Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.213
1.213 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.212 2019/05/03 18:16:57 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
747: filescan(const char *file)
748: {
749: char buf[PATH_MAX];
750: struct stat st;
751: struct mlink *mlink;
752: char *p, *start;
753:
754: assert(use_all);
755:
1.213 ! schwarze 756: if (strncmp(file, "./", 2) == 0)
1.47 schwarze 757: file += 2;
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: */
1.213 ! schwarze 766: if (lstat(file, &st) == -1) {
1.96 schwarze 767: exitcode = (int)MANDOCLEVEL_BADARG;
768: say(file, "&lstat");
769: return;
1.213 ! schwarze 770: } else if ((st.st_mode & (S_IFREG | S_IFLNK)) == 0) {
1.96 schwarze 771: exitcode = (int)MANDOCLEVEL_BADARG;
772: say(file, "Not a regular file");
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: */
1.213 ! schwarze 780: if (realpath(file, buf) == NULL) {
1.47 schwarze 781: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 782: say(file, "&realpath");
1.47 schwarze 783: return;
1.63 schwarze 784: }
785:
1.213 ! schwarze 786: if (op == OP_TEST)
1.63 schwarze 787: start = buf;
1.213 ! schwarze 788: else if (strncmp(buf, basedir, basedir_len) == 0)
! 789: start = buf + basedir_len;
1.63 schwarze 790: else {
1.47 schwarze 791: exitcode = (int)MANDOCLEVEL_BADARG;
792: say("", "%s: outside base directory", buf);
793: return;
1.63 schwarze 794: }
795:
1.96 schwarze 796: /*
797: * Now we are sure the file is inside our tree.
798: * If it is a symbolic link, ignore the real path
799: * and use the original name.
800: * This implies passing stuff like "cat1/../man1/foo.1"
801: * on the command line won't work. So don't do that.
802: * Note the stat(2) can still fail if the link target
803: * doesn't exist.
804: */
1.213 ! schwarze 805: if (st.st_mode & S_IFLNK) {
! 806: if (stat(buf, &st) == -1) {
1.96 schwarze 807: exitcode = (int)MANDOCLEVEL_BADARG;
808: say(file, "&stat");
809: return;
810: }
1.100 schwarze 811: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
812: say(file, "Filename too long");
813: return;
814: }
1.109 schwarze 815: start = buf;
1.213 ! schwarze 816: if (op != OP_TEST && strncmp(buf, basedir, basedir_len) == 0)
! 817: start += basedir_len;
1.1 schwarze 818: }
1.63 schwarze 819:
1.47 schwarze 820: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 821: mlink->dform = FORM_NONE;
1.100 schwarze 822: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
823: sizeof(mlink->file)) {
824: say(start, "Filename too long");
1.134 schwarze 825: free(mlink);
1.100 schwarze 826: return;
1.188 schwarze 827: }
828:
829: /*
830: * In test mode or when the original name is absolute
831: * but outside our tree, guess the base directory.
832: */
833:
834: if (op == OP_TEST || (start == buf && *start == '/')) {
835: if (strncmp(buf, "man/", 4) == 0)
836: start = buf + 4;
837: else if ((start = strstr(buf, "/man/")) != NULL)
838: start += 5;
839: else
840: start = buf;
1.100 schwarze 841: }
1.1 schwarze 842:
1.10 schwarze 843: /*
1.47 schwarze 844: * First try to guess our directory structure.
845: * If we find a separator, try to look for man* or cat*.
846: * If we find one of these and what's underneath is a directory,
847: * assume it's an architecture.
1.10 schwarze 848: */
1.213 ! schwarze 849: if ((p = strchr(start, '/')) != NULL) {
1.47 schwarze 850: *p++ = '\0';
1.213 ! schwarze 851: if (strncmp(start, "man", 3) == 0) {
1.47 schwarze 852: mlink->dform = FORM_SRC;
1.51 schwarze 853: mlink->dsec = start + 3;
1.213 ! schwarze 854: } else if (strncmp(start, "cat", 3) == 0) {
1.47 schwarze 855: mlink->dform = FORM_CAT;
1.51 schwarze 856: mlink->dsec = start + 3;
1.47 schwarze 857: }
1.10 schwarze 858:
1.47 schwarze 859: start = p;
1.213 ! schwarze 860: if (mlink->dsec != NULL && (p = strchr(start, '/')) != NULL) {
1.47 schwarze 861: *p++ = '\0';
1.51 schwarze 862: mlink->arch = start;
1.47 schwarze 863: start = p;
1.41 deraadt 864: }
1.47 schwarze 865: }
1.7 schwarze 866:
1.47 schwarze 867: /*
868: * Now check the file suffix.
869: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
870: */
871: p = strrchr(start, '\0');
1.213 ! schwarze 872: while (p-- > start && *p != '/' && *p != '.')
! 873: continue;
1.47 schwarze 874:
1.213 ! schwarze 875: if (*p == '.') {
1.47 schwarze 876: *p++ = '\0';
1.51 schwarze 877: mlink->fsec = p;
1.47 schwarze 878: }
1.41 deraadt 879:
1.47 schwarze 880: /*
881: * Now try to parse the name.
882: * Use the filename portion of the path.
883: */
884: mlink->name = start;
1.213 ! schwarze 885: if ((p = strrchr(start, '/')) != NULL) {
1.47 schwarze 886: mlink->name = p + 1;
887: *p = '\0';
888: }
889: mlink_add(mlink, &st);
890: }
1.2 schwarze 891:
1.47 schwarze 892: static void
893: mlink_add(struct mlink *mlink, const struct stat *st)
894: {
895: struct inodev inodev;
896: struct mpage *mpage;
897: unsigned int slot;
898:
899: assert(NULL != mlink->file);
900:
1.51 schwarze 901: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
902: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
903: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
904: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 905:
906: if ('0' == *mlink->fsec) {
907: free(mlink->fsec);
908: mlink->fsec = mandoc_strdup(mlink->dsec);
909: mlink->fform = FORM_CAT;
910: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
911: mlink->fform = FORM_SRC;
912: else
913: mlink->fform = FORM_NONE;
1.1 schwarze 914:
1.47 schwarze 915: slot = ohash_qlookup(&mlinks, mlink->file);
916: assert(NULL == ohash_find(&mlinks, slot));
917: ohash_insert(&mlinks, slot, mlink);
918:
1.139 schwarze 919: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.47 schwarze 920: inodev.st_ino = st->st_ino;
921: inodev.st_dev = st->st_dev;
922: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
923: sizeof(struct inodev), inodev.st_ino);
924: mpage = ohash_find(&mpages, slot);
925: if (NULL == mpage) {
926: mpage = mandoc_calloc(1, sizeof(struct mpage));
927: mpage->inodev.st_ino = inodev.st_ino;
928: mpage->inodev.st_dev = inodev.st_dev;
1.173 schwarze 929: mpage->form = FORM_NONE;
1.180 schwarze 930: mpage->next = mpage_head;
931: mpage_head = mpage;
1.47 schwarze 932: ohash_insert(&mpages, slot, mpage);
933: } else
934: mlink->next = mpage->mlinks;
935: mpage->mlinks = mlink;
1.75 schwarze 936: mlink->mpage = mpage;
1.47 schwarze 937: }
1.1 schwarze 938:
1.47 schwarze 939: static void
940: mlink_free(struct mlink *mlink)
941: {
1.1 schwarze 942:
1.47 schwarze 943: free(mlink->dsec);
944: free(mlink->arch);
945: free(mlink->name);
946: free(mlink->fsec);
947: free(mlink);
948: }
1.1 schwarze 949:
1.47 schwarze 950: static void
951: mpages_free(void)
952: {
953: struct mpage *mpage;
954: struct mlink *mlink;
955:
1.180 schwarze 956: while ((mpage = mpage_head) != NULL) {
957: while ((mlink = mpage->mlinks) != NULL) {
1.47 schwarze 958: mpage->mlinks = mlink->next;
959: mlink_free(mlink);
960: }
1.180 schwarze 961: mpage_head = mpage->next;
1.47 schwarze 962: free(mpage->sec);
963: free(mpage->arch);
964: free(mpage->title);
965: free(mpage->desc);
966: free(mpage);
967: }
968: }
1.1 schwarze 969:
1.47 schwarze 970: /*
971: * For each mlink to the mpage, check whether the path looks like
972: * it is formatted, and if it does, check whether a source manual
973: * exists by the same name, ignoring the suffix.
974: * If both conditions hold, drop the mlink.
975: */
976: static void
977: mlinks_undupe(struct mpage *mpage)
978: {
979: char buf[PATH_MAX];
980: struct mlink **prev;
981: struct mlink *mlink;
982: char *bufp;
983:
984: mpage->form = FORM_CAT;
985: prev = &mpage->mlinks;
986: while (NULL != (mlink = *prev)) {
987: if (FORM_CAT != mlink->dform) {
988: mpage->form = FORM_NONE;
989: goto nextlink;
990: }
1.100 schwarze 991: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 992: bufp = strstr(buf, "cat");
993: assert(NULL != bufp);
994: memcpy(bufp, "man", 3);
995: if (NULL != (bufp = strrchr(buf, '.')))
996: *++bufp = '\0';
1.100 schwarze 997: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 998: if (NULL == ohash_find(&mlinks,
1.98 schwarze 999: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 1000: goto nextlink;
1001: if (warnings)
1002: say(mlink->file, "Man source exists: %s", buf);
1003: if (use_all)
1004: goto nextlink;
1005: *prev = mlink->next;
1006: mlink_free(mlink);
1007: continue;
1008: nextlink:
1009: prev = &(*prev)->next;
1.1 schwarze 1010: }
1.47 schwarze 1011: }
1.1 schwarze 1012:
1.88 schwarze 1013: static void
1.50 schwarze 1014: mlink_check(struct mpage *mpage, struct mlink *mlink)
1015: {
1.88 schwarze 1016: struct str *str;
1017: unsigned int slot;
1.50 schwarze 1018:
1019: /*
1020: * Check whether the manual section given in a file
1021: * agrees with the directory where the file is located.
1022: * Some manuals have suffixes like (3p) on their
1023: * section number either inside the file or in the
1024: * directory name, some are linked into more than one
1025: * section, like encrypt(1) = makekey(8).
1026: */
1027:
1028: if (FORM_SRC == mpage->form &&
1.88 schwarze 1029: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1030: say(mlink->file, "Section \"%s\" manual in %s directory",
1031: mpage->sec, mlink->dsec);
1032:
1033: /*
1034: * Manual page directories exist for each kernel
1035: * architecture as returned by machine(1).
1036: * However, many manuals only depend on the
1037: * application architecture as returned by arch(1).
1038: * For example, some (2/ARM) manuals are shared
1039: * across the "armish" and "zaurus" kernel
1040: * architectures.
1041: * A few manuals are even shared across completely
1042: * different architectures, for example fdformat(1)
1.178 schwarze 1043: * on amd64, i386, and sparc64.
1.50 schwarze 1044: */
1045:
1.88 schwarze 1046: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1047: say(mlink->file, "Architecture \"%s\" manual in "
1048: "\"%s\" directory", mpage->arch, mlink->arch);
1049:
1.88 schwarze 1050: /*
1051: * XXX
1.90 schwarze 1052: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1053: */
1054:
1055: if (FORM_CAT == mpage->form)
1056: return;
1057:
1058: /*
1059: * Check whether this mlink
1060: * appears as a name in the NAME section.
1061: */
1.50 schwarze 1062:
1.90 schwarze 1063: slot = ohash_qlookup(&names, mlink->name);
1064: str = ohash_find(&names, slot);
1.88 schwarze 1065: assert(NULL != str);
1.90 schwarze 1066: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1067: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1068: }
1069:
1.47 schwarze 1070: /*
1071: * Run through the files in the global vector "mpages"
1072: * and add them to the database specified in "basedir".
1073: *
1074: * This handles the parsing scheme itself, using the cues of directory
1075: * and filename to determine whether the file is parsable or not.
1076: */
1077: static void
1.173 schwarze 1078: mpages_merge(struct dba *dba, struct mparse *mp)
1.47 schwarze 1079: {
1.180 schwarze 1080: struct mpage *mpage, *mpage_dest;
1.75 schwarze 1081: struct mlink *mlink, *mlink_dest;
1.211 schwarze 1082: struct roff_meta *meta;
1.69 schwarze 1083: char *cp;
1.117 schwarze 1084: int fd;
1.64 schwarze 1085:
1.180 schwarze 1086: for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
1.47 schwarze 1087: mlinks_undupe(mpage);
1.180 schwarze 1088: if ((mlink = mpage->mlinks) == NULL)
1089: continue;
1.1 schwarze 1090:
1.90 schwarze 1091: name_mask = NAME_MASK;
1.156 schwarze 1092: mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1093: mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1.47 schwarze 1094: mparse_reset(mp);
1.211 schwarze 1095: meta = NULL;
1.81 schwarze 1096:
1.166 schwarze 1097: if ((fd = mparse_open(mp, mlink->file)) == -1) {
1.136 schwarze 1098: say(mlink->file, "&open");
1.117 schwarze 1099: goto nextpage;
1.81 schwarze 1100: }
1.11 schwarze 1101:
1102: /*
1.137 schwarze 1103: * Interpret the file as mdoc(7) or man(7) source
1104: * code, unless it is known to be formatted.
1.11 schwarze 1105: */
1.136 schwarze 1106: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.137 schwarze 1107: mparse_readfd(mp, fd, mlink->file);
1.165 schwarze 1108: close(fd);
1.189 schwarze 1109: fd = -1;
1.211 schwarze 1110: meta = mparse_result(mp);
1.47 schwarze 1111: }
1.11 schwarze 1112:
1.211 schwarze 1113: if (meta != NULL && meta->sodest != NULL) {
1.75 schwarze 1114: mlink_dest = ohash_find(&mlinks,
1.211 schwarze 1115: ohash_qlookup(&mlinks, meta->sodest));
1.126 schwarze 1116: if (mlink_dest == NULL) {
1.211 schwarze 1117: mandoc_asprintf(&cp, "%s.gz", meta->sodest);
1.126 schwarze 1118: mlink_dest = ohash_find(&mlinks,
1119: ohash_qlookup(&mlinks, cp));
1120: free(cp);
1121: }
1122: if (mlink_dest != NULL) {
1.75 schwarze 1123:
1124: /* The .so target exists. */
1125:
1126: mpage_dest = mlink_dest->mpage;
1127: while (1) {
1128: mlink->mpage = mpage_dest;
1129:
1130: /*
1131: * If the target was already
1132: * processed, add the links
1133: * to the database now.
1134: * Otherwise, this will
1135: * happen when we come
1136: * to the target.
1137: */
1138:
1.173 schwarze 1139: if (mpage_dest->dba != NULL)
1140: dbadd_mlink(mlink);
1.75 schwarze 1141:
1.126 schwarze 1142: if (mlink->next == NULL)
1.75 schwarze 1143: break;
1144: mlink = mlink->next;
1145: }
1146:
1147: /* Move all links to the target. */
1148:
1149: mlink->next = mlink_dest->next;
1150: mlink_dest->next = mpage->mlinks;
1151: mpage->mlinks = NULL;
1.212 schwarze 1152: goto nextpage;
1.75 schwarze 1153: }
1.212 schwarze 1154: meta->macroset = MACROSET_NONE;
1155: }
1156: if (meta != NULL && meta->macroset == MACROSET_MDOC) {
1.47 schwarze 1157: mpage->form = FORM_SRC;
1.211 schwarze 1158: mpage->sec = meta->msec;
1.112 schwarze 1159: mpage->sec = mandoc_strdup(
1.126 schwarze 1160: mpage->sec == NULL ? "" : mpage->sec);
1.211 schwarze 1161: mpage->arch = meta->arch;
1.47 schwarze 1162: mpage->arch = mandoc_strdup(
1.126 schwarze 1163: mpage->arch == NULL ? "" : mpage->arch);
1.211 schwarze 1164: mpage->title = mandoc_strdup(meta->title);
1165: } else if (meta != NULL && meta->macroset == MACROSET_MAN) {
1166: if (*meta->msec != '\0' || *meta->title != '\0') {
1.189 schwarze 1167: mpage->form = FORM_SRC;
1.211 schwarze 1168: mpage->sec = mandoc_strdup(meta->msec);
1.189 schwarze 1169: mpage->arch = mandoc_strdup(mlink->arch);
1.211 schwarze 1170: mpage->title = mandoc_strdup(meta->title);
1.189 schwarze 1171: } else
1.211 schwarze 1172: meta = NULL;
1.189 schwarze 1173: }
1174:
1175: assert(mpage->desc == NULL);
1.212 schwarze 1176: if (meta == NULL || meta->sodest != NULL) {
1.136 schwarze 1177: mpage->sec = mandoc_strdup(mlink->dsec);
1178: mpage->arch = mandoc_strdup(mlink->arch);
1179: mpage->title = mandoc_strdup(mlink->name);
1.212 schwarze 1180: if (meta == NULL) {
1181: mpage->form = FORM_CAT;
1182: parse_cat(mpage, fd);
1183: } else
1184: mpage->form = FORM_SRC;
1.211 schwarze 1185: } else if (meta->macroset == MACROSET_MDOC)
1186: parse_mdoc(mpage, meta, meta->first);
1.189 schwarze 1187: else
1.211 schwarze 1188: parse_man(mpage, meta, meta->first);
1.191 schwarze 1189: if (mpage->desc == NULL) {
1190: mpage->desc = mandoc_strdup(mlink->name);
1191: if (warnings)
1192: say(mlink->file, "No one-line description, "
1193: "using filename \"%s\"", mlink->name);
1194: }
1.88 schwarze 1195:
1.187 schwarze 1196: for (mlink = mpage->mlinks;
1197: mlink != NULL;
1198: mlink = mlink->next) {
1199: putkey(mpage, mlink->name, NAME_FILE);
1200: if (warnings && !use_all)
1.88 schwarze 1201: mlink_check(mpage, mlink);
1.187 schwarze 1202: }
1.6 schwarze 1203:
1.173 schwarze 1204: dbadd(dba, mpage);
1.81 schwarze 1205:
1206: nextpage:
1.47 schwarze 1207: ohash_delete(&strings);
1.90 schwarze 1208: ohash_delete(&names);
1.47 schwarze 1209: }
1210: }
1.6 schwarze 1211:
1.47 schwarze 1212: static void
1.81 schwarze 1213: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1214: {
1215: FILE *stream;
1.190 schwarze 1216: struct mlink *mlink;
1217: char *line, *p, *title, *sec;
1.162 schwarze 1218: size_t linesz, plen, titlesz;
1219: ssize_t len;
1220: int offs;
1.1 schwarze 1221:
1.190 schwarze 1222: mlink = mpage->mlinks;
1223: stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1224: if (stream == NULL) {
1225: if (fd != -1)
1.114 doug 1226: close(fd);
1.47 schwarze 1227: if (warnings)
1.190 schwarze 1228: say(mlink->file, "&fopen");
1.47 schwarze 1229: return;
1230: }
1.1 schwarze 1231:
1.162 schwarze 1232: line = NULL;
1233: linesz = 0;
1234:
1.190 schwarze 1235: /* Parse the section number from the header line. */
1236:
1237: while (getline(&line, &linesz, stream) != -1) {
1238: if (*line == '\n')
1239: continue;
1240: if ((sec = strchr(line, '(')) == NULL)
1241: break;
1242: if ((p = strchr(++sec, ')')) == NULL)
1243: break;
1244: free(mpage->sec);
1245: mpage->sec = mandoc_strndup(sec, p - sec);
1246: if (warnings && *mlink->dsec != '\0' &&
1247: strcasecmp(mpage->sec, mlink->dsec))
1248: say(mlink->file,
1249: "Section \"%s\" manual in %s directory",
1250: mpage->sec, mlink->dsec);
1251: break;
1252: }
1253:
1.47 schwarze 1254: /* Skip to first blank line. */
1.1 schwarze 1255:
1.190 schwarze 1256: while (line == NULL || *line != '\n')
1257: if (getline(&line, &linesz, stream) == -1)
1.47 schwarze 1258: break;
1.1 schwarze 1259:
1.47 schwarze 1260: /*
1261: * Assume the first line that is not indented
1262: * is the first section header. Skip to it.
1263: */
1.1 schwarze 1264:
1.162 schwarze 1265: while (getline(&line, &linesz, stream) != -1)
1266: if (*line != '\n' && *line != ' ')
1.47 schwarze 1267: break;
1.98 schwarze 1268:
1.47 schwarze 1269: /*
1270: * Read up until the next section into a buffer.
1271: * Strip the leading and trailing newline from each read line,
1272: * appending a trailing space.
1273: * Ignore empty (whitespace-only) lines.
1274: */
1.28 schwarze 1275:
1.47 schwarze 1276: titlesz = 0;
1277: title = NULL;
1.38 schwarze 1278:
1.162 schwarze 1279: while ((len = getline(&line, &linesz, stream)) != -1) {
1280: if (*line != ' ')
1.47 schwarze 1281: break;
1.162 schwarze 1282: offs = 0;
1283: while (isspace((unsigned char)line[offs]))
1284: offs++;
1285: if (line[offs] == '\0')
1.47 schwarze 1286: continue;
1.162 schwarze 1287: title = mandoc_realloc(title, titlesz + len - offs);
1288: memcpy(title + titlesz, line + offs, len - offs);
1289: titlesz += len - offs;
1.47 schwarze 1290: title[titlesz - 1] = ' ';
1291: }
1.162 schwarze 1292: free(line);
1.28 schwarze 1293:
1.47 schwarze 1294: /*
1295: * If no page content can be found, or the input line
1296: * is already the next section header, or there is no
1297: * trailing newline, reuse the page title as the page
1298: * description.
1299: */
1.1 schwarze 1300:
1.47 schwarze 1301: if (NULL == title || '\0' == *title) {
1302: if (warnings)
1.190 schwarze 1303: say(mlink->file, "Cannot find NAME section");
1.47 schwarze 1304: fclose(stream);
1305: free(title);
1306: return;
1307: }
1.24 schwarze 1308:
1.162 schwarze 1309: title[titlesz - 1] = '\0';
1.24 schwarze 1310:
1.47 schwarze 1311: /*
1312: * Skip to the first dash.
1313: * Use the remaining line as the description (no more than 70
1314: * bytes).
1315: */
1.28 schwarze 1316:
1.47 schwarze 1317: if (NULL != (p = strstr(title, "- "))) {
1318: for (p += 2; ' ' == *p || '\b' == *p; p++)
1319: /* Skip to next word. */ ;
1320: } else {
1321: if (warnings)
1.191 schwarze 1322: say(mlink->file, "No dash in title line, "
1323: "reusing \"%s\" as one-line description", title);
1.47 schwarze 1324: p = title;
1325: }
1.1 schwarze 1326:
1.47 schwarze 1327: plen = strlen(p);
1.1 schwarze 1328:
1.47 schwarze 1329: /* Strip backspace-encoding from line. */
1.1 schwarze 1330:
1.47 schwarze 1331: while (NULL != (line = memchr(p, '\b', plen))) {
1332: len = line - p;
1333: if (0 == len) {
1334: memmove(line, line + 1, plen--);
1335: continue;
1.98 schwarze 1336: }
1.47 schwarze 1337: memmove(line - 1, line + 1, plen - len);
1338: plen -= 2;
1339: }
1.1 schwarze 1340:
1.206 schwarze 1341: /*
1342: * Cut off excessive one-line descriptions.
1343: * Bad pages are not worth better heuristics.
1344: */
1345:
1346: mpage->desc = mandoc_strndup(p, 150);
1.47 schwarze 1347: fclose(stream);
1348: free(title);
1349: }
1.16 schwarze 1350:
1.47 schwarze 1351: /*
1352: * Put a type/word pair into the word database for this particular file.
1353: */
1354: static void
1.69 schwarze 1355: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1356: {
1357: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1358: }
1359:
1360: /*
1.47 schwarze 1361: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1362: */
1363: static void
1.47 schwarze 1364: putmdockey(const struct mpage *mpage,
1.184 schwarze 1365: const struct roff_node *n, uint64_t m, int taboo)
1.2 schwarze 1366: {
1.16 schwarze 1367:
1.47 schwarze 1368: for ( ; NULL != n; n = n->next) {
1.184 schwarze 1369: if (n->flags & taboo)
1370: continue;
1.47 schwarze 1371: if (NULL != n->child)
1.184 schwarze 1372: putmdockey(mpage, n->child, m, taboo);
1.141 schwarze 1373: if (n->type == ROFFT_TEXT)
1.47 schwarze 1374: putkey(mpage, n->string, m);
1375: }
1376: }
1.16 schwarze 1377:
1.47 schwarze 1378: static void
1.143 schwarze 1379: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1380: const struct roff_node *n)
1.47 schwarze 1381: {
1.142 schwarze 1382: const struct roff_node *head, *body;
1.78 schwarze 1383: char *start, *title;
1.47 schwarze 1384: char byte;
1.78 schwarze 1385: size_t sz;
1.16 schwarze 1386:
1.168 schwarze 1387: if (n == NULL)
1.47 schwarze 1388: return;
1.16 schwarze 1389:
1.47 schwarze 1390: /*
1391: * We're only searching for one thing: the first text child in
1392: * the BODY of a NAME section. Since we don't keep track of
1393: * sections in -man, run some hoops to find out whether we're in
1394: * the correct section or not.
1395: */
1.16 schwarze 1396:
1.141 schwarze 1397: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1398: body = n;
1.168 schwarze 1399: if ((head = body->parent->head) != NULL &&
1400: (head = head->child) != NULL &&
1401: head->next == NULL &&
1.141 schwarze 1402: head->type == ROFFT_TEXT &&
1.168 schwarze 1403: strcmp(head->string, "NAME") == 0 &&
1404: body->child != NULL) {
1.2 schwarze 1405:
1.47 schwarze 1406: /*
1407: * Suck the entire NAME section into memory.
1408: * Yes, we might run away.
1409: * But too many manuals have big, spread-out
1410: * NAME sections over many lines.
1411: */
1.2 schwarze 1412:
1.78 schwarze 1413: title = NULL;
1.147 schwarze 1414: deroff(&title, body);
1.47 schwarze 1415: if (NULL == title)
1416: return;
1.16 schwarze 1417:
1.98 schwarze 1418: /*
1.47 schwarze 1419: * Go through a special heuristic dance here.
1420: * Conventionally, one or more manual names are
1421: * comma-specified prior to a whitespace, then a
1422: * dash, then a description. Try to puzzle out
1423: * the name parts here.
1424: */
1.16 schwarze 1425:
1.78 schwarze 1426: start = title;
1.47 schwarze 1427: for ( ;; ) {
1428: sz = strcspn(start, " ,");
1429: if ('\0' == start[sz])
1430: break;
1.1 schwarze 1431:
1.47 schwarze 1432: byte = start[sz];
1433: start[sz] = '\0';
1.67 schwarze 1434:
1435: /*
1436: * Assume a stray trailing comma in the
1437: * name list if a name begins with a dash.
1438: */
1439:
1440: if ('-' == start[0] ||
1441: ('\\' == start[0] && '-' == start[1]))
1442: break;
1.1 schwarze 1443:
1.90 schwarze 1444: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1445: if ( ! (mpage->name_head_done ||
1446: strcasecmp(start, meta->title))) {
1447: putkey(mpage, start, NAME_HEAD);
1448: mpage->name_head_done = 1;
1449: }
1.1 schwarze 1450:
1.47 schwarze 1451: if (' ' == byte) {
1452: start += sz + 1;
1453: break;
1454: }
1.1 schwarze 1455:
1.47 schwarze 1456: assert(',' == byte);
1457: start += sz + 1;
1458: while (' ' == *start)
1459: start++;
1460: }
1.1 schwarze 1461:
1.78 schwarze 1462: if (start == title) {
1.90 schwarze 1463: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1464: if ( ! (mpage->name_head_done ||
1465: strcasecmp(start, meta->title))) {
1466: putkey(mpage, start, NAME_HEAD);
1467: mpage->name_head_done = 1;
1468: }
1.47 schwarze 1469: free(title);
1470: return;
1471: }
1.1 schwarze 1472:
1.47 schwarze 1473: while (isspace((unsigned char)*start))
1474: start++;
1.1 schwarze 1475:
1.47 schwarze 1476: if (0 == strncmp(start, "-", 1))
1477: start += 1;
1478: else if (0 == strncmp(start, "\\-\\-", 4))
1479: start += 4;
1480: else if (0 == strncmp(start, "\\-", 2))
1481: start += 2;
1482: else if (0 == strncmp(start, "\\(en", 4))
1483: start += 4;
1484: else if (0 == strncmp(start, "\\(em", 4))
1485: start += 4;
1.1 schwarze 1486:
1.47 schwarze 1487: while (' ' == *start)
1488: start++;
1.1 schwarze 1489:
1.206 schwarze 1490: /*
1491: * Cut off excessive one-line descriptions.
1492: * Bad pages are not worth better heuristics.
1493: */
1494:
1495: mpage->desc = mandoc_strndup(start, 150);
1.47 schwarze 1496: free(title);
1497: return;
1498: }
1499: }
1.1 schwarze 1500:
1.47 schwarze 1501: for (n = n->child; n; n = n->next) {
1502: if (NULL != mpage->desc)
1503: break;
1.129 schwarze 1504: parse_man(mpage, meta, n);
1.1 schwarze 1505: }
1506: }
1507:
1508: static void
1.143 schwarze 1509: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1510: const struct roff_node *n)
1.1 schwarze 1511: {
1.208 schwarze 1512: const struct mdoc_handler *handler;
1.1 schwarze 1513:
1.195 schwarze 1514: for (n = n->child; n != NULL; n = n->next) {
1.208 schwarze 1515: if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX)
1.184 schwarze 1516: continue;
1.196 schwarze 1517: assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1.208 schwarze 1518: handler = mdoc_handlers + (n->tok - MDOC_Dd);
1519: if (n->flags & handler->taboo)
1520: continue;
1521:
1.47 schwarze 1522: switch (n->type) {
1.141 schwarze 1523: case ROFFT_ELEM:
1524: case ROFFT_BLOCK:
1525: case ROFFT_HEAD:
1526: case ROFFT_BODY:
1527: case ROFFT_TAIL:
1.208 schwarze 1528: if (handler->fp != NULL &&
1529: (*handler->fp)(mpage, meta, n) == 0)
1.195 schwarze 1530: break;
1.208 schwarze 1531: if (handler->mask)
1.47 schwarze 1532: putmdockey(mpage, n->child,
1.208 schwarze 1533: handler->mask, handler->taboo);
1.47 schwarze 1534: break;
1535: default:
1536: continue;
1537: }
1538: if (NULL != n->child)
1.127 schwarze 1539: parse_mdoc(mpage, meta, n);
1.1 schwarze 1540: }
1541: }
1542:
1.19 schwarze 1543: static int
1.207 schwarze 1544: parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta,
1545: const struct roff_node *n)
1546: {
1547: uint64_t mask;
1548:
1549: mask = TYPE_Fa;
1550: if (n->sec == SEC_SYNOPSIS)
1551: mask |= TYPE_Vt;
1552:
1553: putmdockey(mpage, n->child, mask, 0);
1554: return 0;
1555: }
1556:
1557: static int
1.143 schwarze 1558: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1559: const struct roff_node *n)
1.1 schwarze 1560: {
1.131 schwarze 1561: char *start, *end;
1.1 schwarze 1562: size_t sz;
1.19 schwarze 1563:
1.47 schwarze 1564: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1565: NULL == (n = n->child) ||
1.141 schwarze 1566: n->type != ROFFT_TEXT)
1.150 schwarze 1567: return 0;
1.1 schwarze 1568:
1569: /*
1570: * Only consider those `Fd' macro fields that begin with an
1571: * "inclusion" token (versus, e.g., #define).
1572: */
1.47 schwarze 1573:
1.1 schwarze 1574: if (strcmp("#include", n->string))
1.150 schwarze 1575: return 0;
1.1 schwarze 1576:
1.141 schwarze 1577: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1578: return 0;
1.1 schwarze 1579:
1580: /*
1581: * Strip away the enclosing angle brackets and make sure we're
1582: * not zero-length.
1583: */
1584:
1585: start = n->string;
1586: if ('<' == *start || '"' == *start)
1587: start++;
1588:
1589: if (0 == (sz = strlen(start)))
1.150 schwarze 1590: return 0;
1.1 schwarze 1591:
1592: end = &start[(int)sz - 1];
1593: if ('>' == *end || '"' == *end)
1594: end--;
1595:
1.47 schwarze 1596: if (end > start)
1597: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1598: return 0;
1.1 schwarze 1599: }
1600:
1.133 schwarze 1601: static void
1.142 schwarze 1602: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1603: {
1.69 schwarze 1604: char *cp;
1.133 schwarze 1605: size_t sz;
1.1 schwarze 1606:
1.141 schwarze 1607: if (n->type != ROFFT_TEXT)
1.133 schwarze 1608: return;
1.19 schwarze 1609:
1.133 schwarze 1610: /* Skip function pointer punctuation. */
1.1 schwarze 1611:
1.133 schwarze 1612: cp = n->string;
1613: while (*cp == '(' || *cp == '*')
1.1 schwarze 1614: cp++;
1.133 schwarze 1615: sz = strcspn(cp, "()");
1.1 schwarze 1616:
1.133 schwarze 1617: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1618: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1619: putkeys(mpage, cp, sz, NAME_SYN);
1620: }
1.19 schwarze 1621:
1.133 schwarze 1622: static int
1.143 schwarze 1623: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1624: const struct roff_node *n)
1.133 schwarze 1625: {
1.207 schwarze 1626: uint64_t mask;
1.133 schwarze 1627:
1628: if (n->child == NULL)
1.150 schwarze 1629: return 0;
1.133 schwarze 1630:
1631: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1632:
1.207 schwarze 1633: n = n->child->next;
1634: if (n != NULL && n->type == ROFFT_TEXT) {
1635: mask = TYPE_Fa;
1636: if (n->sec == SEC_SYNOPSIS)
1637: mask |= TYPE_Vt;
1638: putmdockey(mpage, n, mask, 0);
1639: }
1.19 schwarze 1640:
1.150 schwarze 1641: return 0;
1.128 schwarze 1642: }
1643:
1644: static int
1.143 schwarze 1645: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1646: const struct roff_node *n)
1.128 schwarze 1647: {
1.132 schwarze 1648:
1.141 schwarze 1649: if (n->type != ROFFT_HEAD)
1.150 schwarze 1650: return 1;
1.128 schwarze 1651:
1.133 schwarze 1652: if (n->child != NULL)
1653: parse_mdoc_fname(mpage, n->child);
1654:
1.150 schwarze 1655: return 0;
1.1 schwarze 1656: }
1657:
1.19 schwarze 1658: static int
1.164 schwarze 1659: parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1660: const struct roff_node *n)
1661: {
1662: char *cp;
1663:
1664: if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1665: return 0;
1666:
1.168 schwarze 1667: if (n->child != NULL &&
1668: n->child->next == NULL &&
1669: n->child->type == ROFFT_TEXT)
1.164 schwarze 1670: return 1;
1671:
1672: cp = NULL;
1673: deroff(&cp, n);
1674: if (cp != NULL) {
1675: putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1676: n->type == ROFFT_BODY ? TYPE_Va : 0));
1677: free(cp);
1678: }
1679:
1680: return 0;
1681: }
1682:
1683: static int
1.143 schwarze 1684: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1685: const struct roff_node *n)
1.1 schwarze 1686: {
1.47 schwarze 1687: char *cp;
1.1 schwarze 1688:
1689: if (NULL == (n = n->child))
1.150 schwarze 1690: return 0;
1.1 schwarze 1691:
1.47 schwarze 1692: if (NULL == n->next) {
1693: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1694: return 0;
1.47 schwarze 1695: }
1.1 schwarze 1696:
1.77 schwarze 1697: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1698: putkey(mpage, cp, TYPE_Xr);
1699: free(cp);
1.150 schwarze 1700: return 0;
1.1 schwarze 1701: }
1702:
1.19 schwarze 1703: static int
1.143 schwarze 1704: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1705: const struct roff_node *n)
1.1 schwarze 1706: {
1707:
1.141 schwarze 1708: if (n->type == ROFFT_BODY)
1.147 schwarze 1709: deroff(&mpage->desc, n);
1.150 schwarze 1710: return 0;
1.1 schwarze 1711: }
1712:
1.19 schwarze 1713: static int
1.143 schwarze 1714: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1715: const struct roff_node *n)
1.1 schwarze 1716: {
1717:
1.86 schwarze 1718: if (SEC_NAME == n->sec)
1.184 schwarze 1719: putmdockey(mpage, n->child, NAME_TITLE, 0);
1.141 schwarze 1720: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1721: if (n->child == NULL)
1722: putkey(mpage, meta->name, NAME_SYN);
1723: else
1.184 schwarze 1724: putmdockey(mpage, n->child, NAME_SYN, 0);
1.129 schwarze 1725: }
1726: if ( ! (mpage->name_head_done ||
1727: n->child == NULL || n->child->string == NULL ||
1728: strcasecmp(n->child->string, meta->title))) {
1.175 schwarze 1729: putkey(mpage, n->child->string, NAME_HEAD);
1.129 schwarze 1730: mpage->name_head_done = 1;
1.127 schwarze 1731: }
1.150 schwarze 1732: return 0;
1.1 schwarze 1733: }
1734:
1.19 schwarze 1735: static int
1.143 schwarze 1736: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1737: const struct roff_node *n)
1.1 schwarze 1738: {
1739:
1.150 schwarze 1740: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1741: }
1742:
1.47 schwarze 1743: static int
1.143 schwarze 1744: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1745: const struct roff_node *n)
1.1 schwarze 1746: {
1747:
1.150 schwarze 1748: return n->type == ROFFT_HEAD;
1.1 schwarze 1749: }
1750:
1.47 schwarze 1751: /*
1752: * Add a string to the hash table for the current manual.
1753: * Each string has a bitmask telling which macros it belongs to.
1754: * When we finish the manual, we'll dump the table.
1755: */
1.1 schwarze 1756: static void
1.131 schwarze 1757: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1758: {
1.90 schwarze 1759: struct ohash *htab;
1.47 schwarze 1760: struct str *s;
1.68 schwarze 1761: const char *end;
1.47 schwarze 1762: unsigned int slot;
1.131 schwarze 1763: int i, mustfree;
1.1 schwarze 1764:
1.47 schwarze 1765: if (0 == sz)
1766: return;
1.68 schwarze 1767:
1.131 schwarze 1768: mustfree = render_string(&cp, &sz);
1769:
1.90 schwarze 1770: if (TYPE_Nm & v) {
1771: htab = &names;
1772: v &= name_mask;
1.124 schwarze 1773: if (v & NAME_FIRST)
1774: name_mask &= ~NAME_FIRST;
1.90 schwarze 1775: if (debug > 1)
1776: say(mpage->mlinks->file,
1.183 schwarze 1777: "Adding name %*s, bits=0x%llx", (int)sz, cp,
1778: (unsigned long long)v);
1.90 schwarze 1779: } else {
1780: htab = &strings;
1781: if (debug > 1)
1.173 schwarze 1782: for (i = 0; i < KEY_MAX; i++)
1.118 schwarze 1783: if ((uint64_t)1 << i & v)
1.90 schwarze 1784: say(mpage->mlinks->file,
1785: "Adding key %s=%*s",
1.172 schwarze 1786: mansearch_keynames[i], (int)sz, cp);
1.68 schwarze 1787: }
1.47 schwarze 1788:
1789: end = cp + sz;
1.90 schwarze 1790: slot = ohash_qlookupi(htab, cp, &end);
1791: s = ohash_find(htab, slot);
1.1 schwarze 1792:
1.47 schwarze 1793: if (NULL != s && mpage == s->mpage) {
1794: s->mask |= v;
1.1 schwarze 1795: return;
1.47 schwarze 1796: } else if (NULL == s) {
1.101 schwarze 1797: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1798: memcpy(s->key, cp, sz);
1.90 schwarze 1799: ohash_insert(htab, slot, s);
1.47 schwarze 1800: }
1801: s->mpage = mpage;
1802: s->mask = v;
1.131 schwarze 1803:
1804: if (mustfree)
1805: free(cp);
1.1 schwarze 1806: }
1807:
1808: /*
1.47 schwarze 1809: * Take a Unicode codepoint and produce its UTF-8 encoding.
1810: * This isn't the best way to do this, but it works.
1811: * The magic numbers are from the UTF-8 packaging.
1812: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1813: */
1.47 schwarze 1814: static size_t
1815: utf8(unsigned int cp, char out[7])
1.1 schwarze 1816: {
1.47 schwarze 1817: size_t rc;
1.1 schwarze 1818:
1.47 schwarze 1819: rc = 0;
1820: if (cp <= 0x0000007F) {
1821: rc = 1;
1822: out[0] = (char)cp;
1823: } else if (cp <= 0x000007FF) {
1824: rc = 2;
1825: out[0] = (cp >> 6 & 31) | 192;
1826: out[1] = (cp & 63) | 128;
1827: } else if (cp <= 0x0000FFFF) {
1828: rc = 3;
1829: out[0] = (cp >> 12 & 15) | 224;
1830: out[1] = (cp >> 6 & 63) | 128;
1831: out[2] = (cp & 63) | 128;
1832: } else if (cp <= 0x001FFFFF) {
1833: rc = 4;
1834: out[0] = (cp >> 18 & 7) | 240;
1835: out[1] = (cp >> 12 & 63) | 128;
1836: out[2] = (cp >> 6 & 63) | 128;
1837: out[3] = (cp & 63) | 128;
1838: } else if (cp <= 0x03FFFFFF) {
1839: rc = 5;
1840: out[0] = (cp >> 24 & 3) | 248;
1841: out[1] = (cp >> 18 & 63) | 128;
1842: out[2] = (cp >> 12 & 63) | 128;
1843: out[3] = (cp >> 6 & 63) | 128;
1844: out[4] = (cp & 63) | 128;
1845: } else if (cp <= 0x7FFFFFFF) {
1846: rc = 6;
1847: out[0] = (cp >> 30 & 1) | 252;
1848: out[1] = (cp >> 24 & 63) | 128;
1849: out[2] = (cp >> 18 & 63) | 128;
1850: out[3] = (cp >> 12 & 63) | 128;
1851: out[4] = (cp >> 6 & 63) | 128;
1852: out[5] = (cp & 63) | 128;
1853: } else
1.150 schwarze 1854: return 0;
1.19 schwarze 1855:
1.47 schwarze 1856: out[rc] = '\0';
1.150 schwarze 1857: return rc;
1.1 schwarze 1858: }
1859:
1.47 schwarze 1860: /*
1.131 schwarze 1861: * If the string contains escape sequences,
1862: * replace it with an allocated rendering and return 1,
1863: * such that the caller can free it after use.
1864: * Otherwise, do nothing and return 0.
1.47 schwarze 1865: */
1.131 schwarze 1866: static int
1867: render_string(char **public, size_t *psz)
1.1 schwarze 1868: {
1.131 schwarze 1869: const char *src, *scp, *addcp, *seq;
1870: char *dst;
1871: size_t ssz, dsz, addsz;
1.71 schwarze 1872: char utfbuf[7], res[6];
1.131 schwarze 1873: int seqlen, unicode;
1.47 schwarze 1874:
1875: res[0] = '\\';
1876: res[1] = '\t';
1877: res[2] = ASCII_NBRSP;
1878: res[3] = ASCII_HYPH;
1.71 schwarze 1879: res[4] = ASCII_BREAK;
1880: res[5] = '\0';
1.1 schwarze 1881:
1.131 schwarze 1882: src = scp = *public;
1883: ssz = *psz;
1884: dst = NULL;
1885: dsz = 0;
1886:
1887: while (scp < src + *psz) {
1888:
1889: /* Leave normal characters unchanged. */
1890:
1891: if (strchr(res, *scp) == NULL) {
1892: if (dst != NULL)
1893: dst[dsz++] = *scp;
1894: scp++;
1895: continue;
1896: }
1.1 schwarze 1897:
1.47 schwarze 1898: /*
1.131 schwarze 1899: * Found something that requires replacing,
1900: * make sure we have a destination buffer.
1.47 schwarze 1901: */
1.131 schwarze 1902:
1903: if (dst == NULL) {
1904: dst = mandoc_malloc(ssz + 1);
1905: dsz = scp - src;
1906: memcpy(dst, src, dsz);
1.47 schwarze 1907: }
1.39 schwarze 1908:
1.131 schwarze 1909: /* Handle single-char special characters. */
1910:
1911: switch (*scp) {
1912: case '\\':
1913: break;
1.98 schwarze 1914: case '\t':
1915: case ASCII_NBRSP:
1.131 schwarze 1916: dst[dsz++] = ' ';
1917: scp++;
1918: continue;
1919: case ASCII_HYPH:
1920: dst[dsz++] = '-';
1.71 schwarze 1921: /* FALLTHROUGH */
1.98 schwarze 1922: case ASCII_BREAK:
1.131 schwarze 1923: scp++;
1.47 schwarze 1924: continue;
1.71 schwarze 1925: default:
1.131 schwarze 1926: abort();
1.71 schwarze 1927: }
1.39 schwarze 1928:
1.47 schwarze 1929: /*
1.131 schwarze 1930: * Found an escape sequence.
1931: * Read past the slash, then parse it.
1932: * Ignore everything except characters.
1.47 schwarze 1933: */
1.52 schwarze 1934:
1.131 schwarze 1935: scp++;
1936: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1937: continue;
1.39 schwarze 1938:
1.47 schwarze 1939: /*
1.52 schwarze 1940: * Render the special character
1941: * as either UTF-8 or ASCII.
1.47 schwarze 1942: */
1.52 schwarze 1943:
1944: if (write_utf8) {
1.157 schwarze 1945: unicode = mchars_spec2cp(seq, seqlen);
1.131 schwarze 1946: if (unicode <= 0)
1.52 schwarze 1947: continue;
1.131 schwarze 1948: addsz = utf8(unicode, utfbuf);
1949: if (addsz == 0)
1.52 schwarze 1950: continue;
1.131 schwarze 1951: addcp = utfbuf;
1.52 schwarze 1952: } else {
1.157 schwarze 1953: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.131 schwarze 1954: if (addcp == NULL)
1.52 schwarze 1955: continue;
1.131 schwarze 1956: if (*addcp == ASCII_NBRSP) {
1957: addcp = " ";
1958: addsz = 1;
1.52 schwarze 1959: }
1960: }
1.1 schwarze 1961:
1.47 schwarze 1962: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1963:
1.131 schwarze 1964: ssz += addsz;
1965: dst = mandoc_realloc(dst, ssz + 1);
1966: memcpy(dst + dsz, addcp, addsz);
1967: dsz += addsz;
1968: }
1969: if (dst != NULL) {
1970: *public = dst;
1971: *psz = dsz;
1972: }
1973:
1974: /* Trim trailing whitespace and NUL-terminate. */
1975:
1976: while (*psz > 0 && (*public)[*psz - 1] == ' ')
1977: --*psz;
1978: if (dst != NULL) {
1979: (*public)[*psz] = '\0';
1.150 schwarze 1980: return 1;
1.131 schwarze 1981: } else
1.150 schwarze 1982: return 0;
1.1 schwarze 1983: }
1984:
1.75 schwarze 1985: static void
1986: dbadd_mlink(const struct mlink *mlink)
1987: {
1.173 schwarze 1988: dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
1989: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
1990: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
1991: dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
1992: dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
1.75 schwarze 1993: }
1994:
1.11 schwarze 1995: /*
1.47 schwarze 1996: * Flush the current page's terms (and their bits) into the database.
1.53 schwarze 1997: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 1998: */
1999: static void
1.173 schwarze 2000: dbadd(struct dba *dba, struct mpage *mpage)
1.11 schwarze 2001: {
1.47 schwarze 2002: struct mlink *mlink;
2003: struct str *key;
1.131 schwarze 2004: char *cp;
1.173 schwarze 2005: uint64_t mask;
1.47 schwarze 2006: size_t i;
2007: unsigned int slot;
1.131 schwarze 2008: int mustfree;
1.47 schwarze 2009:
1.85 schwarze 2010: mlink = mpage->mlinks;
1.11 schwarze 2011:
1.85 schwarze 2012: if (nodb) {
1.104 schwarze 2013: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2014: key = ohash_next(&names, &slot))
1.104 schwarze 2015: free(key);
2016: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2017: key = ohash_next(&strings, &slot))
1.104 schwarze 2018: free(key);
1.102 schwarze 2019: if (0 == debug)
2020: return;
1.85 schwarze 2021: while (NULL != mlink) {
2022: fputs(mlink->name, stdout);
2023: if (NULL == mlink->next ||
2024: strcmp(mlink->dsec, mlink->next->dsec) ||
2025: strcmp(mlink->fsec, mlink->next->fsec) ||
2026: strcmp(mlink->arch, mlink->next->arch)) {
2027: putchar('(');
2028: if ('\0' == *mlink->dsec)
2029: fputs(mlink->fsec, stdout);
2030: else
2031: fputs(mlink->dsec, stdout);
2032: if ('\0' != *mlink->arch)
2033: printf("/%s", mlink->arch);
2034: putchar(')');
2035: }
2036: mlink = mlink->next;
2037: if (NULL != mlink)
2038: fputs(", ", stdout);
2039: }
1.89 schwarze 2040: printf(" - %s\n", mpage->desc);
1.11 schwarze 2041: return;
1.85 schwarze 2042: }
2043:
2044: if (debug)
2045: say(mlink->file, "Adding to database");
1.47 schwarze 2046:
1.131 schwarze 2047: cp = mpage->desc;
2048: i = strlen(cp);
2049: mustfree = render_string(&cp, &i);
1.176 schwarze 2050: mpage->dba = dba_page_new(dba->pages,
1.173 schwarze 2051: *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2052: cp, mlink->file, mpage->form);
1.131 schwarze 2053: if (mustfree)
2054: free(cp);
1.176 schwarze 2055: dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
1.47 schwarze 2056:
1.173 schwarze 2057: while (mlink != NULL) {
1.75 schwarze 2058: dbadd_mlink(mlink);
1.85 schwarze 2059: mlink = mlink->next;
2060: }
1.47 schwarze 2061:
1.90 schwarze 2062: for (key = ohash_first(&names, &slot); NULL != key;
2063: key = ohash_next(&names, &slot)) {
2064: assert(key->mpage == mpage);
1.173 schwarze 2065: dba_page_alias(mpage->dba, key->key, key->mask);
1.90 schwarze 2066: free(key);
2067: }
1.47 schwarze 2068: for (key = ohash_first(&strings, &slot); NULL != key;
2069: key = ohash_next(&strings, &slot)) {
2070: assert(key->mpage == mpage);
1.173 schwarze 2071: i = 0;
2072: for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2073: if (key->mask & mask)
2074: dba_macro_add(dba->macros, i,
2075: key->key, mpage->dba);
2076: i++;
2077: }
1.47 schwarze 2078: free(key);
1.33 schwarze 2079: }
1.47 schwarze 2080: }
1.41 deraadt 2081:
1.47 schwarze 2082: static void
1.173 schwarze 2083: dbprune(struct dba *dba)
1.47 schwarze 2084: {
1.173 schwarze 2085: struct dba_array *page, *files;
2086: char *file;
1.11 schwarze 2087:
1.173 schwarze 2088: dba_array_FOREACH(dba->pages, page) {
2089: files = dba_array_get(page, DBP_FILE);
2090: dba_array_FOREACH(files, file) {
2091: if (*file < ' ')
2092: file++;
2093: if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2094: file)) != NULL) {
2095: if (debug)
2096: say(file, "Deleting from database");
2097: dba_array_del(dba->pages);
2098: break;
2099: }
1.63 schwarze 2100: }
1.11 schwarze 2101: }
1.47 schwarze 2102: }
1.22 schwarze 2103:
1.47 schwarze 2104: /*
1.173 schwarze 2105: * Write the database from memory to disk.
1.47 schwarze 2106: */
2107: static void
1.173 schwarze 2108: dbwrite(struct dba *dba)
1.47 schwarze 2109: {
1.204 schwarze 2110: struct stat sb1, sb2;
2111: char tfn[33], *cp1, *cp2;
2112: off_t i;
2113: int fd1, fd2;
1.11 schwarze 2114:
1.199 schwarze 2115: /*
2116: * Do not write empty databases, and delete existing ones
2117: * when makewhatis -u causes them to become empty.
2118: */
2119:
2120: dba_array_start(dba->pages);
2121: if (dba_array_next(dba->pages) == NULL) {
1.201 schwarze 2122: if (unlink(MANDOC_DB) == -1 && errno != ENOENT)
1.199 schwarze 2123: say(MANDOC_DB, "&unlink");
2124: return;
2125: }
2126:
2127: /*
2128: * Build the database in a temporary file,
2129: * then atomically move it into place.
2130: */
2131:
1.173 schwarze 2132: if (dba_write(MANDOC_DB "~", dba) != -1) {
2133: if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2134: exitcode = (int)MANDOCLEVEL_SYSERR;
2135: say(MANDOC_DB, "&rename");
2136: unlink(MANDOC_DB "~");
2137: }
1.47 schwarze 2138: return;
1.28 schwarze 2139: }
1.199 schwarze 2140:
2141: /*
2142: * We lack write permission and cannot replace the database
2143: * file, but let's at least check whether the data changed.
2144: */
1.22 schwarze 2145:
1.173 schwarze 2146: (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2147: if (mkdtemp(tfn) == NULL) {
2148: exitcode = (int)MANDOCLEVEL_SYSERR;
2149: say("", "&%s", tfn);
1.47 schwarze 2150: return;
1.173 schwarze 2151: }
1.205 schwarze 2152: cp1 = cp2 = MAP_FAILED;
1.204 schwarze 2153: fd1 = fd2 = -1;
1.173 schwarze 2154: (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2155: if (dba_write(tfn, dba) == -1) {
2156: say(tfn, "&dba_write");
1.204 schwarze 2157: goto err;
1.72 schwarze 2158: }
1.204 schwarze 2159: if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) {
2160: say(MANDOC_DB, "&open");
2161: goto err;
2162: }
2163: if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) {
2164: say(tfn, "&open");
2165: goto err;
2166: }
2167: if (fstat(fd1, &sb1) == -1) {
2168: say(MANDOC_DB, "&fstat");
2169: goto err;
2170: }
2171: if (fstat(fd2, &sb2) == -1) {
2172: say(tfn, "&fstat");
2173: goto err;
2174: }
2175: if (sb1.st_size != sb2.st_size)
2176: goto err;
2177: if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE,
1.205 schwarze 2178: fd1, 0)) == MAP_FAILED) {
1.204 schwarze 2179: say(MANDOC_DB, "&mmap");
2180: goto err;
2181: }
2182: if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE,
1.205 schwarze 2183: fd2, 0)) == MAP_FAILED) {
1.204 schwarze 2184: say(tfn, "&mmap");
2185: goto err;
2186: }
2187: for (i = 0; i < sb1.st_size; i++)
2188: if (cp1[i] != cp2[i])
2189: goto err;
2190: goto out;
2191:
2192: err:
2193: exitcode = (int)MANDOCLEVEL_SYSERR;
2194: say(MANDOC_DB, "Data changed, but cannot replace database");
1.72 schwarze 2195:
1.173 schwarze 2196: out:
1.205 schwarze 2197: if (cp1 != MAP_FAILED)
1.204 schwarze 2198: munmap(cp1, sb1.st_size);
1.205 schwarze 2199: if (cp2 != MAP_FAILED)
1.204 schwarze 2200: munmap(cp2, sb2.st_size);
2201: if (fd1 != -1)
2202: close(fd1);
2203: if (fd2 != -1)
2204: close(fd2);
1.203 schwarze 2205: unlink(tfn);
1.173 schwarze 2206: *strrchr(tfn, '/') = '\0';
1.203 schwarze 2207: rmdir(tfn);
1.47 schwarze 2208: }
1.6 schwarze 2209:
1.47 schwarze 2210: static int
1.120 schwarze 2211: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2212: {
2213: static char startdir[PATH_MAX];
1.110 schwarze 2214: static int getcwd_status; /* 1 = ok, 2 = failure */
2215: static int chdir_status; /* 1 = changed directory */
1.6 schwarze 2216:
1.47 schwarze 2217: /*
1.110 schwarze 2218: * Remember the original working directory, if possible.
2219: * This will be needed if the second or a later directory
2220: * on the command line is given as a relative path.
2221: * Do not error out if the current directory is not
2222: * searchable: Maybe it won't be needed after all.
2223: */
1.213 ! schwarze 2224: if (getcwd_status == 0) {
! 2225: if (getcwd(startdir, sizeof(startdir)) == NULL) {
1.110 schwarze 2226: getcwd_status = 2;
2227: (void)strlcpy(startdir, strerror(errno),
2228: sizeof(startdir));
2229: } else
2230: getcwd_status = 1;
2231: }
2232:
2233: /*
2234: * We are leaving the old base directory.
2235: * Do not use it any longer, not even for messages.
2236: */
2237: *basedir = '\0';
1.213 ! schwarze 2238: basedir_len = 0;
1.110 schwarze 2239:
2240: /*
2241: * If and only if the directory was changed earlier and
2242: * the next directory to process is given as a relative path,
2243: * first go back, or bail out if that is impossible.
1.47 schwarze 2244: */
1.213 ! schwarze 2245: if (chdir_status && *targetdir != '/') {
! 2246: if (getcwd_status == 2) {
1.47 schwarze 2247: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2248: say("", "getcwd: %s", startdir);
1.150 schwarze 2249: return 0;
1.47 schwarze 2250: }
1.213 ! schwarze 2251: if (chdir(startdir) == -1) {
1.47 schwarze 2252: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2253: say("", "&chdir %s", startdir);
1.150 schwarze 2254: return 0;
1.2 schwarze 2255: }
2256: }
1.110 schwarze 2257:
2258: /*
2259: * Always resolve basedir to the canonicalized absolute
2260: * pathname and append a trailing slash, such that
2261: * we can reliably check whether files are inside.
2262: */
1.213 ! schwarze 2263: if (realpath(targetdir, basedir) == NULL) {
1.120 schwarze 2264: if (report_baddir || errno != ENOENT) {
2265: exitcode = (int)MANDOCLEVEL_BADARG;
2266: say("", "&%s: realpath", targetdir);
2267: }
1.213 ! schwarze 2268: *basedir = '\0';
1.150 schwarze 2269: return 0;
1.213 ! schwarze 2270: } else if (chdir(basedir) == -1) {
1.120 schwarze 2271: if (report_baddir || errno != ENOENT) {
2272: exitcode = (int)MANDOCLEVEL_BADARG;
2273: say("", "&chdir");
2274: }
1.213 ! schwarze 2275: *basedir = '\0';
1.150 schwarze 2276: return 0;
1.109 schwarze 2277: }
1.110 schwarze 2278: chdir_status = 1;
1.213 ! schwarze 2279: basedir_len = strlen(basedir);
! 2280: if (basedir[basedir_len - 1] != '/') {
! 2281: if (basedir_len >= PATH_MAX - 1) {
1.109 schwarze 2282: exitcode = (int)MANDOCLEVEL_SYSERR;
2283: say("", "Filename too long");
1.213 ! schwarze 2284: *basedir = '\0';
! 2285: basedir_len = 0;
1.150 schwarze 2286: return 0;
1.109 schwarze 2287: }
1.213 ! schwarze 2288: basedir[basedir_len++] = '/';
! 2289: basedir[basedir_len] = '\0';
1.47 schwarze 2290: }
1.150 schwarze 2291: return 1;
1.2 schwarze 2292: }
2293:
2294: static void
1.47 schwarze 2295: say(const char *file, const char *format, ...)
1.2 schwarze 2296: {
1.47 schwarze 2297: va_list ap;
1.80 schwarze 2298: int use_errno;
1.2 schwarze 2299:
1.213 ! schwarze 2300: if (*basedir != '\0')
1.47 schwarze 2301: fprintf(stderr, "%s", basedir);
1.213 ! schwarze 2302: if (*basedir != '\0' && *file != '\0')
1.110 schwarze 2303: fputc('/', stderr);
1.213 ! schwarze 2304: if (*file != '\0')
1.47 schwarze 2305: fprintf(stderr, "%s", file);
1.31 schwarze 2306:
1.80 schwarze 2307: use_errno = 1;
1.213 ! schwarze 2308: if (format != NULL) {
1.80 schwarze 2309: switch (*format) {
1.98 schwarze 2310: case '&':
1.80 schwarze 2311: format++;
2312: break;
1.98 schwarze 2313: case '\0':
1.80 schwarze 2314: format = NULL;
2315: break;
2316: default:
2317: use_errno = 0;
2318: break;
2319: }
2320: }
1.213 ! schwarze 2321: if (format != NULL) {
! 2322: if (*basedir != '\0' || *file != '\0')
1.80 schwarze 2323: fputs(": ", stderr);
2324: va_start(ap, format);
2325: vfprintf(stderr, format, ap);
2326: va_end(ap);
2327: }
2328: if (use_errno) {
1.213 ! schwarze 2329: if (*basedir != '\0' || *file != '\0' || format != NULL)
1.80 schwarze 2330: fputs(": ", stderr);
1.47 schwarze 2331: perror(NULL);
1.80 schwarze 2332: } else
2333: fputc('\n', stderr);
1.1 schwarze 2334: }