Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.161
1.161 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.160 2015/10/22 22:05:42 schwarze 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:
1.72 schwarze 180: static char tempfilename[32];
1.59 schwarze 181: static int nodb; /* no database changes */
1.73 schwarze 182: static int mparse_options; /* abort the parse early */
1.98 schwarze 183: static int use_all; /* use all found files */
184: static int debug; /* print what we're doing */
185: static int warnings; /* warn about crap */
1.52 schwarze 186: static int write_utf8; /* write UTF-8 output; else ASCII */
1.47 schwarze 187: static int exitcode; /* to be returned by main */
1.98 schwarze 188: static enum op op; /* operational mode */
1.47 schwarze 189: static char basedir[PATH_MAX]; /* current base directory */
190: static struct ohash mpages; /* table of distinct manual pages */
191: static struct ohash mlinks; /* table of directory entries */
1.90 schwarze 192: static struct ohash names; /* table of all names */
1.47 schwarze 193: static struct ohash strings; /* table of all strings */
194: static sqlite3 *db = NULL; /* current database */
195: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.90 schwarze 196: static uint64_t name_mask;
1.47 schwarze 197:
1.19 schwarze 198: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.47 schwarze 199: { NULL, 0 }, /* Ap */
200: { NULL, 0 }, /* Dd */
201: { NULL, 0 }, /* Dt */
202: { NULL, 0 }, /* Os */
203: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
204: { parse_mdoc_head, TYPE_Ss }, /* Ss */
205: { NULL, 0 }, /* Pp */
206: { NULL, 0 }, /* D1 */
207: { NULL, 0 }, /* Dl */
208: { NULL, 0 }, /* Bd */
209: { NULL, 0 }, /* Ed */
210: { NULL, 0 }, /* Bl */
211: { NULL, 0 }, /* El */
212: { NULL, 0 }, /* It */
213: { NULL, 0 }, /* Ad */
214: { NULL, TYPE_An }, /* An */
215: { NULL, TYPE_Ar }, /* Ar */
216: { NULL, TYPE_Cd }, /* Cd */
217: { NULL, TYPE_Cm }, /* Cm */
218: { NULL, TYPE_Dv }, /* Dv */
219: { NULL, TYPE_Er }, /* Er */
220: { NULL, TYPE_Ev }, /* Ev */
221: { NULL, 0 }, /* Ex */
222: { NULL, TYPE_Fa }, /* Fa */
223: { parse_mdoc_Fd, 0 }, /* Fd */
224: { NULL, TYPE_Fl }, /* Fl */
225: { parse_mdoc_Fn, 0 }, /* Fn */
226: { NULL, TYPE_Ft }, /* Ft */
227: { NULL, TYPE_Ic }, /* Ic */
1.49 schwarze 228: { NULL, TYPE_In }, /* In */
1.47 schwarze 229: { NULL, TYPE_Li }, /* Li */
1.89 schwarze 230: { parse_mdoc_Nd, 0 }, /* Nd */
1.86 schwarze 231: { parse_mdoc_Nm, 0 }, /* Nm */
1.47 schwarze 232: { NULL, 0 }, /* Op */
233: { NULL, 0 }, /* Ot */
234: { NULL, TYPE_Pa }, /* Pa */
235: { NULL, 0 }, /* Rv */
1.49 schwarze 236: { NULL, TYPE_St }, /* St */
1.47 schwarze 237: { NULL, TYPE_Va }, /* Va */
238: { parse_mdoc_body, TYPE_Va }, /* Vt */
239: { parse_mdoc_Xr, 0 }, /* Xr */
240: { NULL, 0 }, /* %A */
241: { NULL, 0 }, /* %B */
242: { NULL, 0 }, /* %D */
243: { NULL, 0 }, /* %I */
244: { NULL, 0 }, /* %J */
245: { NULL, 0 }, /* %N */
246: { NULL, 0 }, /* %O */
247: { NULL, 0 }, /* %P */
248: { NULL, 0 }, /* %R */
249: { NULL, 0 }, /* %T */
250: { NULL, 0 }, /* %V */
251: { NULL, 0 }, /* Ac */
252: { NULL, 0 }, /* Ao */
253: { NULL, 0 }, /* Aq */
254: { NULL, TYPE_At }, /* At */
255: { NULL, 0 }, /* Bc */
256: { NULL, 0 }, /* Bf */
257: { NULL, 0 }, /* Bo */
258: { NULL, 0 }, /* Bq */
259: { NULL, TYPE_Bsx }, /* Bsx */
260: { NULL, TYPE_Bx }, /* Bx */
261: { NULL, 0 }, /* Db */
262: { NULL, 0 }, /* Dc */
263: { NULL, 0 }, /* Do */
264: { NULL, 0 }, /* Dq */
265: { NULL, 0 }, /* Ec */
266: { NULL, 0 }, /* Ef */
267: { NULL, TYPE_Em }, /* Em */
268: { NULL, 0 }, /* Eo */
269: { NULL, TYPE_Fx }, /* Fx */
270: { NULL, TYPE_Ms }, /* Ms */
271: { NULL, 0 }, /* No */
272: { NULL, 0 }, /* Ns */
273: { NULL, TYPE_Nx }, /* Nx */
274: { NULL, TYPE_Ox }, /* Ox */
275: { NULL, 0 }, /* Pc */
276: { NULL, 0 }, /* Pf */
277: { NULL, 0 }, /* Po */
278: { NULL, 0 }, /* Pq */
279: { NULL, 0 }, /* Qc */
280: { NULL, 0 }, /* Ql */
281: { NULL, 0 }, /* Qo */
282: { NULL, 0 }, /* Qq */
283: { NULL, 0 }, /* Re */
284: { NULL, 0 }, /* Rs */
285: { NULL, 0 }, /* Sc */
286: { NULL, 0 }, /* So */
287: { NULL, 0 }, /* Sq */
288: { NULL, 0 }, /* Sm */
289: { NULL, 0 }, /* Sx */
290: { NULL, TYPE_Sy }, /* Sy */
291: { NULL, TYPE_Tn }, /* Tn */
292: { NULL, 0 }, /* Ux */
293: { NULL, 0 }, /* Xc */
294: { NULL, 0 }, /* Xo */
1.128 schwarze 295: { parse_mdoc_Fo, 0 }, /* Fo */
1.47 schwarze 296: { NULL, 0 }, /* Fc */
297: { NULL, 0 }, /* Oo */
298: { NULL, 0 }, /* Oc */
299: { NULL, 0 }, /* Bk */
300: { NULL, 0 }, /* Ek */
301: { NULL, 0 }, /* Bt */
302: { NULL, 0 }, /* Hf */
303: { NULL, 0 }, /* Fr */
304: { NULL, 0 }, /* Ud */
305: { NULL, TYPE_Lb }, /* Lb */
306: { NULL, 0 }, /* Lp */
307: { NULL, TYPE_Lk }, /* Lk */
308: { NULL, TYPE_Mt }, /* Mt */
309: { NULL, 0 }, /* Brq */
310: { NULL, 0 }, /* Bro */
311: { NULL, 0 }, /* Brc */
312: { NULL, 0 }, /* %C */
313: { NULL, 0 }, /* Es */
314: { NULL, 0 }, /* En */
315: { NULL, TYPE_Dx }, /* Dx */
316: { NULL, 0 }, /* %Q */
317: { NULL, 0 }, /* br */
318: { NULL, 0 }, /* sp */
319: { NULL, 0 }, /* %U */
320: { NULL, 0 }, /* Ta */
1.121 schwarze 321: { NULL, 0 }, /* ll */
1.1 schwarze 322: };
323:
1.98 schwarze 324:
1.1 schwarze 325: int
1.3 schwarze 326: mandocdb(int argc, char *argv[])
1.1 schwarze 327: {
1.140 schwarze 328: struct manconf conf;
329: struct mparse *mp;
1.161 ! schwarze 330: const char *path_arg, *progname;
1.140 schwarze 331: size_t j, sz;
1.47 schwarze 332: int ch, i;
333:
1.158 millert 334: if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) {
1.155 schwarze 335: perror("pledge");
336: return (int)MANDOCLEVEL_SYSERR;
337: }
338:
1.140 schwarze 339: memset(&conf, 0, sizeof(conf));
1.47 schwarze 340: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
341:
342: /*
1.98 schwarze 343: * We accept a few different invocations.
1.47 schwarze 344: * The CHECKOP macro makes sure that invocation styles don't
345: * clobber each other.
346: */
347: #define CHECKOP(_op, _ch) do \
348: if (OP_DEFAULT != (_op)) { \
1.151 schwarze 349: warnx("-%c: Conflicting option", (_ch)); \
1.47 schwarze 350: goto usage; \
351: } while (/*CONSTCOND*/0)
1.10 schwarze 352:
1.47 schwarze 353: path_arg = NULL;
1.28 schwarze 354: op = OP_DEFAULT;
1.1 schwarze 355:
1.83 schwarze 356: while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
1.1 schwarze 357: switch (ch) {
1.98 schwarze 358: case 'a':
1.6 schwarze 359: use_all = 1;
360: break;
1.98 schwarze 361: case 'C':
1.47 schwarze 362: CHECKOP(op, ch);
363: path_arg = optarg;
1.28 schwarze 364: op = OP_CONFFILE;
1.25 schwarze 365: break;
1.98 schwarze 366: case 'D':
1.82 schwarze 367: debug++;
368: break;
1.98 schwarze 369: case 'd':
1.47 schwarze 370: CHECKOP(op, ch);
371: path_arg = optarg;
1.2 schwarze 372: op = OP_UPDATE;
1.1 schwarze 373: break;
1.98 schwarze 374: case 'n':
1.47 schwarze 375: nodb = 1;
376: break;
1.98 schwarze 377: case 'p':
1.83 schwarze 378: warnings = 1;
379: break;
1.98 schwarze 380: case 'Q':
1.73 schwarze 381: mparse_options |= MPARSE_QUICK;
1.59 schwarze 382: break;
1.98 schwarze 383: case 'T':
1.52 schwarze 384: if (strcmp(optarg, "utf8")) {
1.151 schwarze 385: warnx("-T%s: Unsupported output format",
386: optarg);
1.52 schwarze 387: goto usage;
388: }
389: write_utf8 = 1;
390: break;
1.98 schwarze 391: case 't':
1.47 schwarze 392: CHECKOP(op, ch);
1.28 schwarze 393: dup2(STDOUT_FILENO, STDERR_FILENO);
394: op = OP_TEST;
1.47 schwarze 395: nodb = warnings = 1;
1.28 schwarze 396: break;
1.98 schwarze 397: case 'u':
1.47 schwarze 398: CHECKOP(op, ch);
399: path_arg = optarg;
1.1 schwarze 400: op = OP_DELETE;
401: break;
1.98 schwarze 402: case 'v':
1.83 schwarze 403: /* Compatibility with espie@'s makewhatis. */
1.28 schwarze 404: break;
1.1 schwarze 405: default:
1.28 schwarze 406: goto usage;
1.1 schwarze 407: }
408:
409: argc -= optind;
410: argv += optind;
411:
1.155 schwarze 412: if (nodb && pledge("stdio rpath", NULL) == -1) {
413: perror("pledge");
414: return (int)MANDOCLEVEL_SYSERR;
415: }
416:
1.28 schwarze 417: if (OP_CONFFILE == op && argc > 0) {
1.151 schwarze 418: warnx("-C: Too many arguments");
1.28 schwarze 419: goto usage;
420: }
421:
1.47 schwarze 422: exitcode = (int)MANDOCLEVEL_OK;
1.157 schwarze 423: mchars_alloc();
424: mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
1.156 schwarze 425: mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
426: mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
1.2 schwarze 427:
1.47 schwarze 428: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.2 schwarze 429:
1.47 schwarze 430: /*
1.109 schwarze 431: * Most of these deal with a specific directory.
1.97 schwarze 432: * Jump into that directory first.
1.47 schwarze 433: */
1.120 schwarze 434: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.47 schwarze 435: goto out;
1.97 schwarze 436:
1.95 schwarze 437: if (dbopen(1)) {
1.97 schwarze 438: /*
439: * The existing database is usable. Process
440: * all files specified on the command-line.
441: */
1.158 millert 442: if (!nodb && pledge("stdio rpath wpath cpath fattr flock",
1.155 schwarze 443: NULL) == -1) {
444: perror("pledge");
445: exitcode = (int)MANDOCLEVEL_SYSERR;
446: goto out;
447: }
1.97 schwarze 448: use_all = 1;
1.95 schwarze 449: for (i = 0; i < argc; i++)
450: filescan(argv[i]);
451: if (OP_TEST != op)
452: dbprune();
453: } else {
454: /*
455: * Database missing or corrupt.
456: * Recreate from scratch.
457: */
1.97 schwarze 458: exitcode = (int)MANDOCLEVEL_OK;
1.95 schwarze 459: op = OP_DEFAULT;
460: if (0 == treescan())
461: goto out;
462: if (0 == dbopen(0))
463: goto out;
464: }
1.47 schwarze 465: if (OP_DELETE != op)
1.131 schwarze 466: mpages_merge(mp);
1.95 schwarze 467: dbclose(OP_DEFAULT == op ? 0 : 1);
1.47 schwarze 468: } else {
469: /*
470: * If we have arguments, use them as our manpaths.
471: * If we don't, grok from manpath(1) or however else
1.140 schwarze 472: * manconf_parse() wants to do it.
1.47 schwarze 473: */
474: if (argc > 0) {
1.140 schwarze 475: conf.manpath.paths = mandoc_reallocarray(NULL,
1.101 schwarze 476: argc, sizeof(char *));
1.140 schwarze 477: conf.manpath.sz = (size_t)argc;
1.47 schwarze 478: for (i = 0; i < argc; i++)
1.140 schwarze 479: conf.manpath.paths[i] = mandoc_strdup(argv[i]);
1.47 schwarze 480: } else
1.140 schwarze 481: manconf_parse(&conf, path_arg, NULL, NULL);
1.84 schwarze 482:
1.140 schwarze 483: if (conf.manpath.sz == 0) {
1.84 schwarze 484: exitcode = (int)MANDOCLEVEL_BADARG;
485: say("", "Empty manpath");
486: }
1.2 schwarze 487:
1.47 schwarze 488: /*
489: * First scan the tree rooted at a base directory, then
490: * build a new database and finally move it into place.
491: * Ignore zero-length directories and strip trailing
492: * slashes.
493: */
1.140 schwarze 494: for (j = 0; j < conf.manpath.sz; j++) {
495: sz = strlen(conf.manpath.paths[j]);
496: if (sz && conf.manpath.paths[j][sz - 1] == '/')
497: conf.manpath.paths[j][--sz] = '\0';
1.47 schwarze 498: if (0 == sz)
499: continue;
1.2 schwarze 500:
1.47 schwarze 501: if (j) {
1.156 schwarze 502: mandoc_ohash_init(&mpages, 6,
503: offsetof(struct mpage, inodev));
504: mandoc_ohash_init(&mlinks, 6,
505: offsetof(struct mlink, file));
1.47 schwarze 506: }
1.1 schwarze 507:
1.140 schwarze 508: if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
1.120 schwarze 509: continue;
1.47 schwarze 510: if (0 == treescan())
1.120 schwarze 511: continue;
1.47 schwarze 512: if (0 == dbopen(0))
1.120 schwarze 513: continue;
1.41 deraadt 514:
1.131 schwarze 515: mpages_merge(mp);
1.104 schwarze 516: if (warnings && !nodb &&
1.87 schwarze 517: ! (MPARSE_QUICK & mparse_options))
518: names_check();
1.47 schwarze 519: dbclose(0);
1.28 schwarze 520:
1.140 schwarze 521: if (j + 1 < conf.manpath.sz) {
1.47 schwarze 522: mpages_free();
523: ohash_delete(&mpages);
524: ohash_delete(&mlinks);
525: }
1.2 schwarze 526: }
1.47 schwarze 527: }
528: out:
1.140 schwarze 529: manconf_free(&conf);
1.123 schwarze 530: mparse_free(mp);
1.157 schwarze 531: mchars_free();
1.47 schwarze 532: mpages_free();
533: ohash_delete(&mpages);
534: ohash_delete(&mlinks);
1.150 schwarze 535: return exitcode;
1.47 schwarze 536: usage:
1.161 ! schwarze 537: progname = getprogname();
1.83 schwarze 538: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
539: " %s [-aDnpQ] [-Tutf8] dir ...\n"
540: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
541: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 542: " %s [-Q] -t file ...\n",
1.161 ! schwarze 543: progname, progname, progname, progname, progname);
1.1 schwarze 544:
1.150 schwarze 545: return (int)MANDOCLEVEL_BADARG;
1.47 schwarze 546: }
1.1 schwarze 547:
1.47 schwarze 548: /*
549: * Scan a directory tree rooted at "basedir" for manpages.
550: * We use fts(), scanning directory parts along the way for clues to our
551: * section and architecture.
552: *
553: * If use_all has been specified, grok all files.
554: * If not, sanitise paths to the following:
555: *
1.98 schwarze 556: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 557: * or
558: * [./]cat<section>[/<arch>]/<name>.0
559: *
560: * TODO: accomodate for multi-language directories.
561: */
562: static int
563: treescan(void)
564: {
1.96 schwarze 565: char buf[PATH_MAX];
1.47 schwarze 566: FTS *f;
567: FTSENT *ff;
568: struct mlink *mlink;
1.81 schwarze 569: int dform, gzip;
1.51 schwarze 570: char *dsec, *arch, *fsec, *cp;
571: const char *path;
1.47 schwarze 572: const char *argv[2];
1.1 schwarze 573:
1.47 schwarze 574: argv[0] = ".";
575: argv[1] = (char *)NULL;
1.1 schwarze 576:
1.96 schwarze 577: f = fts_open((char * const *)argv,
578: FTS_PHYSICAL | FTS_NOCHDIR, NULL);
1.153 schwarze 579: if (f == NULL) {
1.47 schwarze 580: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 581: say("", "&fts_open");
1.150 schwarze 582: return 0;
1.47 schwarze 583: }
1.2 schwarze 584:
1.47 schwarze 585: dsec = arch = NULL;
586: dform = FORM_NONE;
1.2 schwarze 587:
1.153 schwarze 588: while ((ff = fts_read(f)) != NULL) {
1.47 schwarze 589: path = ff->fts_path + 2;
1.96 schwarze 590: switch (ff->fts_info) {
591:
592: /*
593: * Symbolic links require various sanity checks,
594: * then get handled just like regular files.
595: */
1.98 schwarze 596: case FTS_SL:
1.153 schwarze 597: if (realpath(path, buf) == NULL) {
1.96 schwarze 598: if (warnings)
599: say(path, "&realpath");
600: continue;
601: }
602: if (strstr(buf, basedir) != buf) {
603: if (warnings) say("",
604: "%s: outside base directory", buf);
605: continue;
606: }
607: /* Use logical inode to avoid mpages dupe. */
1.153 schwarze 608: if (stat(path, ff->fts_statp) == -1) {
1.96 schwarze 609: if (warnings)
610: say(path, "&stat");
611: continue;
612: }
613: /* FALLTHROUGH */
614:
1.15 schwarze 615: /*
1.47 schwarze 616: * If we're a regular file, add an mlink by using the
617: * stored directory data and handling the filename.
1.15 schwarze 618: */
1.98 schwarze 619: case FTS_F:
1.153 schwarze 620: if ( ! strcmp(path, MANDOC_DB))
1.47 schwarze 621: continue;
622: if ( ! use_all && ff->fts_level < 2) {
623: if (warnings)
624: say(path, "Extraneous file");
625: continue;
1.81 schwarze 626: }
627: gzip = 0;
628: fsec = NULL;
1.153 schwarze 629: while (fsec == NULL) {
1.81 schwarze 630: fsec = strrchr(ff->fts_name, '.');
1.153 schwarze 631: if (fsec == NULL || strcmp(fsec+1, "gz"))
1.81 schwarze 632: break;
633: gzip = 1;
634: *fsec = '\0';
635: fsec = NULL;
636: }
1.153 schwarze 637: if (fsec == NULL) {
1.47 schwarze 638: if ( ! use_all) {
639: if (warnings)
640: say(path,
641: "No filename suffix");
642: continue;
643: }
1.153 schwarze 644: } else if ( ! strcmp(++fsec, "html")) {
1.47 schwarze 645: if (warnings)
646: say(path, "Skip html");
647: continue;
1.153 schwarze 648: } else if ( ! strcmp(fsec, "ps")) {
1.47 schwarze 649: if (warnings)
650: say(path, "Skip ps");
651: continue;
1.153 schwarze 652: } else if ( ! strcmp(fsec, "pdf")) {
1.47 schwarze 653: if (warnings)
654: say(path, "Skip pdf");
655: continue;
656: } else if ( ! use_all &&
1.153 schwarze 657: ((dform == FORM_SRC &&
1.138 schwarze 658: strncmp(fsec, dsec, strlen(dsec))) ||
1.153 schwarze 659: (dform == FORM_CAT && strcmp(fsec, "0")))) {
1.47 schwarze 660: if (warnings)
661: say(path, "Wrong filename suffix");
662: continue;
663: } else
664: fsec[-1] = '\0';
1.51 schwarze 665:
1.47 schwarze 666: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.100 schwarze 667: if (strlcpy(mlink->file, path,
668: sizeof(mlink->file)) >=
669: sizeof(mlink->file)) {
670: say(path, "Filename too long");
671: free(mlink);
672: continue;
673: }
1.47 schwarze 674: mlink->dform = dform;
1.51 schwarze 675: mlink->dsec = dsec;
676: mlink->arch = arch;
677: mlink->name = ff->fts_name;
678: mlink->fsec = fsec;
1.81 schwarze 679: mlink->gzip = gzip;
1.47 schwarze 680: mlink_add(mlink, ff->fts_statp);
681: continue;
1.96 schwarze 682:
1.98 schwarze 683: case FTS_D:
684: case FTS_DP:
1.96 schwarze 685: break;
686:
687: default:
1.47 schwarze 688: if (warnings)
689: say(path, "Not a regular file");
690: continue;
691: }
692:
693: switch (ff->fts_level) {
1.98 schwarze 694: case 0:
1.47 schwarze 695: /* Ignore the root directory. */
696: break;
1.98 schwarze 697: case 1:
1.47 schwarze 698: /*
699: * This might contain manX/ or catX/.
700: * Try to infer this from the name.
701: * If we're not in use_all, enforce it.
702: */
703: cp = ff->fts_name;
1.153 schwarze 704: if (ff->fts_info == FTS_DP) {
705: dform = FORM_NONE;
706: dsec = NULL;
1.47 schwarze 707: break;
1.153 schwarze 708: }
1.15 schwarze 709:
1.153 schwarze 710: if ( ! strncmp(cp, "man", 3)) {
1.47 schwarze 711: dform = FORM_SRC;
712: dsec = cp + 3;
1.153 schwarze 713: } else if ( ! strncmp(cp, "cat", 3)) {
1.47 schwarze 714: dform = FORM_CAT;
715: dsec = cp + 3;
1.51 schwarze 716: } else {
717: dform = FORM_NONE;
718: dsec = NULL;
1.26 schwarze 719: }
1.47 schwarze 720:
1.153 schwarze 721: if (dsec != NULL || use_all)
1.47 schwarze 722: break;
723:
724: if (warnings)
725: say(path, "Unknown directory part");
726: fts_set(f, ff, FTS_SKIP);
727: break;
1.98 schwarze 728: case 2:
1.47 schwarze 729: /*
730: * Possibly our architecture.
731: * If we're descending, keep tabs on it.
732: */
1.153 schwarze 733: if (ff->fts_info != FTS_DP && dsec != NULL)
1.47 schwarze 734: arch = ff->fts_name;
1.51 schwarze 735: else
736: arch = NULL;
1.47 schwarze 737: break;
738: default:
1.153 schwarze 739: if (ff->fts_info == FTS_DP || use_all)
1.47 schwarze 740: break;
741: if (warnings)
742: say(path, "Extraneous directory part");
743: fts_set(f, ff, FTS_SKIP);
744: break;
1.14 schwarze 745: }
1.47 schwarze 746: }
747:
748: fts_close(f);
1.150 schwarze 749: return 1;
1.47 schwarze 750: }
1.1 schwarze 751:
1.47 schwarze 752: /*
753: * Add a file to the mlinks table.
754: * Do not verify that it's a "valid" looking manpage (we'll do that
755: * later).
756: *
757: * Try to infer the manual section, architecture, and page name from the
758: * path, assuming it looks like
759: *
1.98 schwarze 760: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 761: * or
762: * [./]cat<section>[/<arch>]/<name>.0
763: *
764: * See treescan() for the fts(3) version of this.
765: */
766: static void
767: filescan(const char *file)
768: {
769: char buf[PATH_MAX];
770: struct stat st;
771: struct mlink *mlink;
772: char *p, *start;
773:
774: assert(use_all);
775:
776: if (0 == strncmp(file, "./", 2))
777: file += 2;
778:
1.96 schwarze 779: /*
780: * We have to do lstat(2) before realpath(3) loses
781: * the information whether this is a symbolic link.
782: * We need to know that because for symbolic links,
783: * we want to use the orginal file name, while for
784: * regular files, we want to use the real path.
785: */
786: if (-1 == lstat(file, &st)) {
787: exitcode = (int)MANDOCLEVEL_BADARG;
788: say(file, "&lstat");
789: return;
790: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
791: exitcode = (int)MANDOCLEVEL_BADARG;
792: say(file, "Not a regular file");
793: return;
794: }
795:
796: /*
797: * We have to resolve the file name to the real path
798: * in any case for the base directory check.
799: */
1.47 schwarze 800: if (NULL == realpath(file, buf)) {
801: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 802: say(file, "&realpath");
1.47 schwarze 803: return;
1.63 schwarze 804: }
805:
1.109 schwarze 806: if (OP_TEST == op)
1.63 schwarze 807: start = buf;
1.109 schwarze 808: else if (strstr(buf, basedir) == buf)
809: start = buf + strlen(basedir);
1.63 schwarze 810: else {
1.47 schwarze 811: exitcode = (int)MANDOCLEVEL_BADARG;
812: say("", "%s: outside base directory", buf);
813: return;
1.63 schwarze 814: }
815:
1.96 schwarze 816: /*
817: * Now we are sure the file is inside our tree.
818: * If it is a symbolic link, ignore the real path
819: * and use the original name.
820: * This implies passing stuff like "cat1/../man1/foo.1"
821: * on the command line won't work. So don't do that.
822: * Note the stat(2) can still fail if the link target
823: * doesn't exist.
824: */
825: if (S_IFLNK & st.st_mode) {
826: if (-1 == stat(buf, &st)) {
827: exitcode = (int)MANDOCLEVEL_BADARG;
828: say(file, "&stat");
829: return;
830: }
1.100 schwarze 831: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
832: say(file, "Filename too long");
833: return;
834: }
1.109 schwarze 835: start = buf;
836: if (OP_TEST != op && strstr(buf, basedir) == buf)
837: start += strlen(basedir);
1.1 schwarze 838: }
1.63 schwarze 839:
1.47 schwarze 840: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 841: mlink->dform = FORM_NONE;
1.100 schwarze 842: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
843: sizeof(mlink->file)) {
844: say(start, "Filename too long");
1.134 schwarze 845: free(mlink);
1.100 schwarze 846: return;
847: }
1.1 schwarze 848:
1.10 schwarze 849: /*
1.47 schwarze 850: * First try to guess our directory structure.
851: * If we find a separator, try to look for man* or cat*.
852: * If we find one of these and what's underneath is a directory,
853: * assume it's an architecture.
1.10 schwarze 854: */
1.47 schwarze 855: if (NULL != (p = strchr(start, '/'))) {
856: *p++ = '\0';
857: if (0 == strncmp(start, "man", 3)) {
858: mlink->dform = FORM_SRC;
1.51 schwarze 859: mlink->dsec = start + 3;
1.47 schwarze 860: } else if (0 == strncmp(start, "cat", 3)) {
861: mlink->dform = FORM_CAT;
1.51 schwarze 862: mlink->dsec = start + 3;
1.47 schwarze 863: }
1.10 schwarze 864:
1.47 schwarze 865: start = p;
866: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
867: *p++ = '\0';
1.51 schwarze 868: mlink->arch = start;
1.47 schwarze 869: start = p;
1.41 deraadt 870: }
1.47 schwarze 871: }
1.7 schwarze 872:
1.47 schwarze 873: /*
874: * Now check the file suffix.
875: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
876: */
877: p = strrchr(start, '\0');
878: while (p-- > start && '/' != *p && '.' != *p)
879: /* Loop. */ ;
880:
881: if ('.' == *p) {
882: *p++ = '\0';
1.51 schwarze 883: mlink->fsec = p;
1.47 schwarze 884: }
1.41 deraadt 885:
1.47 schwarze 886: /*
887: * Now try to parse the name.
888: * Use the filename portion of the path.
889: */
890: mlink->name = start;
891: if (NULL != (p = strrchr(start, '/'))) {
892: mlink->name = p + 1;
893: *p = '\0';
894: }
895: mlink_add(mlink, &st);
896: }
1.2 schwarze 897:
1.47 schwarze 898: static void
899: mlink_add(struct mlink *mlink, const struct stat *st)
900: {
901: struct inodev inodev;
902: struct mpage *mpage;
903: unsigned int slot;
904:
905: assert(NULL != mlink->file);
906:
1.51 schwarze 907: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
908: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
909: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
910: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 911:
912: if ('0' == *mlink->fsec) {
913: free(mlink->fsec);
914: mlink->fsec = mandoc_strdup(mlink->dsec);
915: mlink->fform = FORM_CAT;
916: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
917: mlink->fform = FORM_SRC;
918: else
919: mlink->fform = FORM_NONE;
1.1 schwarze 920:
1.47 schwarze 921: slot = ohash_qlookup(&mlinks, mlink->file);
922: assert(NULL == ohash_find(&mlinks, slot));
923: ohash_insert(&mlinks, slot, mlink);
924:
1.139 schwarze 925: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.47 schwarze 926: inodev.st_ino = st->st_ino;
927: inodev.st_dev = st->st_dev;
928: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
929: sizeof(struct inodev), inodev.st_ino);
930: mpage = ohash_find(&mpages, slot);
931: if (NULL == mpage) {
932: mpage = mandoc_calloc(1, sizeof(struct mpage));
933: mpage->inodev.st_ino = inodev.st_ino;
934: mpage->inodev.st_dev = inodev.st_dev;
935: ohash_insert(&mpages, slot, mpage);
936: } else
937: mlink->next = mpage->mlinks;
938: mpage->mlinks = mlink;
1.75 schwarze 939: mlink->mpage = mpage;
1.47 schwarze 940: }
1.1 schwarze 941:
1.47 schwarze 942: static void
943: mlink_free(struct mlink *mlink)
944: {
1.1 schwarze 945:
1.47 schwarze 946: free(mlink->dsec);
947: free(mlink->arch);
948: free(mlink->name);
949: free(mlink->fsec);
950: free(mlink);
951: }
1.1 schwarze 952:
1.47 schwarze 953: static void
954: mpages_free(void)
955: {
956: struct mpage *mpage;
957: struct mlink *mlink;
958: unsigned int slot;
959:
960: mpage = ohash_first(&mpages, &slot);
961: while (NULL != mpage) {
962: while (NULL != (mlink = mpage->mlinks)) {
963: mpage->mlinks = mlink->next;
964: mlink_free(mlink);
965: }
966: free(mpage->sec);
967: free(mpage->arch);
968: free(mpage->title);
969: free(mpage->desc);
970: free(mpage);
971: mpage = ohash_next(&mpages, &slot);
972: }
973: }
1.1 schwarze 974:
1.47 schwarze 975: /*
976: * For each mlink to the mpage, check whether the path looks like
977: * it is formatted, and if it does, check whether a source manual
978: * exists by the same name, ignoring the suffix.
979: * If both conditions hold, drop the mlink.
980: */
981: static void
982: mlinks_undupe(struct mpage *mpage)
983: {
984: char buf[PATH_MAX];
985: struct mlink **prev;
986: struct mlink *mlink;
987: char *bufp;
988:
989: mpage->form = FORM_CAT;
990: prev = &mpage->mlinks;
991: while (NULL != (mlink = *prev)) {
992: if (FORM_CAT != mlink->dform) {
993: mpage->form = FORM_NONE;
994: goto nextlink;
995: }
1.100 schwarze 996: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 997: bufp = strstr(buf, "cat");
998: assert(NULL != bufp);
999: memcpy(bufp, "man", 3);
1000: if (NULL != (bufp = strrchr(buf, '.')))
1001: *++bufp = '\0';
1.100 schwarze 1002: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 1003: if (NULL == ohash_find(&mlinks,
1.98 schwarze 1004: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 1005: goto nextlink;
1006: if (warnings)
1007: say(mlink->file, "Man source exists: %s", buf);
1008: if (use_all)
1009: goto nextlink;
1010: *prev = mlink->next;
1011: mlink_free(mlink);
1012: continue;
1013: nextlink:
1014: prev = &(*prev)->next;
1.1 schwarze 1015: }
1.47 schwarze 1016: }
1.1 schwarze 1017:
1.88 schwarze 1018: static void
1.50 schwarze 1019: mlink_check(struct mpage *mpage, struct mlink *mlink)
1020: {
1.88 schwarze 1021: struct str *str;
1022: unsigned int slot;
1.50 schwarze 1023:
1024: /*
1025: * Check whether the manual section given in a file
1026: * agrees with the directory where the file is located.
1027: * Some manuals have suffixes like (3p) on their
1028: * section number either inside the file or in the
1029: * directory name, some are linked into more than one
1030: * section, like encrypt(1) = makekey(8).
1031: */
1032:
1033: if (FORM_SRC == mpage->form &&
1.88 schwarze 1034: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1035: say(mlink->file, "Section \"%s\" manual in %s directory",
1036: mpage->sec, mlink->dsec);
1037:
1038: /*
1039: * Manual page directories exist for each kernel
1040: * architecture as returned by machine(1).
1041: * However, many manuals only depend on the
1042: * application architecture as returned by arch(1).
1043: * For example, some (2/ARM) manuals are shared
1044: * across the "armish" and "zaurus" kernel
1045: * architectures.
1046: * A few manuals are even shared across completely
1047: * different architectures, for example fdformat(1)
1048: * on amd64, i386, sparc, and sparc64.
1049: */
1050:
1.88 schwarze 1051: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1052: say(mlink->file, "Architecture \"%s\" manual in "
1053: "\"%s\" directory", mpage->arch, mlink->arch);
1054:
1.88 schwarze 1055: /*
1056: * XXX
1.90 schwarze 1057: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1058: */
1059:
1060: if (FORM_CAT == mpage->form)
1061: return;
1062:
1063: /*
1064: * Check whether this mlink
1065: * appears as a name in the NAME section.
1066: */
1.50 schwarze 1067:
1.90 schwarze 1068: slot = ohash_qlookup(&names, mlink->name);
1069: str = ohash_find(&names, slot);
1.88 schwarze 1070: assert(NULL != str);
1.90 schwarze 1071: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1072: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1073: }
1074:
1.47 schwarze 1075: /*
1076: * Run through the files in the global vector "mpages"
1077: * and add them to the database specified in "basedir".
1078: *
1079: * This handles the parsing scheme itself, using the cues of directory
1080: * and filename to determine whether the file is parsable or not.
1081: */
1082: static void
1.131 schwarze 1083: mpages_merge(struct mparse *mp)
1.47 schwarze 1084: {
1.70 schwarze 1085: char any[] = "any";
1.75 schwarze 1086: struct mpage *mpage, *mpage_dest;
1087: struct mlink *mlink, *mlink_dest;
1.144 schwarze 1088: struct roff_man *man;
1.75 schwarze 1089: char *sodest;
1.69 schwarze 1090: char *cp;
1.117 schwarze 1091: int fd;
1.58 schwarze 1092: unsigned int pslot;
1.47 schwarze 1093:
1.126 schwarze 1094: if ( ! nodb)
1.64 schwarze 1095: SQL_EXEC("BEGIN TRANSACTION");
1096:
1.47 schwarze 1097: mpage = ohash_first(&mpages, &pslot);
1.126 schwarze 1098: while (mpage != NULL) {
1.47 schwarze 1099: mlinks_undupe(mpage);
1.136 schwarze 1100: if ((mlink = mpage->mlinks) == NULL) {
1.47 schwarze 1101: mpage = ohash_next(&mpages, &pslot);
1102: continue;
1103: }
1.1 schwarze 1104:
1.90 schwarze 1105: name_mask = NAME_MASK;
1.156 schwarze 1106: mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1107: mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1.47 schwarze 1108: mparse_reset(mp);
1109: man = NULL;
1.81 schwarze 1110: sodest = NULL;
1111:
1.136 schwarze 1112: mparse_open(mp, &fd, mlink->file);
1.117 schwarze 1113: if (fd == -1) {
1.136 schwarze 1114: say(mlink->file, "&open");
1.117 schwarze 1115: goto nextpage;
1.81 schwarze 1116: }
1.11 schwarze 1117:
1118: /*
1.137 schwarze 1119: * Interpret the file as mdoc(7) or man(7) source
1120: * code, unless it is known to be formatted.
1.11 schwarze 1121: */
1.136 schwarze 1122: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.137 schwarze 1123: mparse_readfd(mp, fd, mlink->file);
1.145 schwarze 1124: mparse_result(mp, &man, &sodest);
1.47 schwarze 1125: }
1.11 schwarze 1126:
1.126 schwarze 1127: if (sodest != NULL) {
1.75 schwarze 1128: mlink_dest = ohash_find(&mlinks,
1129: ohash_qlookup(&mlinks, sodest));
1.126 schwarze 1130: if (mlink_dest == NULL) {
1131: mandoc_asprintf(&cp, "%s.gz", sodest);
1132: mlink_dest = ohash_find(&mlinks,
1133: ohash_qlookup(&mlinks, cp));
1134: free(cp);
1135: }
1136: if (mlink_dest != NULL) {
1.75 schwarze 1137:
1138: /* The .so target exists. */
1139:
1140: mpage_dest = mlink_dest->mpage;
1141: while (1) {
1142: mlink->mpage = mpage_dest;
1143:
1144: /*
1145: * If the target was already
1146: * processed, add the links
1147: * to the database now.
1148: * Otherwise, this will
1149: * happen when we come
1150: * to the target.
1151: */
1152:
1.94 schwarze 1153: if (mpage_dest->pageid)
1.124 schwarze 1154: dbadd_mlink_name(mlink);
1.75 schwarze 1155:
1.126 schwarze 1156: if (mlink->next == NULL)
1.75 schwarze 1157: break;
1158: mlink = mlink->next;
1159: }
1160:
1161: /* Move all links to the target. */
1162:
1163: mlink->next = mlink_dest->next;
1164: mlink_dest->next = mpage->mlinks;
1165: mpage->mlinks = NULL;
1166: }
1.81 schwarze 1167: goto nextpage;
1.145 schwarze 1168: } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1.159 schwarze 1169: mdoc_validate(man);
1.47 schwarze 1170: mpage->form = FORM_SRC;
1.146 schwarze 1171: mpage->sec = man->meta.msec;
1.112 schwarze 1172: mpage->sec = mandoc_strdup(
1.126 schwarze 1173: mpage->sec == NULL ? "" : mpage->sec);
1.146 schwarze 1174: mpage->arch = man->meta.arch;
1.47 schwarze 1175: mpage->arch = mandoc_strdup(
1.126 schwarze 1176: mpage->arch == NULL ? "" : mpage->arch);
1.146 schwarze 1177: mpage->title = mandoc_strdup(man->meta.title);
1.145 schwarze 1178: } else if (man != NULL && man->macroset == MACROSET_MAN) {
1.160 schwarze 1179: man_validate(man);
1.47 schwarze 1180: mpage->form = FORM_SRC;
1.146 schwarze 1181: mpage->sec = mandoc_strdup(man->meta.msec);
1.136 schwarze 1182: mpage->arch = mandoc_strdup(mlink->arch);
1.146 schwarze 1183: mpage->title = mandoc_strdup(man->meta.title);
1.11 schwarze 1184: } else {
1.47 schwarze 1185: mpage->form = FORM_CAT;
1.136 schwarze 1186: mpage->sec = mandoc_strdup(mlink->dsec);
1187: mpage->arch = mandoc_strdup(mlink->arch);
1188: mpage->title = mandoc_strdup(mlink->name);
1.1 schwarze 1189: }
1.54 schwarze 1190: putkey(mpage, mpage->sec, TYPE_sec);
1.119 schwarze 1191: if (*mpage->arch != '\0')
1192: putkey(mpage, mpage->arch, TYPE_arch);
1.1 schwarze 1193:
1.136 schwarze 1194: for ( ; mlink != NULL; mlink = mlink->next) {
1.54 schwarze 1195: if ('\0' != *mlink->dsec)
1196: putkey(mpage, mlink->dsec, TYPE_sec);
1197: if ('\0' != *mlink->fsec)
1198: putkey(mpage, mlink->fsec, TYPE_sec);
1.55 schwarze 1199: putkey(mpage, '\0' == *mlink->arch ?
1.70 schwarze 1200: any : mlink->arch, TYPE_arch);
1.90 schwarze 1201: putkey(mpage, mlink->name, NAME_FILE);
1.54 schwarze 1202: }
1.41 deraadt 1203:
1.129 schwarze 1204: assert(mpage->desc == NULL);
1.145 schwarze 1205: if (man != NULL && man->macroset == MACROSET_MDOC)
1.146 schwarze 1206: parse_mdoc(mpage, &man->meta, man->first);
1.129 schwarze 1207: else if (man != NULL)
1.146 schwarze 1208: parse_man(mpage, &man->meta, man->first);
1.47 schwarze 1209: else
1.117 schwarze 1210: parse_cat(mpage, fd);
1.129 schwarze 1211: if (mpage->desc == NULL)
1.92 schwarze 1212: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1.88 schwarze 1213:
1214: if (warnings && !use_all)
1215: for (mlink = mpage->mlinks; mlink;
1216: mlink = mlink->next)
1217: mlink_check(mpage, mlink);
1.6 schwarze 1218:
1.131 schwarze 1219: dbadd(mpage);
1.136 schwarze 1220: mlink = mpage->mlinks;
1.81 schwarze 1221:
1222: nextpage:
1.47 schwarze 1223: ohash_delete(&strings);
1.90 schwarze 1224: ohash_delete(&names);
1.47 schwarze 1225: mpage = ohash_next(&mpages, &pslot);
1226: }
1.64 schwarze 1227:
1228: if (0 == nodb)
1229: SQL_EXEC("END TRANSACTION");
1.87 schwarze 1230: }
1231:
1232: static void
1233: names_check(void)
1234: {
1235: sqlite3_stmt *stmt;
1236: const char *name, *sec, *arch, *key;
1237:
1238: sqlite3_prepare_v2(db,
1239: "SELECT name, sec, arch, key FROM ("
1.90 schwarze 1240: "SELECT name AS key, pageid FROM names "
1.87 schwarze 1241: "WHERE bits & ? AND NOT EXISTS ("
1242: "SELECT pageid FROM mlinks "
1.90 schwarze 1243: "WHERE mlinks.pageid == names.pageid "
1244: "AND mlinks.name == names.name"
1.87 schwarze 1245: ")"
1246: ") JOIN ("
1.99 schwarze 1247: "SELECT sec, arch, name, pageid FROM mlinks "
1248: "GROUP BY pageid"
1.87 schwarze 1249: ") USING (pageid);",
1250: -1, &stmt, NULL);
1251:
1.154 schwarze 1252: if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK)
1.91 schwarze 1253: say("", "%s", sqlite3_errmsg(db));
1.87 schwarze 1254:
1.154 schwarze 1255: while (sqlite3_step(stmt) == SQLITE_ROW) {
1.113 schwarze 1256: name = (const char *)sqlite3_column_text(stmt, 0);
1257: sec = (const char *)sqlite3_column_text(stmt, 1);
1258: arch = (const char *)sqlite3_column_text(stmt, 2);
1259: key = (const char *)sqlite3_column_text(stmt, 3);
1.87 schwarze 1260: say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1261: '\0' == *arch ? "" : "/",
1262: '\0' == *arch ? "" : arch, key);
1263: }
1264: sqlite3_finalize(stmt);
1.47 schwarze 1265: }
1.6 schwarze 1266:
1.47 schwarze 1267: static void
1.81 schwarze 1268: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1269: {
1270: FILE *stream;
1271: char *line, *p, *title;
1272: size_t len, plen, titlesz;
1.1 schwarze 1273:
1.81 schwarze 1274: stream = (-1 == fd) ?
1275: fopen(mpage->mlinks->file, "r") :
1276: fdopen(fd, "r");
1277: if (NULL == stream) {
1.114 doug 1278: if (-1 != fd)
1279: close(fd);
1.47 schwarze 1280: if (warnings)
1.80 schwarze 1281: say(mpage->mlinks->file, "&fopen");
1.47 schwarze 1282: return;
1283: }
1.1 schwarze 1284:
1.47 schwarze 1285: /* Skip to first blank line. */
1.1 schwarze 1286:
1.47 schwarze 1287: while (NULL != (line = fgetln(stream, &len)))
1288: if ('\n' == *line)
1289: break;
1.1 schwarze 1290:
1.47 schwarze 1291: /*
1292: * Assume the first line that is not indented
1293: * is the first section header. Skip to it.
1294: */
1.1 schwarze 1295:
1.47 schwarze 1296: while (NULL != (line = fgetln(stream, &len)))
1297: if ('\n' != *line && ' ' != *line)
1298: break;
1.98 schwarze 1299:
1.47 schwarze 1300: /*
1301: * Read up until the next section into a buffer.
1302: * Strip the leading and trailing newline from each read line,
1303: * appending a trailing space.
1304: * Ignore empty (whitespace-only) lines.
1305: */
1.28 schwarze 1306:
1.47 schwarze 1307: titlesz = 0;
1308: title = NULL;
1.38 schwarze 1309:
1.47 schwarze 1310: while (NULL != (line = fgetln(stream, &len))) {
1311: if (' ' != *line || '\n' != line[len - 1])
1312: break;
1313: while (len > 0 && isspace((unsigned char)*line)) {
1314: line++;
1315: len--;
1316: }
1317: if (1 == len)
1318: continue;
1319: title = mandoc_realloc(title, titlesz + len);
1320: memcpy(title + titlesz, line, len);
1321: titlesz += len;
1322: title[titlesz - 1] = ' ';
1323: }
1.28 schwarze 1324:
1.47 schwarze 1325: /*
1326: * If no page content can be found, or the input line
1327: * is already the next section header, or there is no
1328: * trailing newline, reuse the page title as the page
1329: * description.
1330: */
1.1 schwarze 1331:
1.47 schwarze 1332: if (NULL == title || '\0' == *title) {
1333: if (warnings)
1334: say(mpage->mlinks->file,
1335: "Cannot find NAME section");
1336: fclose(stream);
1337: free(title);
1338: return;
1339: }
1.24 schwarze 1340:
1.47 schwarze 1341: title = mandoc_realloc(title, titlesz + 1);
1342: title[titlesz] = '\0';
1.24 schwarze 1343:
1.47 schwarze 1344: /*
1345: * Skip to the first dash.
1346: * Use the remaining line as the description (no more than 70
1347: * bytes).
1348: */
1.28 schwarze 1349:
1.47 schwarze 1350: if (NULL != (p = strstr(title, "- "))) {
1351: for (p += 2; ' ' == *p || '\b' == *p; p++)
1352: /* Skip to next word. */ ;
1353: } else {
1354: if (warnings)
1355: say(mpage->mlinks->file,
1356: "No dash in title line");
1357: p = title;
1358: }
1.1 schwarze 1359:
1.47 schwarze 1360: plen = strlen(p);
1.1 schwarze 1361:
1.47 schwarze 1362: /* Strip backspace-encoding from line. */
1.1 schwarze 1363:
1.47 schwarze 1364: while (NULL != (line = memchr(p, '\b', plen))) {
1365: len = line - p;
1366: if (0 == len) {
1367: memmove(line, line + 1, plen--);
1368: continue;
1.98 schwarze 1369: }
1.47 schwarze 1370: memmove(line - 1, line + 1, plen - len);
1371: plen -= 2;
1372: }
1.1 schwarze 1373:
1.47 schwarze 1374: mpage->desc = mandoc_strdup(p);
1375: fclose(stream);
1376: free(title);
1377: }
1.16 schwarze 1378:
1.47 schwarze 1379: /*
1380: * Put a type/word pair into the word database for this particular file.
1381: */
1382: static void
1.69 schwarze 1383: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1384: {
1.69 schwarze 1385: char *cp;
1.37 schwarze 1386:
1.47 schwarze 1387: assert(NULL != value);
1.69 schwarze 1388: if (TYPE_arch == type)
1389: for (cp = value; *cp; cp++)
1390: if (isupper((unsigned char)*cp))
1391: *cp = _tolower((unsigned char)*cp);
1.47 schwarze 1392: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1393: }
1394:
1395: /*
1.47 schwarze 1396: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1397: */
1398: static void
1.47 schwarze 1399: putmdockey(const struct mpage *mpage,
1.142 schwarze 1400: const struct roff_node *n, uint64_t m)
1.2 schwarze 1401: {
1.16 schwarze 1402:
1.47 schwarze 1403: for ( ; NULL != n; n = n->next) {
1404: if (NULL != n->child)
1405: putmdockey(mpage, n->child, m);
1.141 schwarze 1406: if (n->type == ROFFT_TEXT)
1.47 schwarze 1407: putkey(mpage, n->string, m);
1408: }
1409: }
1.16 schwarze 1410:
1.47 schwarze 1411: static void
1.143 schwarze 1412: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1413: const struct roff_node *n)
1.47 schwarze 1414: {
1.142 schwarze 1415: const struct roff_node *head, *body;
1.78 schwarze 1416: char *start, *title;
1.47 schwarze 1417: char byte;
1.78 schwarze 1418: size_t sz;
1.16 schwarze 1419:
1.47 schwarze 1420: if (NULL == n)
1421: return;
1.16 schwarze 1422:
1.47 schwarze 1423: /*
1424: * We're only searching for one thing: the first text child in
1425: * the BODY of a NAME section. Since we don't keep track of
1426: * sections in -man, run some hoops to find out whether we're in
1427: * the correct section or not.
1428: */
1.16 schwarze 1429:
1.141 schwarze 1430: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1431: body = n;
1432: assert(body->parent);
1433: if (NULL != (head = body->parent->head) &&
1.98 schwarze 1434: 1 == head->nchild &&
1435: NULL != (head = (head->child)) &&
1.141 schwarze 1436: head->type == ROFFT_TEXT &&
1.98 schwarze 1437: 0 == strcmp(head->string, "NAME") &&
1438: NULL != body->child) {
1.2 schwarze 1439:
1.47 schwarze 1440: /*
1441: * Suck the entire NAME section into memory.
1442: * Yes, we might run away.
1443: * But too many manuals have big, spread-out
1444: * NAME sections over many lines.
1445: */
1.2 schwarze 1446:
1.78 schwarze 1447: title = NULL;
1.147 schwarze 1448: deroff(&title, body);
1.47 schwarze 1449: if (NULL == title)
1450: return;
1.16 schwarze 1451:
1.98 schwarze 1452: /*
1.47 schwarze 1453: * Go through a special heuristic dance here.
1454: * Conventionally, one or more manual names are
1455: * comma-specified prior to a whitespace, then a
1456: * dash, then a description. Try to puzzle out
1457: * the name parts here.
1458: */
1.16 schwarze 1459:
1.78 schwarze 1460: start = title;
1.47 schwarze 1461: for ( ;; ) {
1462: sz = strcspn(start, " ,");
1463: if ('\0' == start[sz])
1464: break;
1.1 schwarze 1465:
1.47 schwarze 1466: byte = start[sz];
1467: start[sz] = '\0';
1.67 schwarze 1468:
1469: /*
1470: * Assume a stray trailing comma in the
1471: * name list if a name begins with a dash.
1472: */
1473:
1474: if ('-' == start[0] ||
1475: ('\\' == start[0] && '-' == start[1]))
1476: break;
1.1 schwarze 1477:
1.90 schwarze 1478: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1479: if ( ! (mpage->name_head_done ||
1480: strcasecmp(start, meta->title))) {
1481: putkey(mpage, start, NAME_HEAD);
1482: mpage->name_head_done = 1;
1483: }
1.1 schwarze 1484:
1.47 schwarze 1485: if (' ' == byte) {
1486: start += sz + 1;
1487: break;
1488: }
1.1 schwarze 1489:
1.47 schwarze 1490: assert(',' == byte);
1491: start += sz + 1;
1492: while (' ' == *start)
1493: start++;
1494: }
1.1 schwarze 1495:
1.78 schwarze 1496: if (start == title) {
1.90 schwarze 1497: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1498: if ( ! (mpage->name_head_done ||
1499: strcasecmp(start, meta->title))) {
1500: putkey(mpage, start, NAME_HEAD);
1501: mpage->name_head_done = 1;
1502: }
1.47 schwarze 1503: free(title);
1504: return;
1505: }
1.1 schwarze 1506:
1.47 schwarze 1507: while (isspace((unsigned char)*start))
1508: start++;
1.1 schwarze 1509:
1.47 schwarze 1510: if (0 == strncmp(start, "-", 1))
1511: start += 1;
1512: else if (0 == strncmp(start, "\\-\\-", 4))
1513: start += 4;
1514: else if (0 == strncmp(start, "\\-", 2))
1515: start += 2;
1516: else if (0 == strncmp(start, "\\(en", 4))
1517: start += 4;
1518: else if (0 == strncmp(start, "\\(em", 4))
1519: start += 4;
1.1 schwarze 1520:
1.47 schwarze 1521: while (' ' == *start)
1522: start++;
1.1 schwarze 1523:
1.47 schwarze 1524: mpage->desc = mandoc_strdup(start);
1525: free(title);
1526: return;
1527: }
1528: }
1.1 schwarze 1529:
1.47 schwarze 1530: for (n = n->child; n; n = n->next) {
1531: if (NULL != mpage->desc)
1532: break;
1.129 schwarze 1533: parse_man(mpage, meta, n);
1.1 schwarze 1534: }
1535: }
1536:
1537: static void
1.143 schwarze 1538: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1539: const struct roff_node *n)
1.1 schwarze 1540: {
1541:
1.47 schwarze 1542: assert(NULL != n);
1543: for (n = n->child; NULL != n; n = n->next) {
1544: switch (n->type) {
1.141 schwarze 1545: case ROFFT_ELEM:
1546: case ROFFT_BLOCK:
1547: case ROFFT_HEAD:
1548: case ROFFT_BODY:
1549: case ROFFT_TAIL:
1.47 schwarze 1550: if (NULL != mdocs[n->tok].fp)
1.127 schwarze 1551: if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1.47 schwarze 1552: break;
1553: if (mdocs[n->tok].mask)
1554: putmdockey(mpage, n->child,
1555: mdocs[n->tok].mask);
1556: break;
1557: default:
1.141 schwarze 1558: assert(n->type != ROFFT_ROOT);
1.47 schwarze 1559: continue;
1560: }
1561: if (NULL != n->child)
1.127 schwarze 1562: parse_mdoc(mpage, meta, n);
1.1 schwarze 1563: }
1564: }
1565:
1.19 schwarze 1566: static int
1.143 schwarze 1567: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1568: const struct roff_node *n)
1.1 schwarze 1569: {
1.131 schwarze 1570: char *start, *end;
1.1 schwarze 1571: size_t sz;
1.19 schwarze 1572:
1.47 schwarze 1573: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1574: NULL == (n = n->child) ||
1.141 schwarze 1575: n->type != ROFFT_TEXT)
1.150 schwarze 1576: return 0;
1.1 schwarze 1577:
1578: /*
1579: * Only consider those `Fd' macro fields that begin with an
1580: * "inclusion" token (versus, e.g., #define).
1581: */
1.47 schwarze 1582:
1.1 schwarze 1583: if (strcmp("#include", n->string))
1.150 schwarze 1584: return 0;
1.1 schwarze 1585:
1.141 schwarze 1586: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1587: return 0;
1.1 schwarze 1588:
1589: /*
1590: * Strip away the enclosing angle brackets and make sure we're
1591: * not zero-length.
1592: */
1593:
1594: start = n->string;
1595: if ('<' == *start || '"' == *start)
1596: start++;
1597:
1598: if (0 == (sz = strlen(start)))
1.150 schwarze 1599: return 0;
1.1 schwarze 1600:
1601: end = &start[(int)sz - 1];
1602: if ('>' == *end || '"' == *end)
1603: end--;
1604:
1.47 schwarze 1605: if (end > start)
1606: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1607: return 0;
1.1 schwarze 1608: }
1609:
1.133 schwarze 1610: static void
1.142 schwarze 1611: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1612: {
1.69 schwarze 1613: char *cp;
1.133 schwarze 1614: size_t sz;
1.1 schwarze 1615:
1.141 schwarze 1616: if (n->type != ROFFT_TEXT)
1.133 schwarze 1617: return;
1.19 schwarze 1618:
1.133 schwarze 1619: /* Skip function pointer punctuation. */
1.1 schwarze 1620:
1.133 schwarze 1621: cp = n->string;
1622: while (*cp == '(' || *cp == '*')
1.1 schwarze 1623: cp++;
1.133 schwarze 1624: sz = strcspn(cp, "()");
1.1 schwarze 1625:
1.133 schwarze 1626: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1627: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1628: putkeys(mpage, cp, sz, NAME_SYN);
1629: }
1.19 schwarze 1630:
1.133 schwarze 1631: static int
1.143 schwarze 1632: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1633: const struct roff_node *n)
1.133 schwarze 1634: {
1635:
1636: if (n->child == NULL)
1.150 schwarze 1637: return 0;
1.133 schwarze 1638:
1639: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1640:
1.133 schwarze 1641: for (n = n->child->next; n != NULL; n = n->next)
1.141 schwarze 1642: if (n->type == ROFFT_TEXT)
1.47 schwarze 1643: putkey(mpage, n->string, TYPE_Fa);
1.19 schwarze 1644:
1.150 schwarze 1645: return 0;
1.128 schwarze 1646: }
1647:
1648: static int
1.143 schwarze 1649: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1650: const struct roff_node *n)
1.128 schwarze 1651: {
1.132 schwarze 1652:
1.141 schwarze 1653: if (n->type != ROFFT_HEAD)
1.150 schwarze 1654: return 1;
1.128 schwarze 1655:
1.133 schwarze 1656: if (n->child != NULL)
1657: parse_mdoc_fname(mpage, n->child);
1658:
1.150 schwarze 1659: return 0;
1.1 schwarze 1660: }
1661:
1.19 schwarze 1662: static int
1.143 schwarze 1663: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1664: const struct roff_node *n)
1.1 schwarze 1665: {
1.47 schwarze 1666: char *cp;
1.1 schwarze 1667:
1668: if (NULL == (n = n->child))
1.150 schwarze 1669: return 0;
1.1 schwarze 1670:
1.47 schwarze 1671: if (NULL == n->next) {
1672: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1673: return 0;
1.47 schwarze 1674: }
1.1 schwarze 1675:
1.77 schwarze 1676: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1677: putkey(mpage, cp, TYPE_Xr);
1678: free(cp);
1.150 schwarze 1679: return 0;
1.1 schwarze 1680: }
1681:
1.19 schwarze 1682: static int
1.143 schwarze 1683: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1684: const struct roff_node *n)
1.1 schwarze 1685: {
1686:
1.141 schwarze 1687: if (n->type == ROFFT_BODY)
1.147 schwarze 1688: deroff(&mpage->desc, n);
1.150 schwarze 1689: return 0;
1.1 schwarze 1690: }
1691:
1.19 schwarze 1692: static int
1.143 schwarze 1693: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1694: const struct roff_node *n)
1.1 schwarze 1695: {
1696:
1.86 schwarze 1697: if (SEC_NAME == n->sec)
1.90 schwarze 1698: putmdockey(mpage, n->child, NAME_TITLE);
1.141 schwarze 1699: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1700: if (n->child == NULL)
1701: putkey(mpage, meta->name, NAME_SYN);
1702: else
1703: putmdockey(mpage, n->child, NAME_SYN);
1.129 schwarze 1704: }
1705: if ( ! (mpage->name_head_done ||
1706: n->child == NULL || n->child->string == NULL ||
1707: strcasecmp(n->child->string, meta->title))) {
1.141 schwarze 1708: putkey(mpage, n->child->string, ROFFT_HEAD);
1.129 schwarze 1709: mpage->name_head_done = 1;
1.127 schwarze 1710: }
1.150 schwarze 1711: return 0;
1.1 schwarze 1712: }
1713:
1.19 schwarze 1714: static int
1.143 schwarze 1715: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1716: const struct roff_node *n)
1.1 schwarze 1717: {
1718:
1.150 schwarze 1719: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1720: }
1721:
1.47 schwarze 1722: static int
1.143 schwarze 1723: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1724: const struct roff_node *n)
1.1 schwarze 1725: {
1726:
1.150 schwarze 1727: return n->type == ROFFT_HEAD;
1.47 schwarze 1728: }
1.1 schwarze 1729:
1.47 schwarze 1730: static int
1.143 schwarze 1731: parse_mdoc_body(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1732: const struct roff_node *n)
1.47 schwarze 1733: {
1.1 schwarze 1734:
1.150 schwarze 1735: return n->type == ROFFT_BODY;
1.1 schwarze 1736: }
1737:
1.47 schwarze 1738: /*
1739: * Add a string to the hash table for the current manual.
1740: * Each string has a bitmask telling which macros it belongs to.
1741: * When we finish the manual, we'll dump the table.
1742: */
1.1 schwarze 1743: static void
1.131 schwarze 1744: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1745: {
1.90 schwarze 1746: struct ohash *htab;
1.47 schwarze 1747: struct str *s;
1.68 schwarze 1748: const char *end;
1.47 schwarze 1749: unsigned int slot;
1.131 schwarze 1750: int i, mustfree;
1.1 schwarze 1751:
1.47 schwarze 1752: if (0 == sz)
1753: return;
1.68 schwarze 1754:
1.131 schwarze 1755: mustfree = render_string(&cp, &sz);
1756:
1.90 schwarze 1757: if (TYPE_Nm & v) {
1758: htab = &names;
1759: v &= name_mask;
1.124 schwarze 1760: if (v & NAME_FIRST)
1761: name_mask &= ~NAME_FIRST;
1.90 schwarze 1762: if (debug > 1)
1763: say(mpage->mlinks->file,
1.130 schwarze 1764: "Adding name %*s, bits=%d", sz, cp, v);
1.90 schwarze 1765: } else {
1766: htab = &strings;
1767: if (debug > 1)
1768: for (i = 0; i < mansearch_keymax; i++)
1.118 schwarze 1769: if ((uint64_t)1 << i & v)
1.90 schwarze 1770: say(mpage->mlinks->file,
1771: "Adding key %s=%*s",
1772: mansearch_keynames[i], sz, cp);
1.68 schwarze 1773: }
1.47 schwarze 1774:
1775: end = cp + sz;
1.90 schwarze 1776: slot = ohash_qlookupi(htab, cp, &end);
1777: s = ohash_find(htab, slot);
1.1 schwarze 1778:
1.47 schwarze 1779: if (NULL != s && mpage == s->mpage) {
1780: s->mask |= v;
1.1 schwarze 1781: return;
1.47 schwarze 1782: } else if (NULL == s) {
1.101 schwarze 1783: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1784: memcpy(s->key, cp, sz);
1.90 schwarze 1785: ohash_insert(htab, slot, s);
1.47 schwarze 1786: }
1787: s->mpage = mpage;
1788: s->mask = v;
1.131 schwarze 1789:
1790: if (mustfree)
1791: free(cp);
1.1 schwarze 1792: }
1793:
1794: /*
1.47 schwarze 1795: * Take a Unicode codepoint and produce its UTF-8 encoding.
1796: * This isn't the best way to do this, but it works.
1797: * The magic numbers are from the UTF-8 packaging.
1798: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1799: */
1.47 schwarze 1800: static size_t
1801: utf8(unsigned int cp, char out[7])
1.1 schwarze 1802: {
1.47 schwarze 1803: size_t rc;
1.1 schwarze 1804:
1.47 schwarze 1805: rc = 0;
1806: if (cp <= 0x0000007F) {
1807: rc = 1;
1808: out[0] = (char)cp;
1809: } else if (cp <= 0x000007FF) {
1810: rc = 2;
1811: out[0] = (cp >> 6 & 31) | 192;
1812: out[1] = (cp & 63) | 128;
1813: } else if (cp <= 0x0000FFFF) {
1814: rc = 3;
1815: out[0] = (cp >> 12 & 15) | 224;
1816: out[1] = (cp >> 6 & 63) | 128;
1817: out[2] = (cp & 63) | 128;
1818: } else if (cp <= 0x001FFFFF) {
1819: rc = 4;
1820: out[0] = (cp >> 18 & 7) | 240;
1821: out[1] = (cp >> 12 & 63) | 128;
1822: out[2] = (cp >> 6 & 63) | 128;
1823: out[3] = (cp & 63) | 128;
1824: } else if (cp <= 0x03FFFFFF) {
1825: rc = 5;
1826: out[0] = (cp >> 24 & 3) | 248;
1827: out[1] = (cp >> 18 & 63) | 128;
1828: out[2] = (cp >> 12 & 63) | 128;
1829: out[3] = (cp >> 6 & 63) | 128;
1830: out[4] = (cp & 63) | 128;
1831: } else if (cp <= 0x7FFFFFFF) {
1832: rc = 6;
1833: out[0] = (cp >> 30 & 1) | 252;
1834: out[1] = (cp >> 24 & 63) | 128;
1835: out[2] = (cp >> 18 & 63) | 128;
1836: out[3] = (cp >> 12 & 63) | 128;
1837: out[4] = (cp >> 6 & 63) | 128;
1838: out[5] = (cp & 63) | 128;
1839: } else
1.150 schwarze 1840: return 0;
1.19 schwarze 1841:
1.47 schwarze 1842: out[rc] = '\0';
1.150 schwarze 1843: return rc;
1.1 schwarze 1844: }
1845:
1.47 schwarze 1846: /*
1.131 schwarze 1847: * If the string contains escape sequences,
1848: * replace it with an allocated rendering and return 1,
1849: * such that the caller can free it after use.
1850: * Otherwise, do nothing and return 0.
1.47 schwarze 1851: */
1.131 schwarze 1852: static int
1853: render_string(char **public, size_t *psz)
1.1 schwarze 1854: {
1.131 schwarze 1855: const char *src, *scp, *addcp, *seq;
1856: char *dst;
1857: size_t ssz, dsz, addsz;
1.71 schwarze 1858: char utfbuf[7], res[6];
1.131 schwarze 1859: int seqlen, unicode;
1.47 schwarze 1860:
1861: res[0] = '\\';
1862: res[1] = '\t';
1863: res[2] = ASCII_NBRSP;
1864: res[3] = ASCII_HYPH;
1.71 schwarze 1865: res[4] = ASCII_BREAK;
1866: res[5] = '\0';
1.1 schwarze 1867:
1.131 schwarze 1868: src = scp = *public;
1869: ssz = *psz;
1870: dst = NULL;
1871: dsz = 0;
1872:
1873: while (scp < src + *psz) {
1874:
1875: /* Leave normal characters unchanged. */
1876:
1877: if (strchr(res, *scp) == NULL) {
1878: if (dst != NULL)
1879: dst[dsz++] = *scp;
1880: scp++;
1881: continue;
1882: }
1.1 schwarze 1883:
1.47 schwarze 1884: /*
1.131 schwarze 1885: * Found something that requires replacing,
1886: * make sure we have a destination buffer.
1.47 schwarze 1887: */
1.131 schwarze 1888:
1889: if (dst == NULL) {
1890: dst = mandoc_malloc(ssz + 1);
1891: dsz = scp - src;
1892: memcpy(dst, src, dsz);
1.47 schwarze 1893: }
1.39 schwarze 1894:
1.131 schwarze 1895: /* Handle single-char special characters. */
1896:
1897: switch (*scp) {
1898: case '\\':
1899: break;
1.98 schwarze 1900: case '\t':
1901: case ASCII_NBRSP:
1.131 schwarze 1902: dst[dsz++] = ' ';
1903: scp++;
1904: continue;
1905: case ASCII_HYPH:
1906: dst[dsz++] = '-';
1.71 schwarze 1907: /* FALLTHROUGH */
1.98 schwarze 1908: case ASCII_BREAK:
1.131 schwarze 1909: scp++;
1.47 schwarze 1910: continue;
1.71 schwarze 1911: default:
1.131 schwarze 1912: abort();
1.71 schwarze 1913: }
1.39 schwarze 1914:
1.47 schwarze 1915: /*
1.131 schwarze 1916: * Found an escape sequence.
1917: * Read past the slash, then parse it.
1918: * Ignore everything except characters.
1.47 schwarze 1919: */
1.52 schwarze 1920:
1.131 schwarze 1921: scp++;
1922: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1923: continue;
1.39 schwarze 1924:
1.47 schwarze 1925: /*
1.52 schwarze 1926: * Render the special character
1927: * as either UTF-8 or ASCII.
1.47 schwarze 1928: */
1.52 schwarze 1929:
1930: if (write_utf8) {
1.157 schwarze 1931: unicode = mchars_spec2cp(seq, seqlen);
1.131 schwarze 1932: if (unicode <= 0)
1.52 schwarze 1933: continue;
1.131 schwarze 1934: addsz = utf8(unicode, utfbuf);
1935: if (addsz == 0)
1.52 schwarze 1936: continue;
1.131 schwarze 1937: addcp = utfbuf;
1.52 schwarze 1938: } else {
1.157 schwarze 1939: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.131 schwarze 1940: if (addcp == NULL)
1.52 schwarze 1941: continue;
1.131 schwarze 1942: if (*addcp == ASCII_NBRSP) {
1943: addcp = " ";
1944: addsz = 1;
1.52 schwarze 1945: }
1946: }
1.1 schwarze 1947:
1.47 schwarze 1948: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1949:
1.131 schwarze 1950: ssz += addsz;
1951: dst = mandoc_realloc(dst, ssz + 1);
1952: memcpy(dst + dsz, addcp, addsz);
1953: dsz += addsz;
1954: }
1955: if (dst != NULL) {
1956: *public = dst;
1957: *psz = dsz;
1958: }
1959:
1960: /* Trim trailing whitespace and NUL-terminate. */
1961:
1962: while (*psz > 0 && (*public)[*psz - 1] == ' ')
1963: --*psz;
1964: if (dst != NULL) {
1965: (*public)[*psz] = '\0';
1.150 schwarze 1966: return 1;
1.131 schwarze 1967: } else
1.150 schwarze 1968: return 0;
1.1 schwarze 1969: }
1970:
1.75 schwarze 1971: static void
1972: dbadd_mlink(const struct mlink *mlink)
1973: {
1974: size_t i;
1975:
1976: i = 1;
1977: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1978: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1979: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.94 schwarze 1980: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1.75 schwarze 1981: SQL_STEP(stmts[STMT_INSERT_LINK]);
1982: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1.124 schwarze 1983: }
1984:
1985: static void
1986: dbadd_mlink_name(const struct mlink *mlink)
1987: {
1.130 schwarze 1988: uint64_t bits;
1.124 schwarze 1989: size_t i;
1990:
1991: dbadd_mlink(mlink);
1.115 schwarze 1992:
1993: i = 1;
1.130 schwarze 1994: SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
1995: bits = NAME_FILE & NAME_MASK;
1996: if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
1997: bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
1998: sqlite3_reset(stmts[STMT_SELECT_NAME]);
1999: }
2000:
2001: i = 1;
2002: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
1.115 schwarze 2003: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2004: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2005: SQL_STEP(stmts[STMT_INSERT_NAME]);
2006: sqlite3_reset(stmts[STMT_INSERT_NAME]);
1.75 schwarze 2007: }
2008:
1.11 schwarze 2009: /*
1.47 schwarze 2010: * Flush the current page's terms (and their bits) into the database.
2011: * Wrap the entire set of additions in a transaction to make sqlite be a
2012: * little faster.
1.53 schwarze 2013: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 2014: */
2015: static void
1.131 schwarze 2016: dbadd(struct mpage *mpage)
1.11 schwarze 2017: {
1.47 schwarze 2018: struct mlink *mlink;
2019: struct str *key;
1.131 schwarze 2020: char *cp;
1.47 schwarze 2021: size_t i;
2022: unsigned int slot;
1.131 schwarze 2023: int mustfree;
1.47 schwarze 2024:
1.85 schwarze 2025: mlink = mpage->mlinks;
1.11 schwarze 2026:
1.85 schwarze 2027: if (nodb) {
1.104 schwarze 2028: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2029: key = ohash_next(&names, &slot))
1.104 schwarze 2030: free(key);
2031: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2032: key = ohash_next(&strings, &slot))
1.104 schwarze 2033: free(key);
1.102 schwarze 2034: if (0 == debug)
2035: return;
1.85 schwarze 2036: while (NULL != mlink) {
2037: fputs(mlink->name, stdout);
2038: if (NULL == mlink->next ||
2039: strcmp(mlink->dsec, mlink->next->dsec) ||
2040: strcmp(mlink->fsec, mlink->next->fsec) ||
2041: strcmp(mlink->arch, mlink->next->arch)) {
2042: putchar('(');
2043: if ('\0' == *mlink->dsec)
2044: fputs(mlink->fsec, stdout);
2045: else
2046: fputs(mlink->dsec, stdout);
2047: if ('\0' != *mlink->arch)
2048: printf("/%s", mlink->arch);
2049: putchar(')');
2050: }
2051: mlink = mlink->next;
2052: if (NULL != mlink)
2053: fputs(", ", stdout);
2054: }
1.89 schwarze 2055: printf(" - %s\n", mpage->desc);
1.11 schwarze 2056: return;
1.85 schwarze 2057: }
2058:
2059: if (debug)
2060: say(mlink->file, "Adding to database");
1.47 schwarze 2061:
1.131 schwarze 2062: cp = mpage->desc;
2063: i = strlen(cp);
2064: mustfree = render_string(&cp, &i);
1.47 schwarze 2065: i = 1;
1.131 schwarze 2066: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
1.116 schwarze 2067: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
1.47 schwarze 2068: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.94 schwarze 2069: mpage->pageid = sqlite3_last_insert_rowid(db);
1.47 schwarze 2070: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1.131 schwarze 2071: if (mustfree)
2072: free(cp);
1.47 schwarze 2073:
1.85 schwarze 2074: while (NULL != mlink) {
1.75 schwarze 2075: dbadd_mlink(mlink);
1.85 schwarze 2076: mlink = mlink->next;
2077: }
1.91 schwarze 2078: mlink = mpage->mlinks;
1.47 schwarze 2079:
1.90 schwarze 2080: for (key = ohash_first(&names, &slot); NULL != key;
2081: key = ohash_next(&names, &slot)) {
2082: assert(key->mpage == mpage);
2083: i = 1;
2084: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
1.131 schwarze 2085: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
1.94 schwarze 2086: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
1.90 schwarze 2087: SQL_STEP(stmts[STMT_INSERT_NAME]);
2088: sqlite3_reset(stmts[STMT_INSERT_NAME]);
2089: free(key);
2090: }
1.47 schwarze 2091: for (key = ohash_first(&strings, &slot); NULL != key;
2092: key = ohash_next(&strings, &slot)) {
2093: assert(key->mpage == mpage);
2094: i = 1;
2095: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.131 schwarze 2096: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
1.94 schwarze 2097: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
1.47 schwarze 2098: SQL_STEP(stmts[STMT_INSERT_KEY]);
2099: sqlite3_reset(stmts[STMT_INSERT_KEY]);
2100: free(key);
1.33 schwarze 2101: }
1.47 schwarze 2102: }
1.41 deraadt 2103:
1.47 schwarze 2104: static void
2105: dbprune(void)
2106: {
2107: struct mpage *mpage;
2108: struct mlink *mlink;
2109: size_t i;
2110: unsigned int slot;
1.11 schwarze 2111:
1.63 schwarze 2112: if (0 == nodb)
2113: SQL_EXEC("BEGIN TRANSACTION");
1.47 schwarze 2114:
1.63 schwarze 2115: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2116: mpage = ohash_next(&mpages, &slot)) {
1.47 schwarze 2117: mlink = mpage->mlinks;
1.82 schwarze 2118: if (debug)
1.63 schwarze 2119: say(mlink->file, "Deleting from database");
2120: if (nodb)
2121: continue;
2122: for ( ; NULL != mlink; mlink = mlink->next) {
2123: i = 1;
2124: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2125: i, mlink->dsec);
2126: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2127: i, mlink->arch);
2128: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2129: i, mlink->name);
2130: SQL_STEP(stmts[STMT_DELETE_PAGE]);
2131: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2132: }
1.11 schwarze 2133: }
1.63 schwarze 2134:
2135: if (0 == nodb)
2136: SQL_EXEC("END TRANSACTION");
1.47 schwarze 2137: }
1.22 schwarze 2138:
1.47 schwarze 2139: /*
2140: * Close an existing database and its prepared statements.
2141: * If "real" is not set, rename the temporary file into the real one.
2142: */
2143: static void
2144: dbclose(int real)
2145: {
2146: size_t i;
1.72 schwarze 2147: int status;
2148: pid_t child;
1.11 schwarze 2149:
1.47 schwarze 2150: if (nodb)
2151: return;
1.11 schwarze 2152:
1.47 schwarze 2153: for (i = 0; i < STMT__MAX; i++) {
2154: sqlite3_finalize(stmts[i]);
2155: stmts[i] = NULL;
1.28 schwarze 2156: }
1.22 schwarze 2157:
1.47 schwarze 2158: sqlite3_close(db);
2159: db = NULL;
1.11 schwarze 2160:
1.47 schwarze 2161: if (real)
2162: return;
1.22 schwarze 2163:
1.72 schwarze 2164: if ('\0' == *tempfilename) {
2165: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2166: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2167: say(MANDOC_DB, "&rename");
1.72 schwarze 2168: }
2169: return;
2170: }
2171:
2172: switch (child = fork()) {
1.98 schwarze 2173: case -1:
1.72 schwarze 2174: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2175: say("", "&fork cmp");
1.72 schwarze 2176: return;
1.98 schwarze 2177: case 0:
1.72 schwarze 2178: execlp("cmp", "cmp", "-s",
1.149 schwarze 2179: tempfilename, MANDOC_DB, (char *)NULL);
1.80 schwarze 2180: say("", "&exec cmp");
1.72 schwarze 2181: exit(0);
2182: default:
2183: break;
2184: }
2185: if (-1 == waitpid(child, &status, 0)) {
2186: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2187: say("", "&wait cmp");
1.72 schwarze 2188: } else if (WIFSIGNALED(status)) {
2189: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2190: say("", "cmp died from signal %d", WTERMSIG(status));
1.72 schwarze 2191: } else if (WEXITSTATUS(status)) {
1.47 schwarze 2192: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2193: say(MANDOC_DB,
2194: "Data changed, but cannot replace database");
2195: }
2196:
2197: *strrchr(tempfilename, '/') = '\0';
2198: switch (child = fork()) {
1.98 schwarze 2199: case -1:
1.72 schwarze 2200: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2201: say("", "&fork rm");
1.72 schwarze 2202: return;
1.98 schwarze 2203: case 0:
1.149 schwarze 2204: execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
1.80 schwarze 2205: say("", "&exec rm");
1.72 schwarze 2206: exit((int)MANDOCLEVEL_SYSERR);
2207: default:
2208: break;
2209: }
2210: if (-1 == waitpid(child, &status, 0)) {
2211: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2212: say("", "&wait rm");
1.72 schwarze 2213: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2214: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2215: say("", "%s: Cannot remove temporary directory",
2216: tempfilename);
1.22 schwarze 2217: }
1.11 schwarze 2218: }
2219:
1.47 schwarze 2220: /*
2221: * This is straightforward stuff.
2222: * Open a database connection to a "temporary" database, then open a set
2223: * of prepared statements we'll use over and over again.
2224: * If "real" is set, we use the existing database; if not, we truncate a
2225: * temporary one.
2226: * Must be matched by dbclose().
2227: */
2228: static int
2229: dbopen(int real)
1.2 schwarze 2230: {
1.72 schwarze 2231: const char *sql;
1.47 schwarze 2232: int rc, ofl;
1.6 schwarze 2233:
1.98 schwarze 2234: if (nodb)
1.150 schwarze 2235: return 1;
1.6 schwarze 2236:
1.72 schwarze 2237: *tempfilename = '\0';
1.47 schwarze 2238: ofl = SQLITE_OPEN_READWRITE;
1.72 schwarze 2239:
2240: if (real) {
2241: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2242: if (SQLITE_OK != rc) {
1.47 schwarze 2243: exitcode = (int)MANDOCLEVEL_SYSERR;
1.108 sthen 2244: if (SQLITE_CANTOPEN != rc)
2245: say(MANDOC_DB, "%s", sqlite3_errstr(rc));
1.150 schwarze 2246: return 0;
1.28 schwarze 2247: }
1.72 schwarze 2248: goto prepare_statements;
2249: }
2250:
2251: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.6 schwarze 2252:
1.72 schwarze 2253: remove(MANDOC_DB "~");
2254: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.98 schwarze 2255: if (SQLITE_OK == rc)
1.72 schwarze 2256: goto create_tables;
1.73 schwarze 2257: if (MPARSE_QUICK & mparse_options) {
1.47 schwarze 2258: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2259: say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
1.150 schwarze 2260: return 0;
1.47 schwarze 2261: }
1.6 schwarze 2262:
1.100 schwarze 2263: (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2264: sizeof(tempfilename));
1.72 schwarze 2265: if (NULL == mkdtemp(tempfilename)) {
2266: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2267: say("", "&%s", tempfilename);
1.150 schwarze 2268: return 0;
1.72 schwarze 2269: }
1.100 schwarze 2270: (void)strlcat(tempfilename, "/" MANDOC_DB,
2271: sizeof(tempfilename));
1.72 schwarze 2272: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2273: if (SQLITE_OK != rc) {
1.47 schwarze 2274: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2275: say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
1.150 schwarze 2276: return 0;
1.2 schwarze 2277: }
2278:
1.72 schwarze 2279: create_tables:
1.47 schwarze 2280: sql = "CREATE TABLE \"mpages\" (\n"
1.89 schwarze 2281: " \"desc\" TEXT NOT NULL,\n"
1.47 schwarze 2282: " \"form\" INTEGER NOT NULL,\n"
1.94 schwarze 2283: " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1.47 schwarze 2284: ");\n"
2285: "\n"
2286: "CREATE TABLE \"mlinks\" (\n"
2287: " \"sec\" TEXT NOT NULL,\n"
2288: " \"arch\" TEXT NOT NULL,\n"
2289: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2290: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2291: "ON DELETE CASCADE\n"
1.47 schwarze 2292: ");\n"
1.93 schwarze 2293: "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
1.47 schwarze 2294: "\n"
1.90 schwarze 2295: "CREATE TABLE \"names\" (\n"
2296: " \"bits\" INTEGER NOT NULL,\n"
2297: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2298: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.130 schwarze 2299: "ON DELETE CASCADE,\n"
2300: " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
1.90 schwarze 2301: ");\n"
2302: "\n"
1.47 schwarze 2303: "CREATE TABLE \"keys\" (\n"
2304: " \"bits\" INTEGER NOT NULL,\n"
2305: " \"key\" TEXT NOT NULL,\n"
1.94 schwarze 2306: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2307: "ON DELETE CASCADE\n"
1.93 schwarze 2308: ");\n"
2309: "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
1.47 schwarze 2310:
2311: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2312: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2313: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.103 schwarze 2314: sqlite3_close(db);
1.150 schwarze 2315: return 0;
1.2 schwarze 2316: }
2317:
1.47 schwarze 2318: prepare_statements:
1.103 schwarze 2319: if (SQLITE_OK != sqlite3_exec(db,
2320: "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2321: exitcode = (int)MANDOCLEVEL_SYSERR;
2322: say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2323: sqlite3_errmsg(db));
2324: sqlite3_close(db);
1.150 schwarze 2325: return 0;
1.103 schwarze 2326: }
2327:
1.94 schwarze 2328: sql = "DELETE FROM mpages WHERE pageid IN "
1.63 schwarze 2329: "(SELECT pageid FROM mlinks WHERE "
2330: "sec=? AND arch=? AND name=?)";
1.47 schwarze 2331: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2332: sql = "INSERT INTO mpages "
1.89 schwarze 2333: "(desc,form) VALUES (?,?)";
1.47 schwarze 2334: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2335: sql = "INSERT INTO mlinks "
1.61 schwarze 2336: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.47 schwarze 2337: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.130 schwarze 2338: sql = "SELECT bits FROM names where pageid = ?";
2339: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
1.90 schwarze 2340: sql = "INSERT INTO names "
2341: "(bits,name,pageid) VALUES (?,?,?)";
2342: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
1.47 schwarze 2343: sql = "INSERT INTO keys "
2344: "(bits,key,pageid) VALUES (?,?,?)";
2345: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.6 schwarze 2346:
1.47 schwarze 2347: /*
2348: * When opening a new database, we can turn off
2349: * synchronous mode for much better performance.
2350: */
1.6 schwarze 2351:
1.103 schwarze 2352: if (real && SQLITE_OK != sqlite3_exec(db,
2353: "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2354: exitcode = (int)MANDOCLEVEL_SYSERR;
2355: say(MANDOC_DB, "PRAGMA synchronous: %s",
2356: sqlite3_errmsg(db));
2357: sqlite3_close(db);
1.150 schwarze 2358: return 0;
1.103 schwarze 2359: }
1.11 schwarze 2360:
1.150 schwarze 2361: return 1;
1.47 schwarze 2362: }
1.6 schwarze 2363:
1.47 schwarze 2364: static int
1.120 schwarze 2365: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2366: {
2367: static char startdir[PATH_MAX];
1.110 schwarze 2368: static int getcwd_status; /* 1 = ok, 2 = failure */
2369: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2370: char *cp;
1.6 schwarze 2371:
1.47 schwarze 2372: /*
1.110 schwarze 2373: * Remember the original working directory, if possible.
2374: * This will be needed if the second or a later directory
2375: * on the command line is given as a relative path.
2376: * Do not error out if the current directory is not
2377: * searchable: Maybe it won't be needed after all.
2378: */
2379: if (0 == getcwd_status) {
2380: if (NULL == getcwd(startdir, sizeof(startdir))) {
2381: getcwd_status = 2;
2382: (void)strlcpy(startdir, strerror(errno),
2383: sizeof(startdir));
2384: } else
2385: getcwd_status = 1;
2386: }
2387:
2388: /*
2389: * We are leaving the old base directory.
2390: * Do not use it any longer, not even for messages.
2391: */
2392: *basedir = '\0';
2393:
2394: /*
2395: * If and only if the directory was changed earlier and
2396: * the next directory to process is given as a relative path,
2397: * first go back, or bail out if that is impossible.
1.47 schwarze 2398: */
1.110 schwarze 2399: if (chdir_status && '/' != *targetdir) {
2400: if (2 == getcwd_status) {
1.47 schwarze 2401: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2402: say("", "getcwd: %s", startdir);
1.150 schwarze 2403: return 0;
1.47 schwarze 2404: }
1.110 schwarze 2405: if (-1 == chdir(startdir)) {
1.47 schwarze 2406: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2407: say("", "&chdir %s", startdir);
1.150 schwarze 2408: return 0;
1.2 schwarze 2409: }
2410: }
1.110 schwarze 2411:
2412: /*
2413: * Always resolve basedir to the canonicalized absolute
2414: * pathname and append a trailing slash, such that
2415: * we can reliably check whether files are inside.
2416: */
1.47 schwarze 2417: if (NULL == realpath(targetdir, basedir)) {
1.120 schwarze 2418: if (report_baddir || errno != ENOENT) {
2419: exitcode = (int)MANDOCLEVEL_BADARG;
2420: say("", "&%s: realpath", targetdir);
2421: }
1.150 schwarze 2422: return 0;
1.47 schwarze 2423: } else if (-1 == chdir(basedir)) {
1.120 schwarze 2424: if (report_baddir || errno != ENOENT) {
2425: exitcode = (int)MANDOCLEVEL_BADARG;
2426: say("", "&chdir");
2427: }
1.150 schwarze 2428: return 0;
1.109 schwarze 2429: }
1.110 schwarze 2430: chdir_status = 1;
1.109 schwarze 2431: cp = strchr(basedir, '\0');
2432: if ('/' != cp[-1]) {
2433: if (cp - basedir >= PATH_MAX - 1) {
2434: exitcode = (int)MANDOCLEVEL_SYSERR;
2435: say("", "Filename too long");
1.150 schwarze 2436: return 0;
1.109 schwarze 2437: }
2438: *cp++ = '/';
2439: *cp = '\0';
1.47 schwarze 2440: }
1.150 schwarze 2441: return 1;
1.2 schwarze 2442: }
2443:
2444: static void
1.47 schwarze 2445: say(const char *file, const char *format, ...)
1.2 schwarze 2446: {
1.47 schwarze 2447: va_list ap;
1.80 schwarze 2448: int use_errno;
1.2 schwarze 2449:
1.47 schwarze 2450: if ('\0' != *basedir)
2451: fprintf(stderr, "%s", basedir);
2452: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2453: fputc('/', stderr);
1.47 schwarze 2454: if ('\0' != *file)
2455: fprintf(stderr, "%s", file);
1.31 schwarze 2456:
1.80 schwarze 2457: use_errno = 1;
2458: if (NULL != format) {
2459: switch (*format) {
1.98 schwarze 2460: case '&':
1.80 schwarze 2461: format++;
2462: break;
1.98 schwarze 2463: case '\0':
1.80 schwarze 2464: format = NULL;
2465: break;
2466: default:
2467: use_errno = 0;
2468: break;
2469: }
2470: }
2471: if (NULL != format) {
2472: if ('\0' != *basedir || '\0' != *file)
2473: fputs(": ", stderr);
2474: va_start(ap, format);
2475: vfprintf(stderr, format, ap);
2476: va_end(ap);
2477: }
2478: if (use_errno) {
2479: if ('\0' != *basedir || '\0' != *file || NULL != format)
2480: fputs(": ", stderr);
1.47 schwarze 2481: perror(NULL);
1.80 schwarze 2482: } else
2483: fputc('\n', stderr);
1.1 schwarze 2484: }