]> 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
Use getprogname(3) rather than __progname.
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.206 2015/11/06 16:30:33 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 static char tempfilename[32];
187 static int nodb; /* no database changes */
188 static int mparse_options; /* abort the parse early */
189 static int use_all; /* use all found files */
190 static int debug; /* print what we're doing */
191 static int warnings; /* warn about crap */
192 static int write_utf8; /* write UTF-8 output; else ASCII */
193 static int exitcode; /* to be returned by main */
194 static enum op op; /* operational mode */
195 static char basedir[PATH_MAX]; /* current base directory */
196 static struct ohash mpages; /* table of distinct manual pages */
197 static struct ohash mlinks; /* table of directory entries */
198 static struct ohash names; /* table of all names */
199 static struct ohash strings; /* table of all strings */
200 static sqlite3 *db = NULL; /* current database */
201 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
202 static uint64_t name_mask;
203
204 static const struct mdoc_handler mdocs[MDOC_MAX] = {
205 { NULL, 0 }, /* Ap */
206 { NULL, 0 }, /* Dd */
207 { NULL, 0 }, /* Dt */
208 { NULL, 0 }, /* Os */
209 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
210 { parse_mdoc_head, TYPE_Ss }, /* Ss */
211 { NULL, 0 }, /* Pp */
212 { NULL, 0 }, /* D1 */
213 { NULL, 0 }, /* Dl */
214 { NULL, 0 }, /* Bd */
215 { NULL, 0 }, /* Ed */
216 { NULL, 0 }, /* Bl */
217 { NULL, 0 }, /* El */
218 { NULL, 0 }, /* It */
219 { NULL, 0 }, /* Ad */
220 { NULL, TYPE_An }, /* An */
221 { NULL, TYPE_Ar }, /* Ar */
222 { NULL, TYPE_Cd }, /* Cd */
223 { NULL, TYPE_Cm }, /* Cm */
224 { NULL, TYPE_Dv }, /* Dv */
225 { NULL, TYPE_Er }, /* Er */
226 { NULL, TYPE_Ev }, /* Ev */
227 { NULL, 0 }, /* Ex */
228 { NULL, TYPE_Fa }, /* Fa */
229 { parse_mdoc_Fd, 0 }, /* Fd */
230 { NULL, TYPE_Fl }, /* Fl */
231 { parse_mdoc_Fn, 0 }, /* Fn */
232 { NULL, TYPE_Ft }, /* Ft */
233 { NULL, TYPE_Ic }, /* Ic */
234 { NULL, TYPE_In }, /* In */
235 { NULL, TYPE_Li }, /* Li */
236 { parse_mdoc_Nd, 0 }, /* Nd */
237 { parse_mdoc_Nm, 0 }, /* Nm */
238 { NULL, 0 }, /* Op */
239 { NULL, 0 }, /* Ot */
240 { NULL, TYPE_Pa }, /* Pa */
241 { NULL, 0 }, /* Rv */
242 { NULL, TYPE_St }, /* St */
243 { NULL, TYPE_Va }, /* Va */
244 { parse_mdoc_body, TYPE_Va }, /* Vt */
245 { parse_mdoc_Xr, 0 }, /* Xr */
246 { NULL, 0 }, /* %A */
247 { NULL, 0 }, /* %B */
248 { NULL, 0 }, /* %D */
249 { NULL, 0 }, /* %I */
250 { NULL, 0 }, /* %J */
251 { NULL, 0 }, /* %N */
252 { NULL, 0 }, /* %O */
253 { NULL, 0 }, /* %P */
254 { NULL, 0 }, /* %R */
255 { NULL, 0 }, /* %T */
256 { NULL, 0 }, /* %V */
257 { NULL, 0 }, /* Ac */
258 { NULL, 0 }, /* Ao */
259 { NULL, 0 }, /* Aq */
260 { NULL, TYPE_At }, /* At */
261 { NULL, 0 }, /* Bc */
262 { NULL, 0 }, /* Bf */
263 { NULL, 0 }, /* Bo */
264 { NULL, 0 }, /* Bq */
265 { NULL, TYPE_Bsx }, /* Bsx */
266 { NULL, TYPE_Bx }, /* Bx */
267 { NULL, 0 }, /* Db */
268 { NULL, 0 }, /* Dc */
269 { NULL, 0 }, /* Do */
270 { NULL, 0 }, /* Dq */
271 { NULL, 0 }, /* Ec */
272 { NULL, 0 }, /* Ef */
273 { NULL, TYPE_Em }, /* Em */
274 { NULL, 0 }, /* Eo */
275 { NULL, TYPE_Fx }, /* Fx */
276 { NULL, TYPE_Ms }, /* Ms */
277 { NULL, 0 }, /* No */
278 { NULL, 0 }, /* Ns */
279 { NULL, TYPE_Nx }, /* Nx */
280 { NULL, TYPE_Ox }, /* Ox */
281 { NULL, 0 }, /* Pc */
282 { NULL, 0 }, /* Pf */
283 { NULL, 0 }, /* Po */
284 { NULL, 0 }, /* Pq */
285 { NULL, 0 }, /* Qc */
286 { NULL, 0 }, /* Ql */
287 { NULL, 0 }, /* Qo */
288 { NULL, 0 }, /* Qq */
289 { NULL, 0 }, /* Re */
290 { NULL, 0 }, /* Rs */
291 { NULL, 0 }, /* Sc */
292 { NULL, 0 }, /* So */
293 { NULL, 0 }, /* Sq */
294 { NULL, 0 }, /* Sm */
295 { NULL, 0 }, /* Sx */
296 { NULL, TYPE_Sy }, /* Sy */
297 { NULL, TYPE_Tn }, /* Tn */
298 { NULL, 0 }, /* Ux */
299 { NULL, 0 }, /* Xc */
300 { NULL, 0 }, /* Xo */
301 { parse_mdoc_Fo, 0 }, /* Fo */
302 { NULL, 0 }, /* Fc */
303 { NULL, 0 }, /* Oo */
304 { NULL, 0 }, /* Oc */
305 { NULL, 0 }, /* Bk */
306 { NULL, 0 }, /* Ek */
307 { NULL, 0 }, /* Bt */
308 { NULL, 0 }, /* Hf */
309 { NULL, 0 }, /* Fr */
310 { NULL, 0 }, /* Ud */
311 { NULL, TYPE_Lb }, /* Lb */
312 { NULL, 0 }, /* Lp */
313 { NULL, TYPE_Lk }, /* Lk */
314 { NULL, TYPE_Mt }, /* Mt */
315 { NULL, 0 }, /* Brq */
316 { NULL, 0 }, /* Bro */
317 { NULL, 0 }, /* Brc */
318 { NULL, 0 }, /* %C */
319 { NULL, 0 }, /* Es */
320 { NULL, 0 }, /* En */
321 { NULL, TYPE_Dx }, /* Dx */
322 { NULL, 0 }, /* %Q */
323 { NULL, 0 }, /* br */
324 { NULL, 0 }, /* sp */
325 { NULL, 0 }, /* %U */
326 { NULL, 0 }, /* Ta */
327 { NULL, 0 }, /* ll */
328 };
329
330
331 int
332 mandocdb(int argc, char *argv[])
333 {
334 struct manconf conf;
335 struct mparse *mp;
336 const char *path_arg, *progname;
337 size_t j, sz;
338 int ch, i;
339
340 memset(&conf, 0, sizeof(conf));
341 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
342
343 /*
344 * We accept a few different invocations.
345 * The CHECKOP macro makes sure that invocation styles don't
346 * clobber each other.
347 */
348 #define CHECKOP(_op, _ch) do \
349 if (OP_DEFAULT != (_op)) { \
350 warnx("-%c: Conflicting option", (_ch)); \
351 goto usage; \
352 } while (/*CONSTCOND*/0)
353
354 path_arg = NULL;
355 op = OP_DEFAULT;
356
357 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
358 switch (ch) {
359 case 'a':
360 use_all = 1;
361 break;
362 case 'C':
363 CHECKOP(op, ch);
364 path_arg = optarg;
365 op = OP_CONFFILE;
366 break;
367 case 'D':
368 debug++;
369 break;
370 case 'd':
371 CHECKOP(op, ch);
372 path_arg = optarg;
373 op = OP_UPDATE;
374 break;
375 case 'n':
376 nodb = 1;
377 break;
378 case 'p':
379 warnings = 1;
380 break;
381 case 'Q':
382 mparse_options |= MPARSE_QUICK;
383 break;
384 case 'T':
385 if (strcmp(optarg, "utf8")) {
386 warnx("-T%s: Unsupported output format",
387 optarg);
388 goto usage;
389 }
390 write_utf8 = 1;
391 break;
392 case 't':
393 CHECKOP(op, ch);
394 dup2(STDOUT_FILENO, STDERR_FILENO);
395 op = OP_TEST;
396 nodb = warnings = 1;
397 break;
398 case 'u':
399 CHECKOP(op, ch);
400 path_arg = optarg;
401 op = OP_DELETE;
402 break;
403 case 'v':
404 /* Compatibility with espie@'s makewhatis. */
405 break;
406 default:
407 goto usage;
408 }
409
410 argc -= optind;
411 argv += optind;
412
413 if (OP_CONFFILE == op && argc > 0) {
414 warnx("-C: Too many arguments");
415 goto usage;
416 }
417
418 exitcode = (int)MANDOCLEVEL_OK;
419 mchars_alloc();
420 mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
421 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
422 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
423
424 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
425
426 /*
427 * Most of these deal with a specific directory.
428 * Jump into that directory first.
429 */
430 if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
431 goto out;
432
433 if (dbopen(1)) {
434 /*
435 * The existing database is usable. Process
436 * all files specified on the command-line.
437 */
438 use_all = 1;
439 for (i = 0; i < argc; i++)
440 filescan(argv[i]);
441 if (OP_TEST != op)
442 dbprune();
443 } else {
444 /*
445 * Database missing or corrupt.
446 * Recreate from scratch.
447 */
448 exitcode = (int)MANDOCLEVEL_OK;
449 op = OP_DEFAULT;
450 if (0 == treescan())
451 goto out;
452 if (0 == dbopen(0))
453 goto out;
454 }
455 if (OP_DELETE != op)
456 mpages_merge(mp);
457 dbclose(OP_DEFAULT == op ? 0 : 1);
458 } else {
459 /*
460 * If we have arguments, use them as our manpaths.
461 * If we don't, grok from manpath(1) or however else
462 * manconf_parse() wants to do it.
463 */
464 if (argc > 0) {
465 conf.manpath.paths = mandoc_reallocarray(NULL,
466 argc, sizeof(char *));
467 conf.manpath.sz = (size_t)argc;
468 for (i = 0; i < argc; i++)
469 conf.manpath.paths[i] = mandoc_strdup(argv[i]);
470 } else
471 manconf_parse(&conf, path_arg, NULL, NULL);
472
473 if (conf.manpath.sz == 0) {
474 exitcode = (int)MANDOCLEVEL_BADARG;
475 say("", "Empty manpath");
476 }
477
478 /*
479 * First scan the tree rooted at a base directory, then
480 * build a new database and finally move it into place.
481 * Ignore zero-length directories and strip trailing
482 * slashes.
483 */
484 for (j = 0; j < conf.manpath.sz; j++) {
485 sz = strlen(conf.manpath.paths[j]);
486 if (sz && conf.manpath.paths[j][sz - 1] == '/')
487 conf.manpath.paths[j][--sz] = '\0';
488 if (0 == sz)
489 continue;
490
491 if (j) {
492 mandoc_ohash_init(&mpages, 6,
493 offsetof(struct mpage, inodev));
494 mandoc_ohash_init(&mlinks, 6,
495 offsetof(struct mlink, file));
496 }
497
498 if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
499 continue;
500 if (0 == treescan())
501 continue;
502 if (0 == dbopen(0))
503 continue;
504
505 mpages_merge(mp);
506 if (warnings && !nodb &&
507 ! (MPARSE_QUICK & mparse_options))
508 names_check();
509 dbclose(0);
510
511 if (j + 1 < conf.manpath.sz) {
512 mpages_free();
513 ohash_delete(&mpages);
514 ohash_delete(&mlinks);
515 }
516 }
517 }
518 out:
519 manconf_free(&conf);
520 mparse_free(mp);
521 mchars_free();
522 mpages_free();
523 ohash_delete(&mpages);
524 ohash_delete(&mlinks);
525 return exitcode;
526 usage:
527 progname = getprogname();
528 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
529 " %s [-aDnpQ] [-Tutf8] dir ...\n"
530 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
531 " %s [-Dnp] -u dir [file ...]\n"
532 " %s [-Q] -t file ...\n",
533 progname, progname, progname, progname, progname);
534
535 return (int)MANDOCLEVEL_BADARG;
536 }
537
538 /*
539 * Scan a directory tree rooted at "basedir" for manpages.
540 * We use fts(), scanning directory parts along the way for clues to our
541 * section and architecture.
542 *
543 * If use_all has been specified, grok all files.
544 * If not, sanitise paths to the following:
545 *
546 * [./]man*[/<arch>]/<name>.<section>
547 * or
548 * [./]cat<section>[/<arch>]/<name>.0
549 *
550 * TODO: accomodate for multi-language directories.
551 */
552 static int
553 treescan(void)
554 {
555 char buf[PATH_MAX];
556 FTS *f;
557 FTSENT *ff;
558 struct mlink *mlink;
559 int dform, gzip;
560 char *dsec, *arch, *fsec, *cp;
561 const char *path;
562 const char *argv[2];
563
564 argv[0] = ".";
565 argv[1] = (char *)NULL;
566
567 f = fts_open((char * const *)argv,
568 FTS_PHYSICAL | FTS_NOCHDIR, NULL);
569 if (f == NULL) {
570 exitcode = (int)MANDOCLEVEL_SYSERR;
571 say("", "&fts_open");
572 return 0;
573 }
574
575 dsec = arch = NULL;
576 dform = FORM_NONE;
577
578 while ((ff = fts_read(f)) != NULL) {
579 path = ff->fts_path + 2;
580 switch (ff->fts_info) {
581
582 /*
583 * Symbolic links require various sanity checks,
584 * then get handled just like regular files.
585 */
586 case FTS_SL:
587 if (realpath(path, buf) == NULL) {
588 if (warnings)
589 say(path, "&realpath");
590 continue;
591 }
592 if (strstr(buf, basedir) != buf
593 #ifdef HOMEBREWDIR
594 && strstr(buf, HOMEBREWDIR) != buf
595 #endif
596 ) {
597 if (warnings) say("",
598 "%s: outside base directory", buf);
599 continue;
600 }
601 /* Use logical inode to avoid mpages dupe. */
602 if (stat(path, ff->fts_statp) == -1) {
603 if (warnings)
604 say(path, "&stat");
605 continue;
606 }
607 /* FALLTHROUGH */
608
609 /*
610 * If we're a regular file, add an mlink by using the
611 * stored directory data and handling the filename.
612 */
613 case FTS_F:
614 if ( ! strcmp(path, MANDOC_DB))
615 continue;
616 if ( ! use_all && ff->fts_level < 2) {
617 if (warnings)
618 say(path, "Extraneous file");
619 continue;
620 }
621 gzip = 0;
622 fsec = NULL;
623 while (fsec == NULL) {
624 fsec = strrchr(ff->fts_name, '.');
625 if (fsec == NULL || strcmp(fsec+1, "gz"))
626 break;
627 gzip = 1;
628 *fsec = '\0';
629 fsec = NULL;
630 }
631 if (fsec == NULL) {
632 if ( ! use_all) {
633 if (warnings)
634 say(path,
635 "No filename suffix");
636 continue;
637 }
638 } else if ( ! strcmp(++fsec, "html")) {
639 if (warnings)
640 say(path, "Skip html");
641 continue;
642 } else if ( ! strcmp(fsec, "ps")) {
643 if (warnings)
644 say(path, "Skip ps");
645 continue;
646 } else if ( ! strcmp(fsec, "pdf")) {
647 if (warnings)
648 say(path, "Skip pdf");
649 continue;
650 } else if ( ! use_all &&
651 ((dform == FORM_SRC &&
652 strncmp(fsec, dsec, strlen(dsec))) ||
653 (dform == FORM_CAT && strcmp(fsec, "0")))) {
654 if (warnings)
655 say(path, "Wrong filename suffix");
656 continue;
657 } else
658 fsec[-1] = '\0';
659
660 mlink = mandoc_calloc(1, sizeof(struct mlink));
661 if (strlcpy(mlink->file, path,
662 sizeof(mlink->file)) >=
663 sizeof(mlink->file)) {
664 say(path, "Filename too long");
665 free(mlink);
666 continue;
667 }
668 mlink->dform = dform;
669 mlink->dsec = dsec;
670 mlink->arch = arch;
671 mlink->name = ff->fts_name;
672 mlink->fsec = fsec;
673 mlink->gzip = gzip;
674 mlink_add(mlink, ff->fts_statp);
675 continue;
676
677 case FTS_D:
678 case FTS_DP:
679 break;
680
681 default:
682 if (warnings)
683 say(path, "Not a regular file");
684 continue;
685 }
686
687 switch (ff->fts_level) {
688 case 0:
689 /* Ignore the root directory. */
690 break;
691 case 1:
692 /*
693 * This might contain manX/ or catX/.
694 * Try to infer this from the name.
695 * If we're not in use_all, enforce it.
696 */
697 cp = ff->fts_name;
698 if (ff->fts_info == FTS_DP) {
699 dform = FORM_NONE;
700 dsec = NULL;
701 break;
702 }
703
704 if ( ! strncmp(cp, "man", 3)) {
705 dform = FORM_SRC;
706 dsec = cp + 3;
707 } else if ( ! strncmp(cp, "cat", 3)) {
708 dform = FORM_CAT;
709 dsec = cp + 3;
710 } else {
711 dform = FORM_NONE;
712 dsec = NULL;
713 }
714
715 if (dsec != NULL || use_all)
716 break;
717
718 if (warnings)
719 say(path, "Unknown directory part");
720 fts_set(f, ff, FTS_SKIP);
721 break;
722 case 2:
723 /*
724 * Possibly our architecture.
725 * If we're descending, keep tabs on it.
726 */
727 if (ff->fts_info != FTS_DP && dsec != NULL)
728 arch = ff->fts_name;
729 else
730 arch = NULL;
731 break;
732 default:
733 if (ff->fts_info == FTS_DP || use_all)
734 break;
735 if (warnings)
736 say(path, "Extraneous directory part");
737 fts_set(f, ff, FTS_SKIP);
738 break;
739 }
740 }
741
742 fts_close(f);
743 return 1;
744 }
745
746 /*
747 * Add a file to the mlinks table.
748 * Do not verify that it's a "valid" looking manpage (we'll do that
749 * later).
750 *
751 * Try to infer the manual section, architecture, and page name from the
752 * path, assuming it looks like
753 *
754 * [./]man*[/<arch>]/<name>.<section>
755 * or
756 * [./]cat<section>[/<arch>]/<name>.0
757 *
758 * See treescan() for the fts(3) version of this.
759 */
760 static void
761 filescan(const char *file)
762 {
763 char buf[PATH_MAX];
764 struct stat st;
765 struct mlink *mlink;
766 char *p, *start;
767
768 assert(use_all);
769
770 if (0 == strncmp(file, "./", 2))
771 file += 2;
772
773 /*
774 * We have to do lstat(2) before realpath(3) loses
775 * the information whether this is a symbolic link.
776 * We need to know that because for symbolic links,
777 * we want to use the orginal file name, while for
778 * regular files, we want to use the real path.
779 */
780 if (-1 == lstat(file, &st)) {
781 exitcode = (int)MANDOCLEVEL_BADARG;
782 say(file, "&lstat");
783 return;
784 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
785 exitcode = (int)MANDOCLEVEL_BADARG;
786 say(file, "Not a regular file");
787 return;
788 }
789
790 /*
791 * We have to resolve the file name to the real path
792 * in any case for the base directory check.
793 */
794 if (NULL == realpath(file, buf)) {
795 exitcode = (int)MANDOCLEVEL_BADARG;
796 say(file, "&realpath");
797 return;
798 }
799
800 if (OP_TEST == op)
801 start = buf;
802 else if (strstr(buf, basedir) == buf)
803 start = buf + strlen(basedir);
804 #ifdef HOMEBREWDIR
805 else if (strstr(buf, HOMEBREWDIR) == buf)
806 start = buf;
807 #endif
808 else {
809 exitcode = (int)MANDOCLEVEL_BADARG;
810 say("", "%s: outside base directory", buf);
811 return;
812 }
813
814 /*
815 * Now we are sure the file is inside our tree.
816 * If it is a symbolic link, ignore the real path
817 * and use the original name.
818 * This implies passing stuff like "cat1/../man1/foo.1"
819 * on the command line won't work. So don't do that.
820 * Note the stat(2) can still fail if the link target
821 * doesn't exist.
822 */
823 if (S_IFLNK & st.st_mode) {
824 if (-1 == stat(buf, &st)) {
825 exitcode = (int)MANDOCLEVEL_BADARG;
826 say(file, "&stat");
827 return;
828 }
829 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
830 say(file, "Filename too long");
831 return;
832 }
833 start = buf;
834 if (OP_TEST != op && strstr(buf, basedir) == buf)
835 start += strlen(basedir);
836 }
837
838 mlink = mandoc_calloc(1, sizeof(struct mlink));
839 mlink->dform = FORM_NONE;
840 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
841 sizeof(mlink->file)) {
842 say(start, "Filename too long");
843 free(mlink);
844 return;
845 }
846
847 /*
848 * First try to guess our directory structure.
849 * If we find a separator, try to look for man* or cat*.
850 * If we find one of these and what's underneath is a directory,
851 * assume it's an architecture.
852 */
853 if (NULL != (p = strchr(start, '/'))) {
854 *p++ = '\0';
855 if (0 == strncmp(start, "man", 3)) {
856 mlink->dform = FORM_SRC;
857 mlink->dsec = start + 3;
858 } else if (0 == strncmp(start, "cat", 3)) {
859 mlink->dform = FORM_CAT;
860 mlink->dsec = start + 3;
861 }
862
863 start = p;
864 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
865 *p++ = '\0';
866 mlink->arch = start;
867 start = p;
868 }
869 }
870
871 /*
872 * Now check the file suffix.
873 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
874 */
875 p = strrchr(start, '\0');
876 while (p-- > start && '/' != *p && '.' != *p)
877 /* Loop. */ ;
878
879 if ('.' == *p) {
880 *p++ = '\0';
881 mlink->fsec = p;
882 }
883
884 /*
885 * Now try to parse the name.
886 * Use the filename portion of the path.
887 */
888 mlink->name = start;
889 if (NULL != (p = strrchr(start, '/'))) {
890 mlink->name = p + 1;
891 *p = '\0';
892 }
893 mlink_add(mlink, &st);
894 }
895
896 static void
897 mlink_add(struct mlink *mlink, const struct stat *st)
898 {
899 struct inodev inodev;
900 struct mpage *mpage;
901 unsigned int slot;
902
903 assert(NULL != mlink->file);
904
905 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
906 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
907 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
908 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
909
910 if ('0' == *mlink->fsec) {
911 free(mlink->fsec);
912 mlink->fsec = mandoc_strdup(mlink->dsec);
913 mlink->fform = FORM_CAT;
914 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
915 mlink->fform = FORM_SRC;
916 else
917 mlink->fform = FORM_NONE;
918
919 slot = ohash_qlookup(&mlinks, mlink->file);
920 assert(NULL == ohash_find(&mlinks, slot));
921 ohash_insert(&mlinks, slot, mlink);
922
923 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
924 inodev.st_ino = st->st_ino;
925 inodev.st_dev = st->st_dev;
926 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
927 sizeof(struct inodev), inodev.st_ino);
928 mpage = ohash_find(&mpages, slot);
929 if (NULL == mpage) {
930 mpage = mandoc_calloc(1, sizeof(struct mpage));
931 mpage->inodev.st_ino = inodev.st_ino;
932 mpage->inodev.st_dev = inodev.st_dev;
933 ohash_insert(&mpages, slot, mpage);
934 } else
935 mlink->next = mpage->mlinks;
936 mpage->mlinks = mlink;
937 mlink->mpage = mpage;
938 }
939
940 static void
941 mlink_free(struct mlink *mlink)
942 {
943
944 free(mlink->dsec);
945 free(mlink->arch);
946 free(mlink->name);
947 free(mlink->fsec);
948 free(mlink);
949 }
950
951 static void
952 mpages_free(void)
953 {
954 struct mpage *mpage;
955 struct mlink *mlink;
956 unsigned int slot;
957
958 mpage = ohash_first(&mpages, &slot);
959 while (NULL != mpage) {
960 while (NULL != (mlink = mpage->mlinks)) {
961 mpage->mlinks = mlink->next;
962 mlink_free(mlink);
963 }
964 free(mpage->sec);
965 free(mpage->arch);
966 free(mpage->title);
967 free(mpage->desc);
968 free(mpage);
969 mpage = ohash_next(&mpages, &slot);
970 }
971 }
972
973 /*
974 * For each mlink to the mpage, check whether the path looks like
975 * it is formatted, and if it does, check whether a source manual
976 * exists by the same name, ignoring the suffix.
977 * If both conditions hold, drop the mlink.
978 */
979 static void
980 mlinks_undupe(struct mpage *mpage)
981 {
982 char buf[PATH_MAX];
983 struct mlink **prev;
984 struct mlink *mlink;
985 char *bufp;
986
987 mpage->form = FORM_CAT;
988 prev = &mpage->mlinks;
989 while (NULL != (mlink = *prev)) {
990 if (FORM_CAT != mlink->dform) {
991 mpage->form = FORM_NONE;
992 goto nextlink;
993 }
994 (void)strlcpy(buf, mlink->file, sizeof(buf));
995 bufp = strstr(buf, "cat");
996 assert(NULL != bufp);
997 memcpy(bufp, "man", 3);
998 if (NULL != (bufp = strrchr(buf, '.')))
999 *++bufp = '\0';
1000 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1001 if (NULL == ohash_find(&mlinks,
1002 ohash_qlookup(&mlinks, buf)))
1003 goto nextlink;
1004 if (warnings)
1005 say(mlink->file, "Man source exists: %s", buf);
1006 if (use_all)
1007 goto nextlink;
1008 *prev = mlink->next;
1009 mlink_free(mlink);
1010 continue;
1011 nextlink:
1012 prev = &(*prev)->next;
1013 }
1014 }
1015
1016 static void
1017 mlink_check(struct mpage *mpage, struct mlink *mlink)
1018 {
1019 struct str *str;
1020 unsigned int slot;
1021
1022 /*
1023 * Check whether the manual section given in a file
1024 * agrees with the directory where the file is located.
1025 * Some manuals have suffixes like (3p) on their
1026 * section number either inside the file or in the
1027 * directory name, some are linked into more than one
1028 * section, like encrypt(1) = makekey(8).
1029 */
1030
1031 if (FORM_SRC == mpage->form &&
1032 strcasecmp(mpage->sec, mlink->dsec))
1033 say(mlink->file, "Section \"%s\" manual in %s directory",
1034 mpage->sec, mlink->dsec);
1035
1036 /*
1037 * Manual page directories exist for each kernel
1038 * architecture as returned by machine(1).
1039 * However, many manuals only depend on the
1040 * application architecture as returned by arch(1).
1041 * For example, some (2/ARM) manuals are shared
1042 * across the "armish" and "zaurus" kernel
1043 * architectures.
1044 * A few manuals are even shared across completely
1045 * different architectures, for example fdformat(1)
1046 * on amd64, i386, sparc, and sparc64.
1047 */
1048
1049 if (strcasecmp(mpage->arch, mlink->arch))
1050 say(mlink->file, "Architecture \"%s\" manual in "
1051 "\"%s\" directory", mpage->arch, mlink->arch);
1052
1053 /*
1054 * XXX
1055 * parse_cat() doesn't set NAME_TITLE yet.
1056 */
1057
1058 if (FORM_CAT == mpage->form)
1059 return;
1060
1061 /*
1062 * Check whether this mlink
1063 * appears as a name in the NAME section.
1064 */
1065
1066 slot = ohash_qlookup(&names, mlink->name);
1067 str = ohash_find(&names, slot);
1068 assert(NULL != str);
1069 if ( ! (NAME_TITLE & str->mask))
1070 say(mlink->file, "Name missing in NAME section");
1071 }
1072
1073 /*
1074 * Run through the files in the global vector "mpages"
1075 * and add them to the database specified in "basedir".
1076 *
1077 * This handles the parsing scheme itself, using the cues of directory
1078 * and filename to determine whether the file is parsable or not.
1079 */
1080 static void
1081 mpages_merge(struct mparse *mp)
1082 {
1083 char any[] = "any";
1084 struct mpage *mpage, *mpage_dest;
1085 struct mlink *mlink, *mlink_dest;
1086 struct roff_man *man;
1087 char *sodest;
1088 char *cp;
1089 int fd;
1090 unsigned int pslot;
1091
1092 if ( ! nodb)
1093 SQL_EXEC("BEGIN TRANSACTION");
1094
1095 mpage = ohash_first(&mpages, &pslot);
1096 while (mpage != NULL) {
1097 mlinks_undupe(mpage);
1098 if ((mlink = mpage->mlinks) == NULL) {
1099 mpage = ohash_next(&mpages, &pslot);
1100 continue;
1101 }
1102
1103 name_mask = NAME_MASK;
1104 mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1105 mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1106 mparse_reset(mp);
1107 man = NULL;
1108 sodest = NULL;
1109
1110 mparse_open(mp, &fd, mlink->file);
1111 if (fd == -1) {
1112 say(mlink->file, "&open");
1113 goto nextpage;
1114 }
1115
1116 /*
1117 * Interpret the file as mdoc(7) or man(7) source
1118 * code, unless it is known to be formatted.
1119 */
1120 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1121 mparse_readfd(mp, fd, mlink->file);
1122 mparse_result(mp, &man, &sodest);
1123 }
1124
1125 if (sodest != NULL) {
1126 mlink_dest = ohash_find(&mlinks,
1127 ohash_qlookup(&mlinks, sodest));
1128 if (mlink_dest == NULL) {
1129 mandoc_asprintf(&cp, "%s.gz", sodest);
1130 mlink_dest = ohash_find(&mlinks,
1131 ohash_qlookup(&mlinks, cp));
1132 free(cp);
1133 }
1134 if (mlink_dest != NULL) {
1135
1136 /* The .so target exists. */
1137
1138 mpage_dest = mlink_dest->mpage;
1139 while (1) {
1140 mlink->mpage = mpage_dest;
1141
1142 /*
1143 * If the target was already
1144 * processed, add the links
1145 * to the database now.
1146 * Otherwise, this will
1147 * happen when we come
1148 * to the target.
1149 */
1150
1151 if (mpage_dest->pageid)
1152 dbadd_mlink_name(mlink);
1153
1154 if (mlink->next == NULL)
1155 break;
1156 mlink = mlink->next;
1157 }
1158
1159 /* Move all links to the target. */
1160
1161 mlink->next = mlink_dest->next;
1162 mlink_dest->next = mpage->mlinks;
1163 mpage->mlinks = NULL;
1164 }
1165 goto nextpage;
1166 } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1167 mdoc_validate(man);
1168 mpage->form = FORM_SRC;
1169 mpage->sec = man->meta.msec;
1170 mpage->sec = mandoc_strdup(
1171 mpage->sec == NULL ? "" : mpage->sec);
1172 mpage->arch = man->meta.arch;
1173 mpage->arch = mandoc_strdup(
1174 mpage->arch == NULL ? "" : mpage->arch);
1175 mpage->title = mandoc_strdup(man->meta.title);
1176 } else if (man != NULL && man->macroset == MACROSET_MAN) {
1177 man_validate(man);
1178 mpage->form = FORM_SRC;
1179 mpage->sec = mandoc_strdup(man->meta.msec);
1180 mpage->arch = mandoc_strdup(mlink->arch);
1181 mpage->title = mandoc_strdup(man->meta.title);
1182 } else {
1183 mpage->form = FORM_CAT;
1184 mpage->sec = mandoc_strdup(mlink->dsec);
1185 mpage->arch = mandoc_strdup(mlink->arch);
1186 mpage->title = mandoc_strdup(mlink->name);
1187 }
1188 putkey(mpage, mpage->sec, TYPE_sec);
1189 if (*mpage->arch != '\0')
1190 putkey(mpage, mpage->arch, TYPE_arch);
1191
1192 for ( ; mlink != NULL; mlink = mlink->next) {
1193 if ('\0' != *mlink->dsec)
1194 putkey(mpage, mlink->dsec, TYPE_sec);
1195 if ('\0' != *mlink->fsec)
1196 putkey(mpage, mlink->fsec, TYPE_sec);
1197 putkey(mpage, '\0' == *mlink->arch ?
1198 any : mlink->arch, TYPE_arch);
1199 putkey(mpage, mlink->name, NAME_FILE);
1200 }
1201
1202 assert(mpage->desc == NULL);
1203 if (man != NULL && man->macroset == MACROSET_MDOC)
1204 parse_mdoc(mpage, &man->meta, man->first);
1205 else if (man != NULL)
1206 parse_man(mpage, &man->meta, man->first);
1207 else
1208 parse_cat(mpage, fd);
1209 if (mpage->desc == NULL)
1210 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1211
1212 if (warnings && !use_all)
1213 for (mlink = mpage->mlinks; mlink;
1214 mlink = mlink->next)
1215 mlink_check(mpage, mlink);
1216
1217 dbadd(mpage);
1218 mlink = mpage->mlinks;
1219
1220 nextpage:
1221 ohash_delete(&strings);
1222 ohash_delete(&names);
1223 mpage = ohash_next(&mpages, &pslot);
1224 }
1225
1226 if (0 == nodb)
1227 SQL_EXEC("END TRANSACTION");
1228 }
1229
1230 static void
1231 names_check(void)
1232 {
1233 sqlite3_stmt *stmt;
1234 const char *name, *sec, *arch, *key;
1235
1236 sqlite3_prepare_v2(db,
1237 "SELECT name, sec, arch, key FROM ("
1238 "SELECT name AS key, pageid FROM names "
1239 "WHERE bits & ? AND NOT EXISTS ("
1240 "SELECT pageid FROM mlinks "
1241 "WHERE mlinks.pageid == names.pageid "
1242 "AND mlinks.name == names.name"
1243 ")"
1244 ") JOIN ("
1245 "SELECT sec, arch, name, pageid FROM mlinks "
1246 "GROUP BY pageid"
1247 ") USING (pageid);",
1248 -1, &stmt, NULL);
1249
1250 if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK)
1251 say("", "%s", sqlite3_errmsg(db));
1252
1253 while (sqlite3_step(stmt) == SQLITE_ROW) {
1254 name = (const char *)sqlite3_column_text(stmt, 0);
1255 sec = (const char *)sqlite3_column_text(stmt, 1);
1256 arch = (const char *)sqlite3_column_text(stmt, 2);
1257 key = (const char *)sqlite3_column_text(stmt, 3);
1258 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1259 '\0' == *arch ? "" : "/",
1260 '\0' == *arch ? "" : arch, key);
1261 }
1262 sqlite3_finalize(stmt);
1263 }
1264
1265 static void
1266 parse_cat(struct mpage *mpage, int fd)
1267 {
1268 FILE *stream;
1269 char *line, *p, *title;
1270 size_t len, plen, titlesz;
1271
1272 stream = (-1 == fd) ?
1273 fopen(mpage->mlinks->file, "r") :
1274 fdopen(fd, "r");
1275 if (NULL == stream) {
1276 if (-1 != fd)
1277 close(fd);
1278 if (warnings)
1279 say(mpage->mlinks->file, "&fopen");
1280 return;
1281 }
1282
1283 /* Skip to first blank line. */
1284
1285 while (NULL != (line = fgetln(stream, &len)))
1286 if ('\n' == *line)
1287 break;
1288
1289 /*
1290 * Assume the first line that is not indented
1291 * is the first section header. Skip to it.
1292 */
1293
1294 while (NULL != (line = fgetln(stream, &len)))
1295 if ('\n' != *line && ' ' != *line)
1296 break;
1297
1298 /*
1299 * Read up until the next section into a buffer.
1300 * Strip the leading and trailing newline from each read line,
1301 * appending a trailing space.
1302 * Ignore empty (whitespace-only) lines.
1303 */
1304
1305 titlesz = 0;
1306 title = NULL;
1307
1308 while (NULL != (line = fgetln(stream, &len))) {
1309 if (' ' != *line || '\n' != line[len - 1])
1310 break;
1311 while (len > 0 && isspace((unsigned char)*line)) {
1312 line++;
1313 len--;
1314 }
1315 if (1 == len)
1316 continue;
1317 title = mandoc_realloc(title, titlesz + len);
1318 memcpy(title + titlesz, line, len);
1319 titlesz += len;
1320 title[titlesz - 1] = ' ';
1321 }
1322
1323 /*
1324 * If no page content can be found, or the input line
1325 * is already the next section header, or there is no
1326 * trailing newline, reuse the page title as the page
1327 * description.
1328 */
1329
1330 if (NULL == title || '\0' == *title) {
1331 if (warnings)
1332 say(mpage->mlinks->file,
1333 "Cannot find NAME section");
1334 fclose(stream);
1335 free(title);
1336 return;
1337 }
1338
1339 title = mandoc_realloc(title, titlesz + 1);
1340 title[titlesz] = '\0';
1341
1342 /*
1343 * Skip to the first dash.
1344 * Use the remaining line as the description (no more than 70
1345 * bytes).
1346 */
1347
1348 if (NULL != (p = strstr(title, "- "))) {
1349 for (p += 2; ' ' == *p || '\b' == *p; p++)
1350 /* Skip to next word. */ ;
1351 } else {
1352 if (warnings)
1353 say(mpage->mlinks->file,
1354 "No dash in title line");
1355 p = title;
1356 }
1357
1358 plen = strlen(p);
1359
1360 /* Strip backspace-encoding from line. */
1361
1362 while (NULL != (line = memchr(p, '\b', plen))) {
1363 len = line - p;
1364 if (0 == len) {
1365 memmove(line, line + 1, plen--);
1366 continue;
1367 }
1368 memmove(line - 1, line + 1, plen - len);
1369 plen -= 2;
1370 }
1371
1372 mpage->desc = mandoc_strdup(p);
1373 fclose(stream);
1374 free(title);
1375 }
1376
1377 /*
1378 * Put a type/word pair into the word database for this particular file.
1379 */
1380 static void
1381 putkey(const struct mpage *mpage, char *value, uint64_t type)
1382 {
1383 char *cp;
1384
1385 assert(NULL != value);
1386 if (TYPE_arch == type)
1387 for (cp = value; *cp; cp++)
1388 if (isupper((unsigned char)*cp))
1389 *cp = _tolower((unsigned char)*cp);
1390 putkeys(mpage, value, strlen(value), type);
1391 }
1392
1393 /*
1394 * Grok all nodes at or below a certain mdoc node into putkey().
1395 */
1396 static void
1397 putmdockey(const struct mpage *mpage,
1398 const struct roff_node *n, uint64_t m)
1399 {
1400
1401 for ( ; NULL != n; n = n->next) {
1402 if (NULL != n->child)
1403 putmdockey(mpage, n->child, m);
1404 if (n->type == ROFFT_TEXT)
1405 putkey(mpage, n->string, m);
1406 }
1407 }
1408
1409 static void
1410 parse_man(struct mpage *mpage, const struct roff_meta *meta,
1411 const struct roff_node *n)
1412 {
1413 const struct roff_node *head, *body;
1414 char *start, *title;
1415 char byte;
1416 size_t sz;
1417
1418 if (NULL == n)
1419 return;
1420
1421 /*
1422 * We're only searching for one thing: the first text child in
1423 * the BODY of a NAME section. Since we don't keep track of
1424 * sections in -man, run some hoops to find out whether we're in
1425 * the correct section or not.
1426 */
1427
1428 if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1429 body = n;
1430 assert(body->parent);
1431 if (NULL != (head = body->parent->head) &&
1432 1 == head->nchild &&
1433 NULL != (head = (head->child)) &&
1434 head->type == ROFFT_TEXT &&
1435 0 == strcmp(head->string, "NAME") &&
1436 NULL != body->child) {
1437
1438 /*
1439 * Suck the entire NAME section into memory.
1440 * Yes, we might run away.
1441 * But too many manuals have big, spread-out
1442 * NAME sections over many lines.
1443 */
1444
1445 title = NULL;
1446 deroff(&title, body);
1447 if (NULL == title)
1448 return;
1449
1450 /*
1451 * Go through a special heuristic dance here.
1452 * Conventionally, one or more manual names are
1453 * comma-specified prior to a whitespace, then a
1454 * dash, then a description. Try to puzzle out
1455 * the name parts here.
1456 */
1457
1458 start = title;
1459 for ( ;; ) {
1460 sz = strcspn(start, " ,");
1461 if ('\0' == start[sz])
1462 break;
1463
1464 byte = start[sz];
1465 start[sz] = '\0';
1466
1467 /*
1468 * Assume a stray trailing comma in the
1469 * name list if a name begins with a dash.
1470 */
1471
1472 if ('-' == start[0] ||
1473 ('\\' == start[0] && '-' == start[1]))
1474 break;
1475
1476 putkey(mpage, start, NAME_TITLE);
1477 if ( ! (mpage->name_head_done ||
1478 strcasecmp(start, meta->title))) {
1479 putkey(mpage, start, NAME_HEAD);
1480 mpage->name_head_done = 1;
1481 }
1482
1483 if (' ' == byte) {
1484 start += sz + 1;
1485 break;
1486 }
1487
1488 assert(',' == byte);
1489 start += sz + 1;
1490 while (' ' == *start)
1491 start++;
1492 }
1493
1494 if (start == title) {
1495 putkey(mpage, start, NAME_TITLE);
1496 if ( ! (mpage->name_head_done ||
1497 strcasecmp(start, meta->title))) {
1498 putkey(mpage, start, NAME_HEAD);
1499 mpage->name_head_done = 1;
1500 }
1501 free(title);
1502 return;
1503 }
1504
1505 while (isspace((unsigned char)*start))
1506 start++;
1507
1508 if (0 == strncmp(start, "-", 1))
1509 start += 1;
1510 else if (0 == strncmp(start, "\\-\\-", 4))
1511 start += 4;
1512 else if (0 == strncmp(start, "\\-", 2))
1513 start += 2;
1514 else if (0 == strncmp(start, "\\(en", 4))
1515 start += 4;
1516 else if (0 == strncmp(start, "\\(em", 4))
1517 start += 4;
1518
1519 while (' ' == *start)
1520 start++;
1521
1522 mpage->desc = mandoc_strdup(start);
1523 free(title);
1524 return;
1525 }
1526 }
1527
1528 for (n = n->child; n; n = n->next) {
1529 if (NULL != mpage->desc)
1530 break;
1531 parse_man(mpage, meta, n);
1532 }
1533 }
1534
1535 static void
1536 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1537 const struct roff_node *n)
1538 {
1539
1540 assert(NULL != n);
1541 for (n = n->child; NULL != n; n = n->next) {
1542 switch (n->type) {
1543 case ROFFT_ELEM:
1544 case ROFFT_BLOCK:
1545 case ROFFT_HEAD:
1546 case ROFFT_BODY:
1547 case ROFFT_TAIL:
1548 if (NULL != mdocs[n->tok].fp)
1549 if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1550 break;
1551 if (mdocs[n->tok].mask)
1552 putmdockey(mpage, n->child,
1553 mdocs[n->tok].mask);
1554 break;
1555 default:
1556 assert(n->type != ROFFT_ROOT);
1557 continue;
1558 }
1559 if (NULL != n->child)
1560 parse_mdoc(mpage, meta, n);
1561 }
1562 }
1563
1564 static int
1565 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1566 const struct roff_node *n)
1567 {
1568 char *start, *end;
1569 size_t sz;
1570
1571 if (SEC_SYNOPSIS != n->sec ||
1572 NULL == (n = n->child) ||
1573 n->type != ROFFT_TEXT)
1574 return 0;
1575
1576 /*
1577 * Only consider those `Fd' macro fields that begin with an
1578 * "inclusion" token (versus, e.g., #define).
1579 */
1580
1581 if (strcmp("#include", n->string))
1582 return 0;
1583
1584 if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1585 return 0;
1586
1587 /*
1588 * Strip away the enclosing angle brackets and make sure we're
1589 * not zero-length.
1590 */
1591
1592 start = n->string;
1593 if ('<' == *start || '"' == *start)
1594 start++;
1595
1596 if (0 == (sz = strlen(start)))
1597 return 0;
1598
1599 end = &start[(int)sz - 1];
1600 if ('>' == *end || '"' == *end)
1601 end--;
1602
1603 if (end > start)
1604 putkeys(mpage, start, end - start + 1, TYPE_In);
1605 return 0;
1606 }
1607
1608 static void
1609 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1610 {
1611 char *cp;
1612 size_t sz;
1613
1614 if (n->type != ROFFT_TEXT)
1615 return;
1616
1617 /* Skip function pointer punctuation. */
1618
1619 cp = n->string;
1620 while (*cp == '(' || *cp == '*')
1621 cp++;
1622 sz = strcspn(cp, "()");
1623
1624 putkeys(mpage, cp, sz, TYPE_Fn);
1625 if (n->sec == SEC_SYNOPSIS)
1626 putkeys(mpage, cp, sz, NAME_SYN);
1627 }
1628
1629 static int
1630 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1631 const struct roff_node *n)
1632 {
1633
1634 if (n->child == NULL)
1635 return 0;
1636
1637 parse_mdoc_fname(mpage, n->child);
1638
1639 for (n = n->child->next; n != NULL; n = n->next)
1640 if (n->type == ROFFT_TEXT)
1641 putkey(mpage, n->string, TYPE_Fa);
1642
1643 return 0;
1644 }
1645
1646 static int
1647 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1648 const struct roff_node *n)
1649 {
1650
1651 if (n->type != ROFFT_HEAD)
1652 return 1;
1653
1654 if (n->child != NULL)
1655 parse_mdoc_fname(mpage, n->child);
1656
1657 return 0;
1658 }
1659
1660 static int
1661 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1662 const struct roff_node *n)
1663 {
1664 char *cp;
1665
1666 if (NULL == (n = n->child))
1667 return 0;
1668
1669 if (NULL == n->next) {
1670 putkey(mpage, n->string, TYPE_Xr);
1671 return 0;
1672 }
1673
1674 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1675 putkey(mpage, cp, TYPE_Xr);
1676 free(cp);
1677 return 0;
1678 }
1679
1680 static int
1681 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1682 const struct roff_node *n)
1683 {
1684
1685 if (n->type == ROFFT_BODY)
1686 deroff(&mpage->desc, n);
1687 return 0;
1688 }
1689
1690 static int
1691 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1692 const struct roff_node *n)
1693 {
1694
1695 if (SEC_NAME == n->sec)
1696 putmdockey(mpage, n->child, NAME_TITLE);
1697 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1698 if (n->child == NULL)
1699 putkey(mpage, meta->name, NAME_SYN);
1700 else
1701 putmdockey(mpage, n->child, NAME_SYN);
1702 }
1703 if ( ! (mpage->name_head_done ||
1704 n->child == NULL || n->child->string == NULL ||
1705 strcasecmp(n->child->string, meta->title))) {
1706 putkey(mpage, n->child->string, ROFFT_HEAD);
1707 mpage->name_head_done = 1;
1708 }
1709 return 0;
1710 }
1711
1712 static int
1713 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1714 const struct roff_node *n)
1715 {
1716
1717 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1718 }
1719
1720 static int
1721 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1722 const struct roff_node *n)
1723 {
1724
1725 return n->type == ROFFT_HEAD;
1726 }
1727
1728 static int
1729 parse_mdoc_body(struct mpage *mpage, const struct roff_meta *meta,
1730 const struct roff_node *n)
1731 {
1732
1733 return n->type == ROFFT_BODY;
1734 }
1735
1736 /*
1737 * Add a string to the hash table for the current manual.
1738 * Each string has a bitmask telling which macros it belongs to.
1739 * When we finish the manual, we'll dump the table.
1740 */
1741 static void
1742 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1743 {
1744 struct ohash *htab;
1745 struct str *s;
1746 const char *end;
1747 unsigned int slot;
1748 int i, mustfree;
1749
1750 if (0 == sz)
1751 return;
1752
1753 mustfree = render_string(&cp, &sz);
1754
1755 if (TYPE_Nm & v) {
1756 htab = &names;
1757 v &= name_mask;
1758 if (v & NAME_FIRST)
1759 name_mask &= ~NAME_FIRST;
1760 if (debug > 1)
1761 say(mpage->mlinks->file,
1762 "Adding name %*s, bits=%d", sz, cp, v);
1763 } else {
1764 htab = &strings;
1765 if (debug > 1)
1766 for (i = 0; i < mansearch_keymax; i++)
1767 if ((uint64_t)1 << i & v)
1768 say(mpage->mlinks->file,
1769 "Adding key %s=%*s",
1770 mansearch_keynames[i], sz, cp);
1771 }
1772
1773 end = cp + sz;
1774 slot = ohash_qlookupi(htab, cp, &end);
1775 s = ohash_find(htab, slot);
1776
1777 if (NULL != s && mpage == s->mpage) {
1778 s->mask |= v;
1779 return;
1780 } else if (NULL == s) {
1781 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1782 memcpy(s->key, cp, sz);
1783 ohash_insert(htab, slot, s);
1784 }
1785 s->mpage = mpage;
1786 s->mask = v;
1787
1788 if (mustfree)
1789 free(cp);
1790 }
1791
1792 /*
1793 * Take a Unicode codepoint and produce its UTF-8 encoding.
1794 * This isn't the best way to do this, but it works.
1795 * The magic numbers are from the UTF-8 packaging.
1796 * They're not as scary as they seem: read the UTF-8 spec for details.
1797 */
1798 static size_t
1799 utf8(unsigned int cp, char out[7])
1800 {
1801 size_t rc;
1802
1803 rc = 0;
1804 if (cp <= 0x0000007F) {
1805 rc = 1;
1806 out[0] = (char)cp;
1807 } else if (cp <= 0x000007FF) {
1808 rc = 2;
1809 out[0] = (cp >> 6 & 31) | 192;
1810 out[1] = (cp & 63) | 128;
1811 } else if (cp <= 0x0000FFFF) {
1812 rc = 3;
1813 out[0] = (cp >> 12 & 15) | 224;
1814 out[1] = (cp >> 6 & 63) | 128;
1815 out[2] = (cp & 63) | 128;
1816 } else if (cp <= 0x001FFFFF) {
1817 rc = 4;
1818 out[0] = (cp >> 18 & 7) | 240;
1819 out[1] = (cp >> 12 & 63) | 128;
1820 out[2] = (cp >> 6 & 63) | 128;
1821 out[3] = (cp & 63) | 128;
1822 } else if (cp <= 0x03FFFFFF) {
1823 rc = 5;
1824 out[0] = (cp >> 24 & 3) | 248;
1825 out[1] = (cp >> 18 & 63) | 128;
1826 out[2] = (cp >> 12 & 63) | 128;
1827 out[3] = (cp >> 6 & 63) | 128;
1828 out[4] = (cp & 63) | 128;
1829 } else if (cp <= 0x7FFFFFFF) {
1830 rc = 6;
1831 out[0] = (cp >> 30 & 1) | 252;
1832 out[1] = (cp >> 24 & 63) | 128;
1833 out[2] = (cp >> 18 & 63) | 128;
1834 out[3] = (cp >> 12 & 63) | 128;
1835 out[4] = (cp >> 6 & 63) | 128;
1836 out[5] = (cp & 63) | 128;
1837 } else
1838 return 0;
1839
1840 out[rc] = '\0';
1841 return rc;
1842 }
1843
1844 /*
1845 * If the string contains escape sequences,
1846 * replace it with an allocated rendering and return 1,
1847 * such that the caller can free it after use.
1848 * Otherwise, do nothing and return 0.
1849 */
1850 static int
1851 render_string(char **public, size_t *psz)
1852 {
1853 const char *src, *scp, *addcp, *seq;
1854 char *dst;
1855 size_t ssz, dsz, addsz;
1856 char utfbuf[7], res[6];
1857 int seqlen, unicode;
1858
1859 res[0] = '\\';
1860 res[1] = '\t';
1861 res[2] = ASCII_NBRSP;
1862 res[3] = ASCII_HYPH;
1863 res[4] = ASCII_BREAK;
1864 res[5] = '\0';
1865
1866 src = scp = *public;
1867 ssz = *psz;
1868 dst = NULL;
1869 dsz = 0;
1870
1871 while (scp < src + *psz) {
1872
1873 /* Leave normal characters unchanged. */
1874
1875 if (strchr(res, *scp) == NULL) {
1876 if (dst != NULL)
1877 dst[dsz++] = *scp;
1878 scp++;
1879 continue;
1880 }
1881
1882 /*
1883 * Found something that requires replacing,
1884 * make sure we have a destination buffer.
1885 */
1886
1887 if (dst == NULL) {
1888 dst = mandoc_malloc(ssz + 1);
1889 dsz = scp - src;
1890 memcpy(dst, src, dsz);
1891 }
1892
1893 /* Handle single-char special characters. */
1894
1895 switch (*scp) {
1896 case '\\':
1897 break;
1898 case '\t':
1899 case ASCII_NBRSP:
1900 dst[dsz++] = ' ';
1901 scp++;
1902 continue;
1903 case ASCII_HYPH:
1904 dst[dsz++] = '-';
1905 /* FALLTHROUGH */
1906 case ASCII_BREAK:
1907 scp++;
1908 continue;
1909 default:
1910 abort();
1911 }
1912
1913 /*
1914 * Found an escape sequence.
1915 * Read past the slash, then parse it.
1916 * Ignore everything except characters.
1917 */
1918
1919 scp++;
1920 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1921 continue;
1922
1923 /*
1924 * Render the special character
1925 * as either UTF-8 or ASCII.
1926 */
1927
1928 if (write_utf8) {
1929 unicode = mchars_spec2cp(seq, seqlen);
1930 if (unicode <= 0)
1931 continue;
1932 addsz = utf8(unicode, utfbuf);
1933 if (addsz == 0)
1934 continue;
1935 addcp = utfbuf;
1936 } else {
1937 addcp = mchars_spec2str(seq, seqlen, &addsz);
1938 if (addcp == NULL)
1939 continue;
1940 if (*addcp == ASCII_NBRSP) {
1941 addcp = " ";
1942 addsz = 1;
1943 }
1944 }
1945
1946 /* Copy the rendered glyph into the stream. */
1947
1948 ssz += addsz;
1949 dst = mandoc_realloc(dst, ssz + 1);
1950 memcpy(dst + dsz, addcp, addsz);
1951 dsz += addsz;
1952 }
1953 if (dst != NULL) {
1954 *public = dst;
1955 *psz = dsz;
1956 }
1957
1958 /* Trim trailing whitespace and NUL-terminate. */
1959
1960 while (*psz > 0 && (*public)[*psz - 1] == ' ')
1961 --*psz;
1962 if (dst != NULL) {
1963 (*public)[*psz] = '\0';
1964 return 1;
1965 } else
1966 return 0;
1967 }
1968
1969 static void
1970 dbadd_mlink(const struct mlink *mlink)
1971 {
1972 size_t i;
1973
1974 i = 1;
1975 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1976 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1977 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1978 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1979 SQL_STEP(stmts[STMT_INSERT_LINK]);
1980 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1981 }
1982
1983 static void
1984 dbadd_mlink_name(const struct mlink *mlink)
1985 {
1986 uint64_t bits;
1987 size_t i;
1988
1989 dbadd_mlink(mlink);
1990
1991 i = 1;
1992 SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
1993 bits = NAME_FILE & NAME_MASK;
1994 if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
1995 bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
1996 sqlite3_reset(stmts[STMT_SELECT_NAME]);
1997 }
1998
1999 i = 1;
2000 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
2001 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2002 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2003 SQL_STEP(stmts[STMT_INSERT_NAME]);
2004 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2005 }
2006
2007 /*
2008 * Flush the current page's terms (and their bits) into the database.
2009 * Wrap the entire set of additions in a transaction to make sqlite be a
2010 * little faster.
2011 * Also, handle escape sequences at the last possible moment.
2012 */
2013 static void
2014 dbadd(struct mpage *mpage)
2015 {
2016 struct mlink *mlink;
2017 struct str *key;
2018 char *cp;
2019 size_t i;
2020 unsigned int slot;
2021 int mustfree;
2022
2023 mlink = mpage->mlinks;
2024
2025 if (nodb) {
2026 for (key = ohash_first(&names, &slot); NULL != key;
2027 key = ohash_next(&names, &slot))
2028 free(key);
2029 for (key = ohash_first(&strings, &slot); NULL != key;
2030 key = ohash_next(&strings, &slot))
2031 free(key);
2032 if (0 == debug)
2033 return;
2034 while (NULL != mlink) {
2035 fputs(mlink->name, stdout);
2036 if (NULL == mlink->next ||
2037 strcmp(mlink->dsec, mlink->next->dsec) ||
2038 strcmp(mlink->fsec, mlink->next->fsec) ||
2039 strcmp(mlink->arch, mlink->next->arch)) {
2040 putchar('(');
2041 if ('\0' == *mlink->dsec)
2042 fputs(mlink->fsec, stdout);
2043 else
2044 fputs(mlink->dsec, stdout);
2045 if ('\0' != *mlink->arch)
2046 printf("/%s", mlink->arch);
2047 putchar(')');
2048 }
2049 mlink = mlink->next;
2050 if (NULL != mlink)
2051 fputs(", ", stdout);
2052 }
2053 printf(" - %s\n", mpage->desc);
2054 return;
2055 }
2056
2057 if (debug)
2058 say(mlink->file, "Adding to database");
2059
2060 cp = mpage->desc;
2061 i = strlen(cp);
2062 mustfree = render_string(&cp, &i);
2063 i = 1;
2064 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
2065 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
2066 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2067 mpage->pageid = sqlite3_last_insert_rowid(db);
2068 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2069 if (mustfree)
2070 free(cp);
2071
2072 while (NULL != mlink) {
2073 dbadd_mlink(mlink);
2074 mlink = mlink->next;
2075 }
2076 mlink = mpage->mlinks;
2077
2078 for (key = ohash_first(&names, &slot); NULL != key;
2079 key = ohash_next(&names, &slot)) {
2080 assert(key->mpage == mpage);
2081 i = 1;
2082 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2083 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
2084 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2085 SQL_STEP(stmts[STMT_INSERT_NAME]);
2086 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2087 free(key);
2088 }
2089 for (key = ohash_first(&strings, &slot); NULL != key;
2090 key = ohash_next(&strings, &slot)) {
2091 assert(key->mpage == mpage);
2092 i = 1;
2093 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2094 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
2095 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2096 SQL_STEP(stmts[STMT_INSERT_KEY]);
2097 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2098 free(key);
2099 }
2100 }
2101
2102 static void
2103 dbprune(void)
2104 {
2105 struct mpage *mpage;
2106 struct mlink *mlink;
2107 size_t i;
2108 unsigned int slot;
2109
2110 if (0 == nodb)
2111 SQL_EXEC("BEGIN TRANSACTION");
2112
2113 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2114 mpage = ohash_next(&mpages, &slot)) {
2115 mlink = mpage->mlinks;
2116 if (debug)
2117 say(mlink->file, "Deleting from database");
2118 if (nodb)
2119 continue;
2120 for ( ; NULL != mlink; mlink = mlink->next) {
2121 i = 1;
2122 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2123 i, mlink->dsec);
2124 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2125 i, mlink->arch);
2126 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2127 i, mlink->name);
2128 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2129 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2130 }
2131 }
2132
2133 if (0 == nodb)
2134 SQL_EXEC("END TRANSACTION");
2135 }
2136
2137 /*
2138 * Close an existing database and its prepared statements.
2139 * If "real" is not set, rename the temporary file into the real one.
2140 */
2141 static void
2142 dbclose(int real)
2143 {
2144 size_t i;
2145 int status;
2146 pid_t child;
2147
2148 if (nodb)
2149 return;
2150
2151 for (i = 0; i < STMT__MAX; i++) {
2152 sqlite3_finalize(stmts[i]);
2153 stmts[i] = NULL;
2154 }
2155
2156 sqlite3_close(db);
2157 db = NULL;
2158
2159 if (real)
2160 return;
2161
2162 if ('\0' == *tempfilename) {
2163 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2164 exitcode = (int)MANDOCLEVEL_SYSERR;
2165 say(MANDOC_DB, "&rename");
2166 }
2167 return;
2168 }
2169
2170 switch (child = fork()) {
2171 case -1:
2172 exitcode = (int)MANDOCLEVEL_SYSERR;
2173 say("", "&fork cmp");
2174 return;
2175 case 0:
2176 execlp("cmp", "cmp", "-s",
2177 tempfilename, MANDOC_DB, (char *)NULL);
2178 say("", "&exec cmp");
2179 exit(0);
2180 default:
2181 break;
2182 }
2183 if (-1 == waitpid(child, &status, 0)) {
2184 exitcode = (int)MANDOCLEVEL_SYSERR;
2185 say("", "&wait cmp");
2186 } else if (WIFSIGNALED(status)) {
2187 exitcode = (int)MANDOCLEVEL_SYSERR;
2188 say("", "cmp died from signal %d", WTERMSIG(status));
2189 } else if (WEXITSTATUS(status)) {
2190 exitcode = (int)MANDOCLEVEL_SYSERR;
2191 say(MANDOC_DB,
2192 "Data changed, but cannot replace database");
2193 }
2194
2195 *strrchr(tempfilename, '/') = '\0';
2196 switch (child = fork()) {
2197 case -1:
2198 exitcode = (int)MANDOCLEVEL_SYSERR;
2199 say("", "&fork rm");
2200 return;
2201 case 0:
2202 execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
2203 say("", "&exec rm");
2204 exit((int)MANDOCLEVEL_SYSERR);
2205 default:
2206 break;
2207 }
2208 if (-1 == waitpid(child, &status, 0)) {
2209 exitcode = (int)MANDOCLEVEL_SYSERR;
2210 say("", "&wait rm");
2211 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2212 exitcode = (int)MANDOCLEVEL_SYSERR;
2213 say("", "%s: Cannot remove temporary directory",
2214 tempfilename);
2215 }
2216 }
2217
2218 /*
2219 * This is straightforward stuff.
2220 * Open a database connection to a "temporary" database, then open a set
2221 * of prepared statements we'll use over and over again.
2222 * If "real" is set, we use the existing database; if not, we truncate a
2223 * temporary one.
2224 * Must be matched by dbclose().
2225 */
2226 static int
2227 dbopen(int real)
2228 {
2229 const char *sql;
2230 int rc, ofl;
2231
2232 if (nodb)
2233 return 1;
2234
2235 *tempfilename = '\0';
2236 ofl = SQLITE_OPEN_READWRITE;
2237
2238 if (real) {
2239 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2240 if (SQLITE_OK != rc) {
2241 exitcode = (int)MANDOCLEVEL_SYSERR;
2242 if (SQLITE_CANTOPEN != rc)
2243 say(MANDOC_DB, "%s", sqlite3_errstr(rc));
2244 return 0;
2245 }
2246 goto prepare_statements;
2247 }
2248
2249 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2250
2251 remove(MANDOC_DB "~");
2252 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2253 if (SQLITE_OK == rc)
2254 goto create_tables;
2255 if (MPARSE_QUICK & mparse_options) {
2256 exitcode = (int)MANDOCLEVEL_SYSERR;
2257 say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
2258 return 0;
2259 }
2260
2261 (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2262 sizeof(tempfilename));
2263 if (NULL == mkdtemp(tempfilename)) {
2264 exitcode = (int)MANDOCLEVEL_SYSERR;
2265 say("", "&%s", tempfilename);
2266 return 0;
2267 }
2268 (void)strlcat(tempfilename, "/" MANDOC_DB,
2269 sizeof(tempfilename));
2270 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2271 if (SQLITE_OK != rc) {
2272 exitcode = (int)MANDOCLEVEL_SYSERR;
2273 say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
2274 return 0;
2275 }
2276
2277 create_tables:
2278 sql = "CREATE TABLE \"mpages\" (\n"
2279 " \"desc\" TEXT NOT NULL,\n"
2280 " \"form\" INTEGER NOT NULL,\n"
2281 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2282 ");\n"
2283 "\n"
2284 "CREATE TABLE \"mlinks\" (\n"
2285 " \"sec\" TEXT NOT NULL,\n"
2286 " \"arch\" TEXT NOT NULL,\n"
2287 " \"name\" TEXT NOT NULL,\n"
2288 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2289 "ON DELETE CASCADE\n"
2290 ");\n"
2291 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2292 "\n"
2293 "CREATE TABLE \"names\" (\n"
2294 " \"bits\" INTEGER NOT NULL,\n"
2295 " \"name\" TEXT NOT NULL,\n"
2296 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2297 "ON DELETE CASCADE,\n"
2298 " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
2299 ");\n"
2300 "\n"
2301 "CREATE TABLE \"keys\" (\n"
2302 " \"bits\" INTEGER NOT NULL,\n"
2303 " \"key\" TEXT NOT NULL,\n"
2304 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2305 "ON DELETE CASCADE\n"
2306 ");\n"
2307 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2308
2309 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2310 exitcode = (int)MANDOCLEVEL_SYSERR;
2311 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2312 sqlite3_close(db);
2313 return 0;
2314 }
2315
2316 prepare_statements:
2317 if (SQLITE_OK != sqlite3_exec(db,
2318 "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2319 exitcode = (int)MANDOCLEVEL_SYSERR;
2320 say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2321 sqlite3_errmsg(db));
2322 sqlite3_close(db);
2323 return 0;
2324 }
2325
2326 sql = "DELETE FROM mpages WHERE pageid IN "
2327 "(SELECT pageid FROM mlinks WHERE "
2328 "sec=? AND arch=? AND name=?)";
2329 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2330 sql = "INSERT INTO mpages "
2331 "(desc,form) VALUES (?,?)";
2332 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2333 sql = "INSERT INTO mlinks "
2334 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2335 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2336 sql = "SELECT bits FROM names where pageid = ?";
2337 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
2338 sql = "INSERT INTO names "
2339 "(bits,name,pageid) VALUES (?,?,?)";
2340 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2341 sql = "INSERT INTO keys "
2342 "(bits,key,pageid) VALUES (?,?,?)";
2343 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2344
2345 #ifndef __APPLE__
2346 /*
2347 * When opening a new database, we can turn off
2348 * synchronous mode for much better performance.
2349 */
2350
2351 if (real && SQLITE_OK != sqlite3_exec(db,
2352 "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2353 exitcode = (int)MANDOCLEVEL_SYSERR;
2354 say(MANDOC_DB, "PRAGMA synchronous: %s",
2355 sqlite3_errmsg(db));
2356 sqlite3_close(db);
2357 return 0;
2358 }
2359 #endif
2360
2361 return 1;
2362 }
2363
2364 static int
2365 set_basedir(const char *targetdir, int report_baddir)
2366 {
2367 static char startdir[PATH_MAX];
2368 static int getcwd_status; /* 1 = ok, 2 = failure */
2369 static int chdir_status; /* 1 = changed directory */
2370 char *cp;
2371
2372 /*
2373 * Remember the original working directory, if possible.
2374 * This will be needed if the second or a later directory
2375 * on the command line is given as a relative path.
2376 * Do not error out if the current directory is not
2377 * searchable: Maybe it won't be needed after all.
2378 */
2379 if (0 == getcwd_status) {
2380 if (NULL == getcwd(startdir, sizeof(startdir))) {
2381 getcwd_status = 2;
2382 (void)strlcpy(startdir, strerror(errno),
2383 sizeof(startdir));
2384 } else
2385 getcwd_status = 1;
2386 }
2387
2388 /*
2389 * We are leaving the old base directory.
2390 * Do not use it any longer, not even for messages.
2391 */
2392 *basedir = '\0';
2393
2394 /*
2395 * If and only if the directory was changed earlier and
2396 * the next directory to process is given as a relative path,
2397 * first go back, or bail out if that is impossible.
2398 */
2399 if (chdir_status && '/' != *targetdir) {
2400 if (2 == getcwd_status) {
2401 exitcode = (int)MANDOCLEVEL_SYSERR;
2402 say("", "getcwd: %s", startdir);
2403 return 0;
2404 }
2405 if (-1 == chdir(startdir)) {
2406 exitcode = (int)MANDOCLEVEL_SYSERR;
2407 say("", "&chdir %s", startdir);
2408 return 0;
2409 }
2410 }
2411
2412 /*
2413 * Always resolve basedir to the canonicalized absolute
2414 * pathname and append a trailing slash, such that
2415 * we can reliably check whether files are inside.
2416 */
2417 if (NULL == realpath(targetdir, basedir)) {
2418 if (report_baddir || errno != ENOENT) {
2419 exitcode = (int)MANDOCLEVEL_BADARG;
2420 say("", "&%s: realpath", targetdir);
2421 }
2422 return 0;
2423 } else if (-1 == chdir(basedir)) {
2424 if (report_baddir || errno != ENOENT) {
2425 exitcode = (int)MANDOCLEVEL_BADARG;
2426 say("", "&chdir");
2427 }
2428 return 0;
2429 }
2430 chdir_status = 1;
2431 cp = strchr(basedir, '\0');
2432 if ('/' != cp[-1]) {
2433 if (cp - basedir >= PATH_MAX - 1) {
2434 exitcode = (int)MANDOCLEVEL_SYSERR;
2435 say("", "Filename too long");
2436 return 0;
2437 }
2438 *cp++ = '/';
2439 *cp = '\0';
2440 }
2441 return 1;
2442 }
2443
2444 static void
2445 say(const char *file, const char *format, ...)
2446 {
2447 va_list ap;
2448 int use_errno;
2449
2450 if ('\0' != *basedir)
2451 fprintf(stderr, "%s", basedir);
2452 if ('\0' != *basedir && '\0' != *file)
2453 fputc('/', stderr);
2454 if ('\0' != *file)
2455 fprintf(stderr, "%s", file);
2456
2457 use_errno = 1;
2458 if (NULL != format) {
2459 switch (*format) {
2460 case '&':
2461 format++;
2462 break;
2463 case '\0':
2464 format = NULL;
2465 break;
2466 default:
2467 use_errno = 0;
2468 break;
2469 }
2470 }
2471 if (NULL != format) {
2472 if ('\0' != *basedir || '\0' != *file)
2473 fputs(": ", stderr);
2474 va_start(ap, format);
2475 vfprintf(stderr, format, ap);
2476 va_end(ap);
2477 }
2478 if (use_errno) {
2479 if ('\0' != *basedir || '\0' != *file || NULL != format)
2480 fputs(": ", stderr);
2481 perror(NULL);
2482 } else
2483 fputc('\n', stderr);
2484 }
A git conversion mirror of mandoc.bsd.lv