Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.163
1.163 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.162 2015/11/07 17:58:52 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:
1.163 ! schwarze 412: if (nodb) {
! 413: if (pledge("stdio rpath", NULL) == -1) {
! 414: perror("pledge");
! 415: return (int)MANDOCLEVEL_SYSERR;
! 416: }
1.155 schwarze 417: }
418:
1.28 schwarze 419: if (OP_CONFFILE == op && argc > 0) {
1.151 schwarze 420: warnx("-C: Too many arguments");
1.28 schwarze 421: goto usage;
422: }
423:
1.47 schwarze 424: exitcode = (int)MANDOCLEVEL_OK;
1.157 schwarze 425: mchars_alloc();
426: mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
1.156 schwarze 427: mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
428: mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
1.2 schwarze 429:
1.47 schwarze 430: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.2 schwarze 431:
1.47 schwarze 432: /*
1.109 schwarze 433: * Most of these deal with a specific directory.
1.97 schwarze 434: * Jump into that directory first.
1.47 schwarze 435: */
1.120 schwarze 436: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.47 schwarze 437: goto out;
1.97 schwarze 438:
1.95 schwarze 439: if (dbopen(1)) {
1.97 schwarze 440: /*
441: * The existing database is usable. Process
442: * all files specified on the command-line.
443: */
1.163 ! schwarze 444: if (!nodb) {
! 445: if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) {
! 446: perror("pledge");
! 447: exitcode = (int)MANDOCLEVEL_SYSERR;
! 448: goto out;
! 449: }
1.155 schwarze 450: }
1.97 schwarze 451: use_all = 1;
1.95 schwarze 452: for (i = 0; i < argc; i++)
453: filescan(argv[i]);
454: if (OP_TEST != op)
455: dbprune();
456: } else {
457: /*
458: * Database missing or corrupt.
459: * Recreate from scratch.
460: */
1.97 schwarze 461: exitcode = (int)MANDOCLEVEL_OK;
1.95 schwarze 462: op = OP_DEFAULT;
463: if (0 == treescan())
464: goto out;
465: if (0 == dbopen(0))
466: goto out;
467: }
1.47 schwarze 468: if (OP_DELETE != op)
1.131 schwarze 469: mpages_merge(mp);
1.95 schwarze 470: dbclose(OP_DEFAULT == op ? 0 : 1);
1.47 schwarze 471: } else {
472: /*
473: * If we have arguments, use them as our manpaths.
474: * If we don't, grok from manpath(1) or however else
1.140 schwarze 475: * manconf_parse() wants to do it.
1.47 schwarze 476: */
477: if (argc > 0) {
1.140 schwarze 478: conf.manpath.paths = mandoc_reallocarray(NULL,
1.101 schwarze 479: argc, sizeof(char *));
1.140 schwarze 480: conf.manpath.sz = (size_t)argc;
1.47 schwarze 481: for (i = 0; i < argc; i++)
1.140 schwarze 482: conf.manpath.paths[i] = mandoc_strdup(argv[i]);
1.47 schwarze 483: } else
1.140 schwarze 484: manconf_parse(&conf, path_arg, NULL, NULL);
1.84 schwarze 485:
1.140 schwarze 486: if (conf.manpath.sz == 0) {
1.84 schwarze 487: exitcode = (int)MANDOCLEVEL_BADARG;
488: say("", "Empty manpath");
489: }
1.2 schwarze 490:
1.47 schwarze 491: /*
492: * First scan the tree rooted at a base directory, then
493: * build a new database and finally move it into place.
494: * Ignore zero-length directories and strip trailing
495: * slashes.
496: */
1.140 schwarze 497: for (j = 0; j < conf.manpath.sz; j++) {
498: sz = strlen(conf.manpath.paths[j]);
499: if (sz && conf.manpath.paths[j][sz - 1] == '/')
500: conf.manpath.paths[j][--sz] = '\0';
1.47 schwarze 501: if (0 == sz)
502: continue;
1.2 schwarze 503:
1.47 schwarze 504: if (j) {
1.156 schwarze 505: mandoc_ohash_init(&mpages, 6,
506: offsetof(struct mpage, inodev));
507: mandoc_ohash_init(&mlinks, 6,
508: offsetof(struct mlink, file));
1.47 schwarze 509: }
1.1 schwarze 510:
1.140 schwarze 511: if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
1.120 schwarze 512: continue;
1.47 schwarze 513: if (0 == treescan())
1.120 schwarze 514: continue;
1.47 schwarze 515: if (0 == dbopen(0))
1.120 schwarze 516: continue;
1.41 deraadt 517:
1.131 schwarze 518: mpages_merge(mp);
1.104 schwarze 519: if (warnings && !nodb &&
1.87 schwarze 520: ! (MPARSE_QUICK & mparse_options))
521: names_check();
1.47 schwarze 522: dbclose(0);
1.28 schwarze 523:
1.140 schwarze 524: if (j + 1 < conf.manpath.sz) {
1.47 schwarze 525: mpages_free();
526: ohash_delete(&mpages);
527: ohash_delete(&mlinks);
528: }
1.2 schwarze 529: }
1.47 schwarze 530: }
531: out:
1.140 schwarze 532: manconf_free(&conf);
1.123 schwarze 533: mparse_free(mp);
1.157 schwarze 534: mchars_free();
1.47 schwarze 535: mpages_free();
536: ohash_delete(&mpages);
537: ohash_delete(&mlinks);
1.150 schwarze 538: return exitcode;
1.47 schwarze 539: usage:
1.161 schwarze 540: progname = getprogname();
1.83 schwarze 541: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
542: " %s [-aDnpQ] [-Tutf8] dir ...\n"
543: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
544: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 545: " %s [-Q] -t file ...\n",
1.161 schwarze 546: progname, progname, progname, progname, progname);
1.1 schwarze 547:
1.150 schwarze 548: return (int)MANDOCLEVEL_BADARG;
1.47 schwarze 549: }
1.1 schwarze 550:
1.47 schwarze 551: /*
552: * Scan a directory tree rooted at "basedir" for manpages.
553: * We use fts(), scanning directory parts along the way for clues to our
554: * section and architecture.
555: *
556: * If use_all has been specified, grok all files.
557: * If not, sanitise paths to the following:
558: *
1.98 schwarze 559: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 560: * or
561: * [./]cat<section>[/<arch>]/<name>.0
562: *
563: * TODO: accomodate for multi-language directories.
564: */
565: static int
566: treescan(void)
567: {
1.96 schwarze 568: char buf[PATH_MAX];
1.47 schwarze 569: FTS *f;
570: FTSENT *ff;
571: struct mlink *mlink;
1.81 schwarze 572: int dform, gzip;
1.51 schwarze 573: char *dsec, *arch, *fsec, *cp;
574: const char *path;
1.47 schwarze 575: const char *argv[2];
1.1 schwarze 576:
1.47 schwarze 577: argv[0] = ".";
578: argv[1] = (char *)NULL;
1.1 schwarze 579:
1.96 schwarze 580: f = fts_open((char * const *)argv,
581: FTS_PHYSICAL | FTS_NOCHDIR, NULL);
1.153 schwarze 582: if (f == NULL) {
1.47 schwarze 583: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 584: say("", "&fts_open");
1.150 schwarze 585: return 0;
1.47 schwarze 586: }
1.2 schwarze 587:
1.47 schwarze 588: dsec = arch = NULL;
589: dform = FORM_NONE;
1.2 schwarze 590:
1.153 schwarze 591: while ((ff = fts_read(f)) != NULL) {
1.47 schwarze 592: path = ff->fts_path + 2;
1.96 schwarze 593: switch (ff->fts_info) {
594:
595: /*
596: * Symbolic links require various sanity checks,
597: * then get handled just like regular files.
598: */
1.98 schwarze 599: case FTS_SL:
1.153 schwarze 600: if (realpath(path, buf) == NULL) {
1.96 schwarze 601: if (warnings)
602: say(path, "&realpath");
603: continue;
604: }
605: if (strstr(buf, basedir) != buf) {
606: if (warnings) say("",
607: "%s: outside base directory", buf);
608: continue;
609: }
610: /* Use logical inode to avoid mpages dupe. */
1.153 schwarze 611: if (stat(path, ff->fts_statp) == -1) {
1.96 schwarze 612: if (warnings)
613: say(path, "&stat");
614: continue;
615: }
616: /* FALLTHROUGH */
617:
1.15 schwarze 618: /*
1.47 schwarze 619: * If we're a regular file, add an mlink by using the
620: * stored directory data and handling the filename.
1.15 schwarze 621: */
1.98 schwarze 622: case FTS_F:
1.153 schwarze 623: if ( ! strcmp(path, MANDOC_DB))
1.47 schwarze 624: continue;
625: if ( ! use_all && ff->fts_level < 2) {
626: if (warnings)
627: say(path, "Extraneous file");
628: continue;
1.81 schwarze 629: }
630: gzip = 0;
631: fsec = NULL;
1.153 schwarze 632: while (fsec == NULL) {
1.81 schwarze 633: fsec = strrchr(ff->fts_name, '.');
1.153 schwarze 634: if (fsec == NULL || strcmp(fsec+1, "gz"))
1.81 schwarze 635: break;
636: gzip = 1;
637: *fsec = '\0';
638: fsec = NULL;
639: }
1.153 schwarze 640: if (fsec == NULL) {
1.47 schwarze 641: if ( ! use_all) {
642: if (warnings)
643: say(path,
644: "No filename suffix");
645: continue;
646: }
1.153 schwarze 647: } else if ( ! strcmp(++fsec, "html")) {
1.47 schwarze 648: if (warnings)
649: say(path, "Skip html");
650: continue;
1.153 schwarze 651: } else if ( ! strcmp(fsec, "ps")) {
1.47 schwarze 652: if (warnings)
653: say(path, "Skip ps");
654: continue;
1.153 schwarze 655: } else if ( ! strcmp(fsec, "pdf")) {
1.47 schwarze 656: if (warnings)
657: say(path, "Skip pdf");
658: continue;
659: } else if ( ! use_all &&
1.153 schwarze 660: ((dform == FORM_SRC &&
1.138 schwarze 661: strncmp(fsec, dsec, strlen(dsec))) ||
1.153 schwarze 662: (dform == FORM_CAT && strcmp(fsec, "0")))) {
1.47 schwarze 663: if (warnings)
664: say(path, "Wrong filename suffix");
665: continue;
666: } else
667: fsec[-1] = '\0';
1.51 schwarze 668:
1.47 schwarze 669: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.100 schwarze 670: if (strlcpy(mlink->file, path,
671: sizeof(mlink->file)) >=
672: sizeof(mlink->file)) {
673: say(path, "Filename too long");
674: free(mlink);
675: continue;
676: }
1.47 schwarze 677: mlink->dform = dform;
1.51 schwarze 678: mlink->dsec = dsec;
679: mlink->arch = arch;
680: mlink->name = ff->fts_name;
681: mlink->fsec = fsec;
1.81 schwarze 682: mlink->gzip = gzip;
1.47 schwarze 683: mlink_add(mlink, ff->fts_statp);
684: continue;
1.96 schwarze 685:
1.98 schwarze 686: case FTS_D:
687: case FTS_DP:
1.96 schwarze 688: break;
689:
690: default:
1.47 schwarze 691: if (warnings)
692: say(path, "Not a regular file");
693: continue;
694: }
695:
696: switch (ff->fts_level) {
1.98 schwarze 697: case 0:
1.47 schwarze 698: /* Ignore the root directory. */
699: break;
1.98 schwarze 700: case 1:
1.47 schwarze 701: /*
702: * This might contain manX/ or catX/.
703: * Try to infer this from the name.
704: * If we're not in use_all, enforce it.
705: */
706: cp = ff->fts_name;
1.153 schwarze 707: if (ff->fts_info == FTS_DP) {
708: dform = FORM_NONE;
709: dsec = NULL;
1.47 schwarze 710: break;
1.153 schwarze 711: }
1.15 schwarze 712:
1.153 schwarze 713: if ( ! strncmp(cp, "man", 3)) {
1.47 schwarze 714: dform = FORM_SRC;
715: dsec = cp + 3;
1.153 schwarze 716: } else if ( ! strncmp(cp, "cat", 3)) {
1.47 schwarze 717: dform = FORM_CAT;
718: dsec = cp + 3;
1.51 schwarze 719: } else {
720: dform = FORM_NONE;
721: dsec = NULL;
1.26 schwarze 722: }
1.47 schwarze 723:
1.153 schwarze 724: if (dsec != NULL || use_all)
1.47 schwarze 725: break;
726:
727: if (warnings)
728: say(path, "Unknown directory part");
729: fts_set(f, ff, FTS_SKIP);
730: break;
1.98 schwarze 731: case 2:
1.47 schwarze 732: /*
733: * Possibly our architecture.
734: * If we're descending, keep tabs on it.
735: */
1.153 schwarze 736: if (ff->fts_info != FTS_DP && dsec != NULL)
1.47 schwarze 737: arch = ff->fts_name;
1.51 schwarze 738: else
739: arch = NULL;
1.47 schwarze 740: break;
741: default:
1.153 schwarze 742: if (ff->fts_info == FTS_DP || use_all)
1.47 schwarze 743: break;
744: if (warnings)
745: say(path, "Extraneous directory part");
746: fts_set(f, ff, FTS_SKIP);
747: break;
1.14 schwarze 748: }
1.47 schwarze 749: }
750:
751: fts_close(f);
1.150 schwarze 752: return 1;
1.47 schwarze 753: }
1.1 schwarze 754:
1.47 schwarze 755: /*
756: * Add a file to the mlinks table.
757: * Do not verify that it's a "valid" looking manpage (we'll do that
758: * later).
759: *
760: * Try to infer the manual section, architecture, and page name from the
761: * path, assuming it looks like
762: *
1.98 schwarze 763: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 764: * or
765: * [./]cat<section>[/<arch>]/<name>.0
766: *
767: * See treescan() for the fts(3) version of this.
768: */
769: static void
770: filescan(const char *file)
771: {
772: char buf[PATH_MAX];
773: struct stat st;
774: struct mlink *mlink;
775: char *p, *start;
776:
777: assert(use_all);
778:
779: if (0 == strncmp(file, "./", 2))
780: file += 2;
781:
1.96 schwarze 782: /*
783: * We have to do lstat(2) before realpath(3) loses
784: * the information whether this is a symbolic link.
785: * We need to know that because for symbolic links,
786: * we want to use the orginal file name, while for
787: * regular files, we want to use the real path.
788: */
789: if (-1 == lstat(file, &st)) {
790: exitcode = (int)MANDOCLEVEL_BADARG;
791: say(file, "&lstat");
792: return;
793: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
794: exitcode = (int)MANDOCLEVEL_BADARG;
795: say(file, "Not a regular file");
796: return;
797: }
798:
799: /*
800: * We have to resolve the file name to the real path
801: * in any case for the base directory check.
802: */
1.47 schwarze 803: if (NULL == realpath(file, buf)) {
804: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 805: say(file, "&realpath");
1.47 schwarze 806: return;
1.63 schwarze 807: }
808:
1.109 schwarze 809: if (OP_TEST == op)
1.63 schwarze 810: start = buf;
1.109 schwarze 811: else if (strstr(buf, basedir) == buf)
812: start = buf + strlen(basedir);
1.63 schwarze 813: else {
1.47 schwarze 814: exitcode = (int)MANDOCLEVEL_BADARG;
815: say("", "%s: outside base directory", buf);
816: return;
1.63 schwarze 817: }
818:
1.96 schwarze 819: /*
820: * Now we are sure the file is inside our tree.
821: * If it is a symbolic link, ignore the real path
822: * and use the original name.
823: * This implies passing stuff like "cat1/../man1/foo.1"
824: * on the command line won't work. So don't do that.
825: * Note the stat(2) can still fail if the link target
826: * doesn't exist.
827: */
828: if (S_IFLNK & st.st_mode) {
829: if (-1 == stat(buf, &st)) {
830: exitcode = (int)MANDOCLEVEL_BADARG;
831: say(file, "&stat");
832: return;
833: }
1.100 schwarze 834: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
835: say(file, "Filename too long");
836: return;
837: }
1.109 schwarze 838: start = buf;
839: if (OP_TEST != op && strstr(buf, basedir) == buf)
840: start += strlen(basedir);
1.1 schwarze 841: }
1.63 schwarze 842:
1.47 schwarze 843: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 844: mlink->dform = FORM_NONE;
1.100 schwarze 845: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
846: sizeof(mlink->file)) {
847: say(start, "Filename too long");
1.134 schwarze 848: free(mlink);
1.100 schwarze 849: return;
850: }
1.1 schwarze 851:
1.10 schwarze 852: /*
1.47 schwarze 853: * First try to guess our directory structure.
854: * If we find a separator, try to look for man* or cat*.
855: * If we find one of these and what's underneath is a directory,
856: * assume it's an architecture.
1.10 schwarze 857: */
1.47 schwarze 858: if (NULL != (p = strchr(start, '/'))) {
859: *p++ = '\0';
860: if (0 == strncmp(start, "man", 3)) {
861: mlink->dform = FORM_SRC;
1.51 schwarze 862: mlink->dsec = start + 3;
1.47 schwarze 863: } else if (0 == strncmp(start, "cat", 3)) {
864: mlink->dform = FORM_CAT;
1.51 schwarze 865: mlink->dsec = start + 3;
1.47 schwarze 866: }
1.10 schwarze 867:
1.47 schwarze 868: start = p;
869: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
870: *p++ = '\0';
1.51 schwarze 871: mlink->arch = start;
1.47 schwarze 872: start = p;
1.41 deraadt 873: }
1.47 schwarze 874: }
1.7 schwarze 875:
1.47 schwarze 876: /*
877: * Now check the file suffix.
878: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
879: */
880: p = strrchr(start, '\0');
881: while (p-- > start && '/' != *p && '.' != *p)
882: /* Loop. */ ;
883:
884: if ('.' == *p) {
885: *p++ = '\0';
1.51 schwarze 886: mlink->fsec = p;
1.47 schwarze 887: }
1.41 deraadt 888:
1.47 schwarze 889: /*
890: * Now try to parse the name.
891: * Use the filename portion of the path.
892: */
893: mlink->name = start;
894: if (NULL != (p = strrchr(start, '/'))) {
895: mlink->name = p + 1;
896: *p = '\0';
897: }
898: mlink_add(mlink, &st);
899: }
1.2 schwarze 900:
1.47 schwarze 901: static void
902: mlink_add(struct mlink *mlink, const struct stat *st)
903: {
904: struct inodev inodev;
905: struct mpage *mpage;
906: unsigned int slot;
907:
908: assert(NULL != mlink->file);
909:
1.51 schwarze 910: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
911: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
912: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
913: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 914:
915: if ('0' == *mlink->fsec) {
916: free(mlink->fsec);
917: mlink->fsec = mandoc_strdup(mlink->dsec);
918: mlink->fform = FORM_CAT;
919: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
920: mlink->fform = FORM_SRC;
921: else
922: mlink->fform = FORM_NONE;
1.1 schwarze 923:
1.47 schwarze 924: slot = ohash_qlookup(&mlinks, mlink->file);
925: assert(NULL == ohash_find(&mlinks, slot));
926: ohash_insert(&mlinks, slot, mlink);
927:
1.139 schwarze 928: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.47 schwarze 929: inodev.st_ino = st->st_ino;
930: inodev.st_dev = st->st_dev;
931: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
932: sizeof(struct inodev), inodev.st_ino);
933: mpage = ohash_find(&mpages, slot);
934: if (NULL == mpage) {
935: mpage = mandoc_calloc(1, sizeof(struct mpage));
936: mpage->inodev.st_ino = inodev.st_ino;
937: mpage->inodev.st_dev = inodev.st_dev;
938: ohash_insert(&mpages, slot, mpage);
939: } else
940: mlink->next = mpage->mlinks;
941: mpage->mlinks = mlink;
1.75 schwarze 942: mlink->mpage = mpage;
1.47 schwarze 943: }
1.1 schwarze 944:
1.47 schwarze 945: static void
946: mlink_free(struct mlink *mlink)
947: {
1.1 schwarze 948:
1.47 schwarze 949: free(mlink->dsec);
950: free(mlink->arch);
951: free(mlink->name);
952: free(mlink->fsec);
953: free(mlink);
954: }
1.1 schwarze 955:
1.47 schwarze 956: static void
957: mpages_free(void)
958: {
959: struct mpage *mpage;
960: struct mlink *mlink;
961: unsigned int slot;
962:
963: mpage = ohash_first(&mpages, &slot);
964: while (NULL != mpage) {
965: while (NULL != (mlink = mpage->mlinks)) {
966: mpage->mlinks = mlink->next;
967: mlink_free(mlink);
968: }
969: free(mpage->sec);
970: free(mpage->arch);
971: free(mpage->title);
972: free(mpage->desc);
973: free(mpage);
974: mpage = ohash_next(&mpages, &slot);
975: }
976: }
1.1 schwarze 977:
1.47 schwarze 978: /*
979: * For each mlink to the mpage, check whether the path looks like
980: * it is formatted, and if it does, check whether a source manual
981: * exists by the same name, ignoring the suffix.
982: * If both conditions hold, drop the mlink.
983: */
984: static void
985: mlinks_undupe(struct mpage *mpage)
986: {
987: char buf[PATH_MAX];
988: struct mlink **prev;
989: struct mlink *mlink;
990: char *bufp;
991:
992: mpage->form = FORM_CAT;
993: prev = &mpage->mlinks;
994: while (NULL != (mlink = *prev)) {
995: if (FORM_CAT != mlink->dform) {
996: mpage->form = FORM_NONE;
997: goto nextlink;
998: }
1.100 schwarze 999: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 1000: bufp = strstr(buf, "cat");
1001: assert(NULL != bufp);
1002: memcpy(bufp, "man", 3);
1003: if (NULL != (bufp = strrchr(buf, '.')))
1004: *++bufp = '\0';
1.100 schwarze 1005: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 1006: if (NULL == ohash_find(&mlinks,
1.98 schwarze 1007: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 1008: goto nextlink;
1009: if (warnings)
1010: say(mlink->file, "Man source exists: %s", buf);
1011: if (use_all)
1012: goto nextlink;
1013: *prev = mlink->next;
1014: mlink_free(mlink);
1015: continue;
1016: nextlink:
1017: prev = &(*prev)->next;
1.1 schwarze 1018: }
1.47 schwarze 1019: }
1.1 schwarze 1020:
1.88 schwarze 1021: static void
1.50 schwarze 1022: mlink_check(struct mpage *mpage, struct mlink *mlink)
1023: {
1.88 schwarze 1024: struct str *str;
1025: unsigned int slot;
1.50 schwarze 1026:
1027: /*
1028: * Check whether the manual section given in a file
1029: * agrees with the directory where the file is located.
1030: * Some manuals have suffixes like (3p) on their
1031: * section number either inside the file or in the
1032: * directory name, some are linked into more than one
1033: * section, like encrypt(1) = makekey(8).
1034: */
1035:
1036: if (FORM_SRC == mpage->form &&
1.88 schwarze 1037: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1038: say(mlink->file, "Section \"%s\" manual in %s directory",
1039: mpage->sec, mlink->dsec);
1040:
1041: /*
1042: * Manual page directories exist for each kernel
1043: * architecture as returned by machine(1).
1044: * However, many manuals only depend on the
1045: * application architecture as returned by arch(1).
1046: * For example, some (2/ARM) manuals are shared
1047: * across the "armish" and "zaurus" kernel
1048: * architectures.
1049: * A few manuals are even shared across completely
1050: * different architectures, for example fdformat(1)
1051: * on amd64, i386, sparc, and sparc64.
1052: */
1053:
1.88 schwarze 1054: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1055: say(mlink->file, "Architecture \"%s\" manual in "
1056: "\"%s\" directory", mpage->arch, mlink->arch);
1057:
1.88 schwarze 1058: /*
1059: * XXX
1.90 schwarze 1060: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1061: */
1062:
1063: if (FORM_CAT == mpage->form)
1064: return;
1065:
1066: /*
1067: * Check whether this mlink
1068: * appears as a name in the NAME section.
1069: */
1.50 schwarze 1070:
1.90 schwarze 1071: slot = ohash_qlookup(&names, mlink->name);
1072: str = ohash_find(&names, slot);
1.88 schwarze 1073: assert(NULL != str);
1.90 schwarze 1074: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1075: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1076: }
1077:
1.47 schwarze 1078: /*
1079: * Run through the files in the global vector "mpages"
1080: * and add them to the database specified in "basedir".
1081: *
1082: * This handles the parsing scheme itself, using the cues of directory
1083: * and filename to determine whether the file is parsable or not.
1084: */
1085: static void
1.131 schwarze 1086: mpages_merge(struct mparse *mp)
1.47 schwarze 1087: {
1.70 schwarze 1088: char any[] = "any";
1.75 schwarze 1089: struct mpage *mpage, *mpage_dest;
1090: struct mlink *mlink, *mlink_dest;
1.144 schwarze 1091: struct roff_man *man;
1.75 schwarze 1092: char *sodest;
1.69 schwarze 1093: char *cp;
1.117 schwarze 1094: int fd;
1.58 schwarze 1095: unsigned int pslot;
1.47 schwarze 1096:
1.126 schwarze 1097: if ( ! nodb)
1.64 schwarze 1098: SQL_EXEC("BEGIN TRANSACTION");
1099:
1.47 schwarze 1100: mpage = ohash_first(&mpages, &pslot);
1.126 schwarze 1101: while (mpage != NULL) {
1.47 schwarze 1102: mlinks_undupe(mpage);
1.136 schwarze 1103: if ((mlink = mpage->mlinks) == NULL) {
1.47 schwarze 1104: mpage = ohash_next(&mpages, &pslot);
1105: continue;
1106: }
1.1 schwarze 1107:
1.90 schwarze 1108: name_mask = NAME_MASK;
1.156 schwarze 1109: mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1110: mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1.47 schwarze 1111: mparse_reset(mp);
1112: man = NULL;
1.81 schwarze 1113: sodest = NULL;
1114:
1.136 schwarze 1115: mparse_open(mp, &fd, mlink->file);
1.117 schwarze 1116: if (fd == -1) {
1.136 schwarze 1117: say(mlink->file, "&open");
1.117 schwarze 1118: goto nextpage;
1.81 schwarze 1119: }
1.11 schwarze 1120:
1121: /*
1.137 schwarze 1122: * Interpret the file as mdoc(7) or man(7) source
1123: * code, unless it is known to be formatted.
1.11 schwarze 1124: */
1.136 schwarze 1125: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.137 schwarze 1126: mparse_readfd(mp, fd, mlink->file);
1.145 schwarze 1127: mparse_result(mp, &man, &sodest);
1.47 schwarze 1128: }
1.11 schwarze 1129:
1.126 schwarze 1130: if (sodest != NULL) {
1.75 schwarze 1131: mlink_dest = ohash_find(&mlinks,
1132: ohash_qlookup(&mlinks, sodest));
1.126 schwarze 1133: if (mlink_dest == NULL) {
1134: mandoc_asprintf(&cp, "%s.gz", sodest);
1135: mlink_dest = ohash_find(&mlinks,
1136: ohash_qlookup(&mlinks, cp));
1137: free(cp);
1138: }
1139: if (mlink_dest != NULL) {
1.75 schwarze 1140:
1141: /* The .so target exists. */
1142:
1143: mpage_dest = mlink_dest->mpage;
1144: while (1) {
1145: mlink->mpage = mpage_dest;
1146:
1147: /*
1148: * If the target was already
1149: * processed, add the links
1150: * to the database now.
1151: * Otherwise, this will
1152: * happen when we come
1153: * to the target.
1154: */
1155:
1.94 schwarze 1156: if (mpage_dest->pageid)
1.124 schwarze 1157: dbadd_mlink_name(mlink);
1.75 schwarze 1158:
1.126 schwarze 1159: if (mlink->next == NULL)
1.75 schwarze 1160: break;
1161: mlink = mlink->next;
1162: }
1163:
1164: /* Move all links to the target. */
1165:
1166: mlink->next = mlink_dest->next;
1167: mlink_dest->next = mpage->mlinks;
1168: mpage->mlinks = NULL;
1169: }
1.81 schwarze 1170: goto nextpage;
1.145 schwarze 1171: } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1.159 schwarze 1172: mdoc_validate(man);
1.47 schwarze 1173: mpage->form = FORM_SRC;
1.146 schwarze 1174: mpage->sec = man->meta.msec;
1.112 schwarze 1175: mpage->sec = mandoc_strdup(
1.126 schwarze 1176: mpage->sec == NULL ? "" : mpage->sec);
1.146 schwarze 1177: mpage->arch = man->meta.arch;
1.47 schwarze 1178: mpage->arch = mandoc_strdup(
1.126 schwarze 1179: mpage->arch == NULL ? "" : mpage->arch);
1.146 schwarze 1180: mpage->title = mandoc_strdup(man->meta.title);
1.145 schwarze 1181: } else if (man != NULL && man->macroset == MACROSET_MAN) {
1.160 schwarze 1182: man_validate(man);
1.47 schwarze 1183: mpage->form = FORM_SRC;
1.146 schwarze 1184: mpage->sec = mandoc_strdup(man->meta.msec);
1.136 schwarze 1185: mpage->arch = mandoc_strdup(mlink->arch);
1.146 schwarze 1186: mpage->title = mandoc_strdup(man->meta.title);
1.11 schwarze 1187: } else {
1.47 schwarze 1188: mpage->form = FORM_CAT;
1.136 schwarze 1189: mpage->sec = mandoc_strdup(mlink->dsec);
1190: mpage->arch = mandoc_strdup(mlink->arch);
1191: mpage->title = mandoc_strdup(mlink->name);
1.1 schwarze 1192: }
1.54 schwarze 1193: putkey(mpage, mpage->sec, TYPE_sec);
1.119 schwarze 1194: if (*mpage->arch != '\0')
1195: putkey(mpage, mpage->arch, TYPE_arch);
1.1 schwarze 1196:
1.136 schwarze 1197: for ( ; mlink != NULL; mlink = mlink->next) {
1.54 schwarze 1198: if ('\0' != *mlink->dsec)
1199: putkey(mpage, mlink->dsec, TYPE_sec);
1200: if ('\0' != *mlink->fsec)
1201: putkey(mpage, mlink->fsec, TYPE_sec);
1.55 schwarze 1202: putkey(mpage, '\0' == *mlink->arch ?
1.70 schwarze 1203: any : mlink->arch, TYPE_arch);
1.90 schwarze 1204: putkey(mpage, mlink->name, NAME_FILE);
1.54 schwarze 1205: }
1.41 deraadt 1206:
1.129 schwarze 1207: assert(mpage->desc == NULL);
1.145 schwarze 1208: if (man != NULL && man->macroset == MACROSET_MDOC)
1.146 schwarze 1209: parse_mdoc(mpage, &man->meta, man->first);
1.129 schwarze 1210: else if (man != NULL)
1.146 schwarze 1211: parse_man(mpage, &man->meta, man->first);
1.47 schwarze 1212: else
1.117 schwarze 1213: parse_cat(mpage, fd);
1.129 schwarze 1214: if (mpage->desc == NULL)
1.92 schwarze 1215: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1.88 schwarze 1216:
1217: if (warnings && !use_all)
1218: for (mlink = mpage->mlinks; mlink;
1219: mlink = mlink->next)
1220: mlink_check(mpage, mlink);
1.6 schwarze 1221:
1.131 schwarze 1222: dbadd(mpage);
1.136 schwarze 1223: mlink = mpage->mlinks;
1.81 schwarze 1224:
1225: nextpage:
1.47 schwarze 1226: ohash_delete(&strings);
1.90 schwarze 1227: ohash_delete(&names);
1.47 schwarze 1228: mpage = ohash_next(&mpages, &pslot);
1229: }
1.64 schwarze 1230:
1231: if (0 == nodb)
1232: SQL_EXEC("END TRANSACTION");
1.87 schwarze 1233: }
1234:
1235: static void
1236: names_check(void)
1237: {
1238: sqlite3_stmt *stmt;
1239: const char *name, *sec, *arch, *key;
1240:
1241: sqlite3_prepare_v2(db,
1242: "SELECT name, sec, arch, key FROM ("
1.90 schwarze 1243: "SELECT name AS key, pageid FROM names "
1.87 schwarze 1244: "WHERE bits & ? AND NOT EXISTS ("
1245: "SELECT pageid FROM mlinks "
1.90 schwarze 1246: "WHERE mlinks.pageid == names.pageid "
1247: "AND mlinks.name == names.name"
1.87 schwarze 1248: ")"
1249: ") JOIN ("
1.99 schwarze 1250: "SELECT sec, arch, name, pageid FROM mlinks "
1251: "GROUP BY pageid"
1.87 schwarze 1252: ") USING (pageid);",
1253: -1, &stmt, NULL);
1254:
1.154 schwarze 1255: if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK)
1.91 schwarze 1256: say("", "%s", sqlite3_errmsg(db));
1.87 schwarze 1257:
1.154 schwarze 1258: while (sqlite3_step(stmt) == SQLITE_ROW) {
1.113 schwarze 1259: name = (const char *)sqlite3_column_text(stmt, 0);
1260: sec = (const char *)sqlite3_column_text(stmt, 1);
1261: arch = (const char *)sqlite3_column_text(stmt, 2);
1262: key = (const char *)sqlite3_column_text(stmt, 3);
1.87 schwarze 1263: say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1264: '\0' == *arch ? "" : "/",
1265: '\0' == *arch ? "" : arch, key);
1266: }
1267: sqlite3_finalize(stmt);
1.47 schwarze 1268: }
1.6 schwarze 1269:
1.47 schwarze 1270: static void
1.81 schwarze 1271: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1272: {
1273: FILE *stream;
1274: char *line, *p, *title;
1.162 schwarze 1275: size_t linesz, plen, titlesz;
1276: ssize_t len;
1277: int offs;
1.1 schwarze 1278:
1.81 schwarze 1279: stream = (-1 == fd) ?
1280: fopen(mpage->mlinks->file, "r") :
1281: fdopen(fd, "r");
1282: if (NULL == stream) {
1.114 doug 1283: if (-1 != fd)
1284: close(fd);
1.47 schwarze 1285: if (warnings)
1.80 schwarze 1286: say(mpage->mlinks->file, "&fopen");
1.47 schwarze 1287: return;
1288: }
1.1 schwarze 1289:
1.162 schwarze 1290: line = NULL;
1291: linesz = 0;
1292:
1.47 schwarze 1293: /* Skip to first blank line. */
1.1 schwarze 1294:
1.162 schwarze 1295: while (getline(&line, &linesz, stream) != -1)
1296: if (*line == '\n')
1.47 schwarze 1297: break;
1.1 schwarze 1298:
1.47 schwarze 1299: /*
1300: * Assume the first line that is not indented
1301: * is the first section header. Skip to it.
1302: */
1.1 schwarze 1303:
1.162 schwarze 1304: while (getline(&line, &linesz, stream) != -1)
1305: if (*line != '\n' && *line != ' ')
1.47 schwarze 1306: break;
1.98 schwarze 1307:
1.47 schwarze 1308: /*
1309: * Read up until the next section into a buffer.
1310: * Strip the leading and trailing newline from each read line,
1311: * appending a trailing space.
1312: * Ignore empty (whitespace-only) lines.
1313: */
1.28 schwarze 1314:
1.47 schwarze 1315: titlesz = 0;
1316: title = NULL;
1.38 schwarze 1317:
1.162 schwarze 1318: while ((len = getline(&line, &linesz, stream)) != -1) {
1319: if (*line != ' ')
1.47 schwarze 1320: break;
1.162 schwarze 1321: offs = 0;
1322: while (isspace((unsigned char)line[offs]))
1323: offs++;
1324: if (line[offs] == '\0')
1.47 schwarze 1325: continue;
1.162 schwarze 1326: title = mandoc_realloc(title, titlesz + len - offs);
1327: memcpy(title + titlesz, line + offs, len - offs);
1328: titlesz += len - offs;
1.47 schwarze 1329: title[titlesz - 1] = ' ';
1330: }
1.162 schwarze 1331: free(line);
1.28 schwarze 1332:
1.47 schwarze 1333: /*
1334: * If no page content can be found, or the input line
1335: * is already the next section header, or there is no
1336: * trailing newline, reuse the page title as the page
1337: * description.
1338: */
1.1 schwarze 1339:
1.47 schwarze 1340: if (NULL == title || '\0' == *title) {
1341: if (warnings)
1342: say(mpage->mlinks->file,
1343: "Cannot find NAME section");
1344: fclose(stream);
1345: free(title);
1346: return;
1347: }
1.24 schwarze 1348:
1.162 schwarze 1349: title[titlesz - 1] = '\0';
1.24 schwarze 1350:
1.47 schwarze 1351: /*
1352: * Skip to the first dash.
1353: * Use the remaining line as the description (no more than 70
1354: * bytes).
1355: */
1.28 schwarze 1356:
1.47 schwarze 1357: if (NULL != (p = strstr(title, "- "))) {
1358: for (p += 2; ' ' == *p || '\b' == *p; p++)
1359: /* Skip to next word. */ ;
1360: } else {
1361: if (warnings)
1362: say(mpage->mlinks->file,
1363: "No dash in title line");
1364: p = title;
1365: }
1.1 schwarze 1366:
1.47 schwarze 1367: plen = strlen(p);
1.1 schwarze 1368:
1.47 schwarze 1369: /* Strip backspace-encoding from line. */
1.1 schwarze 1370:
1.47 schwarze 1371: while (NULL != (line = memchr(p, '\b', plen))) {
1372: len = line - p;
1373: if (0 == len) {
1374: memmove(line, line + 1, plen--);
1375: continue;
1.98 schwarze 1376: }
1.47 schwarze 1377: memmove(line - 1, line + 1, plen - len);
1378: plen -= 2;
1379: }
1.1 schwarze 1380:
1.47 schwarze 1381: mpage->desc = mandoc_strdup(p);
1382: fclose(stream);
1383: free(title);
1384: }
1.16 schwarze 1385:
1.47 schwarze 1386: /*
1387: * Put a type/word pair into the word database for this particular file.
1388: */
1389: static void
1.69 schwarze 1390: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1391: {
1.69 schwarze 1392: char *cp;
1.37 schwarze 1393:
1.47 schwarze 1394: assert(NULL != value);
1.69 schwarze 1395: if (TYPE_arch == type)
1396: for (cp = value; *cp; cp++)
1397: if (isupper((unsigned char)*cp))
1398: *cp = _tolower((unsigned char)*cp);
1.47 schwarze 1399: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1400: }
1401:
1402: /*
1.47 schwarze 1403: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1404: */
1405: static void
1.47 schwarze 1406: putmdockey(const struct mpage *mpage,
1.142 schwarze 1407: const struct roff_node *n, uint64_t m)
1.2 schwarze 1408: {
1.16 schwarze 1409:
1.47 schwarze 1410: for ( ; NULL != n; n = n->next) {
1411: if (NULL != n->child)
1412: putmdockey(mpage, n->child, m);
1.141 schwarze 1413: if (n->type == ROFFT_TEXT)
1.47 schwarze 1414: putkey(mpage, n->string, m);
1415: }
1416: }
1.16 schwarze 1417:
1.47 schwarze 1418: static void
1.143 schwarze 1419: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1420: const struct roff_node *n)
1.47 schwarze 1421: {
1.142 schwarze 1422: const struct roff_node *head, *body;
1.78 schwarze 1423: char *start, *title;
1.47 schwarze 1424: char byte;
1.78 schwarze 1425: size_t sz;
1.16 schwarze 1426:
1.47 schwarze 1427: if (NULL == n)
1428: return;
1.16 schwarze 1429:
1.47 schwarze 1430: /*
1431: * We're only searching for one thing: the first text child in
1432: * the BODY of a NAME section. Since we don't keep track of
1433: * sections in -man, run some hoops to find out whether we're in
1434: * the correct section or not.
1435: */
1.16 schwarze 1436:
1.141 schwarze 1437: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1438: body = n;
1439: assert(body->parent);
1440: if (NULL != (head = body->parent->head) &&
1.98 schwarze 1441: 1 == head->nchild &&
1442: NULL != (head = (head->child)) &&
1.141 schwarze 1443: head->type == ROFFT_TEXT &&
1.98 schwarze 1444: 0 == strcmp(head->string, "NAME") &&
1445: NULL != body->child) {
1.2 schwarze 1446:
1.47 schwarze 1447: /*
1448: * Suck the entire NAME section into memory.
1449: * Yes, we might run away.
1450: * But too many manuals have big, spread-out
1451: * NAME sections over many lines.
1452: */
1.2 schwarze 1453:
1.78 schwarze 1454: title = NULL;
1.147 schwarze 1455: deroff(&title, body);
1.47 schwarze 1456: if (NULL == title)
1457: return;
1.16 schwarze 1458:
1.98 schwarze 1459: /*
1.47 schwarze 1460: * Go through a special heuristic dance here.
1461: * Conventionally, one or more manual names are
1462: * comma-specified prior to a whitespace, then a
1463: * dash, then a description. Try to puzzle out
1464: * the name parts here.
1465: */
1.16 schwarze 1466:
1.78 schwarze 1467: start = title;
1.47 schwarze 1468: for ( ;; ) {
1469: sz = strcspn(start, " ,");
1470: if ('\0' == start[sz])
1471: break;
1.1 schwarze 1472:
1.47 schwarze 1473: byte = start[sz];
1474: start[sz] = '\0';
1.67 schwarze 1475:
1476: /*
1477: * Assume a stray trailing comma in the
1478: * name list if a name begins with a dash.
1479: */
1480:
1481: if ('-' == start[0] ||
1482: ('\\' == start[0] && '-' == start[1]))
1483: break;
1.1 schwarze 1484:
1.90 schwarze 1485: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1486: if ( ! (mpage->name_head_done ||
1487: strcasecmp(start, meta->title))) {
1488: putkey(mpage, start, NAME_HEAD);
1489: mpage->name_head_done = 1;
1490: }
1.1 schwarze 1491:
1.47 schwarze 1492: if (' ' == byte) {
1493: start += sz + 1;
1494: break;
1495: }
1.1 schwarze 1496:
1.47 schwarze 1497: assert(',' == byte);
1498: start += sz + 1;
1499: while (' ' == *start)
1500: start++;
1501: }
1.1 schwarze 1502:
1.78 schwarze 1503: if (start == title) {
1.90 schwarze 1504: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1505: if ( ! (mpage->name_head_done ||
1506: strcasecmp(start, meta->title))) {
1507: putkey(mpage, start, NAME_HEAD);
1508: mpage->name_head_done = 1;
1509: }
1.47 schwarze 1510: free(title);
1511: return;
1512: }
1.1 schwarze 1513:
1.47 schwarze 1514: while (isspace((unsigned char)*start))
1515: start++;
1.1 schwarze 1516:
1.47 schwarze 1517: if (0 == strncmp(start, "-", 1))
1518: start += 1;
1519: else if (0 == strncmp(start, "\\-\\-", 4))
1520: start += 4;
1521: else if (0 == strncmp(start, "\\-", 2))
1522: start += 2;
1523: else if (0 == strncmp(start, "\\(en", 4))
1524: start += 4;
1525: else if (0 == strncmp(start, "\\(em", 4))
1526: start += 4;
1.1 schwarze 1527:
1.47 schwarze 1528: while (' ' == *start)
1529: start++;
1.1 schwarze 1530:
1.47 schwarze 1531: mpage->desc = mandoc_strdup(start);
1532: free(title);
1533: return;
1534: }
1535: }
1.1 schwarze 1536:
1.47 schwarze 1537: for (n = n->child; n; n = n->next) {
1538: if (NULL != mpage->desc)
1539: break;
1.129 schwarze 1540: parse_man(mpage, meta, n);
1.1 schwarze 1541: }
1542: }
1543:
1544: static void
1.143 schwarze 1545: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1546: const struct roff_node *n)
1.1 schwarze 1547: {
1548:
1.47 schwarze 1549: assert(NULL != n);
1550: for (n = n->child; NULL != n; n = n->next) {
1551: switch (n->type) {
1.141 schwarze 1552: case ROFFT_ELEM:
1553: case ROFFT_BLOCK:
1554: case ROFFT_HEAD:
1555: case ROFFT_BODY:
1556: case ROFFT_TAIL:
1.47 schwarze 1557: if (NULL != mdocs[n->tok].fp)
1.127 schwarze 1558: if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1.47 schwarze 1559: break;
1560: if (mdocs[n->tok].mask)
1561: putmdockey(mpage, n->child,
1562: mdocs[n->tok].mask);
1563: break;
1564: default:
1.141 schwarze 1565: assert(n->type != ROFFT_ROOT);
1.47 schwarze 1566: continue;
1567: }
1568: if (NULL != n->child)
1.127 schwarze 1569: parse_mdoc(mpage, meta, n);
1.1 schwarze 1570: }
1571: }
1572:
1.19 schwarze 1573: static int
1.143 schwarze 1574: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1575: const struct roff_node *n)
1.1 schwarze 1576: {
1.131 schwarze 1577: char *start, *end;
1.1 schwarze 1578: size_t sz;
1.19 schwarze 1579:
1.47 schwarze 1580: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1581: NULL == (n = n->child) ||
1.141 schwarze 1582: n->type != ROFFT_TEXT)
1.150 schwarze 1583: return 0;
1.1 schwarze 1584:
1585: /*
1586: * Only consider those `Fd' macro fields that begin with an
1587: * "inclusion" token (versus, e.g., #define).
1588: */
1.47 schwarze 1589:
1.1 schwarze 1590: if (strcmp("#include", n->string))
1.150 schwarze 1591: return 0;
1.1 schwarze 1592:
1.141 schwarze 1593: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1594: return 0;
1.1 schwarze 1595:
1596: /*
1597: * Strip away the enclosing angle brackets and make sure we're
1598: * not zero-length.
1599: */
1600:
1601: start = n->string;
1602: if ('<' == *start || '"' == *start)
1603: start++;
1604:
1605: if (0 == (sz = strlen(start)))
1.150 schwarze 1606: return 0;
1.1 schwarze 1607:
1608: end = &start[(int)sz - 1];
1609: if ('>' == *end || '"' == *end)
1610: end--;
1611:
1.47 schwarze 1612: if (end > start)
1613: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1614: return 0;
1.1 schwarze 1615: }
1616:
1.133 schwarze 1617: static void
1.142 schwarze 1618: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1619: {
1.69 schwarze 1620: char *cp;
1.133 schwarze 1621: size_t sz;
1.1 schwarze 1622:
1.141 schwarze 1623: if (n->type != ROFFT_TEXT)
1.133 schwarze 1624: return;
1.19 schwarze 1625:
1.133 schwarze 1626: /* Skip function pointer punctuation. */
1.1 schwarze 1627:
1.133 schwarze 1628: cp = n->string;
1629: while (*cp == '(' || *cp == '*')
1.1 schwarze 1630: cp++;
1.133 schwarze 1631: sz = strcspn(cp, "()");
1.1 schwarze 1632:
1.133 schwarze 1633: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1634: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1635: putkeys(mpage, cp, sz, NAME_SYN);
1636: }
1.19 schwarze 1637:
1.133 schwarze 1638: static int
1.143 schwarze 1639: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1640: const struct roff_node *n)
1.133 schwarze 1641: {
1642:
1643: if (n->child == NULL)
1.150 schwarze 1644: return 0;
1.133 schwarze 1645:
1646: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1647:
1.133 schwarze 1648: for (n = n->child->next; n != NULL; n = n->next)
1.141 schwarze 1649: if (n->type == ROFFT_TEXT)
1.47 schwarze 1650: putkey(mpage, n->string, TYPE_Fa);
1.19 schwarze 1651:
1.150 schwarze 1652: return 0;
1.128 schwarze 1653: }
1654:
1655: static int
1.143 schwarze 1656: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1657: const struct roff_node *n)
1.128 schwarze 1658: {
1.132 schwarze 1659:
1.141 schwarze 1660: if (n->type != ROFFT_HEAD)
1.150 schwarze 1661: return 1;
1.128 schwarze 1662:
1.133 schwarze 1663: if (n->child != NULL)
1664: parse_mdoc_fname(mpage, n->child);
1665:
1.150 schwarze 1666: return 0;
1.1 schwarze 1667: }
1668:
1.19 schwarze 1669: static int
1.143 schwarze 1670: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1671: const struct roff_node *n)
1.1 schwarze 1672: {
1.47 schwarze 1673: char *cp;
1.1 schwarze 1674:
1675: if (NULL == (n = n->child))
1.150 schwarze 1676: return 0;
1.1 schwarze 1677:
1.47 schwarze 1678: if (NULL == n->next) {
1679: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1680: return 0;
1.47 schwarze 1681: }
1.1 schwarze 1682:
1.77 schwarze 1683: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1684: putkey(mpage, cp, TYPE_Xr);
1685: free(cp);
1.150 schwarze 1686: return 0;
1.1 schwarze 1687: }
1688:
1.19 schwarze 1689: static int
1.143 schwarze 1690: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1691: const struct roff_node *n)
1.1 schwarze 1692: {
1693:
1.141 schwarze 1694: if (n->type == ROFFT_BODY)
1.147 schwarze 1695: deroff(&mpage->desc, n);
1.150 schwarze 1696: return 0;
1.1 schwarze 1697: }
1698:
1.19 schwarze 1699: static int
1.143 schwarze 1700: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1701: const struct roff_node *n)
1.1 schwarze 1702: {
1703:
1.86 schwarze 1704: if (SEC_NAME == n->sec)
1.90 schwarze 1705: putmdockey(mpage, n->child, NAME_TITLE);
1.141 schwarze 1706: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1707: if (n->child == NULL)
1708: putkey(mpage, meta->name, NAME_SYN);
1709: else
1710: putmdockey(mpage, n->child, NAME_SYN);
1.129 schwarze 1711: }
1712: if ( ! (mpage->name_head_done ||
1713: n->child == NULL || n->child->string == NULL ||
1714: strcasecmp(n->child->string, meta->title))) {
1.141 schwarze 1715: putkey(mpage, n->child->string, ROFFT_HEAD);
1.129 schwarze 1716: mpage->name_head_done = 1;
1.127 schwarze 1717: }
1.150 schwarze 1718: return 0;
1.1 schwarze 1719: }
1720:
1.19 schwarze 1721: static int
1.143 schwarze 1722: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1723: const struct roff_node *n)
1.1 schwarze 1724: {
1725:
1.150 schwarze 1726: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1727: }
1728:
1.47 schwarze 1729: static int
1.143 schwarze 1730: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1731: const struct roff_node *n)
1.1 schwarze 1732: {
1733:
1.150 schwarze 1734: return n->type == ROFFT_HEAD;
1.47 schwarze 1735: }
1.1 schwarze 1736:
1.47 schwarze 1737: static int
1.143 schwarze 1738: parse_mdoc_body(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1739: const struct roff_node *n)
1.47 schwarze 1740: {
1.1 schwarze 1741:
1.150 schwarze 1742: return n->type == ROFFT_BODY;
1.1 schwarze 1743: }
1744:
1.47 schwarze 1745: /*
1746: * Add a string to the hash table for the current manual.
1747: * Each string has a bitmask telling which macros it belongs to.
1748: * When we finish the manual, we'll dump the table.
1749: */
1.1 schwarze 1750: static void
1.131 schwarze 1751: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1752: {
1.90 schwarze 1753: struct ohash *htab;
1.47 schwarze 1754: struct str *s;
1.68 schwarze 1755: const char *end;
1.47 schwarze 1756: unsigned int slot;
1.131 schwarze 1757: int i, mustfree;
1.1 schwarze 1758:
1.47 schwarze 1759: if (0 == sz)
1760: return;
1.68 schwarze 1761:
1.131 schwarze 1762: mustfree = render_string(&cp, &sz);
1763:
1.90 schwarze 1764: if (TYPE_Nm & v) {
1765: htab = &names;
1766: v &= name_mask;
1.124 schwarze 1767: if (v & NAME_FIRST)
1768: name_mask &= ~NAME_FIRST;
1.90 schwarze 1769: if (debug > 1)
1770: say(mpage->mlinks->file,
1.130 schwarze 1771: "Adding name %*s, bits=%d", sz, cp, v);
1.90 schwarze 1772: } else {
1773: htab = &strings;
1774: if (debug > 1)
1775: for (i = 0; i < mansearch_keymax; i++)
1.118 schwarze 1776: if ((uint64_t)1 << i & v)
1.90 schwarze 1777: say(mpage->mlinks->file,
1778: "Adding key %s=%*s",
1779: mansearch_keynames[i], sz, cp);
1.68 schwarze 1780: }
1.47 schwarze 1781:
1782: end = cp + sz;
1.90 schwarze 1783: slot = ohash_qlookupi(htab, cp, &end);
1784: s = ohash_find(htab, slot);
1.1 schwarze 1785:
1.47 schwarze 1786: if (NULL != s && mpage == s->mpage) {
1787: s->mask |= v;
1.1 schwarze 1788: return;
1.47 schwarze 1789: } else if (NULL == s) {
1.101 schwarze 1790: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1791: memcpy(s->key, cp, sz);
1.90 schwarze 1792: ohash_insert(htab, slot, s);
1.47 schwarze 1793: }
1794: s->mpage = mpage;
1795: s->mask = v;
1.131 schwarze 1796:
1797: if (mustfree)
1798: free(cp);
1.1 schwarze 1799: }
1800:
1801: /*
1.47 schwarze 1802: * Take a Unicode codepoint and produce its UTF-8 encoding.
1803: * This isn't the best way to do this, but it works.
1804: * The magic numbers are from the UTF-8 packaging.
1805: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1806: */
1.47 schwarze 1807: static size_t
1808: utf8(unsigned int cp, char out[7])
1.1 schwarze 1809: {
1.47 schwarze 1810: size_t rc;
1.1 schwarze 1811:
1.47 schwarze 1812: rc = 0;
1813: if (cp <= 0x0000007F) {
1814: rc = 1;
1815: out[0] = (char)cp;
1816: } else if (cp <= 0x000007FF) {
1817: rc = 2;
1818: out[0] = (cp >> 6 & 31) | 192;
1819: out[1] = (cp & 63) | 128;
1820: } else if (cp <= 0x0000FFFF) {
1821: rc = 3;
1822: out[0] = (cp >> 12 & 15) | 224;
1823: out[1] = (cp >> 6 & 63) | 128;
1824: out[2] = (cp & 63) | 128;
1825: } else if (cp <= 0x001FFFFF) {
1826: rc = 4;
1827: out[0] = (cp >> 18 & 7) | 240;
1828: out[1] = (cp >> 12 & 63) | 128;
1829: out[2] = (cp >> 6 & 63) | 128;
1830: out[3] = (cp & 63) | 128;
1831: } else if (cp <= 0x03FFFFFF) {
1832: rc = 5;
1833: out[0] = (cp >> 24 & 3) | 248;
1834: out[1] = (cp >> 18 & 63) | 128;
1835: out[2] = (cp >> 12 & 63) | 128;
1836: out[3] = (cp >> 6 & 63) | 128;
1837: out[4] = (cp & 63) | 128;
1838: } else if (cp <= 0x7FFFFFFF) {
1839: rc = 6;
1840: out[0] = (cp >> 30 & 1) | 252;
1841: out[1] = (cp >> 24 & 63) | 128;
1842: out[2] = (cp >> 18 & 63) | 128;
1843: out[3] = (cp >> 12 & 63) | 128;
1844: out[4] = (cp >> 6 & 63) | 128;
1845: out[5] = (cp & 63) | 128;
1846: } else
1.150 schwarze 1847: return 0;
1.19 schwarze 1848:
1.47 schwarze 1849: out[rc] = '\0';
1.150 schwarze 1850: return rc;
1.1 schwarze 1851: }
1852:
1.47 schwarze 1853: /*
1.131 schwarze 1854: * If the string contains escape sequences,
1855: * replace it with an allocated rendering and return 1,
1856: * such that the caller can free it after use.
1857: * Otherwise, do nothing and return 0.
1.47 schwarze 1858: */
1.131 schwarze 1859: static int
1860: render_string(char **public, size_t *psz)
1.1 schwarze 1861: {
1.131 schwarze 1862: const char *src, *scp, *addcp, *seq;
1863: char *dst;
1864: size_t ssz, dsz, addsz;
1.71 schwarze 1865: char utfbuf[7], res[6];
1.131 schwarze 1866: int seqlen, unicode;
1.47 schwarze 1867:
1868: res[0] = '\\';
1869: res[1] = '\t';
1870: res[2] = ASCII_NBRSP;
1871: res[3] = ASCII_HYPH;
1.71 schwarze 1872: res[4] = ASCII_BREAK;
1873: res[5] = '\0';
1.1 schwarze 1874:
1.131 schwarze 1875: src = scp = *public;
1876: ssz = *psz;
1877: dst = NULL;
1878: dsz = 0;
1879:
1880: while (scp < src + *psz) {
1881:
1882: /* Leave normal characters unchanged. */
1883:
1884: if (strchr(res, *scp) == NULL) {
1885: if (dst != NULL)
1886: dst[dsz++] = *scp;
1887: scp++;
1888: continue;
1889: }
1.1 schwarze 1890:
1.47 schwarze 1891: /*
1.131 schwarze 1892: * Found something that requires replacing,
1893: * make sure we have a destination buffer.
1.47 schwarze 1894: */
1.131 schwarze 1895:
1896: if (dst == NULL) {
1897: dst = mandoc_malloc(ssz + 1);
1898: dsz = scp - src;
1899: memcpy(dst, src, dsz);
1.47 schwarze 1900: }
1.39 schwarze 1901:
1.131 schwarze 1902: /* Handle single-char special characters. */
1903:
1904: switch (*scp) {
1905: case '\\':
1906: break;
1.98 schwarze 1907: case '\t':
1908: case ASCII_NBRSP:
1.131 schwarze 1909: dst[dsz++] = ' ';
1910: scp++;
1911: continue;
1912: case ASCII_HYPH:
1913: dst[dsz++] = '-';
1.71 schwarze 1914: /* FALLTHROUGH */
1.98 schwarze 1915: case ASCII_BREAK:
1.131 schwarze 1916: scp++;
1.47 schwarze 1917: continue;
1.71 schwarze 1918: default:
1.131 schwarze 1919: abort();
1.71 schwarze 1920: }
1.39 schwarze 1921:
1.47 schwarze 1922: /*
1.131 schwarze 1923: * Found an escape sequence.
1924: * Read past the slash, then parse it.
1925: * Ignore everything except characters.
1.47 schwarze 1926: */
1.52 schwarze 1927:
1.131 schwarze 1928: scp++;
1929: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1930: continue;
1.39 schwarze 1931:
1.47 schwarze 1932: /*
1.52 schwarze 1933: * Render the special character
1934: * as either UTF-8 or ASCII.
1.47 schwarze 1935: */
1.52 schwarze 1936:
1937: if (write_utf8) {
1.157 schwarze 1938: unicode = mchars_spec2cp(seq, seqlen);
1.131 schwarze 1939: if (unicode <= 0)
1.52 schwarze 1940: continue;
1.131 schwarze 1941: addsz = utf8(unicode, utfbuf);
1942: if (addsz == 0)
1.52 schwarze 1943: continue;
1.131 schwarze 1944: addcp = utfbuf;
1.52 schwarze 1945: } else {
1.157 schwarze 1946: addcp = mchars_spec2str(seq, seqlen, &addsz);
1.131 schwarze 1947: if (addcp == NULL)
1.52 schwarze 1948: continue;
1.131 schwarze 1949: if (*addcp == ASCII_NBRSP) {
1950: addcp = " ";
1951: addsz = 1;
1.52 schwarze 1952: }
1953: }
1.1 schwarze 1954:
1.47 schwarze 1955: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1956:
1.131 schwarze 1957: ssz += addsz;
1958: dst = mandoc_realloc(dst, ssz + 1);
1959: memcpy(dst + dsz, addcp, addsz);
1960: dsz += addsz;
1961: }
1962: if (dst != NULL) {
1963: *public = dst;
1964: *psz = dsz;
1965: }
1966:
1967: /* Trim trailing whitespace and NUL-terminate. */
1968:
1969: while (*psz > 0 && (*public)[*psz - 1] == ' ')
1970: --*psz;
1971: if (dst != NULL) {
1972: (*public)[*psz] = '\0';
1.150 schwarze 1973: return 1;
1.131 schwarze 1974: } else
1.150 schwarze 1975: return 0;
1.1 schwarze 1976: }
1977:
1.75 schwarze 1978: static void
1979: dbadd_mlink(const struct mlink *mlink)
1980: {
1981: size_t i;
1982:
1983: i = 1;
1984: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1985: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1986: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.94 schwarze 1987: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1.75 schwarze 1988: SQL_STEP(stmts[STMT_INSERT_LINK]);
1989: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1.124 schwarze 1990: }
1991:
1992: static void
1993: dbadd_mlink_name(const struct mlink *mlink)
1994: {
1.130 schwarze 1995: uint64_t bits;
1.124 schwarze 1996: size_t i;
1997:
1998: dbadd_mlink(mlink);
1.115 schwarze 1999:
2000: i = 1;
1.130 schwarze 2001: SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
2002: bits = NAME_FILE & NAME_MASK;
2003: if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
2004: bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
2005: sqlite3_reset(stmts[STMT_SELECT_NAME]);
2006: }
2007:
2008: i = 1;
2009: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
1.115 schwarze 2010: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2011: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2012: SQL_STEP(stmts[STMT_INSERT_NAME]);
2013: sqlite3_reset(stmts[STMT_INSERT_NAME]);
1.75 schwarze 2014: }
2015:
1.11 schwarze 2016: /*
1.47 schwarze 2017: * Flush the current page's terms (and their bits) into the database.
2018: * Wrap the entire set of additions in a transaction to make sqlite be a
2019: * little faster.
1.53 schwarze 2020: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 2021: */
2022: static void
1.131 schwarze 2023: dbadd(struct mpage *mpage)
1.11 schwarze 2024: {
1.47 schwarze 2025: struct mlink *mlink;
2026: struct str *key;
1.131 schwarze 2027: char *cp;
1.47 schwarze 2028: size_t i;
2029: unsigned int slot;
1.131 schwarze 2030: int mustfree;
1.47 schwarze 2031:
1.85 schwarze 2032: mlink = mpage->mlinks;
1.11 schwarze 2033:
1.85 schwarze 2034: if (nodb) {
1.104 schwarze 2035: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2036: key = ohash_next(&names, &slot))
1.104 schwarze 2037: free(key);
2038: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2039: key = ohash_next(&strings, &slot))
1.104 schwarze 2040: free(key);
1.102 schwarze 2041: if (0 == debug)
2042: return;
1.85 schwarze 2043: while (NULL != mlink) {
2044: fputs(mlink->name, stdout);
2045: if (NULL == mlink->next ||
2046: strcmp(mlink->dsec, mlink->next->dsec) ||
2047: strcmp(mlink->fsec, mlink->next->fsec) ||
2048: strcmp(mlink->arch, mlink->next->arch)) {
2049: putchar('(');
2050: if ('\0' == *mlink->dsec)
2051: fputs(mlink->fsec, stdout);
2052: else
2053: fputs(mlink->dsec, stdout);
2054: if ('\0' != *mlink->arch)
2055: printf("/%s", mlink->arch);
2056: putchar(')');
2057: }
2058: mlink = mlink->next;
2059: if (NULL != mlink)
2060: fputs(", ", stdout);
2061: }
1.89 schwarze 2062: printf(" - %s\n", mpage->desc);
1.11 schwarze 2063: return;
1.85 schwarze 2064: }
2065:
2066: if (debug)
2067: say(mlink->file, "Adding to database");
1.47 schwarze 2068:
1.131 schwarze 2069: cp = mpage->desc;
2070: i = strlen(cp);
2071: mustfree = render_string(&cp, &i);
1.47 schwarze 2072: i = 1;
1.131 schwarze 2073: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
1.116 schwarze 2074: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
1.47 schwarze 2075: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.94 schwarze 2076: mpage->pageid = sqlite3_last_insert_rowid(db);
1.47 schwarze 2077: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1.131 schwarze 2078: if (mustfree)
2079: free(cp);
1.47 schwarze 2080:
1.85 schwarze 2081: while (NULL != mlink) {
1.75 schwarze 2082: dbadd_mlink(mlink);
1.85 schwarze 2083: mlink = mlink->next;
2084: }
1.91 schwarze 2085: mlink = mpage->mlinks;
1.47 schwarze 2086:
1.90 schwarze 2087: for (key = ohash_first(&names, &slot); NULL != key;
2088: key = ohash_next(&names, &slot)) {
2089: assert(key->mpage == mpage);
2090: i = 1;
2091: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
1.131 schwarze 2092: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
1.94 schwarze 2093: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
1.90 schwarze 2094: SQL_STEP(stmts[STMT_INSERT_NAME]);
2095: sqlite3_reset(stmts[STMT_INSERT_NAME]);
2096: free(key);
2097: }
1.47 schwarze 2098: for (key = ohash_first(&strings, &slot); NULL != key;
2099: key = ohash_next(&strings, &slot)) {
2100: assert(key->mpage == mpage);
2101: i = 1;
2102: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.131 schwarze 2103: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
1.94 schwarze 2104: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
1.47 schwarze 2105: SQL_STEP(stmts[STMT_INSERT_KEY]);
2106: sqlite3_reset(stmts[STMT_INSERT_KEY]);
2107: free(key);
1.33 schwarze 2108: }
1.47 schwarze 2109: }
1.41 deraadt 2110:
1.47 schwarze 2111: static void
2112: dbprune(void)
2113: {
2114: struct mpage *mpage;
2115: struct mlink *mlink;
2116: size_t i;
2117: unsigned int slot;
1.11 schwarze 2118:
1.63 schwarze 2119: if (0 == nodb)
2120: SQL_EXEC("BEGIN TRANSACTION");
1.47 schwarze 2121:
1.63 schwarze 2122: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2123: mpage = ohash_next(&mpages, &slot)) {
1.47 schwarze 2124: mlink = mpage->mlinks;
1.82 schwarze 2125: if (debug)
1.63 schwarze 2126: say(mlink->file, "Deleting from database");
2127: if (nodb)
2128: continue;
2129: for ( ; NULL != mlink; mlink = mlink->next) {
2130: i = 1;
2131: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2132: i, mlink->dsec);
2133: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2134: i, mlink->arch);
2135: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2136: i, mlink->name);
2137: SQL_STEP(stmts[STMT_DELETE_PAGE]);
2138: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2139: }
1.11 schwarze 2140: }
1.63 schwarze 2141:
2142: if (0 == nodb)
2143: SQL_EXEC("END TRANSACTION");
1.47 schwarze 2144: }
1.22 schwarze 2145:
1.47 schwarze 2146: /*
2147: * Close an existing database and its prepared statements.
2148: * If "real" is not set, rename the temporary file into the real one.
2149: */
2150: static void
2151: dbclose(int real)
2152: {
2153: size_t i;
1.72 schwarze 2154: int status;
2155: pid_t child;
1.11 schwarze 2156:
1.47 schwarze 2157: if (nodb)
2158: return;
1.11 schwarze 2159:
1.47 schwarze 2160: for (i = 0; i < STMT__MAX; i++) {
2161: sqlite3_finalize(stmts[i]);
2162: stmts[i] = NULL;
1.28 schwarze 2163: }
1.22 schwarze 2164:
1.47 schwarze 2165: sqlite3_close(db);
2166: db = NULL;
1.11 schwarze 2167:
1.47 schwarze 2168: if (real)
2169: return;
1.22 schwarze 2170:
1.72 schwarze 2171: if ('\0' == *tempfilename) {
2172: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2173: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2174: say(MANDOC_DB, "&rename");
1.72 schwarze 2175: }
2176: return;
2177: }
2178:
2179: switch (child = fork()) {
1.98 schwarze 2180: case -1:
1.72 schwarze 2181: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2182: say("", "&fork cmp");
1.72 schwarze 2183: return;
1.98 schwarze 2184: case 0:
1.72 schwarze 2185: execlp("cmp", "cmp", "-s",
1.149 schwarze 2186: tempfilename, MANDOC_DB, (char *)NULL);
1.80 schwarze 2187: say("", "&exec cmp");
1.72 schwarze 2188: exit(0);
2189: default:
2190: break;
2191: }
2192: if (-1 == waitpid(child, &status, 0)) {
2193: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2194: say("", "&wait cmp");
1.72 schwarze 2195: } else if (WIFSIGNALED(status)) {
2196: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2197: say("", "cmp died from signal %d", WTERMSIG(status));
1.72 schwarze 2198: } else if (WEXITSTATUS(status)) {
1.47 schwarze 2199: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2200: say(MANDOC_DB,
2201: "Data changed, but cannot replace database");
2202: }
2203:
2204: *strrchr(tempfilename, '/') = '\0';
2205: switch (child = fork()) {
1.98 schwarze 2206: case -1:
1.72 schwarze 2207: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2208: say("", "&fork rm");
1.72 schwarze 2209: return;
1.98 schwarze 2210: case 0:
1.149 schwarze 2211: execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
1.80 schwarze 2212: say("", "&exec rm");
1.72 schwarze 2213: exit((int)MANDOCLEVEL_SYSERR);
2214: default:
2215: break;
2216: }
2217: if (-1 == waitpid(child, &status, 0)) {
2218: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2219: say("", "&wait rm");
1.72 schwarze 2220: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2221: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2222: say("", "%s: Cannot remove temporary directory",
2223: tempfilename);
1.22 schwarze 2224: }
1.11 schwarze 2225: }
2226:
1.47 schwarze 2227: /*
2228: * This is straightforward stuff.
2229: * Open a database connection to a "temporary" database, then open a set
2230: * of prepared statements we'll use over and over again.
2231: * If "real" is set, we use the existing database; if not, we truncate a
2232: * temporary one.
2233: * Must be matched by dbclose().
2234: */
2235: static int
2236: dbopen(int real)
1.2 schwarze 2237: {
1.72 schwarze 2238: const char *sql;
1.47 schwarze 2239: int rc, ofl;
1.6 schwarze 2240:
1.98 schwarze 2241: if (nodb)
1.150 schwarze 2242: return 1;
1.6 schwarze 2243:
1.72 schwarze 2244: *tempfilename = '\0';
1.47 schwarze 2245: ofl = SQLITE_OPEN_READWRITE;
1.72 schwarze 2246:
2247: if (real) {
2248: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2249: if (SQLITE_OK != rc) {
1.47 schwarze 2250: exitcode = (int)MANDOCLEVEL_SYSERR;
1.108 sthen 2251: if (SQLITE_CANTOPEN != rc)
2252: say(MANDOC_DB, "%s", sqlite3_errstr(rc));
1.150 schwarze 2253: return 0;
1.28 schwarze 2254: }
1.72 schwarze 2255: goto prepare_statements;
2256: }
2257:
2258: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.6 schwarze 2259:
1.72 schwarze 2260: remove(MANDOC_DB "~");
2261: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.98 schwarze 2262: if (SQLITE_OK == rc)
1.72 schwarze 2263: goto create_tables;
1.73 schwarze 2264: if (MPARSE_QUICK & mparse_options) {
1.47 schwarze 2265: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2266: say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
1.150 schwarze 2267: return 0;
1.47 schwarze 2268: }
1.6 schwarze 2269:
1.100 schwarze 2270: (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2271: sizeof(tempfilename));
1.72 schwarze 2272: if (NULL == mkdtemp(tempfilename)) {
2273: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2274: say("", "&%s", tempfilename);
1.150 schwarze 2275: return 0;
1.72 schwarze 2276: }
1.100 schwarze 2277: (void)strlcat(tempfilename, "/" MANDOC_DB,
2278: sizeof(tempfilename));
1.72 schwarze 2279: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2280: if (SQLITE_OK != rc) {
1.47 schwarze 2281: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2282: say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
1.150 schwarze 2283: return 0;
1.2 schwarze 2284: }
2285:
1.72 schwarze 2286: create_tables:
1.47 schwarze 2287: sql = "CREATE TABLE \"mpages\" (\n"
1.89 schwarze 2288: " \"desc\" TEXT NOT NULL,\n"
1.47 schwarze 2289: " \"form\" INTEGER NOT NULL,\n"
1.94 schwarze 2290: " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1.47 schwarze 2291: ");\n"
2292: "\n"
2293: "CREATE TABLE \"mlinks\" (\n"
2294: " \"sec\" TEXT NOT NULL,\n"
2295: " \"arch\" TEXT NOT NULL,\n"
2296: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2297: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2298: "ON DELETE CASCADE\n"
1.47 schwarze 2299: ");\n"
1.93 schwarze 2300: "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
1.47 schwarze 2301: "\n"
1.90 schwarze 2302: "CREATE TABLE \"names\" (\n"
2303: " \"bits\" INTEGER NOT NULL,\n"
2304: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2305: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.130 schwarze 2306: "ON DELETE CASCADE,\n"
2307: " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
1.90 schwarze 2308: ");\n"
2309: "\n"
1.47 schwarze 2310: "CREATE TABLE \"keys\" (\n"
2311: " \"bits\" INTEGER NOT NULL,\n"
2312: " \"key\" TEXT NOT NULL,\n"
1.94 schwarze 2313: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2314: "ON DELETE CASCADE\n"
1.93 schwarze 2315: ");\n"
2316: "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
1.47 schwarze 2317:
2318: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2319: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2320: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.103 schwarze 2321: sqlite3_close(db);
1.150 schwarze 2322: return 0;
1.2 schwarze 2323: }
2324:
1.47 schwarze 2325: prepare_statements:
1.103 schwarze 2326: if (SQLITE_OK != sqlite3_exec(db,
2327: "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2328: exitcode = (int)MANDOCLEVEL_SYSERR;
2329: say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2330: sqlite3_errmsg(db));
2331: sqlite3_close(db);
1.150 schwarze 2332: return 0;
1.103 schwarze 2333: }
2334:
1.94 schwarze 2335: sql = "DELETE FROM mpages WHERE pageid IN "
1.63 schwarze 2336: "(SELECT pageid FROM mlinks WHERE "
2337: "sec=? AND arch=? AND name=?)";
1.47 schwarze 2338: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2339: sql = "INSERT INTO mpages "
1.89 schwarze 2340: "(desc,form) VALUES (?,?)";
1.47 schwarze 2341: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2342: sql = "INSERT INTO mlinks "
1.61 schwarze 2343: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.47 schwarze 2344: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.130 schwarze 2345: sql = "SELECT bits FROM names where pageid = ?";
2346: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
1.90 schwarze 2347: sql = "INSERT INTO names "
2348: "(bits,name,pageid) VALUES (?,?,?)";
2349: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
1.47 schwarze 2350: sql = "INSERT INTO keys "
2351: "(bits,key,pageid) VALUES (?,?,?)";
2352: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.6 schwarze 2353:
1.47 schwarze 2354: /*
2355: * When opening a new database, we can turn off
2356: * synchronous mode for much better performance.
2357: */
1.6 schwarze 2358:
1.103 schwarze 2359: if (real && SQLITE_OK != sqlite3_exec(db,
2360: "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2361: exitcode = (int)MANDOCLEVEL_SYSERR;
2362: say(MANDOC_DB, "PRAGMA synchronous: %s",
2363: sqlite3_errmsg(db));
2364: sqlite3_close(db);
1.150 schwarze 2365: return 0;
1.103 schwarze 2366: }
1.11 schwarze 2367:
1.150 schwarze 2368: return 1;
1.47 schwarze 2369: }
1.6 schwarze 2370:
1.47 schwarze 2371: static int
1.120 schwarze 2372: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2373: {
2374: static char startdir[PATH_MAX];
1.110 schwarze 2375: static int getcwd_status; /* 1 = ok, 2 = failure */
2376: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2377: char *cp;
1.6 schwarze 2378:
1.47 schwarze 2379: /*
1.110 schwarze 2380: * Remember the original working directory, if possible.
2381: * This will be needed if the second or a later directory
2382: * on the command line is given as a relative path.
2383: * Do not error out if the current directory is not
2384: * searchable: Maybe it won't be needed after all.
2385: */
2386: if (0 == getcwd_status) {
2387: if (NULL == getcwd(startdir, sizeof(startdir))) {
2388: getcwd_status = 2;
2389: (void)strlcpy(startdir, strerror(errno),
2390: sizeof(startdir));
2391: } else
2392: getcwd_status = 1;
2393: }
2394:
2395: /*
2396: * We are leaving the old base directory.
2397: * Do not use it any longer, not even for messages.
2398: */
2399: *basedir = '\0';
2400:
2401: /*
2402: * If and only if the directory was changed earlier and
2403: * the next directory to process is given as a relative path,
2404: * first go back, or bail out if that is impossible.
1.47 schwarze 2405: */
1.110 schwarze 2406: if (chdir_status && '/' != *targetdir) {
2407: if (2 == getcwd_status) {
1.47 schwarze 2408: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2409: say("", "getcwd: %s", startdir);
1.150 schwarze 2410: return 0;
1.47 schwarze 2411: }
1.110 schwarze 2412: if (-1 == chdir(startdir)) {
1.47 schwarze 2413: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2414: say("", "&chdir %s", startdir);
1.150 schwarze 2415: return 0;
1.2 schwarze 2416: }
2417: }
1.110 schwarze 2418:
2419: /*
2420: * Always resolve basedir to the canonicalized absolute
2421: * pathname and append a trailing slash, such that
2422: * we can reliably check whether files are inside.
2423: */
1.47 schwarze 2424: if (NULL == realpath(targetdir, basedir)) {
1.120 schwarze 2425: if (report_baddir || errno != ENOENT) {
2426: exitcode = (int)MANDOCLEVEL_BADARG;
2427: say("", "&%s: realpath", targetdir);
2428: }
1.150 schwarze 2429: return 0;
1.47 schwarze 2430: } else if (-1 == chdir(basedir)) {
1.120 schwarze 2431: if (report_baddir || errno != ENOENT) {
2432: exitcode = (int)MANDOCLEVEL_BADARG;
2433: say("", "&chdir");
2434: }
1.150 schwarze 2435: return 0;
1.109 schwarze 2436: }
1.110 schwarze 2437: chdir_status = 1;
1.109 schwarze 2438: cp = strchr(basedir, '\0');
2439: if ('/' != cp[-1]) {
2440: if (cp - basedir >= PATH_MAX - 1) {
2441: exitcode = (int)MANDOCLEVEL_SYSERR;
2442: say("", "Filename too long");
1.150 schwarze 2443: return 0;
1.109 schwarze 2444: }
2445: *cp++ = '/';
2446: *cp = '\0';
1.47 schwarze 2447: }
1.150 schwarze 2448: return 1;
1.2 schwarze 2449: }
2450:
2451: static void
1.47 schwarze 2452: say(const char *file, const char *format, ...)
1.2 schwarze 2453: {
1.47 schwarze 2454: va_list ap;
1.80 schwarze 2455: int use_errno;
1.2 schwarze 2456:
1.47 schwarze 2457: if ('\0' != *basedir)
2458: fprintf(stderr, "%s", basedir);
2459: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2460: fputc('/', stderr);
1.47 schwarze 2461: if ('\0' != *file)
2462: fprintf(stderr, "%s", file);
1.31 schwarze 2463:
1.80 schwarze 2464: use_errno = 1;
2465: if (NULL != format) {
2466: switch (*format) {
1.98 schwarze 2467: case '&':
1.80 schwarze 2468: format++;
2469: break;
1.98 schwarze 2470: case '\0':
1.80 schwarze 2471: format = NULL;
2472: break;
2473: default:
2474: use_errno = 0;
2475: break;
2476: }
2477: }
2478: if (NULL != format) {
2479: if ('\0' != *basedir || '\0' != *file)
2480: fputs(": ", stderr);
2481: va_start(ap, format);
2482: vfprintf(stderr, format, ap);
2483: va_end(ap);
2484: }
2485: if (use_errno) {
2486: if ('\0' != *basedir || '\0' != *file || NULL != format)
2487: fputs(": ", stderr);
1.47 schwarze 2488: perror(NULL);
1.80 schwarze 2489: } else
2490: fputc('\n', stderr);
1.1 schwarze 2491: }