Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.209
1.209 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.208 2018/08/17 20:31:52 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"
1.209 ! 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 */
1.180 schwarze 167: static struct mpage *mpage_head; /* list of distinct manual pages */
1.47 schwarze 168: static struct ohash mpages; /* table of distinct manual pages */
169: static struct ohash mlinks; /* table of directory entries */
1.90 schwarze 170: static struct ohash names; /* table of all names */
1.47 schwarze 171: static struct ohash strings; /* table of all strings */
1.90 schwarze 172: static uint64_t name_mask;
1.47 schwarze 173:
1.208 schwarze 174: static const struct mdoc_handler mdoc_handlers[MDOC_MAX - MDOC_Dd] = {
1.184 schwarze 175: { NULL, 0, NODE_NOPRT }, /* Dd */
176: { NULL, 0, NODE_NOPRT }, /* Dt */
177: { NULL, 0, NODE_NOPRT }, /* Os */
178: { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
179: { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
180: { NULL, 0, 0 }, /* Pp */
181: { NULL, 0, 0 }, /* D1 */
182: { NULL, 0, 0 }, /* Dl */
183: { NULL, 0, 0 }, /* Bd */
184: { NULL, 0, 0 }, /* Ed */
185: { NULL, 0, 0 }, /* Bl */
186: { NULL, 0, 0 }, /* El */
187: { NULL, 0, 0 }, /* It */
188: { NULL, 0, 0 }, /* Ad */
189: { NULL, TYPE_An, 0 }, /* An */
1.195 schwarze 190: { NULL, 0, 0 }, /* Ap */
1.184 schwarze 191: { NULL, TYPE_Ar, 0 }, /* Ar */
192: { NULL, TYPE_Cd, 0 }, /* Cd */
193: { NULL, TYPE_Cm, 0 }, /* Cm */
194: { NULL, TYPE_Dv, 0 }, /* Dv */
195: { NULL, TYPE_Er, 0 }, /* Er */
196: { NULL, TYPE_Ev, 0 }, /* Ev */
197: { NULL, 0, 0 }, /* Ex */
1.207 schwarze 198: { parse_mdoc_Fa, 0, 0 }, /* Fa */
1.184 schwarze 199: { parse_mdoc_Fd, 0, 0 }, /* Fd */
200: { NULL, TYPE_Fl, 0 }, /* Fl */
201: { parse_mdoc_Fn, 0, 0 }, /* Fn */
1.207 schwarze 202: { NULL, TYPE_Ft | TYPE_Vt, 0 }, /* Ft */
1.184 schwarze 203: { NULL, TYPE_Ic, 0 }, /* Ic */
204: { NULL, TYPE_In, 0 }, /* In */
205: { NULL, TYPE_Li, 0 }, /* Li */
206: { parse_mdoc_Nd, 0, 0 }, /* Nd */
207: { parse_mdoc_Nm, 0, 0 }, /* Nm */
208: { NULL, 0, 0 }, /* Op */
209: { NULL, 0, 0 }, /* Ot */
210: { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
211: { NULL, 0, 0 }, /* Rv */
212: { NULL, TYPE_St, 0 }, /* St */
213: { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
214: { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
215: { parse_mdoc_Xr, 0, 0 }, /* Xr */
216: { NULL, 0, 0 }, /* %A */
217: { NULL, 0, 0 }, /* %B */
218: { NULL, 0, 0 }, /* %D */
219: { NULL, 0, 0 }, /* %I */
220: { NULL, 0, 0 }, /* %J */
221: { NULL, 0, 0 }, /* %N */
222: { NULL, 0, 0 }, /* %O */
223: { NULL, 0, 0 }, /* %P */
224: { NULL, 0, 0 }, /* %R */
225: { NULL, 0, 0 }, /* %T */
226: { NULL, 0, 0 }, /* %V */
227: { NULL, 0, 0 }, /* Ac */
228: { NULL, 0, 0 }, /* Ao */
229: { NULL, 0, 0 }, /* Aq */
1.186 schwarze 230: { NULL, TYPE_At, 0 }, /* At */
1.184 schwarze 231: { NULL, 0, 0 }, /* Bc */
232: { NULL, 0, 0 }, /* Bf */
233: { NULL, 0, 0 }, /* Bo */
234: { NULL, 0, 0 }, /* Bq */
235: { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
1.185 schwarze 236: { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
1.184 schwarze 237: { NULL, 0, 0 }, /* Db */
238: { NULL, 0, 0 }, /* Dc */
239: { NULL, 0, 0 }, /* Do */
240: { NULL, 0, 0 }, /* Dq */
241: { NULL, 0, 0 }, /* Ec */
242: { NULL, 0, 0 }, /* Ef */
243: { NULL, TYPE_Em, 0 }, /* Em */
244: { NULL, 0, 0 }, /* Eo */
245: { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
246: { NULL, TYPE_Ms, 0 }, /* Ms */
247: { NULL, 0, 0 }, /* No */
248: { NULL, 0, 0 }, /* Ns */
249: { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
250: { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
251: { NULL, 0, 0 }, /* Pc */
252: { NULL, 0, 0 }, /* Pf */
253: { NULL, 0, 0 }, /* Po */
254: { NULL, 0, 0 }, /* Pq */
255: { NULL, 0, 0 }, /* Qc */
256: { NULL, 0, 0 }, /* Ql */
257: { NULL, 0, 0 }, /* Qo */
258: { NULL, 0, 0 }, /* Qq */
259: { NULL, 0, 0 }, /* Re */
260: { NULL, 0, 0 }, /* Rs */
261: { NULL, 0, 0 }, /* Sc */
262: { NULL, 0, 0 }, /* So */
263: { NULL, 0, 0 }, /* Sq */
264: { NULL, 0, 0 }, /* Sm */
265: { NULL, 0, 0 }, /* Sx */
266: { NULL, TYPE_Sy, 0 }, /* Sy */
267: { NULL, TYPE_Tn, 0 }, /* Tn */
268: { NULL, 0, NODE_NOSRC }, /* Ux */
269: { NULL, 0, 0 }, /* Xc */
270: { NULL, 0, 0 }, /* Xo */
271: { parse_mdoc_Fo, 0, 0 }, /* Fo */
272: { NULL, 0, 0 }, /* Fc */
273: { NULL, 0, 0 }, /* Oo */
274: { NULL, 0, 0 }, /* Oc */
275: { NULL, 0, 0 }, /* Bk */
276: { NULL, 0, 0 }, /* Ek */
277: { NULL, 0, 0 }, /* Bt */
278: { NULL, 0, 0 }, /* Hf */
279: { NULL, 0, 0 }, /* Fr */
280: { NULL, 0, 0 }, /* Ud */
1.186 schwarze 281: { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */
1.184 schwarze 282: { NULL, 0, 0 }, /* Lp */
283: { NULL, TYPE_Lk, 0 }, /* Lk */
284: { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
285: { NULL, 0, 0 }, /* Brq */
286: { NULL, 0, 0 }, /* Bro */
287: { NULL, 0, 0 }, /* Brc */
288: { NULL, 0, 0 }, /* %C */
289: { NULL, 0, 0 }, /* Es */
290: { NULL, 0, 0 }, /* En */
291: { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
292: { NULL, 0, 0 }, /* %Q */
293: { NULL, 0, 0 }, /* %U */
294: { NULL, 0, 0 }, /* Ta */
1.1 schwarze 295: };
296:
1.98 schwarze 297:
1.1 schwarze 298: int
1.3 schwarze 299: mandocdb(int argc, char *argv[])
1.1 schwarze 300: {
1.140 schwarze 301: struct manconf conf;
302: struct mparse *mp;
1.173 schwarze 303: struct dba *dba;
1.161 schwarze 304: const char *path_arg, *progname;
1.140 schwarze 305: size_t j, sz;
1.47 schwarze 306: int ch, i;
307:
1.204 schwarze 308: if (pledge("stdio rpath wpath cpath", NULL) == -1) {
1.167 schwarze 309: warn("pledge");
1.155 schwarze 310: return (int)MANDOCLEVEL_SYSERR;
311: }
312:
1.140 schwarze 313: memset(&conf, 0, sizeof(conf));
1.47 schwarze 314:
315: /*
1.98 schwarze 316: * We accept a few different invocations.
1.47 schwarze 317: * The CHECKOP macro makes sure that invocation styles don't
318: * clobber each other.
319: */
320: #define CHECKOP(_op, _ch) do \
321: if (OP_DEFAULT != (_op)) { \
1.151 schwarze 322: warnx("-%c: Conflicting option", (_ch)); \
1.47 schwarze 323: goto usage; \
324: } while (/*CONSTCOND*/0)
1.10 schwarze 325:
1.47 schwarze 326: path_arg = NULL;
1.28 schwarze 327: op = OP_DEFAULT;
1.1 schwarze 328:
1.83 schwarze 329: while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
1.1 schwarze 330: switch (ch) {
1.98 schwarze 331: case 'a':
1.6 schwarze 332: use_all = 1;
333: break;
1.98 schwarze 334: case 'C':
1.47 schwarze 335: CHECKOP(op, ch);
336: path_arg = optarg;
1.28 schwarze 337: op = OP_CONFFILE;
1.25 schwarze 338: break;
1.98 schwarze 339: case 'D':
1.82 schwarze 340: debug++;
341: break;
1.98 schwarze 342: case 'd':
1.47 schwarze 343: CHECKOP(op, ch);
344: path_arg = optarg;
1.2 schwarze 345: op = OP_UPDATE;
1.1 schwarze 346: break;
1.98 schwarze 347: case 'n':
1.47 schwarze 348: nodb = 1;
349: break;
1.98 schwarze 350: case 'p':
1.83 schwarze 351: warnings = 1;
352: break;
1.98 schwarze 353: case 'Q':
1.73 schwarze 354: mparse_options |= MPARSE_QUICK;
1.59 schwarze 355: break;
1.98 schwarze 356: case 'T':
1.52 schwarze 357: if (strcmp(optarg, "utf8")) {
1.151 schwarze 358: warnx("-T%s: Unsupported output format",
359: optarg);
1.52 schwarze 360: goto usage;
361: }
362: write_utf8 = 1;
363: break;
1.98 schwarze 364: case 't':
1.47 schwarze 365: CHECKOP(op, ch);
1.28 schwarze 366: dup2(STDOUT_FILENO, STDERR_FILENO);
367: op = OP_TEST;
1.47 schwarze 368: nodb = warnings = 1;
1.28 schwarze 369: break;
1.98 schwarze 370: case 'u':
1.47 schwarze 371: CHECKOP(op, ch);
372: path_arg = optarg;
1.1 schwarze 373: op = OP_DELETE;
374: break;
1.98 schwarze 375: case 'v':
1.83 schwarze 376: /* Compatibility with espie@'s makewhatis. */
1.28 schwarze 377: break;
1.1 schwarze 378: default:
1.28 schwarze 379: goto usage;
1.1 schwarze 380: }
381:
382: argc -= optind;
383: argv += optind;
384:
1.163 schwarze 385: if (nodb) {
386: if (pledge("stdio rpath", NULL) == -1) {
1.167 schwarze 387: warn("pledge");
1.163 schwarze 388: return (int)MANDOCLEVEL_SYSERR;
389: }
1.155 schwarze 390: }
391:
1.28 schwarze 392: if (OP_CONFFILE == op && argc > 0) {
1.151 schwarze 393: warnx("-C: Too many arguments");
1.28 schwarze 394: goto usage;
395: }
396:
1.47 schwarze 397: exitcode = (int)MANDOCLEVEL_OK;
1.157 schwarze 398: mchars_alloc();
1.202 schwarze 399: mp = mparse_alloc(mparse_options, MANDOCERR_MAX, NULL,
1.200 schwarze 400: MANDOC_OS_OTHER, NULL);
1.156 schwarze 401: mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
402: mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
1.2 schwarze 403:
1.47 schwarze 404: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.2 schwarze 405:
1.47 schwarze 406: /*
1.109 schwarze 407: * Most of these deal with a specific directory.
1.97 schwarze 408: * Jump into that directory first.
1.47 schwarze 409: */
1.120 schwarze 410: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.47 schwarze 411: goto out;
1.97 schwarze 412:
1.174 schwarze 413: dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
414: if (dba != NULL) {
1.97 schwarze 415: /*
416: * The existing database is usable. Process
417: * all files specified on the command-line.
418: */
419: use_all = 1;
1.95 schwarze 420: for (i = 0; i < argc; i++)
421: filescan(argv[i]);
1.174 schwarze 422: if (nodb == 0)
1.173 schwarze 423: dbprune(dba);
1.95 schwarze 424: } else {
1.177 schwarze 425: /* Database missing or corrupt. */
1.179 schwarze 426: if (op != OP_UPDATE || errno != ENOENT)
427: say(MANDOC_DB, "%s: Automatically recreating"
428: " from scratch", strerror(errno));
1.97 schwarze 429: exitcode = (int)MANDOCLEVEL_OK;
1.95 schwarze 430: op = OP_DEFAULT;
431: if (0 == treescan())
432: goto out;
1.173 schwarze 433: dba = dba_new(128);
1.95 schwarze 434: }
1.47 schwarze 435: if (OP_DELETE != op)
1.173 schwarze 436: mpages_merge(dba, mp);
437: if (nodb == 0)
438: dbwrite(dba);
439: dba_free(dba);
1.47 schwarze 440: } else {
441: /*
442: * If we have arguments, use them as our manpaths.
1.181 schwarze 443: * If we don't, use man.conf(5).
1.47 schwarze 444: */
445: if (argc > 0) {
1.140 schwarze 446: conf.manpath.paths = mandoc_reallocarray(NULL,
1.101 schwarze 447: argc, sizeof(char *));
1.140 schwarze 448: conf.manpath.sz = (size_t)argc;
1.47 schwarze 449: for (i = 0; i < argc; i++)
1.140 schwarze 450: conf.manpath.paths[i] = mandoc_strdup(argv[i]);
1.47 schwarze 451: } else
1.140 schwarze 452: manconf_parse(&conf, path_arg, NULL, NULL);
1.84 schwarze 453:
1.140 schwarze 454: if (conf.manpath.sz == 0) {
1.84 schwarze 455: exitcode = (int)MANDOCLEVEL_BADARG;
456: say("", "Empty manpath");
457: }
1.2 schwarze 458:
1.47 schwarze 459: /*
460: * First scan the tree rooted at a base directory, then
461: * build a new database and finally move it into place.
462: * Ignore zero-length directories and strip trailing
463: * slashes.
464: */
1.140 schwarze 465: for (j = 0; j < conf.manpath.sz; j++) {
466: sz = strlen(conf.manpath.paths[j]);
467: if (sz && conf.manpath.paths[j][sz - 1] == '/')
468: conf.manpath.paths[j][--sz] = '\0';
1.47 schwarze 469: if (0 == sz)
470: continue;
1.2 schwarze 471:
1.47 schwarze 472: if (j) {
1.156 schwarze 473: mandoc_ohash_init(&mpages, 6,
474: offsetof(struct mpage, inodev));
475: mandoc_ohash_init(&mlinks, 6,
476: offsetof(struct mlink, file));
1.47 schwarze 477: }
1.1 schwarze 478:
1.140 schwarze 479: if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
1.120 schwarze 480: continue;
1.47 schwarze 481: if (0 == treescan())
1.120 schwarze 482: continue;
1.173 schwarze 483: dba = dba_new(128);
484: mpages_merge(dba, mp);
485: if (nodb == 0)
486: dbwrite(dba);
487: dba_free(dba);
1.28 schwarze 488:
1.140 schwarze 489: if (j + 1 < conf.manpath.sz) {
1.47 schwarze 490: mpages_free();
491: ohash_delete(&mpages);
492: ohash_delete(&mlinks);
493: }
1.2 schwarze 494: }
1.47 schwarze 495: }
496: out:
1.140 schwarze 497: manconf_free(&conf);
1.123 schwarze 498: mparse_free(mp);
1.157 schwarze 499: mchars_free();
1.47 schwarze 500: mpages_free();
501: ohash_delete(&mpages);
502: ohash_delete(&mlinks);
1.150 schwarze 503: return exitcode;
1.47 schwarze 504: usage:
1.161 schwarze 505: progname = getprogname();
1.83 schwarze 506: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
507: " %s [-aDnpQ] [-Tutf8] dir ...\n"
508: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
509: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 510: " %s [-Q] -t file ...\n",
1.161 schwarze 511: progname, progname, progname, progname, progname);
1.1 schwarze 512:
1.150 schwarze 513: return (int)MANDOCLEVEL_BADARG;
1.47 schwarze 514: }
1.1 schwarze 515:
1.47 schwarze 516: /*
1.180 schwarze 517: * To get a singly linked list in alpha order while inserting entries
518: * at the beginning, process directory entries in reverse alpha order.
519: */
520: static int
521: fts_compare(const FTSENT **a, const FTSENT **b)
522: {
523: return -strcmp((*a)->fts_name, (*b)->fts_name);
524: }
525:
526: /*
1.47 schwarze 527: * Scan a directory tree rooted at "basedir" for manpages.
528: * We use fts(), scanning directory parts along the way for clues to our
529: * section and architecture.
530: *
531: * If use_all has been specified, grok all files.
532: * If not, sanitise paths to the following:
533: *
1.98 schwarze 534: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 535: * or
536: * [./]cat<section>[/<arch>]/<name>.0
537: *
1.169 krw 538: * TODO: accommodate for multi-language directories.
1.47 schwarze 539: */
540: static int
541: treescan(void)
542: {
1.96 schwarze 543: char buf[PATH_MAX];
1.47 schwarze 544: FTS *f;
545: FTSENT *ff;
546: struct mlink *mlink;
1.173 schwarze 547: int gzip;
548: enum form dform;
1.51 schwarze 549: char *dsec, *arch, *fsec, *cp;
550: const char *path;
1.47 schwarze 551: const char *argv[2];
1.1 schwarze 552:
1.47 schwarze 553: argv[0] = ".";
1.193 schwarze 554: argv[1] = NULL;
1.1 schwarze 555:
1.180 schwarze 556: f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR,
557: fts_compare);
1.153 schwarze 558: if (f == NULL) {
1.47 schwarze 559: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 560: say("", "&fts_open");
1.150 schwarze 561: return 0;
1.47 schwarze 562: }
1.2 schwarze 563:
1.47 schwarze 564: dsec = arch = NULL;
565: dform = FORM_NONE;
1.2 schwarze 566:
1.153 schwarze 567: while ((ff = fts_read(f)) != NULL) {
1.47 schwarze 568: path = ff->fts_path + 2;
1.96 schwarze 569: switch (ff->fts_info) {
570:
571: /*
572: * Symbolic links require various sanity checks,
573: * then get handled just like regular files.
574: */
1.98 schwarze 575: case FTS_SL:
1.153 schwarze 576: if (realpath(path, buf) == NULL) {
1.96 schwarze 577: if (warnings)
578: say(path, "&realpath");
579: continue;
580: }
581: if (strstr(buf, basedir) != buf) {
582: if (warnings) say("",
583: "%s: outside base directory", buf);
584: continue;
585: }
586: /* Use logical inode to avoid mpages dupe. */
1.153 schwarze 587: if (stat(path, ff->fts_statp) == -1) {
1.96 schwarze 588: if (warnings)
589: say(path, "&stat");
590: continue;
591: }
592: /* FALLTHROUGH */
593:
1.15 schwarze 594: /*
1.47 schwarze 595: * If we're a regular file, add an mlink by using the
596: * stored directory data and handling the filename.
1.15 schwarze 597: */
1.98 schwarze 598: case FTS_F:
1.153 schwarze 599: if ( ! strcmp(path, MANDOC_DB))
1.47 schwarze 600: continue;
601: if ( ! use_all && ff->fts_level < 2) {
602: if (warnings)
603: say(path, "Extraneous file");
604: continue;
1.81 schwarze 605: }
606: gzip = 0;
607: fsec = NULL;
1.153 schwarze 608: while (fsec == NULL) {
1.81 schwarze 609: fsec = strrchr(ff->fts_name, '.');
1.153 schwarze 610: if (fsec == NULL || strcmp(fsec+1, "gz"))
1.81 schwarze 611: break;
612: gzip = 1;
613: *fsec = '\0';
614: fsec = NULL;
615: }
1.153 schwarze 616: if (fsec == NULL) {
1.47 schwarze 617: if ( ! use_all) {
618: if (warnings)
619: say(path,
620: "No filename suffix");
621: continue;
622: }
1.153 schwarze 623: } else if ( ! strcmp(++fsec, "html")) {
1.47 schwarze 624: if (warnings)
625: say(path, "Skip html");
626: continue;
1.153 schwarze 627: } else if ( ! strcmp(fsec, "ps")) {
1.47 schwarze 628: if (warnings)
629: say(path, "Skip ps");
630: continue;
1.153 schwarze 631: } else if ( ! strcmp(fsec, "pdf")) {
1.47 schwarze 632: if (warnings)
633: say(path, "Skip pdf");
634: continue;
635: } else if ( ! use_all &&
1.153 schwarze 636: ((dform == FORM_SRC &&
1.138 schwarze 637: strncmp(fsec, dsec, strlen(dsec))) ||
1.153 schwarze 638: (dform == FORM_CAT && strcmp(fsec, "0")))) {
1.47 schwarze 639: if (warnings)
640: say(path, "Wrong filename suffix");
641: continue;
642: } else
643: fsec[-1] = '\0';
1.51 schwarze 644:
1.47 schwarze 645: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.100 schwarze 646: if (strlcpy(mlink->file, path,
647: sizeof(mlink->file)) >=
648: sizeof(mlink->file)) {
649: say(path, "Filename too long");
650: free(mlink);
651: continue;
652: }
1.47 schwarze 653: mlink->dform = dform;
1.51 schwarze 654: mlink->dsec = dsec;
655: mlink->arch = arch;
656: mlink->name = ff->fts_name;
657: mlink->fsec = fsec;
1.81 schwarze 658: mlink->gzip = gzip;
1.47 schwarze 659: mlink_add(mlink, ff->fts_statp);
660: continue;
1.96 schwarze 661:
1.98 schwarze 662: case FTS_D:
663: case FTS_DP:
1.96 schwarze 664: break;
665:
666: default:
1.47 schwarze 667: if (warnings)
668: say(path, "Not a regular file");
669: continue;
670: }
671:
672: switch (ff->fts_level) {
1.98 schwarze 673: case 0:
1.47 schwarze 674: /* Ignore the root directory. */
675: break;
1.98 schwarze 676: case 1:
1.47 schwarze 677: /*
678: * This might contain manX/ or catX/.
679: * Try to infer this from the name.
680: * If we're not in use_all, enforce it.
681: */
682: cp = ff->fts_name;
1.153 schwarze 683: if (ff->fts_info == FTS_DP) {
684: dform = FORM_NONE;
685: dsec = NULL;
1.47 schwarze 686: break;
1.153 schwarze 687: }
1.15 schwarze 688:
1.153 schwarze 689: if ( ! strncmp(cp, "man", 3)) {
1.47 schwarze 690: dform = FORM_SRC;
691: dsec = cp + 3;
1.153 schwarze 692: } else if ( ! strncmp(cp, "cat", 3)) {
1.47 schwarze 693: dform = FORM_CAT;
694: dsec = cp + 3;
1.51 schwarze 695: } else {
696: dform = FORM_NONE;
697: dsec = NULL;
1.26 schwarze 698: }
1.47 schwarze 699:
1.153 schwarze 700: if (dsec != NULL || use_all)
1.47 schwarze 701: break;
702:
703: if (warnings)
704: say(path, "Unknown directory part");
705: fts_set(f, ff, FTS_SKIP);
706: break;
1.98 schwarze 707: case 2:
1.47 schwarze 708: /*
709: * Possibly our architecture.
710: * If we're descending, keep tabs on it.
711: */
1.153 schwarze 712: if (ff->fts_info != FTS_DP && dsec != NULL)
1.47 schwarze 713: arch = ff->fts_name;
1.51 schwarze 714: else
715: arch = NULL;
1.47 schwarze 716: break;
717: default:
1.153 schwarze 718: if (ff->fts_info == FTS_DP || use_all)
1.47 schwarze 719: break;
720: if (warnings)
721: say(path, "Extraneous directory part");
722: fts_set(f, ff, FTS_SKIP);
723: break;
1.14 schwarze 724: }
1.47 schwarze 725: }
726:
727: fts_close(f);
1.150 schwarze 728: return 1;
1.47 schwarze 729: }
1.1 schwarze 730:
1.47 schwarze 731: /*
732: * Add a file to the mlinks table.
733: * Do not verify that it's a "valid" looking manpage (we'll do that
734: * later).
735: *
736: * Try to infer the manual section, architecture, and page name from the
737: * path, assuming it looks like
738: *
1.98 schwarze 739: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 740: * or
741: * [./]cat<section>[/<arch>]/<name>.0
742: *
743: * See treescan() for the fts(3) version of this.
744: */
745: static void
746: filescan(const char *file)
747: {
748: char buf[PATH_MAX];
749: struct stat st;
750: struct mlink *mlink;
751: char *p, *start;
752:
753: assert(use_all);
754:
755: if (0 == strncmp(file, "./", 2))
756: file += 2;
757:
1.96 schwarze 758: /*
759: * We have to do lstat(2) before realpath(3) loses
760: * the information whether this is a symbolic link.
761: * We need to know that because for symbolic links,
762: * we want to use the orginal file name, while for
763: * regular files, we want to use the real path.
764: */
765: if (-1 == lstat(file, &st)) {
766: exitcode = (int)MANDOCLEVEL_BADARG;
767: say(file, "&lstat");
768: return;
769: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
770: exitcode = (int)MANDOCLEVEL_BADARG;
771: say(file, "Not a regular file");
772: return;
773: }
774:
775: /*
776: * We have to resolve the file name to the real path
777: * in any case for the base directory check.
778: */
1.47 schwarze 779: if (NULL == realpath(file, buf)) {
780: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 781: say(file, "&realpath");
1.47 schwarze 782: return;
1.63 schwarze 783: }
784:
1.109 schwarze 785: if (OP_TEST == op)
1.63 schwarze 786: start = buf;
1.109 schwarze 787: else if (strstr(buf, basedir) == buf)
788: start = buf + strlen(basedir);
1.63 schwarze 789: else {
1.47 schwarze 790: exitcode = (int)MANDOCLEVEL_BADARG;
791: say("", "%s: outside base directory", buf);
792: return;
1.63 schwarze 793: }
794:
1.96 schwarze 795: /*
796: * Now we are sure the file is inside our tree.
797: * If it is a symbolic link, ignore the real path
798: * and use the original name.
799: * This implies passing stuff like "cat1/../man1/foo.1"
800: * on the command line won't work. So don't do that.
801: * Note the stat(2) can still fail if the link target
802: * doesn't exist.
803: */
804: if (S_IFLNK & st.st_mode) {
805: if (-1 == stat(buf, &st)) {
806: exitcode = (int)MANDOCLEVEL_BADARG;
807: say(file, "&stat");
808: return;
809: }
1.100 schwarze 810: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
811: say(file, "Filename too long");
812: return;
813: }
1.109 schwarze 814: start = buf;
815: if (OP_TEST != op && strstr(buf, basedir) == buf)
816: start += strlen(basedir);
1.1 schwarze 817: }
1.63 schwarze 818:
1.47 schwarze 819: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 820: mlink->dform = FORM_NONE;
1.100 schwarze 821: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
822: sizeof(mlink->file)) {
823: say(start, "Filename too long");
1.134 schwarze 824: free(mlink);
1.100 schwarze 825: return;
1.188 schwarze 826: }
827:
828: /*
829: * In test mode or when the original name is absolute
830: * but outside our tree, guess the base directory.
831: */
832:
833: if (op == OP_TEST || (start == buf && *start == '/')) {
834: if (strncmp(buf, "man/", 4) == 0)
835: start = buf + 4;
836: else if ((start = strstr(buf, "/man/")) != NULL)
837: start += 5;
838: else
839: start = buf;
1.100 schwarze 840: }
1.1 schwarze 841:
1.10 schwarze 842: /*
1.47 schwarze 843: * First try to guess our directory structure.
844: * If we find a separator, try to look for man* or cat*.
845: * If we find one of these and what's underneath is a directory,
846: * assume it's an architecture.
1.10 schwarze 847: */
1.47 schwarze 848: if (NULL != (p = strchr(start, '/'))) {
849: *p++ = '\0';
850: if (0 == strncmp(start, "man", 3)) {
851: mlink->dform = FORM_SRC;
1.51 schwarze 852: mlink->dsec = start + 3;
1.47 schwarze 853: } else if (0 == strncmp(start, "cat", 3)) {
854: mlink->dform = FORM_CAT;
1.51 schwarze 855: mlink->dsec = start + 3;
1.47 schwarze 856: }
1.10 schwarze 857:
1.47 schwarze 858: start = p;
859: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
860: *p++ = '\0';
1.51 schwarze 861: mlink->arch = start;
1.47 schwarze 862: start = p;
1.41 deraadt 863: }
1.47 schwarze 864: }
1.7 schwarze 865:
1.47 schwarze 866: /*
867: * Now check the file suffix.
868: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
869: */
870: p = strrchr(start, '\0');
871: while (p-- > start && '/' != *p && '.' != *p)
872: /* Loop. */ ;
873:
874: if ('.' == *p) {
875: *p++ = '\0';
1.51 schwarze 876: mlink->fsec = p;
1.47 schwarze 877: }
1.41 deraadt 878:
1.47 schwarze 879: /*
880: * Now try to parse the name.
881: * Use the filename portion of the path.
882: */
883: mlink->name = start;
884: if (NULL != (p = strrchr(start, '/'))) {
885: mlink->name = p + 1;
886: *p = '\0';
887: }
888: mlink_add(mlink, &st);
889: }
1.2 schwarze 890:
1.47 schwarze 891: static void
892: mlink_add(struct mlink *mlink, const struct stat *st)
893: {
894: struct inodev inodev;
895: struct mpage *mpage;
896: unsigned int slot;
897:
898: assert(NULL != mlink->file);
899:
1.51 schwarze 900: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
901: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
902: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
903: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 904:
905: if ('0' == *mlink->fsec) {
906: free(mlink->fsec);
907: mlink->fsec = mandoc_strdup(mlink->dsec);
908: mlink->fform = FORM_CAT;
909: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
910: mlink->fform = FORM_SRC;
911: else
912: mlink->fform = FORM_NONE;
1.1 schwarze 913:
1.47 schwarze 914: slot = ohash_qlookup(&mlinks, mlink->file);
915: assert(NULL == ohash_find(&mlinks, slot));
916: ohash_insert(&mlinks, slot, mlink);
917:
1.139 schwarze 918: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.47 schwarze 919: inodev.st_ino = st->st_ino;
920: inodev.st_dev = st->st_dev;
921: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
922: sizeof(struct inodev), inodev.st_ino);
923: mpage = ohash_find(&mpages, slot);
924: if (NULL == mpage) {
925: mpage = mandoc_calloc(1, sizeof(struct mpage));
926: mpage->inodev.st_ino = inodev.st_ino;
927: mpage->inodev.st_dev = inodev.st_dev;
1.173 schwarze 928: mpage->form = FORM_NONE;
1.180 schwarze 929: mpage->next = mpage_head;
930: mpage_head = mpage;
1.47 schwarze 931: ohash_insert(&mpages, slot, mpage);
932: } else
933: mlink->next = mpage->mlinks;
934: mpage->mlinks = mlink;
1.75 schwarze 935: mlink->mpage = mpage;
1.47 schwarze 936: }
1.1 schwarze 937:
1.47 schwarze 938: static void
939: mlink_free(struct mlink *mlink)
940: {
1.1 schwarze 941:
1.47 schwarze 942: free(mlink->dsec);
943: free(mlink->arch);
944: free(mlink->name);
945: free(mlink->fsec);
946: free(mlink);
947: }
1.1 schwarze 948:
1.47 schwarze 949: static void
950: mpages_free(void)
951: {
952: struct mpage *mpage;
953: struct mlink *mlink;
954:
1.180 schwarze 955: while ((mpage = mpage_head) != NULL) {
956: while ((mlink = mpage->mlinks) != NULL) {
1.47 schwarze 957: mpage->mlinks = mlink->next;
958: mlink_free(mlink);
959: }
1.180 schwarze 960: mpage_head = mpage->next;
1.47 schwarze 961: free(mpage->sec);
962: free(mpage->arch);
963: free(mpage->title);
964: free(mpage->desc);
965: free(mpage);
966: }
967: }
1.1 schwarze 968:
1.47 schwarze 969: /*
970: * For each mlink to the mpage, check whether the path looks like
971: * it is formatted, and if it does, check whether a source manual
972: * exists by the same name, ignoring the suffix.
973: * If both conditions hold, drop the mlink.
974: */
975: static void
976: mlinks_undupe(struct mpage *mpage)
977: {
978: char buf[PATH_MAX];
979: struct mlink **prev;
980: struct mlink *mlink;
981: char *bufp;
982:
983: mpage->form = FORM_CAT;
984: prev = &mpage->mlinks;
985: while (NULL != (mlink = *prev)) {
986: if (FORM_CAT != mlink->dform) {
987: mpage->form = FORM_NONE;
988: goto nextlink;
989: }
1.100 schwarze 990: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 991: bufp = strstr(buf, "cat");
992: assert(NULL != bufp);
993: memcpy(bufp, "man", 3);
994: if (NULL != (bufp = strrchr(buf, '.')))
995: *++bufp = '\0';
1.100 schwarze 996: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 997: if (NULL == ohash_find(&mlinks,
1.98 schwarze 998: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 999: goto nextlink;
1000: if (warnings)
1001: say(mlink->file, "Man source exists: %s", buf);
1002: if (use_all)
1003: goto nextlink;
1004: *prev = mlink->next;
1005: mlink_free(mlink);
1006: continue;
1007: nextlink:
1008: prev = &(*prev)->next;
1.1 schwarze 1009: }
1.47 schwarze 1010: }
1.1 schwarze 1011:
1.88 schwarze 1012: static void
1.50 schwarze 1013: mlink_check(struct mpage *mpage, struct mlink *mlink)
1014: {
1.88 schwarze 1015: struct str *str;
1016: unsigned int slot;
1.50 schwarze 1017:
1018: /*
1019: * Check whether the manual section given in a file
1020: * agrees with the directory where the file is located.
1021: * Some manuals have suffixes like (3p) on their
1022: * section number either inside the file or in the
1023: * directory name, some are linked into more than one
1024: * section, like encrypt(1) = makekey(8).
1025: */
1026:
1027: if (FORM_SRC == mpage->form &&
1.88 schwarze 1028: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1029: say(mlink->file, "Section \"%s\" manual in %s directory",
1030: mpage->sec, mlink->dsec);
1031:
1032: /*
1033: * Manual page directories exist for each kernel
1034: * architecture as returned by machine(1).
1035: * However, many manuals only depend on the
1036: * application architecture as returned by arch(1).
1037: * For example, some (2/ARM) manuals are shared
1038: * across the "armish" and "zaurus" kernel
1039: * architectures.
1040: * A few manuals are even shared across completely
1041: * different architectures, for example fdformat(1)
1.178 schwarze 1042: * on amd64, i386, and sparc64.
1.50 schwarze 1043: */
1044:
1.88 schwarze 1045: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1046: say(mlink->file, "Architecture \"%s\" manual in "
1047: "\"%s\" directory", mpage->arch, mlink->arch);
1048:
1.88 schwarze 1049: /*
1050: * XXX
1.90 schwarze 1051: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1052: */
1053:
1054: if (FORM_CAT == mpage->form)
1055: return;
1056:
1057: /*
1058: * Check whether this mlink
1059: * appears as a name in the NAME section.
1060: */
1.50 schwarze 1061:
1.90 schwarze 1062: slot = ohash_qlookup(&names, mlink->name);
1063: str = ohash_find(&names, slot);
1.88 schwarze 1064: assert(NULL != str);
1.90 schwarze 1065: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1066: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1067: }
1068:
1.47 schwarze 1069: /*
1070: * Run through the files in the global vector "mpages"
1071: * and add them to the database specified in "basedir".
1072: *
1073: * This handles the parsing scheme itself, using the cues of directory
1074: * and filename to determine whether the file is parsable or not.
1075: */
1076: static void
1.173 schwarze 1077: mpages_merge(struct dba *dba, struct mparse *mp)
1.47 schwarze 1078: {
1.180 schwarze 1079: struct mpage *mpage, *mpage_dest;
1.75 schwarze 1080: struct mlink *mlink, *mlink_dest;
1.144 schwarze 1081: struct roff_man *man;
1.75 schwarze 1082: char *sodest;
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);
1095: man = NULL;
1.81 schwarze 1096: sodest = NULL;
1097:
1.166 schwarze 1098: if ((fd = mparse_open(mp, mlink->file)) == -1) {
1.136 schwarze 1099: say(mlink->file, "&open");
1.117 schwarze 1100: goto nextpage;
1.81 schwarze 1101: }
1.11 schwarze 1102:
1103: /*
1.137 schwarze 1104: * Interpret the file as mdoc(7) or man(7) source
1105: * code, unless it is known to be formatted.
1.11 schwarze 1106: */
1.136 schwarze 1107: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.137 schwarze 1108: mparse_readfd(mp, fd, mlink->file);
1.165 schwarze 1109: close(fd);
1.189 schwarze 1110: fd = -1;
1.145 schwarze 1111: mparse_result(mp, &man, &sodest);
1.47 schwarze 1112: }
1.11 schwarze 1113:
1.126 schwarze 1114: if (sodest != NULL) {
1.75 schwarze 1115: mlink_dest = ohash_find(&mlinks,
1116: ohash_qlookup(&mlinks, sodest));
1.126 schwarze 1117: if (mlink_dest == NULL) {
1118: mandoc_asprintf(&cp, "%s.gz", sodest);
1119: mlink_dest = ohash_find(&mlinks,
1120: ohash_qlookup(&mlinks, cp));
1121: free(cp);
1122: }
1123: if (mlink_dest != NULL) {
1.75 schwarze 1124:
1125: /* The .so target exists. */
1126:
1127: mpage_dest = mlink_dest->mpage;
1128: while (1) {
1129: mlink->mpage = mpage_dest;
1130:
1131: /*
1132: * If the target was already
1133: * processed, add the links
1134: * to the database now.
1135: * Otherwise, this will
1136: * happen when we come
1137: * to the target.
1138: */
1139:
1.173 schwarze 1140: if (mpage_dest->dba != NULL)
1141: dbadd_mlink(mlink);
1.75 schwarze 1142:
1.126 schwarze 1143: if (mlink->next == NULL)
1.75 schwarze 1144: break;
1145: mlink = mlink->next;
1146: }
1147:
1148: /* Move all links to the target. */
1149:
1150: mlink->next = mlink_dest->next;
1151: mlink_dest->next = mpage->mlinks;
1152: mpage->mlinks = NULL;
1153: }
1.81 schwarze 1154: goto nextpage;
1.145 schwarze 1155: } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1.159 schwarze 1156: mdoc_validate(man);
1.47 schwarze 1157: mpage->form = FORM_SRC;
1.146 schwarze 1158: mpage->sec = man->meta.msec;
1.112 schwarze 1159: mpage->sec = mandoc_strdup(
1.126 schwarze 1160: mpage->sec == NULL ? "" : mpage->sec);
1.146 schwarze 1161: mpage->arch = man->meta.arch;
1.47 schwarze 1162: mpage->arch = mandoc_strdup(
1.126 schwarze 1163: mpage->arch == NULL ? "" : mpage->arch);
1.146 schwarze 1164: mpage->title = mandoc_strdup(man->meta.title);
1.145 schwarze 1165: } else if (man != NULL && man->macroset == MACROSET_MAN) {
1.160 schwarze 1166: man_validate(man);
1.189 schwarze 1167: if (*man->meta.msec != '\0' ||
1.194 schwarze 1168: *man->meta.title != '\0') {
1.189 schwarze 1169: mpage->form = FORM_SRC;
1170: mpage->sec = mandoc_strdup(man->meta.msec);
1171: mpage->arch = mandoc_strdup(mlink->arch);
1172: mpage->title = mandoc_strdup(man->meta.title);
1173: } else
1174: man = NULL;
1175: }
1176:
1177: assert(mpage->desc == NULL);
1178: if (man == NULL) {
1.47 schwarze 1179: mpage->form = FORM_CAT;
1.136 schwarze 1180: mpage->sec = mandoc_strdup(mlink->dsec);
1181: mpage->arch = mandoc_strdup(mlink->arch);
1182: mpage->title = mandoc_strdup(mlink->name);
1.189 schwarze 1183: parse_cat(mpage, fd);
1184: } else if (man->macroset == MACROSET_MDOC)
1.146 schwarze 1185: parse_mdoc(mpage, &man->meta, man->first);
1.189 schwarze 1186: else
1.146 schwarze 1187: parse_man(mpage, &man->meta, man->first);
1.191 schwarze 1188: if (mpage->desc == NULL) {
1189: mpage->desc = mandoc_strdup(mlink->name);
1190: if (warnings)
1191: say(mlink->file, "No one-line description, "
1192: "using filename \"%s\"", mlink->name);
1193: }
1.88 schwarze 1194:
1.187 schwarze 1195: for (mlink = mpage->mlinks;
1196: mlink != NULL;
1197: mlink = mlink->next) {
1198: putkey(mpage, mlink->name, NAME_FILE);
1199: if (warnings && !use_all)
1.88 schwarze 1200: mlink_check(mpage, mlink);
1.187 schwarze 1201: }
1.6 schwarze 1202:
1.173 schwarze 1203: dbadd(dba, mpage);
1.81 schwarze 1204:
1205: nextpage:
1.47 schwarze 1206: ohash_delete(&strings);
1.90 schwarze 1207: ohash_delete(&names);
1.47 schwarze 1208: }
1209: }
1.6 schwarze 1210:
1.47 schwarze 1211: static void
1.81 schwarze 1212: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1213: {
1214: FILE *stream;
1.190 schwarze 1215: struct mlink *mlink;
1216: char *line, *p, *title, *sec;
1.162 schwarze 1217: size_t linesz, plen, titlesz;
1218: ssize_t len;
1219: int offs;
1.1 schwarze 1220:
1.190 schwarze 1221: mlink = mpage->mlinks;
1222: stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1223: if (stream == NULL) {
1224: if (fd != -1)
1.114 doug 1225: close(fd);
1.47 schwarze 1226: if (warnings)
1.190 schwarze 1227: say(mlink->file, "&fopen");
1.47 schwarze 1228: return;
1229: }
1.1 schwarze 1230:
1.162 schwarze 1231: line = NULL;
1232: linesz = 0;
1233:
1.190 schwarze 1234: /* Parse the section number from the header line. */
1235:
1236: while (getline(&line, &linesz, stream) != -1) {
1237: if (*line == '\n')
1238: continue;
1239: if ((sec = strchr(line, '(')) == NULL)
1240: break;
1241: if ((p = strchr(++sec, ')')) == NULL)
1242: break;
1243: free(mpage->sec);
1244: mpage->sec = mandoc_strndup(sec, p - sec);
1245: if (warnings && *mlink->dsec != '\0' &&
1246: strcasecmp(mpage->sec, mlink->dsec))
1247: say(mlink->file,
1248: "Section \"%s\" manual in %s directory",
1249: mpage->sec, mlink->dsec);
1250: break;
1251: }
1252:
1.47 schwarze 1253: /* Skip to first blank line. */
1.1 schwarze 1254:
1.190 schwarze 1255: while (line == NULL || *line != '\n')
1256: if (getline(&line, &linesz, stream) == -1)
1.47 schwarze 1257: break;
1.1 schwarze 1258:
1.47 schwarze 1259: /*
1260: * Assume the first line that is not indented
1261: * is the first section header. Skip to it.
1262: */
1.1 schwarze 1263:
1.162 schwarze 1264: while (getline(&line, &linesz, stream) != -1)
1265: if (*line != '\n' && *line != ' ')
1.47 schwarze 1266: break;
1.98 schwarze 1267:
1.47 schwarze 1268: /*
1269: * Read up until the next section into a buffer.
1270: * Strip the leading and trailing newline from each read line,
1271: * appending a trailing space.
1272: * Ignore empty (whitespace-only) lines.
1273: */
1.28 schwarze 1274:
1.47 schwarze 1275: titlesz = 0;
1276: title = NULL;
1.38 schwarze 1277:
1.162 schwarze 1278: while ((len = getline(&line, &linesz, stream)) != -1) {
1279: if (*line != ' ')
1.47 schwarze 1280: break;
1.162 schwarze 1281: offs = 0;
1282: while (isspace((unsigned char)line[offs]))
1283: offs++;
1284: if (line[offs] == '\0')
1.47 schwarze 1285: continue;
1.162 schwarze 1286: title = mandoc_realloc(title, titlesz + len - offs);
1287: memcpy(title + titlesz, line + offs, len - offs);
1288: titlesz += len - offs;
1.47 schwarze 1289: title[titlesz - 1] = ' ';
1290: }
1.162 schwarze 1291: free(line);
1.28 schwarze 1292:
1.47 schwarze 1293: /*
1294: * If no page content can be found, or the input line
1295: * is already the next section header, or there is no
1296: * trailing newline, reuse the page title as the page
1297: * description.
1298: */
1.1 schwarze 1299:
1.47 schwarze 1300: if (NULL == title || '\0' == *title) {
1301: if (warnings)
1.190 schwarze 1302: say(mlink->file, "Cannot find NAME section");
1.47 schwarze 1303: fclose(stream);
1304: free(title);
1305: return;
1306: }
1.24 schwarze 1307:
1.162 schwarze 1308: title[titlesz - 1] = '\0';
1.24 schwarze 1309:
1.47 schwarze 1310: /*
1311: * Skip to the first dash.
1312: * Use the remaining line as the description (no more than 70
1313: * bytes).
1314: */
1.28 schwarze 1315:
1.47 schwarze 1316: if (NULL != (p = strstr(title, "- "))) {
1317: for (p += 2; ' ' == *p || '\b' == *p; p++)
1318: /* Skip to next word. */ ;
1319: } else {
1320: if (warnings)
1.191 schwarze 1321: say(mlink->file, "No dash in title line, "
1322: "reusing \"%s\" as one-line description", title);
1.47 schwarze 1323: p = title;
1324: }
1.1 schwarze 1325:
1.47 schwarze 1326: plen = strlen(p);
1.1 schwarze 1327:
1.47 schwarze 1328: /* Strip backspace-encoding from line. */
1.1 schwarze 1329:
1.47 schwarze 1330: while (NULL != (line = memchr(p, '\b', plen))) {
1331: len = line - p;
1332: if (0 == len) {
1333: memmove(line, line + 1, plen--);
1334: continue;
1.98 schwarze 1335: }
1.47 schwarze 1336: memmove(line - 1, line + 1, plen - len);
1337: plen -= 2;
1338: }
1.1 schwarze 1339:
1.206 schwarze 1340: /*
1341: * Cut off excessive one-line descriptions.
1342: * Bad pages are not worth better heuristics.
1343: */
1344:
1345: mpage->desc = mandoc_strndup(p, 150);
1.47 schwarze 1346: fclose(stream);
1347: free(title);
1348: }
1.16 schwarze 1349:
1.47 schwarze 1350: /*
1351: * Put a type/word pair into the word database for this particular file.
1352: */
1353: static void
1.69 schwarze 1354: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1355: {
1356: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1357: }
1358:
1359: /*
1.47 schwarze 1360: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1361: */
1362: static void
1.47 schwarze 1363: putmdockey(const struct mpage *mpage,
1.184 schwarze 1364: const struct roff_node *n, uint64_t m, int taboo)
1.2 schwarze 1365: {
1.16 schwarze 1366:
1.47 schwarze 1367: for ( ; NULL != n; n = n->next) {
1.184 schwarze 1368: if (n->flags & taboo)
1369: continue;
1.47 schwarze 1370: if (NULL != n->child)
1.184 schwarze 1371: putmdockey(mpage, n->child, m, taboo);
1.141 schwarze 1372: if (n->type == ROFFT_TEXT)
1.47 schwarze 1373: putkey(mpage, n->string, m);
1374: }
1375: }
1.16 schwarze 1376:
1.47 schwarze 1377: static void
1.143 schwarze 1378: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1379: const struct roff_node *n)
1.47 schwarze 1380: {
1.142 schwarze 1381: const struct roff_node *head, *body;
1.78 schwarze 1382: char *start, *title;
1.47 schwarze 1383: char byte;
1.78 schwarze 1384: size_t sz;
1.16 schwarze 1385:
1.168 schwarze 1386: if (n == NULL)
1.47 schwarze 1387: return;
1.16 schwarze 1388:
1.47 schwarze 1389: /*
1390: * We're only searching for one thing: the first text child in
1391: * the BODY of a NAME section. Since we don't keep track of
1392: * sections in -man, run some hoops to find out whether we're in
1393: * the correct section or not.
1394: */
1.16 schwarze 1395:
1.141 schwarze 1396: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1397: body = n;
1.168 schwarze 1398: if ((head = body->parent->head) != NULL &&
1399: (head = head->child) != NULL &&
1400: head->next == NULL &&
1.141 schwarze 1401: head->type == ROFFT_TEXT &&
1.168 schwarze 1402: strcmp(head->string, "NAME") == 0 &&
1403: body->child != NULL) {
1.2 schwarze 1404:
1.47 schwarze 1405: /*
1406: * Suck the entire NAME section into memory.
1407: * Yes, we might run away.
1408: * But too many manuals have big, spread-out
1409: * NAME sections over many lines.
1410: */
1.2 schwarze 1411:
1.78 schwarze 1412: title = NULL;
1.147 schwarze 1413: deroff(&title, body);
1.47 schwarze 1414: if (NULL == title)
1415: return;
1.16 schwarze 1416:
1.98 schwarze 1417: /*
1.47 schwarze 1418: * Go through a special heuristic dance here.
1419: * Conventionally, one or more manual names are
1420: * comma-specified prior to a whitespace, then a
1421: * dash, then a description. Try to puzzle out
1422: * the name parts here.
1423: */
1.16 schwarze 1424:
1.78 schwarze 1425: start = title;
1.47 schwarze 1426: for ( ;; ) {
1427: sz = strcspn(start, " ,");
1428: if ('\0' == start[sz])
1429: break;
1.1 schwarze 1430:
1.47 schwarze 1431: byte = start[sz];
1432: start[sz] = '\0';
1.67 schwarze 1433:
1434: /*
1435: * Assume a stray trailing comma in the
1436: * name list if a name begins with a dash.
1437: */
1438:
1439: if ('-' == start[0] ||
1440: ('\\' == start[0] && '-' == start[1]))
1441: break;
1.1 schwarze 1442:
1.90 schwarze 1443: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1444: if ( ! (mpage->name_head_done ||
1445: strcasecmp(start, meta->title))) {
1446: putkey(mpage, start, NAME_HEAD);
1447: mpage->name_head_done = 1;
1448: }
1.1 schwarze 1449:
1.47 schwarze 1450: if (' ' == byte) {
1451: start += sz + 1;
1452: break;
1453: }
1.1 schwarze 1454:
1.47 schwarze 1455: assert(',' == byte);
1456: start += sz + 1;
1457: while (' ' == *start)
1458: start++;
1459: }
1.1 schwarze 1460:
1.78 schwarze 1461: if (start == title) {
1.90 schwarze 1462: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1463: if ( ! (mpage->name_head_done ||
1464: strcasecmp(start, meta->title))) {
1465: putkey(mpage, start, NAME_HEAD);
1466: mpage->name_head_done = 1;
1467: }
1.47 schwarze 1468: free(title);
1469: return;
1470: }
1.1 schwarze 1471:
1.47 schwarze 1472: while (isspace((unsigned char)*start))
1473: start++;
1.1 schwarze 1474:
1.47 schwarze 1475: if (0 == strncmp(start, "-", 1))
1476: start += 1;
1477: else if (0 == strncmp(start, "\\-\\-", 4))
1478: start += 4;
1479: else if (0 == strncmp(start, "\\-", 2))
1480: start += 2;
1481: else if (0 == strncmp(start, "\\(en", 4))
1482: start += 4;
1483: else if (0 == strncmp(start, "\\(em", 4))
1484: start += 4;
1.1 schwarze 1485:
1.47 schwarze 1486: while (' ' == *start)
1487: start++;
1.1 schwarze 1488:
1.206 schwarze 1489: /*
1490: * Cut off excessive one-line descriptions.
1491: * Bad pages are not worth better heuristics.
1492: */
1493:
1494: mpage->desc = mandoc_strndup(start, 150);
1.47 schwarze 1495: free(title);
1496: return;
1497: }
1498: }
1.1 schwarze 1499:
1.47 schwarze 1500: for (n = n->child; n; n = n->next) {
1501: if (NULL != mpage->desc)
1502: break;
1.129 schwarze 1503: parse_man(mpage, meta, n);
1.1 schwarze 1504: }
1505: }
1506:
1507: static void
1.143 schwarze 1508: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1509: const struct roff_node *n)
1.1 schwarze 1510: {
1.208 schwarze 1511: const struct mdoc_handler *handler;
1.1 schwarze 1512:
1.195 schwarze 1513: for (n = n->child; n != NULL; n = n->next) {
1.208 schwarze 1514: if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX)
1.184 schwarze 1515: continue;
1.196 schwarze 1516: assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1.208 schwarze 1517: handler = mdoc_handlers + (n->tok - MDOC_Dd);
1518: if (n->flags & handler->taboo)
1519: continue;
1520:
1.47 schwarze 1521: switch (n->type) {
1.141 schwarze 1522: case ROFFT_ELEM:
1523: case ROFFT_BLOCK:
1524: case ROFFT_HEAD:
1525: case ROFFT_BODY:
1526: case ROFFT_TAIL:
1.208 schwarze 1527: if (handler->fp != NULL &&
1528: (*handler->fp)(mpage, meta, n) == 0)
1.195 schwarze 1529: break;
1.208 schwarze 1530: if (handler->mask)
1.47 schwarze 1531: putmdockey(mpage, n->child,
1.208 schwarze 1532: handler->mask, handler->taboo);
1.47 schwarze 1533: break;
1534: default:
1535: continue;
1536: }
1537: if (NULL != n->child)
1.127 schwarze 1538: parse_mdoc(mpage, meta, n);
1.1 schwarze 1539: }
1540: }
1541:
1.19 schwarze 1542: static int
1.207 schwarze 1543: parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta,
1544: const struct roff_node *n)
1545: {
1546: uint64_t mask;
1547:
1548: mask = TYPE_Fa;
1549: if (n->sec == SEC_SYNOPSIS)
1550: mask |= TYPE_Vt;
1551:
1552: putmdockey(mpage, n->child, mask, 0);
1553: return 0;
1554: }
1555:
1556: static int
1.143 schwarze 1557: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1558: const struct roff_node *n)
1.1 schwarze 1559: {
1.131 schwarze 1560: char *start, *end;
1.1 schwarze 1561: size_t sz;
1.19 schwarze 1562:
1.47 schwarze 1563: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1564: NULL == (n = n->child) ||
1.141 schwarze 1565: n->type != ROFFT_TEXT)
1.150 schwarze 1566: return 0;
1.1 schwarze 1567:
1568: /*
1569: * Only consider those `Fd' macro fields that begin with an
1570: * "inclusion" token (versus, e.g., #define).
1571: */
1.47 schwarze 1572:
1.1 schwarze 1573: if (strcmp("#include", n->string))
1.150 schwarze 1574: return 0;
1.1 schwarze 1575:
1.141 schwarze 1576: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1577: return 0;
1.1 schwarze 1578:
1579: /*
1580: * Strip away the enclosing angle brackets and make sure we're
1581: * not zero-length.
1582: */
1583:
1584: start = n->string;
1585: if ('<' == *start || '"' == *start)
1586: start++;
1587:
1588: if (0 == (sz = strlen(start)))
1.150 schwarze 1589: return 0;
1.1 schwarze 1590:
1591: end = &start[(int)sz - 1];
1592: if ('>' == *end || '"' == *end)
1593: end--;
1594:
1.47 schwarze 1595: if (end > start)
1596: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1597: return 0;
1.1 schwarze 1598: }
1599:
1.133 schwarze 1600: static void
1.142 schwarze 1601: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1602: {
1.69 schwarze 1603: char *cp;
1.133 schwarze 1604: size_t sz;
1.1 schwarze 1605:
1.141 schwarze 1606: if (n->type != ROFFT_TEXT)
1.133 schwarze 1607: return;
1.19 schwarze 1608:
1.133 schwarze 1609: /* Skip function pointer punctuation. */
1.1 schwarze 1610:
1.133 schwarze 1611: cp = n->string;
1612: while (*cp == '(' || *cp == '*')
1.1 schwarze 1613: cp++;
1.133 schwarze 1614: sz = strcspn(cp, "()");
1.1 schwarze 1615:
1.133 schwarze 1616: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1617: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1618: putkeys(mpage, cp, sz, NAME_SYN);
1619: }
1.19 schwarze 1620:
1.133 schwarze 1621: static int
1.143 schwarze 1622: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1623: const struct roff_node *n)
1.133 schwarze 1624: {
1.207 schwarze 1625: uint64_t mask;
1.133 schwarze 1626:
1627: if (n->child == NULL)
1.150 schwarze 1628: return 0;
1.133 schwarze 1629:
1630: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1631:
1.207 schwarze 1632: n = n->child->next;
1633: if (n != NULL && n->type == ROFFT_TEXT) {
1634: mask = TYPE_Fa;
1635: if (n->sec == SEC_SYNOPSIS)
1636: mask |= TYPE_Vt;
1637: putmdockey(mpage, n, mask, 0);
1638: }
1.19 schwarze 1639:
1.150 schwarze 1640: return 0;
1.128 schwarze 1641: }
1642:
1643: static int
1.143 schwarze 1644: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1645: const struct roff_node *n)
1.128 schwarze 1646: {
1.132 schwarze 1647:
1.141 schwarze 1648: if (n->type != ROFFT_HEAD)
1.150 schwarze 1649: return 1;
1.128 schwarze 1650:
1.133 schwarze 1651: if (n->child != NULL)
1652: parse_mdoc_fname(mpage, n->child);
1653:
1.150 schwarze 1654: return 0;
1.1 schwarze 1655: }
1656:
1.19 schwarze 1657: static int
1.164 schwarze 1658: parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1659: const struct roff_node *n)
1660: {
1661: char *cp;
1662:
1663: if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1664: return 0;
1665:
1.168 schwarze 1666: if (n->child != NULL &&
1667: n->child->next == NULL &&
1668: n->child->type == ROFFT_TEXT)
1.164 schwarze 1669: return 1;
1670:
1671: cp = NULL;
1672: deroff(&cp, n);
1673: if (cp != NULL) {
1674: putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1675: n->type == ROFFT_BODY ? TYPE_Va : 0));
1676: free(cp);
1677: }
1678:
1679: return 0;
1680: }
1681:
1682: static int
1.143 schwarze 1683: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1684: const struct roff_node *n)
1.1 schwarze 1685: {
1.47 schwarze 1686: char *cp;
1.1 schwarze 1687:
1688: if (NULL == (n = n->child))
1.150 schwarze 1689: return 0;
1.1 schwarze 1690:
1.47 schwarze 1691: if (NULL == n->next) {
1692: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1693: return 0;
1.47 schwarze 1694: }
1.1 schwarze 1695:
1.77 schwarze 1696: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1697: putkey(mpage, cp, TYPE_Xr);
1698: free(cp);
1.150 schwarze 1699: return 0;
1.1 schwarze 1700: }
1701:
1.19 schwarze 1702: static int
1.143 schwarze 1703: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1704: const struct roff_node *n)
1.1 schwarze 1705: {
1706:
1.141 schwarze 1707: if (n->type == ROFFT_BODY)
1.147 schwarze 1708: deroff(&mpage->desc, n);
1.150 schwarze 1709: return 0;
1.1 schwarze 1710: }
1711:
1.19 schwarze 1712: static int
1.143 schwarze 1713: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1714: const struct roff_node *n)
1.1 schwarze 1715: {
1716:
1.86 schwarze 1717: if (SEC_NAME == n->sec)
1.184 schwarze 1718: putmdockey(mpage, n->child, NAME_TITLE, 0);
1.141 schwarze 1719: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1720: if (n->child == NULL)
1721: putkey(mpage, meta->name, NAME_SYN);
1722: else
1.184 schwarze 1723: putmdockey(mpage, n->child, NAME_SYN, 0);
1.129 schwarze 1724: }
1725: if ( ! (mpage->name_head_done ||
1726: n->child == NULL || n->child->string == NULL ||
1727: strcasecmp(n->child->string, meta->title))) {
1.175 schwarze 1728: putkey(mpage, n->child->string, NAME_HEAD);
1.129 schwarze 1729: mpage->name_head_done = 1;
1.127 schwarze 1730: }
1.150 schwarze 1731: return 0;
1.1 schwarze 1732: }
1733:
1.19 schwarze 1734: static int
1.143 schwarze 1735: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1736: const struct roff_node *n)
1.1 schwarze 1737: {
1738:
1.150 schwarze 1739: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1740: }
1741:
1.47 schwarze 1742: static int
1.143 schwarze 1743: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1744: const struct roff_node *n)
1.1 schwarze 1745: {
1746:
1.150 schwarze 1747: return n->type == ROFFT_HEAD;
1.1 schwarze 1748: }
1749:
1.47 schwarze 1750: /*
1751: * Add a string to the hash table for the current manual.
1752: * Each string has a bitmask telling which macros it belongs to.
1753: * When we finish the manual, we'll dump the table.
1754: */
1.1 schwarze 1755: static void
1.131 schwarze 1756: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1757: {
1.90 schwarze 1758: struct ohash *htab;
1.47 schwarze 1759: struct str *s;
1.68 schwarze 1760: const char *end;
1.47 schwarze 1761: unsigned int slot;
1.131 schwarze 1762: int i, mustfree;
1.1 schwarze 1763:
1.47 schwarze 1764: if (0 == sz)
1765: return;
1.68 schwarze 1766:
1.131 schwarze 1767: mustfree = render_string(&cp, &sz);
1768:
1.90 schwarze 1769: if (TYPE_Nm & v) {
1770: htab = &names;
1771: v &= name_mask;
1.124 schwarze 1772: if (v & NAME_FIRST)
1773: name_mask &= ~NAME_FIRST;
1.90 schwarze 1774: if (debug > 1)
1775: say(mpage->mlinks->file,
1.183 schwarze 1776: "Adding name %*s, bits=0x%llx", (int)sz, cp,
1777: (unsigned long long)v);
1.90 schwarze 1778: } else {
1779: htab = &strings;
1780: if (debug > 1)
1.173 schwarze 1781: for (i = 0; i < KEY_MAX; i++)
1.118 schwarze 1782: if ((uint64_t)1 << i & v)
1.90 schwarze 1783: say(mpage->mlinks->file,
1784: "Adding key %s=%*s",
1.172 schwarze 1785: mansearch_keynames[i], (int)sz, cp);
1.68 schwarze 1786: }
1.47 schwarze 1787:
1788: end = cp + sz;
1.90 schwarze 1789: slot = ohash_qlookupi(htab, cp, &end);
1790: s = ohash_find(htab, slot);
1.1 schwarze 1791:
1.47 schwarze 1792: if (NULL != s && mpage == s->mpage) {
1793: s->mask |= v;
1.1 schwarze 1794: return;
1.47 schwarze 1795: } else if (NULL == s) {
1.101 schwarze 1796: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1797: memcpy(s->key, cp, sz);
1.90 schwarze 1798: ohash_insert(htab, slot, s);
1.47 schwarze 1799: }
1800: s->mpage = mpage;
1801: s->mask = v;
1.131 schwarze 1802:
1803: if (mustfree)
1804: free(cp);
1.1 schwarze 1805: }
1806:
1807: /*
1.47 schwarze 1808: * Take a Unicode codepoint and produce its UTF-8 encoding.
1809: * This isn't the best way to do this, but it works.
1810: * The magic numbers are from the UTF-8 packaging.
1811: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1812: */
1.47 schwarze 1813: static size_t
1814: utf8(unsigned int cp, char out[7])
1.1 schwarze 1815: {
1.47 schwarze 1816: size_t rc;
1.1 schwarze 1817:
1.47 schwarze 1818: rc = 0;
1819: if (cp <= 0x0000007F) {
1820: rc = 1;
1821: out[0] = (char)cp;
1822: } else if (cp <= 0x000007FF) {
1823: rc = 2;
1824: out[0] = (cp >> 6 & 31) | 192;
1825: out[1] = (cp & 63) | 128;
1826: } else if (cp <= 0x0000FFFF) {
1827: rc = 3;
1828: out[0] = (cp >> 12 & 15) | 224;
1829: out[1] = (cp >> 6 & 63) | 128;
1830: out[2] = (cp & 63) | 128;
1831: } else if (cp <= 0x001FFFFF) {
1832: rc = 4;
1833: out[0] = (cp >> 18 & 7) | 240;
1834: out[1] = (cp >> 12 & 63) | 128;
1835: out[2] = (cp >> 6 & 63) | 128;
1836: out[3] = (cp & 63) | 128;
1837: } else if (cp <= 0x03FFFFFF) {
1838: rc = 5;
1839: out[0] = (cp >> 24 & 3) | 248;
1840: out[1] = (cp >> 18 & 63) | 128;
1841: out[2] = (cp >> 12 & 63) | 128;
1842: out[3] = (cp >> 6 & 63) | 128;
1843: out[4] = (cp & 63) | 128;
1844: } else if (cp <= 0x7FFFFFFF) {
1845: rc = 6;
1846: out[0] = (cp >> 30 & 1) | 252;
1847: out[1] = (cp >> 24 & 63) | 128;
1848: out[2] = (cp >> 18 & 63) | 128;
1849: out[3] = (cp >> 12 & 63) | 128;
1850: out[4] = (cp >> 6 & 63) | 128;
1851: out[5] = (cp & 63) | 128;
1852: } else
1.150 schwarze 1853: return 0;
1.19 schwarze 1854:
1.47 schwarze 1855: out[rc] = '\0';
1.150 schwarze 1856: return rc;
1.1 schwarze 1857: }
1858:
1.47 schwarze 1859: /*
1.131 schwarze 1860: * If the string contains escape sequences,
1861: * replace it with an allocated rendering and return 1,
1862: * such that the caller can free it after use.
1863: * Otherwise, do nothing and return 0.
1.47 schwarze 1864: */
1.131 schwarze 1865: static int
1866: render_string(char **public, size_t *psz)
1.1 schwarze 1867: {
1.131 schwarze 1868: const char *src, *scp, *addcp, *seq;
1869: char *dst;
1870: size_t ssz, dsz, addsz;
1.71 schwarze 1871: char utfbuf[7], res[6];
1.131 schwarze 1872: int seqlen, unicode;
1.47 schwarze 1873:
1874: res[0] = '\\';
1875: res[1] = '\t';
1876: res[2] = ASCII_NBRSP;
1877: res[3] = ASCII_HYPH;
1.71 schwarze 1878: res[4] = ASCII_BREAK;
1879: res[5] = '\0';
1.1 schwarze 1880:
1.131 schwarze 1881: src = scp = *public;
1882: ssz = *psz;
1883: dst = NULL;
1884: dsz = 0;
1885:
1886: while (scp < src + *psz) {
1887:
1888: /* Leave normal characters unchanged. */
1889:
1890: if (strchr(res, *scp) == NULL) {
1891: if (dst != NULL)
1892: dst[dsz++] = *scp;
1893: scp++;
1894: continue;
1895: }
1.1 schwarze 1896:
1.47 schwarze 1897: /*
1.131 schwarze 1898: * Found something that requires replacing,
1899: * make sure we have a destination buffer.
1.47 schwarze 1900: */
1.131 schwarze 1901:
1902: if (dst == NULL) {
1903: dst = mandoc_malloc(ssz + 1);
1904: dsz = scp - src;
1905: memcpy(dst, src, dsz);
1.47 schwarze 1906: }
1.39 schwarze 1907:
1.131 schwarze 1908: /* Handle single-char special characters. */
1909:
1910: switch (*scp) {
1911: case '\\':
1912: break;
1.98 schwarze 1913: case '\t':
1914: case ASCII_NBRSP:
1.131 schwarze 1915: dst[dsz++] = ' ';
1916: scp++;
1917: continue;
1918: case ASCII_HYPH:
1919: dst[dsz++] = '-';
1.71 schwarze 1920: /* FALLTHROUGH */
1.98 schwarze 1921: case ASCII_BREAK:
1.131 schwarze 1922: scp++;
1.47 schwarze 1923: continue;
1.71 schwarze 1924: default:
1.131 schwarze 1925: abort();
1.71 schwarze 1926: }
1.39 schwarze 1927:
1.47 schwarze 1928: /*
1.131 schwarze 1929: * Found an escape sequence.
1930: * Read past the slash, then parse it.
1931: * Ignore everything except characters.
1.47 schwarze 1932: */
1.52 schwarze 1933:
1.131 schwarze 1934: scp++;
1935: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1936: continue;
1.39 schwarze 1937:
1.47 schwarze 1938: /*
1.52 schwarze 1939: * Render the special character
1940: * as either UTF-8 or ASCII.
1.47 schwarze 1941: */
1.52 schwarze 1942:
1943: if (write_utf8) {
1.157 schwarze 1944: unicode = mchars_spec2cp(seq, seqlen);
1.131 schwarze 1945: if (unicode <= 0)
1.52 schwarze 1946: continue;
1.131 schwarze 1947: addsz = utf8(unicode, utfbuf);
1948: if (addsz == 0)
1.52 schwarze 1949: continue;
1.131 schwarze 1950: addcp = utfbuf;
1.52 schwarze 1951: } else {
1.157 schwarze 1952: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.131 schwarze 1953: if (addcp == NULL)
1.52 schwarze 1954: continue;
1.131 schwarze 1955: if (*addcp == ASCII_NBRSP) {
1956: addcp = " ";
1957: addsz = 1;
1.52 schwarze 1958: }
1959: }
1.1 schwarze 1960:
1.47 schwarze 1961: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1962:
1.131 schwarze 1963: ssz += addsz;
1964: dst = mandoc_realloc(dst, ssz + 1);
1965: memcpy(dst + dsz, addcp, addsz);
1966: dsz += addsz;
1967: }
1968: if (dst != NULL) {
1969: *public = dst;
1970: *psz = dsz;
1971: }
1972:
1973: /* Trim trailing whitespace and NUL-terminate. */
1974:
1975: while (*psz > 0 && (*public)[*psz - 1] == ' ')
1976: --*psz;
1977: if (dst != NULL) {
1978: (*public)[*psz] = '\0';
1.150 schwarze 1979: return 1;
1.131 schwarze 1980: } else
1.150 schwarze 1981: return 0;
1.1 schwarze 1982: }
1983:
1.75 schwarze 1984: static void
1985: dbadd_mlink(const struct mlink *mlink)
1986: {
1.173 schwarze 1987: dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
1988: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
1989: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
1990: dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
1991: dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
1.75 schwarze 1992: }
1993:
1.11 schwarze 1994: /*
1.47 schwarze 1995: * Flush the current page's terms (and their bits) into the database.
1.53 schwarze 1996: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 1997: */
1998: static void
1.173 schwarze 1999: dbadd(struct dba *dba, struct mpage *mpage)
1.11 schwarze 2000: {
1.47 schwarze 2001: struct mlink *mlink;
2002: struct str *key;
1.131 schwarze 2003: char *cp;
1.173 schwarze 2004: uint64_t mask;
1.47 schwarze 2005: size_t i;
2006: unsigned int slot;
1.131 schwarze 2007: int mustfree;
1.47 schwarze 2008:
1.85 schwarze 2009: mlink = mpage->mlinks;
1.11 schwarze 2010:
1.85 schwarze 2011: if (nodb) {
1.104 schwarze 2012: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2013: key = ohash_next(&names, &slot))
1.104 schwarze 2014: free(key);
2015: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2016: key = ohash_next(&strings, &slot))
1.104 schwarze 2017: free(key);
1.102 schwarze 2018: if (0 == debug)
2019: return;
1.85 schwarze 2020: while (NULL != mlink) {
2021: fputs(mlink->name, stdout);
2022: if (NULL == mlink->next ||
2023: strcmp(mlink->dsec, mlink->next->dsec) ||
2024: strcmp(mlink->fsec, mlink->next->fsec) ||
2025: strcmp(mlink->arch, mlink->next->arch)) {
2026: putchar('(');
2027: if ('\0' == *mlink->dsec)
2028: fputs(mlink->fsec, stdout);
2029: else
2030: fputs(mlink->dsec, stdout);
2031: if ('\0' != *mlink->arch)
2032: printf("/%s", mlink->arch);
2033: putchar(')');
2034: }
2035: mlink = mlink->next;
2036: if (NULL != mlink)
2037: fputs(", ", stdout);
2038: }
1.89 schwarze 2039: printf(" - %s\n", mpage->desc);
1.11 schwarze 2040: return;
1.85 schwarze 2041: }
2042:
2043: if (debug)
2044: say(mlink->file, "Adding to database");
1.47 schwarze 2045:
1.131 schwarze 2046: cp = mpage->desc;
2047: i = strlen(cp);
2048: mustfree = render_string(&cp, &i);
1.176 schwarze 2049: mpage->dba = dba_page_new(dba->pages,
1.173 schwarze 2050: *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2051: cp, mlink->file, mpage->form);
1.131 schwarze 2052: if (mustfree)
2053: free(cp);
1.176 schwarze 2054: dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
1.47 schwarze 2055:
1.173 schwarze 2056: while (mlink != NULL) {
1.75 schwarze 2057: dbadd_mlink(mlink);
1.85 schwarze 2058: mlink = mlink->next;
2059: }
1.47 schwarze 2060:
1.90 schwarze 2061: for (key = ohash_first(&names, &slot); NULL != key;
2062: key = ohash_next(&names, &slot)) {
2063: assert(key->mpage == mpage);
1.173 schwarze 2064: dba_page_alias(mpage->dba, key->key, key->mask);
1.90 schwarze 2065: free(key);
2066: }
1.47 schwarze 2067: for (key = ohash_first(&strings, &slot); NULL != key;
2068: key = ohash_next(&strings, &slot)) {
2069: assert(key->mpage == mpage);
1.173 schwarze 2070: i = 0;
2071: for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2072: if (key->mask & mask)
2073: dba_macro_add(dba->macros, i,
2074: key->key, mpage->dba);
2075: i++;
2076: }
1.47 schwarze 2077: free(key);
1.33 schwarze 2078: }
1.47 schwarze 2079: }
1.41 deraadt 2080:
1.47 schwarze 2081: static void
1.173 schwarze 2082: dbprune(struct dba *dba)
1.47 schwarze 2083: {
1.173 schwarze 2084: struct dba_array *page, *files;
2085: char *file;
1.11 schwarze 2086:
1.173 schwarze 2087: dba_array_FOREACH(dba->pages, page) {
2088: files = dba_array_get(page, DBP_FILE);
2089: dba_array_FOREACH(files, file) {
2090: if (*file < ' ')
2091: file++;
2092: if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2093: file)) != NULL) {
2094: if (debug)
2095: say(file, "Deleting from database");
2096: dba_array_del(dba->pages);
2097: break;
2098: }
1.63 schwarze 2099: }
1.11 schwarze 2100: }
1.47 schwarze 2101: }
1.22 schwarze 2102:
1.47 schwarze 2103: /*
1.173 schwarze 2104: * Write the database from memory to disk.
1.47 schwarze 2105: */
2106: static void
1.173 schwarze 2107: dbwrite(struct dba *dba)
1.47 schwarze 2108: {
1.204 schwarze 2109: struct stat sb1, sb2;
2110: char tfn[33], *cp1, *cp2;
2111: off_t i;
2112: int fd1, fd2;
1.11 schwarze 2113:
1.199 schwarze 2114: /*
2115: * Do not write empty databases, and delete existing ones
2116: * when makewhatis -u causes them to become empty.
2117: */
2118:
2119: dba_array_start(dba->pages);
2120: if (dba_array_next(dba->pages) == NULL) {
1.201 schwarze 2121: if (unlink(MANDOC_DB) == -1 && errno != ENOENT)
1.199 schwarze 2122: say(MANDOC_DB, "&unlink");
2123: return;
2124: }
2125:
2126: /*
2127: * Build the database in a temporary file,
2128: * then atomically move it into place.
2129: */
2130:
1.173 schwarze 2131: if (dba_write(MANDOC_DB "~", dba) != -1) {
2132: if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2133: exitcode = (int)MANDOCLEVEL_SYSERR;
2134: say(MANDOC_DB, "&rename");
2135: unlink(MANDOC_DB "~");
2136: }
1.47 schwarze 2137: return;
1.28 schwarze 2138: }
1.199 schwarze 2139:
2140: /*
2141: * We lack write permission and cannot replace the database
2142: * file, but let's at least check whether the data changed.
2143: */
1.22 schwarze 2144:
1.173 schwarze 2145: (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2146: if (mkdtemp(tfn) == NULL) {
2147: exitcode = (int)MANDOCLEVEL_SYSERR;
2148: say("", "&%s", tfn);
1.47 schwarze 2149: return;
1.173 schwarze 2150: }
1.205 schwarze 2151: cp1 = cp2 = MAP_FAILED;
1.204 schwarze 2152: fd1 = fd2 = -1;
1.173 schwarze 2153: (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2154: if (dba_write(tfn, dba) == -1) {
2155: say(tfn, "&dba_write");
1.204 schwarze 2156: goto err;
1.72 schwarze 2157: }
1.204 schwarze 2158: if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) {
2159: say(MANDOC_DB, "&open");
2160: goto err;
2161: }
2162: if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) {
2163: say(tfn, "&open");
2164: goto err;
2165: }
2166: if (fstat(fd1, &sb1) == -1) {
2167: say(MANDOC_DB, "&fstat");
2168: goto err;
2169: }
2170: if (fstat(fd2, &sb2) == -1) {
2171: say(tfn, "&fstat");
2172: goto err;
2173: }
2174: if (sb1.st_size != sb2.st_size)
2175: goto err;
2176: if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE,
1.205 schwarze 2177: fd1, 0)) == MAP_FAILED) {
1.204 schwarze 2178: say(MANDOC_DB, "&mmap");
2179: goto err;
2180: }
2181: if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE,
1.205 schwarze 2182: fd2, 0)) == MAP_FAILED) {
1.204 schwarze 2183: say(tfn, "&mmap");
2184: goto err;
2185: }
2186: for (i = 0; i < sb1.st_size; i++)
2187: if (cp1[i] != cp2[i])
2188: goto err;
2189: goto out;
2190:
2191: err:
2192: exitcode = (int)MANDOCLEVEL_SYSERR;
2193: say(MANDOC_DB, "Data changed, but cannot replace database");
1.72 schwarze 2194:
1.173 schwarze 2195: out:
1.205 schwarze 2196: if (cp1 != MAP_FAILED)
1.204 schwarze 2197: munmap(cp1, sb1.st_size);
1.205 schwarze 2198: if (cp2 != MAP_FAILED)
1.204 schwarze 2199: munmap(cp2, sb2.st_size);
2200: if (fd1 != -1)
2201: close(fd1);
2202: if (fd2 != -1)
2203: close(fd2);
1.203 schwarze 2204: unlink(tfn);
1.173 schwarze 2205: *strrchr(tfn, '/') = '\0';
1.203 schwarze 2206: rmdir(tfn);
1.47 schwarze 2207: }
1.6 schwarze 2208:
1.47 schwarze 2209: static int
1.120 schwarze 2210: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2211: {
2212: static char startdir[PATH_MAX];
1.110 schwarze 2213: static int getcwd_status; /* 1 = ok, 2 = failure */
2214: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2215: char *cp;
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: */
2224: if (0 == getcwd_status) {
2225: if (NULL == getcwd(startdir, sizeof(startdir))) {
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';
2238:
2239: /*
2240: * If and only if the directory was changed earlier and
2241: * the next directory to process is given as a relative path,
2242: * first go back, or bail out if that is impossible.
1.47 schwarze 2243: */
1.110 schwarze 2244: if (chdir_status && '/' != *targetdir) {
2245: if (2 == getcwd_status) {
1.47 schwarze 2246: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2247: say("", "getcwd: %s", startdir);
1.150 schwarze 2248: return 0;
1.47 schwarze 2249: }
1.110 schwarze 2250: if (-1 == chdir(startdir)) {
1.47 schwarze 2251: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2252: say("", "&chdir %s", startdir);
1.150 schwarze 2253: return 0;
1.2 schwarze 2254: }
2255: }
1.110 schwarze 2256:
2257: /*
2258: * Always resolve basedir to the canonicalized absolute
2259: * pathname and append a trailing slash, such that
2260: * we can reliably check whether files are inside.
2261: */
1.47 schwarze 2262: if (NULL == realpath(targetdir, basedir)) {
1.120 schwarze 2263: if (report_baddir || errno != ENOENT) {
2264: exitcode = (int)MANDOCLEVEL_BADARG;
2265: say("", "&%s: realpath", targetdir);
2266: }
1.150 schwarze 2267: return 0;
1.47 schwarze 2268: } else if (-1 == chdir(basedir)) {
1.120 schwarze 2269: if (report_baddir || errno != ENOENT) {
2270: exitcode = (int)MANDOCLEVEL_BADARG;
2271: say("", "&chdir");
2272: }
1.150 schwarze 2273: return 0;
1.109 schwarze 2274: }
1.110 schwarze 2275: chdir_status = 1;
1.109 schwarze 2276: cp = strchr(basedir, '\0');
2277: if ('/' != cp[-1]) {
2278: if (cp - basedir >= PATH_MAX - 1) {
2279: exitcode = (int)MANDOCLEVEL_SYSERR;
2280: say("", "Filename too long");
1.150 schwarze 2281: return 0;
1.109 schwarze 2282: }
2283: *cp++ = '/';
2284: *cp = '\0';
1.47 schwarze 2285: }
1.150 schwarze 2286: return 1;
1.2 schwarze 2287: }
2288:
2289: static void
1.47 schwarze 2290: say(const char *file, const char *format, ...)
1.2 schwarze 2291: {
1.47 schwarze 2292: va_list ap;
1.80 schwarze 2293: int use_errno;
1.2 schwarze 2294:
1.47 schwarze 2295: if ('\0' != *basedir)
2296: fprintf(stderr, "%s", basedir);
2297: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2298: fputc('/', stderr);
1.47 schwarze 2299: if ('\0' != *file)
2300: fprintf(stderr, "%s", file);
1.31 schwarze 2301:
1.80 schwarze 2302: use_errno = 1;
2303: if (NULL != format) {
2304: switch (*format) {
1.98 schwarze 2305: case '&':
1.80 schwarze 2306: format++;
2307: break;
1.98 schwarze 2308: case '\0':
1.80 schwarze 2309: format = NULL;
2310: break;
2311: default:
2312: use_errno = 0;
2313: break;
2314: }
2315: }
2316: if (NULL != format) {
2317: if ('\0' != *basedir || '\0' != *file)
2318: fputs(": ", stderr);
2319: va_start(ap, format);
2320: vfprintf(stderr, format, ap);
2321: va_end(ap);
2322: }
2323: if (use_errno) {
2324: if ('\0' != *basedir || '\0' != *file || NULL != format)
2325: fputs(": ", stderr);
1.47 schwarze 2326: perror(NULL);
1.80 schwarze 2327: } else
2328: fputc('\n', stderr);
1.1 schwarze 2329: }
![[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