Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.171
1.171 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.170 2016/07/09 15:23:36 schwarze 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: };
1.171 ! schwarze 130:
! 131:
! 132: int mandocdb(int, char *[]);
1.19 schwarze 133:
1.47 schwarze 134: static void dbclose(int);
1.131 schwarze 135: static void dbadd(struct mpage *);
1.75 schwarze 136: static void dbadd_mlink(const struct mlink *mlink);
1.124 schwarze 137: static void dbadd_mlink_name(const struct mlink *mlink);
1.47 schwarze 138: static int dbopen(int);
139: static void dbprune(void);
140: static void filescan(const char *);
141: static void mlink_add(struct mlink *, const struct stat *);
1.88 schwarze 142: static void mlink_check(struct mpage *, struct mlink *);
1.47 schwarze 143: static void mlink_free(struct mlink *);
144: static void mlinks_undupe(struct mpage *);
145: static void mpages_free(void);
1.131 schwarze 146: static void mpages_merge(struct mparse *);
1.87 schwarze 147: static void names_check(void);
1.81 schwarze 148: static void parse_cat(struct mpage *, int);
1.143 schwarze 149: static void parse_man(struct mpage *, const struct roff_meta *,
1.142 schwarze 150: const struct roff_node *);
1.143 schwarze 151: static void parse_mdoc(struct mpage *, const struct roff_meta *,
1.142 schwarze 152: const struct roff_node *);
1.143 schwarze 153: static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
1.142 schwarze 154: const struct roff_node *);
1.143 schwarze 155: static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
1.142 schwarze 156: const struct roff_node *);
157: static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
1.143 schwarze 158: static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
1.142 schwarze 159: const struct roff_node *);
1.143 schwarze 160: static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
1.142 schwarze 161: const struct roff_node *);
1.143 schwarze 162: static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
1.142 schwarze 163: const struct roff_node *);
1.143 schwarze 164: static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
1.142 schwarze 165: const struct roff_node *);
1.143 schwarze 166: static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
1.142 schwarze 167: const struct roff_node *);
1.164 schwarze 168: static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
169: const struct roff_node *);
1.143 schwarze 170: static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
1.142 schwarze 171: const struct roff_node *);
1.69 schwarze 172: static void putkey(const struct mpage *, char *, uint64_t);
1.131 schwarze 173: static void putkeys(const struct mpage *, char *, size_t, uint64_t);
1.47 schwarze 174: static void putmdockey(const struct mpage *,
1.142 schwarze 175: const struct roff_node *, uint64_t);
1.131 schwarze 176: static int render_string(char **, size_t *);
1.47 schwarze 177: static void say(const char *, const char *, ...);
1.120 schwarze 178: static int set_basedir(const char *, int);
1.47 schwarze 179: static int treescan(void);
180: static size_t utf8(unsigned int, char [7]);
181:
1.72 schwarze 182: static char tempfilename[32];
1.59 schwarze 183: static int nodb; /* no database changes */
1.73 schwarze 184: static int mparse_options; /* abort the parse early */
1.98 schwarze 185: static int use_all; /* use all found files */
186: static int debug; /* print what we're doing */
187: static int warnings; /* warn about crap */
1.52 schwarze 188: static int write_utf8; /* write UTF-8 output; else ASCII */
1.47 schwarze 189: static int exitcode; /* to be returned by main */
1.98 schwarze 190: static enum op op; /* operational mode */
1.47 schwarze 191: static char basedir[PATH_MAX]; /* current base directory */
192: static struct ohash mpages; /* table of distinct manual pages */
193: static struct ohash mlinks; /* table of directory entries */
1.90 schwarze 194: static struct ohash names; /* table of all names */
1.47 schwarze 195: static struct ohash strings; /* table of all strings */
196: static sqlite3 *db = NULL; /* current database */
197: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.90 schwarze 198: static uint64_t name_mask;
1.47 schwarze 199:
1.19 schwarze 200: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.47 schwarze 201: { NULL, 0 }, /* Ap */
202: { NULL, 0 }, /* Dd */
203: { NULL, 0 }, /* Dt */
204: { NULL, 0 }, /* Os */
205: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
206: { parse_mdoc_head, TYPE_Ss }, /* Ss */
207: { NULL, 0 }, /* Pp */
208: { NULL, 0 }, /* D1 */
209: { NULL, 0 }, /* Dl */
210: { NULL, 0 }, /* Bd */
211: { NULL, 0 }, /* Ed */
212: { NULL, 0 }, /* Bl */
213: { NULL, 0 }, /* El */
214: { NULL, 0 }, /* It */
215: { NULL, 0 }, /* Ad */
216: { NULL, TYPE_An }, /* An */
217: { NULL, TYPE_Ar }, /* Ar */
218: { NULL, TYPE_Cd }, /* Cd */
219: { NULL, TYPE_Cm }, /* Cm */
220: { NULL, TYPE_Dv }, /* Dv */
221: { NULL, TYPE_Er }, /* Er */
222: { NULL, TYPE_Ev }, /* Ev */
223: { NULL, 0 }, /* Ex */
224: { NULL, TYPE_Fa }, /* Fa */
225: { parse_mdoc_Fd, 0 }, /* Fd */
226: { NULL, TYPE_Fl }, /* Fl */
227: { parse_mdoc_Fn, 0 }, /* Fn */
228: { NULL, TYPE_Ft }, /* Ft */
229: { NULL, TYPE_Ic }, /* Ic */
1.49 schwarze 230: { NULL, TYPE_In }, /* In */
1.47 schwarze 231: { NULL, TYPE_Li }, /* Li */
1.89 schwarze 232: { parse_mdoc_Nd, 0 }, /* Nd */
1.86 schwarze 233: { parse_mdoc_Nm, 0 }, /* Nm */
1.47 schwarze 234: { NULL, 0 }, /* Op */
235: { NULL, 0 }, /* Ot */
236: { NULL, TYPE_Pa }, /* Pa */
237: { NULL, 0 }, /* Rv */
1.49 schwarze 238: { NULL, TYPE_St }, /* St */
1.164 schwarze 239: { parse_mdoc_Va, TYPE_Va }, /* Va */
240: { parse_mdoc_Va, TYPE_Vt }, /* Vt */
1.47 schwarze 241: { parse_mdoc_Xr, 0 }, /* Xr */
242: { NULL, 0 }, /* %A */
243: { NULL, 0 }, /* %B */
244: { NULL, 0 }, /* %D */
245: { NULL, 0 }, /* %I */
246: { NULL, 0 }, /* %J */
247: { NULL, 0 }, /* %N */
248: { NULL, 0 }, /* %O */
249: { NULL, 0 }, /* %P */
250: { NULL, 0 }, /* %R */
251: { NULL, 0 }, /* %T */
252: { NULL, 0 }, /* %V */
253: { NULL, 0 }, /* Ac */
254: { NULL, 0 }, /* Ao */
255: { NULL, 0 }, /* Aq */
256: { NULL, TYPE_At }, /* At */
257: { NULL, 0 }, /* Bc */
258: { NULL, 0 }, /* Bf */
259: { NULL, 0 }, /* Bo */
260: { NULL, 0 }, /* Bq */
261: { NULL, TYPE_Bsx }, /* Bsx */
262: { NULL, TYPE_Bx }, /* Bx */
263: { NULL, 0 }, /* Db */
264: { NULL, 0 }, /* Dc */
265: { NULL, 0 }, /* Do */
266: { NULL, 0 }, /* Dq */
267: { NULL, 0 }, /* Ec */
268: { NULL, 0 }, /* Ef */
269: { NULL, TYPE_Em }, /* Em */
270: { NULL, 0 }, /* Eo */
271: { NULL, TYPE_Fx }, /* Fx */
272: { NULL, TYPE_Ms }, /* Ms */
273: { NULL, 0 }, /* No */
274: { NULL, 0 }, /* Ns */
275: { NULL, TYPE_Nx }, /* Nx */
276: { NULL, TYPE_Ox }, /* Ox */
277: { NULL, 0 }, /* Pc */
278: { NULL, 0 }, /* Pf */
279: { NULL, 0 }, /* Po */
280: { NULL, 0 }, /* Pq */
281: { NULL, 0 }, /* Qc */
282: { NULL, 0 }, /* Ql */
283: { NULL, 0 }, /* Qo */
284: { NULL, 0 }, /* Qq */
285: { NULL, 0 }, /* Re */
286: { NULL, 0 }, /* Rs */
287: { NULL, 0 }, /* Sc */
288: { NULL, 0 }, /* So */
289: { NULL, 0 }, /* Sq */
290: { NULL, 0 }, /* Sm */
291: { NULL, 0 }, /* Sx */
292: { NULL, TYPE_Sy }, /* Sy */
293: { NULL, TYPE_Tn }, /* Tn */
294: { NULL, 0 }, /* Ux */
295: { NULL, 0 }, /* Xc */
296: { NULL, 0 }, /* Xo */
1.128 schwarze 297: { parse_mdoc_Fo, 0 }, /* Fo */
1.47 schwarze 298: { NULL, 0 }, /* Fc */
299: { NULL, 0 }, /* Oo */
300: { NULL, 0 }, /* Oc */
301: { NULL, 0 }, /* Bk */
302: { NULL, 0 }, /* Ek */
303: { NULL, 0 }, /* Bt */
304: { NULL, 0 }, /* Hf */
305: { NULL, 0 }, /* Fr */
306: { NULL, 0 }, /* Ud */
307: { NULL, TYPE_Lb }, /* Lb */
308: { NULL, 0 }, /* Lp */
309: { NULL, TYPE_Lk }, /* Lk */
310: { NULL, TYPE_Mt }, /* Mt */
311: { NULL, 0 }, /* Brq */
312: { NULL, 0 }, /* Bro */
313: { NULL, 0 }, /* Brc */
314: { NULL, 0 }, /* %C */
315: { NULL, 0 }, /* Es */
316: { NULL, 0 }, /* En */
317: { NULL, TYPE_Dx }, /* Dx */
318: { NULL, 0 }, /* %Q */
319: { NULL, 0 }, /* br */
320: { NULL, 0 }, /* sp */
321: { NULL, 0 }, /* %U */
322: { NULL, 0 }, /* Ta */
1.121 schwarze 323: { NULL, 0 }, /* ll */
1.1 schwarze 324: };
325:
1.98 schwarze 326:
1.1 schwarze 327: int
1.3 schwarze 328: mandocdb(int argc, char *argv[])
1.1 schwarze 329: {
1.140 schwarze 330: struct manconf conf;
331: struct mparse *mp;
1.161 schwarze 332: const char *path_arg, *progname;
1.140 schwarze 333: size_t j, sz;
1.47 schwarze 334: int ch, i;
335:
1.158 millert 336: if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) {
1.167 schwarze 337: warn("pledge");
1.155 schwarze 338: return (int)MANDOCLEVEL_SYSERR;
339: }
340:
1.140 schwarze 341: memset(&conf, 0, sizeof(conf));
1.47 schwarze 342: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
343:
344: /*
1.98 schwarze 345: * We accept a few different invocations.
1.47 schwarze 346: * The CHECKOP macro makes sure that invocation styles don't
347: * clobber each other.
348: */
349: #define CHECKOP(_op, _ch) do \
350: if (OP_DEFAULT != (_op)) { \
1.151 schwarze 351: warnx("-%c: Conflicting option", (_ch)); \
1.47 schwarze 352: goto usage; \
353: } while (/*CONSTCOND*/0)
1.10 schwarze 354:
1.47 schwarze 355: path_arg = NULL;
1.28 schwarze 356: op = OP_DEFAULT;
1.1 schwarze 357:
1.83 schwarze 358: while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
1.1 schwarze 359: switch (ch) {
1.98 schwarze 360: case 'a':
1.6 schwarze 361: use_all = 1;
362: break;
1.98 schwarze 363: case 'C':
1.47 schwarze 364: CHECKOP(op, ch);
365: path_arg = optarg;
1.28 schwarze 366: op = OP_CONFFILE;
1.25 schwarze 367: break;
1.98 schwarze 368: case 'D':
1.82 schwarze 369: debug++;
370: break;
1.98 schwarze 371: case 'd':
1.47 schwarze 372: CHECKOP(op, ch);
373: path_arg = optarg;
1.2 schwarze 374: op = OP_UPDATE;
1.1 schwarze 375: break;
1.98 schwarze 376: case 'n':
1.47 schwarze 377: nodb = 1;
378: break;
1.98 schwarze 379: case 'p':
1.83 schwarze 380: warnings = 1;
381: break;
1.98 schwarze 382: case 'Q':
1.73 schwarze 383: mparse_options |= MPARSE_QUICK;
1.59 schwarze 384: break;
1.98 schwarze 385: case 'T':
1.52 schwarze 386: if (strcmp(optarg, "utf8")) {
1.151 schwarze 387: warnx("-T%s: Unsupported output format",
388: optarg);
1.52 schwarze 389: goto usage;
390: }
391: write_utf8 = 1;
392: break;
1.98 schwarze 393: case 't':
1.47 schwarze 394: CHECKOP(op, ch);
1.28 schwarze 395: dup2(STDOUT_FILENO, STDERR_FILENO);
396: op = OP_TEST;
1.47 schwarze 397: nodb = warnings = 1;
1.28 schwarze 398: break;
1.98 schwarze 399: case 'u':
1.47 schwarze 400: CHECKOP(op, ch);
401: path_arg = optarg;
1.1 schwarze 402: op = OP_DELETE;
403: break;
1.98 schwarze 404: case 'v':
1.83 schwarze 405: /* Compatibility with espie@'s makewhatis. */
1.28 schwarze 406: break;
1.1 schwarze 407: default:
1.28 schwarze 408: goto usage;
1.1 schwarze 409: }
410:
411: argc -= optind;
412: argv += optind;
413:
1.163 schwarze 414: if (nodb) {
415: if (pledge("stdio rpath", NULL) == -1) {
1.167 schwarze 416: warn("pledge");
1.163 schwarze 417: return (int)MANDOCLEVEL_SYSERR;
418: }
1.155 schwarze 419: }
420:
1.28 schwarze 421: if (OP_CONFFILE == op && argc > 0) {
1.151 schwarze 422: warnx("-C: Too many arguments");
1.28 schwarze 423: goto usage;
424: }
425:
1.47 schwarze 426: exitcode = (int)MANDOCLEVEL_OK;
1.157 schwarze 427: mchars_alloc();
428: mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
1.156 schwarze 429: mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
430: mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
1.2 schwarze 431:
1.47 schwarze 432: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.2 schwarze 433:
1.47 schwarze 434: /*
1.109 schwarze 435: * Most of these deal with a specific directory.
1.97 schwarze 436: * Jump into that directory first.
1.47 schwarze 437: */
1.120 schwarze 438: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.47 schwarze 439: goto out;
1.97 schwarze 440:
1.95 schwarze 441: if (dbopen(1)) {
1.97 schwarze 442: /*
443: * The existing database is usable. Process
444: * all files specified on the command-line.
445: */
1.163 schwarze 446: if (!nodb) {
447: if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) {
1.167 schwarze 448: warn("pledge");
1.163 schwarze 449: exitcode = (int)MANDOCLEVEL_SYSERR;
450: goto out;
451: }
1.155 schwarze 452: }
1.97 schwarze 453: use_all = 1;
1.95 schwarze 454: for (i = 0; i < argc; i++)
455: filescan(argv[i]);
456: if (OP_TEST != op)
457: dbprune();
458: } else {
459: /*
460: * Database missing or corrupt.
461: * Recreate from scratch.
462: */
1.97 schwarze 463: exitcode = (int)MANDOCLEVEL_OK;
1.95 schwarze 464: op = OP_DEFAULT;
465: if (0 == treescan())
466: goto out;
467: if (0 == dbopen(0))
468: goto out;
469: }
1.47 schwarze 470: if (OP_DELETE != op)
1.131 schwarze 471: mpages_merge(mp);
1.95 schwarze 472: dbclose(OP_DEFAULT == op ? 0 : 1);
1.47 schwarze 473: } else {
474: /*
475: * If we have arguments, use them as our manpaths.
476: * If we don't, grok from manpath(1) or however else
1.140 schwarze 477: * manconf_parse() wants to do it.
1.47 schwarze 478: */
479: if (argc > 0) {
1.140 schwarze 480: conf.manpath.paths = mandoc_reallocarray(NULL,
1.101 schwarze 481: argc, sizeof(char *));
1.140 schwarze 482: conf.manpath.sz = (size_t)argc;
1.47 schwarze 483: for (i = 0; i < argc; i++)
1.140 schwarze 484: conf.manpath.paths[i] = mandoc_strdup(argv[i]);
1.47 schwarze 485: } else
1.140 schwarze 486: manconf_parse(&conf, path_arg, NULL, NULL);
1.84 schwarze 487:
1.140 schwarze 488: if (conf.manpath.sz == 0) {
1.84 schwarze 489: exitcode = (int)MANDOCLEVEL_BADARG;
490: say("", "Empty manpath");
491: }
1.2 schwarze 492:
1.47 schwarze 493: /*
494: * First scan the tree rooted at a base directory, then
495: * build a new database and finally move it into place.
496: * Ignore zero-length directories and strip trailing
497: * slashes.
498: */
1.140 schwarze 499: for (j = 0; j < conf.manpath.sz; j++) {
500: sz = strlen(conf.manpath.paths[j]);
501: if (sz && conf.manpath.paths[j][sz - 1] == '/')
502: conf.manpath.paths[j][--sz] = '\0';
1.47 schwarze 503: if (0 == sz)
504: continue;
1.2 schwarze 505:
1.47 schwarze 506: if (j) {
1.156 schwarze 507: mandoc_ohash_init(&mpages, 6,
508: offsetof(struct mpage, inodev));
509: mandoc_ohash_init(&mlinks, 6,
510: offsetof(struct mlink, file));
1.47 schwarze 511: }
1.1 schwarze 512:
1.140 schwarze 513: if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
1.120 schwarze 514: continue;
1.47 schwarze 515: if (0 == treescan())
1.120 schwarze 516: continue;
1.47 schwarze 517: if (0 == dbopen(0))
1.120 schwarze 518: continue;
1.41 deraadt 519:
1.131 schwarze 520: mpages_merge(mp);
1.104 schwarze 521: if (warnings && !nodb &&
1.87 schwarze 522: ! (MPARSE_QUICK & mparse_options))
523: names_check();
1.47 schwarze 524: dbclose(0);
1.28 schwarze 525:
1.140 schwarze 526: if (j + 1 < conf.manpath.sz) {
1.47 schwarze 527: mpages_free();
528: ohash_delete(&mpages);
529: ohash_delete(&mlinks);
530: }
1.2 schwarze 531: }
1.47 schwarze 532: }
533: out:
1.140 schwarze 534: manconf_free(&conf);
1.123 schwarze 535: mparse_free(mp);
1.157 schwarze 536: mchars_free();
1.47 schwarze 537: mpages_free();
538: ohash_delete(&mpages);
539: ohash_delete(&mlinks);
1.150 schwarze 540: return exitcode;
1.47 schwarze 541: usage:
1.161 schwarze 542: progname = getprogname();
1.83 schwarze 543: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
544: " %s [-aDnpQ] [-Tutf8] dir ...\n"
545: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
546: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 547: " %s [-Q] -t file ...\n",
1.161 schwarze 548: progname, progname, progname, progname, progname);
1.1 schwarze 549:
1.150 schwarze 550: return (int)MANDOCLEVEL_BADARG;
1.47 schwarze 551: }
1.1 schwarze 552:
1.47 schwarze 553: /*
554: * Scan a directory tree rooted at "basedir" for manpages.
555: * We use fts(), scanning directory parts along the way for clues to our
556: * section and architecture.
557: *
558: * If use_all has been specified, grok all files.
559: * If not, sanitise paths to the following:
560: *
1.98 schwarze 561: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 562: * or
563: * [./]cat<section>[/<arch>]/<name>.0
564: *
1.169 krw 565: * TODO: accommodate for multi-language directories.
1.47 schwarze 566: */
567: static int
568: treescan(void)
569: {
1.96 schwarze 570: char buf[PATH_MAX];
1.47 schwarze 571: FTS *f;
572: FTSENT *ff;
573: struct mlink *mlink;
1.81 schwarze 574: int dform, gzip;
1.51 schwarze 575: char *dsec, *arch, *fsec, *cp;
576: const char *path;
1.47 schwarze 577: const char *argv[2];
1.1 schwarze 578:
1.47 schwarze 579: argv[0] = ".";
580: argv[1] = (char *)NULL;
1.1 schwarze 581:
1.96 schwarze 582: f = fts_open((char * const *)argv,
583: FTS_PHYSICAL | FTS_NOCHDIR, NULL);
1.153 schwarze 584: if (f == NULL) {
1.47 schwarze 585: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 586: say("", "&fts_open");
1.150 schwarze 587: return 0;
1.47 schwarze 588: }
1.2 schwarze 589:
1.47 schwarze 590: dsec = arch = NULL;
591: dform = FORM_NONE;
1.2 schwarze 592:
1.153 schwarze 593: while ((ff = fts_read(f)) != NULL) {
1.47 schwarze 594: path = ff->fts_path + 2;
1.96 schwarze 595: switch (ff->fts_info) {
596:
597: /*
598: * Symbolic links require various sanity checks,
599: * then get handled just like regular files.
600: */
1.98 schwarze 601: case FTS_SL:
1.153 schwarze 602: if (realpath(path, buf) == NULL) {
1.96 schwarze 603: if (warnings)
604: say(path, "&realpath");
605: continue;
606: }
607: if (strstr(buf, basedir) != buf) {
608: if (warnings) say("",
609: "%s: outside base directory", buf);
610: continue;
611: }
612: /* Use logical inode to avoid mpages dupe. */
1.153 schwarze 613: if (stat(path, ff->fts_statp) == -1) {
1.96 schwarze 614: if (warnings)
615: say(path, "&stat");
616: continue;
617: }
618: /* FALLTHROUGH */
619:
1.15 schwarze 620: /*
1.47 schwarze 621: * If we're a regular file, add an mlink by using the
622: * stored directory data and handling the filename.
1.15 schwarze 623: */
1.98 schwarze 624: case FTS_F:
1.153 schwarze 625: if ( ! strcmp(path, MANDOC_DB))
1.47 schwarze 626: continue;
627: if ( ! use_all && ff->fts_level < 2) {
628: if (warnings)
629: say(path, "Extraneous file");
630: continue;
1.81 schwarze 631: }
632: gzip = 0;
633: fsec = NULL;
1.153 schwarze 634: while (fsec == NULL) {
1.81 schwarze 635: fsec = strrchr(ff->fts_name, '.');
1.153 schwarze 636: if (fsec == NULL || strcmp(fsec+1, "gz"))
1.81 schwarze 637: break;
638: gzip = 1;
639: *fsec = '\0';
640: fsec = NULL;
641: }
1.153 schwarze 642: if (fsec == NULL) {
1.47 schwarze 643: if ( ! use_all) {
644: if (warnings)
645: say(path,
646: "No filename suffix");
647: continue;
648: }
1.153 schwarze 649: } else if ( ! strcmp(++fsec, "html")) {
1.47 schwarze 650: if (warnings)
651: say(path, "Skip html");
652: continue;
1.153 schwarze 653: } else if ( ! strcmp(fsec, "ps")) {
1.47 schwarze 654: if (warnings)
655: say(path, "Skip ps");
656: continue;
1.153 schwarze 657: } else if ( ! strcmp(fsec, "pdf")) {
1.47 schwarze 658: if (warnings)
659: say(path, "Skip pdf");
660: continue;
661: } else if ( ! use_all &&
1.153 schwarze 662: ((dform == FORM_SRC &&
1.138 schwarze 663: strncmp(fsec, dsec, strlen(dsec))) ||
1.153 schwarze 664: (dform == FORM_CAT && strcmp(fsec, "0")))) {
1.47 schwarze 665: if (warnings)
666: say(path, "Wrong filename suffix");
667: continue;
668: } else
669: fsec[-1] = '\0';
1.51 schwarze 670:
1.47 schwarze 671: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.100 schwarze 672: if (strlcpy(mlink->file, path,
673: sizeof(mlink->file)) >=
674: sizeof(mlink->file)) {
675: say(path, "Filename too long");
676: free(mlink);
677: continue;
678: }
1.47 schwarze 679: mlink->dform = dform;
1.51 schwarze 680: mlink->dsec = dsec;
681: mlink->arch = arch;
682: mlink->name = ff->fts_name;
683: mlink->fsec = fsec;
1.81 schwarze 684: mlink->gzip = gzip;
1.47 schwarze 685: mlink_add(mlink, ff->fts_statp);
686: continue;
1.96 schwarze 687:
1.98 schwarze 688: case FTS_D:
689: case FTS_DP:
1.96 schwarze 690: break;
691:
692: default:
1.47 schwarze 693: if (warnings)
694: say(path, "Not a regular file");
695: continue;
696: }
697:
698: switch (ff->fts_level) {
1.98 schwarze 699: case 0:
1.47 schwarze 700: /* Ignore the root directory. */
701: break;
1.98 schwarze 702: case 1:
1.47 schwarze 703: /*
704: * This might contain manX/ or catX/.
705: * Try to infer this from the name.
706: * If we're not in use_all, enforce it.
707: */
708: cp = ff->fts_name;
1.153 schwarze 709: if (ff->fts_info == FTS_DP) {
710: dform = FORM_NONE;
711: dsec = NULL;
1.47 schwarze 712: break;
1.153 schwarze 713: }
1.15 schwarze 714:
1.153 schwarze 715: if ( ! strncmp(cp, "man", 3)) {
1.47 schwarze 716: dform = FORM_SRC;
717: dsec = cp + 3;
1.153 schwarze 718: } else if ( ! strncmp(cp, "cat", 3)) {
1.47 schwarze 719: dform = FORM_CAT;
720: dsec = cp + 3;
1.51 schwarze 721: } else {
722: dform = FORM_NONE;
723: dsec = NULL;
1.26 schwarze 724: }
1.47 schwarze 725:
1.153 schwarze 726: if (dsec != NULL || use_all)
1.47 schwarze 727: break;
728:
729: if (warnings)
730: say(path, "Unknown directory part");
731: fts_set(f, ff, FTS_SKIP);
732: break;
1.98 schwarze 733: case 2:
1.47 schwarze 734: /*
735: * Possibly our architecture.
736: * If we're descending, keep tabs on it.
737: */
1.153 schwarze 738: if (ff->fts_info != FTS_DP && dsec != NULL)
1.47 schwarze 739: arch = ff->fts_name;
1.51 schwarze 740: else
741: arch = NULL;
1.47 schwarze 742: break;
743: default:
1.153 schwarze 744: if (ff->fts_info == FTS_DP || use_all)
1.47 schwarze 745: break;
746: if (warnings)
747: say(path, "Extraneous directory part");
748: fts_set(f, ff, FTS_SKIP);
749: break;
1.14 schwarze 750: }
1.47 schwarze 751: }
752:
753: fts_close(f);
1.150 schwarze 754: return 1;
1.47 schwarze 755: }
1.1 schwarze 756:
1.47 schwarze 757: /*
758: * Add a file to the mlinks table.
759: * Do not verify that it's a "valid" looking manpage (we'll do that
760: * later).
761: *
762: * Try to infer the manual section, architecture, and page name from the
763: * path, assuming it looks like
764: *
1.98 schwarze 765: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 766: * or
767: * [./]cat<section>[/<arch>]/<name>.0
768: *
769: * See treescan() for the fts(3) version of this.
770: */
771: static void
772: filescan(const char *file)
773: {
774: char buf[PATH_MAX];
775: struct stat st;
776: struct mlink *mlink;
777: char *p, *start;
778:
779: assert(use_all);
780:
781: if (0 == strncmp(file, "./", 2))
782: file += 2;
783:
1.96 schwarze 784: /*
785: * We have to do lstat(2) before realpath(3) loses
786: * the information whether this is a symbolic link.
787: * We need to know that because for symbolic links,
788: * we want to use the orginal file name, while for
789: * regular files, we want to use the real path.
790: */
791: if (-1 == lstat(file, &st)) {
792: exitcode = (int)MANDOCLEVEL_BADARG;
793: say(file, "&lstat");
794: return;
795: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
796: exitcode = (int)MANDOCLEVEL_BADARG;
797: say(file, "Not a regular file");
798: return;
799: }
800:
801: /*
802: * We have to resolve the file name to the real path
803: * in any case for the base directory check.
804: */
1.47 schwarze 805: if (NULL == realpath(file, buf)) {
806: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 807: say(file, "&realpath");
1.47 schwarze 808: return;
1.63 schwarze 809: }
810:
1.109 schwarze 811: if (OP_TEST == op)
1.63 schwarze 812: start = buf;
1.109 schwarze 813: else if (strstr(buf, basedir) == buf)
814: start = buf + strlen(basedir);
1.63 schwarze 815: else {
1.47 schwarze 816: exitcode = (int)MANDOCLEVEL_BADARG;
817: say("", "%s: outside base directory", buf);
818: return;
1.63 schwarze 819: }
820:
1.96 schwarze 821: /*
822: * Now we are sure the file is inside our tree.
823: * If it is a symbolic link, ignore the real path
824: * and use the original name.
825: * This implies passing stuff like "cat1/../man1/foo.1"
826: * on the command line won't work. So don't do that.
827: * Note the stat(2) can still fail if the link target
828: * doesn't exist.
829: */
830: if (S_IFLNK & st.st_mode) {
831: if (-1 == stat(buf, &st)) {
832: exitcode = (int)MANDOCLEVEL_BADARG;
833: say(file, "&stat");
834: return;
835: }
1.100 schwarze 836: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
837: say(file, "Filename too long");
838: return;
839: }
1.109 schwarze 840: start = buf;
841: if (OP_TEST != op && strstr(buf, basedir) == buf)
842: start += strlen(basedir);
1.1 schwarze 843: }
1.63 schwarze 844:
1.47 schwarze 845: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 846: mlink->dform = FORM_NONE;
1.100 schwarze 847: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
848: sizeof(mlink->file)) {
849: say(start, "Filename too long");
1.134 schwarze 850: free(mlink);
1.100 schwarze 851: return;
852: }
1.1 schwarze 853:
1.10 schwarze 854: /*
1.47 schwarze 855: * First try to guess our directory structure.
856: * If we find a separator, try to look for man* or cat*.
857: * If we find one of these and what's underneath is a directory,
858: * assume it's an architecture.
1.10 schwarze 859: */
1.47 schwarze 860: if (NULL != (p = strchr(start, '/'))) {
861: *p++ = '\0';
862: if (0 == strncmp(start, "man", 3)) {
863: mlink->dform = FORM_SRC;
1.51 schwarze 864: mlink->dsec = start + 3;
1.47 schwarze 865: } else if (0 == strncmp(start, "cat", 3)) {
866: mlink->dform = FORM_CAT;
1.51 schwarze 867: mlink->dsec = start + 3;
1.47 schwarze 868: }
1.10 schwarze 869:
1.47 schwarze 870: start = p;
871: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
872: *p++ = '\0';
1.51 schwarze 873: mlink->arch = start;
1.47 schwarze 874: start = p;
1.41 deraadt 875: }
1.47 schwarze 876: }
1.7 schwarze 877:
1.47 schwarze 878: /*
879: * Now check the file suffix.
880: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
881: */
882: p = strrchr(start, '\0');
883: while (p-- > start && '/' != *p && '.' != *p)
884: /* Loop. */ ;
885:
886: if ('.' == *p) {
887: *p++ = '\0';
1.51 schwarze 888: mlink->fsec = p;
1.47 schwarze 889: }
1.41 deraadt 890:
1.47 schwarze 891: /*
892: * Now try to parse the name.
893: * Use the filename portion of the path.
894: */
895: mlink->name = start;
896: if (NULL != (p = strrchr(start, '/'))) {
897: mlink->name = p + 1;
898: *p = '\0';
899: }
900: mlink_add(mlink, &st);
901: }
1.2 schwarze 902:
1.47 schwarze 903: static void
904: mlink_add(struct mlink *mlink, const struct stat *st)
905: {
906: struct inodev inodev;
907: struct mpage *mpage;
908: unsigned int slot;
909:
910: assert(NULL != mlink->file);
911:
1.51 schwarze 912: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
913: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
914: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
915: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 916:
917: if ('0' == *mlink->fsec) {
918: free(mlink->fsec);
919: mlink->fsec = mandoc_strdup(mlink->dsec);
920: mlink->fform = FORM_CAT;
921: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
922: mlink->fform = FORM_SRC;
923: else
924: mlink->fform = FORM_NONE;
1.1 schwarze 925:
1.47 schwarze 926: slot = ohash_qlookup(&mlinks, mlink->file);
927: assert(NULL == ohash_find(&mlinks, slot));
928: ohash_insert(&mlinks, slot, mlink);
929:
1.139 schwarze 930: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.47 schwarze 931: inodev.st_ino = st->st_ino;
932: inodev.st_dev = st->st_dev;
933: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
934: sizeof(struct inodev), inodev.st_ino);
935: mpage = ohash_find(&mpages, slot);
936: if (NULL == mpage) {
937: mpage = mandoc_calloc(1, sizeof(struct mpage));
938: mpage->inodev.st_ino = inodev.st_ino;
939: mpage->inodev.st_dev = inodev.st_dev;
940: ohash_insert(&mpages, slot, mpage);
941: } else
942: mlink->next = mpage->mlinks;
943: mpage->mlinks = mlink;
1.75 schwarze 944: mlink->mpage = mpage;
1.47 schwarze 945: }
1.1 schwarze 946:
1.47 schwarze 947: static void
948: mlink_free(struct mlink *mlink)
949: {
1.1 schwarze 950:
1.47 schwarze 951: free(mlink->dsec);
952: free(mlink->arch);
953: free(mlink->name);
954: free(mlink->fsec);
955: free(mlink);
956: }
1.1 schwarze 957:
1.47 schwarze 958: static void
959: mpages_free(void)
960: {
961: struct mpage *mpage;
962: struct mlink *mlink;
963: unsigned int slot;
964:
965: mpage = ohash_first(&mpages, &slot);
966: while (NULL != mpage) {
967: while (NULL != (mlink = mpage->mlinks)) {
968: mpage->mlinks = mlink->next;
969: mlink_free(mlink);
970: }
971: free(mpage->sec);
972: free(mpage->arch);
973: free(mpage->title);
974: free(mpage->desc);
975: free(mpage);
976: mpage = ohash_next(&mpages, &slot);
977: }
978: }
1.1 schwarze 979:
1.47 schwarze 980: /*
981: * For each mlink to the mpage, check whether the path looks like
982: * it is formatted, and if it does, check whether a source manual
983: * exists by the same name, ignoring the suffix.
984: * If both conditions hold, drop the mlink.
985: */
986: static void
987: mlinks_undupe(struct mpage *mpage)
988: {
989: char buf[PATH_MAX];
990: struct mlink **prev;
991: struct mlink *mlink;
992: char *bufp;
993:
994: mpage->form = FORM_CAT;
995: prev = &mpage->mlinks;
996: while (NULL != (mlink = *prev)) {
997: if (FORM_CAT != mlink->dform) {
998: mpage->form = FORM_NONE;
999: goto nextlink;
1000: }
1.100 schwarze 1001: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 1002: bufp = strstr(buf, "cat");
1003: assert(NULL != bufp);
1004: memcpy(bufp, "man", 3);
1005: if (NULL != (bufp = strrchr(buf, '.')))
1006: *++bufp = '\0';
1.100 schwarze 1007: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 1008: if (NULL == ohash_find(&mlinks,
1.98 schwarze 1009: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 1010: goto nextlink;
1011: if (warnings)
1012: say(mlink->file, "Man source exists: %s", buf);
1013: if (use_all)
1014: goto nextlink;
1015: *prev = mlink->next;
1016: mlink_free(mlink);
1017: continue;
1018: nextlink:
1019: prev = &(*prev)->next;
1.1 schwarze 1020: }
1.47 schwarze 1021: }
1.1 schwarze 1022:
1.88 schwarze 1023: static void
1.50 schwarze 1024: mlink_check(struct mpage *mpage, struct mlink *mlink)
1025: {
1.88 schwarze 1026: struct str *str;
1027: unsigned int slot;
1.50 schwarze 1028:
1029: /*
1030: * Check whether the manual section given in a file
1031: * agrees with the directory where the file is located.
1032: * Some manuals have suffixes like (3p) on their
1033: * section number either inside the file or in the
1034: * directory name, some are linked into more than one
1035: * section, like encrypt(1) = makekey(8).
1036: */
1037:
1038: if (FORM_SRC == mpage->form &&
1.88 schwarze 1039: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1040: say(mlink->file, "Section \"%s\" manual in %s directory",
1041: mpage->sec, mlink->dsec);
1042:
1043: /*
1044: * Manual page directories exist for each kernel
1045: * architecture as returned by machine(1).
1046: * However, many manuals only depend on the
1047: * application architecture as returned by arch(1).
1048: * For example, some (2/ARM) manuals are shared
1049: * across the "armish" and "zaurus" kernel
1050: * architectures.
1051: * A few manuals are even shared across completely
1052: * different architectures, for example fdformat(1)
1053: * on amd64, i386, sparc, and sparc64.
1054: */
1055:
1.88 schwarze 1056: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1057: say(mlink->file, "Architecture \"%s\" manual in "
1058: "\"%s\" directory", mpage->arch, mlink->arch);
1059:
1.88 schwarze 1060: /*
1061: * XXX
1.90 schwarze 1062: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1063: */
1064:
1065: if (FORM_CAT == mpage->form)
1066: return;
1067:
1068: /*
1069: * Check whether this mlink
1070: * appears as a name in the NAME section.
1071: */
1.50 schwarze 1072:
1.90 schwarze 1073: slot = ohash_qlookup(&names, mlink->name);
1074: str = ohash_find(&names, slot);
1.88 schwarze 1075: assert(NULL != str);
1.90 schwarze 1076: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1077: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1078: }
1079:
1.47 schwarze 1080: /*
1081: * Run through the files in the global vector "mpages"
1082: * and add them to the database specified in "basedir".
1083: *
1084: * This handles the parsing scheme itself, using the cues of directory
1085: * and filename to determine whether the file is parsable or not.
1086: */
1087: static void
1.131 schwarze 1088: mpages_merge(struct mparse *mp)
1.47 schwarze 1089: {
1.70 schwarze 1090: char any[] = "any";
1.75 schwarze 1091: struct mpage *mpage, *mpage_dest;
1092: struct mlink *mlink, *mlink_dest;
1.144 schwarze 1093: struct roff_man *man;
1.75 schwarze 1094: char *sodest;
1.69 schwarze 1095: char *cp;
1.117 schwarze 1096: int fd;
1.58 schwarze 1097: unsigned int pslot;
1.47 schwarze 1098:
1.126 schwarze 1099: if ( ! nodb)
1.64 schwarze 1100: SQL_EXEC("BEGIN TRANSACTION");
1101:
1.47 schwarze 1102: mpage = ohash_first(&mpages, &pslot);
1.126 schwarze 1103: while (mpage != NULL) {
1.47 schwarze 1104: mlinks_undupe(mpage);
1.136 schwarze 1105: if ((mlink = mpage->mlinks) == NULL) {
1.47 schwarze 1106: mpage = ohash_next(&mpages, &pslot);
1107: continue;
1108: }
1.1 schwarze 1109:
1.90 schwarze 1110: name_mask = NAME_MASK;
1.156 schwarze 1111: mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1112: mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1.47 schwarze 1113: mparse_reset(mp);
1114: man = NULL;
1.81 schwarze 1115: sodest = NULL;
1116:
1.166 schwarze 1117: if ((fd = mparse_open(mp, mlink->file)) == -1) {
1.136 schwarze 1118: say(mlink->file, "&open");
1.117 schwarze 1119: goto nextpage;
1.81 schwarze 1120: }
1.11 schwarze 1121:
1122: /*
1.137 schwarze 1123: * Interpret the file as mdoc(7) or man(7) source
1124: * code, unless it is known to be formatted.
1.11 schwarze 1125: */
1.136 schwarze 1126: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.137 schwarze 1127: mparse_readfd(mp, fd, mlink->file);
1.165 schwarze 1128: close(fd);
1.145 schwarze 1129: mparse_result(mp, &man, &sodest);
1.47 schwarze 1130: }
1.11 schwarze 1131:
1.126 schwarze 1132: if (sodest != NULL) {
1.75 schwarze 1133: mlink_dest = ohash_find(&mlinks,
1134: ohash_qlookup(&mlinks, sodest));
1.126 schwarze 1135: if (mlink_dest == NULL) {
1136: mandoc_asprintf(&cp, "%s.gz", sodest);
1137: mlink_dest = ohash_find(&mlinks,
1138: ohash_qlookup(&mlinks, cp));
1139: free(cp);
1140: }
1141: if (mlink_dest != NULL) {
1.75 schwarze 1142:
1143: /* The .so target exists. */
1144:
1145: mpage_dest = mlink_dest->mpage;
1146: while (1) {
1147: mlink->mpage = mpage_dest;
1148:
1149: /*
1150: * If the target was already
1151: * processed, add the links
1152: * to the database now.
1153: * Otherwise, this will
1154: * happen when we come
1155: * to the target.
1156: */
1157:
1.94 schwarze 1158: if (mpage_dest->pageid)
1.124 schwarze 1159: dbadd_mlink_name(mlink);
1.75 schwarze 1160:
1.126 schwarze 1161: if (mlink->next == NULL)
1.75 schwarze 1162: break;
1163: mlink = mlink->next;
1164: }
1165:
1166: /* Move all links to the target. */
1167:
1168: mlink->next = mlink_dest->next;
1169: mlink_dest->next = mpage->mlinks;
1170: mpage->mlinks = NULL;
1171: }
1.81 schwarze 1172: goto nextpage;
1.145 schwarze 1173: } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1.159 schwarze 1174: mdoc_validate(man);
1.47 schwarze 1175: mpage->form = FORM_SRC;
1.146 schwarze 1176: mpage->sec = man->meta.msec;
1.112 schwarze 1177: mpage->sec = mandoc_strdup(
1.126 schwarze 1178: mpage->sec == NULL ? "" : mpage->sec);
1.146 schwarze 1179: mpage->arch = man->meta.arch;
1.47 schwarze 1180: mpage->arch = mandoc_strdup(
1.126 schwarze 1181: mpage->arch == NULL ? "" : mpage->arch);
1.146 schwarze 1182: mpage->title = mandoc_strdup(man->meta.title);
1.145 schwarze 1183: } else if (man != NULL && man->macroset == MACROSET_MAN) {
1.160 schwarze 1184: man_validate(man);
1.47 schwarze 1185: mpage->form = FORM_SRC;
1.146 schwarze 1186: mpage->sec = mandoc_strdup(man->meta.msec);
1.136 schwarze 1187: mpage->arch = mandoc_strdup(mlink->arch);
1.146 schwarze 1188: mpage->title = mandoc_strdup(man->meta.title);
1.11 schwarze 1189: } else {
1.47 schwarze 1190: mpage->form = FORM_CAT;
1.136 schwarze 1191: mpage->sec = mandoc_strdup(mlink->dsec);
1192: mpage->arch = mandoc_strdup(mlink->arch);
1193: mpage->title = mandoc_strdup(mlink->name);
1.1 schwarze 1194: }
1.54 schwarze 1195: putkey(mpage, mpage->sec, TYPE_sec);
1.119 schwarze 1196: if (*mpage->arch != '\0')
1197: putkey(mpage, mpage->arch, TYPE_arch);
1.1 schwarze 1198:
1.136 schwarze 1199: for ( ; mlink != NULL; mlink = mlink->next) {
1.54 schwarze 1200: if ('\0' != *mlink->dsec)
1201: putkey(mpage, mlink->dsec, TYPE_sec);
1202: if ('\0' != *mlink->fsec)
1203: putkey(mpage, mlink->fsec, TYPE_sec);
1.55 schwarze 1204: putkey(mpage, '\0' == *mlink->arch ?
1.70 schwarze 1205: any : mlink->arch, TYPE_arch);
1.90 schwarze 1206: putkey(mpage, mlink->name, NAME_FILE);
1.54 schwarze 1207: }
1.41 deraadt 1208:
1.129 schwarze 1209: assert(mpage->desc == NULL);
1.145 schwarze 1210: if (man != NULL && man->macroset == MACROSET_MDOC)
1.146 schwarze 1211: parse_mdoc(mpage, &man->meta, man->first);
1.129 schwarze 1212: else if (man != NULL)
1.146 schwarze 1213: parse_man(mpage, &man->meta, man->first);
1.47 schwarze 1214: else
1.117 schwarze 1215: parse_cat(mpage, fd);
1.129 schwarze 1216: if (mpage->desc == NULL)
1.92 schwarze 1217: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1.88 schwarze 1218:
1219: if (warnings && !use_all)
1220: for (mlink = mpage->mlinks; mlink;
1221: mlink = mlink->next)
1222: mlink_check(mpage, mlink);
1.6 schwarze 1223:
1.131 schwarze 1224: dbadd(mpage);
1.136 schwarze 1225: mlink = mpage->mlinks;
1.81 schwarze 1226:
1227: nextpage:
1.47 schwarze 1228: ohash_delete(&strings);
1.90 schwarze 1229: ohash_delete(&names);
1.47 schwarze 1230: mpage = ohash_next(&mpages, &pslot);
1231: }
1.64 schwarze 1232:
1233: if (0 == nodb)
1234: SQL_EXEC("END TRANSACTION");
1.87 schwarze 1235: }
1236:
1237: static void
1238: names_check(void)
1239: {
1240: sqlite3_stmt *stmt;
1241: const char *name, *sec, *arch, *key;
1242:
1243: sqlite3_prepare_v2(db,
1244: "SELECT name, sec, arch, key FROM ("
1.90 schwarze 1245: "SELECT name AS key, pageid FROM names "
1.87 schwarze 1246: "WHERE bits & ? AND NOT EXISTS ("
1247: "SELECT pageid FROM mlinks "
1.90 schwarze 1248: "WHERE mlinks.pageid == names.pageid "
1249: "AND mlinks.name == names.name"
1.87 schwarze 1250: ")"
1251: ") JOIN ("
1.99 schwarze 1252: "SELECT sec, arch, name, pageid FROM mlinks "
1253: "GROUP BY pageid"
1.87 schwarze 1254: ") USING (pageid);",
1255: -1, &stmt, NULL);
1256:
1.154 schwarze 1257: if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK)
1.91 schwarze 1258: say("", "%s", sqlite3_errmsg(db));
1.87 schwarze 1259:
1.154 schwarze 1260: while (sqlite3_step(stmt) == SQLITE_ROW) {
1.113 schwarze 1261: name = (const char *)sqlite3_column_text(stmt, 0);
1262: sec = (const char *)sqlite3_column_text(stmt, 1);
1263: arch = (const char *)sqlite3_column_text(stmt, 2);
1264: key = (const char *)sqlite3_column_text(stmt, 3);
1.87 schwarze 1265: say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1266: '\0' == *arch ? "" : "/",
1267: '\0' == *arch ? "" : arch, key);
1268: }
1269: sqlite3_finalize(stmt);
1.47 schwarze 1270: }
1.6 schwarze 1271:
1.47 schwarze 1272: static void
1.81 schwarze 1273: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1274: {
1275: FILE *stream;
1276: char *line, *p, *title;
1.162 schwarze 1277: size_t linesz, plen, titlesz;
1278: ssize_t len;
1279: int offs;
1.1 schwarze 1280:
1.81 schwarze 1281: stream = (-1 == fd) ?
1282: fopen(mpage->mlinks->file, "r") :
1283: fdopen(fd, "r");
1284: if (NULL == stream) {
1.114 doug 1285: if (-1 != fd)
1286: close(fd);
1.47 schwarze 1287: if (warnings)
1.80 schwarze 1288: say(mpage->mlinks->file, "&fopen");
1.47 schwarze 1289: return;
1290: }
1.1 schwarze 1291:
1.162 schwarze 1292: line = NULL;
1293: linesz = 0;
1294:
1.47 schwarze 1295: /* Skip to first blank line. */
1.1 schwarze 1296:
1.162 schwarze 1297: while (getline(&line, &linesz, stream) != -1)
1298: if (*line == '\n')
1.47 schwarze 1299: break;
1.1 schwarze 1300:
1.47 schwarze 1301: /*
1302: * Assume the first line that is not indented
1303: * is the first section header. Skip to it.
1304: */
1.1 schwarze 1305:
1.162 schwarze 1306: while (getline(&line, &linesz, stream) != -1)
1307: if (*line != '\n' && *line != ' ')
1.47 schwarze 1308: break;
1.98 schwarze 1309:
1.47 schwarze 1310: /*
1311: * Read up until the next section into a buffer.
1312: * Strip the leading and trailing newline from each read line,
1313: * appending a trailing space.
1314: * Ignore empty (whitespace-only) lines.
1315: */
1.28 schwarze 1316:
1.47 schwarze 1317: titlesz = 0;
1318: title = NULL;
1.38 schwarze 1319:
1.162 schwarze 1320: while ((len = getline(&line, &linesz, stream)) != -1) {
1321: if (*line != ' ')
1.47 schwarze 1322: break;
1.162 schwarze 1323: offs = 0;
1324: while (isspace((unsigned char)line[offs]))
1325: offs++;
1326: if (line[offs] == '\0')
1.47 schwarze 1327: continue;
1.162 schwarze 1328: title = mandoc_realloc(title, titlesz + len - offs);
1329: memcpy(title + titlesz, line + offs, len - offs);
1330: titlesz += len - offs;
1.47 schwarze 1331: title[titlesz - 1] = ' ';
1332: }
1.162 schwarze 1333: free(line);
1.28 schwarze 1334:
1.47 schwarze 1335: /*
1336: * If no page content can be found, or the input line
1337: * is already the next section header, or there is no
1338: * trailing newline, reuse the page title as the page
1339: * description.
1340: */
1.1 schwarze 1341:
1.47 schwarze 1342: if (NULL == title || '\0' == *title) {
1343: if (warnings)
1344: say(mpage->mlinks->file,
1345: "Cannot find NAME section");
1346: fclose(stream);
1347: free(title);
1348: return;
1349: }
1.24 schwarze 1350:
1.162 schwarze 1351: title[titlesz - 1] = '\0';
1.24 schwarze 1352:
1.47 schwarze 1353: /*
1354: * Skip to the first dash.
1355: * Use the remaining line as the description (no more than 70
1356: * bytes).
1357: */
1.28 schwarze 1358:
1.47 schwarze 1359: if (NULL != (p = strstr(title, "- "))) {
1360: for (p += 2; ' ' == *p || '\b' == *p; p++)
1361: /* Skip to next word. */ ;
1362: } else {
1363: if (warnings)
1364: say(mpage->mlinks->file,
1365: "No dash in title line");
1366: p = title;
1367: }
1.1 schwarze 1368:
1.47 schwarze 1369: plen = strlen(p);
1.1 schwarze 1370:
1.47 schwarze 1371: /* Strip backspace-encoding from line. */
1.1 schwarze 1372:
1.47 schwarze 1373: while (NULL != (line = memchr(p, '\b', plen))) {
1374: len = line - p;
1375: if (0 == len) {
1376: memmove(line, line + 1, plen--);
1377: continue;
1.98 schwarze 1378: }
1.47 schwarze 1379: memmove(line - 1, line + 1, plen - len);
1380: plen -= 2;
1381: }
1.1 schwarze 1382:
1.47 schwarze 1383: mpage->desc = mandoc_strdup(p);
1384: fclose(stream);
1385: free(title);
1386: }
1.16 schwarze 1387:
1.47 schwarze 1388: /*
1389: * Put a type/word pair into the word database for this particular file.
1390: */
1391: static void
1.69 schwarze 1392: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1393: {
1.69 schwarze 1394: char *cp;
1.37 schwarze 1395:
1.47 schwarze 1396: assert(NULL != value);
1.69 schwarze 1397: if (TYPE_arch == type)
1398: for (cp = value; *cp; cp++)
1399: if (isupper((unsigned char)*cp))
1400: *cp = _tolower((unsigned char)*cp);
1.47 schwarze 1401: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1402: }
1403:
1404: /*
1.47 schwarze 1405: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1406: */
1407: static void
1.47 schwarze 1408: putmdockey(const struct mpage *mpage,
1.142 schwarze 1409: const struct roff_node *n, uint64_t m)
1.2 schwarze 1410: {
1.16 schwarze 1411:
1.47 schwarze 1412: for ( ; NULL != n; n = n->next) {
1413: if (NULL != n->child)
1414: putmdockey(mpage, n->child, m);
1.141 schwarze 1415: if (n->type == ROFFT_TEXT)
1.47 schwarze 1416: putkey(mpage, n->string, m);
1417: }
1418: }
1.16 schwarze 1419:
1.47 schwarze 1420: static void
1.143 schwarze 1421: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1422: const struct roff_node *n)
1.47 schwarze 1423: {
1.142 schwarze 1424: const struct roff_node *head, *body;
1.78 schwarze 1425: char *start, *title;
1.47 schwarze 1426: char byte;
1.78 schwarze 1427: size_t sz;
1.16 schwarze 1428:
1.168 schwarze 1429: if (n == NULL)
1.47 schwarze 1430: return;
1.16 schwarze 1431:
1.47 schwarze 1432: /*
1433: * We're only searching for one thing: the first text child in
1434: * the BODY of a NAME section. Since we don't keep track of
1435: * sections in -man, run some hoops to find out whether we're in
1436: * the correct section or not.
1437: */
1.16 schwarze 1438:
1.141 schwarze 1439: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1440: body = n;
1.168 schwarze 1441: if ((head = body->parent->head) != NULL &&
1442: (head = head->child) != NULL &&
1443: head->next == NULL &&
1.141 schwarze 1444: head->type == ROFFT_TEXT &&
1.168 schwarze 1445: strcmp(head->string, "NAME") == 0 &&
1446: body->child != NULL) {
1.2 schwarze 1447:
1.47 schwarze 1448: /*
1449: * Suck the entire NAME section into memory.
1450: * Yes, we might run away.
1451: * But too many manuals have big, spread-out
1452: * NAME sections over many lines.
1453: */
1.2 schwarze 1454:
1.78 schwarze 1455: title = NULL;
1.147 schwarze 1456: deroff(&title, body);
1.47 schwarze 1457: if (NULL == title)
1458: return;
1.16 schwarze 1459:
1.98 schwarze 1460: /*
1.47 schwarze 1461: * Go through a special heuristic dance here.
1462: * Conventionally, one or more manual names are
1463: * comma-specified prior to a whitespace, then a
1464: * dash, then a description. Try to puzzle out
1465: * the name parts here.
1466: */
1.16 schwarze 1467:
1.78 schwarze 1468: start = title;
1.47 schwarze 1469: for ( ;; ) {
1470: sz = strcspn(start, " ,");
1471: if ('\0' == start[sz])
1472: break;
1.1 schwarze 1473:
1.47 schwarze 1474: byte = start[sz];
1475: start[sz] = '\0';
1.67 schwarze 1476:
1477: /*
1478: * Assume a stray trailing comma in the
1479: * name list if a name begins with a dash.
1480: */
1481:
1482: if ('-' == start[0] ||
1483: ('\\' == start[0] && '-' == start[1]))
1484: break;
1.1 schwarze 1485:
1.90 schwarze 1486: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1487: if ( ! (mpage->name_head_done ||
1488: strcasecmp(start, meta->title))) {
1489: putkey(mpage, start, NAME_HEAD);
1490: mpage->name_head_done = 1;
1491: }
1.1 schwarze 1492:
1.47 schwarze 1493: if (' ' == byte) {
1494: start += sz + 1;
1495: break;
1496: }
1.1 schwarze 1497:
1.47 schwarze 1498: assert(',' == byte);
1499: start += sz + 1;
1500: while (' ' == *start)
1501: start++;
1502: }
1.1 schwarze 1503:
1.78 schwarze 1504: if (start == title) {
1.90 schwarze 1505: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1506: if ( ! (mpage->name_head_done ||
1507: strcasecmp(start, meta->title))) {
1508: putkey(mpage, start, NAME_HEAD);
1509: mpage->name_head_done = 1;
1510: }
1.47 schwarze 1511: free(title);
1512: return;
1513: }
1.1 schwarze 1514:
1.47 schwarze 1515: while (isspace((unsigned char)*start))
1516: start++;
1.1 schwarze 1517:
1.47 schwarze 1518: if (0 == strncmp(start, "-", 1))
1519: start += 1;
1520: else if (0 == strncmp(start, "\\-\\-", 4))
1521: start += 4;
1522: else if (0 == strncmp(start, "\\-", 2))
1523: start += 2;
1524: else if (0 == strncmp(start, "\\(en", 4))
1525: start += 4;
1526: else if (0 == strncmp(start, "\\(em", 4))
1527: start += 4;
1.1 schwarze 1528:
1.47 schwarze 1529: while (' ' == *start)
1530: start++;
1.1 schwarze 1531:
1.47 schwarze 1532: mpage->desc = mandoc_strdup(start);
1533: free(title);
1534: return;
1535: }
1536: }
1.1 schwarze 1537:
1.47 schwarze 1538: for (n = n->child; n; n = n->next) {
1539: if (NULL != mpage->desc)
1540: break;
1.129 schwarze 1541: parse_man(mpage, meta, n);
1.1 schwarze 1542: }
1543: }
1544:
1545: static void
1.143 schwarze 1546: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1547: const struct roff_node *n)
1.1 schwarze 1548: {
1549:
1.47 schwarze 1550: assert(NULL != n);
1551: for (n = n->child; NULL != n; n = n->next) {
1552: switch (n->type) {
1.141 schwarze 1553: case ROFFT_ELEM:
1554: case ROFFT_BLOCK:
1555: case ROFFT_HEAD:
1556: case ROFFT_BODY:
1557: case ROFFT_TAIL:
1.47 schwarze 1558: if (NULL != mdocs[n->tok].fp)
1.127 schwarze 1559: if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1.47 schwarze 1560: break;
1561: if (mdocs[n->tok].mask)
1562: putmdockey(mpage, n->child,
1563: mdocs[n->tok].mask);
1564: break;
1565: default:
1.141 schwarze 1566: assert(n->type != ROFFT_ROOT);
1.47 schwarze 1567: continue;
1568: }
1569: if (NULL != n->child)
1.127 schwarze 1570: parse_mdoc(mpage, meta, n);
1.1 schwarze 1571: }
1572: }
1573:
1.19 schwarze 1574: static int
1.143 schwarze 1575: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1576: const struct roff_node *n)
1.1 schwarze 1577: {
1.131 schwarze 1578: char *start, *end;
1.1 schwarze 1579: size_t sz;
1.19 schwarze 1580:
1.47 schwarze 1581: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1582: NULL == (n = n->child) ||
1.141 schwarze 1583: n->type != ROFFT_TEXT)
1.150 schwarze 1584: return 0;
1.1 schwarze 1585:
1586: /*
1587: * Only consider those `Fd' macro fields that begin with an
1588: * "inclusion" token (versus, e.g., #define).
1589: */
1.47 schwarze 1590:
1.1 schwarze 1591: if (strcmp("#include", n->string))
1.150 schwarze 1592: return 0;
1.1 schwarze 1593:
1.141 schwarze 1594: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1595: return 0;
1.1 schwarze 1596:
1597: /*
1598: * Strip away the enclosing angle brackets and make sure we're
1599: * not zero-length.
1600: */
1601:
1602: start = n->string;
1603: if ('<' == *start || '"' == *start)
1604: start++;
1605:
1606: if (0 == (sz = strlen(start)))
1.150 schwarze 1607: return 0;
1.1 schwarze 1608:
1609: end = &start[(int)sz - 1];
1610: if ('>' == *end || '"' == *end)
1611: end--;
1612:
1.47 schwarze 1613: if (end > start)
1614: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1615: return 0;
1.1 schwarze 1616: }
1617:
1.133 schwarze 1618: static void
1.142 schwarze 1619: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1620: {
1.69 schwarze 1621: char *cp;
1.133 schwarze 1622: size_t sz;
1.1 schwarze 1623:
1.141 schwarze 1624: if (n->type != ROFFT_TEXT)
1.133 schwarze 1625: return;
1.19 schwarze 1626:
1.133 schwarze 1627: /* Skip function pointer punctuation. */
1.1 schwarze 1628:
1.133 schwarze 1629: cp = n->string;
1630: while (*cp == '(' || *cp == '*')
1.1 schwarze 1631: cp++;
1.133 schwarze 1632: sz = strcspn(cp, "()");
1.1 schwarze 1633:
1.133 schwarze 1634: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1635: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1636: putkeys(mpage, cp, sz, NAME_SYN);
1637: }
1.19 schwarze 1638:
1.133 schwarze 1639: static int
1.143 schwarze 1640: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1641: const struct roff_node *n)
1.133 schwarze 1642: {
1643:
1644: if (n->child == NULL)
1.150 schwarze 1645: return 0;
1.133 schwarze 1646:
1647: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1648:
1.133 schwarze 1649: for (n = n->child->next; n != NULL; n = n->next)
1.141 schwarze 1650: if (n->type == ROFFT_TEXT)
1.47 schwarze 1651: putkey(mpage, n->string, TYPE_Fa);
1.19 schwarze 1652:
1.150 schwarze 1653: return 0;
1.128 schwarze 1654: }
1655:
1656: static int
1.143 schwarze 1657: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1658: const struct roff_node *n)
1.128 schwarze 1659: {
1.132 schwarze 1660:
1.141 schwarze 1661: if (n->type != ROFFT_HEAD)
1.150 schwarze 1662: return 1;
1.128 schwarze 1663:
1.133 schwarze 1664: if (n->child != NULL)
1665: parse_mdoc_fname(mpage, n->child);
1666:
1.150 schwarze 1667: return 0;
1.1 schwarze 1668: }
1669:
1.19 schwarze 1670: static int
1.164 schwarze 1671: parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1672: const struct roff_node *n)
1673: {
1674: char *cp;
1675:
1676: if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1677: return 0;
1678:
1.168 schwarze 1679: if (n->child != NULL &&
1680: n->child->next == NULL &&
1681: n->child->type == ROFFT_TEXT)
1.164 schwarze 1682: return 1;
1683:
1684: cp = NULL;
1685: deroff(&cp, n);
1686: if (cp != NULL) {
1687: putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1688: n->type == ROFFT_BODY ? TYPE_Va : 0));
1689: free(cp);
1690: }
1691:
1692: return 0;
1693: }
1694:
1695: static int
1.143 schwarze 1696: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1697: const struct roff_node *n)
1.1 schwarze 1698: {
1.47 schwarze 1699: char *cp;
1.1 schwarze 1700:
1701: if (NULL == (n = n->child))
1.150 schwarze 1702: return 0;
1.1 schwarze 1703:
1.47 schwarze 1704: if (NULL == n->next) {
1705: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1706: return 0;
1.47 schwarze 1707: }
1.1 schwarze 1708:
1.77 schwarze 1709: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1710: putkey(mpage, cp, TYPE_Xr);
1711: free(cp);
1.150 schwarze 1712: return 0;
1.1 schwarze 1713: }
1714:
1.19 schwarze 1715: static int
1.143 schwarze 1716: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1717: const struct roff_node *n)
1.1 schwarze 1718: {
1719:
1.141 schwarze 1720: if (n->type == ROFFT_BODY)
1.147 schwarze 1721: deroff(&mpage->desc, n);
1.150 schwarze 1722: return 0;
1.1 schwarze 1723: }
1724:
1.19 schwarze 1725: static int
1.143 schwarze 1726: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1727: const struct roff_node *n)
1.1 schwarze 1728: {
1729:
1.86 schwarze 1730: if (SEC_NAME == n->sec)
1.90 schwarze 1731: putmdockey(mpage, n->child, NAME_TITLE);
1.141 schwarze 1732: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1733: if (n->child == NULL)
1734: putkey(mpage, meta->name, NAME_SYN);
1735: else
1736: putmdockey(mpage, n->child, NAME_SYN);
1.129 schwarze 1737: }
1738: if ( ! (mpage->name_head_done ||
1739: n->child == NULL || n->child->string == NULL ||
1740: strcasecmp(n->child->string, meta->title))) {
1.141 schwarze 1741: putkey(mpage, n->child->string, ROFFT_HEAD);
1.129 schwarze 1742: mpage->name_head_done = 1;
1.127 schwarze 1743: }
1.150 schwarze 1744: return 0;
1.1 schwarze 1745: }
1746:
1.19 schwarze 1747: static int
1.143 schwarze 1748: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1749: const struct roff_node *n)
1.1 schwarze 1750: {
1751:
1.150 schwarze 1752: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1753: }
1754:
1.47 schwarze 1755: static int
1.143 schwarze 1756: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1757: const struct roff_node *n)
1.1 schwarze 1758: {
1759:
1.150 schwarze 1760: return n->type == ROFFT_HEAD;
1.1 schwarze 1761: }
1762:
1.47 schwarze 1763: /*
1764: * Add a string to the hash table for the current manual.
1765: * Each string has a bitmask telling which macros it belongs to.
1766: * When we finish the manual, we'll dump the table.
1767: */
1.1 schwarze 1768: static void
1.131 schwarze 1769: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1770: {
1.90 schwarze 1771: struct ohash *htab;
1.47 schwarze 1772: struct str *s;
1.68 schwarze 1773: const char *end;
1.47 schwarze 1774: unsigned int slot;
1.131 schwarze 1775: int i, mustfree;
1.1 schwarze 1776:
1.47 schwarze 1777: if (0 == sz)
1778: return;
1.68 schwarze 1779:
1.131 schwarze 1780: mustfree = render_string(&cp, &sz);
1781:
1.90 schwarze 1782: if (TYPE_Nm & v) {
1783: htab = &names;
1784: v &= name_mask;
1.124 schwarze 1785: if (v & NAME_FIRST)
1786: name_mask &= ~NAME_FIRST;
1.90 schwarze 1787: if (debug > 1)
1788: say(mpage->mlinks->file,
1.130 schwarze 1789: "Adding name %*s, bits=%d", sz, cp, v);
1.90 schwarze 1790: } else {
1791: htab = &strings;
1792: if (debug > 1)
1793: for (i = 0; i < mansearch_keymax; i++)
1.118 schwarze 1794: if ((uint64_t)1 << i & v)
1.90 schwarze 1795: say(mpage->mlinks->file,
1796: "Adding key %s=%*s",
1797: mansearch_keynames[i], sz, cp);
1.68 schwarze 1798: }
1.47 schwarze 1799:
1800: end = cp + sz;
1.90 schwarze 1801: slot = ohash_qlookupi(htab, cp, &end);
1802: s = ohash_find(htab, slot);
1.1 schwarze 1803:
1.47 schwarze 1804: if (NULL != s && mpage == s->mpage) {
1805: s->mask |= v;
1.1 schwarze 1806: return;
1.47 schwarze 1807: } else if (NULL == s) {
1.101 schwarze 1808: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1809: memcpy(s->key, cp, sz);
1.90 schwarze 1810: ohash_insert(htab, slot, s);
1.47 schwarze 1811: }
1812: s->mpage = mpage;
1813: s->mask = v;
1.131 schwarze 1814:
1815: if (mustfree)
1816: free(cp);
1.1 schwarze 1817: }
1818:
1819: /*
1.47 schwarze 1820: * Take a Unicode codepoint and produce its UTF-8 encoding.
1821: * This isn't the best way to do this, but it works.
1822: * The magic numbers are from the UTF-8 packaging.
1823: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1824: */
1.47 schwarze 1825: static size_t
1826: utf8(unsigned int cp, char out[7])
1.1 schwarze 1827: {
1.47 schwarze 1828: size_t rc;
1.1 schwarze 1829:
1.47 schwarze 1830: rc = 0;
1831: if (cp <= 0x0000007F) {
1832: rc = 1;
1833: out[0] = (char)cp;
1834: } else if (cp <= 0x000007FF) {
1835: rc = 2;
1836: out[0] = (cp >> 6 & 31) | 192;
1837: out[1] = (cp & 63) | 128;
1838: } else if (cp <= 0x0000FFFF) {
1839: rc = 3;
1840: out[0] = (cp >> 12 & 15) | 224;
1841: out[1] = (cp >> 6 & 63) | 128;
1842: out[2] = (cp & 63) | 128;
1843: } else if (cp <= 0x001FFFFF) {
1844: rc = 4;
1845: out[0] = (cp >> 18 & 7) | 240;
1846: out[1] = (cp >> 12 & 63) | 128;
1847: out[2] = (cp >> 6 & 63) | 128;
1848: out[3] = (cp & 63) | 128;
1849: } else if (cp <= 0x03FFFFFF) {
1850: rc = 5;
1851: out[0] = (cp >> 24 & 3) | 248;
1852: out[1] = (cp >> 18 & 63) | 128;
1853: out[2] = (cp >> 12 & 63) | 128;
1854: out[3] = (cp >> 6 & 63) | 128;
1855: out[4] = (cp & 63) | 128;
1856: } else if (cp <= 0x7FFFFFFF) {
1857: rc = 6;
1858: out[0] = (cp >> 30 & 1) | 252;
1859: out[1] = (cp >> 24 & 63) | 128;
1860: out[2] = (cp >> 18 & 63) | 128;
1861: out[3] = (cp >> 12 & 63) | 128;
1862: out[4] = (cp >> 6 & 63) | 128;
1863: out[5] = (cp & 63) | 128;
1864: } else
1.150 schwarze 1865: return 0;
1.19 schwarze 1866:
1.47 schwarze 1867: out[rc] = '\0';
1.150 schwarze 1868: return rc;
1.1 schwarze 1869: }
1870:
1.47 schwarze 1871: /*
1.131 schwarze 1872: * If the string contains escape sequences,
1873: * replace it with an allocated rendering and return 1,
1874: * such that the caller can free it after use.
1875: * Otherwise, do nothing and return 0.
1.47 schwarze 1876: */
1.131 schwarze 1877: static int
1878: render_string(char **public, size_t *psz)
1.1 schwarze 1879: {
1.131 schwarze 1880: const char *src, *scp, *addcp, *seq;
1881: char *dst;
1882: size_t ssz, dsz, addsz;
1.71 schwarze 1883: char utfbuf[7], res[6];
1.131 schwarze 1884: int seqlen, unicode;
1.47 schwarze 1885:
1886: res[0] = '\\';
1887: res[1] = '\t';
1888: res[2] = ASCII_NBRSP;
1889: res[3] = ASCII_HYPH;
1.71 schwarze 1890: res[4] = ASCII_BREAK;
1891: res[5] = '\0';
1.1 schwarze 1892:
1.131 schwarze 1893: src = scp = *public;
1894: ssz = *psz;
1895: dst = NULL;
1896: dsz = 0;
1897:
1898: while (scp < src + *psz) {
1899:
1900: /* Leave normal characters unchanged. */
1901:
1902: if (strchr(res, *scp) == NULL) {
1903: if (dst != NULL)
1904: dst[dsz++] = *scp;
1905: scp++;
1906: continue;
1907: }
1.1 schwarze 1908:
1.47 schwarze 1909: /*
1.131 schwarze 1910: * Found something that requires replacing,
1911: * make sure we have a destination buffer.
1.47 schwarze 1912: */
1.131 schwarze 1913:
1914: if (dst == NULL) {
1915: dst = mandoc_malloc(ssz + 1);
1916: dsz = scp - src;
1917: memcpy(dst, src, dsz);
1.47 schwarze 1918: }
1.39 schwarze 1919:
1.131 schwarze 1920: /* Handle single-char special characters. */
1921:
1922: switch (*scp) {
1923: case '\\':
1924: break;
1.98 schwarze 1925: case '\t':
1926: case ASCII_NBRSP:
1.131 schwarze 1927: dst[dsz++] = ' ';
1928: scp++;
1929: continue;
1930: case ASCII_HYPH:
1931: dst[dsz++] = '-';
1.71 schwarze 1932: /* FALLTHROUGH */
1.98 schwarze 1933: case ASCII_BREAK:
1.131 schwarze 1934: scp++;
1.47 schwarze 1935: continue;
1.71 schwarze 1936: default:
1.131 schwarze 1937: abort();
1.71 schwarze 1938: }
1.39 schwarze 1939:
1.47 schwarze 1940: /*
1.131 schwarze 1941: * Found an escape sequence.
1942: * Read past the slash, then parse it.
1943: * Ignore everything except characters.
1.47 schwarze 1944: */
1.52 schwarze 1945:
1.131 schwarze 1946: scp++;
1947: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1948: continue;
1.39 schwarze 1949:
1.47 schwarze 1950: /*
1.52 schwarze 1951: * Render the special character
1952: * as either UTF-8 or ASCII.
1.47 schwarze 1953: */
1.52 schwarze 1954:
1955: if (write_utf8) {
1.157 schwarze 1956: unicode = mchars_spec2cp(seq, seqlen);
1.131 schwarze 1957: if (unicode <= 0)
1.52 schwarze 1958: continue;
1.131 schwarze 1959: addsz = utf8(unicode, utfbuf);
1960: if (addsz == 0)
1.52 schwarze 1961: continue;
1.131 schwarze 1962: addcp = utfbuf;
1.52 schwarze 1963: } else {
1.157 schwarze 1964: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.131 schwarze 1965: if (addcp == NULL)
1.52 schwarze 1966: continue;
1.131 schwarze 1967: if (*addcp == ASCII_NBRSP) {
1968: addcp = " ";
1969: addsz = 1;
1.52 schwarze 1970: }
1971: }
1.1 schwarze 1972:
1.47 schwarze 1973: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1974:
1.131 schwarze 1975: ssz += addsz;
1976: dst = mandoc_realloc(dst, ssz + 1);
1977: memcpy(dst + dsz, addcp, addsz);
1978: dsz += addsz;
1979: }
1980: if (dst != NULL) {
1981: *public = dst;
1982: *psz = dsz;
1983: }
1984:
1985: /* Trim trailing whitespace and NUL-terminate. */
1986:
1987: while (*psz > 0 && (*public)[*psz - 1] == ' ')
1988: --*psz;
1989: if (dst != NULL) {
1990: (*public)[*psz] = '\0';
1.150 schwarze 1991: return 1;
1.131 schwarze 1992: } else
1.150 schwarze 1993: return 0;
1.1 schwarze 1994: }
1995:
1.75 schwarze 1996: static void
1997: dbadd_mlink(const struct mlink *mlink)
1998: {
1999: size_t i;
2000:
2001: i = 1;
2002: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
2003: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
2004: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.94 schwarze 2005: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1.75 schwarze 2006: SQL_STEP(stmts[STMT_INSERT_LINK]);
2007: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1.124 schwarze 2008: }
2009:
2010: static void
2011: dbadd_mlink_name(const struct mlink *mlink)
2012: {
1.130 schwarze 2013: uint64_t bits;
1.124 schwarze 2014: size_t i;
2015:
2016: dbadd_mlink(mlink);
1.115 schwarze 2017:
2018: i = 1;
1.130 schwarze 2019: SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
2020: bits = NAME_FILE & NAME_MASK;
2021: if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
2022: bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
2023: sqlite3_reset(stmts[STMT_SELECT_NAME]);
2024: }
2025:
2026: i = 1;
2027: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
1.115 schwarze 2028: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2029: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2030: SQL_STEP(stmts[STMT_INSERT_NAME]);
2031: sqlite3_reset(stmts[STMT_INSERT_NAME]);
1.75 schwarze 2032: }
2033:
1.11 schwarze 2034: /*
1.47 schwarze 2035: * Flush the current page's terms (and their bits) into the database.
2036: * Wrap the entire set of additions in a transaction to make sqlite be a
2037: * little faster.
1.53 schwarze 2038: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 2039: */
2040: static void
1.131 schwarze 2041: dbadd(struct mpage *mpage)
1.11 schwarze 2042: {
1.47 schwarze 2043: struct mlink *mlink;
2044: struct str *key;
1.131 schwarze 2045: char *cp;
1.47 schwarze 2046: size_t i;
2047: unsigned int slot;
1.131 schwarze 2048: int mustfree;
1.47 schwarze 2049:
1.85 schwarze 2050: mlink = mpage->mlinks;
1.11 schwarze 2051:
1.85 schwarze 2052: if (nodb) {
1.104 schwarze 2053: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2054: key = ohash_next(&names, &slot))
1.104 schwarze 2055: free(key);
2056: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2057: key = ohash_next(&strings, &slot))
1.104 schwarze 2058: free(key);
1.102 schwarze 2059: if (0 == debug)
2060: return;
1.85 schwarze 2061: while (NULL != mlink) {
2062: fputs(mlink->name, stdout);
2063: if (NULL == mlink->next ||
2064: strcmp(mlink->dsec, mlink->next->dsec) ||
2065: strcmp(mlink->fsec, mlink->next->fsec) ||
2066: strcmp(mlink->arch, mlink->next->arch)) {
2067: putchar('(');
2068: if ('\0' == *mlink->dsec)
2069: fputs(mlink->fsec, stdout);
2070: else
2071: fputs(mlink->dsec, stdout);
2072: if ('\0' != *mlink->arch)
2073: printf("/%s", mlink->arch);
2074: putchar(')');
2075: }
2076: mlink = mlink->next;
2077: if (NULL != mlink)
2078: fputs(", ", stdout);
2079: }
1.89 schwarze 2080: printf(" - %s\n", mpage->desc);
1.11 schwarze 2081: return;
1.85 schwarze 2082: }
2083:
2084: if (debug)
2085: say(mlink->file, "Adding to database");
1.47 schwarze 2086:
1.131 schwarze 2087: cp = mpage->desc;
2088: i = strlen(cp);
2089: mustfree = render_string(&cp, &i);
1.47 schwarze 2090: i = 1;
1.131 schwarze 2091: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
1.116 schwarze 2092: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
1.47 schwarze 2093: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.94 schwarze 2094: mpage->pageid = sqlite3_last_insert_rowid(db);
1.47 schwarze 2095: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1.131 schwarze 2096: if (mustfree)
2097: free(cp);
1.47 schwarze 2098:
1.85 schwarze 2099: while (NULL != mlink) {
1.75 schwarze 2100: dbadd_mlink(mlink);
1.85 schwarze 2101: mlink = mlink->next;
2102: }
1.91 schwarze 2103: mlink = mpage->mlinks;
1.47 schwarze 2104:
1.90 schwarze 2105: for (key = ohash_first(&names, &slot); NULL != key;
2106: key = ohash_next(&names, &slot)) {
2107: assert(key->mpage == mpage);
2108: i = 1;
2109: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
1.131 schwarze 2110: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
1.94 schwarze 2111: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
1.90 schwarze 2112: SQL_STEP(stmts[STMT_INSERT_NAME]);
2113: sqlite3_reset(stmts[STMT_INSERT_NAME]);
2114: free(key);
2115: }
1.47 schwarze 2116: for (key = ohash_first(&strings, &slot); NULL != key;
2117: key = ohash_next(&strings, &slot)) {
2118: assert(key->mpage == mpage);
2119: i = 1;
2120: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.131 schwarze 2121: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
1.94 schwarze 2122: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
1.47 schwarze 2123: SQL_STEP(stmts[STMT_INSERT_KEY]);
2124: sqlite3_reset(stmts[STMT_INSERT_KEY]);
2125: free(key);
1.33 schwarze 2126: }
1.47 schwarze 2127: }
1.41 deraadt 2128:
1.47 schwarze 2129: static void
2130: dbprune(void)
2131: {
2132: struct mpage *mpage;
2133: struct mlink *mlink;
2134: size_t i;
2135: unsigned int slot;
1.11 schwarze 2136:
1.63 schwarze 2137: if (0 == nodb)
2138: SQL_EXEC("BEGIN TRANSACTION");
1.47 schwarze 2139:
1.63 schwarze 2140: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2141: mpage = ohash_next(&mpages, &slot)) {
1.47 schwarze 2142: mlink = mpage->mlinks;
1.82 schwarze 2143: if (debug)
1.63 schwarze 2144: say(mlink->file, "Deleting from database");
2145: if (nodb)
2146: continue;
2147: for ( ; NULL != mlink; mlink = mlink->next) {
2148: i = 1;
2149: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2150: i, mlink->dsec);
2151: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2152: i, mlink->arch);
2153: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2154: i, mlink->name);
2155: SQL_STEP(stmts[STMT_DELETE_PAGE]);
2156: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2157: }
1.11 schwarze 2158: }
1.63 schwarze 2159:
2160: if (0 == nodb)
2161: SQL_EXEC("END TRANSACTION");
1.47 schwarze 2162: }
1.22 schwarze 2163:
1.47 schwarze 2164: /*
2165: * Close an existing database and its prepared statements.
2166: * If "real" is not set, rename the temporary file into the real one.
2167: */
2168: static void
2169: dbclose(int real)
2170: {
2171: size_t i;
1.72 schwarze 2172: int status;
2173: pid_t child;
1.11 schwarze 2174:
1.47 schwarze 2175: if (nodb)
2176: return;
1.11 schwarze 2177:
1.47 schwarze 2178: for (i = 0; i < STMT__MAX; i++) {
2179: sqlite3_finalize(stmts[i]);
2180: stmts[i] = NULL;
1.28 schwarze 2181: }
1.22 schwarze 2182:
1.47 schwarze 2183: sqlite3_close(db);
2184: db = NULL;
1.11 schwarze 2185:
1.47 schwarze 2186: if (real)
2187: return;
1.22 schwarze 2188:
1.72 schwarze 2189: if ('\0' == *tempfilename) {
2190: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2191: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2192: say(MANDOC_DB, "&rename");
1.72 schwarze 2193: }
2194: return;
2195: }
2196:
2197: switch (child = fork()) {
1.98 schwarze 2198: case -1:
1.72 schwarze 2199: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2200: say("", "&fork cmp");
1.72 schwarze 2201: return;
1.98 schwarze 2202: case 0:
1.72 schwarze 2203: execlp("cmp", "cmp", "-s",
1.149 schwarze 2204: tempfilename, MANDOC_DB, (char *)NULL);
1.80 schwarze 2205: say("", "&exec cmp");
1.72 schwarze 2206: exit(0);
2207: default:
2208: break;
2209: }
2210: if (-1 == waitpid(child, &status, 0)) {
2211: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2212: say("", "&wait cmp");
1.72 schwarze 2213: } else if (WIFSIGNALED(status)) {
2214: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2215: say("", "cmp died from signal %d", WTERMSIG(status));
1.72 schwarze 2216: } else if (WEXITSTATUS(status)) {
1.47 schwarze 2217: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2218: say(MANDOC_DB,
2219: "Data changed, but cannot replace database");
2220: }
2221:
2222: *strrchr(tempfilename, '/') = '\0';
2223: switch (child = fork()) {
1.98 schwarze 2224: case -1:
1.72 schwarze 2225: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2226: say("", "&fork rm");
1.72 schwarze 2227: return;
1.98 schwarze 2228: case 0:
1.149 schwarze 2229: execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
1.80 schwarze 2230: say("", "&exec rm");
1.72 schwarze 2231: exit((int)MANDOCLEVEL_SYSERR);
2232: default:
2233: break;
2234: }
2235: if (-1 == waitpid(child, &status, 0)) {
2236: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2237: say("", "&wait rm");
1.72 schwarze 2238: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2239: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2240: say("", "%s: Cannot remove temporary directory",
2241: tempfilename);
1.22 schwarze 2242: }
1.11 schwarze 2243: }
2244:
1.47 schwarze 2245: /*
2246: * This is straightforward stuff.
2247: * Open a database connection to a "temporary" database, then open a set
2248: * of prepared statements we'll use over and over again.
2249: * If "real" is set, we use the existing database; if not, we truncate a
2250: * temporary one.
2251: * Must be matched by dbclose().
2252: */
2253: static int
2254: dbopen(int real)
1.2 schwarze 2255: {
1.72 schwarze 2256: const char *sql;
1.47 schwarze 2257: int rc, ofl;
1.6 schwarze 2258:
1.98 schwarze 2259: if (nodb)
1.150 schwarze 2260: return 1;
1.6 schwarze 2261:
1.72 schwarze 2262: *tempfilename = '\0';
1.47 schwarze 2263: ofl = SQLITE_OPEN_READWRITE;
1.72 schwarze 2264:
2265: if (real) {
2266: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2267: if (SQLITE_OK != rc) {
1.47 schwarze 2268: exitcode = (int)MANDOCLEVEL_SYSERR;
1.108 sthen 2269: if (SQLITE_CANTOPEN != rc)
2270: say(MANDOC_DB, "%s", sqlite3_errstr(rc));
1.150 schwarze 2271: return 0;
1.28 schwarze 2272: }
1.72 schwarze 2273: goto prepare_statements;
2274: }
2275:
2276: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.6 schwarze 2277:
1.72 schwarze 2278: remove(MANDOC_DB "~");
2279: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.98 schwarze 2280: if (SQLITE_OK == rc)
1.72 schwarze 2281: goto create_tables;
1.73 schwarze 2282: if (MPARSE_QUICK & mparse_options) {
1.47 schwarze 2283: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2284: say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
1.150 schwarze 2285: return 0;
1.47 schwarze 2286: }
1.6 schwarze 2287:
1.100 schwarze 2288: (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2289: sizeof(tempfilename));
1.72 schwarze 2290: if (NULL == mkdtemp(tempfilename)) {
2291: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2292: say("", "&%s", tempfilename);
1.150 schwarze 2293: return 0;
1.72 schwarze 2294: }
1.100 schwarze 2295: (void)strlcat(tempfilename, "/" MANDOC_DB,
2296: sizeof(tempfilename));
1.72 schwarze 2297: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2298: if (SQLITE_OK != rc) {
1.47 schwarze 2299: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2300: say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
1.150 schwarze 2301: return 0;
1.2 schwarze 2302: }
2303:
1.72 schwarze 2304: create_tables:
1.47 schwarze 2305: sql = "CREATE TABLE \"mpages\" (\n"
1.89 schwarze 2306: " \"desc\" TEXT NOT NULL,\n"
1.47 schwarze 2307: " \"form\" INTEGER NOT NULL,\n"
1.94 schwarze 2308: " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1.47 schwarze 2309: ");\n"
2310: "\n"
2311: "CREATE TABLE \"mlinks\" (\n"
2312: " \"sec\" TEXT NOT NULL,\n"
2313: " \"arch\" TEXT NOT NULL,\n"
2314: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2315: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2316: "ON DELETE CASCADE\n"
1.47 schwarze 2317: ");\n"
1.93 schwarze 2318: "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
1.47 schwarze 2319: "\n"
1.90 schwarze 2320: "CREATE TABLE \"names\" (\n"
2321: " \"bits\" INTEGER NOT NULL,\n"
2322: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2323: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.130 schwarze 2324: "ON DELETE CASCADE,\n"
2325: " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
1.90 schwarze 2326: ");\n"
2327: "\n"
1.47 schwarze 2328: "CREATE TABLE \"keys\" (\n"
2329: " \"bits\" INTEGER NOT NULL,\n"
2330: " \"key\" TEXT NOT NULL,\n"
1.94 schwarze 2331: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2332: "ON DELETE CASCADE\n"
1.93 schwarze 2333: ");\n"
2334: "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
1.47 schwarze 2335:
2336: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2337: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2338: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.103 schwarze 2339: sqlite3_close(db);
1.150 schwarze 2340: return 0;
1.2 schwarze 2341: }
2342:
1.47 schwarze 2343: prepare_statements:
1.103 schwarze 2344: if (SQLITE_OK != sqlite3_exec(db,
2345: "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2346: exitcode = (int)MANDOCLEVEL_SYSERR;
2347: say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2348: sqlite3_errmsg(db));
2349: sqlite3_close(db);
1.150 schwarze 2350: return 0;
1.103 schwarze 2351: }
2352:
1.94 schwarze 2353: sql = "DELETE FROM mpages WHERE pageid IN "
1.63 schwarze 2354: "(SELECT pageid FROM mlinks WHERE "
2355: "sec=? AND arch=? AND name=?)";
1.47 schwarze 2356: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2357: sql = "INSERT INTO mpages "
1.89 schwarze 2358: "(desc,form) VALUES (?,?)";
1.47 schwarze 2359: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2360: sql = "INSERT INTO mlinks "
1.61 schwarze 2361: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.47 schwarze 2362: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.130 schwarze 2363: sql = "SELECT bits FROM names where pageid = ?";
2364: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
1.90 schwarze 2365: sql = "INSERT INTO names "
2366: "(bits,name,pageid) VALUES (?,?,?)";
2367: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
1.47 schwarze 2368: sql = "INSERT INTO keys "
2369: "(bits,key,pageid) VALUES (?,?,?)";
2370: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.6 schwarze 2371:
1.47 schwarze 2372: /*
2373: * When opening a new database, we can turn off
2374: * synchronous mode for much better performance.
2375: */
1.6 schwarze 2376:
1.103 schwarze 2377: if (real && SQLITE_OK != sqlite3_exec(db,
2378: "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2379: exitcode = (int)MANDOCLEVEL_SYSERR;
2380: say(MANDOC_DB, "PRAGMA synchronous: %s",
2381: sqlite3_errmsg(db));
2382: sqlite3_close(db);
1.150 schwarze 2383: return 0;
1.103 schwarze 2384: }
1.11 schwarze 2385:
1.150 schwarze 2386: return 1;
1.47 schwarze 2387: }
1.6 schwarze 2388:
1.47 schwarze 2389: static int
1.120 schwarze 2390: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2391: {
2392: static char startdir[PATH_MAX];
1.110 schwarze 2393: static int getcwd_status; /* 1 = ok, 2 = failure */
2394: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2395: char *cp;
1.6 schwarze 2396:
1.47 schwarze 2397: /*
1.110 schwarze 2398: * Remember the original working directory, if possible.
2399: * This will be needed if the second or a later directory
2400: * on the command line is given as a relative path.
2401: * Do not error out if the current directory is not
2402: * searchable: Maybe it won't be needed after all.
2403: */
2404: if (0 == getcwd_status) {
2405: if (NULL == getcwd(startdir, sizeof(startdir))) {
2406: getcwd_status = 2;
2407: (void)strlcpy(startdir, strerror(errno),
2408: sizeof(startdir));
2409: } else
2410: getcwd_status = 1;
2411: }
2412:
2413: /*
2414: * We are leaving the old base directory.
2415: * Do not use it any longer, not even for messages.
2416: */
2417: *basedir = '\0';
2418:
2419: /*
2420: * If and only if the directory was changed earlier and
2421: * the next directory to process is given as a relative path,
2422: * first go back, or bail out if that is impossible.
1.47 schwarze 2423: */
1.110 schwarze 2424: if (chdir_status && '/' != *targetdir) {
2425: if (2 == getcwd_status) {
1.47 schwarze 2426: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2427: say("", "getcwd: %s", startdir);
1.150 schwarze 2428: return 0;
1.47 schwarze 2429: }
1.110 schwarze 2430: if (-1 == chdir(startdir)) {
1.47 schwarze 2431: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2432: say("", "&chdir %s", startdir);
1.150 schwarze 2433: return 0;
1.2 schwarze 2434: }
2435: }
1.110 schwarze 2436:
2437: /*
2438: * Always resolve basedir to the canonicalized absolute
2439: * pathname and append a trailing slash, such that
2440: * we can reliably check whether files are inside.
2441: */
1.47 schwarze 2442: if (NULL == realpath(targetdir, basedir)) {
1.120 schwarze 2443: if (report_baddir || errno != ENOENT) {
2444: exitcode = (int)MANDOCLEVEL_BADARG;
2445: say("", "&%s: realpath", targetdir);
2446: }
1.150 schwarze 2447: return 0;
1.47 schwarze 2448: } else if (-1 == chdir(basedir)) {
1.120 schwarze 2449: if (report_baddir || errno != ENOENT) {
2450: exitcode = (int)MANDOCLEVEL_BADARG;
2451: say("", "&chdir");
2452: }
1.150 schwarze 2453: return 0;
1.109 schwarze 2454: }
1.110 schwarze 2455: chdir_status = 1;
1.109 schwarze 2456: cp = strchr(basedir, '\0');
2457: if ('/' != cp[-1]) {
2458: if (cp - basedir >= PATH_MAX - 1) {
2459: exitcode = (int)MANDOCLEVEL_SYSERR;
2460: say("", "Filename too long");
1.150 schwarze 2461: return 0;
1.109 schwarze 2462: }
2463: *cp++ = '/';
2464: *cp = '\0';
1.47 schwarze 2465: }
1.150 schwarze 2466: return 1;
1.2 schwarze 2467: }
2468:
2469: static void
1.47 schwarze 2470: say(const char *file, const char *format, ...)
1.2 schwarze 2471: {
1.47 schwarze 2472: va_list ap;
1.80 schwarze 2473: int use_errno;
1.2 schwarze 2474:
1.47 schwarze 2475: if ('\0' != *basedir)
2476: fprintf(stderr, "%s", basedir);
2477: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2478: fputc('/', stderr);
1.47 schwarze 2479: if ('\0' != *file)
2480: fprintf(stderr, "%s", file);
1.31 schwarze 2481:
1.80 schwarze 2482: use_errno = 1;
2483: if (NULL != format) {
2484: switch (*format) {
1.98 schwarze 2485: case '&':
1.80 schwarze 2486: format++;
2487: break;
1.98 schwarze 2488: case '\0':
1.80 schwarze 2489: format = NULL;
2490: break;
2491: default:
2492: use_errno = 0;
2493: break;
2494: }
2495: }
2496: if (NULL != format) {
2497: if ('\0' != *basedir || '\0' != *file)
2498: fputs(": ", stderr);
2499: va_start(ap, format);
2500: vfprintf(stderr, format, ap);
2501: va_end(ap);
2502: }
2503: if (use_errno) {
2504: if ('\0' != *basedir || '\0' != *file || NULL != format)
2505: fputs(": ", stderr);
1.47 schwarze 2506: perror(NULL);
1.80 schwarze 2507: } else
2508: fputc('\n', stderr);
1.1 schwarze 2509: }
![[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