Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.16
1.16 ! schwarze 1: /* $Id: mandocdb.c,v 1.15 2011/11/29 22:30:56 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);
106: static void pmdoc_An(MDOC_ARGS);
107: static void pmdoc_Cd(MDOC_ARGS);
108: static void pmdoc_Er(MDOC_ARGS);
109: static void pmdoc_Ev(MDOC_ARGS);
110: static void pmdoc_Fd(MDOC_ARGS);
111: static void pmdoc_In(MDOC_ARGS);
112: static void pmdoc_Fn(MDOC_ARGS);
113: static void pmdoc_Fo(MDOC_ARGS);
114: static void pmdoc_Nd(MDOC_ARGS);
115: static void pmdoc_Nm(MDOC_ARGS);
116: static void pmdoc_Pa(MDOC_ARGS);
117: static void pmdoc_St(MDOC_ARGS);
118: static void pmdoc_Vt(MDOC_ARGS);
119: static void pmdoc_Xr(MDOC_ARGS);
120: static void usage(void);
121:
122: typedef void (*pmdoc_nf)(MDOC_ARGS);
123:
124: static const pmdoc_nf mdocs[MDOC_MAX] = {
125: NULL, /* Ap */
126: NULL, /* Dd */
127: NULL, /* Dt */
128: NULL, /* Os */
129: NULL, /* Sh */
130: NULL, /* Ss */
131: NULL, /* Pp */
132: NULL, /* D1 */
133: NULL, /* Dl */
134: NULL, /* Bd */
135: NULL, /* Ed */
136: NULL, /* Bl */
137: NULL, /* El */
138: NULL, /* It */
139: NULL, /* Ad */
140: pmdoc_An, /* An */
141: NULL, /* Ar */
142: pmdoc_Cd, /* Cd */
143: NULL, /* Cm */
144: NULL, /* Dv */
145: pmdoc_Er, /* Er */
146: pmdoc_Ev, /* Ev */
147: NULL, /* Ex */
148: NULL, /* Fa */
149: pmdoc_Fd, /* Fd */
150: NULL, /* Fl */
151: pmdoc_Fn, /* Fn */
152: NULL, /* Ft */
153: NULL, /* Ic */
154: pmdoc_In, /* In */
155: NULL, /* Li */
156: pmdoc_Nd, /* Nd */
157: pmdoc_Nm, /* Nm */
158: NULL, /* Op */
159: NULL, /* Ot */
160: pmdoc_Pa, /* Pa */
161: NULL, /* Rv */
162: pmdoc_St, /* St */
163: pmdoc_Vt, /* Va */
164: pmdoc_Vt, /* Vt */
165: pmdoc_Xr, /* Xr */
166: NULL, /* %A */
167: NULL, /* %B */
168: NULL, /* %D */
169: NULL, /* %I */
170: NULL, /* %J */
171: NULL, /* %N */
172: NULL, /* %O */
173: NULL, /* %P */
174: NULL, /* %R */
175: NULL, /* %T */
176: NULL, /* %V */
177: NULL, /* Ac */
178: NULL, /* Ao */
179: NULL, /* Aq */
180: NULL, /* At */
181: NULL, /* Bc */
182: NULL, /* Bf */
183: NULL, /* Bo */
184: NULL, /* Bq */
185: NULL, /* Bsx */
186: NULL, /* Bx */
187: NULL, /* Db */
188: NULL, /* Dc */
189: NULL, /* Do */
190: NULL, /* Dq */
191: NULL, /* Ec */
192: NULL, /* Ef */
193: NULL, /* Em */
194: NULL, /* Eo */
195: NULL, /* Fx */
196: NULL, /* Ms */
197: NULL, /* No */
198: NULL, /* Ns */
199: NULL, /* Nx */
200: NULL, /* Ox */
201: NULL, /* Pc */
202: NULL, /* Pf */
203: NULL, /* Po */
204: NULL, /* Pq */
205: NULL, /* Qc */
206: NULL, /* Ql */
207: NULL, /* Qo */
208: NULL, /* Qq */
209: NULL, /* Re */
210: NULL, /* Rs */
211: NULL, /* Sc */
212: NULL, /* So */
213: NULL, /* Sq */
214: NULL, /* Sm */
215: NULL, /* Sx */
216: NULL, /* Sy */
217: NULL, /* Tn */
218: NULL, /* Ux */
219: NULL, /* Xc */
220: NULL, /* Xo */
221: pmdoc_Fo, /* Fo */
222: NULL, /* Fc */
223: NULL, /* Oo */
224: NULL, /* Oc */
225: NULL, /* Bk */
226: NULL, /* Ek */
227: NULL, /* Bt */
228: NULL, /* Hf */
229: NULL, /* Fr */
230: NULL, /* Ud */
231: NULL, /* Lb */
232: NULL, /* Lp */
233: NULL, /* Lk */
234: NULL, /* Mt */
235: NULL, /* Brq */
236: NULL, /* Bro */
237: NULL, /* Brc */
238: NULL, /* %C */
239: NULL, /* Es */
240: NULL, /* En */
241: NULL, /* Dx */
242: NULL, /* %Q */
243: NULL, /* br */
244: NULL, /* sp */
245: NULL, /* %U */
246: NULL, /* Ta */
247: };
248:
249: static const char *progname;
1.13 schwarze 250: static int use_all; /* Use all directories and files. */
251: static int verb; /* Output verbosity level. */
1.1 schwarze 252:
253: int
1.3 schwarze 254: mandocdb(int argc, char *argv[])
1.1 schwarze 255: {
256: struct mparse *mp; /* parse sequence */
1.10 schwarze 257: struct manpaths dirs;
1.1 schwarze 258: enum op op; /* current operation */
1.2 schwarze 259: const char *dir;
1.1 schwarze 260: char ibuf[MAXPATHLEN], /* index fname */
1.2 schwarze 261: fbuf[MAXPATHLEN]; /* btree fname */
1.13 schwarze 262: int ch, i, flags;
1.1 schwarze 263: DB *idx, /* index database */
264: *db, /* keyword database */
265: *hash; /* temporary keyword hashtable */
266: BTREEINFO info; /* btree configuration */
1.8 schwarze 267: recno_t maxrec; /* last record number in the index */
268: recno_t *recs; /* the numbers of all empty records */
1.2 schwarze 269: size_t sz1, sz2,
1.8 schwarze 270: recsz, /* number of allocated slots in recs */
271: reccur; /* current number of empty records */
1.1 schwarze 272: struct buf buf, /* keyword buffer */
273: dbuf; /* description buffer */
1.2 schwarze 274: struct of *of; /* list of files for processing */
1.1 schwarze 275: extern int optind;
276: extern char *optarg;
277:
278: progname = strrchr(argv[0], '/');
279: if (progname == NULL)
280: progname = argv[0];
281: else
282: ++progname;
283:
1.10 schwarze 284: memset(&dirs, 0, sizeof(struct manpaths));
285:
1.1 schwarze 286: verb = 0;
1.6 schwarze 287: use_all = 0;
1.2 schwarze 288: of = NULL;
1.1 schwarze 289: db = idx = NULL;
290: mp = NULL;
291: hash = NULL;
292: recs = NULL;
293: recsz = reccur = 0;
294: maxrec = 0;
295: op = OP_NEW;
1.2 schwarze 296: dir = NULL;
1.1 schwarze 297:
1.6 schwarze 298: while (-1 != (ch = getopt(argc, argv, "ad:u:v")))
1.1 schwarze 299: switch (ch) {
1.6 schwarze 300: case ('a'):
301: use_all = 1;
302: break;
1.1 schwarze 303: case ('d'):
304: dir = optarg;
1.2 schwarze 305: op = OP_UPDATE;
1.1 schwarze 306: break;
1.2 schwarze 307: case ('u'):
308: dir = optarg;
1.1 schwarze 309: op = OP_DELETE;
310: break;
311: case ('v'):
312: verb++;
313: break;
314: default:
315: usage();
316: return((int)MANDOCLEVEL_BADARG);
317: }
318:
319: argc -= optind;
320: argv += optind;
321:
1.2 schwarze 322: memset(&info, 0, sizeof(BTREEINFO));
323: info.flags = R_DUP;
324:
325: mp = mparse_alloc(MPARSE_AUTO, MANDOCLEVEL_FATAL, NULL, NULL);
326:
327: memset(&buf, 0, sizeof(struct buf));
328: memset(&dbuf, 0, sizeof(struct buf));
329:
330: buf.size = dbuf.size = MANDOC_BUFSZ;
331:
332: buf.cp = mandoc_malloc(buf.size);
333: dbuf.cp = mandoc_malloc(dbuf.size);
334:
335: flags = OP_NEW == op ? O_CREAT|O_TRUNC|O_RDWR : O_CREAT|O_RDWR;
1.1 schwarze 336:
1.2 schwarze 337: if (OP_UPDATE == op || OP_DELETE == op) {
338: ibuf[0] = fbuf[0] = '\0';
1.1 schwarze 339:
1.2 schwarze 340: strlcat(fbuf, dir, MAXPATHLEN);
341: strlcat(fbuf, "/", MAXPATHLEN);
342: sz1 = strlcat(fbuf, MANDOC_DB, MAXPATHLEN);
343:
344: strlcat(ibuf, dir, MAXPATHLEN);
345: strlcat(ibuf, "/", MAXPATHLEN);
346: sz2 = strlcat(ibuf, MANDOC_IDX, MAXPATHLEN);
347:
348: if (sz1 >= MAXPATHLEN || sz2 >= MAXPATHLEN) {
349: fprintf(stderr, "%s: Path too long\n", dir);
350: exit((int)MANDOCLEVEL_BADARG);
351: }
1.1 schwarze 352:
1.2 schwarze 353: db = dbopen(fbuf, flags, 0644, DB_BTREE, &info);
354: idx = dbopen(ibuf, flags, 0644, DB_RECNO, NULL);
1.1 schwarze 355:
1.2 schwarze 356: if (NULL == db) {
357: perror(fbuf);
358: exit((int)MANDOCLEVEL_SYSERR);
1.8 schwarze 359: } else if (NULL == idx) {
1.2 schwarze 360: perror(ibuf);
361: exit((int)MANDOCLEVEL_SYSERR);
362: }
1.1 schwarze 363:
1.2 schwarze 364: if (verb > 2) {
365: printf("%s: Opened\n", fbuf);
366: printf("%s: Opened\n", ibuf);
367: }
1.1 schwarze 368:
1.13 schwarze 369: ofile_argbuild(argc, argv, &of);
1.2 schwarze 370: if (NULL == of)
371: goto out;
372:
373: of = of->first;
374:
1.13 schwarze 375: index_prune(of, db, fbuf, idx, ibuf,
1.2 schwarze 376: &maxrec, &recs, &recsz);
377:
1.15 schwarze 378: /*
379: * Go to the root of the respective manual tree
380: * such that .so links work. In case of failure,
381: * just prod on, even though .so links won't work.
382: */
383:
1.14 schwarze 384: if (OP_UPDATE == op) {
385: chdir(dir);
1.10 schwarze 386: index_merge(of, mp, &dbuf, &buf, hash,
1.13 schwarze 387: db, fbuf, idx, ibuf,
388: maxrec, recs, reccur);
1.14 schwarze 389: }
1.1 schwarze 390:
391: goto out;
392: }
393:
1.10 schwarze 394: /*
395: * Configure the directories we're going to scan.
396: * If we have command-line arguments, use them.
397: * If not, we use man(1)'s method (see mandocdb.8).
398: */
399:
400: if (argc > 0) {
401: dirs.paths = mandoc_malloc(argc * sizeof(char *));
402: dirs.sz = argc;
403: for (i = 0; i < argc; i++)
404: dirs.paths[i] = mandoc_strdup(argv[i]);
405: } else
406: manpath_parse(&dirs, NULL, NULL);
1.7 schwarze 407:
1.10 schwarze 408: for (i = 0; i < dirs.sz; i++) {
1.2 schwarze 409: ibuf[0] = fbuf[0] = '\0';
410:
1.10 schwarze 411: strlcat(fbuf, dirs.paths[i], MAXPATHLEN);
1.2 schwarze 412: strlcat(fbuf, "/", MAXPATHLEN);
413: sz1 = strlcat(fbuf, MANDOC_DB, MAXPATHLEN);
414:
1.10 schwarze 415: strlcat(ibuf, dirs.paths[i], MAXPATHLEN);
1.2 schwarze 416: strlcat(ibuf, "/", MAXPATHLEN);
417: sz2 = strlcat(ibuf, MANDOC_IDX, MAXPATHLEN);
418:
419: if (sz1 >= MAXPATHLEN || sz2 >= MAXPATHLEN) {
1.10 schwarze 420: fprintf(stderr, "%s: Path too long\n",
421: dirs.paths[i]);
1.2 schwarze 422: exit((int)MANDOCLEVEL_BADARG);
423: }
424:
1.7 schwarze 425: if (db)
426: (*db->close)(db);
427: if (idx)
428: (*idx->close)(idx);
429:
1.2 schwarze 430: db = dbopen(fbuf, flags, 0644, DB_BTREE, &info);
431: idx = dbopen(ibuf, flags, 0644, DB_RECNO, NULL);
1.1 schwarze 432:
1.2 schwarze 433: if (NULL == db) {
434: perror(fbuf);
435: exit((int)MANDOCLEVEL_SYSERR);
1.8 schwarze 436: } else if (NULL == idx) {
1.2 schwarze 437: perror(ibuf);
438: exit((int)MANDOCLEVEL_SYSERR);
439: }
1.1 schwarze 440:
1.2 schwarze 441: if (verb > 2) {
442: printf("%s: Truncated\n", fbuf);
443: printf("%s: Truncated\n", ibuf);
444: }
1.1 schwarze 445:
1.2 schwarze 446: ofile_free(of);
447: of = NULL;
1.1 schwarze 448:
1.10 schwarze 449: if ( ! ofile_dirbuild(dirs.paths[i], NULL, NULL,
1.13 schwarze 450: 0, &of))
1.2 schwarze 451: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 452:
1.2 schwarze 453: if (NULL == of)
454: continue;
1.1 schwarze 455:
1.2 schwarze 456: of = of->first;
1.15 schwarze 457:
458: /*
459: * Go to the root of the respective manual tree
460: * such that .so links work. In case of failure,
461: * just prod on, even though .so links won't work.
462: */
1.1 schwarze 463:
1.14 schwarze 464: chdir(dirs.paths[i]);
1.10 schwarze 465: index_merge(of, mp, &dbuf, &buf, hash, db, fbuf,
1.13 schwarze 466: idx, ibuf, maxrec, recs, reccur);
1.1 schwarze 467: }
468:
1.2 schwarze 469: out:
470: if (db)
471: (*db->close)(db);
472: if (idx)
473: (*idx->close)(idx);
474: if (hash)
475: (*hash->close)(hash);
476: if (mp)
477: mparse_free(mp);
1.1 schwarze 478:
1.10 schwarze 479: manpath_free(&dirs);
1.2 schwarze 480: ofile_free(of);
481: free(buf.cp);
482: free(dbuf.cp);
483: free(recs);
1.1 schwarze 484:
1.2 schwarze 485: return(MANDOCLEVEL_OK);
486: }
1.1 schwarze 487:
1.2 schwarze 488: void
489: index_merge(const struct of *of, struct mparse *mp,
1.13 schwarze 490: struct buf *dbuf, struct buf *buf, DB *hash,
491: DB *db, const char *dbf, DB *idx, const char *idxf,
1.2 schwarze 492: recno_t maxrec, const recno_t *recs, size_t reccur)
493: {
494: recno_t rec;
495: int ch;
496: DBT key, val;
497: struct mdoc *mdoc;
498: struct man *man;
499: const char *fn, *msec, *mtitle, *arch;
500: size_t sv;
501: unsigned seq;
1.9 schwarze 502: struct db_val vbuf;
1.1 schwarze 503:
1.2 schwarze 504: for (rec = 0; of; of = of->next) {
505: fn = of->fname;
1.11 schwarze 506:
507: /*
508: * Reclaim an empty index record, if available.
509: */
510:
1.2 schwarze 511: if (reccur > 0) {
512: --reccur;
513: rec = recs[(int)reccur];
514: } else if (maxrec > 0) {
515: rec = maxrec;
516: maxrec = 0;
1.1 schwarze 517: } else
518: rec++;
519:
520: mparse_reset(mp);
521: hash_reset(&hash);
1.11 schwarze 522: mdoc = NULL;
523: man = NULL;
1.1 schwarze 524:
1.11 schwarze 525: /*
526: * Try interpreting the file as mdoc(7) or man(7)
527: * source code, unless it is already known to be
528: * formatted. Fall back to formatted mode.
529: */
530:
531: if ((MANDOC_SRC & of->src_form ||
532: ! (MANDOC_FORM & of->src_form)) &&
533: MANDOCLEVEL_FATAL > mparse_readfd(mp, -1, fn))
534: mparse_result(mp, &mdoc, &man);
535:
536: if (NULL != mdoc) {
537: msec = mdoc_meta(mdoc)->msec;
538: arch = mdoc_meta(mdoc)->arch;
539: mtitle = mdoc_meta(mdoc)->title;
540: } else if (NULL != man) {
541: msec = man_meta(man)->msec;
542: arch = NULL;
543: mtitle = man_meta(man)->title;
544: } else {
545: msec = of->sec;
546: arch = of->arch;
547: mtitle = of->title;
1.1 schwarze 548: }
549:
1.6 schwarze 550: /*
1.8 schwarze 551: * By default, skip a file if the manual section
552: * and architecture given in the file disagree
553: * with the directory where the file is located.
1.6 schwarze 554: */
555:
556: if (0 == use_all) {
557: assert(of->sec);
558: assert(msec);
559: if (strcmp(msec, of->sec))
560: continue;
561:
562: if (NULL == arch) {
563: if (NULL != of->arch)
564: continue;
565: } else if (NULL == of->arch ||
566: strcmp(arch, of->arch))
567: continue;
568: }
569:
1.1 schwarze 570: if (NULL == arch)
571: arch = "";
572:
573: /*
1.8 schwarze 574: * By default, skip a file if the title given
575: * in the file disagrees with the file name.
576: * If both agree, use the file name as the title,
577: * because the one in the file usually is all caps.
1.6 schwarze 578: */
579:
580: assert(of->title);
581: assert(mtitle);
582:
583: if (0 == strcasecmp(mtitle, of->title))
584: mtitle = of->title;
585: else if (0 == use_all)
586: continue;
587:
588: /*
1.1 schwarze 589: * The index record value consists of a nil-terminated
590: * filename, a nil-terminated manual section, and a
591: * nil-terminated description. Since the description
592: * may not be set, we set a sentinel to see if we're
593: * going to write a nil byte in its place.
594: */
595:
1.2 schwarze 596: dbuf->len = 0;
1.12 schwarze 597: buf_append(dbuf, mdoc ? "mdoc" : (man ? "man" : "cat"));
1.2 schwarze 598: buf_appendb(dbuf, fn, strlen(fn) + 1);
599: buf_appendb(dbuf, msec, strlen(msec) + 1);
600: buf_appendb(dbuf, mtitle, strlen(mtitle) + 1);
601: buf_appendb(dbuf, arch, strlen(arch) + 1);
1.1 schwarze 602:
1.2 schwarze 603: sv = dbuf->len;
1.1 schwarze 604:
605: /* Fix the record number in the btree value. */
606:
607: if (mdoc)
1.2 schwarze 608: pmdoc_node(hash, buf, dbuf,
1.1 schwarze 609: mdoc_node(mdoc), mdoc_meta(mdoc));
1.11 schwarze 610: else if (man)
1.2 schwarze 611: pman_node(hash, buf, dbuf, man_node(man));
1.11 schwarze 612: else
613: pformatted(hash, buf, dbuf, of);
1.1 schwarze 614:
615: /*
616: * Copy from the in-memory hashtable of pending keywords
617: * into the database.
618: */
619:
1.9 schwarze 620: vbuf.rec = rec;
1.1 schwarze 621: seq = R_FIRST;
622: while (0 == (ch = (*hash->seq)(hash, &key, &val, seq))) {
623: seq = R_NEXT;
624:
1.9 schwarze 625: vbuf.mask = *(uint64_t *)val.data;
626: val.size = sizeof(struct db_val);
627: val.data = &vbuf;
1.1 schwarze 628:
1.2 schwarze 629: dbt_put(db, dbf, &key, &val);
1.1 schwarze 630: }
631: if (ch < 0) {
632: perror("hash");
633: exit((int)MANDOCLEVEL_SYSERR);
634: }
635:
636: /*
637: * Apply to the index. If we haven't had a description
638: * set, put an empty one in now.
639: */
640:
1.2 schwarze 641: if (dbuf->len == sv)
642: buf_appendb(dbuf, "", 1);
1.1 schwarze 643:
644: key.data = &rec;
645: key.size = sizeof(recno_t);
646:
1.2 schwarze 647: val.data = dbuf->cp;
648: val.size = dbuf->len;
1.1 schwarze 649:
1.2 schwarze 650: if (verb)
1.1 schwarze 651: printf("%s: Added index\n", fn);
1.16 ! schwarze 652:
1.2 schwarze 653: dbt_put(idx, idxf, &key, &val);
654: }
655: }
656:
657: /*
658: * Scan through all entries in the index file `idx' and prune those
659: * entries in `ofile'.
660: * Pruning consists of removing from `db', then invalidating the entry
661: * in `idx' (zeroing its value size).
662: */
663: static void
664: index_prune(const struct of *ofile, DB *db, const char *dbf,
1.13 schwarze 665: DB *idx, const char *idxf,
1.2 schwarze 666: recno_t *maxrec, recno_t **recs, size_t *recsz)
667: {
668: const struct of *of;
1.16 ! schwarze 669: const char *fn, *cp;
1.9 schwarze 670: struct db_val *vbuf;
1.2 schwarze 671: unsigned seq, sseq;
672: DBT key, val;
673: size_t reccur;
674: int ch;
675:
676: reccur = 0;
677: seq = R_FIRST;
678: while (0 == (ch = (*idx->seq)(idx, &key, &val, seq))) {
679: seq = R_NEXT;
680: *maxrec = *(recno_t *)key.data;
1.16 ! schwarze 681: cp = val.data;
! 682:
! 683: /* Deleted records are zero-sized. Skip them. */
! 684:
! 685: if (0 == val.size)
! 686: goto cont;
! 687:
! 688: /*
! 689: * Make sure we're sane.
! 690: * Read past our mdoc/man/cat type to the next string,
! 691: * then make sure it's bounded by a NUL.
! 692: * Failing any of these, we go into our error handler.
! 693: */
! 694:
! 695: if (NULL == (fn = memchr(cp, '\0', val.size)))
! 696: break;
! 697: if (++fn - cp >= (int)val.size)
! 698: break;
! 699: if (NULL == memchr(fn, '\0', val.size - (fn - cp)))
! 700: break;
! 701:
! 702: /*
! 703: * Search for the file in those we care about.
! 704: * XXX: build this into a tree. Too slow.
! 705: */
1.2 schwarze 706:
707: for (of = ofile; of; of = of->next)
708: if (0 == strcmp(fn, of->fname))
709: break;
710:
711: if (NULL == of)
712: continue;
713:
1.16 ! schwarze 714: /*
! 715: * Search through the keyword database, throwing out all
! 716: * references to our file.
! 717: */
! 718:
1.2 schwarze 719: sseq = R_FIRST;
720: while (0 == (ch = (*db->seq)(db, &key, &val, sseq))) {
721: sseq = R_NEXT;
1.16 ! schwarze 722: if (sizeof(struct db_val) != val.size)
! 723: break;
! 724:
1.9 schwarze 725: vbuf = val.data;
726: if (*maxrec != vbuf->rec)
1.2 schwarze 727: continue;
1.16 ! schwarze 728:
! 729: if ((ch = (*db->del)(db, &key, R_CURSOR)) < 0)
1.2 schwarze 730: break;
731: }
1.16 ! schwarze 732:
1.2 schwarze 733: if (ch < 0) {
734: perror(dbf);
735: exit((int)MANDOCLEVEL_SYSERR);
1.16 ! schwarze 736: } else if (1 != ch) {
! 737: fprintf(stderr, "%s: Corrupt database\n", dbf);
! 738: exit((int)MANDOCLEVEL_SYSERR);
1.2 schwarze 739: }
1.1 schwarze 740:
1.2 schwarze 741: if (verb)
742: printf("%s: Deleted index\n", fn);
1.1 schwarze 743:
1.2 schwarze 744: val.size = 0;
745: ch = (*idx->put)(idx, &key, &val, R_CURSOR);
1.1 schwarze 746:
1.16 ! schwarze 747: if (ch < 0)
! 748: break;
! 749: cont:
1.2 schwarze 750: if (reccur >= *recsz) {
751: *recsz += MANDOC_SLOP;
752: *recs = mandoc_realloc
753: (*recs, *recsz * sizeof(recno_t));
754: }
1.1 schwarze 755:
1.2 schwarze 756: (*recs)[(int)reccur] = *maxrec;
757: reccur++;
758: }
1.16 ! schwarze 759:
! 760: if (ch < 0) {
! 761: perror(idxf);
! 762: exit((int)MANDOCLEVEL_SYSERR);
! 763: } else if (1 != ch) {
! 764: fprintf(stderr, "%s: Corrupt index\n", idxf);
! 765: exit((int)MANDOCLEVEL_SYSERR);
! 766: }
! 767:
1.2 schwarze 768: (*maxrec)++;
1.1 schwarze 769: }
770:
771: /*
772: * Grow the buffer (if necessary) and copy in a binary string.
773: */
774: static void
775: buf_appendb(struct buf *buf, const void *cp, size_t sz)
776: {
777:
778: /* Overshoot by MANDOC_BUFSZ. */
779:
780: while (buf->len + sz >= buf->size) {
781: buf->size = buf->len + sz + MANDOC_BUFSZ;
782: buf->cp = mandoc_realloc(buf->cp, buf->size);
783: }
784:
785: memcpy(buf->cp + (int)buf->len, cp, sz);
786: buf->len += sz;
787: }
788:
789: /*
790: * Append a nil-terminated string to the buffer.
791: * This can be invoked multiple times.
792: * The buffer string will be nil-terminated.
793: * If invoked multiple times, a space is put between strings.
794: */
795: static void
796: buf_append(struct buf *buf, const char *cp)
797: {
798: size_t sz;
799:
800: if (0 == (sz = strlen(cp)))
801: return;
802:
803: if (buf->len)
804: buf->cp[(int)buf->len - 1] = ' ';
805:
806: buf_appendb(buf, cp, sz + 1);
807: }
808:
809: /*
810: * Recursively add all text from a given node.
811: * This is optimised for general mdoc nodes in this context, which do
812: * not consist of subexpressions and having a recursive call for n->next
813: * would be wasteful.
814: * The "f" variable should be 0 unless called from pmdoc_Nd for the
815: * description buffer, which does not start at the beginning of the
816: * buffer.
817: */
818: static void
819: buf_appendmdoc(struct buf *buf, const struct mdoc_node *n, int f)
820: {
821:
822: for ( ; n; n = n->next) {
823: if (n->child)
824: buf_appendmdoc(buf, n->child, f);
825:
826: if (MDOC_TEXT == n->type && f) {
827: f = 0;
828: buf_appendb(buf, n->string,
829: strlen(n->string) + 1);
830: } else if (MDOC_TEXT == n->type)
831: buf_append(buf, n->string);
832:
833: }
834: }
835:
836: /* ARGSUSED */
837: static void
838: pmdoc_An(MDOC_ARGS)
839: {
840:
841: if (SEC_AUTHORS != n->sec)
842: return;
843:
844: buf_appendmdoc(buf, n->child, 0);
1.5 schwarze 845: hash_put(hash, buf, TYPE_An);
1.1 schwarze 846: }
847:
848: static void
849: hash_reset(DB **db)
850: {
851: DB *hash;
852:
853: if (NULL != (hash = *db))
854: (*hash->close)(hash);
855:
1.2 schwarze 856: *db = dbopen(NULL, O_CREAT|O_RDWR, 0644, DB_HASH, NULL);
1.1 schwarze 857: if (NULL == *db) {
858: perror("hash");
859: exit((int)MANDOCLEVEL_SYSERR);
860: }
861: }
862:
863: /* ARGSUSED */
864: static void
865: pmdoc_Fd(MDOC_ARGS)
866: {
867: const char *start, *end;
868: size_t sz;
869:
870: if (SEC_SYNOPSIS != n->sec)
871: return;
872: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
873: return;
874:
875: /*
876: * Only consider those `Fd' macro fields that begin with an
877: * "inclusion" token (versus, e.g., #define).
878: */
879: if (strcmp("#include", n->string))
880: return;
881:
882: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
883: return;
884:
885: /*
886: * Strip away the enclosing angle brackets and make sure we're
887: * not zero-length.
888: */
889:
890: start = n->string;
891: if ('<' == *start || '"' == *start)
892: start++;
893:
894: if (0 == (sz = strlen(start)))
895: return;
896:
897: end = &start[(int)sz - 1];
898: if ('>' == *end || '"' == *end)
899: end--;
900:
901: assert(end >= start);
902:
903: buf_appendb(buf, start, (size_t)(end - start + 1));
904: buf_appendb(buf, "", 1);
905:
1.5 schwarze 906: hash_put(hash, buf, TYPE_In);
1.1 schwarze 907: }
908:
909: /* ARGSUSED */
910: static void
911: pmdoc_Cd(MDOC_ARGS)
912: {
913:
914: if (SEC_SYNOPSIS != n->sec)
915: return;
916:
917: buf_appendmdoc(buf, n->child, 0);
1.5 schwarze 918: hash_put(hash, buf, TYPE_Cd);
1.1 schwarze 919: }
920:
921: /* ARGSUSED */
922: static void
923: pmdoc_In(MDOC_ARGS)
924: {
925:
926: if (SEC_SYNOPSIS != n->sec)
927: return;
928: if (NULL == n->child || MDOC_TEXT != n->child->type)
929: return;
930:
931: buf_append(buf, n->child->string);
1.5 schwarze 932: hash_put(hash, buf, TYPE_In);
1.1 schwarze 933: }
934:
935: /* ARGSUSED */
936: static void
937: pmdoc_Fn(MDOC_ARGS)
938: {
939: const char *cp;
940:
941: if (SEC_SYNOPSIS != n->sec)
942: return;
943: if (NULL == n->child || MDOC_TEXT != n->child->type)
944: return;
945:
946: /* .Fn "struct type *arg" "foo" */
947:
948: cp = strrchr(n->child->string, ' ');
949: if (NULL == cp)
950: cp = n->child->string;
951:
952: /* Strip away pointer symbol. */
953:
954: while ('*' == *cp)
955: cp++;
956:
957: buf_append(buf, cp);
1.5 schwarze 958: hash_put(hash, buf, TYPE_Fn);
1.1 schwarze 959: }
960:
961: /* ARGSUSED */
962: static void
963: pmdoc_St(MDOC_ARGS)
964: {
965:
966: if (SEC_STANDARDS != n->sec)
967: return;
968: if (NULL == n->child || MDOC_TEXT != n->child->type)
969: return;
970:
971: buf_append(buf, n->child->string);
1.5 schwarze 972: hash_put(hash, buf, TYPE_St);
1.1 schwarze 973: }
974:
975: /* ARGSUSED */
976: static void
977: pmdoc_Xr(MDOC_ARGS)
978: {
979:
980: if (NULL == (n = n->child))
981: return;
982:
983: buf_appendb(buf, n->string, strlen(n->string));
984:
985: if (NULL != (n = n->next)) {
986: buf_appendb(buf, ".", 1);
987: buf_appendb(buf, n->string, strlen(n->string) + 1);
988: } else
989: buf_appendb(buf, ".", 2);
990:
1.5 schwarze 991: hash_put(hash, buf, TYPE_Xr);
1.1 schwarze 992: }
993:
994: /* ARGSUSED */
995: static void
996: pmdoc_Vt(MDOC_ARGS)
997: {
998: const char *start;
999: size_t sz;
1000:
1001: if (SEC_SYNOPSIS != n->sec)
1002: return;
1003: if (MDOC_Vt == n->tok && MDOC_BODY != n->type)
1004: return;
1005: if (NULL == n->last || MDOC_TEXT != n->last->type)
1006: return;
1007:
1008: /*
1009: * Strip away leading pointer symbol '*' and trailing ';'.
1010: */
1011:
1012: start = n->last->string;
1013:
1014: while ('*' == *start)
1015: start++;
1016:
1017: if (0 == (sz = strlen(start)))
1018: return;
1019:
1020: if (';' == start[(int)sz - 1])
1021: sz--;
1022:
1023: if (0 == sz)
1024: return;
1025:
1026: buf_appendb(buf, start, sz);
1027: buf_appendb(buf, "", 1);
1.5 schwarze 1028: hash_put(hash, buf, TYPE_Va);
1.1 schwarze 1029: }
1030:
1031: /* ARGSUSED */
1032: static void
1033: pmdoc_Fo(MDOC_ARGS)
1034: {
1035:
1036: if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
1037: return;
1038: if (NULL == n->child || MDOC_TEXT != n->child->type)
1039: return;
1040:
1041: buf_append(buf, n->child->string);
1.5 schwarze 1042: hash_put(hash, buf, TYPE_Fn);
1.1 schwarze 1043: }
1044:
1045:
1046: /* ARGSUSED */
1047: static void
1048: pmdoc_Nd(MDOC_ARGS)
1049: {
1050:
1051: if (MDOC_BODY != n->type)
1052: return;
1053:
1054: buf_appendmdoc(dbuf, n->child, 1);
1055: buf_appendmdoc(buf, n->child, 0);
1056:
1.5 schwarze 1057: hash_put(hash, buf, TYPE_Nd);
1.1 schwarze 1058: }
1059:
1060: /* ARGSUSED */
1061: static void
1062: pmdoc_Er(MDOC_ARGS)
1063: {
1064:
1065: if (SEC_ERRORS != n->sec)
1066: return;
1067:
1068: buf_appendmdoc(buf, n->child, 0);
1.5 schwarze 1069: hash_put(hash, buf, TYPE_Er);
1.1 schwarze 1070: }
1071:
1072: /* ARGSUSED */
1073: static void
1074: pmdoc_Ev(MDOC_ARGS)
1075: {
1076:
1077: if (SEC_ENVIRONMENT != n->sec)
1078: return;
1079:
1080: buf_appendmdoc(buf, n->child, 0);
1.5 schwarze 1081: hash_put(hash, buf, TYPE_Ev);
1.1 schwarze 1082: }
1083:
1084: /* ARGSUSED */
1085: static void
1086: pmdoc_Pa(MDOC_ARGS)
1087: {
1088:
1089: if (SEC_FILES != n->sec)
1090: return;
1091:
1092: buf_appendmdoc(buf, n->child, 0);
1.5 schwarze 1093: hash_put(hash, buf, TYPE_Pa);
1.1 schwarze 1094: }
1095:
1096: /* ARGSUSED */
1097: static void
1098: pmdoc_Nm(MDOC_ARGS)
1099: {
1100:
1101: if (SEC_NAME == n->sec) {
1102: buf_appendmdoc(buf, n->child, 0);
1.5 schwarze 1103: hash_put(hash, buf, TYPE_Nm);
1.1 schwarze 1104: return;
1105: } else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
1106: return;
1107:
1108: if (NULL == n->child)
1109: buf_append(buf, m->name);
1110:
1111: buf_appendmdoc(buf, n->child, 0);
1.5 schwarze 1112: hash_put(hash, buf, TYPE_Nm);
1.1 schwarze 1113: }
1114:
1115: static void
1.9 schwarze 1116: hash_put(DB *db, const struct buf *buf, uint64_t mask)
1.1 schwarze 1117: {
1118: DBT key, val;
1119: int rc;
1120:
1121: if (buf->len < 2)
1122: return;
1123:
1124: key.data = buf->cp;
1125: key.size = buf->len;
1126:
1127: if ((rc = (*db->get)(db, &key, &val, 0)) < 0) {
1128: perror("hash");
1129: exit((int)MANDOCLEVEL_SYSERR);
1130: } else if (0 == rc)
1.9 schwarze 1131: mask |= *(uint64_t *)val.data;
1.1 schwarze 1132:
1133: val.data = &mask;
1.9 schwarze 1134: val.size = sizeof(uint64_t);
1.1 schwarze 1135:
1136: if ((rc = (*db->put)(db, &key, &val, 0)) < 0) {
1137: perror("hash");
1138: exit((int)MANDOCLEVEL_SYSERR);
1139: }
1140: }
1141:
1142: static void
1143: dbt_put(DB *db, const char *dbn, DBT *key, DBT *val)
1144: {
1145:
1146: assert(key->size);
1147: assert(val->size);
1148:
1149: if (0 == (*db->put)(db, key, val, 0))
1150: return;
1151:
1152: perror(dbn);
1153: exit((int)MANDOCLEVEL_SYSERR);
1154: /* NOTREACHED */
1155: }
1156:
1157: /*
1158: * Call out to per-macro handlers after clearing the persistent database
1159: * key. If the macro sets the database key, flush it to the database.
1160: */
1161: static void
1162: pmdoc_node(MDOC_ARGS)
1163: {
1164:
1165: if (NULL == n)
1166: return;
1167:
1168: switch (n->type) {
1169: case (MDOC_HEAD):
1170: /* FALLTHROUGH */
1171: case (MDOC_BODY):
1172: /* FALLTHROUGH */
1173: case (MDOC_TAIL):
1174: /* FALLTHROUGH */
1175: case (MDOC_BLOCK):
1176: /* FALLTHROUGH */
1177: case (MDOC_ELEM):
1178: if (NULL == mdocs[n->tok])
1179: break;
1180:
1181: buf->len = 0;
1182: (*mdocs[n->tok])(hash, buf, dbuf, n, m);
1183: break;
1184: default:
1185: break;
1186: }
1187:
1188: pmdoc_node(hash, buf, dbuf, n->child, m);
1189: pmdoc_node(hash, buf, dbuf, n->next, m);
1190: }
1191:
1192: static int
1193: pman_node(MAN_ARGS)
1194: {
1195: const struct man_node *head, *body;
1196: const char *start, *sv;
1197: size_t sz;
1198:
1199: if (NULL == n)
1200: return(0);
1201:
1202: /*
1203: * We're only searching for one thing: the first text child in
1204: * the BODY of a NAME section. Since we don't keep track of
1205: * sections in -man, run some hoops to find out whether we're in
1206: * the correct section or not.
1207: */
1208:
1209: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1210: body = n;
1211: assert(body->parent);
1212: if (NULL != (head = body->parent->head) &&
1213: 1 == head->nchild &&
1214: NULL != (head = (head->child)) &&
1215: MAN_TEXT == head->type &&
1216: 0 == strcmp(head->string, "NAME") &&
1217: NULL != (body = body->child) &&
1218: MAN_TEXT == body->type) {
1219:
1220: assert(body->string);
1221: start = sv = body->string;
1222:
1223: /*
1224: * Go through a special heuristic dance here.
1225: * This is why -man manuals are great!
1226: * (I'm being sarcastic: my eyes are bleeding.)
1227: * Conventionally, one or more manual names are
1228: * comma-specified prior to a whitespace, then a
1229: * dash, then a description. Try to puzzle out
1230: * the name parts here.
1231: */
1232:
1233: for ( ;; ) {
1234: sz = strcspn(start, " ,");
1235: if ('\0' == start[(int)sz])
1236: break;
1237:
1238: buf->len = 0;
1239: buf_appendb(buf, start, sz);
1240: buf_appendb(buf, "", 1);
1241:
1.5 schwarze 1242: hash_put(hash, buf, TYPE_Nm);
1.1 schwarze 1243:
1244: if (' ' == start[(int)sz]) {
1245: start += (int)sz + 1;
1246: break;
1247: }
1248:
1249: assert(',' == start[(int)sz]);
1250: start += (int)sz + 1;
1251: while (' ' == *start)
1252: start++;
1253: }
1254:
1255: buf->len = 0;
1256:
1257: if (sv == start) {
1258: buf_append(buf, start);
1259: return(1);
1260: }
1261:
1262: while (' ' == *start)
1263: start++;
1264:
1265: if (0 == strncmp(start, "-", 1))
1266: start += 1;
1267: else if (0 == strncmp(start, "\\-", 2))
1268: start += 2;
1269: else if (0 == strncmp(start, "\\(en", 4))
1270: start += 4;
1271: else if (0 == strncmp(start, "\\(em", 4))
1272: start += 4;
1273:
1274: while (' ' == *start)
1275: start++;
1276:
1277: sz = strlen(start) + 1;
1278: buf_appendb(dbuf, start, sz);
1279: buf_appendb(buf, start, sz);
1280:
1.5 schwarze 1281: hash_put(hash, buf, TYPE_Nd);
1.1 schwarze 1282: }
1283: }
1284:
1.4 schwarze 1285: for (n = n->child; n; n = n->next)
1286: if (pman_node(hash, buf, dbuf, n))
1287: return(1);
1.1 schwarze 1288:
1289: return(0);
1290: }
1291:
1.11 schwarze 1292: /*
1293: * Parse a formatted manual page.
1294: * By necessity, this involves rather crude guesswork.
1295: */
1296: static void
1297: pformatted(DB *hash, struct buf *buf, struct buf *dbuf,
1298: const struct of *of)
1299: {
1300: FILE *stream;
1301: char *line, *p;
1302: size_t len, plen;
1303:
1304: if (NULL == (stream = fopen(of->fname, "r"))) {
1305: perror(of->fname);
1306: return;
1307: }
1308:
1309: /*
1310: * Always use the title derived from the filename up front,
1311: * do not even try to find it in the file. This also makes
1312: * sure we don't end up with an orphan index record, even if
1313: * the file content turns out to be completely unintelligible.
1314: */
1315:
1316: buf->len = 0;
1317: buf_append(buf, of->title);
1318: hash_put(hash, buf, TYPE_Nm);
1319:
1320: while (NULL != (line = fgetln(stream, &len)) && '\n' != *line)
1321: /* Skip to first blank line. */ ;
1322:
1323: while (NULL != (line = fgetln(stream, &len)) &&
1324: ('\n' == *line || ' ' == *line))
1325: /* Skip to first section header. */ ;
1326:
1327: /*
1328: * If no page content can be found,
1329: * reuse the page title as the page description.
1330: */
1331:
1332: if (NULL == (line = fgetln(stream, &len))) {
1333: buf_appendb(dbuf, buf->cp, buf->size);
1334: hash_put(hash, buf, TYPE_Nd);
1335: fclose(stream);
1336: return;
1337: }
1338: fclose(stream);
1339:
1340: /*
1341: * If there is a dash, skip to the text following it.
1342: */
1343:
1344: for (p = line, plen = len; plen; p++, plen--)
1345: if ('-' == *p)
1346: break;
1347: for ( ; plen; p++, plen--)
1348: if ('-' != *p && ' ' != *p && 8 != *p)
1349: break;
1350: if (0 == plen) {
1351: p = line;
1352: plen = len;
1353: }
1354:
1355: /*
1356: * Copy the rest of the line, but no more than 70 bytes.
1357: */
1358:
1359: if (70 < plen)
1360: plen = 70;
1361: p[plen-1] = '\0';
1362: buf_appendb(dbuf, p, plen);
1363: buf->len = 0;
1364: buf_appendb(buf, p, plen);
1365: hash_put(hash, buf, TYPE_Nd);
1366: }
1367:
1.1 schwarze 1368: static void
1.13 schwarze 1369: ofile_argbuild(int argc, char *argv[], struct of **of)
1.2 schwarze 1370: {
1.6 schwarze 1371: char buf[MAXPATHLEN];
1372: char *sec, *arch, *title, *p;
1.11 schwarze 1373: int i, src_form;
1.2 schwarze 1374: struct of *nof;
1375:
1376: for (i = 0; i < argc; i++) {
1.6 schwarze 1377:
1378: /*
1.8 schwarze 1379: * Try to infer the manual section, architecture and
1380: * page title from the path, assuming it looks like
1.11 schwarze 1381: * man*[/<arch>]/<title>.<section> or
1382: * cat<section>[/<arch>]/<title>.0
1.6 schwarze 1383: */
1384:
1385: if (strlcpy(buf, argv[i], sizeof(buf)) >= sizeof(buf)) {
1386: fprintf(stderr, "%s: Path too long\n", argv[i]);
1387: continue;
1388: }
1389: sec = arch = title = NULL;
1.11 schwarze 1390: src_form = 0;
1.6 schwarze 1391: p = strrchr(buf, '\0');
1392: while (p-- > buf) {
1393: if (NULL == sec && '.' == *p) {
1394: sec = p + 1;
1395: *p = '\0';
1.11 schwarze 1396: if ('0' == *sec)
1397: src_form |= MANDOC_FORM;
1398: else if ('1' <= *sec && '9' >= *sec)
1399: src_form |= MANDOC_SRC;
1.6 schwarze 1400: continue;
1401: }
1402: if ('/' != *p)
1403: continue;
1404: if (NULL == title) {
1405: title = p + 1;
1406: *p = '\0';
1407: continue;
1408: }
1.11 schwarze 1409: if (strncmp("man", p + 1, 3)) {
1410: src_form |= MANDOC_SRC;
1411: arch = p + 1;
1412: } else if (strncmp("cat", p + 1, 3)) {
1413: src_form |= MANDOC_FORM;
1.6 schwarze 1414: arch = p + 1;
1.11 schwarze 1415: }
1.6 schwarze 1416: break;
1417: }
1418: if (NULL == title)
1419: title = buf;
1420:
1421: /*
1422: * Build the file structure.
1423: */
1424:
1.2 schwarze 1425: nof = mandoc_calloc(1, sizeof(struct of));
1.6 schwarze 1426: nof->fname = mandoc_strdup(argv[i]);
1427: if (NULL != sec)
1428: nof->sec = mandoc_strdup(sec);
1429: if (NULL != arch)
1430: nof->arch = mandoc_strdup(arch);
1431: nof->title = mandoc_strdup(title);
1.11 schwarze 1432: nof->src_form = src_form;
1.6 schwarze 1433:
1434: /*
1435: * Add the structure to the list.
1436: */
1437:
1.2 schwarze 1438: if (verb > 2)
1439: printf("%s: Scheduling\n", argv[i]);
1440: if (NULL == *of) {
1441: *of = nof;
1442: (*of)->first = nof;
1443: } else {
1444: nof->first = (*of)->first;
1445: (*of)->next = nof;
1446: *of = nof;
1447: }
1448: }
1449: }
1450:
1451: /*
1452: * Recursively build up a list of files to parse.
1453: * We use this instead of ftw() and so on because I don't want global
1454: * variables hanging around.
1455: * This ignores the mandoc.db and mandoc.index files, but assumes that
1456: * everything else is a manual.
1457: * Pass in a pointer to a NULL structure for the first invocation.
1458: */
1459: static int
1.6 schwarze 1460: ofile_dirbuild(const char *dir, const char* psec, const char *parch,
1.13 schwarze 1461: int p_src_form, struct of **of)
1.2 schwarze 1462: {
1463: char buf[MAXPATHLEN];
1.11 schwarze 1464: struct stat sb;
1.2 schwarze 1465: size_t sz;
1466: DIR *d;
1.6 schwarze 1467: const char *fn, *sec, *arch;
1.11 schwarze 1468: char *p, *q, *suffix;
1.2 schwarze 1469: struct of *nof;
1470: struct dirent *dp;
1.11 schwarze 1471: int src_form;
1.2 schwarze 1472:
1473: if (NULL == (d = opendir(dir))) {
1474: perror(dir);
1475: return(0);
1476: }
1477:
1478: while (NULL != (dp = readdir(d))) {
1479: fn = dp->d_name;
1.6 schwarze 1480:
1481: if ('.' == *fn)
1482: continue;
1483:
1.11 schwarze 1484: src_form = p_src_form;
1485:
1.2 schwarze 1486: if (DT_DIR == dp->d_type) {
1.6 schwarze 1487: sec = psec;
1488: arch = parch;
1489:
1490: /*
1.8 schwarze 1491: * By default, only use directories called:
1.11 schwarze 1492: * man<section>/[<arch>/] or
1493: * cat<section>/[<arch>/]
1.6 schwarze 1494: */
1495:
1496: if (NULL == sec) {
1.11 schwarze 1497: if(0 == strncmp("man", fn, 3)) {
1498: src_form |= MANDOC_SRC;
1.6 schwarze 1499: sec = fn + 3;
1.11 schwarze 1500: } else if (0 == strncmp("cat", fn, 3)) {
1501: src_form |= MANDOC_FORM;
1502: sec = fn + 3;
1503: } else if (use_all)
1.6 schwarze 1504: sec = fn;
1505: else
1506: continue;
1507: } else if (NULL == arch && (use_all ||
1508: NULL == strchr(fn, '.')))
1509: arch = fn;
1510: else if (0 == use_all)
1.2 schwarze 1511: continue;
1512:
1513: buf[0] = '\0';
1514: strlcat(buf, dir, MAXPATHLEN);
1515: strlcat(buf, "/", MAXPATHLEN);
1516: sz = strlcat(buf, fn, MAXPATHLEN);
1517:
1.6 schwarze 1518: if (MAXPATHLEN <= sz) {
1519: fprintf(stderr, "%s: Path too long\n", dir);
1520: return(0);
1521: }
1522:
1523: if (verb > 2)
1524: printf("%s: Scanning\n", buf);
1525:
1526: if ( ! ofile_dirbuild(buf, sec, arch,
1.13 schwarze 1527: src_form, of))
1.6 schwarze 1528: return(0);
1529: }
1530: if (DT_REG != dp->d_type ||
1531: (NULL == psec && !use_all) ||
1532: !strcmp(MANDOC_DB, fn) ||
1533: !strcmp(MANDOC_IDX, fn))
1534: continue;
1535:
1536: /*
1.8 schwarze 1537: * By default, skip files where the file name suffix
1538: * does not agree with the section directory
1539: * they are located in.
1.6 schwarze 1540: */
1541:
1542: suffix = strrchr(fn, '.');
1543: if (0 == use_all) {
1544: if (NULL == suffix)
1.2 schwarze 1545: continue;
1.11 schwarze 1546: if ((MANDOC_SRC & src_form &&
1547: strcmp(suffix + 1, psec)) ||
1548: (MANDOC_FORM & src_form &&
1549: strcmp(suffix + 1, "0")))
1550: continue;
1551: }
1552: if (NULL != suffix) {
1553: if ('0' == suffix[1])
1554: src_form |= MANDOC_FORM;
1555: else if ('1' <= suffix[1] && '9' >= suffix[1])
1556: src_form |= MANDOC_SRC;
1557: }
1558:
1559:
1560: /*
1561: * Skip formatted manuals if a source version is
1562: * available. Ignore the age: it is very unlikely
1563: * that people install newer formatted base manuals
1564: * when they used to have source manuals before,
1565: * and in ports, old manuals get removed on update.
1566: */
1567: if (0 == use_all && MANDOC_FORM & src_form &&
1568: NULL != psec) {
1569: buf[0] = '\0';
1570: strlcat(buf, dir, MAXPATHLEN);
1571: p = strrchr(buf, '/');
1572: if (NULL == p)
1573: p = buf;
1574: else
1575: p++;
1576: if (0 == strncmp("cat", p, 3))
1577: memcpy(p, "man", 3);
1578: strlcat(buf, "/", MAXPATHLEN);
1579: sz = strlcat(buf, fn, MAXPATHLEN);
1580: if (sz >= MAXPATHLEN) {
1581: fprintf(stderr, "%s: Path too long\n", buf);
1.2 schwarze 1582: continue;
1.11 schwarze 1583: }
1584: q = strrchr(buf, '.');
1585: if (NULL != q && p < q++) {
1586: *q = '\0';
1587: sz = strlcat(buf, psec, MAXPATHLEN);
1588: if (sz >= MAXPATHLEN) {
1589: fprintf(stderr,
1590: "%s: Path too long\n", buf);
1591: continue;
1592: }
1593: if (0 == stat(buf, &sb))
1594: continue;
1595: }
1.2 schwarze 1596: }
1597:
1598: buf[0] = '\0';
1599: strlcat(buf, dir, MAXPATHLEN);
1600: strlcat(buf, "/", MAXPATHLEN);
1601: sz = strlcat(buf, fn, MAXPATHLEN);
1602: if (sz >= MAXPATHLEN) {
1603: fprintf(stderr, "%s: Path too long\n", dir);
1.11 schwarze 1604: continue;
1.2 schwarze 1605: }
1606:
1607: nof = mandoc_calloc(1, sizeof(struct of));
1608: nof->fname = mandoc_strdup(buf);
1.6 schwarze 1609: if (NULL != psec)
1610: nof->sec = mandoc_strdup(psec);
1611: if (NULL != parch)
1612: nof->arch = mandoc_strdup(parch);
1.11 schwarze 1613: nof->src_form = src_form;
1.8 schwarze 1614:
1615: /*
1616: * Remember the file name without the extension,
1617: * to be used as the page title in the database.
1618: */
1619:
1.6 schwarze 1620: if (NULL != suffix)
1621: *suffix = '\0';
1622: nof->title = mandoc_strdup(fn);
1.2 schwarze 1623:
1.11 schwarze 1624: /*
1625: * Add the structure to the list.
1626: */
1627:
1.2 schwarze 1628: if (verb > 2)
1629: printf("%s: Scheduling\n", buf);
1630: if (NULL == *of) {
1631: *of = nof;
1632: (*of)->first = nof;
1633: } else {
1634: nof->first = (*of)->first;
1635: (*of)->next = nof;
1636: *of = nof;
1637: }
1638: }
1639:
1.4 schwarze 1640: closedir(d);
1.2 schwarze 1641: return(1);
1642: }
1643:
1644: static void
1645: ofile_free(struct of *of)
1646: {
1647: struct of *nof;
1648:
1649: while (of) {
1650: nof = of->next;
1651: free(of->fname);
1.6 schwarze 1652: free(of->sec);
1653: free(of->arch);
1654: free(of->title);
1.2 schwarze 1655: free(of);
1656: of = nof;
1657: }
1658: }
1659:
1660: static void
1.1 schwarze 1661: usage(void)
1662: {
1663:
1.2 schwarze 1664: fprintf(stderr, "usage: %s [-v] "
1665: "[-d dir [files...] |"
1666: " -u dir [files...] |"
1667: " dir...]\n", progname);
1.1 schwarze 1668: }