Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.19
1.19 ! schwarze 1: /* $Id: mandocdb.c,v 1.18 2011/12/04 14:11:55 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>
20: #include <sys/stat.h>
1.1 schwarze 21:
22: #include <assert.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:
1.2 schwarze 45: /* Tiny list for files. No need to bring in QUEUE. */
46:
47: struct of {
48: char *fname; /* heap-allocated */
1.6 schwarze 49: char *sec;
50: char *arch;
51: char *title;
1.11 schwarze 52: int src_form;
1.2 schwarze 53: struct of *next; /* NULL for last one */
54: struct of *first; /* first in list */
55: };
56:
1.1 schwarze 57: /* Buffer for storing growable data. */
58:
59: struct buf {
60: char *cp;
1.2 schwarze 61: size_t len; /* current length */
62: size_t size; /* total buffer size */
1.1 schwarze 63: };
64:
65: /* Operation we're going to perform. */
66:
67: enum op {
68: OP_NEW = 0, /* new database */
1.2 schwarze 69: OP_UPDATE, /* delete/add entries in existing database */
1.1 schwarze 70: OP_DELETE /* delete entries from existing database */
71: };
72:
73: #define MAN_ARGS DB *hash, \
74: struct buf *buf, \
75: struct buf *dbuf, \
76: const struct man_node *n
77: #define MDOC_ARGS DB *hash, \
78: struct buf *buf, \
79: struct buf *dbuf, \
80: const struct mdoc_node *n, \
81: const struct mdoc_meta *m
82:
83: static void buf_appendmdoc(struct buf *,
84: const struct mdoc_node *, int);
85: static void buf_append(struct buf *, const char *);
86: static void buf_appendb(struct buf *,
87: const void *, size_t);
88: static void dbt_put(DB *, const char *, DBT *, DBT *);
1.9 schwarze 89: static void hash_put(DB *, const struct buf *, uint64_t);
1.1 schwarze 90: static void hash_reset(DB **);
1.2 schwarze 91: static void index_merge(const struct of *, struct mparse *,
1.13 schwarze 92: struct buf *, struct buf *, DB *,
93: DB *, const char *, DB *, const char *,
1.2 schwarze 94: recno_t, const recno_t *, size_t);
95: static void index_prune(const struct of *, DB *,
96: const char *, DB *, const char *,
1.13 schwarze 97: recno_t *, recno_t **, size_t *);
98: static void ofile_argbuild(int, char *[], struct of **);
1.6 schwarze 99: static int ofile_dirbuild(const char *, const char *,
1.13 schwarze 100: const char *, int, struct of **);
1.2 schwarze 101: static void ofile_free(struct of *);
1.11 schwarze 102: static void pformatted(DB *, struct buf *, struct buf *,
103: const struct of *);
1.1 schwarze 104: static int pman_node(MAN_ARGS);
105: static void pmdoc_node(MDOC_ARGS);
1.19 ! schwarze 106: static int pmdoc_head(MDOC_ARGS);
! 107: static int pmdoc_body(MDOC_ARGS);
! 108: static int pmdoc_Fd(MDOC_ARGS);
! 109: static int pmdoc_In(MDOC_ARGS);
! 110: static int pmdoc_Fn(MDOC_ARGS);
! 111: static int pmdoc_Nd(MDOC_ARGS);
! 112: static int pmdoc_Nm(MDOC_ARGS);
! 113: static int pmdoc_Sh(MDOC_ARGS);
! 114: static int pmdoc_St(MDOC_ARGS);
! 115: static int pmdoc_Xr(MDOC_ARGS);
1.1 schwarze 116: static void usage(void);
117:
1.19 ! schwarze 118: #define MDOCF_CHILD 0x01 /* Automatically index child nodes. */
1.1 schwarze 119:
1.19 ! schwarze 120: struct mdoc_handler {
! 121: int (*fp)(MDOC_ARGS); /* Optional handler. */
! 122: uint64_t mask; /* Set unless handler returns 0. */
! 123: int flags; /* For use by pmdoc_node. */
! 124: };
! 125:
! 126: static const struct mdoc_handler mdocs[MDOC_MAX] = {
! 127: { NULL, 0, 0 }, /* Ap */
! 128: { NULL, 0, 0 }, /* Dd */
! 129: { NULL, 0, 0 }, /* Dt */
! 130: { NULL, 0, 0 }, /* Os */
! 131: { pmdoc_Sh, TYPE_Sh, MDOCF_CHILD }, /* Sh */
! 132: { pmdoc_head, TYPE_Ss, MDOCF_CHILD }, /* Ss */
! 133: { NULL, 0, 0 }, /* Pp */
! 134: { NULL, 0, 0 }, /* D1 */
! 135: { NULL, 0, 0 }, /* Dl */
! 136: { NULL, 0, 0 }, /* Bd */
! 137: { NULL, 0, 0 }, /* Ed */
! 138: { NULL, 0, 0 }, /* Bl */
! 139: { NULL, 0, 0 }, /* El */
! 140: { NULL, 0, 0 }, /* It */
! 141: { NULL, 0, 0 }, /* Ad */
! 142: { NULL, TYPE_An, MDOCF_CHILD }, /* An */
! 143: { NULL, TYPE_Ar, MDOCF_CHILD }, /* Ar */
! 144: { NULL, TYPE_Cd, MDOCF_CHILD }, /* Cd */
! 145: { NULL, TYPE_Cm, MDOCF_CHILD }, /* Cm */
! 146: { NULL, TYPE_Dv, MDOCF_CHILD }, /* Dv */
! 147: { NULL, TYPE_Er, MDOCF_CHILD }, /* Er */
! 148: { NULL, TYPE_Ev, MDOCF_CHILD }, /* Ev */
! 149: { NULL, 0, 0 }, /* Ex */
! 150: { NULL, TYPE_Fa, MDOCF_CHILD }, /* Fa */
! 151: { pmdoc_Fd, TYPE_In, 0 }, /* Fd */
! 152: { NULL, TYPE_Fl, MDOCF_CHILD }, /* Fl */
! 153: { pmdoc_Fn, 0, 0 }, /* Fn */
! 154: { NULL, TYPE_Ft, MDOCF_CHILD }, /* Ft */
! 155: { NULL, TYPE_Ic, MDOCF_CHILD }, /* Ic */
! 156: { pmdoc_In, TYPE_In, 0 }, /* In */
! 157: { NULL, TYPE_Li, MDOCF_CHILD }, /* Li */
! 158: { pmdoc_Nd, TYPE_Nd, MDOCF_CHILD }, /* Nd */
! 159: { pmdoc_Nm, TYPE_Nm, MDOCF_CHILD }, /* Nm */
! 160: { NULL, 0, 0 }, /* Op */
! 161: { NULL, 0, 0 }, /* Ot */
! 162: { NULL, TYPE_Pa, MDOCF_CHILD }, /* Pa */
! 163: { NULL, 0, 0 }, /* Rv */
! 164: { pmdoc_St, TYPE_St, 0 }, /* St */
! 165: { NULL, TYPE_Va, MDOCF_CHILD }, /* Va */
! 166: { pmdoc_body, TYPE_Va, MDOCF_CHILD }, /* Vt */
! 167: { pmdoc_Xr, TYPE_Xr, 0 }, /* Xr */
! 168: { NULL, 0, 0 }, /* %A */
! 169: { NULL, 0, 0 }, /* %B */
! 170: { NULL, 0, 0 }, /* %D */
! 171: { NULL, 0, 0 }, /* %I */
! 172: { NULL, 0, 0 }, /* %J */
! 173: { NULL, 0, 0 }, /* %N */
! 174: { NULL, 0, 0 }, /* %O */
! 175: { NULL, 0, 0 }, /* %P */
! 176: { NULL, 0, 0 }, /* %R */
! 177: { NULL, 0, 0 }, /* %T */
! 178: { NULL, 0, 0 }, /* %V */
! 179: { NULL, 0, 0 }, /* Ac */
! 180: { NULL, 0, 0 }, /* Ao */
! 181: { NULL, 0, 0 }, /* Aq */
! 182: { NULL, TYPE_At, MDOCF_CHILD }, /* At */
! 183: { NULL, 0, 0 }, /* Bc */
! 184: { NULL, 0, 0 }, /* Bf */
! 185: { NULL, 0, 0 }, /* Bo */
! 186: { NULL, 0, 0 }, /* Bq */
! 187: { NULL, TYPE_Bsx, MDOCF_CHILD }, /* Bsx */
! 188: { NULL, TYPE_Bx, MDOCF_CHILD }, /* Bx */
! 189: { NULL, 0, 0 }, /* Db */
! 190: { NULL, 0, 0 }, /* Dc */
! 191: { NULL, 0, 0 }, /* Do */
! 192: { NULL, 0, 0 }, /* Dq */
! 193: { NULL, 0, 0 }, /* Ec */
! 194: { NULL, 0, 0 }, /* Ef */
! 195: { NULL, TYPE_Em, MDOCF_CHILD }, /* Em */
! 196: { NULL, 0, 0 }, /* Eo */
! 197: { NULL, TYPE_Fx, MDOCF_CHILD }, /* Fx */
! 198: { NULL, TYPE_Ms, MDOCF_CHILD }, /* Ms */
! 199: { NULL, 0, 0 }, /* No */
! 200: { NULL, 0, 0 }, /* Ns */
! 201: { NULL, TYPE_Nx, MDOCF_CHILD }, /* Nx */
! 202: { NULL, TYPE_Ox, MDOCF_CHILD }, /* Ox */
! 203: { NULL, 0, 0 }, /* Pc */
! 204: { NULL, 0, 0 }, /* Pf */
! 205: { NULL, 0, 0 }, /* Po */
! 206: { NULL, 0, 0 }, /* Pq */
! 207: { NULL, 0, 0 }, /* Qc */
! 208: { NULL, 0, 0 }, /* Ql */
! 209: { NULL, 0, 0 }, /* Qo */
! 210: { NULL, 0, 0 }, /* Qq */
! 211: { NULL, 0, 0 }, /* Re */
! 212: { NULL, 0, 0 }, /* Rs */
! 213: { NULL, 0, 0 }, /* Sc */
! 214: { NULL, 0, 0 }, /* So */
! 215: { NULL, 0, 0 }, /* Sq */
! 216: { NULL, 0, 0 }, /* Sm */
! 217: { NULL, 0, 0 }, /* Sx */
! 218: { NULL, TYPE_Sy, MDOCF_CHILD }, /* Sy */
! 219: { NULL, TYPE_Tn, MDOCF_CHILD }, /* Tn */
! 220: { NULL, 0, 0 }, /* Ux */
! 221: { NULL, 0, 0 }, /* Xc */
! 222: { NULL, 0, 0 }, /* Xo */
! 223: { pmdoc_head, TYPE_Fn, 0 }, /* Fo */
! 224: { NULL, 0, 0 }, /* Fc */
! 225: { NULL, 0, 0 }, /* Oo */
! 226: { NULL, 0, 0 }, /* Oc */
! 227: { NULL, 0, 0 }, /* Bk */
! 228: { NULL, 0, 0 }, /* Ek */
! 229: { NULL, 0, 0 }, /* Bt */
! 230: { NULL, 0, 0 }, /* Hf */
! 231: { NULL, 0, 0 }, /* Fr */
! 232: { NULL, 0, 0 }, /* Ud */
! 233: { NULL, TYPE_Lb, MDOCF_CHILD }, /* Lb */
! 234: { NULL, 0, 0 }, /* Lp */
! 235: { NULL, TYPE_Lk, MDOCF_CHILD }, /* Lk */
! 236: { NULL, TYPE_Mt, MDOCF_CHILD }, /* Mt */
! 237: { NULL, 0, 0 }, /* Brq */
! 238: { NULL, 0, 0 }, /* Bro */
! 239: { NULL, 0, 0 }, /* Brc */
! 240: { NULL, 0, 0 }, /* %C */
! 241: { NULL, 0, 0 }, /* Es */
! 242: { NULL, 0, 0 }, /* En */
! 243: { NULL, TYPE_Dx, MDOCF_CHILD }, /* Dx */
! 244: { NULL, 0, 0 }, /* %Q */
! 245: { NULL, 0, 0 }, /* br */
! 246: { NULL, 0, 0 }, /* sp */
! 247: { NULL, 0, 0 }, /* %U */
! 248: { NULL, 0, 0 }, /* Ta */
1.1 schwarze 249: };
250:
251: static const char *progname;
1.13 schwarze 252: static int use_all; /* Use all directories and files. */
253: static int verb; /* Output verbosity level. */
1.1 schwarze 254:
255: int
1.3 schwarze 256: mandocdb(int argc, char *argv[])
1.1 schwarze 257: {
258: struct mparse *mp; /* parse sequence */
1.10 schwarze 259: struct manpaths dirs;
1.1 schwarze 260: enum op op; /* current operation */
1.2 schwarze 261: const char *dir;
1.1 schwarze 262: char ibuf[MAXPATHLEN], /* index fname */
1.2 schwarze 263: fbuf[MAXPATHLEN]; /* btree fname */
1.13 schwarze 264: int ch, i, flags;
1.1 schwarze 265: DB *idx, /* index database */
266: *db, /* keyword database */
267: *hash; /* temporary keyword hashtable */
268: BTREEINFO info; /* btree configuration */
1.8 schwarze 269: recno_t maxrec; /* last record number in the index */
270: recno_t *recs; /* the numbers of all empty records */
1.2 schwarze 271: size_t sz1, sz2,
1.8 schwarze 272: recsz, /* number of allocated slots in recs */
273: reccur; /* current number of empty records */
1.1 schwarze 274: struct buf buf, /* keyword buffer */
275: dbuf; /* description buffer */
1.2 schwarze 276: struct of *of; /* list of files for processing */
1.1 schwarze 277: extern int optind;
278: extern char *optarg;
279:
280: progname = strrchr(argv[0], '/');
281: if (progname == NULL)
282: progname = argv[0];
283: else
284: ++progname;
285:
1.10 schwarze 286: memset(&dirs, 0, sizeof(struct manpaths));
287:
1.1 schwarze 288: verb = 0;
1.6 schwarze 289: use_all = 0;
1.2 schwarze 290: of = NULL;
1.1 schwarze 291: db = idx = NULL;
292: mp = NULL;
293: hash = NULL;
294: recs = NULL;
295: recsz = reccur = 0;
296: maxrec = 0;
297: op = OP_NEW;
1.2 schwarze 298: dir = NULL;
1.1 schwarze 299:
1.6 schwarze 300: while (-1 != (ch = getopt(argc, argv, "ad:u:v")))
1.1 schwarze 301: switch (ch) {
1.6 schwarze 302: case ('a'):
303: use_all = 1;
304: break;
1.1 schwarze 305: case ('d'):
306: dir = optarg;
1.2 schwarze 307: op = OP_UPDATE;
1.1 schwarze 308: break;
1.2 schwarze 309: case ('u'):
310: dir = optarg;
1.1 schwarze 311: op = OP_DELETE;
312: break;
313: case ('v'):
314: verb++;
315: break;
316: default:
317: usage();
318: return((int)MANDOCLEVEL_BADARG);
319: }
320:
321: argc -= optind;
322: argv += optind;
323:
1.2 schwarze 324: memset(&info, 0, sizeof(BTREEINFO));
1.17 schwarze 325: info.lorder = 4321;
1.2 schwarze 326: info.flags = R_DUP;
327:
328: mp = mparse_alloc(MPARSE_AUTO, MANDOCLEVEL_FATAL, NULL, NULL);
329:
330: memset(&buf, 0, sizeof(struct buf));
331: memset(&dbuf, 0, sizeof(struct buf));
332:
333: buf.size = dbuf.size = MANDOC_BUFSZ;
334:
335: buf.cp = mandoc_malloc(buf.size);
336: dbuf.cp = mandoc_malloc(dbuf.size);
337:
338: flags = OP_NEW == op ? O_CREAT|O_TRUNC|O_RDWR : O_CREAT|O_RDWR;
1.1 schwarze 339:
1.2 schwarze 340: if (OP_UPDATE == op || OP_DELETE == op) {
341: ibuf[0] = fbuf[0] = '\0';
1.1 schwarze 342:
1.2 schwarze 343: strlcat(fbuf, dir, MAXPATHLEN);
344: strlcat(fbuf, "/", MAXPATHLEN);
345: sz1 = strlcat(fbuf, MANDOC_DB, MAXPATHLEN);
346:
347: strlcat(ibuf, dir, MAXPATHLEN);
348: strlcat(ibuf, "/", MAXPATHLEN);
349: sz2 = strlcat(ibuf, MANDOC_IDX, MAXPATHLEN);
350:
351: if (sz1 >= MAXPATHLEN || sz2 >= MAXPATHLEN) {
352: fprintf(stderr, "%s: Path too long\n", dir);
353: exit((int)MANDOCLEVEL_BADARG);
354: }
1.1 schwarze 355:
1.2 schwarze 356: db = dbopen(fbuf, flags, 0644, DB_BTREE, &info);
357: idx = dbopen(ibuf, flags, 0644, DB_RECNO, NULL);
1.1 schwarze 358:
1.2 schwarze 359: if (NULL == db) {
360: perror(fbuf);
361: exit((int)MANDOCLEVEL_SYSERR);
1.8 schwarze 362: } else if (NULL == idx) {
1.2 schwarze 363: perror(ibuf);
364: exit((int)MANDOCLEVEL_SYSERR);
365: }
1.1 schwarze 366:
1.2 schwarze 367: if (verb > 2) {
368: printf("%s: Opened\n", fbuf);
369: printf("%s: Opened\n", ibuf);
370: }
1.1 schwarze 371:
1.13 schwarze 372: ofile_argbuild(argc, argv, &of);
1.2 schwarze 373: if (NULL == of)
374: goto out;
375:
376: of = of->first;
377:
1.13 schwarze 378: index_prune(of, db, fbuf, idx, ibuf,
1.2 schwarze 379: &maxrec, &recs, &recsz);
380:
1.15 schwarze 381: /*
382: * Go to the root of the respective manual tree
383: * such that .so links work. In case of failure,
384: * just prod on, even though .so links won't work.
385: */
386:
1.14 schwarze 387: if (OP_UPDATE == op) {
388: chdir(dir);
1.10 schwarze 389: index_merge(of, mp, &dbuf, &buf, hash,
1.13 schwarze 390: db, fbuf, idx, ibuf,
391: maxrec, recs, reccur);
1.14 schwarze 392: }
1.1 schwarze 393:
394: goto out;
395: }
396:
1.10 schwarze 397: /*
398: * Configure the directories we're going to scan.
399: * If we have command-line arguments, use them.
400: * If not, we use man(1)'s method (see mandocdb.8).
401: */
402:
403: if (argc > 0) {
404: dirs.paths = mandoc_malloc(argc * sizeof(char *));
405: dirs.sz = argc;
406: for (i = 0; i < argc; i++)
407: dirs.paths[i] = mandoc_strdup(argv[i]);
408: } else
409: manpath_parse(&dirs, NULL, NULL);
1.7 schwarze 410:
1.10 schwarze 411: for (i = 0; i < dirs.sz; i++) {
1.2 schwarze 412: ibuf[0] = fbuf[0] = '\0';
413:
1.10 schwarze 414: strlcat(fbuf, dirs.paths[i], MAXPATHLEN);
1.2 schwarze 415: strlcat(fbuf, "/", MAXPATHLEN);
416: sz1 = strlcat(fbuf, MANDOC_DB, MAXPATHLEN);
417:
1.10 schwarze 418: strlcat(ibuf, dirs.paths[i], MAXPATHLEN);
1.2 schwarze 419: strlcat(ibuf, "/", MAXPATHLEN);
420: sz2 = strlcat(ibuf, MANDOC_IDX, MAXPATHLEN);
421:
422: if (sz1 >= MAXPATHLEN || sz2 >= MAXPATHLEN) {
1.10 schwarze 423: fprintf(stderr, "%s: Path too long\n",
424: dirs.paths[i]);
1.2 schwarze 425: exit((int)MANDOCLEVEL_BADARG);
426: }
427:
1.7 schwarze 428: if (db)
429: (*db->close)(db);
430: if (idx)
431: (*idx->close)(idx);
432:
1.2 schwarze 433: db = dbopen(fbuf, flags, 0644, DB_BTREE, &info);
434: idx = dbopen(ibuf, flags, 0644, DB_RECNO, NULL);
1.1 schwarze 435:
1.2 schwarze 436: if (NULL == db) {
437: perror(fbuf);
438: exit((int)MANDOCLEVEL_SYSERR);
1.8 schwarze 439: } else if (NULL == idx) {
1.2 schwarze 440: perror(ibuf);
441: exit((int)MANDOCLEVEL_SYSERR);
442: }
1.1 schwarze 443:
1.2 schwarze 444: if (verb > 2) {
445: printf("%s: Truncated\n", fbuf);
446: printf("%s: Truncated\n", ibuf);
447: }
1.1 schwarze 448:
1.2 schwarze 449: ofile_free(of);
450: of = NULL;
1.1 schwarze 451:
1.10 schwarze 452: if ( ! ofile_dirbuild(dirs.paths[i], NULL, NULL,
1.13 schwarze 453: 0, &of))
1.2 schwarze 454: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 455:
1.2 schwarze 456: if (NULL == of)
457: continue;
1.1 schwarze 458:
1.2 schwarze 459: of = of->first;
1.15 schwarze 460:
461: /*
462: * Go to the root of the respective manual tree
463: * such that .so links work. In case of failure,
464: * just prod on, even though .so links won't work.
465: */
1.1 schwarze 466:
1.14 schwarze 467: chdir(dirs.paths[i]);
1.10 schwarze 468: index_merge(of, mp, &dbuf, &buf, hash, db, fbuf,
1.13 schwarze 469: idx, ibuf, maxrec, recs, reccur);
1.1 schwarze 470: }
471:
1.2 schwarze 472: out:
473: if (db)
474: (*db->close)(db);
475: if (idx)
476: (*idx->close)(idx);
477: if (hash)
478: (*hash->close)(hash);
479: if (mp)
480: mparse_free(mp);
1.1 schwarze 481:
1.10 schwarze 482: manpath_free(&dirs);
1.2 schwarze 483: ofile_free(of);
484: free(buf.cp);
485: free(dbuf.cp);
486: free(recs);
1.1 schwarze 487:
1.2 schwarze 488: return(MANDOCLEVEL_OK);
489: }
1.1 schwarze 490:
1.2 schwarze 491: void
492: index_merge(const struct of *of, struct mparse *mp,
1.13 schwarze 493: struct buf *dbuf, struct buf *buf, DB *hash,
494: DB *db, const char *dbf, DB *idx, const char *idxf,
1.2 schwarze 495: recno_t maxrec, const recno_t *recs, size_t reccur)
496: {
497: recno_t rec;
498: int ch;
499: DBT key, val;
500: struct mdoc *mdoc;
501: struct man *man;
502: const char *fn, *msec, *mtitle, *arch;
503: size_t sv;
504: unsigned seq;
1.9 schwarze 505: struct db_val vbuf;
1.1 schwarze 506:
1.2 schwarze 507: for (rec = 0; of; of = of->next) {
508: fn = of->fname;
1.11 schwarze 509:
510: /*
511: * Reclaim an empty index record, if available.
512: */
513:
1.2 schwarze 514: if (reccur > 0) {
515: --reccur;
516: rec = recs[(int)reccur];
517: } else if (maxrec > 0) {
518: rec = maxrec;
519: maxrec = 0;
1.1 schwarze 520: } else
521: rec++;
522:
523: mparse_reset(mp);
524: hash_reset(&hash);
1.11 schwarze 525: mdoc = NULL;
526: man = NULL;
1.1 schwarze 527:
1.11 schwarze 528: /*
529: * Try interpreting the file as mdoc(7) or man(7)
530: * source code, unless it is already known to be
531: * formatted. Fall back to formatted mode.
532: */
533:
534: if ((MANDOC_SRC & of->src_form ||
535: ! (MANDOC_FORM & of->src_form)) &&
536: MANDOCLEVEL_FATAL > mparse_readfd(mp, -1, fn))
537: mparse_result(mp, &mdoc, &man);
538:
539: if (NULL != mdoc) {
540: msec = mdoc_meta(mdoc)->msec;
541: arch = mdoc_meta(mdoc)->arch;
542: mtitle = mdoc_meta(mdoc)->title;
543: } else if (NULL != man) {
544: msec = man_meta(man)->msec;
545: arch = NULL;
546: mtitle = man_meta(man)->title;
547: } else {
548: msec = of->sec;
549: arch = of->arch;
550: mtitle = of->title;
1.1 schwarze 551: }
552:
1.6 schwarze 553: /*
1.8 schwarze 554: * By default, skip a file if the manual section
555: * and architecture given in the file disagree
556: * with the directory where the file is located.
1.6 schwarze 557: */
558:
559: if (0 == use_all) {
560: assert(of->sec);
561: assert(msec);
562: if (strcmp(msec, of->sec))
563: continue;
564:
565: if (NULL == arch) {
566: if (NULL != of->arch)
567: continue;
568: } else if (NULL == of->arch ||
569: strcmp(arch, of->arch))
570: continue;
571: }
572:
1.1 schwarze 573: if (NULL == arch)
574: arch = "";
575:
576: /*
1.8 schwarze 577: * By default, skip a file if the title given
578: * in the file disagrees with the file name.
579: * If both agree, use the file name as the title,
580: * because the one in the file usually is all caps.
1.6 schwarze 581: */
582:
583: assert(of->title);
584: assert(mtitle);
585:
586: if (0 == strcasecmp(mtitle, of->title))
587: mtitle = of->title;
588: else if (0 == use_all)
589: continue;
590:
591: /*
1.1 schwarze 592: * The index record value consists of a nil-terminated
593: * filename, a nil-terminated manual section, and a
594: * nil-terminated description. Since the description
595: * may not be set, we set a sentinel to see if we're
596: * going to write a nil byte in its place.
597: */
598:
1.2 schwarze 599: dbuf->len = 0;
1.12 schwarze 600: buf_append(dbuf, mdoc ? "mdoc" : (man ? "man" : "cat"));
1.2 schwarze 601: buf_appendb(dbuf, fn, strlen(fn) + 1);
602: buf_appendb(dbuf, msec, strlen(msec) + 1);
603: buf_appendb(dbuf, mtitle, strlen(mtitle) + 1);
604: buf_appendb(dbuf, arch, strlen(arch) + 1);
1.1 schwarze 605:
1.2 schwarze 606: sv = dbuf->len;
1.1 schwarze 607:
608: /* Fix the record number in the btree value. */
609:
610: if (mdoc)
1.2 schwarze 611: pmdoc_node(hash, buf, dbuf,
1.1 schwarze 612: mdoc_node(mdoc), mdoc_meta(mdoc));
1.11 schwarze 613: else if (man)
1.2 schwarze 614: pman_node(hash, buf, dbuf, man_node(man));
1.11 schwarze 615: else
616: pformatted(hash, buf, dbuf, of);
1.1 schwarze 617:
618: /*
619: * Copy from the in-memory hashtable of pending keywords
620: * into the database.
621: */
622:
1.17 schwarze 623: vbuf.rec = htobe32(rec);
1.1 schwarze 624: seq = R_FIRST;
625: while (0 == (ch = (*hash->seq)(hash, &key, &val, seq))) {
626: seq = R_NEXT;
1.17 schwarze 627: vbuf.mask = htobe64(*(uint64_t *)val.data);
1.9 schwarze 628: val.size = sizeof(struct db_val);
629: val.data = &vbuf;
1.2 schwarze 630: dbt_put(db, dbf, &key, &val);
1.1 schwarze 631: }
632: if (ch < 0) {
633: perror("hash");
634: exit((int)MANDOCLEVEL_SYSERR);
635: }
636:
637: /*
638: * Apply to the index. If we haven't had a description
639: * set, put an empty one in now.
640: */
641:
1.2 schwarze 642: if (dbuf->len == sv)
643: buf_appendb(dbuf, "", 1);
1.1 schwarze 644:
645: key.data = &rec;
646: key.size = sizeof(recno_t);
647:
1.2 schwarze 648: val.data = dbuf->cp;
649: val.size = dbuf->len;
1.1 schwarze 650:
1.2 schwarze 651: if (verb)
1.1 schwarze 652: printf("%s: Added index\n", fn);
1.16 schwarze 653:
1.2 schwarze 654: dbt_put(idx, idxf, &key, &val);
655: }
656: }
657:
658: /*
659: * Scan through all entries in the index file `idx' and prune those
660: * entries in `ofile'.
661: * Pruning consists of removing from `db', then invalidating the entry
662: * in `idx' (zeroing its value size).
663: */
664: static void
665: index_prune(const struct of *ofile, DB *db, const char *dbf,
1.13 schwarze 666: DB *idx, const char *idxf,
1.2 schwarze 667: recno_t *maxrec, recno_t **recs, size_t *recsz)
668: {
669: const struct of *of;
1.16 schwarze 670: const char *fn, *cp;
1.9 schwarze 671: struct db_val *vbuf;
1.2 schwarze 672: unsigned seq, sseq;
673: DBT key, val;
674: size_t reccur;
675: int ch;
676:
677: reccur = 0;
678: seq = R_FIRST;
679: while (0 == (ch = (*idx->seq)(idx, &key, &val, seq))) {
680: seq = R_NEXT;
681: *maxrec = *(recno_t *)key.data;
1.16 schwarze 682: cp = val.data;
683:
684: /* Deleted records are zero-sized. Skip them. */
685:
686: if (0 == val.size)
687: goto cont;
688:
689: /*
690: * Make sure we're sane.
691: * Read past our mdoc/man/cat type to the next string,
692: * then make sure it's bounded by a NUL.
693: * Failing any of these, we go into our error handler.
694: */
695:
696: if (NULL == (fn = memchr(cp, '\0', val.size)))
697: break;
698: if (++fn - cp >= (int)val.size)
699: break;
700: if (NULL == memchr(fn, '\0', val.size - (fn - cp)))
701: break;
702:
703: /*
704: * Search for the file in those we care about.
705: * XXX: build this into a tree. Too slow.
706: */
1.2 schwarze 707:
708: for (of = ofile; of; of = of->next)
709: if (0 == strcmp(fn, of->fname))
710: break;
711:
712: if (NULL == of)
713: continue;
714:
1.16 schwarze 715: /*
716: * Search through the keyword database, throwing out all
717: * references to our file.
718: */
719:
1.2 schwarze 720: sseq = R_FIRST;
721: while (0 == (ch = (*db->seq)(db, &key, &val, sseq))) {
722: sseq = R_NEXT;
1.16 schwarze 723: if (sizeof(struct db_val) != val.size)
724: break;
725:
1.9 schwarze 726: vbuf = val.data;
1.17 schwarze 727: if (*maxrec != betoh32(vbuf->rec))
1.2 schwarze 728: continue;
1.16 schwarze 729:
730: if ((ch = (*db->del)(db, &key, R_CURSOR)) < 0)
1.2 schwarze 731: break;
732: }
1.16 schwarze 733:
1.2 schwarze 734: if (ch < 0) {
735: perror(dbf);
736: exit((int)MANDOCLEVEL_SYSERR);
1.16 schwarze 737: } else if (1 != ch) {
738: fprintf(stderr, "%s: Corrupt database\n", dbf);
739: exit((int)MANDOCLEVEL_SYSERR);
1.2 schwarze 740: }
1.1 schwarze 741:
1.2 schwarze 742: if (verb)
743: printf("%s: Deleted index\n", fn);
1.1 schwarze 744:
1.2 schwarze 745: val.size = 0;
746: ch = (*idx->put)(idx, &key, &val, R_CURSOR);
1.1 schwarze 747:
1.16 schwarze 748: if (ch < 0)
749: break;
750: cont:
1.2 schwarze 751: if (reccur >= *recsz) {
752: *recsz += MANDOC_SLOP;
753: *recs = mandoc_realloc
754: (*recs, *recsz * sizeof(recno_t));
755: }
1.1 schwarze 756:
1.2 schwarze 757: (*recs)[(int)reccur] = *maxrec;
758: reccur++;
759: }
1.16 schwarze 760:
761: if (ch < 0) {
762: perror(idxf);
763: exit((int)MANDOCLEVEL_SYSERR);
764: } else if (1 != ch) {
765: fprintf(stderr, "%s: Corrupt index\n", idxf);
766: exit((int)MANDOCLEVEL_SYSERR);
767: }
768:
1.2 schwarze 769: (*maxrec)++;
1.1 schwarze 770: }
771:
772: /*
773: * Grow the buffer (if necessary) and copy in a binary string.
774: */
775: static void
776: buf_appendb(struct buf *buf, const void *cp, size_t sz)
777: {
778:
779: /* Overshoot by MANDOC_BUFSZ. */
780:
781: while (buf->len + sz >= buf->size) {
782: buf->size = buf->len + sz + MANDOC_BUFSZ;
783: buf->cp = mandoc_realloc(buf->cp, buf->size);
784: }
785:
786: memcpy(buf->cp + (int)buf->len, cp, sz);
787: buf->len += sz;
788: }
789:
790: /*
791: * Append a nil-terminated string to the buffer.
792: * This can be invoked multiple times.
793: * The buffer string will be nil-terminated.
794: * If invoked multiple times, a space is put between strings.
795: */
796: static void
797: buf_append(struct buf *buf, const char *cp)
798: {
799: size_t sz;
800:
801: if (0 == (sz = strlen(cp)))
802: return;
803:
804: if (buf->len)
805: buf->cp[(int)buf->len - 1] = ' ';
806:
807: buf_appendb(buf, cp, sz + 1);
808: }
809:
810: /*
811: * Recursively add all text from a given node.
812: * This is optimised for general mdoc nodes in this context, which do
813: * not consist of subexpressions and having a recursive call for n->next
814: * would be wasteful.
815: * The "f" variable should be 0 unless called from pmdoc_Nd for the
816: * description buffer, which does not start at the beginning of the
817: * buffer.
818: */
819: static void
820: buf_appendmdoc(struct buf *buf, const struct mdoc_node *n, int f)
821: {
822:
823: for ( ; n; n = n->next) {
824: if (n->child)
825: buf_appendmdoc(buf, n->child, f);
826:
827: if (MDOC_TEXT == n->type && f) {
828: f = 0;
829: buf_appendb(buf, n->string,
830: strlen(n->string) + 1);
831: } else if (MDOC_TEXT == n->type)
832: buf_append(buf, n->string);
833:
834: }
835: }
836:
837: static void
838: hash_reset(DB **db)
839: {
840: DB *hash;
841:
842: if (NULL != (hash = *db))
843: (*hash->close)(hash);
844:
1.2 schwarze 845: *db = dbopen(NULL, O_CREAT|O_RDWR, 0644, DB_HASH, NULL);
1.1 schwarze 846: if (NULL == *db) {
847: perror("hash");
848: exit((int)MANDOCLEVEL_SYSERR);
849: }
850: }
851:
852: /* ARGSUSED */
1.19 ! schwarze 853: static int
! 854: pmdoc_head(MDOC_ARGS)
! 855: {
! 856:
! 857: return(MDOC_HEAD == n->type);
! 858: }
! 859:
! 860: /* ARGSUSED */
! 861: static int
! 862: pmdoc_body(MDOC_ARGS)
! 863: {
! 864:
! 865: return(MDOC_BODY == n->type);
! 866: }
! 867:
! 868: /* ARGSUSED */
! 869: static int
1.1 schwarze 870: pmdoc_Fd(MDOC_ARGS)
871: {
872: const char *start, *end;
873: size_t sz;
1.19 ! schwarze 874:
1.1 schwarze 875: if (SEC_SYNOPSIS != n->sec)
1.19 ! schwarze 876: return(0);
1.1 schwarze 877: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.19 ! schwarze 878: return(0);
1.1 schwarze 879:
880: /*
881: * Only consider those `Fd' macro fields that begin with an
882: * "inclusion" token (versus, e.g., #define).
883: */
884: if (strcmp("#include", n->string))
1.19 ! schwarze 885: return(0);
1.1 schwarze 886:
887: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.19 ! schwarze 888: return(0);
1.1 schwarze 889:
890: /*
891: * Strip away the enclosing angle brackets and make sure we're
892: * not zero-length.
893: */
894:
895: start = n->string;
896: if ('<' == *start || '"' == *start)
897: start++;
898:
899: if (0 == (sz = strlen(start)))
1.19 ! schwarze 900: return(0);
1.1 schwarze 901:
902: end = &start[(int)sz - 1];
903: if ('>' == *end || '"' == *end)
904: end--;
905:
906: assert(end >= start);
907:
908: buf_appendb(buf, start, (size_t)(end - start + 1));
909: buf_appendb(buf, "", 1);
1.19 ! schwarze 910: return(1);
1.1 schwarze 911: }
912:
913: /* ARGSUSED */
1.19 ! schwarze 914: static int
! 915: pmdoc_In(MDOC_ARGS)
1.1 schwarze 916: {
917:
918: if (NULL == n->child || MDOC_TEXT != n->child->type)
1.19 ! schwarze 919: return(0);
1.1 schwarze 920:
921: buf_append(buf, n->child->string);
1.19 ! schwarze 922: return(1);
1.1 schwarze 923: }
924:
925: /* ARGSUSED */
1.19 ! schwarze 926: static int
1.1 schwarze 927: pmdoc_Fn(MDOC_ARGS)
928: {
1.19 ! schwarze 929: struct mdoc_node *nn;
1.1 schwarze 930: const char *cp;
931:
1.19 ! schwarze 932: nn = n->child;
! 933:
! 934: if (NULL == nn || MDOC_TEXT != nn->type)
! 935: return(0);
! 936:
! 937: /* .Fn "struct type *name" "char *arg" */
1.1 schwarze 938:
1.19 ! schwarze 939: cp = strrchr(nn->string, ' ');
1.1 schwarze 940: if (NULL == cp)
1.19 ! schwarze 941: cp = nn->string;
1.1 schwarze 942:
943: /* Strip away pointer symbol. */
944:
945: while ('*' == *cp)
946: cp++;
947:
1.19 ! schwarze 948: /* Store the function name. */
! 949:
1.1 schwarze 950: buf_append(buf, cp);
1.5 schwarze 951: hash_put(hash, buf, TYPE_Fn);
1.19 ! schwarze 952:
! 953: /* Store the function type. */
! 954:
! 955: if (nn->string < cp) {
! 956: buf->len = 0;
! 957: buf_appendb(buf, nn->string, cp - nn->string);
! 958: buf_appendb(buf, "", 1);
! 959: hash_put(hash, buf, TYPE_Ft);
! 960: }
! 961:
! 962: /* Store the arguments. */
! 963:
! 964: for (nn = nn->next; nn; nn = nn->next) {
! 965: if (MDOC_TEXT != nn->type)
! 966: continue;
! 967: buf->len = 0;
! 968: buf_append(buf, nn->string);
! 969: hash_put(hash, buf, TYPE_Fa);
! 970: }
! 971:
! 972: return(0);
1.1 schwarze 973: }
974:
975: /* ARGSUSED */
1.19 ! schwarze 976: static int
1.1 schwarze 977: pmdoc_St(MDOC_ARGS)
978: {
1.19 ! schwarze 979:
1.1 schwarze 980: if (NULL == n->child || MDOC_TEXT != n->child->type)
1.19 ! schwarze 981: return(0);
1.1 schwarze 982:
983: buf_append(buf, n->child->string);
1.19 ! schwarze 984: return(1);
1.1 schwarze 985: }
986:
987: /* ARGSUSED */
1.19 ! schwarze 988: static int
1.1 schwarze 989: pmdoc_Xr(MDOC_ARGS)
990: {
991:
992: if (NULL == (n = n->child))
1.19 ! schwarze 993: return(0);
1.1 schwarze 994:
995: buf_appendb(buf, n->string, strlen(n->string));
996:
997: if (NULL != (n = n->next)) {
998: buf_appendb(buf, ".", 1);
999: buf_appendb(buf, n->string, strlen(n->string) + 1);
1000: } else
1001: buf_appendb(buf, ".", 2);
1002:
1.19 ! schwarze 1003: return(1);
1.1 schwarze 1004: }
1005:
1006: /* ARGSUSED */
1.19 ! schwarze 1007: static int
1.1 schwarze 1008: pmdoc_Nd(MDOC_ARGS)
1009: {
1010:
1011: if (MDOC_BODY != n->type)
1.19 ! schwarze 1012: return(0);
1.1 schwarze 1013:
1014: buf_appendmdoc(dbuf, n->child, 1);
1.19 ! schwarze 1015: return(1);
1.1 schwarze 1016: }
1017:
1018: /* ARGSUSED */
1.19 ! schwarze 1019: static int
! 1020: pmdoc_Nm(MDOC_ARGS)
1.1 schwarze 1021: {
1022:
1.19 ! schwarze 1023: if (SEC_NAME == n->sec)
! 1024: return(1);
! 1025: else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
! 1026: return(0);
1.1 schwarze 1027:
1.19 ! schwarze 1028: if (NULL == n->child)
! 1029: buf_append(buf, m->name);
1.1 schwarze 1030:
1.19 ! schwarze 1031: return(1);
1.1 schwarze 1032: }
1033:
1034: /* ARGSUSED */
1.19 ! schwarze 1035: static int
! 1036: pmdoc_Sh(MDOC_ARGS)
1.1 schwarze 1037: {
1038:
1.19 ! schwarze 1039: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 schwarze 1040: }
1041:
1042: static void
1.9 schwarze 1043: hash_put(DB *db, const struct buf *buf, uint64_t mask)
1.1 schwarze 1044: {
1045: DBT key, val;
1046: int rc;
1047:
1048: if (buf->len < 2)
1049: return;
1050:
1051: key.data = buf->cp;
1052: key.size = buf->len;
1053:
1054: if ((rc = (*db->get)(db, &key, &val, 0)) < 0) {
1055: perror("hash");
1056: exit((int)MANDOCLEVEL_SYSERR);
1057: } else if (0 == rc)
1.9 schwarze 1058: mask |= *(uint64_t *)val.data;
1.1 schwarze 1059:
1060: val.data = &mask;
1.9 schwarze 1061: val.size = sizeof(uint64_t);
1.1 schwarze 1062:
1063: if ((rc = (*db->put)(db, &key, &val, 0)) < 0) {
1064: perror("hash");
1065: exit((int)MANDOCLEVEL_SYSERR);
1066: }
1067: }
1068:
1069: static void
1070: dbt_put(DB *db, const char *dbn, DBT *key, DBT *val)
1071: {
1072:
1073: assert(key->size);
1074: assert(val->size);
1075:
1076: if (0 == (*db->put)(db, key, val, 0))
1077: return;
1078:
1079: perror(dbn);
1080: exit((int)MANDOCLEVEL_SYSERR);
1081: /* NOTREACHED */
1082: }
1083:
1084: /*
1085: * Call out to per-macro handlers after clearing the persistent database
1086: * key. If the macro sets the database key, flush it to the database.
1087: */
1088: static void
1089: pmdoc_node(MDOC_ARGS)
1090: {
1091:
1092: if (NULL == n)
1093: return;
1094:
1095: switch (n->type) {
1096: case (MDOC_HEAD):
1097: /* FALLTHROUGH */
1098: case (MDOC_BODY):
1099: /* FALLTHROUGH */
1100: case (MDOC_TAIL):
1101: /* FALLTHROUGH */
1102: case (MDOC_BLOCK):
1103: /* FALLTHROUGH */
1104: case (MDOC_ELEM):
1.19 ! schwarze 1105: buf->len = 0;
! 1106:
! 1107: /*
! 1108: * Both NULL handlers and handlers returning true
! 1109: * request using the data. Only skip the element
! 1110: * when the handler returns false.
! 1111: */
! 1112:
! 1113: if (NULL != mdocs[n->tok].fp &&
! 1114: 0 == (*mdocs[n->tok].fp)(hash, buf, dbuf, n, m))
1.1 schwarze 1115: break;
1116:
1.19 ! schwarze 1117: /*
! 1118: * For many macros, use the text from all children.
! 1119: * Set zero flags for macros not needing this.
! 1120: * In that case, the handler must fill the buffer.
! 1121: */
! 1122:
! 1123: if (MDOCF_CHILD & mdocs[n->tok].flags)
! 1124: buf_appendmdoc(buf, n->child, 0);
! 1125:
! 1126: /*
! 1127: * Cover the most common case:
! 1128: * Automatically stage one string per element.
! 1129: * Set a zero mask for macros not needing this.
! 1130: * Additional staging can be done in the handler.
! 1131: */
! 1132:
! 1133: if (mdocs[n->tok].mask)
! 1134: hash_put(hash, buf, mdocs[n->tok].mask);
1.1 schwarze 1135: break;
1136: default:
1137: break;
1138: }
1139:
1140: pmdoc_node(hash, buf, dbuf, n->child, m);
1141: pmdoc_node(hash, buf, dbuf, n->next, m);
1142: }
1143:
1144: static int
1145: pman_node(MAN_ARGS)
1146: {
1147: const struct man_node *head, *body;
1148: const char *start, *sv;
1149: size_t sz;
1150:
1151: if (NULL == n)
1152: return(0);
1153:
1154: /*
1155: * We're only searching for one thing: the first text child in
1156: * the BODY of a NAME section. Since we don't keep track of
1157: * sections in -man, run some hoops to find out whether we're in
1158: * the correct section or not.
1159: */
1160:
1161: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1162: body = n;
1163: assert(body->parent);
1164: if (NULL != (head = body->parent->head) &&
1165: 1 == head->nchild &&
1166: NULL != (head = (head->child)) &&
1167: MAN_TEXT == head->type &&
1168: 0 == strcmp(head->string, "NAME") &&
1169: NULL != (body = body->child) &&
1170: MAN_TEXT == body->type) {
1171:
1172: assert(body->string);
1173: start = sv = body->string;
1174:
1175: /*
1176: * Go through a special heuristic dance here.
1177: * This is why -man manuals are great!
1178: * (I'm being sarcastic: my eyes are bleeding.)
1179: * Conventionally, one or more manual names are
1180: * comma-specified prior to a whitespace, then a
1181: * dash, then a description. Try to puzzle out
1182: * the name parts here.
1183: */
1184:
1185: for ( ;; ) {
1186: sz = strcspn(start, " ,");
1187: if ('\0' == start[(int)sz])
1188: break;
1189:
1190: buf->len = 0;
1191: buf_appendb(buf, start, sz);
1192: buf_appendb(buf, "", 1);
1193:
1.5 schwarze 1194: hash_put(hash, buf, TYPE_Nm);
1.1 schwarze 1195:
1196: if (' ' == start[(int)sz]) {
1197: start += (int)sz + 1;
1198: break;
1199: }
1200:
1201: assert(',' == start[(int)sz]);
1202: start += (int)sz + 1;
1203: while (' ' == *start)
1204: start++;
1205: }
1206:
1207: buf->len = 0;
1208:
1209: if (sv == start) {
1210: buf_append(buf, start);
1211: return(1);
1212: }
1213:
1214: while (' ' == *start)
1215: start++;
1216:
1217: if (0 == strncmp(start, "-", 1))
1218: start += 1;
1219: else if (0 == strncmp(start, "\\-", 2))
1220: start += 2;
1221: else if (0 == strncmp(start, "\\(en", 4))
1222: start += 4;
1223: else if (0 == strncmp(start, "\\(em", 4))
1224: start += 4;
1225:
1226: while (' ' == *start)
1227: start++;
1228:
1229: sz = strlen(start) + 1;
1230: buf_appendb(dbuf, start, sz);
1231: buf_appendb(buf, start, sz);
1232:
1.5 schwarze 1233: hash_put(hash, buf, TYPE_Nd);
1.1 schwarze 1234: }
1235: }
1236:
1.4 schwarze 1237: for (n = n->child; n; n = n->next)
1238: if (pman_node(hash, buf, dbuf, n))
1239: return(1);
1.1 schwarze 1240:
1241: return(0);
1242: }
1243:
1.11 schwarze 1244: /*
1245: * Parse a formatted manual page.
1246: * By necessity, this involves rather crude guesswork.
1247: */
1248: static void
1249: pformatted(DB *hash, struct buf *buf, struct buf *dbuf,
1250: const struct of *of)
1251: {
1252: FILE *stream;
1253: char *line, *p;
1254: size_t len, plen;
1255:
1256: if (NULL == (stream = fopen(of->fname, "r"))) {
1257: perror(of->fname);
1258: return;
1259: }
1260:
1261: /*
1262: * Always use the title derived from the filename up front,
1263: * do not even try to find it in the file. This also makes
1264: * sure we don't end up with an orphan index record, even if
1265: * the file content turns out to be completely unintelligible.
1266: */
1267:
1268: buf->len = 0;
1269: buf_append(buf, of->title);
1270: hash_put(hash, buf, TYPE_Nm);
1271:
1272: while (NULL != (line = fgetln(stream, &len)) && '\n' != *line)
1273: /* Skip to first blank line. */ ;
1274:
1275: while (NULL != (line = fgetln(stream, &len)) &&
1276: ('\n' == *line || ' ' == *line))
1277: /* Skip to first section header. */ ;
1278:
1279: /*
1280: * If no page content can be found,
1281: * reuse the page title as the page description.
1282: */
1283:
1284: if (NULL == (line = fgetln(stream, &len))) {
1285: buf_appendb(dbuf, buf->cp, buf->size);
1286: hash_put(hash, buf, TYPE_Nd);
1287: fclose(stream);
1288: return;
1289: }
1290: fclose(stream);
1291:
1292: /*
1293: * If there is a dash, skip to the text following it.
1294: */
1295:
1296: for (p = line, plen = len; plen; p++, plen--)
1297: if ('-' == *p)
1298: break;
1299: for ( ; plen; p++, plen--)
1300: if ('-' != *p && ' ' != *p && 8 != *p)
1301: break;
1302: if (0 == plen) {
1303: p = line;
1304: plen = len;
1305: }
1306:
1307: /*
1308: * Copy the rest of the line, but no more than 70 bytes.
1309: */
1310:
1311: if (70 < plen)
1312: plen = 70;
1313: p[plen-1] = '\0';
1314: buf_appendb(dbuf, p, plen);
1315: buf->len = 0;
1316: buf_appendb(buf, p, plen);
1317: hash_put(hash, buf, TYPE_Nd);
1318: }
1319:
1.1 schwarze 1320: static void
1.13 schwarze 1321: ofile_argbuild(int argc, char *argv[], struct of **of)
1.2 schwarze 1322: {
1.6 schwarze 1323: char buf[MAXPATHLEN];
1324: char *sec, *arch, *title, *p;
1.11 schwarze 1325: int i, src_form;
1.2 schwarze 1326: struct of *nof;
1327:
1328: for (i = 0; i < argc; i++) {
1.6 schwarze 1329:
1330: /*
1.8 schwarze 1331: * Try to infer the manual section, architecture and
1332: * page title from the path, assuming it looks like
1.11 schwarze 1333: * man*[/<arch>]/<title>.<section> or
1334: * cat<section>[/<arch>]/<title>.0
1.6 schwarze 1335: */
1336:
1337: if (strlcpy(buf, argv[i], sizeof(buf)) >= sizeof(buf)) {
1338: fprintf(stderr, "%s: Path too long\n", argv[i]);
1339: continue;
1340: }
1341: sec = arch = title = NULL;
1.11 schwarze 1342: src_form = 0;
1.6 schwarze 1343: p = strrchr(buf, '\0');
1344: while (p-- > buf) {
1345: if (NULL == sec && '.' == *p) {
1346: sec = p + 1;
1347: *p = '\0';
1.11 schwarze 1348: if ('0' == *sec)
1349: src_form |= MANDOC_FORM;
1350: else if ('1' <= *sec && '9' >= *sec)
1351: src_form |= MANDOC_SRC;
1.6 schwarze 1352: continue;
1353: }
1354: if ('/' != *p)
1355: continue;
1356: if (NULL == title) {
1357: title = p + 1;
1358: *p = '\0';
1359: continue;
1360: }
1.18 schwarze 1361: if (0 == strncmp("man", p + 1, 3))
1.11 schwarze 1362: src_form |= MANDOC_SRC;
1.18 schwarze 1363: else if (0 == strncmp("cat", p + 1, 3))
1.11 schwarze 1364: src_form |= MANDOC_FORM;
1.18 schwarze 1365: else
1.6 schwarze 1366: arch = p + 1;
1367: break;
1368: }
1369: if (NULL == title)
1370: title = buf;
1371:
1372: /*
1373: * Build the file structure.
1374: */
1375:
1.2 schwarze 1376: nof = mandoc_calloc(1, sizeof(struct of));
1.6 schwarze 1377: nof->fname = mandoc_strdup(argv[i]);
1378: if (NULL != sec)
1379: nof->sec = mandoc_strdup(sec);
1380: if (NULL != arch)
1381: nof->arch = mandoc_strdup(arch);
1382: nof->title = mandoc_strdup(title);
1.11 schwarze 1383: nof->src_form = src_form;
1.6 schwarze 1384:
1385: /*
1386: * Add the structure to the list.
1387: */
1388:
1.2 schwarze 1389: if (verb > 2)
1390: printf("%s: Scheduling\n", argv[i]);
1391: if (NULL == *of) {
1392: *of = nof;
1393: (*of)->first = nof;
1394: } else {
1395: nof->first = (*of)->first;
1396: (*of)->next = nof;
1397: *of = nof;
1398: }
1399: }
1400: }
1401:
1402: /*
1403: * Recursively build up a list of files to parse.
1404: * We use this instead of ftw() and so on because I don't want global
1405: * variables hanging around.
1406: * This ignores the mandoc.db and mandoc.index files, but assumes that
1407: * everything else is a manual.
1408: * Pass in a pointer to a NULL structure for the first invocation.
1409: */
1410: static int
1.6 schwarze 1411: ofile_dirbuild(const char *dir, const char* psec, const char *parch,
1.13 schwarze 1412: int p_src_form, struct of **of)
1.2 schwarze 1413: {
1414: char buf[MAXPATHLEN];
1.11 schwarze 1415: struct stat sb;
1.2 schwarze 1416: size_t sz;
1417: DIR *d;
1.6 schwarze 1418: const char *fn, *sec, *arch;
1.11 schwarze 1419: char *p, *q, *suffix;
1.2 schwarze 1420: struct of *nof;
1421: struct dirent *dp;
1.11 schwarze 1422: int src_form;
1.2 schwarze 1423:
1424: if (NULL == (d = opendir(dir))) {
1425: perror(dir);
1426: return(0);
1427: }
1428:
1429: while (NULL != (dp = readdir(d))) {
1430: fn = dp->d_name;
1.6 schwarze 1431:
1432: if ('.' == *fn)
1433: continue;
1434:
1.11 schwarze 1435: src_form = p_src_form;
1436:
1.2 schwarze 1437: if (DT_DIR == dp->d_type) {
1.6 schwarze 1438: sec = psec;
1439: arch = parch;
1440:
1441: /*
1.8 schwarze 1442: * By default, only use directories called:
1.11 schwarze 1443: * man<section>/[<arch>/] or
1444: * cat<section>/[<arch>/]
1.6 schwarze 1445: */
1446:
1447: if (NULL == sec) {
1.11 schwarze 1448: if(0 == strncmp("man", fn, 3)) {
1449: src_form |= MANDOC_SRC;
1.6 schwarze 1450: sec = fn + 3;
1.11 schwarze 1451: } else if (0 == strncmp("cat", fn, 3)) {
1452: src_form |= MANDOC_FORM;
1453: sec = fn + 3;
1454: } else if (use_all)
1.6 schwarze 1455: sec = fn;
1456: else
1457: continue;
1458: } else if (NULL == arch && (use_all ||
1459: NULL == strchr(fn, '.')))
1460: arch = fn;
1461: else if (0 == use_all)
1.2 schwarze 1462: continue;
1463:
1464: buf[0] = '\0';
1465: strlcat(buf, dir, MAXPATHLEN);
1466: strlcat(buf, "/", MAXPATHLEN);
1467: sz = strlcat(buf, fn, MAXPATHLEN);
1468:
1.6 schwarze 1469: if (MAXPATHLEN <= sz) {
1470: fprintf(stderr, "%s: Path too long\n", dir);
1471: return(0);
1472: }
1473:
1474: if (verb > 2)
1475: printf("%s: Scanning\n", buf);
1476:
1477: if ( ! ofile_dirbuild(buf, sec, arch,
1.13 schwarze 1478: src_form, of))
1.6 schwarze 1479: return(0);
1480: }
1481: if (DT_REG != dp->d_type ||
1482: (NULL == psec && !use_all) ||
1483: !strcmp(MANDOC_DB, fn) ||
1484: !strcmp(MANDOC_IDX, fn))
1485: continue;
1486:
1487: /*
1.8 schwarze 1488: * By default, skip files where the file name suffix
1489: * does not agree with the section directory
1490: * they are located in.
1.6 schwarze 1491: */
1492:
1493: suffix = strrchr(fn, '.');
1494: if (0 == use_all) {
1495: if (NULL == suffix)
1.2 schwarze 1496: continue;
1.11 schwarze 1497: if ((MANDOC_SRC & src_form &&
1498: strcmp(suffix + 1, psec)) ||
1499: (MANDOC_FORM & src_form &&
1500: strcmp(suffix + 1, "0")))
1501: continue;
1502: }
1503: if (NULL != suffix) {
1504: if ('0' == suffix[1])
1505: src_form |= MANDOC_FORM;
1506: else if ('1' <= suffix[1] && '9' >= suffix[1])
1507: src_form |= MANDOC_SRC;
1508: }
1509:
1510:
1511: /*
1512: * Skip formatted manuals if a source version is
1513: * available. Ignore the age: it is very unlikely
1514: * that people install newer formatted base manuals
1515: * when they used to have source manuals before,
1516: * and in ports, old manuals get removed on update.
1517: */
1518: if (0 == use_all && MANDOC_FORM & src_form &&
1519: NULL != psec) {
1520: buf[0] = '\0';
1521: strlcat(buf, dir, MAXPATHLEN);
1522: p = strrchr(buf, '/');
1523: if (NULL == p)
1524: p = buf;
1525: else
1526: p++;
1527: if (0 == strncmp("cat", p, 3))
1528: memcpy(p, "man", 3);
1529: strlcat(buf, "/", MAXPATHLEN);
1530: sz = strlcat(buf, fn, MAXPATHLEN);
1531: if (sz >= MAXPATHLEN) {
1532: fprintf(stderr, "%s: Path too long\n", buf);
1.2 schwarze 1533: continue;
1.11 schwarze 1534: }
1535: q = strrchr(buf, '.');
1536: if (NULL != q && p < q++) {
1537: *q = '\0';
1538: sz = strlcat(buf, psec, MAXPATHLEN);
1539: if (sz >= MAXPATHLEN) {
1540: fprintf(stderr,
1541: "%s: Path too long\n", buf);
1542: continue;
1543: }
1544: if (0 == stat(buf, &sb))
1545: continue;
1546: }
1.2 schwarze 1547: }
1548:
1549: buf[0] = '\0';
1550: strlcat(buf, dir, MAXPATHLEN);
1551: strlcat(buf, "/", MAXPATHLEN);
1552: sz = strlcat(buf, fn, MAXPATHLEN);
1553: if (sz >= MAXPATHLEN) {
1554: fprintf(stderr, "%s: Path too long\n", dir);
1.11 schwarze 1555: continue;
1.2 schwarze 1556: }
1557:
1558: nof = mandoc_calloc(1, sizeof(struct of));
1559: nof->fname = mandoc_strdup(buf);
1.6 schwarze 1560: if (NULL != psec)
1561: nof->sec = mandoc_strdup(psec);
1562: if (NULL != parch)
1563: nof->arch = mandoc_strdup(parch);
1.11 schwarze 1564: nof->src_form = src_form;
1.8 schwarze 1565:
1566: /*
1567: * Remember the file name without the extension,
1568: * to be used as the page title in the database.
1569: */
1570:
1.6 schwarze 1571: if (NULL != suffix)
1572: *suffix = '\0';
1573: nof->title = mandoc_strdup(fn);
1.2 schwarze 1574:
1.11 schwarze 1575: /*
1576: * Add the structure to the list.
1577: */
1578:
1.2 schwarze 1579: if (verb > 2)
1580: printf("%s: Scheduling\n", buf);
1581: if (NULL == *of) {
1582: *of = nof;
1583: (*of)->first = nof;
1584: } else {
1585: nof->first = (*of)->first;
1586: (*of)->next = nof;
1587: *of = nof;
1588: }
1589: }
1590:
1.4 schwarze 1591: closedir(d);
1.2 schwarze 1592: return(1);
1593: }
1594:
1595: static void
1596: ofile_free(struct of *of)
1597: {
1598: struct of *nof;
1599:
1600: while (of) {
1601: nof = of->next;
1602: free(of->fname);
1.6 schwarze 1603: free(of->sec);
1604: free(of->arch);
1605: free(of->title);
1.2 schwarze 1606: free(of);
1607: of = nof;
1608: }
1609: }
1610:
1611: static void
1.1 schwarze 1612: usage(void)
1613: {
1614:
1.2 schwarze 1615: fprintf(stderr, "usage: %s [-v] "
1616: "[-d dir [files...] |"
1617: " -u dir [files...] |"
1618: " dir...]\n", progname);
1.1 schwarze 1619: }