Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.170
1.170 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.169 2016/03/15 20:50:23 krw Exp $ */
1.1 schwarze 2: /*
1.47 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.164 schwarze 4: * Copyright (c) 2011-2016 Ingo Schwarze <schwarze@openbsd.org>
1.1 schwarze 5: *
6: * Permission to use, copy, modify, and distribute this software for any
7: * purpose with or without fee is hereby granted, provided that the above
8: * copyright notice and this permission notice appear in all copies.
9: *
1.140 schwarze 10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
1.1 schwarze 11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
1.140 schwarze 12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
1.1 schwarze 13: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17: */
1.117 schwarze 18: #include <sys/types.h>
1.47 schwarze 19: #include <sys/stat.h>
1.72 schwarze 20: #include <sys/wait.h>
1.1 schwarze 21:
22: #include <assert.h>
1.33 schwarze 23: #include <ctype.h>
1.151 schwarze 24: #include <err.h>
1.34 schwarze 25: #include <errno.h>
1.1 schwarze 26: #include <fcntl.h>
1.47 schwarze 27: #include <fts.h>
1.44 schwarze 28: #include <limits.h>
1.47 schwarze 29: #include <stddef.h>
1.1 schwarze 30: #include <stdio.h>
31: #include <stdint.h>
32: #include <stdlib.h>
33: #include <string.h>
1.14 schwarze 34: #include <unistd.h>
1.1 schwarze 35:
1.47 schwarze 36: #include <sqlite3.h>
37:
1.141 schwarze 38: #include "mandoc_aux.h"
1.156 schwarze 39: #include "mandoc_ohash.h"
1.141 schwarze 40: #include "mandoc.h"
41: #include "roff.h"
1.47 schwarze 42: #include "mdoc.h"
1.1 schwarze 43: #include "man.h"
1.140 schwarze 44: #include "manconf.h"
1.47 schwarze 45: #include "mansearch.h"
1.1 schwarze 46:
1.68 schwarze 47: extern int mansearch_keymax;
48: extern const char *const mansearch_keynames[];
49:
1.47 schwarze 50: #define SQL_EXEC(_v) \
51: if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
1.91 schwarze 52: say("", "%s: %s", (_v), sqlite3_errmsg(db))
1.47 schwarze 53: #define SQL_BIND_TEXT(_s, _i, _v) \
54: if (SQLITE_OK != sqlite3_bind_text \
55: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.91 schwarze 56: say(mlink->file, "%s", sqlite3_errmsg(db))
1.47 schwarze 57: #define SQL_BIND_INT(_s, _i, _v) \
58: if (SQLITE_OK != sqlite3_bind_int \
59: ((_s), (_i)++, (_v))) \
1.91 schwarze 60: say(mlink->file, "%s", sqlite3_errmsg(db))
1.47 schwarze 61: #define SQL_BIND_INT64(_s, _i, _v) \
62: if (SQLITE_OK != sqlite3_bind_int64 \
63: ((_s), (_i)++, (_v))) \
1.91 schwarze 64: say(mlink->file, "%s", sqlite3_errmsg(db))
1.47 schwarze 65: #define SQL_STEP(_s) \
66: if (SQLITE_DONE != sqlite3_step((_s))) \
1.91 schwarze 67: say(mlink->file, "%s", sqlite3_errmsg(db))
1.1 schwarze 68:
1.47 schwarze 69: enum op {
70: OP_DEFAULT = 0, /* new dbs from dir list or default config */
71: OP_CONFFILE, /* new databases from custom config file */
72: OP_UPDATE, /* delete/add entries in existing database */
73: OP_DELETE, /* delete entries from existing database */
74: OP_TEST /* change no databases, report potential problems */
75: };
1.11 schwarze 76:
1.47 schwarze 77: struct str {
78: const struct mpage *mpage; /* if set, the owning parse */
79: uint64_t mask; /* bitmask in sequence */
1.131 schwarze 80: char key[]; /* rendered text */
1.28 schwarze 81: };
82:
1.47 schwarze 83: struct inodev {
84: ino_t st_ino;
85: dev_t st_dev;
86: };
1.28 schwarze 87:
1.47 schwarze 88: struct mpage {
89: struct inodev inodev; /* used for hashing routine */
1.94 schwarze 90: int64_t pageid; /* pageid in mpages SQL table */
1.47 schwarze 91: char *sec; /* section from file content */
92: char *arch; /* architecture from file content */
93: char *title; /* title from file content */
94: char *desc; /* description from file content */
95: struct mlink *mlinks; /* singly linked list */
1.116 schwarze 96: int form; /* format from file content */
1.129 schwarze 97: int name_head_done;
1.28 schwarze 98: };
99:
1.47 schwarze 100: struct mlink {
101: char file[PATH_MAX]; /* filename rel. to manpath */
102: char *dsec; /* section from directory */
103: char *arch; /* architecture from directory */
104: char *name; /* name from file name (not empty) */
105: char *fsec; /* section from file name suffix */
106: struct mlink *next; /* singly linked list */
1.75 schwarze 107: struct mpage *mpage; /* parent */
1.116 schwarze 108: int dform; /* format from directory */
109: int fform; /* format from file name suffix */
1.81 schwarze 110: int gzip; /* filename has a .gz suffix */
1.2 schwarze 111: };
112:
1.47 schwarze 113: enum stmt {
114: STMT_DELETE_PAGE = 0, /* delete mpage */
115: STMT_INSERT_PAGE, /* insert mpage */
116: STMT_INSERT_LINK, /* insert mlink */
1.90 schwarze 117: STMT_INSERT_NAME, /* insert name */
1.130 schwarze 118: STMT_SELECT_NAME, /* retrieve existing name flags */
1.47 schwarze 119: STMT_INSERT_KEY, /* insert parsed key */
120: STMT__MAX
1.1 schwarze 121: };
122:
1.143 schwarze 123: typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
1.142 schwarze 124: const struct roff_node *);
1.1 schwarze 125:
1.19 schwarze 126: struct mdoc_handler {
1.47 schwarze 127: mdoc_fp fp; /* optional handler */
128: uint64_t mask; /* set unless handler returns 0 */
1.19 schwarze 129: };
130:
1.47 schwarze 131: static void dbclose(int);
1.131 schwarze 132: static void dbadd(struct mpage *);
1.75 schwarze 133: static void dbadd_mlink(const struct mlink *mlink);
1.124 schwarze 134: static void dbadd_mlink_name(const struct mlink *mlink);
1.47 schwarze 135: static int dbopen(int);
136: static void dbprune(void);
137: static void filescan(const char *);
138: static void mlink_add(struct mlink *, const struct stat *);
1.88 schwarze 139: static void mlink_check(struct mpage *, struct mlink *);
1.47 schwarze 140: static void mlink_free(struct mlink *);
141: static void mlinks_undupe(struct mpage *);
142: static void mpages_free(void);
1.131 schwarze 143: static void mpages_merge(struct mparse *);
1.87 schwarze 144: static void names_check(void);
1.81 schwarze 145: static void parse_cat(struct mpage *, int);
1.143 schwarze 146: static void parse_man(struct mpage *, const struct roff_meta *,
1.142 schwarze 147: const struct roff_node *);
1.143 schwarze 148: static void parse_mdoc(struct mpage *, const struct roff_meta *,
1.142 schwarze 149: const struct roff_node *);
1.143 schwarze 150: static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
1.142 schwarze 151: const struct roff_node *);
1.143 schwarze 152: static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
1.142 schwarze 153: const struct roff_node *);
154: static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
1.143 schwarze 155: static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
1.142 schwarze 156: const struct roff_node *);
1.143 schwarze 157: static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
1.142 schwarze 158: const struct roff_node *);
1.143 schwarze 159: static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
1.142 schwarze 160: const struct roff_node *);
1.143 schwarze 161: static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
1.142 schwarze 162: const struct roff_node *);
1.143 schwarze 163: static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
1.142 schwarze 164: const struct roff_node *);
1.164 schwarze 165: static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
166: const struct roff_node *);
1.143 schwarze 167: static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
1.142 schwarze 168: const struct roff_node *);
1.69 schwarze 169: static void putkey(const struct mpage *, char *, uint64_t);
1.131 schwarze 170: static void putkeys(const struct mpage *, char *, size_t, uint64_t);
1.47 schwarze 171: static void putmdockey(const struct mpage *,
1.142 schwarze 172: const struct roff_node *, uint64_t);
1.131 schwarze 173: static int render_string(char **, size_t *);
1.47 schwarze 174: static void say(const char *, const char *, ...);
1.120 schwarze 175: static int set_basedir(const char *, int);
1.47 schwarze 176: static int treescan(void);
177: static size_t utf8(unsigned int, char [7]);
178:
1.72 schwarze 179: static char tempfilename[32];
1.59 schwarze 180: static int nodb; /* no database changes */
1.73 schwarze 181: static int mparse_options; /* abort the parse early */
1.98 schwarze 182: static int use_all; /* use all found files */
183: static int debug; /* print what we're doing */
184: static int warnings; /* warn about crap */
1.52 schwarze 185: static int write_utf8; /* write UTF-8 output; else ASCII */
1.47 schwarze 186: static int exitcode; /* to be returned by main */
1.98 schwarze 187: static enum op op; /* operational mode */
1.47 schwarze 188: static char basedir[PATH_MAX]; /* current base directory */
189: static struct ohash mpages; /* table of distinct manual pages */
190: static struct ohash mlinks; /* table of directory entries */
1.90 schwarze 191: static struct ohash names; /* table of all names */
1.47 schwarze 192: static struct ohash strings; /* table of all strings */
193: static sqlite3 *db = NULL; /* current database */
194: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.90 schwarze 195: static uint64_t name_mask;
1.47 schwarze 196:
1.19 schwarze 197: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.47 schwarze 198: { NULL, 0 }, /* Ap */
199: { NULL, 0 }, /* Dd */
200: { NULL, 0 }, /* Dt */
201: { NULL, 0 }, /* Os */
202: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
203: { parse_mdoc_head, TYPE_Ss }, /* Ss */
204: { NULL, 0 }, /* Pp */
205: { NULL, 0 }, /* D1 */
206: { NULL, 0 }, /* Dl */
207: { NULL, 0 }, /* Bd */
208: { NULL, 0 }, /* Ed */
209: { NULL, 0 }, /* Bl */
210: { NULL, 0 }, /* El */
211: { NULL, 0 }, /* It */
212: { NULL, 0 }, /* Ad */
213: { NULL, TYPE_An }, /* An */
214: { NULL, TYPE_Ar }, /* Ar */
215: { NULL, TYPE_Cd }, /* Cd */
216: { NULL, TYPE_Cm }, /* Cm */
217: { NULL, TYPE_Dv }, /* Dv */
218: { NULL, TYPE_Er }, /* Er */
219: { NULL, TYPE_Ev }, /* Ev */
220: { NULL, 0 }, /* Ex */
221: { NULL, TYPE_Fa }, /* Fa */
222: { parse_mdoc_Fd, 0 }, /* Fd */
223: { NULL, TYPE_Fl }, /* Fl */
224: { parse_mdoc_Fn, 0 }, /* Fn */
225: { NULL, TYPE_Ft }, /* Ft */
226: { NULL, TYPE_Ic }, /* Ic */
1.49 schwarze 227: { NULL, TYPE_In }, /* In */
1.47 schwarze 228: { NULL, TYPE_Li }, /* Li */
1.89 schwarze 229: { parse_mdoc_Nd, 0 }, /* Nd */
1.86 schwarze 230: { parse_mdoc_Nm, 0 }, /* Nm */
1.47 schwarze 231: { NULL, 0 }, /* Op */
232: { NULL, 0 }, /* Ot */
233: { NULL, TYPE_Pa }, /* Pa */
234: { NULL, 0 }, /* Rv */
1.49 schwarze 235: { NULL, TYPE_St }, /* St */
1.164 schwarze 236: { parse_mdoc_Va, TYPE_Va }, /* Va */
237: { parse_mdoc_Va, TYPE_Vt }, /* Vt */
1.47 schwarze 238: { parse_mdoc_Xr, 0 }, /* Xr */
239: { NULL, 0 }, /* %A */
240: { NULL, 0 }, /* %B */
241: { NULL, 0 }, /* %D */
242: { NULL, 0 }, /* %I */
243: { NULL, 0 }, /* %J */
244: { NULL, 0 }, /* %N */
245: { NULL, 0 }, /* %O */
246: { NULL, 0 }, /* %P */
247: { NULL, 0 }, /* %R */
248: { NULL, 0 }, /* %T */
249: { NULL, 0 }, /* %V */
250: { NULL, 0 }, /* Ac */
251: { NULL, 0 }, /* Ao */
252: { NULL, 0 }, /* Aq */
253: { NULL, TYPE_At }, /* At */
254: { NULL, 0 }, /* Bc */
255: { NULL, 0 }, /* Bf */
256: { NULL, 0 }, /* Bo */
257: { NULL, 0 }, /* Bq */
258: { NULL, TYPE_Bsx }, /* Bsx */
259: { NULL, TYPE_Bx }, /* Bx */
260: { NULL, 0 }, /* Db */
261: { NULL, 0 }, /* Dc */
262: { NULL, 0 }, /* Do */
263: { NULL, 0 }, /* Dq */
264: { NULL, 0 }, /* Ec */
265: { NULL, 0 }, /* Ef */
266: { NULL, TYPE_Em }, /* Em */
267: { NULL, 0 }, /* Eo */
268: { NULL, TYPE_Fx }, /* Fx */
269: { NULL, TYPE_Ms }, /* Ms */
270: { NULL, 0 }, /* No */
271: { NULL, 0 }, /* Ns */
272: { NULL, TYPE_Nx }, /* Nx */
273: { NULL, TYPE_Ox }, /* Ox */
274: { NULL, 0 }, /* Pc */
275: { NULL, 0 }, /* Pf */
276: { NULL, 0 }, /* Po */
277: { NULL, 0 }, /* Pq */
278: { NULL, 0 }, /* Qc */
279: { NULL, 0 }, /* Ql */
280: { NULL, 0 }, /* Qo */
281: { NULL, 0 }, /* Qq */
282: { NULL, 0 }, /* Re */
283: { NULL, 0 }, /* Rs */
284: { NULL, 0 }, /* Sc */
285: { NULL, 0 }, /* So */
286: { NULL, 0 }, /* Sq */
287: { NULL, 0 }, /* Sm */
288: { NULL, 0 }, /* Sx */
289: { NULL, TYPE_Sy }, /* Sy */
290: { NULL, TYPE_Tn }, /* Tn */
291: { NULL, 0 }, /* Ux */
292: { NULL, 0 }, /* Xc */
293: { NULL, 0 }, /* Xo */
1.128 schwarze 294: { parse_mdoc_Fo, 0 }, /* Fo */
1.47 schwarze 295: { NULL, 0 }, /* Fc */
296: { NULL, 0 }, /* Oo */
297: { NULL, 0 }, /* Oc */
298: { NULL, 0 }, /* Bk */
299: { NULL, 0 }, /* Ek */
300: { NULL, 0 }, /* Bt */
301: { NULL, 0 }, /* Hf */
302: { NULL, 0 }, /* Fr */
303: { NULL, 0 }, /* Ud */
304: { NULL, TYPE_Lb }, /* Lb */
305: { NULL, 0 }, /* Lp */
306: { NULL, TYPE_Lk }, /* Lk */
307: { NULL, TYPE_Mt }, /* Mt */
308: { NULL, 0 }, /* Brq */
309: { NULL, 0 }, /* Bro */
310: { NULL, 0 }, /* Brc */
311: { NULL, 0 }, /* %C */
312: { NULL, 0 }, /* Es */
313: { NULL, 0 }, /* En */
314: { NULL, TYPE_Dx }, /* Dx */
315: { NULL, 0 }, /* %Q */
316: { NULL, 0 }, /* br */
317: { NULL, 0 }, /* sp */
318: { NULL, 0 }, /* %U */
319: { NULL, 0 }, /* Ta */
1.121 schwarze 320: { NULL, 0 }, /* ll */
1.1 schwarze 321: };
322:
1.98 schwarze 323:
1.1 schwarze 324: int
1.3 schwarze 325: mandocdb(int argc, char *argv[])
1.1 schwarze 326: {
1.140 schwarze 327: struct manconf conf;
328: struct mparse *mp;
1.161 schwarze 329: const char *path_arg, *progname;
1.140 schwarze 330: size_t j, sz;
1.47 schwarze 331: int ch, i;
332:
1.158 millert 333: if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) {
1.167 schwarze 334: warn("pledge");
1.155 schwarze 335: return (int)MANDOCLEVEL_SYSERR;
336: }
337:
1.140 schwarze 338: memset(&conf, 0, sizeof(conf));
1.47 schwarze 339: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
340:
341: /*
1.98 schwarze 342: * We accept a few different invocations.
1.47 schwarze 343: * The CHECKOP macro makes sure that invocation styles don't
344: * clobber each other.
345: */
346: #define CHECKOP(_op, _ch) do \
347: if (OP_DEFAULT != (_op)) { \
1.151 schwarze 348: warnx("-%c: Conflicting option", (_ch)); \
1.47 schwarze 349: goto usage; \
350: } while (/*CONSTCOND*/0)
1.10 schwarze 351:
1.47 schwarze 352: path_arg = NULL;
1.28 schwarze 353: op = OP_DEFAULT;
1.1 schwarze 354:
1.83 schwarze 355: while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
1.1 schwarze 356: switch (ch) {
1.98 schwarze 357: case 'a':
1.6 schwarze 358: use_all = 1;
359: break;
1.98 schwarze 360: case 'C':
1.47 schwarze 361: CHECKOP(op, ch);
362: path_arg = optarg;
1.28 schwarze 363: op = OP_CONFFILE;
1.25 schwarze 364: break;
1.98 schwarze 365: case 'D':
1.82 schwarze 366: debug++;
367: break;
1.98 schwarze 368: case 'd':
1.47 schwarze 369: CHECKOP(op, ch);
370: path_arg = optarg;
1.2 schwarze 371: op = OP_UPDATE;
1.1 schwarze 372: break;
1.98 schwarze 373: case 'n':
1.47 schwarze 374: nodb = 1;
375: break;
1.98 schwarze 376: case 'p':
1.83 schwarze 377: warnings = 1;
378: break;
1.98 schwarze 379: case 'Q':
1.73 schwarze 380: mparse_options |= MPARSE_QUICK;
1.59 schwarze 381: break;
1.98 schwarze 382: case 'T':
1.52 schwarze 383: if (strcmp(optarg, "utf8")) {
1.151 schwarze 384: warnx("-T%s: Unsupported output format",
385: optarg);
1.52 schwarze 386: goto usage;
387: }
388: write_utf8 = 1;
389: break;
1.98 schwarze 390: case 't':
1.47 schwarze 391: CHECKOP(op, ch);
1.28 schwarze 392: dup2(STDOUT_FILENO, STDERR_FILENO);
393: op = OP_TEST;
1.47 schwarze 394: nodb = warnings = 1;
1.28 schwarze 395: break;
1.98 schwarze 396: case 'u':
1.47 schwarze 397: CHECKOP(op, ch);
398: path_arg = optarg;
1.1 schwarze 399: op = OP_DELETE;
400: break;
1.98 schwarze 401: case 'v':
1.83 schwarze 402: /* Compatibility with espie@'s makewhatis. */
1.28 schwarze 403: break;
1.1 schwarze 404: default:
1.28 schwarze 405: goto usage;
1.1 schwarze 406: }
407:
408: argc -= optind;
409: argv += optind;
410:
1.163 schwarze 411: if (nodb) {
412: if (pledge("stdio rpath", NULL) == -1) {
1.167 schwarze 413: warn("pledge");
1.163 schwarze 414: return (int)MANDOCLEVEL_SYSERR;
415: }
1.155 schwarze 416: }
417:
1.28 schwarze 418: if (OP_CONFFILE == op && argc > 0) {
1.151 schwarze 419: warnx("-C: Too many arguments");
1.28 schwarze 420: goto usage;
421: }
422:
1.47 schwarze 423: exitcode = (int)MANDOCLEVEL_OK;
1.157 schwarze 424: mchars_alloc();
425: mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
1.156 schwarze 426: mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
427: mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
1.2 schwarze 428:
1.47 schwarze 429: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.2 schwarze 430:
1.47 schwarze 431: /*
1.109 schwarze 432: * Most of these deal with a specific directory.
1.97 schwarze 433: * Jump into that directory first.
1.47 schwarze 434: */
1.120 schwarze 435: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.47 schwarze 436: goto out;
1.97 schwarze 437:
1.95 schwarze 438: if (dbopen(1)) {
1.97 schwarze 439: /*
440: * The existing database is usable. Process
441: * all files specified on the command-line.
442: */
1.163 schwarze 443: if (!nodb) {
444: if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) {
1.167 schwarze 445: warn("pledge");
1.163 schwarze 446: exitcode = (int)MANDOCLEVEL_SYSERR;
447: goto out;
448: }
1.155 schwarze 449: }
1.97 schwarze 450: use_all = 1;
1.95 schwarze 451: for (i = 0; i < argc; i++)
452: filescan(argv[i]);
453: if (OP_TEST != op)
454: dbprune();
455: } else {
456: /*
457: * Database missing or corrupt.
458: * Recreate from scratch.
459: */
1.97 schwarze 460: exitcode = (int)MANDOCLEVEL_OK;
1.95 schwarze 461: op = OP_DEFAULT;
462: if (0 == treescan())
463: goto out;
464: if (0 == dbopen(0))
465: goto out;
466: }
1.47 schwarze 467: if (OP_DELETE != op)
1.131 schwarze 468: mpages_merge(mp);
1.95 schwarze 469: dbclose(OP_DEFAULT == op ? 0 : 1);
1.47 schwarze 470: } else {
471: /*
472: * If we have arguments, use them as our manpaths.
473: * If we don't, grok from manpath(1) or however else
1.140 schwarze 474: * manconf_parse() wants to do it.
1.47 schwarze 475: */
476: if (argc > 0) {
1.140 schwarze 477: conf.manpath.paths = mandoc_reallocarray(NULL,
1.101 schwarze 478: argc, sizeof(char *));
1.140 schwarze 479: conf.manpath.sz = (size_t)argc;
1.47 schwarze 480: for (i = 0; i < argc; i++)
1.140 schwarze 481: conf.manpath.paths[i] = mandoc_strdup(argv[i]);
1.47 schwarze 482: } else
1.140 schwarze 483: manconf_parse(&conf, path_arg, NULL, NULL);
1.84 schwarze 484:
1.140 schwarze 485: if (conf.manpath.sz == 0) {
1.84 schwarze 486: exitcode = (int)MANDOCLEVEL_BADARG;
487: say("", "Empty manpath");
488: }
1.2 schwarze 489:
1.47 schwarze 490: /*
491: * First scan the tree rooted at a base directory, then
492: * build a new database and finally move it into place.
493: * Ignore zero-length directories and strip trailing
494: * slashes.
495: */
1.140 schwarze 496: for (j = 0; j < conf.manpath.sz; j++) {
497: sz = strlen(conf.manpath.paths[j]);
498: if (sz && conf.manpath.paths[j][sz - 1] == '/')
499: conf.manpath.paths[j][--sz] = '\0';
1.47 schwarze 500: if (0 == sz)
501: continue;
1.2 schwarze 502:
1.47 schwarze 503: if (j) {
1.156 schwarze 504: mandoc_ohash_init(&mpages, 6,
505: offsetof(struct mpage, inodev));
506: mandoc_ohash_init(&mlinks, 6,
507: offsetof(struct mlink, file));
1.47 schwarze 508: }
1.1 schwarze 509:
1.140 schwarze 510: if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
1.120 schwarze 511: continue;
1.47 schwarze 512: if (0 == treescan())
1.120 schwarze 513: continue;
1.47 schwarze 514: if (0 == dbopen(0))
1.120 schwarze 515: continue;
1.41 deraadt 516:
1.131 schwarze 517: mpages_merge(mp);
1.104 schwarze 518: if (warnings && !nodb &&
1.87 schwarze 519: ! (MPARSE_QUICK & mparse_options))
520: names_check();
1.47 schwarze 521: dbclose(0);
1.28 schwarze 522:
1.140 schwarze 523: if (j + 1 < conf.manpath.sz) {
1.47 schwarze 524: mpages_free();
525: ohash_delete(&mpages);
526: ohash_delete(&mlinks);
527: }
1.2 schwarze 528: }
1.47 schwarze 529: }
530: out:
1.140 schwarze 531: manconf_free(&conf);
1.123 schwarze 532: mparse_free(mp);
1.157 schwarze 533: mchars_free();
1.47 schwarze 534: mpages_free();
535: ohash_delete(&mpages);
536: ohash_delete(&mlinks);
1.150 schwarze 537: return exitcode;
1.47 schwarze 538: usage:
1.161 schwarze 539: progname = getprogname();
1.83 schwarze 540: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
541: " %s [-aDnpQ] [-Tutf8] dir ...\n"
542: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
543: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 544: " %s [-Q] -t file ...\n",
1.161 schwarze 545: progname, progname, progname, progname, progname);
1.1 schwarze 546:
1.150 schwarze 547: return (int)MANDOCLEVEL_BADARG;
1.47 schwarze 548: }
1.1 schwarze 549:
1.47 schwarze 550: /*
551: * Scan a directory tree rooted at "basedir" for manpages.
552: * We use fts(), scanning directory parts along the way for clues to our
553: * section and architecture.
554: *
555: * If use_all has been specified, grok all files.
556: * If not, sanitise paths to the following:
557: *
1.98 schwarze 558: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 559: * or
560: * [./]cat<section>[/<arch>]/<name>.0
561: *
1.169 krw 562: * TODO: accommodate for multi-language directories.
1.47 schwarze 563: */
564: static int
565: treescan(void)
566: {
1.96 schwarze 567: char buf[PATH_MAX];
1.47 schwarze 568: FTS *f;
569: FTSENT *ff;
570: struct mlink *mlink;
1.81 schwarze 571: int dform, gzip;
1.51 schwarze 572: char *dsec, *arch, *fsec, *cp;
573: const char *path;
1.47 schwarze 574: const char *argv[2];
1.1 schwarze 575:
1.47 schwarze 576: argv[0] = ".";
577: argv[1] = (char *)NULL;
1.1 schwarze 578:
1.96 schwarze 579: f = fts_open((char * const *)argv,
580: FTS_PHYSICAL | FTS_NOCHDIR, NULL);
1.153 schwarze 581: if (f == NULL) {
1.47 schwarze 582: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 583: say("", "&fts_open");
1.150 schwarze 584: return 0;
1.47 schwarze 585: }
1.2 schwarze 586:
1.47 schwarze 587: dsec = arch = NULL;
588: dform = FORM_NONE;
1.2 schwarze 589:
1.153 schwarze 590: while ((ff = fts_read(f)) != NULL) {
1.47 schwarze 591: path = ff->fts_path + 2;
1.96 schwarze 592: switch (ff->fts_info) {
593:
594: /*
595: * Symbolic links require various sanity checks,
596: * then get handled just like regular files.
597: */
1.98 schwarze 598: case FTS_SL:
1.153 schwarze 599: if (realpath(path, buf) == NULL) {
1.96 schwarze 600: if (warnings)
601: say(path, "&realpath");
602: continue;
603: }
604: if (strstr(buf, basedir) != buf) {
605: if (warnings) say("",
606: "%s: outside base directory", buf);
607: continue;
608: }
609: /* Use logical inode to avoid mpages dupe. */
1.153 schwarze 610: if (stat(path, ff->fts_statp) == -1) {
1.96 schwarze 611: if (warnings)
612: say(path, "&stat");
613: continue;
614: }
615: /* FALLTHROUGH */
616:
1.15 schwarze 617: /*
1.47 schwarze 618: * If we're a regular file, add an mlink by using the
619: * stored directory data and handling the filename.
1.15 schwarze 620: */
1.98 schwarze 621: case FTS_F:
1.153 schwarze 622: if ( ! strcmp(path, MANDOC_DB))
1.47 schwarze 623: continue;
624: if ( ! use_all && ff->fts_level < 2) {
625: if (warnings)
626: say(path, "Extraneous file");
627: continue;
1.81 schwarze 628: }
629: gzip = 0;
630: fsec = NULL;
1.153 schwarze 631: while (fsec == NULL) {
1.81 schwarze 632: fsec = strrchr(ff->fts_name, '.');
1.153 schwarze 633: if (fsec == NULL || strcmp(fsec+1, "gz"))
1.81 schwarze 634: break;
635: gzip = 1;
636: *fsec = '\0';
637: fsec = NULL;
638: }
1.153 schwarze 639: if (fsec == NULL) {
1.47 schwarze 640: if ( ! use_all) {
641: if (warnings)
642: say(path,
643: "No filename suffix");
644: continue;
645: }
1.153 schwarze 646: } else if ( ! strcmp(++fsec, "html")) {
1.47 schwarze 647: if (warnings)
648: say(path, "Skip html");
649: continue;
1.153 schwarze 650: } else if ( ! strcmp(fsec, "ps")) {
1.47 schwarze 651: if (warnings)
652: say(path, "Skip ps");
653: continue;
1.153 schwarze 654: } else if ( ! strcmp(fsec, "pdf")) {
1.47 schwarze 655: if (warnings)
656: say(path, "Skip pdf");
657: continue;
658: } else if ( ! use_all &&
1.153 schwarze 659: ((dform == FORM_SRC &&
1.138 schwarze 660: strncmp(fsec, dsec, strlen(dsec))) ||
1.153 schwarze 661: (dform == FORM_CAT && strcmp(fsec, "0")))) {
1.47 schwarze 662: if (warnings)
663: say(path, "Wrong filename suffix");
664: continue;
665: } else
666: fsec[-1] = '\0';
1.51 schwarze 667:
1.47 schwarze 668: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.100 schwarze 669: if (strlcpy(mlink->file, path,
670: sizeof(mlink->file)) >=
671: sizeof(mlink->file)) {
672: say(path, "Filename too long");
673: free(mlink);
674: continue;
675: }
1.47 schwarze 676: mlink->dform = dform;
1.51 schwarze 677: mlink->dsec = dsec;
678: mlink->arch = arch;
679: mlink->name = ff->fts_name;
680: mlink->fsec = fsec;
1.81 schwarze 681: mlink->gzip = gzip;
1.47 schwarze 682: mlink_add(mlink, ff->fts_statp);
683: continue;
1.96 schwarze 684:
1.98 schwarze 685: case FTS_D:
686: case FTS_DP:
1.96 schwarze 687: break;
688:
689: default:
1.47 schwarze 690: if (warnings)
691: say(path, "Not a regular file");
692: continue;
693: }
694:
695: switch (ff->fts_level) {
1.98 schwarze 696: case 0:
1.47 schwarze 697: /* Ignore the root directory. */
698: break;
1.98 schwarze 699: case 1:
1.47 schwarze 700: /*
701: * This might contain manX/ or catX/.
702: * Try to infer this from the name.
703: * If we're not in use_all, enforce it.
704: */
705: cp = ff->fts_name;
1.153 schwarze 706: if (ff->fts_info == FTS_DP) {
707: dform = FORM_NONE;
708: dsec = NULL;
1.47 schwarze 709: break;
1.153 schwarze 710: }
1.15 schwarze 711:
1.153 schwarze 712: if ( ! strncmp(cp, "man", 3)) {
1.47 schwarze 713: dform = FORM_SRC;
714: dsec = cp + 3;
1.153 schwarze 715: } else if ( ! strncmp(cp, "cat", 3)) {
1.47 schwarze 716: dform = FORM_CAT;
717: dsec = cp + 3;
1.51 schwarze 718: } else {
719: dform = FORM_NONE;
720: dsec = NULL;
1.26 schwarze 721: }
1.47 schwarze 722:
1.153 schwarze 723: if (dsec != NULL || use_all)
1.47 schwarze 724: break;
725:
726: if (warnings)
727: say(path, "Unknown directory part");
728: fts_set(f, ff, FTS_SKIP);
729: break;
1.98 schwarze 730: case 2:
1.47 schwarze 731: /*
732: * Possibly our architecture.
733: * If we're descending, keep tabs on it.
734: */
1.153 schwarze 735: if (ff->fts_info != FTS_DP && dsec != NULL)
1.47 schwarze 736: arch = ff->fts_name;
1.51 schwarze 737: else
738: arch = NULL;
1.47 schwarze 739: break;
740: default:
1.153 schwarze 741: if (ff->fts_info == FTS_DP || use_all)
1.47 schwarze 742: break;
743: if (warnings)
744: say(path, "Extraneous directory part");
745: fts_set(f, ff, FTS_SKIP);
746: break;
1.14 schwarze 747: }
1.47 schwarze 748: }
749:
750: fts_close(f);
1.150 schwarze 751: return 1;
1.47 schwarze 752: }
1.1 schwarze 753:
1.47 schwarze 754: /*
755: * Add a file to the mlinks table.
756: * Do not verify that it's a "valid" looking manpage (we'll do that
757: * later).
758: *
759: * Try to infer the manual section, architecture, and page name from the
760: * path, assuming it looks like
761: *
1.98 schwarze 762: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 763: * or
764: * [./]cat<section>[/<arch>]/<name>.0
765: *
766: * See treescan() for the fts(3) version of this.
767: */
768: static void
769: filescan(const char *file)
770: {
771: char buf[PATH_MAX];
772: struct stat st;
773: struct mlink *mlink;
774: char *p, *start;
775:
776: assert(use_all);
777:
778: if (0 == strncmp(file, "./", 2))
779: file += 2;
780:
1.96 schwarze 781: /*
782: * We have to do lstat(2) before realpath(3) loses
783: * the information whether this is a symbolic link.
784: * We need to know that because for symbolic links,
785: * we want to use the orginal file name, while for
786: * regular files, we want to use the real path.
787: */
788: if (-1 == lstat(file, &st)) {
789: exitcode = (int)MANDOCLEVEL_BADARG;
790: say(file, "&lstat");
791: return;
792: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
793: exitcode = (int)MANDOCLEVEL_BADARG;
794: say(file, "Not a regular file");
795: return;
796: }
797:
798: /*
799: * We have to resolve the file name to the real path
800: * in any case for the base directory check.
801: */
1.47 schwarze 802: if (NULL == realpath(file, buf)) {
803: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 804: say(file, "&realpath");
1.47 schwarze 805: return;
1.63 schwarze 806: }
807:
1.109 schwarze 808: if (OP_TEST == op)
1.63 schwarze 809: start = buf;
1.109 schwarze 810: else if (strstr(buf, basedir) == buf)
811: start = buf + strlen(basedir);
1.63 schwarze 812: else {
1.47 schwarze 813: exitcode = (int)MANDOCLEVEL_BADARG;
814: say("", "%s: outside base directory", buf);
815: return;
1.63 schwarze 816: }
817:
1.96 schwarze 818: /*
819: * Now we are sure the file is inside our tree.
820: * If it is a symbolic link, ignore the real path
821: * and use the original name.
822: * This implies passing stuff like "cat1/../man1/foo.1"
823: * on the command line won't work. So don't do that.
824: * Note the stat(2) can still fail if the link target
825: * doesn't exist.
826: */
827: if (S_IFLNK & st.st_mode) {
828: if (-1 == stat(buf, &st)) {
829: exitcode = (int)MANDOCLEVEL_BADARG;
830: say(file, "&stat");
831: return;
832: }
1.100 schwarze 833: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
834: say(file, "Filename too long");
835: return;
836: }
1.109 schwarze 837: start = buf;
838: if (OP_TEST != op && strstr(buf, basedir) == buf)
839: start += strlen(basedir);
1.1 schwarze 840: }
1.63 schwarze 841:
1.47 schwarze 842: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 843: mlink->dform = FORM_NONE;
1.100 schwarze 844: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
845: sizeof(mlink->file)) {
846: say(start, "Filename too long");
1.134 schwarze 847: free(mlink);
1.100 schwarze 848: return;
849: }
1.1 schwarze 850:
1.10 schwarze 851: /*
1.47 schwarze 852: * First try to guess our directory structure.
853: * If we find a separator, try to look for man* or cat*.
854: * If we find one of these and what's underneath is a directory,
855: * assume it's an architecture.
1.10 schwarze 856: */
1.47 schwarze 857: if (NULL != (p = strchr(start, '/'))) {
858: *p++ = '\0';
859: if (0 == strncmp(start, "man", 3)) {
860: mlink->dform = FORM_SRC;
1.51 schwarze 861: mlink->dsec = start + 3;
1.47 schwarze 862: } else if (0 == strncmp(start, "cat", 3)) {
863: mlink->dform = FORM_CAT;
1.51 schwarze 864: mlink->dsec = start + 3;
1.47 schwarze 865: }
1.10 schwarze 866:
1.47 schwarze 867: start = p;
868: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
869: *p++ = '\0';
1.51 schwarze 870: mlink->arch = start;
1.47 schwarze 871: start = p;
1.41 deraadt 872: }
1.47 schwarze 873: }
1.7 schwarze 874:
1.47 schwarze 875: /*
876: * Now check the file suffix.
877: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
878: */
879: p = strrchr(start, '\0');
880: while (p-- > start && '/' != *p && '.' != *p)
881: /* Loop. */ ;
882:
883: if ('.' == *p) {
884: *p++ = '\0';
1.51 schwarze 885: mlink->fsec = p;
1.47 schwarze 886: }
1.41 deraadt 887:
1.47 schwarze 888: /*
889: * Now try to parse the name.
890: * Use the filename portion of the path.
891: */
892: mlink->name = start;
893: if (NULL != (p = strrchr(start, '/'))) {
894: mlink->name = p + 1;
895: *p = '\0';
896: }
897: mlink_add(mlink, &st);
898: }
1.2 schwarze 899:
1.47 schwarze 900: static void
901: mlink_add(struct mlink *mlink, const struct stat *st)
902: {
903: struct inodev inodev;
904: struct mpage *mpage;
905: unsigned int slot;
906:
907: assert(NULL != mlink->file);
908:
1.51 schwarze 909: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
910: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
911: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
912: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 913:
914: if ('0' == *mlink->fsec) {
915: free(mlink->fsec);
916: mlink->fsec = mandoc_strdup(mlink->dsec);
917: mlink->fform = FORM_CAT;
918: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
919: mlink->fform = FORM_SRC;
920: else
921: mlink->fform = FORM_NONE;
1.1 schwarze 922:
1.47 schwarze 923: slot = ohash_qlookup(&mlinks, mlink->file);
924: assert(NULL == ohash_find(&mlinks, slot));
925: ohash_insert(&mlinks, slot, mlink);
926:
1.139 schwarze 927: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.47 schwarze 928: inodev.st_ino = st->st_ino;
929: inodev.st_dev = st->st_dev;
930: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
931: sizeof(struct inodev), inodev.st_ino);
932: mpage = ohash_find(&mpages, slot);
933: if (NULL == mpage) {
934: mpage = mandoc_calloc(1, sizeof(struct mpage));
935: mpage->inodev.st_ino = inodev.st_ino;
936: mpage->inodev.st_dev = inodev.st_dev;
937: ohash_insert(&mpages, slot, mpage);
938: } else
939: mlink->next = mpage->mlinks;
940: mpage->mlinks = mlink;
1.75 schwarze 941: mlink->mpage = mpage;
1.47 schwarze 942: }
1.1 schwarze 943:
1.47 schwarze 944: static void
945: mlink_free(struct mlink *mlink)
946: {
1.1 schwarze 947:
1.47 schwarze 948: free(mlink->dsec);
949: free(mlink->arch);
950: free(mlink->name);
951: free(mlink->fsec);
952: free(mlink);
953: }
1.1 schwarze 954:
1.47 schwarze 955: static void
956: mpages_free(void)
957: {
958: struct mpage *mpage;
959: struct mlink *mlink;
960: unsigned int slot;
961:
962: mpage = ohash_first(&mpages, &slot);
963: while (NULL != mpage) {
964: while (NULL != (mlink = mpage->mlinks)) {
965: mpage->mlinks = mlink->next;
966: mlink_free(mlink);
967: }
968: free(mpage->sec);
969: free(mpage->arch);
970: free(mpage->title);
971: free(mpage->desc);
972: free(mpage);
973: mpage = ohash_next(&mpages, &slot);
974: }
975: }
1.1 schwarze 976:
1.47 schwarze 977: /*
978: * For each mlink to the mpage, check whether the path looks like
979: * it is formatted, and if it does, check whether a source manual
980: * exists by the same name, ignoring the suffix.
981: * If both conditions hold, drop the mlink.
982: */
983: static void
984: mlinks_undupe(struct mpage *mpage)
985: {
986: char buf[PATH_MAX];
987: struct mlink **prev;
988: struct mlink *mlink;
989: char *bufp;
990:
991: mpage->form = FORM_CAT;
992: prev = &mpage->mlinks;
993: while (NULL != (mlink = *prev)) {
994: if (FORM_CAT != mlink->dform) {
995: mpage->form = FORM_NONE;
996: goto nextlink;
997: }
1.100 schwarze 998: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 999: bufp = strstr(buf, "cat");
1000: assert(NULL != bufp);
1001: memcpy(bufp, "man", 3);
1002: if (NULL != (bufp = strrchr(buf, '.')))
1003: *++bufp = '\0';
1.100 schwarze 1004: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 1005: if (NULL == ohash_find(&mlinks,
1.98 schwarze 1006: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 1007: goto nextlink;
1008: if (warnings)
1009: say(mlink->file, "Man source exists: %s", buf);
1010: if (use_all)
1011: goto nextlink;
1012: *prev = mlink->next;
1013: mlink_free(mlink);
1014: continue;
1015: nextlink:
1016: prev = &(*prev)->next;
1.1 schwarze 1017: }
1.47 schwarze 1018: }
1.1 schwarze 1019:
1.88 schwarze 1020: static void
1.50 schwarze 1021: mlink_check(struct mpage *mpage, struct mlink *mlink)
1022: {
1.88 schwarze 1023: struct str *str;
1024: unsigned int slot;
1.50 schwarze 1025:
1026: /*
1027: * Check whether the manual section given in a file
1028: * agrees with the directory where the file is located.
1029: * Some manuals have suffixes like (3p) on their
1030: * section number either inside the file or in the
1031: * directory name, some are linked into more than one
1032: * section, like encrypt(1) = makekey(8).
1033: */
1034:
1035: if (FORM_SRC == mpage->form &&
1.88 schwarze 1036: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1037: say(mlink->file, "Section \"%s\" manual in %s directory",
1038: mpage->sec, mlink->dsec);
1039:
1040: /*
1041: * Manual page directories exist for each kernel
1042: * architecture as returned by machine(1).
1043: * However, many manuals only depend on the
1044: * application architecture as returned by arch(1).
1045: * For example, some (2/ARM) manuals are shared
1046: * across the "armish" and "zaurus" kernel
1047: * architectures.
1048: * A few manuals are even shared across completely
1049: * different architectures, for example fdformat(1)
1050: * on amd64, i386, sparc, and sparc64.
1051: */
1052:
1.88 schwarze 1053: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1054: say(mlink->file, "Architecture \"%s\" manual in "
1055: "\"%s\" directory", mpage->arch, mlink->arch);
1056:
1.88 schwarze 1057: /*
1058: * XXX
1.90 schwarze 1059: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1060: */
1061:
1062: if (FORM_CAT == mpage->form)
1063: return;
1064:
1065: /*
1066: * Check whether this mlink
1067: * appears as a name in the NAME section.
1068: */
1.50 schwarze 1069:
1.90 schwarze 1070: slot = ohash_qlookup(&names, mlink->name);
1071: str = ohash_find(&names, slot);
1.88 schwarze 1072: assert(NULL != str);
1.90 schwarze 1073: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1074: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1075: }
1076:
1.47 schwarze 1077: /*
1078: * Run through the files in the global vector "mpages"
1079: * and add them to the database specified in "basedir".
1080: *
1081: * This handles the parsing scheme itself, using the cues of directory
1082: * and filename to determine whether the file is parsable or not.
1083: */
1084: static void
1.131 schwarze 1085: mpages_merge(struct mparse *mp)
1.47 schwarze 1086: {
1.70 schwarze 1087: char any[] = "any";
1.75 schwarze 1088: struct mpage *mpage, *mpage_dest;
1089: struct mlink *mlink, *mlink_dest;
1.144 schwarze 1090: struct roff_man *man;
1.75 schwarze 1091: char *sodest;
1.69 schwarze 1092: char *cp;
1.117 schwarze 1093: int fd;
1.58 schwarze 1094: unsigned int pslot;
1.47 schwarze 1095:
1.126 schwarze 1096: if ( ! nodb)
1.64 schwarze 1097: SQL_EXEC("BEGIN TRANSACTION");
1098:
1.47 schwarze 1099: mpage = ohash_first(&mpages, &pslot);
1.126 schwarze 1100: while (mpage != NULL) {
1.47 schwarze 1101: mlinks_undupe(mpage);
1.136 schwarze 1102: if ((mlink = mpage->mlinks) == NULL) {
1.47 schwarze 1103: mpage = ohash_next(&mpages, &pslot);
1104: continue;
1105: }
1.1 schwarze 1106:
1.90 schwarze 1107: name_mask = NAME_MASK;
1.156 schwarze 1108: mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1109: mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1.47 schwarze 1110: mparse_reset(mp);
1111: man = NULL;
1.81 schwarze 1112: sodest = NULL;
1113:
1.166 schwarze 1114: if ((fd = mparse_open(mp, mlink->file)) == -1) {
1.136 schwarze 1115: say(mlink->file, "&open");
1.117 schwarze 1116: goto nextpage;
1.81 schwarze 1117: }
1.11 schwarze 1118:
1119: /*
1.137 schwarze 1120: * Interpret the file as mdoc(7) or man(7) source
1121: * code, unless it is known to be formatted.
1.11 schwarze 1122: */
1.136 schwarze 1123: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.137 schwarze 1124: mparse_readfd(mp, fd, mlink->file);
1.165 schwarze 1125: close(fd);
1.145 schwarze 1126: mparse_result(mp, &man, &sodest);
1.47 schwarze 1127: }
1.11 schwarze 1128:
1.126 schwarze 1129: if (sodest != NULL) {
1.75 schwarze 1130: mlink_dest = ohash_find(&mlinks,
1131: ohash_qlookup(&mlinks, sodest));
1.126 schwarze 1132: if (mlink_dest == NULL) {
1133: mandoc_asprintf(&cp, "%s.gz", sodest);
1134: mlink_dest = ohash_find(&mlinks,
1135: ohash_qlookup(&mlinks, cp));
1136: free(cp);
1137: }
1138: if (mlink_dest != NULL) {
1.75 schwarze 1139:
1140: /* The .so target exists. */
1141:
1142: mpage_dest = mlink_dest->mpage;
1143: while (1) {
1144: mlink->mpage = mpage_dest;
1145:
1146: /*
1147: * If the target was already
1148: * processed, add the links
1149: * to the database now.
1150: * Otherwise, this will
1151: * happen when we come
1152: * to the target.
1153: */
1154:
1.94 schwarze 1155: if (mpage_dest->pageid)
1.124 schwarze 1156: dbadd_mlink_name(mlink);
1.75 schwarze 1157:
1.126 schwarze 1158: if (mlink->next == NULL)
1.75 schwarze 1159: break;
1160: mlink = mlink->next;
1161: }
1162:
1163: /* Move all links to the target. */
1164:
1165: mlink->next = mlink_dest->next;
1166: mlink_dest->next = mpage->mlinks;
1167: mpage->mlinks = NULL;
1168: }
1.81 schwarze 1169: goto nextpage;
1.145 schwarze 1170: } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1.159 schwarze 1171: mdoc_validate(man);
1.47 schwarze 1172: mpage->form = FORM_SRC;
1.146 schwarze 1173: mpage->sec = man->meta.msec;
1.112 schwarze 1174: mpage->sec = mandoc_strdup(
1.126 schwarze 1175: mpage->sec == NULL ? "" : mpage->sec);
1.146 schwarze 1176: mpage->arch = man->meta.arch;
1.47 schwarze 1177: mpage->arch = mandoc_strdup(
1.126 schwarze 1178: mpage->arch == NULL ? "" : mpage->arch);
1.146 schwarze 1179: mpage->title = mandoc_strdup(man->meta.title);
1.145 schwarze 1180: } else if (man != NULL && man->macroset == MACROSET_MAN) {
1.160 schwarze 1181: man_validate(man);
1.47 schwarze 1182: mpage->form = FORM_SRC;
1.146 schwarze 1183: mpage->sec = mandoc_strdup(man->meta.msec);
1.136 schwarze 1184: mpage->arch = mandoc_strdup(mlink->arch);
1.146 schwarze 1185: mpage->title = mandoc_strdup(man->meta.title);
1.11 schwarze 1186: } else {
1.47 schwarze 1187: mpage->form = FORM_CAT;
1.136 schwarze 1188: mpage->sec = mandoc_strdup(mlink->dsec);
1189: mpage->arch = mandoc_strdup(mlink->arch);
1190: mpage->title = mandoc_strdup(mlink->name);
1.1 schwarze 1191: }
1.54 schwarze 1192: putkey(mpage, mpage->sec, TYPE_sec);
1.119 schwarze 1193: if (*mpage->arch != '\0')
1194: putkey(mpage, mpage->arch, TYPE_arch);
1.1 schwarze 1195:
1.136 schwarze 1196: for ( ; mlink != NULL; mlink = mlink->next) {
1.54 schwarze 1197: if ('\0' != *mlink->dsec)
1198: putkey(mpage, mlink->dsec, TYPE_sec);
1199: if ('\0' != *mlink->fsec)
1200: putkey(mpage, mlink->fsec, TYPE_sec);
1.55 schwarze 1201: putkey(mpage, '\0' == *mlink->arch ?
1.70 schwarze 1202: any : mlink->arch, TYPE_arch);
1.90 schwarze 1203: putkey(mpage, mlink->name, NAME_FILE);
1.54 schwarze 1204: }
1.41 deraadt 1205:
1.129 schwarze 1206: assert(mpage->desc == NULL);
1.145 schwarze 1207: if (man != NULL && man->macroset == MACROSET_MDOC)
1.146 schwarze 1208: parse_mdoc(mpage, &man->meta, man->first);
1.129 schwarze 1209: else if (man != NULL)
1.146 schwarze 1210: parse_man(mpage, &man->meta, man->first);
1.47 schwarze 1211: else
1.117 schwarze 1212: parse_cat(mpage, fd);
1.129 schwarze 1213: if (mpage->desc == NULL)
1.92 schwarze 1214: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1.88 schwarze 1215:
1216: if (warnings && !use_all)
1217: for (mlink = mpage->mlinks; mlink;
1218: mlink = mlink->next)
1219: mlink_check(mpage, mlink);
1.6 schwarze 1220:
1.131 schwarze 1221: dbadd(mpage);
1.136 schwarze 1222: mlink = mpage->mlinks;
1.81 schwarze 1223:
1224: nextpage:
1.47 schwarze 1225: ohash_delete(&strings);
1.90 schwarze 1226: ohash_delete(&names);
1.47 schwarze 1227: mpage = ohash_next(&mpages, &pslot);
1228: }
1.64 schwarze 1229:
1230: if (0 == nodb)
1231: SQL_EXEC("END TRANSACTION");
1.87 schwarze 1232: }
1233:
1234: static void
1235: names_check(void)
1236: {
1237: sqlite3_stmt *stmt;
1238: const char *name, *sec, *arch, *key;
1239:
1240: sqlite3_prepare_v2(db,
1241: "SELECT name, sec, arch, key FROM ("
1.90 schwarze 1242: "SELECT name AS key, pageid FROM names "
1.87 schwarze 1243: "WHERE bits & ? AND NOT EXISTS ("
1244: "SELECT pageid FROM mlinks "
1.90 schwarze 1245: "WHERE mlinks.pageid == names.pageid "
1246: "AND mlinks.name == names.name"
1.87 schwarze 1247: ")"
1248: ") JOIN ("
1.99 schwarze 1249: "SELECT sec, arch, name, pageid FROM mlinks "
1250: "GROUP BY pageid"
1.87 schwarze 1251: ") USING (pageid);",
1252: -1, &stmt, NULL);
1253:
1.154 schwarze 1254: if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK)
1.91 schwarze 1255: say("", "%s", sqlite3_errmsg(db));
1.87 schwarze 1256:
1.154 schwarze 1257: while (sqlite3_step(stmt) == SQLITE_ROW) {
1.113 schwarze 1258: name = (const char *)sqlite3_column_text(stmt, 0);
1259: sec = (const char *)sqlite3_column_text(stmt, 1);
1260: arch = (const char *)sqlite3_column_text(stmt, 2);
1261: key = (const char *)sqlite3_column_text(stmt, 3);
1.87 schwarze 1262: say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1263: '\0' == *arch ? "" : "/",
1264: '\0' == *arch ? "" : arch, key);
1265: }
1266: sqlite3_finalize(stmt);
1.47 schwarze 1267: }
1.6 schwarze 1268:
1.47 schwarze 1269: static void
1.81 schwarze 1270: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1271: {
1272: FILE *stream;
1273: char *line, *p, *title;
1.162 schwarze 1274: size_t linesz, plen, titlesz;
1275: ssize_t len;
1276: int offs;
1.1 schwarze 1277:
1.81 schwarze 1278: stream = (-1 == fd) ?
1279: fopen(mpage->mlinks->file, "r") :
1280: fdopen(fd, "r");
1281: if (NULL == stream) {
1.114 doug 1282: if (-1 != fd)
1283: close(fd);
1.47 schwarze 1284: if (warnings)
1.80 schwarze 1285: say(mpage->mlinks->file, "&fopen");
1.47 schwarze 1286: return;
1287: }
1.1 schwarze 1288:
1.162 schwarze 1289: line = NULL;
1290: linesz = 0;
1291:
1.47 schwarze 1292: /* Skip to first blank line. */
1.1 schwarze 1293:
1.162 schwarze 1294: while (getline(&line, &linesz, stream) != -1)
1295: if (*line == '\n')
1.47 schwarze 1296: break;
1.1 schwarze 1297:
1.47 schwarze 1298: /*
1299: * Assume the first line that is not indented
1300: * is the first section header. Skip to it.
1301: */
1.1 schwarze 1302:
1.162 schwarze 1303: while (getline(&line, &linesz, stream) != -1)
1304: if (*line != '\n' && *line != ' ')
1.47 schwarze 1305: break;
1.98 schwarze 1306:
1.47 schwarze 1307: /*
1308: * Read up until the next section into a buffer.
1309: * Strip the leading and trailing newline from each read line,
1310: * appending a trailing space.
1311: * Ignore empty (whitespace-only) lines.
1312: */
1.28 schwarze 1313:
1.47 schwarze 1314: titlesz = 0;
1315: title = NULL;
1.38 schwarze 1316:
1.162 schwarze 1317: while ((len = getline(&line, &linesz, stream)) != -1) {
1318: if (*line != ' ')
1.47 schwarze 1319: break;
1.162 schwarze 1320: offs = 0;
1321: while (isspace((unsigned char)line[offs]))
1322: offs++;
1323: if (line[offs] == '\0')
1.47 schwarze 1324: continue;
1.162 schwarze 1325: title = mandoc_realloc(title, titlesz + len - offs);
1326: memcpy(title + titlesz, line + offs, len - offs);
1327: titlesz += len - offs;
1.47 schwarze 1328: title[titlesz - 1] = ' ';
1329: }
1.162 schwarze 1330: free(line);
1.28 schwarze 1331:
1.47 schwarze 1332: /*
1333: * If no page content can be found, or the input line
1334: * is already the next section header, or there is no
1335: * trailing newline, reuse the page title as the page
1336: * description.
1337: */
1.1 schwarze 1338:
1.47 schwarze 1339: if (NULL == title || '\0' == *title) {
1340: if (warnings)
1341: say(mpage->mlinks->file,
1342: "Cannot find NAME section");
1343: fclose(stream);
1344: free(title);
1345: return;
1346: }
1.24 schwarze 1347:
1.162 schwarze 1348: title[titlesz - 1] = '\0';
1.24 schwarze 1349:
1.47 schwarze 1350: /*
1351: * Skip to the first dash.
1352: * Use the remaining line as the description (no more than 70
1353: * bytes).
1354: */
1.28 schwarze 1355:
1.47 schwarze 1356: if (NULL != (p = strstr(title, "- "))) {
1357: for (p += 2; ' ' == *p || '\b' == *p; p++)
1358: /* Skip to next word. */ ;
1359: } else {
1360: if (warnings)
1361: say(mpage->mlinks->file,
1362: "No dash in title line");
1363: p = title;
1364: }
1.1 schwarze 1365:
1.47 schwarze 1366: plen = strlen(p);
1.1 schwarze 1367:
1.47 schwarze 1368: /* Strip backspace-encoding from line. */
1.1 schwarze 1369:
1.47 schwarze 1370: while (NULL != (line = memchr(p, '\b', plen))) {
1371: len = line - p;
1372: if (0 == len) {
1373: memmove(line, line + 1, plen--);
1374: continue;
1.98 schwarze 1375: }
1.47 schwarze 1376: memmove(line - 1, line + 1, plen - len);
1377: plen -= 2;
1378: }
1.1 schwarze 1379:
1.47 schwarze 1380: mpage->desc = mandoc_strdup(p);
1381: fclose(stream);
1382: free(title);
1383: }
1.16 schwarze 1384:
1.47 schwarze 1385: /*
1386: * Put a type/word pair into the word database for this particular file.
1387: */
1388: static void
1.69 schwarze 1389: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1390: {
1.69 schwarze 1391: char *cp;
1.37 schwarze 1392:
1.47 schwarze 1393: assert(NULL != value);
1.69 schwarze 1394: if (TYPE_arch == type)
1395: for (cp = value; *cp; cp++)
1396: if (isupper((unsigned char)*cp))
1397: *cp = _tolower((unsigned char)*cp);
1.47 schwarze 1398: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1399: }
1400:
1401: /*
1.47 schwarze 1402: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1403: */
1404: static void
1.47 schwarze 1405: putmdockey(const struct mpage *mpage,
1.142 schwarze 1406: const struct roff_node *n, uint64_t m)
1.2 schwarze 1407: {
1.16 schwarze 1408:
1.47 schwarze 1409: for ( ; NULL != n; n = n->next) {
1410: if (NULL != n->child)
1411: putmdockey(mpage, n->child, m);
1.141 schwarze 1412: if (n->type == ROFFT_TEXT)
1.47 schwarze 1413: putkey(mpage, n->string, m);
1414: }
1415: }
1.16 schwarze 1416:
1.47 schwarze 1417: static void
1.143 schwarze 1418: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1419: const struct roff_node *n)
1.47 schwarze 1420: {
1.142 schwarze 1421: const struct roff_node *head, *body;
1.78 schwarze 1422: char *start, *title;
1.47 schwarze 1423: char byte;
1.78 schwarze 1424: size_t sz;
1.16 schwarze 1425:
1.168 schwarze 1426: if (n == NULL)
1.47 schwarze 1427: return;
1.16 schwarze 1428:
1.47 schwarze 1429: /*
1430: * We're only searching for one thing: the first text child in
1431: * the BODY of a NAME section. Since we don't keep track of
1432: * sections in -man, run some hoops to find out whether we're in
1433: * the correct section or not.
1434: */
1.16 schwarze 1435:
1.141 schwarze 1436: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1437: body = n;
1.168 schwarze 1438: if ((head = body->parent->head) != NULL &&
1439: (head = head->child) != NULL &&
1440: head->next == NULL &&
1.141 schwarze 1441: head->type == ROFFT_TEXT &&
1.168 schwarze 1442: strcmp(head->string, "NAME") == 0 &&
1443: body->child != NULL) {
1.2 schwarze 1444:
1.47 schwarze 1445: /*
1446: * Suck the entire NAME section into memory.
1447: * Yes, we might run away.
1448: * But too many manuals have big, spread-out
1449: * NAME sections over many lines.
1450: */
1.2 schwarze 1451:
1.78 schwarze 1452: title = NULL;
1.147 schwarze 1453: deroff(&title, body);
1.47 schwarze 1454: if (NULL == title)
1455: return;
1.16 schwarze 1456:
1.98 schwarze 1457: /*
1.47 schwarze 1458: * Go through a special heuristic dance here.
1459: * Conventionally, one or more manual names are
1460: * comma-specified prior to a whitespace, then a
1461: * dash, then a description. Try to puzzle out
1462: * the name parts here.
1463: */
1.16 schwarze 1464:
1.78 schwarze 1465: start = title;
1.47 schwarze 1466: for ( ;; ) {
1467: sz = strcspn(start, " ,");
1468: if ('\0' == start[sz])
1469: break;
1.1 schwarze 1470:
1.47 schwarze 1471: byte = start[sz];
1472: start[sz] = '\0';
1.67 schwarze 1473:
1474: /*
1475: * Assume a stray trailing comma in the
1476: * name list if a name begins with a dash.
1477: */
1478:
1479: if ('-' == start[0] ||
1480: ('\\' == start[0] && '-' == start[1]))
1481: break;
1.1 schwarze 1482:
1.90 schwarze 1483: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1484: if ( ! (mpage->name_head_done ||
1485: strcasecmp(start, meta->title))) {
1486: putkey(mpage, start, NAME_HEAD);
1487: mpage->name_head_done = 1;
1488: }
1.1 schwarze 1489:
1.47 schwarze 1490: if (' ' == byte) {
1491: start += sz + 1;
1492: break;
1493: }
1.1 schwarze 1494:
1.47 schwarze 1495: assert(',' == byte);
1496: start += sz + 1;
1497: while (' ' == *start)
1498: start++;
1499: }
1.1 schwarze 1500:
1.78 schwarze 1501: if (start == title) {
1.90 schwarze 1502: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1503: if ( ! (mpage->name_head_done ||
1504: strcasecmp(start, meta->title))) {
1505: putkey(mpage, start, NAME_HEAD);
1506: mpage->name_head_done = 1;
1507: }
1.47 schwarze 1508: free(title);
1509: return;
1510: }
1.1 schwarze 1511:
1.47 schwarze 1512: while (isspace((unsigned char)*start))
1513: start++;
1.1 schwarze 1514:
1.47 schwarze 1515: if (0 == strncmp(start, "-", 1))
1516: start += 1;
1517: else if (0 == strncmp(start, "\\-\\-", 4))
1518: start += 4;
1519: else if (0 == strncmp(start, "\\-", 2))
1520: start += 2;
1521: else if (0 == strncmp(start, "\\(en", 4))
1522: start += 4;
1523: else if (0 == strncmp(start, "\\(em", 4))
1524: start += 4;
1.1 schwarze 1525:
1.47 schwarze 1526: while (' ' == *start)
1527: start++;
1.1 schwarze 1528:
1.47 schwarze 1529: mpage->desc = mandoc_strdup(start);
1530: free(title);
1531: return;
1532: }
1533: }
1.1 schwarze 1534:
1.47 schwarze 1535: for (n = n->child; n; n = n->next) {
1536: if (NULL != mpage->desc)
1537: break;
1.129 schwarze 1538: parse_man(mpage, meta, n);
1.1 schwarze 1539: }
1540: }
1541:
1542: static void
1.143 schwarze 1543: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1544: const struct roff_node *n)
1.1 schwarze 1545: {
1546:
1.47 schwarze 1547: assert(NULL != n);
1548: for (n = n->child; NULL != n; n = n->next) {
1549: switch (n->type) {
1.141 schwarze 1550: case ROFFT_ELEM:
1551: case ROFFT_BLOCK:
1552: case ROFFT_HEAD:
1553: case ROFFT_BODY:
1554: case ROFFT_TAIL:
1.47 schwarze 1555: if (NULL != mdocs[n->tok].fp)
1.127 schwarze 1556: if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1.47 schwarze 1557: break;
1558: if (mdocs[n->tok].mask)
1559: putmdockey(mpage, n->child,
1560: mdocs[n->tok].mask);
1561: break;
1562: default:
1.141 schwarze 1563: assert(n->type != ROFFT_ROOT);
1.47 schwarze 1564: continue;
1565: }
1566: if (NULL != n->child)
1.127 schwarze 1567: parse_mdoc(mpage, meta, n);
1.1 schwarze 1568: }
1569: }
1570:
1.19 schwarze 1571: static int
1.143 schwarze 1572: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1573: const struct roff_node *n)
1.1 schwarze 1574: {
1.131 schwarze 1575: char *start, *end;
1.1 schwarze 1576: size_t sz;
1.19 schwarze 1577:
1.47 schwarze 1578: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1579: NULL == (n = n->child) ||
1.141 schwarze 1580: n->type != ROFFT_TEXT)
1.150 schwarze 1581: return 0;
1.1 schwarze 1582:
1583: /*
1584: * Only consider those `Fd' macro fields that begin with an
1585: * "inclusion" token (versus, e.g., #define).
1586: */
1.47 schwarze 1587:
1.1 schwarze 1588: if (strcmp("#include", n->string))
1.150 schwarze 1589: return 0;
1.1 schwarze 1590:
1.141 schwarze 1591: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1592: return 0;
1.1 schwarze 1593:
1594: /*
1595: * Strip away the enclosing angle brackets and make sure we're
1596: * not zero-length.
1597: */
1598:
1599: start = n->string;
1600: if ('<' == *start || '"' == *start)
1601: start++;
1602:
1603: if (0 == (sz = strlen(start)))
1.150 schwarze 1604: return 0;
1.1 schwarze 1605:
1606: end = &start[(int)sz - 1];
1607: if ('>' == *end || '"' == *end)
1608: end--;
1609:
1.47 schwarze 1610: if (end > start)
1611: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1612: return 0;
1.1 schwarze 1613: }
1614:
1.133 schwarze 1615: static void
1.142 schwarze 1616: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1617: {
1.69 schwarze 1618: char *cp;
1.133 schwarze 1619: size_t sz;
1.1 schwarze 1620:
1.141 schwarze 1621: if (n->type != ROFFT_TEXT)
1.133 schwarze 1622: return;
1.19 schwarze 1623:
1.133 schwarze 1624: /* Skip function pointer punctuation. */
1.1 schwarze 1625:
1.133 schwarze 1626: cp = n->string;
1627: while (*cp == '(' || *cp == '*')
1.1 schwarze 1628: cp++;
1.133 schwarze 1629: sz = strcspn(cp, "()");
1.1 schwarze 1630:
1.133 schwarze 1631: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1632: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1633: putkeys(mpage, cp, sz, NAME_SYN);
1634: }
1.19 schwarze 1635:
1.133 schwarze 1636: static int
1.143 schwarze 1637: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1638: const struct roff_node *n)
1.133 schwarze 1639: {
1640:
1641: if (n->child == NULL)
1.150 schwarze 1642: return 0;
1.133 schwarze 1643:
1644: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1645:
1.133 schwarze 1646: for (n = n->child->next; n != NULL; n = n->next)
1.141 schwarze 1647: if (n->type == ROFFT_TEXT)
1.47 schwarze 1648: putkey(mpage, n->string, TYPE_Fa);
1.19 schwarze 1649:
1.150 schwarze 1650: return 0;
1.128 schwarze 1651: }
1652:
1653: static int
1.143 schwarze 1654: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1655: const struct roff_node *n)
1.128 schwarze 1656: {
1.132 schwarze 1657:
1.141 schwarze 1658: if (n->type != ROFFT_HEAD)
1.150 schwarze 1659: return 1;
1.128 schwarze 1660:
1.133 schwarze 1661: if (n->child != NULL)
1662: parse_mdoc_fname(mpage, n->child);
1663:
1.150 schwarze 1664: return 0;
1.1 schwarze 1665: }
1666:
1.19 schwarze 1667: static int
1.164 schwarze 1668: parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1669: const struct roff_node *n)
1670: {
1671: char *cp;
1672:
1673: if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1674: return 0;
1675:
1.168 schwarze 1676: if (n->child != NULL &&
1677: n->child->next == NULL &&
1678: n->child->type == ROFFT_TEXT)
1.164 schwarze 1679: return 1;
1680:
1681: cp = NULL;
1682: deroff(&cp, n);
1683: if (cp != NULL) {
1684: putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1685: n->type == ROFFT_BODY ? TYPE_Va : 0));
1686: free(cp);
1687: }
1688:
1689: return 0;
1690: }
1691:
1692: static int
1.143 schwarze 1693: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1694: const struct roff_node *n)
1.1 schwarze 1695: {
1.47 schwarze 1696: char *cp;
1.1 schwarze 1697:
1698: if (NULL == (n = n->child))
1.150 schwarze 1699: return 0;
1.1 schwarze 1700:
1.47 schwarze 1701: if (NULL == n->next) {
1702: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1703: return 0;
1.47 schwarze 1704: }
1.1 schwarze 1705:
1.77 schwarze 1706: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1707: putkey(mpage, cp, TYPE_Xr);
1708: free(cp);
1.150 schwarze 1709: return 0;
1.1 schwarze 1710: }
1711:
1.19 schwarze 1712: static int
1.143 schwarze 1713: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1714: const struct roff_node *n)
1.1 schwarze 1715: {
1716:
1.141 schwarze 1717: if (n->type == ROFFT_BODY)
1.147 schwarze 1718: deroff(&mpage->desc, n);
1.150 schwarze 1719: return 0;
1.1 schwarze 1720: }
1721:
1.19 schwarze 1722: static int
1.143 schwarze 1723: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1724: const struct roff_node *n)
1.1 schwarze 1725: {
1726:
1.86 schwarze 1727: if (SEC_NAME == n->sec)
1.90 schwarze 1728: putmdockey(mpage, n->child, NAME_TITLE);
1.141 schwarze 1729: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1730: if (n->child == NULL)
1731: putkey(mpage, meta->name, NAME_SYN);
1732: else
1733: putmdockey(mpage, n->child, NAME_SYN);
1.129 schwarze 1734: }
1735: if ( ! (mpage->name_head_done ||
1736: n->child == NULL || n->child->string == NULL ||
1737: strcasecmp(n->child->string, meta->title))) {
1.141 schwarze 1738: putkey(mpage, n->child->string, ROFFT_HEAD);
1.129 schwarze 1739: mpage->name_head_done = 1;
1.127 schwarze 1740: }
1.150 schwarze 1741: return 0;
1.1 schwarze 1742: }
1743:
1.19 schwarze 1744: static int
1.143 schwarze 1745: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1746: const struct roff_node *n)
1.1 schwarze 1747: {
1748:
1.150 schwarze 1749: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1750: }
1751:
1.47 schwarze 1752: static int
1.143 schwarze 1753: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1754: const struct roff_node *n)
1.1 schwarze 1755: {
1756:
1.150 schwarze 1757: return n->type == ROFFT_HEAD;
1.1 schwarze 1758: }
1759:
1.47 schwarze 1760: /*
1761: * Add a string to the hash table for the current manual.
1762: * Each string has a bitmask telling which macros it belongs to.
1763: * When we finish the manual, we'll dump the table.
1764: */
1.1 schwarze 1765: static void
1.131 schwarze 1766: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1767: {
1.90 schwarze 1768: struct ohash *htab;
1.47 schwarze 1769: struct str *s;
1.68 schwarze 1770: const char *end;
1.47 schwarze 1771: unsigned int slot;
1.131 schwarze 1772: int i, mustfree;
1.1 schwarze 1773:
1.47 schwarze 1774: if (0 == sz)
1775: return;
1.68 schwarze 1776:
1.131 schwarze 1777: mustfree = render_string(&cp, &sz);
1778:
1.90 schwarze 1779: if (TYPE_Nm & v) {
1780: htab = &names;
1781: v &= name_mask;
1.124 schwarze 1782: if (v & NAME_FIRST)
1783: name_mask &= ~NAME_FIRST;
1.90 schwarze 1784: if (debug > 1)
1785: say(mpage->mlinks->file,
1.130 schwarze 1786: "Adding name %*s, bits=%d", sz, cp, v);
1.90 schwarze 1787: } else {
1788: htab = &strings;
1789: if (debug > 1)
1790: for (i = 0; i < mansearch_keymax; i++)
1.118 schwarze 1791: if ((uint64_t)1 << i & v)
1.90 schwarze 1792: say(mpage->mlinks->file,
1793: "Adding key %s=%*s",
1794: mansearch_keynames[i], sz, cp);
1.68 schwarze 1795: }
1.47 schwarze 1796:
1797: end = cp + sz;
1.90 schwarze 1798: slot = ohash_qlookupi(htab, cp, &end);
1799: s = ohash_find(htab, slot);
1.1 schwarze 1800:
1.47 schwarze 1801: if (NULL != s && mpage == s->mpage) {
1802: s->mask |= v;
1.1 schwarze 1803: return;
1.47 schwarze 1804: } else if (NULL == s) {
1.101 schwarze 1805: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1806: memcpy(s->key, cp, sz);
1.90 schwarze 1807: ohash_insert(htab, slot, s);
1.47 schwarze 1808: }
1809: s->mpage = mpage;
1810: s->mask = v;
1.131 schwarze 1811:
1812: if (mustfree)
1813: free(cp);
1.1 schwarze 1814: }
1815:
1816: /*
1.47 schwarze 1817: * Take a Unicode codepoint and produce its UTF-8 encoding.
1818: * This isn't the best way to do this, but it works.
1819: * The magic numbers are from the UTF-8 packaging.
1820: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1821: */
1.47 schwarze 1822: static size_t
1823: utf8(unsigned int cp, char out[7])
1.1 schwarze 1824: {
1.47 schwarze 1825: size_t rc;
1.1 schwarze 1826:
1.47 schwarze 1827: rc = 0;
1828: if (cp <= 0x0000007F) {
1829: rc = 1;
1830: out[0] = (char)cp;
1831: } else if (cp <= 0x000007FF) {
1832: rc = 2;
1833: out[0] = (cp >> 6 & 31) | 192;
1834: out[1] = (cp & 63) | 128;
1835: } else if (cp <= 0x0000FFFF) {
1836: rc = 3;
1837: out[0] = (cp >> 12 & 15) | 224;
1838: out[1] = (cp >> 6 & 63) | 128;
1839: out[2] = (cp & 63) | 128;
1840: } else if (cp <= 0x001FFFFF) {
1841: rc = 4;
1842: out[0] = (cp >> 18 & 7) | 240;
1843: out[1] = (cp >> 12 & 63) | 128;
1844: out[2] = (cp >> 6 & 63) | 128;
1845: out[3] = (cp & 63) | 128;
1846: } else if (cp <= 0x03FFFFFF) {
1847: rc = 5;
1848: out[0] = (cp >> 24 & 3) | 248;
1849: out[1] = (cp >> 18 & 63) | 128;
1850: out[2] = (cp >> 12 & 63) | 128;
1851: out[3] = (cp >> 6 & 63) | 128;
1852: out[4] = (cp & 63) | 128;
1853: } else if (cp <= 0x7FFFFFFF) {
1854: rc = 6;
1855: out[0] = (cp >> 30 & 1) | 252;
1856: out[1] = (cp >> 24 & 63) | 128;
1857: out[2] = (cp >> 18 & 63) | 128;
1858: out[3] = (cp >> 12 & 63) | 128;
1859: out[4] = (cp >> 6 & 63) | 128;
1860: out[5] = (cp & 63) | 128;
1861: } else
1.150 schwarze 1862: return 0;
1.19 schwarze 1863:
1.47 schwarze 1864: out[rc] = '\0';
1.150 schwarze 1865: return rc;
1.1 schwarze 1866: }
1867:
1.47 schwarze 1868: /*
1.131 schwarze 1869: * If the string contains escape sequences,
1870: * replace it with an allocated rendering and return 1,
1871: * such that the caller can free it after use.
1872: * Otherwise, do nothing and return 0.
1.47 schwarze 1873: */
1.131 schwarze 1874: static int
1875: render_string(char **public, size_t *psz)
1.1 schwarze 1876: {
1.131 schwarze 1877: const char *src, *scp, *addcp, *seq;
1878: char *dst;
1879: size_t ssz, dsz, addsz;
1.71 schwarze 1880: char utfbuf[7], res[6];
1.131 schwarze 1881: int seqlen, unicode;
1.47 schwarze 1882:
1883: res[0] = '\\';
1884: res[1] = '\t';
1885: res[2] = ASCII_NBRSP;
1886: res[3] = ASCII_HYPH;
1.71 schwarze 1887: res[4] = ASCII_BREAK;
1888: res[5] = '\0';
1.1 schwarze 1889:
1.131 schwarze 1890: src = scp = *public;
1891: ssz = *psz;
1892: dst = NULL;
1893: dsz = 0;
1894:
1895: while (scp < src + *psz) {
1896:
1897: /* Leave normal characters unchanged. */
1898:
1899: if (strchr(res, *scp) == NULL) {
1900: if (dst != NULL)
1901: dst[dsz++] = *scp;
1902: scp++;
1903: continue;
1904: }
1.1 schwarze 1905:
1.47 schwarze 1906: /*
1.131 schwarze 1907: * Found something that requires replacing,
1908: * make sure we have a destination buffer.
1.47 schwarze 1909: */
1.131 schwarze 1910:
1911: if (dst == NULL) {
1912: dst = mandoc_malloc(ssz + 1);
1913: dsz = scp - src;
1914: memcpy(dst, src, dsz);
1.47 schwarze 1915: }
1.39 schwarze 1916:
1.131 schwarze 1917: /* Handle single-char special characters. */
1918:
1919: switch (*scp) {
1920: case '\\':
1921: break;
1.98 schwarze 1922: case '\t':
1923: case ASCII_NBRSP:
1.131 schwarze 1924: dst[dsz++] = ' ';
1925: scp++;
1926: continue;
1927: case ASCII_HYPH:
1928: dst[dsz++] = '-';
1.71 schwarze 1929: /* FALLTHROUGH */
1.98 schwarze 1930: case ASCII_BREAK:
1.131 schwarze 1931: scp++;
1.47 schwarze 1932: continue;
1.71 schwarze 1933: default:
1.131 schwarze 1934: abort();
1.71 schwarze 1935: }
1.39 schwarze 1936:
1.47 schwarze 1937: /*
1.131 schwarze 1938: * Found an escape sequence.
1939: * Read past the slash, then parse it.
1940: * Ignore everything except characters.
1.47 schwarze 1941: */
1.52 schwarze 1942:
1.131 schwarze 1943: scp++;
1944: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1945: continue;
1.39 schwarze 1946:
1.47 schwarze 1947: /*
1.52 schwarze 1948: * Render the special character
1949: * as either UTF-8 or ASCII.
1.47 schwarze 1950: */
1.52 schwarze 1951:
1952: if (write_utf8) {
1.157 schwarze 1953: unicode = mchars_spec2cp(seq, seqlen);
1.131 schwarze 1954: if (unicode <= 0)
1.52 schwarze 1955: continue;
1.131 schwarze 1956: addsz = utf8(unicode, utfbuf);
1957: if (addsz == 0)
1.52 schwarze 1958: continue;
1.131 schwarze 1959: addcp = utfbuf;
1.52 schwarze 1960: } else {
1.157 schwarze 1961: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.131 schwarze 1962: if (addcp == NULL)
1.52 schwarze 1963: continue;
1.131 schwarze 1964: if (*addcp == ASCII_NBRSP) {
1965: addcp = " ";
1966: addsz = 1;
1.52 schwarze 1967: }
1968: }
1.1 schwarze 1969:
1.47 schwarze 1970: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1971:
1.131 schwarze 1972: ssz += addsz;
1973: dst = mandoc_realloc(dst, ssz + 1);
1974: memcpy(dst + dsz, addcp, addsz);
1975: dsz += addsz;
1976: }
1977: if (dst != NULL) {
1978: *public = dst;
1979: *psz = dsz;
1980: }
1981:
1982: /* Trim trailing whitespace and NUL-terminate. */
1983:
1984: while (*psz > 0 && (*public)[*psz - 1] == ' ')
1985: --*psz;
1986: if (dst != NULL) {
1987: (*public)[*psz] = '\0';
1.150 schwarze 1988: return 1;
1.131 schwarze 1989: } else
1.150 schwarze 1990: return 0;
1.1 schwarze 1991: }
1992:
1.75 schwarze 1993: static void
1994: dbadd_mlink(const struct mlink *mlink)
1995: {
1996: size_t i;
1997:
1998: i = 1;
1999: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
2000: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
2001: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.94 schwarze 2002: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1.75 schwarze 2003: SQL_STEP(stmts[STMT_INSERT_LINK]);
2004: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1.124 schwarze 2005: }
2006:
2007: static void
2008: dbadd_mlink_name(const struct mlink *mlink)
2009: {
1.130 schwarze 2010: uint64_t bits;
1.124 schwarze 2011: size_t i;
2012:
2013: dbadd_mlink(mlink);
1.115 schwarze 2014:
2015: i = 1;
1.130 schwarze 2016: SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
2017: bits = NAME_FILE & NAME_MASK;
2018: if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
2019: bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
2020: sqlite3_reset(stmts[STMT_SELECT_NAME]);
2021: }
2022:
2023: i = 1;
2024: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
1.115 schwarze 2025: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2026: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2027: SQL_STEP(stmts[STMT_INSERT_NAME]);
2028: sqlite3_reset(stmts[STMT_INSERT_NAME]);
1.75 schwarze 2029: }
2030:
1.11 schwarze 2031: /*
1.47 schwarze 2032: * Flush the current page's terms (and their bits) into the database.
2033: * Wrap the entire set of additions in a transaction to make sqlite be a
2034: * little faster.
1.53 schwarze 2035: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 2036: */
2037: static void
1.131 schwarze 2038: dbadd(struct mpage *mpage)
1.11 schwarze 2039: {
1.47 schwarze 2040: struct mlink *mlink;
2041: struct str *key;
1.131 schwarze 2042: char *cp;
1.47 schwarze 2043: size_t i;
2044: unsigned int slot;
1.131 schwarze 2045: int mustfree;
1.47 schwarze 2046:
1.85 schwarze 2047: mlink = mpage->mlinks;
1.11 schwarze 2048:
1.85 schwarze 2049: if (nodb) {
1.104 schwarze 2050: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2051: key = ohash_next(&names, &slot))
1.104 schwarze 2052: free(key);
2053: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2054: key = ohash_next(&strings, &slot))
1.104 schwarze 2055: free(key);
1.102 schwarze 2056: if (0 == debug)
2057: return;
1.85 schwarze 2058: while (NULL != mlink) {
2059: fputs(mlink->name, stdout);
2060: if (NULL == mlink->next ||
2061: strcmp(mlink->dsec, mlink->next->dsec) ||
2062: strcmp(mlink->fsec, mlink->next->fsec) ||
2063: strcmp(mlink->arch, mlink->next->arch)) {
2064: putchar('(');
2065: if ('\0' == *mlink->dsec)
2066: fputs(mlink->fsec, stdout);
2067: else
2068: fputs(mlink->dsec, stdout);
2069: if ('\0' != *mlink->arch)
2070: printf("/%s", mlink->arch);
2071: putchar(')');
2072: }
2073: mlink = mlink->next;
2074: if (NULL != mlink)
2075: fputs(", ", stdout);
2076: }
1.89 schwarze 2077: printf(" - %s\n", mpage->desc);
1.11 schwarze 2078: return;
1.85 schwarze 2079: }
2080:
2081: if (debug)
2082: say(mlink->file, "Adding to database");
1.47 schwarze 2083:
1.131 schwarze 2084: cp = mpage->desc;
2085: i = strlen(cp);
2086: mustfree = render_string(&cp, &i);
1.47 schwarze 2087: i = 1;
1.131 schwarze 2088: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
1.116 schwarze 2089: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
1.47 schwarze 2090: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.94 schwarze 2091: mpage->pageid = sqlite3_last_insert_rowid(db);
1.47 schwarze 2092: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1.131 schwarze 2093: if (mustfree)
2094: free(cp);
1.47 schwarze 2095:
1.85 schwarze 2096: while (NULL != mlink) {
1.75 schwarze 2097: dbadd_mlink(mlink);
1.85 schwarze 2098: mlink = mlink->next;
2099: }
1.91 schwarze 2100: mlink = mpage->mlinks;
1.47 schwarze 2101:
1.90 schwarze 2102: for (key = ohash_first(&names, &slot); NULL != key;
2103: key = ohash_next(&names, &slot)) {
2104: assert(key->mpage == mpage);
2105: i = 1;
2106: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
1.131 schwarze 2107: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
1.94 schwarze 2108: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
1.90 schwarze 2109: SQL_STEP(stmts[STMT_INSERT_NAME]);
2110: sqlite3_reset(stmts[STMT_INSERT_NAME]);
2111: free(key);
2112: }
1.47 schwarze 2113: for (key = ohash_first(&strings, &slot); NULL != key;
2114: key = ohash_next(&strings, &slot)) {
2115: assert(key->mpage == mpage);
2116: i = 1;
2117: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.131 schwarze 2118: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
1.94 schwarze 2119: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
1.47 schwarze 2120: SQL_STEP(stmts[STMT_INSERT_KEY]);
2121: sqlite3_reset(stmts[STMT_INSERT_KEY]);
2122: free(key);
1.33 schwarze 2123: }
1.47 schwarze 2124: }
1.41 deraadt 2125:
1.47 schwarze 2126: static void
2127: dbprune(void)
2128: {
2129: struct mpage *mpage;
2130: struct mlink *mlink;
2131: size_t i;
2132: unsigned int slot;
1.11 schwarze 2133:
1.63 schwarze 2134: if (0 == nodb)
2135: SQL_EXEC("BEGIN TRANSACTION");
1.47 schwarze 2136:
1.63 schwarze 2137: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2138: mpage = ohash_next(&mpages, &slot)) {
1.47 schwarze 2139: mlink = mpage->mlinks;
1.82 schwarze 2140: if (debug)
1.63 schwarze 2141: say(mlink->file, "Deleting from database");
2142: if (nodb)
2143: continue;
2144: for ( ; NULL != mlink; mlink = mlink->next) {
2145: i = 1;
2146: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2147: i, mlink->dsec);
2148: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2149: i, mlink->arch);
2150: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2151: i, mlink->name);
2152: SQL_STEP(stmts[STMT_DELETE_PAGE]);
2153: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2154: }
1.11 schwarze 2155: }
1.63 schwarze 2156:
2157: if (0 == nodb)
2158: SQL_EXEC("END TRANSACTION");
1.47 schwarze 2159: }
1.22 schwarze 2160:
1.47 schwarze 2161: /*
2162: * Close an existing database and its prepared statements.
2163: * If "real" is not set, rename the temporary file into the real one.
2164: */
2165: static void
2166: dbclose(int real)
2167: {
2168: size_t i;
1.72 schwarze 2169: int status;
2170: pid_t child;
1.11 schwarze 2171:
1.47 schwarze 2172: if (nodb)
2173: return;
1.11 schwarze 2174:
1.47 schwarze 2175: for (i = 0; i < STMT__MAX; i++) {
2176: sqlite3_finalize(stmts[i]);
2177: stmts[i] = NULL;
1.28 schwarze 2178: }
1.22 schwarze 2179:
1.47 schwarze 2180: sqlite3_close(db);
2181: db = NULL;
1.11 schwarze 2182:
1.47 schwarze 2183: if (real)
2184: return;
1.22 schwarze 2185:
1.72 schwarze 2186: if ('\0' == *tempfilename) {
2187: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2188: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2189: say(MANDOC_DB, "&rename");
1.72 schwarze 2190: }
2191: return;
2192: }
2193:
2194: switch (child = fork()) {
1.98 schwarze 2195: case -1:
1.72 schwarze 2196: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2197: say("", "&fork cmp");
1.72 schwarze 2198: return;
1.98 schwarze 2199: case 0:
1.72 schwarze 2200: execlp("cmp", "cmp", "-s",
1.149 schwarze 2201: tempfilename, MANDOC_DB, (char *)NULL);
1.80 schwarze 2202: say("", "&exec cmp");
1.72 schwarze 2203: exit(0);
2204: default:
2205: break;
2206: }
2207: if (-1 == waitpid(child, &status, 0)) {
2208: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2209: say("", "&wait cmp");
1.72 schwarze 2210: } else if (WIFSIGNALED(status)) {
2211: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2212: say("", "cmp died from signal %d", WTERMSIG(status));
1.72 schwarze 2213: } else if (WEXITSTATUS(status)) {
1.47 schwarze 2214: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2215: say(MANDOC_DB,
2216: "Data changed, but cannot replace database");
2217: }
2218:
2219: *strrchr(tempfilename, '/') = '\0';
2220: switch (child = fork()) {
1.98 schwarze 2221: case -1:
1.72 schwarze 2222: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2223: say("", "&fork rm");
1.72 schwarze 2224: return;
1.98 schwarze 2225: case 0:
1.149 schwarze 2226: execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
1.80 schwarze 2227: say("", "&exec rm");
1.72 schwarze 2228: exit((int)MANDOCLEVEL_SYSERR);
2229: default:
2230: break;
2231: }
2232: if (-1 == waitpid(child, &status, 0)) {
2233: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2234: say("", "&wait rm");
1.72 schwarze 2235: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2236: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2237: say("", "%s: Cannot remove temporary directory",
2238: tempfilename);
1.22 schwarze 2239: }
1.11 schwarze 2240: }
2241:
1.47 schwarze 2242: /*
2243: * This is straightforward stuff.
2244: * Open a database connection to a "temporary" database, then open a set
2245: * of prepared statements we'll use over and over again.
2246: * If "real" is set, we use the existing database; if not, we truncate a
2247: * temporary one.
2248: * Must be matched by dbclose().
2249: */
2250: static int
2251: dbopen(int real)
1.2 schwarze 2252: {
1.72 schwarze 2253: const char *sql;
1.47 schwarze 2254: int rc, ofl;
1.6 schwarze 2255:
1.98 schwarze 2256: if (nodb)
1.150 schwarze 2257: return 1;
1.6 schwarze 2258:
1.72 schwarze 2259: *tempfilename = '\0';
1.47 schwarze 2260: ofl = SQLITE_OPEN_READWRITE;
1.72 schwarze 2261:
2262: if (real) {
2263: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2264: if (SQLITE_OK != rc) {
1.47 schwarze 2265: exitcode = (int)MANDOCLEVEL_SYSERR;
1.108 sthen 2266: if (SQLITE_CANTOPEN != rc)
2267: say(MANDOC_DB, "%s", sqlite3_errstr(rc));
1.150 schwarze 2268: return 0;
1.28 schwarze 2269: }
1.72 schwarze 2270: goto prepare_statements;
2271: }
2272:
2273: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.6 schwarze 2274:
1.72 schwarze 2275: remove(MANDOC_DB "~");
2276: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.98 schwarze 2277: if (SQLITE_OK == rc)
1.72 schwarze 2278: goto create_tables;
1.73 schwarze 2279: if (MPARSE_QUICK & mparse_options) {
1.47 schwarze 2280: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2281: say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
1.150 schwarze 2282: return 0;
1.47 schwarze 2283: }
1.6 schwarze 2284:
1.100 schwarze 2285: (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2286: sizeof(tempfilename));
1.72 schwarze 2287: if (NULL == mkdtemp(tempfilename)) {
2288: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2289: say("", "&%s", tempfilename);
1.150 schwarze 2290: return 0;
1.72 schwarze 2291: }
1.100 schwarze 2292: (void)strlcat(tempfilename, "/" MANDOC_DB,
2293: sizeof(tempfilename));
1.72 schwarze 2294: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2295: if (SQLITE_OK != rc) {
1.47 schwarze 2296: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2297: say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
1.150 schwarze 2298: return 0;
1.2 schwarze 2299: }
2300:
1.72 schwarze 2301: create_tables:
1.47 schwarze 2302: sql = "CREATE TABLE \"mpages\" (\n"
1.89 schwarze 2303: " \"desc\" TEXT NOT NULL,\n"
1.47 schwarze 2304: " \"form\" INTEGER NOT NULL,\n"
1.94 schwarze 2305: " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1.47 schwarze 2306: ");\n"
2307: "\n"
2308: "CREATE TABLE \"mlinks\" (\n"
2309: " \"sec\" TEXT NOT NULL,\n"
2310: " \"arch\" TEXT NOT NULL,\n"
2311: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2312: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2313: "ON DELETE CASCADE\n"
1.47 schwarze 2314: ");\n"
1.93 schwarze 2315: "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
1.47 schwarze 2316: "\n"
1.90 schwarze 2317: "CREATE TABLE \"names\" (\n"
2318: " \"bits\" INTEGER NOT NULL,\n"
2319: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2320: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.130 schwarze 2321: "ON DELETE CASCADE,\n"
2322: " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
1.90 schwarze 2323: ");\n"
2324: "\n"
1.47 schwarze 2325: "CREATE TABLE \"keys\" (\n"
2326: " \"bits\" INTEGER NOT NULL,\n"
2327: " \"key\" TEXT NOT NULL,\n"
1.94 schwarze 2328: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2329: "ON DELETE CASCADE\n"
1.93 schwarze 2330: ");\n"
2331: "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
1.47 schwarze 2332:
2333: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2334: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2335: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.103 schwarze 2336: sqlite3_close(db);
1.150 schwarze 2337: return 0;
1.2 schwarze 2338: }
2339:
1.47 schwarze 2340: prepare_statements:
1.103 schwarze 2341: if (SQLITE_OK != sqlite3_exec(db,
2342: "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2343: exitcode = (int)MANDOCLEVEL_SYSERR;
2344: say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2345: sqlite3_errmsg(db));
2346: sqlite3_close(db);
1.150 schwarze 2347: return 0;
1.103 schwarze 2348: }
2349:
1.94 schwarze 2350: sql = "DELETE FROM mpages WHERE pageid IN "
1.63 schwarze 2351: "(SELECT pageid FROM mlinks WHERE "
2352: "sec=? AND arch=? AND name=?)";
1.47 schwarze 2353: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2354: sql = "INSERT INTO mpages "
1.89 schwarze 2355: "(desc,form) VALUES (?,?)";
1.47 schwarze 2356: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2357: sql = "INSERT INTO mlinks "
1.61 schwarze 2358: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.47 schwarze 2359: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.130 schwarze 2360: sql = "SELECT bits FROM names where pageid = ?";
2361: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
1.90 schwarze 2362: sql = "INSERT INTO names "
2363: "(bits,name,pageid) VALUES (?,?,?)";
2364: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
1.47 schwarze 2365: sql = "INSERT INTO keys "
2366: "(bits,key,pageid) VALUES (?,?,?)";
2367: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.6 schwarze 2368:
1.47 schwarze 2369: /*
2370: * When opening a new database, we can turn off
2371: * synchronous mode for much better performance.
2372: */
1.6 schwarze 2373:
1.103 schwarze 2374: if (real && SQLITE_OK != sqlite3_exec(db,
2375: "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2376: exitcode = (int)MANDOCLEVEL_SYSERR;
2377: say(MANDOC_DB, "PRAGMA synchronous: %s",
2378: sqlite3_errmsg(db));
2379: sqlite3_close(db);
1.150 schwarze 2380: return 0;
1.103 schwarze 2381: }
1.11 schwarze 2382:
1.150 schwarze 2383: return 1;
1.47 schwarze 2384: }
1.6 schwarze 2385:
1.47 schwarze 2386: static int
1.120 schwarze 2387: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2388: {
2389: static char startdir[PATH_MAX];
1.110 schwarze 2390: static int getcwd_status; /* 1 = ok, 2 = failure */
2391: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2392: char *cp;
1.6 schwarze 2393:
1.47 schwarze 2394: /*
1.110 schwarze 2395: * Remember the original working directory, if possible.
2396: * This will be needed if the second or a later directory
2397: * on the command line is given as a relative path.
2398: * Do not error out if the current directory is not
2399: * searchable: Maybe it won't be needed after all.
2400: */
2401: if (0 == getcwd_status) {
2402: if (NULL == getcwd(startdir, sizeof(startdir))) {
2403: getcwd_status = 2;
2404: (void)strlcpy(startdir, strerror(errno),
2405: sizeof(startdir));
2406: } else
2407: getcwd_status = 1;
2408: }
2409:
2410: /*
2411: * We are leaving the old base directory.
2412: * Do not use it any longer, not even for messages.
2413: */
2414: *basedir = '\0';
2415:
2416: /*
2417: * If and only if the directory was changed earlier and
2418: * the next directory to process is given as a relative path,
2419: * first go back, or bail out if that is impossible.
1.47 schwarze 2420: */
1.110 schwarze 2421: if (chdir_status && '/' != *targetdir) {
2422: if (2 == getcwd_status) {
1.47 schwarze 2423: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2424: say("", "getcwd: %s", startdir);
1.150 schwarze 2425: return 0;
1.47 schwarze 2426: }
1.110 schwarze 2427: if (-1 == chdir(startdir)) {
1.47 schwarze 2428: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2429: say("", "&chdir %s", startdir);
1.150 schwarze 2430: return 0;
1.2 schwarze 2431: }
2432: }
1.110 schwarze 2433:
2434: /*
2435: * Always resolve basedir to the canonicalized absolute
2436: * pathname and append a trailing slash, such that
2437: * we can reliably check whether files are inside.
2438: */
1.47 schwarze 2439: if (NULL == realpath(targetdir, basedir)) {
1.120 schwarze 2440: if (report_baddir || errno != ENOENT) {
2441: exitcode = (int)MANDOCLEVEL_BADARG;
2442: say("", "&%s: realpath", targetdir);
2443: }
1.150 schwarze 2444: return 0;
1.47 schwarze 2445: } else if (-1 == chdir(basedir)) {
1.120 schwarze 2446: if (report_baddir || errno != ENOENT) {
2447: exitcode = (int)MANDOCLEVEL_BADARG;
2448: say("", "&chdir");
2449: }
1.150 schwarze 2450: return 0;
1.109 schwarze 2451: }
1.110 schwarze 2452: chdir_status = 1;
1.109 schwarze 2453: cp = strchr(basedir, '\0');
2454: if ('/' != cp[-1]) {
2455: if (cp - basedir >= PATH_MAX - 1) {
2456: exitcode = (int)MANDOCLEVEL_SYSERR;
2457: say("", "Filename too long");
1.150 schwarze 2458: return 0;
1.109 schwarze 2459: }
2460: *cp++ = '/';
2461: *cp = '\0';
1.47 schwarze 2462: }
1.150 schwarze 2463: return 1;
1.2 schwarze 2464: }
2465:
2466: static void
1.47 schwarze 2467: say(const char *file, const char *format, ...)
1.2 schwarze 2468: {
1.47 schwarze 2469: va_list ap;
1.80 schwarze 2470: int use_errno;
1.2 schwarze 2471:
1.47 schwarze 2472: if ('\0' != *basedir)
2473: fprintf(stderr, "%s", basedir);
2474: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2475: fputc('/', stderr);
1.47 schwarze 2476: if ('\0' != *file)
2477: fprintf(stderr, "%s", file);
1.31 schwarze 2478:
1.80 schwarze 2479: use_errno = 1;
2480: if (NULL != format) {
2481: switch (*format) {
1.98 schwarze 2482: case '&':
1.80 schwarze 2483: format++;
2484: break;
1.98 schwarze 2485: case '\0':
1.80 schwarze 2486: format = NULL;
2487: break;
2488: default:
2489: use_errno = 0;
2490: break;
2491: }
2492: }
2493: if (NULL != format) {
2494: if ('\0' != *basedir || '\0' != *file)
2495: fputs(": ", stderr);
2496: va_start(ap, format);
2497: vfprintf(stderr, format, ap);
2498: va_end(ap);
2499: }
2500: if (use_errno) {
2501: if ('\0' != *basedir || '\0' != *file || NULL != format)
2502: fputs(": ", stderr);
1.47 schwarze 2503: perror(NULL);
1.80 schwarze 2504: } else
2505: fputc('\n', stderr);
1.1 schwarze 2506: }
![[BACK]](/icons/back.gif){kind=icon}
Return to mandocdb.c CVS log ![[TXT]](/icons/text.gif){kind=icon}
![[DIR]](/icons/dir.gif){kind=icon}
Up to [local] / src / usr.bin / mandoc