Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.208
1.208 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.207 2018/02/23 18:24:41 schwarze Exp $ */
1.1 schwarze 2: /*
1.47 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.208 ! schwarze 4: * Copyright (c) 2011-2018 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"
1.140 schwarze 44: #include "manconf.h"
1.47 schwarze 45: #include "mansearch.h"
1.173 schwarze 46: #include "dba_array.h"
47: #include "dba.h"
1.1 schwarze 48:
1.68 schwarze 49: extern const char *const mansearch_keynames[];
50:
1.47 schwarze 51: enum op {
52: OP_DEFAULT = 0, /* new dbs from dir list or default config */
53: OP_CONFFILE, /* new databases from custom config file */
54: OP_UPDATE, /* delete/add entries in existing database */
55: OP_DELETE, /* delete entries from existing database */
56: OP_TEST /* change no databases, report potential problems */
57: };
1.11 schwarze 58:
1.47 schwarze 59: struct str {
60: const struct mpage *mpage; /* if set, the owning parse */
61: uint64_t mask; /* bitmask in sequence */
1.131 schwarze 62: char key[]; /* rendered text */
1.28 schwarze 63: };
64:
1.47 schwarze 65: struct inodev {
66: ino_t st_ino;
67: dev_t st_dev;
68: };
1.28 schwarze 69:
1.47 schwarze 70: struct mpage {
71: struct inodev inodev; /* used for hashing routine */
1.173 schwarze 72: struct dba_array *dba;
1.47 schwarze 73: char *sec; /* section from file content */
74: char *arch; /* architecture from file content */
75: char *title; /* title from file content */
76: char *desc; /* description from file content */
1.180 schwarze 77: struct mpage *next; /* singly linked list */
1.47 schwarze 78: struct mlink *mlinks; /* singly linked list */
1.129 schwarze 79: int name_head_done;
1.173 schwarze 80: enum form form; /* format from file content */
1.28 schwarze 81: };
82:
1.47 schwarze 83: struct mlink {
84: char file[PATH_MAX]; /* filename rel. to manpath */
85: char *dsec; /* section from directory */
86: char *arch; /* architecture from directory */
87: char *name; /* name from file name (not empty) */
88: char *fsec; /* section from file name suffix */
89: struct mlink *next; /* singly linked list */
1.75 schwarze 90: struct mpage *mpage; /* parent */
1.81 schwarze 91: int gzip; /* filename has a .gz suffix */
1.173 schwarze 92: enum form dform; /* format from directory */
93: enum form fform; /* format from file name suffix */
1.1 schwarze 94: };
95:
1.143 schwarze 96: typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
1.142 schwarze 97: const struct roff_node *);
1.1 schwarze 98:
1.19 schwarze 99: struct mdoc_handler {
1.47 schwarze 100: mdoc_fp fp; /* optional handler */
101: uint64_t mask; /* set unless handler returns 0 */
1.184 schwarze 102: int taboo; /* node flags that must not be set */
1.19 schwarze 103: };
1.171 schwarze 104:
105:
106: int mandocdb(int, char *[]);
1.19 schwarze 107:
1.173 schwarze 108: static void dbadd(struct dba *, struct mpage *);
1.75 schwarze 109: static void dbadd_mlink(const struct mlink *mlink);
1.173 schwarze 110: static void dbprune(struct dba *);
111: static void dbwrite(struct dba *);
1.47 schwarze 112: static void filescan(const char *);
1.180 schwarze 113: static int fts_compare(const FTSENT **, const FTSENT **);
1.47 schwarze 114: static void mlink_add(struct mlink *, const struct stat *);
1.88 schwarze 115: static void mlink_check(struct mpage *, struct mlink *);
1.47 schwarze 116: static void mlink_free(struct mlink *);
117: static void mlinks_undupe(struct mpage *);
118: static void mpages_free(void);
1.173 schwarze 119: static void mpages_merge(struct dba *, struct mparse *);
1.81 schwarze 120: static void parse_cat(struct mpage *, int);
1.143 schwarze 121: static void parse_man(struct mpage *, const struct roff_meta *,
1.142 schwarze 122: const struct roff_node *);
1.143 schwarze 123: static void parse_mdoc(struct mpage *, const struct roff_meta *,
1.142 schwarze 124: const struct roff_node *);
1.143 schwarze 125: static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
1.142 schwarze 126: const struct roff_node *);
1.207 schwarze 127: static int parse_mdoc_Fa(struct mpage *, const struct roff_meta *,
128: const struct roff_node *);
1.143 schwarze 129: static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
1.142 schwarze 130: const struct roff_node *);
131: static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
1.143 schwarze 132: static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
1.142 schwarze 133: const struct roff_node *);
1.143 schwarze 134: static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
1.142 schwarze 135: const struct roff_node *);
1.143 schwarze 136: static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
1.142 schwarze 137: const struct roff_node *);
1.143 schwarze 138: static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
1.142 schwarze 139: const struct roff_node *);
1.143 schwarze 140: static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
1.142 schwarze 141: const struct roff_node *);
1.164 schwarze 142: static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
143: const struct roff_node *);
1.143 schwarze 144: static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
1.142 schwarze 145: const struct roff_node *);
1.69 schwarze 146: static void putkey(const struct mpage *, char *, uint64_t);
1.131 schwarze 147: static void putkeys(const struct mpage *, char *, size_t, uint64_t);
1.47 schwarze 148: static void putmdockey(const struct mpage *,
1.184 schwarze 149: const struct roff_node *, uint64_t, int);
1.131 schwarze 150: static int render_string(char **, size_t *);
1.172 schwarze 151: static void say(const char *, const char *, ...)
1.192 schwarze 152: __attribute__((__format__ (__printf__, 2, 3)));
1.120 schwarze 153: static int set_basedir(const char *, int);
1.47 schwarze 154: static int treescan(void);
155: static size_t utf8(unsigned int, char [7]);
156:
1.59 schwarze 157: static int nodb; /* no database changes */
1.73 schwarze 158: static int mparse_options; /* abort the parse early */
1.98 schwarze 159: static int use_all; /* use all found files */
160: static int debug; /* print what we're doing */
161: static int warnings; /* warn about crap */
1.52 schwarze 162: static int write_utf8; /* write UTF-8 output; else ASCII */
1.47 schwarze 163: static int exitcode; /* to be returned by main */
1.98 schwarze 164: static enum op op; /* operational mode */
1.47 schwarze 165: static char basedir[PATH_MAX]; /* current base directory */
1.180 schwarze 166: static struct mpage *mpage_head; /* list of distinct manual pages */
1.47 schwarze 167: static struct ohash mpages; /* table of distinct manual pages */
168: static struct ohash mlinks; /* table of directory entries */
1.90 schwarze 169: static struct ohash names; /* table of all names */
1.47 schwarze 170: static struct ohash strings; /* table of all strings */
1.90 schwarze 171: static uint64_t name_mask;
1.47 schwarze 172:
1.208 ! schwarze 173: static const struct mdoc_handler mdoc_handlers[MDOC_MAX - MDOC_Dd] = {
1.184 schwarze 174: { NULL, 0, NODE_NOPRT }, /* Dd */
175: { NULL, 0, NODE_NOPRT }, /* Dt */
176: { NULL, 0, NODE_NOPRT }, /* Os */
177: { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
178: { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
179: { NULL, 0, 0 }, /* Pp */
180: { NULL, 0, 0 }, /* D1 */
181: { NULL, 0, 0 }, /* Dl */
182: { NULL, 0, 0 }, /* Bd */
183: { NULL, 0, 0 }, /* Ed */
184: { NULL, 0, 0 }, /* Bl */
185: { NULL, 0, 0 }, /* El */
186: { NULL, 0, 0 }, /* It */
187: { NULL, 0, 0 }, /* Ad */
188: { NULL, TYPE_An, 0 }, /* An */
1.195 schwarze 189: { NULL, 0, 0 }, /* Ap */
1.184 schwarze 190: { NULL, TYPE_Ar, 0 }, /* Ar */
191: { NULL, TYPE_Cd, 0 }, /* Cd */
192: { NULL, TYPE_Cm, 0 }, /* Cm */
193: { NULL, TYPE_Dv, 0 }, /* Dv */
194: { NULL, TYPE_Er, 0 }, /* Er */
195: { NULL, TYPE_Ev, 0 }, /* Ev */
196: { NULL, 0, 0 }, /* Ex */
1.207 schwarze 197: { parse_mdoc_Fa, 0, 0 }, /* Fa */
1.184 schwarze 198: { parse_mdoc_Fd, 0, 0 }, /* Fd */
199: { NULL, TYPE_Fl, 0 }, /* Fl */
200: { parse_mdoc_Fn, 0, 0 }, /* Fn */
1.207 schwarze 201: { NULL, TYPE_Ft | TYPE_Vt, 0 }, /* Ft */
1.184 schwarze 202: { NULL, TYPE_Ic, 0 }, /* Ic */
203: { NULL, TYPE_In, 0 }, /* In */
204: { NULL, TYPE_Li, 0 }, /* Li */
205: { parse_mdoc_Nd, 0, 0 }, /* Nd */
206: { parse_mdoc_Nm, 0, 0 }, /* Nm */
207: { NULL, 0, 0 }, /* Op */
208: { NULL, 0, 0 }, /* Ot */
209: { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
210: { NULL, 0, 0 }, /* Rv */
211: { NULL, TYPE_St, 0 }, /* St */
212: { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
213: { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
214: { parse_mdoc_Xr, 0, 0 }, /* Xr */
215: { NULL, 0, 0 }, /* %A */
216: { NULL, 0, 0 }, /* %B */
217: { NULL, 0, 0 }, /* %D */
218: { NULL, 0, 0 }, /* %I */
219: { NULL, 0, 0 }, /* %J */
220: { NULL, 0, 0 }, /* %N */
221: { NULL, 0, 0 }, /* %O */
222: { NULL, 0, 0 }, /* %P */
223: { NULL, 0, 0 }, /* %R */
224: { NULL, 0, 0 }, /* %T */
225: { NULL, 0, 0 }, /* %V */
226: { NULL, 0, 0 }, /* Ac */
227: { NULL, 0, 0 }, /* Ao */
228: { NULL, 0, 0 }, /* Aq */
1.186 schwarze 229: { NULL, TYPE_At, 0 }, /* At */
1.184 schwarze 230: { NULL, 0, 0 }, /* Bc */
231: { NULL, 0, 0 }, /* Bf */
232: { NULL, 0, 0 }, /* Bo */
233: { NULL, 0, 0 }, /* Bq */
234: { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
1.185 schwarze 235: { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
1.184 schwarze 236: { NULL, 0, 0 }, /* Db */
237: { NULL, 0, 0 }, /* Dc */
238: { NULL, 0, 0 }, /* Do */
239: { NULL, 0, 0 }, /* Dq */
240: { NULL, 0, 0 }, /* Ec */
241: { NULL, 0, 0 }, /* Ef */
242: { NULL, TYPE_Em, 0 }, /* Em */
243: { NULL, 0, 0 }, /* Eo */
244: { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
245: { NULL, TYPE_Ms, 0 }, /* Ms */
246: { NULL, 0, 0 }, /* No */
247: { NULL, 0, 0 }, /* Ns */
248: { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
249: { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
250: { NULL, 0, 0 }, /* Pc */
251: { NULL, 0, 0 }, /* Pf */
252: { NULL, 0, 0 }, /* Po */
253: { NULL, 0, 0 }, /* Pq */
254: { NULL, 0, 0 }, /* Qc */
255: { NULL, 0, 0 }, /* Ql */
256: { NULL, 0, 0 }, /* Qo */
257: { NULL, 0, 0 }, /* Qq */
258: { NULL, 0, 0 }, /* Re */
259: { NULL, 0, 0 }, /* Rs */
260: { NULL, 0, 0 }, /* Sc */
261: { NULL, 0, 0 }, /* So */
262: { NULL, 0, 0 }, /* Sq */
263: { NULL, 0, 0 }, /* Sm */
264: { NULL, 0, 0 }, /* Sx */
265: { NULL, TYPE_Sy, 0 }, /* Sy */
266: { NULL, TYPE_Tn, 0 }, /* Tn */
267: { NULL, 0, NODE_NOSRC }, /* Ux */
268: { NULL, 0, 0 }, /* Xc */
269: { NULL, 0, 0 }, /* Xo */
270: { parse_mdoc_Fo, 0, 0 }, /* Fo */
271: { NULL, 0, 0 }, /* Fc */
272: { NULL, 0, 0 }, /* Oo */
273: { NULL, 0, 0 }, /* Oc */
274: { NULL, 0, 0 }, /* Bk */
275: { NULL, 0, 0 }, /* Ek */
276: { NULL, 0, 0 }, /* Bt */
277: { NULL, 0, 0 }, /* Hf */
278: { NULL, 0, 0 }, /* Fr */
279: { NULL, 0, 0 }, /* Ud */
1.186 schwarze 280: { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */
1.184 schwarze 281: { NULL, 0, 0 }, /* Lp */
282: { NULL, TYPE_Lk, 0 }, /* Lk */
283: { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
284: { NULL, 0, 0 }, /* Brq */
285: { NULL, 0, 0 }, /* Bro */
286: { NULL, 0, 0 }, /* Brc */
287: { NULL, 0, 0 }, /* %C */
288: { NULL, 0, 0 }, /* Es */
289: { NULL, 0, 0 }, /* En */
290: { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
291: { NULL, 0, 0 }, /* %Q */
292: { NULL, 0, 0 }, /* %U */
293: { NULL, 0, 0 }, /* Ta */
1.1 schwarze 294: };
295:
1.98 schwarze 296:
1.1 schwarze 297: int
1.3 schwarze 298: mandocdb(int argc, char *argv[])
1.1 schwarze 299: {
1.140 schwarze 300: struct manconf conf;
301: struct mparse *mp;
1.173 schwarze 302: struct dba *dba;
1.161 schwarze 303: const char *path_arg, *progname;
1.140 schwarze 304: size_t j, sz;
1.47 schwarze 305: int ch, i;
306:
1.204 schwarze 307: if (pledge("stdio rpath wpath cpath", NULL) == -1) {
1.167 schwarze 308: warn("pledge");
1.155 schwarze 309: return (int)MANDOCLEVEL_SYSERR;
310: }
311:
1.140 schwarze 312: memset(&conf, 0, sizeof(conf));
1.47 schwarze 313:
314: /*
1.98 schwarze 315: * We accept a few different invocations.
1.47 schwarze 316: * The CHECKOP macro makes sure that invocation styles don't
317: * clobber each other.
318: */
319: #define CHECKOP(_op, _ch) do \
320: if (OP_DEFAULT != (_op)) { \
1.151 schwarze 321: warnx("-%c: Conflicting option", (_ch)); \
1.47 schwarze 322: goto usage; \
323: } while (/*CONSTCOND*/0)
1.10 schwarze 324:
1.47 schwarze 325: path_arg = NULL;
1.28 schwarze 326: op = OP_DEFAULT;
1.1 schwarze 327:
1.83 schwarze 328: while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
1.1 schwarze 329: switch (ch) {
1.98 schwarze 330: case 'a':
1.6 schwarze 331: use_all = 1;
332: break;
1.98 schwarze 333: case 'C':
1.47 schwarze 334: CHECKOP(op, ch);
335: path_arg = optarg;
1.28 schwarze 336: op = OP_CONFFILE;
1.25 schwarze 337: break;
1.98 schwarze 338: case 'D':
1.82 schwarze 339: debug++;
340: break;
1.98 schwarze 341: case 'd':
1.47 schwarze 342: CHECKOP(op, ch);
343: path_arg = optarg;
1.2 schwarze 344: op = OP_UPDATE;
1.1 schwarze 345: break;
1.98 schwarze 346: case 'n':
1.47 schwarze 347: nodb = 1;
348: break;
1.98 schwarze 349: case 'p':
1.83 schwarze 350: warnings = 1;
351: break;
1.98 schwarze 352: case 'Q':
1.73 schwarze 353: mparse_options |= MPARSE_QUICK;
1.59 schwarze 354: break;
1.98 schwarze 355: case 'T':
1.52 schwarze 356: if (strcmp(optarg, "utf8")) {
1.151 schwarze 357: warnx("-T%s: Unsupported output format",
358: optarg);
1.52 schwarze 359: goto usage;
360: }
361: write_utf8 = 1;
362: break;
1.98 schwarze 363: case 't':
1.47 schwarze 364: CHECKOP(op, ch);
1.28 schwarze 365: dup2(STDOUT_FILENO, STDERR_FILENO);
366: op = OP_TEST;
1.47 schwarze 367: nodb = warnings = 1;
1.28 schwarze 368: break;
1.98 schwarze 369: case 'u':
1.47 schwarze 370: CHECKOP(op, ch);
371: path_arg = optarg;
1.1 schwarze 372: op = OP_DELETE;
373: break;
1.98 schwarze 374: case 'v':
1.83 schwarze 375: /* Compatibility with espie@'s makewhatis. */
1.28 schwarze 376: break;
1.1 schwarze 377: default:
1.28 schwarze 378: goto usage;
1.1 schwarze 379: }
380:
381: argc -= optind;
382: argv += optind;
383:
1.163 schwarze 384: if (nodb) {
385: if (pledge("stdio rpath", NULL) == -1) {
1.167 schwarze 386: warn("pledge");
1.163 schwarze 387: return (int)MANDOCLEVEL_SYSERR;
388: }
1.155 schwarze 389: }
390:
1.28 schwarze 391: if (OP_CONFFILE == op && argc > 0) {
1.151 schwarze 392: warnx("-C: Too many arguments");
1.28 schwarze 393: goto usage;
394: }
395:
1.47 schwarze 396: exitcode = (int)MANDOCLEVEL_OK;
1.157 schwarze 397: mchars_alloc();
1.202 schwarze 398: mp = mparse_alloc(mparse_options, MANDOCERR_MAX, NULL,
1.200 schwarze 399: MANDOC_OS_OTHER, NULL);
1.156 schwarze 400: mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
401: mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
1.2 schwarze 402:
1.47 schwarze 403: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.2 schwarze 404:
1.47 schwarze 405: /*
1.109 schwarze 406: * Most of these deal with a specific directory.
1.97 schwarze 407: * Jump into that directory first.
1.47 schwarze 408: */
1.120 schwarze 409: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.47 schwarze 410: goto out;
1.97 schwarze 411:
1.174 schwarze 412: dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
413: if (dba != NULL) {
1.97 schwarze 414: /*
415: * The existing database is usable. Process
416: * all files specified on the command-line.
417: */
418: use_all = 1;
1.95 schwarze 419: for (i = 0; i < argc; i++)
420: filescan(argv[i]);
1.174 schwarze 421: if (nodb == 0)
1.173 schwarze 422: dbprune(dba);
1.95 schwarze 423: } else {
1.177 schwarze 424: /* Database missing or corrupt. */
1.179 schwarze 425: if (op != OP_UPDATE || errno != ENOENT)
426: say(MANDOC_DB, "%s: Automatically recreating"
427: " from scratch", strerror(errno));
1.97 schwarze 428: exitcode = (int)MANDOCLEVEL_OK;
1.95 schwarze 429: op = OP_DEFAULT;
430: if (0 == treescan())
431: goto out;
1.173 schwarze 432: dba = dba_new(128);
1.95 schwarze 433: }
1.47 schwarze 434: if (OP_DELETE != op)
1.173 schwarze 435: mpages_merge(dba, mp);
436: if (nodb == 0)
437: dbwrite(dba);
438: dba_free(dba);
1.47 schwarze 439: } else {
440: /*
441: * If we have arguments, use them as our manpaths.
1.181 schwarze 442: * If we don't, use man.conf(5).
1.47 schwarze 443: */
444: if (argc > 0) {
1.140 schwarze 445: conf.manpath.paths = mandoc_reallocarray(NULL,
1.101 schwarze 446: argc, sizeof(char *));
1.140 schwarze 447: conf.manpath.sz = (size_t)argc;
1.47 schwarze 448: for (i = 0; i < argc; i++)
1.140 schwarze 449: conf.manpath.paths[i] = mandoc_strdup(argv[i]);
1.47 schwarze 450: } else
1.140 schwarze 451: manconf_parse(&conf, path_arg, NULL, NULL);
1.84 schwarze 452:
1.140 schwarze 453: if (conf.manpath.sz == 0) {
1.84 schwarze 454: exitcode = (int)MANDOCLEVEL_BADARG;
455: say("", "Empty manpath");
456: }
1.2 schwarze 457:
1.47 schwarze 458: /*
459: * First scan the tree rooted at a base directory, then
460: * build a new database and finally move it into place.
461: * Ignore zero-length directories and strip trailing
462: * slashes.
463: */
1.140 schwarze 464: for (j = 0; j < conf.manpath.sz; j++) {
465: sz = strlen(conf.manpath.paths[j]);
466: if (sz && conf.manpath.paths[j][sz - 1] == '/')
467: conf.manpath.paths[j][--sz] = '\0';
1.47 schwarze 468: if (0 == sz)
469: continue;
1.2 schwarze 470:
1.47 schwarze 471: if (j) {
1.156 schwarze 472: mandoc_ohash_init(&mpages, 6,
473: offsetof(struct mpage, inodev));
474: mandoc_ohash_init(&mlinks, 6,
475: offsetof(struct mlink, file));
1.47 schwarze 476: }
1.1 schwarze 477:
1.140 schwarze 478: if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
1.120 schwarze 479: continue;
1.47 schwarze 480: if (0 == treescan())
1.120 schwarze 481: continue;
1.173 schwarze 482: dba = dba_new(128);
483: mpages_merge(dba, mp);
484: if (nodb == 0)
485: dbwrite(dba);
486: dba_free(dba);
1.28 schwarze 487:
1.140 schwarze 488: if (j + 1 < conf.manpath.sz) {
1.47 schwarze 489: mpages_free();
490: ohash_delete(&mpages);
491: ohash_delete(&mlinks);
492: }
1.2 schwarze 493: }
1.47 schwarze 494: }
495: out:
1.140 schwarze 496: manconf_free(&conf);
1.123 schwarze 497: mparse_free(mp);
1.157 schwarze 498: mchars_free();
1.47 schwarze 499: mpages_free();
500: ohash_delete(&mpages);
501: ohash_delete(&mlinks);
1.150 schwarze 502: return exitcode;
1.47 schwarze 503: usage:
1.161 schwarze 504: progname = getprogname();
1.83 schwarze 505: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
506: " %s [-aDnpQ] [-Tutf8] dir ...\n"
507: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
508: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 509: " %s [-Q] -t file ...\n",
1.161 schwarze 510: progname, progname, progname, progname, progname);
1.1 schwarze 511:
1.150 schwarze 512: return (int)MANDOCLEVEL_BADARG;
1.47 schwarze 513: }
1.1 schwarze 514:
1.47 schwarze 515: /*
1.180 schwarze 516: * To get a singly linked list in alpha order while inserting entries
517: * at the beginning, process directory entries in reverse alpha order.
518: */
519: static int
520: fts_compare(const FTSENT **a, const FTSENT **b)
521: {
522: return -strcmp((*a)->fts_name, (*b)->fts_name);
523: }
524:
525: /*
1.47 schwarze 526: * Scan a directory tree rooted at "basedir" for manpages.
527: * We use fts(), scanning directory parts along the way for clues to our
528: * section and architecture.
529: *
530: * If use_all has been specified, grok all files.
531: * If not, sanitise paths to the following:
532: *
1.98 schwarze 533: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 534: * or
535: * [./]cat<section>[/<arch>]/<name>.0
536: *
1.169 krw 537: * TODO: accommodate for multi-language directories.
1.47 schwarze 538: */
539: static int
540: treescan(void)
541: {
1.96 schwarze 542: char buf[PATH_MAX];
1.47 schwarze 543: FTS *f;
544: FTSENT *ff;
545: struct mlink *mlink;
1.173 schwarze 546: int gzip;
547: enum form dform;
1.51 schwarze 548: char *dsec, *arch, *fsec, *cp;
549: const char *path;
1.47 schwarze 550: const char *argv[2];
1.1 schwarze 551:
1.47 schwarze 552: argv[0] = ".";
1.193 schwarze 553: argv[1] = NULL;
1.1 schwarze 554:
1.180 schwarze 555: f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR,
556: fts_compare);
1.153 schwarze 557: if (f == NULL) {
1.47 schwarze 558: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 559: say("", "&fts_open");
1.150 schwarze 560: return 0;
1.47 schwarze 561: }
1.2 schwarze 562:
1.47 schwarze 563: dsec = arch = NULL;
564: dform = FORM_NONE;
1.2 schwarze 565:
1.153 schwarze 566: while ((ff = fts_read(f)) != NULL) {
1.47 schwarze 567: path = ff->fts_path + 2;
1.96 schwarze 568: switch (ff->fts_info) {
569:
570: /*
571: * Symbolic links require various sanity checks,
572: * then get handled just like regular files.
573: */
1.98 schwarze 574: case FTS_SL:
1.153 schwarze 575: if (realpath(path, buf) == NULL) {
1.96 schwarze 576: if (warnings)
577: say(path, "&realpath");
578: continue;
579: }
580: if (strstr(buf, basedir) != buf) {
581: if (warnings) say("",
582: "%s: outside base directory", buf);
583: continue;
584: }
585: /* Use logical inode to avoid mpages dupe. */
1.153 schwarze 586: if (stat(path, ff->fts_statp) == -1) {
1.96 schwarze 587: if (warnings)
588: say(path, "&stat");
589: continue;
590: }
591: /* FALLTHROUGH */
592:
1.15 schwarze 593: /*
1.47 schwarze 594: * If we're a regular file, add an mlink by using the
595: * stored directory data and handling the filename.
1.15 schwarze 596: */
1.98 schwarze 597: case FTS_F:
1.153 schwarze 598: if ( ! strcmp(path, MANDOC_DB))
1.47 schwarze 599: continue;
600: if ( ! use_all && ff->fts_level < 2) {
601: if (warnings)
602: say(path, "Extraneous file");
603: continue;
1.81 schwarze 604: }
605: gzip = 0;
606: fsec = NULL;
1.153 schwarze 607: while (fsec == NULL) {
1.81 schwarze 608: fsec = strrchr(ff->fts_name, '.');
1.153 schwarze 609: if (fsec == NULL || strcmp(fsec+1, "gz"))
1.81 schwarze 610: break;
611: gzip = 1;
612: *fsec = '\0';
613: fsec = NULL;
614: }
1.153 schwarze 615: if (fsec == NULL) {
1.47 schwarze 616: if ( ! use_all) {
617: if (warnings)
618: say(path,
619: "No filename suffix");
620: continue;
621: }
1.153 schwarze 622: } else if ( ! strcmp(++fsec, "html")) {
1.47 schwarze 623: if (warnings)
624: say(path, "Skip html");
625: continue;
1.153 schwarze 626: } else if ( ! strcmp(fsec, "ps")) {
1.47 schwarze 627: if (warnings)
628: say(path, "Skip ps");
629: continue;
1.153 schwarze 630: } else if ( ! strcmp(fsec, "pdf")) {
1.47 schwarze 631: if (warnings)
632: say(path, "Skip pdf");
633: continue;
634: } else if ( ! use_all &&
1.153 schwarze 635: ((dform == FORM_SRC &&
1.138 schwarze 636: strncmp(fsec, dsec, strlen(dsec))) ||
1.153 schwarze 637: (dform == FORM_CAT && strcmp(fsec, "0")))) {
1.47 schwarze 638: if (warnings)
639: say(path, "Wrong filename suffix");
640: continue;
641: } else
642: fsec[-1] = '\0';
1.51 schwarze 643:
1.47 schwarze 644: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.100 schwarze 645: if (strlcpy(mlink->file, path,
646: sizeof(mlink->file)) >=
647: sizeof(mlink->file)) {
648: say(path, "Filename too long");
649: free(mlink);
650: continue;
651: }
1.47 schwarze 652: mlink->dform = dform;
1.51 schwarze 653: mlink->dsec = dsec;
654: mlink->arch = arch;
655: mlink->name = ff->fts_name;
656: mlink->fsec = fsec;
1.81 schwarze 657: mlink->gzip = gzip;
1.47 schwarze 658: mlink_add(mlink, ff->fts_statp);
659: continue;
1.96 schwarze 660:
1.98 schwarze 661: case FTS_D:
662: case FTS_DP:
1.96 schwarze 663: break;
664:
665: default:
1.47 schwarze 666: if (warnings)
667: say(path, "Not a regular file");
668: continue;
669: }
670:
671: switch (ff->fts_level) {
1.98 schwarze 672: case 0:
1.47 schwarze 673: /* Ignore the root directory. */
674: break;
1.98 schwarze 675: case 1:
1.47 schwarze 676: /*
677: * This might contain manX/ or catX/.
678: * Try to infer this from the name.
679: * If we're not in use_all, enforce it.
680: */
681: cp = ff->fts_name;
1.153 schwarze 682: if (ff->fts_info == FTS_DP) {
683: dform = FORM_NONE;
684: dsec = NULL;
1.47 schwarze 685: break;
1.153 schwarze 686: }
1.15 schwarze 687:
1.153 schwarze 688: if ( ! strncmp(cp, "man", 3)) {
1.47 schwarze 689: dform = FORM_SRC;
690: dsec = cp + 3;
1.153 schwarze 691: } else if ( ! strncmp(cp, "cat", 3)) {
1.47 schwarze 692: dform = FORM_CAT;
693: dsec = cp + 3;
1.51 schwarze 694: } else {
695: dform = FORM_NONE;
696: dsec = NULL;
1.26 schwarze 697: }
1.47 schwarze 698:
1.153 schwarze 699: if (dsec != NULL || use_all)
1.47 schwarze 700: break;
701:
702: if (warnings)
703: say(path, "Unknown directory part");
704: fts_set(f, ff, FTS_SKIP);
705: break;
1.98 schwarze 706: case 2:
1.47 schwarze 707: /*
708: * Possibly our architecture.
709: * If we're descending, keep tabs on it.
710: */
1.153 schwarze 711: if (ff->fts_info != FTS_DP && dsec != NULL)
1.47 schwarze 712: arch = ff->fts_name;
1.51 schwarze 713: else
714: arch = NULL;
1.47 schwarze 715: break;
716: default:
1.153 schwarze 717: if (ff->fts_info == FTS_DP || use_all)
1.47 schwarze 718: break;
719: if (warnings)
720: say(path, "Extraneous directory part");
721: fts_set(f, ff, FTS_SKIP);
722: break;
1.14 schwarze 723: }
1.47 schwarze 724: }
725:
726: fts_close(f);
1.150 schwarze 727: return 1;
1.47 schwarze 728: }
1.1 schwarze 729:
1.47 schwarze 730: /*
731: * Add a file to the mlinks table.
732: * Do not verify that it's a "valid" looking manpage (we'll do that
733: * later).
734: *
735: * Try to infer the manual section, architecture, and page name from the
736: * path, assuming it looks like
737: *
1.98 schwarze 738: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 739: * or
740: * [./]cat<section>[/<arch>]/<name>.0
741: *
742: * See treescan() for the fts(3) version of this.
743: */
744: static void
745: filescan(const char *file)
746: {
747: char buf[PATH_MAX];
748: struct stat st;
749: struct mlink *mlink;
750: char *p, *start;
751:
752: assert(use_all);
753:
754: if (0 == strncmp(file, "./", 2))
755: file += 2;
756:
1.96 schwarze 757: /*
758: * We have to do lstat(2) before realpath(3) loses
759: * the information whether this is a symbolic link.
760: * We need to know that because for symbolic links,
761: * we want to use the orginal file name, while for
762: * regular files, we want to use the real path.
763: */
764: if (-1 == lstat(file, &st)) {
765: exitcode = (int)MANDOCLEVEL_BADARG;
766: say(file, "&lstat");
767: return;
768: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
769: exitcode = (int)MANDOCLEVEL_BADARG;
770: say(file, "Not a regular file");
771: return;
772: }
773:
774: /*
775: * We have to resolve the file name to the real path
776: * in any case for the base directory check.
777: */
1.47 schwarze 778: if (NULL == realpath(file, buf)) {
779: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 780: say(file, "&realpath");
1.47 schwarze 781: return;
1.63 schwarze 782: }
783:
1.109 schwarze 784: if (OP_TEST == op)
1.63 schwarze 785: start = buf;
1.109 schwarze 786: else if (strstr(buf, basedir) == buf)
787: start = buf + strlen(basedir);
1.63 schwarze 788: else {
1.47 schwarze 789: exitcode = (int)MANDOCLEVEL_BADARG;
790: say("", "%s: outside base directory", buf);
791: return;
1.63 schwarze 792: }
793:
1.96 schwarze 794: /*
795: * Now we are sure the file is inside our tree.
796: * If it is a symbolic link, ignore the real path
797: * and use the original name.
798: * This implies passing stuff like "cat1/../man1/foo.1"
799: * on the command line won't work. So don't do that.
800: * Note the stat(2) can still fail if the link target
801: * doesn't exist.
802: */
803: if (S_IFLNK & st.st_mode) {
804: if (-1 == stat(buf, &st)) {
805: exitcode = (int)MANDOCLEVEL_BADARG;
806: say(file, "&stat");
807: return;
808: }
1.100 schwarze 809: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
810: say(file, "Filename too long");
811: return;
812: }
1.109 schwarze 813: start = buf;
814: if (OP_TEST != op && strstr(buf, basedir) == buf)
815: start += strlen(basedir);
1.1 schwarze 816: }
1.63 schwarze 817:
1.47 schwarze 818: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 819: mlink->dform = FORM_NONE;
1.100 schwarze 820: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
821: sizeof(mlink->file)) {
822: say(start, "Filename too long");
1.134 schwarze 823: free(mlink);
1.100 schwarze 824: return;
1.188 schwarze 825: }
826:
827: /*
828: * In test mode or when the original name is absolute
829: * but outside our tree, guess the base directory.
830: */
831:
832: if (op == OP_TEST || (start == buf && *start == '/')) {
833: if (strncmp(buf, "man/", 4) == 0)
834: start = buf + 4;
835: else if ((start = strstr(buf, "/man/")) != NULL)
836: start += 5;
837: else
838: start = buf;
1.100 schwarze 839: }
1.1 schwarze 840:
1.10 schwarze 841: /*
1.47 schwarze 842: * First try to guess our directory structure.
843: * If we find a separator, try to look for man* or cat*.
844: * If we find one of these and what's underneath is a directory,
845: * assume it's an architecture.
1.10 schwarze 846: */
1.47 schwarze 847: if (NULL != (p = strchr(start, '/'))) {
848: *p++ = '\0';
849: if (0 == strncmp(start, "man", 3)) {
850: mlink->dform = FORM_SRC;
1.51 schwarze 851: mlink->dsec = start + 3;
1.47 schwarze 852: } else if (0 == strncmp(start, "cat", 3)) {
853: mlink->dform = FORM_CAT;
1.51 schwarze 854: mlink->dsec = start + 3;
1.47 schwarze 855: }
1.10 schwarze 856:
1.47 schwarze 857: start = p;
858: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
859: *p++ = '\0';
1.51 schwarze 860: mlink->arch = start;
1.47 schwarze 861: start = p;
1.41 deraadt 862: }
1.47 schwarze 863: }
1.7 schwarze 864:
1.47 schwarze 865: /*
866: * Now check the file suffix.
867: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
868: */
869: p = strrchr(start, '\0');
870: while (p-- > start && '/' != *p && '.' != *p)
871: /* Loop. */ ;
872:
873: if ('.' == *p) {
874: *p++ = '\0';
1.51 schwarze 875: mlink->fsec = p;
1.47 schwarze 876: }
1.41 deraadt 877:
1.47 schwarze 878: /*
879: * Now try to parse the name.
880: * Use the filename portion of the path.
881: */
882: mlink->name = start;
883: if (NULL != (p = strrchr(start, '/'))) {
884: mlink->name = p + 1;
885: *p = '\0';
886: }
887: mlink_add(mlink, &st);
888: }
1.2 schwarze 889:
1.47 schwarze 890: static void
891: mlink_add(struct mlink *mlink, const struct stat *st)
892: {
893: struct inodev inodev;
894: struct mpage *mpage;
895: unsigned int slot;
896:
897: assert(NULL != mlink->file);
898:
1.51 schwarze 899: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
900: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
901: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
902: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 903:
904: if ('0' == *mlink->fsec) {
905: free(mlink->fsec);
906: mlink->fsec = mandoc_strdup(mlink->dsec);
907: mlink->fform = FORM_CAT;
908: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
909: mlink->fform = FORM_SRC;
910: else
911: mlink->fform = FORM_NONE;
1.1 schwarze 912:
1.47 schwarze 913: slot = ohash_qlookup(&mlinks, mlink->file);
914: assert(NULL == ohash_find(&mlinks, slot));
915: ohash_insert(&mlinks, slot, mlink);
916:
1.139 schwarze 917: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.47 schwarze 918: inodev.st_ino = st->st_ino;
919: inodev.st_dev = st->st_dev;
920: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
921: sizeof(struct inodev), inodev.st_ino);
922: mpage = ohash_find(&mpages, slot);
923: if (NULL == mpage) {
924: mpage = mandoc_calloc(1, sizeof(struct mpage));
925: mpage->inodev.st_ino = inodev.st_ino;
926: mpage->inodev.st_dev = inodev.st_dev;
1.173 schwarze 927: mpage->form = FORM_NONE;
1.180 schwarze 928: mpage->next = mpage_head;
929: mpage_head = mpage;
1.47 schwarze 930: ohash_insert(&mpages, slot, mpage);
931: } else
932: mlink->next = mpage->mlinks;
933: mpage->mlinks = mlink;
1.75 schwarze 934: mlink->mpage = mpage;
1.47 schwarze 935: }
1.1 schwarze 936:
1.47 schwarze 937: static void
938: mlink_free(struct mlink *mlink)
939: {
1.1 schwarze 940:
1.47 schwarze 941: free(mlink->dsec);
942: free(mlink->arch);
943: free(mlink->name);
944: free(mlink->fsec);
945: free(mlink);
946: }
1.1 schwarze 947:
1.47 schwarze 948: static void
949: mpages_free(void)
950: {
951: struct mpage *mpage;
952: struct mlink *mlink;
953:
1.180 schwarze 954: while ((mpage = mpage_head) != NULL) {
955: while ((mlink = mpage->mlinks) != NULL) {
1.47 schwarze 956: mpage->mlinks = mlink->next;
957: mlink_free(mlink);
958: }
1.180 schwarze 959: mpage_head = mpage->next;
1.47 schwarze 960: free(mpage->sec);
961: free(mpage->arch);
962: free(mpage->title);
963: free(mpage->desc);
964: free(mpage);
965: }
966: }
1.1 schwarze 967:
1.47 schwarze 968: /*
969: * For each mlink to the mpage, check whether the path looks like
970: * it is formatted, and if it does, check whether a source manual
971: * exists by the same name, ignoring the suffix.
972: * If both conditions hold, drop the mlink.
973: */
974: static void
975: mlinks_undupe(struct mpage *mpage)
976: {
977: char buf[PATH_MAX];
978: struct mlink **prev;
979: struct mlink *mlink;
980: char *bufp;
981:
982: mpage->form = FORM_CAT;
983: prev = &mpage->mlinks;
984: while (NULL != (mlink = *prev)) {
985: if (FORM_CAT != mlink->dform) {
986: mpage->form = FORM_NONE;
987: goto nextlink;
988: }
1.100 schwarze 989: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 990: bufp = strstr(buf, "cat");
991: assert(NULL != bufp);
992: memcpy(bufp, "man", 3);
993: if (NULL != (bufp = strrchr(buf, '.')))
994: *++bufp = '\0';
1.100 schwarze 995: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 996: if (NULL == ohash_find(&mlinks,
1.98 schwarze 997: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 998: goto nextlink;
999: if (warnings)
1000: say(mlink->file, "Man source exists: %s", buf);
1001: if (use_all)
1002: goto nextlink;
1003: *prev = mlink->next;
1004: mlink_free(mlink);
1005: continue;
1006: nextlink:
1007: prev = &(*prev)->next;
1.1 schwarze 1008: }
1.47 schwarze 1009: }
1.1 schwarze 1010:
1.88 schwarze 1011: static void
1.50 schwarze 1012: mlink_check(struct mpage *mpage, struct mlink *mlink)
1013: {
1.88 schwarze 1014: struct str *str;
1015: unsigned int slot;
1.50 schwarze 1016:
1017: /*
1018: * Check whether the manual section given in a file
1019: * agrees with the directory where the file is located.
1020: * Some manuals have suffixes like (3p) on their
1021: * section number either inside the file or in the
1022: * directory name, some are linked into more than one
1023: * section, like encrypt(1) = makekey(8).
1024: */
1025:
1026: if (FORM_SRC == mpage->form &&
1.88 schwarze 1027: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1028: say(mlink->file, "Section \"%s\" manual in %s directory",
1029: mpage->sec, mlink->dsec);
1030:
1031: /*
1032: * Manual page directories exist for each kernel
1033: * architecture as returned by machine(1).
1034: * However, many manuals only depend on the
1035: * application architecture as returned by arch(1).
1036: * For example, some (2/ARM) manuals are shared
1037: * across the "armish" and "zaurus" kernel
1038: * architectures.
1039: * A few manuals are even shared across completely
1040: * different architectures, for example fdformat(1)
1.178 schwarze 1041: * on amd64, i386, and sparc64.
1.50 schwarze 1042: */
1043:
1.88 schwarze 1044: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1045: say(mlink->file, "Architecture \"%s\" manual in "
1046: "\"%s\" directory", mpage->arch, mlink->arch);
1047:
1.88 schwarze 1048: /*
1049: * XXX
1.90 schwarze 1050: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1051: */
1052:
1053: if (FORM_CAT == mpage->form)
1054: return;
1055:
1056: /*
1057: * Check whether this mlink
1058: * appears as a name in the NAME section.
1059: */
1.50 schwarze 1060:
1.90 schwarze 1061: slot = ohash_qlookup(&names, mlink->name);
1062: str = ohash_find(&names, slot);
1.88 schwarze 1063: assert(NULL != str);
1.90 schwarze 1064: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1065: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1066: }
1067:
1.47 schwarze 1068: /*
1069: * Run through the files in the global vector "mpages"
1070: * and add them to the database specified in "basedir".
1071: *
1072: * This handles the parsing scheme itself, using the cues of directory
1073: * and filename to determine whether the file is parsable or not.
1074: */
1075: static void
1.173 schwarze 1076: mpages_merge(struct dba *dba, struct mparse *mp)
1.47 schwarze 1077: {
1.180 schwarze 1078: struct mpage *mpage, *mpage_dest;
1.75 schwarze 1079: struct mlink *mlink, *mlink_dest;
1.144 schwarze 1080: struct roff_man *man;
1.75 schwarze 1081: char *sodest;
1.69 schwarze 1082: char *cp;
1.117 schwarze 1083: int fd;
1.64 schwarze 1084:
1.180 schwarze 1085: for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
1.47 schwarze 1086: mlinks_undupe(mpage);
1.180 schwarze 1087: if ((mlink = mpage->mlinks) == NULL)
1088: continue;
1.1 schwarze 1089:
1.90 schwarze 1090: name_mask = NAME_MASK;
1.156 schwarze 1091: mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1092: mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1.47 schwarze 1093: mparse_reset(mp);
1094: man = NULL;
1.81 schwarze 1095: sodest = NULL;
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;
1.145 schwarze 1110: mparse_result(mp, &man, &sodest);
1.47 schwarze 1111: }
1.11 schwarze 1112:
1.126 schwarze 1113: if (sodest != NULL) {
1.75 schwarze 1114: mlink_dest = ohash_find(&mlinks,
1115: ohash_qlookup(&mlinks, sodest));
1.126 schwarze 1116: if (mlink_dest == NULL) {
1117: mandoc_asprintf(&cp, "%s.gz", sodest);
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;
1152: }
1.81 schwarze 1153: goto nextpage;
1.145 schwarze 1154: } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1.159 schwarze 1155: mdoc_validate(man);
1.47 schwarze 1156: mpage->form = FORM_SRC;
1.146 schwarze 1157: mpage->sec = man->meta.msec;
1.112 schwarze 1158: mpage->sec = mandoc_strdup(
1.126 schwarze 1159: mpage->sec == NULL ? "" : mpage->sec);
1.146 schwarze 1160: mpage->arch = man->meta.arch;
1.47 schwarze 1161: mpage->arch = mandoc_strdup(
1.126 schwarze 1162: mpage->arch == NULL ? "" : mpage->arch);
1.146 schwarze 1163: mpage->title = mandoc_strdup(man->meta.title);
1.145 schwarze 1164: } else if (man != NULL && man->macroset == MACROSET_MAN) {
1.160 schwarze 1165: man_validate(man);
1.189 schwarze 1166: if (*man->meta.msec != '\0' ||
1.194 schwarze 1167: *man->meta.title != '\0') {
1.189 schwarze 1168: mpage->form = FORM_SRC;
1169: mpage->sec = mandoc_strdup(man->meta.msec);
1170: mpage->arch = mandoc_strdup(mlink->arch);
1171: mpage->title = mandoc_strdup(man->meta.title);
1172: } else
1173: man = NULL;
1174: }
1175:
1176: assert(mpage->desc == NULL);
1177: if (man == NULL) {
1.47 schwarze 1178: mpage->form = FORM_CAT;
1.136 schwarze 1179: mpage->sec = mandoc_strdup(mlink->dsec);
1180: mpage->arch = mandoc_strdup(mlink->arch);
1181: mpage->title = mandoc_strdup(mlink->name);
1.189 schwarze 1182: parse_cat(mpage, fd);
1183: } else if (man->macroset == MACROSET_MDOC)
1.146 schwarze 1184: parse_mdoc(mpage, &man->meta, man->first);
1.189 schwarze 1185: else
1.146 schwarze 1186: parse_man(mpage, &man->meta, man->first);
1.191 schwarze 1187: if (mpage->desc == NULL) {
1188: mpage->desc = mandoc_strdup(mlink->name);
1189: if (warnings)
1190: say(mlink->file, "No one-line description, "
1191: "using filename \"%s\"", mlink->name);
1192: }
1.88 schwarze 1193:
1.187 schwarze 1194: for (mlink = mpage->mlinks;
1195: mlink != NULL;
1196: mlink = mlink->next) {
1197: putkey(mpage, mlink->name, NAME_FILE);
1198: if (warnings && !use_all)
1.88 schwarze 1199: mlink_check(mpage, mlink);
1.187 schwarze 1200: }
1.6 schwarze 1201:
1.173 schwarze 1202: dbadd(dba, mpage);
1.81 schwarze 1203:
1204: nextpage:
1.47 schwarze 1205: ohash_delete(&strings);
1.90 schwarze 1206: ohash_delete(&names);
1.47 schwarze 1207: }
1208: }
1.6 schwarze 1209:
1.47 schwarze 1210: static void
1.81 schwarze 1211: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1212: {
1213: FILE *stream;
1.190 schwarze 1214: struct mlink *mlink;
1215: char *line, *p, *title, *sec;
1.162 schwarze 1216: size_t linesz, plen, titlesz;
1217: ssize_t len;
1218: int offs;
1.1 schwarze 1219:
1.190 schwarze 1220: mlink = mpage->mlinks;
1221: stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1222: if (stream == NULL) {
1223: if (fd != -1)
1.114 doug 1224: close(fd);
1.47 schwarze 1225: if (warnings)
1.190 schwarze 1226: say(mlink->file, "&fopen");
1.47 schwarze 1227: return;
1228: }
1.1 schwarze 1229:
1.162 schwarze 1230: line = NULL;
1231: linesz = 0;
1232:
1.190 schwarze 1233: /* Parse the section number from the header line. */
1234:
1235: while (getline(&line, &linesz, stream) != -1) {
1236: if (*line == '\n')
1237: continue;
1238: if ((sec = strchr(line, '(')) == NULL)
1239: break;
1240: if ((p = strchr(++sec, ')')) == NULL)
1241: break;
1242: free(mpage->sec);
1243: mpage->sec = mandoc_strndup(sec, p - sec);
1244: if (warnings && *mlink->dsec != '\0' &&
1245: strcasecmp(mpage->sec, mlink->dsec))
1246: say(mlink->file,
1247: "Section \"%s\" manual in %s directory",
1248: mpage->sec, mlink->dsec);
1249: break;
1250: }
1251:
1.47 schwarze 1252: /* Skip to first blank line. */
1.1 schwarze 1253:
1.190 schwarze 1254: while (line == NULL || *line != '\n')
1255: if (getline(&line, &linesz, stream) == -1)
1.47 schwarze 1256: break;
1.1 schwarze 1257:
1.47 schwarze 1258: /*
1259: * Assume the first line that is not indented
1260: * is the first section header. Skip to it.
1261: */
1.1 schwarze 1262:
1.162 schwarze 1263: while (getline(&line, &linesz, stream) != -1)
1264: if (*line != '\n' && *line != ' ')
1.47 schwarze 1265: break;
1.98 schwarze 1266:
1.47 schwarze 1267: /*
1268: * Read up until the next section into a buffer.
1269: * Strip the leading and trailing newline from each read line,
1270: * appending a trailing space.
1271: * Ignore empty (whitespace-only) lines.
1272: */
1.28 schwarze 1273:
1.47 schwarze 1274: titlesz = 0;
1275: title = NULL;
1.38 schwarze 1276:
1.162 schwarze 1277: while ((len = getline(&line, &linesz, stream)) != -1) {
1278: if (*line != ' ')
1.47 schwarze 1279: break;
1.162 schwarze 1280: offs = 0;
1281: while (isspace((unsigned char)line[offs]))
1282: offs++;
1283: if (line[offs] == '\0')
1.47 schwarze 1284: continue;
1.162 schwarze 1285: title = mandoc_realloc(title, titlesz + len - offs);
1286: memcpy(title + titlesz, line + offs, len - offs);
1287: titlesz += len - offs;
1.47 schwarze 1288: title[titlesz - 1] = ' ';
1289: }
1.162 schwarze 1290: free(line);
1.28 schwarze 1291:
1.47 schwarze 1292: /*
1293: * If no page content can be found, or the input line
1294: * is already the next section header, or there is no
1295: * trailing newline, reuse the page title as the page
1296: * description.
1297: */
1.1 schwarze 1298:
1.47 schwarze 1299: if (NULL == title || '\0' == *title) {
1300: if (warnings)
1.190 schwarze 1301: say(mlink->file, "Cannot find NAME section");
1.47 schwarze 1302: fclose(stream);
1303: free(title);
1304: return;
1305: }
1.24 schwarze 1306:
1.162 schwarze 1307: title[titlesz - 1] = '\0';
1.24 schwarze 1308:
1.47 schwarze 1309: /*
1310: * Skip to the first dash.
1311: * Use the remaining line as the description (no more than 70
1312: * bytes).
1313: */
1.28 schwarze 1314:
1.47 schwarze 1315: if (NULL != (p = strstr(title, "- "))) {
1316: for (p += 2; ' ' == *p || '\b' == *p; p++)
1317: /* Skip to next word. */ ;
1318: } else {
1319: if (warnings)
1.191 schwarze 1320: say(mlink->file, "No dash in title line, "
1321: "reusing \"%s\" as one-line description", title);
1.47 schwarze 1322: p = title;
1323: }
1.1 schwarze 1324:
1.47 schwarze 1325: plen = strlen(p);
1.1 schwarze 1326:
1.47 schwarze 1327: /* Strip backspace-encoding from line. */
1.1 schwarze 1328:
1.47 schwarze 1329: while (NULL != (line = memchr(p, '\b', plen))) {
1330: len = line - p;
1331: if (0 == len) {
1332: memmove(line, line + 1, plen--);
1333: continue;
1.98 schwarze 1334: }
1.47 schwarze 1335: memmove(line - 1, line + 1, plen - len);
1336: plen -= 2;
1337: }
1.1 schwarze 1338:
1.206 schwarze 1339: /*
1340: * Cut off excessive one-line descriptions.
1341: * Bad pages are not worth better heuristics.
1342: */
1343:
1344: mpage->desc = mandoc_strndup(p, 150);
1.47 schwarze 1345: fclose(stream);
1346: free(title);
1347: }
1.16 schwarze 1348:
1.47 schwarze 1349: /*
1350: * Put a type/word pair into the word database for this particular file.
1351: */
1352: static void
1.69 schwarze 1353: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1354: {
1355: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1356: }
1357:
1358: /*
1.47 schwarze 1359: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1360: */
1361: static void
1.47 schwarze 1362: putmdockey(const struct mpage *mpage,
1.184 schwarze 1363: const struct roff_node *n, uint64_t m, int taboo)
1.2 schwarze 1364: {
1.16 schwarze 1365:
1.47 schwarze 1366: for ( ; NULL != n; n = n->next) {
1.184 schwarze 1367: if (n->flags & taboo)
1368: continue;
1.47 schwarze 1369: if (NULL != n->child)
1.184 schwarze 1370: putmdockey(mpage, n->child, m, taboo);
1.141 schwarze 1371: if (n->type == ROFFT_TEXT)
1.47 schwarze 1372: putkey(mpage, n->string, m);
1373: }
1374: }
1.16 schwarze 1375:
1.47 schwarze 1376: static void
1.143 schwarze 1377: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1378: const struct roff_node *n)
1.47 schwarze 1379: {
1.142 schwarze 1380: const struct roff_node *head, *body;
1.78 schwarze 1381: char *start, *title;
1.47 schwarze 1382: char byte;
1.78 schwarze 1383: size_t sz;
1.16 schwarze 1384:
1.168 schwarze 1385: if (n == NULL)
1.47 schwarze 1386: return;
1.16 schwarze 1387:
1.47 schwarze 1388: /*
1389: * We're only searching for one thing: the first text child in
1390: * the BODY of a NAME section. Since we don't keep track of
1391: * sections in -man, run some hoops to find out whether we're in
1392: * the correct section or not.
1393: */
1.16 schwarze 1394:
1.141 schwarze 1395: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1396: body = n;
1.168 schwarze 1397: if ((head = body->parent->head) != NULL &&
1398: (head = head->child) != NULL &&
1399: head->next == NULL &&
1.141 schwarze 1400: head->type == ROFFT_TEXT &&
1.168 schwarze 1401: strcmp(head->string, "NAME") == 0 &&
1402: body->child != NULL) {
1.2 schwarze 1403:
1.47 schwarze 1404: /*
1405: * Suck the entire NAME section into memory.
1406: * Yes, we might run away.
1407: * But too many manuals have big, spread-out
1408: * NAME sections over many lines.
1409: */
1.2 schwarze 1410:
1.78 schwarze 1411: title = NULL;
1.147 schwarze 1412: deroff(&title, body);
1.47 schwarze 1413: if (NULL == title)
1414: return;
1.16 schwarze 1415:
1.98 schwarze 1416: /*
1.47 schwarze 1417: * Go through a special heuristic dance here.
1418: * Conventionally, one or more manual names are
1419: * comma-specified prior to a whitespace, then a
1420: * dash, then a description. Try to puzzle out
1421: * the name parts here.
1422: */
1.16 schwarze 1423:
1.78 schwarze 1424: start = title;
1.47 schwarze 1425: for ( ;; ) {
1426: sz = strcspn(start, " ,");
1427: if ('\0' == start[sz])
1428: break;
1.1 schwarze 1429:
1.47 schwarze 1430: byte = start[sz];
1431: start[sz] = '\0';
1.67 schwarze 1432:
1433: /*
1434: * Assume a stray trailing comma in the
1435: * name list if a name begins with a dash.
1436: */
1437:
1438: if ('-' == start[0] ||
1439: ('\\' == start[0] && '-' == start[1]))
1440: break;
1.1 schwarze 1441:
1.90 schwarze 1442: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1443: if ( ! (mpage->name_head_done ||
1444: strcasecmp(start, meta->title))) {
1445: putkey(mpage, start, NAME_HEAD);
1446: mpage->name_head_done = 1;
1447: }
1.1 schwarze 1448:
1.47 schwarze 1449: if (' ' == byte) {
1450: start += sz + 1;
1451: break;
1452: }
1.1 schwarze 1453:
1.47 schwarze 1454: assert(',' == byte);
1455: start += sz + 1;
1456: while (' ' == *start)
1457: start++;
1458: }
1.1 schwarze 1459:
1.78 schwarze 1460: if (start == title) {
1.90 schwarze 1461: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1462: if ( ! (mpage->name_head_done ||
1463: strcasecmp(start, meta->title))) {
1464: putkey(mpage, start, NAME_HEAD);
1465: mpage->name_head_done = 1;
1466: }
1.47 schwarze 1467: free(title);
1468: return;
1469: }
1.1 schwarze 1470:
1.47 schwarze 1471: while (isspace((unsigned char)*start))
1472: start++;
1.1 schwarze 1473:
1.47 schwarze 1474: if (0 == strncmp(start, "-", 1))
1475: start += 1;
1476: else if (0 == strncmp(start, "\\-\\-", 4))
1477: start += 4;
1478: else if (0 == strncmp(start, "\\-", 2))
1479: start += 2;
1480: else if (0 == strncmp(start, "\\(en", 4))
1481: start += 4;
1482: else if (0 == strncmp(start, "\\(em", 4))
1483: start += 4;
1.1 schwarze 1484:
1.47 schwarze 1485: while (' ' == *start)
1486: start++;
1.1 schwarze 1487:
1.206 schwarze 1488: /*
1489: * Cut off excessive one-line descriptions.
1490: * Bad pages are not worth better heuristics.
1491: */
1492:
1493: mpage->desc = mandoc_strndup(start, 150);
1.47 schwarze 1494: free(title);
1495: return;
1496: }
1497: }
1.1 schwarze 1498:
1.47 schwarze 1499: for (n = n->child; n; n = n->next) {
1500: if (NULL != mpage->desc)
1501: break;
1.129 schwarze 1502: parse_man(mpage, meta, n);
1.1 schwarze 1503: }
1504: }
1505:
1506: static void
1.143 schwarze 1507: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1508: const struct roff_node *n)
1.1 schwarze 1509: {
1.208 ! schwarze 1510: const struct mdoc_handler *handler;
1.1 schwarze 1511:
1.195 schwarze 1512: for (n = n->child; n != NULL; n = n->next) {
1.208 ! schwarze 1513: if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX)
1.184 schwarze 1514: continue;
1.196 schwarze 1515: assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1.208 ! schwarze 1516: handler = mdoc_handlers + (n->tok - MDOC_Dd);
! 1517: if (n->flags & handler->taboo)
! 1518: continue;
! 1519:
1.47 schwarze 1520: switch (n->type) {
1.141 schwarze 1521: case ROFFT_ELEM:
1522: case ROFFT_BLOCK:
1523: case ROFFT_HEAD:
1524: case ROFFT_BODY:
1525: case ROFFT_TAIL:
1.208 ! schwarze 1526: if (handler->fp != NULL &&
! 1527: (*handler->fp)(mpage, meta, n) == 0)
1.195 schwarze 1528: break;
1.208 ! schwarze 1529: if (handler->mask)
1.47 schwarze 1530: putmdockey(mpage, n->child,
1.208 ! schwarze 1531: handler->mask, handler->taboo);
1.47 schwarze 1532: break;
1533: default:
1534: continue;
1535: }
1536: if (NULL != n->child)
1.127 schwarze 1537: parse_mdoc(mpage, meta, n);
1.1 schwarze 1538: }
1539: }
1540:
1.19 schwarze 1541: static int
1.207 schwarze 1542: parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta,
1543: const struct roff_node *n)
1544: {
1545: uint64_t mask;
1546:
1547: mask = TYPE_Fa;
1548: if (n->sec == SEC_SYNOPSIS)
1549: mask |= TYPE_Vt;
1550:
1551: putmdockey(mpage, n->child, mask, 0);
1552: return 0;
1553: }
1554:
1555: static int
1.143 schwarze 1556: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1557: const struct roff_node *n)
1.1 schwarze 1558: {
1.131 schwarze 1559: char *start, *end;
1.1 schwarze 1560: size_t sz;
1.19 schwarze 1561:
1.47 schwarze 1562: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1563: NULL == (n = n->child) ||
1.141 schwarze 1564: n->type != ROFFT_TEXT)
1.150 schwarze 1565: return 0;
1.1 schwarze 1566:
1567: /*
1568: * Only consider those `Fd' macro fields that begin with an
1569: * "inclusion" token (versus, e.g., #define).
1570: */
1.47 schwarze 1571:
1.1 schwarze 1572: if (strcmp("#include", n->string))
1.150 schwarze 1573: return 0;
1.1 schwarze 1574:
1.141 schwarze 1575: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1576: return 0;
1.1 schwarze 1577:
1578: /*
1579: * Strip away the enclosing angle brackets and make sure we're
1580: * not zero-length.
1581: */
1582:
1583: start = n->string;
1584: if ('<' == *start || '"' == *start)
1585: start++;
1586:
1587: if (0 == (sz = strlen(start)))
1.150 schwarze 1588: return 0;
1.1 schwarze 1589:
1590: end = &start[(int)sz - 1];
1591: if ('>' == *end || '"' == *end)
1592: end--;
1593:
1.47 schwarze 1594: if (end > start)
1595: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1596: return 0;
1.1 schwarze 1597: }
1598:
1.133 schwarze 1599: static void
1.142 schwarze 1600: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1601: {
1.69 schwarze 1602: char *cp;
1.133 schwarze 1603: size_t sz;
1.1 schwarze 1604:
1.141 schwarze 1605: if (n->type != ROFFT_TEXT)
1.133 schwarze 1606: return;
1.19 schwarze 1607:
1.133 schwarze 1608: /* Skip function pointer punctuation. */
1.1 schwarze 1609:
1.133 schwarze 1610: cp = n->string;
1611: while (*cp == '(' || *cp == '*')
1.1 schwarze 1612: cp++;
1.133 schwarze 1613: sz = strcspn(cp, "()");
1.1 schwarze 1614:
1.133 schwarze 1615: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1616: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1617: putkeys(mpage, cp, sz, NAME_SYN);
1618: }
1.19 schwarze 1619:
1.133 schwarze 1620: static int
1.143 schwarze 1621: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1622: const struct roff_node *n)
1.133 schwarze 1623: {
1.207 schwarze 1624: uint64_t mask;
1.133 schwarze 1625:
1626: if (n->child == NULL)
1.150 schwarze 1627: return 0;
1.133 schwarze 1628:
1629: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1630:
1.207 schwarze 1631: n = n->child->next;
1632: if (n != NULL && n->type == ROFFT_TEXT) {
1633: mask = TYPE_Fa;
1634: if (n->sec == SEC_SYNOPSIS)
1635: mask |= TYPE_Vt;
1636: putmdockey(mpage, n, mask, 0);
1637: }
1.19 schwarze 1638:
1.150 schwarze 1639: return 0;
1.128 schwarze 1640: }
1641:
1642: static int
1.143 schwarze 1643: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1644: const struct roff_node *n)
1.128 schwarze 1645: {
1.132 schwarze 1646:
1.141 schwarze 1647: if (n->type != ROFFT_HEAD)
1.150 schwarze 1648: return 1;
1.128 schwarze 1649:
1.133 schwarze 1650: if (n->child != NULL)
1651: parse_mdoc_fname(mpage, n->child);
1652:
1.150 schwarze 1653: return 0;
1.1 schwarze 1654: }
1655:
1.19 schwarze 1656: static int
1.164 schwarze 1657: parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1658: const struct roff_node *n)
1659: {
1660: char *cp;
1661:
1662: if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1663: return 0;
1664:
1.168 schwarze 1665: if (n->child != NULL &&
1666: n->child->next == NULL &&
1667: n->child->type == ROFFT_TEXT)
1.164 schwarze 1668: return 1;
1669:
1670: cp = NULL;
1671: deroff(&cp, n);
1672: if (cp != NULL) {
1673: putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1674: n->type == ROFFT_BODY ? TYPE_Va : 0));
1675: free(cp);
1676: }
1677:
1678: return 0;
1679: }
1680:
1681: static int
1.143 schwarze 1682: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1683: const struct roff_node *n)
1.1 schwarze 1684: {
1.47 schwarze 1685: char *cp;
1.1 schwarze 1686:
1687: if (NULL == (n = n->child))
1.150 schwarze 1688: return 0;
1.1 schwarze 1689:
1.47 schwarze 1690: if (NULL == n->next) {
1691: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1692: return 0;
1.47 schwarze 1693: }
1.1 schwarze 1694:
1.77 schwarze 1695: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1696: putkey(mpage, cp, TYPE_Xr);
1697: free(cp);
1.150 schwarze 1698: return 0;
1.1 schwarze 1699: }
1700:
1.19 schwarze 1701: static int
1.143 schwarze 1702: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1703: const struct roff_node *n)
1.1 schwarze 1704: {
1705:
1.141 schwarze 1706: if (n->type == ROFFT_BODY)
1.147 schwarze 1707: deroff(&mpage->desc, n);
1.150 schwarze 1708: return 0;
1.1 schwarze 1709: }
1710:
1.19 schwarze 1711: static int
1.143 schwarze 1712: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1713: const struct roff_node *n)
1.1 schwarze 1714: {
1715:
1.86 schwarze 1716: if (SEC_NAME == n->sec)
1.184 schwarze 1717: putmdockey(mpage, n->child, NAME_TITLE, 0);
1.141 schwarze 1718: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1719: if (n->child == NULL)
1720: putkey(mpage, meta->name, NAME_SYN);
1721: else
1.184 schwarze 1722: putmdockey(mpage, n->child, NAME_SYN, 0);
1.129 schwarze 1723: }
1724: if ( ! (mpage->name_head_done ||
1725: n->child == NULL || n->child->string == NULL ||
1726: strcasecmp(n->child->string, meta->title))) {
1.175 schwarze 1727: putkey(mpage, n->child->string, NAME_HEAD);
1.129 schwarze 1728: mpage->name_head_done = 1;
1.127 schwarze 1729: }
1.150 schwarze 1730: return 0;
1.1 schwarze 1731: }
1732:
1.19 schwarze 1733: static int
1.143 schwarze 1734: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1735: const struct roff_node *n)
1.1 schwarze 1736: {
1737:
1.150 schwarze 1738: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1739: }
1740:
1.47 schwarze 1741: static int
1.143 schwarze 1742: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1743: const struct roff_node *n)
1.1 schwarze 1744: {
1745:
1.150 schwarze 1746: return n->type == ROFFT_HEAD;
1.1 schwarze 1747: }
1748:
1.47 schwarze 1749: /*
1750: * Add a string to the hash table for the current manual.
1751: * Each string has a bitmask telling which macros it belongs to.
1752: * When we finish the manual, we'll dump the table.
1753: */
1.1 schwarze 1754: static void
1.131 schwarze 1755: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1756: {
1.90 schwarze 1757: struct ohash *htab;
1.47 schwarze 1758: struct str *s;
1.68 schwarze 1759: const char *end;
1.47 schwarze 1760: unsigned int slot;
1.131 schwarze 1761: int i, mustfree;
1.1 schwarze 1762:
1.47 schwarze 1763: if (0 == sz)
1764: return;
1.68 schwarze 1765:
1.131 schwarze 1766: mustfree = render_string(&cp, &sz);
1767:
1.90 schwarze 1768: if (TYPE_Nm & v) {
1769: htab = &names;
1770: v &= name_mask;
1.124 schwarze 1771: if (v & NAME_FIRST)
1772: name_mask &= ~NAME_FIRST;
1.90 schwarze 1773: if (debug > 1)
1774: say(mpage->mlinks->file,
1.183 schwarze 1775: "Adding name %*s, bits=0x%llx", (int)sz, cp,
1776: (unsigned long long)v);
1.90 schwarze 1777: } else {
1778: htab = &strings;
1779: if (debug > 1)
1.173 schwarze 1780: for (i = 0; i < KEY_MAX; i++)
1.118 schwarze 1781: if ((uint64_t)1 << i & v)
1.90 schwarze 1782: say(mpage->mlinks->file,
1783: "Adding key %s=%*s",
1.172 schwarze 1784: mansearch_keynames[i], (int)sz, cp);
1.68 schwarze 1785: }
1.47 schwarze 1786:
1787: end = cp + sz;
1.90 schwarze 1788: slot = ohash_qlookupi(htab, cp, &end);
1789: s = ohash_find(htab, slot);
1.1 schwarze 1790:
1.47 schwarze 1791: if (NULL != s && mpage == s->mpage) {
1792: s->mask |= v;
1.1 schwarze 1793: return;
1.47 schwarze 1794: } else if (NULL == s) {
1.101 schwarze 1795: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1796: memcpy(s->key, cp, sz);
1.90 schwarze 1797: ohash_insert(htab, slot, s);
1.47 schwarze 1798: }
1799: s->mpage = mpage;
1800: s->mask = v;
1.131 schwarze 1801:
1802: if (mustfree)
1803: free(cp);
1.1 schwarze 1804: }
1805:
1806: /*
1.47 schwarze 1807: * Take a Unicode codepoint and produce its UTF-8 encoding.
1808: * This isn't the best way to do this, but it works.
1809: * The magic numbers are from the UTF-8 packaging.
1810: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1811: */
1.47 schwarze 1812: static size_t
1813: utf8(unsigned int cp, char out[7])
1.1 schwarze 1814: {
1.47 schwarze 1815: size_t rc;
1.1 schwarze 1816:
1.47 schwarze 1817: rc = 0;
1818: if (cp <= 0x0000007F) {
1819: rc = 1;
1820: out[0] = (char)cp;
1821: } else if (cp <= 0x000007FF) {
1822: rc = 2;
1823: out[0] = (cp >> 6 & 31) | 192;
1824: out[1] = (cp & 63) | 128;
1825: } else if (cp <= 0x0000FFFF) {
1826: rc = 3;
1827: out[0] = (cp >> 12 & 15) | 224;
1828: out[1] = (cp >> 6 & 63) | 128;
1829: out[2] = (cp & 63) | 128;
1830: } else if (cp <= 0x001FFFFF) {
1831: rc = 4;
1832: out[0] = (cp >> 18 & 7) | 240;
1833: out[1] = (cp >> 12 & 63) | 128;
1834: out[2] = (cp >> 6 & 63) | 128;
1835: out[3] = (cp & 63) | 128;
1836: } else if (cp <= 0x03FFFFFF) {
1837: rc = 5;
1838: out[0] = (cp >> 24 & 3) | 248;
1839: out[1] = (cp >> 18 & 63) | 128;
1840: out[2] = (cp >> 12 & 63) | 128;
1841: out[3] = (cp >> 6 & 63) | 128;
1842: out[4] = (cp & 63) | 128;
1843: } else if (cp <= 0x7FFFFFFF) {
1844: rc = 6;
1845: out[0] = (cp >> 30 & 1) | 252;
1846: out[1] = (cp >> 24 & 63) | 128;
1847: out[2] = (cp >> 18 & 63) | 128;
1848: out[3] = (cp >> 12 & 63) | 128;
1849: out[4] = (cp >> 6 & 63) | 128;
1850: out[5] = (cp & 63) | 128;
1851: } else
1.150 schwarze 1852: return 0;
1.19 schwarze 1853:
1.47 schwarze 1854: out[rc] = '\0';
1.150 schwarze 1855: return rc;
1.1 schwarze 1856: }
1857:
1.47 schwarze 1858: /*
1.131 schwarze 1859: * If the string contains escape sequences,
1860: * replace it with an allocated rendering and return 1,
1861: * such that the caller can free it after use.
1862: * Otherwise, do nothing and return 0.
1.47 schwarze 1863: */
1.131 schwarze 1864: static int
1865: render_string(char **public, size_t *psz)
1.1 schwarze 1866: {
1.131 schwarze 1867: const char *src, *scp, *addcp, *seq;
1868: char *dst;
1869: size_t ssz, dsz, addsz;
1.71 schwarze 1870: char utfbuf[7], res[6];
1.131 schwarze 1871: int seqlen, unicode;
1.47 schwarze 1872:
1873: res[0] = '\\';
1874: res[1] = '\t';
1875: res[2] = ASCII_NBRSP;
1876: res[3] = ASCII_HYPH;
1.71 schwarze 1877: res[4] = ASCII_BREAK;
1878: res[5] = '\0';
1.1 schwarze 1879:
1.131 schwarze 1880: src = scp = *public;
1881: ssz = *psz;
1882: dst = NULL;
1883: dsz = 0;
1884:
1885: while (scp < src + *psz) {
1886:
1887: /* Leave normal characters unchanged. */
1888:
1889: if (strchr(res, *scp) == NULL) {
1890: if (dst != NULL)
1891: dst[dsz++] = *scp;
1892: scp++;
1893: continue;
1894: }
1.1 schwarze 1895:
1.47 schwarze 1896: /*
1.131 schwarze 1897: * Found something that requires replacing,
1898: * make sure we have a destination buffer.
1.47 schwarze 1899: */
1.131 schwarze 1900:
1901: if (dst == NULL) {
1902: dst = mandoc_malloc(ssz + 1);
1903: dsz = scp - src;
1904: memcpy(dst, src, dsz);
1.47 schwarze 1905: }
1.39 schwarze 1906:
1.131 schwarze 1907: /* Handle single-char special characters. */
1908:
1909: switch (*scp) {
1910: case '\\':
1911: break;
1.98 schwarze 1912: case '\t':
1913: case ASCII_NBRSP:
1.131 schwarze 1914: dst[dsz++] = ' ';
1915: scp++;
1916: continue;
1917: case ASCII_HYPH:
1918: dst[dsz++] = '-';
1.71 schwarze 1919: /* FALLTHROUGH */
1.98 schwarze 1920: case ASCII_BREAK:
1.131 schwarze 1921: scp++;
1.47 schwarze 1922: continue;
1.71 schwarze 1923: default:
1.131 schwarze 1924: abort();
1.71 schwarze 1925: }
1.39 schwarze 1926:
1.47 schwarze 1927: /*
1.131 schwarze 1928: * Found an escape sequence.
1929: * Read past the slash, then parse it.
1930: * Ignore everything except characters.
1.47 schwarze 1931: */
1.52 schwarze 1932:
1.131 schwarze 1933: scp++;
1934: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1935: continue;
1.39 schwarze 1936:
1.47 schwarze 1937: /*
1.52 schwarze 1938: * Render the special character
1939: * as either UTF-8 or ASCII.
1.47 schwarze 1940: */
1.52 schwarze 1941:
1942: if (write_utf8) {
1.157 schwarze 1943: unicode = mchars_spec2cp(seq, seqlen);
1.131 schwarze 1944: if (unicode <= 0)
1.52 schwarze 1945: continue;
1.131 schwarze 1946: addsz = utf8(unicode, utfbuf);
1947: if (addsz == 0)
1.52 schwarze 1948: continue;
1.131 schwarze 1949: addcp = utfbuf;
1.52 schwarze 1950: } else {
1.157 schwarze 1951: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.131 schwarze 1952: if (addcp == NULL)
1.52 schwarze 1953: continue;
1.131 schwarze 1954: if (*addcp == ASCII_NBRSP) {
1955: addcp = " ";
1956: addsz = 1;
1.52 schwarze 1957: }
1958: }
1.1 schwarze 1959:
1.47 schwarze 1960: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1961:
1.131 schwarze 1962: ssz += addsz;
1963: dst = mandoc_realloc(dst, ssz + 1);
1964: memcpy(dst + dsz, addcp, addsz);
1965: dsz += addsz;
1966: }
1967: if (dst != NULL) {
1968: *public = dst;
1969: *psz = dsz;
1970: }
1971:
1972: /* Trim trailing whitespace and NUL-terminate. */
1973:
1974: while (*psz > 0 && (*public)[*psz - 1] == ' ')
1975: --*psz;
1976: if (dst != NULL) {
1977: (*public)[*psz] = '\0';
1.150 schwarze 1978: return 1;
1.131 schwarze 1979: } else
1.150 schwarze 1980: return 0;
1.1 schwarze 1981: }
1982:
1.75 schwarze 1983: static void
1984: dbadd_mlink(const struct mlink *mlink)
1985: {
1.173 schwarze 1986: dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
1987: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
1988: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
1989: dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
1990: dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
1.75 schwarze 1991: }
1992:
1.11 schwarze 1993: /*
1.47 schwarze 1994: * Flush the current page's terms (and their bits) into the database.
1.53 schwarze 1995: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 1996: */
1997: static void
1.173 schwarze 1998: dbadd(struct dba *dba, struct mpage *mpage)
1.11 schwarze 1999: {
1.47 schwarze 2000: struct mlink *mlink;
2001: struct str *key;
1.131 schwarze 2002: char *cp;
1.173 schwarze 2003: uint64_t mask;
1.47 schwarze 2004: size_t i;
2005: unsigned int slot;
1.131 schwarze 2006: int mustfree;
1.47 schwarze 2007:
1.85 schwarze 2008: mlink = mpage->mlinks;
1.11 schwarze 2009:
1.85 schwarze 2010: if (nodb) {
1.104 schwarze 2011: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2012: key = ohash_next(&names, &slot))
1.104 schwarze 2013: free(key);
2014: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2015: key = ohash_next(&strings, &slot))
1.104 schwarze 2016: free(key);
1.102 schwarze 2017: if (0 == debug)
2018: return;
1.85 schwarze 2019: while (NULL != mlink) {
2020: fputs(mlink->name, stdout);
2021: if (NULL == mlink->next ||
2022: strcmp(mlink->dsec, mlink->next->dsec) ||
2023: strcmp(mlink->fsec, mlink->next->fsec) ||
2024: strcmp(mlink->arch, mlink->next->arch)) {
2025: putchar('(');
2026: if ('\0' == *mlink->dsec)
2027: fputs(mlink->fsec, stdout);
2028: else
2029: fputs(mlink->dsec, stdout);
2030: if ('\0' != *mlink->arch)
2031: printf("/%s", mlink->arch);
2032: putchar(')');
2033: }
2034: mlink = mlink->next;
2035: if (NULL != mlink)
2036: fputs(", ", stdout);
2037: }
1.89 schwarze 2038: printf(" - %s\n", mpage->desc);
1.11 schwarze 2039: return;
1.85 schwarze 2040: }
2041:
2042: if (debug)
2043: say(mlink->file, "Adding to database");
1.47 schwarze 2044:
1.131 schwarze 2045: cp = mpage->desc;
2046: i = strlen(cp);
2047: mustfree = render_string(&cp, &i);
1.176 schwarze 2048: mpage->dba = dba_page_new(dba->pages,
1.173 schwarze 2049: *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2050: cp, mlink->file, mpage->form);
1.131 schwarze 2051: if (mustfree)
2052: free(cp);
1.176 schwarze 2053: dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
1.47 schwarze 2054:
1.173 schwarze 2055: while (mlink != NULL) {
1.75 schwarze 2056: dbadd_mlink(mlink);
1.85 schwarze 2057: mlink = mlink->next;
2058: }
1.47 schwarze 2059:
1.90 schwarze 2060: for (key = ohash_first(&names, &slot); NULL != key;
2061: key = ohash_next(&names, &slot)) {
2062: assert(key->mpage == mpage);
1.173 schwarze 2063: dba_page_alias(mpage->dba, key->key, key->mask);
1.90 schwarze 2064: free(key);
2065: }
1.47 schwarze 2066: for (key = ohash_first(&strings, &slot); NULL != key;
2067: key = ohash_next(&strings, &slot)) {
2068: assert(key->mpage == mpage);
1.173 schwarze 2069: i = 0;
2070: for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2071: if (key->mask & mask)
2072: dba_macro_add(dba->macros, i,
2073: key->key, mpage->dba);
2074: i++;
2075: }
1.47 schwarze 2076: free(key);
1.33 schwarze 2077: }
1.47 schwarze 2078: }
1.41 deraadt 2079:
1.47 schwarze 2080: static void
1.173 schwarze 2081: dbprune(struct dba *dba)
1.47 schwarze 2082: {
1.173 schwarze 2083: struct dba_array *page, *files;
2084: char *file;
1.11 schwarze 2085:
1.173 schwarze 2086: dba_array_FOREACH(dba->pages, page) {
2087: files = dba_array_get(page, DBP_FILE);
2088: dba_array_FOREACH(files, file) {
2089: if (*file < ' ')
2090: file++;
2091: if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2092: file)) != NULL) {
2093: if (debug)
2094: say(file, "Deleting from database");
2095: dba_array_del(dba->pages);
2096: break;
2097: }
1.63 schwarze 2098: }
1.11 schwarze 2099: }
1.47 schwarze 2100: }
1.22 schwarze 2101:
1.47 schwarze 2102: /*
1.173 schwarze 2103: * Write the database from memory to disk.
1.47 schwarze 2104: */
2105: static void
1.173 schwarze 2106: dbwrite(struct dba *dba)
1.47 schwarze 2107: {
1.204 schwarze 2108: struct stat sb1, sb2;
2109: char tfn[33], *cp1, *cp2;
2110: off_t i;
2111: int fd1, fd2;
1.11 schwarze 2112:
1.199 schwarze 2113: /*
2114: * Do not write empty databases, and delete existing ones
2115: * when makewhatis -u causes them to become empty.
2116: */
2117:
2118: dba_array_start(dba->pages);
2119: if (dba_array_next(dba->pages) == NULL) {
1.201 schwarze 2120: if (unlink(MANDOC_DB) == -1 && errno != ENOENT)
1.199 schwarze 2121: say(MANDOC_DB, "&unlink");
2122: return;
2123: }
2124:
2125: /*
2126: * Build the database in a temporary file,
2127: * then atomically move it into place.
2128: */
2129:
1.173 schwarze 2130: if (dba_write(MANDOC_DB "~", dba) != -1) {
2131: if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2132: exitcode = (int)MANDOCLEVEL_SYSERR;
2133: say(MANDOC_DB, "&rename");
2134: unlink(MANDOC_DB "~");
2135: }
1.47 schwarze 2136: return;
1.28 schwarze 2137: }
1.199 schwarze 2138:
2139: /*
2140: * We lack write permission and cannot replace the database
2141: * file, but let's at least check whether the data changed.
2142: */
1.22 schwarze 2143:
1.173 schwarze 2144: (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2145: if (mkdtemp(tfn) == NULL) {
2146: exitcode = (int)MANDOCLEVEL_SYSERR;
2147: say("", "&%s", tfn);
1.47 schwarze 2148: return;
1.173 schwarze 2149: }
1.205 schwarze 2150: cp1 = cp2 = MAP_FAILED;
1.204 schwarze 2151: fd1 = fd2 = -1;
1.173 schwarze 2152: (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2153: if (dba_write(tfn, dba) == -1) {
2154: say(tfn, "&dba_write");
1.204 schwarze 2155: goto err;
1.72 schwarze 2156: }
1.204 schwarze 2157: if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) {
2158: say(MANDOC_DB, "&open");
2159: goto err;
2160: }
2161: if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) {
2162: say(tfn, "&open");
2163: goto err;
2164: }
2165: if (fstat(fd1, &sb1) == -1) {
2166: say(MANDOC_DB, "&fstat");
2167: goto err;
2168: }
2169: if (fstat(fd2, &sb2) == -1) {
2170: say(tfn, "&fstat");
2171: goto err;
2172: }
2173: if (sb1.st_size != sb2.st_size)
2174: goto err;
2175: if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE,
1.205 schwarze 2176: fd1, 0)) == MAP_FAILED) {
1.204 schwarze 2177: say(MANDOC_DB, "&mmap");
2178: goto err;
2179: }
2180: if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE,
1.205 schwarze 2181: fd2, 0)) == MAP_FAILED) {
1.204 schwarze 2182: say(tfn, "&mmap");
2183: goto err;
2184: }
2185: for (i = 0; i < sb1.st_size; i++)
2186: if (cp1[i] != cp2[i])
2187: goto err;
2188: goto out;
2189:
2190: err:
2191: exitcode = (int)MANDOCLEVEL_SYSERR;
2192: say(MANDOC_DB, "Data changed, but cannot replace database");
1.72 schwarze 2193:
1.173 schwarze 2194: out:
1.205 schwarze 2195: if (cp1 != MAP_FAILED)
1.204 schwarze 2196: munmap(cp1, sb1.st_size);
1.205 schwarze 2197: if (cp2 != MAP_FAILED)
1.204 schwarze 2198: munmap(cp2, sb2.st_size);
2199: if (fd1 != -1)
2200: close(fd1);
2201: if (fd2 != -1)
2202: close(fd2);
1.203 schwarze 2203: unlink(tfn);
1.173 schwarze 2204: *strrchr(tfn, '/') = '\0';
1.203 schwarze 2205: rmdir(tfn);
1.47 schwarze 2206: }
1.6 schwarze 2207:
1.47 schwarze 2208: static int
1.120 schwarze 2209: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2210: {
2211: static char startdir[PATH_MAX];
1.110 schwarze 2212: static int getcwd_status; /* 1 = ok, 2 = failure */
2213: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2214: char *cp;
1.6 schwarze 2215:
1.47 schwarze 2216: /*
1.110 schwarze 2217: * Remember the original working directory, if possible.
2218: * This will be needed if the second or a later directory
2219: * on the command line is given as a relative path.
2220: * Do not error out if the current directory is not
2221: * searchable: Maybe it won't be needed after all.
2222: */
2223: if (0 == getcwd_status) {
2224: if (NULL == getcwd(startdir, sizeof(startdir))) {
2225: getcwd_status = 2;
2226: (void)strlcpy(startdir, strerror(errno),
2227: sizeof(startdir));
2228: } else
2229: getcwd_status = 1;
2230: }
2231:
2232: /*
2233: * We are leaving the old base directory.
2234: * Do not use it any longer, not even for messages.
2235: */
2236: *basedir = '\0';
2237:
2238: /*
2239: * If and only if the directory was changed earlier and
2240: * the next directory to process is given as a relative path,
2241: * first go back, or bail out if that is impossible.
1.47 schwarze 2242: */
1.110 schwarze 2243: if (chdir_status && '/' != *targetdir) {
2244: if (2 == getcwd_status) {
1.47 schwarze 2245: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2246: say("", "getcwd: %s", startdir);
1.150 schwarze 2247: return 0;
1.47 schwarze 2248: }
1.110 schwarze 2249: if (-1 == chdir(startdir)) {
1.47 schwarze 2250: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2251: say("", "&chdir %s", startdir);
1.150 schwarze 2252: return 0;
1.2 schwarze 2253: }
2254: }
1.110 schwarze 2255:
2256: /*
2257: * Always resolve basedir to the canonicalized absolute
2258: * pathname and append a trailing slash, such that
2259: * we can reliably check whether files are inside.
2260: */
1.47 schwarze 2261: if (NULL == realpath(targetdir, basedir)) {
1.120 schwarze 2262: if (report_baddir || errno != ENOENT) {
2263: exitcode = (int)MANDOCLEVEL_BADARG;
2264: say("", "&%s: realpath", targetdir);
2265: }
1.150 schwarze 2266: return 0;
1.47 schwarze 2267: } else if (-1 == chdir(basedir)) {
1.120 schwarze 2268: if (report_baddir || errno != ENOENT) {
2269: exitcode = (int)MANDOCLEVEL_BADARG;
2270: say("", "&chdir");
2271: }
1.150 schwarze 2272: return 0;
1.109 schwarze 2273: }
1.110 schwarze 2274: chdir_status = 1;
1.109 schwarze 2275: cp = strchr(basedir, '\0');
2276: if ('/' != cp[-1]) {
2277: if (cp - basedir >= PATH_MAX - 1) {
2278: exitcode = (int)MANDOCLEVEL_SYSERR;
2279: say("", "Filename too long");
1.150 schwarze 2280: return 0;
1.109 schwarze 2281: }
2282: *cp++ = '/';
2283: *cp = '\0';
1.47 schwarze 2284: }
1.150 schwarze 2285: return 1;
1.2 schwarze 2286: }
2287:
2288: static void
1.47 schwarze 2289: say(const char *file, const char *format, ...)
1.2 schwarze 2290: {
1.47 schwarze 2291: va_list ap;
1.80 schwarze 2292: int use_errno;
1.2 schwarze 2293:
1.47 schwarze 2294: if ('\0' != *basedir)
2295: fprintf(stderr, "%s", basedir);
2296: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2297: fputc('/', stderr);
1.47 schwarze 2298: if ('\0' != *file)
2299: fprintf(stderr, "%s", file);
1.31 schwarze 2300:
1.80 schwarze 2301: use_errno = 1;
2302: if (NULL != format) {
2303: switch (*format) {
1.98 schwarze 2304: case '&':
1.80 schwarze 2305: format++;
2306: break;
1.98 schwarze 2307: case '\0':
1.80 schwarze 2308: format = NULL;
2309: break;
2310: default:
2311: use_errno = 0;
2312: break;
2313: }
2314: }
2315: if (NULL != format) {
2316: if ('\0' != *basedir || '\0' != *file)
2317: fputs(": ", stderr);
2318: va_start(ap, format);
2319: vfprintf(stderr, format, ap);
2320: va_end(ap);
2321: }
2322: if (use_errno) {
2323: if ('\0' != *basedir || '\0' != *file || NULL != format)
2324: fputs(": ", stderr);
1.47 schwarze 2325: perror(NULL);
1.80 schwarze 2326: } else
2327: fputc('\n', stderr);
1.1 schwarze 2328: }
![[BACK]](/icons/back.gif){kind=icon}
Return to mandocdb.c CVS log ![[TXT]](/icons/text.gif){kind=icon}
![[DIR]](/icons/dir.gif){kind=icon}
Up to [local] / src / usr.bin / mandoc