Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.33
1.33 ! schwarze 1: /* $Id: mandocdb.c,v 1.32 2011/12/26 11:44:31 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.10 schwarze 4: * Copyright (c) 2011 Ingo Schwarze <schwarze@openbsd.org>
1.1 schwarze 5: *
6: * Permission to use, copy, modify, and distribute this software for any
7: * purpose with or without fee is hereby granted, provided that the above
8: * copyright notice and this permission notice appear in all copies.
9: *
10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17: */
18: #include <sys/param.h>
1.11 schwarze 19: #include <sys/types.h>
1.1 schwarze 20:
21: #include <assert.h>
1.33 ! schwarze 22: #include <ctype.h>
1.2 schwarze 23: #include <dirent.h>
1.1 schwarze 24: #include <fcntl.h>
25: #include <getopt.h>
26: #include <stdio.h>
27: #include <stdint.h>
28: #include <stdlib.h>
29: #include <string.h>
1.14 schwarze 30: #include <unistd.h>
1.1 schwarze 31: #include <db.h>
32:
33: #include "man.h"
34: #include "mdoc.h"
35: #include "mandoc.h"
1.5 schwarze 36: #include "mandocdb.h"
1.10 schwarze 37: #include "manpath.h"
1.1 schwarze 38:
39: #define MANDOC_BUFSZ BUFSIZ
40: #define MANDOC_SLOP 1024
41:
1.11 schwarze 42: #define MANDOC_SRC 0x1
43: #define MANDOC_FORM 0x2
44:
1.28 schwarze 45: /* Access to the mandoc database on disk. */
46:
47: struct mdb {
48: char idxn[MAXPATHLEN]; /* index db filename */
49: char dbn[MAXPATHLEN]; /* keyword db filename */
50: DB *idx; /* index recno database */
51: DB *db; /* keyword btree database */
52: };
53:
54: /* Stack of temporarily unused index records. */
55:
56: struct recs {
57: recno_t *stack; /* pointer to a malloc'ed array */
58: size_t size; /* number of allocated slots */
59: size_t cur; /* current number of empty records */
60: recno_t last; /* last record number in the index */
61: };
62:
1.2 schwarze 63: /* Tiny list for files. No need to bring in QUEUE. */
64:
65: struct of {
66: char *fname; /* heap-allocated */
1.6 schwarze 67: char *sec;
68: char *arch;
69: char *title;
1.11 schwarze 70: int src_form;
1.2 schwarze 71: struct of *next; /* NULL for last one */
72: struct of *first; /* first in list */
73: };
74:
1.1 schwarze 75: /* Buffer for storing growable data. */
76:
77: struct buf {
78: char *cp;
1.2 schwarze 79: size_t len; /* current length */
80: size_t size; /* total buffer size */
1.1 schwarze 81: };
82:
83: /* Operation we're going to perform. */
84:
85: enum op {
1.28 schwarze 86: OP_DEFAULT = 0, /* new dbs from dir list or default config */
87: OP_CONFFILE, /* new databases from custom config file */
1.2 schwarze 88: OP_UPDATE, /* delete/add entries in existing database */
1.28 schwarze 89: OP_DELETE, /* delete entries from existing database */
90: OP_TEST /* change no databases, report potential problems */
1.1 schwarze 91: };
92:
93: #define MAN_ARGS DB *hash, \
94: struct buf *buf, \
95: struct buf *dbuf, \
96: const struct man_node *n
97: #define MDOC_ARGS DB *hash, \
98: struct buf *buf, \
99: struct buf *dbuf, \
100: const struct mdoc_node *n, \
101: const struct mdoc_meta *m
102:
103: static void buf_appendmdoc(struct buf *,
104: const struct mdoc_node *, int);
105: static void buf_append(struct buf *, const char *);
106: static void buf_appendb(struct buf *,
107: const void *, size_t);
108: static void dbt_put(DB *, const char *, DBT *, DBT *);
1.9 schwarze 109: static void hash_put(DB *, const struct buf *, uint64_t);
1.1 schwarze 110: static void hash_reset(DB **);
1.2 schwarze 111: static void index_merge(const struct of *, struct mparse *,
1.13 schwarze 112: struct buf *, struct buf *, DB *,
1.28 schwarze 113: struct mdb *, struct recs *);
114: static void index_prune(const struct of *, struct mdb *,
115: struct recs *);
1.13 schwarze 116: static void ofile_argbuild(int, char *[], struct of **);
1.26 schwarze 117: static void ofile_dirbuild(const char *, const char *,
1.13 schwarze 118: const char *, int, struct of **);
1.2 schwarze 119: static void ofile_free(struct of *);
1.33 ! schwarze 120: static void pformatted(DB *, struct buf *,
! 121: struct buf *, const struct of *);
1.1 schwarze 122: static int pman_node(MAN_ARGS);
123: static void pmdoc_node(MDOC_ARGS);
1.19 schwarze 124: static int pmdoc_head(MDOC_ARGS);
125: static int pmdoc_body(MDOC_ARGS);
126: static int pmdoc_Fd(MDOC_ARGS);
127: static int pmdoc_In(MDOC_ARGS);
128: static int pmdoc_Fn(MDOC_ARGS);
129: static int pmdoc_Nd(MDOC_ARGS);
130: static int pmdoc_Nm(MDOC_ARGS);
131: static int pmdoc_Sh(MDOC_ARGS);
132: static int pmdoc_St(MDOC_ARGS);
133: static int pmdoc_Xr(MDOC_ARGS);
1.1 schwarze 134:
1.19 schwarze 135: #define MDOCF_CHILD 0x01 /* Automatically index child nodes. */
1.1 schwarze 136:
1.19 schwarze 137: struct mdoc_handler {
138: int (*fp)(MDOC_ARGS); /* Optional handler. */
139: uint64_t mask; /* Set unless handler returns 0. */
140: int flags; /* For use by pmdoc_node. */
141: };
142:
143: static const struct mdoc_handler mdocs[MDOC_MAX] = {
144: { NULL, 0, 0 }, /* Ap */
145: { NULL, 0, 0 }, /* Dd */
146: { NULL, 0, 0 }, /* Dt */
147: { NULL, 0, 0 }, /* Os */
148: { pmdoc_Sh, TYPE_Sh, MDOCF_CHILD }, /* Sh */
149: { pmdoc_head, TYPE_Ss, MDOCF_CHILD }, /* Ss */
150: { NULL, 0, 0 }, /* Pp */
151: { NULL, 0, 0 }, /* D1 */
152: { NULL, 0, 0 }, /* Dl */
153: { NULL, 0, 0 }, /* Bd */
154: { NULL, 0, 0 }, /* Ed */
155: { NULL, 0, 0 }, /* Bl */
156: { NULL, 0, 0 }, /* El */
157: { NULL, 0, 0 }, /* It */
158: { NULL, 0, 0 }, /* Ad */
159: { NULL, TYPE_An, MDOCF_CHILD }, /* An */
160: { NULL, TYPE_Ar, MDOCF_CHILD }, /* Ar */
161: { NULL, TYPE_Cd, MDOCF_CHILD }, /* Cd */
162: { NULL, TYPE_Cm, MDOCF_CHILD }, /* Cm */
163: { NULL, TYPE_Dv, MDOCF_CHILD }, /* Dv */
164: { NULL, TYPE_Er, MDOCF_CHILD }, /* Er */
165: { NULL, TYPE_Ev, MDOCF_CHILD }, /* Ev */
166: { NULL, 0, 0 }, /* Ex */
167: { NULL, TYPE_Fa, MDOCF_CHILD }, /* Fa */
168: { pmdoc_Fd, TYPE_In, 0 }, /* Fd */
169: { NULL, TYPE_Fl, MDOCF_CHILD }, /* Fl */
170: { pmdoc_Fn, 0, 0 }, /* Fn */
171: { NULL, TYPE_Ft, MDOCF_CHILD }, /* Ft */
172: { NULL, TYPE_Ic, MDOCF_CHILD }, /* Ic */
173: { pmdoc_In, TYPE_In, 0 }, /* In */
174: { NULL, TYPE_Li, MDOCF_CHILD }, /* Li */
175: { pmdoc_Nd, TYPE_Nd, MDOCF_CHILD }, /* Nd */
176: { pmdoc_Nm, TYPE_Nm, MDOCF_CHILD }, /* Nm */
177: { NULL, 0, 0 }, /* Op */
178: { NULL, 0, 0 }, /* Ot */
179: { NULL, TYPE_Pa, MDOCF_CHILD }, /* Pa */
180: { NULL, 0, 0 }, /* Rv */
181: { pmdoc_St, TYPE_St, 0 }, /* St */
182: { NULL, TYPE_Va, MDOCF_CHILD }, /* Va */
183: { pmdoc_body, TYPE_Va, MDOCF_CHILD }, /* Vt */
184: { pmdoc_Xr, TYPE_Xr, 0 }, /* Xr */
185: { NULL, 0, 0 }, /* %A */
186: { NULL, 0, 0 }, /* %B */
187: { NULL, 0, 0 }, /* %D */
188: { NULL, 0, 0 }, /* %I */
189: { NULL, 0, 0 }, /* %J */
190: { NULL, 0, 0 }, /* %N */
191: { NULL, 0, 0 }, /* %O */
192: { NULL, 0, 0 }, /* %P */
193: { NULL, 0, 0 }, /* %R */
194: { NULL, 0, 0 }, /* %T */
195: { NULL, 0, 0 }, /* %V */
196: { NULL, 0, 0 }, /* Ac */
197: { NULL, 0, 0 }, /* Ao */
198: { NULL, 0, 0 }, /* Aq */
199: { NULL, TYPE_At, MDOCF_CHILD }, /* At */
200: { NULL, 0, 0 }, /* Bc */
201: { NULL, 0, 0 }, /* Bf */
202: { NULL, 0, 0 }, /* Bo */
203: { NULL, 0, 0 }, /* Bq */
204: { NULL, TYPE_Bsx, MDOCF_CHILD }, /* Bsx */
205: { NULL, TYPE_Bx, MDOCF_CHILD }, /* Bx */
206: { NULL, 0, 0 }, /* Db */
207: { NULL, 0, 0 }, /* Dc */
208: { NULL, 0, 0 }, /* Do */
209: { NULL, 0, 0 }, /* Dq */
210: { NULL, 0, 0 }, /* Ec */
211: { NULL, 0, 0 }, /* Ef */
212: { NULL, TYPE_Em, MDOCF_CHILD }, /* Em */
213: { NULL, 0, 0 }, /* Eo */
214: { NULL, TYPE_Fx, MDOCF_CHILD }, /* Fx */
215: { NULL, TYPE_Ms, MDOCF_CHILD }, /* Ms */
216: { NULL, 0, 0 }, /* No */
217: { NULL, 0, 0 }, /* Ns */
218: { NULL, TYPE_Nx, MDOCF_CHILD }, /* Nx */
219: { NULL, TYPE_Ox, MDOCF_CHILD }, /* Ox */
220: { NULL, 0, 0 }, /* Pc */
221: { NULL, 0, 0 }, /* Pf */
222: { NULL, 0, 0 }, /* Po */
223: { NULL, 0, 0 }, /* Pq */
224: { NULL, 0, 0 }, /* Qc */
225: { NULL, 0, 0 }, /* Ql */
226: { NULL, 0, 0 }, /* Qo */
227: { NULL, 0, 0 }, /* Qq */
228: { NULL, 0, 0 }, /* Re */
229: { NULL, 0, 0 }, /* Rs */
230: { NULL, 0, 0 }, /* Sc */
231: { NULL, 0, 0 }, /* So */
232: { NULL, 0, 0 }, /* Sq */
233: { NULL, 0, 0 }, /* Sm */
234: { NULL, 0, 0 }, /* Sx */
235: { NULL, TYPE_Sy, MDOCF_CHILD }, /* Sy */
236: { NULL, TYPE_Tn, MDOCF_CHILD }, /* Tn */
237: { NULL, 0, 0 }, /* Ux */
238: { NULL, 0, 0 }, /* Xc */
239: { NULL, 0, 0 }, /* Xo */
240: { pmdoc_head, TYPE_Fn, 0 }, /* Fo */
241: { NULL, 0, 0 }, /* Fc */
242: { NULL, 0, 0 }, /* Oo */
243: { NULL, 0, 0 }, /* Oc */
244: { NULL, 0, 0 }, /* Bk */
245: { NULL, 0, 0 }, /* Ek */
246: { NULL, 0, 0 }, /* Bt */
247: { NULL, 0, 0 }, /* Hf */
248: { NULL, 0, 0 }, /* Fr */
249: { NULL, 0, 0 }, /* Ud */
250: { NULL, TYPE_Lb, MDOCF_CHILD }, /* Lb */
251: { NULL, 0, 0 }, /* Lp */
252: { NULL, TYPE_Lk, MDOCF_CHILD }, /* Lk */
253: { NULL, TYPE_Mt, MDOCF_CHILD }, /* Mt */
254: { NULL, 0, 0 }, /* Brq */
255: { NULL, 0, 0 }, /* Bro */
256: { NULL, 0, 0 }, /* Brc */
257: { NULL, 0, 0 }, /* %C */
258: { NULL, 0, 0 }, /* Es */
259: { NULL, 0, 0 }, /* En */
260: { NULL, TYPE_Dx, MDOCF_CHILD }, /* Dx */
261: { NULL, 0, 0 }, /* %Q */
262: { NULL, 0, 0 }, /* br */
263: { NULL, 0, 0 }, /* sp */
264: { NULL, 0, 0 }, /* %U */
265: { NULL, 0, 0 }, /* Ta */
1.1 schwarze 266: };
267:
268: static const char *progname;
1.13 schwarze 269: static int use_all; /* Use all directories and files. */
270: static int verb; /* Output verbosity level. */
1.28 schwarze 271: static int warnings; /* Potential problems in manuals. */
1.1 schwarze 272:
273: int
1.3 schwarze 274: mandocdb(int argc, char *argv[])
1.1 schwarze 275: {
276: struct mparse *mp; /* parse sequence */
1.10 schwarze 277: struct manpaths dirs;
1.28 schwarze 278: struct mdb mdb;
279: struct recs recs;
1.1 schwarze 280: enum op op; /* current operation */
1.2 schwarze 281: const char *dir;
1.21 schwarze 282: char *cp;
1.28 schwarze 283: char pbuf[PATH_MAX];
1.13 schwarze 284: int ch, i, flags;
1.28 schwarze 285: DB *hash; /* temporary keyword hashtable */
1.1 schwarze 286: BTREEINFO info; /* btree configuration */
1.28 schwarze 287: size_t sz1, sz2;
1.1 schwarze 288: struct buf buf, /* keyword buffer */
289: dbuf; /* description buffer */
1.2 schwarze 290: struct of *of; /* list of files for processing */
1.1 schwarze 291: extern int optind;
292: extern char *optarg;
293:
294: progname = strrchr(argv[0], '/');
295: if (progname == NULL)
296: progname = argv[0];
297: else
298: ++progname;
299:
1.10 schwarze 300: memset(&dirs, 0, sizeof(struct manpaths));
1.28 schwarze 301: memset(&mdb, 0, sizeof(struct mdb));
302: memset(&recs, 0, sizeof(struct recs));
1.10 schwarze 303:
1.2 schwarze 304: of = NULL;
1.1 schwarze 305: mp = NULL;
306: hash = NULL;
1.28 schwarze 307: op = OP_DEFAULT;
1.2 schwarze 308: dir = NULL;
1.1 schwarze 309:
1.28 schwarze 310: while (-1 != (ch = getopt(argc, argv, "aC:d:tu:vW")))
1.1 schwarze 311: switch (ch) {
1.6 schwarze 312: case ('a'):
313: use_all = 1;
314: break;
1.25 schwarze 315: case ('C'):
1.28 schwarze 316: if (op) {
317: fprintf(stderr,
318: "-C: conflicting options\n");
319: goto usage;
320: }
321: dir = optarg;
322: op = OP_CONFFILE;
1.25 schwarze 323: break;
1.1 schwarze 324: case ('d'):
1.28 schwarze 325: if (op) {
326: fprintf(stderr,
327: "-d: conflicting options\n");
328: goto usage;
329: }
1.1 schwarze 330: dir = optarg;
1.2 schwarze 331: op = OP_UPDATE;
1.1 schwarze 332: break;
1.28 schwarze 333: case ('t'):
334: dup2(STDOUT_FILENO, STDERR_FILENO);
335: if (op) {
336: fprintf(stderr,
337: "-t: conflicting options\n");
338: goto usage;
339: }
340: op = OP_TEST;
341: use_all = 1;
342: warnings = 1;
343: break;
1.2 schwarze 344: case ('u'):
1.28 schwarze 345: if (op) {
346: fprintf(stderr,
347: "-u: conflicting options\n");
348: goto usage;
349: }
1.2 schwarze 350: dir = optarg;
1.1 schwarze 351: op = OP_DELETE;
352: break;
353: case ('v'):
354: verb++;
355: break;
1.28 schwarze 356: case ('W'):
357: warnings = 1;
358: break;
1.1 schwarze 359: default:
1.28 schwarze 360: goto usage;
1.1 schwarze 361: }
362:
363: argc -= optind;
364: argv += optind;
365:
1.28 schwarze 366: if (OP_CONFFILE == op && argc > 0) {
367: fprintf(stderr, "-C: too many arguments\n");
368: goto usage;
369: }
370:
1.2 schwarze 371: memset(&info, 0, sizeof(BTREEINFO));
1.17 schwarze 372: info.lorder = 4321;
1.2 schwarze 373: info.flags = R_DUP;
374:
375: mp = mparse_alloc(MPARSE_AUTO, MANDOCLEVEL_FATAL, NULL, NULL);
376:
377: memset(&buf, 0, sizeof(struct buf));
378: memset(&dbuf, 0, sizeof(struct buf));
379:
380: buf.size = dbuf.size = MANDOC_BUFSZ;
381:
382: buf.cp = mandoc_malloc(buf.size);
383: dbuf.cp = mandoc_malloc(dbuf.size);
384:
1.28 schwarze 385: flags = O_CREAT | O_RDWR;
386: if (OP_DEFAULT == op || OP_CONFFILE == op)
387: flags |= O_TRUNC;
388:
389: if (OP_TEST == op) {
390: ofile_argbuild(argc, argv, &of);
391: if (NULL == of)
392: goto out;
393: index_merge(of, mp, &dbuf, &buf, hash, &mdb, &recs);
394: goto out;
395: }
1.1 schwarze 396:
1.2 schwarze 397: if (OP_UPDATE == op || OP_DELETE == op) {
1.28 schwarze 398: strlcat(mdb.dbn, dir, MAXPATHLEN);
399: strlcat(mdb.dbn, "/", MAXPATHLEN);
400: sz1 = strlcat(mdb.dbn, MANDOC_DB, MAXPATHLEN);
401:
402: strlcat(mdb.idxn, dir, MAXPATHLEN);
403: strlcat(mdb.idxn, "/", MAXPATHLEN);
404: sz2 = strlcat(mdb.idxn, MANDOC_IDX, MAXPATHLEN);
1.2 schwarze 405:
406: if (sz1 >= MAXPATHLEN || sz2 >= MAXPATHLEN) {
1.28 schwarze 407: fprintf(stderr, "%s: path too long\n", dir);
1.2 schwarze 408: exit((int)MANDOCLEVEL_BADARG);
409: }
1.1 schwarze 410:
1.28 schwarze 411: mdb.db = dbopen(mdb.dbn, flags, 0644, DB_BTREE, &info);
412: mdb.idx = dbopen(mdb.idxn, flags, 0644, DB_RECNO, NULL);
1.1 schwarze 413:
1.28 schwarze 414: if (NULL == mdb.db) {
415: perror(mdb.dbn);
1.2 schwarze 416: exit((int)MANDOCLEVEL_SYSERR);
1.28 schwarze 417: } else if (NULL == mdb.idx) {
418: perror(mdb.idxn);
1.2 schwarze 419: exit((int)MANDOCLEVEL_SYSERR);
420: }
1.1 schwarze 421:
1.26 schwarze 422: ofile_argbuild(argc, argv, &of);
1.1 schwarze 423:
1.2 schwarze 424: if (NULL == of)
425: goto out;
426:
1.28 schwarze 427: index_prune(of, &mdb, &recs);
1.2 schwarze 428:
1.15 schwarze 429: /*
1.26 schwarze 430: * Go to the root of the respective manual tree.
431: * This must work or no manuals may be found (they're
432: * indexed relative to the root).
1.15 schwarze 433: */
434:
1.14 schwarze 435: if (OP_UPDATE == op) {
1.26 schwarze 436: if (-1 == chdir(dir)) {
437: perror(dir);
438: exit((int)MANDOCLEVEL_SYSERR);
439: }
1.10 schwarze 440: index_merge(of, mp, &dbuf, &buf, hash,
1.28 schwarze 441: &mdb, &recs);
1.14 schwarze 442: }
1.1 schwarze 443:
444: goto out;
445: }
446:
1.10 schwarze 447: /*
448: * Configure the directories we're going to scan.
449: * If we have command-line arguments, use them.
450: * If not, we use man(1)'s method (see mandocdb.8).
451: */
452:
453: if (argc > 0) {
1.21 schwarze 454: dirs.paths = mandoc_calloc(argc, sizeof(char *));
1.10 schwarze 455: dirs.sz = argc;
1.21 schwarze 456: for (i = 0; i < argc; i++) {
457: if (NULL == (cp = realpath(argv[i], pbuf))) {
458: perror(argv[i]);
459: goto out;
460: }
461: dirs.paths[i] = mandoc_strdup(cp);
462: }
1.10 schwarze 463: } else
1.28 schwarze 464: manpath_parse(&dirs, dir, NULL, NULL);
1.7 schwarze 465:
1.10 schwarze 466: for (i = 0; i < dirs.sz; i++) {
1.28 schwarze 467: mdb.idxn[0] = mdb.dbn[0] = '\0';
1.2 schwarze 468:
1.28 schwarze 469: strlcat(mdb.dbn, dirs.paths[i], MAXPATHLEN);
470: strlcat(mdb.dbn, "/", MAXPATHLEN);
471: sz1 = strlcat(mdb.dbn, MANDOC_DB, MAXPATHLEN);
472:
473: strlcat(mdb.idxn, dirs.paths[i], MAXPATHLEN);
474: strlcat(mdb.idxn, "/", MAXPATHLEN);
475: sz2 = strlcat(mdb.idxn, MANDOC_IDX, MAXPATHLEN);
1.2 schwarze 476:
477: if (sz1 >= MAXPATHLEN || sz2 >= MAXPATHLEN) {
1.28 schwarze 478: fprintf(stderr, "%s: path too long\n",
1.10 schwarze 479: dirs.paths[i]);
1.2 schwarze 480: exit((int)MANDOCLEVEL_BADARG);
481: }
482:
1.28 schwarze 483: if (mdb.db)
484: (*mdb.db->close)(mdb.db);
485: if (mdb.idx)
486: (*mdb.idx->close)(mdb.idx);
1.7 schwarze 487:
1.28 schwarze 488: mdb.db = dbopen(mdb.dbn, flags, 0644, DB_BTREE, &info);
489: mdb.idx = dbopen(mdb.idxn, flags, 0644, DB_RECNO, NULL);
1.1 schwarze 490:
1.28 schwarze 491: if (NULL == mdb.db) {
492: perror(mdb.dbn);
1.2 schwarze 493: exit((int)MANDOCLEVEL_SYSERR);
1.28 schwarze 494: } else if (NULL == mdb.idx) {
495: perror(mdb.idxn);
1.2 schwarze 496: exit((int)MANDOCLEVEL_SYSERR);
497: }
1.1 schwarze 498:
1.2 schwarze 499: ofile_free(of);
500: of = NULL;
1.1 schwarze 501:
1.26 schwarze 502: if (-1 == chdir(dirs.paths[i])) {
503: perror(dirs.paths[i]);
1.2 schwarze 504: exit((int)MANDOCLEVEL_SYSERR);
1.28 schwarze 505: }
1.1 schwarze 506:
1.28 schwarze 507: ofile_dirbuild(".", "", "", 0, &of);
1.2 schwarze 508: if (NULL == of)
509: continue;
1.1 schwarze 510:
1.15 schwarze 511: /*
1.28 schwarze 512: * Go to the root of the respective manual tree.
1.26 schwarze 513: * This must work or no manuals may be found (they're
514: * indexed relative to the root).
1.15 schwarze 515: */
1.1 schwarze 516:
1.26 schwarze 517: if (-1 == chdir(dirs.paths[i])) {
518: perror(dirs.paths[i]);
519: exit((int)MANDOCLEVEL_SYSERR);
520: }
521:
1.28 schwarze 522: index_merge(of, mp, &dbuf, &buf, hash, &mdb, &recs);
1.1 schwarze 523: }
524:
1.2 schwarze 525: out:
1.28 schwarze 526: if (mdb.db)
527: (*mdb.db->close)(mdb.db);
528: if (mdb.idx)
529: (*mdb.idx->close)(mdb.idx);
1.2 schwarze 530: if (hash)
531: (*hash->close)(hash);
532: if (mp)
533: mparse_free(mp);
1.1 schwarze 534:
1.10 schwarze 535: manpath_free(&dirs);
1.2 schwarze 536: ofile_free(of);
537: free(buf.cp);
538: free(dbuf.cp);
1.28 schwarze 539: free(recs.stack);
1.1 schwarze 540:
1.2 schwarze 541: return(MANDOCLEVEL_OK);
1.28 schwarze 542:
543: usage:
544: fprintf(stderr,
545: "usage: %s [-avvv] [-C file] | dir ... | -t file ...\n"
546: " -d dir [file ...] | "
547: "-u dir [file ...]\n",
548: progname);
549:
550: return((int)MANDOCLEVEL_BADARG);
1.2 schwarze 551: }
1.1 schwarze 552:
1.2 schwarze 553: void
554: index_merge(const struct of *of, struct mparse *mp,
1.13 schwarze 555: struct buf *dbuf, struct buf *buf, DB *hash,
1.28 schwarze 556: struct mdb *mdb, struct recs *recs)
1.2 schwarze 557: {
558: recno_t rec;
1.28 schwarze 559: int ch, skip;
1.2 schwarze 560: DBT key, val;
561: struct mdoc *mdoc;
562: struct man *man;
1.28 schwarze 563: const char *fn, *msec, *march, *mtitle;
1.27 schwarze 564: uint64_t mask;
1.2 schwarze 565: size_t sv;
566: unsigned seq;
1.29 schwarze 567: uint64_t vbuf[2];
1.26 schwarze 568: char type;
1.1 schwarze 569:
1.28 schwarze 570: rec = 0;
571: for (of = of->first; of; of = of->next) {
1.2 schwarze 572: fn = of->fname;
1.11 schwarze 573:
574: /*
1.24 schwarze 575: * Try interpreting the file as mdoc(7) or man(7)
576: * source code, unless it is already known to be
577: * formatted. Fall back to formatted mode.
1.11 schwarze 578: */
579:
1.1 schwarze 580: mparse_reset(mp);
1.11 schwarze 581: mdoc = NULL;
582: man = NULL;
1.1 schwarze 583:
1.11 schwarze 584: if ((MANDOC_SRC & of->src_form ||
585: ! (MANDOC_FORM & of->src_form)) &&
586: MANDOCLEVEL_FATAL > mparse_readfd(mp, -1, fn))
587: mparse_result(mp, &mdoc, &man);
588:
589: if (NULL != mdoc) {
590: msec = mdoc_meta(mdoc)->msec;
1.28 schwarze 591: march = mdoc_meta(mdoc)->arch;
592: if (NULL == march)
593: march = "";
1.11 schwarze 594: mtitle = mdoc_meta(mdoc)->title;
595: } else if (NULL != man) {
596: msec = man_meta(man)->msec;
1.28 schwarze 597: march = "";
1.11 schwarze 598: mtitle = man_meta(man)->title;
599: } else {
600: msec = of->sec;
1.28 schwarze 601: march = of->arch;
1.11 schwarze 602: mtitle = of->title;
1.1 schwarze 603: }
604:
1.6 schwarze 605: /*
1.8 schwarze 606: * By default, skip a file if the manual section
1.32 schwarze 607: * given in the file disagrees with the directory
608: * where the file is located.
1.6 schwarze 609: */
610:
1.28 schwarze 611: skip = 0;
612: assert(of->sec);
613: assert(msec);
614: if (strcasecmp(msec, of->sec)) {
615: if (warnings)
616: fprintf(stderr, "%s: "
617: "section \"%s\" manual "
618: "in \"%s\" directory\n",
619: fn, msec, of->sec);
620: skip = 1;
621: }
622:
1.32 schwarze 623: /*
624: * Manual page directories exist for each kernel
625: * architecture as returned by machine(1).
626: * However, many manuals only depend on the
627: * application architecture as returned by arch(1).
628: * For example, some (2/ARM) manuals are shared
629: * across the "armish" and "zaurus" kernel
630: * architectures.
631: * A few manuals are even shared across completely
632: * different architectures, for example fdformat(1)
633: * on amd64, i386, sparc, and sparc64.
634: * Thus, warn about architecture mismatches,
635: * but don't skip manuals for this reason.
636: */
637:
1.28 schwarze 638: assert(of->arch);
639: assert(march);
640: if (strcasecmp(march, of->arch)) {
641: if (warnings)
642: fprintf(stderr, "%s: "
643: "architecture \"%s\" manual "
644: "in \"%s\" directory\n",
645: fn, march, of->arch);
1.32 schwarze 646: march = of->arch;
1.6 schwarze 647: }
648:
1.28 schwarze 649: /*
1.8 schwarze 650: * By default, skip a file if the title given
651: * in the file disagrees with the file name.
652: * If both agree, use the file name as the title,
653: * because the one in the file usually is all caps.
1.6 schwarze 654: */
655:
656: assert(of->title);
657: assert(mtitle);
1.28 schwarze 658: if (strcasecmp(mtitle, of->title)) {
659: if (warnings)
660: fprintf(stderr, "%s: "
661: "title \"%s\" in file "
662: "but \"%s\" in filename\n",
663: fn, mtitle, of->title);
664: skip = 1;
665: } else
666: mtitle = of->title;
1.6 schwarze 667:
1.28 schwarze 668: if (skip && !use_all)
1.6 schwarze 669: continue;
670:
1.28 schwarze 671: /*
1.1 schwarze 672: * The index record value consists of a nil-terminated
673: * filename, a nil-terminated manual section, and a
674: * nil-terminated description. Since the description
675: * may not be set, we set a sentinel to see if we're
676: * going to write a nil byte in its place.
677: */
678:
1.2 schwarze 679: dbuf->len = 0;
1.26 schwarze 680: type = mdoc ? 'd' : (man ? 'a' : 'c');
681: buf_appendb(dbuf, &type, 1);
1.2 schwarze 682: buf_appendb(dbuf, fn, strlen(fn) + 1);
683: buf_appendb(dbuf, msec, strlen(msec) + 1);
684: buf_appendb(dbuf, mtitle, strlen(mtitle) + 1);
1.28 schwarze 685: buf_appendb(dbuf, march, strlen(march) + 1);
1.1 schwarze 686:
1.2 schwarze 687: sv = dbuf->len;
1.1 schwarze 688:
1.24 schwarze 689: /*
690: * Collect keyword/mask pairs.
691: * Each pair will become a new btree node.
692: */
1.1 schwarze 693:
1.24 schwarze 694: hash_reset(&hash);
1.1 schwarze 695: if (mdoc)
1.2 schwarze 696: pmdoc_node(hash, buf, dbuf,
1.1 schwarze 697: mdoc_node(mdoc), mdoc_meta(mdoc));
1.11 schwarze 698: else if (man)
1.2 schwarze 699: pman_node(hash, buf, dbuf, man_node(man));
1.11 schwarze 700: else
701: pformatted(hash, buf, dbuf, of);
1.1 schwarze 702:
1.28 schwarze 703: /* Test mode, do not access any database. */
704:
705: if (NULL == mdb->db || NULL == mdb->idx)
706: continue;
707:
1.1 schwarze 708: /*
1.24 schwarze 709: * Reclaim an empty index record, if available.
710: * Use its record number for all new btree nodes.
1.1 schwarze 711: */
712:
1.28 schwarze 713: if (recs->cur > 0) {
714: recs->cur--;
715: rec = recs->stack[(int)recs->cur];
716: } else if (recs->last > 0) {
717: rec = recs->last;
718: recs->last = 0;
1.24 schwarze 719: } else
720: rec++;
1.29 schwarze 721: vbuf[1] = htobe64(rec);
1.24 schwarze 722:
723: /*
724: * Copy from the in-memory hashtable of pending
725: * keyword/mask pairs into the database.
726: */
727:
1.1 schwarze 728: seq = R_FIRST;
729: while (0 == (ch = (*hash->seq)(hash, &key, &val, seq))) {
730: seq = R_NEXT;
1.27 schwarze 731: assert(sizeof(uint64_t) == val.size);
732: memcpy(&mask, val.data, val.size);
1.29 schwarze 733: vbuf[0] = htobe64(mask);
734: val.size = sizeof(vbuf);
1.9 schwarze 735: val.data = &vbuf;
1.28 schwarze 736: dbt_put(mdb->db, mdb->dbn, &key, &val);
1.1 schwarze 737: }
738: if (ch < 0) {
739: perror("hash");
740: exit((int)MANDOCLEVEL_SYSERR);
741: }
1.28 schwarze 742:
1.1 schwarze 743: /*
744: * Apply to the index. If we haven't had a description
745: * set, put an empty one in now.
746: */
747:
1.2 schwarze 748: if (dbuf->len == sv)
749: buf_appendb(dbuf, "", 1);
1.1 schwarze 750:
751: key.data = &rec;
752: key.size = sizeof(recno_t);
753:
1.2 schwarze 754: val.data = dbuf->cp;
755: val.size = dbuf->len;
1.1 schwarze 756:
1.2 schwarze 757: if (verb)
1.28 schwarze 758: printf("%s: adding to index\n", fn);
1.16 schwarze 759:
1.28 schwarze 760: dbt_put(mdb->idx, mdb->idxn, &key, &val);
1.2 schwarze 761: }
762: }
763:
764: /*
765: * Scan through all entries in the index file `idx' and prune those
766: * entries in `ofile'.
767: * Pruning consists of removing from `db', then invalidating the entry
768: * in `idx' (zeroing its value size).
769: */
770: static void
1.28 schwarze 771: index_prune(const struct of *ofile, struct mdb *mdb, struct recs *recs)
1.2 schwarze 772: {
773: const struct of *of;
1.26 schwarze 774: const char *fn;
1.29 schwarze 775: uint64_t vbuf[2];
1.2 schwarze 776: unsigned seq, sseq;
777: DBT key, val;
778: int ch;
779:
1.28 schwarze 780: recs->cur = 0;
1.2 schwarze 781: seq = R_FIRST;
1.28 schwarze 782: while (0 == (ch = (*mdb->idx->seq)(mdb->idx, &key, &val, seq))) {
1.2 schwarze 783: seq = R_NEXT;
1.27 schwarze 784: assert(sizeof(recno_t) == key.size);
1.28 schwarze 785: memcpy(&recs->last, key.data, key.size);
1.16 schwarze 786:
787: /* Deleted records are zero-sized. Skip them. */
788:
789: if (0 == val.size)
790: goto cont;
791:
792: /*
793: * Make sure we're sane.
794: * Read past our mdoc/man/cat type to the next string,
795: * then make sure it's bounded by a NUL.
796: * Failing any of these, we go into our error handler.
797: */
798:
1.26 schwarze 799: fn = (char *)val.data + 1;
800: if (NULL == memchr(fn, '\0', val.size - 1))
1.16 schwarze 801: break;
802:
1.28 schwarze 803: /*
1.16 schwarze 804: * Search for the file in those we care about.
805: * XXX: build this into a tree. Too slow.
806: */
1.2 schwarze 807:
1.28 schwarze 808: for (of = ofile->first; of; of = of->next)
1.2 schwarze 809: if (0 == strcmp(fn, of->fname))
810: break;
811:
812: if (NULL == of)
813: continue;
814:
1.16 schwarze 815: /*
816: * Search through the keyword database, throwing out all
817: * references to our file.
818: */
819:
1.2 schwarze 820: sseq = R_FIRST;
1.28 schwarze 821: while (0 == (ch = (*mdb->db->seq)(mdb->db,
822: &key, &val, sseq))) {
1.2 schwarze 823: sseq = R_NEXT;
1.29 schwarze 824: if (sizeof(vbuf) != val.size)
1.16 schwarze 825: break;
826:
1.29 schwarze 827: memcpy(vbuf, val.data, val.size);
828: if (recs->last != betoh64(vbuf[1]))
1.2 schwarze 829: continue;
1.16 schwarze 830:
1.28 schwarze 831: if ((ch = (*mdb->db->del)(mdb->db,
832: &key, R_CURSOR)) < 0)
1.2 schwarze 833: break;
834: }
1.16 schwarze 835:
1.2 schwarze 836: if (ch < 0) {
1.28 schwarze 837: perror(mdb->dbn);
1.2 schwarze 838: exit((int)MANDOCLEVEL_SYSERR);
1.16 schwarze 839: } else if (1 != ch) {
1.28 schwarze 840: fprintf(stderr, "%s: corrupt database\n",
841: mdb->dbn);
1.16 schwarze 842: exit((int)MANDOCLEVEL_SYSERR);
1.2 schwarze 843: }
1.1 schwarze 844:
1.2 schwarze 845: if (verb)
1.28 schwarze 846: printf("%s: deleting from index\n", fn);
1.1 schwarze 847:
1.2 schwarze 848: val.size = 0;
1.28 schwarze 849: ch = (*mdb->idx->put)(mdb->idx, &key, &val, R_CURSOR);
1.1 schwarze 850:
1.16 schwarze 851: if (ch < 0)
852: break;
853: cont:
1.28 schwarze 854: if (recs->cur >= recs->size) {
855: recs->size += MANDOC_SLOP;
856: recs->stack = mandoc_realloc(recs->stack,
857: recs->size * sizeof(recno_t));
1.2 schwarze 858: }
1.1 schwarze 859:
1.28 schwarze 860: recs->stack[(int)recs->cur] = recs->last;
861: recs->cur++;
1.2 schwarze 862: }
1.16 schwarze 863:
864: if (ch < 0) {
1.28 schwarze 865: perror(mdb->idxn);
1.16 schwarze 866: exit((int)MANDOCLEVEL_SYSERR);
867: } else if (1 != ch) {
1.28 schwarze 868: fprintf(stderr, "%s: corrupt index\n", mdb->idxn);
1.16 schwarze 869: exit((int)MANDOCLEVEL_SYSERR);
870: }
871:
1.28 schwarze 872: recs->last++;
1.1 schwarze 873: }
874:
875: /*
876: * Grow the buffer (if necessary) and copy in a binary string.
877: */
878: static void
879: buf_appendb(struct buf *buf, const void *cp, size_t sz)
880: {
881:
882: /* Overshoot by MANDOC_BUFSZ. */
883:
884: while (buf->len + sz >= buf->size) {
885: buf->size = buf->len + sz + MANDOC_BUFSZ;
886: buf->cp = mandoc_realloc(buf->cp, buf->size);
887: }
888:
889: memcpy(buf->cp + (int)buf->len, cp, sz);
890: buf->len += sz;
891: }
892:
893: /*
894: * Append a nil-terminated string to the buffer.
895: * This can be invoked multiple times.
896: * The buffer string will be nil-terminated.
897: * If invoked multiple times, a space is put between strings.
898: */
899: static void
900: buf_append(struct buf *buf, const char *cp)
901: {
902: size_t sz;
903:
904: if (0 == (sz = strlen(cp)))
905: return;
906:
907: if (buf->len)
908: buf->cp[(int)buf->len - 1] = ' ';
909:
910: buf_appendb(buf, cp, sz + 1);
911: }
912:
913: /*
914: * Recursively add all text from a given node.
915: * This is optimised for general mdoc nodes in this context, which do
916: * not consist of subexpressions and having a recursive call for n->next
917: * would be wasteful.
918: * The "f" variable should be 0 unless called from pmdoc_Nd for the
919: * description buffer, which does not start at the beginning of the
920: * buffer.
921: */
922: static void
923: buf_appendmdoc(struct buf *buf, const struct mdoc_node *n, int f)
924: {
925:
926: for ( ; n; n = n->next) {
927: if (n->child)
928: buf_appendmdoc(buf, n->child, f);
929:
930: if (MDOC_TEXT == n->type && f) {
931: f = 0;
932: buf_appendb(buf, n->string,
933: strlen(n->string) + 1);
934: } else if (MDOC_TEXT == n->type)
935: buf_append(buf, n->string);
936:
937: }
938: }
939:
940: static void
941: hash_reset(DB **db)
942: {
943: DB *hash;
944:
945: if (NULL != (hash = *db))
946: (*hash->close)(hash);
947:
1.2 schwarze 948: *db = dbopen(NULL, O_CREAT|O_RDWR, 0644, DB_HASH, NULL);
1.1 schwarze 949: if (NULL == *db) {
950: perror("hash");
951: exit((int)MANDOCLEVEL_SYSERR);
952: }
953: }
954:
955: /* ARGSUSED */
1.19 schwarze 956: static int
957: pmdoc_head(MDOC_ARGS)
958: {
959:
960: return(MDOC_HEAD == n->type);
961: }
962:
963: /* ARGSUSED */
964: static int
965: pmdoc_body(MDOC_ARGS)
966: {
967:
968: return(MDOC_BODY == n->type);
969: }
970:
971: /* ARGSUSED */
972: static int
1.1 schwarze 973: pmdoc_Fd(MDOC_ARGS)
974: {
975: const char *start, *end;
976: size_t sz;
1.19 schwarze 977:
1.1 schwarze 978: if (SEC_SYNOPSIS != n->sec)
1.19 schwarze 979: return(0);
1.1 schwarze 980: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.19 schwarze 981: return(0);
1.1 schwarze 982:
983: /*
984: * Only consider those `Fd' macro fields that begin with an
985: * "inclusion" token (versus, e.g., #define).
986: */
987: if (strcmp("#include", n->string))
1.19 schwarze 988: return(0);
1.1 schwarze 989:
990: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.19 schwarze 991: return(0);
1.1 schwarze 992:
993: /*
994: * Strip away the enclosing angle brackets and make sure we're
995: * not zero-length.
996: */
997:
998: start = n->string;
999: if ('<' == *start || '"' == *start)
1000: start++;
1001:
1002: if (0 == (sz = strlen(start)))
1.19 schwarze 1003: return(0);
1.1 schwarze 1004:
1005: end = &start[(int)sz - 1];
1006: if ('>' == *end || '"' == *end)
1007: end--;
1008:
1009: assert(end >= start);
1010:
1011: buf_appendb(buf, start, (size_t)(end - start + 1));
1012: buf_appendb(buf, "", 1);
1.19 schwarze 1013: return(1);
1.1 schwarze 1014: }
1015:
1016: /* ARGSUSED */
1.19 schwarze 1017: static int
1018: pmdoc_In(MDOC_ARGS)
1.1 schwarze 1019: {
1020:
1021: if (NULL == n->child || MDOC_TEXT != n->child->type)
1.19 schwarze 1022: return(0);
1.1 schwarze 1023:
1024: buf_append(buf, n->child->string);
1.19 schwarze 1025: return(1);
1.1 schwarze 1026: }
1027:
1028: /* ARGSUSED */
1.19 schwarze 1029: static int
1.1 schwarze 1030: pmdoc_Fn(MDOC_ARGS)
1031: {
1.19 schwarze 1032: struct mdoc_node *nn;
1.1 schwarze 1033: const char *cp;
1034:
1.19 schwarze 1035: nn = n->child;
1036:
1037: if (NULL == nn || MDOC_TEXT != nn->type)
1038: return(0);
1039:
1040: /* .Fn "struct type *name" "char *arg" */
1.1 schwarze 1041:
1.19 schwarze 1042: cp = strrchr(nn->string, ' ');
1.1 schwarze 1043: if (NULL == cp)
1.19 schwarze 1044: cp = nn->string;
1.1 schwarze 1045:
1046: /* Strip away pointer symbol. */
1047:
1048: while ('*' == *cp)
1049: cp++;
1050:
1.19 schwarze 1051: /* Store the function name. */
1052:
1.1 schwarze 1053: buf_append(buf, cp);
1.5 schwarze 1054: hash_put(hash, buf, TYPE_Fn);
1.19 schwarze 1055:
1056: /* Store the function type. */
1057:
1058: if (nn->string < cp) {
1059: buf->len = 0;
1060: buf_appendb(buf, nn->string, cp - nn->string);
1061: buf_appendb(buf, "", 1);
1062: hash_put(hash, buf, TYPE_Ft);
1063: }
1064:
1065: /* Store the arguments. */
1066:
1067: for (nn = nn->next; nn; nn = nn->next) {
1068: if (MDOC_TEXT != nn->type)
1069: continue;
1070: buf->len = 0;
1071: buf_append(buf, nn->string);
1072: hash_put(hash, buf, TYPE_Fa);
1073: }
1074:
1075: return(0);
1.1 schwarze 1076: }
1077:
1078: /* ARGSUSED */
1.19 schwarze 1079: static int
1.1 schwarze 1080: pmdoc_St(MDOC_ARGS)
1081: {
1.19 schwarze 1082:
1.1 schwarze 1083: if (NULL == n->child || MDOC_TEXT != n->child->type)
1.19 schwarze 1084: return(0);
1.1 schwarze 1085:
1086: buf_append(buf, n->child->string);
1.19 schwarze 1087: return(1);
1.1 schwarze 1088: }
1089:
1090: /* ARGSUSED */
1.19 schwarze 1091: static int
1.1 schwarze 1092: pmdoc_Xr(MDOC_ARGS)
1093: {
1094:
1095: if (NULL == (n = n->child))
1.19 schwarze 1096: return(0);
1.1 schwarze 1097:
1098: buf_appendb(buf, n->string, strlen(n->string));
1099:
1100: if (NULL != (n = n->next)) {
1101: buf_appendb(buf, ".", 1);
1102: buf_appendb(buf, n->string, strlen(n->string) + 1);
1103: } else
1104: buf_appendb(buf, ".", 2);
1105:
1.19 schwarze 1106: return(1);
1.1 schwarze 1107: }
1108:
1109: /* ARGSUSED */
1.19 schwarze 1110: static int
1.1 schwarze 1111: pmdoc_Nd(MDOC_ARGS)
1112: {
1113:
1114: if (MDOC_BODY != n->type)
1.19 schwarze 1115: return(0);
1.1 schwarze 1116:
1117: buf_appendmdoc(dbuf, n->child, 1);
1.19 schwarze 1118: return(1);
1.1 schwarze 1119: }
1120:
1121: /* ARGSUSED */
1.19 schwarze 1122: static int
1123: pmdoc_Nm(MDOC_ARGS)
1.1 schwarze 1124: {
1125:
1.19 schwarze 1126: if (SEC_NAME == n->sec)
1127: return(1);
1128: else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
1129: return(0);
1.1 schwarze 1130:
1.19 schwarze 1131: if (NULL == n->child)
1132: buf_append(buf, m->name);
1.1 schwarze 1133:
1.19 schwarze 1134: return(1);
1.1 schwarze 1135: }
1136:
1137: /* ARGSUSED */
1.19 schwarze 1138: static int
1139: pmdoc_Sh(MDOC_ARGS)
1.1 schwarze 1140: {
1141:
1.19 schwarze 1142: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 schwarze 1143: }
1144:
1145: static void
1.9 schwarze 1146: hash_put(DB *db, const struct buf *buf, uint64_t mask)
1.1 schwarze 1147: {
1.27 schwarze 1148: uint64_t oldmask;
1.1 schwarze 1149: DBT key, val;
1150: int rc;
1151:
1152: if (buf->len < 2)
1153: return;
1154:
1155: key.data = buf->cp;
1156: key.size = buf->len;
1157:
1158: if ((rc = (*db->get)(db, &key, &val, 0)) < 0) {
1159: perror("hash");
1160: exit((int)MANDOCLEVEL_SYSERR);
1.27 schwarze 1161: } else if (0 == rc) {
1162: assert(sizeof(uint64_t) == val.size);
1163: memcpy(&oldmask, val.data, val.size);
1164: mask |= oldmask;
1165: }
1.1 schwarze 1166:
1167: val.data = &mask;
1.9 schwarze 1168: val.size = sizeof(uint64_t);
1.1 schwarze 1169:
1170: if ((rc = (*db->put)(db, &key, &val, 0)) < 0) {
1171: perror("hash");
1172: exit((int)MANDOCLEVEL_SYSERR);
1173: }
1174: }
1175:
1176: static void
1177: dbt_put(DB *db, const char *dbn, DBT *key, DBT *val)
1178: {
1179:
1180: assert(key->size);
1181: assert(val->size);
1182:
1183: if (0 == (*db->put)(db, key, val, 0))
1184: return;
1185:
1186: perror(dbn);
1187: exit((int)MANDOCLEVEL_SYSERR);
1188: /* NOTREACHED */
1189: }
1190:
1191: /*
1192: * Call out to per-macro handlers after clearing the persistent database
1193: * key. If the macro sets the database key, flush it to the database.
1194: */
1195: static void
1196: pmdoc_node(MDOC_ARGS)
1197: {
1198:
1199: if (NULL == n)
1200: return;
1201:
1202: switch (n->type) {
1203: case (MDOC_HEAD):
1204: /* FALLTHROUGH */
1205: case (MDOC_BODY):
1206: /* FALLTHROUGH */
1207: case (MDOC_TAIL):
1208: /* FALLTHROUGH */
1209: case (MDOC_BLOCK):
1210: /* FALLTHROUGH */
1211: case (MDOC_ELEM):
1.19 schwarze 1212: buf->len = 0;
1213:
1214: /*
1215: * Both NULL handlers and handlers returning true
1216: * request using the data. Only skip the element
1217: * when the handler returns false.
1218: */
1219:
1220: if (NULL != mdocs[n->tok].fp &&
1221: 0 == (*mdocs[n->tok].fp)(hash, buf, dbuf, n, m))
1.1 schwarze 1222: break;
1223:
1.19 schwarze 1224: /*
1225: * For many macros, use the text from all children.
1226: * Set zero flags for macros not needing this.
1227: * In that case, the handler must fill the buffer.
1228: */
1229:
1230: if (MDOCF_CHILD & mdocs[n->tok].flags)
1231: buf_appendmdoc(buf, n->child, 0);
1232:
1233: /*
1234: * Cover the most common case:
1235: * Automatically stage one string per element.
1236: * Set a zero mask for macros not needing this.
1237: * Additional staging can be done in the handler.
1238: */
1239:
1240: if (mdocs[n->tok].mask)
1241: hash_put(hash, buf, mdocs[n->tok].mask);
1.1 schwarze 1242: break;
1243: default:
1244: break;
1245: }
1246:
1247: pmdoc_node(hash, buf, dbuf, n->child, m);
1248: pmdoc_node(hash, buf, dbuf, n->next, m);
1249: }
1250:
1251: static int
1252: pman_node(MAN_ARGS)
1253: {
1254: const struct man_node *head, *body;
1255: const char *start, *sv;
1256: size_t sz;
1257:
1258: if (NULL == n)
1259: return(0);
1260:
1261: /*
1262: * We're only searching for one thing: the first text child in
1263: * the BODY of a NAME section. Since we don't keep track of
1264: * sections in -man, run some hoops to find out whether we're in
1265: * the correct section or not.
1266: */
1267:
1268: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1269: body = n;
1270: assert(body->parent);
1271: if (NULL != (head = body->parent->head) &&
1272: 1 == head->nchild &&
1273: NULL != (head = (head->child)) &&
1274: MAN_TEXT == head->type &&
1275: 0 == strcmp(head->string, "NAME") &&
1276: NULL != (body = body->child) &&
1277: MAN_TEXT == body->type) {
1278:
1279: assert(body->string);
1280: start = sv = body->string;
1281:
1282: /*
1283: * Go through a special heuristic dance here.
1284: * This is why -man manuals are great!
1285: * (I'm being sarcastic: my eyes are bleeding.)
1286: * Conventionally, one or more manual names are
1287: * comma-specified prior to a whitespace, then a
1288: * dash, then a description. Try to puzzle out
1289: * the name parts here.
1290: */
1291:
1292: for ( ;; ) {
1293: sz = strcspn(start, " ,");
1294: if ('\0' == start[(int)sz])
1295: break;
1296:
1297: buf->len = 0;
1298: buf_appendb(buf, start, sz);
1299: buf_appendb(buf, "", 1);
1300:
1.5 schwarze 1301: hash_put(hash, buf, TYPE_Nm);
1.1 schwarze 1302:
1303: if (' ' == start[(int)sz]) {
1304: start += (int)sz + 1;
1305: break;
1306: }
1307:
1308: assert(',' == start[(int)sz]);
1309: start += (int)sz + 1;
1310: while (' ' == *start)
1311: start++;
1312: }
1313:
1314: buf->len = 0;
1315:
1316: if (sv == start) {
1317: buf_append(buf, start);
1318: return(1);
1319: }
1320:
1321: while (' ' == *start)
1322: start++;
1323:
1324: if (0 == strncmp(start, "-", 1))
1325: start += 1;
1.33 ! schwarze 1326: else if (0 == strncmp(start, "\\-\\-", 4))
! 1327: start += 4;
1.1 schwarze 1328: else if (0 == strncmp(start, "\\-", 2))
1329: start += 2;
1330: else if (0 == strncmp(start, "\\(en", 4))
1331: start += 4;
1332: else if (0 == strncmp(start, "\\(em", 4))
1333: start += 4;
1334:
1335: while (' ' == *start)
1336: start++;
1337:
1338: sz = strlen(start) + 1;
1339: buf_appendb(dbuf, start, sz);
1340: buf_appendb(buf, start, sz);
1341:
1.5 schwarze 1342: hash_put(hash, buf, TYPE_Nd);
1.1 schwarze 1343: }
1344: }
1345:
1.4 schwarze 1346: for (n = n->child; n; n = n->next)
1347: if (pman_node(hash, buf, dbuf, n))
1348: return(1);
1.1 schwarze 1349:
1350: return(0);
1351: }
1352:
1.11 schwarze 1353: /*
1354: * Parse a formatted manual page.
1355: * By necessity, this involves rather crude guesswork.
1356: */
1357: static void
1.33 ! schwarze 1358: pformatted(DB *hash, struct buf *buf,
! 1359: struct buf *dbuf, const struct of *of)
1.11 schwarze 1360: {
1361: FILE *stream;
1.33 ! schwarze 1362: char *line, *p, *title;
! 1363: size_t len, plen, titlesz;
1.11 schwarze 1364:
1365: if (NULL == (stream = fopen(of->fname, "r"))) {
1.28 schwarze 1366: if (warnings)
1367: perror(of->fname);
1.11 schwarze 1368: return;
1369: }
1370:
1371: /*
1372: * Always use the title derived from the filename up front,
1373: * do not even try to find it in the file. This also makes
1374: * sure we don't end up with an orphan index record, even if
1375: * the file content turns out to be completely unintelligible.
1376: */
1377:
1378: buf->len = 0;
1379: buf_append(buf, of->title);
1380: hash_put(hash, buf, TYPE_Nm);
1381:
1.22 schwarze 1382: /* Skip to first blank line. */
1.11 schwarze 1383:
1.22 schwarze 1384: while (NULL != (line = fgetln(stream, &len)))
1385: if ('\n' == *line)
1386: break;
1387:
1388: /*
1389: * Assume the first line that is not indented
1390: * is the first section header. Skip to it.
1391: */
1392:
1393: while (NULL != (line = fgetln(stream, &len)))
1394: if ('\n' != *line && ' ' != *line)
1395: break;
1.33 ! schwarze 1396:
! 1397: /*
! 1398: * Read up until the next section into a buffer.
! 1399: * Strip the leading and trailing newline from each read line,
! 1400: * appending a trailing space.
! 1401: * Ignore empty (whitespace-only) lines.
! 1402: */
! 1403:
! 1404: titlesz = 0;
! 1405: title = NULL;
! 1406:
! 1407: while (NULL != (line = fgetln(stream, &len))) {
! 1408: if (' ' != *line || '\n' != line[(int)len - 1])
! 1409: break;
! 1410: while (len > 0 && isspace((unsigned char)*line)) {
! 1411: line++;
! 1412: len--;
! 1413: }
! 1414: if (1 == len)
! 1415: continue;
! 1416: title = mandoc_realloc(title, titlesz + len);
! 1417: memcpy(title + titlesz, line, len);
! 1418: titlesz += len;
! 1419: title[(int)titlesz - 1] = ' ';
! 1420: }
! 1421:
1.11 schwarze 1422:
1423: /*
1.22 schwarze 1424: * If no page content can be found, or the input line
1425: * is already the next section header, or there is no
1426: * trailing newline, reuse the page title as the page
1427: * description.
1.11 schwarze 1428: */
1429:
1.33 ! schwarze 1430: if (NULL == title || '\0' == *title) {
1.28 schwarze 1431: if (warnings)
1432: fprintf(stderr, "%s: cannot find NAME section\n",
1433: of->fname);
1.11 schwarze 1434: buf_appendb(dbuf, buf->cp, buf->size);
1435: hash_put(hash, buf, TYPE_Nd);
1436: fclose(stream);
1.33 ! schwarze 1437: free(title);
1.11 schwarze 1438: return;
1439: }
1.22 schwarze 1440:
1.33 ! schwarze 1441: title = mandoc_realloc(title, titlesz + 1);
! 1442: title[(int)titlesz] = '\0';
1.11 schwarze 1443:
1444: /*
1.22 schwarze 1445: * Skip to the first dash.
1446: * Use the remaining line as the description (no more than 70
1447: * bytes).
1.11 schwarze 1448: */
1449:
1.33 ! schwarze 1450: if (NULL != (p = strstr(title, "- "))) {
1.22 schwarze 1451: for (p += 2; ' ' == *p || '\b' == *p; p++)
1452: /* Skip to next word. */ ;
1.28 schwarze 1453: } else {
1454: if (warnings)
1455: fprintf(stderr, "%s: no dash in title line\n",
1456: of->fname);
1.33 ! schwarze 1457: p = title;
1.28 schwarze 1458: }
1.22 schwarze 1459:
1.33 ! schwarze 1460: plen = strlen(p);
1.11 schwarze 1461:
1.22 schwarze 1462: /* Strip backspace-encoding from line. */
1463:
1464: while (NULL != (line = memchr(p, '\b', plen))) {
1465: len = line - p;
1466: if (0 == len) {
1467: memmove(line, line + 1, plen--);
1468: continue;
1469: }
1470: memmove(line - 1, line + 1, plen - len);
1471: plen -= 2;
1472: }
1.11 schwarze 1473:
1.22 schwarze 1474: buf_appendb(dbuf, p, plen + 1);
1.11 schwarze 1475: buf->len = 0;
1.22 schwarze 1476: buf_appendb(buf, p, plen + 1);
1.11 schwarze 1477: hash_put(hash, buf, TYPE_Nd);
1.22 schwarze 1478: fclose(stream);
1.33 ! schwarze 1479: free(title);
1.11 schwarze 1480: }
1481:
1.1 schwarze 1482: static void
1.13 schwarze 1483: ofile_argbuild(int argc, char *argv[], struct of **of)
1.2 schwarze 1484: {
1.6 schwarze 1485: char buf[MAXPATHLEN];
1.31 schwarze 1486: const char *sec, *arch, *title;
1487: char *p;
1.11 schwarze 1488: int i, src_form;
1.2 schwarze 1489: struct of *nof;
1490:
1491: for (i = 0; i < argc; i++) {
1.6 schwarze 1492:
1493: /*
1.8 schwarze 1494: * Try to infer the manual section, architecture and
1495: * page title from the path, assuming it looks like
1.11 schwarze 1496: * man*[/<arch>]/<title>.<section> or
1497: * cat<section>[/<arch>]/<title>.0
1.6 schwarze 1498: */
1499:
1500: if (strlcpy(buf, argv[i], sizeof(buf)) >= sizeof(buf)) {
1.28 schwarze 1501: fprintf(stderr, "%s: path too long\n", argv[i]);
1.6 schwarze 1502: continue;
1503: }
1.28 schwarze 1504: sec = arch = title = "";
1.11 schwarze 1505: src_form = 0;
1.6 schwarze 1506: p = strrchr(buf, '\0');
1507: while (p-- > buf) {
1.28 schwarze 1508: if ('\0' == *sec && '.' == *p) {
1.6 schwarze 1509: sec = p + 1;
1510: *p = '\0';
1.11 schwarze 1511: if ('0' == *sec)
1512: src_form |= MANDOC_FORM;
1513: else if ('1' <= *sec && '9' >= *sec)
1514: src_form |= MANDOC_SRC;
1.6 schwarze 1515: continue;
1516: }
1517: if ('/' != *p)
1518: continue;
1.28 schwarze 1519: if ('\0' == *title) {
1.6 schwarze 1520: title = p + 1;
1521: *p = '\0';
1522: continue;
1523: }
1.18 schwarze 1524: if (0 == strncmp("man", p + 1, 3))
1.11 schwarze 1525: src_form |= MANDOC_SRC;
1.18 schwarze 1526: else if (0 == strncmp("cat", p + 1, 3))
1.11 schwarze 1527: src_form |= MANDOC_FORM;
1.18 schwarze 1528: else
1.6 schwarze 1529: arch = p + 1;
1530: break;
1531: }
1.28 schwarze 1532: if ('\0' == *title) {
1533: if (warnings)
1534: fprintf(stderr,
1535: "%s: cannot deduce title "
1536: "from filename\n",
1537: argv[i]);
1.6 schwarze 1538: title = buf;
1.28 schwarze 1539: }
1.6 schwarze 1540:
1541: /*
1542: * Build the file structure.
1543: */
1544:
1.2 schwarze 1545: nof = mandoc_calloc(1, sizeof(struct of));
1.6 schwarze 1546: nof->fname = mandoc_strdup(argv[i]);
1.28 schwarze 1547: nof->sec = mandoc_strdup(sec);
1548: nof->arch = mandoc_strdup(arch);
1.6 schwarze 1549: nof->title = mandoc_strdup(title);
1.11 schwarze 1550: nof->src_form = src_form;
1.6 schwarze 1551:
1552: /*
1553: * Add the structure to the list.
1554: */
1555:
1.28 schwarze 1556: if (verb > 1)
1557: printf("%s: scheduling\n", argv[i]);
1.2 schwarze 1558: if (NULL == *of) {
1559: *of = nof;
1560: (*of)->first = nof;
1561: } else {
1562: nof->first = (*of)->first;
1563: (*of)->next = nof;
1564: *of = nof;
1565: }
1566: }
1567: }
1568:
1569: /*
1570: * Recursively build up a list of files to parse.
1571: * We use this instead of ftw() and so on because I don't want global
1572: * variables hanging around.
1.30 schwarze 1573: * This ignores the whatis.db and whatis.index files, but assumes that
1.2 schwarze 1574: * everything else is a manual.
1575: * Pass in a pointer to a NULL structure for the first invocation.
1576: */
1.26 schwarze 1577: static void
1.6 schwarze 1578: ofile_dirbuild(const char *dir, const char* psec, const char *parch,
1.13 schwarze 1579: int p_src_form, struct of **of)
1.2 schwarze 1580: {
1581: char buf[MAXPATHLEN];
1582: size_t sz;
1583: DIR *d;
1.6 schwarze 1584: const char *fn, *sec, *arch;
1.11 schwarze 1585: char *p, *q, *suffix;
1.2 schwarze 1586: struct of *nof;
1587: struct dirent *dp;
1.11 schwarze 1588: int src_form;
1.2 schwarze 1589:
1590: if (NULL == (d = opendir(dir))) {
1.28 schwarze 1591: if (warnings)
1592: perror(dir);
1593: return;
1.2 schwarze 1594: }
1595:
1596: while (NULL != (dp = readdir(d))) {
1597: fn = dp->d_name;
1.6 schwarze 1598:
1599: if ('.' == *fn)
1600: continue;
1601:
1.11 schwarze 1602: src_form = p_src_form;
1603:
1.2 schwarze 1604: if (DT_DIR == dp->d_type) {
1.6 schwarze 1605: sec = psec;
1606: arch = parch;
1607:
1608: /*
1.8 schwarze 1609: * By default, only use directories called:
1.11 schwarze 1610: * man<section>/[<arch>/] or
1611: * cat<section>/[<arch>/]
1.6 schwarze 1612: */
1613:
1.28 schwarze 1614: if ('\0' == *sec) {
1.11 schwarze 1615: if(0 == strncmp("man", fn, 3)) {
1616: src_form |= MANDOC_SRC;
1.6 schwarze 1617: sec = fn + 3;
1.11 schwarze 1618: } else if (0 == strncmp("cat", fn, 3)) {
1619: src_form |= MANDOC_FORM;
1620: sec = fn + 3;
1.28 schwarze 1621: } else {
1622: if (warnings) fprintf(stderr,
1623: "%s/%s: bad section\n",
1624: dir, fn);
1625: if (use_all)
1626: sec = fn;
1627: else
1628: continue;
1629: }
1630: } else if ('\0' == *arch) {
1631: if (NULL != strchr(fn, '.')) {
1632: if (warnings) fprintf(stderr,
1633: "%s/%s: bad architecture\n",
1634: dir, fn);
1635: if (0 == use_all)
1636: continue;
1637: }
1638: arch = fn;
1639: } else {
1640: if (warnings) fprintf(stderr, "%s/%s: "
1641: "excessive subdirectory\n", dir, fn);
1642: if (0 == use_all)
1.6 schwarze 1643: continue;
1.28 schwarze 1644: }
1.2 schwarze 1645:
1646: buf[0] = '\0';
1647: strlcat(buf, dir, MAXPATHLEN);
1648: strlcat(buf, "/", MAXPATHLEN);
1649: sz = strlcat(buf, fn, MAXPATHLEN);
1650:
1.6 schwarze 1651: if (MAXPATHLEN <= sz) {
1.28 schwarze 1652: if (warnings) fprintf(stderr, "%s/%s: "
1653: "path too long\n", dir, fn);
1654: continue;
1.6 schwarze 1655: }
1.28 schwarze 1656:
1657: if (verb > 1)
1658: printf("%s: scanning\n", buf);
1.6 schwarze 1659:
1.26 schwarze 1660: ofile_dirbuild(buf, sec, arch, src_form, of);
1.28 schwarze 1661: continue;
1.6 schwarze 1662: }
1.26 schwarze 1663:
1.28 schwarze 1664: if (DT_REG != dp->d_type) {
1665: if (warnings)
1666: fprintf(stderr,
1667: "%s/%s: not a regular file\n",
1668: dir, fn);
1669: continue;
1670: }
1671: if (!strcmp(MANDOC_DB, fn) || !strcmp(MANDOC_IDX, fn))
1.6 schwarze 1672: continue;
1.28 schwarze 1673: if ('\0' == *psec) {
1674: if (warnings)
1675: fprintf(stderr,
1676: "%s/%s: file outside section\n",
1677: dir, fn);
1678: if (0 == use_all)
1679: continue;
1680: }
1.6 schwarze 1681:
1682: /*
1.8 schwarze 1683: * By default, skip files where the file name suffix
1684: * does not agree with the section directory
1685: * they are located in.
1.6 schwarze 1686: */
1687:
1688: suffix = strrchr(fn, '.');
1.28 schwarze 1689: if (NULL == suffix) {
1690: if (warnings)
1691: fprintf(stderr,
1692: "%s/%s: no filename suffix\n",
1693: dir, fn);
1694: if (0 == use_all)
1.2 schwarze 1695: continue;
1.28 schwarze 1696: } else if ((MANDOC_SRC & src_form &&
1697: strcmp(suffix + 1, psec)) ||
1.11 schwarze 1698: (MANDOC_FORM & src_form &&
1.28 schwarze 1699: strcmp(suffix + 1, "0"))) {
1700: if (warnings)
1701: fprintf(stderr,
1702: "%s/%s: wrong filename suffix\n",
1703: dir, fn);
1704: if (0 == use_all)
1705: continue;
1.11 schwarze 1706: if ('0' == suffix[1])
1707: src_form |= MANDOC_FORM;
1708: else if ('1' <= suffix[1] && '9' >= suffix[1])
1709: src_form |= MANDOC_SRC;
1710: }
1711:
1712: /*
1713: * Skip formatted manuals if a source version is
1714: * available. Ignore the age: it is very unlikely
1715: * that people install newer formatted base manuals
1716: * when they used to have source manuals before,
1717: * and in ports, old manuals get removed on update.
1718: */
1719: if (0 == use_all && MANDOC_FORM & src_form &&
1.28 schwarze 1720: '\0' != *psec) {
1.11 schwarze 1721: buf[0] = '\0';
1722: strlcat(buf, dir, MAXPATHLEN);
1723: p = strrchr(buf, '/');
1.28 schwarze 1724: if ('\0' != *parch && NULL != p)
1.23 schwarze 1725: for (p--; p > buf; p--)
1726: if ('/' == *p)
1727: break;
1.11 schwarze 1728: if (NULL == p)
1729: p = buf;
1730: else
1731: p++;
1732: if (0 == strncmp("cat", p, 3))
1733: memcpy(p, "man", 3);
1734: strlcat(buf, "/", MAXPATHLEN);
1735: sz = strlcat(buf, fn, MAXPATHLEN);
1736: if (sz >= MAXPATHLEN) {
1.28 schwarze 1737: if (warnings) fprintf(stderr,
1738: "%s/%s: path too long\n",
1739: dir, fn);
1.2 schwarze 1740: continue;
1.11 schwarze 1741: }
1742: q = strrchr(buf, '.');
1743: if (NULL != q && p < q++) {
1744: *q = '\0';
1745: sz = strlcat(buf, psec, MAXPATHLEN);
1746: if (sz >= MAXPATHLEN) {
1.28 schwarze 1747: if (warnings) fprintf(stderr,
1748: "%s/%s: path too long\n",
1749: dir, fn);
1.11 schwarze 1750: continue;
1751: }
1.26 schwarze 1752: if (0 == access(buf, R_OK))
1.11 schwarze 1753: continue;
1754: }
1.2 schwarze 1755: }
1756:
1.28 schwarze 1757: buf[0] = '\0';
1.26 schwarze 1758: assert('.' == dir[0]);
1.28 schwarze 1759: if ('/' == dir[1]) {
1760: strlcat(buf, dir + 2, MAXPATHLEN);
1761: strlcat(buf, "/", MAXPATHLEN);
1762: }
1.2 schwarze 1763: sz = strlcat(buf, fn, MAXPATHLEN);
1764: if (sz >= MAXPATHLEN) {
1.28 schwarze 1765: if (warnings) fprintf(stderr,
1766: "%s/%s: path too long\n", dir, fn);
1.11 schwarze 1767: continue;
1.2 schwarze 1768: }
1769:
1770: nof = mandoc_calloc(1, sizeof(struct of));
1771: nof->fname = mandoc_strdup(buf);
1.28 schwarze 1772: nof->sec = mandoc_strdup(psec);
1773: nof->arch = mandoc_strdup(parch);
1.11 schwarze 1774: nof->src_form = src_form;
1.8 schwarze 1775:
1776: /*
1777: * Remember the file name without the extension,
1778: * to be used as the page title in the database.
1779: */
1780:
1.6 schwarze 1781: if (NULL != suffix)
1782: *suffix = '\0';
1783: nof->title = mandoc_strdup(fn);
1.2 schwarze 1784:
1.11 schwarze 1785: /*
1786: * Add the structure to the list.
1787: */
1788:
1.28 schwarze 1789: if (verb > 1)
1790: printf("%s: scheduling\n", buf);
1.31 schwarze 1791:
1.2 schwarze 1792: if (NULL == *of) {
1793: *of = nof;
1794: (*of)->first = nof;
1795: } else {
1796: nof->first = (*of)->first;
1797: (*of)->next = nof;
1798: *of = nof;
1799: }
1800: }
1801:
1.4 schwarze 1802: closedir(d);
1.2 schwarze 1803: }
1804:
1805: static void
1806: ofile_free(struct of *of)
1807: {
1808: struct of *nof;
1809:
1.31 schwarze 1810: if (NULL != of)
1811: of = of->first;
1812:
1813: while (NULL != of) {
1.2 schwarze 1814: nof = of->next;
1815: free(of->fname);
1.6 schwarze 1816: free(of->sec);
1817: free(of->arch);
1818: free(of->title);
1.2 schwarze 1819: free(of);
1820: of = nof;
1821: }
1.1 schwarze 1822: }