Annotation of mandoc/mandocdb.c, Revision 1.259
1.259 ! schwarze 1: /* $Id: mandocdb.c,v 1.258 2018/02/23 18:25:57 schwarze Exp $ */
1.1 kristaps 2: /*
1.48 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.259 ! schwarze 4: * Copyright (c) 2011-2018 Ingo Schwarze <schwarze@openbsd.org>
1.231 schwarze 5: * Copyright (c) 2016 Ed Maste <emaste@freebsd.org>
1.1 kristaps 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: *
1.187 schwarze 11: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
1.1 kristaps 12: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
1.187 schwarze 13: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
1.1 kristaps 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: */
19: #include "config.h"
20:
1.162 schwarze 21: #include <sys/types.h>
1.255 schwarze 22: #include <sys/mman.h>
1.50 kristaps 23: #include <sys/stat.h>
1.1 kristaps 24:
25: #include <assert.h>
1.43 kristaps 26: #include <ctype.h>
1.208 schwarze 27: #if HAVE_ERR
1.198 schwarze 28: #include <err.h>
1.208 schwarze 29: #endif
1.63 schwarze 30: #include <errno.h>
1.1 kristaps 31: #include <fcntl.h>
1.158 schwarze 32: #if HAVE_FTS
1.50 kristaps 33: #include <fts.h>
1.157 schwarze 34: #else
35: #include "compat_fts.h"
36: #endif
1.58 schwarze 37: #include <limits.h>
1.218 kristaps 38: #if HAVE_SANDBOX_INIT
39: #include <sandbox.h>
40: #endif
1.221 schwarze 41: #include <stdarg.h>
1.50 kristaps 42: #include <stddef.h>
1.59 schwarze 43: #include <stdio.h>
1.1 kristaps 44: #include <stdint.h>
45: #include <stdlib.h>
46: #include <string.h>
1.17 schwarze 47: #include <unistd.h>
1.1 kristaps 48:
1.188 schwarze 49: #include "mandoc_aux.h"
1.202 schwarze 50: #include "mandoc_ohash.h"
1.188 schwarze 51: #include "mandoc.h"
52: #include "roff.h"
1.50 kristaps 53: #include "mdoc.h"
1.1 kristaps 54: #include "man.h"
1.187 schwarze 55: #include "manconf.h"
1.55 kristaps 56: #include "mansearch.h"
1.221 schwarze 57: #include "dba_array.h"
58: #include "dba.h"
1.1 kristaps 59:
1.111 schwarze 60: extern const char *const mansearch_keynames[];
61:
1.50 kristaps 62: enum op {
63: OP_DEFAULT = 0, /* new dbs from dir list or default config */
64: OP_CONFFILE, /* new databases from custom config file */
65: OP_UPDATE, /* delete/add entries in existing database */
66: OP_DELETE, /* delete entries from existing database */
67: OP_TEST /* change no databases, report potential problems */
1.38 schwarze 68: };
69:
1.50 kristaps 70: struct str {
1.78 schwarze 71: const struct mpage *mpage; /* if set, the owning parse */
1.50 kristaps 72: uint64_t mask; /* bitmask in sequence */
1.176 schwarze 73: char key[]; /* rendered text */
1.38 schwarze 74: };
75:
1.79 schwarze 76: struct inodev {
77: ino_t st_ino;
78: dev_t st_dev;
1.50 kristaps 79: };
1.5 kristaps 80:
1.78 schwarze 81: struct mpage {
1.79 schwarze 82: struct inodev inodev; /* used for hashing routine */
1.221 schwarze 83: struct dba_array *dba;
1.82 schwarze 84: char *sec; /* section from file content */
85: char *arch; /* architecture from file content */
86: char *title; /* title from file content */
87: char *desc; /* description from file content */
1.231 schwarze 88: struct mpage *next; /* singly linked list */
1.82 schwarze 89: struct mlink *mlinks; /* singly linked list */
1.174 schwarze 90: int name_head_done;
1.221 schwarze 91: enum form form; /* format from file content */
1.82 schwarze 92: };
93:
94: struct mlink {
1.58 schwarze 95: char file[PATH_MAX]; /* filename rel. to manpath */
1.82 schwarze 96: char *dsec; /* section from directory */
97: char *arch; /* architecture from directory */
98: char *name; /* name from file name (not empty) */
99: char *fsec; /* section from file name suffix */
1.85 schwarze 100: struct mlink *next; /* singly linked list */
1.118 schwarze 101: struct mpage *mpage; /* parent */
1.124 schwarze 102: int gzip; /* filename has a .gz suffix */
1.221 schwarze 103: enum form dform; /* format from directory */
104: enum form fform; /* format from file name suffix */
1.1 kristaps 105: };
106:
1.190 schwarze 107: typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
1.189 schwarze 108: const struct roff_node *);
1.1 kristaps 109:
1.50 kristaps 110: struct mdoc_handler {
111: mdoc_fp fp; /* optional handler */
112: uint64_t mask; /* set unless handler returns 0 */
1.235 schwarze 113: int taboo; /* node flags that must not be set */
1.1 kristaps 114: };
1.219 schwarze 115:
116:
117: int mandocdb(int, char *[]);
1.1 kristaps 118:
1.221 schwarze 119: static void dbadd(struct dba *, struct mpage *);
1.118 schwarze 120: static void dbadd_mlink(const struct mlink *mlink);
1.221 schwarze 121: static void dbprune(struct dba *);
122: static void dbwrite(struct dba *);
1.59 schwarze 123: static void filescan(const char *);
1.232 schwarze 124: #if HAVE_FTS_COMPARE_CONST
125: static int fts_compare(const FTSENT *const *, const FTSENT *const *);
126: #else
1.231 schwarze 127: static int fts_compare(const FTSENT **, const FTSENT **);
1.232 schwarze 128: #endif
1.84 schwarze 129: static void mlink_add(struct mlink *, const struct stat *);
1.131 schwarze 130: static void mlink_check(struct mpage *, struct mlink *);
1.82 schwarze 131: static void mlink_free(struct mlink *);
1.89 schwarze 132: static void mlinks_undupe(struct mpage *);
1.78 schwarze 133: static void mpages_free(void);
1.221 schwarze 134: static void mpages_merge(struct dba *, struct mparse *);
1.124 schwarze 135: static void parse_cat(struct mpage *, int);
1.190 schwarze 136: static void parse_man(struct mpage *, const struct roff_meta *,
1.189 schwarze 137: const struct roff_node *);
1.190 schwarze 138: static void parse_mdoc(struct mpage *, const struct roff_meta *,
1.189 schwarze 139: const struct roff_node *);
1.190 schwarze 140: static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
1.189 schwarze 141: const struct roff_node *);
1.258 schwarze 142: static int parse_mdoc_Fa(struct mpage *, const struct roff_meta *,
143: const struct roff_node *);
1.190 schwarze 144: static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
1.189 schwarze 145: const struct roff_node *);
146: static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
1.190 schwarze 147: static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
1.189 schwarze 148: const struct roff_node *);
1.190 schwarze 149: static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
1.189 schwarze 150: const struct roff_node *);
1.190 schwarze 151: static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
1.189 schwarze 152: const struct roff_node *);
1.190 schwarze 153: static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
1.189 schwarze 154: const struct roff_node *);
1.190 schwarze 155: static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
1.189 schwarze 156: const struct roff_node *);
1.211 schwarze 157: static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
158: const struct roff_node *);
1.190 schwarze 159: static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
1.189 schwarze 160: const struct roff_node *);
1.112 schwarze 161: static void putkey(const struct mpage *, char *, uint64_t);
1.176 schwarze 162: static void putkeys(const struct mpage *, char *, size_t, uint64_t);
1.78 schwarze 163: static void putmdockey(const struct mpage *,
1.235 schwarze 164: const struct roff_node *, uint64_t, int);
1.176 schwarze 165: static int render_string(char **, size_t *);
1.220 schwarze 166: static void say(const char *, const char *, ...)
1.243 schwarze 167: __attribute__((__format__ (__printf__, 2, 3)));
1.165 schwarze 168: static int set_basedir(const char *, int);
1.59 schwarze 169: static int treescan(void);
1.50 kristaps 170: static size_t utf8(unsigned int, char [7]);
171:
1.102 schwarze 172: static int nodb; /* no database changes */
1.116 schwarze 173: static int mparse_options; /* abort the parse early */
1.141 schwarze 174: static int use_all; /* use all found files */
175: static int debug; /* print what we're doing */
176: static int warnings; /* warn about crap */
1.95 schwarze 177: static int write_utf8; /* write UTF-8 output; else ASCII */
1.59 schwarze 178: static int exitcode; /* to be returned by main */
1.141 schwarze 179: static enum op op; /* operational mode */
1.59 schwarze 180: static char basedir[PATH_MAX]; /* current base directory */
1.231 schwarze 181: static struct mpage *mpage_head; /* list of distinct manual pages */
1.78 schwarze 182: static struct ohash mpages; /* table of distinct manual pages */
1.83 schwarze 183: static struct ohash mlinks; /* table of directory entries */
1.133 schwarze 184: static struct ohash names; /* table of all names */
1.50 kristaps 185: static struct ohash strings; /* table of all strings */
1.133 schwarze 186: static uint64_t name_mask;
1.25 schwarze 187:
1.259 ! schwarze 188: static const struct mdoc_handler mdoc_handlers[MDOC_MAX - MDOC_Dd] = {
1.235 schwarze 189: { NULL, 0, NODE_NOPRT }, /* Dd */
190: { NULL, 0, NODE_NOPRT }, /* Dt */
191: { NULL, 0, NODE_NOPRT }, /* Os */
192: { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
193: { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
194: { NULL, 0, 0 }, /* Pp */
195: { NULL, 0, 0 }, /* D1 */
196: { NULL, 0, 0 }, /* Dl */
197: { NULL, 0, 0 }, /* Bd */
198: { NULL, 0, 0 }, /* Ed */
199: { NULL, 0, 0 }, /* Bl */
200: { NULL, 0, 0 }, /* El */
201: { NULL, 0, 0 }, /* It */
202: { NULL, 0, 0 }, /* Ad */
203: { NULL, TYPE_An, 0 }, /* An */
1.246 schwarze 204: { NULL, 0, 0 }, /* Ap */
1.235 schwarze 205: { NULL, TYPE_Ar, 0 }, /* Ar */
206: { NULL, TYPE_Cd, 0 }, /* Cd */
207: { NULL, TYPE_Cm, 0 }, /* Cm */
208: { NULL, TYPE_Dv, 0 }, /* Dv */
209: { NULL, TYPE_Er, 0 }, /* Er */
210: { NULL, TYPE_Ev, 0 }, /* Ev */
211: { NULL, 0, 0 }, /* Ex */
1.258 schwarze 212: { parse_mdoc_Fa, 0, 0 }, /* Fa */
1.235 schwarze 213: { parse_mdoc_Fd, 0, 0 }, /* Fd */
214: { NULL, TYPE_Fl, 0 }, /* Fl */
215: { parse_mdoc_Fn, 0, 0 }, /* Fn */
1.258 schwarze 216: { NULL, TYPE_Ft | TYPE_Vt, 0 }, /* Ft */
1.235 schwarze 217: { NULL, TYPE_Ic, 0 }, /* Ic */
218: { NULL, TYPE_In, 0 }, /* In */
219: { NULL, TYPE_Li, 0 }, /* Li */
220: { parse_mdoc_Nd, 0, 0 }, /* Nd */
221: { parse_mdoc_Nm, 0, 0 }, /* Nm */
222: { NULL, 0, 0 }, /* Op */
223: { NULL, 0, 0 }, /* Ot */
224: { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
225: { NULL, 0, 0 }, /* Rv */
226: { NULL, TYPE_St, 0 }, /* St */
227: { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
228: { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
229: { parse_mdoc_Xr, 0, 0 }, /* Xr */
230: { NULL, 0, 0 }, /* %A */
231: { NULL, 0, 0 }, /* %B */
232: { NULL, 0, 0 }, /* %D */
233: { NULL, 0, 0 }, /* %I */
234: { NULL, 0, 0 }, /* %J */
235: { NULL, 0, 0 }, /* %N */
236: { NULL, 0, 0 }, /* %O */
237: { NULL, 0, 0 }, /* %P */
238: { NULL, 0, 0 }, /* %R */
239: { NULL, 0, 0 }, /* %T */
240: { NULL, 0, 0 }, /* %V */
241: { NULL, 0, 0 }, /* Ac */
242: { NULL, 0, 0 }, /* Ao */
243: { NULL, 0, 0 }, /* Aq */
1.237 schwarze 244: { NULL, TYPE_At, 0 }, /* At */
1.235 schwarze 245: { NULL, 0, 0 }, /* Bc */
246: { NULL, 0, 0 }, /* Bf */
247: { NULL, 0, 0 }, /* Bo */
248: { NULL, 0, 0 }, /* Bq */
249: { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
1.236 schwarze 250: { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
1.235 schwarze 251: { NULL, 0, 0 }, /* Db */
252: { NULL, 0, 0 }, /* Dc */
253: { NULL, 0, 0 }, /* Do */
254: { NULL, 0, 0 }, /* Dq */
255: { NULL, 0, 0 }, /* Ec */
256: { NULL, 0, 0 }, /* Ef */
257: { NULL, TYPE_Em, 0 }, /* Em */
258: { NULL, 0, 0 }, /* Eo */
259: { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
260: { NULL, TYPE_Ms, 0 }, /* Ms */
261: { NULL, 0, 0 }, /* No */
262: { NULL, 0, 0 }, /* Ns */
263: { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
264: { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
265: { NULL, 0, 0 }, /* Pc */
266: { NULL, 0, 0 }, /* Pf */
267: { NULL, 0, 0 }, /* Po */
268: { NULL, 0, 0 }, /* Pq */
269: { NULL, 0, 0 }, /* Qc */
270: { NULL, 0, 0 }, /* Ql */
271: { NULL, 0, 0 }, /* Qo */
272: { NULL, 0, 0 }, /* Qq */
273: { NULL, 0, 0 }, /* Re */
274: { NULL, 0, 0 }, /* Rs */
275: { NULL, 0, 0 }, /* Sc */
276: { NULL, 0, 0 }, /* So */
277: { NULL, 0, 0 }, /* Sq */
278: { NULL, 0, 0 }, /* Sm */
279: { NULL, 0, 0 }, /* Sx */
280: { NULL, TYPE_Sy, 0 }, /* Sy */
281: { NULL, TYPE_Tn, 0 }, /* Tn */
282: { NULL, 0, NODE_NOSRC }, /* Ux */
283: { NULL, 0, 0 }, /* Xc */
284: { NULL, 0, 0 }, /* Xo */
285: { parse_mdoc_Fo, 0, 0 }, /* Fo */
286: { NULL, 0, 0 }, /* Fc */
287: { NULL, 0, 0 }, /* Oo */
288: { NULL, 0, 0 }, /* Oc */
289: { NULL, 0, 0 }, /* Bk */
290: { NULL, 0, 0 }, /* Ek */
291: { NULL, 0, 0 }, /* Bt */
292: { NULL, 0, 0 }, /* Hf */
293: { NULL, 0, 0 }, /* Fr */
294: { NULL, 0, 0 }, /* Ud */
1.237 schwarze 295: { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */
1.235 schwarze 296: { NULL, 0, 0 }, /* Lp */
297: { NULL, TYPE_Lk, 0 }, /* Lk */
298: { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
299: { NULL, 0, 0 }, /* Brq */
300: { NULL, 0, 0 }, /* Bro */
301: { NULL, 0, 0 }, /* Brc */
302: { NULL, 0, 0 }, /* %C */
303: { NULL, 0, 0 }, /* Es */
304: { NULL, 0, 0 }, /* En */
305: { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
306: { NULL, 0, 0 }, /* %Q */
307: { NULL, 0, 0 }, /* %U */
308: { NULL, 0, 0 }, /* Ta */
1.1 kristaps 309: };
310:
1.141 schwarze 311:
1.1 kristaps 312: int
1.179 schwarze 313: mandocdb(int argc, char *argv[])
1.1 kristaps 314: {
1.187 schwarze 315: struct manconf conf;
316: struct mparse *mp;
1.221 schwarze 317: struct dba *dba;
1.206 schwarze 318: const char *path_arg, *progname;
1.187 schwarze 319: size_t j, sz;
1.59 schwarze 320: int ch, i;
1.50 kristaps 321:
1.207 schwarze 322: #if HAVE_PLEDGE
1.255 schwarze 323: if (pledge("stdio rpath wpath cpath", NULL) == -1) {
1.214 schwarze 324: warn("pledge");
1.218 kristaps 325: return (int)MANDOCLEVEL_SYSERR;
326: }
327: #endif
328:
329: #if HAVE_SANDBOX_INIT
330: if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) {
331: warnx("sandbox_init");
1.207 schwarze 332: return (int)MANDOCLEVEL_SYSERR;
333: }
334: #endif
335:
1.187 schwarze 336: memset(&conf, 0, sizeof(conf));
1.50 kristaps 337:
338: /*
1.141 schwarze 339: * We accept a few different invocations.
1.50 kristaps 340: * The CHECKOP macro makes sure that invocation styles don't
341: * clobber each other.
342: */
343: #define CHECKOP(_op, _ch) do \
344: if (OP_DEFAULT != (_op)) { \
1.198 schwarze 345: warnx("-%c: Conflicting option", (_ch)); \
1.50 kristaps 346: goto usage; \
347: } while (/*CONSTCOND*/0)
1.10 kristaps 348:
1.59 schwarze 349: path_arg = NULL;
1.38 schwarze 350: op = OP_DEFAULT;
1.1 kristaps 351:
1.126 schwarze 352: while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
1.1 kristaps 353: switch (ch) {
1.141 schwarze 354: case 'a':
1.12 schwarze 355: use_all = 1;
356: break;
1.141 schwarze 357: case 'C':
1.50 kristaps 358: CHECKOP(op, ch);
1.59 schwarze 359: path_arg = optarg;
1.38 schwarze 360: op = OP_CONFFILE;
1.34 schwarze 361: break;
1.141 schwarze 362: case 'D':
1.125 schwarze 363: debug++;
364: break;
1.141 schwarze 365: case 'd':
1.50 kristaps 366: CHECKOP(op, ch);
1.59 schwarze 367: path_arg = optarg;
1.5 kristaps 368: op = OP_UPDATE;
369: break;
1.141 schwarze 370: case 'n':
1.50 kristaps 371: nodb = 1;
372: break;
1.141 schwarze 373: case 'p':
1.126 schwarze 374: warnings = 1;
375: break;
1.141 schwarze 376: case 'Q':
1.116 schwarze 377: mparse_options |= MPARSE_QUICK;
1.102 schwarze 378: break;
1.141 schwarze 379: case 'T':
1.95 schwarze 380: if (strcmp(optarg, "utf8")) {
1.198 schwarze 381: warnx("-T%s: Unsupported output format",
382: optarg);
1.95 schwarze 383: goto usage;
384: }
385: write_utf8 = 1;
386: break;
1.141 schwarze 387: case 't':
1.50 kristaps 388: CHECKOP(op, ch);
1.38 schwarze 389: dup2(STDOUT_FILENO, STDERR_FILENO);
390: op = OP_TEST;
1.50 kristaps 391: nodb = warnings = 1;
1.38 schwarze 392: break;
1.141 schwarze 393: case 'u':
1.50 kristaps 394: CHECKOP(op, ch);
1.59 schwarze 395: path_arg = optarg;
1.5 kristaps 396: op = OP_DELETE;
397: break;
1.141 schwarze 398: case 'v':
1.126 schwarze 399: /* Compatibility with espie@'s makewhatis. */
1.38 schwarze 400: break;
1.1 kristaps 401: default:
1.38 schwarze 402: goto usage;
1.1 kristaps 403: }
404:
405: argc -= optind;
406: argv += optind;
407:
1.207 schwarze 408: #if HAVE_PLEDGE
1.210 schwarze 409: if (nodb) {
410: if (pledge("stdio rpath", NULL) == -1) {
1.214 schwarze 411: warn("pledge");
1.210 schwarze 412: return (int)MANDOCLEVEL_SYSERR;
413: }
1.207 schwarze 414: }
415: #endif
416:
1.38 schwarze 417: if (OP_CONFFILE == op && argc > 0) {
1.198 schwarze 418: warnx("-C: Too many arguments");
1.38 schwarze 419: goto usage;
420: }
421:
1.59 schwarze 422: exitcode = (int)MANDOCLEVEL_OK;
1.203 schwarze 423: mchars_alloc();
1.253 schwarze 424: mp = mparse_alloc(mparse_options, MANDOCERR_MAX, NULL,
1.251 schwarze 425: MANDOC_OS_OTHER, NULL);
1.202 schwarze 426: mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
427: mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
1.50 kristaps 428:
429: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.59 schwarze 430:
1.50 kristaps 431: /*
1.150 schwarze 432: * Most of these deal with a specific directory.
1.140 schwarze 433: * Jump into that directory first.
1.50 kristaps 434: */
1.165 schwarze 435: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.50 kristaps 436: goto out;
1.140 schwarze 437:
1.225 schwarze 438: dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
439: if (dba != NULL) {
1.140 schwarze 440: /*
441: * The existing database is usable. Process
442: * all files specified on the command-line.
443: */
444: use_all = 1;
1.138 schwarze 445: for (i = 0; i < argc; i++)
446: filescan(argv[i]);
1.225 schwarze 447: if (nodb == 0)
1.221 schwarze 448: dbprune(dba);
1.138 schwarze 449: } else {
1.228 schwarze 450: /* Database missing or corrupt. */
1.230 schwarze 451: if (op != OP_UPDATE || errno != ENOENT)
452: say(MANDOC_DB, "%s: Automatically recreating"
453: " from scratch", strerror(errno));
1.140 schwarze 454: exitcode = (int)MANDOCLEVEL_OK;
1.138 schwarze 455: op = OP_DEFAULT;
456: if (0 == treescan())
457: goto out;
1.221 schwarze 458: dba = dba_new(128);
1.138 schwarze 459: }
1.50 kristaps 460: if (OP_DELETE != op)
1.221 schwarze 461: mpages_merge(dba, mp);
462: if (nodb == 0)
463: dbwrite(dba);
464: dba_free(dba);
1.50 kristaps 465: } else {
466: /*
467: * If we have arguments, use them as our manpaths.
1.222 schwarze 468: * If we don't, use man.conf(5).
1.50 kristaps 469: */
470: if (argc > 0) {
1.187 schwarze 471: conf.manpath.paths = mandoc_reallocarray(NULL,
1.144 schwarze 472: argc, sizeof(char *));
1.187 schwarze 473: conf.manpath.sz = (size_t)argc;
1.50 kristaps 474: for (i = 0; i < argc; i++)
1.187 schwarze 475: conf.manpath.paths[i] = mandoc_strdup(argv[i]);
1.50 kristaps 476: } else
1.187 schwarze 477: manconf_parse(&conf, path_arg, NULL, NULL);
1.127 schwarze 478:
1.187 schwarze 479: if (conf.manpath.sz == 0) {
1.127 schwarze 480: exitcode = (int)MANDOCLEVEL_BADARG;
481: say("", "Empty manpath");
482: }
1.50 kristaps 483:
484: /*
1.69 schwarze 485: * First scan the tree rooted at a base directory, then
486: * build a new database and finally move it into place.
1.50 kristaps 487: * Ignore zero-length directories and strip trailing
488: * slashes.
489: */
1.187 schwarze 490: for (j = 0; j < conf.manpath.sz; j++) {
491: sz = strlen(conf.manpath.paths[j]);
492: if (sz && conf.manpath.paths[j][sz - 1] == '/')
493: conf.manpath.paths[j][--sz] = '\0';
1.50 kristaps 494: if (0 == sz)
495: continue;
1.66 schwarze 496:
497: if (j) {
1.202 schwarze 498: mandoc_ohash_init(&mpages, 6,
499: offsetof(struct mpage, inodev));
500: mandoc_ohash_init(&mlinks, 6,
501: offsetof(struct mlink, file));
1.66 schwarze 502: }
503:
1.187 schwarze 504: if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
1.165 schwarze 505: continue;
1.59 schwarze 506: if (0 == treescan())
1.165 schwarze 507: continue;
1.221 schwarze 508: dba = dba_new(128);
509: mpages_merge(dba, mp);
510: if (nodb == 0)
511: dbwrite(dba);
512: dba_free(dba);
1.66 schwarze 513:
1.187 schwarze 514: if (j + 1 < conf.manpath.sz) {
1.80 schwarze 515: mpages_free();
1.78 schwarze 516: ohash_delete(&mpages);
1.83 schwarze 517: ohash_delete(&mlinks);
1.66 schwarze 518: }
1.50 kristaps 519: }
520: }
521: out:
1.187 schwarze 522: manconf_free(&conf);
1.168 schwarze 523: mparse_free(mp);
1.203 schwarze 524: mchars_free();
1.80 schwarze 525: mpages_free();
1.78 schwarze 526: ohash_delete(&mpages);
1.83 schwarze 527: ohash_delete(&mlinks);
1.197 schwarze 528: return exitcode;
1.50 kristaps 529: usage:
1.206 schwarze 530: progname = getprogname();
1.126 schwarze 531: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
532: " %s [-aDnpQ] [-Tutf8] dir ...\n"
533: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
534: " %s [-Dnp] -u dir [file ...]\n"
1.102 schwarze 535: " %s [-Q] -t file ...\n",
1.206 schwarze 536: progname, progname, progname, progname, progname);
1.50 kristaps 537:
1.197 schwarze 538: return (int)MANDOCLEVEL_BADARG;
1.50 kristaps 539: }
540:
541: /*
1.231 schwarze 542: * To get a singly linked list in alpha order while inserting entries
543: * at the beginning, process directory entries in reverse alpha order.
544: */
545: static int
1.232 schwarze 546: #if HAVE_FTS_COMPARE_CONST
547: fts_compare(const FTSENT *const *a, const FTSENT *const *b)
548: #else
1.231 schwarze 549: fts_compare(const FTSENT **a, const FTSENT **b)
1.232 schwarze 550: #endif
1.231 schwarze 551: {
552: return -strcmp((*a)->fts_name, (*b)->fts_name);
553: }
554:
555: /*
1.59 schwarze 556: * Scan a directory tree rooted at "basedir" for manpages.
1.50 kristaps 557: * We use fts(), scanning directory parts along the way for clues to our
558: * section and architecture.
559: *
560: * If use_all has been specified, grok all files.
561: * If not, sanitise paths to the following:
562: *
1.141 schwarze 563: * [./]man*[/<arch>]/<name>.<section>
1.50 kristaps 564: * or
565: * [./]cat<section>[/<arch>]/<name>.0
566: *
1.216 schwarze 567: * TODO: accommodate for multi-language directories.
1.50 kristaps 568: */
569: static int
1.59 schwarze 570: treescan(void)
1.50 kristaps 571: {
1.139 schwarze 572: char buf[PATH_MAX];
1.50 kristaps 573: FTS *f;
574: FTSENT *ff;
1.84 schwarze 575: struct mlink *mlink;
1.221 schwarze 576: int gzip;
577: enum form dform;
1.94 schwarze 578: char *dsec, *arch, *fsec, *cp;
579: const char *path;
1.50 kristaps 580: const char *argv[2];
581:
582: argv[0] = ".";
1.244 schwarze 583: argv[1] = NULL;
1.50 kristaps 584:
1.231 schwarze 585: f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR,
586: fts_compare);
1.200 schwarze 587: if (f == NULL) {
1.59 schwarze 588: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 589: say("", "&fts_open");
1.197 schwarze 590: return 0;
1.50 kristaps 591: }
592:
593: dsec = arch = NULL;
594: dform = FORM_NONE;
595:
1.200 schwarze 596: while ((ff = fts_read(f)) != NULL) {
1.50 kristaps 597: path = ff->fts_path + 2;
1.139 schwarze 598: switch (ff->fts_info) {
599:
600: /*
601: * Symbolic links require various sanity checks,
602: * then get handled just like regular files.
603: */
1.141 schwarze 604: case FTS_SL:
1.200 schwarze 605: if (realpath(path, buf) == NULL) {
1.139 schwarze 606: if (warnings)
607: say(path, "&realpath");
608: continue;
609: }
1.184 schwarze 610: if (strstr(buf, basedir) != buf
611: #ifdef HOMEBREWDIR
612: && strstr(buf, HOMEBREWDIR) != buf
613: #endif
614: ) {
1.139 schwarze 615: if (warnings) say("",
616: "%s: outside base directory", buf);
617: continue;
618: }
619: /* Use logical inode to avoid mpages dupe. */
1.200 schwarze 620: if (stat(path, ff->fts_statp) == -1) {
1.139 schwarze 621: if (warnings)
622: say(path, "&stat");
623: continue;
624: }
625: /* FALLTHROUGH */
626:
1.50 kristaps 627: /*
1.83 schwarze 628: * If we're a regular file, add an mlink by using the
1.50 kristaps 629: * stored directory data and handling the filename.
630: */
1.141 schwarze 631: case FTS_F:
1.200 schwarze 632: if ( ! strcmp(path, MANDOC_DB))
1.60 schwarze 633: continue;
1.50 kristaps 634: if ( ! use_all && ff->fts_level < 2) {
1.56 schwarze 635: if (warnings)
1.59 schwarze 636: say(path, "Extraneous file");
1.50 kristaps 637: continue;
1.124 schwarze 638: }
639: gzip = 0;
640: fsec = NULL;
1.200 schwarze 641: while (fsec == NULL) {
1.124 schwarze 642: fsec = strrchr(ff->fts_name, '.');
1.200 schwarze 643: if (fsec == NULL || strcmp(fsec+1, "gz"))
1.124 schwarze 644: break;
645: gzip = 1;
646: *fsec = '\0';
647: fsec = NULL;
648: }
1.200 schwarze 649: if (fsec == NULL) {
1.60 schwarze 650: if ( ! use_all) {
1.56 schwarze 651: if (warnings)
1.60 schwarze 652: say(path,
653: "No filename suffix");
1.50 kristaps 654: continue;
655: }
1.200 schwarze 656: } else if ( ! strcmp(++fsec, "html")) {
1.60 schwarze 657: if (warnings)
658: say(path, "Skip html");
659: continue;
1.200 schwarze 660: } else if ( ! strcmp(fsec, "ps")) {
1.60 schwarze 661: if (warnings)
662: say(path, "Skip ps");
663: continue;
1.200 schwarze 664: } else if ( ! strcmp(fsec, "pdf")) {
1.60 schwarze 665: if (warnings)
666: say(path, "Skip pdf");
667: continue;
668: } else if ( ! use_all &&
1.200 schwarze 669: ((dform == FORM_SRC &&
1.185 schwarze 670: strncmp(fsec, dsec, strlen(dsec))) ||
1.200 schwarze 671: (dform == FORM_CAT && strcmp(fsec, "0")))) {
1.60 schwarze 672: if (warnings)
673: say(path, "Wrong filename suffix");
674: continue;
1.66 schwarze 675: } else
1.84 schwarze 676: fsec[-1] = '\0';
1.94 schwarze 677:
1.84 schwarze 678: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.143 schwarze 679: if (strlcpy(mlink->file, path,
680: sizeof(mlink->file)) >=
681: sizeof(mlink->file)) {
682: say(path, "Filename too long");
683: free(mlink);
684: continue;
685: }
1.84 schwarze 686: mlink->dform = dform;
1.94 schwarze 687: mlink->dsec = dsec;
688: mlink->arch = arch;
689: mlink->name = ff->fts_name;
690: mlink->fsec = fsec;
1.124 schwarze 691: mlink->gzip = gzip;
1.84 schwarze 692: mlink_add(mlink, ff->fts_statp);
1.50 kristaps 693: continue;
1.139 schwarze 694:
1.141 schwarze 695: case FTS_D:
696: case FTS_DP:
1.139 schwarze 697: break;
698:
699: default:
1.60 schwarze 700: if (warnings)
701: say(path, "Not a regular file");
1.50 kristaps 702: continue;
1.60 schwarze 703: }
1.1 kristaps 704:
1.50 kristaps 705: switch (ff->fts_level) {
1.141 schwarze 706: case 0:
1.50 kristaps 707: /* Ignore the root directory. */
708: break;
1.141 schwarze 709: case 1:
1.50 kristaps 710: /*
711: * This might contain manX/ or catX/.
712: * Try to infer this from the name.
713: * If we're not in use_all, enforce it.
714: */
715: cp = ff->fts_name;
1.200 schwarze 716: if (ff->fts_info == FTS_DP) {
717: dform = FORM_NONE;
718: dsec = NULL;
1.50 kristaps 719: break;
1.200 schwarze 720: }
1.1 kristaps 721:
1.200 schwarze 722: if ( ! strncmp(cp, "man", 3)) {
1.50 kristaps 723: dform = FORM_SRC;
1.66 schwarze 724: dsec = cp + 3;
1.200 schwarze 725: } else if ( ! strncmp(cp, "cat", 3)) {
1.50 kristaps 726: dform = FORM_CAT;
1.66 schwarze 727: dsec = cp + 3;
1.94 schwarze 728: } else {
729: dform = FORM_NONE;
730: dsec = NULL;
1.50 kristaps 731: }
1.1 kristaps 732:
1.200 schwarze 733: if (dsec != NULL || use_all)
1.50 kristaps 734: break;
1.1 kristaps 735:
1.56 schwarze 736: if (warnings)
1.59 schwarze 737: say(path, "Unknown directory part");
1.50 kristaps 738: fts_set(f, ff, FTS_SKIP);
739: break;
1.141 schwarze 740: case 2:
1.50 kristaps 741: /*
742: * Possibly our architecture.
743: * If we're descending, keep tabs on it.
744: */
1.200 schwarze 745: if (ff->fts_info != FTS_DP && dsec != NULL)
1.66 schwarze 746: arch = ff->fts_name;
1.94 schwarze 747: else
748: arch = NULL;
1.50 kristaps 749: break;
750: default:
1.200 schwarze 751: if (ff->fts_info == FTS_DP || use_all)
1.50 kristaps 752: break;
1.56 schwarze 753: if (warnings)
1.59 schwarze 754: say(path, "Extraneous directory part");
1.50 kristaps 755: fts_set(f, ff, FTS_SKIP);
756: break;
1.5 kristaps 757: }
1.50 kristaps 758: }
759:
760: fts_close(f);
1.197 schwarze 761: return 1;
1.50 kristaps 762: }
1.5 kristaps 763:
1.50 kristaps 764: /*
1.89 schwarze 765: * Add a file to the mlinks table.
1.50 kristaps 766: * Do not verify that it's a "valid" looking manpage (we'll do that
767: * later).
768: *
769: * Try to infer the manual section, architecture, and page name from the
770: * path, assuming it looks like
771: *
1.141 schwarze 772: * [./]man*[/<arch>]/<name>.<section>
1.50 kristaps 773: * or
774: * [./]cat<section>[/<arch>]/<name>.0
775: *
776: * See treescan() for the fts(3) version of this.
777: */
778: static void
1.59 schwarze 779: filescan(const char *file)
1.50 kristaps 780: {
1.59 schwarze 781: char buf[PATH_MAX];
1.84 schwarze 782: struct stat st;
783: struct mlink *mlink;
1.59 schwarze 784: char *p, *start;
1.5 kristaps 785:
1.50 kristaps 786: assert(use_all);
1.5 kristaps 787:
1.50 kristaps 788: if (0 == strncmp(file, "./", 2))
789: file += 2;
1.5 kristaps 790:
1.139 schwarze 791: /*
792: * We have to do lstat(2) before realpath(3) loses
793: * the information whether this is a symbolic link.
794: * We need to know that because for symbolic links,
795: * we want to use the orginal file name, while for
796: * regular files, we want to use the real path.
797: */
798: if (-1 == lstat(file, &st)) {
799: exitcode = (int)MANDOCLEVEL_BADARG;
800: say(file, "&lstat");
801: return;
802: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
803: exitcode = (int)MANDOCLEVEL_BADARG;
804: say(file, "Not a regular file");
805: return;
806: }
807:
808: /*
809: * We have to resolve the file name to the real path
810: * in any case for the base directory check.
811: */
1.59 schwarze 812: if (NULL == realpath(file, buf)) {
813: exitcode = (int)MANDOCLEVEL_BADARG;
1.123 schwarze 814: say(file, "&realpath");
1.59 schwarze 815: return;
1.106 schwarze 816: }
817:
1.150 schwarze 818: if (OP_TEST == op)
1.106 schwarze 819: start = buf;
1.150 schwarze 820: else if (strstr(buf, basedir) == buf)
821: start = buf + strlen(basedir);
1.184 schwarze 822: #ifdef HOMEBREWDIR
823: else if (strstr(buf, HOMEBREWDIR) == buf)
824: start = buf;
825: #endif
1.106 schwarze 826: else {
1.59 schwarze 827: exitcode = (int)MANDOCLEVEL_BADARG;
828: say("", "%s: outside base directory", buf);
829: return;
1.106 schwarze 830: }
831:
1.139 schwarze 832: /*
833: * Now we are sure the file is inside our tree.
834: * If it is a symbolic link, ignore the real path
835: * and use the original name.
836: * This implies passing stuff like "cat1/../man1/foo.1"
837: * on the command line won't work. So don't do that.
838: * Note the stat(2) can still fail if the link target
839: * doesn't exist.
840: */
841: if (S_IFLNK & st.st_mode) {
842: if (-1 == stat(buf, &st)) {
843: exitcode = (int)MANDOCLEVEL_BADARG;
844: say(file, "&stat");
845: return;
846: }
1.143 schwarze 847: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
848: say(file, "Filename too long");
849: return;
850: }
1.150 schwarze 851: start = buf;
852: if (OP_TEST != op && strstr(buf, basedir) == buf)
853: start += strlen(basedir);
1.50 kristaps 854: }
1.106 schwarze 855:
1.84 schwarze 856: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.161 schwarze 857: mlink->dform = FORM_NONE;
1.143 schwarze 858: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
859: sizeof(mlink->file)) {
860: say(start, "Filename too long");
1.180 schwarze 861: free(mlink);
1.143 schwarze 862: return;
1.239 schwarze 863: }
864:
865: /*
866: * In test mode or when the original name is absolute
867: * but outside our tree, guess the base directory.
868: */
869:
870: if (op == OP_TEST || (start == buf && *start == '/')) {
871: if (strncmp(buf, "man/", 4) == 0)
872: start = buf + 4;
873: else if ((start = strstr(buf, "/man/")) != NULL)
874: start += 5;
875: else
876: start = buf;
1.143 schwarze 877: }
1.17 schwarze 878:
1.50 kristaps 879: /*
880: * First try to guess our directory structure.
881: * If we find a separator, try to look for man* or cat*.
882: * If we find one of these and what's underneath is a directory,
883: * assume it's an architecture.
884: */
885: if (NULL != (p = strchr(start, '/'))) {
886: *p++ = '\0';
887: if (0 == strncmp(start, "man", 3)) {
1.84 schwarze 888: mlink->dform = FORM_SRC;
1.94 schwarze 889: mlink->dsec = start + 3;
1.50 kristaps 890: } else if (0 == strncmp(start, "cat", 3)) {
1.84 schwarze 891: mlink->dform = FORM_CAT;
1.94 schwarze 892: mlink->dsec = start + 3;
1.17 schwarze 893: }
1.5 kristaps 894:
1.50 kristaps 895: start = p;
1.84 schwarze 896: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
1.50 kristaps 897: *p++ = '\0';
1.94 schwarze 898: mlink->arch = start;
1.50 kristaps 899: start = p;
1.84 schwarze 900: }
1.50 kristaps 901: }
902:
903: /*
904: * Now check the file suffix.
905: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
906: */
907: p = strrchr(start, '\0');
908: while (p-- > start && '/' != *p && '.' != *p)
909: /* Loop. */ ;
910:
911: if ('.' == *p) {
912: *p++ = '\0';
1.94 schwarze 913: mlink->fsec = p;
1.5 kristaps 914: }
915:
1.10 kristaps 916: /*
1.50 kristaps 917: * Now try to parse the name.
918: * Use the filename portion of the path.
1.10 kristaps 919: */
1.84 schwarze 920: mlink->name = start;
1.50 kristaps 921: if (NULL != (p = strrchr(start, '/'))) {
1.84 schwarze 922: mlink->name = p + 1;
1.50 kristaps 923: *p = '\0';
1.84 schwarze 924: }
925: mlink_add(mlink, &st);
1.50 kristaps 926: }
1.5 kristaps 927:
1.50 kristaps 928: static void
1.84 schwarze 929: mlink_add(struct mlink *mlink, const struct stat *st)
1.50 kristaps 930: {
1.83 schwarze 931: struct inodev inodev;
1.78 schwarze 932: struct mpage *mpage;
1.83 schwarze 933: unsigned int slot;
1.50 kristaps 934:
1.84 schwarze 935: assert(NULL != mlink->file);
1.50 kristaps 936:
1.94 schwarze 937: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
938: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
939: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
940: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.84 schwarze 941:
942: if ('0' == *mlink->fsec) {
943: free(mlink->fsec);
944: mlink->fsec = mandoc_strdup(mlink->dsec);
945: mlink->fform = FORM_CAT;
946: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
947: mlink->fform = FORM_SRC;
1.74 schwarze 948: else
1.84 schwarze 949: mlink->fform = FORM_NONE;
1.82 schwarze 950:
1.83 schwarze 951: slot = ohash_qlookup(&mlinks, mlink->file);
952: assert(NULL == ohash_find(&mlinks, slot));
953: ohash_insert(&mlinks, slot, mlink);
954:
1.186 schwarze 955: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.83 schwarze 956: inodev.st_ino = st->st_ino;
957: inodev.st_dev = st->st_dev;
958: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
959: sizeof(struct inodev), inodev.st_ino);
960: mpage = ohash_find(&mpages, slot);
961: if (NULL == mpage) {
962: mpage = mandoc_calloc(1, sizeof(struct mpage));
963: mpage->inodev.st_ino = inodev.st_ino;
964: mpage->inodev.st_dev = inodev.st_dev;
1.221 schwarze 965: mpage->form = FORM_NONE;
1.231 schwarze 966: mpage->next = mpage_head;
967: mpage_head = mpage;
1.83 schwarze 968: ohash_insert(&mpages, slot, mpage);
969: } else
1.85 schwarze 970: mlink->next = mpage->mlinks;
1.82 schwarze 971: mpage->mlinks = mlink;
1.118 schwarze 972: mlink->mpage = mpage;
1.82 schwarze 973: }
974:
975: static void
976: mlink_free(struct mlink *mlink)
977: {
978:
979: free(mlink->dsec);
980: free(mlink->arch);
981: free(mlink->name);
982: free(mlink->fsec);
983: free(mlink);
1.50 kristaps 984: }
1.3 kristaps 985:
1.50 kristaps 986: static void
1.78 schwarze 987: mpages_free(void)
1.50 kristaps 988: {
1.78 schwarze 989: struct mpage *mpage;
1.82 schwarze 990: struct mlink *mlink;
1.3 kristaps 991:
1.231 schwarze 992: while ((mpage = mpage_head) != NULL) {
993: while ((mlink = mpage->mlinks) != NULL) {
1.85 schwarze 994: mpage->mlinks = mlink->next;
1.82 schwarze 995: mlink_free(mlink);
996: }
1.231 schwarze 997: mpage_head = mpage->next;
1.78 schwarze 998: free(mpage->sec);
999: free(mpage->arch);
1.82 schwarze 1000: free(mpage->title);
1001: free(mpage->desc);
1.78 schwarze 1002: free(mpage);
1.50 kristaps 1003: }
1004: }
1.38 schwarze 1005:
1.50 kristaps 1006: /*
1.89 schwarze 1007: * For each mlink to the mpage, check whether the path looks like
1008: * it is formatted, and if it does, check whether a source manual
1009: * exists by the same name, ignoring the suffix.
1010: * If both conditions hold, drop the mlink.
1011: */
1012: static void
1013: mlinks_undupe(struct mpage *mpage)
1014: {
1015: char buf[PATH_MAX];
1016: struct mlink **prev;
1017: struct mlink *mlink;
1018: char *bufp;
1019:
1020: mpage->form = FORM_CAT;
1.90 schwarze 1021: prev = &mpage->mlinks;
1022: while (NULL != (mlink = *prev)) {
1.89 schwarze 1023: if (FORM_CAT != mlink->dform) {
1024: mpage->form = FORM_NONE;
1.90 schwarze 1025: goto nextlink;
1.89 schwarze 1026: }
1.143 schwarze 1027: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.89 schwarze 1028: bufp = strstr(buf, "cat");
1029: assert(NULL != bufp);
1030: memcpy(bufp, "man", 3);
1031: if (NULL != (bufp = strrchr(buf, '.')))
1032: *++bufp = '\0';
1.143 schwarze 1033: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.89 schwarze 1034: if (NULL == ohash_find(&mlinks,
1.141 schwarze 1035: ohash_qlookup(&mlinks, buf)))
1.90 schwarze 1036: goto nextlink;
1.89 schwarze 1037: if (warnings)
1038: say(mlink->file, "Man source exists: %s", buf);
1039: if (use_all)
1.90 schwarze 1040: goto nextlink;
1.89 schwarze 1041: *prev = mlink->next;
1042: mlink_free(mlink);
1.90 schwarze 1043: continue;
1044: nextlink:
1045: prev = &(*prev)->next;
1.89 schwarze 1046: }
1047: }
1048:
1.131 schwarze 1049: static void
1.93 schwarze 1050: mlink_check(struct mpage *mpage, struct mlink *mlink)
1051: {
1.131 schwarze 1052: struct str *str;
1053: unsigned int slot;
1.93 schwarze 1054:
1055: /*
1056: * Check whether the manual section given in a file
1057: * agrees with the directory where the file is located.
1058: * Some manuals have suffixes like (3p) on their
1059: * section number either inside the file or in the
1060: * directory name, some are linked into more than one
1061: * section, like encrypt(1) = makekey(8).
1062: */
1063:
1064: if (FORM_SRC == mpage->form &&
1.131 schwarze 1065: strcasecmp(mpage->sec, mlink->dsec))
1.93 schwarze 1066: say(mlink->file, "Section \"%s\" manual in %s directory",
1067: mpage->sec, mlink->dsec);
1068:
1069: /*
1070: * Manual page directories exist for each kernel
1071: * architecture as returned by machine(1).
1072: * However, many manuals only depend on the
1073: * application architecture as returned by arch(1).
1074: * For example, some (2/ARM) manuals are shared
1075: * across the "armish" and "zaurus" kernel
1076: * architectures.
1077: * A few manuals are even shared across completely
1078: * different architectures, for example fdformat(1)
1.229 schwarze 1079: * on amd64, i386, and sparc64.
1.93 schwarze 1080: */
1081:
1.131 schwarze 1082: if (strcasecmp(mpage->arch, mlink->arch))
1.93 schwarze 1083: say(mlink->file, "Architecture \"%s\" manual in "
1084: "\"%s\" directory", mpage->arch, mlink->arch);
1085:
1.131 schwarze 1086: /*
1087: * XXX
1.133 schwarze 1088: * parse_cat() doesn't set NAME_TITLE yet.
1.131 schwarze 1089: */
1090:
1091: if (FORM_CAT == mpage->form)
1092: return;
1093:
1094: /*
1095: * Check whether this mlink
1096: * appears as a name in the NAME section.
1097: */
1.93 schwarze 1098:
1.133 schwarze 1099: slot = ohash_qlookup(&names, mlink->name);
1100: str = ohash_find(&names, slot);
1.131 schwarze 1101: assert(NULL != str);
1.133 schwarze 1102: if ( ! (NAME_TITLE & str->mask))
1.131 schwarze 1103: say(mlink->file, "Name missing in NAME section");
1.93 schwarze 1104: }
1105:
1.89 schwarze 1106: /*
1.80 schwarze 1107: * Run through the files in the global vector "mpages"
1108: * and add them to the database specified in "basedir".
1.50 kristaps 1109: *
1110: * This handles the parsing scheme itself, using the cues of directory
1111: * and filename to determine whether the file is parsable or not.
1112: */
1.59 schwarze 1113: static void
1.221 schwarze 1114: mpages_merge(struct dba *dba, struct mparse *mp)
1.50 kristaps 1115: {
1.231 schwarze 1116: struct mpage *mpage, *mpage_dest;
1.118 schwarze 1117: struct mlink *mlink, *mlink_dest;
1.191 schwarze 1118: struct roff_man *man;
1.118 schwarze 1119: char *sodest;
1.112 schwarze 1120: char *cp;
1.162 schwarze 1121: int fd;
1.69 schwarze 1122:
1.231 schwarze 1123: for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
1.89 schwarze 1124: mlinks_undupe(mpage);
1.231 schwarze 1125: if ((mlink = mpage->mlinks) == NULL)
1126: continue;
1.50 kristaps 1127:
1.133 schwarze 1128: name_mask = NAME_MASK;
1.202 schwarze 1129: mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1130: mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1.50 kristaps 1131: mparse_reset(mp);
1132: man = NULL;
1.124 schwarze 1133: sodest = NULL;
1134:
1.213 schwarze 1135: if ((fd = mparse_open(mp, mlink->file)) == -1) {
1.182 schwarze 1136: say(mlink->file, "&open");
1.162 schwarze 1137: goto nextpage;
1.124 schwarze 1138: }
1.14 schwarze 1139:
1140: /*
1.183 schwarze 1141: * Interpret the file as mdoc(7) or man(7) source
1142: * code, unless it is known to be formatted.
1.14 schwarze 1143: */
1.182 schwarze 1144: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.183 schwarze 1145: mparse_readfd(mp, fd, mlink->file);
1.212 schwarze 1146: close(fd);
1.240 schwarze 1147: fd = -1;
1.192 schwarze 1148: mparse_result(mp, &man, &sodest);
1.76 schwarze 1149: }
1.14 schwarze 1150:
1.171 schwarze 1151: if (sodest != NULL) {
1.118 schwarze 1152: mlink_dest = ohash_find(&mlinks,
1153: ohash_qlookup(&mlinks, sodest));
1.171 schwarze 1154: if (mlink_dest == NULL) {
1155: mandoc_asprintf(&cp, "%s.gz", sodest);
1156: mlink_dest = ohash_find(&mlinks,
1157: ohash_qlookup(&mlinks, cp));
1158: free(cp);
1159: }
1160: if (mlink_dest != NULL) {
1.118 schwarze 1161:
1162: /* The .so target exists. */
1163:
1164: mpage_dest = mlink_dest->mpage;
1165: while (1) {
1166: mlink->mpage = mpage_dest;
1167:
1168: /*
1169: * If the target was already
1170: * processed, add the links
1171: * to the database now.
1172: * Otherwise, this will
1173: * happen when we come
1174: * to the target.
1175: */
1176:
1.221 schwarze 1177: if (mpage_dest->dba != NULL)
1178: dbadd_mlink(mlink);
1.118 schwarze 1179:
1.171 schwarze 1180: if (mlink->next == NULL)
1.118 schwarze 1181: break;
1182: mlink = mlink->next;
1183: }
1184:
1185: /* Move all links to the target. */
1186:
1187: mlink->next = mlink_dest->next;
1188: mlink_dest->next = mpage->mlinks;
1189: mpage->mlinks = NULL;
1190: }
1.124 schwarze 1191: goto nextpage;
1.192 schwarze 1192: } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1.204 schwarze 1193: mdoc_validate(man);
1.82 schwarze 1194: mpage->form = FORM_SRC;
1.193 schwarze 1195: mpage->sec = man->meta.msec;
1.155 schwarze 1196: mpage->sec = mandoc_strdup(
1.171 schwarze 1197: mpage->sec == NULL ? "" : mpage->sec);
1.193 schwarze 1198: mpage->arch = man->meta.arch;
1.84 schwarze 1199: mpage->arch = mandoc_strdup(
1.171 schwarze 1200: mpage->arch == NULL ? "" : mpage->arch);
1.193 schwarze 1201: mpage->title = mandoc_strdup(man->meta.title);
1.192 schwarze 1202: } else if (man != NULL && man->macroset == MACROSET_MAN) {
1.205 schwarze 1203: man_validate(man);
1.240 schwarze 1204: if (*man->meta.msec != '\0' ||
1.245 schwarze 1205: *man->meta.title != '\0') {
1.240 schwarze 1206: mpage->form = FORM_SRC;
1207: mpage->sec = mandoc_strdup(man->meta.msec);
1208: mpage->arch = mandoc_strdup(mlink->arch);
1209: mpage->title = mandoc_strdup(man->meta.title);
1210: } else
1211: man = NULL;
1212: }
1213:
1214: assert(mpage->desc == NULL);
1215: if (man == NULL) {
1.82 schwarze 1216: mpage->form = FORM_CAT;
1.182 schwarze 1217: mpage->sec = mandoc_strdup(mlink->dsec);
1218: mpage->arch = mandoc_strdup(mlink->arch);
1219: mpage->title = mandoc_strdup(mlink->name);
1.240 schwarze 1220: parse_cat(mpage, fd);
1221: } else if (man->macroset == MACROSET_MDOC)
1.193 schwarze 1222: parse_mdoc(mpage, &man->meta, man->first);
1.240 schwarze 1223: else
1.193 schwarze 1224: parse_man(mpage, &man->meta, man->first);
1.242 schwarze 1225: if (mpage->desc == NULL) {
1226: mpage->desc = mandoc_strdup(mlink->name);
1227: if (warnings)
1228: say(mlink->file, "No one-line description, "
1229: "using filename \"%s\"", mlink->name);
1230: }
1.131 schwarze 1231:
1.238 schwarze 1232: for (mlink = mpage->mlinks;
1233: mlink != NULL;
1234: mlink = mlink->next) {
1235: putkey(mpage, mlink->name, NAME_FILE);
1236: if (warnings && !use_all)
1.131 schwarze 1237: mlink_check(mpage, mlink);
1.238 schwarze 1238: }
1.44 kristaps 1239:
1.221 schwarze 1240: dbadd(dba, mpage);
1.124 schwarze 1241:
1242: nextpage:
1.66 schwarze 1243: ohash_delete(&strings);
1.133 schwarze 1244: ohash_delete(&names);
1.50 kristaps 1245: }
1246: }
1.12 schwarze 1247:
1.50 kristaps 1248: static void
1.124 schwarze 1249: parse_cat(struct mpage *mpage, int fd)
1.50 kristaps 1250: {
1251: FILE *stream;
1.241 schwarze 1252: struct mlink *mlink;
1253: char *line, *p, *title, *sec;
1.209 schwarze 1254: size_t linesz, plen, titlesz;
1255: ssize_t len;
1256: int offs;
1.12 schwarze 1257:
1.241 schwarze 1258: mlink = mpage->mlinks;
1259: stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1260: if (stream == NULL) {
1261: if (fd != -1)
1.159 schwarze 1262: close(fd);
1.56 schwarze 1263: if (warnings)
1.241 schwarze 1264: say(mlink->file, "&fopen");
1.50 kristaps 1265: return;
1266: }
1.1 kristaps 1267:
1.209 schwarze 1268: line = NULL;
1269: linesz = 0;
1270:
1.241 schwarze 1271: /* Parse the section number from the header line. */
1272:
1273: while (getline(&line, &linesz, stream) != -1) {
1274: if (*line == '\n')
1275: continue;
1276: if ((sec = strchr(line, '(')) == NULL)
1277: break;
1278: if ((p = strchr(++sec, ')')) == NULL)
1279: break;
1280: free(mpage->sec);
1281: mpage->sec = mandoc_strndup(sec, p - sec);
1282: if (warnings && *mlink->dsec != '\0' &&
1283: strcasecmp(mpage->sec, mlink->dsec))
1284: say(mlink->file,
1285: "Section \"%s\" manual in %s directory",
1286: mpage->sec, mlink->dsec);
1287: break;
1288: }
1289:
1.50 kristaps 1290: /* Skip to first blank line. */
1.1 kristaps 1291:
1.241 schwarze 1292: while (line == NULL || *line != '\n')
1293: if (getline(&line, &linesz, stream) == -1)
1.50 kristaps 1294: break;
1.1 kristaps 1295:
1.50 kristaps 1296: /*
1297: * Assume the first line that is not indented
1298: * is the first section header. Skip to it.
1299: */
1.1 kristaps 1300:
1.209 schwarze 1301: while (getline(&line, &linesz, stream) != -1)
1302: if (*line != '\n' && *line != ' ')
1.50 kristaps 1303: break;
1.141 schwarze 1304:
1.50 kristaps 1305: /*
1306: * Read up until the next section into a buffer.
1307: * Strip the leading and trailing newline from each read line,
1308: * appending a trailing space.
1309: * Ignore empty (whitespace-only) lines.
1310: */
1.1 kristaps 1311:
1.50 kristaps 1312: titlesz = 0;
1313: title = NULL;
1.38 schwarze 1314:
1.209 schwarze 1315: while ((len = getline(&line, &linesz, stream)) != -1) {
1316: if (*line != ' ')
1.50 kristaps 1317: break;
1.209 schwarze 1318: offs = 0;
1319: while (isspace((unsigned char)line[offs]))
1320: offs++;
1321: if (line[offs] == '\0')
1.38 schwarze 1322: continue;
1.209 schwarze 1323: title = mandoc_realloc(title, titlesz + len - offs);
1324: memcpy(title + titlesz, line + offs, len - offs);
1325: titlesz += len - offs;
1.50 kristaps 1326: title[titlesz - 1] = ' ';
1327: }
1.209 schwarze 1328: free(line);
1.38 schwarze 1329:
1.50 kristaps 1330: /*
1331: * If no page content can be found, or the input line
1332: * is already the next section header, or there is no
1333: * trailing newline, reuse the page title as the page
1334: * description.
1335: */
1.44 kristaps 1336:
1.50 kristaps 1337: if (NULL == title || '\0' == *title) {
1.56 schwarze 1338: if (warnings)
1.241 schwarze 1339: say(mlink->file, "Cannot find NAME section");
1.50 kristaps 1340: fclose(stream);
1341: free(title);
1342: return;
1343: }
1.1 kristaps 1344:
1.209 schwarze 1345: title[titlesz - 1] = '\0';
1.33 schwarze 1346:
1.50 kristaps 1347: /*
1348: * Skip to the first dash.
1349: * Use the remaining line as the description (no more than 70
1350: * bytes).
1351: */
1.33 schwarze 1352:
1.50 kristaps 1353: if (NULL != (p = strstr(title, "- "))) {
1354: for (p += 2; ' ' == *p || '\b' == *p; p++)
1355: /* Skip to next word. */ ;
1356: } else {
1.56 schwarze 1357: if (warnings)
1.242 schwarze 1358: say(mlink->file, "No dash in title line, "
1359: "reusing \"%s\" as one-line description", title);
1.50 kristaps 1360: p = title;
1361: }
1.38 schwarze 1362:
1.50 kristaps 1363: plen = strlen(p);
1.1 kristaps 1364:
1.50 kristaps 1365: /* Strip backspace-encoding from line. */
1.1 kristaps 1366:
1.50 kristaps 1367: while (NULL != (line = memchr(p, '\b', plen))) {
1368: len = line - p;
1369: if (0 == len) {
1370: memmove(line, line + 1, plen--);
1371: continue;
1.141 schwarze 1372: }
1.50 kristaps 1373: memmove(line - 1, line + 1, plen - len);
1374: plen -= 2;
1375: }
1.1 kristaps 1376:
1.257 schwarze 1377: /*
1378: * Cut off excessive one-line descriptions.
1379: * Bad pages are not worth better heuristics.
1380: */
1381:
1382: mpage->desc = mandoc_strndup(p, 150);
1.50 kristaps 1383: fclose(stream);
1384: free(title);
1385: }
1.1 kristaps 1386:
1.50 kristaps 1387: /*
1388: * Put a type/word pair into the word database for this particular file.
1389: */
1390: static void
1.112 schwarze 1391: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.50 kristaps 1392: {
1.78 schwarze 1393: putkeys(mpage, value, strlen(value), type);
1.3 kristaps 1394: }
1395:
1396: /*
1.50 kristaps 1397: * Grok all nodes at or below a certain mdoc node into putkey().
1.3 kristaps 1398: */
1399: static void
1.78 schwarze 1400: putmdockey(const struct mpage *mpage,
1.235 schwarze 1401: const struct roff_node *n, uint64_t m, int taboo)
1.3 kristaps 1402: {
1.18 kristaps 1403:
1.50 kristaps 1404: for ( ; NULL != n; n = n->next) {
1.235 schwarze 1405: if (n->flags & taboo)
1406: continue;
1.50 kristaps 1407: if (NULL != n->child)
1.235 schwarze 1408: putmdockey(mpage, n->child, m, taboo);
1.188 schwarze 1409: if (n->type == ROFFT_TEXT)
1.78 schwarze 1410: putkey(mpage, n->string, m);
1.50 kristaps 1411: }
1412: }
1.18 kristaps 1413:
1.61 schwarze 1414: static void
1.190 schwarze 1415: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1416: const struct roff_node *n)
1.50 kristaps 1417: {
1.189 schwarze 1418: const struct roff_node *head, *body;
1.121 schwarze 1419: char *start, *title;
1.50 kristaps 1420: char byte;
1.121 schwarze 1421: size_t sz;
1.18 kristaps 1422:
1.215 schwarze 1423: if (n == NULL)
1.61 schwarze 1424: return;
1.18 kristaps 1425:
1.50 kristaps 1426: /*
1427: * We're only searching for one thing: the first text child in
1428: * the BODY of a NAME section. Since we don't keep track of
1429: * sections in -man, run some hoops to find out whether we're in
1430: * the correct section or not.
1431: */
1.18 kristaps 1432:
1.188 schwarze 1433: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.50 kristaps 1434: body = n;
1.215 schwarze 1435: if ((head = body->parent->head) != NULL &&
1436: (head = head->child) != NULL &&
1437: head->next == NULL &&
1.188 schwarze 1438: head->type == ROFFT_TEXT &&
1.215 schwarze 1439: strcmp(head->string, "NAME") == 0 &&
1440: body->child != NULL) {
1.3 kristaps 1441:
1.50 kristaps 1442: /*
1443: * Suck the entire NAME section into memory.
1444: * Yes, we might run away.
1445: * But too many manuals have big, spread-out
1446: * NAME sections over many lines.
1447: */
1.3 kristaps 1448:
1.121 schwarze 1449: title = NULL;
1.194 schwarze 1450: deroff(&title, body);
1.50 kristaps 1451: if (NULL == title)
1.61 schwarze 1452: return;
1.18 kristaps 1453:
1.141 schwarze 1454: /*
1.50 kristaps 1455: * Go through a special heuristic dance here.
1456: * Conventionally, one or more manual names are
1457: * comma-specified prior to a whitespace, then a
1458: * dash, then a description. Try to puzzle out
1459: * the name parts here.
1460: */
1.18 kristaps 1461:
1.121 schwarze 1462: start = title;
1.50 kristaps 1463: for ( ;; ) {
1464: sz = strcspn(start, " ,");
1465: if ('\0' == start[sz])
1466: break;
1.1 kristaps 1467:
1.50 kristaps 1468: byte = start[sz];
1469: start[sz] = '\0';
1.110 schwarze 1470:
1471: /*
1472: * Assume a stray trailing comma in the
1473: * name list if a name begins with a dash.
1474: */
1475:
1476: if ('-' == start[0] ||
1477: ('\\' == start[0] && '-' == start[1]))
1478: break;
1.1 kristaps 1479:
1.133 schwarze 1480: putkey(mpage, start, NAME_TITLE);
1.174 schwarze 1481: if ( ! (mpage->name_head_done ||
1482: strcasecmp(start, meta->title))) {
1483: putkey(mpage, start, NAME_HEAD);
1484: mpage->name_head_done = 1;
1485: }
1.1 kristaps 1486:
1.50 kristaps 1487: if (' ' == byte) {
1488: start += sz + 1;
1489: break;
1490: }
1.1 kristaps 1491:
1.50 kristaps 1492: assert(',' == byte);
1493: start += sz + 1;
1494: while (' ' == *start)
1495: start++;
1496: }
1.1 kristaps 1497:
1.121 schwarze 1498: if (start == title) {
1.133 schwarze 1499: putkey(mpage, start, NAME_TITLE);
1.174 schwarze 1500: if ( ! (mpage->name_head_done ||
1501: strcasecmp(start, meta->title))) {
1502: putkey(mpage, start, NAME_HEAD);
1503: mpage->name_head_done = 1;
1504: }
1.50 kristaps 1505: free(title);
1.61 schwarze 1506: return;
1.50 kristaps 1507: }
1.1 kristaps 1508:
1.50 kristaps 1509: while (isspace((unsigned char)*start))
1510: start++;
1.1 kristaps 1511:
1.50 kristaps 1512: if (0 == strncmp(start, "-", 1))
1513: start += 1;
1514: else if (0 == strncmp(start, "\\-\\-", 4))
1515: start += 4;
1516: else if (0 == strncmp(start, "\\-", 2))
1517: start += 2;
1518: else if (0 == strncmp(start, "\\(en", 4))
1519: start += 4;
1520: else if (0 == strncmp(start, "\\(em", 4))
1521: start += 4;
1.1 kristaps 1522:
1.50 kristaps 1523: while (' ' == *start)
1524: start++;
1.1 kristaps 1525:
1.257 schwarze 1526: /*
1527: * Cut off excessive one-line descriptions.
1528: * Bad pages are not worth better heuristics.
1529: */
1530:
1531: mpage->desc = mandoc_strndup(start, 150);
1.50 kristaps 1532: free(title);
1.61 schwarze 1533: return;
1.50 kristaps 1534: }
1535: }
1.1 kristaps 1536:
1.77 schwarze 1537: for (n = n->child; n; n = n->next) {
1.78 schwarze 1538: if (NULL != mpage->desc)
1.77 schwarze 1539: break;
1.174 schwarze 1540: parse_man(mpage, meta, n);
1.77 schwarze 1541: }
1.1 kristaps 1542: }
1543:
1544: static void
1.190 schwarze 1545: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1546: const struct roff_node *n)
1.1 kristaps 1547: {
1.259 ! schwarze 1548: const struct mdoc_handler *handler;
1.1 kristaps 1549:
1.246 schwarze 1550: for (n = n->child; n != NULL; n = n->next) {
1.259 ! schwarze 1551: if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX)
1.235 schwarze 1552: continue;
1.247 schwarze 1553: assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1.259 ! schwarze 1554: handler = mdoc_handlers + (n->tok - MDOC_Dd);
! 1555: if (n->flags & handler->taboo)
! 1556: continue;
! 1557:
1.50 kristaps 1558: switch (n->type) {
1.188 schwarze 1559: case ROFFT_ELEM:
1560: case ROFFT_BLOCK:
1561: case ROFFT_HEAD:
1562: case ROFFT_BODY:
1563: case ROFFT_TAIL:
1.259 ! schwarze 1564: if (handler->fp != NULL &&
! 1565: (*handler->fp)(mpage, meta, n) == 0)
1.246 schwarze 1566: break;
1.259 ! schwarze 1567: if (handler->mask)
1.78 schwarze 1568: putmdockey(mpage, n->child,
1.259 ! schwarze 1569: handler->mask, handler->taboo);
1.50 kristaps 1570: break;
1571: default:
1572: continue;
1573: }
1574: if (NULL != n->child)
1.172 schwarze 1575: parse_mdoc(mpage, meta, n);
1.1 kristaps 1576: }
1577: }
1578:
1.25 schwarze 1579: static int
1.258 schwarze 1580: parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta,
1581: const struct roff_node *n)
1582: {
1583: uint64_t mask;
1584:
1585: mask = TYPE_Fa;
1586: if (n->sec == SEC_SYNOPSIS)
1587: mask |= TYPE_Vt;
1588:
1589: putmdockey(mpage, n->child, mask, 0);
1590: return 0;
1591: }
1592:
1593: static int
1.190 schwarze 1594: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1595: const struct roff_node *n)
1.1 kristaps 1596: {
1.176 schwarze 1597: char *start, *end;
1.1 kristaps 1598: size_t sz;
1.25 schwarze 1599:
1.50 kristaps 1600: if (SEC_SYNOPSIS != n->sec ||
1.141 schwarze 1601: NULL == (n = n->child) ||
1.188 schwarze 1602: n->type != ROFFT_TEXT)
1.197 schwarze 1603: return 0;
1.1 kristaps 1604:
1605: /*
1606: * Only consider those `Fd' macro fields that begin with an
1607: * "inclusion" token (versus, e.g., #define).
1608: */
1.50 kristaps 1609:
1.1 kristaps 1610: if (strcmp("#include", n->string))
1.197 schwarze 1611: return 0;
1.1 kristaps 1612:
1.188 schwarze 1613: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.197 schwarze 1614: return 0;
1.1 kristaps 1615:
1616: /*
1617: * Strip away the enclosing angle brackets and make sure we're
1618: * not zero-length.
1619: */
1620:
1621: start = n->string;
1622: if ('<' == *start || '"' == *start)
1623: start++;
1624:
1625: if (0 == (sz = strlen(start)))
1.197 schwarze 1626: return 0;
1.1 kristaps 1627:
1628: end = &start[(int)sz - 1];
1629: if ('>' == *end || '"' == *end)
1630: end--;
1631:
1.50 kristaps 1632: if (end > start)
1.78 schwarze 1633: putkeys(mpage, start, end - start + 1, TYPE_In);
1.197 schwarze 1634: return 0;
1.1 kristaps 1635: }
1636:
1.178 schwarze 1637: static void
1.189 schwarze 1638: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 kristaps 1639: {
1.112 schwarze 1640: char *cp;
1.178 schwarze 1641: size_t sz;
1.1 kristaps 1642:
1.188 schwarze 1643: if (n->type != ROFFT_TEXT)
1.178 schwarze 1644: return;
1.25 schwarze 1645:
1.178 schwarze 1646: /* Skip function pointer punctuation. */
1.1 kristaps 1647:
1.178 schwarze 1648: cp = n->string;
1649: while (*cp == '(' || *cp == '*')
1.1 kristaps 1650: cp++;
1.178 schwarze 1651: sz = strcspn(cp, "()");
1.1 kristaps 1652:
1.178 schwarze 1653: putkeys(mpage, cp, sz, TYPE_Fn);
1.173 schwarze 1654: if (n->sec == SEC_SYNOPSIS)
1.178 schwarze 1655: putkeys(mpage, cp, sz, NAME_SYN);
1656: }
1.25 schwarze 1657:
1.178 schwarze 1658: static int
1.190 schwarze 1659: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1660: const struct roff_node *n)
1.178 schwarze 1661: {
1.258 schwarze 1662: uint64_t mask;
1.178 schwarze 1663:
1664: if (n->child == NULL)
1.197 schwarze 1665: return 0;
1.178 schwarze 1666:
1667: parse_mdoc_fname(mpage, n->child);
1.25 schwarze 1668:
1.258 schwarze 1669: n = n->child->next;
1670: if (n != NULL && n->type == ROFFT_TEXT) {
1671: mask = TYPE_Fa;
1672: if (n->sec == SEC_SYNOPSIS)
1673: mask |= TYPE_Vt;
1674: putmdockey(mpage, n, mask, 0);
1675: }
1.25 schwarze 1676:
1.197 schwarze 1677: return 0;
1.173 schwarze 1678: }
1679:
1680: static int
1.190 schwarze 1681: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1682: const struct roff_node *n)
1.173 schwarze 1683: {
1.177 schwarze 1684:
1.188 schwarze 1685: if (n->type != ROFFT_HEAD)
1.197 schwarze 1686: return 1;
1.173 schwarze 1687:
1.178 schwarze 1688: if (n->child != NULL)
1689: parse_mdoc_fname(mpage, n->child);
1690:
1.197 schwarze 1691: return 0;
1.1 kristaps 1692: }
1693:
1.25 schwarze 1694: static int
1.211 schwarze 1695: parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1696: const struct roff_node *n)
1697: {
1698: char *cp;
1699:
1700: if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1701: return 0;
1702:
1.215 schwarze 1703: if (n->child != NULL &&
1704: n->child->next == NULL &&
1705: n->child->type == ROFFT_TEXT)
1.211 schwarze 1706: return 1;
1707:
1708: cp = NULL;
1709: deroff(&cp, n);
1710: if (cp != NULL) {
1711: putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1712: n->type == ROFFT_BODY ? TYPE_Va : 0));
1713: free(cp);
1714: }
1715:
1716: return 0;
1717: }
1718:
1719: static int
1.190 schwarze 1720: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1721: const struct roff_node *n)
1.1 kristaps 1722: {
1.67 schwarze 1723: char *cp;
1.1 kristaps 1724:
1725: if (NULL == (n = n->child))
1.197 schwarze 1726: return 0;
1.1 kristaps 1727:
1.67 schwarze 1728: if (NULL == n->next) {
1.78 schwarze 1729: putkey(mpage, n->string, TYPE_Xr);
1.197 schwarze 1730: return 0;
1.67 schwarze 1731: }
1732:
1.120 schwarze 1733: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.78 schwarze 1734: putkey(mpage, cp, TYPE_Xr);
1.67 schwarze 1735: free(cp);
1.197 schwarze 1736: return 0;
1.1 kristaps 1737: }
1738:
1.25 schwarze 1739: static int
1.190 schwarze 1740: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1741: const struct roff_node *n)
1.1 kristaps 1742: {
1743:
1.188 schwarze 1744: if (n->type == ROFFT_BODY)
1.194 schwarze 1745: deroff(&mpage->desc, n);
1.197 schwarze 1746: return 0;
1.1 kristaps 1747: }
1748:
1.25 schwarze 1749: static int
1.190 schwarze 1750: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1751: const struct roff_node *n)
1.1 kristaps 1752: {
1753:
1.129 schwarze 1754: if (SEC_NAME == n->sec)
1.235 schwarze 1755: putmdockey(mpage, n->child, NAME_TITLE, 0);
1.188 schwarze 1756: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.172 schwarze 1757: if (n->child == NULL)
1758: putkey(mpage, meta->name, NAME_SYN);
1759: else
1.235 schwarze 1760: putmdockey(mpage, n->child, NAME_SYN, 0);
1.174 schwarze 1761: }
1762: if ( ! (mpage->name_head_done ||
1763: n->child == NULL || n->child->string == NULL ||
1764: strcasecmp(n->child->string, meta->title))) {
1.226 schwarze 1765: putkey(mpage, n->child->string, NAME_HEAD);
1.174 schwarze 1766: mpage->name_head_done = 1;
1.172 schwarze 1767: }
1.197 schwarze 1768: return 0;
1.1 kristaps 1769: }
1770:
1.25 schwarze 1771: static int
1.190 schwarze 1772: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1773: const struct roff_node *n)
1.1 kristaps 1774: {
1775:
1.197 schwarze 1776: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 kristaps 1777: }
1778:
1.50 kristaps 1779: static int
1.190 schwarze 1780: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1781: const struct roff_node *n)
1.1 kristaps 1782: {
1783:
1.197 schwarze 1784: return n->type == ROFFT_HEAD;
1.1 kristaps 1785: }
1786:
1.50 kristaps 1787: /*
1.66 schwarze 1788: * Add a string to the hash table for the current manual.
1789: * Each string has a bitmask telling which macros it belongs to.
1790: * When we finish the manual, we'll dump the table.
1.50 kristaps 1791: */
1792: static void
1.176 schwarze 1793: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.50 kristaps 1794: {
1.133 schwarze 1795: struct ohash *htab;
1.50 kristaps 1796: struct str *s;
1.111 schwarze 1797: const char *end;
1.65 schwarze 1798: unsigned int slot;
1.176 schwarze 1799: int i, mustfree;
1.25 schwarze 1800:
1.50 kristaps 1801: if (0 == sz)
1802: return;
1.111 schwarze 1803:
1.176 schwarze 1804: mustfree = render_string(&cp, &sz);
1805:
1.133 schwarze 1806: if (TYPE_Nm & v) {
1807: htab = &names;
1808: v &= name_mask;
1.169 schwarze 1809: if (v & NAME_FIRST)
1810: name_mask &= ~NAME_FIRST;
1.133 schwarze 1811: if (debug > 1)
1812: say(mpage->mlinks->file,
1.234 schwarze 1813: "Adding name %*s, bits=0x%llx", (int)sz, cp,
1814: (unsigned long long)v);
1.133 schwarze 1815: } else {
1816: htab = &strings;
1817: if (debug > 1)
1.221 schwarze 1818: for (i = 0; i < KEY_MAX; i++)
1.163 schwarze 1819: if ((uint64_t)1 << i & v)
1.133 schwarze 1820: say(mpage->mlinks->file,
1821: "Adding key %s=%*s",
1.220 schwarze 1822: mansearch_keynames[i], (int)sz, cp);
1.111 schwarze 1823: }
1.25 schwarze 1824:
1.65 schwarze 1825: end = cp + sz;
1.133 schwarze 1826: slot = ohash_qlookupi(htab, cp, &end);
1827: s = ohash_find(htab, slot);
1.25 schwarze 1828:
1.78 schwarze 1829: if (NULL != s && mpage == s->mpage) {
1.50 kristaps 1830: s->mask |= v;
1831: return;
1832: } else if (NULL == s) {
1.144 schwarze 1833: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.50 kristaps 1834: memcpy(s->key, cp, sz);
1.133 schwarze 1835: ohash_insert(htab, slot, s);
1.1 kristaps 1836: }
1.78 schwarze 1837: s->mpage = mpage;
1.50 kristaps 1838: s->mask = v;
1.176 schwarze 1839:
1840: if (mustfree)
1841: free(cp);
1.1 kristaps 1842: }
1843:
1.50 kristaps 1844: /*
1845: * Take a Unicode codepoint and produce its UTF-8 encoding.
1846: * This isn't the best way to do this, but it works.
1847: * The magic numbers are from the UTF-8 packaging.
1848: * They're not as scary as they seem: read the UTF-8 spec for details.
1849: */
1850: static size_t
1851: utf8(unsigned int cp, char out[7])
1.1 kristaps 1852: {
1.50 kristaps 1853: size_t rc;
1.1 kristaps 1854:
1.50 kristaps 1855: rc = 0;
1856: if (cp <= 0x0000007F) {
1857: rc = 1;
1858: out[0] = (char)cp;
1859: } else if (cp <= 0x000007FF) {
1860: rc = 2;
1861: out[0] = (cp >> 6 & 31) | 192;
1862: out[1] = (cp & 63) | 128;
1863: } else if (cp <= 0x0000FFFF) {
1864: rc = 3;
1865: out[0] = (cp >> 12 & 15) | 224;
1866: out[1] = (cp >> 6 & 63) | 128;
1867: out[2] = (cp & 63) | 128;
1868: } else if (cp <= 0x001FFFFF) {
1869: rc = 4;
1870: out[0] = (cp >> 18 & 7) | 240;
1871: out[1] = (cp >> 12 & 63) | 128;
1872: out[2] = (cp >> 6 & 63) | 128;
1873: out[3] = (cp & 63) | 128;
1874: } else if (cp <= 0x03FFFFFF) {
1875: rc = 5;
1876: out[0] = (cp >> 24 & 3) | 248;
1877: out[1] = (cp >> 18 & 63) | 128;
1878: out[2] = (cp >> 12 & 63) | 128;
1879: out[3] = (cp >> 6 & 63) | 128;
1880: out[4] = (cp & 63) | 128;
1881: } else if (cp <= 0x7FFFFFFF) {
1882: rc = 6;
1883: out[0] = (cp >> 30 & 1) | 252;
1884: out[1] = (cp >> 24 & 63) | 128;
1885: out[2] = (cp >> 18 & 63) | 128;
1886: out[3] = (cp >> 12 & 63) | 128;
1887: out[4] = (cp >> 6 & 63) | 128;
1888: out[5] = (cp & 63) | 128;
1889: } else
1.197 schwarze 1890: return 0;
1.1 kristaps 1891:
1.50 kristaps 1892: out[rc] = '\0';
1.197 schwarze 1893: return rc;
1.50 kristaps 1894: }
1.1 kristaps 1895:
1.50 kristaps 1896: /*
1.176 schwarze 1897: * If the string contains escape sequences,
1898: * replace it with an allocated rendering and return 1,
1899: * such that the caller can free it after use.
1900: * Otherwise, do nothing and return 0.
1.50 kristaps 1901: */
1.176 schwarze 1902: static int
1903: render_string(char **public, size_t *psz)
1.50 kristaps 1904: {
1.176 schwarze 1905: const char *src, *scp, *addcp, *seq;
1906: char *dst;
1907: size_t ssz, dsz, addsz;
1.114 schwarze 1908: char utfbuf[7], res[6];
1.176 schwarze 1909: int seqlen, unicode;
1.50 kristaps 1910:
1911: res[0] = '\\';
1912: res[1] = '\t';
1913: res[2] = ASCII_NBRSP;
1914: res[3] = ASCII_HYPH;
1.114 schwarze 1915: res[4] = ASCII_BREAK;
1916: res[5] = '\0';
1.1 kristaps 1917:
1.176 schwarze 1918: src = scp = *public;
1919: ssz = *psz;
1920: dst = NULL;
1921: dsz = 0;
1922:
1923: while (scp < src + *psz) {
1924:
1925: /* Leave normal characters unchanged. */
1926:
1927: if (strchr(res, *scp) == NULL) {
1928: if (dst != NULL)
1929: dst[dsz++] = *scp;
1930: scp++;
1931: continue;
1932: }
1.46 kristaps 1933:
1.50 kristaps 1934: /*
1.176 schwarze 1935: * Found something that requires replacing,
1936: * make sure we have a destination buffer.
1.50 kristaps 1937: */
1.176 schwarze 1938:
1939: if (dst == NULL) {
1940: dst = mandoc_malloc(ssz + 1);
1941: dsz = scp - src;
1942: memcpy(dst, src, dsz);
1.50 kristaps 1943: }
1.46 kristaps 1944:
1.176 schwarze 1945: /* Handle single-char special characters. */
1946:
1947: switch (*scp) {
1948: case '\\':
1949: break;
1.141 schwarze 1950: case '\t':
1951: case ASCII_NBRSP:
1.176 schwarze 1952: dst[dsz++] = ' ';
1953: scp++;
1954: continue;
1955: case ASCII_HYPH:
1956: dst[dsz++] = '-';
1.114 schwarze 1957: /* FALLTHROUGH */
1.141 schwarze 1958: case ASCII_BREAK:
1.176 schwarze 1959: scp++;
1.50 kristaps 1960: continue;
1.114 schwarze 1961: default:
1.176 schwarze 1962: abort();
1.114 schwarze 1963: }
1.46 kristaps 1964:
1.50 kristaps 1965: /*
1.176 schwarze 1966: * Found an escape sequence.
1967: * Read past the slash, then parse it.
1968: * Ignore everything except characters.
1.50 kristaps 1969: */
1.95 schwarze 1970:
1.176 schwarze 1971: scp++;
1972: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.50 kristaps 1973: continue;
1.1 kristaps 1974:
1.50 kristaps 1975: /*
1.95 schwarze 1976: * Render the special character
1977: * as either UTF-8 or ASCII.
1.50 kristaps 1978: */
1.95 schwarze 1979:
1980: if (write_utf8) {
1.203 schwarze 1981: unicode = mchars_spec2cp(seq, seqlen);
1.176 schwarze 1982: if (unicode <= 0)
1.95 schwarze 1983: continue;
1.176 schwarze 1984: addsz = utf8(unicode, utfbuf);
1985: if (addsz == 0)
1.95 schwarze 1986: continue;
1.176 schwarze 1987: addcp = utfbuf;
1.95 schwarze 1988: } else {
1.203 schwarze 1989: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.176 schwarze 1990: if (addcp == NULL)
1.95 schwarze 1991: continue;
1.176 schwarze 1992: if (*addcp == ASCII_NBRSP) {
1993: addcp = " ";
1994: addsz = 1;
1.95 schwarze 1995: }
1996: }
1.1 kristaps 1997:
1.50 kristaps 1998: /* Copy the rendered glyph into the stream. */
1.1 kristaps 1999:
1.176 schwarze 2000: ssz += addsz;
2001: dst = mandoc_realloc(dst, ssz + 1);
2002: memcpy(dst + dsz, addcp, addsz);
2003: dsz += addsz;
2004: }
2005: if (dst != NULL) {
2006: *public = dst;
2007: *psz = dsz;
2008: }
2009:
2010: /* Trim trailing whitespace and NUL-terminate. */
2011:
2012: while (*psz > 0 && (*public)[*psz - 1] == ' ')
2013: --*psz;
2014: if (dst != NULL) {
2015: (*public)[*psz] = '\0';
1.197 schwarze 2016: return 1;
1.176 schwarze 2017: } else
1.197 schwarze 2018: return 0;
1.1 kristaps 2019: }
2020:
1.118 schwarze 2021: static void
2022: dbadd_mlink(const struct mlink *mlink)
2023: {
1.221 schwarze 2024: dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
2025: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
1.223 schwarze 2026: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
1.221 schwarze 2027: dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
2028: dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
1.118 schwarze 2029: }
2030:
1.14 schwarze 2031: /*
1.50 kristaps 2032: * Flush the current page's terms (and their bits) into the database.
1.96 schwarze 2033: * Also, handle escape sequences at the last possible moment.
1.14 schwarze 2034: */
2035: static void
1.221 schwarze 2036: dbadd(struct dba *dba, struct mpage *mpage)
1.14 schwarze 2037: {
1.87 schwarze 2038: struct mlink *mlink;
1.50 kristaps 2039: struct str *key;
1.176 schwarze 2040: char *cp;
1.221 schwarze 2041: uint64_t mask;
1.52 kristaps 2042: size_t i;
1.66 schwarze 2043: unsigned int slot;
1.176 schwarze 2044: int mustfree;
1.14 schwarze 2045:
1.128 schwarze 2046: mlink = mpage->mlinks;
1.43 kristaps 2047:
1.128 schwarze 2048: if (nodb) {
1.147 schwarze 2049: for (key = ohash_first(&names, &slot); NULL != key;
1.176 schwarze 2050: key = ohash_next(&names, &slot))
1.147 schwarze 2051: free(key);
2052: for (key = ohash_first(&strings, &slot); NULL != key;
1.176 schwarze 2053: key = ohash_next(&strings, &slot))
1.147 schwarze 2054: free(key);
1.145 schwarze 2055: if (0 == debug)
2056: return;
1.128 schwarze 2057: while (NULL != mlink) {
2058: fputs(mlink->name, stdout);
2059: if (NULL == mlink->next ||
2060: strcmp(mlink->dsec, mlink->next->dsec) ||
2061: strcmp(mlink->fsec, mlink->next->fsec) ||
2062: strcmp(mlink->arch, mlink->next->arch)) {
2063: putchar('(');
2064: if ('\0' == *mlink->dsec)
2065: fputs(mlink->fsec, stdout);
2066: else
2067: fputs(mlink->dsec, stdout);
2068: if ('\0' != *mlink->arch)
2069: printf("/%s", mlink->arch);
2070: putchar(')');
2071: }
2072: mlink = mlink->next;
2073: if (NULL != mlink)
2074: fputs(", ", stdout);
2075: }
1.132 schwarze 2076: printf(" - %s\n", mpage->desc);
1.14 schwarze 2077: return;
1.128 schwarze 2078: }
2079:
2080: if (debug)
2081: say(mlink->file, "Adding to database");
1.28 kristaps 2082:
1.176 schwarze 2083: cp = mpage->desc;
2084: i = strlen(cp);
2085: mustfree = render_string(&cp, &i);
1.227 schwarze 2086: mpage->dba = dba_page_new(dba->pages,
1.223 schwarze 2087: *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2088: cp, mlink->file, mpage->form);
1.176 schwarze 2089: if (mustfree)
2090: free(cp);
1.227 schwarze 2091: dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
1.81 schwarze 2092:
1.223 schwarze 2093: while (mlink != NULL) {
1.118 schwarze 2094: dbadd_mlink(mlink);
1.223 schwarze 2095: mlink = mlink->next;
2096: }
1.50 kristaps 2097:
1.133 schwarze 2098: for (key = ohash_first(&names, &slot); NULL != key;
2099: key = ohash_next(&names, &slot)) {
2100: assert(key->mpage == mpage);
1.221 schwarze 2101: dba_page_alias(mpage->dba, key->key, key->mask);
1.133 schwarze 2102: free(key);
2103: }
1.66 schwarze 2104: for (key = ohash_first(&strings, &slot); NULL != key;
2105: key = ohash_next(&strings, &slot)) {
1.78 schwarze 2106: assert(key->mpage == mpage);
1.221 schwarze 2107: i = 0;
2108: for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2109: if (key->mask & mask)
2110: dba_macro_add(dba->macros, i,
2111: key->key, mpage->dba);
2112: i++;
2113: }
1.66 schwarze 2114: free(key);
1.38 schwarze 2115: }
1.14 schwarze 2116: }
2117:
1.5 kristaps 2118: static void
1.221 schwarze 2119: dbprune(struct dba *dba)
1.5 kristaps 2120: {
1.221 schwarze 2121: struct dba_array *page, *files;
2122: char *file;
1.5 kristaps 2123:
1.221 schwarze 2124: dba_array_FOREACH(dba->pages, page) {
2125: files = dba_array_get(page, DBP_FILE);
2126: dba_array_FOREACH(files, file) {
2127: if (*file < ' ')
2128: file++;
2129: if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2130: file)) != NULL) {
2131: if (debug)
2132: say(file, "Deleting from database");
2133: dba_array_del(dba->pages);
2134: break;
2135: }
1.106 schwarze 2136: }
1.5 kristaps 2137: }
2138: }
2139:
1.4 kristaps 2140: /*
1.221 schwarze 2141: * Write the database from memory to disk.
1.4 kristaps 2142: */
1.35 kristaps 2143: static void
1.221 schwarze 2144: dbwrite(struct dba *dba)
1.4 kristaps 2145: {
1.255 schwarze 2146: struct stat sb1, sb2;
2147: char tfn[33], *cp1, *cp2;
2148: off_t i;
2149: int fd1, fd2;
1.4 kristaps 2150:
1.250 schwarze 2151: /*
2152: * Do not write empty databases, and delete existing ones
2153: * when makewhatis -u causes them to become empty.
2154: */
2155:
2156: dba_array_start(dba->pages);
2157: if (dba_array_next(dba->pages) == NULL) {
1.252 schwarze 2158: if (unlink(MANDOC_DB) == -1 && errno != ENOENT)
1.250 schwarze 2159: say(MANDOC_DB, "&unlink");
2160: return;
2161: }
2162:
2163: /*
2164: * Build the database in a temporary file,
2165: * then atomically move it into place.
2166: */
2167:
1.221 schwarze 2168: if (dba_write(MANDOC_DB "~", dba) != -1) {
2169: if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2170: exitcode = (int)MANDOCLEVEL_SYSERR;
2171: say(MANDOC_DB, "&rename");
2172: unlink(MANDOC_DB "~");
2173: }
1.38 schwarze 2174: return;
1.4 kristaps 2175: }
1.250 schwarze 2176:
2177: /*
2178: * We lack write permission and cannot replace the database
2179: * file, but let's at least check whether the data changed.
2180: */
1.4 kristaps 2181:
1.221 schwarze 2182: (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2183: if (mkdtemp(tfn) == NULL) {
2184: exitcode = (int)MANDOCLEVEL_SYSERR;
2185: say("", "&%s", tfn);
1.50 kristaps 2186: return;
1.221 schwarze 2187: }
1.256 schwarze 2188: cp1 = cp2 = MAP_FAILED;
1.255 schwarze 2189: fd1 = fd2 = -1;
1.221 schwarze 2190: (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2191: if (dba_write(tfn, dba) == -1) {
2192: say(tfn, "&dba_write");
1.255 schwarze 2193: goto err;
1.115 schwarze 2194: }
1.255 schwarze 2195: if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) {
2196: say(MANDOC_DB, "&open");
2197: goto err;
2198: }
2199: if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) {
2200: say(tfn, "&open");
2201: goto err;
2202: }
2203: if (fstat(fd1, &sb1) == -1) {
2204: say(MANDOC_DB, "&fstat");
2205: goto err;
2206: }
2207: if (fstat(fd2, &sb2) == -1) {
2208: say(tfn, "&fstat");
2209: goto err;
2210: }
2211: if (sb1.st_size != sb2.st_size)
2212: goto err;
2213: if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE,
1.256 schwarze 2214: fd1, 0)) == MAP_FAILED) {
1.255 schwarze 2215: say(MANDOC_DB, "&mmap");
2216: goto err;
2217: }
2218: if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE,
1.256 schwarze 2219: fd2, 0)) == MAP_FAILED) {
1.255 schwarze 2220: say(tfn, "&mmap");
2221: goto err;
2222: }
2223: for (i = 0; i < sb1.st_size; i++)
2224: if (cp1[i] != cp2[i])
2225: goto err;
2226: goto out;
2227:
2228: err:
2229: exitcode = (int)MANDOCLEVEL_SYSERR;
2230: say(MANDOC_DB, "Data changed, but cannot replace database");
1.115 schwarze 2231:
1.221 schwarze 2232: out:
1.256 schwarze 2233: if (cp1 != MAP_FAILED)
1.255 schwarze 2234: munmap(cp1, sb1.st_size);
1.256 schwarze 2235: if (cp2 != MAP_FAILED)
1.255 schwarze 2236: munmap(cp2, sb2.st_size);
2237: if (fd1 != -1)
2238: close(fd1);
2239: if (fd2 != -1)
2240: close(fd2);
1.254 schwarze 2241: unlink(tfn);
1.221 schwarze 2242: *strrchr(tfn, '/') = '\0';
1.254 schwarze 2243: rmdir(tfn);
1.4 kristaps 2244: }
2245:
1.50 kristaps 2246: static int
1.165 schwarze 2247: set_basedir(const char *targetdir, int report_baddir)
1.4 kristaps 2248: {
1.59 schwarze 2249: static char startdir[PATH_MAX];
1.151 schwarze 2250: static int getcwd_status; /* 1 = ok, 2 = failure */
2251: static int chdir_status; /* 1 = changed directory */
1.150 schwarze 2252: char *cp;
1.4 kristaps 2253:
1.59 schwarze 2254: /*
1.151 schwarze 2255: * Remember the original working directory, if possible.
2256: * This will be needed if the second or a later directory
2257: * on the command line is given as a relative path.
2258: * Do not error out if the current directory is not
2259: * searchable: Maybe it won't be needed after all.
2260: */
2261: if (0 == getcwd_status) {
2262: if (NULL == getcwd(startdir, sizeof(startdir))) {
2263: getcwd_status = 2;
2264: (void)strlcpy(startdir, strerror(errno),
2265: sizeof(startdir));
2266: } else
2267: getcwd_status = 1;
2268: }
2269:
2270: /*
2271: * We are leaving the old base directory.
2272: * Do not use it any longer, not even for messages.
2273: */
2274: *basedir = '\0';
2275:
2276: /*
2277: * If and only if the directory was changed earlier and
2278: * the next directory to process is given as a relative path,
2279: * first go back, or bail out if that is impossible.
1.59 schwarze 2280: */
1.151 schwarze 2281: if (chdir_status && '/' != *targetdir) {
2282: if (2 == getcwd_status) {
1.59 schwarze 2283: exitcode = (int)MANDOCLEVEL_SYSERR;
1.151 schwarze 2284: say("", "getcwd: %s", startdir);
1.197 schwarze 2285: return 0;
1.59 schwarze 2286: }
1.151 schwarze 2287: if (-1 == chdir(startdir)) {
1.59 schwarze 2288: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2289: say("", "&chdir %s", startdir);
1.197 schwarze 2290: return 0;
1.59 schwarze 2291: }
2292: }
1.151 schwarze 2293:
2294: /*
2295: * Always resolve basedir to the canonicalized absolute
2296: * pathname and append a trailing slash, such that
2297: * we can reliably check whether files are inside.
2298: */
1.59 schwarze 2299: if (NULL == realpath(targetdir, basedir)) {
1.165 schwarze 2300: if (report_baddir || errno != ENOENT) {
2301: exitcode = (int)MANDOCLEVEL_BADARG;
2302: say("", "&%s: realpath", targetdir);
2303: }
1.197 schwarze 2304: return 0;
1.59 schwarze 2305: } else if (-1 == chdir(basedir)) {
1.165 schwarze 2306: if (report_baddir || errno != ENOENT) {
2307: exitcode = (int)MANDOCLEVEL_BADARG;
2308: say("", "&chdir");
2309: }
1.197 schwarze 2310: return 0;
1.150 schwarze 2311: }
1.151 schwarze 2312: chdir_status = 1;
1.150 schwarze 2313: cp = strchr(basedir, '\0');
2314: if ('/' != cp[-1]) {
2315: if (cp - basedir >= PATH_MAX - 1) {
2316: exitcode = (int)MANDOCLEVEL_SYSERR;
2317: say("", "Filename too long");
1.197 schwarze 2318: return 0;
1.150 schwarze 2319: }
2320: *cp++ = '/';
2321: *cp = '\0';
1.4 kristaps 2322: }
1.197 schwarze 2323: return 1;
1.56 schwarze 2324: }
2325:
2326: static void
1.59 schwarze 2327: say(const char *file, const char *format, ...)
1.56 schwarze 2328: {
2329: va_list ap;
1.123 schwarze 2330: int use_errno;
1.56 schwarze 2331:
1.59 schwarze 2332: if ('\0' != *basedir)
2333: fprintf(stderr, "%s", basedir);
2334: if ('\0' != *basedir && '\0' != *file)
1.151 schwarze 2335: fputc('/', stderr);
1.56 schwarze 2336: if ('\0' != *file)
1.59 schwarze 2337: fprintf(stderr, "%s", file);
2338:
1.123 schwarze 2339: use_errno = 1;
2340: if (NULL != format) {
2341: switch (*format) {
1.141 schwarze 2342: case '&':
1.123 schwarze 2343: format++;
2344: break;
1.141 schwarze 2345: case '\0':
1.123 schwarze 2346: format = NULL;
2347: break;
2348: default:
2349: use_errno = 0;
2350: break;
2351: }
2352: }
2353: if (NULL != format) {
2354: if ('\0' != *basedir || '\0' != *file)
2355: fputs(": ", stderr);
2356: va_start(ap, format);
2357: vfprintf(stderr, format, ap);
2358: va_end(ap);
2359: }
2360: if (use_errno) {
2361: if ('\0' != *basedir || '\0' != *file || NULL != format)
2362: fputs(": ", stderr);
1.59 schwarze 2363: perror(NULL);
1.123 schwarze 2364: } else
2365: fputc('\n', stderr);
1.1 kristaps 2366: }
CVSweb