Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.151
1.151 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.150 2015/10/06 18:30:43 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 <ohash.h>
38: #include <sqlite3.h>
39:
1.141 schwarze 40: #include "mandoc_aux.h"
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 *hash_alloc(size_t, void *);
1.106 espie 140: static void hash_free(void *, void *);
141: static void *hash_calloc(size_t, size_t, void *);
1.47 schwarze 142: static void mlink_add(struct mlink *, const struct stat *);
1.88 schwarze 143: static void mlink_check(struct mpage *, struct mlink *);
1.47 schwarze 144: static void mlink_free(struct mlink *);
145: static void mlinks_undupe(struct mpage *);
146: static void mpages_free(void);
1.131 schwarze 147: static void mpages_merge(struct mparse *);
1.87 schwarze 148: static void names_check(void);
1.81 schwarze 149: static void parse_cat(struct mpage *, int);
1.143 schwarze 150: static void parse_man(struct mpage *, const struct roff_meta *,
1.142 schwarze 151: const struct roff_node *);
1.143 schwarze 152: static void parse_mdoc(struct mpage *, const struct roff_meta *,
1.142 schwarze 153: const struct roff_node *);
1.143 schwarze 154: static int parse_mdoc_body(struct mpage *, const struct roff_meta *,
1.142 schwarze 155: const struct roff_node *);
1.143 schwarze 156: static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
1.142 schwarze 157: const struct roff_node *);
1.143 schwarze 158: static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
1.142 schwarze 159: const struct roff_node *);
160: static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
1.143 schwarze 161: static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
1.142 schwarze 162: const struct roff_node *);
1.143 schwarze 163: static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
1.142 schwarze 164: const struct roff_node *);
1.143 schwarze 165: static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
1.142 schwarze 166: const struct roff_node *);
1.143 schwarze 167: static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
1.142 schwarze 168: const struct roff_node *);
1.143 schwarze 169: static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
1.142 schwarze 170: const struct roff_node *);
1.143 schwarze 171: static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
1.142 schwarze 172: const struct roff_node *);
1.69 schwarze 173: static void putkey(const struct mpage *, char *, uint64_t);
1.131 schwarze 174: static void putkeys(const struct mpage *, char *, size_t, uint64_t);
1.47 schwarze 175: static void putmdockey(const struct mpage *,
1.142 schwarze 176: const struct roff_node *, uint64_t);
1.131 schwarze 177: static int render_string(char **, size_t *);
1.47 schwarze 178: static void say(const char *, const char *, ...);
1.120 schwarze 179: static int set_basedir(const char *, int);
1.47 schwarze 180: static int treescan(void);
181: static size_t utf8(unsigned int, char [7]);
182:
1.151 ! schwarze 183: extern char *__progname;
! 184:
1.72 schwarze 185: static char tempfilename[32];
1.59 schwarze 186: static int nodb; /* no database changes */
1.73 schwarze 187: static int mparse_options; /* abort the parse early */
1.98 schwarze 188: static int use_all; /* use all found files */
189: static int debug; /* print what we're doing */
190: static int warnings; /* warn about crap */
1.52 schwarze 191: static int write_utf8; /* write UTF-8 output; else ASCII */
1.47 schwarze 192: static int exitcode; /* to be returned by main */
1.98 schwarze 193: static enum op op; /* operational mode */
1.47 schwarze 194: static char basedir[PATH_MAX]; /* current base directory */
1.131 schwarze 195: static struct mchars *mchars; /* table of named characters */
1.47 schwarze 196: static struct ohash mpages; /* table of distinct manual pages */
197: static struct ohash mlinks; /* table of directory entries */
1.90 schwarze 198: static struct ohash names; /* table of all names */
1.47 schwarze 199: static struct ohash strings; /* table of all strings */
200: static sqlite3 *db = NULL; /* current database */
201: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.90 schwarze 202: static uint64_t name_mask;
1.47 schwarze 203:
1.19 schwarze 204: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.47 schwarze 205: { NULL, 0 }, /* Ap */
206: { NULL, 0 }, /* Dd */
207: { NULL, 0 }, /* Dt */
208: { NULL, 0 }, /* Os */
209: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
210: { parse_mdoc_head, TYPE_Ss }, /* Ss */
211: { NULL, 0 }, /* Pp */
212: { NULL, 0 }, /* D1 */
213: { NULL, 0 }, /* Dl */
214: { NULL, 0 }, /* Bd */
215: { NULL, 0 }, /* Ed */
216: { NULL, 0 }, /* Bl */
217: { NULL, 0 }, /* El */
218: { NULL, 0 }, /* It */
219: { NULL, 0 }, /* Ad */
220: { NULL, TYPE_An }, /* An */
221: { NULL, TYPE_Ar }, /* Ar */
222: { NULL, TYPE_Cd }, /* Cd */
223: { NULL, TYPE_Cm }, /* Cm */
224: { NULL, TYPE_Dv }, /* Dv */
225: { NULL, TYPE_Er }, /* Er */
226: { NULL, TYPE_Ev }, /* Ev */
227: { NULL, 0 }, /* Ex */
228: { NULL, TYPE_Fa }, /* Fa */
229: { parse_mdoc_Fd, 0 }, /* Fd */
230: { NULL, TYPE_Fl }, /* Fl */
231: { parse_mdoc_Fn, 0 }, /* Fn */
232: { NULL, TYPE_Ft }, /* Ft */
233: { NULL, TYPE_Ic }, /* Ic */
1.49 schwarze 234: { NULL, TYPE_In }, /* In */
1.47 schwarze 235: { NULL, TYPE_Li }, /* Li */
1.89 schwarze 236: { parse_mdoc_Nd, 0 }, /* Nd */
1.86 schwarze 237: { parse_mdoc_Nm, 0 }, /* Nm */
1.47 schwarze 238: { NULL, 0 }, /* Op */
239: { NULL, 0 }, /* Ot */
240: { NULL, TYPE_Pa }, /* Pa */
241: { NULL, 0 }, /* Rv */
1.49 schwarze 242: { NULL, TYPE_St }, /* St */
1.47 schwarze 243: { NULL, TYPE_Va }, /* Va */
244: { parse_mdoc_body, TYPE_Va }, /* Vt */
245: { parse_mdoc_Xr, 0 }, /* Xr */
246: { NULL, 0 }, /* %A */
247: { NULL, 0 }, /* %B */
248: { NULL, 0 }, /* %D */
249: { NULL, 0 }, /* %I */
250: { NULL, 0 }, /* %J */
251: { NULL, 0 }, /* %N */
252: { NULL, 0 }, /* %O */
253: { NULL, 0 }, /* %P */
254: { NULL, 0 }, /* %R */
255: { NULL, 0 }, /* %T */
256: { NULL, 0 }, /* %V */
257: { NULL, 0 }, /* Ac */
258: { NULL, 0 }, /* Ao */
259: { NULL, 0 }, /* Aq */
260: { NULL, TYPE_At }, /* At */
261: { NULL, 0 }, /* Bc */
262: { NULL, 0 }, /* Bf */
263: { NULL, 0 }, /* Bo */
264: { NULL, 0 }, /* Bq */
265: { NULL, TYPE_Bsx }, /* Bsx */
266: { NULL, TYPE_Bx }, /* Bx */
267: { NULL, 0 }, /* Db */
268: { NULL, 0 }, /* Dc */
269: { NULL, 0 }, /* Do */
270: { NULL, 0 }, /* Dq */
271: { NULL, 0 }, /* Ec */
272: { NULL, 0 }, /* Ef */
273: { NULL, TYPE_Em }, /* Em */
274: { NULL, 0 }, /* Eo */
275: { NULL, TYPE_Fx }, /* Fx */
276: { NULL, TYPE_Ms }, /* Ms */
277: { NULL, 0 }, /* No */
278: { NULL, 0 }, /* Ns */
279: { NULL, TYPE_Nx }, /* Nx */
280: { NULL, TYPE_Ox }, /* Ox */
281: { NULL, 0 }, /* Pc */
282: { NULL, 0 }, /* Pf */
283: { NULL, 0 }, /* Po */
284: { NULL, 0 }, /* Pq */
285: { NULL, 0 }, /* Qc */
286: { NULL, 0 }, /* Ql */
287: { NULL, 0 }, /* Qo */
288: { NULL, 0 }, /* Qq */
289: { NULL, 0 }, /* Re */
290: { NULL, 0 }, /* Rs */
291: { NULL, 0 }, /* Sc */
292: { NULL, 0 }, /* So */
293: { NULL, 0 }, /* Sq */
294: { NULL, 0 }, /* Sm */
295: { NULL, 0 }, /* Sx */
296: { NULL, TYPE_Sy }, /* Sy */
297: { NULL, TYPE_Tn }, /* Tn */
298: { NULL, 0 }, /* Ux */
299: { NULL, 0 }, /* Xc */
300: { NULL, 0 }, /* Xo */
1.128 schwarze 301: { parse_mdoc_Fo, 0 }, /* Fo */
1.47 schwarze 302: { NULL, 0 }, /* Fc */
303: { NULL, 0 }, /* Oo */
304: { NULL, 0 }, /* Oc */
305: { NULL, 0 }, /* Bk */
306: { NULL, 0 }, /* Ek */
307: { NULL, 0 }, /* Bt */
308: { NULL, 0 }, /* Hf */
309: { NULL, 0 }, /* Fr */
310: { NULL, 0 }, /* Ud */
311: { NULL, TYPE_Lb }, /* Lb */
312: { NULL, 0 }, /* Lp */
313: { NULL, TYPE_Lk }, /* Lk */
314: { NULL, TYPE_Mt }, /* Mt */
315: { NULL, 0 }, /* Brq */
316: { NULL, 0 }, /* Bro */
317: { NULL, 0 }, /* Brc */
318: { NULL, 0 }, /* %C */
319: { NULL, 0 }, /* Es */
320: { NULL, 0 }, /* En */
321: { NULL, TYPE_Dx }, /* Dx */
322: { NULL, 0 }, /* %Q */
323: { NULL, 0 }, /* br */
324: { NULL, 0 }, /* sp */
325: { NULL, 0 }, /* %U */
326: { NULL, 0 }, /* Ta */
1.121 schwarze 327: { NULL, 0 }, /* ll */
1.1 schwarze 328: };
329:
1.98 schwarze 330:
1.1 schwarze 331: int
1.3 schwarze 332: mandocdb(int argc, char *argv[])
1.1 schwarze 333: {
1.140 schwarze 334: struct manconf conf;
335: struct ohash_info mpages_info, mlinks_info;
336: struct mparse *mp;
337: const char *path_arg;
338: size_t j, sz;
1.47 schwarze 339: int ch, i;
340:
1.140 schwarze 341: memset(&conf, 0, sizeof(conf));
1.47 schwarze 342: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
343:
344: mpages_info.alloc = mlinks_info.alloc = hash_alloc;
1.106 espie 345: mpages_info.calloc = mlinks_info.calloc = hash_calloc;
1.135 schwarze 346: mpages_info.free = mlinks_info.free = hash_free;
347: mpages_info.data = mlinks_info.data = NULL;
1.47 schwarze 348:
349: mpages_info.key_offset = offsetof(struct mpage, inodev);
350: mlinks_info.key_offset = offsetof(struct mlink, file);
1.1 schwarze 351:
1.47 schwarze 352: /*
1.98 schwarze 353: * We accept a few different invocations.
1.47 schwarze 354: * The CHECKOP macro makes sure that invocation styles don't
355: * clobber each other.
356: */
357: #define CHECKOP(_op, _ch) do \
358: if (OP_DEFAULT != (_op)) { \
1.151 ! schwarze 359: warnx("-%c: Conflicting option", (_ch)); \
1.47 schwarze 360: goto usage; \
361: } while (/*CONSTCOND*/0)
1.10 schwarze 362:
1.47 schwarze 363: path_arg = NULL;
1.28 schwarze 364: op = OP_DEFAULT;
1.1 schwarze 365:
1.83 schwarze 366: while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
1.1 schwarze 367: switch (ch) {
1.98 schwarze 368: case 'a':
1.6 schwarze 369: use_all = 1;
370: break;
1.98 schwarze 371: case 'C':
1.47 schwarze 372: CHECKOP(op, ch);
373: path_arg = optarg;
1.28 schwarze 374: op = OP_CONFFILE;
1.25 schwarze 375: break;
1.98 schwarze 376: case 'D':
1.82 schwarze 377: debug++;
378: break;
1.98 schwarze 379: case 'd':
1.47 schwarze 380: CHECKOP(op, ch);
381: path_arg = optarg;
1.2 schwarze 382: op = OP_UPDATE;
1.1 schwarze 383: break;
1.98 schwarze 384: case 'n':
1.47 schwarze 385: nodb = 1;
386: break;
1.98 schwarze 387: case 'p':
1.83 schwarze 388: warnings = 1;
389: break;
1.98 schwarze 390: case 'Q':
1.73 schwarze 391: mparse_options |= MPARSE_QUICK;
1.59 schwarze 392: break;
1.98 schwarze 393: case 'T':
1.52 schwarze 394: if (strcmp(optarg, "utf8")) {
1.151 ! schwarze 395: warnx("-T%s: Unsupported output format",
! 396: optarg);
1.52 schwarze 397: goto usage;
398: }
399: write_utf8 = 1;
400: break;
1.98 schwarze 401: case 't':
1.47 schwarze 402: CHECKOP(op, ch);
1.28 schwarze 403: dup2(STDOUT_FILENO, STDERR_FILENO);
404: op = OP_TEST;
1.47 schwarze 405: nodb = warnings = 1;
1.28 schwarze 406: break;
1.98 schwarze 407: case 'u':
1.47 schwarze 408: CHECKOP(op, ch);
409: path_arg = optarg;
1.1 schwarze 410: op = OP_DELETE;
411: break;
1.98 schwarze 412: case 'v':
1.83 schwarze 413: /* Compatibility with espie@'s makewhatis. */
1.28 schwarze 414: break;
1.1 schwarze 415: default:
1.28 schwarze 416: goto usage;
1.1 schwarze 417: }
418:
419: argc -= optind;
420: argv += optind;
421:
1.28 schwarze 422: if (OP_CONFFILE == op && argc > 0) {
1.151 ! schwarze 423: warnx("-C: Too many arguments");
1.28 schwarze 424: goto usage;
425: }
426:
1.47 schwarze 427: exitcode = (int)MANDOCLEVEL_OK;
1.131 schwarze 428: mchars = mchars_alloc();
1.137 schwarze 429: mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL,
1.131 schwarze 430: mchars, NULL);
1.47 schwarze 431: ohash_init(&mpages, 6, &mpages_info);
432: ohash_init(&mlinks, 6, &mlinks_info);
1.2 schwarze 433:
1.47 schwarze 434: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.2 schwarze 435:
1.47 schwarze 436: /*
1.109 schwarze 437: * Most of these deal with a specific directory.
1.97 schwarze 438: * Jump into that directory first.
1.47 schwarze 439: */
1.120 schwarze 440: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.47 schwarze 441: goto out;
1.97 schwarze 442:
1.95 schwarze 443: if (dbopen(1)) {
1.97 schwarze 444: /*
445: * The existing database is usable. Process
446: * all files specified on the command-line.
447: */
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) {
502: ohash_init(&mpages, 6, &mpages_info);
503: ohash_init(&mlinks, 6, &mlinks_info);
504: }
1.1 schwarze 505:
1.140 schwarze 506: if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
1.120 schwarze 507: continue;
1.47 schwarze 508: if (0 == treescan())
1.120 schwarze 509: continue;
1.47 schwarze 510: if (0 == dbopen(0))
1.120 schwarze 511: continue;
1.41 deraadt 512:
1.131 schwarze 513: mpages_merge(mp);
1.104 schwarze 514: if (warnings && !nodb &&
1.87 schwarze 515: ! (MPARSE_QUICK & mparse_options))
516: names_check();
1.47 schwarze 517: dbclose(0);
1.28 schwarze 518:
1.140 schwarze 519: if (j + 1 < conf.manpath.sz) {
1.47 schwarze 520: mpages_free();
521: ohash_delete(&mpages);
522: ohash_delete(&mlinks);
523: }
1.2 schwarze 524: }
1.47 schwarze 525: }
526: out:
1.140 schwarze 527: manconf_free(&conf);
1.123 schwarze 528: mparse_free(mp);
1.131 schwarze 529: mchars_free(mchars);
1.47 schwarze 530: mpages_free();
531: ohash_delete(&mpages);
532: ohash_delete(&mlinks);
1.150 schwarze 533: return exitcode;
1.47 schwarze 534: usage:
1.83 schwarze 535: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
536: " %s [-aDnpQ] [-Tutf8] dir ...\n"
537: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
538: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 539: " %s [-Q] -t file ...\n",
1.151 ! schwarze 540: __progname, __progname, __progname,
! 541: __progname, __progname);
1.1 schwarze 542:
1.150 schwarze 543: return (int)MANDOCLEVEL_BADARG;
1.47 schwarze 544: }
1.1 schwarze 545:
1.47 schwarze 546: /*
547: * Scan a directory tree rooted at "basedir" for manpages.
548: * We use fts(), scanning directory parts along the way for clues to our
549: * section and architecture.
550: *
551: * If use_all has been specified, grok all files.
552: * If not, sanitise paths to the following:
553: *
1.98 schwarze 554: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 555: * or
556: * [./]cat<section>[/<arch>]/<name>.0
557: *
558: * TODO: accomodate for multi-language directories.
559: */
560: static int
561: treescan(void)
562: {
1.96 schwarze 563: char buf[PATH_MAX];
1.47 schwarze 564: FTS *f;
565: FTSENT *ff;
566: struct mlink *mlink;
1.81 schwarze 567: int dform, gzip;
1.51 schwarze 568: char *dsec, *arch, *fsec, *cp;
569: const char *path;
1.47 schwarze 570: const char *argv[2];
1.1 schwarze 571:
1.47 schwarze 572: argv[0] = ".";
573: argv[1] = (char *)NULL;
1.1 schwarze 574:
1.96 schwarze 575: f = fts_open((char * const *)argv,
576: FTS_PHYSICAL | FTS_NOCHDIR, NULL);
1.47 schwarze 577: if (NULL == f) {
578: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 579: say("", "&fts_open");
1.150 schwarze 580: return 0;
1.47 schwarze 581: }
1.2 schwarze 582:
1.47 schwarze 583: dsec = arch = NULL;
584: dform = FORM_NONE;
1.2 schwarze 585:
1.47 schwarze 586: while (NULL != (ff = fts_read(f))) {
587: path = ff->fts_path + 2;
1.96 schwarze 588: switch (ff->fts_info) {
589:
590: /*
591: * Symbolic links require various sanity checks,
592: * then get handled just like regular files.
593: */
1.98 schwarze 594: case FTS_SL:
1.96 schwarze 595: if (NULL == realpath(path, buf)) {
596: if (warnings)
597: say(path, "&realpath");
598: continue;
599: }
600: if (strstr(buf, basedir) != buf) {
601: if (warnings) say("",
602: "%s: outside base directory", buf);
603: continue;
604: }
605: /* Use logical inode to avoid mpages dupe. */
606: if (-1 == stat(path, ff->fts_statp)) {
607: if (warnings)
608: say(path, "&stat");
609: continue;
610: }
611: /* FALLTHROUGH */
612:
1.15 schwarze 613: /*
1.47 schwarze 614: * If we're a regular file, add an mlink by using the
615: * stored directory data and handling the filename.
1.15 schwarze 616: */
1.98 schwarze 617: case FTS_F:
1.47 schwarze 618: if (0 == strcmp(path, MANDOC_DB))
619: continue;
620: if ( ! use_all && ff->fts_level < 2) {
621: if (warnings)
622: say(path, "Extraneous file");
623: continue;
1.81 schwarze 624: }
625: gzip = 0;
626: fsec = NULL;
627: while (NULL == fsec) {
628: fsec = strrchr(ff->fts_name, '.');
629: if (NULL == fsec || strcmp(fsec+1, "gz"))
630: break;
631: gzip = 1;
632: *fsec = '\0';
633: fsec = NULL;
634: }
635: if (NULL == fsec) {
1.47 schwarze 636: if ( ! use_all) {
637: if (warnings)
638: say(path,
639: "No filename suffix");
640: continue;
641: }
642: } else if (0 == strcmp(++fsec, "html")) {
643: if (warnings)
644: say(path, "Skip html");
645: continue;
646: } else if (0 == strcmp(fsec, "ps")) {
647: if (warnings)
648: say(path, "Skip ps");
649: continue;
650: } else if (0 == strcmp(fsec, "pdf")) {
651: if (warnings)
652: say(path, "Skip pdf");
653: continue;
654: } else if ( ! use_all &&
1.138 schwarze 655: ((FORM_SRC == dform &&
656: strncmp(fsec, dsec, strlen(dsec))) ||
1.47 schwarze 657: (FORM_CAT == dform && strcmp(fsec, "0")))) {
658: if (warnings)
659: say(path, "Wrong filename suffix");
660: continue;
661: } else
662: fsec[-1] = '\0';
1.51 schwarze 663:
1.47 schwarze 664: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.100 schwarze 665: if (strlcpy(mlink->file, path,
666: sizeof(mlink->file)) >=
667: sizeof(mlink->file)) {
668: say(path, "Filename too long");
669: free(mlink);
670: continue;
671: }
1.47 schwarze 672: mlink->dform = dform;
1.51 schwarze 673: mlink->dsec = dsec;
674: mlink->arch = arch;
675: mlink->name = ff->fts_name;
676: mlink->fsec = fsec;
1.81 schwarze 677: mlink->gzip = gzip;
1.47 schwarze 678: mlink_add(mlink, ff->fts_statp);
679: continue;
1.96 schwarze 680:
1.98 schwarze 681: case FTS_D:
1.96 schwarze 682: /* FALLTHROUGH */
1.98 schwarze 683: case FTS_DP:
1.96 schwarze 684: break;
685:
686: default:
1.47 schwarze 687: if (warnings)
688: say(path, "Not a regular file");
689: continue;
690: }
691:
692: switch (ff->fts_level) {
1.98 schwarze 693: case 0:
1.47 schwarze 694: /* Ignore the root directory. */
695: break;
1.98 schwarze 696: case 1:
1.47 schwarze 697: /*
698: * This might contain manX/ or catX/.
699: * Try to infer this from the name.
700: * If we're not in use_all, enforce it.
701: */
702: cp = ff->fts_name;
703: if (FTS_DP == ff->fts_info)
704: break;
1.15 schwarze 705:
1.47 schwarze 706: if (0 == strncmp(cp, "man", 3)) {
707: dform = FORM_SRC;
708: dsec = cp + 3;
709: } else if (0 == strncmp(cp, "cat", 3)) {
710: dform = FORM_CAT;
711: dsec = cp + 3;
1.51 schwarze 712: } else {
713: dform = FORM_NONE;
714: dsec = NULL;
1.26 schwarze 715: }
1.47 schwarze 716:
1.98 schwarze 717: if (NULL != dsec || use_all)
1.47 schwarze 718: break;
719:
720: if (warnings)
721: say(path, "Unknown directory part");
722: fts_set(f, ff, FTS_SKIP);
723: break;
1.98 schwarze 724: case 2:
1.47 schwarze 725: /*
726: * Possibly our architecture.
727: * If we're descending, keep tabs on it.
728: */
729: if (FTS_DP != ff->fts_info && NULL != dsec)
730: arch = ff->fts_name;
1.51 schwarze 731: else
732: arch = NULL;
1.47 schwarze 733: break;
734: default:
735: if (FTS_DP == ff->fts_info || use_all)
736: break;
737: if (warnings)
738: say(path, "Extraneous directory part");
739: fts_set(f, ff, FTS_SKIP);
740: break;
1.14 schwarze 741: }
1.47 schwarze 742: }
743:
744: fts_close(f);
1.150 schwarze 745: return 1;
1.47 schwarze 746: }
1.1 schwarze 747:
1.47 schwarze 748: /*
749: * Add a file to the mlinks table.
750: * Do not verify that it's a "valid" looking manpage (we'll do that
751: * later).
752: *
753: * Try to infer the manual section, architecture, and page name from the
754: * path, assuming it looks like
755: *
1.98 schwarze 756: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 757: * or
758: * [./]cat<section>[/<arch>]/<name>.0
759: *
760: * See treescan() for the fts(3) version of this.
761: */
762: static void
763: filescan(const char *file)
764: {
765: char buf[PATH_MAX];
766: struct stat st;
767: struct mlink *mlink;
768: char *p, *start;
769:
770: assert(use_all);
771:
772: if (0 == strncmp(file, "./", 2))
773: file += 2;
774:
1.96 schwarze 775: /*
776: * We have to do lstat(2) before realpath(3) loses
777: * the information whether this is a symbolic link.
778: * We need to know that because for symbolic links,
779: * we want to use the orginal file name, while for
780: * regular files, we want to use the real path.
781: */
782: if (-1 == lstat(file, &st)) {
783: exitcode = (int)MANDOCLEVEL_BADARG;
784: say(file, "&lstat");
785: return;
786: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
787: exitcode = (int)MANDOCLEVEL_BADARG;
788: say(file, "Not a regular file");
789: return;
790: }
791:
792: /*
793: * We have to resolve the file name to the real path
794: * in any case for the base directory check.
795: */
1.47 schwarze 796: if (NULL == realpath(file, buf)) {
797: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 798: say(file, "&realpath");
1.47 schwarze 799: return;
1.63 schwarze 800: }
801:
1.109 schwarze 802: if (OP_TEST == op)
1.63 schwarze 803: start = buf;
1.109 schwarze 804: else if (strstr(buf, basedir) == buf)
805: start = buf + strlen(basedir);
1.63 schwarze 806: else {
1.47 schwarze 807: exitcode = (int)MANDOCLEVEL_BADARG;
808: say("", "%s: outside base directory", buf);
809: return;
1.63 schwarze 810: }
811:
1.96 schwarze 812: /*
813: * Now we are sure the file is inside our tree.
814: * If it is a symbolic link, ignore the real path
815: * and use the original name.
816: * This implies passing stuff like "cat1/../man1/foo.1"
817: * on the command line won't work. So don't do that.
818: * Note the stat(2) can still fail if the link target
819: * doesn't exist.
820: */
821: if (S_IFLNK & st.st_mode) {
822: if (-1 == stat(buf, &st)) {
823: exitcode = (int)MANDOCLEVEL_BADARG;
824: say(file, "&stat");
825: return;
826: }
1.100 schwarze 827: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
828: say(file, "Filename too long");
829: return;
830: }
1.109 schwarze 831: start = buf;
832: if (OP_TEST != op && strstr(buf, basedir) == buf)
833: start += strlen(basedir);
1.1 schwarze 834: }
1.63 schwarze 835:
1.47 schwarze 836: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 837: mlink->dform = FORM_NONE;
1.100 schwarze 838: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
839: sizeof(mlink->file)) {
840: say(start, "Filename too long");
1.134 schwarze 841: free(mlink);
1.100 schwarze 842: return;
843: }
1.1 schwarze 844:
1.10 schwarze 845: /*
1.47 schwarze 846: * First try to guess our directory structure.
847: * If we find a separator, try to look for man* or cat*.
848: * If we find one of these and what's underneath is a directory,
849: * assume it's an architecture.
1.10 schwarze 850: */
1.47 schwarze 851: if (NULL != (p = strchr(start, '/'))) {
852: *p++ = '\0';
853: if (0 == strncmp(start, "man", 3)) {
854: mlink->dform = FORM_SRC;
1.51 schwarze 855: mlink->dsec = start + 3;
1.47 schwarze 856: } else if (0 == strncmp(start, "cat", 3)) {
857: mlink->dform = FORM_CAT;
1.51 schwarze 858: mlink->dsec = start + 3;
1.47 schwarze 859: }
1.10 schwarze 860:
1.47 schwarze 861: start = p;
862: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
863: *p++ = '\0';
1.51 schwarze 864: mlink->arch = start;
1.47 schwarze 865: start = p;
1.41 deraadt 866: }
1.47 schwarze 867: }
1.7 schwarze 868:
1.47 schwarze 869: /*
870: * Now check the file suffix.
871: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
872: */
873: p = strrchr(start, '\0');
874: while (p-- > start && '/' != *p && '.' != *p)
875: /* Loop. */ ;
876:
877: if ('.' == *p) {
878: *p++ = '\0';
1.51 schwarze 879: mlink->fsec = p;
1.47 schwarze 880: }
1.41 deraadt 881:
1.47 schwarze 882: /*
883: * Now try to parse the name.
884: * Use the filename portion of the path.
885: */
886: mlink->name = start;
887: if (NULL != (p = strrchr(start, '/'))) {
888: mlink->name = p + 1;
889: *p = '\0';
890: }
891: mlink_add(mlink, &st);
892: }
1.2 schwarze 893:
1.47 schwarze 894: static void
895: mlink_add(struct mlink *mlink, const struct stat *st)
896: {
897: struct inodev inodev;
898: struct mpage *mpage;
899: unsigned int slot;
900:
901: assert(NULL != mlink->file);
902:
1.51 schwarze 903: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
904: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
905: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
906: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 907:
908: if ('0' == *mlink->fsec) {
909: free(mlink->fsec);
910: mlink->fsec = mandoc_strdup(mlink->dsec);
911: mlink->fform = FORM_CAT;
912: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
913: mlink->fform = FORM_SRC;
914: else
915: mlink->fform = FORM_NONE;
1.1 schwarze 916:
1.47 schwarze 917: slot = ohash_qlookup(&mlinks, mlink->file);
918: assert(NULL == ohash_find(&mlinks, slot));
919: ohash_insert(&mlinks, slot, mlink);
920:
1.139 schwarze 921: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.47 schwarze 922: inodev.st_ino = st->st_ino;
923: inodev.st_dev = st->st_dev;
924: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
925: sizeof(struct inodev), inodev.st_ino);
926: mpage = ohash_find(&mpages, slot);
927: if (NULL == mpage) {
928: mpage = mandoc_calloc(1, sizeof(struct mpage));
929: mpage->inodev.st_ino = inodev.st_ino;
930: mpage->inodev.st_dev = inodev.st_dev;
931: ohash_insert(&mpages, slot, mpage);
932: } else
933: mlink->next = mpage->mlinks;
934: mpage->mlinks = mlink;
1.75 schwarze 935: mlink->mpage = mpage;
1.47 schwarze 936: }
1.1 schwarze 937:
1.47 schwarze 938: static void
939: mlink_free(struct mlink *mlink)
940: {
1.1 schwarze 941:
1.47 schwarze 942: free(mlink->dsec);
943: free(mlink->arch);
944: free(mlink->name);
945: free(mlink->fsec);
946: free(mlink);
947: }
1.1 schwarze 948:
1.47 schwarze 949: static void
950: mpages_free(void)
951: {
952: struct mpage *mpage;
953: struct mlink *mlink;
954: unsigned int slot;
955:
956: mpage = ohash_first(&mpages, &slot);
957: while (NULL != mpage) {
958: while (NULL != (mlink = mpage->mlinks)) {
959: mpage->mlinks = mlink->next;
960: mlink_free(mlink);
961: }
962: free(mpage->sec);
963: free(mpage->arch);
964: free(mpage->title);
965: free(mpage->desc);
966: free(mpage);
967: mpage = ohash_next(&mpages, &slot);
968: }
969: }
1.1 schwarze 970:
1.47 schwarze 971: /*
972: * For each mlink to the mpage, check whether the path looks like
973: * it is formatted, and if it does, check whether a source manual
974: * exists by the same name, ignoring the suffix.
975: * If both conditions hold, drop the mlink.
976: */
977: static void
978: mlinks_undupe(struct mpage *mpage)
979: {
980: char buf[PATH_MAX];
981: struct mlink **prev;
982: struct mlink *mlink;
983: char *bufp;
984:
985: mpage->form = FORM_CAT;
986: prev = &mpage->mlinks;
987: while (NULL != (mlink = *prev)) {
988: if (FORM_CAT != mlink->dform) {
989: mpage->form = FORM_NONE;
990: goto nextlink;
991: }
1.100 schwarze 992: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 993: bufp = strstr(buf, "cat");
994: assert(NULL != bufp);
995: memcpy(bufp, "man", 3);
996: if (NULL != (bufp = strrchr(buf, '.')))
997: *++bufp = '\0';
1.100 schwarze 998: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 999: if (NULL == ohash_find(&mlinks,
1.98 schwarze 1000: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 1001: goto nextlink;
1002: if (warnings)
1003: say(mlink->file, "Man source exists: %s", buf);
1004: if (use_all)
1005: goto nextlink;
1006: *prev = mlink->next;
1007: mlink_free(mlink);
1008: continue;
1009: nextlink:
1010: prev = &(*prev)->next;
1.1 schwarze 1011: }
1.47 schwarze 1012: }
1.1 schwarze 1013:
1.88 schwarze 1014: static void
1.50 schwarze 1015: mlink_check(struct mpage *mpage, struct mlink *mlink)
1016: {
1.88 schwarze 1017: struct str *str;
1018: unsigned int slot;
1.50 schwarze 1019:
1020: /*
1021: * Check whether the manual section given in a file
1022: * agrees with the directory where the file is located.
1023: * Some manuals have suffixes like (3p) on their
1024: * section number either inside the file or in the
1025: * directory name, some are linked into more than one
1026: * section, like encrypt(1) = makekey(8).
1027: */
1028:
1029: if (FORM_SRC == mpage->form &&
1.88 schwarze 1030: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1031: say(mlink->file, "Section \"%s\" manual in %s directory",
1032: mpage->sec, mlink->dsec);
1033:
1034: /*
1035: * Manual page directories exist for each kernel
1036: * architecture as returned by machine(1).
1037: * However, many manuals only depend on the
1038: * application architecture as returned by arch(1).
1039: * For example, some (2/ARM) manuals are shared
1040: * across the "armish" and "zaurus" kernel
1041: * architectures.
1042: * A few manuals are even shared across completely
1043: * different architectures, for example fdformat(1)
1044: * on amd64, i386, sparc, and sparc64.
1045: */
1046:
1.88 schwarze 1047: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1048: say(mlink->file, "Architecture \"%s\" manual in "
1049: "\"%s\" directory", mpage->arch, mlink->arch);
1050:
1.88 schwarze 1051: /*
1052: * XXX
1.90 schwarze 1053: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1054: */
1055:
1056: if (FORM_CAT == mpage->form)
1057: return;
1058:
1059: /*
1060: * Check whether this mlink
1061: * appears as a name in the NAME section.
1062: */
1.50 schwarze 1063:
1.90 schwarze 1064: slot = ohash_qlookup(&names, mlink->name);
1065: str = ohash_find(&names, slot);
1.88 schwarze 1066: assert(NULL != str);
1.90 schwarze 1067: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1068: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1069: }
1070:
1.47 schwarze 1071: /*
1072: * Run through the files in the global vector "mpages"
1073: * and add them to the database specified in "basedir".
1074: *
1075: * This handles the parsing scheme itself, using the cues of directory
1076: * and filename to determine whether the file is parsable or not.
1077: */
1078: static void
1.131 schwarze 1079: mpages_merge(struct mparse *mp)
1.47 schwarze 1080: {
1.70 schwarze 1081: char any[] = "any";
1.58 schwarze 1082: struct ohash_info str_info;
1.75 schwarze 1083: struct mpage *mpage, *mpage_dest;
1084: struct mlink *mlink, *mlink_dest;
1.144 schwarze 1085: struct roff_man *man;
1.75 schwarze 1086: char *sodest;
1.69 schwarze 1087: char *cp;
1.117 schwarze 1088: int fd;
1.58 schwarze 1089: unsigned int pslot;
1.47 schwarze 1090:
1091: str_info.alloc = hash_alloc;
1.106 espie 1092: str_info.calloc = hash_calloc;
1093: str_info.free = hash_free;
1.135 schwarze 1094: str_info.data = NULL;
1.47 schwarze 1095: str_info.key_offset = offsetof(struct str, key);
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;
1109: ohash_init(&names, 4, &str_info);
1.47 schwarze 1110: ohash_init(&strings, 6, &str_info);
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) {
1.47 schwarze 1172: mpage->form = FORM_SRC;
1.146 schwarze 1173: mpage->sec = man->meta.msec;
1.112 schwarze 1174: mpage->sec = mandoc_strdup(
1.126 schwarze 1175: mpage->sec == NULL ? "" : mpage->sec);
1.146 schwarze 1176: mpage->arch = man->meta.arch;
1.47 schwarze 1177: mpage->arch = mandoc_strdup(
1.126 schwarze 1178: mpage->arch == NULL ? "" : mpage->arch);
1.146 schwarze 1179: mpage->title = mandoc_strdup(man->meta.title);
1.145 schwarze 1180: } else if (man != NULL && man->macroset == MACROSET_MAN) {
1.47 schwarze 1181: mpage->form = FORM_SRC;
1.146 schwarze 1182: mpage->sec = mandoc_strdup(man->meta.msec);
1.136 schwarze 1183: mpage->arch = mandoc_strdup(mlink->arch);
1.146 schwarze 1184: mpage->title = mandoc_strdup(man->meta.title);
1.11 schwarze 1185: } else {
1.47 schwarze 1186: mpage->form = FORM_CAT;
1.136 schwarze 1187: mpage->sec = mandoc_strdup(mlink->dsec);
1188: mpage->arch = mandoc_strdup(mlink->arch);
1189: mpage->title = mandoc_strdup(mlink->name);
1.1 schwarze 1190: }
1.54 schwarze 1191: putkey(mpage, mpage->sec, TYPE_sec);
1.119 schwarze 1192: if (*mpage->arch != '\0')
1193: putkey(mpage, mpage->arch, TYPE_arch);
1.1 schwarze 1194:
1.136 schwarze 1195: for ( ; mlink != NULL; mlink = mlink->next) {
1.54 schwarze 1196: if ('\0' != *mlink->dsec)
1197: putkey(mpage, mlink->dsec, TYPE_sec);
1198: if ('\0' != *mlink->fsec)
1199: putkey(mpage, mlink->fsec, TYPE_sec);
1.55 schwarze 1200: putkey(mpage, '\0' == *mlink->arch ?
1.70 schwarze 1201: any : mlink->arch, TYPE_arch);
1.90 schwarze 1202: putkey(mpage, mlink->name, NAME_FILE);
1.54 schwarze 1203: }
1.41 deraadt 1204:
1.129 schwarze 1205: assert(mpage->desc == NULL);
1.145 schwarze 1206: if (man != NULL && man->macroset == MACROSET_MDOC)
1.146 schwarze 1207: parse_mdoc(mpage, &man->meta, man->first);
1.129 schwarze 1208: else if (man != NULL)
1.146 schwarze 1209: parse_man(mpage, &man->meta, man->first);
1.47 schwarze 1210: else
1.117 schwarze 1211: parse_cat(mpage, fd);
1.129 schwarze 1212: if (mpage->desc == NULL)
1.92 schwarze 1213: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1.88 schwarze 1214:
1215: if (warnings && !use_all)
1216: for (mlink = mpage->mlinks; mlink;
1217: mlink = mlink->next)
1218: mlink_check(mpage, mlink);
1.6 schwarze 1219:
1.131 schwarze 1220: dbadd(mpage);
1.136 schwarze 1221: mlink = mpage->mlinks;
1.81 schwarze 1222:
1223: nextpage:
1.47 schwarze 1224: ohash_delete(&strings);
1.90 schwarze 1225: ohash_delete(&names);
1.47 schwarze 1226: mpage = ohash_next(&mpages, &pslot);
1227: }
1.64 schwarze 1228:
1229: if (0 == nodb)
1230: SQL_EXEC("END TRANSACTION");
1.87 schwarze 1231: }
1232:
1233: static void
1234: names_check(void)
1235: {
1236: sqlite3_stmt *stmt;
1237: const char *name, *sec, *arch, *key;
1238: int irc;
1239:
1240: sqlite3_prepare_v2(db,
1241: "SELECT name, sec, arch, key FROM ("
1.90 schwarze 1242: "SELECT name AS key, pageid FROM names "
1.87 schwarze 1243: "WHERE bits & ? AND NOT EXISTS ("
1244: "SELECT pageid FROM mlinks "
1.90 schwarze 1245: "WHERE mlinks.pageid == names.pageid "
1246: "AND mlinks.name == names.name"
1.87 schwarze 1247: ")"
1248: ") JOIN ("
1.99 schwarze 1249: "SELECT sec, arch, name, pageid FROM mlinks "
1250: "GROUP BY pageid"
1.87 schwarze 1251: ") USING (pageid);",
1252: -1, &stmt, NULL);
1253:
1.91 schwarze 1254: if (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE))
1255: say("", "%s", sqlite3_errmsg(db));
1.87 schwarze 1256:
1257: while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1.113 schwarze 1258: name = (const char *)sqlite3_column_text(stmt, 0);
1259: sec = (const char *)sqlite3_column_text(stmt, 1);
1260: arch = (const char *)sqlite3_column_text(stmt, 2);
1261: key = (const char *)sqlite3_column_text(stmt, 3);
1.87 schwarze 1262: say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1263: '\0' == *arch ? "" : "/",
1264: '\0' == *arch ? "" : arch, key);
1265: }
1266: sqlite3_finalize(stmt);
1.47 schwarze 1267: }
1.6 schwarze 1268:
1.47 schwarze 1269: static void
1.81 schwarze 1270: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1271: {
1272: FILE *stream;
1273: char *line, *p, *title;
1274: size_t len, plen, titlesz;
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:
1.47 schwarze 1287: /* Skip to first blank line. */
1.1 schwarze 1288:
1.47 schwarze 1289: while (NULL != (line = fgetln(stream, &len)))
1290: if ('\n' == *line)
1291: break;
1.1 schwarze 1292:
1.47 schwarze 1293: /*
1294: * Assume the first line that is not indented
1295: * is the first section header. Skip to it.
1296: */
1.1 schwarze 1297:
1.47 schwarze 1298: while (NULL != (line = fgetln(stream, &len)))
1299: if ('\n' != *line && ' ' != *line)
1300: break;
1.98 schwarze 1301:
1.47 schwarze 1302: /*
1303: * Read up until the next section into a buffer.
1304: * Strip the leading and trailing newline from each read line,
1305: * appending a trailing space.
1306: * Ignore empty (whitespace-only) lines.
1307: */
1.28 schwarze 1308:
1.47 schwarze 1309: titlesz = 0;
1310: title = NULL;
1.38 schwarze 1311:
1.47 schwarze 1312: while (NULL != (line = fgetln(stream, &len))) {
1313: if (' ' != *line || '\n' != line[len - 1])
1314: break;
1315: while (len > 0 && isspace((unsigned char)*line)) {
1316: line++;
1317: len--;
1318: }
1319: if (1 == len)
1320: continue;
1321: title = mandoc_realloc(title, titlesz + len);
1322: memcpy(title + titlesz, line, len);
1323: titlesz += len;
1324: title[titlesz - 1] = ' ';
1325: }
1.28 schwarze 1326:
1.47 schwarze 1327: /*
1328: * If no page content can be found, or the input line
1329: * is already the next section header, or there is no
1330: * trailing newline, reuse the page title as the page
1331: * description.
1332: */
1.1 schwarze 1333:
1.47 schwarze 1334: if (NULL == title || '\0' == *title) {
1335: if (warnings)
1336: say(mpage->mlinks->file,
1337: "Cannot find NAME section");
1338: fclose(stream);
1339: free(title);
1340: return;
1341: }
1.24 schwarze 1342:
1.47 schwarze 1343: title = mandoc_realloc(title, titlesz + 1);
1344: title[titlesz] = '\0';
1.24 schwarze 1345:
1.47 schwarze 1346: /*
1347: * Skip to the first dash.
1348: * Use the remaining line as the description (no more than 70
1349: * bytes).
1350: */
1.28 schwarze 1351:
1.47 schwarze 1352: if (NULL != (p = strstr(title, "- "))) {
1353: for (p += 2; ' ' == *p || '\b' == *p; p++)
1354: /* Skip to next word. */ ;
1355: } else {
1356: if (warnings)
1357: say(mpage->mlinks->file,
1358: "No dash in title line");
1359: p = title;
1360: }
1.1 schwarze 1361:
1.47 schwarze 1362: plen = strlen(p);
1.1 schwarze 1363:
1.47 schwarze 1364: /* Strip backspace-encoding from line. */
1.1 schwarze 1365:
1.47 schwarze 1366: while (NULL != (line = memchr(p, '\b', plen))) {
1367: len = line - p;
1368: if (0 == len) {
1369: memmove(line, line + 1, plen--);
1370: continue;
1.98 schwarze 1371: }
1.47 schwarze 1372: memmove(line - 1, line + 1, plen - len);
1373: plen -= 2;
1374: }
1.1 schwarze 1375:
1.47 schwarze 1376: mpage->desc = mandoc_strdup(p);
1377: fclose(stream);
1378: free(title);
1379: }
1.16 schwarze 1380:
1.47 schwarze 1381: /*
1382: * Put a type/word pair into the word database for this particular file.
1383: */
1384: static void
1.69 schwarze 1385: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1386: {
1.69 schwarze 1387: char *cp;
1.37 schwarze 1388:
1.47 schwarze 1389: assert(NULL != value);
1.69 schwarze 1390: if (TYPE_arch == type)
1391: for (cp = value; *cp; cp++)
1392: if (isupper((unsigned char)*cp))
1393: *cp = _tolower((unsigned char)*cp);
1.47 schwarze 1394: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1395: }
1396:
1397: /*
1.47 schwarze 1398: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1399: */
1400: static void
1.47 schwarze 1401: putmdockey(const struct mpage *mpage,
1.142 schwarze 1402: const struct roff_node *n, uint64_t m)
1.2 schwarze 1403: {
1.16 schwarze 1404:
1.47 schwarze 1405: for ( ; NULL != n; n = n->next) {
1406: if (NULL != n->child)
1407: putmdockey(mpage, n->child, m);
1.141 schwarze 1408: if (n->type == ROFFT_TEXT)
1.47 schwarze 1409: putkey(mpage, n->string, m);
1410: }
1411: }
1.16 schwarze 1412:
1.47 schwarze 1413: static void
1.143 schwarze 1414: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1415: const struct roff_node *n)
1.47 schwarze 1416: {
1.142 schwarze 1417: const struct roff_node *head, *body;
1.78 schwarze 1418: char *start, *title;
1.47 schwarze 1419: char byte;
1.78 schwarze 1420: size_t sz;
1.16 schwarze 1421:
1.47 schwarze 1422: if (NULL == n)
1423: return;
1.16 schwarze 1424:
1.47 schwarze 1425: /*
1426: * We're only searching for one thing: the first text child in
1427: * the BODY of a NAME section. Since we don't keep track of
1428: * sections in -man, run some hoops to find out whether we're in
1429: * the correct section or not.
1430: */
1.16 schwarze 1431:
1.141 schwarze 1432: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1433: body = n;
1434: assert(body->parent);
1435: if (NULL != (head = body->parent->head) &&
1.98 schwarze 1436: 1 == head->nchild &&
1437: NULL != (head = (head->child)) &&
1.141 schwarze 1438: head->type == ROFFT_TEXT &&
1.98 schwarze 1439: 0 == strcmp(head->string, "NAME") &&
1440: NULL != body->child) {
1.2 schwarze 1441:
1.47 schwarze 1442: /*
1443: * Suck the entire NAME section into memory.
1444: * Yes, we might run away.
1445: * But too many manuals have big, spread-out
1446: * NAME sections over many lines.
1447: */
1.2 schwarze 1448:
1.78 schwarze 1449: title = NULL;
1.147 schwarze 1450: deroff(&title, body);
1.47 schwarze 1451: if (NULL == title)
1452: return;
1.16 schwarze 1453:
1.98 schwarze 1454: /*
1.47 schwarze 1455: * Go through a special heuristic dance here.
1456: * Conventionally, one or more manual names are
1457: * comma-specified prior to a whitespace, then a
1458: * dash, then a description. Try to puzzle out
1459: * the name parts here.
1460: */
1.16 schwarze 1461:
1.78 schwarze 1462: start = title;
1.47 schwarze 1463: for ( ;; ) {
1464: sz = strcspn(start, " ,");
1465: if ('\0' == start[sz])
1466: break;
1.1 schwarze 1467:
1.47 schwarze 1468: byte = start[sz];
1469: start[sz] = '\0';
1.67 schwarze 1470:
1471: /*
1472: * Assume a stray trailing comma in the
1473: * name list if a name begins with a dash.
1474: */
1475:
1476: if ('-' == start[0] ||
1477: ('\\' == start[0] && '-' == start[1]))
1478: break;
1.1 schwarze 1479:
1.90 schwarze 1480: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1481: if ( ! (mpage->name_head_done ||
1482: strcasecmp(start, meta->title))) {
1483: putkey(mpage, start, NAME_HEAD);
1484: mpage->name_head_done = 1;
1485: }
1.1 schwarze 1486:
1.47 schwarze 1487: if (' ' == byte) {
1488: start += sz + 1;
1489: break;
1490: }
1.1 schwarze 1491:
1.47 schwarze 1492: assert(',' == byte);
1493: start += sz + 1;
1494: while (' ' == *start)
1495: start++;
1496: }
1.1 schwarze 1497:
1.78 schwarze 1498: if (start == title) {
1.90 schwarze 1499: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1500: if ( ! (mpage->name_head_done ||
1501: strcasecmp(start, meta->title))) {
1502: putkey(mpage, start, NAME_HEAD);
1503: mpage->name_head_done = 1;
1504: }
1.47 schwarze 1505: free(title);
1506: return;
1507: }
1.1 schwarze 1508:
1.47 schwarze 1509: while (isspace((unsigned char)*start))
1510: start++;
1.1 schwarze 1511:
1.47 schwarze 1512: if (0 == strncmp(start, "-", 1))
1513: start += 1;
1514: else if (0 == strncmp(start, "\\-\\-", 4))
1515: start += 4;
1516: else if (0 == strncmp(start, "\\-", 2))
1517: start += 2;
1518: else if (0 == strncmp(start, "\\(en", 4))
1519: start += 4;
1520: else if (0 == strncmp(start, "\\(em", 4))
1521: start += 4;
1.1 schwarze 1522:
1.47 schwarze 1523: while (' ' == *start)
1524: start++;
1.1 schwarze 1525:
1.47 schwarze 1526: mpage->desc = mandoc_strdup(start);
1527: free(title);
1528: return;
1529: }
1530: }
1.1 schwarze 1531:
1.47 schwarze 1532: for (n = n->child; n; n = n->next) {
1533: if (NULL != mpage->desc)
1534: break;
1.129 schwarze 1535: parse_man(mpage, meta, n);
1.1 schwarze 1536: }
1537: }
1538:
1539: static void
1.143 schwarze 1540: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1541: const struct roff_node *n)
1.1 schwarze 1542: {
1543:
1.47 schwarze 1544: assert(NULL != n);
1545: for (n = n->child; NULL != n; n = n->next) {
1546: switch (n->type) {
1.141 schwarze 1547: case ROFFT_ELEM:
1.47 schwarze 1548: /* FALLTHROUGH */
1.141 schwarze 1549: case ROFFT_BLOCK:
1.47 schwarze 1550: /* FALLTHROUGH */
1.141 schwarze 1551: case ROFFT_HEAD:
1.47 schwarze 1552: /* FALLTHROUGH */
1.141 schwarze 1553: case ROFFT_BODY:
1.47 schwarze 1554: /* FALLTHROUGH */
1.141 schwarze 1555: case ROFFT_TAIL:
1.47 schwarze 1556: if (NULL != mdocs[n->tok].fp)
1.127 schwarze 1557: if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1.47 schwarze 1558: break;
1559: if (mdocs[n->tok].mask)
1560: putmdockey(mpage, n->child,
1561: mdocs[n->tok].mask);
1562: break;
1563: default:
1.141 schwarze 1564: assert(n->type != ROFFT_ROOT);
1.47 schwarze 1565: continue;
1566: }
1567: if (NULL != n->child)
1.127 schwarze 1568: parse_mdoc(mpage, meta, n);
1.1 schwarze 1569: }
1570: }
1571:
1.19 schwarze 1572: static int
1.143 schwarze 1573: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1574: const struct roff_node *n)
1.1 schwarze 1575: {
1.131 schwarze 1576: char *start, *end;
1.1 schwarze 1577: size_t sz;
1.19 schwarze 1578:
1.47 schwarze 1579: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1580: NULL == (n = n->child) ||
1.141 schwarze 1581: n->type != ROFFT_TEXT)
1.150 schwarze 1582: return 0;
1.1 schwarze 1583:
1584: /*
1585: * Only consider those `Fd' macro fields that begin with an
1586: * "inclusion" token (versus, e.g., #define).
1587: */
1.47 schwarze 1588:
1.1 schwarze 1589: if (strcmp("#include", n->string))
1.150 schwarze 1590: return 0;
1.1 schwarze 1591:
1.141 schwarze 1592: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1593: return 0;
1.1 schwarze 1594:
1595: /*
1596: * Strip away the enclosing angle brackets and make sure we're
1597: * not zero-length.
1598: */
1599:
1600: start = n->string;
1601: if ('<' == *start || '"' == *start)
1602: start++;
1603:
1604: if (0 == (sz = strlen(start)))
1.150 schwarze 1605: return 0;
1.1 schwarze 1606:
1607: end = &start[(int)sz - 1];
1608: if ('>' == *end || '"' == *end)
1609: end--;
1610:
1.47 schwarze 1611: if (end > start)
1612: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1613: return 0;
1.1 schwarze 1614: }
1615:
1.133 schwarze 1616: static void
1.142 schwarze 1617: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1618: {
1.69 schwarze 1619: char *cp;
1.133 schwarze 1620: size_t sz;
1.1 schwarze 1621:
1.141 schwarze 1622: if (n->type != ROFFT_TEXT)
1.133 schwarze 1623: return;
1.19 schwarze 1624:
1.133 schwarze 1625: /* Skip function pointer punctuation. */
1.1 schwarze 1626:
1.133 schwarze 1627: cp = n->string;
1628: while (*cp == '(' || *cp == '*')
1.1 schwarze 1629: cp++;
1.133 schwarze 1630: sz = strcspn(cp, "()");
1.1 schwarze 1631:
1.133 schwarze 1632: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1633: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1634: putkeys(mpage, cp, sz, NAME_SYN);
1635: }
1.19 schwarze 1636:
1.133 schwarze 1637: static int
1.143 schwarze 1638: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1639: const struct roff_node *n)
1.133 schwarze 1640: {
1641:
1642: if (n->child == NULL)
1.150 schwarze 1643: return 0;
1.133 schwarze 1644:
1645: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1646:
1.133 schwarze 1647: for (n = n->child->next; n != NULL; n = n->next)
1.141 schwarze 1648: if (n->type == ROFFT_TEXT)
1.47 schwarze 1649: putkey(mpage, n->string, TYPE_Fa);
1.19 schwarze 1650:
1.150 schwarze 1651: return 0;
1.128 schwarze 1652: }
1653:
1654: static int
1.143 schwarze 1655: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1656: const struct roff_node *n)
1.128 schwarze 1657: {
1.132 schwarze 1658:
1.141 schwarze 1659: if (n->type != ROFFT_HEAD)
1.150 schwarze 1660: return 1;
1.128 schwarze 1661:
1.133 schwarze 1662: if (n->child != NULL)
1663: parse_mdoc_fname(mpage, n->child);
1664:
1.150 schwarze 1665: return 0;
1.1 schwarze 1666: }
1667:
1.19 schwarze 1668: static int
1.143 schwarze 1669: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1670: const struct roff_node *n)
1.1 schwarze 1671: {
1.47 schwarze 1672: char *cp;
1.1 schwarze 1673:
1674: if (NULL == (n = n->child))
1.150 schwarze 1675: return 0;
1.1 schwarze 1676:
1.47 schwarze 1677: if (NULL == n->next) {
1678: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1679: return 0;
1.47 schwarze 1680: }
1.1 schwarze 1681:
1.77 schwarze 1682: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1683: putkey(mpage, cp, TYPE_Xr);
1684: free(cp);
1.150 schwarze 1685: return 0;
1.1 schwarze 1686: }
1687:
1.19 schwarze 1688: static int
1.143 schwarze 1689: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1690: const struct roff_node *n)
1.1 schwarze 1691: {
1692:
1.141 schwarze 1693: if (n->type == ROFFT_BODY)
1.147 schwarze 1694: deroff(&mpage->desc, n);
1.150 schwarze 1695: return 0;
1.1 schwarze 1696: }
1697:
1.19 schwarze 1698: static int
1.143 schwarze 1699: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1700: const struct roff_node *n)
1.1 schwarze 1701: {
1702:
1.86 schwarze 1703: if (SEC_NAME == n->sec)
1.90 schwarze 1704: putmdockey(mpage, n->child, NAME_TITLE);
1.141 schwarze 1705: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1706: if (n->child == NULL)
1707: putkey(mpage, meta->name, NAME_SYN);
1708: else
1709: putmdockey(mpage, n->child, NAME_SYN);
1.129 schwarze 1710: }
1711: if ( ! (mpage->name_head_done ||
1712: n->child == NULL || n->child->string == NULL ||
1713: strcasecmp(n->child->string, meta->title))) {
1.141 schwarze 1714: putkey(mpage, n->child->string, ROFFT_HEAD);
1.129 schwarze 1715: mpage->name_head_done = 1;
1.127 schwarze 1716: }
1.150 schwarze 1717: return 0;
1.1 schwarze 1718: }
1719:
1.19 schwarze 1720: static int
1.143 schwarze 1721: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1722: const struct roff_node *n)
1.1 schwarze 1723: {
1724:
1.150 schwarze 1725: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1726: }
1727:
1.47 schwarze 1728: static int
1.143 schwarze 1729: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1730: const struct roff_node *n)
1.1 schwarze 1731: {
1732:
1.150 schwarze 1733: return n->type == ROFFT_HEAD;
1.47 schwarze 1734: }
1.1 schwarze 1735:
1.47 schwarze 1736: static int
1.143 schwarze 1737: parse_mdoc_body(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1738: const struct roff_node *n)
1.47 schwarze 1739: {
1.1 schwarze 1740:
1.150 schwarze 1741: return n->type == ROFFT_BODY;
1.1 schwarze 1742: }
1743:
1.47 schwarze 1744: /*
1745: * Add a string to the hash table for the current manual.
1746: * Each string has a bitmask telling which macros it belongs to.
1747: * When we finish the manual, we'll dump the table.
1748: */
1.1 schwarze 1749: static void
1.131 schwarze 1750: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1751: {
1.90 schwarze 1752: struct ohash *htab;
1.47 schwarze 1753: struct str *s;
1.68 schwarze 1754: const char *end;
1.47 schwarze 1755: unsigned int slot;
1.131 schwarze 1756: int i, mustfree;
1.1 schwarze 1757:
1.47 schwarze 1758: if (0 == sz)
1759: return;
1.68 schwarze 1760:
1.131 schwarze 1761: mustfree = render_string(&cp, &sz);
1762:
1.90 schwarze 1763: if (TYPE_Nm & v) {
1764: htab = &names;
1765: v &= name_mask;
1.124 schwarze 1766: if (v & NAME_FIRST)
1767: name_mask &= ~NAME_FIRST;
1.90 schwarze 1768: if (debug > 1)
1769: say(mpage->mlinks->file,
1.130 schwarze 1770: "Adding name %*s, bits=%d", sz, cp, v);
1.90 schwarze 1771: } else {
1772: htab = &strings;
1773: if (debug > 1)
1774: for (i = 0; i < mansearch_keymax; i++)
1.118 schwarze 1775: if ((uint64_t)1 << i & v)
1.90 schwarze 1776: say(mpage->mlinks->file,
1777: "Adding key %s=%*s",
1778: mansearch_keynames[i], sz, cp);
1.68 schwarze 1779: }
1.47 schwarze 1780:
1781: end = cp + sz;
1.90 schwarze 1782: slot = ohash_qlookupi(htab, cp, &end);
1783: s = ohash_find(htab, slot);
1.1 schwarze 1784:
1.47 schwarze 1785: if (NULL != s && mpage == s->mpage) {
1786: s->mask |= v;
1.1 schwarze 1787: return;
1.47 schwarze 1788: } else if (NULL == s) {
1.101 schwarze 1789: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1790: memcpy(s->key, cp, sz);
1.90 schwarze 1791: ohash_insert(htab, slot, s);
1.47 schwarze 1792: }
1793: s->mpage = mpage;
1794: s->mask = v;
1.131 schwarze 1795:
1796: if (mustfree)
1797: free(cp);
1.1 schwarze 1798: }
1799:
1800: /*
1.47 schwarze 1801: * Take a Unicode codepoint and produce its UTF-8 encoding.
1802: * This isn't the best way to do this, but it works.
1803: * The magic numbers are from the UTF-8 packaging.
1804: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1805: */
1.47 schwarze 1806: static size_t
1807: utf8(unsigned int cp, char out[7])
1.1 schwarze 1808: {
1.47 schwarze 1809: size_t rc;
1.1 schwarze 1810:
1.47 schwarze 1811: rc = 0;
1812: if (cp <= 0x0000007F) {
1813: rc = 1;
1814: out[0] = (char)cp;
1815: } else if (cp <= 0x000007FF) {
1816: rc = 2;
1817: out[0] = (cp >> 6 & 31) | 192;
1818: out[1] = (cp & 63) | 128;
1819: } else if (cp <= 0x0000FFFF) {
1820: rc = 3;
1821: out[0] = (cp >> 12 & 15) | 224;
1822: out[1] = (cp >> 6 & 63) | 128;
1823: out[2] = (cp & 63) | 128;
1824: } else if (cp <= 0x001FFFFF) {
1825: rc = 4;
1826: out[0] = (cp >> 18 & 7) | 240;
1827: out[1] = (cp >> 12 & 63) | 128;
1828: out[2] = (cp >> 6 & 63) | 128;
1829: out[3] = (cp & 63) | 128;
1830: } else if (cp <= 0x03FFFFFF) {
1831: rc = 5;
1832: out[0] = (cp >> 24 & 3) | 248;
1833: out[1] = (cp >> 18 & 63) | 128;
1834: out[2] = (cp >> 12 & 63) | 128;
1835: out[3] = (cp >> 6 & 63) | 128;
1836: out[4] = (cp & 63) | 128;
1837: } else if (cp <= 0x7FFFFFFF) {
1838: rc = 6;
1839: out[0] = (cp >> 30 & 1) | 252;
1840: out[1] = (cp >> 24 & 63) | 128;
1841: out[2] = (cp >> 18 & 63) | 128;
1842: out[3] = (cp >> 12 & 63) | 128;
1843: out[4] = (cp >> 6 & 63) | 128;
1844: out[5] = (cp & 63) | 128;
1845: } else
1.150 schwarze 1846: return 0;
1.19 schwarze 1847:
1.47 schwarze 1848: out[rc] = '\0';
1.150 schwarze 1849: return rc;
1.1 schwarze 1850: }
1851:
1.47 schwarze 1852: /*
1.131 schwarze 1853: * If the string contains escape sequences,
1854: * replace it with an allocated rendering and return 1,
1855: * such that the caller can free it after use.
1856: * Otherwise, do nothing and return 0.
1.47 schwarze 1857: */
1.131 schwarze 1858: static int
1859: render_string(char **public, size_t *psz)
1.1 schwarze 1860: {
1.131 schwarze 1861: const char *src, *scp, *addcp, *seq;
1862: char *dst;
1863: size_t ssz, dsz, addsz;
1.71 schwarze 1864: char utfbuf[7], res[6];
1.131 schwarze 1865: int seqlen, unicode;
1.47 schwarze 1866:
1867: res[0] = '\\';
1868: res[1] = '\t';
1869: res[2] = ASCII_NBRSP;
1870: res[3] = ASCII_HYPH;
1.71 schwarze 1871: res[4] = ASCII_BREAK;
1872: res[5] = '\0';
1.1 schwarze 1873:
1.131 schwarze 1874: src = scp = *public;
1875: ssz = *psz;
1876: dst = NULL;
1877: dsz = 0;
1878:
1879: while (scp < src + *psz) {
1880:
1881: /* Leave normal characters unchanged. */
1882:
1883: if (strchr(res, *scp) == NULL) {
1884: if (dst != NULL)
1885: dst[dsz++] = *scp;
1886: scp++;
1887: continue;
1888: }
1.1 schwarze 1889:
1.47 schwarze 1890: /*
1.131 schwarze 1891: * Found something that requires replacing,
1892: * make sure we have a destination buffer.
1.47 schwarze 1893: */
1.131 schwarze 1894:
1895: if (dst == NULL) {
1896: dst = mandoc_malloc(ssz + 1);
1897: dsz = scp - src;
1898: memcpy(dst, src, dsz);
1.47 schwarze 1899: }
1.39 schwarze 1900:
1.131 schwarze 1901: /* Handle single-char special characters. */
1902:
1903: switch (*scp) {
1904: case '\\':
1905: break;
1.98 schwarze 1906: case '\t':
1.71 schwarze 1907: /* FALLTHROUGH */
1.98 schwarze 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) {
1.131 schwarze 1938: unicode = mchars_spec2cp(mchars, seq, seqlen);
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 {
1.131 schwarze 1946: addcp = mchars_spec2str(mchars, seq, seqlen, &addsz);
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 void *
1.106 espie 2372: hash_calloc(size_t n, size_t sz, void *arg)
1.47 schwarze 2373: {
1.6 schwarze 2374:
1.150 schwarze 2375: return mandoc_calloc(n, sz);
1.47 schwarze 2376: }
1.2 schwarze 2377:
1.47 schwarze 2378: static void *
2379: hash_alloc(size_t sz, void *arg)
2380: {
1.28 schwarze 2381:
1.150 schwarze 2382: return mandoc_malloc(sz);
1.47 schwarze 2383: }
1.6 schwarze 2384:
1.47 schwarze 2385: static void
1.106 espie 2386: hash_free(void *p, void *arg)
1.47 schwarze 2387: {
1.26 schwarze 2388:
1.47 schwarze 2389: free(p);
2390: }
1.6 schwarze 2391:
1.47 schwarze 2392: static int
1.120 schwarze 2393: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2394: {
2395: static char startdir[PATH_MAX];
1.110 schwarze 2396: static int getcwd_status; /* 1 = ok, 2 = failure */
2397: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2398: char *cp;
1.6 schwarze 2399:
1.47 schwarze 2400: /*
1.110 schwarze 2401: * Remember the original working directory, if possible.
2402: * This will be needed if the second or a later directory
2403: * on the command line is given as a relative path.
2404: * Do not error out if the current directory is not
2405: * searchable: Maybe it won't be needed after all.
2406: */
2407: if (0 == getcwd_status) {
2408: if (NULL == getcwd(startdir, sizeof(startdir))) {
2409: getcwd_status = 2;
2410: (void)strlcpy(startdir, strerror(errno),
2411: sizeof(startdir));
2412: } else
2413: getcwd_status = 1;
2414: }
2415:
2416: /*
2417: * We are leaving the old base directory.
2418: * Do not use it any longer, not even for messages.
2419: */
2420: *basedir = '\0';
2421:
2422: /*
2423: * If and only if the directory was changed earlier and
2424: * the next directory to process is given as a relative path,
2425: * first go back, or bail out if that is impossible.
1.47 schwarze 2426: */
1.110 schwarze 2427: if (chdir_status && '/' != *targetdir) {
2428: if (2 == getcwd_status) {
1.47 schwarze 2429: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2430: say("", "getcwd: %s", startdir);
1.150 schwarze 2431: return 0;
1.47 schwarze 2432: }
1.110 schwarze 2433: if (-1 == chdir(startdir)) {
1.47 schwarze 2434: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2435: say("", "&chdir %s", startdir);
1.150 schwarze 2436: return 0;
1.2 schwarze 2437: }
2438: }
1.110 schwarze 2439:
2440: /*
2441: * Always resolve basedir to the canonicalized absolute
2442: * pathname and append a trailing slash, such that
2443: * we can reliably check whether files are inside.
2444: */
1.47 schwarze 2445: if (NULL == realpath(targetdir, basedir)) {
1.120 schwarze 2446: if (report_baddir || errno != ENOENT) {
2447: exitcode = (int)MANDOCLEVEL_BADARG;
2448: say("", "&%s: realpath", targetdir);
2449: }
1.150 schwarze 2450: return 0;
1.47 schwarze 2451: } else if (-1 == chdir(basedir)) {
1.120 schwarze 2452: if (report_baddir || errno != ENOENT) {
2453: exitcode = (int)MANDOCLEVEL_BADARG;
2454: say("", "&chdir");
2455: }
1.150 schwarze 2456: return 0;
1.109 schwarze 2457: }
1.110 schwarze 2458: chdir_status = 1;
1.109 schwarze 2459: cp = strchr(basedir, '\0');
2460: if ('/' != cp[-1]) {
2461: if (cp - basedir >= PATH_MAX - 1) {
2462: exitcode = (int)MANDOCLEVEL_SYSERR;
2463: say("", "Filename too long");
1.150 schwarze 2464: return 0;
1.109 schwarze 2465: }
2466: *cp++ = '/';
2467: *cp = '\0';
1.47 schwarze 2468: }
1.150 schwarze 2469: return 1;
1.2 schwarze 2470: }
2471:
2472: static void
1.47 schwarze 2473: say(const char *file, const char *format, ...)
1.2 schwarze 2474: {
1.47 schwarze 2475: va_list ap;
1.80 schwarze 2476: int use_errno;
1.2 schwarze 2477:
1.47 schwarze 2478: if ('\0' != *basedir)
2479: fprintf(stderr, "%s", basedir);
2480: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2481: fputc('/', stderr);
1.47 schwarze 2482: if ('\0' != *file)
2483: fprintf(stderr, "%s", file);
1.31 schwarze 2484:
1.80 schwarze 2485: use_errno = 1;
2486: if (NULL != format) {
2487: switch (*format) {
1.98 schwarze 2488: case '&':
1.80 schwarze 2489: format++;
2490: break;
1.98 schwarze 2491: case '\0':
1.80 schwarze 2492: format = NULL;
2493: break;
2494: default:
2495: use_errno = 0;
2496: break;
2497: }
2498: }
2499: if (NULL != format) {
2500: if ('\0' != *basedir || '\0' != *file)
2501: fputs(": ", stderr);
2502: va_start(ap, format);
2503: vfprintf(stderr, format, ap);
2504: va_end(ap);
2505: }
2506: if (use_errno) {
2507: if ('\0' != *basedir || '\0' != *file || NULL != format)
2508: fputs(": ", stderr);
1.47 schwarze 2509: perror(NULL);
1.80 schwarze 2510: } else
2511: fputc('\n', stderr);
1.1 schwarze 2512: }