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