Annotation of mandoc/mandocdb.c, Revision 1.269
1.269 ! schwarze 1: /* $Id: mandocdb.c,v 1.268 2021/08/07 13:02:10 schwarze Exp $ */
1.1 kristaps 2: /*
1.267 schwarze 3: * Copyright (c) 2011-2020 Ingo Schwarze <schwarze@openbsd.org>
1.48 schwarze 4: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
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.
1.267 schwarze 18: *
19: * Implementation of the makewhatis(8) program.
1.1 kristaps 20: */
21: #include "config.h"
22:
1.162 schwarze 23: #include <sys/types.h>
1.255 schwarze 24: #include <sys/mman.h>
1.50 kristaps 25: #include <sys/stat.h>
1.1 kristaps 26:
27: #include <assert.h>
1.43 kristaps 28: #include <ctype.h>
1.208 schwarze 29: #if HAVE_ERR
1.198 schwarze 30: #include <err.h>
1.208 schwarze 31: #endif
1.63 schwarze 32: #include <errno.h>
1.1 kristaps 33: #include <fcntl.h>
1.158 schwarze 34: #if HAVE_FTS
1.50 kristaps 35: #include <fts.h>
1.157 schwarze 36: #else
37: #include "compat_fts.h"
38: #endif
1.58 schwarze 39: #include <limits.h>
1.218 kristaps 40: #if HAVE_SANDBOX_INIT
41: #include <sandbox.h>
42: #endif
1.221 schwarze 43: #include <stdarg.h>
1.50 kristaps 44: #include <stddef.h>
1.59 schwarze 45: #include <stdio.h>
1.1 kristaps 46: #include <stdint.h>
47: #include <stdlib.h>
48: #include <string.h>
1.17 schwarze 49: #include <unistd.h>
1.1 kristaps 50:
1.188 schwarze 51: #include "mandoc_aux.h"
1.202 schwarze 52: #include "mandoc_ohash.h"
1.188 schwarze 53: #include "mandoc.h"
54: #include "roff.h"
1.50 kristaps 55: #include "mdoc.h"
1.1 kristaps 56: #include "man.h"
1.260 schwarze 57: #include "mandoc_parse.h"
1.187 schwarze 58: #include "manconf.h"
1.55 kristaps 59: #include "mansearch.h"
1.221 schwarze 60: #include "dba_array.h"
61: #include "dba.h"
1.1 kristaps 62:
1.111 schwarze 63: extern const char *const mansearch_keynames[];
64:
1.50 kristaps 65: enum op {
66: OP_DEFAULT = 0, /* new dbs from dir list or default config */
67: OP_CONFFILE, /* new databases from custom config file */
68: OP_UPDATE, /* delete/add entries in existing database */
69: OP_DELETE, /* delete entries from existing database */
70: OP_TEST /* change no databases, report potential problems */
1.38 schwarze 71: };
72:
1.50 kristaps 73: struct str {
1.78 schwarze 74: const struct mpage *mpage; /* if set, the owning parse */
1.50 kristaps 75: uint64_t mask; /* bitmask in sequence */
1.176 schwarze 76: char key[]; /* rendered text */
1.38 schwarze 77: };
78:
1.79 schwarze 79: struct inodev {
80: ino_t st_ino;
81: dev_t st_dev;
1.50 kristaps 82: };
1.5 kristaps 83:
1.78 schwarze 84: struct mpage {
1.79 schwarze 85: struct inodev inodev; /* used for hashing routine */
1.221 schwarze 86: struct dba_array *dba;
1.82 schwarze 87: char *sec; /* section from file content */
88: char *arch; /* architecture from file content */
89: char *title; /* title from file content */
90: char *desc; /* description from file content */
1.231 schwarze 91: struct mpage *next; /* singly linked list */
1.82 schwarze 92: struct mlink *mlinks; /* singly linked list */
1.174 schwarze 93: int name_head_done;
1.221 schwarze 94: enum form form; /* format from file content */
1.82 schwarze 95: };
96:
97: struct mlink {
1.58 schwarze 98: char file[PATH_MAX]; /* filename rel. to manpath */
1.82 schwarze 99: char *dsec; /* section from directory */
100: char *arch; /* architecture from directory */
101: char *name; /* name from file name (not empty) */
102: char *fsec; /* section from file name suffix */
1.85 schwarze 103: struct mlink *next; /* singly linked list */
1.118 schwarze 104: struct mpage *mpage; /* parent */
1.124 schwarze 105: int gzip; /* filename has a .gz suffix */
1.221 schwarze 106: enum form dform; /* format from directory */
107: enum form fform; /* format from file name suffix */
1.1 kristaps 108: };
109:
1.190 schwarze 110: typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
1.189 schwarze 111: const struct roff_node *);
1.1 kristaps 112:
1.50 kristaps 113: struct mdoc_handler {
114: mdoc_fp fp; /* optional handler */
115: uint64_t mask; /* set unless handler returns 0 */
1.235 schwarze 116: int taboo; /* node flags that must not be set */
1.1 kristaps 117: };
1.219 schwarze 118:
119:
120: int mandocdb(int, char *[]);
1.1 kristaps 121:
1.221 schwarze 122: static void dbadd(struct dba *, struct mpage *);
1.267 schwarze 123: static void dbadd_mlink(const struct mlink *);
1.221 schwarze 124: static void dbprune(struct dba *);
125: static void dbwrite(struct dba *);
1.59 schwarze 126: static void filescan(const char *);
1.232 schwarze 127: #if HAVE_FTS_COMPARE_CONST
128: static int fts_compare(const FTSENT *const *, const FTSENT *const *);
129: #else
1.231 schwarze 130: static int fts_compare(const FTSENT **, const FTSENT **);
1.232 schwarze 131: #endif
1.84 schwarze 132: static void mlink_add(struct mlink *, const struct stat *);
1.131 schwarze 133: static void mlink_check(struct mpage *, struct mlink *);
1.82 schwarze 134: static void mlink_free(struct mlink *);
1.89 schwarze 135: static void mlinks_undupe(struct mpage *);
1.78 schwarze 136: static void mpages_free(void);
1.221 schwarze 137: static void mpages_merge(struct dba *, struct mparse *);
1.124 schwarze 138: static void parse_cat(struct mpage *, int);
1.190 schwarze 139: static void parse_man(struct mpage *, const struct roff_meta *,
1.189 schwarze 140: const struct roff_node *);
1.190 schwarze 141: static void parse_mdoc(struct mpage *, const struct roff_meta *,
1.189 schwarze 142: const struct roff_node *);
1.190 schwarze 143: static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
1.189 schwarze 144: const struct roff_node *);
1.258 schwarze 145: static int parse_mdoc_Fa(struct mpage *, const struct roff_meta *,
146: const struct roff_node *);
1.190 schwarze 147: static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
1.189 schwarze 148: const struct roff_node *);
149: static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
1.190 schwarze 150: static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
1.189 schwarze 151: const struct roff_node *);
1.190 schwarze 152: static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
1.189 schwarze 153: const struct roff_node *);
1.190 schwarze 154: static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
1.189 schwarze 155: const struct roff_node *);
1.190 schwarze 156: static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
1.189 schwarze 157: const struct roff_node *);
1.190 schwarze 158: static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
1.189 schwarze 159: const struct roff_node *);
1.211 schwarze 160: static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
161: const struct roff_node *);
1.190 schwarze 162: static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
1.189 schwarze 163: const struct roff_node *);
1.112 schwarze 164: static void putkey(const struct mpage *, char *, uint64_t);
1.176 schwarze 165: static void putkeys(const struct mpage *, char *, size_t, uint64_t);
1.78 schwarze 166: static void putmdockey(const struct mpage *,
1.235 schwarze 167: const struct roff_node *, uint64_t, int);
1.268 schwarze 168: #ifdef READ_ALLOWED_PATH
169: static int read_allowed(const char *);
170: #endif
1.176 schwarze 171: static int render_string(char **, size_t *);
1.220 schwarze 172: static void say(const char *, const char *, ...)
1.243 schwarze 173: __attribute__((__format__ (__printf__, 2, 3)));
1.165 schwarze 174: static int set_basedir(const char *, int);
1.59 schwarze 175: static int treescan(void);
1.50 kristaps 176: static size_t utf8(unsigned int, char [7]);
177:
1.102 schwarze 178: static int nodb; /* no database changes */
1.116 schwarze 179: static int mparse_options; /* abort the parse early */
1.141 schwarze 180: static int use_all; /* use all found files */
181: static int debug; /* print what we're doing */
182: static int warnings; /* warn about crap */
1.95 schwarze 183: static int write_utf8; /* write UTF-8 output; else ASCII */
1.59 schwarze 184: static int exitcode; /* to be returned by main */
1.141 schwarze 185: static enum op op; /* operational mode */
1.59 schwarze 186: static char basedir[PATH_MAX]; /* current base directory */
1.264 schwarze 187: static size_t basedir_len; /* strlen(basedir) */
1.231 schwarze 188: static struct mpage *mpage_head; /* list of distinct manual pages */
1.78 schwarze 189: static struct ohash mpages; /* table of distinct manual pages */
1.83 schwarze 190: static struct ohash mlinks; /* table of directory entries */
1.133 schwarze 191: static struct ohash names; /* table of all names */
1.50 kristaps 192: static struct ohash strings; /* table of all strings */
1.133 schwarze 193: static uint64_t name_mask;
1.25 schwarze 194:
1.259 schwarze 195: static const struct mdoc_handler mdoc_handlers[MDOC_MAX - MDOC_Dd] = {
1.235 schwarze 196: { NULL, 0, NODE_NOPRT }, /* Dd */
197: { NULL, 0, NODE_NOPRT }, /* Dt */
198: { NULL, 0, NODE_NOPRT }, /* Os */
199: { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
200: { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
201: { NULL, 0, 0 }, /* Pp */
202: { NULL, 0, 0 }, /* D1 */
203: { NULL, 0, 0 }, /* Dl */
204: { NULL, 0, 0 }, /* Bd */
205: { NULL, 0, 0 }, /* Ed */
206: { NULL, 0, 0 }, /* Bl */
207: { NULL, 0, 0 }, /* El */
208: { NULL, 0, 0 }, /* It */
209: { NULL, 0, 0 }, /* Ad */
210: { NULL, TYPE_An, 0 }, /* An */
1.246 schwarze 211: { NULL, 0, 0 }, /* Ap */
1.235 schwarze 212: { NULL, TYPE_Ar, 0 }, /* Ar */
213: { NULL, TYPE_Cd, 0 }, /* Cd */
214: { NULL, TYPE_Cm, 0 }, /* Cm */
215: { NULL, TYPE_Dv, 0 }, /* Dv */
216: { NULL, TYPE_Er, 0 }, /* Er */
217: { NULL, TYPE_Ev, 0 }, /* Ev */
218: { NULL, 0, 0 }, /* Ex */
1.258 schwarze 219: { parse_mdoc_Fa, 0, 0 }, /* Fa */
1.235 schwarze 220: { parse_mdoc_Fd, 0, 0 }, /* Fd */
221: { NULL, TYPE_Fl, 0 }, /* Fl */
222: { parse_mdoc_Fn, 0, 0 }, /* Fn */
1.258 schwarze 223: { NULL, TYPE_Ft | TYPE_Vt, 0 }, /* Ft */
1.235 schwarze 224: { NULL, TYPE_Ic, 0 }, /* Ic */
225: { NULL, TYPE_In, 0 }, /* In */
226: { NULL, TYPE_Li, 0 }, /* Li */
227: { parse_mdoc_Nd, 0, 0 }, /* Nd */
228: { parse_mdoc_Nm, 0, 0 }, /* Nm */
229: { NULL, 0, 0 }, /* Op */
230: { NULL, 0, 0 }, /* Ot */
231: { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
232: { NULL, 0, 0 }, /* Rv */
233: { NULL, TYPE_St, 0 }, /* St */
234: { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
235: { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
236: { parse_mdoc_Xr, 0, 0 }, /* Xr */
237: { NULL, 0, 0 }, /* %A */
238: { NULL, 0, 0 }, /* %B */
239: { NULL, 0, 0 }, /* %D */
240: { NULL, 0, 0 }, /* %I */
241: { NULL, 0, 0 }, /* %J */
242: { NULL, 0, 0 }, /* %N */
243: { NULL, 0, 0 }, /* %O */
244: { NULL, 0, 0 }, /* %P */
245: { NULL, 0, 0 }, /* %R */
246: { NULL, 0, 0 }, /* %T */
247: { NULL, 0, 0 }, /* %V */
248: { NULL, 0, 0 }, /* Ac */
249: { NULL, 0, 0 }, /* Ao */
250: { NULL, 0, 0 }, /* Aq */
1.237 schwarze 251: { NULL, TYPE_At, 0 }, /* At */
1.235 schwarze 252: { NULL, 0, 0 }, /* Bc */
253: { NULL, 0, 0 }, /* Bf */
254: { NULL, 0, 0 }, /* Bo */
255: { NULL, 0, 0 }, /* Bq */
256: { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
1.236 schwarze 257: { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
1.235 schwarze 258: { NULL, 0, 0 }, /* Db */
259: { NULL, 0, 0 }, /* Dc */
260: { NULL, 0, 0 }, /* Do */
261: { NULL, 0, 0 }, /* Dq */
262: { NULL, 0, 0 }, /* Ec */
263: { NULL, 0, 0 }, /* Ef */
264: { NULL, TYPE_Em, 0 }, /* Em */
265: { NULL, 0, 0 }, /* Eo */
266: { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
267: { NULL, TYPE_Ms, 0 }, /* Ms */
268: { NULL, 0, 0 }, /* No */
269: { NULL, 0, 0 }, /* Ns */
270: { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
271: { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
272: { NULL, 0, 0 }, /* Pc */
273: { NULL, 0, 0 }, /* Pf */
274: { NULL, 0, 0 }, /* Po */
275: { NULL, 0, 0 }, /* Pq */
276: { NULL, 0, 0 }, /* Qc */
277: { NULL, 0, 0 }, /* Ql */
278: { NULL, 0, 0 }, /* Qo */
279: { NULL, 0, 0 }, /* Qq */
280: { NULL, 0, 0 }, /* Re */
281: { NULL, 0, 0 }, /* Rs */
282: { NULL, 0, 0 }, /* Sc */
283: { NULL, 0, 0 }, /* So */
284: { NULL, 0, 0 }, /* Sq */
285: { NULL, 0, 0 }, /* Sm */
286: { NULL, 0, 0 }, /* Sx */
287: { NULL, TYPE_Sy, 0 }, /* Sy */
288: { NULL, TYPE_Tn, 0 }, /* Tn */
289: { NULL, 0, NODE_NOSRC }, /* Ux */
290: { NULL, 0, 0 }, /* Xc */
291: { NULL, 0, 0 }, /* Xo */
292: { parse_mdoc_Fo, 0, 0 }, /* Fo */
293: { NULL, 0, 0 }, /* Fc */
294: { NULL, 0, 0 }, /* Oo */
295: { NULL, 0, 0 }, /* Oc */
296: { NULL, 0, 0 }, /* Bk */
297: { NULL, 0, 0 }, /* Ek */
298: { NULL, 0, 0 }, /* Bt */
299: { NULL, 0, 0 }, /* Hf */
300: { NULL, 0, 0 }, /* Fr */
301: { NULL, 0, 0 }, /* Ud */
1.237 schwarze 302: { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */
1.235 schwarze 303: { NULL, 0, 0 }, /* Lp */
304: { NULL, TYPE_Lk, 0 }, /* Lk */
305: { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
306: { NULL, 0, 0 }, /* Brq */
307: { NULL, 0, 0 }, /* Bro */
308: { NULL, 0, 0 }, /* Brc */
309: { NULL, 0, 0 }, /* %C */
310: { NULL, 0, 0 }, /* Es */
311: { NULL, 0, 0 }, /* En */
312: { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
313: { NULL, 0, 0 }, /* %Q */
314: { NULL, 0, 0 }, /* %U */
315: { NULL, 0, 0 }, /* Ta */
1.1 kristaps 316: };
317:
1.141 schwarze 318:
1.1 kristaps 319: int
1.179 schwarze 320: mandocdb(int argc, char *argv[])
1.1 kristaps 321: {
1.187 schwarze 322: struct manconf conf;
323: struct mparse *mp;
1.221 schwarze 324: struct dba *dba;
1.206 schwarze 325: const char *path_arg, *progname;
1.187 schwarze 326: size_t j, sz;
1.59 schwarze 327: int ch, i;
1.50 kristaps 328:
1.207 schwarze 329: #if HAVE_PLEDGE
1.255 schwarze 330: if (pledge("stdio rpath wpath cpath", NULL) == -1) {
1.214 schwarze 331: warn("pledge");
1.218 kristaps 332: return (int)MANDOCLEVEL_SYSERR;
333: }
334: #endif
335:
336: #if HAVE_SANDBOX_INIT
337: if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) {
338: warnx("sandbox_init");
1.207 schwarze 339: return (int)MANDOCLEVEL_SYSERR;
340: }
341: #endif
342:
1.187 schwarze 343: memset(&conf, 0, sizeof(conf));
1.50 kristaps 344:
345: /*
1.141 schwarze 346: * We accept a few different invocations.
1.50 kristaps 347: * The CHECKOP macro makes sure that invocation styles don't
348: * clobber each other.
349: */
350: #define CHECKOP(_op, _ch) do \
1.264 schwarze 351: if ((_op) != OP_DEFAULT) { \
1.198 schwarze 352: warnx("-%c: Conflicting option", (_ch)); \
1.50 kristaps 353: goto usage; \
354: } while (/*CONSTCOND*/0)
1.10 kristaps 355:
1.262 schwarze 356: mparse_options = MPARSE_VALIDATE;
1.59 schwarze 357: path_arg = NULL;
1.38 schwarze 358: op = OP_DEFAULT;
1.1 kristaps 359:
1.264 schwarze 360: while ((ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")) != -1)
1.1 kristaps 361: switch (ch) {
1.141 schwarze 362: case 'a':
1.12 schwarze 363: use_all = 1;
364: break;
1.141 schwarze 365: case 'C':
1.50 kristaps 366: CHECKOP(op, ch);
1.59 schwarze 367: path_arg = optarg;
1.38 schwarze 368: op = OP_CONFFILE;
1.34 schwarze 369: break;
1.141 schwarze 370: case 'D':
1.125 schwarze 371: debug++;
372: break;
1.141 schwarze 373: case 'd':
1.50 kristaps 374: CHECKOP(op, ch);
1.59 schwarze 375: path_arg = optarg;
1.5 kristaps 376: op = OP_UPDATE;
377: break;
1.141 schwarze 378: case 'n':
1.50 kristaps 379: nodb = 1;
380: break;
1.141 schwarze 381: case 'p':
1.126 schwarze 382: warnings = 1;
383: break;
1.141 schwarze 384: case 'Q':
1.116 schwarze 385: mparse_options |= MPARSE_QUICK;
1.102 schwarze 386: break;
1.141 schwarze 387: case 'T':
1.264 schwarze 388: if (strcmp(optarg, "utf8") != 0) {
1.198 schwarze 389: warnx("-T%s: Unsupported output format",
390: optarg);
1.95 schwarze 391: goto usage;
392: }
393: write_utf8 = 1;
394: break;
1.141 schwarze 395: case 't':
1.50 kristaps 396: CHECKOP(op, ch);
1.38 schwarze 397: dup2(STDOUT_FILENO, STDERR_FILENO);
398: op = OP_TEST;
1.50 kristaps 399: nodb = warnings = 1;
1.38 schwarze 400: break;
1.141 schwarze 401: case 'u':
1.50 kristaps 402: CHECKOP(op, ch);
1.59 schwarze 403: path_arg = optarg;
1.5 kristaps 404: op = OP_DELETE;
405: break;
1.141 schwarze 406: case 'v':
1.126 schwarze 407: /* Compatibility with espie@'s makewhatis. */
1.38 schwarze 408: break;
1.1 kristaps 409: default:
1.38 schwarze 410: goto usage;
1.1 kristaps 411: }
412:
413: argc -= optind;
414: argv += optind;
415:
1.207 schwarze 416: #if HAVE_PLEDGE
1.210 schwarze 417: if (nodb) {
418: if (pledge("stdio rpath", NULL) == -1) {
1.214 schwarze 419: warn("pledge");
1.210 schwarze 420: return (int)MANDOCLEVEL_SYSERR;
421: }
1.207 schwarze 422: }
423: #endif
424:
1.264 schwarze 425: if (op == OP_CONFFILE && argc > 0) {
1.198 schwarze 426: warnx("-C: Too many arguments");
1.38 schwarze 427: goto usage;
428: }
429:
1.59 schwarze 430: exitcode = (int)MANDOCLEVEL_OK;
1.203 schwarze 431: mchars_alloc();
1.261 schwarze 432: mp = mparse_alloc(mparse_options, MANDOC_OS_OTHER, NULL);
1.202 schwarze 433: mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
434: mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
1.50 kristaps 435:
1.264 schwarze 436: if (op == OP_UPDATE || op == OP_DELETE || op == OP_TEST) {
1.59 schwarze 437:
1.50 kristaps 438: /*
1.150 schwarze 439: * Most of these deal with a specific directory.
1.140 schwarze 440: * Jump into that directory first.
1.50 kristaps 441: */
1.264 schwarze 442: if (op != OP_TEST && set_basedir(path_arg, 1) == 0)
1.50 kristaps 443: goto out;
1.140 schwarze 444:
1.225 schwarze 445: dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
446: if (dba != NULL) {
1.140 schwarze 447: /*
448: * The existing database is usable. Process
449: * all files specified on the command-line.
450: */
451: use_all = 1;
1.138 schwarze 452: for (i = 0; i < argc; i++)
453: filescan(argv[i]);
1.225 schwarze 454: if (nodb == 0)
1.221 schwarze 455: dbprune(dba);
1.138 schwarze 456: } else {
1.228 schwarze 457: /* Database missing or corrupt. */
1.230 schwarze 458: if (op != OP_UPDATE || errno != ENOENT)
459: say(MANDOC_DB, "%s: Automatically recreating"
460: " from scratch", strerror(errno));
1.140 schwarze 461: exitcode = (int)MANDOCLEVEL_OK;
1.138 schwarze 462: op = OP_DEFAULT;
1.264 schwarze 463: if (treescan() == 0)
1.138 schwarze 464: goto out;
1.221 schwarze 465: dba = dba_new(128);
1.138 schwarze 466: }
1.264 schwarze 467: if (op != OP_DELETE)
1.221 schwarze 468: mpages_merge(dba, mp);
469: if (nodb == 0)
470: dbwrite(dba);
471: dba_free(dba);
1.50 kristaps 472: } else {
473: /*
474: * If we have arguments, use them as our manpaths.
1.222 schwarze 475: * If we don't, use man.conf(5).
1.50 kristaps 476: */
477: if (argc > 0) {
1.187 schwarze 478: conf.manpath.paths = mandoc_reallocarray(NULL,
1.144 schwarze 479: argc, sizeof(char *));
1.187 schwarze 480: conf.manpath.sz = (size_t)argc;
1.50 kristaps 481: for (i = 0; i < argc; i++)
1.187 schwarze 482: conf.manpath.paths[i] = mandoc_strdup(argv[i]);
1.50 kristaps 483: } else
1.187 schwarze 484: manconf_parse(&conf, path_arg, NULL, NULL);
1.127 schwarze 485:
1.187 schwarze 486: if (conf.manpath.sz == 0) {
1.127 schwarze 487: exitcode = (int)MANDOCLEVEL_BADARG;
488: say("", "Empty manpath");
489: }
1.50 kristaps 490:
491: /*
1.69 schwarze 492: * First scan the tree rooted at a base directory, then
493: * build a new database and finally move it into place.
1.50 kristaps 494: * Ignore zero-length directories and strip trailing
495: * slashes.
496: */
1.187 schwarze 497: for (j = 0; j < conf.manpath.sz; j++) {
498: sz = strlen(conf.manpath.paths[j]);
499: if (sz && conf.manpath.paths[j][sz - 1] == '/')
500: conf.manpath.paths[j][--sz] = '\0';
1.264 schwarze 501: if (sz == 0)
1.50 kristaps 502: continue;
1.66 schwarze 503:
504: if (j) {
1.202 schwarze 505: mandoc_ohash_init(&mpages, 6,
506: offsetof(struct mpage, inodev));
507: mandoc_ohash_init(&mlinks, 6,
508: offsetof(struct mlink, file));
1.66 schwarze 509: }
510:
1.264 schwarze 511: if (set_basedir(conf.manpath.paths[j], argc > 0) == 0)
1.165 schwarze 512: continue;
1.264 schwarze 513: if (treescan() == 0)
1.165 schwarze 514: continue;
1.221 schwarze 515: dba = dba_new(128);
516: mpages_merge(dba, mp);
517: if (nodb == 0)
518: dbwrite(dba);
519: dba_free(dba);
1.66 schwarze 520:
1.187 schwarze 521: if (j + 1 < conf.manpath.sz) {
1.80 schwarze 522: mpages_free();
1.78 schwarze 523: ohash_delete(&mpages);
1.83 schwarze 524: ohash_delete(&mlinks);
1.66 schwarze 525: }
1.50 kristaps 526: }
527: }
528: out:
1.187 schwarze 529: manconf_free(&conf);
1.168 schwarze 530: mparse_free(mp);
1.203 schwarze 531: mchars_free();
1.80 schwarze 532: mpages_free();
1.78 schwarze 533: ohash_delete(&mpages);
1.83 schwarze 534: ohash_delete(&mlinks);
1.197 schwarze 535: return exitcode;
1.50 kristaps 536: usage:
1.206 schwarze 537: progname = getprogname();
1.126 schwarze 538: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
539: " %s [-aDnpQ] [-Tutf8] dir ...\n"
540: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
541: " %s [-Dnp] -u dir [file ...]\n"
1.102 schwarze 542: " %s [-Q] -t file ...\n",
1.206 schwarze 543: progname, progname, progname, progname, progname);
1.50 kristaps 544:
1.197 schwarze 545: return (int)MANDOCLEVEL_BADARG;
1.50 kristaps 546: }
547:
548: /*
1.231 schwarze 549: * To get a singly linked list in alpha order while inserting entries
550: * at the beginning, process directory entries in reverse alpha order.
551: */
552: static int
1.232 schwarze 553: #if HAVE_FTS_COMPARE_CONST
554: fts_compare(const FTSENT *const *a, const FTSENT *const *b)
555: #else
1.231 schwarze 556: fts_compare(const FTSENT **a, const FTSENT **b)
1.232 schwarze 557: #endif
1.231 schwarze 558: {
559: return -strcmp((*a)->fts_name, (*b)->fts_name);
560: }
561:
562: /*
1.59 schwarze 563: * Scan a directory tree rooted at "basedir" for manpages.
1.50 kristaps 564: * We use fts(), scanning directory parts along the way for clues to our
565: * section and architecture.
566: *
567: * If use_all has been specified, grok all files.
568: * If not, sanitise paths to the following:
569: *
1.141 schwarze 570: * [./]man*[/<arch>]/<name>.<section>
1.50 kristaps 571: * or
572: * [./]cat<section>[/<arch>]/<name>.0
573: *
1.216 schwarze 574: * TODO: accommodate for multi-language directories.
1.50 kristaps 575: */
576: static int
1.59 schwarze 577: treescan(void)
1.50 kristaps 578: {
1.139 schwarze 579: char buf[PATH_MAX];
1.50 kristaps 580: FTS *f;
581: FTSENT *ff;
1.84 schwarze 582: struct mlink *mlink;
1.221 schwarze 583: int gzip;
584: enum form dform;
1.94 schwarze 585: char *dsec, *arch, *fsec, *cp;
586: const char *path;
1.50 kristaps 587: const char *argv[2];
588:
589: argv[0] = ".";
1.244 schwarze 590: argv[1] = NULL;
1.50 kristaps 591:
1.231 schwarze 592: f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR,
593: fts_compare);
1.200 schwarze 594: if (f == NULL) {
1.59 schwarze 595: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 596: say("", "&fts_open");
1.197 schwarze 597: return 0;
1.50 kristaps 598: }
599:
600: dsec = arch = NULL;
601: dform = FORM_NONE;
602:
1.200 schwarze 603: while ((ff = fts_read(f)) != NULL) {
1.50 kristaps 604: path = ff->fts_path + 2;
1.139 schwarze 605: switch (ff->fts_info) {
606:
607: /*
608: * Symbolic links require various sanity checks,
609: * then get handled just like regular files.
610: */
1.141 schwarze 611: case FTS_SL:
1.200 schwarze 612: if (realpath(path, buf) == NULL) {
1.139 schwarze 613: if (warnings)
614: say(path, "&realpath");
615: continue;
616: }
1.264 schwarze 617: if (strncmp(buf, basedir, basedir_len) != 0
1.268 schwarze 618: #ifdef READ_ALLOWED_PATH
619: && !read_allowed(buf)
1.184 schwarze 620: #endif
621: ) {
1.139 schwarze 622: if (warnings) say("",
623: "%s: outside base directory", buf);
624: continue;
625: }
626: /* Use logical inode to avoid mpages dupe. */
1.200 schwarze 627: if (stat(path, ff->fts_statp) == -1) {
1.139 schwarze 628: if (warnings)
629: say(path, "&stat");
630: continue;
631: }
1.269 ! schwarze 632: if ((ff->fts_statp->st_mode & S_IFMT) != S_IFREG)
! 633: continue;
1.139 schwarze 634: /* FALLTHROUGH */
635:
1.50 kristaps 636: /*
1.83 schwarze 637: * If we're a regular file, add an mlink by using the
1.50 kristaps 638: * stored directory data and handling the filename.
639: */
1.141 schwarze 640: case FTS_F:
1.200 schwarze 641: if ( ! strcmp(path, MANDOC_DB))
1.60 schwarze 642: continue;
1.50 kristaps 643: if ( ! use_all && ff->fts_level < 2) {
1.56 schwarze 644: if (warnings)
1.59 schwarze 645: say(path, "Extraneous file");
1.50 kristaps 646: continue;
1.124 schwarze 647: }
648: gzip = 0;
649: fsec = NULL;
1.200 schwarze 650: while (fsec == NULL) {
1.124 schwarze 651: fsec = strrchr(ff->fts_name, '.');
1.200 schwarze 652: if (fsec == NULL || strcmp(fsec+1, "gz"))
1.124 schwarze 653: break;
654: gzip = 1;
655: *fsec = '\0';
656: fsec = NULL;
657: }
1.200 schwarze 658: if (fsec == NULL) {
1.60 schwarze 659: if ( ! use_all) {
1.56 schwarze 660: if (warnings)
1.60 schwarze 661: say(path,
662: "No filename suffix");
1.50 kristaps 663: continue;
664: }
1.200 schwarze 665: } else if ( ! strcmp(++fsec, "html")) {
1.60 schwarze 666: if (warnings)
667: say(path, "Skip html");
668: continue;
1.200 schwarze 669: } else if ( ! strcmp(fsec, "ps")) {
1.60 schwarze 670: if (warnings)
671: say(path, "Skip ps");
672: continue;
1.200 schwarze 673: } else if ( ! strcmp(fsec, "pdf")) {
1.60 schwarze 674: if (warnings)
675: say(path, "Skip pdf");
676: continue;
677: } else if ( ! use_all &&
1.200 schwarze 678: ((dform == FORM_SRC &&
1.185 schwarze 679: strncmp(fsec, dsec, strlen(dsec))) ||
1.200 schwarze 680: (dform == FORM_CAT && strcmp(fsec, "0")))) {
1.60 schwarze 681: if (warnings)
682: say(path, "Wrong filename suffix");
683: continue;
1.66 schwarze 684: } else
1.84 schwarze 685: fsec[-1] = '\0';
1.94 schwarze 686:
1.84 schwarze 687: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.143 schwarze 688: if (strlcpy(mlink->file, path,
689: sizeof(mlink->file)) >=
690: sizeof(mlink->file)) {
691: say(path, "Filename too long");
692: free(mlink);
693: continue;
694: }
1.84 schwarze 695: mlink->dform = dform;
1.94 schwarze 696: mlink->dsec = dsec;
697: mlink->arch = arch;
698: mlink->name = ff->fts_name;
699: mlink->fsec = fsec;
1.124 schwarze 700: mlink->gzip = gzip;
1.84 schwarze 701: mlink_add(mlink, ff->fts_statp);
1.50 kristaps 702: continue;
1.139 schwarze 703:
1.141 schwarze 704: case FTS_D:
705: case FTS_DP:
1.139 schwarze 706: break;
707:
708: default:
1.60 schwarze 709: if (warnings)
710: say(path, "Not a regular file");
1.50 kristaps 711: continue;
1.60 schwarze 712: }
1.1 kristaps 713:
1.50 kristaps 714: switch (ff->fts_level) {
1.141 schwarze 715: case 0:
1.50 kristaps 716: /* Ignore the root directory. */
717: break;
1.141 schwarze 718: case 1:
1.50 kristaps 719: /*
720: * This might contain manX/ or catX/.
721: * Try to infer this from the name.
722: * If we're not in use_all, enforce it.
723: */
724: cp = ff->fts_name;
1.200 schwarze 725: if (ff->fts_info == FTS_DP) {
726: dform = FORM_NONE;
727: dsec = NULL;
1.50 kristaps 728: break;
1.200 schwarze 729: }
1.1 kristaps 730:
1.200 schwarze 731: if ( ! strncmp(cp, "man", 3)) {
1.50 kristaps 732: dform = FORM_SRC;
1.66 schwarze 733: dsec = cp + 3;
1.200 schwarze 734: } else if ( ! strncmp(cp, "cat", 3)) {
1.50 kristaps 735: dform = FORM_CAT;
1.66 schwarze 736: dsec = cp + 3;
1.94 schwarze 737: } else {
738: dform = FORM_NONE;
739: dsec = NULL;
1.50 kristaps 740: }
1.1 kristaps 741:
1.200 schwarze 742: if (dsec != NULL || use_all)
1.50 kristaps 743: break;
1.1 kristaps 744:
1.56 schwarze 745: if (warnings)
1.59 schwarze 746: say(path, "Unknown directory part");
1.50 kristaps 747: fts_set(f, ff, FTS_SKIP);
748: break;
1.141 schwarze 749: case 2:
1.50 kristaps 750: /*
751: * Possibly our architecture.
752: * If we're descending, keep tabs on it.
753: */
1.200 schwarze 754: if (ff->fts_info != FTS_DP && dsec != NULL)
1.66 schwarze 755: arch = ff->fts_name;
1.94 schwarze 756: else
757: arch = NULL;
1.50 kristaps 758: break;
759: default:
1.200 schwarze 760: if (ff->fts_info == FTS_DP || use_all)
1.50 kristaps 761: break;
1.56 schwarze 762: if (warnings)
1.59 schwarze 763: say(path, "Extraneous directory part");
1.50 kristaps 764: fts_set(f, ff, FTS_SKIP);
765: break;
1.5 kristaps 766: }
1.50 kristaps 767: }
768:
769: fts_close(f);
1.197 schwarze 770: return 1;
1.50 kristaps 771: }
1.5 kristaps 772:
1.50 kristaps 773: /*
1.89 schwarze 774: * Add a file to the mlinks table.
1.50 kristaps 775: * Do not verify that it's a "valid" looking manpage (we'll do that
776: * later).
777: *
778: * Try to infer the manual section, architecture, and page name from the
779: * path, assuming it looks like
780: *
1.141 schwarze 781: * [./]man*[/<arch>]/<name>.<section>
1.50 kristaps 782: * or
783: * [./]cat<section>[/<arch>]/<name>.0
784: *
785: * See treescan() for the fts(3) version of this.
786: */
787: static void
1.266 schwarze 788: filescan(const char *infile)
1.50 kristaps 789: {
1.84 schwarze 790: struct stat st;
791: struct mlink *mlink;
1.266 schwarze 792: char *linkfile, *p, *realdir, *start, *usefile;
793: size_t realdir_len;
1.5 kristaps 794:
1.50 kristaps 795: assert(use_all);
1.5 kristaps 796:
1.266 schwarze 797: if (strncmp(infile, "./", 2) == 0)
798: infile += 2;
1.5 kristaps 799:
1.139 schwarze 800: /*
801: * We have to do lstat(2) before realpath(3) loses
802: * the information whether this is a symbolic link.
803: * We need to know that because for symbolic links,
804: * we want to use the orginal file name, while for
805: * regular files, we want to use the real path.
806: */
1.266 schwarze 807: if (lstat(infile, &st) == -1) {
1.139 schwarze 808: exitcode = (int)MANDOCLEVEL_BADARG;
1.266 schwarze 809: say(infile, "&lstat");
1.139 schwarze 810: return;
1.265 schwarze 811: } else if (S_ISREG(st.st_mode) == 0 && S_ISLNK(st.st_mode) == 0) {
1.139 schwarze 812: exitcode = (int)MANDOCLEVEL_BADARG;
1.266 schwarze 813: say(infile, "Not a regular file");
1.139 schwarze 814: return;
815: }
816:
817: /*
818: * We have to resolve the file name to the real path
819: * in any case for the base directory check.
820: */
1.266 schwarze 821: if ((usefile = realpath(infile, NULL)) == NULL) {
1.59 schwarze 822: exitcode = (int)MANDOCLEVEL_BADARG;
1.266 schwarze 823: say(infile, "&realpath");
1.59 schwarze 824: return;
1.106 schwarze 825: }
826:
1.264 schwarze 827: if (op == OP_TEST)
1.266 schwarze 828: start = usefile;
829: else if (strncmp(usefile, basedir, basedir_len) == 0)
830: start = usefile + basedir_len;
1.268 schwarze 831: #ifdef READ_ALLOWED_PATH
832: else if (read_allowed(usefile))
1.266 schwarze 833: start = usefile;
1.184 schwarze 834: #endif
1.106 schwarze 835: else {
1.59 schwarze 836: exitcode = (int)MANDOCLEVEL_BADARG;
1.266 schwarze 837: say("", "%s: outside base directory", infile);
838: free(usefile);
1.59 schwarze 839: return;
1.106 schwarze 840: }
841:
1.139 schwarze 842: /*
843: * Now we are sure the file is inside our tree.
844: * If it is a symbolic link, ignore the real path
845: * and use the original name.
846: */
1.266 schwarze 847: do {
848: if (S_ISLNK(st.st_mode) == 0)
849: break;
850:
851: /*
852: * Some implementations of realpath(3) may succeed
853: * even if the target of the link does not exist,
854: * so check again for extra safety.
855: */
856: if (stat(usefile, &st) == -1) {
1.139 schwarze 857: exitcode = (int)MANDOCLEVEL_BADARG;
1.266 schwarze 858: say(infile, "&stat");
859: free(usefile);
1.139 schwarze 860: return;
861: }
1.266 schwarze 862: linkfile = mandoc_strdup(infile);
863: if (op == OP_TEST) {
864: free(usefile);
865: start = usefile = linkfile;
866: break;
867: }
868: if (strncmp(infile, basedir, basedir_len) == 0) {
869: free(usefile);
870: usefile = linkfile;
871: start = usefile + basedir_len;
872: break;
873: }
874:
875: /*
876: * This symbolic link points into the basedir
877: * from the outside. Let's see whether any of
878: * the parent directories resolve to the basedir.
879: */
880: p = strchr(linkfile, '\0');
881: do {
882: while (*--p != '/')
883: continue;
884: *p = '\0';
885: if ((realdir = realpath(linkfile, NULL)) == NULL) {
886: exitcode = (int)MANDOCLEVEL_BADARG;
887: say(infile, "&realpath");
888: free(linkfile);
889: free(usefile);
890: return;
891: }
892: realdir_len = strlen(realdir) + 1;
893: free(realdir);
894: *p = '/';
895: } while (realdir_len > basedir_len);
896:
897: /*
898: * If one of the directories resolves to the basedir,
899: * use the rest of the original name.
900: * Otherwise, the best we can do
901: * is to use the filename pointed to.
902: */
903: if (realdir_len == basedir_len) {
904: free(usefile);
905: usefile = linkfile;
906: start = p + 1;
907: } else {
908: free(linkfile);
909: start = usefile + basedir_len;
1.143 schwarze 910: }
1.266 schwarze 911: } while (/* CONSTCOND */ 0);
1.106 schwarze 912:
1.84 schwarze 913: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.161 schwarze 914: mlink->dform = FORM_NONE;
1.143 schwarze 915: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
916: sizeof(mlink->file)) {
917: say(start, "Filename too long");
1.180 schwarze 918: free(mlink);
1.266 schwarze 919: free(usefile);
1.143 schwarze 920: return;
1.239 schwarze 921: }
922:
923: /*
924: * In test mode or when the original name is absolute
925: * but outside our tree, guess the base directory.
926: */
927:
1.266 schwarze 928: if (op == OP_TEST || (start == usefile && *start == '/')) {
929: if (strncmp(usefile, "man/", 4) == 0)
930: start = usefile + 4;
931: else if ((start = strstr(usefile, "/man/")) != NULL)
1.239 schwarze 932: start += 5;
933: else
1.266 schwarze 934: start = usefile;
1.143 schwarze 935: }
1.17 schwarze 936:
1.50 kristaps 937: /*
938: * First try to guess our directory structure.
939: * If we find a separator, try to look for man* or cat*.
940: * If we find one of these and what's underneath is a directory,
941: * assume it's an architecture.
942: */
1.264 schwarze 943: if ((p = strchr(start, '/')) != NULL) {
1.50 kristaps 944: *p++ = '\0';
1.264 schwarze 945: if (strncmp(start, "man", 3) == 0) {
1.84 schwarze 946: mlink->dform = FORM_SRC;
1.94 schwarze 947: mlink->dsec = start + 3;
1.264 schwarze 948: } else if (strncmp(start, "cat", 3) == 0) {
1.84 schwarze 949: mlink->dform = FORM_CAT;
1.94 schwarze 950: mlink->dsec = start + 3;
1.17 schwarze 951: }
1.5 kristaps 952:
1.50 kristaps 953: start = p;
1.264 schwarze 954: if (mlink->dsec != NULL && (p = strchr(start, '/')) != NULL) {
1.50 kristaps 955: *p++ = '\0';
1.94 schwarze 956: mlink->arch = start;
1.50 kristaps 957: start = p;
1.84 schwarze 958: }
1.50 kristaps 959: }
960:
961: /*
962: * Now check the file suffix.
963: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
964: */
965: p = strrchr(start, '\0');
1.264 schwarze 966: while (p-- > start && *p != '/' && *p != '.')
967: continue;
1.50 kristaps 968:
1.264 schwarze 969: if (*p == '.') {
1.50 kristaps 970: *p++ = '\0';
1.94 schwarze 971: mlink->fsec = p;
1.5 kristaps 972: }
973:
1.10 kristaps 974: /*
1.50 kristaps 975: * Now try to parse the name.
976: * Use the filename portion of the path.
1.10 kristaps 977: */
1.84 schwarze 978: mlink->name = start;
1.264 schwarze 979: if ((p = strrchr(start, '/')) != NULL) {
1.84 schwarze 980: mlink->name = p + 1;
1.50 kristaps 981: *p = '\0';
1.84 schwarze 982: }
983: mlink_add(mlink, &st);
1.266 schwarze 984: free(usefile);
1.50 kristaps 985: }
1.5 kristaps 986:
1.50 kristaps 987: static void
1.84 schwarze 988: mlink_add(struct mlink *mlink, const struct stat *st)
1.50 kristaps 989: {
1.83 schwarze 990: struct inodev inodev;
1.78 schwarze 991: struct mpage *mpage;
1.83 schwarze 992: unsigned int slot;
1.50 kristaps 993:
1.84 schwarze 994: assert(NULL != mlink->file);
1.50 kristaps 995:
1.94 schwarze 996: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
997: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
998: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
999: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.84 schwarze 1000:
1001: if ('0' == *mlink->fsec) {
1002: free(mlink->fsec);
1003: mlink->fsec = mandoc_strdup(mlink->dsec);
1004: mlink->fform = FORM_CAT;
1005: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
1006: mlink->fform = FORM_SRC;
1.74 schwarze 1007: else
1.84 schwarze 1008: mlink->fform = FORM_NONE;
1.82 schwarze 1009:
1.83 schwarze 1010: slot = ohash_qlookup(&mlinks, mlink->file);
1011: assert(NULL == ohash_find(&mlinks, slot));
1012: ohash_insert(&mlinks, slot, mlink);
1013:
1.186 schwarze 1014: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.83 schwarze 1015: inodev.st_ino = st->st_ino;
1016: inodev.st_dev = st->st_dev;
1017: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
1018: sizeof(struct inodev), inodev.st_ino);
1019: mpage = ohash_find(&mpages, slot);
1020: if (NULL == mpage) {
1021: mpage = mandoc_calloc(1, sizeof(struct mpage));
1022: mpage->inodev.st_ino = inodev.st_ino;
1023: mpage->inodev.st_dev = inodev.st_dev;
1.221 schwarze 1024: mpage->form = FORM_NONE;
1.231 schwarze 1025: mpage->next = mpage_head;
1026: mpage_head = mpage;
1.83 schwarze 1027: ohash_insert(&mpages, slot, mpage);
1028: } else
1.85 schwarze 1029: mlink->next = mpage->mlinks;
1.82 schwarze 1030: mpage->mlinks = mlink;
1.118 schwarze 1031: mlink->mpage = mpage;
1.82 schwarze 1032: }
1033:
1034: static void
1035: mlink_free(struct mlink *mlink)
1036: {
1037:
1038: free(mlink->dsec);
1039: free(mlink->arch);
1040: free(mlink->name);
1041: free(mlink->fsec);
1042: free(mlink);
1.50 kristaps 1043: }
1.3 kristaps 1044:
1.50 kristaps 1045: static void
1.78 schwarze 1046: mpages_free(void)
1.50 kristaps 1047: {
1.78 schwarze 1048: struct mpage *mpage;
1.82 schwarze 1049: struct mlink *mlink;
1.3 kristaps 1050:
1.231 schwarze 1051: while ((mpage = mpage_head) != NULL) {
1052: while ((mlink = mpage->mlinks) != NULL) {
1.85 schwarze 1053: mpage->mlinks = mlink->next;
1.82 schwarze 1054: mlink_free(mlink);
1055: }
1.231 schwarze 1056: mpage_head = mpage->next;
1.78 schwarze 1057: free(mpage->sec);
1058: free(mpage->arch);
1.82 schwarze 1059: free(mpage->title);
1060: free(mpage->desc);
1.78 schwarze 1061: free(mpage);
1.50 kristaps 1062: }
1063: }
1.38 schwarze 1064:
1.50 kristaps 1065: /*
1.89 schwarze 1066: * For each mlink to the mpage, check whether the path looks like
1067: * it is formatted, and if it does, check whether a source manual
1068: * exists by the same name, ignoring the suffix.
1069: * If both conditions hold, drop the mlink.
1070: */
1071: static void
1072: mlinks_undupe(struct mpage *mpage)
1073: {
1074: char buf[PATH_MAX];
1075: struct mlink **prev;
1076: struct mlink *mlink;
1077: char *bufp;
1078:
1079: mpage->form = FORM_CAT;
1.90 schwarze 1080: prev = &mpage->mlinks;
1081: while (NULL != (mlink = *prev)) {
1.89 schwarze 1082: if (FORM_CAT != mlink->dform) {
1083: mpage->form = FORM_NONE;
1.90 schwarze 1084: goto nextlink;
1.89 schwarze 1085: }
1.143 schwarze 1086: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.89 schwarze 1087: bufp = strstr(buf, "cat");
1088: assert(NULL != bufp);
1089: memcpy(bufp, "man", 3);
1090: if (NULL != (bufp = strrchr(buf, '.')))
1091: *++bufp = '\0';
1.143 schwarze 1092: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.89 schwarze 1093: if (NULL == ohash_find(&mlinks,
1.141 schwarze 1094: ohash_qlookup(&mlinks, buf)))
1.90 schwarze 1095: goto nextlink;
1.89 schwarze 1096: if (warnings)
1097: say(mlink->file, "Man source exists: %s", buf);
1098: if (use_all)
1.90 schwarze 1099: goto nextlink;
1.89 schwarze 1100: *prev = mlink->next;
1101: mlink_free(mlink);
1.90 schwarze 1102: continue;
1103: nextlink:
1104: prev = &(*prev)->next;
1.89 schwarze 1105: }
1106: }
1107:
1.131 schwarze 1108: static void
1.93 schwarze 1109: mlink_check(struct mpage *mpage, struct mlink *mlink)
1110: {
1.131 schwarze 1111: struct str *str;
1112: unsigned int slot;
1.93 schwarze 1113:
1114: /*
1115: * Check whether the manual section given in a file
1116: * agrees with the directory where the file is located.
1117: * Some manuals have suffixes like (3p) on their
1118: * section number either inside the file or in the
1119: * directory name, some are linked into more than one
1120: * section, like encrypt(1) = makekey(8).
1121: */
1122:
1123: if (FORM_SRC == mpage->form &&
1.131 schwarze 1124: strcasecmp(mpage->sec, mlink->dsec))
1.93 schwarze 1125: say(mlink->file, "Section \"%s\" manual in %s directory",
1126: mpage->sec, mlink->dsec);
1127:
1128: /*
1129: * Manual page directories exist for each kernel
1130: * architecture as returned by machine(1).
1131: * However, many manuals only depend on the
1132: * application architecture as returned by arch(1).
1133: * For example, some (2/ARM) manuals are shared
1134: * across the "armish" and "zaurus" kernel
1135: * architectures.
1136: * A few manuals are even shared across completely
1137: * different architectures, for example fdformat(1)
1.229 schwarze 1138: * on amd64, i386, and sparc64.
1.93 schwarze 1139: */
1140:
1.131 schwarze 1141: if (strcasecmp(mpage->arch, mlink->arch))
1.93 schwarze 1142: say(mlink->file, "Architecture \"%s\" manual in "
1143: "\"%s\" directory", mpage->arch, mlink->arch);
1144:
1.131 schwarze 1145: /*
1146: * XXX
1.133 schwarze 1147: * parse_cat() doesn't set NAME_TITLE yet.
1.131 schwarze 1148: */
1149:
1150: if (FORM_CAT == mpage->form)
1151: return;
1152:
1153: /*
1154: * Check whether this mlink
1155: * appears as a name in the NAME section.
1156: */
1.93 schwarze 1157:
1.133 schwarze 1158: slot = ohash_qlookup(&names, mlink->name);
1159: str = ohash_find(&names, slot);
1.131 schwarze 1160: assert(NULL != str);
1.133 schwarze 1161: if ( ! (NAME_TITLE & str->mask))
1.131 schwarze 1162: say(mlink->file, "Name missing in NAME section");
1.93 schwarze 1163: }
1164:
1.89 schwarze 1165: /*
1.80 schwarze 1166: * Run through the files in the global vector "mpages"
1167: * and add them to the database specified in "basedir".
1.50 kristaps 1168: *
1169: * This handles the parsing scheme itself, using the cues of directory
1170: * and filename to determine whether the file is parsable or not.
1171: */
1.59 schwarze 1172: static void
1.221 schwarze 1173: mpages_merge(struct dba *dba, struct mparse *mp)
1.50 kristaps 1174: {
1.231 schwarze 1175: struct mpage *mpage, *mpage_dest;
1.118 schwarze 1176: struct mlink *mlink, *mlink_dest;
1.262 schwarze 1177: struct roff_meta *meta;
1.112 schwarze 1178: char *cp;
1.162 schwarze 1179: int fd;
1.69 schwarze 1180:
1.231 schwarze 1181: for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
1.89 schwarze 1182: mlinks_undupe(mpage);
1.231 schwarze 1183: if ((mlink = mpage->mlinks) == NULL)
1184: continue;
1.50 kristaps 1185:
1.133 schwarze 1186: name_mask = NAME_MASK;
1.202 schwarze 1187: mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1188: mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1.50 kristaps 1189: mparse_reset(mp);
1.262 schwarze 1190: meta = NULL;
1.124 schwarze 1191:
1.213 schwarze 1192: if ((fd = mparse_open(mp, mlink->file)) == -1) {
1.182 schwarze 1193: say(mlink->file, "&open");
1.162 schwarze 1194: goto nextpage;
1.124 schwarze 1195: }
1.14 schwarze 1196:
1197: /*
1.183 schwarze 1198: * Interpret the file as mdoc(7) or man(7) source
1199: * code, unless it is known to be formatted.
1.14 schwarze 1200: */
1.182 schwarze 1201: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.183 schwarze 1202: mparse_readfd(mp, fd, mlink->file);
1.212 schwarze 1203: close(fd);
1.240 schwarze 1204: fd = -1;
1.262 schwarze 1205: meta = mparse_result(mp);
1.76 schwarze 1206: }
1.14 schwarze 1207:
1.262 schwarze 1208: if (meta != NULL && meta->sodest != NULL) {
1.118 schwarze 1209: mlink_dest = ohash_find(&mlinks,
1.262 schwarze 1210: ohash_qlookup(&mlinks, meta->sodest));
1.171 schwarze 1211: if (mlink_dest == NULL) {
1.262 schwarze 1212: mandoc_asprintf(&cp, "%s.gz", meta->sodest);
1.171 schwarze 1213: mlink_dest = ohash_find(&mlinks,
1214: ohash_qlookup(&mlinks, cp));
1215: free(cp);
1216: }
1217: if (mlink_dest != NULL) {
1.118 schwarze 1218:
1219: /* The .so target exists. */
1220:
1221: mpage_dest = mlink_dest->mpage;
1222: while (1) {
1223: mlink->mpage = mpage_dest;
1224:
1225: /*
1226: * If the target was already
1227: * processed, add the links
1228: * to the database now.
1229: * Otherwise, this will
1230: * happen when we come
1231: * to the target.
1232: */
1233:
1.221 schwarze 1234: if (mpage_dest->dba != NULL)
1235: dbadd_mlink(mlink);
1.118 schwarze 1236:
1.171 schwarze 1237: if (mlink->next == NULL)
1.118 schwarze 1238: break;
1239: mlink = mlink->next;
1240: }
1241:
1242: /* Move all links to the target. */
1243:
1244: mlink->next = mlink_dest->next;
1245: mlink_dest->next = mpage->mlinks;
1246: mpage->mlinks = NULL;
1.263 schwarze 1247: goto nextpage;
1.118 schwarze 1248: }
1.263 schwarze 1249: meta->macroset = MACROSET_NONE;
1250: }
1251: if (meta != NULL && meta->macroset == MACROSET_MDOC) {
1.82 schwarze 1252: mpage->form = FORM_SRC;
1.262 schwarze 1253: mpage->sec = meta->msec;
1.155 schwarze 1254: mpage->sec = mandoc_strdup(
1.171 schwarze 1255: mpage->sec == NULL ? "" : mpage->sec);
1.262 schwarze 1256: mpage->arch = meta->arch;
1.84 schwarze 1257: mpage->arch = mandoc_strdup(
1.171 schwarze 1258: mpage->arch == NULL ? "" : mpage->arch);
1.262 schwarze 1259: mpage->title = mandoc_strdup(meta->title);
1260: } else if (meta != NULL && meta->macroset == MACROSET_MAN) {
1261: if (*meta->msec != '\0' || *meta->title != '\0') {
1.240 schwarze 1262: mpage->form = FORM_SRC;
1.262 schwarze 1263: mpage->sec = mandoc_strdup(meta->msec);
1.240 schwarze 1264: mpage->arch = mandoc_strdup(mlink->arch);
1.262 schwarze 1265: mpage->title = mandoc_strdup(meta->title);
1.240 schwarze 1266: } else
1.262 schwarze 1267: meta = NULL;
1.240 schwarze 1268: }
1269:
1270: assert(mpage->desc == NULL);
1.263 schwarze 1271: if (meta == NULL || meta->sodest != NULL) {
1.182 schwarze 1272: mpage->sec = mandoc_strdup(mlink->dsec);
1273: mpage->arch = mandoc_strdup(mlink->arch);
1274: mpage->title = mandoc_strdup(mlink->name);
1.263 schwarze 1275: if (meta == NULL) {
1276: mpage->form = FORM_CAT;
1277: parse_cat(mpage, fd);
1278: } else
1279: mpage->form = FORM_SRC;
1.262 schwarze 1280: } else if (meta->macroset == MACROSET_MDOC)
1281: parse_mdoc(mpage, meta, meta->first);
1.240 schwarze 1282: else
1.262 schwarze 1283: parse_man(mpage, meta, meta->first);
1.242 schwarze 1284: if (mpage->desc == NULL) {
1285: mpage->desc = mandoc_strdup(mlink->name);
1286: if (warnings)
1287: say(mlink->file, "No one-line description, "
1288: "using filename \"%s\"", mlink->name);
1289: }
1.131 schwarze 1290:
1.238 schwarze 1291: for (mlink = mpage->mlinks;
1292: mlink != NULL;
1293: mlink = mlink->next) {
1294: putkey(mpage, mlink->name, NAME_FILE);
1295: if (warnings && !use_all)
1.131 schwarze 1296: mlink_check(mpage, mlink);
1.238 schwarze 1297: }
1.44 kristaps 1298:
1.221 schwarze 1299: dbadd(dba, mpage);
1.124 schwarze 1300:
1301: nextpage:
1.66 schwarze 1302: ohash_delete(&strings);
1.133 schwarze 1303: ohash_delete(&names);
1.50 kristaps 1304: }
1305: }
1.12 schwarze 1306:
1.50 kristaps 1307: static void
1.124 schwarze 1308: parse_cat(struct mpage *mpage, int fd)
1.50 kristaps 1309: {
1310: FILE *stream;
1.241 schwarze 1311: struct mlink *mlink;
1312: char *line, *p, *title, *sec;
1.209 schwarze 1313: size_t linesz, plen, titlesz;
1314: ssize_t len;
1315: int offs;
1.12 schwarze 1316:
1.241 schwarze 1317: mlink = mpage->mlinks;
1318: stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1319: if (stream == NULL) {
1320: if (fd != -1)
1.159 schwarze 1321: close(fd);
1.56 schwarze 1322: if (warnings)
1.241 schwarze 1323: say(mlink->file, "&fopen");
1.50 kristaps 1324: return;
1325: }
1.1 kristaps 1326:
1.209 schwarze 1327: line = NULL;
1328: linesz = 0;
1329:
1.241 schwarze 1330: /* Parse the section number from the header line. */
1331:
1332: while (getline(&line, &linesz, stream) != -1) {
1333: if (*line == '\n')
1334: continue;
1335: if ((sec = strchr(line, '(')) == NULL)
1336: break;
1337: if ((p = strchr(++sec, ')')) == NULL)
1338: break;
1339: free(mpage->sec);
1340: mpage->sec = mandoc_strndup(sec, p - sec);
1341: if (warnings && *mlink->dsec != '\0' &&
1342: strcasecmp(mpage->sec, mlink->dsec))
1343: say(mlink->file,
1344: "Section \"%s\" manual in %s directory",
1345: mpage->sec, mlink->dsec);
1346: break;
1347: }
1348:
1.50 kristaps 1349: /* Skip to first blank line. */
1.1 kristaps 1350:
1.241 schwarze 1351: while (line == NULL || *line != '\n')
1352: if (getline(&line, &linesz, stream) == -1)
1.50 kristaps 1353: break;
1.1 kristaps 1354:
1.50 kristaps 1355: /*
1356: * Assume the first line that is not indented
1357: * is the first section header. Skip to it.
1358: */
1.1 kristaps 1359:
1.209 schwarze 1360: while (getline(&line, &linesz, stream) != -1)
1361: if (*line != '\n' && *line != ' ')
1.50 kristaps 1362: break;
1.141 schwarze 1363:
1.50 kristaps 1364: /*
1365: * Read up until the next section into a buffer.
1366: * Strip the leading and trailing newline from each read line,
1367: * appending a trailing space.
1368: * Ignore empty (whitespace-only) lines.
1369: */
1.1 kristaps 1370:
1.50 kristaps 1371: titlesz = 0;
1372: title = NULL;
1.38 schwarze 1373:
1.209 schwarze 1374: while ((len = getline(&line, &linesz, stream)) != -1) {
1375: if (*line != ' ')
1.50 kristaps 1376: break;
1.209 schwarze 1377: offs = 0;
1378: while (isspace((unsigned char)line[offs]))
1379: offs++;
1380: if (line[offs] == '\0')
1.38 schwarze 1381: continue;
1.209 schwarze 1382: title = mandoc_realloc(title, titlesz + len - offs);
1383: memcpy(title + titlesz, line + offs, len - offs);
1384: titlesz += len - offs;
1.50 kristaps 1385: title[titlesz - 1] = ' ';
1386: }
1.209 schwarze 1387: free(line);
1.38 schwarze 1388:
1.50 kristaps 1389: /*
1390: * If no page content can be found, or the input line
1391: * is already the next section header, or there is no
1392: * trailing newline, reuse the page title as the page
1393: * description.
1394: */
1.44 kristaps 1395:
1.50 kristaps 1396: if (NULL == title || '\0' == *title) {
1.56 schwarze 1397: if (warnings)
1.241 schwarze 1398: say(mlink->file, "Cannot find NAME section");
1.50 kristaps 1399: fclose(stream);
1400: free(title);
1401: return;
1402: }
1.1 kristaps 1403:
1.209 schwarze 1404: title[titlesz - 1] = '\0';
1.33 schwarze 1405:
1.50 kristaps 1406: /*
1407: * Skip to the first dash.
1408: * Use the remaining line as the description (no more than 70
1409: * bytes).
1410: */
1.33 schwarze 1411:
1.50 kristaps 1412: if (NULL != (p = strstr(title, "- "))) {
1413: for (p += 2; ' ' == *p || '\b' == *p; p++)
1414: /* Skip to next word. */ ;
1415: } else {
1.56 schwarze 1416: if (warnings)
1.242 schwarze 1417: say(mlink->file, "No dash in title line, "
1418: "reusing \"%s\" as one-line description", title);
1.50 kristaps 1419: p = title;
1420: }
1.38 schwarze 1421:
1.50 kristaps 1422: plen = strlen(p);
1.1 kristaps 1423:
1.50 kristaps 1424: /* Strip backspace-encoding from line. */
1.1 kristaps 1425:
1.50 kristaps 1426: while (NULL != (line = memchr(p, '\b', plen))) {
1427: len = line - p;
1428: if (0 == len) {
1429: memmove(line, line + 1, plen--);
1430: continue;
1.141 schwarze 1431: }
1.50 kristaps 1432: memmove(line - 1, line + 1, plen - len);
1433: plen -= 2;
1434: }
1.1 kristaps 1435:
1.257 schwarze 1436: /*
1437: * Cut off excessive one-line descriptions.
1438: * Bad pages are not worth better heuristics.
1439: */
1440:
1441: mpage->desc = mandoc_strndup(p, 150);
1.50 kristaps 1442: fclose(stream);
1443: free(title);
1444: }
1.1 kristaps 1445:
1.50 kristaps 1446: /*
1447: * Put a type/word pair into the word database for this particular file.
1448: */
1449: static void
1.112 schwarze 1450: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.50 kristaps 1451: {
1.78 schwarze 1452: putkeys(mpage, value, strlen(value), type);
1.3 kristaps 1453: }
1454:
1455: /*
1.50 kristaps 1456: * Grok all nodes at or below a certain mdoc node into putkey().
1.3 kristaps 1457: */
1458: static void
1.78 schwarze 1459: putmdockey(const struct mpage *mpage,
1.235 schwarze 1460: const struct roff_node *n, uint64_t m, int taboo)
1.3 kristaps 1461: {
1.18 kristaps 1462:
1.50 kristaps 1463: for ( ; NULL != n; n = n->next) {
1.235 schwarze 1464: if (n->flags & taboo)
1465: continue;
1.50 kristaps 1466: if (NULL != n->child)
1.235 schwarze 1467: putmdockey(mpage, n->child, m, taboo);
1.188 schwarze 1468: if (n->type == ROFFT_TEXT)
1.78 schwarze 1469: putkey(mpage, n->string, m);
1.50 kristaps 1470: }
1471: }
1.18 kristaps 1472:
1.61 schwarze 1473: static void
1.190 schwarze 1474: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1475: const struct roff_node *n)
1.50 kristaps 1476: {
1.189 schwarze 1477: const struct roff_node *head, *body;
1.121 schwarze 1478: char *start, *title;
1.50 kristaps 1479: char byte;
1.121 schwarze 1480: size_t sz;
1.18 kristaps 1481:
1.215 schwarze 1482: if (n == NULL)
1.61 schwarze 1483: return;
1.18 kristaps 1484:
1.50 kristaps 1485: /*
1486: * We're only searching for one thing: the first text child in
1487: * the BODY of a NAME section. Since we don't keep track of
1488: * sections in -man, run some hoops to find out whether we're in
1489: * the correct section or not.
1490: */
1.18 kristaps 1491:
1.188 schwarze 1492: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.50 kristaps 1493: body = n;
1.215 schwarze 1494: if ((head = body->parent->head) != NULL &&
1495: (head = head->child) != NULL &&
1496: head->next == NULL &&
1.188 schwarze 1497: head->type == ROFFT_TEXT &&
1.215 schwarze 1498: strcmp(head->string, "NAME") == 0 &&
1499: body->child != NULL) {
1.3 kristaps 1500:
1.50 kristaps 1501: /*
1502: * Suck the entire NAME section into memory.
1503: * Yes, we might run away.
1504: * But too many manuals have big, spread-out
1505: * NAME sections over many lines.
1506: */
1.3 kristaps 1507:
1.121 schwarze 1508: title = NULL;
1.194 schwarze 1509: deroff(&title, body);
1.50 kristaps 1510: if (NULL == title)
1.61 schwarze 1511: return;
1.18 kristaps 1512:
1.141 schwarze 1513: /*
1.50 kristaps 1514: * Go through a special heuristic dance here.
1515: * Conventionally, one or more manual names are
1516: * comma-specified prior to a whitespace, then a
1517: * dash, then a description. Try to puzzle out
1518: * the name parts here.
1519: */
1.18 kristaps 1520:
1.121 schwarze 1521: start = title;
1.50 kristaps 1522: for ( ;; ) {
1523: sz = strcspn(start, " ,");
1524: if ('\0' == start[sz])
1525: break;
1.1 kristaps 1526:
1.50 kristaps 1527: byte = start[sz];
1528: start[sz] = '\0';
1.110 schwarze 1529:
1530: /*
1531: * Assume a stray trailing comma in the
1532: * name list if a name begins with a dash.
1533: */
1534:
1535: if ('-' == start[0] ||
1536: ('\\' == start[0] && '-' == start[1]))
1537: break;
1.1 kristaps 1538:
1.133 schwarze 1539: putkey(mpage, start, NAME_TITLE);
1.174 schwarze 1540: if ( ! (mpage->name_head_done ||
1541: strcasecmp(start, meta->title))) {
1542: putkey(mpage, start, NAME_HEAD);
1543: mpage->name_head_done = 1;
1544: }
1.1 kristaps 1545:
1.50 kristaps 1546: if (' ' == byte) {
1547: start += sz + 1;
1548: break;
1549: }
1.1 kristaps 1550:
1.50 kristaps 1551: assert(',' == byte);
1552: start += sz + 1;
1553: while (' ' == *start)
1554: start++;
1555: }
1.1 kristaps 1556:
1.121 schwarze 1557: if (start == title) {
1.133 schwarze 1558: putkey(mpage, start, NAME_TITLE);
1.174 schwarze 1559: if ( ! (mpage->name_head_done ||
1560: strcasecmp(start, meta->title))) {
1561: putkey(mpage, start, NAME_HEAD);
1562: mpage->name_head_done = 1;
1563: }
1.50 kristaps 1564: free(title);
1.61 schwarze 1565: return;
1.50 kristaps 1566: }
1.1 kristaps 1567:
1.50 kristaps 1568: while (isspace((unsigned char)*start))
1569: start++;
1.1 kristaps 1570:
1.50 kristaps 1571: if (0 == strncmp(start, "-", 1))
1572: start += 1;
1573: else if (0 == strncmp(start, "\\-\\-", 4))
1574: start += 4;
1575: else if (0 == strncmp(start, "\\-", 2))
1576: start += 2;
1577: else if (0 == strncmp(start, "\\(en", 4))
1578: start += 4;
1579: else if (0 == strncmp(start, "\\(em", 4))
1580: start += 4;
1.1 kristaps 1581:
1.50 kristaps 1582: while (' ' == *start)
1583: start++;
1.1 kristaps 1584:
1.257 schwarze 1585: /*
1586: * Cut off excessive one-line descriptions.
1587: * Bad pages are not worth better heuristics.
1588: */
1589:
1590: mpage->desc = mandoc_strndup(start, 150);
1.50 kristaps 1591: free(title);
1.61 schwarze 1592: return;
1.50 kristaps 1593: }
1594: }
1.1 kristaps 1595:
1.77 schwarze 1596: for (n = n->child; n; n = n->next) {
1.78 schwarze 1597: if (NULL != mpage->desc)
1.77 schwarze 1598: break;
1.174 schwarze 1599: parse_man(mpage, meta, n);
1.77 schwarze 1600: }
1.1 kristaps 1601: }
1602:
1603: static void
1.190 schwarze 1604: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1605: const struct roff_node *n)
1.1 kristaps 1606: {
1.259 schwarze 1607: const struct mdoc_handler *handler;
1.1 kristaps 1608:
1.246 schwarze 1609: for (n = n->child; n != NULL; n = n->next) {
1.259 schwarze 1610: if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX)
1.235 schwarze 1611: continue;
1.247 schwarze 1612: assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1.259 schwarze 1613: handler = mdoc_handlers + (n->tok - MDOC_Dd);
1614: if (n->flags & handler->taboo)
1615: continue;
1616:
1.50 kristaps 1617: switch (n->type) {
1.188 schwarze 1618: case ROFFT_ELEM:
1619: case ROFFT_BLOCK:
1620: case ROFFT_HEAD:
1621: case ROFFT_BODY:
1622: case ROFFT_TAIL:
1.259 schwarze 1623: if (handler->fp != NULL &&
1624: (*handler->fp)(mpage, meta, n) == 0)
1.246 schwarze 1625: break;
1.259 schwarze 1626: if (handler->mask)
1.78 schwarze 1627: putmdockey(mpage, n->child,
1.259 schwarze 1628: handler->mask, handler->taboo);
1.50 kristaps 1629: break;
1630: default:
1631: continue;
1632: }
1633: if (NULL != n->child)
1.172 schwarze 1634: parse_mdoc(mpage, meta, n);
1.1 kristaps 1635: }
1636: }
1637:
1.25 schwarze 1638: static int
1.258 schwarze 1639: parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta,
1640: const struct roff_node *n)
1641: {
1642: uint64_t mask;
1643:
1644: mask = TYPE_Fa;
1645: if (n->sec == SEC_SYNOPSIS)
1646: mask |= TYPE_Vt;
1647:
1648: putmdockey(mpage, n->child, mask, 0);
1649: return 0;
1650: }
1651:
1652: static int
1.190 schwarze 1653: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1654: const struct roff_node *n)
1.1 kristaps 1655: {
1.176 schwarze 1656: char *start, *end;
1.1 kristaps 1657: size_t sz;
1.25 schwarze 1658:
1.50 kristaps 1659: if (SEC_SYNOPSIS != n->sec ||
1.141 schwarze 1660: NULL == (n = n->child) ||
1.188 schwarze 1661: n->type != ROFFT_TEXT)
1.197 schwarze 1662: return 0;
1.1 kristaps 1663:
1664: /*
1665: * Only consider those `Fd' macro fields that begin with an
1666: * "inclusion" token (versus, e.g., #define).
1667: */
1.50 kristaps 1668:
1.1 kristaps 1669: if (strcmp("#include", n->string))
1.197 schwarze 1670: return 0;
1.1 kristaps 1671:
1.188 schwarze 1672: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.197 schwarze 1673: return 0;
1.1 kristaps 1674:
1675: /*
1676: * Strip away the enclosing angle brackets and make sure we're
1677: * not zero-length.
1678: */
1679:
1680: start = n->string;
1681: if ('<' == *start || '"' == *start)
1682: start++;
1683:
1684: if (0 == (sz = strlen(start)))
1.197 schwarze 1685: return 0;
1.1 kristaps 1686:
1687: end = &start[(int)sz - 1];
1688: if ('>' == *end || '"' == *end)
1689: end--;
1690:
1.50 kristaps 1691: if (end > start)
1.78 schwarze 1692: putkeys(mpage, start, end - start + 1, TYPE_In);
1.197 schwarze 1693: return 0;
1.1 kristaps 1694: }
1695:
1.178 schwarze 1696: static void
1.189 schwarze 1697: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 kristaps 1698: {
1.112 schwarze 1699: char *cp;
1.178 schwarze 1700: size_t sz;
1.1 kristaps 1701:
1.188 schwarze 1702: if (n->type != ROFFT_TEXT)
1.178 schwarze 1703: return;
1.25 schwarze 1704:
1.178 schwarze 1705: /* Skip function pointer punctuation. */
1.1 kristaps 1706:
1.178 schwarze 1707: cp = n->string;
1708: while (*cp == '(' || *cp == '*')
1.1 kristaps 1709: cp++;
1.178 schwarze 1710: sz = strcspn(cp, "()");
1.1 kristaps 1711:
1.178 schwarze 1712: putkeys(mpage, cp, sz, TYPE_Fn);
1.173 schwarze 1713: if (n->sec == SEC_SYNOPSIS)
1.178 schwarze 1714: putkeys(mpage, cp, sz, NAME_SYN);
1715: }
1.25 schwarze 1716:
1.178 schwarze 1717: static int
1.190 schwarze 1718: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1719: const struct roff_node *n)
1.178 schwarze 1720: {
1.258 schwarze 1721: uint64_t mask;
1.178 schwarze 1722:
1723: if (n->child == NULL)
1.197 schwarze 1724: return 0;
1.178 schwarze 1725:
1726: parse_mdoc_fname(mpage, n->child);
1.25 schwarze 1727:
1.258 schwarze 1728: n = n->child->next;
1729: if (n != NULL && n->type == ROFFT_TEXT) {
1730: mask = TYPE_Fa;
1731: if (n->sec == SEC_SYNOPSIS)
1732: mask |= TYPE_Vt;
1733: putmdockey(mpage, n, mask, 0);
1734: }
1.25 schwarze 1735:
1.197 schwarze 1736: return 0;
1.173 schwarze 1737: }
1738:
1739: static int
1.190 schwarze 1740: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1741: const struct roff_node *n)
1.173 schwarze 1742: {
1.177 schwarze 1743:
1.188 schwarze 1744: if (n->type != ROFFT_HEAD)
1.197 schwarze 1745: return 1;
1.173 schwarze 1746:
1.178 schwarze 1747: if (n->child != NULL)
1748: parse_mdoc_fname(mpage, n->child);
1749:
1.197 schwarze 1750: return 0;
1.1 kristaps 1751: }
1752:
1.25 schwarze 1753: static int
1.211 schwarze 1754: parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1755: const struct roff_node *n)
1756: {
1757: char *cp;
1758:
1759: if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1760: return 0;
1761:
1.215 schwarze 1762: if (n->child != NULL &&
1763: n->child->next == NULL &&
1764: n->child->type == ROFFT_TEXT)
1.211 schwarze 1765: return 1;
1766:
1767: cp = NULL;
1768: deroff(&cp, n);
1769: if (cp != NULL) {
1770: putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1771: n->type == ROFFT_BODY ? TYPE_Va : 0));
1772: free(cp);
1773: }
1774:
1775: return 0;
1776: }
1777:
1778: static int
1.190 schwarze 1779: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1780: const struct roff_node *n)
1.1 kristaps 1781: {
1.67 schwarze 1782: char *cp;
1.1 kristaps 1783:
1784: if (NULL == (n = n->child))
1.197 schwarze 1785: return 0;
1.1 kristaps 1786:
1.67 schwarze 1787: if (NULL == n->next) {
1.78 schwarze 1788: putkey(mpage, n->string, TYPE_Xr);
1.197 schwarze 1789: return 0;
1.67 schwarze 1790: }
1791:
1.120 schwarze 1792: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.78 schwarze 1793: putkey(mpage, cp, TYPE_Xr);
1.67 schwarze 1794: free(cp);
1.197 schwarze 1795: return 0;
1.1 kristaps 1796: }
1797:
1.25 schwarze 1798: static int
1.190 schwarze 1799: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1800: const struct roff_node *n)
1.1 kristaps 1801: {
1802:
1.188 schwarze 1803: if (n->type == ROFFT_BODY)
1.194 schwarze 1804: deroff(&mpage->desc, n);
1.197 schwarze 1805: return 0;
1.1 kristaps 1806: }
1807:
1.25 schwarze 1808: static int
1.190 schwarze 1809: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1810: const struct roff_node *n)
1.1 kristaps 1811: {
1812:
1.129 schwarze 1813: if (SEC_NAME == n->sec)
1.235 schwarze 1814: putmdockey(mpage, n->child, NAME_TITLE, 0);
1.188 schwarze 1815: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.172 schwarze 1816: if (n->child == NULL)
1817: putkey(mpage, meta->name, NAME_SYN);
1818: else
1.235 schwarze 1819: putmdockey(mpage, n->child, NAME_SYN, 0);
1.174 schwarze 1820: }
1821: if ( ! (mpage->name_head_done ||
1822: n->child == NULL || n->child->string == NULL ||
1823: strcasecmp(n->child->string, meta->title))) {
1.226 schwarze 1824: putkey(mpage, n->child->string, NAME_HEAD);
1.174 schwarze 1825: mpage->name_head_done = 1;
1.172 schwarze 1826: }
1.197 schwarze 1827: return 0;
1.1 kristaps 1828: }
1829:
1.25 schwarze 1830: static int
1.190 schwarze 1831: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1832: const struct roff_node *n)
1.1 kristaps 1833: {
1834:
1.197 schwarze 1835: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 kristaps 1836: }
1837:
1.50 kristaps 1838: static int
1.190 schwarze 1839: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.189 schwarze 1840: const struct roff_node *n)
1.1 kristaps 1841: {
1842:
1.197 schwarze 1843: return n->type == ROFFT_HEAD;
1.1 kristaps 1844: }
1845:
1.50 kristaps 1846: /*
1.66 schwarze 1847: * Add a string to the hash table for the current manual.
1848: * Each string has a bitmask telling which macros it belongs to.
1849: * When we finish the manual, we'll dump the table.
1.50 kristaps 1850: */
1851: static void
1.176 schwarze 1852: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.50 kristaps 1853: {
1.133 schwarze 1854: struct ohash *htab;
1.50 kristaps 1855: struct str *s;
1.111 schwarze 1856: const char *end;
1.65 schwarze 1857: unsigned int slot;
1.176 schwarze 1858: int i, mustfree;
1.25 schwarze 1859:
1.50 kristaps 1860: if (0 == sz)
1861: return;
1.111 schwarze 1862:
1.176 schwarze 1863: mustfree = render_string(&cp, &sz);
1864:
1.133 schwarze 1865: if (TYPE_Nm & v) {
1866: htab = &names;
1867: v &= name_mask;
1.169 schwarze 1868: if (v & NAME_FIRST)
1869: name_mask &= ~NAME_FIRST;
1.133 schwarze 1870: if (debug > 1)
1871: say(mpage->mlinks->file,
1.234 schwarze 1872: "Adding name %*s, bits=0x%llx", (int)sz, cp,
1873: (unsigned long long)v);
1.133 schwarze 1874: } else {
1875: htab = &strings;
1876: if (debug > 1)
1.221 schwarze 1877: for (i = 0; i < KEY_MAX; i++)
1.163 schwarze 1878: if ((uint64_t)1 << i & v)
1.133 schwarze 1879: say(mpage->mlinks->file,
1880: "Adding key %s=%*s",
1.220 schwarze 1881: mansearch_keynames[i], (int)sz, cp);
1.111 schwarze 1882: }
1.25 schwarze 1883:
1.65 schwarze 1884: end = cp + sz;
1.133 schwarze 1885: slot = ohash_qlookupi(htab, cp, &end);
1886: s = ohash_find(htab, slot);
1.25 schwarze 1887:
1.78 schwarze 1888: if (NULL != s && mpage == s->mpage) {
1.50 kristaps 1889: s->mask |= v;
1890: return;
1891: } else if (NULL == s) {
1.144 schwarze 1892: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.50 kristaps 1893: memcpy(s->key, cp, sz);
1.133 schwarze 1894: ohash_insert(htab, slot, s);
1.1 kristaps 1895: }
1.78 schwarze 1896: s->mpage = mpage;
1.50 kristaps 1897: s->mask = v;
1.176 schwarze 1898:
1899: if (mustfree)
1900: free(cp);
1.1 kristaps 1901: }
1902:
1.50 kristaps 1903: /*
1904: * Take a Unicode codepoint and produce its UTF-8 encoding.
1905: * This isn't the best way to do this, but it works.
1906: * The magic numbers are from the UTF-8 packaging.
1907: * They're not as scary as they seem: read the UTF-8 spec for details.
1908: */
1909: static size_t
1910: utf8(unsigned int cp, char out[7])
1.1 kristaps 1911: {
1.50 kristaps 1912: size_t rc;
1.1 kristaps 1913:
1.50 kristaps 1914: rc = 0;
1915: if (cp <= 0x0000007F) {
1916: rc = 1;
1917: out[0] = (char)cp;
1918: } else if (cp <= 0x000007FF) {
1919: rc = 2;
1920: out[0] = (cp >> 6 & 31) | 192;
1921: out[1] = (cp & 63) | 128;
1922: } else if (cp <= 0x0000FFFF) {
1923: rc = 3;
1924: out[0] = (cp >> 12 & 15) | 224;
1925: out[1] = (cp >> 6 & 63) | 128;
1926: out[2] = (cp & 63) | 128;
1927: } else if (cp <= 0x001FFFFF) {
1928: rc = 4;
1929: out[0] = (cp >> 18 & 7) | 240;
1930: out[1] = (cp >> 12 & 63) | 128;
1931: out[2] = (cp >> 6 & 63) | 128;
1932: out[3] = (cp & 63) | 128;
1933: } else if (cp <= 0x03FFFFFF) {
1934: rc = 5;
1935: out[0] = (cp >> 24 & 3) | 248;
1936: out[1] = (cp >> 18 & 63) | 128;
1937: out[2] = (cp >> 12 & 63) | 128;
1938: out[3] = (cp >> 6 & 63) | 128;
1939: out[4] = (cp & 63) | 128;
1940: } else if (cp <= 0x7FFFFFFF) {
1941: rc = 6;
1942: out[0] = (cp >> 30 & 1) | 252;
1943: out[1] = (cp >> 24 & 63) | 128;
1944: out[2] = (cp >> 18 & 63) | 128;
1945: out[3] = (cp >> 12 & 63) | 128;
1946: out[4] = (cp >> 6 & 63) | 128;
1947: out[5] = (cp & 63) | 128;
1948: } else
1.197 schwarze 1949: return 0;
1.1 kristaps 1950:
1.50 kristaps 1951: out[rc] = '\0';
1.197 schwarze 1952: return rc;
1.50 kristaps 1953: }
1.1 kristaps 1954:
1.50 kristaps 1955: /*
1.176 schwarze 1956: * If the string contains escape sequences,
1957: * replace it with an allocated rendering and return 1,
1958: * such that the caller can free it after use.
1959: * Otherwise, do nothing and return 0.
1.50 kristaps 1960: */
1.176 schwarze 1961: static int
1962: render_string(char **public, size_t *psz)
1.50 kristaps 1963: {
1.176 schwarze 1964: const char *src, *scp, *addcp, *seq;
1965: char *dst;
1966: size_t ssz, dsz, addsz;
1.114 schwarze 1967: char utfbuf[7], res[6];
1.176 schwarze 1968: int seqlen, unicode;
1.50 kristaps 1969:
1970: res[0] = '\\';
1971: res[1] = '\t';
1972: res[2] = ASCII_NBRSP;
1973: res[3] = ASCII_HYPH;
1.114 schwarze 1974: res[4] = ASCII_BREAK;
1975: res[5] = '\0';
1.1 kristaps 1976:
1.176 schwarze 1977: src = scp = *public;
1978: ssz = *psz;
1979: dst = NULL;
1980: dsz = 0;
1981:
1982: while (scp < src + *psz) {
1983:
1984: /* Leave normal characters unchanged. */
1985:
1986: if (strchr(res, *scp) == NULL) {
1987: if (dst != NULL)
1988: dst[dsz++] = *scp;
1989: scp++;
1990: continue;
1991: }
1.46 kristaps 1992:
1.50 kristaps 1993: /*
1.176 schwarze 1994: * Found something that requires replacing,
1995: * make sure we have a destination buffer.
1.50 kristaps 1996: */
1.176 schwarze 1997:
1998: if (dst == NULL) {
1999: dst = mandoc_malloc(ssz + 1);
2000: dsz = scp - src;
2001: memcpy(dst, src, dsz);
1.50 kristaps 2002: }
1.46 kristaps 2003:
1.176 schwarze 2004: /* Handle single-char special characters. */
2005:
2006: switch (*scp) {
2007: case '\\':
2008: break;
1.141 schwarze 2009: case '\t':
2010: case ASCII_NBRSP:
1.176 schwarze 2011: dst[dsz++] = ' ';
2012: scp++;
2013: continue;
2014: case ASCII_HYPH:
2015: dst[dsz++] = '-';
1.114 schwarze 2016: /* FALLTHROUGH */
1.141 schwarze 2017: case ASCII_BREAK:
1.176 schwarze 2018: scp++;
1.50 kristaps 2019: continue;
1.114 schwarze 2020: default:
1.176 schwarze 2021: abort();
1.114 schwarze 2022: }
1.46 kristaps 2023:
1.50 kristaps 2024: /*
1.176 schwarze 2025: * Found an escape sequence.
2026: * Read past the slash, then parse it.
2027: * Ignore everything except characters.
1.50 kristaps 2028: */
1.95 schwarze 2029:
1.176 schwarze 2030: scp++;
2031: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.50 kristaps 2032: continue;
1.1 kristaps 2033:
1.50 kristaps 2034: /*
1.95 schwarze 2035: * Render the special character
2036: * as either UTF-8 or ASCII.
1.50 kristaps 2037: */
1.95 schwarze 2038:
2039: if (write_utf8) {
1.203 schwarze 2040: unicode = mchars_spec2cp(seq, seqlen);
1.176 schwarze 2041: if (unicode <= 0)
1.95 schwarze 2042: continue;
1.176 schwarze 2043: addsz = utf8(unicode, utfbuf);
2044: if (addsz == 0)
1.95 schwarze 2045: continue;
1.176 schwarze 2046: addcp = utfbuf;
1.95 schwarze 2047: } else {
1.203 schwarze 2048: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.176 schwarze 2049: if (addcp == NULL)
1.95 schwarze 2050: continue;
1.176 schwarze 2051: if (*addcp == ASCII_NBRSP) {
2052: addcp = " ";
2053: addsz = 1;
1.95 schwarze 2054: }
2055: }
1.1 kristaps 2056:
1.50 kristaps 2057: /* Copy the rendered glyph into the stream. */
1.1 kristaps 2058:
1.176 schwarze 2059: ssz += addsz;
2060: dst = mandoc_realloc(dst, ssz + 1);
2061: memcpy(dst + dsz, addcp, addsz);
2062: dsz += addsz;
2063: }
2064: if (dst != NULL) {
2065: *public = dst;
2066: *psz = dsz;
2067: }
2068:
2069: /* Trim trailing whitespace and NUL-terminate. */
2070:
2071: while (*psz > 0 && (*public)[*psz - 1] == ' ')
2072: --*psz;
2073: if (dst != NULL) {
2074: (*public)[*psz] = '\0';
1.197 schwarze 2075: return 1;
1.176 schwarze 2076: } else
1.197 schwarze 2077: return 0;
1.1 kristaps 2078: }
2079:
1.118 schwarze 2080: static void
2081: dbadd_mlink(const struct mlink *mlink)
2082: {
1.221 schwarze 2083: dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
2084: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
1.223 schwarze 2085: dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
1.221 schwarze 2086: dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
2087: dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
1.118 schwarze 2088: }
2089:
1.14 schwarze 2090: /*
1.50 kristaps 2091: * Flush the current page's terms (and their bits) into the database.
1.96 schwarze 2092: * Also, handle escape sequences at the last possible moment.
1.14 schwarze 2093: */
2094: static void
1.221 schwarze 2095: dbadd(struct dba *dba, struct mpage *mpage)
1.14 schwarze 2096: {
1.87 schwarze 2097: struct mlink *mlink;
1.50 kristaps 2098: struct str *key;
1.176 schwarze 2099: char *cp;
1.221 schwarze 2100: uint64_t mask;
1.52 kristaps 2101: size_t i;
1.66 schwarze 2102: unsigned int slot;
1.176 schwarze 2103: int mustfree;
1.14 schwarze 2104:
1.128 schwarze 2105: mlink = mpage->mlinks;
1.43 kristaps 2106:
1.128 schwarze 2107: if (nodb) {
1.147 schwarze 2108: for (key = ohash_first(&names, &slot); NULL != key;
1.176 schwarze 2109: key = ohash_next(&names, &slot))
1.147 schwarze 2110: free(key);
2111: for (key = ohash_first(&strings, &slot); NULL != key;
1.176 schwarze 2112: key = ohash_next(&strings, &slot))
1.147 schwarze 2113: free(key);
1.145 schwarze 2114: if (0 == debug)
2115: return;
1.128 schwarze 2116: while (NULL != mlink) {
2117: fputs(mlink->name, stdout);
2118: if (NULL == mlink->next ||
2119: strcmp(mlink->dsec, mlink->next->dsec) ||
2120: strcmp(mlink->fsec, mlink->next->fsec) ||
2121: strcmp(mlink->arch, mlink->next->arch)) {
2122: putchar('(');
2123: if ('\0' == *mlink->dsec)
2124: fputs(mlink->fsec, stdout);
2125: else
2126: fputs(mlink->dsec, stdout);
2127: if ('\0' != *mlink->arch)
2128: printf("/%s", mlink->arch);
2129: putchar(')');
2130: }
2131: mlink = mlink->next;
2132: if (NULL != mlink)
2133: fputs(", ", stdout);
2134: }
1.132 schwarze 2135: printf(" - %s\n", mpage->desc);
1.14 schwarze 2136: return;
1.128 schwarze 2137: }
2138:
2139: if (debug)
2140: say(mlink->file, "Adding to database");
1.28 kristaps 2141:
1.176 schwarze 2142: cp = mpage->desc;
2143: i = strlen(cp);
2144: mustfree = render_string(&cp, &i);
1.227 schwarze 2145: mpage->dba = dba_page_new(dba->pages,
1.223 schwarze 2146: *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2147: cp, mlink->file, mpage->form);
1.176 schwarze 2148: if (mustfree)
2149: free(cp);
1.227 schwarze 2150: dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
1.81 schwarze 2151:
1.223 schwarze 2152: while (mlink != NULL) {
1.118 schwarze 2153: dbadd_mlink(mlink);
1.223 schwarze 2154: mlink = mlink->next;
2155: }
1.50 kristaps 2156:
1.133 schwarze 2157: for (key = ohash_first(&names, &slot); NULL != key;
2158: key = ohash_next(&names, &slot)) {
2159: assert(key->mpage == mpage);
1.221 schwarze 2160: dba_page_alias(mpage->dba, key->key, key->mask);
1.133 schwarze 2161: free(key);
2162: }
1.66 schwarze 2163: for (key = ohash_first(&strings, &slot); NULL != key;
2164: key = ohash_next(&strings, &slot)) {
1.78 schwarze 2165: assert(key->mpage == mpage);
1.221 schwarze 2166: i = 0;
2167: for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2168: if (key->mask & mask)
2169: dba_macro_add(dba->macros, i,
2170: key->key, mpage->dba);
2171: i++;
2172: }
1.66 schwarze 2173: free(key);
1.38 schwarze 2174: }
1.14 schwarze 2175: }
2176:
1.5 kristaps 2177: static void
1.221 schwarze 2178: dbprune(struct dba *dba)
1.5 kristaps 2179: {
1.221 schwarze 2180: struct dba_array *page, *files;
2181: char *file;
1.5 kristaps 2182:
1.221 schwarze 2183: dba_array_FOREACH(dba->pages, page) {
2184: files = dba_array_get(page, DBP_FILE);
2185: dba_array_FOREACH(files, file) {
2186: if (*file < ' ')
2187: file++;
2188: if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2189: file)) != NULL) {
2190: if (debug)
2191: say(file, "Deleting from database");
2192: dba_array_del(dba->pages);
2193: break;
2194: }
1.106 schwarze 2195: }
1.5 kristaps 2196: }
2197: }
2198:
1.4 kristaps 2199: /*
1.221 schwarze 2200: * Write the database from memory to disk.
1.4 kristaps 2201: */
1.35 kristaps 2202: static void
1.221 schwarze 2203: dbwrite(struct dba *dba)
1.4 kristaps 2204: {
1.255 schwarze 2205: struct stat sb1, sb2;
2206: char tfn[33], *cp1, *cp2;
2207: off_t i;
2208: int fd1, fd2;
1.4 kristaps 2209:
1.250 schwarze 2210: /*
2211: * Do not write empty databases, and delete existing ones
2212: * when makewhatis -u causes them to become empty.
2213: */
2214:
2215: dba_array_start(dba->pages);
2216: if (dba_array_next(dba->pages) == NULL) {
1.252 schwarze 2217: if (unlink(MANDOC_DB) == -1 && errno != ENOENT)
1.250 schwarze 2218: say(MANDOC_DB, "&unlink");
2219: return;
2220: }
2221:
2222: /*
2223: * Build the database in a temporary file,
2224: * then atomically move it into place.
2225: */
2226:
1.221 schwarze 2227: if (dba_write(MANDOC_DB "~", dba) != -1) {
2228: if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2229: exitcode = (int)MANDOCLEVEL_SYSERR;
2230: say(MANDOC_DB, "&rename");
2231: unlink(MANDOC_DB "~");
2232: }
1.38 schwarze 2233: return;
1.4 kristaps 2234: }
1.250 schwarze 2235:
2236: /*
2237: * We lack write permission and cannot replace the database
2238: * file, but let's at least check whether the data changed.
2239: */
1.4 kristaps 2240:
1.221 schwarze 2241: (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2242: if (mkdtemp(tfn) == NULL) {
2243: exitcode = (int)MANDOCLEVEL_SYSERR;
2244: say("", "&%s", tfn);
1.50 kristaps 2245: return;
1.221 schwarze 2246: }
1.256 schwarze 2247: cp1 = cp2 = MAP_FAILED;
1.255 schwarze 2248: fd1 = fd2 = -1;
1.221 schwarze 2249: (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2250: if (dba_write(tfn, dba) == -1) {
2251: say(tfn, "&dba_write");
1.255 schwarze 2252: goto err;
1.115 schwarze 2253: }
1.255 schwarze 2254: if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) {
2255: say(MANDOC_DB, "&open");
2256: goto err;
2257: }
2258: if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) {
2259: say(tfn, "&open");
2260: goto err;
2261: }
2262: if (fstat(fd1, &sb1) == -1) {
2263: say(MANDOC_DB, "&fstat");
2264: goto err;
2265: }
2266: if (fstat(fd2, &sb2) == -1) {
2267: say(tfn, "&fstat");
2268: goto err;
2269: }
2270: if (sb1.st_size != sb2.st_size)
2271: goto err;
2272: if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE,
1.256 schwarze 2273: fd1, 0)) == MAP_FAILED) {
1.255 schwarze 2274: say(MANDOC_DB, "&mmap");
2275: goto err;
2276: }
2277: if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE,
1.256 schwarze 2278: fd2, 0)) == MAP_FAILED) {
1.255 schwarze 2279: say(tfn, "&mmap");
2280: goto err;
2281: }
2282: for (i = 0; i < sb1.st_size; i++)
2283: if (cp1[i] != cp2[i])
2284: goto err;
2285: goto out;
2286:
2287: err:
2288: exitcode = (int)MANDOCLEVEL_SYSERR;
2289: say(MANDOC_DB, "Data changed, but cannot replace database");
1.115 schwarze 2290:
1.221 schwarze 2291: out:
1.256 schwarze 2292: if (cp1 != MAP_FAILED)
1.255 schwarze 2293: munmap(cp1, sb1.st_size);
1.256 schwarze 2294: if (cp2 != MAP_FAILED)
1.255 schwarze 2295: munmap(cp2, sb2.st_size);
2296: if (fd1 != -1)
2297: close(fd1);
2298: if (fd2 != -1)
2299: close(fd2);
1.254 schwarze 2300: unlink(tfn);
1.221 schwarze 2301: *strrchr(tfn, '/') = '\0';
1.254 schwarze 2302: rmdir(tfn);
1.4 kristaps 2303: }
2304:
1.50 kristaps 2305: static int
1.165 schwarze 2306: set_basedir(const char *targetdir, int report_baddir)
1.4 kristaps 2307: {
1.59 schwarze 2308: static char startdir[PATH_MAX];
1.151 schwarze 2309: static int getcwd_status; /* 1 = ok, 2 = failure */
2310: static int chdir_status; /* 1 = changed directory */
1.4 kristaps 2311:
1.59 schwarze 2312: /*
1.151 schwarze 2313: * Remember the original working directory, if possible.
2314: * This will be needed if the second or a later directory
2315: * on the command line is given as a relative path.
2316: * Do not error out if the current directory is not
2317: * searchable: Maybe it won't be needed after all.
2318: */
1.264 schwarze 2319: if (getcwd_status == 0) {
2320: if (getcwd(startdir, sizeof(startdir)) == NULL) {
1.151 schwarze 2321: getcwd_status = 2;
2322: (void)strlcpy(startdir, strerror(errno),
2323: sizeof(startdir));
2324: } else
2325: getcwd_status = 1;
2326: }
2327:
2328: /*
2329: * We are leaving the old base directory.
2330: * Do not use it any longer, not even for messages.
2331: */
2332: *basedir = '\0';
1.264 schwarze 2333: basedir_len = 0;
1.151 schwarze 2334:
2335: /*
2336: * If and only if the directory was changed earlier and
2337: * the next directory to process is given as a relative path,
2338: * first go back, or bail out if that is impossible.
1.59 schwarze 2339: */
1.264 schwarze 2340: if (chdir_status && *targetdir != '/') {
2341: if (getcwd_status == 2) {
1.59 schwarze 2342: exitcode = (int)MANDOCLEVEL_SYSERR;
1.151 schwarze 2343: say("", "getcwd: %s", startdir);
1.197 schwarze 2344: return 0;
1.59 schwarze 2345: }
1.264 schwarze 2346: if (chdir(startdir) == -1) {
1.59 schwarze 2347: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2348: say("", "&chdir %s", startdir);
1.197 schwarze 2349: return 0;
1.59 schwarze 2350: }
2351: }
1.151 schwarze 2352:
2353: /*
2354: * Always resolve basedir to the canonicalized absolute
2355: * pathname and append a trailing slash, such that
2356: * we can reliably check whether files are inside.
2357: */
1.264 schwarze 2358: if (realpath(targetdir, basedir) == NULL) {
1.165 schwarze 2359: if (report_baddir || errno != ENOENT) {
2360: exitcode = (int)MANDOCLEVEL_BADARG;
2361: say("", "&%s: realpath", targetdir);
2362: }
1.264 schwarze 2363: *basedir = '\0';
1.197 schwarze 2364: return 0;
1.264 schwarze 2365: } else if (chdir(basedir) == -1) {
1.165 schwarze 2366: if (report_baddir || errno != ENOENT) {
2367: exitcode = (int)MANDOCLEVEL_BADARG;
2368: say("", "&chdir");
2369: }
1.264 schwarze 2370: *basedir = '\0';
1.197 schwarze 2371: return 0;
1.150 schwarze 2372: }
1.151 schwarze 2373: chdir_status = 1;
1.264 schwarze 2374: basedir_len = strlen(basedir);
2375: if (basedir[basedir_len - 1] != '/') {
2376: if (basedir_len >= PATH_MAX - 1) {
1.150 schwarze 2377: exitcode = (int)MANDOCLEVEL_SYSERR;
2378: say("", "Filename too long");
1.264 schwarze 2379: *basedir = '\0';
2380: basedir_len = 0;
1.197 schwarze 2381: return 0;
1.150 schwarze 2382: }
1.264 schwarze 2383: basedir[basedir_len++] = '/';
2384: basedir[basedir_len] = '\0';
1.4 kristaps 2385: }
1.197 schwarze 2386: return 1;
1.56 schwarze 2387: }
1.268 schwarze 2388:
2389: #ifdef READ_ALLOWED_PATH
2390: static int
2391: read_allowed(const char *candidate)
2392: {
2393: const char *cp;
2394: size_t len;
2395:
2396: for (cp = READ_ALLOWED_PATH;; cp += len) {
2397: while (*cp == ':')
2398: cp++;
2399: if (*cp == '\0')
2400: return 0;
2401: len = strcspn(cp, ":");
2402: if (strncmp(candidate, cp, len) == 0)
2403: return 1;
2404: }
2405: }
2406: #endif
1.56 schwarze 2407:
2408: static void
1.59 schwarze 2409: say(const char *file, const char *format, ...)
1.56 schwarze 2410: {
2411: va_list ap;
1.123 schwarze 2412: int use_errno;
1.56 schwarze 2413:
1.264 schwarze 2414: if (*basedir != '\0')
1.59 schwarze 2415: fprintf(stderr, "%s", basedir);
1.264 schwarze 2416: if (*basedir != '\0' && *file != '\0')
1.151 schwarze 2417: fputc('/', stderr);
1.264 schwarze 2418: if (*file != '\0')
1.59 schwarze 2419: fprintf(stderr, "%s", file);
2420:
1.123 schwarze 2421: use_errno = 1;
1.264 schwarze 2422: if (format != NULL) {
1.123 schwarze 2423: switch (*format) {
1.141 schwarze 2424: case '&':
1.123 schwarze 2425: format++;
2426: break;
1.141 schwarze 2427: case '\0':
1.123 schwarze 2428: format = NULL;
2429: break;
2430: default:
2431: use_errno = 0;
2432: break;
2433: }
2434: }
1.264 schwarze 2435: if (format != NULL) {
2436: if (*basedir != '\0' || *file != '\0')
1.123 schwarze 2437: fputs(": ", stderr);
2438: va_start(ap, format);
2439: vfprintf(stderr, format, ap);
2440: va_end(ap);
2441: }
2442: if (use_errno) {
1.264 schwarze 2443: if (*basedir != '\0' || *file != '\0' || format != NULL)
1.123 schwarze 2444: fputs(": ", stderr);
1.59 schwarze 2445: perror(NULL);
1.123 schwarze 2446: } else
2447: fputc('\n', stderr);
1.1 kristaps 2448: }
CVSweb