Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.215
1.215 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.214 2020/01/26 11:15:49 schwarze Exp $ */
1.1 schwarze 2: /*
1.47 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.213 schwarze 4: * Copyright (c) 2011-2020 Ingo Schwarze <schwarze@openbsd.org>
1.180 schwarze 5: * Copyright (c) 2016 Ed Maste <emaste@freebsd.org>
1.1 schwarze 6: *
7: * Permission to use, copy, modify, and distribute this software for any
8: * purpose with or without fee is hereby granted, provided that the above
9: * copyright notice and this permission notice appear in all copies.
10: *
1.140 schwarze 11: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
1.1 schwarze 12: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
1.140 schwarze 13: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
1.1 schwarze 14: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18: */
1.117 schwarze 19: #include <sys/types.h>
1.204 schwarze 20: #include <sys/mman.h>
1.47 schwarze 21: #include <sys/stat.h>
1.1 schwarze 22:
23: #include <assert.h>
1.33 schwarze 24: #include <ctype.h>
1.151 schwarze 25: #include <err.h>
1.34 schwarze 26: #include <errno.h>
1.1 schwarze 27: #include <fcntl.h>
1.47 schwarze 28: #include <fts.h>
1.44 schwarze 29: #include <limits.h>
1.173 schwarze 30: #include <stdarg.h>
1.47 schwarze 31: #include <stddef.h>
1.1 schwarze 32: #include <stdio.h>
33: #include <stdint.h>
34: #include <stdlib.h>
35: #include <string.h>
1.14 schwarze 36: #include <unistd.h>
1.1 schwarze 37:
1.141 schwarze 38: #include "mandoc_aux.h"
1.156 schwarze 39: #include "mandoc_ohash.h"
1.141 schwarze 40: #include "mandoc.h"
41: #include "roff.h"
1.47 schwarze 42: #include "mdoc.h"
1.1 schwarze 43: #include "man.h"
1.209 schwarze 44: #include "mandoc_parse.h"
1.140 schwarze 45: #include "manconf.h"
1.47 schwarze 46: #include "mansearch.h"
1.173 schwarze 47: #include "dba_array.h"
48: #include "dba.h"
1.1 schwarze 49:
1.68 schwarze 50: extern const char *const mansearch_keynames[];
51:
1.47 schwarze 52: enum op {
53: OP_DEFAULT = 0, /* new dbs from dir list or default config */
54: OP_CONFFILE, /* new databases from custom config file */
55: OP_UPDATE, /* delete/add entries in existing database */
56: OP_DELETE, /* delete entries from existing database */
57: OP_TEST /* change no databases, report potential problems */
58: };
1.11 schwarze 59:
1.47 schwarze 60: struct str {
61: const struct mpage *mpage; /* if set, the owning parse */
62: uint64_t mask; /* bitmask in sequence */
1.131 schwarze 63: char key[]; /* rendered text */
1.28 schwarze 64: };
65:
1.47 schwarze 66: struct inodev {
67: ino_t st_ino;
68: dev_t st_dev;
69: };
1.28 schwarze 70:
1.47 schwarze 71: struct mpage {
72: struct inodev inodev; /* used for hashing routine */
1.173 schwarze 73: struct dba_array *dba;
1.47 schwarze 74: char *sec; /* section from file content */
75: char *arch; /* architecture from file content */
76: char *title; /* title from file content */
77: char *desc; /* description from file content */
1.180 schwarze 78: struct mpage *next; /* singly linked list */
1.47 schwarze 79: struct mlink *mlinks; /* singly linked list */
1.129 schwarze 80: int name_head_done;
1.173 schwarze 81: enum form form; /* format from file content */
1.28 schwarze 82: };
83:
1.47 schwarze 84: struct mlink {
85: char file[PATH_MAX]; /* filename rel. to manpath */
86: char *dsec; /* section from directory */
87: char *arch; /* architecture from directory */
88: char *name; /* name from file name (not empty) */
89: char *fsec; /* section from file name suffix */
90: struct mlink *next; /* singly linked list */
1.75 schwarze 91: struct mpage *mpage; /* parent */
1.81 schwarze 92: int gzip; /* filename has a .gz suffix */
1.173 schwarze 93: enum form dform; /* format from directory */
94: enum form fform; /* format from file name suffix */
1.1 schwarze 95: };
96:
1.143 schwarze 97: typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
1.142 schwarze 98: const struct roff_node *);
1.1 schwarze 99:
1.19 schwarze 100: struct mdoc_handler {
1.47 schwarze 101: mdoc_fp fp; /* optional handler */
102: uint64_t mask; /* set unless handler returns 0 */
1.184 schwarze 103: int taboo; /* node flags that must not be set */
1.19 schwarze 104: };
1.171 schwarze 105:
106:
107: int mandocdb(int, char *[]);
1.19 schwarze 108:
1.173 schwarze 109: static void dbadd(struct dba *, struct mpage *);
1.75 schwarze 110: static void dbadd_mlink(const struct mlink *mlink);
1.173 schwarze 111: static void dbprune(struct dba *);
112: static void dbwrite(struct dba *);
1.47 schwarze 113: static void filescan(const char *);
1.180 schwarze 114: static int fts_compare(const FTSENT **, const FTSENT **);
1.47 schwarze 115: static void mlink_add(struct mlink *, const struct stat *);
1.88 schwarze 116: static void mlink_check(struct mpage *, struct mlink *);
1.47 schwarze 117: static void mlink_free(struct mlink *);
118: static void mlinks_undupe(struct mpage *);
119: static void mpages_free(void);
1.173 schwarze 120: static void mpages_merge(struct dba *, struct mparse *);
1.81 schwarze 121: static void parse_cat(struct mpage *, int);
1.143 schwarze 122: static void parse_man(struct mpage *, const struct roff_meta *,
1.142 schwarze 123: const struct roff_node *);
1.143 schwarze 124: static void parse_mdoc(struct mpage *, const struct roff_meta *,
1.142 schwarze 125: const struct roff_node *);
1.143 schwarze 126: static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
1.142 schwarze 127: const struct roff_node *);
1.207 schwarze 128: static int parse_mdoc_Fa(struct mpage *, const struct roff_meta *,
129: const struct roff_node *);
1.143 schwarze 130: static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
1.142 schwarze 131: const struct roff_node *);
132: static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
1.143 schwarze 133: static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
1.142 schwarze 134: const struct roff_node *);
1.143 schwarze 135: static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
1.142 schwarze 136: const struct roff_node *);
1.143 schwarze 137: static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
1.142 schwarze 138: const struct roff_node *);
1.143 schwarze 139: static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
1.142 schwarze 140: const struct roff_node *);
1.143 schwarze 141: static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
1.142 schwarze 142: const struct roff_node *);
1.164 schwarze 143: static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
144: const struct roff_node *);
1.143 schwarze 145: static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
1.142 schwarze 146: const struct roff_node *);
1.69 schwarze 147: static void putkey(const struct mpage *, char *, uint64_t);
1.131 schwarze 148: static void putkeys(const struct mpage *, char *, size_t, uint64_t);
1.47 schwarze 149: static void putmdockey(const struct mpage *,
1.184 schwarze 150: const struct roff_node *, uint64_t, int);
1.131 schwarze 151: static int render_string(char **, size_t *);
1.172 schwarze 152: static void say(const char *, const char *, ...)
1.192 schwarze 153: __attribute__((__format__ (__printf__, 2, 3)));
1.120 schwarze 154: static int set_basedir(const char *, int);
1.47 schwarze 155: static int treescan(void);
156: static size_t utf8(unsigned int, char [7]);
157:
1.59 schwarze 158: static int nodb; /* no database changes */
1.73 schwarze 159: static int mparse_options; /* abort the parse early */
1.98 schwarze 160: static int use_all; /* use all found files */
161: static int debug; /* print what we're doing */
162: static int warnings; /* warn about crap */
1.52 schwarze 163: static int write_utf8; /* write UTF-8 output; else ASCII */
1.47 schwarze 164: static int exitcode; /* to be returned by main */
1.98 schwarze 165: static enum op op; /* operational mode */
1.47 schwarze 166: static char basedir[PATH_MAX]; /* current base directory */
1.213 schwarze 167: static size_t basedir_len; /* strlen(basedir) */
1.180 schwarze 168: static struct mpage *mpage_head; /* list of distinct manual pages */
1.47 schwarze 169: static struct ohash mpages; /* table of distinct manual pages */
170: static struct ohash mlinks; /* table of directory entries */
1.90 schwarze 171: static struct ohash names; /* table of all names */
1.47 schwarze 172: static struct ohash strings; /* table of all strings */
1.90 schwarze 173: static uint64_t name_mask;
1.47 schwarze 174:
1.208 schwarze 175: static const struct mdoc_handler mdoc_handlers[MDOC_MAX - MDOC_Dd] = {
1.184 schwarze 176: { NULL, 0, NODE_NOPRT }, /* Dd */
177: { NULL, 0, NODE_NOPRT }, /* Dt */
178: { NULL, 0, NODE_NOPRT }, /* Os */
179: { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
180: { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
181: { NULL, 0, 0 }, /* Pp */
182: { NULL, 0, 0 }, /* D1 */
183: { NULL, 0, 0 }, /* Dl */
184: { NULL, 0, 0 }, /* Bd */
185: { NULL, 0, 0 }, /* Ed */
186: { NULL, 0, 0 }, /* Bl */
187: { NULL, 0, 0 }, /* El */
188: { NULL, 0, 0 }, /* It */
189: { NULL, 0, 0 }, /* Ad */
190: { NULL, TYPE_An, 0 }, /* An */
1.195 schwarze 191: { NULL, 0, 0 }, /* Ap */
1.184 schwarze 192: { NULL, TYPE_Ar, 0 }, /* Ar */
193: { NULL, TYPE_Cd, 0 }, /* Cd */
194: { NULL, TYPE_Cm, 0 }, /* Cm */
195: { NULL, TYPE_Dv, 0 }, /* Dv */
196: { NULL, TYPE_Er, 0 }, /* Er */
197: { NULL, TYPE_Ev, 0 }, /* Ev */
198: { NULL, 0, 0 }, /* Ex */
1.207 schwarze 199: { parse_mdoc_Fa, 0, 0 }, /* Fa */
1.184 schwarze 200: { parse_mdoc_Fd, 0, 0 }, /* Fd */
201: { NULL, TYPE_Fl, 0 }, /* Fl */
202: { parse_mdoc_Fn, 0, 0 }, /* Fn */
1.207 schwarze 203: { NULL, TYPE_Ft | TYPE_Vt, 0 }, /* Ft */
1.184 schwarze 204: { NULL, TYPE_Ic, 0 }, /* Ic */
205: { NULL, TYPE_In, 0 }, /* In */
206: { NULL, TYPE_Li, 0 }, /* Li */
207: { parse_mdoc_Nd, 0, 0 }, /* Nd */
208: { parse_mdoc_Nm, 0, 0 }, /* Nm */
209: { NULL, 0, 0 }, /* Op */
210: { NULL, 0, 0 }, /* Ot */
211: { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
212: { NULL, 0, 0 }, /* Rv */
213: { NULL, TYPE_St, 0 }, /* St */
214: { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
215: { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
216: { parse_mdoc_Xr, 0, 0 }, /* Xr */
217: { NULL, 0, 0 }, /* %A */
218: { NULL, 0, 0 }, /* %B */
219: { NULL, 0, 0 }, /* %D */
220: { NULL, 0, 0 }, /* %I */
221: { NULL, 0, 0 }, /* %J */
222: { NULL, 0, 0 }, /* %N */
223: { NULL, 0, 0 }, /* %O */
224: { NULL, 0, 0 }, /* %P */
225: { NULL, 0, 0 }, /* %R */
226: { NULL, 0, 0 }, /* %T */
227: { NULL, 0, 0 }, /* %V */
228: { NULL, 0, 0 }, /* Ac */
229: { NULL, 0, 0 }, /* Ao */
230: { NULL, 0, 0 }, /* Aq */
1.186 schwarze 231: { NULL, TYPE_At, 0 }, /* At */
1.184 schwarze 232: { NULL, 0, 0 }, /* Bc */
233: { NULL, 0, 0 }, /* Bf */
234: { NULL, 0, 0 }, /* Bo */
235: { NULL, 0, 0 }, /* Bq */
236: { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
1.185 schwarze 237: { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
1.184 schwarze 238: { NULL, 0, 0 }, /* Db */
239: { NULL, 0, 0 }, /* Dc */
240: { NULL, 0, 0 }, /* Do */
241: { NULL, 0, 0 }, /* Dq */
242: { NULL, 0, 0 }, /* Ec */
243: { NULL, 0, 0 }, /* Ef */
244: { NULL, TYPE_Em, 0 }, /* Em */
245: { NULL, 0, 0 }, /* Eo */
246: { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
247: { NULL, TYPE_Ms, 0 }, /* Ms */
248: { NULL, 0, 0 }, /* No */
249: { NULL, 0, 0 }, /* Ns */
250: { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
251: { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
252: { NULL, 0, 0 }, /* Pc */
253: { NULL, 0, 0 }, /* Pf */
254: { NULL, 0, 0 }, /* Po */
255: { NULL, 0, 0 }, /* Pq */
256: { NULL, 0, 0 }, /* Qc */
257: { NULL, 0, 0 }, /* Ql */
258: { NULL, 0, 0 }, /* Qo */
259: { NULL, 0, 0 }, /* Qq */
260: { NULL, 0, 0 }, /* Re */
261: { NULL, 0, 0 }, /* Rs */
262: { NULL, 0, 0 }, /* Sc */
263: { NULL, 0, 0 }, /* So */
264: { NULL, 0, 0 }, /* Sq */
265: { NULL, 0, 0 }, /* Sm */
266: { NULL, 0, 0 }, /* Sx */
267: { NULL, TYPE_Sy, 0 }, /* Sy */
268: { NULL, TYPE_Tn, 0 }, /* Tn */
269: { NULL, 0, NODE_NOSRC }, /* Ux */
270: { NULL, 0, 0 }, /* Xc */
271: { NULL, 0, 0 }, /* Xo */
272: { parse_mdoc_Fo, 0, 0 }, /* Fo */
273: { NULL, 0, 0 }, /* Fc */
274: { NULL, 0, 0 }, /* Oo */
275: { NULL, 0, 0 }, /* Oc */
276: { NULL, 0, 0 }, /* Bk */
277: { NULL, 0, 0 }, /* Ek */
278: { NULL, 0, 0 }, /* Bt */
279: { NULL, 0, 0 }, /* Hf */
280: { NULL, 0, 0 }, /* Fr */
281: { NULL, 0, 0 }, /* Ud */
1.186 schwarze 282: { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */
1.184 schwarze 283: { NULL, 0, 0 }, /* Lp */
284: { NULL, TYPE_Lk, 0 }, /* Lk */
285: { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
286: { NULL, 0, 0 }, /* Brq */
287: { NULL, 0, 0 }, /* Bro */
288: { NULL, 0, 0 }, /* Brc */
289: { NULL, 0, 0 }, /* %C */
290: { NULL, 0, 0 }, /* Es */
291: { NULL, 0, 0 }, /* En */
292: { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
293: { NULL, 0, 0 }, /* %Q */
294: { NULL, 0, 0 }, /* %U */
295: { NULL, 0, 0 }, /* Ta */
1.1 schwarze 296: };
297:
1.98 schwarze 298:
1.1 schwarze 299: int
1.3 schwarze 300: mandocdb(int argc, char *argv[])
1.1 schwarze 301: {
1.140 schwarze 302: struct manconf conf;
303: struct mparse *mp;
1.173 schwarze 304: struct dba *dba;
1.161 schwarze 305: const char *path_arg, *progname;
1.140 schwarze 306: size_t j, sz;
1.47 schwarze 307: int ch, i;
308:
1.204 schwarze 309: if (pledge("stdio rpath wpath cpath", NULL) == -1) {
1.167 schwarze 310: warn("pledge");
1.155 schwarze 311: return (int)MANDOCLEVEL_SYSERR;
312: }
313:
1.140 schwarze 314: memset(&conf, 0, sizeof(conf));
1.47 schwarze 315:
316: /*
1.98 schwarze 317: * We accept a few different invocations.
1.47 schwarze 318: * The CHECKOP macro makes sure that invocation styles don't
319: * clobber each other.
320: */
321: #define CHECKOP(_op, _ch) do \
1.213 schwarze 322: if ((_op) != OP_DEFAULT) { \
1.151 schwarze 323: warnx("-%c: Conflicting option", (_ch)); \
1.47 schwarze 324: goto usage; \
325: } while (/*CONSTCOND*/0)
1.10 schwarze 326:
1.211 schwarze 327: mparse_options = MPARSE_VALIDATE;
1.47 schwarze 328: path_arg = NULL;
1.28 schwarze 329: op = OP_DEFAULT;
1.1 schwarze 330:
1.213 schwarze 331: while ((ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")) != -1)
1.1 schwarze 332: switch (ch) {
1.98 schwarze 333: case 'a':
1.6 schwarze 334: use_all = 1;
335: break;
1.98 schwarze 336: case 'C':
1.47 schwarze 337: CHECKOP(op, ch);
338: path_arg = optarg;
1.28 schwarze 339: op = OP_CONFFILE;
1.25 schwarze 340: break;
1.98 schwarze 341: case 'D':
1.82 schwarze 342: debug++;
343: break;
1.98 schwarze 344: case 'd':
1.47 schwarze 345: CHECKOP(op, ch);
346: path_arg = optarg;
1.2 schwarze 347: op = OP_UPDATE;
1.1 schwarze 348: break;
1.98 schwarze 349: case 'n':
1.47 schwarze 350: nodb = 1;
351: break;
1.98 schwarze 352: case 'p':
1.83 schwarze 353: warnings = 1;
354: break;
1.98 schwarze 355: case 'Q':
1.73 schwarze 356: mparse_options |= MPARSE_QUICK;
1.59 schwarze 357: break;
1.98 schwarze 358: case 'T':
1.213 schwarze 359: if (strcmp(optarg, "utf8") != 0) {
1.151 schwarze 360: warnx("-T%s: Unsupported output format",
361: optarg);
1.52 schwarze 362: goto usage;
363: }
364: write_utf8 = 1;
365: break;
1.98 schwarze 366: case 't':
1.47 schwarze 367: CHECKOP(op, ch);
1.28 schwarze 368: dup2(STDOUT_FILENO, STDERR_FILENO);
369: op = OP_TEST;
1.47 schwarze 370: nodb = warnings = 1;
1.28 schwarze 371: break;
1.98 schwarze 372: case 'u':
1.47 schwarze 373: CHECKOP(op, ch);
374: path_arg = optarg;
1.1 schwarze 375: op = OP_DELETE;
376: break;
1.98 schwarze 377: case 'v':
1.83 schwarze 378: /* Compatibility with espie@'s makewhatis. */
1.28 schwarze 379: break;
1.1 schwarze 380: default:
1.28 schwarze 381: goto usage;
1.1 schwarze 382: }
383:
384: argc -= optind;
385: argv += optind;
386:
1.163 schwarze 387: if (nodb) {
388: if (pledge("stdio rpath", NULL) == -1) {
1.167 schwarze 389: warn("pledge");
1.163 schwarze 390: return (int)MANDOCLEVEL_SYSERR;
391: }
1.155 schwarze 392: }
393:
1.213 schwarze 394: if (op == OP_CONFFILE && argc > 0) {
1.151 schwarze 395: warnx("-C: Too many arguments");
1.28 schwarze 396: goto usage;
397: }
398:
1.47 schwarze 399: exitcode = (int)MANDOCLEVEL_OK;
1.157 schwarze 400: mchars_alloc();
1.210 schwarze 401: mp = mparse_alloc(mparse_options, MANDOC_OS_OTHER, NULL);
1.156 schwarze 402: mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
403: mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
1.2 schwarze 404:
1.213 schwarze 405: if (op == OP_UPDATE || op == OP_DELETE || op == OP_TEST) {
1.2 schwarze 406:
1.47 schwarze 407: /*
1.109 schwarze 408: * Most of these deal with a specific directory.
1.97 schwarze 409: * Jump into that directory first.
1.47 schwarze 410: */
1.213 schwarze 411: if (op != OP_TEST && set_basedir(path_arg, 1) == 0)
1.47 schwarze 412: goto out;
1.97 schwarze 413:
1.174 schwarze 414: dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
415: if (dba != NULL) {
1.97 schwarze 416: /*
417: * The existing database is usable. Process
418: * all files specified on the command-line.
419: */
420: use_all = 1;
1.95 schwarze 421: for (i = 0; i < argc; i++)
422: filescan(argv[i]);
1.174 schwarze 423: if (nodb == 0)
1.173 schwarze 424: dbprune(dba);
1.95 schwarze 425: } else {
1.177 schwarze 426: /* Database missing or corrupt. */
1.179 schwarze 427: if (op != OP_UPDATE || errno != ENOENT)
428: say(MANDOC_DB, "%s: Automatically recreating"
429: " from scratch", strerror(errno));
1.97 schwarze 430: exitcode = (int)MANDOCLEVEL_OK;
1.95 schwarze 431: op = OP_DEFAULT;
1.213 schwarze 432: if (treescan() == 0)
1.95 schwarze 433: goto out;
1.173 schwarze 434: dba = dba_new(128);
1.95 schwarze 435: }
1.213 schwarze 436: if (op != OP_DELETE)
1.173 schwarze 437: mpages_merge(dba, mp);
438: if (nodb == 0)
439: dbwrite(dba);
440: dba_free(dba);
1.47 schwarze 441: } else {
442: /*
443: * If we have arguments, use them as our manpaths.
1.181 schwarze 444: * If we don't, use man.conf(5).
1.47 schwarze 445: */
446: if (argc > 0) {
1.140 schwarze 447: conf.manpath.paths = mandoc_reallocarray(NULL,
1.101 schwarze 448: argc, sizeof(char *));
1.140 schwarze 449: conf.manpath.sz = (size_t)argc;
1.47 schwarze 450: for (i = 0; i < argc; i++)
1.140 schwarze 451: conf.manpath.paths[i] = mandoc_strdup(argv[i]);
1.47 schwarze 452: } else
1.140 schwarze 453: manconf_parse(&conf, path_arg, NULL, NULL);
1.84 schwarze 454:
1.140 schwarze 455: if (conf.manpath.sz == 0) {
1.84 schwarze 456: exitcode = (int)MANDOCLEVEL_BADARG;
457: say("", "Empty manpath");
458: }
1.2 schwarze 459:
1.47 schwarze 460: /*
461: * First scan the tree rooted at a base directory, then
462: * build a new database and finally move it into place.
463: * Ignore zero-length directories and strip trailing
464: * slashes.
465: */
1.140 schwarze 466: for (j = 0; j < conf.manpath.sz; j++) {
467: sz = strlen(conf.manpath.paths[j]);
468: if (sz && conf.manpath.paths[j][sz - 1] == '/')
469: conf.manpath.paths[j][--sz] = '\0';
1.213 schwarze 470: if (sz == 0)
1.47 schwarze 471: continue;
1.2 schwarze 472:
1.47 schwarze 473: if (j) {
1.156 schwarze 474: mandoc_ohash_init(&mpages, 6,
475: offsetof(struct mpage, inodev));
476: mandoc_ohash_init(&mlinks, 6,
477: offsetof(struct mlink, file));
1.47 schwarze 478: }
1.1 schwarze 479:
1.213 schwarze 480: if (set_basedir(conf.manpath.paths[j], argc > 0) == 0)
1.120 schwarze 481: continue;
1.213 schwarze 482: if (treescan() == 0)
1.120 schwarze 483: continue;
1.173 schwarze 484: dba = dba_new(128);
485: mpages_merge(dba, mp);
486: if (nodb == 0)
487: dbwrite(dba);
488: dba_free(dba);
1.28 schwarze 489:
1.140 schwarze 490: if (j + 1 < conf.manpath.sz) {
1.47 schwarze 491: mpages_free();
492: ohash_delete(&mpages);
493: ohash_delete(&mlinks);
494: }
1.2 schwarze 495: }
1.47 schwarze 496: }
497: out:
1.140 schwarze 498: manconf_free(&conf);
1.123 schwarze 499: mparse_free(mp);
1.157 schwarze 500: mchars_free();
1.47 schwarze 501: mpages_free();
502: ohash_delete(&mpages);
503: ohash_delete(&mlinks);
1.150 schwarze 504: return exitcode;
1.47 schwarze 505: usage:
1.161 schwarze 506: progname = getprogname();
1.83 schwarze 507: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
508: " %s [-aDnpQ] [-Tutf8] dir ...\n"
509: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
510: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 511: " %s [-Q] -t file ...\n",
1.161 schwarze 512: progname, progname, progname, progname, progname);
1.1 schwarze 513:
1.150 schwarze 514: return (int)MANDOCLEVEL_BADARG;
1.47 schwarze 515: }
1.1 schwarze 516:
1.47 schwarze 517: /*
1.180 schwarze 518: * To get a singly linked list in alpha order while inserting entries
519: * at the beginning, process directory entries in reverse alpha order.
520: */
521: static int
522: fts_compare(const FTSENT **a, const FTSENT **b)
523: {
524: return -strcmp((*a)->fts_name, (*b)->fts_name);
525: }
526:
527: /*
1.47 schwarze 528: * Scan a directory tree rooted at "basedir" for manpages.
529: * We use fts(), scanning directory parts along the way for clues to our
530: * section and architecture.
531: *
532: * If use_all has been specified, grok all files.
533: * If not, sanitise paths to the following:
534: *
1.98 schwarze 535: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 536: * or
537: * [./]cat<section>[/<arch>]/<name>.0
538: *
1.169 krw 539: * TODO: accommodate for multi-language directories.
1.47 schwarze 540: */
541: static int
542: treescan(void)
543: {
1.96 schwarze 544: char buf[PATH_MAX];
1.47 schwarze 545: FTS *f;
546: FTSENT *ff;
547: struct mlink *mlink;
1.173 schwarze 548: int gzip;
549: enum form dform;
1.51 schwarze 550: char *dsec, *arch, *fsec, *cp;
551: const char *path;
1.47 schwarze 552: const char *argv[2];
1.1 schwarze 553:
1.47 schwarze 554: argv[0] = ".";
1.193 schwarze 555: argv[1] = NULL;
1.1 schwarze 556:
1.180 schwarze 557: f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR,
558: fts_compare);
1.153 schwarze 559: if (f == NULL) {
1.47 schwarze 560: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 561: say("", "&fts_open");
1.150 schwarze 562: return 0;
1.47 schwarze 563: }
1.2 schwarze 564:
1.47 schwarze 565: dsec = arch = NULL;
566: dform = FORM_NONE;
1.2 schwarze 567:
1.153 schwarze 568: while ((ff = fts_read(f)) != NULL) {
1.47 schwarze 569: path = ff->fts_path + 2;
1.96 schwarze 570: switch (ff->fts_info) {
571:
572: /*
573: * Symbolic links require various sanity checks,
574: * then get handled just like regular files.
575: */
1.98 schwarze 576: case FTS_SL:
1.153 schwarze 577: if (realpath(path, buf) == NULL) {
1.96 schwarze 578: if (warnings)
579: say(path, "&realpath");
580: continue;
581: }
1.213 schwarze 582: if (strncmp(buf, basedir, basedir_len) != 0) {
1.96 schwarze 583: if (warnings) say("",
584: "%s: outside base directory", buf);
585: continue;
586: }
587: /* Use logical inode to avoid mpages dupe. */
1.153 schwarze 588: if (stat(path, ff->fts_statp) == -1) {
1.96 schwarze 589: if (warnings)
590: say(path, "&stat");
591: continue;
592: }
593: /* FALLTHROUGH */
594:
1.15 schwarze 595: /*
1.47 schwarze 596: * If we're a regular file, add an mlink by using the
597: * stored directory data and handling the filename.
1.15 schwarze 598: */
1.98 schwarze 599: case FTS_F:
1.153 schwarze 600: if ( ! strcmp(path, MANDOC_DB))
1.47 schwarze 601: continue;
602: if ( ! use_all && ff->fts_level < 2) {
603: if (warnings)
604: say(path, "Extraneous file");
605: continue;
1.81 schwarze 606: }
607: gzip = 0;
608: fsec = NULL;
1.153 schwarze 609: while (fsec == NULL) {
1.81 schwarze 610: fsec = strrchr(ff->fts_name, '.');
1.153 schwarze 611: if (fsec == NULL || strcmp(fsec+1, "gz"))
1.81 schwarze 612: break;
613: gzip = 1;
614: *fsec = '\0';
615: fsec = NULL;
616: }
1.153 schwarze 617: if (fsec == NULL) {
1.47 schwarze 618: if ( ! use_all) {
619: if (warnings)
620: say(path,
621: "No filename suffix");
622: continue;
623: }
1.153 schwarze 624: } else if ( ! strcmp(++fsec, "html")) {
1.47 schwarze 625: if (warnings)
626: say(path, "Skip html");
627: continue;
1.153 schwarze 628: } else if ( ! strcmp(fsec, "ps")) {
1.47 schwarze 629: if (warnings)
630: say(path, "Skip ps");
631: continue;
1.153 schwarze 632: } else if ( ! strcmp(fsec, "pdf")) {
1.47 schwarze 633: if (warnings)
634: say(path, "Skip pdf");
635: continue;
636: } else if ( ! use_all &&
1.153 schwarze 637: ((dform == FORM_SRC &&
1.138 schwarze 638: strncmp(fsec, dsec, strlen(dsec))) ||
1.153 schwarze 639: (dform == FORM_CAT && strcmp(fsec, "0")))) {
1.47 schwarze 640: if (warnings)
641: say(path, "Wrong filename suffix");
642: continue;
643: } else
644: fsec[-1] = '\0';
1.51 schwarze 645:
1.47 schwarze 646: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.100 schwarze 647: if (strlcpy(mlink->file, path,
648: sizeof(mlink->file)) >=
649: sizeof(mlink->file)) {
650: say(path, "Filename too long");
651: free(mlink);
652: continue;
653: }
1.47 schwarze 654: mlink->dform = dform;
1.51 schwarze 655: mlink->dsec = dsec;
656: mlink->arch = arch;
657: mlink->name = ff->fts_name;
658: mlink->fsec = fsec;
1.81 schwarze 659: mlink->gzip = gzip;
1.47 schwarze 660: mlink_add(mlink, ff->fts_statp);
661: continue;
1.96 schwarze 662:
1.98 schwarze 663: case FTS_D:
664: case FTS_DP:
1.96 schwarze 665: break;
666:
667: default:
1.47 schwarze 668: if (warnings)
669: say(path, "Not a regular file");
670: continue;
671: }
672:
673: switch (ff->fts_level) {
1.98 schwarze 674: case 0:
1.47 schwarze 675: /* Ignore the root directory. */
676: break;
1.98 schwarze 677: case 1:
1.47 schwarze 678: /*
679: * This might contain manX/ or catX/.
680: * Try to infer this from the name.
681: * If we're not in use_all, enforce it.
682: */
683: cp = ff->fts_name;
1.153 schwarze 684: if (ff->fts_info == FTS_DP) {
685: dform = FORM_NONE;
686: dsec = NULL;
1.47 schwarze 687: break;
1.153 schwarze 688: }
1.15 schwarze 689:
1.153 schwarze 690: if ( ! strncmp(cp, "man", 3)) {
1.47 schwarze 691: dform = FORM_SRC;
692: dsec = cp + 3;
1.153 schwarze 693: } else if ( ! strncmp(cp, "cat", 3)) {
1.47 schwarze 694: dform = FORM_CAT;
695: dsec = cp + 3;
1.51 schwarze 696: } else {
697: dform = FORM_NONE;
698: dsec = NULL;
1.26 schwarze 699: }
1.47 schwarze 700:
1.153 schwarze 701: if (dsec != NULL || use_all)
1.47 schwarze 702: break;
703:
704: if (warnings)
705: say(path, "Unknown directory part");
706: fts_set(f, ff, FTS_SKIP);
707: break;
1.98 schwarze 708: case 2:
1.47 schwarze 709: /*
710: * Possibly our architecture.
711: * If we're descending, keep tabs on it.
712: */
1.153 schwarze 713: if (ff->fts_info != FTS_DP && dsec != NULL)
1.47 schwarze 714: arch = ff->fts_name;
1.51 schwarze 715: else
716: arch = NULL;
1.47 schwarze 717: break;
718: default:
1.153 schwarze 719: if (ff->fts_info == FTS_DP || use_all)
1.47 schwarze 720: break;
721: if (warnings)
722: say(path, "Extraneous directory part");
723: fts_set(f, ff, FTS_SKIP);
724: break;
1.14 schwarze 725: }
1.47 schwarze 726: }
727:
728: fts_close(f);
1.150 schwarze 729: return 1;
1.47 schwarze 730: }
1.1 schwarze 731:
1.47 schwarze 732: /*
733: * Add a file to the mlinks table.
734: * Do not verify that it's a "valid" looking manpage (we'll do that
735: * later).
736: *
737: * Try to infer the manual section, architecture, and page name from the
738: * path, assuming it looks like
739: *
1.98 schwarze 740: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 741: * or
742: * [./]cat<section>[/<arch>]/<name>.0
743: *
744: * See treescan() for the fts(3) version of this.
745: */
746: static void
1.215 ! schwarze 747: filescan(const char *infile)
1.47 schwarze 748: {
749: struct stat st;
750: struct mlink *mlink;
1.215 ! schwarze 751: char *linkfile, *p, *realdir, *start, *usefile;
! 752: size_t realdir_len;
1.47 schwarze 753:
754: assert(use_all);
755:
1.215 ! schwarze 756: if (strncmp(infile, "./", 2) == 0)
! 757: infile += 2;
1.47 schwarze 758:
1.96 schwarze 759: /*
760: * We have to do lstat(2) before realpath(3) loses
761: * the information whether this is a symbolic link.
762: * We need to know that because for symbolic links,
763: * we want to use the orginal file name, while for
764: * regular files, we want to use the real path.
765: */
1.215 ! schwarze 766: if (lstat(infile, &st) == -1) {
1.96 schwarze 767: exitcode = (int)MANDOCLEVEL_BADARG;
1.215 ! schwarze 768: say(infile, "&lstat");
1.96 schwarze 769: return;
1.214 schwarze 770: } else if (S_ISREG(st.st_mode) == 0 && S_ISLNK(st.st_mode) == 0) {
1.96 schwarze 771: exitcode = (int)MANDOCLEVEL_BADARG;
1.215 ! schwarze 772: say(infile, "Not a regular file");
1.96 schwarze 773: return;
774: }
775:
776: /*
777: * We have to resolve the file name to the real path
778: * in any case for the base directory check.
779: */
1.215 ! schwarze 780: if ((usefile = realpath(infile, NULL)) == NULL) {
1.47 schwarze 781: exitcode = (int)MANDOCLEVEL_BADARG;
1.215 ! schwarze 782: say(infile, "&realpath");
1.47 schwarze 783: return;
1.63 schwarze 784: }
785:
1.213 schwarze 786: if (op == OP_TEST)
1.215 ! schwarze 787: start = usefile;
! 788: else if (strncmp(usefile, basedir, basedir_len) == 0)
! 789: start = usefile + basedir_len;
1.63 schwarze 790: else {
1.47 schwarze 791: exitcode = (int)MANDOCLEVEL_BADARG;
1.215 ! schwarze 792: say("", "%s: outside base directory", infile);
! 793: free(usefile);
1.47 schwarze 794: return;
1.63 schwarze 795: }
796:
1.96 schwarze 797: /*
798: * Now we are sure the file is inside our tree.
799: * If it is a symbolic link, ignore the real path
800: * and use the original name.
801: */
1.215 ! schwarze 802: do {
! 803: if (S_ISLNK(st.st_mode) == 0)
! 804: break;
! 805:
! 806: /*
! 807: * Some implementations of realpath(3) may succeed
! 808: * even if the target of the link does not exist,
! 809: * so check again for extra safety.
! 810: */
! 811: if (stat(usefile, &st) == -1) {
1.96 schwarze 812: exitcode = (int)MANDOCLEVEL_BADARG;
1.215 ! schwarze 813: say(infile, "&stat");
! 814: free(usefile);
1.96 schwarze 815: return;
816: }
1.215 ! schwarze 817: linkfile = mandoc_strdup(infile);
! 818: if (op == OP_TEST) {
! 819: free(usefile);
! 820: start = usefile = linkfile;
! 821: break;
! 822: }
! 823: if (strncmp(infile, basedir, basedir_len) == 0) {
! 824: free(usefile);
! 825: usefile = linkfile;
! 826: start = usefile + basedir_len;
! 827: break;
! 828: }
! 829:
! 830: /*
! 831: * This symbolic link points into the basedir
! 832: * from the outside. Let's see whether any of
! 833: * the parent directories resolve to the basedir.
! 834: */
! 835: p = strchr(linkfile, '\0');
! 836: do {
! 837: while (*--p != '/')
! 838: continue;
! 839: *p = '\0';
! 840: if ((realdir = realpath(linkfile, NULL)) == NULL) {
! 841: exitcode = (int)MANDOCLEVEL_BADARG;
! 842: say(infile, "&realpath");
! 843: free(linkfile);
! 844: free(usefile);
! 845: return;
! 846: }
! 847: realdir_len = strlen(realdir) + 1;
! 848: free(realdir);
! 849: *p = '/';
! 850: } while (realdir_len > basedir_len);
! 851:
! 852: /*
! 853: * If one of the directories resolves to the basedir,
! 854: * use the rest of the original name.
! 855: * Otherwise, the best we can do
! 856: * is to use the filename pointed to.
! 857: */
! 858: if (realdir_len == basedir_len) {
! 859: free(usefile);
! 860: usefile = linkfile;
! 861: start = p + 1;
! 862: } else {
! 863: free(linkfile);
! 864: start = usefile + basedir_len;
1.100 schwarze 865: }
1.215 ! schwarze 866: } while (/* CONSTCOND */ 0);
1.63 schwarze 867:
1.47 schwarze 868: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 869: mlink->dform = FORM_NONE;
1.100 schwarze 870: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
871: sizeof(mlink->file)) {
872: say(start, "Filename too long");
1.134 schwarze 873: free(mlink);
1.215 ! schwarze 874: free(usefile);
1.100 schwarze 875: return;
1.188 schwarze 876: }
877:
878: /*
879: * In test mode or when the original name is absolute
880: * but outside our tree, guess the base directory.
881: */
882:
1.215 ! schwarze 883: if (op == OP_TEST || (start == usefile && *start == '/')) {
! 884: if (strncmp(usefile, "man/", 4) == 0)
! 885: start = usefile + 4;
! 886: else if ((start = strstr(usefile, "/man/")) != NULL)
1.188 schwarze 887: start += 5;
888: else
1.215 ! schwarze 889: start = usefile;
1.100 schwarze 890: }
1.1 schwarze 891:
1.10 schwarze 892: /*
1.47 schwarze 893: * First try to guess our directory structure.
894: * If we find a separator, try to look for man* or cat*.
895: * If we find one of these and what's underneath is a directory,
896: * assume it's an architecture.
1.10 schwarze 897: */
1.213 schwarze 898: if ((p = strchr(start, '/')) != NULL) {
1.47 schwarze 899: *p++ = '\0';
1.213 schwarze 900: if (strncmp(start, "man", 3) == 0) {
1.47 schwarze 901: mlink->dform = FORM_SRC;
1.51 schwarze 902: mlink->dsec = start + 3;
1.213 schwarze 903: } else if (strncmp(start, "cat", 3) == 0) {
1.47 schwarze 904: mlink->dform = FORM_CAT;
1.51 schwarze 905: mlink->dsec = start + 3;
1.47 schwarze 906: }
1.10 schwarze 907:
1.47 schwarze 908: start = p;
1.213 schwarze 909: if (mlink->dsec != NULL && (p = strchr(start, '/')) != NULL) {
1.47 schwarze 910: *p++ = '\0';
1.51 schwarze 911: mlink->arch = start;
1.47 schwarze 912: start = p;
1.41 deraadt 913: }
1.47 schwarze 914: }
1.7 schwarze 915:
1.47 schwarze 916: /*
917: * Now check the file suffix.
918: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
919: */
920: p = strrchr(start, '\0');
1.213 schwarze 921: while (p-- > start && *p != '/' && *p != '.')
922: continue;
1.47 schwarze 923:
1.213 schwarze 924: if (*p == '.') {
1.47 schwarze 925: *p++ = '\0';
1.51 schwarze 926: mlink->fsec = p;
1.47 schwarze 927: }
1.41 deraadt 928:
1.47 schwarze 929: /*
930: * Now try to parse the name.
931: * Use the filename portion of the path.
932: */
933: mlink->name = start;
1.213 schwarze 934: if ((p = strrchr(start, '/')) != NULL) {
1.47 schwarze 935: mlink->name = p + 1;
936: *p = '\0';
937: }
938: mlink_add(mlink, &st);
1.215 ! schwarze 939: free(usefile);
1.47 schwarze 940: }
1.2 schwarze 941:
1.47 schwarze 942: static void
943: mlink_add(struct mlink *mlink, const struct stat *st)
944: {
945: struct inodev inodev;
946: struct mpage *mpage;
947: unsigned int slot;
948:
949: assert(NULL != mlink->file);
950:
1.51 schwarze 951: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
952: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
953: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
954: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 955:
956: if ('0' == *mlink->fsec) {
957: free(mlink->fsec);
958: mlink->fsec = mandoc_strdup(mlink->dsec);
959: mlink->fform = FORM_CAT;
960: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
961: mlink->fform = FORM_SRC;
962: else
963: mlink->fform = FORM_NONE;
1.1 schwarze 964:
1.47 schwarze 965: slot = ohash_qlookup(&mlinks, mlink->file);
966: assert(NULL == ohash_find(&mlinks, slot));
967: ohash_insert(&mlinks, slot, mlink);
968:
1.139 schwarze 969: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.47 schwarze 970: inodev.st_ino = st->st_ino;
971: inodev.st_dev = st->st_dev;
972: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
973: sizeof(struct inodev), inodev.st_ino);
974: mpage = ohash_find(&mpages, slot);
975: if (NULL == mpage) {
976: mpage = mandoc_calloc(1, sizeof(struct mpage));
977: mpage->inodev.st_ino = inodev.st_ino;
978: mpage->inodev.st_dev = inodev.st_dev;
1.173 schwarze 979: mpage->form = FORM_NONE;
1.180 schwarze 980: mpage->next = mpage_head;
981: mpage_head = mpage;
1.47 schwarze 982: ohash_insert(&mpages, slot, mpage);
983: } else
984: mlink->next = mpage->mlinks;
985: mpage->mlinks = mlink;
1.75 schwarze 986: mlink->mpage = mpage;
1.47 schwarze 987: }
1.1 schwarze 988:
1.47 schwarze 989: static void
990: mlink_free(struct mlink *mlink)
991: {
1.1 schwarze 992:
1.47 schwarze 993: free(mlink->dsec);
994: free(mlink->arch);
995: free(mlink->name);
996: free(mlink->fsec);
997: free(mlink);
998: }
1.1 schwarze 999:
1.47 schwarze 1000: static void
1001: mpages_free(void)
1002: {
1003: struct mpage *mpage;
1004: struct mlink *mlink;
1005:
1.180 schwarze 1006: while ((mpage = mpage_head) != NULL) {
1007: while ((mlink = mpage->mlinks) != NULL) {
1.47 schwarze 1008: mpage->mlinks = mlink->next;
1009: mlink_free(mlink);
1010: }
1.180 schwarze 1011: mpage_head = mpage->next;
1.47 schwarze 1012: free(mpage->sec);
1013: free(mpage->arch);
1014: free(mpage->title);
1015: free(mpage->desc);
1016: free(mpage);
1017: }
1018: }
1.1 schwarze 1019:
1.47 schwarze 1020: /*
1021: * For each mlink to the mpage, check whether the path looks like
1022: * it is formatted, and if it does, check whether a source manual
1023: * exists by the same name, ignoring the suffix.
1024: * If both conditions hold, drop the mlink.
1025: */
1026: static void
1027: mlinks_undupe(struct mpage *mpage)
1028: {
1029: char buf[PATH_MAX];
1030: struct mlink **prev;
1031: struct mlink *mlink;
1032: char *bufp;
1033:
1034: mpage->form = FORM_CAT;
1035: prev = &mpage->mlinks;
1036: while (NULL != (mlink = *prev)) {
1037: if (FORM_CAT != mlink->dform) {
1038: mpage->form = FORM_NONE;
1039: goto nextlink;
1040: }
1.100 schwarze 1041: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 1042: bufp = strstr(buf, "cat");
1043: assert(NULL != bufp);
1044: memcpy(bufp, "man", 3);
1045: if (NULL != (bufp = strrchr(buf, '.')))
1046: *++bufp = '\0';
1.100 schwarze 1047: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 1048: if (NULL == ohash_find(&mlinks,
1.98 schwarze 1049: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 1050: goto nextlink;
1051: if (warnings)
1052: say(mlink->file, "Man source exists: %s", buf);
1053: if (use_all)
1054: goto nextlink;
1055: *prev = mlink->next;
1056: mlink_free(mlink);
1057: continue;
1058: nextlink:
1059: prev = &(*prev)->next;
1.1 schwarze 1060: }
1.47 schwarze 1061: }
1.1 schwarze 1062:
1.88 schwarze 1063: static void
1.50 schwarze 1064: mlink_check(struct mpage *mpage, struct mlink *mlink)
1065: {
1.88 schwarze 1066: struct str *str;
1067: unsigned int slot;
1.50 schwarze 1068:
1069: /*
1070: * Check whether the manual section given in a file
1071: * agrees with the directory where the file is located.
1072: * Some manuals have suffixes like (3p) on their
1073: * section number either inside the file or in the
1074: * directory name, some are linked into more than one
1075: * section, like encrypt(1) = makekey(8).
1076: */
1077:
1078: if (FORM_SRC == mpage->form &&
1.88 schwarze 1079: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1080: say(mlink->file, "Section \"%s\" manual in %s directory",
1081: mpage->sec, mlink->dsec);
1082:
1083: /*
1084: * Manual page directories exist for each kernel
1085: * architecture as returned by machine(1).
1086: * However, many manuals only depend on the
1087: * application architecture as returned by arch(1).
1088: * For example, some (2/ARM) manuals are shared
1089: * across the "armish" and "zaurus" kernel
1090: * architectures.
1091: * A few manuals are even shared across completely
1092: * different architectures, for example fdformat(1)
1.178 schwarze 1093: * on amd64, i386, and sparc64.
1.50 schwarze 1094: */
1095:
1.88 schwarze 1096: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1097: say(mlink->file, "Architecture \"%s\" manual in "
1098: "\"%s\" directory", mpage->arch, mlink->arch);
1099:
1.88 schwarze 1100: /*
1101: * XXX
1.90 schwarze 1102: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1103: */
1104:
1105: if (FORM_CAT == mpage->form)
1106: return;
1107:
1108: /*
1109: * Check whether this mlink
1110: * appears as a name in the NAME section.
1111: */
1.50 schwarze 1112:
1.90 schwarze 1113: slot = ohash_qlookup(&names, mlink->name);
1114: str = ohash_find(&names, slot);
1.88 schwarze 1115: assert(NULL != str);
1.90 schwarze 1116: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1117: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1118: }
1119:
1.47 schwarze 1120: /*
1121: * Run through the files in the global vector "mpages"
1122: * and add them to the database specified in "basedir".
1123: *
1124: * This handles the parsing scheme itself, using the cues of directory
1125: * and filename to determine whether the file is parsable or not.
1126: */
1127: static void
1.173 schwarze 1128: mpages_merge(struct dba *dba, struct mparse *mp)
1.47 schwarze 1129: {
1.180 schwarze 1130: struct mpage *mpage, *mpage_dest;
1.75 schwarze 1131: struct mlink *mlink, *mlink_dest;
1.211 schwarze 1132: struct roff_meta *meta;
1.69 schwarze 1133: char *cp;
1.117 schwarze 1134: int fd;
1.64 schwarze 1135:
1.180 schwarze 1136: for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
1.47 schwarze 1137: mlinks_undupe(mpage);
1.180 schwarze 1138: if ((mlink = mpage->mlinks) == NULL)
1139: continue;
1.1 schwarze 1140:
1.90 schwarze 1141: name_mask = NAME_MASK;
1.156 schwarze 1142: mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1143: mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1.47 schwarze 1144: mparse_reset(mp);
1.211 schwarze 1145: meta = NULL;
1.81 schwarze 1146:
1.166 schwarze 1147: if ((fd = mparse_open(mp, mlink->file)) == -1) {
1.136 schwarze 1148: say(mlink->file, "&open");
1.117 schwarze 1149: goto nextpage;
1.81 schwarze 1150: }
1.11 schwarze 1151:
1152: /*
1.137 schwarze 1153: * Interpret the file as mdoc(7) or man(7) source
1154: * code, unless it is known to be formatted.
1.11 schwarze 1155: */
1.136 schwarze 1156: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.137 schwarze 1157: mparse_readfd(mp, fd, mlink->file);
1.165 schwarze 1158: close(fd);
1.189 schwarze 1159: fd = -1;
1.211 schwarze 1160: meta = mparse_result(mp);
1.47 schwarze 1161: }
1.11 schwarze 1162:
1.211 schwarze 1163: if (meta != NULL && meta->sodest != NULL) {
1.75 schwarze 1164: mlink_dest = ohash_find(&mlinks,
1.211 schwarze 1165: ohash_qlookup(&mlinks, meta->sodest));
1.126 schwarze 1166: if (mlink_dest == NULL) {
1.211 schwarze 1167: mandoc_asprintf(&cp, "%s.gz", meta->sodest);
1.126 schwarze 1168: mlink_dest = ohash_find(&mlinks,
1169: ohash_qlookup(&mlinks, cp));
1170: free(cp);
1171: }
1172: if (mlink_dest != NULL) {
1.75 schwarze 1173:
1174: /* The .so target exists. */
1175:
1176: mpage_dest = mlink_dest->mpage;
1177: while (1) {
1178: mlink->mpage = mpage_dest;
1179:
1180: /*
1181: * If the target was already
1182: * processed, add the links
1183: * to the database now.
1184: * Otherwise, this will
1185: * happen when we come
1186: * to the target.
1187: */
1188:
1.173 schwarze 1189: if (mpage_dest->dba != NULL)
1190: dbadd_mlink(mlink);
1.75 schwarze 1191:
1.126 schwarze 1192: if (mlink->next == NULL)
1.75 schwarze 1193: break;
1194: mlink = mlink->next;
1195: }
1196:
1197: /* Move all links to the target. */
1198:
1199: mlink->next = mlink_dest->next;
1200: mlink_dest->next = mpage->mlinks;
1201: mpage->mlinks = NULL;
1.212 schwarze 1202: goto nextpage;
1.75 schwarze 1203: }
1.212 schwarze 1204: meta->macroset = MACROSET_NONE;
1205: }
1206: if (meta != NULL && meta->macroset == MACROSET_MDOC) {
1.47 schwarze 1207: mpage->form = FORM_SRC;
1.211 schwarze 1208: mpage->sec = meta->msec;
1.112 schwarze 1209: mpage->sec = mandoc_strdup(
1.126 schwarze 1210: mpage->sec == NULL ? "" : mpage->sec);
1.211 schwarze 1211: mpage->arch = meta->arch;
1.47 schwarze 1212: mpage->arch = mandoc_strdup(
1.126 schwarze 1213: mpage->arch == NULL ? "" : mpage->arch);
1.211 schwarze 1214: mpage->title = mandoc_strdup(meta->title);
1215: } else if (meta != NULL && meta->macroset == MACROSET_MAN) {
1216: if (*meta->msec != '\0' || *meta->title != '\0') {
1.189 schwarze 1217: mpage->form = FORM_SRC;
1.211 schwarze 1218: mpage->sec = mandoc_strdup(meta->msec);
1.189 schwarze 1219: mpage->arch = mandoc_strdup(mlink->arch);
1.211 schwarze 1220: mpage->title = mandoc_strdup(meta->title);
1.189 schwarze 1221: } else
1.211 schwarze 1222: meta = NULL;
1.189 schwarze 1223: }
1224:
1225: assert(mpage->desc == NULL);
1.212 schwarze 1226: if (meta == NULL || meta->sodest != NULL) {
1.136 schwarze 1227: mpage->sec = mandoc_strdup(mlink->dsec);
1228: mpage->arch = mandoc_strdup(mlink->arch);
1229: mpage->title = mandoc_strdup(mlink->name);
1.212 schwarze 1230: if (meta == NULL) {
1231: mpage->form = FORM_CAT;
1232: parse_cat(mpage, fd);
1233: } else
1234: mpage->form = FORM_SRC;
1.211 schwarze 1235: } else if (meta->macroset == MACROSET_MDOC)
1236: parse_mdoc(mpage, meta, meta->first);
1.189 schwarze 1237: else
1.211 schwarze 1238: parse_man(mpage, meta, meta->first);
1.191 schwarze 1239: if (mpage->desc == NULL) {
1240: mpage->desc = mandoc_strdup(mlink->name);
1241: if (warnings)
1242: say(mlink->file, "No one-line description, "
1243: "using filename \"%s\"", mlink->name);
1244: }
1.88 schwarze 1245:
1.187 schwarze 1246: for (mlink = mpage->mlinks;
1247: mlink != NULL;
1248: mlink = mlink->next) {
1249: putkey(mpage, mlink->name, NAME_FILE);
1250: if (warnings && !use_all)
1.88 schwarze 1251: mlink_check(mpage, mlink);
1.187 schwarze 1252: }
1.6 schwarze 1253:
1.173 schwarze 1254: dbadd(dba, mpage);
1.81 schwarze 1255:
1256: nextpage:
1.47 schwarze 1257: ohash_delete(&strings);
1.90 schwarze 1258: ohash_delete(&names);
1.47 schwarze 1259: }
1260: }
1.6 schwarze 1261:
1.47 schwarze 1262: static void
1.81 schwarze 1263: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1264: {
1265: FILE *stream;
1.190 schwarze 1266: struct mlink *mlink;
1267: char *line, *p, *title, *sec;
1.162 schwarze 1268: size_t linesz, plen, titlesz;
1269: ssize_t len;
1270: int offs;
1.1 schwarze 1271:
1.190 schwarze 1272: mlink = mpage->mlinks;
1273: stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1274: if (stream == NULL) {
1275: if (fd != -1)
1.114 doug 1276: close(fd);
1.47 schwarze 1277: if (warnings)
1.190 schwarze 1278: say(mlink->file, "&fopen");
1.47 schwarze 1279: return;
1280: }
1.1 schwarze 1281:
1.162 schwarze 1282: line = NULL;
1283: linesz = 0;
1284:
1.190 schwarze 1285: /* Parse the section number from the header line. */
1286:
1287: while (getline(&line, &linesz, stream) != -1) {
1288: if (*line == '\n')
1289: continue;
1290: if ((sec = strchr(line, '(')) == NULL)
1291: break;
1292: if ((p = strchr(++sec, ')')) == NULL)
1293: break;
1294: free(mpage->sec);
1295: mpage->sec = mandoc_strndup(sec, p - sec);
1296: if (warnings && *mlink->dsec != '\0' &&
1297: strcasecmp(mpage->sec, mlink->dsec))
1298: say(mlink->file,
1299: "Section \"%s\" manual in %s directory",
1300: mpage->sec, mlink->dsec);
1301: break;
1302: }
1303:
1.47 schwarze 1304: /* Skip to first blank line. */
1.1 schwarze 1305:
1.190 schwarze 1306: while (line == NULL || *line != '\n')
1307: if (getline(&line, &linesz, stream) == -1)
1.47 schwarze 1308: break;
1.1 schwarze 1309:
1.47 schwarze 1310: /*
1311: * Assume the first line that is not indented
1312: * is the first section header. Skip to it.
1313: */
1.1 schwarze 1314:
1.162 schwarze 1315: while (getline(&line, &linesz, stream) != -1)
1316: if (*line != '\n' && *line != ' ')
1.47 schwarze 1317: break;
1.98 schwarze 1318:
1.47 schwarze 1319: /*
1320: * Read up until the next section into a buffer.
1321: * Strip the leading and trailing newline from each read line,
1322: * appending a trailing space.
1323: * Ignore empty (whitespace-only) lines.
1324: */
1.28 schwarze 1325:
1.47 schwarze 1326: titlesz = 0;
1327: title = NULL;
1.38 schwarze 1328:
1.162 schwarze 1329: while ((len = getline(&line, &linesz, stream)) != -1) {
1330: if (*line != ' ')
1.47 schwarze 1331: break;
1.162 schwarze 1332: offs = 0;
1333: while (isspace((unsigned char)line[offs]))
1334: offs++;
1335: if (line[offs] == '\0')
1.47 schwarze 1336: continue;
1.162 schwarze 1337: title = mandoc_realloc(title, titlesz + len - offs);
1338: memcpy(title + titlesz, line + offs, len - offs);
1339: titlesz += len - offs;
1.47 schwarze 1340: title[titlesz - 1] = ' ';
1341: }
1.162 schwarze 1342: free(line);
1.28 schwarze 1343:
1.47 schwarze 1344: /*
1345: * If no page content can be found, or the input line
1346: * is already the next section header, or there is no
1347: * trailing newline, reuse the page title as the page
1348: * description.
1349: */
1.1 schwarze 1350:
1.47 schwarze 1351: if (NULL == title || '\0' == *title) {
1352: if (warnings)
1.190 schwarze 1353: say(mlink->file, "Cannot find NAME section");
1.47 schwarze 1354: fclose(stream);
1355: free(title);
1356: return;
1357: }
1.24 schwarze 1358:
1.162 schwarze 1359: title[titlesz - 1] = '\0';
1.24 schwarze 1360:
1.47 schwarze 1361: /*
1362: * Skip to the first dash.
1363: * Use the remaining line as the description (no more than 70
1364: * bytes).
1365: */
1.28 schwarze 1366:
1.47 schwarze 1367: if (NULL != (p = strstr(title, "- "))) {
1368: for (p += 2; ' ' == *p || '\b' == *p; p++)
1369: /* Skip to next word. */ ;
1370: } else {
1371: if (warnings)
1.191 schwarze 1372: say(mlink->file, "No dash in title line, "
1373: "reusing \"%s\" as one-line description", title);
1.47 schwarze 1374: p = title;
1375: }
1.1 schwarze 1376:
1.47 schwarze 1377: plen = strlen(p);
1.1 schwarze 1378:
1.47 schwarze 1379: /* Strip backspace-encoding from line. */
1.1 schwarze 1380:
1.47 schwarze 1381: while (NULL != (line = memchr(p, '\b', plen))) {
1382: len = line - p;
1383: if (0 == len) {
1384: memmove(line, line + 1, plen--);
1385: continue;
1.98 schwarze 1386: }
1.47 schwarze 1387: memmove(line - 1, line + 1, plen - len);
1388: plen -= 2;
1389: }
1.1 schwarze 1390:
1.206 schwarze 1391: /*
1392: * Cut off excessive one-line descriptions.
1393: * Bad pages are not worth better heuristics.
1394: */
1395:
1396: mpage->desc = mandoc_strndup(p, 150);
1.47 schwarze 1397: fclose(stream);
1398: free(title);
1399: }
1.16 schwarze 1400:
1.47 schwarze 1401: /*
1402: * Put a type/word pair into the word database for this particular file.
1403: */
1404: static void
1.69 schwarze 1405: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1406: {
1407: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1408: }
1409:
1410: /*
1.47 schwarze 1411: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1412: */
1413: static void
1.47 schwarze 1414: putmdockey(const struct mpage *mpage,
1.184 schwarze 1415: const struct roff_node *n, uint64_t m, int taboo)
1.2 schwarze 1416: {
1.16 schwarze 1417:
1.47 schwarze 1418: for ( ; NULL != n; n = n->next) {
1.184 schwarze 1419: if (n->flags & taboo)
1420: continue;
1.47 schwarze 1421: if (NULL != n->child)
1.184 schwarze 1422: putmdockey(mpage, n->child, m, taboo);
1.141 schwarze 1423: if (n->type == ROFFT_TEXT)
1.47 schwarze 1424: putkey(mpage, n->string, m);
1425: }
1426: }
1.16 schwarze 1427:
1.47 schwarze 1428: static void
1.143 schwarze 1429: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1430: const struct roff_node *n)
1.47 schwarze 1431: {
1.142 schwarze 1432: const struct roff_node *head, *body;
1.78 schwarze 1433: char *start, *title;
1.47 schwarze 1434: char byte;
1.78 schwarze 1435: size_t sz;
1.16 schwarze 1436:
1.168 schwarze 1437: if (n == NULL)
1.47 schwarze 1438: return;
1.16 schwarze 1439:
1.47 schwarze 1440: /*
1441: * We're only searching for one thing: the first text child in
1442: * the BODY of a NAME section. Since we don't keep track of
1443: * sections in -man, run some hoops to find out whether we're in
1444: * the correct section or not.
1445: */
1.16 schwarze 1446:
1.141 schwarze 1447: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1448: body = n;
1.168 schwarze 1449: if ((head = body->parent->head) != NULL &&
1450: (head = head->child) != NULL &&
1451: head->next == NULL &&
1.141 schwarze 1452: head->type == ROFFT_TEXT &&
1.168 schwarze 1453: strcmp(head->string, "NAME") == 0 &&
1454: body->child != NULL) {
1.2 schwarze 1455:
1.47 schwarze 1456: /*
1457: * Suck the entire NAME section into memory.
1458: * Yes, we might run away.
1459: * But too many manuals have big, spread-out
1460: * NAME sections over many lines.
1461: */
1.2 schwarze 1462:
1.78 schwarze 1463: title = NULL;
1.147 schwarze 1464: deroff(&title, body);
1.47 schwarze 1465: if (NULL == title)
1466: return;
1.16 schwarze 1467:
1.98 schwarze 1468: /*
1.47 schwarze 1469: * Go through a special heuristic dance here.
1470: * Conventionally, one or more manual names are
1471: * comma-specified prior to a whitespace, then a
1472: * dash, then a description. Try to puzzle out
1473: * the name parts here.
1474: */
1.16 schwarze 1475:
1.78 schwarze 1476: start = title;
1.47 schwarze 1477: for ( ;; ) {
1478: sz = strcspn(start, " ,");
1479: if ('\0' == start[sz])
1480: break;
1.1 schwarze 1481:
1.47 schwarze 1482: byte = start[sz];
1483: start[sz] = '\0';
1.67 schwarze 1484:
1485: /*
1486: * Assume a stray trailing comma in the
1487: * name list if a name begins with a dash.
1488: */
1489:
1490: if ('-' == start[0] ||
1491: ('\\' == start[0] && '-' == start[1]))
1492: break;
1.1 schwarze 1493:
1.90 schwarze 1494: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1495: if ( ! (mpage->name_head_done ||
1496: strcasecmp(start, meta->title))) {
1497: putkey(mpage, start, NAME_HEAD);
1498: mpage->name_head_done = 1;
1499: }
1.1 schwarze 1500:
1.47 schwarze 1501: if (' ' == byte) {
1502: start += sz + 1;
1503: break;
1504: }
1.1 schwarze 1505:
1.47 schwarze 1506: assert(',' == byte);
1507: start += sz + 1;
1508: while (' ' == *start)
1509: start++;
1510: }
1.1 schwarze 1511:
1.78 schwarze 1512: if (start == title) {
1.90 schwarze 1513: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1514: if ( ! (mpage->name_head_done ||
1515: strcasecmp(start, meta->title))) {
1516: putkey(mpage, start, NAME_HEAD);
1517: mpage->name_head_done = 1;
1518: }
1.47 schwarze 1519: free(title);
1520: return;
1521: }
1.1 schwarze 1522:
1.47 schwarze 1523: while (isspace((unsigned char)*start))
1524: start++;
1.1 schwarze 1525:
1.47 schwarze 1526: if (0 == strncmp(start, "-", 1))
1527: start += 1;
1528: else if (0 == strncmp(start, "\\-\\-", 4))
1529: start += 4;
1530: else if (0 == strncmp(start, "\\-", 2))
1531: start += 2;
1532: else if (0 == strncmp(start, "\\(en", 4))
1533: start += 4;
1534: else if (0 == strncmp(start, "\\(em", 4))
1535: start += 4;
1.1 schwarze 1536:
1.47 schwarze 1537: while (' ' == *start)
1538: start++;
1.1 schwarze 1539:
1.206 schwarze 1540: /*
1541: * Cut off excessive one-line descriptions.
1542: * Bad pages are not worth better heuristics.
1543: */
1544:
1545: mpage->desc = mandoc_strndup(start, 150);
1.47 schwarze 1546: free(title);
1547: return;
1548: }
1549: }
1.1 schwarze 1550:
1.47 schwarze 1551: for (n = n->child; n; n = n->next) {
1552: if (NULL != mpage->desc)
1553: break;
1.129 schwarze 1554: parse_man(mpage, meta, n);
1.1 schwarze 1555: }
1556: }
1557:
1558: static void
1.143 schwarze 1559: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1560: const struct roff_node *n)
1.1 schwarze 1561: {
1.208 schwarze 1562: const struct mdoc_handler *handler;
1.1 schwarze 1563:
1.195 schwarze 1564: for (n = n->child; n != NULL; n = n->next) {
1.208 schwarze 1565: if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX)
1.184 schwarze 1566: continue;
1.196 schwarze 1567: assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1.208 schwarze 1568: handler = mdoc_handlers + (n->tok - MDOC_Dd);
1569: if (n->flags & handler->taboo)
1570: continue;
1571:
1.47 schwarze 1572: switch (n->type) {
1.141 schwarze 1573: case ROFFT_ELEM:
1574: case ROFFT_BLOCK:
1575: case ROFFT_HEAD:
1576: case ROFFT_BODY:
1577: case ROFFT_TAIL:
1.208 schwarze 1578: if (handler->fp != NULL &&
1579: (*handler->fp)(mpage, meta, n) == 0)
1.195 schwarze 1580: break;
1.208 schwarze 1581: if (handler->mask)
1.47 schwarze 1582: putmdockey(mpage, n->child,
1.208 schwarze 1583: handler->mask, handler->taboo);
1.47 schwarze 1584: break;
1585: default:
1586: continue;
1587: }
1588: if (NULL != n->child)
1.127 schwarze 1589: parse_mdoc(mpage, meta, n);
1.1 schwarze 1590: }
1591: }
1592:
1.19 schwarze 1593: static int
1.207 schwarze 1594: parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta,
1595: const struct roff_node *n)
1596: {
1597: uint64_t mask;
1598:
1599: mask = TYPE_Fa;
1600: if (n->sec == SEC_SYNOPSIS)
1601: mask |= TYPE_Vt;
1602:
1603: putmdockey(mpage, n->child, mask, 0);
1604: return 0;
1605: }
1606:
1607: static int
1.143 schwarze 1608: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1609: const struct roff_node *n)
1.1 schwarze 1610: {
1.131 schwarze 1611: char *start, *end;
1.1 schwarze 1612: size_t sz;
1.19 schwarze 1613:
1.47 schwarze 1614: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1615: NULL == (n = n->child) ||
1.141 schwarze 1616: n->type != ROFFT_TEXT)
1.150 schwarze 1617: return 0;
1.1 schwarze 1618:
1619: /*
1620: * Only consider those `Fd' macro fields that begin with an
1621: * "inclusion" token (versus, e.g., #define).
1622: */
1.47 schwarze 1623:
1.1 schwarze 1624: if (strcmp("#include", n->string))
1.150 schwarze 1625: return 0;
1.1 schwarze 1626:
1.141 schwarze 1627: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1628: return 0;
1.1 schwarze 1629:
1630: /*
1631: * Strip away the enclosing angle brackets and make sure we're
1632: * not zero-length.
1633: */
1634:
1635: start = n->string;
1636: if ('<' == *start || '"' == *start)
1637: start++;
1638:
1639: if (0 == (sz = strlen(start)))
1.150 schwarze 1640: return 0;
1.1 schwarze 1641:
1642: end = &start[(int)sz - 1];
1643: if ('>' == *end || '"' == *end)
1644: end--;
1645:
1.47 schwarze 1646: if (end > start)
1647: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1648: return 0;
1.1 schwarze 1649: }
1650:
1.133 schwarze 1651: static void
1.142 schwarze 1652: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1653: {
1.69 schwarze 1654: char *cp;
1.133 schwarze 1655: size_t sz;
1.1 schwarze 1656:
1.141 schwarze 1657: if (n->type != ROFFT_TEXT)
1.133 schwarze 1658: return;
1.19 schwarze 1659:
1.133 schwarze 1660: /* Skip function pointer punctuation. */
1.1 schwarze 1661:
1.133 schwarze 1662: cp = n->string;
1663: while (*cp == '(' || *cp == '*')
1.1 schwarze 1664: cp++;
1.133 schwarze 1665: sz = strcspn(cp, "()");
1.1 schwarze 1666:
1.133 schwarze 1667: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1668: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1669: putkeys(mpage, cp, sz, NAME_SYN);
1670: }
1.19 schwarze 1671:
1.133 schwarze 1672: static int
1.143 schwarze 1673: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1674: const struct roff_node *n)
1.133 schwarze 1675: {
1.207 schwarze 1676: uint64_t mask;
1.133 schwarze 1677:
1678: if (n->child == NULL)
1.150 schwarze 1679: return 0;
1.133 schwarze 1680:
1681: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1682:
1.207 schwarze 1683: n = n->child->next;
1684: if (n != NULL && n->type == ROFFT_TEXT) {
1685: mask = TYPE_Fa;
1686: if (n->sec == SEC_SYNOPSIS)
1687: mask |= TYPE_Vt;
1688: putmdockey(mpage, n, mask, 0);
1689: }
1.19 schwarze 1690:
1.150 schwarze 1691: return 0;
1.128 schwarze 1692: }
1693:
1694: static int
1.143 schwarze 1695: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1696: const struct roff_node *n)
1.128 schwarze 1697: {
1.132 schwarze 1698:
1.141 schwarze 1699: if (n->type != ROFFT_HEAD)
1.150 schwarze 1700: return 1;
1.128 schwarze 1701:
1.133 schwarze 1702: if (n->child != NULL)
1703: parse_mdoc_fname(mpage, n->child);
1704:
1.150 schwarze 1705: return 0;
1.1 schwarze 1706: }
1707:
1.19 schwarze 1708: static int
1.164 schwarze 1709: parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1710: const struct roff_node *n)
1711: {
1712: char *cp;
1713:
1714: if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1715: return 0;
1716:
1.168 schwarze 1717: if (n->child != NULL &&
1718: n->child->next == NULL &&
1719: n->child->type == ROFFT_TEXT)
1.164 schwarze 1720: return 1;
1721:
1722: cp = NULL;
1723: deroff(&cp, n);
1724: if (cp != NULL) {
1725: putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1726: n->type == ROFFT_BODY ? TYPE_Va : 0));
1727: free(cp);
1728: }
1729:
1730: return 0;
1731: }
1732:
1733: static int
1.143 schwarze 1734: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1735: const struct roff_node *n)
1.1 schwarze 1736: {
1.47 schwarze 1737: char *cp;
1.1 schwarze 1738:
1739: if (NULL == (n = n->child))
1.150 schwarze 1740: return 0;
1.1 schwarze 1741:
1.47 schwarze 1742: if (NULL == n->next) {
1743: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1744: return 0;
1.47 schwarze 1745: }
1.1 schwarze 1746:
1.77 schwarze 1747: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1748: putkey(mpage, cp, TYPE_Xr);
1749: free(cp);
1.150 schwarze 1750: return 0;
1.1 schwarze 1751: }
1752:
1.19 schwarze 1753: static int
1.143 schwarze 1754: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1755: const struct roff_node *n)
1.1 schwarze 1756: {
1757:
1.141 schwarze 1758: if (n->type == ROFFT_BODY)
1.147 schwarze 1759: deroff(&mpage->desc, n);
1.150 schwarze 1760: return 0;
1.1 schwarze 1761: }
1762:
1.19 schwarze 1763: static int
1.143 schwarze 1764: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1765: const struct roff_node *n)
1.1 schwarze 1766: {
1767:
1.86 schwarze 1768: if (SEC_NAME == n->sec)
1.184 schwarze 1769: putmdockey(mpage, n->child, NAME_TITLE, 0);
1.141 schwarze 1770: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1771: if (n->child == NULL)
1772: putkey(mpage, meta->name, NAME_SYN);
1773: else
1.184 schwarze 1774: putmdockey(mpage, n->child, NAME_SYN, 0);
1.129 schwarze 1775: }
1776: if ( ! (mpage->name_head_done ||
1777: n->child == NULL || n->child->string == NULL ||
1778: strcasecmp(n->child->string, meta->title))) {
1.175 schwarze 1779: putkey(mpage, n->child->string, NAME_HEAD);
1.129 schwarze 1780: mpage->name_head_done = 1;
1.127 schwarze 1781: }
1.150 schwarze 1782: return 0;
1.1 schwarze 1783: }
1784:
1.19 schwarze 1785: static int
1.143 schwarze 1786: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1787: const struct roff_node *n)
1.1 schwarze 1788: {
1789:
1.150 schwarze 1790: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1791: }
1792:
1.47 schwarze 1793: static int
1.143 schwarze 1794: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1795: const struct roff_node *n)
1.1 schwarze 1796: {
1797:
1.150 schwarze 1798: return n->type == ROFFT_HEAD;
1.1 schwarze 1799: }
1800:
1.47 schwarze 1801: /*
1802: * Add a string to the hash table for the current manual.
1803: * Each string has a bitmask telling which macros it belongs to.
1804: * When we finish the manual, we'll dump the table.
1805: */
1.1 schwarze 1806: static void
1.131 schwarze 1807: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1808: {
1.90 schwarze 1809: struct ohash *htab;
1.47 schwarze 1810: struct str *s;
1.68 schwarze 1811: const char *end;
1.47 schwarze 1812: unsigned int slot;
1.131 schwarze 1813: int i, mustfree;
1.1 schwarze 1814:
1.47 schwarze 1815: if (0 == sz)
1816: return;
1.68 schwarze 1817:
1.131 schwarze 1818: mustfree = render_string(&cp, &sz);
1819:
1.90 schwarze 1820: if (TYPE_Nm & v) {
1821: htab = &names;
1822: v &= name_mask;
1.124 schwarze 1823: if (v & NAME_FIRST)
1824: name_mask &= ~NAME_FIRST;
1.90 schwarze 1825: if (debug > 1)
1826: say(mpage->mlinks->file,
1.183 schwarze 1827: "Adding name %*s, bits=0x%llx", (int)sz, cp,
1828: (unsigned long long)v);
1.90 schwarze 1829: } else {
1830: htab = &strings;
1831: if (debug > 1)
1.173 schwarze 1832: for (i = 0; i < KEY_MAX; i++)
1.118 schwarze 1833: if ((uint64_t)1 << i & v)
1.90 schwarze 1834: say(mpage->mlinks->file,
1835: "Adding key %s=%*s",
1.172 schwarze 1836: mansearch_keynames[i], (int)sz, cp);
1.68 schwarze 1837: }
1.47 schwarze 1838:
1839: end = cp + sz;
1.90 schwarze 1840: slot = ohash_qlookupi(htab, cp, &end);
1841: s = ohash_find(htab, slot);
1.1 schwarze 1842:
1.47 schwarze 1843: if (NULL != s && mpage == s->mpage) {
1844: s->mask |= v;
1.1 schwarze 1845: return;
1.47 schwarze 1846: } else if (NULL == s) {
1.101 schwarze 1847: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1848: memcpy(s->key, cp, sz);
1.90 schwarze 1849: ohash_insert(htab, slot, s);
1.47 schwarze 1850: }
1851: s->mpage = mpage;
1852: s->mask = v;
1.131 schwarze 1853:
1854: if (mustfree)
1855: free(cp);
1.1 schwarze 1856: }
1857:
1858: /*
1.47 schwarze 1859: * Take a Unicode codepoint and produce its UTF-8 encoding.
1860: * This isn't the best way to do this, but it works.
1861: * The magic numbers are from the UTF-8 packaging.
1862: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1863: */
1.47 schwarze 1864: static size_t
1865: utf8(unsigned int cp, char out[7])
1.1 schwarze 1866: {
1.47 schwarze 1867: size_t rc;
1.1 schwarze 1868:
1.47 schwarze 1869: rc = 0;
1870: if (cp <= 0x0000007F) {
1871: rc = 1;
1872: out[0] = (char)cp;
1873: } else if (cp <= 0x000007FF) {
1874: rc = 2;
1875: out[0] = (cp >> 6 & 31) | 192;
1876: out[1] = (cp & 63) | 128;
1877: } else if (cp <= 0x0000FFFF) {
1878: rc = 3;
1879: out[0] = (cp >> 12 & 15) | 224;
1880: out[1] = (cp >> 6 & 63) | 128;
1881: out[2] = (cp & 63) | 128;
1882: } else if (cp <= 0x001FFFFF) {
1883: rc = 4;
1884: out[0] = (cp >> 18 & 7) | 240;
1885: out[1] = (cp >> 12 & 63) | 128;
1886: out[2] = (cp >> 6 & 63) | 128;
1887: out[3] = (cp & 63) | 128;
1888: } else if (cp <= 0x03FFFFFF) {
1889: rc = 5;
1890: out[0] = (cp >> 24 & 3) | 248;
1891: out[1] = (cp >> 18 & 63) | 128;
1892: out[2] = (cp >> 12 & 63) | 128;
1893: out[3] = (cp >> 6 & 63) | 128;
1894: out[4] = (cp & 63) | 128;
1895: } else if (cp <= 0x7FFFFFFF) {
1896: rc = 6;
1897: out[0] = (cp >> 30 & 1) | 252;
1898: out[1] = (cp >> 24 & 63) | 128;
1899: out[2] = (cp >> 18 & 63) | 128;
1900: out[3] = (cp >> 12 & 63) | 128;
1901: out[4] = (cp >> 6 & 63) | 128;
1902: out[5] = (cp & 63) | 128;
1903: } else
1.150 schwarze 1904: return 0;
1.19 schwarze 1905:
1.47 schwarze 1906: out[rc] = '\0';
1.150 schwarze 1907: return rc;
1.1 schwarze 1908: }
1909:
1.47 schwarze 1910: /*
1.131 schwarze 1911: * If the string contains escape sequences,
1912: * replace it with an allocated rendering and return 1,
1913: * such that the caller can free it after use.
1914: * Otherwise, do nothing and return 0.
1.47 schwarze 1915: */
1.131 schwarze 1916: static int
1917: render_string(char **public, size_t *psz)
1.1 schwarze 1918: {
1.131 schwarze 1919: const char *src, *scp, *addcp, *seq;
1920: char *dst;
1921: size_t ssz, dsz, addsz;
1.71 schwarze 1922: char utfbuf[7], res[6];
1.131 schwarze 1923: int seqlen, unicode;
1.47 schwarze 1924:
1925: res[0] = '\\';
1926: res[1] = '\t';
1927: res[2] = ASCII_NBRSP;
1928: res[3] = ASCII_HYPH;
1.71 schwarze 1929: res[4] = ASCII_BREAK;
1930: res[5] = '\0';
1.1 schwarze 1931:
1.131 schwarze 1932: src = scp = *public;
1933: ssz = *psz;
1934: dst = NULL;
1935: dsz = 0;
1936:
1937: while (scp < src + *psz) {
1938:
1939: /* Leave normal characters unchanged. */
1940:
1941: if (strchr(res, *scp) == NULL) {
1942: if (dst != NULL)
1943: dst[dsz++] = *scp;
1944: scp++;
1945: continue;
1946: }
1.1 schwarze 1947:
1.47 schwarze 1948: /*
1.131 schwarze 1949: * Found something that requires replacing,
1950: * make sure we have a destination buffer.
1.47 schwarze 1951: */
1.131 schwarze 1952:
1953: if (dst == NULL) {
1954: dst = mandoc_malloc(ssz + 1);
1955: dsz = scp - src;
1956: memcpy(dst, src, dsz);
1.47 schwarze 1957: }
1.39 schwarze 1958:
1.131 schwarze 1959: /* Handle single-char special characters. */
1960:
1961: switch (*scp) {
1962: case '\\':
1963: break;
1.98 schwarze 1964: case '\t':
1965: case ASCII_NBRSP:
1.131 schwarze 1966: dst[dsz++] = ' ';
1967: scp++;
1968: continue;
1969: case ASCII_HYPH:
1970: dst[dsz++] = '-';
1.71 schwarze 1971: /* FALLTHROUGH */
1.98 schwarze 1972: case ASCII_BREAK:
1.131 schwarze 1973: scp++;
1.47 schwarze 1974: continue;
1.71 schwarze 1975: default:
1.131 schwarze 1976: abort();
1.71 schwarze 1977: }
1.39 schwarze 1978:
1.47 schwarze 1979: /*
1.131 schwarze 1980: * Found an escape sequence.
1981: * Read past the slash, then parse it.
1982: * Ignore everything except characters.
1.47 schwarze 1983: */
1.52 schwarze 1984:
1.131 schwarze 1985: scp++;
1986: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1987: continue;
1.39 schwarze 1988:
1.47 schwarze 1989: /*
1.52 schwarze 1990: * Render the special character
1991: * as either UTF-8 or ASCII.
1.47 schwarze 1992: */
1.52 schwarze 1993:
1994: if (write_utf8) {
1.157 schwarze 1995: unicode = mchars_spec2cp(seq, seqlen);
1.131 schwarze 1996: if (unicode <= 0)
1.52 schwarze 1997: continue;
1.131 schwarze 1998: addsz = utf8(unicode, utfbuf);
1999: if (addsz == 0)
1.52 schwarze 2000: continue;
1.131 schwarze 2001: addcp = utfbuf;
1.52 schwarze 2002: } else {
1.157 schwarze 2003: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.131 schwarze 2004: if (addcp == NULL)
1.52 schwarze 2005: continue;
1.131 schwarze 2006: if (*addcp == ASCII_NBRSP) {
2007: addcp = " ";
2008: addsz = 1;
1.52 schwarze 2009: }
2010: }
1.1 schwarze 2011:
1.47 schwarze 2012: /* Copy the rendered glyph into the stream. */
1.1 schwarze 2013:
1.131 schwarze 2014: ssz += addsz;
2015: dst = mandoc_realloc(dst, ssz + 1);
2016: memcpy(dst + dsz, addcp, addsz);
2017: dsz += addsz;
2018: }
2019: if (dst != NULL) {
2020: *public = dst;
2021: *psz = dsz;
2022: }
2023:
2024: /* Trim trailing whitespace and NUL-terminate. */
2025:
2026: while (*psz > 0 && (*public)[*psz - 1] == ' ')
2027: --*psz;
2028: if (dst != NULL) {
2029: (*public)[*psz] = '\0';
1.150 schwarze 2030: return 1;
1.131 schwarze 2031: } else
1.150 schwarze 2032: return 0;
1.1 schwarze 2033: }
2034:
1.75 schwarze 2035: static void
2036: dbadd_mlink(const struct mlink *mlink)
2037: {
1.173 schwarze 2038: dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
2039: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
2040: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
2041: dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
2042: dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
1.75 schwarze 2043: }
2044:
1.11 schwarze 2045: /*
1.47 schwarze 2046: * Flush the current page's terms (and their bits) into the database.
1.53 schwarze 2047: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 2048: */
2049: static void
1.173 schwarze 2050: dbadd(struct dba *dba, struct mpage *mpage)
1.11 schwarze 2051: {
1.47 schwarze 2052: struct mlink *mlink;
2053: struct str *key;
1.131 schwarze 2054: char *cp;
1.173 schwarze 2055: uint64_t mask;
1.47 schwarze 2056: size_t i;
2057: unsigned int slot;
1.131 schwarze 2058: int mustfree;
1.47 schwarze 2059:
1.85 schwarze 2060: mlink = mpage->mlinks;
1.11 schwarze 2061:
1.85 schwarze 2062: if (nodb) {
1.104 schwarze 2063: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2064: key = ohash_next(&names, &slot))
1.104 schwarze 2065: free(key);
2066: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2067: key = ohash_next(&strings, &slot))
1.104 schwarze 2068: free(key);
1.102 schwarze 2069: if (0 == debug)
2070: return;
1.85 schwarze 2071: while (NULL != mlink) {
2072: fputs(mlink->name, stdout);
2073: if (NULL == mlink->next ||
2074: strcmp(mlink->dsec, mlink->next->dsec) ||
2075: strcmp(mlink->fsec, mlink->next->fsec) ||
2076: strcmp(mlink->arch, mlink->next->arch)) {
2077: putchar('(');
2078: if ('\0' == *mlink->dsec)
2079: fputs(mlink->fsec, stdout);
2080: else
2081: fputs(mlink->dsec, stdout);
2082: if ('\0' != *mlink->arch)
2083: printf("/%s", mlink->arch);
2084: putchar(')');
2085: }
2086: mlink = mlink->next;
2087: if (NULL != mlink)
2088: fputs(", ", stdout);
2089: }
1.89 schwarze 2090: printf(" - %s\n", mpage->desc);
1.11 schwarze 2091: return;
1.85 schwarze 2092: }
2093:
2094: if (debug)
2095: say(mlink->file, "Adding to database");
1.47 schwarze 2096:
1.131 schwarze 2097: cp = mpage->desc;
2098: i = strlen(cp);
2099: mustfree = render_string(&cp, &i);
1.176 schwarze 2100: mpage->dba = dba_page_new(dba->pages,
1.173 schwarze 2101: *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2102: cp, mlink->file, mpage->form);
1.131 schwarze 2103: if (mustfree)
2104: free(cp);
1.176 schwarze 2105: dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
1.47 schwarze 2106:
1.173 schwarze 2107: while (mlink != NULL) {
1.75 schwarze 2108: dbadd_mlink(mlink);
1.85 schwarze 2109: mlink = mlink->next;
2110: }
1.47 schwarze 2111:
1.90 schwarze 2112: for (key = ohash_first(&names, &slot); NULL != key;
2113: key = ohash_next(&names, &slot)) {
2114: assert(key->mpage == mpage);
1.173 schwarze 2115: dba_page_alias(mpage->dba, key->key, key->mask);
1.90 schwarze 2116: free(key);
2117: }
1.47 schwarze 2118: for (key = ohash_first(&strings, &slot); NULL != key;
2119: key = ohash_next(&strings, &slot)) {
2120: assert(key->mpage == mpage);
1.173 schwarze 2121: i = 0;
2122: for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2123: if (key->mask & mask)
2124: dba_macro_add(dba->macros, i,
2125: key->key, mpage->dba);
2126: i++;
2127: }
1.47 schwarze 2128: free(key);
1.33 schwarze 2129: }
1.47 schwarze 2130: }
1.41 deraadt 2131:
1.47 schwarze 2132: static void
1.173 schwarze 2133: dbprune(struct dba *dba)
1.47 schwarze 2134: {
1.173 schwarze 2135: struct dba_array *page, *files;
2136: char *file;
1.11 schwarze 2137:
1.173 schwarze 2138: dba_array_FOREACH(dba->pages, page) {
2139: files = dba_array_get(page, DBP_FILE);
2140: dba_array_FOREACH(files, file) {
2141: if (*file < ' ')
2142: file++;
2143: if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2144: file)) != NULL) {
2145: if (debug)
2146: say(file, "Deleting from database");
2147: dba_array_del(dba->pages);
2148: break;
2149: }
1.63 schwarze 2150: }
1.11 schwarze 2151: }
1.47 schwarze 2152: }
1.22 schwarze 2153:
1.47 schwarze 2154: /*
1.173 schwarze 2155: * Write the database from memory to disk.
1.47 schwarze 2156: */
2157: static void
1.173 schwarze 2158: dbwrite(struct dba *dba)
1.47 schwarze 2159: {
1.204 schwarze 2160: struct stat sb1, sb2;
2161: char tfn[33], *cp1, *cp2;
2162: off_t i;
2163: int fd1, fd2;
1.11 schwarze 2164:
1.199 schwarze 2165: /*
2166: * Do not write empty databases, and delete existing ones
2167: * when makewhatis -u causes them to become empty.
2168: */
2169:
2170: dba_array_start(dba->pages);
2171: if (dba_array_next(dba->pages) == NULL) {
1.201 schwarze 2172: if (unlink(MANDOC_DB) == -1 && errno != ENOENT)
1.199 schwarze 2173: say(MANDOC_DB, "&unlink");
2174: return;
2175: }
2176:
2177: /*
2178: * Build the database in a temporary file,
2179: * then atomically move it into place.
2180: */
2181:
1.173 schwarze 2182: if (dba_write(MANDOC_DB "~", dba) != -1) {
2183: if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2184: exitcode = (int)MANDOCLEVEL_SYSERR;
2185: say(MANDOC_DB, "&rename");
2186: unlink(MANDOC_DB "~");
2187: }
1.47 schwarze 2188: return;
1.28 schwarze 2189: }
1.199 schwarze 2190:
2191: /*
2192: * We lack write permission and cannot replace the database
2193: * file, but let's at least check whether the data changed.
2194: */
1.22 schwarze 2195:
1.173 schwarze 2196: (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2197: if (mkdtemp(tfn) == NULL) {
2198: exitcode = (int)MANDOCLEVEL_SYSERR;
2199: say("", "&%s", tfn);
1.47 schwarze 2200: return;
1.173 schwarze 2201: }
1.205 schwarze 2202: cp1 = cp2 = MAP_FAILED;
1.204 schwarze 2203: fd1 = fd2 = -1;
1.173 schwarze 2204: (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2205: if (dba_write(tfn, dba) == -1) {
2206: say(tfn, "&dba_write");
1.204 schwarze 2207: goto err;
1.72 schwarze 2208: }
1.204 schwarze 2209: if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) {
2210: say(MANDOC_DB, "&open");
2211: goto err;
2212: }
2213: if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) {
2214: say(tfn, "&open");
2215: goto err;
2216: }
2217: if (fstat(fd1, &sb1) == -1) {
2218: say(MANDOC_DB, "&fstat");
2219: goto err;
2220: }
2221: if (fstat(fd2, &sb2) == -1) {
2222: say(tfn, "&fstat");
2223: goto err;
2224: }
2225: if (sb1.st_size != sb2.st_size)
2226: goto err;
2227: if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE,
1.205 schwarze 2228: fd1, 0)) == MAP_FAILED) {
1.204 schwarze 2229: say(MANDOC_DB, "&mmap");
2230: goto err;
2231: }
2232: if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE,
1.205 schwarze 2233: fd2, 0)) == MAP_FAILED) {
1.204 schwarze 2234: say(tfn, "&mmap");
2235: goto err;
2236: }
2237: for (i = 0; i < sb1.st_size; i++)
2238: if (cp1[i] != cp2[i])
2239: goto err;
2240: goto out;
2241:
2242: err:
2243: exitcode = (int)MANDOCLEVEL_SYSERR;
2244: say(MANDOC_DB, "Data changed, but cannot replace database");
1.72 schwarze 2245:
1.173 schwarze 2246: out:
1.205 schwarze 2247: if (cp1 != MAP_FAILED)
1.204 schwarze 2248: munmap(cp1, sb1.st_size);
1.205 schwarze 2249: if (cp2 != MAP_FAILED)
1.204 schwarze 2250: munmap(cp2, sb2.st_size);
2251: if (fd1 != -1)
2252: close(fd1);
2253: if (fd2 != -1)
2254: close(fd2);
1.203 schwarze 2255: unlink(tfn);
1.173 schwarze 2256: *strrchr(tfn, '/') = '\0';
1.203 schwarze 2257: rmdir(tfn);
1.47 schwarze 2258: }
1.6 schwarze 2259:
1.47 schwarze 2260: static int
1.120 schwarze 2261: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2262: {
2263: static char startdir[PATH_MAX];
1.110 schwarze 2264: static int getcwd_status; /* 1 = ok, 2 = failure */
2265: static int chdir_status; /* 1 = changed directory */
1.6 schwarze 2266:
1.47 schwarze 2267: /*
1.110 schwarze 2268: * Remember the original working directory, if possible.
2269: * This will be needed if the second or a later directory
2270: * on the command line is given as a relative path.
2271: * Do not error out if the current directory is not
2272: * searchable: Maybe it won't be needed after all.
2273: */
1.213 schwarze 2274: if (getcwd_status == 0) {
2275: if (getcwd(startdir, sizeof(startdir)) == NULL) {
1.110 schwarze 2276: getcwd_status = 2;
2277: (void)strlcpy(startdir, strerror(errno),
2278: sizeof(startdir));
2279: } else
2280: getcwd_status = 1;
2281: }
2282:
2283: /*
2284: * We are leaving the old base directory.
2285: * Do not use it any longer, not even for messages.
2286: */
2287: *basedir = '\0';
1.213 schwarze 2288: basedir_len = 0;
1.110 schwarze 2289:
2290: /*
2291: * If and only if the directory was changed earlier and
2292: * the next directory to process is given as a relative path,
2293: * first go back, or bail out if that is impossible.
1.47 schwarze 2294: */
1.213 schwarze 2295: if (chdir_status && *targetdir != '/') {
2296: if (getcwd_status == 2) {
1.47 schwarze 2297: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2298: say("", "getcwd: %s", startdir);
1.150 schwarze 2299: return 0;
1.47 schwarze 2300: }
1.213 schwarze 2301: if (chdir(startdir) == -1) {
1.47 schwarze 2302: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2303: say("", "&chdir %s", startdir);
1.150 schwarze 2304: return 0;
1.2 schwarze 2305: }
2306: }
1.110 schwarze 2307:
2308: /*
2309: * Always resolve basedir to the canonicalized absolute
2310: * pathname and append a trailing slash, such that
2311: * we can reliably check whether files are inside.
2312: */
1.213 schwarze 2313: if (realpath(targetdir, basedir) == NULL) {
1.120 schwarze 2314: if (report_baddir || errno != ENOENT) {
2315: exitcode = (int)MANDOCLEVEL_BADARG;
2316: say("", "&%s: realpath", targetdir);
2317: }
1.213 schwarze 2318: *basedir = '\0';
1.150 schwarze 2319: return 0;
1.213 schwarze 2320: } else if (chdir(basedir) == -1) {
1.120 schwarze 2321: if (report_baddir || errno != ENOENT) {
2322: exitcode = (int)MANDOCLEVEL_BADARG;
2323: say("", "&chdir");
2324: }
1.213 schwarze 2325: *basedir = '\0';
1.150 schwarze 2326: return 0;
1.109 schwarze 2327: }
1.110 schwarze 2328: chdir_status = 1;
1.213 schwarze 2329: basedir_len = strlen(basedir);
2330: if (basedir[basedir_len - 1] != '/') {
2331: if (basedir_len >= PATH_MAX - 1) {
1.109 schwarze 2332: exitcode = (int)MANDOCLEVEL_SYSERR;
2333: say("", "Filename too long");
1.213 schwarze 2334: *basedir = '\0';
2335: basedir_len = 0;
1.150 schwarze 2336: return 0;
1.109 schwarze 2337: }
1.213 schwarze 2338: basedir[basedir_len++] = '/';
2339: basedir[basedir_len] = '\0';
1.47 schwarze 2340: }
1.150 schwarze 2341: return 1;
1.2 schwarze 2342: }
2343:
2344: static void
1.47 schwarze 2345: say(const char *file, const char *format, ...)
1.2 schwarze 2346: {
1.47 schwarze 2347: va_list ap;
1.80 schwarze 2348: int use_errno;
1.2 schwarze 2349:
1.213 schwarze 2350: if (*basedir != '\0')
1.47 schwarze 2351: fprintf(stderr, "%s", basedir);
1.213 schwarze 2352: if (*basedir != '\0' && *file != '\0')
1.110 schwarze 2353: fputc('/', stderr);
1.213 schwarze 2354: if (*file != '\0')
1.47 schwarze 2355: fprintf(stderr, "%s", file);
1.31 schwarze 2356:
1.80 schwarze 2357: use_errno = 1;
1.213 schwarze 2358: if (format != NULL) {
1.80 schwarze 2359: switch (*format) {
1.98 schwarze 2360: case '&':
1.80 schwarze 2361: format++;
2362: break;
1.98 schwarze 2363: case '\0':
1.80 schwarze 2364: format = NULL;
2365: break;
2366: default:
2367: use_errno = 0;
2368: break;
2369: }
2370: }
1.213 schwarze 2371: if (format != NULL) {
2372: if (*basedir != '\0' || *file != '\0')
1.80 schwarze 2373: fputs(": ", stderr);
2374: va_start(ap, format);
2375: vfprintf(stderr, format, ap);
2376: va_end(ap);
2377: }
2378: if (use_errno) {
1.213 schwarze 2379: if (*basedir != '\0' || *file != '\0' || format != NULL)
1.80 schwarze 2380: fputs(": ", stderr);
1.47 schwarze 2381: perror(NULL);
1.80 schwarze 2382: } else
2383: fputc('\n', stderr);
1.1 schwarze 2384: }