpaths.
463 * If we don't, grok from manpath(1) or however else
464 * manconf_parse() wants to do it.
465 */
466 if (argc > 0) {
467 conf.manpath.paths = mandoc_reallocarray(NULL,
468 argc, sizeof(char *));
469 conf.manpath.sz = (size_t)argc;
470 for (i = 0; i < argc; i++)
471 conf.manpath.paths[i] = mandoc_strdup(argv[i]);
472 } else
473 manconf_parse(&conf, path_arg, NULL, NULL);
474
475 if (conf.manpath.sz == 0) {
476 exitcode = (int)MANDOCLEVEL_BADARG;
477 say("", "Empty manpath");
478 }
479
480 /*
481 * First scan the tree rooted at a base directory, then
482 * build a new database and finally move it into place.
483 * Ignore zero-length directories and strip trailing
484 * slashes.
485 */
486 for (j = 0; j < conf.manpath.sz; j++) {
487 sz = strlen(conf.manpath.paths[j]);
488 if (sz && conf.manpath.paths[j][sz - 1] == '/')
489 conf.manpath.paths[j][--sz] = '\0';
490 if (0 == sz)
491 continue;
492
493 if (j) {
494 mandoc_ohash_init(&mpages, 6,
495 offsetof(struct mpage, inodev));
496 mandoc_ohash_init(&mlinks, 6,
497 offsetof(struct mlink, file));
498 }
499
500 if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
501 continue;
502 if (0 == treescan())
503 continue;
504 if (0 == dbopen(0))
505 continue;
506
507 mpages_merge(mp);
508 if (warnings && !nodb &&
509 ! (MPARSE_QUICK & mparse_options))
510 names_check();
511 dbclose(0);
512
513 if (j + 1 < conf.manpath.sz) {
514 mpages_free();
515 ohash_delete(&mpages);
516 ohash_delete(&mlinks);
517 }
518 }
519 }
520 out:
521 manconf_free(&conf);
522 mparse_free(mp);
523 mchars_free();
524 mpages_free();
525 ohash_delete(&mpages);
526 ohash_delete(&mlinks);
527 return exitcode;
528 usage:
529 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
530 " %s [-aDnpQ] [-Tutf8] dir ...\n"
531 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
532 " %s [-Dnp] -u dir [file ...]\n"
533 " %s [-Q] -t file ...\n",
534 __progname, __progname, __progname,
535 __progname, __progname);
536
537 return (int)MANDOCLEVEL_BADARG;
538 }
539
540 /*
541 * Scan a directory tree rooted at "basedir" for manpages.
542 * We use fts(), scanning directory parts along the way for clues to our
543 * section and architecture.
544 *
545 * If use_all has been specified, grok all files.
546 * If not, sanitise paths to the following:
547 *
548 * [./]man*[/<arch>]/<name>.<section>
549 * or
550 * [./]cat<section>[/<arch>]/<name>.0
551 *
552 * TODO: accomodate for multi-language directories.
553 */
554 static int
555 treescan(void)
556 {
557 char buf[PATH_MAX];
558 FTS *f;
559 FTSENT *ff;
560 struct mlink *mlink;
561 int dform, gzip;
562 char *dsec, *arch, *fsec, *cp;
563 const char *path;
564 const char *argv[2];
565
566 argv[0] = ".";
567 argv[1] = (char *)NULL;
568
569 f = fts_open((char * const *)argv,
570 FTS_PHYSICAL | FTS_NOCHDIR, NULL);
571 if (f == NULL) {
572 exitcode = (int)MANDOCLEVEL_SYSERR;
573 say("", "&fts_open");
574 return 0;
575 }
576
577 dsec = arch = NULL;
578 dform = FORM_NONE;
579
580 while ((ff = fts_read(f)) != NULL) {
581 path = ff->fts_path + 2;
582 switch (ff->fts_info) {
583
584 /*
585 * Symbolic links require various sanity checks,
586 * then get handled just like regular files.
587 */
588 case FTS_SL:
589 if (realpath(path, buf) == NULL) {
590 if (warnings)
591 say(path, "&realpath");
592 continue;
593 }
594 if (strstr(buf, basedir) != buf
595 #ifdef HOMEBREWDIR
596 && strstr(buf, HOMEBREWDIR) != buf
597 #endif
598 ) {
599 if (warnings) say("",
600 "%s: outside base directory", buf);
601 continue;
602 }
603 /* Use logical inode to avoid mpages dupe. */
604 if (stat(path, ff->fts_statp) == -1) {
605 if (warnings)
606 say(path, "&stat");
607 continue;
608 }
609 /* FALLTHROUGH */
610
611 /*
612 * If we're a regular file, add an mlink by using the
613 * stored directory data and handling the filename.
614 */
615 case FTS_F:
616 if ( ! strcmp(path, MANDOC_DB))
617 continue;
618 if ( ! use_all && ff->fts_level < 2) {
619 if (warnings)
620 say(path, "Extraneous file");
621 continue;
622 }
623 gzip = 0;
624 fsec = NULL;
625 while (fsec == NULL) {
626 fsec = strrchr(ff->fts_name, '.');
627 if (fsec == NULL || strcmp(fsec+1, "gz"))
628 break;
629 gzip = 1;
630 *fsec = '\0';
631 fsec = NULL;
632 }
633 if (fsec == NULL) {
634 if ( ! use_all) {
635 if (warnings)
636 say(path,
637 "No filename suffix");
638 continue;
639 }
640 } else if ( ! strcmp(++fsec, "html")) {
641 if (warnings)
642 say(path, "Skip html");
643 continue;
644 } else if ( ! strcmp(fsec, "ps")) {
645 if (warnings)
646 say(path, "Skip ps");
647 continue;
648 } else if ( ! strcmp(fsec, "pdf")) {
649 if (warnings)
650 say(path, "Skip pdf");
651 continue;
652 } else if ( ! use_all &&
653 ((dform == FORM_SRC &&
654 strncmp(fsec, dsec, strlen(dsec))) ||
655 (dform == FORM_CAT && strcmp(fsec, "0")))) {
656 if (warnings)
657 say(path, "Wrong filename suffix");
658 continue;
659 } else
660 fsec[-1] = '\0';
661
662 mlink = mandoc_calloc(1, sizeof(struct mlink));
663 if (strlcpy(mlink->file, path,
664 sizeof(mlink->file)) >=
665 sizeof(mlink->file)) {
666 say(path, "Filename too long");
667 free(mlink);
668 continue;
669 }
670 mlink->dform = dform;
671 mlink->dsec = dsec;
672 mlink->arch = arch;
673 mlink->name = ff->fts_name;
674 mlink->fsec = fsec;
675 mlink->gzip = gzip;
676 mlink_add(mlink, ff->fts_statp);
677 continue;
678
679 case FTS_D:
680 case FTS_DP:
681 break;
682
683 default:
684 if (warnings)
685 say(path, "Not a regular file");
686 continue;
687 }
688
689 switch (ff->fts_level) {
690 case 0:
691 /* Ignore the root directory. */
692 break;
693 case 1:
694 /*
695 * This might contain manX/ or catX/.
696 * Try to infer this from the name.
697 * If we're not in use_all, enforce it.
698 */
699 cp = ff->fts_name;
700 if (ff->fts_info == FTS_DP) {
701 dform = FORM_NONE;
702 dsec = NULL;
703 break;
704 }
705
706 if ( ! strncmp(cp, "man", 3)) {
707 dform = FORM_SRC;
708 dsec = cp + 3;
709 } else if ( ! strncmp(cp, "cat", 3)) {
710 dform = FORM_CAT;
711 dsec = cp + 3;
712 } else {
713 dform = FORM_NONE;
714 dsec = NULL;
715 }
716
717 if (dsec != NULL || use_all)
718 break;
719
720 if (warnings)
721 say(path, "Unknown directory part");
722 fts_set(f, ff, FTS_SKIP);
723 break;
724 case 2:
725 /*
726 * Possibly our architecture.
727 * If we're descending, keep tabs on it.
728 */
729 if (ff->fts_info != FTS_DP && dsec != NULL)
730 arch = ff->fts_name;
731 else
732 arch = NULL;
733 break;
734 default:
735 if (ff->fts_info == FTS_DP || use_all)
736 break;
737 if (warnings)
738 say(path, "Extraneous directory part");
739 fts_set(f, ff, FTS_SKIP);
740 break;
741 }
742 }
743
744 fts_close(f);
745 return 1;
746 }
747
748 /*
749 * Add a file to the mlinks table.
750 * Do not verify that it's a "valid" looking manpage (we'll do that
751 * later).
752 *
753 * Try to infer the manual section, architecture, and page name from the
754 * path, assuming it looks like
755 *
756 * [./]man*[/<arch>]/<name>.<section>
757 * or
758 * [./]cat<section>[/<arch>]/<name>.0
759 *
760 * See treescan() for the fts(3) version of this.
761 */
762 static void
763 filescan(const char *file)
764 {
765 char buf[PATH_MAX];
766 struct stat st;
767 struct mlink *mlink;
768 char *p, *start;
769
770 assert(use_all);
771
772 if (0 == strncmp(file, "./", 2))
773 file += 2;
774
775 /*
776 * We have to do lstat(2) before realpath(3) loses
777 * the information whether this is a symbolic link.
778 * We need to know that because for symbolic links,
779 * we want to use the orginal file name, while for
780 * regular files, we want to use the real path.
781 */
782 if (-1 == lstat(file, &st)) {
783 exitcode = (int)MANDOCLEVEL_BADARG;
784 say(file, "&lstat");
785 return;
786 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
787 exitcode = (int)MANDOCLEVEL_BADARG;
788 say(file, "Not a regular file");
789 return;
790 }
791
792 /*
793 * We have to resolve the file name to the real path
794 * in any case for the base directory check.
795 */
796 if (NULL == realpath(file, buf)) {
797 exitcode = (int)MANDOCLEVEL_BADARG;
798 say(file, "&realpath");
799 return;
800 }
801
802 if (OP_TEST == op)
803 start = buf;
804 else if (strstr(buf, basedir) == buf)
805 start = buf + strlen(basedir);
806 #ifdef HOMEBREWDIR
807 else if (strstr(buf, HOMEBREWDIR) == buf)
808 start = buf;
809 #endif
810 else {
811 exitcode = (int)MANDOCLEVEL_BADARG;
812 say("", "%s: outside base directory", buf);
813 return;
814 }
815
816 /*
817 * Now we are sure the file is inside our tree.
818 * If it is a symbolic link, ignore the real path
819 * and use the original name.
820 * This implies passing stuff like "cat1/../man1/foo.1"
821 * on the command line won't work. So don't do that.
822 * Note the stat(2) can still fail if the link target
823 * doesn't exist.
824 */
825 if (S_IFLNK & st.st_mode) {
826 if (-1 == stat(buf, &st)) {
827 exitcode = (int)MANDOCLEVEL_BADARG;
828 say(file, "&stat");
829 return;
830 }
831 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
832 say(file, "Filename too long");
833 return;
834 }
835 start = buf;
836 if (OP_TEST != op && strstr(buf, basedir) == buf)
837 start += strlen(basedir);
838 }
839
840 mlink = mandoc_calloc(1, sizeof(struct mlink));
841 mlink->dform = FORM_NONE;
842 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
843 sizeof(mlink->file)) {
844 say(start, "Filename too long");
845 free(mlink);
846 return;
847 }
848
849 /*
850 * First try to guess our directory structure.
851 * If we find a separator, try to look for man* or cat*.
852 * If we find one of these and what's underneath is a directory,
853 * assume it's an architecture.
854 */
855 if (NULL != (p = strchr(start, '/'))) {
856 *p++ = '\0';
857 if (0 == strncmp(start, "man", 3)) {
858 mlink->dform = FORM_SRC;
859 mlink->dsec = start + 3;
860 } else if (0 == strncmp(start, "cat", 3)) {
861 mlink->dform = FORM_CAT;
862 mlink->dsec = start + 3;
863 }
864
865 start = p;
866 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
867 *p++ = '\0';
868 mlink->arch = start;
869 start = p;
870 }
871 }
872
873 /*
874 * Now check the file suffix.
875 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
876 */
877 p = strrchr(start, '\0');
878 while (p-- > start && '/' != *p && '.' != *p)
879 /* Loop. */ ;
880
881 if ('.' == *p) {
882 *p++ = '\0';
883 mlink->fsec = p;
884 }
885
886 /*
887 * Now try to parse the name.
888 * Use the filename portion of the path.
889 */
890 mlink->name = start;
891 if (NULL != (p = strrchr(start, '/'))) {
892 mlink->name = p + 1;
893 *p = '\0';
894 }
895 mlink_add(mlink, &st);
896 }
897
898 static void
899 mlink_add(struct mlink *mlink, const struct stat *st)
900 {
901 struct inodev inodev;
902 struct mpage *mpage;
903 unsigned int slot;
904
905 assert(NULL != mlink->file);
906
907 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
908 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
909 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
910 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
911
912 if ('0' == *mlink->fsec) {
913 free(mlink->fsec);
914 mlink->fsec = mandoc_strdup(mlink->dsec);
915 mlink->fform = FORM_CAT;
916 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
917 mlink->fform = FORM_SRC;
918 else
919 mlink->fform = FORM_NONE;
920
921 slot = ohash_qlookup(&mlinks, mlink->file);
922 assert(NULL == ohash_find(&mlinks, slot));
923 ohash_insert(&mlinks, slot, mlink);
924
925 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
926 inodev.st_ino = st->st_ino;
927 inodev.st_dev = st->st_dev;
928 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
929 sizeof(struct inodev), inodev.st_ino);
930 mpage = ohash_find(&mpages, slot);
931 if (NULL == mpage) {
932 mpage = mandoc_calloc(1, sizeof(struct mpage));
933 mpage->inodev.st_ino = inodev.st_ino;
934 mpage->inodev.st_dev = inodev.st_dev;
935 ohash_insert(&mpages, slot, mpage);
936 } else
937 mlink->next = mpage->mlinks;
938 mpage->mlinks = mlink;
939 mlink->mpage = mpage;
940 }
941
942 static void
943 mlink_free(struct mlink *mlink)
944 {
945
946 free(mlink->dsec);
947 free(mlink->arch);
948 free(mlink->name);
949 free(mlink->fsec);
950 free(mlink);
951 }
952
953 static void
954 mpages_free(void)
955 {
956 struct mpage *mpage;
957 struct mlink *mlink;
958 unsigned int slot;
959
960 mpage = ohash_first(&mpages, &slot);
961 while (NULL != mpage) {
962 while (NULL != (mlink = mpage->mlinks)) {
963 mpage->mlinks = mlink->next;
964 mlink_free(mlink);
965 }
966 free(mpage->sec);
967 free(mpage->arch);
968 free(mpage->title);
969 free(mpage->desc);
970 free(mpage);
971 mpage = ohash_next(&mpages, &slot);
972 }
973 }
974
975 /*
976 * For each mlink to the mpage, check whether the path looks like
977 * it is formatted, and if it does, check whether a source manual
978 * exists by the same name, ignoring the suffix.
979 * If both conditions hold, drop the mlink.
980 */
981 static void
982 mlinks_undupe(struct mpage *mpage)
983 {
984 char buf[PATH_MAX];
985 struct mlink **prev;
986 struct mlink *mlink;
987 char *bufp;
988
989 mpage->form = FORM_CAT;
990 prev = &mpage->mlinks;
991 while (NULL != (mlink = *prev)) {
992 if (FORM_CAT != mlink->dform) {
993 mpage->form = FORM_NONE;
994 goto nextlink;
995 }
996 (void)strlcpy(buf, mlink->file, sizeof(buf));
997 bufp = strstr(buf, "cat");
998 assert(NULL != bufp);
999 memcpy(bufp, "man", 3);
1000 if (NULL != (bufp = strrchr(buf, '.')))
1001 *++bufp = '\0';
1002 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1003 if (NULL == ohash_find(&mlinks,
1004 ohash_qlookup(&mlinks, buf)))
1005 goto nextlink;
1006 if (warnings)
1007 say(mlink->file, "Man source exists: %s", buf);
1008 if (use_all)
1009 goto nextlink;
1010 *prev = mlink->next;
1011 mlink_free(mlink);
1012 continue;
1013 nextlink:
1014 prev = &(*prev)->next;
1015 }
1016 }
1017
1018 static void
1019 mlink_check(struct mpage *mpage, struct mlink *mlink)
1020 {
1021 struct str *str;
1022 unsigned int slot;
1023
1024 /*
1025 * Check whether the manual section given in a file
1026 * agrees with the directory where the file is located.
1027 * Some manuals have suffixes like (3p) on their
1028 * section number either inside the file or in the
1029 * directory name, some are linked into more than one
1030 * section, like encrypt(1) = makekey(8).
1031 */
1032
1033 if (FORM_SRC == mpage->form &&
1034 strcasecmp(mpage->sec, mlink->dsec))
1035 say(mlink->file, "Section \"%s\" manual in %s directory",
1036 mpage->sec, mlink->dsec);
1037
1038 /*
1039 * Manual page directories exist for each kernel
1040 * architecture as returned by machine(1).
1041 * However, many manuals only depend on the
1042 * application architecture as returned by arch(1).
1043 * For example, some (2/ARM) manuals are shared
1044 * across the "armish" and "zaurus" kernel
1045 * architectures.
1046 * A few manuals are even shared across completely
1047 * different architectures, for example fdformat(1)
1048 * on amd64, i386, sparc, and sparc64.
1049 */
1050
1051 if (strcasecmp(mpage->arch, mlink->arch))
1052 say(mlink->file, "Architecture \"%s\" manual in "
1053 "\"%s\" directory", mpage->arch, mlink->arch);
1054
1055 /*
1056 * XXX
1057 * parse_cat() doesn't set NAME_TITLE yet.
1058 */
1059
1060 if (FORM_CAT == mpage->form)
1061 return;
1062
1063 /*
1064 * Check whether this mlink
1065 * appears as a name in the NAME section.
1066 */
1067
1068 slot = ohash_qlookup(&names, mlink->name);
1069 str = ohash_find(&names, slot);
1070 assert(NULL != str);
1071 if ( ! (NAME_TITLE & str->mask))
1072 say(mlink->file, "Name missing in NAME section");
1073 }
1074
1075 /*
1076 * Run through the files in the global vector "mpages"
1077 * and add them to the database specified in "basedir".
1078 *
1079 * This handles the parsing scheme itself, using the cues of directory
1080 * and filename to determine whether the file is parsable or not.
1081 */
1082 static void
1083 mpages_merge(struct mparse *mp)
1084 {
1085 char any[] = "any";
1086 struct mpage *mpage, *mpage_dest;
1087 struct mlink *mlink, *mlink_dest;
1088 struct roff_man *man;
1089 char *sodest;
1090 char *cp;
1091 int fd;
1092 unsigned int pslot;
1093
1094 if ( ! nodb)
1095 SQL_EXEC("BEGIN TRANSACTION");
1096
1097 mpage = ohash_first(&mpages, &pslot);
1098 while (mpage != NULL) {
1099 mlinks_undupe(mpage);
1100 if ((mlink = mpage->mlinks) == NULL) {
1101 mpage = ohash_next(&mpages, &pslot);
1102 continue;
1103 }
1104
1105 name_mask = NAME_MASK;
1106 mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1107 mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1108 mparse_reset(mp);
1109 man = NULL;
1110 sodest = NULL;
1111
1112 mparse_open(mp, &fd, mlink->file);
1113 if (fd == -1) {
1114 say(mlink->file, "&open");
1115 goto nextpage;
1116 }
1117
1118 /*
1119 * Interpret the file as mdoc(7) or man(7) source
1120 * code, unless it is known to be formatted.
1121 */
1122 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1123 mparse_readfd(mp, fd, mlink->file);
1124 mparse_result(mp, &man, &sodest);
1125 }
1126
1127 if (sodest != NULL) {
1128 mlink_dest = ohash_find(&mlinks,
1129 ohash_qlookup(&mlinks, sodest));
1130 if (mlink_dest == NULL) {
1131 mandoc_asprintf(&cp, "%s.gz", sodest);
1132 mlink_dest = ohash_find(&mlinks,
1133 ohash_qlookup(&mlinks, cp));
1134 free(cp);
1135 }
1136 if (mlink_dest != NULL) {
1137
1138 /* The .so target exists. */
1139
1140 mpage_dest = mlink_dest->mpage;
1141 while (1) {
1142 mlink->mpage = mpage_dest;
1143
1144 /*
1145 * If the target was already
1146 * processed, add the links
1147 * to the database now.
1148 * Otherwise, this will
1149 * happen when we come
1150 * to the target.
1151 */
1152
1153 if (mpage_dest->pageid)
1154 dbadd_mlink_name(mlink);
1155
1156 if (mlink->next == NULL)
1157 break;
1158 mlink = mlink->next;
1159 }
1160
1161 /* Move all links to the target. */
1162
1163 mlink->next = mlink_dest->next;
1164 mlink_dest->next = mpage->mlinks;
1165 mpage->mlinks = NULL;
1166 }
1167 goto nextpage;
1168 } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1169 mdoc_validate(man);
1170 mpage->form = FORM_SRC;
1171 mpage->sec = man->meta.msec;
1172 mpage->sec = mandoc_strdup(
1173 mpage->sec == NULL ? "" : mpage->sec);
1174 mpage->arch = man->meta.arch;
1175 mpage->arch = mandoc_strdup(
1176 mpage->arch == NULL ? "" : mpage->arch);
1177 mpage->title = mandoc_strdup(man->meta.title);
1178 } else if (man != NULL && man->macroset == MACROSET_MAN) {
1179 mpage->form = FORM_SRC;
1180 mpage->sec = mandoc_strdup(man->meta.msec);
1181 mpage->arch = mandoc_strdup(mlink->arch);
1182 mpage->title = mandoc_strdup(man->meta.title);
1183 } else {
1184 mpage->form = FORM_CAT;
1185 mpage->sec = mandoc_strdup(mlink->dsec);
1186 mpage->arch = mandoc_strdup(mlink->arch);
1187 mpage->title = mandoc_strdup(mlink->name);
1188 }
1189 putkey(mpage, mpage->sec, TYPE_sec);
1190 if (*mpage->arch != '\0')
1191 putkey(mpage, mpage->arch, TYPE_arch);
1192
1193 for ( ; mlink != NULL; mlink = mlink->next) {
1194 if ('\0' != *mlink->dsec)
1195 putkey(mpage, mlink->dsec, TYPE_sec);
1196 if ('\0' != *mlink->fsec)
1197 putkey(mpage, mlink->fsec, TYPE_sec);
1198 putkey(mpage, '\0' == *mlink->arch ?
1199 any : mlink->arch, TYPE_arch);
1200 putkey(mpage, mlink->name, NAME_FILE);
1201 }
1202
1203 assert(mpage->desc == NULL);
1204 if (man != NULL && man->macroset == MACROSET_MDOC)
1205 parse_mdoc(mpage, &man->meta, man->first);
1206 else if (man != NULL)
1207 parse_man(mpage, &man->meta, man->first);
1208 else
1209 parse_cat(mpage, fd);
1210 if (mpage->desc == NULL)
1211 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1212
1213 if (warnings && !use_all)
1214 for (mlink = mpage->mlinks; mlink;
1215 mlink = mlink->next)
1216 mlink_check(mpage, mlink);
1217
1218 dbadd(mpage);
1219 mlink = mpage->mlinks;
1220
1221 nextpage:
1222 ohash_delete(&strings);
1223 ohash_delete(&names);
1224 mpage = ohash_next(&mpages, &pslot);
1225 }
1226
1227 if (0 == nodb)
1228 SQL_EXEC("END TRANSACTION");
1229 }
1230
1231 static void
1232 names_check(void)
1233 {
1234 sqlite3_stmt *stmt;
1235 const char *name, *sec, *arch, *key;
1236
1237 sqlite3_prepare_v2(db,
1238 "SELECT name, sec, arch, key FROM ("
1239 "SELECT name AS key, pageid FROM names "
1240 "WHERE bits & ? AND NOT EXISTS ("
1241 "SELECT pageid FROM mlinks "
1242 "WHERE mlinks.pageid == names.pageid "
1243 "AND mlinks.name == names.name"
1244 ")"
1245 ") JOIN ("
1246 "SELECT sec, arch, name, pageid FROM mlinks "
1247 "GROUP BY pageid"
1248 ") USING (pageid);",
1249 -1, &stmt, NULL);
1250
1251 if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK)
1252 say("", "%s", sqlite3_errmsg(db));
1253
1254 while (sqlite3_step(stmt) == SQLITE_ROW) {
1255 name = (const char *)sqlite3_column_text(stmt, 0);
1256 sec = (const char *)sqlite3_column_text(stmt, 1);
1257 arch = (const char *)sqlite3_column_text(stmt, 2);
1258 key = (const char *)sqlite3_column_text(stmt, 3);
1259 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1260 '\0' == *arch ? "" : "/",
1261 '\0' == *arch ? "" : arch, key);
1262 }
1263 sqlite3_finalize(stmt);
1264 }
1265
1266 static void
1267 parse_cat(struct mpage *mpage, int fd)
1268 {
1269 FILE *stream;
1270 char *line, *p, *title;
1271 size_t len, plen, titlesz;
1272
1273 stream = (-1 == fd) ?
1274 fopen(mpage->mlinks->file, "r") :
1275 fdopen(fd, "r");
1276 if (NULL == stream) {
1277 if (-1 != fd)
1278 close(fd);
1279 if (warnings)
1280 say(mpage->mlinks->file, "&fopen");
1281 return;
1282 }
1283
1284 /* Skip to first blank line. */
1285
1286 while (NULL != (line = fgetln(stream, &len)))
1287 if ('\n' == *line)
1288 break;
1289
1290 /*
1291 * Assume the first line that is not indented
1292 * is the first section header. Skip to it.
1293 */
1294
1295 while (NULL != (line = fgetln(stream, &len)))
1296 if ('\n' != *line && ' ' != *line)
1297 break;
1298
1299 /*
1300 * Read up until the next section into a buffer.
1301 * Strip the leading and trailing newline from each read line,
1302 * appending a trailing space.
1303 * Ignore empty (whitespace-only) lines.
1304 */
1305
1306 titlesz = 0;
1307 title = NULL;
1308
1309 while (NULL != (line = fgetln(stream, &len))) {
1310 if (' ' != *line || '\n' != line[len - 1])
1311 break;
1312 while (len > 0 && isspace((unsigned char)*line)) {
1313 line++;
1314 len--;
1315 }
1316 if (1 == len)
1317 continue;
1318 title = mandoc_realloc(title, titlesz + len);
1319 memcpy(title + titlesz, line, len);
1320 titlesz += len;
1321 title[titlesz - 1] = ' ';
1322 }
1323
1324 /*
1325 * If no page content can be found, or the input line
1326 * is already the next section header, or there is no
1327 * trailing newline, reuse the page title as the page
1328 * description.
1329 */
1330
1331 if (NULL == title || '\0' == *title) {
1332 if (warnings)
1333 say(mpage->mlinks->file,
1334 "Cannot find NAME section");
1335 fclose(stream);
1336 free(title);
1337 return;
1338 }
1339
1340 title = mandoc_realloc(title, titlesz + 1);
1341 title[titlesz] = '\0';
1342
1343 /*
1344 * Skip to the first dash.
1345 * Use the remaining line as the description (no more than 70
1346 * bytes).
1347 */
1348
1349 if (NULL != (p = strstr(title, "- "))) {
1350 for (p += 2; ' ' == *p || '\b' == *p; p++)
1351 /* Skip to next word. */ ;
1352 } else {
1353 if (warnings)
1354 say(mpage->mlinks->file,
1355 "No dash in title line");
1356 p = title;
1357 }
1358
1359 plen = strlen(p);
1360
1361 /* Strip backspace-encoding from line. */
1362
1363 while (NULL != (line = memchr(p, '\b', plen))) {
1364 len = line - p;
1365 if (0 == len) {
1366 memmove(line, line + 1, plen--);
1367 continue;
1368 }
1369 memmove(line - 1, line + 1, plen - len);
1370 plen -= 2;
1371 }
1372
1373 mpage->desc = mandoc_strdup(p);
1374 fclose(stream);
1375 free(title);
1376 }
1377
1378 /*
1379 * Put a type/word pair into the word database for this particular file.
1380 */
1381 static void
1382 putkey(const struct mpage *mpage, char *value, uint64_t type)
1383 {
1384 char *cp;
1385
1386 assert(NULL != value);
1387 if (TYPE_arch == type)
1388 for (cp = value; *cp; cp++)
1389 if (isupper((unsigned char)*cp))
1390 *cp = _tolower((unsigned char)*cp);
1391 putkeys(mpage, value, strlen(value), type);
1392 }
1393
1394 /*
1395 * Grok all nodes at or below a certain mdoc node into putkey().
1396 */
1397 static void
1398 putmdockey(const struct mpage *mpage,
1399 const struct roff_node *n, uint64_t m)
1400 {
1401
1402 for ( ; NULL != n; n = n->next) {
1403 if (NULL != n->child)
1404 putmdockey(mpage, n->child, m);
1405 if (n->type == ROFFT_TEXT)
1406 putkey(mpage, n->string, m);
1407 }
1408 }
1409
1410 static void
1411 parse_man(struct mpage *mpage, const struct roff_meta *meta,
1412 const struct roff_node *n)
1413 {
1414 const struct roff_node *head, *body;
1415 char *start, *title;
1416 char byte;
1417 size_t sz;
1418
1419 if (NULL == n)
1420 return;
1421
1422 /*
1423 * We're only searching for one thing: the first text child in
1424 * the BODY of a NAME section. Since we don't keep track of
1425 * sections in -man, run some hoops to find out whether we're in
1426 * the correct section or not.
1427 */
1428
1429 if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1430 body = n;
1431 assert(body->parent);
1432 if (NULL != (head = body->parent->head) &&
1433 1 == head->nchild &&
1434 NULL != (head = (head->child)) &&
1435 head->type == ROFFT_TEXT &&
1436 0 == strcmp(head->string, "NAME") &&
1437 NULL != body->child) {
1438
1439 /*
1440 * Suck the entire NAME section into memory.
1441 * Yes, we might run away.
1442 * But too many manuals have big, spread-out
1443 * NAME sections over many lines.
1444 */
1445
1446 title = NULL;
1447 deroff(&title, body);
1448 if (NULL == title)
1449 return;
1450
1451 /*
1452 * Go through a special heuristic dance here.
1453 * Conventionally, one or more manual names are
1454 * comma-specified prior to a whitespace, then a
1455 * dash, then a description. Try to puzzle out
1456 * the name parts here.
1457 */
1458
1459 start = title;
1460 for ( ;; ) {
1461 sz = strcspn(start, " ,");
1462 if ('\0' == start[sz])
1463 break;
1464
1465 byte = start[sz];
1466 start[sz] = '\0';
1467
1468 /*
1469 * Assume a stray trailing comma in the
1470 * name list if a name begins with a dash.
1471 */
1472
1473 if ('-' == start[0] ||
1474 ('\\' == start[0] && '-' == start[1]))
1475 break;
1476
1477 putkey(mpage, start, NAME_TITLE);
1478 if ( ! (mpage->name_head_done ||
1479 strcasecmp(start, meta->title))) {
1480 putkey(mpage, start, NAME_HEAD);
1481 mpage->name_head_done = 1;
1482 }
1483
1484 if (' ' == byte) {
1485 start += sz + 1;
1486 break;
1487 }
1488
1489 assert(',' == byte);
1490 start += sz + 1;
1491 while (' ' == *start)
1492 start++;
1493 }
1494
1495 if (start == title) {
1496 putkey(mpage, start, NAME_TITLE);
1497 if ( ! (mpage->name_head_done ||
1498 strcasecmp(start, meta->title))) {
1499 putkey(mpage, start, NAME_HEAD);
1500 mpage->name_head_done = 1;
1501 }
1502 free(title);
1503 return;
1504 }
1505
1506 while (isspace((unsigned char)*start))
1507 start++;
1508
1509 if (0 == strncmp(start, "-", 1))
1510 start += 1;
1511 else if (0 == strncmp(start, "\\-\\-", 4))
1512 start += 4;
1513 else if (0 == strncmp(start, "\\-", 2))
1514 start += 2;
1515 else if (0 == strncmp(start, "\\(en", 4))
1516 start += 4;
1517 else if (0 == strncmp(start, "\\(em", 4))
1518 start += 4;
1519
1520 while (' ' == *start)
1521 start++;
1522
1523 mpage->desc = mandoc_strdup(start);
1524 free(title);
1525 return;
1526 }
1527 }
1528
1529 for (n = n->child; n; n = n->next) {
1530 if (NULL != mpage->desc)
1531 break;
1532 parse_man(mpage, meta, n);
1533 }
1534 }
1535
1536 static void
1537 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1538 const struct roff_node *n)
1539 {
1540
1541 assert(NULL != n);
1542 for (n = n->child; NULL != n; n = n->next) {
1543 switch (n->type) {
1544 case ROFFT_ELEM:
1545 case ROFFT_BLOCK:
1546 case ROFFT_HEAD:
1547 case ROFFT_BODY:
1548 case ROFFT_TAIL:
1549 if (NULL != mdocs[n->tok].fp)
1550 if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1551 break;
1552 if (mdocs[n->tok].mask)
1553 putmdockey(mpage, n->child,
1554 mdocs[n->tok].mask);
1555 break;
1556 default:
1557 assert(n->type != ROFFT_ROOT);
1558 continue;
1559 }
1560 if (NULL != n->child)
1561 parse_mdoc(mpage, meta, n);
1562 }
1563 }
1564
1565 static int
1566 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1567 const struct roff_node *n)
1568 {
1569 char *start, *end;
1570 size_t sz;
1571
1572 if (SEC_SYNOPSIS != n->sec ||
1573 NULL == (n = n->child) ||
1574 n->type != ROFFT_TEXT)
1575 return 0;
1576
1577 /*
1578 * Only consider those `Fd' macro fields that begin with an
1579 * "inclusion" token (versus, e.g., #define).
1580 */
1581
1582 if (strcmp("#include", n->string))
1583 return 0;
1584
1585 if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1586 return 0;
1587
1588 /*
1589 * Strip away the enclosing angle brackets and make sure we're
1590 * not zero-length.
1591 */
1592
1593 start = n->string;
1594 if ('<' == *start || '"' == *start)
1595 start++;
1596
1597 if (0 == (sz = strlen(start)))
1598 return 0;
1599
1600 end = &start[(int)sz - 1];
1601 if ('>' == *end || '"' == *end)
1602 end--;
1603
1604 if (end > start)
1605 putkeys(mpage, start, end - start + 1, TYPE_In);
1606 return 0;
1607 }
1608
1609 static void
1610 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1611 {
1612 char *cp;
1613 size_t sz;
1614
1615 if (n->type != ROFFT_TEXT)
1616 return;
1617
1618 /* Skip function pointer punctuation. */
1619
1620 cp = n->string;
1621 while (*cp == '(' || *cp == '*')
1622 cp++;
1623 sz = strcspn(cp, "()");
1624
1625 putkeys(mpage, cp, sz, TYPE_Fn);
1626 if (n->sec == SEC_SYNOPSIS)
1627 putkeys(mpage, cp, sz, NAME_SYN);
1628 }
1629
1630 static int
1631 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1632 const struct roff_node *n)
1633 {
1634
1635 if (n->child == NULL)
1636 return 0;
1637
1638 parse_mdoc_fname(mpage, n->child);
1639
1640 for (n = n->child->next; n != NULL; n = n->next)
1641 if (n->type == ROFFT_TEXT)
1642 putkey(mpage, n->string, TYPE_Fa);
1643
1644 return 0;
1645 }
1646
1647 static int
1648 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1649 const struct roff_node *n)
1650 {
1651
1652 if (n->type != ROFFT_HEAD)
1653 return 1;
1654
1655 if (n->child != NULL)
1656 parse_mdoc_fname(mpage, n->child);
1657
1658 return 0;
1659 }
1660
1661 static int
1662 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1663 const struct roff_node *n)
1664 {
1665 char *cp;
1666
1667 if (NULL == (n = n->child))
1668 return 0;
1669
1670 if (NULL == n->next) {
1671 putkey(mpage, n->string, TYPE_Xr);
1672 return 0;
1673 }
1674
1675 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1676 putkey(mpage, cp, TYPE_Xr);
1677 free(cp);
1678 return 0;
1679 }
1680
1681 static int
1682 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1683 const struct roff_node *n)
1684 {
1685
1686 if (n->type == ROFFT_BODY)
1687 deroff(&mpage->desc, n);
1688 return 0;
1689 }
1690
1691 static int
1692 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1693 const struct roff_node *n)
1694 {
1695
1696 if (SEC_NAME == n->sec)
1697 putmdockey(mpage, n->child, NAME_TITLE);
1698 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1699 if (n->child == NULL)
1700 putkey(mpage, meta->name, NAME_SYN);
1701 else
1702 putmdockey(mpage, n->child, NAME_SYN);
1703 }
1704 if ( ! (mpage->name_head_done ||
1705 n->child == NULL || n->child->string == NULL ||
1706 strcasecmp(n->child->string, meta->title))) {
1707 putkey(mpage, n->child->string, ROFFT_HEAD);
1708 mpage->name_head_done = 1;
1709 }
1710 return 0;
1711 }
1712
1713 static int
1714 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1715 const struct roff_node *n)
1716 {
1717
1718 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1719 }
1720
1721 static int
1722 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1723 const struct roff_node *n)
1724 {
1725
1726 return n->type == ROFFT_HEAD;
1727 }
1728
1729 static int
1730 parse_mdoc_body(struct mpage *mpage, const struct roff_meta *meta,
1731 const struct roff_node *n)
1732 {
1733
1734 return n->type == ROFFT_BODY;
1735 }
1736
1737 /*
1738 * Add a string to the hash table for the current manual.
1739 * Each string has a bitmask telling which macros it belongs to.
1740 * When we finish the manual, we'll dump the table.
1741 */
1742 static void
1743 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1744 {
1745 struct ohash *htab;
1746 struct str *s;
1747 const char *end;
1748 unsigned int slot;
1749 int i, mustfree;
1750
1751 if (0 == sz)
1752 return;
1753
1754 mustfree = render_string(&cp, &sz);
1755
1756 if (TYPE_Nm & v) {
1757 htab = &names;
1758 v &= name_mask;
1759 if (v & NAME_FIRST)
1760 name_mask &= ~NAME_FIRST;
1761 if (debug > 1)
1762 say(mpage->mlinks->file,
1763 "Adding name %*s, bits=%d", sz, cp, v);
1764 } else {
1765 htab = &strings;
1766 if (debug > 1)
1767 for (i = 0; i < mansearch_keymax; i++)
1768 if ((uint64_t)1 << i & v)
1769 say(mpage->mlinks->file,
1770 "Adding key %s=%*s",
1771 mansearch_keynames[i], sz, cp);
1772 }
1773
1774 end = cp + sz;
1775 slot = ohash_qlookupi(htab, cp, &end);
1776 s = ohash_find(htab, slot);
1777
1778 if (NULL != s && mpage == s->mpage) {
1779 s->mask |= v;
1780 return;
1781 } else if (NULL == s) {
1782 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1783 memcpy(s->key, cp, sz);
1784 ohash_insert(htab, slot, s);
1785 }
1786 s->mpage = mpage;
1787 s->mask = v;
1788
1789 if (mustfree)
1790 free(cp);
1791 }
1792
1793 /*
1794 * Take a Unicode codepoint and produce its UTF-8 encoding.
1795 * This isn't the best way to do this, but it works.
1796 * The magic numbers are from the UTF-8 packaging.
1797 * They're not as scary as they seem: read the UTF-8 spec for details.
1798 */
1799 static size_t
1800 utf8(unsigned int cp, char out[7])
1801 {
1802 size_t rc;
1803
1804 rc = 0;
1805 if (cp <= 0x0000007F) {
1806 rc = 1;
1807 out[0] = (char)cp;
1808 } else if (cp <= 0x000007FF) {
1809 rc = 2;
1810 out[0] = (cp >> 6 & 31) | 192;
1811 out[1] = (cp & 63) | 128;
1812 } else if (cp <= 0x0000FFFF) {
1813 rc = 3;
1814 out[0] = (cp >> 12 & 15) | 224;
1815 out[1] = (cp >> 6 & 63) | 128;
1816 out[2] = (cp & 63) | 128;
1817 } else if (cp <= 0x001FFFFF) {
1818 rc = 4;
1819 out[0] = (cp >> 18 & 7) | 240;
1820 out[1] = (cp >> 12 & 63) | 128;
1821 out[2] = (cp >> 6 & 63) | 128;
1822 out[3] = (cp & 63) | 128;
1823 } else if (cp <= 0x03FFFFFF) {
1824 rc = 5;
1825 out[0] = (cp >> 24 & 3) | 248;
1826 out[1] = (cp >> 18 & 63) | 128;
1827 out[2] = (cp >> 12 & 63) | 128;
1828 out[3] = (cp >> 6 & 63) | 128;
1829 out[4] = (cp & 63) | 128;
1830 } else if (cp <= 0x7FFFFFFF) {
1831 rc = 6;
1832 out[0] = (cp >> 30 & 1) | 252;
1833 out[1] = (cp >> 24 & 63) | 128;
1834 out[2] = (cp >> 18 & 63) | 128;
1835 out[3] = (cp >> 12 & 63) | 128;
1836 out[4] = (cp >> 6 & 63) | 128;
1837 out[5] = (cp & 63) | 128;
1838 } else
1839 return 0;
1840
1841 out[rc] = '\0';
1842 return rc;
1843 }
1844
1845 /*
1846 * If the string contains escape sequences,
1847 * replace it with an allocated rendering and return 1,
1848 * such that the caller can free it after use.
1849 * Otherwise, do nothing and return 0.
1850 */
1851 static int
1852 render_string(char **public, size_t *psz)
1853 {
1854 const char *src, *scp, *addcp, *seq;
1855 char *dst;
1856 size_t ssz, dsz, addsz;
1857 char utfbuf[7], res[6];
1858 int seqlen, unicode;
1859
1860 res[0] = '\\';
1861 res[1] = '\t';
1862 res[2] = ASCII_NBRSP;
1863 res[3] = ASCII_HYPH;
1864 res[4] = ASCII_BREAK;
1865 res[5] = '\0';
1866
1867 src = scp = *public;
1868 ssz = *psz;
1869 dst = NULL;
1870 dsz = 0;
1871
1872 while (scp < src + *psz) {
1873
1874 /* Leave normal characters unchanged. */
1875
1876 if (strchr(res, *scp) == NULL) {
1877 if (dst != NULL)
1878 dst[dsz++] = *scp;
1879 scp++;
1880 continue;
1881 }
1882
1883 /*
1884 * Found something that requires replacing,
1885 * make sure we have a destination buffer.
1886 */
1887
1888 if (dst == NULL) {
1889 dst = mandoc_malloc(ssz + 1);
1890 dsz = scp - src;
1891 memcpy(dst, src, dsz);
1892 }
1893
1894 /* Handle single-char special characters. */
1895
1896 switch (*scp) {
1897 case '\\':
1898 break;
1899 case '\t':
1900 case ASCII_NBRSP:
1901 dst[dsz++] = ' ';
1902 scp++;
1903 continue;
1904 case ASCII_HYPH:
1905 dst[dsz++] = '-';
1906 /* FALLTHROUGH */
1907 case ASCII_BREAK:
1908 scp++;
1909 continue;
1910 default:
1911 abort();
1912 }
1913
1914 /*
1915 * Found an escape sequence.
1916 * Read past the slash, then parse it.
1917 * Ignore everything except characters.
1918 */
1919
1920 scp++;
1921 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1922 continue;
1923
1924 /*
1925 * Render the special character
1926 * as either UTF-8 or ASCII.
1927 */
1928
1929 if (write_utf8) {
1930 unicode = mchars_spec2cp(seq, seqlen);
1931 if (unicode <= 0)
1932 continue;
1933 addsz = utf8(unicode, utfbuf);
1934 if (addsz == 0)
1935 continue;
1936 addcp = utfbuf;
1937 } else {
1938 addcp = mchars_spec2str(seq, seqlen, &addsz);
1939 if (addcp == NULL)
1940 continue;
1941 if (*addcp == ASCII_NBRSP) {
1942 addcp = " ";
1943 addsz = 1;
1944 }
1945 }
1946
1947 /* Copy the rendered glyph into the stream. */
1948
1949 ssz += addsz;
1950 dst = mandoc_realloc(dst, ssz + 1);
1951 memcpy(dst + dsz, addcp, addsz);
1952 dsz += addsz;
1953 }
1954 if (dst != NULL) {
1955 *public = dst;
1956 *psz = dsz;
1957 }
1958
1959 /* Trim trailing whitespace and NUL-terminate. */
1960
1961 while (*psz > 0 && (*public)[*psz - 1] == ' ')
1962 --*psz;
1963 if (dst != NULL) {
1964 (*public)[*psz] = '\0';
1965 return 1;
1966 } else
1967 return 0;
1968 }
1969
1970 static void
1971 dbadd_mlink(const struct mlink *mlink)
1972 {
1973 size_t i;
1974
1975 i = 1;
1976 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1977 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1978 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1979 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1980 SQL_STEP(stmts[STMT_INSERT_LINK]);
1981 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1982 }
1983
1984 static void
1985 dbadd_mlink_name(const struct mlink *mlink)
1986 {
1987 uint64_t bits;
1988 size_t i;
1989
1990 dbadd_mlink(mlink);
1991
1992 i = 1;
1993 SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
1994 bits = NAME_FILE & NAME_MASK;
1995 if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
1996 bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
1997 sqlite3_reset(stmts[STMT_SELECT_NAME]);
1998 }
1999
2000 i = 1;
2001 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
2002 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2003 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2004 SQL_STEP(stmts[STMT_INSERT_NAME]);
2005 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2006 }
2007
2008 /*
2009 * Flush the current page's terms (and their bits) into the database.
2010 * Wrap the entire set of additions in a transaction to make sqlite be a
2011 * little faster.
2012 * Also, handle escape sequences at the last possible moment.
2013 */
2014 static void
2015 dbadd(struct mpage *mpage)
2016 {
2017 struct mlink *mlink;
2018 struct str *key;
2019 char *cp;
2020 size_t i;
2021 unsigned int slot;
2022 int mustfree;
2023
2024 mlink = mpage->mlinks;
2025
2026 if (nodb) {
2027 for (key = ohash_first(&names, &slot); NULL != key;
2028 key = ohash_next(&names, &slot))
2029 free(key);
2030 for (key = ohash_first(&strings, &slot); NULL != key;
2031 key = ohash_next(&strings, &slot))
2032 free(key);
2033 if (0 == debug)
2034 return;
2035 while (NULL != mlink) {
2036 fputs(mlink->name, stdout);
2037 if (NULL == mlink->next ||
2038 strcmp(mlink->dsec, mlink->next->dsec) ||
2039 strcmp(mlink->fsec, mlink->next->fsec) ||
2040 strcmp(mlink->arch, mlink->next->arch)) {
2041 putchar('(');
2042 if ('\0' == *mlink->dsec)
2043 fputs(mlink->fsec, stdout);
2044 else
2045 fputs(mlink->dsec, stdout);
2046 if ('\0' != *mlink->arch)
2047 printf("/%s", mlink->arch);
2048 putchar(')');
2049 }
2050 mlink = mlink->next;
2051 if (NULL != mlink)
2052 fputs(", ", stdout);
2053 }
2054 printf(" - %s\n", mpage->desc);
2055 return;
2056 }
2057
2058 if (debug)
2059 say(mlink->file, "Adding to database");
2060
2061 cp = mpage->desc;
2062 i = strlen(cp);
2063 mustfree = render_string(&cp, &i);
2064 i = 1;
2065 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
2066 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
2067 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2068 mpage->pageid = sqlite3_last_insert_rowid(db);
2069 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2070 if (mustfree)
2071 free(cp);
2072
2073 while (NULL != mlink) {
2074 dbadd_mlink(mlink);
2075 mlink = mlink->next;
2076 }
2077 mlink = mpage->mlinks;
2078
2079 for (key = ohash_first(&names, &slot); NULL != key;
2080 key = ohash_next(&names, &slot)) {
2081 assert(key->mpage == mpage);
2082 i = 1;
2083 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2084 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
2085 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2086 SQL_STEP(stmts[STMT_INSERT_NAME]);
2087 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2088 free(key);
2089 }
2090 for (key = ohash_first(&strings, &slot); NULL != key;
2091 key = ohash_next(&strings, &slot)) {
2092 assert(key->mpage == mpage);
2093 i = 1;
2094 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2095 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
2096 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2097 SQL_STEP(stmts[STMT_INSERT_KEY]);
2098 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2099 free(key);
2100 }
2101 }
2102
2103 static void
2104 dbprune(void)
2105 {
2106 struct mpage *mpage;
2107 struct mlink *mlink;
2108 size_t i;
2109 unsigned int slot;
2110
2111 if (0 == nodb)
2112 SQL_EXEC("BEGIN TRANSACTION");
2113
2114 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2115 mpage = ohash_next(&mpages, &slot)) {
2116 mlink = mpage->mlinks;
2117 if (debug)
2118 say(mlink->file, "Deleting from database");
2119 if (nodb)
2120 continue;
2121 for ( ; NULL != mlink; mlink = mlink->next) {
2122 i = 1;
2123 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2124 i, mlink->dsec);
2125 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2126 i, mlink->arch);
2127 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2128 i, mlink->name);
2129 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2130 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2131 }
2132 }
2133
2134 if (0 == nodb)
2135 SQL_EXEC("END TRANSACTION");
2136 }
2137
2138 /*
2139 * Close an existing database and its prepared statements.
2140 * If "real" is not set, rename the temporary file into the real one.
2141 */
2142 static void
2143 dbclose(int real)
2144 {
2145 size_t i;
2146 int status;
2147 pid_t child;
2148
2149 if (nodb)
2150 return;
2151
2152 for (i = 0; i < STMT__MAX; i++) {
2153 sqlite3_finalize(stmts[i]);
2154 stmts[i] = NULL;
2155 }
2156
2157 sqlite3_close(db);
2158 db = NULL;
2159
2160 if (real)
2161 return;
2162
2163 if ('\0' == *tempfilename) {
2164 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2165 exitcode = (int)MANDOCLEVEL_SYSERR;
2166 say(MANDOC_DB, "&rename");
2167 }
2168 return;
2169 }
2170
2171 switch (child = fork()) {
2172 case -1:
2173 exitcode = (int)MANDOCLEVEL_SYSERR;
2174 say("", "&fork cmp");
2175 return;
2176 case 0:
2177 execlp("cmp", "cmp", "-s",
2178 tempfilename, MANDOC_DB, (char *)NULL);
2179 say("", "&exec cmp");
2180 exit(0);
2181 default:
2182 break;
2183 }
2184 if (-1 == waitpid(child, &status, 0)) {
2185 exitcode = (int)MANDOCLEVEL_SYSERR;
2186 say("", "&wait cmp");
2187 } else if (WIFSIGNALED(status)) {
2188 exitcode = (int)MANDOCLEVEL_SYSERR;
2189 say("", "cmp died from signal %d", WTERMSIG(status));
2190 } else if (WEXITSTATUS(status)) {
2191 exitcode = (int)MANDOCLEVEL_SYSERR;
2192 say(MANDOC_DB,
2193 "Data changed, but cannot replace database");
2194 }
2195
2196 *strrchr(tempfilename, '/') = '\0';
2197 switch (child = fork()) {
2198 case -1:
2199 exitcode = (int)MANDOCLEVEL_SYSERR;
2200 say("", "&fork rm");
2201 return;
2202 case 0:
2203 execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
2204 say("", "&exec rm");
2205 exit((int)MANDOCLEVEL_SYSERR);
2206 default:
2207 break;
2208 }
2209 if (-1 == waitpid(child, &status, 0)) {
2210 exitcode = (int)MANDOCLEVEL_SYSERR;
2211 say("", "&wait rm");
2212 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2213 exitcode = (int)MANDOCLEVEL_SYSERR;
2214 say("", "%s: Cannot remove temporary directory",
2215 tempfilename);
2216 }
2217 }
2218
2219 /*
2220 * This is straightforward stuff.
2221 * Open a database connection to a "temporary" database, then open a set
2222 * of prepared statements we'll use over and over again.
2223 * If "real" is set, we use the existing database; if not, we truncate a
2224 * temporary one.
2225 * Must be matched by dbclose().
2226 */
2227 static int
2228 dbopen(int real)
2229 {
2230 const char *sql;
2231 int rc, ofl;
2232
2233 if (nodb)
2234 return 1;
2235
2236 *tempfilename = '\0';
2237 ofl = SQLITE_OPEN_READWRITE;
2238
2239 if (real) {
2240 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2241 if (SQLITE_OK != rc) {
2242 exitcode = (int)MANDOCLEVEL_SYSERR;
2243 if (SQLITE_CANTOPEN != rc)
2244 say(MANDOC_DB, "%s", sqlite3_errstr(rc));
2245 return 0;
2246 }
2247 goto prepare_statements;
2248 }
2249
2250 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2251
2252 remove(MANDOC_DB "~");
2253 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2254 if (SQLITE_OK == rc)
2255 goto create_tables;
2256 if (MPARSE_QUICK & mparse_options) {
2257 exitcode = (int)MANDOCLEVEL_SYSERR;
2258 say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
2259 return 0;
2260 }
2261
2262 (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2263 sizeof(tempfilename));
2264 if (NULL == mkdtemp(tempfilename)) {
2265 exitcode = (int)MANDOCLEVEL_SYSERR;
2266 say("", "&%s", tempfilename);
2267 return 0;
2268 }
2269 (void)strlcat(tempfilename, "/" MANDOC_DB,
2270 sizeof(tempfilename));
2271 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2272 if (SQLITE_OK != rc) {
2273 exitcode = (int)MANDOCLEVEL_SYSERR;
2274 say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
2275 return 0;
2276 }
2277
2278 create_tables:
2279 sql = "CREATE TABLE \"mpages\" (\n"
2280 " \"desc\" TEXT NOT NULL,\n"
2281 " \"form\" INTEGER NOT NULL,\n"
2282 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2283 ");\n"
2284 "\n"
2285 "CREATE TABLE \"mlinks\" (\n"
2286 " \"sec\" TEXT NOT NULL,\n"
2287 " \"arch\" TEXT NOT NULL,\n"
2288 " \"name\" TEXT NOT NULL,\n"
2289 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2290 "ON DELETE CASCADE\n"
2291 ");\n"
2292 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2293 "\n"
2294 "CREATE TABLE \"names\" (\n"
2295 " \"bits\" INTEGER NOT NULL,\n"
2296 " \"name\" TEXT NOT NULL,\n"
2297 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2298 "ON DELETE CASCADE,\n"
2299 " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
2300 ");\n"
2301 "\n"
2302 "CREATE TABLE \"keys\" (\n"
2303 " \"bits\" INTEGER NOT NULL,\n"
2304 " \"key\" TEXT NOT NULL,\n"
2305 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2306 "ON DELETE CASCADE\n"
2307 ");\n"
2308 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2309
2310 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2311 exitcode = (int)MANDOCLEVEL_SYSERR;
2312 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2313 sqlite3_close(db);
2314 return 0;
2315 }
2316
2317 prepare_statements:
2318 if (SQLITE_OK != sqlite3_exec(db,
2319 "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2320 exitcode = (int)MANDOCLEVEL_SYSERR;
2321 say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2322 sqlite3_errmsg(db));
2323 sqlite3_close(db);
2324 return 0;
2325 }
2326
2327 sql = "DELETE FROM mpages WHERE pageid IN "
2328 "(SELECT pageid FROM mlinks WHERE "
2329 "sec=? AND arch=? AND name=?)";
2330 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2331 sql = "INSERT INTO mpages "
2332 "(desc,form) VALUES (?,?)";
2333 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2334 sql = "INSERT INTO mlinks "
2335 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2336 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2337 sql = "SELECT bits FROM names where pageid = ?";
2338 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
2339 sql = "INSERT INTO names "
2340 "(bits,name,pageid) VALUES (?,?,?)";
2341 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2342 sql = "INSERT INTO keys "
2343 "(bits,key,pageid) VALUES (?,?,?)";
2344 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2345
2346 #ifndef __APPLE__
2347 /*
2348 * When opening a new database, we can turn off
2349 * synchronous mode for much better performance.
2350 */
2351
2352 if (real && SQLITE_OK != sqlite3_exec(db,
2353 "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2354 exitcode = (int)MANDOCLEVEL_SYSERR;
2355 say(MANDOC_DB, "PRAGMA synchronous: %s",
2356 sqlite3_errmsg(db));
2357 sqlite3_close(db);
2358 return 0;
2359 }
2360 #endif
2361
2362 return 1;
2363 }
2364
2365 static int
2366 set_basedir(const char *targetdir, int report_baddir)
2367 {
2368 static char startdir[PATH_MAX];
2369 static int getcwd_status; /* 1 = ok, 2 = failure */
2370 static int chdir_status; /* 1 = changed directory */
2371 char *cp;
2372
2373 /*
2374 * Remember the original working directory, if possible.
2375 * This will be needed if the second or a later directory
2376 * on the command line is given as a relative path.
2377 * Do not error out if the current directory is not
2378 * searchable: Maybe it won't be needed after all.
2379 */
2380 if (0 == getcwd_status) {
2381 if (NULL == getcwd(startdir, sizeof(startdir))) {
2382 getcwd_status = 2;
2383 (void)strlcpy(startdir, strerror(errno),
2384 sizeof(startdir));
2385 } else
2386 getcwd_status = 1;
2387 }
2388
2389 /*
2390 * We are leaving the old base directory.
2391 * Do not use it any longer, not even for messages.
2392 */
2393 *basedir = '\0';
2394
2395 /*
2396 * If and only if the directory was changed earlier and
2397 * the next directory to process is given as a relative path,
2398 * first go back, or bail out if that is impossible.
2399 */
2400 if (chdir_status && '/' != *targetdir) {
2401 if (2 == getcwd_status) {
2402 exitcode = (int)MANDOCLEVEL_SYSERR;
2403 say("", "getcwd: %s", startdir);
2404 return 0;
2405 }
2406 if (-1 == chdir(startdir)) {
2407 exitcode = (int)MANDOCLEVEL_SYSERR;
2408 say("", "&chdir %s", startdir);
2409 return 0;
2410 }
2411 }
2412
2413 /*
2414 * Always resolve basedir to the canonicalized absolute
2415 * pathname and append a trailing slash, such that
2416 * we can reliably check whether files are inside.
2417 */
2418 if (NULL == realpath(targetdir, basedir)) {
2419 if (report_baddir || errno != ENOENT) {
2420 exitcode = (int)MANDOCLEVEL_BADARG;
2421 say("", "&%s: realpath", targetdir);
2422 }
2423 return 0;
2424 } else if (-1 == chdir(basedir)) {
2425 if (report_baddir || errno != ENOENT) {
2426 exitcode = (int)MANDOCLEVEL_BADARG;
2427 say("", "&chdir");
2428 }
2429 return 0;
2430 }
2431 chdir_status = 1;
2432 cp = strchr(basedir, '\0');
2433 if ('/' != cp[-1]) {
2434 if (cp - basedir >= PATH_MAX - 1) {
2435 exitcode = (int)MANDOCLEVEL_SYSERR;
2436 say("", "Filename too long");
2437 return 0;
2438 }
2439 *cp++ = '/';
2440 *cp = '\0';
2441 }
2442 return 1;
2443 }
2444
2445 static void
2446 say(const char *file, const char *format, ...)
2447 {
2448 va_list ap;
2449 int use_errno;
2450
2451 if ('\0' != *basedir)
2452 fprintf(stderr, "%s", basedir);
2453 if ('\0' != *basedir && '\0' != *file)
2454 fputc('/', stderr);
2455 if ('\0' != *file)
2456 fprintf(stderr, "%s", file);
2457
2458 use_errno = 1;
2459 if (NULL != format) {
2460 switch (*format) {
2461 case '&':
2462 format++;
2463 break;
2464 case '\0':
2465 format = NULL;
2466 break;
2467 default:
2468 use_errno = 0;
2469 break;
2470 }
2471 }
2472 if (NULL != format) {
2473 if ('\0' != *basedir || '\0' != *file)
2474 fputs(": ", stderr);
2475 va_start(ap, format);
2476 vfprintf(stderr, format, ap);
2477 va_end(ap);
2478 }
2479 if (use_errno) {
2480 if ('\0' != *basedir || '\0' != *file || NULL != format)
2481 fputs(": ", stderr);
2482 perror(NULL);
2483 } else
2484 fputc('\n', stderr);
2485 }
A git conversion mirror of mandoc.bsd.lv