Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.159
1.159 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.158 2015/10/16 13:37:44 millert Exp $ */
1.1 schwarze 2: /*
1.47 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.136 schwarze 4: * Copyright (c) 2011-2015 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>
1.1 schwarze 28: #include <getopt.h>
1.44 schwarze 29: #include <limits.h>
1.47 schwarze 30: #include <stddef.h>
1.1 schwarze 31: #include <stdio.h>
32: #include <stdint.h>
33: #include <stdlib.h>
34: #include <string.h>
1.14 schwarze 35: #include <unistd.h>
1.1 schwarze 36:
1.47 schwarze 37: #include <sqlite3.h>
38:
1.141 schwarze 39: #include "mandoc_aux.h"
1.156 schwarze 40: #include "mandoc_ohash.h"
1.141 schwarze 41: #include "mandoc.h"
42: #include "roff.h"
1.47 schwarze 43: #include "mdoc.h"
1.1 schwarze 44: #include "man.h"
1.140 schwarze 45: #include "manconf.h"
1.47 schwarze 46: #include "mansearch.h"
1.1 schwarze 47:
1.68 schwarze 48: extern int mansearch_keymax;
49: extern const char *const mansearch_keynames[];
50:
1.47 schwarze 51: #define SQL_EXEC(_v) \
52: if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
1.91 schwarze 53: say("", "%s: %s", (_v), sqlite3_errmsg(db))
1.47 schwarze 54: #define SQL_BIND_TEXT(_s, _i, _v) \
55: if (SQLITE_OK != sqlite3_bind_text \
56: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.91 schwarze 57: say(mlink->file, "%s", sqlite3_errmsg(db))
1.47 schwarze 58: #define SQL_BIND_INT(_s, _i, _v) \
59: if (SQLITE_OK != sqlite3_bind_int \
60: ((_s), (_i)++, (_v))) \
1.91 schwarze 61: say(mlink->file, "%s", sqlite3_errmsg(db))
1.47 schwarze 62: #define SQL_BIND_INT64(_s, _i, _v) \
63: if (SQLITE_OK != sqlite3_bind_int64 \
64: ((_s), (_i)++, (_v))) \
1.91 schwarze 65: say(mlink->file, "%s", sqlite3_errmsg(db))
1.47 schwarze 66: #define SQL_STEP(_s) \
67: if (SQLITE_DONE != sqlite3_step((_s))) \
1.91 schwarze 68: say(mlink->file, "%s", sqlite3_errmsg(db))
1.1 schwarze 69:
1.47 schwarze 70: enum op {
71: OP_DEFAULT = 0, /* new dbs from dir list or default config */
72: OP_CONFFILE, /* new databases from custom config file */
73: OP_UPDATE, /* delete/add entries in existing database */
74: OP_DELETE, /* delete entries from existing database */
75: OP_TEST /* change no databases, report potential problems */
76: };
1.11 schwarze 77:
1.47 schwarze 78: struct str {
79: const struct mpage *mpage; /* if set, the owning parse */
80: uint64_t mask; /* bitmask in sequence */
1.131 schwarze 81: char key[]; /* rendered text */
1.28 schwarze 82: };
83:
1.47 schwarze 84: struct inodev {
85: ino_t st_ino;
86: dev_t st_dev;
87: };
1.28 schwarze 88:
1.47 schwarze 89: struct mpage {
90: struct inodev inodev; /* used for hashing routine */
1.94 schwarze 91: int64_t pageid; /* pageid in mpages SQL table */
1.47 schwarze 92: char *sec; /* section from file content */
93: char *arch; /* architecture from file content */
94: char *title; /* title from file content */
95: char *desc; /* description from file content */
96: struct mlink *mlinks; /* singly linked list */
1.116 schwarze 97: int form; /* format from file content */
1.129 schwarze 98: int name_head_done;
1.28 schwarze 99: };
100:
1.47 schwarze 101: struct mlink {
102: char file[PATH_MAX]; /* filename rel. to manpath */
103: char *dsec; /* section from directory */
104: char *arch; /* architecture from directory */
105: char *name; /* name from file name (not empty) */
106: char *fsec; /* section from file name suffix */
107: struct mlink *next; /* singly linked list */
1.75 schwarze 108: struct mpage *mpage; /* parent */
1.116 schwarze 109: int dform; /* format from directory */
110: int fform; /* format from file name suffix */
1.81 schwarze 111: int gzip; /* filename has a .gz suffix */
1.2 schwarze 112: };
113:
1.47 schwarze 114: enum stmt {
115: STMT_DELETE_PAGE = 0, /* delete mpage */
116: STMT_INSERT_PAGE, /* insert mpage */
117: STMT_INSERT_LINK, /* insert mlink */
1.90 schwarze 118: STMT_INSERT_NAME, /* insert name */
1.130 schwarze 119: STMT_SELECT_NAME, /* retrieve existing name flags */
1.47 schwarze 120: STMT_INSERT_KEY, /* insert parsed key */
121: STMT__MAX
1.1 schwarze 122: };
123:
1.143 schwarze 124: typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
1.142 schwarze 125: const struct roff_node *);
1.1 schwarze 126:
1.19 schwarze 127: struct mdoc_handler {
1.47 schwarze 128: mdoc_fp fp; /* optional handler */
129: uint64_t mask; /* set unless handler returns 0 */
1.19 schwarze 130: };
131:
1.47 schwarze 132: static void dbclose(int);
1.131 schwarze 133: static void dbadd(struct mpage *);
1.75 schwarze 134: static void dbadd_mlink(const struct mlink *mlink);
1.124 schwarze 135: static void dbadd_mlink_name(const struct mlink *mlink);
1.47 schwarze 136: static int dbopen(int);
137: static void dbprune(void);
138: static void filescan(const char *);
139: static void mlink_add(struct mlink *, const struct stat *);
1.88 schwarze 140: static void mlink_check(struct mpage *, struct mlink *);
1.47 schwarze 141: static void mlink_free(struct mlink *);
142: static void mlinks_undupe(struct mpage *);
143: static void mpages_free(void);
1.131 schwarze 144: static void mpages_merge(struct mparse *);
1.87 schwarze 145: static void names_check(void);
1.81 schwarze 146: static void parse_cat(struct mpage *, int);
1.143 schwarze 147: static void parse_man(struct mpage *, const struct roff_meta *,
1.142 schwarze 148: const struct roff_node *);
1.143 schwarze 149: static void parse_mdoc(struct mpage *, const struct roff_meta *,
1.142 schwarze 150: const struct roff_node *);
1.143 schwarze 151: static int parse_mdoc_body(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 *);
1.143 schwarze 168: static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
1.142 schwarze 169: const struct roff_node *);
1.69 schwarze 170: static void putkey(const struct mpage *, char *, uint64_t);
1.131 schwarze 171: static void putkeys(const struct mpage *, char *, size_t, uint64_t);
1.47 schwarze 172: static void putmdockey(const struct mpage *,
1.142 schwarze 173: const struct roff_node *, uint64_t);
1.131 schwarze 174: static int render_string(char **, size_t *);
1.47 schwarze 175: static void say(const char *, const char *, ...);
1.120 schwarze 176: static int set_basedir(const char *, int);
1.47 schwarze 177: static int treescan(void);
178: static size_t utf8(unsigned int, char [7]);
179:
1.151 schwarze 180: extern char *__progname;
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 */
1.47 schwarze 239: { NULL, TYPE_Va }, /* Va */
240: { parse_mdoc_body, TYPE_Va }, /* Vt */
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;
332: const char *path_arg;
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) {
1.155 schwarze 337: perror("pledge");
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:
1.155 schwarze 414: if (nodb && pledge("stdio rpath", NULL) == -1) {
415: perror("pledge");
416: return (int)MANDOCLEVEL_SYSERR;
417: }
418:
1.28 schwarze 419: if (OP_CONFFILE == op && argc > 0) {
1.151 schwarze 420: warnx("-C: Too many arguments");
1.28 schwarze 421: goto usage;
422: }
423:
1.47 schwarze 424: exitcode = (int)MANDOCLEVEL_OK;
1.157 schwarze 425: mchars_alloc();
426: mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
1.156 schwarze 427: mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
428: mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
1.2 schwarze 429:
1.47 schwarze 430: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.2 schwarze 431:
1.47 schwarze 432: /*
1.109 schwarze 433: * Most of these deal with a specific directory.
1.97 schwarze 434: * Jump into that directory first.
1.47 schwarze 435: */
1.120 schwarze 436: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.47 schwarze 437: goto out;
1.97 schwarze 438:
1.95 schwarze 439: if (dbopen(1)) {
1.97 schwarze 440: /*
441: * The existing database is usable. Process
442: * all files specified on the command-line.
443: */
1.158 millert 444: if (!nodb && pledge("stdio rpath wpath cpath fattr flock",
1.155 schwarze 445: NULL) == -1) {
446: perror("pledge");
447: exitcode = (int)MANDOCLEVEL_SYSERR;
448: goto out;
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:
1.83 schwarze 539: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
540: " %s [-aDnpQ] [-Tutf8] dir ...\n"
541: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
542: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 543: " %s [-Q] -t file ...\n",
1.151 schwarze 544: __progname, __progname, __progname,
545: __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: *
562: * TODO: accomodate for multi-language directories.
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:
1.136 schwarze 1114: mparse_open(mp, &fd, mlink->file);
1.117 schwarze 1115: if (fd == -1) {
1.136 schwarze 1116: say(mlink->file, "&open");
1.117 schwarze 1117: goto nextpage;
1.81 schwarze 1118: }
1.11 schwarze 1119:
1120: /*
1.137 schwarze 1121: * Interpret the file as mdoc(7) or man(7) source
1122: * code, unless it is known to be formatted.
1.11 schwarze 1123: */
1.136 schwarze 1124: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.137 schwarze 1125: mparse_readfd(mp, fd, mlink->file);
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) {
1.47 schwarze 1181: mpage->form = FORM_SRC;
1.146 schwarze 1182: mpage->sec = mandoc_strdup(man->meta.msec);
1.136 schwarze 1183: mpage->arch = mandoc_strdup(mlink->arch);
1.146 schwarze 1184: mpage->title = mandoc_strdup(man->meta.title);
1.11 schwarze 1185: } else {
1.47 schwarze 1186: mpage->form = FORM_CAT;
1.136 schwarze 1187: mpage->sec = mandoc_strdup(mlink->dsec);
1188: mpage->arch = mandoc_strdup(mlink->arch);
1189: mpage->title = mandoc_strdup(mlink->name);
1.1 schwarze 1190: }
1.54 schwarze 1191: putkey(mpage, mpage->sec, TYPE_sec);
1.119 schwarze 1192: if (*mpage->arch != '\0')
1193: putkey(mpage, mpage->arch, TYPE_arch);
1.1 schwarze 1194:
1.136 schwarze 1195: for ( ; mlink != NULL; mlink = mlink->next) {
1.54 schwarze 1196: if ('\0' != *mlink->dsec)
1197: putkey(mpage, mlink->dsec, TYPE_sec);
1198: if ('\0' != *mlink->fsec)
1199: putkey(mpage, mlink->fsec, TYPE_sec);
1.55 schwarze 1200: putkey(mpage, '\0' == *mlink->arch ?
1.70 schwarze 1201: any : mlink->arch, TYPE_arch);
1.90 schwarze 1202: putkey(mpage, mlink->name, NAME_FILE);
1.54 schwarze 1203: }
1.41 deraadt 1204:
1.129 schwarze 1205: assert(mpage->desc == NULL);
1.145 schwarze 1206: if (man != NULL && man->macroset == MACROSET_MDOC)
1.146 schwarze 1207: parse_mdoc(mpage, &man->meta, man->first);
1.129 schwarze 1208: else if (man != NULL)
1.146 schwarze 1209: parse_man(mpage, &man->meta, man->first);
1.47 schwarze 1210: else
1.117 schwarze 1211: parse_cat(mpage, fd);
1.129 schwarze 1212: if (mpage->desc == NULL)
1.92 schwarze 1213: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1.88 schwarze 1214:
1215: if (warnings && !use_all)
1216: for (mlink = mpage->mlinks; mlink;
1217: mlink = mlink->next)
1218: mlink_check(mpage, mlink);
1.6 schwarze 1219:
1.131 schwarze 1220: dbadd(mpage);
1.136 schwarze 1221: mlink = mpage->mlinks;
1.81 schwarze 1222:
1223: nextpage:
1.47 schwarze 1224: ohash_delete(&strings);
1.90 schwarze 1225: ohash_delete(&names);
1.47 schwarze 1226: mpage = ohash_next(&mpages, &pslot);
1227: }
1.64 schwarze 1228:
1229: if (0 == nodb)
1230: SQL_EXEC("END TRANSACTION");
1.87 schwarze 1231: }
1232:
1233: static void
1234: names_check(void)
1235: {
1236: sqlite3_stmt *stmt;
1237: const char *name, *sec, *arch, *key;
1238:
1239: sqlite3_prepare_v2(db,
1240: "SELECT name, sec, arch, key FROM ("
1.90 schwarze 1241: "SELECT name AS key, pageid FROM names "
1.87 schwarze 1242: "WHERE bits & ? AND NOT EXISTS ("
1243: "SELECT pageid FROM mlinks "
1.90 schwarze 1244: "WHERE mlinks.pageid == names.pageid "
1245: "AND mlinks.name == names.name"
1.87 schwarze 1246: ")"
1247: ") JOIN ("
1.99 schwarze 1248: "SELECT sec, arch, name, pageid FROM mlinks "
1249: "GROUP BY pageid"
1.87 schwarze 1250: ") USING (pageid);",
1251: -1, &stmt, NULL);
1252:
1.154 schwarze 1253: if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK)
1.91 schwarze 1254: say("", "%s", sqlite3_errmsg(db));
1.87 schwarze 1255:
1.154 schwarze 1256: while (sqlite3_step(stmt) == SQLITE_ROW) {
1.113 schwarze 1257: name = (const char *)sqlite3_column_text(stmt, 0);
1258: sec = (const char *)sqlite3_column_text(stmt, 1);
1259: arch = (const char *)sqlite3_column_text(stmt, 2);
1260: key = (const char *)sqlite3_column_text(stmt, 3);
1.87 schwarze 1261: say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1262: '\0' == *arch ? "" : "/",
1263: '\0' == *arch ? "" : arch, key);
1264: }
1265: sqlite3_finalize(stmt);
1.47 schwarze 1266: }
1.6 schwarze 1267:
1.47 schwarze 1268: static void
1.81 schwarze 1269: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1270: {
1271: FILE *stream;
1272: char *line, *p, *title;
1273: size_t len, plen, titlesz;
1.1 schwarze 1274:
1.81 schwarze 1275: stream = (-1 == fd) ?
1276: fopen(mpage->mlinks->file, "r") :
1277: fdopen(fd, "r");
1278: if (NULL == stream) {
1.114 doug 1279: if (-1 != fd)
1280: close(fd);
1.47 schwarze 1281: if (warnings)
1.80 schwarze 1282: say(mpage->mlinks->file, "&fopen");
1.47 schwarze 1283: return;
1284: }
1.1 schwarze 1285:
1.47 schwarze 1286: /* Skip to first blank line. */
1.1 schwarze 1287:
1.47 schwarze 1288: while (NULL != (line = fgetln(stream, &len)))
1289: if ('\n' == *line)
1290: break;
1.1 schwarze 1291:
1.47 schwarze 1292: /*
1293: * Assume the first line that is not indented
1294: * is the first section header. Skip to it.
1295: */
1.1 schwarze 1296:
1.47 schwarze 1297: while (NULL != (line = fgetln(stream, &len)))
1298: if ('\n' != *line && ' ' != *line)
1299: break;
1.98 schwarze 1300:
1.47 schwarze 1301: /*
1302: * Read up until the next section into a buffer.
1303: * Strip the leading and trailing newline from each read line,
1304: * appending a trailing space.
1305: * Ignore empty (whitespace-only) lines.
1306: */
1.28 schwarze 1307:
1.47 schwarze 1308: titlesz = 0;
1309: title = NULL;
1.38 schwarze 1310:
1.47 schwarze 1311: while (NULL != (line = fgetln(stream, &len))) {
1312: if (' ' != *line || '\n' != line[len - 1])
1313: break;
1314: while (len > 0 && isspace((unsigned char)*line)) {
1315: line++;
1316: len--;
1317: }
1318: if (1 == len)
1319: continue;
1320: title = mandoc_realloc(title, titlesz + len);
1321: memcpy(title + titlesz, line, len);
1322: titlesz += len;
1323: title[titlesz - 1] = ' ';
1324: }
1.28 schwarze 1325:
1.47 schwarze 1326: /*
1327: * If no page content can be found, or the input line
1328: * is already the next section header, or there is no
1329: * trailing newline, reuse the page title as the page
1330: * description.
1331: */
1.1 schwarze 1332:
1.47 schwarze 1333: if (NULL == title || '\0' == *title) {
1334: if (warnings)
1335: say(mpage->mlinks->file,
1336: "Cannot find NAME section");
1337: fclose(stream);
1338: free(title);
1339: return;
1340: }
1.24 schwarze 1341:
1.47 schwarze 1342: title = mandoc_realloc(title, titlesz + 1);
1343: title[titlesz] = '\0';
1.24 schwarze 1344:
1.47 schwarze 1345: /*
1346: * Skip to the first dash.
1347: * Use the remaining line as the description (no more than 70
1348: * bytes).
1349: */
1.28 schwarze 1350:
1.47 schwarze 1351: if (NULL != (p = strstr(title, "- "))) {
1352: for (p += 2; ' ' == *p || '\b' == *p; p++)
1353: /* Skip to next word. */ ;
1354: } else {
1355: if (warnings)
1356: say(mpage->mlinks->file,
1357: "No dash in title line");
1358: p = title;
1359: }
1.1 schwarze 1360:
1.47 schwarze 1361: plen = strlen(p);
1.1 schwarze 1362:
1.47 schwarze 1363: /* Strip backspace-encoding from line. */
1.1 schwarze 1364:
1.47 schwarze 1365: while (NULL != (line = memchr(p, '\b', plen))) {
1366: len = line - p;
1367: if (0 == len) {
1368: memmove(line, line + 1, plen--);
1369: continue;
1.98 schwarze 1370: }
1.47 schwarze 1371: memmove(line - 1, line + 1, plen - len);
1372: plen -= 2;
1373: }
1.1 schwarze 1374:
1.47 schwarze 1375: mpage->desc = mandoc_strdup(p);
1376: fclose(stream);
1377: free(title);
1378: }
1.16 schwarze 1379:
1.47 schwarze 1380: /*
1381: * Put a type/word pair into the word database for this particular file.
1382: */
1383: static void
1.69 schwarze 1384: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1385: {
1.69 schwarze 1386: char *cp;
1.37 schwarze 1387:
1.47 schwarze 1388: assert(NULL != value);
1.69 schwarze 1389: if (TYPE_arch == type)
1390: for (cp = value; *cp; cp++)
1391: if (isupper((unsigned char)*cp))
1392: *cp = _tolower((unsigned char)*cp);
1.47 schwarze 1393: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1394: }
1395:
1396: /*
1.47 schwarze 1397: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1398: */
1399: static void
1.47 schwarze 1400: putmdockey(const struct mpage *mpage,
1.142 schwarze 1401: const struct roff_node *n, uint64_t m)
1.2 schwarze 1402: {
1.16 schwarze 1403:
1.47 schwarze 1404: for ( ; NULL != n; n = n->next) {
1405: if (NULL != n->child)
1406: putmdockey(mpage, n->child, m);
1.141 schwarze 1407: if (n->type == ROFFT_TEXT)
1.47 schwarze 1408: putkey(mpage, n->string, m);
1409: }
1410: }
1.16 schwarze 1411:
1.47 schwarze 1412: static void
1.143 schwarze 1413: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1414: const struct roff_node *n)
1.47 schwarze 1415: {
1.142 schwarze 1416: const struct roff_node *head, *body;
1.78 schwarze 1417: char *start, *title;
1.47 schwarze 1418: char byte;
1.78 schwarze 1419: size_t sz;
1.16 schwarze 1420:
1.47 schwarze 1421: if (NULL == n)
1422: return;
1.16 schwarze 1423:
1.47 schwarze 1424: /*
1425: * We're only searching for one thing: the first text child in
1426: * the BODY of a NAME section. Since we don't keep track of
1427: * sections in -man, run some hoops to find out whether we're in
1428: * the correct section or not.
1429: */
1.16 schwarze 1430:
1.141 schwarze 1431: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1432: body = n;
1433: assert(body->parent);
1434: if (NULL != (head = body->parent->head) &&
1.98 schwarze 1435: 1 == head->nchild &&
1436: NULL != (head = (head->child)) &&
1.141 schwarze 1437: head->type == ROFFT_TEXT &&
1.98 schwarze 1438: 0 == strcmp(head->string, "NAME") &&
1439: NULL != body->child) {
1.2 schwarze 1440:
1.47 schwarze 1441: /*
1442: * Suck the entire NAME section into memory.
1443: * Yes, we might run away.
1444: * But too many manuals have big, spread-out
1445: * NAME sections over many lines.
1446: */
1.2 schwarze 1447:
1.78 schwarze 1448: title = NULL;
1.147 schwarze 1449: deroff(&title, body);
1.47 schwarze 1450: if (NULL == title)
1451: return;
1.16 schwarze 1452:
1.98 schwarze 1453: /*
1.47 schwarze 1454: * Go through a special heuristic dance here.
1455: * Conventionally, one or more manual names are
1456: * comma-specified prior to a whitespace, then a
1457: * dash, then a description. Try to puzzle out
1458: * the name parts here.
1459: */
1.16 schwarze 1460:
1.78 schwarze 1461: start = title;
1.47 schwarze 1462: for ( ;; ) {
1463: sz = strcspn(start, " ,");
1464: if ('\0' == start[sz])
1465: break;
1.1 schwarze 1466:
1.47 schwarze 1467: byte = start[sz];
1468: start[sz] = '\0';
1.67 schwarze 1469:
1470: /*
1471: * Assume a stray trailing comma in the
1472: * name list if a name begins with a dash.
1473: */
1474:
1475: if ('-' == start[0] ||
1476: ('\\' == start[0] && '-' == start[1]))
1477: break;
1.1 schwarze 1478:
1.90 schwarze 1479: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1480: if ( ! (mpage->name_head_done ||
1481: strcasecmp(start, meta->title))) {
1482: putkey(mpage, start, NAME_HEAD);
1483: mpage->name_head_done = 1;
1484: }
1.1 schwarze 1485:
1.47 schwarze 1486: if (' ' == byte) {
1487: start += sz + 1;
1488: break;
1489: }
1.1 schwarze 1490:
1.47 schwarze 1491: assert(',' == byte);
1492: start += sz + 1;
1493: while (' ' == *start)
1494: start++;
1495: }
1.1 schwarze 1496:
1.78 schwarze 1497: if (start == title) {
1.90 schwarze 1498: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1499: if ( ! (mpage->name_head_done ||
1500: strcasecmp(start, meta->title))) {
1501: putkey(mpage, start, NAME_HEAD);
1502: mpage->name_head_done = 1;
1503: }
1.47 schwarze 1504: free(title);
1505: return;
1506: }
1.1 schwarze 1507:
1.47 schwarze 1508: while (isspace((unsigned char)*start))
1509: start++;
1.1 schwarze 1510:
1.47 schwarze 1511: if (0 == strncmp(start, "-", 1))
1512: start += 1;
1513: else if (0 == strncmp(start, "\\-\\-", 4))
1514: start += 4;
1515: else if (0 == strncmp(start, "\\-", 2))
1516: start += 2;
1517: else if (0 == strncmp(start, "\\(en", 4))
1518: start += 4;
1519: else if (0 == strncmp(start, "\\(em", 4))
1520: start += 4;
1.1 schwarze 1521:
1.47 schwarze 1522: while (' ' == *start)
1523: start++;
1.1 schwarze 1524:
1.47 schwarze 1525: mpage->desc = mandoc_strdup(start);
1526: free(title);
1527: return;
1528: }
1529: }
1.1 schwarze 1530:
1.47 schwarze 1531: for (n = n->child; n; n = n->next) {
1532: if (NULL != mpage->desc)
1533: break;
1.129 schwarze 1534: parse_man(mpage, meta, n);
1.1 schwarze 1535: }
1536: }
1537:
1538: static void
1.143 schwarze 1539: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1540: const struct roff_node *n)
1.1 schwarze 1541: {
1542:
1.47 schwarze 1543: assert(NULL != n);
1544: for (n = n->child; NULL != n; n = n->next) {
1545: switch (n->type) {
1.141 schwarze 1546: case ROFFT_ELEM:
1547: case ROFFT_BLOCK:
1548: case ROFFT_HEAD:
1549: case ROFFT_BODY:
1550: case ROFFT_TAIL:
1.47 schwarze 1551: if (NULL != mdocs[n->tok].fp)
1.127 schwarze 1552: if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1.47 schwarze 1553: break;
1554: if (mdocs[n->tok].mask)
1555: putmdockey(mpage, n->child,
1556: mdocs[n->tok].mask);
1557: break;
1558: default:
1.141 schwarze 1559: assert(n->type != ROFFT_ROOT);
1.47 schwarze 1560: continue;
1561: }
1562: if (NULL != n->child)
1.127 schwarze 1563: parse_mdoc(mpage, meta, n);
1.1 schwarze 1564: }
1565: }
1566:
1.19 schwarze 1567: static int
1.143 schwarze 1568: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1569: const struct roff_node *n)
1.1 schwarze 1570: {
1.131 schwarze 1571: char *start, *end;
1.1 schwarze 1572: size_t sz;
1.19 schwarze 1573:
1.47 schwarze 1574: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1575: NULL == (n = n->child) ||
1.141 schwarze 1576: n->type != ROFFT_TEXT)
1.150 schwarze 1577: return 0;
1.1 schwarze 1578:
1579: /*
1580: * Only consider those `Fd' macro fields that begin with an
1581: * "inclusion" token (versus, e.g., #define).
1582: */
1.47 schwarze 1583:
1.1 schwarze 1584: if (strcmp("#include", n->string))
1.150 schwarze 1585: return 0;
1.1 schwarze 1586:
1.141 schwarze 1587: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1588: return 0;
1.1 schwarze 1589:
1590: /*
1591: * Strip away the enclosing angle brackets and make sure we're
1592: * not zero-length.
1593: */
1594:
1595: start = n->string;
1596: if ('<' == *start || '"' == *start)
1597: start++;
1598:
1599: if (0 == (sz = strlen(start)))
1.150 schwarze 1600: return 0;
1.1 schwarze 1601:
1602: end = &start[(int)sz - 1];
1603: if ('>' == *end || '"' == *end)
1604: end--;
1605:
1.47 schwarze 1606: if (end > start)
1607: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1608: return 0;
1.1 schwarze 1609: }
1610:
1.133 schwarze 1611: static void
1.142 schwarze 1612: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1613: {
1.69 schwarze 1614: char *cp;
1.133 schwarze 1615: size_t sz;
1.1 schwarze 1616:
1.141 schwarze 1617: if (n->type != ROFFT_TEXT)
1.133 schwarze 1618: return;
1.19 schwarze 1619:
1.133 schwarze 1620: /* Skip function pointer punctuation. */
1.1 schwarze 1621:
1.133 schwarze 1622: cp = n->string;
1623: while (*cp == '(' || *cp == '*')
1.1 schwarze 1624: cp++;
1.133 schwarze 1625: sz = strcspn(cp, "()");
1.1 schwarze 1626:
1.133 schwarze 1627: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1628: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1629: putkeys(mpage, cp, sz, NAME_SYN);
1630: }
1.19 schwarze 1631:
1.133 schwarze 1632: static int
1.143 schwarze 1633: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1634: const struct roff_node *n)
1.133 schwarze 1635: {
1636:
1637: if (n->child == NULL)
1.150 schwarze 1638: return 0;
1.133 schwarze 1639:
1640: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1641:
1.133 schwarze 1642: for (n = n->child->next; n != NULL; n = n->next)
1.141 schwarze 1643: if (n->type == ROFFT_TEXT)
1.47 schwarze 1644: putkey(mpage, n->string, TYPE_Fa);
1.19 schwarze 1645:
1.150 schwarze 1646: return 0;
1.128 schwarze 1647: }
1648:
1649: static int
1.143 schwarze 1650: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1651: const struct roff_node *n)
1.128 schwarze 1652: {
1.132 schwarze 1653:
1.141 schwarze 1654: if (n->type != ROFFT_HEAD)
1.150 schwarze 1655: return 1;
1.128 schwarze 1656:
1.133 schwarze 1657: if (n->child != NULL)
1658: parse_mdoc_fname(mpage, n->child);
1659:
1.150 schwarze 1660: return 0;
1.1 schwarze 1661: }
1662:
1.19 schwarze 1663: static int
1.143 schwarze 1664: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1665: const struct roff_node *n)
1.1 schwarze 1666: {
1.47 schwarze 1667: char *cp;
1.1 schwarze 1668:
1669: if (NULL == (n = n->child))
1.150 schwarze 1670: return 0;
1.1 schwarze 1671:
1.47 schwarze 1672: if (NULL == n->next) {
1673: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1674: return 0;
1.47 schwarze 1675: }
1.1 schwarze 1676:
1.77 schwarze 1677: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1678: putkey(mpage, cp, TYPE_Xr);
1679: free(cp);
1.150 schwarze 1680: return 0;
1.1 schwarze 1681: }
1682:
1.19 schwarze 1683: static int
1.143 schwarze 1684: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1685: const struct roff_node *n)
1.1 schwarze 1686: {
1687:
1.141 schwarze 1688: if (n->type == ROFFT_BODY)
1.147 schwarze 1689: deroff(&mpage->desc, n);
1.150 schwarze 1690: return 0;
1.1 schwarze 1691: }
1692:
1.19 schwarze 1693: static int
1.143 schwarze 1694: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1695: const struct roff_node *n)
1.1 schwarze 1696: {
1697:
1.86 schwarze 1698: if (SEC_NAME == n->sec)
1.90 schwarze 1699: putmdockey(mpage, n->child, NAME_TITLE);
1.141 schwarze 1700: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1701: if (n->child == NULL)
1702: putkey(mpage, meta->name, NAME_SYN);
1703: else
1704: putmdockey(mpage, n->child, NAME_SYN);
1.129 schwarze 1705: }
1706: if ( ! (mpage->name_head_done ||
1707: n->child == NULL || n->child->string == NULL ||
1708: strcasecmp(n->child->string, meta->title))) {
1.141 schwarze 1709: putkey(mpage, n->child->string, ROFFT_HEAD);
1.129 schwarze 1710: mpage->name_head_done = 1;
1.127 schwarze 1711: }
1.150 schwarze 1712: return 0;
1.1 schwarze 1713: }
1714:
1.19 schwarze 1715: static int
1.143 schwarze 1716: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1717: const struct roff_node *n)
1.1 schwarze 1718: {
1719:
1.150 schwarze 1720: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1721: }
1722:
1.47 schwarze 1723: static int
1.143 schwarze 1724: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1725: const struct roff_node *n)
1.1 schwarze 1726: {
1727:
1.150 schwarze 1728: return n->type == ROFFT_HEAD;
1.47 schwarze 1729: }
1.1 schwarze 1730:
1.47 schwarze 1731: static int
1.143 schwarze 1732: parse_mdoc_body(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1733: const struct roff_node *n)
1.47 schwarze 1734: {
1.1 schwarze 1735:
1.150 schwarze 1736: return n->type == ROFFT_BODY;
1.1 schwarze 1737: }
1738:
1.47 schwarze 1739: /*
1740: * Add a string to the hash table for the current manual.
1741: * Each string has a bitmask telling which macros it belongs to.
1742: * When we finish the manual, we'll dump the table.
1743: */
1.1 schwarze 1744: static void
1.131 schwarze 1745: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1746: {
1.90 schwarze 1747: struct ohash *htab;
1.47 schwarze 1748: struct str *s;
1.68 schwarze 1749: const char *end;
1.47 schwarze 1750: unsigned int slot;
1.131 schwarze 1751: int i, mustfree;
1.1 schwarze 1752:
1.47 schwarze 1753: if (0 == sz)
1754: return;
1.68 schwarze 1755:
1.131 schwarze 1756: mustfree = render_string(&cp, &sz);
1757:
1.90 schwarze 1758: if (TYPE_Nm & v) {
1759: htab = &names;
1760: v &= name_mask;
1.124 schwarze 1761: if (v & NAME_FIRST)
1762: name_mask &= ~NAME_FIRST;
1.90 schwarze 1763: if (debug > 1)
1764: say(mpage->mlinks->file,
1.130 schwarze 1765: "Adding name %*s, bits=%d", sz, cp, v);
1.90 schwarze 1766: } else {
1767: htab = &strings;
1768: if (debug > 1)
1769: for (i = 0; i < mansearch_keymax; i++)
1.118 schwarze 1770: if ((uint64_t)1 << i & v)
1.90 schwarze 1771: say(mpage->mlinks->file,
1772: "Adding key %s=%*s",
1773: mansearch_keynames[i], sz, cp);
1.68 schwarze 1774: }
1.47 schwarze 1775:
1776: end = cp + sz;
1.90 schwarze 1777: slot = ohash_qlookupi(htab, cp, &end);
1778: s = ohash_find(htab, slot);
1.1 schwarze 1779:
1.47 schwarze 1780: if (NULL != s && mpage == s->mpage) {
1781: s->mask |= v;
1.1 schwarze 1782: return;
1.47 schwarze 1783: } else if (NULL == s) {
1.101 schwarze 1784: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1785: memcpy(s->key, cp, sz);
1.90 schwarze 1786: ohash_insert(htab, slot, s);
1.47 schwarze 1787: }
1788: s->mpage = mpage;
1789: s->mask = v;
1.131 schwarze 1790:
1791: if (mustfree)
1792: free(cp);
1.1 schwarze 1793: }
1794:
1795: /*
1.47 schwarze 1796: * Take a Unicode codepoint and produce its UTF-8 encoding.
1797: * This isn't the best way to do this, but it works.
1798: * The magic numbers are from the UTF-8 packaging.
1799: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1800: */
1.47 schwarze 1801: static size_t
1802: utf8(unsigned int cp, char out[7])
1.1 schwarze 1803: {
1.47 schwarze 1804: size_t rc;
1.1 schwarze 1805:
1.47 schwarze 1806: rc = 0;
1807: if (cp <= 0x0000007F) {
1808: rc = 1;
1809: out[0] = (char)cp;
1810: } else if (cp <= 0x000007FF) {
1811: rc = 2;
1812: out[0] = (cp >> 6 & 31) | 192;
1813: out[1] = (cp & 63) | 128;
1814: } else if (cp <= 0x0000FFFF) {
1815: rc = 3;
1816: out[0] = (cp >> 12 & 15) | 224;
1817: out[1] = (cp >> 6 & 63) | 128;
1818: out[2] = (cp & 63) | 128;
1819: } else if (cp <= 0x001FFFFF) {
1820: rc = 4;
1821: out[0] = (cp >> 18 & 7) | 240;
1822: out[1] = (cp >> 12 & 63) | 128;
1823: out[2] = (cp >> 6 & 63) | 128;
1824: out[3] = (cp & 63) | 128;
1825: } else if (cp <= 0x03FFFFFF) {
1826: rc = 5;
1827: out[0] = (cp >> 24 & 3) | 248;
1828: out[1] = (cp >> 18 & 63) | 128;
1829: out[2] = (cp >> 12 & 63) | 128;
1830: out[3] = (cp >> 6 & 63) | 128;
1831: out[4] = (cp & 63) | 128;
1832: } else if (cp <= 0x7FFFFFFF) {
1833: rc = 6;
1834: out[0] = (cp >> 30 & 1) | 252;
1835: out[1] = (cp >> 24 & 63) | 128;
1836: out[2] = (cp >> 18 & 63) | 128;
1837: out[3] = (cp >> 12 & 63) | 128;
1838: out[4] = (cp >> 6 & 63) | 128;
1839: out[5] = (cp & 63) | 128;
1840: } else
1.150 schwarze 1841: return 0;
1.19 schwarze 1842:
1.47 schwarze 1843: out[rc] = '\0';
1.150 schwarze 1844: return rc;
1.1 schwarze 1845: }
1846:
1.47 schwarze 1847: /*
1.131 schwarze 1848: * If the string contains escape sequences,
1849: * replace it with an allocated rendering and return 1,
1850: * such that the caller can free it after use.
1851: * Otherwise, do nothing and return 0.
1.47 schwarze 1852: */
1.131 schwarze 1853: static int
1854: render_string(char **public, size_t *psz)
1.1 schwarze 1855: {
1.131 schwarze 1856: const char *src, *scp, *addcp, *seq;
1857: char *dst;
1858: size_t ssz, dsz, addsz;
1.71 schwarze 1859: char utfbuf[7], res[6];
1.131 schwarze 1860: int seqlen, unicode;
1.47 schwarze 1861:
1862: res[0] = '\\';
1863: res[1] = '\t';
1864: res[2] = ASCII_NBRSP;
1865: res[3] = ASCII_HYPH;
1.71 schwarze 1866: res[4] = ASCII_BREAK;
1867: res[5] = '\0';
1.1 schwarze 1868:
1.131 schwarze 1869: src = scp = *public;
1870: ssz = *psz;
1871: dst = NULL;
1872: dsz = 0;
1873:
1874: while (scp < src + *psz) {
1875:
1876: /* Leave normal characters unchanged. */
1877:
1878: if (strchr(res, *scp) == NULL) {
1879: if (dst != NULL)
1880: dst[dsz++] = *scp;
1881: scp++;
1882: continue;
1883: }
1.1 schwarze 1884:
1.47 schwarze 1885: /*
1.131 schwarze 1886: * Found something that requires replacing,
1887: * make sure we have a destination buffer.
1.47 schwarze 1888: */
1.131 schwarze 1889:
1890: if (dst == NULL) {
1891: dst = mandoc_malloc(ssz + 1);
1892: dsz = scp - src;
1893: memcpy(dst, src, dsz);
1.47 schwarze 1894: }
1.39 schwarze 1895:
1.131 schwarze 1896: /* Handle single-char special characters. */
1897:
1898: switch (*scp) {
1899: case '\\':
1900: break;
1.98 schwarze 1901: case '\t':
1902: case ASCII_NBRSP:
1.131 schwarze 1903: dst[dsz++] = ' ';
1904: scp++;
1905: continue;
1906: case ASCII_HYPH:
1907: dst[dsz++] = '-';
1.71 schwarze 1908: /* FALLTHROUGH */
1.98 schwarze 1909: case ASCII_BREAK:
1.131 schwarze 1910: scp++;
1.47 schwarze 1911: continue;
1.71 schwarze 1912: default:
1.131 schwarze 1913: abort();
1.71 schwarze 1914: }
1.39 schwarze 1915:
1.47 schwarze 1916: /*
1.131 schwarze 1917: * Found an escape sequence.
1918: * Read past the slash, then parse it.
1919: * Ignore everything except characters.
1.47 schwarze 1920: */
1.52 schwarze 1921:
1.131 schwarze 1922: scp++;
1923: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1924: continue;
1.39 schwarze 1925:
1.47 schwarze 1926: /*
1.52 schwarze 1927: * Render the special character
1928: * as either UTF-8 or ASCII.
1.47 schwarze 1929: */
1.52 schwarze 1930:
1931: if (write_utf8) {
1.157 schwarze 1932: unicode = mchars_spec2cp(seq, seqlen);
1.131 schwarze 1933: if (unicode <= 0)
1.52 schwarze 1934: continue;
1.131 schwarze 1935: addsz = utf8(unicode, utfbuf);
1936: if (addsz == 0)
1.52 schwarze 1937: continue;
1.131 schwarze 1938: addcp = utfbuf;
1.52 schwarze 1939: } else {
1.157 schwarze 1940: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.131 schwarze 1941: if (addcp == NULL)
1.52 schwarze 1942: continue;
1.131 schwarze 1943: if (*addcp == ASCII_NBRSP) {
1944: addcp = " ";
1945: addsz = 1;
1.52 schwarze 1946: }
1947: }
1.1 schwarze 1948:
1.47 schwarze 1949: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1950:
1.131 schwarze 1951: ssz += addsz;
1952: dst = mandoc_realloc(dst, ssz + 1);
1953: memcpy(dst + dsz, addcp, addsz);
1954: dsz += addsz;
1955: }
1956: if (dst != NULL) {
1957: *public = dst;
1958: *psz = dsz;
1959: }
1960:
1961: /* Trim trailing whitespace and NUL-terminate. */
1962:
1963: while (*psz > 0 && (*public)[*psz - 1] == ' ')
1964: --*psz;
1965: if (dst != NULL) {
1966: (*public)[*psz] = '\0';
1.150 schwarze 1967: return 1;
1.131 schwarze 1968: } else
1.150 schwarze 1969: return 0;
1.1 schwarze 1970: }
1971:
1.75 schwarze 1972: static void
1973: dbadd_mlink(const struct mlink *mlink)
1974: {
1975: size_t i;
1976:
1977: i = 1;
1978: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1979: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1980: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.94 schwarze 1981: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1.75 schwarze 1982: SQL_STEP(stmts[STMT_INSERT_LINK]);
1983: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1.124 schwarze 1984: }
1985:
1986: static void
1987: dbadd_mlink_name(const struct mlink *mlink)
1988: {
1.130 schwarze 1989: uint64_t bits;
1.124 schwarze 1990: size_t i;
1991:
1992: dbadd_mlink(mlink);
1.115 schwarze 1993:
1994: i = 1;
1.130 schwarze 1995: SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
1996: bits = NAME_FILE & NAME_MASK;
1997: if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
1998: bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
1999: sqlite3_reset(stmts[STMT_SELECT_NAME]);
2000: }
2001:
2002: i = 1;
2003: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
1.115 schwarze 2004: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2005: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2006: SQL_STEP(stmts[STMT_INSERT_NAME]);
2007: sqlite3_reset(stmts[STMT_INSERT_NAME]);
1.75 schwarze 2008: }
2009:
1.11 schwarze 2010: /*
1.47 schwarze 2011: * Flush the current page's terms (and their bits) into the database.
2012: * Wrap the entire set of additions in a transaction to make sqlite be a
2013: * little faster.
1.53 schwarze 2014: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 2015: */
2016: static void
1.131 schwarze 2017: dbadd(struct mpage *mpage)
1.11 schwarze 2018: {
1.47 schwarze 2019: struct mlink *mlink;
2020: struct str *key;
1.131 schwarze 2021: char *cp;
1.47 schwarze 2022: size_t i;
2023: unsigned int slot;
1.131 schwarze 2024: int mustfree;
1.47 schwarze 2025:
1.85 schwarze 2026: mlink = mpage->mlinks;
1.11 schwarze 2027:
1.85 schwarze 2028: if (nodb) {
1.104 schwarze 2029: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2030: key = ohash_next(&names, &slot))
1.104 schwarze 2031: free(key);
2032: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2033: key = ohash_next(&strings, &slot))
1.104 schwarze 2034: free(key);
1.102 schwarze 2035: if (0 == debug)
2036: return;
1.85 schwarze 2037: while (NULL != mlink) {
2038: fputs(mlink->name, stdout);
2039: if (NULL == mlink->next ||
2040: strcmp(mlink->dsec, mlink->next->dsec) ||
2041: strcmp(mlink->fsec, mlink->next->fsec) ||
2042: strcmp(mlink->arch, mlink->next->arch)) {
2043: putchar('(');
2044: if ('\0' == *mlink->dsec)
2045: fputs(mlink->fsec, stdout);
2046: else
2047: fputs(mlink->dsec, stdout);
2048: if ('\0' != *mlink->arch)
2049: printf("/%s", mlink->arch);
2050: putchar(')');
2051: }
2052: mlink = mlink->next;
2053: if (NULL != mlink)
2054: fputs(", ", stdout);
2055: }
1.89 schwarze 2056: printf(" - %s\n", mpage->desc);
1.11 schwarze 2057: return;
1.85 schwarze 2058: }
2059:
2060: if (debug)
2061: say(mlink->file, "Adding to database");
1.47 schwarze 2062:
1.131 schwarze 2063: cp = mpage->desc;
2064: i = strlen(cp);
2065: mustfree = render_string(&cp, &i);
1.47 schwarze 2066: i = 1;
1.131 schwarze 2067: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
1.116 schwarze 2068: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
1.47 schwarze 2069: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.94 schwarze 2070: mpage->pageid = sqlite3_last_insert_rowid(db);
1.47 schwarze 2071: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1.131 schwarze 2072: if (mustfree)
2073: free(cp);
1.47 schwarze 2074:
1.85 schwarze 2075: while (NULL != mlink) {
1.75 schwarze 2076: dbadd_mlink(mlink);
1.85 schwarze 2077: mlink = mlink->next;
2078: }
1.91 schwarze 2079: mlink = mpage->mlinks;
1.47 schwarze 2080:
1.90 schwarze 2081: for (key = ohash_first(&names, &slot); NULL != key;
2082: key = ohash_next(&names, &slot)) {
2083: assert(key->mpage == mpage);
2084: i = 1;
2085: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
1.131 schwarze 2086: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
1.94 schwarze 2087: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
1.90 schwarze 2088: SQL_STEP(stmts[STMT_INSERT_NAME]);
2089: sqlite3_reset(stmts[STMT_INSERT_NAME]);
2090: free(key);
2091: }
1.47 schwarze 2092: for (key = ohash_first(&strings, &slot); NULL != key;
2093: key = ohash_next(&strings, &slot)) {
2094: assert(key->mpage == mpage);
2095: i = 1;
2096: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.131 schwarze 2097: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
1.94 schwarze 2098: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
1.47 schwarze 2099: SQL_STEP(stmts[STMT_INSERT_KEY]);
2100: sqlite3_reset(stmts[STMT_INSERT_KEY]);
2101: free(key);
1.33 schwarze 2102: }
1.47 schwarze 2103: }
1.41 deraadt 2104:
1.47 schwarze 2105: static void
2106: dbprune(void)
2107: {
2108: struct mpage *mpage;
2109: struct mlink *mlink;
2110: size_t i;
2111: unsigned int slot;
1.11 schwarze 2112:
1.63 schwarze 2113: if (0 == nodb)
2114: SQL_EXEC("BEGIN TRANSACTION");
1.47 schwarze 2115:
1.63 schwarze 2116: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2117: mpage = ohash_next(&mpages, &slot)) {
1.47 schwarze 2118: mlink = mpage->mlinks;
1.82 schwarze 2119: if (debug)
1.63 schwarze 2120: say(mlink->file, "Deleting from database");
2121: if (nodb)
2122: continue;
2123: for ( ; NULL != mlink; mlink = mlink->next) {
2124: i = 1;
2125: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2126: i, mlink->dsec);
2127: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2128: i, mlink->arch);
2129: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2130: i, mlink->name);
2131: SQL_STEP(stmts[STMT_DELETE_PAGE]);
2132: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2133: }
1.11 schwarze 2134: }
1.63 schwarze 2135:
2136: if (0 == nodb)
2137: SQL_EXEC("END TRANSACTION");
1.47 schwarze 2138: }
1.22 schwarze 2139:
1.47 schwarze 2140: /*
2141: * Close an existing database and its prepared statements.
2142: * If "real" is not set, rename the temporary file into the real one.
2143: */
2144: static void
2145: dbclose(int real)
2146: {
2147: size_t i;
1.72 schwarze 2148: int status;
2149: pid_t child;
1.11 schwarze 2150:
1.47 schwarze 2151: if (nodb)
2152: return;
1.11 schwarze 2153:
1.47 schwarze 2154: for (i = 0; i < STMT__MAX; i++) {
2155: sqlite3_finalize(stmts[i]);
2156: stmts[i] = NULL;
1.28 schwarze 2157: }
1.22 schwarze 2158:
1.47 schwarze 2159: sqlite3_close(db);
2160: db = NULL;
1.11 schwarze 2161:
1.47 schwarze 2162: if (real)
2163: return;
1.22 schwarze 2164:
1.72 schwarze 2165: if ('\0' == *tempfilename) {
2166: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2167: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2168: say(MANDOC_DB, "&rename");
1.72 schwarze 2169: }
2170: return;
2171: }
2172:
2173: switch (child = fork()) {
1.98 schwarze 2174: case -1:
1.72 schwarze 2175: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2176: say("", "&fork cmp");
1.72 schwarze 2177: return;
1.98 schwarze 2178: case 0:
1.72 schwarze 2179: execlp("cmp", "cmp", "-s",
1.149 schwarze 2180: tempfilename, MANDOC_DB, (char *)NULL);
1.80 schwarze 2181: say("", "&exec cmp");
1.72 schwarze 2182: exit(0);
2183: default:
2184: break;
2185: }
2186: if (-1 == waitpid(child, &status, 0)) {
2187: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2188: say("", "&wait cmp");
1.72 schwarze 2189: } else if (WIFSIGNALED(status)) {
2190: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2191: say("", "cmp died from signal %d", WTERMSIG(status));
1.72 schwarze 2192: } else if (WEXITSTATUS(status)) {
1.47 schwarze 2193: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2194: say(MANDOC_DB,
2195: "Data changed, but cannot replace database");
2196: }
2197:
2198: *strrchr(tempfilename, '/') = '\0';
2199: switch (child = fork()) {
1.98 schwarze 2200: case -1:
1.72 schwarze 2201: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2202: say("", "&fork rm");
1.72 schwarze 2203: return;
1.98 schwarze 2204: case 0:
1.149 schwarze 2205: execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
1.80 schwarze 2206: say("", "&exec rm");
1.72 schwarze 2207: exit((int)MANDOCLEVEL_SYSERR);
2208: default:
2209: break;
2210: }
2211: if (-1 == waitpid(child, &status, 0)) {
2212: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2213: say("", "&wait rm");
1.72 schwarze 2214: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2215: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2216: say("", "%s: Cannot remove temporary directory",
2217: tempfilename);
1.22 schwarze 2218: }
1.11 schwarze 2219: }
2220:
1.47 schwarze 2221: /*
2222: * This is straightforward stuff.
2223: * Open a database connection to a "temporary" database, then open a set
2224: * of prepared statements we'll use over and over again.
2225: * If "real" is set, we use the existing database; if not, we truncate a
2226: * temporary one.
2227: * Must be matched by dbclose().
2228: */
2229: static int
2230: dbopen(int real)
1.2 schwarze 2231: {
1.72 schwarze 2232: const char *sql;
1.47 schwarze 2233: int rc, ofl;
1.6 schwarze 2234:
1.98 schwarze 2235: if (nodb)
1.150 schwarze 2236: return 1;
1.6 schwarze 2237:
1.72 schwarze 2238: *tempfilename = '\0';
1.47 schwarze 2239: ofl = SQLITE_OPEN_READWRITE;
1.72 schwarze 2240:
2241: if (real) {
2242: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2243: if (SQLITE_OK != rc) {
1.47 schwarze 2244: exitcode = (int)MANDOCLEVEL_SYSERR;
1.108 sthen 2245: if (SQLITE_CANTOPEN != rc)
2246: say(MANDOC_DB, "%s", sqlite3_errstr(rc));
1.150 schwarze 2247: return 0;
1.28 schwarze 2248: }
1.72 schwarze 2249: goto prepare_statements;
2250: }
2251:
2252: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.6 schwarze 2253:
1.72 schwarze 2254: remove(MANDOC_DB "~");
2255: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.98 schwarze 2256: if (SQLITE_OK == rc)
1.72 schwarze 2257: goto create_tables;
1.73 schwarze 2258: if (MPARSE_QUICK & mparse_options) {
1.47 schwarze 2259: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2260: say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
1.150 schwarze 2261: return 0;
1.47 schwarze 2262: }
1.6 schwarze 2263:
1.100 schwarze 2264: (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2265: sizeof(tempfilename));
1.72 schwarze 2266: if (NULL == mkdtemp(tempfilename)) {
2267: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2268: say("", "&%s", tempfilename);
1.150 schwarze 2269: return 0;
1.72 schwarze 2270: }
1.100 schwarze 2271: (void)strlcat(tempfilename, "/" MANDOC_DB,
2272: sizeof(tempfilename));
1.72 schwarze 2273: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2274: if (SQLITE_OK != rc) {
1.47 schwarze 2275: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2276: say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
1.150 schwarze 2277: return 0;
1.2 schwarze 2278: }
2279:
1.72 schwarze 2280: create_tables:
1.47 schwarze 2281: sql = "CREATE TABLE \"mpages\" (\n"
1.89 schwarze 2282: " \"desc\" TEXT NOT NULL,\n"
1.47 schwarze 2283: " \"form\" INTEGER NOT NULL,\n"
1.94 schwarze 2284: " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1.47 schwarze 2285: ");\n"
2286: "\n"
2287: "CREATE TABLE \"mlinks\" (\n"
2288: " \"sec\" TEXT NOT NULL,\n"
2289: " \"arch\" TEXT NOT NULL,\n"
2290: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2291: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2292: "ON DELETE CASCADE\n"
1.47 schwarze 2293: ");\n"
1.93 schwarze 2294: "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
1.47 schwarze 2295: "\n"
1.90 schwarze 2296: "CREATE TABLE \"names\" (\n"
2297: " \"bits\" INTEGER NOT NULL,\n"
2298: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2299: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.130 schwarze 2300: "ON DELETE CASCADE,\n"
2301: " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
1.90 schwarze 2302: ");\n"
2303: "\n"
1.47 schwarze 2304: "CREATE TABLE \"keys\" (\n"
2305: " \"bits\" INTEGER NOT NULL,\n"
2306: " \"key\" TEXT NOT NULL,\n"
1.94 schwarze 2307: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2308: "ON DELETE CASCADE\n"
1.93 schwarze 2309: ");\n"
2310: "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
1.47 schwarze 2311:
2312: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2313: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2314: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.103 schwarze 2315: sqlite3_close(db);
1.150 schwarze 2316: return 0;
1.2 schwarze 2317: }
2318:
1.47 schwarze 2319: prepare_statements:
1.103 schwarze 2320: if (SQLITE_OK != sqlite3_exec(db,
2321: "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2322: exitcode = (int)MANDOCLEVEL_SYSERR;
2323: say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2324: sqlite3_errmsg(db));
2325: sqlite3_close(db);
1.150 schwarze 2326: return 0;
1.103 schwarze 2327: }
2328:
1.94 schwarze 2329: sql = "DELETE FROM mpages WHERE pageid IN "
1.63 schwarze 2330: "(SELECT pageid FROM mlinks WHERE "
2331: "sec=? AND arch=? AND name=?)";
1.47 schwarze 2332: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2333: sql = "INSERT INTO mpages "
1.89 schwarze 2334: "(desc,form) VALUES (?,?)";
1.47 schwarze 2335: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2336: sql = "INSERT INTO mlinks "
1.61 schwarze 2337: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.47 schwarze 2338: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.130 schwarze 2339: sql = "SELECT bits FROM names where pageid = ?";
2340: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
1.90 schwarze 2341: sql = "INSERT INTO names "
2342: "(bits,name,pageid) VALUES (?,?,?)";
2343: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
1.47 schwarze 2344: sql = "INSERT INTO keys "
2345: "(bits,key,pageid) VALUES (?,?,?)";
2346: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.6 schwarze 2347:
1.47 schwarze 2348: /*
2349: * When opening a new database, we can turn off
2350: * synchronous mode for much better performance.
2351: */
1.6 schwarze 2352:
1.103 schwarze 2353: if (real && SQLITE_OK != sqlite3_exec(db,
2354: "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2355: exitcode = (int)MANDOCLEVEL_SYSERR;
2356: say(MANDOC_DB, "PRAGMA synchronous: %s",
2357: sqlite3_errmsg(db));
2358: sqlite3_close(db);
1.150 schwarze 2359: return 0;
1.103 schwarze 2360: }
1.11 schwarze 2361:
1.150 schwarze 2362: return 1;
1.47 schwarze 2363: }
1.6 schwarze 2364:
1.47 schwarze 2365: static int
1.120 schwarze 2366: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2367: {
2368: static char startdir[PATH_MAX];
1.110 schwarze 2369: static int getcwd_status; /* 1 = ok, 2 = failure */
2370: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2371: char *cp;
1.6 schwarze 2372:
1.47 schwarze 2373: /*
1.110 schwarze 2374: * Remember the original working directory, if possible.
2375: * This will be needed if the second or a later directory
2376: * on the command line is given as a relative path.
2377: * Do not error out if the current directory is not
2378: * searchable: Maybe it won't be needed after all.
2379: */
2380: if (0 == getcwd_status) {
2381: if (NULL == getcwd(startdir, sizeof(startdir))) {
2382: getcwd_status = 2;
2383: (void)strlcpy(startdir, strerror(errno),
2384: sizeof(startdir));
2385: } else
2386: getcwd_status = 1;
2387: }
2388:
2389: /*
2390: * We are leaving the old base directory.
2391: * Do not use it any longer, not even for messages.
2392: */
2393: *basedir = '\0';
2394:
2395: /*
2396: * If and only if the directory was changed earlier and
2397: * the next directory to process is given as a relative path,
2398: * first go back, or bail out if that is impossible.
1.47 schwarze 2399: */
1.110 schwarze 2400: if (chdir_status && '/' != *targetdir) {
2401: if (2 == getcwd_status) {
1.47 schwarze 2402: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2403: say("", "getcwd: %s", startdir);
1.150 schwarze 2404: return 0;
1.47 schwarze 2405: }
1.110 schwarze 2406: if (-1 == chdir(startdir)) {
1.47 schwarze 2407: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2408: say("", "&chdir %s", startdir);
1.150 schwarze 2409: return 0;
1.2 schwarze 2410: }
2411: }
1.110 schwarze 2412:
2413: /*
2414: * Always resolve basedir to the canonicalized absolute
2415: * pathname and append a trailing slash, such that
2416: * we can reliably check whether files are inside.
2417: */
1.47 schwarze 2418: if (NULL == realpath(targetdir, basedir)) {
1.120 schwarze 2419: if (report_baddir || errno != ENOENT) {
2420: exitcode = (int)MANDOCLEVEL_BADARG;
2421: say("", "&%s: realpath", targetdir);
2422: }
1.150 schwarze 2423: return 0;
1.47 schwarze 2424: } else if (-1 == chdir(basedir)) {
1.120 schwarze 2425: if (report_baddir || errno != ENOENT) {
2426: exitcode = (int)MANDOCLEVEL_BADARG;
2427: say("", "&chdir");
2428: }
1.150 schwarze 2429: return 0;
1.109 schwarze 2430: }
1.110 schwarze 2431: chdir_status = 1;
1.109 schwarze 2432: cp = strchr(basedir, '\0');
2433: if ('/' != cp[-1]) {
2434: if (cp - basedir >= PATH_MAX - 1) {
2435: exitcode = (int)MANDOCLEVEL_SYSERR;
2436: say("", "Filename too long");
1.150 schwarze 2437: return 0;
1.109 schwarze 2438: }
2439: *cp++ = '/';
2440: *cp = '\0';
1.47 schwarze 2441: }
1.150 schwarze 2442: return 1;
1.2 schwarze 2443: }
2444:
2445: static void
1.47 schwarze 2446: say(const char *file, const char *format, ...)
1.2 schwarze 2447: {
1.47 schwarze 2448: va_list ap;
1.80 schwarze 2449: int use_errno;
1.2 schwarze 2450:
1.47 schwarze 2451: if ('\0' != *basedir)
2452: fprintf(stderr, "%s", basedir);
2453: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2454: fputc('/', stderr);
1.47 schwarze 2455: if ('\0' != *file)
2456: fprintf(stderr, "%s", file);
1.31 schwarze 2457:
1.80 schwarze 2458: use_errno = 1;
2459: if (NULL != format) {
2460: switch (*format) {
1.98 schwarze 2461: case '&':
1.80 schwarze 2462: format++;
2463: break;
1.98 schwarze 2464: case '\0':
1.80 schwarze 2465: format = NULL;
2466: break;
2467: default:
2468: use_errno = 0;
2469: break;
2470: }
2471: }
2472: if (NULL != format) {
2473: if ('\0' != *basedir || '\0' != *file)
2474: fputs(": ", stderr);
2475: va_start(ap, format);
2476: vfprintf(stderr, format, ap);
2477: va_end(ap);
2478: }
2479: if (use_errno) {
2480: if ('\0' != *basedir || '\0' != *file || NULL != format)
2481: fputs(": ", stderr);
1.47 schwarze 2482: perror(NULL);
1.80 schwarze 2483: } else
2484: fputc('\n', stderr);
1.1 schwarze 2485: }