Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.162
1.162 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.161 2015/11/06 16:27:13 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;
1.162 ! schwarze 1272: size_t linesz, plen, titlesz;
! 1273: ssize_t len;
! 1274: int offs;
1.1 schwarze 1275:
1.81 schwarze 1276: stream = (-1 == fd) ?
1277: fopen(mpage->mlinks->file, "r") :
1278: fdopen(fd, "r");
1279: if (NULL == stream) {
1.114 doug 1280: if (-1 != fd)
1281: close(fd);
1.47 schwarze 1282: if (warnings)
1.80 schwarze 1283: say(mpage->mlinks->file, "&fopen");
1.47 schwarze 1284: return;
1285: }
1.1 schwarze 1286:
1.162 ! schwarze 1287: line = NULL;
! 1288: linesz = 0;
! 1289:
1.47 schwarze 1290: /* Skip to first blank line. */
1.1 schwarze 1291:
1.162 ! schwarze 1292: while (getline(&line, &linesz, stream) != -1)
! 1293: if (*line == '\n')
1.47 schwarze 1294: break;
1.1 schwarze 1295:
1.47 schwarze 1296: /*
1297: * Assume the first line that is not indented
1298: * is the first section header. Skip to it.
1299: */
1.1 schwarze 1300:
1.162 ! schwarze 1301: while (getline(&line, &linesz, stream) != -1)
! 1302: if (*line != '\n' && *line != ' ')
1.47 schwarze 1303: break;
1.98 schwarze 1304:
1.47 schwarze 1305: /*
1306: * Read up until the next section into a buffer.
1307: * Strip the leading and trailing newline from each read line,
1308: * appending a trailing space.
1309: * Ignore empty (whitespace-only) lines.
1310: */
1.28 schwarze 1311:
1.47 schwarze 1312: titlesz = 0;
1313: title = NULL;
1.38 schwarze 1314:
1.162 ! schwarze 1315: while ((len = getline(&line, &linesz, stream)) != -1) {
! 1316: if (*line != ' ')
1.47 schwarze 1317: break;
1.162 ! schwarze 1318: offs = 0;
! 1319: while (isspace((unsigned char)line[offs]))
! 1320: offs++;
! 1321: if (line[offs] == '\0')
1.47 schwarze 1322: continue;
1.162 ! schwarze 1323: title = mandoc_realloc(title, titlesz + len - offs);
! 1324: memcpy(title + titlesz, line + offs, len - offs);
! 1325: titlesz += len - offs;
1.47 schwarze 1326: title[titlesz - 1] = ' ';
1327: }
1.162 ! schwarze 1328: free(line);
1.28 schwarze 1329:
1.47 schwarze 1330: /*
1331: * If no page content can be found, or the input line
1332: * is already the next section header, or there is no
1333: * trailing newline, reuse the page title as the page
1334: * description.
1335: */
1.1 schwarze 1336:
1.47 schwarze 1337: if (NULL == title || '\0' == *title) {
1338: if (warnings)
1339: say(mpage->mlinks->file,
1340: "Cannot find NAME section");
1341: fclose(stream);
1342: free(title);
1343: return;
1344: }
1.24 schwarze 1345:
1.162 ! schwarze 1346: title[titlesz - 1] = '\0';
1.24 schwarze 1347:
1.47 schwarze 1348: /*
1349: * Skip to the first dash.
1350: * Use the remaining line as the description (no more than 70
1351: * bytes).
1352: */
1.28 schwarze 1353:
1.47 schwarze 1354: if (NULL != (p = strstr(title, "- "))) {
1355: for (p += 2; ' ' == *p || '\b' == *p; p++)
1356: /* Skip to next word. */ ;
1357: } else {
1358: if (warnings)
1359: say(mpage->mlinks->file,
1360: "No dash in title line");
1361: p = title;
1362: }
1.1 schwarze 1363:
1.47 schwarze 1364: plen = strlen(p);
1.1 schwarze 1365:
1.47 schwarze 1366: /* Strip backspace-encoding from line. */
1.1 schwarze 1367:
1.47 schwarze 1368: while (NULL != (line = memchr(p, '\b', plen))) {
1369: len = line - p;
1370: if (0 == len) {
1371: memmove(line, line + 1, plen--);
1372: continue;
1.98 schwarze 1373: }
1.47 schwarze 1374: memmove(line - 1, line + 1, plen - len);
1375: plen -= 2;
1376: }
1.1 schwarze 1377:
1.47 schwarze 1378: mpage->desc = mandoc_strdup(p);
1379: fclose(stream);
1380: free(title);
1381: }
1.16 schwarze 1382:
1.47 schwarze 1383: /*
1384: * Put a type/word pair into the word database for this particular file.
1385: */
1386: static void
1.69 schwarze 1387: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1388: {
1.69 schwarze 1389: char *cp;
1.37 schwarze 1390:
1.47 schwarze 1391: assert(NULL != value);
1.69 schwarze 1392: if (TYPE_arch == type)
1393: for (cp = value; *cp; cp++)
1394: if (isupper((unsigned char)*cp))
1395: *cp = _tolower((unsigned char)*cp);
1.47 schwarze 1396: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1397: }
1398:
1399: /*
1.47 schwarze 1400: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1401: */
1402: static void
1.47 schwarze 1403: putmdockey(const struct mpage *mpage,
1.142 schwarze 1404: const struct roff_node *n, uint64_t m)
1.2 schwarze 1405: {
1.16 schwarze 1406:
1.47 schwarze 1407: for ( ; NULL != n; n = n->next) {
1408: if (NULL != n->child)
1409: putmdockey(mpage, n->child, m);
1.141 schwarze 1410: if (n->type == ROFFT_TEXT)
1.47 schwarze 1411: putkey(mpage, n->string, m);
1412: }
1413: }
1.16 schwarze 1414:
1.47 schwarze 1415: static void
1.143 schwarze 1416: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1417: const struct roff_node *n)
1.47 schwarze 1418: {
1.142 schwarze 1419: const struct roff_node *head, *body;
1.78 schwarze 1420: char *start, *title;
1.47 schwarze 1421: char byte;
1.78 schwarze 1422: size_t sz;
1.16 schwarze 1423:
1.47 schwarze 1424: if (NULL == n)
1425: return;
1.16 schwarze 1426:
1.47 schwarze 1427: /*
1428: * We're only searching for one thing: the first text child in
1429: * the BODY of a NAME section. Since we don't keep track of
1430: * sections in -man, run some hoops to find out whether we're in
1431: * the correct section or not.
1432: */
1.16 schwarze 1433:
1.141 schwarze 1434: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1435: body = n;
1436: assert(body->parent);
1437: if (NULL != (head = body->parent->head) &&
1.98 schwarze 1438: 1 == head->nchild &&
1439: NULL != (head = (head->child)) &&
1.141 schwarze 1440: head->type == ROFFT_TEXT &&
1.98 schwarze 1441: 0 == strcmp(head->string, "NAME") &&
1442: NULL != body->child) {
1.2 schwarze 1443:
1.47 schwarze 1444: /*
1445: * Suck the entire NAME section into memory.
1446: * Yes, we might run away.
1447: * But too many manuals have big, spread-out
1448: * NAME sections over many lines.
1449: */
1.2 schwarze 1450:
1.78 schwarze 1451: title = NULL;
1.147 schwarze 1452: deroff(&title, body);
1.47 schwarze 1453: if (NULL == title)
1454: return;
1.16 schwarze 1455:
1.98 schwarze 1456: /*
1.47 schwarze 1457: * Go through a special heuristic dance here.
1458: * Conventionally, one or more manual names are
1459: * comma-specified prior to a whitespace, then a
1460: * dash, then a description. Try to puzzle out
1461: * the name parts here.
1462: */
1.16 schwarze 1463:
1.78 schwarze 1464: start = title;
1.47 schwarze 1465: for ( ;; ) {
1466: sz = strcspn(start, " ,");
1467: if ('\0' == start[sz])
1468: break;
1.1 schwarze 1469:
1.47 schwarze 1470: byte = start[sz];
1471: start[sz] = '\0';
1.67 schwarze 1472:
1473: /*
1474: * Assume a stray trailing comma in the
1475: * name list if a name begins with a dash.
1476: */
1477:
1478: if ('-' == start[0] ||
1479: ('\\' == start[0] && '-' == start[1]))
1480: break;
1.1 schwarze 1481:
1.90 schwarze 1482: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1483: if ( ! (mpage->name_head_done ||
1484: strcasecmp(start, meta->title))) {
1485: putkey(mpage, start, NAME_HEAD);
1486: mpage->name_head_done = 1;
1487: }
1.1 schwarze 1488:
1.47 schwarze 1489: if (' ' == byte) {
1490: start += sz + 1;
1491: break;
1492: }
1.1 schwarze 1493:
1.47 schwarze 1494: assert(',' == byte);
1495: start += sz + 1;
1496: while (' ' == *start)
1497: start++;
1498: }
1.1 schwarze 1499:
1.78 schwarze 1500: if (start == title) {
1.90 schwarze 1501: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1502: if ( ! (mpage->name_head_done ||
1503: strcasecmp(start, meta->title))) {
1504: putkey(mpage, start, NAME_HEAD);
1505: mpage->name_head_done = 1;
1506: }
1.47 schwarze 1507: free(title);
1508: return;
1509: }
1.1 schwarze 1510:
1.47 schwarze 1511: while (isspace((unsigned char)*start))
1512: start++;
1.1 schwarze 1513:
1.47 schwarze 1514: if (0 == strncmp(start, "-", 1))
1515: start += 1;
1516: else if (0 == strncmp(start, "\\-\\-", 4))
1517: start += 4;
1518: else if (0 == strncmp(start, "\\-", 2))
1519: start += 2;
1520: else if (0 == strncmp(start, "\\(en", 4))
1521: start += 4;
1522: else if (0 == strncmp(start, "\\(em", 4))
1523: start += 4;
1.1 schwarze 1524:
1.47 schwarze 1525: while (' ' == *start)
1526: start++;
1.1 schwarze 1527:
1.47 schwarze 1528: mpage->desc = mandoc_strdup(start);
1529: free(title);
1530: return;
1531: }
1532: }
1.1 schwarze 1533:
1.47 schwarze 1534: for (n = n->child; n; n = n->next) {
1535: if (NULL != mpage->desc)
1536: break;
1.129 schwarze 1537: parse_man(mpage, meta, n);
1.1 schwarze 1538: }
1539: }
1540:
1541: static void
1.143 schwarze 1542: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1543: const struct roff_node *n)
1.1 schwarze 1544: {
1545:
1.47 schwarze 1546: assert(NULL != n);
1547: for (n = n->child; NULL != n; n = n->next) {
1548: switch (n->type) {
1.141 schwarze 1549: case ROFFT_ELEM:
1550: case ROFFT_BLOCK:
1551: case ROFFT_HEAD:
1552: case ROFFT_BODY:
1553: case ROFFT_TAIL:
1.47 schwarze 1554: if (NULL != mdocs[n->tok].fp)
1.127 schwarze 1555: if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1.47 schwarze 1556: break;
1557: if (mdocs[n->tok].mask)
1558: putmdockey(mpage, n->child,
1559: mdocs[n->tok].mask);
1560: break;
1561: default:
1.141 schwarze 1562: assert(n->type != ROFFT_ROOT);
1.47 schwarze 1563: continue;
1564: }
1565: if (NULL != n->child)
1.127 schwarze 1566: parse_mdoc(mpage, meta, n);
1.1 schwarze 1567: }
1568: }
1569:
1.19 schwarze 1570: static int
1.143 schwarze 1571: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1572: const struct roff_node *n)
1.1 schwarze 1573: {
1.131 schwarze 1574: char *start, *end;
1.1 schwarze 1575: size_t sz;
1.19 schwarze 1576:
1.47 schwarze 1577: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1578: NULL == (n = n->child) ||
1.141 schwarze 1579: n->type != ROFFT_TEXT)
1.150 schwarze 1580: return 0;
1.1 schwarze 1581:
1582: /*
1583: * Only consider those `Fd' macro fields that begin with an
1584: * "inclusion" token (versus, e.g., #define).
1585: */
1.47 schwarze 1586:
1.1 schwarze 1587: if (strcmp("#include", n->string))
1.150 schwarze 1588: return 0;
1.1 schwarze 1589:
1.141 schwarze 1590: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1591: return 0;
1.1 schwarze 1592:
1593: /*
1594: * Strip away the enclosing angle brackets and make sure we're
1595: * not zero-length.
1596: */
1597:
1598: start = n->string;
1599: if ('<' == *start || '"' == *start)
1600: start++;
1601:
1602: if (0 == (sz = strlen(start)))
1.150 schwarze 1603: return 0;
1.1 schwarze 1604:
1605: end = &start[(int)sz - 1];
1606: if ('>' == *end || '"' == *end)
1607: end--;
1608:
1.47 schwarze 1609: if (end > start)
1610: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1611: return 0;
1.1 schwarze 1612: }
1613:
1.133 schwarze 1614: static void
1.142 schwarze 1615: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1616: {
1.69 schwarze 1617: char *cp;
1.133 schwarze 1618: size_t sz;
1.1 schwarze 1619:
1.141 schwarze 1620: if (n->type != ROFFT_TEXT)
1.133 schwarze 1621: return;
1.19 schwarze 1622:
1.133 schwarze 1623: /* Skip function pointer punctuation. */
1.1 schwarze 1624:
1.133 schwarze 1625: cp = n->string;
1626: while (*cp == '(' || *cp == '*')
1.1 schwarze 1627: cp++;
1.133 schwarze 1628: sz = strcspn(cp, "()");
1.1 schwarze 1629:
1.133 schwarze 1630: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1631: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1632: putkeys(mpage, cp, sz, NAME_SYN);
1633: }
1.19 schwarze 1634:
1.133 schwarze 1635: static int
1.143 schwarze 1636: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1637: const struct roff_node *n)
1.133 schwarze 1638: {
1639:
1640: if (n->child == NULL)
1.150 schwarze 1641: return 0;
1.133 schwarze 1642:
1643: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1644:
1.133 schwarze 1645: for (n = n->child->next; n != NULL; n = n->next)
1.141 schwarze 1646: if (n->type == ROFFT_TEXT)
1.47 schwarze 1647: putkey(mpage, n->string, TYPE_Fa);
1.19 schwarze 1648:
1.150 schwarze 1649: return 0;
1.128 schwarze 1650: }
1651:
1652: static int
1.143 schwarze 1653: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1654: const struct roff_node *n)
1.128 schwarze 1655: {
1.132 schwarze 1656:
1.141 schwarze 1657: if (n->type != ROFFT_HEAD)
1.150 schwarze 1658: return 1;
1.128 schwarze 1659:
1.133 schwarze 1660: if (n->child != NULL)
1661: parse_mdoc_fname(mpage, n->child);
1662:
1.150 schwarze 1663: return 0;
1.1 schwarze 1664: }
1665:
1.19 schwarze 1666: static int
1.143 schwarze 1667: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1668: const struct roff_node *n)
1.1 schwarze 1669: {
1.47 schwarze 1670: char *cp;
1.1 schwarze 1671:
1672: if (NULL == (n = n->child))
1.150 schwarze 1673: return 0;
1.1 schwarze 1674:
1.47 schwarze 1675: if (NULL == n->next) {
1676: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1677: return 0;
1.47 schwarze 1678: }
1.1 schwarze 1679:
1.77 schwarze 1680: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1681: putkey(mpage, cp, TYPE_Xr);
1682: free(cp);
1.150 schwarze 1683: return 0;
1.1 schwarze 1684: }
1685:
1.19 schwarze 1686: static int
1.143 schwarze 1687: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1688: const struct roff_node *n)
1.1 schwarze 1689: {
1690:
1.141 schwarze 1691: if (n->type == ROFFT_BODY)
1.147 schwarze 1692: deroff(&mpage->desc, n);
1.150 schwarze 1693: return 0;
1.1 schwarze 1694: }
1695:
1.19 schwarze 1696: static int
1.143 schwarze 1697: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1698: const struct roff_node *n)
1.1 schwarze 1699: {
1700:
1.86 schwarze 1701: if (SEC_NAME == n->sec)
1.90 schwarze 1702: putmdockey(mpage, n->child, NAME_TITLE);
1.141 schwarze 1703: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1704: if (n->child == NULL)
1705: putkey(mpage, meta->name, NAME_SYN);
1706: else
1707: putmdockey(mpage, n->child, NAME_SYN);
1.129 schwarze 1708: }
1709: if ( ! (mpage->name_head_done ||
1710: n->child == NULL || n->child->string == NULL ||
1711: strcasecmp(n->child->string, meta->title))) {
1.141 schwarze 1712: putkey(mpage, n->child->string, ROFFT_HEAD);
1.129 schwarze 1713: mpage->name_head_done = 1;
1.127 schwarze 1714: }
1.150 schwarze 1715: return 0;
1.1 schwarze 1716: }
1717:
1.19 schwarze 1718: static int
1.143 schwarze 1719: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1720: const struct roff_node *n)
1.1 schwarze 1721: {
1722:
1.150 schwarze 1723: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1724: }
1725:
1.47 schwarze 1726: static int
1.143 schwarze 1727: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1728: const struct roff_node *n)
1.1 schwarze 1729: {
1730:
1.150 schwarze 1731: return n->type == ROFFT_HEAD;
1.47 schwarze 1732: }
1.1 schwarze 1733:
1.47 schwarze 1734: static int
1.143 schwarze 1735: parse_mdoc_body(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1736: const struct roff_node *n)
1.47 schwarze 1737: {
1.1 schwarze 1738:
1.150 schwarze 1739: return n->type == ROFFT_BODY;
1.1 schwarze 1740: }
1741:
1.47 schwarze 1742: /*
1743: * Add a string to the hash table for the current manual.
1744: * Each string has a bitmask telling which macros it belongs to.
1745: * When we finish the manual, we'll dump the table.
1746: */
1.1 schwarze 1747: static void
1.131 schwarze 1748: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1749: {
1.90 schwarze 1750: struct ohash *htab;
1.47 schwarze 1751: struct str *s;
1.68 schwarze 1752: const char *end;
1.47 schwarze 1753: unsigned int slot;
1.131 schwarze 1754: int i, mustfree;
1.1 schwarze 1755:
1.47 schwarze 1756: if (0 == sz)
1757: return;
1.68 schwarze 1758:
1.131 schwarze 1759: mustfree = render_string(&cp, &sz);
1760:
1.90 schwarze 1761: if (TYPE_Nm & v) {
1762: htab = &names;
1763: v &= name_mask;
1.124 schwarze 1764: if (v & NAME_FIRST)
1765: name_mask &= ~NAME_FIRST;
1.90 schwarze 1766: if (debug > 1)
1767: say(mpage->mlinks->file,
1.130 schwarze 1768: "Adding name %*s, bits=%d", sz, cp, v);
1.90 schwarze 1769: } else {
1770: htab = &strings;
1771: if (debug > 1)
1772: for (i = 0; i < mansearch_keymax; i++)
1.118 schwarze 1773: if ((uint64_t)1 << i & v)
1.90 schwarze 1774: say(mpage->mlinks->file,
1775: "Adding key %s=%*s",
1776: mansearch_keynames[i], sz, cp);
1.68 schwarze 1777: }
1.47 schwarze 1778:
1779: end = cp + sz;
1.90 schwarze 1780: slot = ohash_qlookupi(htab, cp, &end);
1781: s = ohash_find(htab, slot);
1.1 schwarze 1782:
1.47 schwarze 1783: if (NULL != s && mpage == s->mpage) {
1784: s->mask |= v;
1.1 schwarze 1785: return;
1.47 schwarze 1786: } else if (NULL == s) {
1.101 schwarze 1787: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1788: memcpy(s->key, cp, sz);
1.90 schwarze 1789: ohash_insert(htab, slot, s);
1.47 schwarze 1790: }
1791: s->mpage = mpage;
1792: s->mask = v;
1.131 schwarze 1793:
1794: if (mustfree)
1795: free(cp);
1.1 schwarze 1796: }
1797:
1798: /*
1.47 schwarze 1799: * Take a Unicode codepoint and produce its UTF-8 encoding.
1800: * This isn't the best way to do this, but it works.
1801: * The magic numbers are from the UTF-8 packaging.
1802: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1803: */
1.47 schwarze 1804: static size_t
1805: utf8(unsigned int cp, char out[7])
1.1 schwarze 1806: {
1.47 schwarze 1807: size_t rc;
1.1 schwarze 1808:
1.47 schwarze 1809: rc = 0;
1810: if (cp <= 0x0000007F) {
1811: rc = 1;
1812: out[0] = (char)cp;
1813: } else if (cp <= 0x000007FF) {
1814: rc = 2;
1815: out[0] = (cp >> 6 & 31) | 192;
1816: out[1] = (cp & 63) | 128;
1817: } else if (cp <= 0x0000FFFF) {
1818: rc = 3;
1819: out[0] = (cp >> 12 & 15) | 224;
1820: out[1] = (cp >> 6 & 63) | 128;
1821: out[2] = (cp & 63) | 128;
1822: } else if (cp <= 0x001FFFFF) {
1823: rc = 4;
1824: out[0] = (cp >> 18 & 7) | 240;
1825: out[1] = (cp >> 12 & 63) | 128;
1826: out[2] = (cp >> 6 & 63) | 128;
1827: out[3] = (cp & 63) | 128;
1828: } else if (cp <= 0x03FFFFFF) {
1829: rc = 5;
1830: out[0] = (cp >> 24 & 3) | 248;
1831: out[1] = (cp >> 18 & 63) | 128;
1832: out[2] = (cp >> 12 & 63) | 128;
1833: out[3] = (cp >> 6 & 63) | 128;
1834: out[4] = (cp & 63) | 128;
1835: } else if (cp <= 0x7FFFFFFF) {
1836: rc = 6;
1837: out[0] = (cp >> 30 & 1) | 252;
1838: out[1] = (cp >> 24 & 63) | 128;
1839: out[2] = (cp >> 18 & 63) | 128;
1840: out[3] = (cp >> 12 & 63) | 128;
1841: out[4] = (cp >> 6 & 63) | 128;
1842: out[5] = (cp & 63) | 128;
1843: } else
1.150 schwarze 1844: return 0;
1.19 schwarze 1845:
1.47 schwarze 1846: out[rc] = '\0';
1.150 schwarze 1847: return rc;
1.1 schwarze 1848: }
1849:
1.47 schwarze 1850: /*
1.131 schwarze 1851: * If the string contains escape sequences,
1852: * replace it with an allocated rendering and return 1,
1853: * such that the caller can free it after use.
1854: * Otherwise, do nothing and return 0.
1.47 schwarze 1855: */
1.131 schwarze 1856: static int
1857: render_string(char **public, size_t *psz)
1.1 schwarze 1858: {
1.131 schwarze 1859: const char *src, *scp, *addcp, *seq;
1860: char *dst;
1861: size_t ssz, dsz, addsz;
1.71 schwarze 1862: char utfbuf[7], res[6];
1.131 schwarze 1863: int seqlen, unicode;
1.47 schwarze 1864:
1865: res[0] = '\\';
1866: res[1] = '\t';
1867: res[2] = ASCII_NBRSP;
1868: res[3] = ASCII_HYPH;
1.71 schwarze 1869: res[4] = ASCII_BREAK;
1870: res[5] = '\0';
1.1 schwarze 1871:
1.131 schwarze 1872: src = scp = *public;
1873: ssz = *psz;
1874: dst = NULL;
1875: dsz = 0;
1876:
1877: while (scp < src + *psz) {
1878:
1879: /* Leave normal characters unchanged. */
1880:
1881: if (strchr(res, *scp) == NULL) {
1882: if (dst != NULL)
1883: dst[dsz++] = *scp;
1884: scp++;
1885: continue;
1886: }
1.1 schwarze 1887:
1.47 schwarze 1888: /*
1.131 schwarze 1889: * Found something that requires replacing,
1890: * make sure we have a destination buffer.
1.47 schwarze 1891: */
1.131 schwarze 1892:
1893: if (dst == NULL) {
1894: dst = mandoc_malloc(ssz + 1);
1895: dsz = scp - src;
1896: memcpy(dst, src, dsz);
1.47 schwarze 1897: }
1.39 schwarze 1898:
1.131 schwarze 1899: /* Handle single-char special characters. */
1900:
1901: switch (*scp) {
1902: case '\\':
1903: break;
1.98 schwarze 1904: case '\t':
1905: case ASCII_NBRSP:
1.131 schwarze 1906: dst[dsz++] = ' ';
1907: scp++;
1908: continue;
1909: case ASCII_HYPH:
1910: dst[dsz++] = '-';
1.71 schwarze 1911: /* FALLTHROUGH */
1.98 schwarze 1912: case ASCII_BREAK:
1.131 schwarze 1913: scp++;
1.47 schwarze 1914: continue;
1.71 schwarze 1915: default:
1.131 schwarze 1916: abort();
1.71 schwarze 1917: }
1.39 schwarze 1918:
1.47 schwarze 1919: /*
1.131 schwarze 1920: * Found an escape sequence.
1921: * Read past the slash, then parse it.
1922: * Ignore everything except characters.
1.47 schwarze 1923: */
1.52 schwarze 1924:
1.131 schwarze 1925: scp++;
1926: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1927: continue;
1.39 schwarze 1928:
1.47 schwarze 1929: /*
1.52 schwarze 1930: * Render the special character
1931: * as either UTF-8 or ASCII.
1.47 schwarze 1932: */
1.52 schwarze 1933:
1934: if (write_utf8) {
1.157 schwarze 1935: unicode = mchars_spec2cp(seq, seqlen);
1.131 schwarze 1936: if (unicode <= 0)
1.52 schwarze 1937: continue;
1.131 schwarze 1938: addsz = utf8(unicode, utfbuf);
1939: if (addsz == 0)
1.52 schwarze 1940: continue;
1.131 schwarze 1941: addcp = utfbuf;
1.52 schwarze 1942: } else {
1.157 schwarze 1943: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.131 schwarze 1944: if (addcp == NULL)
1.52 schwarze 1945: continue;
1.131 schwarze 1946: if (*addcp == ASCII_NBRSP) {
1947: addcp = " ";
1948: addsz = 1;
1.52 schwarze 1949: }
1950: }
1.1 schwarze 1951:
1.47 schwarze 1952: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1953:
1.131 schwarze 1954: ssz += addsz;
1955: dst = mandoc_realloc(dst, ssz + 1);
1956: memcpy(dst + dsz, addcp, addsz);
1957: dsz += addsz;
1958: }
1959: if (dst != NULL) {
1960: *public = dst;
1961: *psz = dsz;
1962: }
1963:
1964: /* Trim trailing whitespace and NUL-terminate. */
1965:
1966: while (*psz > 0 && (*public)[*psz - 1] == ' ')
1967: --*psz;
1968: if (dst != NULL) {
1969: (*public)[*psz] = '\0';
1.150 schwarze 1970: return 1;
1.131 schwarze 1971: } else
1.150 schwarze 1972: return 0;
1.1 schwarze 1973: }
1974:
1.75 schwarze 1975: static void
1976: dbadd_mlink(const struct mlink *mlink)
1977: {
1978: size_t i;
1979:
1980: i = 1;
1981: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1982: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1983: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.94 schwarze 1984: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1.75 schwarze 1985: SQL_STEP(stmts[STMT_INSERT_LINK]);
1986: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1.124 schwarze 1987: }
1988:
1989: static void
1990: dbadd_mlink_name(const struct mlink *mlink)
1991: {
1.130 schwarze 1992: uint64_t bits;
1.124 schwarze 1993: size_t i;
1994:
1995: dbadd_mlink(mlink);
1.115 schwarze 1996:
1997: i = 1;
1.130 schwarze 1998: SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
1999: bits = NAME_FILE & NAME_MASK;
2000: if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
2001: bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
2002: sqlite3_reset(stmts[STMT_SELECT_NAME]);
2003: }
2004:
2005: i = 1;
2006: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
1.115 schwarze 2007: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2008: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2009: SQL_STEP(stmts[STMT_INSERT_NAME]);
2010: sqlite3_reset(stmts[STMT_INSERT_NAME]);
1.75 schwarze 2011: }
2012:
1.11 schwarze 2013: /*
1.47 schwarze 2014: * Flush the current page's terms (and their bits) into the database.
2015: * Wrap the entire set of additions in a transaction to make sqlite be a
2016: * little faster.
1.53 schwarze 2017: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 2018: */
2019: static void
1.131 schwarze 2020: dbadd(struct mpage *mpage)
1.11 schwarze 2021: {
1.47 schwarze 2022: struct mlink *mlink;
2023: struct str *key;
1.131 schwarze 2024: char *cp;
1.47 schwarze 2025: size_t i;
2026: unsigned int slot;
1.131 schwarze 2027: int mustfree;
1.47 schwarze 2028:
1.85 schwarze 2029: mlink = mpage->mlinks;
1.11 schwarze 2030:
1.85 schwarze 2031: if (nodb) {
1.104 schwarze 2032: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2033: key = ohash_next(&names, &slot))
1.104 schwarze 2034: free(key);
2035: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2036: key = ohash_next(&strings, &slot))
1.104 schwarze 2037: free(key);
1.102 schwarze 2038: if (0 == debug)
2039: return;
1.85 schwarze 2040: while (NULL != mlink) {
2041: fputs(mlink->name, stdout);
2042: if (NULL == mlink->next ||
2043: strcmp(mlink->dsec, mlink->next->dsec) ||
2044: strcmp(mlink->fsec, mlink->next->fsec) ||
2045: strcmp(mlink->arch, mlink->next->arch)) {
2046: putchar('(');
2047: if ('\0' == *mlink->dsec)
2048: fputs(mlink->fsec, stdout);
2049: else
2050: fputs(mlink->dsec, stdout);
2051: if ('\0' != *mlink->arch)
2052: printf("/%s", mlink->arch);
2053: putchar(')');
2054: }
2055: mlink = mlink->next;
2056: if (NULL != mlink)
2057: fputs(", ", stdout);
2058: }
1.89 schwarze 2059: printf(" - %s\n", mpage->desc);
1.11 schwarze 2060: return;
1.85 schwarze 2061: }
2062:
2063: if (debug)
2064: say(mlink->file, "Adding to database");
1.47 schwarze 2065:
1.131 schwarze 2066: cp = mpage->desc;
2067: i = strlen(cp);
2068: mustfree = render_string(&cp, &i);
1.47 schwarze 2069: i = 1;
1.131 schwarze 2070: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
1.116 schwarze 2071: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
1.47 schwarze 2072: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.94 schwarze 2073: mpage->pageid = sqlite3_last_insert_rowid(db);
1.47 schwarze 2074: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1.131 schwarze 2075: if (mustfree)
2076: free(cp);
1.47 schwarze 2077:
1.85 schwarze 2078: while (NULL != mlink) {
1.75 schwarze 2079: dbadd_mlink(mlink);
1.85 schwarze 2080: mlink = mlink->next;
2081: }
1.91 schwarze 2082: mlink = mpage->mlinks;
1.47 schwarze 2083:
1.90 schwarze 2084: for (key = ohash_first(&names, &slot); NULL != key;
2085: key = ohash_next(&names, &slot)) {
2086: assert(key->mpage == mpage);
2087: i = 1;
2088: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
1.131 schwarze 2089: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
1.94 schwarze 2090: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
1.90 schwarze 2091: SQL_STEP(stmts[STMT_INSERT_NAME]);
2092: sqlite3_reset(stmts[STMT_INSERT_NAME]);
2093: free(key);
2094: }
1.47 schwarze 2095: for (key = ohash_first(&strings, &slot); NULL != key;
2096: key = ohash_next(&strings, &slot)) {
2097: assert(key->mpage == mpage);
2098: i = 1;
2099: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.131 schwarze 2100: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
1.94 schwarze 2101: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
1.47 schwarze 2102: SQL_STEP(stmts[STMT_INSERT_KEY]);
2103: sqlite3_reset(stmts[STMT_INSERT_KEY]);
2104: free(key);
1.33 schwarze 2105: }
1.47 schwarze 2106: }
1.41 deraadt 2107:
1.47 schwarze 2108: static void
2109: dbprune(void)
2110: {
2111: struct mpage *mpage;
2112: struct mlink *mlink;
2113: size_t i;
2114: unsigned int slot;
1.11 schwarze 2115:
1.63 schwarze 2116: if (0 == nodb)
2117: SQL_EXEC("BEGIN TRANSACTION");
1.47 schwarze 2118:
1.63 schwarze 2119: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2120: mpage = ohash_next(&mpages, &slot)) {
1.47 schwarze 2121: mlink = mpage->mlinks;
1.82 schwarze 2122: if (debug)
1.63 schwarze 2123: say(mlink->file, "Deleting from database");
2124: if (nodb)
2125: continue;
2126: for ( ; NULL != mlink; mlink = mlink->next) {
2127: i = 1;
2128: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2129: i, mlink->dsec);
2130: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2131: i, mlink->arch);
2132: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2133: i, mlink->name);
2134: SQL_STEP(stmts[STMT_DELETE_PAGE]);
2135: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2136: }
1.11 schwarze 2137: }
1.63 schwarze 2138:
2139: if (0 == nodb)
2140: SQL_EXEC("END TRANSACTION");
1.47 schwarze 2141: }
1.22 schwarze 2142:
1.47 schwarze 2143: /*
2144: * Close an existing database and its prepared statements.
2145: * If "real" is not set, rename the temporary file into the real one.
2146: */
2147: static void
2148: dbclose(int real)
2149: {
2150: size_t i;
1.72 schwarze 2151: int status;
2152: pid_t child;
1.11 schwarze 2153:
1.47 schwarze 2154: if (nodb)
2155: return;
1.11 schwarze 2156:
1.47 schwarze 2157: for (i = 0; i < STMT__MAX; i++) {
2158: sqlite3_finalize(stmts[i]);
2159: stmts[i] = NULL;
1.28 schwarze 2160: }
1.22 schwarze 2161:
1.47 schwarze 2162: sqlite3_close(db);
2163: db = NULL;
1.11 schwarze 2164:
1.47 schwarze 2165: if (real)
2166: return;
1.22 schwarze 2167:
1.72 schwarze 2168: if ('\0' == *tempfilename) {
2169: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2170: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2171: say(MANDOC_DB, "&rename");
1.72 schwarze 2172: }
2173: return;
2174: }
2175:
2176: switch (child = fork()) {
1.98 schwarze 2177: case -1:
1.72 schwarze 2178: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2179: say("", "&fork cmp");
1.72 schwarze 2180: return;
1.98 schwarze 2181: case 0:
1.72 schwarze 2182: execlp("cmp", "cmp", "-s",
1.149 schwarze 2183: tempfilename, MANDOC_DB, (char *)NULL);
1.80 schwarze 2184: say("", "&exec cmp");
1.72 schwarze 2185: exit(0);
2186: default:
2187: break;
2188: }
2189: if (-1 == waitpid(child, &status, 0)) {
2190: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2191: say("", "&wait cmp");
1.72 schwarze 2192: } else if (WIFSIGNALED(status)) {
2193: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2194: say("", "cmp died from signal %d", WTERMSIG(status));
1.72 schwarze 2195: } else if (WEXITSTATUS(status)) {
1.47 schwarze 2196: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2197: say(MANDOC_DB,
2198: "Data changed, but cannot replace database");
2199: }
2200:
2201: *strrchr(tempfilename, '/') = '\0';
2202: switch (child = fork()) {
1.98 schwarze 2203: case -1:
1.72 schwarze 2204: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2205: say("", "&fork rm");
1.72 schwarze 2206: return;
1.98 schwarze 2207: case 0:
1.149 schwarze 2208: execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
1.80 schwarze 2209: say("", "&exec rm");
1.72 schwarze 2210: exit((int)MANDOCLEVEL_SYSERR);
2211: default:
2212: break;
2213: }
2214: if (-1 == waitpid(child, &status, 0)) {
2215: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2216: say("", "&wait rm");
1.72 schwarze 2217: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2218: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2219: say("", "%s: Cannot remove temporary directory",
2220: tempfilename);
1.22 schwarze 2221: }
1.11 schwarze 2222: }
2223:
1.47 schwarze 2224: /*
2225: * This is straightforward stuff.
2226: * Open a database connection to a "temporary" database, then open a set
2227: * of prepared statements we'll use over and over again.
2228: * If "real" is set, we use the existing database; if not, we truncate a
2229: * temporary one.
2230: * Must be matched by dbclose().
2231: */
2232: static int
2233: dbopen(int real)
1.2 schwarze 2234: {
1.72 schwarze 2235: const char *sql;
1.47 schwarze 2236: int rc, ofl;
1.6 schwarze 2237:
1.98 schwarze 2238: if (nodb)
1.150 schwarze 2239: return 1;
1.6 schwarze 2240:
1.72 schwarze 2241: *tempfilename = '\0';
1.47 schwarze 2242: ofl = SQLITE_OPEN_READWRITE;
1.72 schwarze 2243:
2244: if (real) {
2245: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2246: if (SQLITE_OK != rc) {
1.47 schwarze 2247: exitcode = (int)MANDOCLEVEL_SYSERR;
1.108 sthen 2248: if (SQLITE_CANTOPEN != rc)
2249: say(MANDOC_DB, "%s", sqlite3_errstr(rc));
1.150 schwarze 2250: return 0;
1.28 schwarze 2251: }
1.72 schwarze 2252: goto prepare_statements;
2253: }
2254:
2255: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.6 schwarze 2256:
1.72 schwarze 2257: remove(MANDOC_DB "~");
2258: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.98 schwarze 2259: if (SQLITE_OK == rc)
1.72 schwarze 2260: goto create_tables;
1.73 schwarze 2261: if (MPARSE_QUICK & mparse_options) {
1.47 schwarze 2262: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2263: say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
1.150 schwarze 2264: return 0;
1.47 schwarze 2265: }
1.6 schwarze 2266:
1.100 schwarze 2267: (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2268: sizeof(tempfilename));
1.72 schwarze 2269: if (NULL == mkdtemp(tempfilename)) {
2270: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2271: say("", "&%s", tempfilename);
1.150 schwarze 2272: return 0;
1.72 schwarze 2273: }
1.100 schwarze 2274: (void)strlcat(tempfilename, "/" MANDOC_DB,
2275: sizeof(tempfilename));
1.72 schwarze 2276: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2277: if (SQLITE_OK != rc) {
1.47 schwarze 2278: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2279: say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
1.150 schwarze 2280: return 0;
1.2 schwarze 2281: }
2282:
1.72 schwarze 2283: create_tables:
1.47 schwarze 2284: sql = "CREATE TABLE \"mpages\" (\n"
1.89 schwarze 2285: " \"desc\" TEXT NOT NULL,\n"
1.47 schwarze 2286: " \"form\" INTEGER NOT NULL,\n"
1.94 schwarze 2287: " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1.47 schwarze 2288: ");\n"
2289: "\n"
2290: "CREATE TABLE \"mlinks\" (\n"
2291: " \"sec\" TEXT NOT NULL,\n"
2292: " \"arch\" TEXT NOT NULL,\n"
2293: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2294: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2295: "ON DELETE CASCADE\n"
1.47 schwarze 2296: ");\n"
1.93 schwarze 2297: "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
1.47 schwarze 2298: "\n"
1.90 schwarze 2299: "CREATE TABLE \"names\" (\n"
2300: " \"bits\" INTEGER NOT NULL,\n"
2301: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2302: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.130 schwarze 2303: "ON DELETE CASCADE,\n"
2304: " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
1.90 schwarze 2305: ");\n"
2306: "\n"
1.47 schwarze 2307: "CREATE TABLE \"keys\" (\n"
2308: " \"bits\" INTEGER NOT NULL,\n"
2309: " \"key\" TEXT NOT NULL,\n"
1.94 schwarze 2310: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2311: "ON DELETE CASCADE\n"
1.93 schwarze 2312: ");\n"
2313: "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
1.47 schwarze 2314:
2315: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2316: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2317: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.103 schwarze 2318: sqlite3_close(db);
1.150 schwarze 2319: return 0;
1.2 schwarze 2320: }
2321:
1.47 schwarze 2322: prepare_statements:
1.103 schwarze 2323: if (SQLITE_OK != sqlite3_exec(db,
2324: "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2325: exitcode = (int)MANDOCLEVEL_SYSERR;
2326: say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2327: sqlite3_errmsg(db));
2328: sqlite3_close(db);
1.150 schwarze 2329: return 0;
1.103 schwarze 2330: }
2331:
1.94 schwarze 2332: sql = "DELETE FROM mpages WHERE pageid IN "
1.63 schwarze 2333: "(SELECT pageid FROM mlinks WHERE "
2334: "sec=? AND arch=? AND name=?)";
1.47 schwarze 2335: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2336: sql = "INSERT INTO mpages "
1.89 schwarze 2337: "(desc,form) VALUES (?,?)";
1.47 schwarze 2338: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2339: sql = "INSERT INTO mlinks "
1.61 schwarze 2340: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.47 schwarze 2341: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.130 schwarze 2342: sql = "SELECT bits FROM names where pageid = ?";
2343: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
1.90 schwarze 2344: sql = "INSERT INTO names "
2345: "(bits,name,pageid) VALUES (?,?,?)";
2346: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
1.47 schwarze 2347: sql = "INSERT INTO keys "
2348: "(bits,key,pageid) VALUES (?,?,?)";
2349: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.6 schwarze 2350:
1.47 schwarze 2351: /*
2352: * When opening a new database, we can turn off
2353: * synchronous mode for much better performance.
2354: */
1.6 schwarze 2355:
1.103 schwarze 2356: if (real && SQLITE_OK != sqlite3_exec(db,
2357: "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2358: exitcode = (int)MANDOCLEVEL_SYSERR;
2359: say(MANDOC_DB, "PRAGMA synchronous: %s",
2360: sqlite3_errmsg(db));
2361: sqlite3_close(db);
1.150 schwarze 2362: return 0;
1.103 schwarze 2363: }
1.11 schwarze 2364:
1.150 schwarze 2365: return 1;
1.47 schwarze 2366: }
1.6 schwarze 2367:
1.47 schwarze 2368: static int
1.120 schwarze 2369: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2370: {
2371: static char startdir[PATH_MAX];
1.110 schwarze 2372: static int getcwd_status; /* 1 = ok, 2 = failure */
2373: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2374: char *cp;
1.6 schwarze 2375:
1.47 schwarze 2376: /*
1.110 schwarze 2377: * Remember the original working directory, if possible.
2378: * This will be needed if the second or a later directory
2379: * on the command line is given as a relative path.
2380: * Do not error out if the current directory is not
2381: * searchable: Maybe it won't be needed after all.
2382: */
2383: if (0 == getcwd_status) {
2384: if (NULL == getcwd(startdir, sizeof(startdir))) {
2385: getcwd_status = 2;
2386: (void)strlcpy(startdir, strerror(errno),
2387: sizeof(startdir));
2388: } else
2389: getcwd_status = 1;
2390: }
2391:
2392: /*
2393: * We are leaving the old base directory.
2394: * Do not use it any longer, not even for messages.
2395: */
2396: *basedir = '\0';
2397:
2398: /*
2399: * If and only if the directory was changed earlier and
2400: * the next directory to process is given as a relative path,
2401: * first go back, or bail out if that is impossible.
1.47 schwarze 2402: */
1.110 schwarze 2403: if (chdir_status && '/' != *targetdir) {
2404: if (2 == getcwd_status) {
1.47 schwarze 2405: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2406: say("", "getcwd: %s", startdir);
1.150 schwarze 2407: return 0;
1.47 schwarze 2408: }
1.110 schwarze 2409: if (-1 == chdir(startdir)) {
1.47 schwarze 2410: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2411: say("", "&chdir %s", startdir);
1.150 schwarze 2412: return 0;
1.2 schwarze 2413: }
2414: }
1.110 schwarze 2415:
2416: /*
2417: * Always resolve basedir to the canonicalized absolute
2418: * pathname and append a trailing slash, such that
2419: * we can reliably check whether files are inside.
2420: */
1.47 schwarze 2421: if (NULL == realpath(targetdir, basedir)) {
1.120 schwarze 2422: if (report_baddir || errno != ENOENT) {
2423: exitcode = (int)MANDOCLEVEL_BADARG;
2424: say("", "&%s: realpath", targetdir);
2425: }
1.150 schwarze 2426: return 0;
1.47 schwarze 2427: } else if (-1 == chdir(basedir)) {
1.120 schwarze 2428: if (report_baddir || errno != ENOENT) {
2429: exitcode = (int)MANDOCLEVEL_BADARG;
2430: say("", "&chdir");
2431: }
1.150 schwarze 2432: return 0;
1.109 schwarze 2433: }
1.110 schwarze 2434: chdir_status = 1;
1.109 schwarze 2435: cp = strchr(basedir, '\0');
2436: if ('/' != cp[-1]) {
2437: if (cp - basedir >= PATH_MAX - 1) {
2438: exitcode = (int)MANDOCLEVEL_SYSERR;
2439: say("", "Filename too long");
1.150 schwarze 2440: return 0;
1.109 schwarze 2441: }
2442: *cp++ = '/';
2443: *cp = '\0';
1.47 schwarze 2444: }
1.150 schwarze 2445: return 1;
1.2 schwarze 2446: }
2447:
2448: static void
1.47 schwarze 2449: say(const char *file, const char *format, ...)
1.2 schwarze 2450: {
1.47 schwarze 2451: va_list ap;
1.80 schwarze 2452: int use_errno;
1.2 schwarze 2453:
1.47 schwarze 2454: if ('\0' != *basedir)
2455: fprintf(stderr, "%s", basedir);
2456: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2457: fputc('/', stderr);
1.47 schwarze 2458: if ('\0' != *file)
2459: fprintf(stderr, "%s", file);
1.31 schwarze 2460:
1.80 schwarze 2461: use_errno = 1;
2462: if (NULL != format) {
2463: switch (*format) {
1.98 schwarze 2464: case '&':
1.80 schwarze 2465: format++;
2466: break;
1.98 schwarze 2467: case '\0':
1.80 schwarze 2468: format = NULL;
2469: break;
2470: default:
2471: use_errno = 0;
2472: break;
2473: }
2474: }
2475: if (NULL != format) {
2476: if ('\0' != *basedir || '\0' != *file)
2477: fputs(": ", stderr);
2478: va_start(ap, format);
2479: vfprintf(stderr, format, ap);
2480: va_end(ap);
2481: }
2482: if (use_errno) {
2483: if ('\0' != *basedir || '\0' != *file || NULL != format)
2484: fputs(": ", stderr);
1.47 schwarze 2485: perror(NULL);
1.80 schwarze 2486: } else
2487: fputc('\n', stderr);
1.1 schwarze 2488: }
![[BACK]](/icons/back.gif){kind=icon}
Return to mandocdb.c CVS log ![[TXT]](/icons/text.gif){kind=icon}
![[DIR]](/icons/dir.gif){kind=icon}
Up to [local] / src / usr.bin / mandoc