Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.117
1.117 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.116 2014/09/03 18:08:26 schwarze Exp $ */
1.1 schwarze 2: /*
1.47 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.52 schwarze 4: * Copyright (c) 2011, 2012, 2013, 2014 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: *
10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
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>
1.34 schwarze 24: #include <errno.h>
1.1 schwarze 25: #include <fcntl.h>
1.47 schwarze 26: #include <fts.h>
1.1 schwarze 27: #include <getopt.h>
1.44 schwarze 28: #include <limits.h>
1.47 schwarze 29: #include <stddef.h>
1.1 schwarze 30: #include <stdio.h>
31: #include <stdint.h>
32: #include <stdlib.h>
33: #include <string.h>
1.14 schwarze 34: #include <unistd.h>
1.1 schwarze 35:
1.47 schwarze 36: #include <ohash.h>
37: #include <sqlite3.h>
38:
39: #include "mdoc.h"
1.1 schwarze 40: #include "man.h"
41: #include "mandoc.h"
1.76 schwarze 42: #include "mandoc_aux.h"
1.10 schwarze 43: #include "manpath.h"
1.47 schwarze 44: #include "mansearch.h"
1.1 schwarze 45:
1.68 schwarze 46: extern int mansearch_keymax;
47: extern const char *const mansearch_keynames[];
48:
1.47 schwarze 49: #define SQL_EXEC(_v) \
50: if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
1.91 schwarze 51: say("", "%s: %s", (_v), sqlite3_errmsg(db))
1.47 schwarze 52: #define SQL_BIND_TEXT(_s, _i, _v) \
53: if (SQLITE_OK != sqlite3_bind_text \
54: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.91 schwarze 55: say(mlink->file, "%s", sqlite3_errmsg(db))
1.47 schwarze 56: #define SQL_BIND_INT(_s, _i, _v) \
57: if (SQLITE_OK != sqlite3_bind_int \
58: ((_s), (_i)++, (_v))) \
1.91 schwarze 59: say(mlink->file, "%s", sqlite3_errmsg(db))
1.47 schwarze 60: #define SQL_BIND_INT64(_s, _i, _v) \
61: if (SQLITE_OK != sqlite3_bind_int64 \
62: ((_s), (_i)++, (_v))) \
1.91 schwarze 63: say(mlink->file, "%s", sqlite3_errmsg(db))
1.47 schwarze 64: #define SQL_STEP(_s) \
65: if (SQLITE_DONE != sqlite3_step((_s))) \
1.91 schwarze 66: say(mlink->file, "%s", sqlite3_errmsg(db))
1.1 schwarze 67:
1.47 schwarze 68: enum op {
69: OP_DEFAULT = 0, /* new dbs from dir list or default config */
70: OP_CONFFILE, /* new databases from custom config file */
71: OP_UPDATE, /* delete/add entries in existing database */
72: OP_DELETE, /* delete entries from existing database */
73: OP_TEST /* change no databases, report potential problems */
74: };
1.11 schwarze 75:
1.47 schwarze 76: struct str {
1.53 schwarze 77: char *rendered; /* key in UTF-8 or ASCII form */
1.47 schwarze 78: const struct mpage *mpage; /* if set, the owning parse */
79: uint64_t mask; /* bitmask in sequence */
1.53 schwarze 80: char key[]; /* may contain escape sequences */
1.28 schwarze 81: };
82:
1.47 schwarze 83: struct inodev {
84: ino_t st_ino;
85: dev_t st_dev;
86: };
1.28 schwarze 87:
1.47 schwarze 88: struct mpage {
89: struct inodev inodev; /* used for hashing routine */
1.94 schwarze 90: int64_t pageid; /* pageid in mpages SQL table */
1.47 schwarze 91: char *sec; /* section from file content */
92: char *arch; /* architecture from file content */
93: char *title; /* title from file content */
94: char *desc; /* description from file content */
95: struct mlink *mlinks; /* singly linked list */
1.116 schwarze 96: int form; /* format from file content */
1.28 schwarze 97: };
98:
1.47 schwarze 99: struct mlink {
100: char file[PATH_MAX]; /* filename rel. to manpath */
101: char *dsec; /* section from directory */
102: char *arch; /* architecture from directory */
103: char *name; /* name from file name (not empty) */
104: char *fsec; /* section from file name suffix */
105: struct mlink *next; /* singly linked list */
1.75 schwarze 106: struct mpage *mpage; /* parent */
1.116 schwarze 107: int dform; /* format from directory */
108: int fform; /* format from file name suffix */
1.81 schwarze 109: int gzip; /* filename has a .gz suffix */
1.2 schwarze 110: };
111:
1.47 schwarze 112: enum stmt {
113: STMT_DELETE_PAGE = 0, /* delete mpage */
114: STMT_INSERT_PAGE, /* insert mpage */
115: STMT_INSERT_LINK, /* insert mlink */
1.90 schwarze 116: STMT_INSERT_NAME, /* insert name */
1.47 schwarze 117: STMT_INSERT_KEY, /* insert parsed key */
118: STMT__MAX
1.1 schwarze 119: };
120:
1.47 schwarze 121: typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
1.1 schwarze 122:
1.19 schwarze 123: struct mdoc_handler {
1.47 schwarze 124: mdoc_fp fp; /* optional handler */
125: uint64_t mask; /* set unless handler returns 0 */
1.19 schwarze 126: };
127:
1.47 schwarze 128: static void dbclose(int);
1.75 schwarze 129: static void dbadd(struct mpage *, struct mchars *);
130: static void dbadd_mlink(const struct mlink *mlink);
1.47 schwarze 131: static int dbopen(int);
132: static void dbprune(void);
133: static void filescan(const char *);
134: static void *hash_alloc(size_t, void *);
1.106 espie 135: static void hash_free(void *, void *);
136: static void *hash_calloc(size_t, size_t, void *);
1.47 schwarze 137: static void mlink_add(struct mlink *, const struct stat *);
1.88 schwarze 138: static void mlink_check(struct mpage *, struct mlink *);
1.47 schwarze 139: static void mlink_free(struct mlink *);
140: static void mlinks_undupe(struct mpage *);
141: static void mpages_free(void);
1.58 schwarze 142: static void mpages_merge(struct mchars *, struct mparse *);
1.87 schwarze 143: static void names_check(void);
1.81 schwarze 144: static void parse_cat(struct mpage *, int);
1.47 schwarze 145: static void parse_man(struct mpage *, const struct man_node *);
146: static void parse_mdoc(struct mpage *, const struct mdoc_node *);
147: static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
148: static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
149: static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
150: static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
151: static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
152: static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
153: static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
154: static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
1.69 schwarze 155: static void putkey(const struct mpage *, char *, uint64_t);
1.47 schwarze 156: static void putkeys(const struct mpage *,
157: const char *, size_t, uint64_t);
158: static void putmdockey(const struct mpage *,
159: const struct mdoc_node *, uint64_t);
1.53 schwarze 160: static void render_key(struct mchars *, struct str *);
1.47 schwarze 161: static void say(const char *, const char *, ...);
162: static int set_basedir(const char *);
163: static int treescan(void);
164: static size_t utf8(unsigned int, char [7]);
165:
1.72 schwarze 166: static char tempfilename[32];
1.47 schwarze 167: static char *progname;
1.59 schwarze 168: static int nodb; /* no database changes */
1.73 schwarze 169: static int mparse_options; /* abort the parse early */
1.98 schwarze 170: static int use_all; /* use all found files */
171: static int debug; /* print what we're doing */
172: static int warnings; /* warn about crap */
1.52 schwarze 173: static int write_utf8; /* write UTF-8 output; else ASCII */
1.47 schwarze 174: static int exitcode; /* to be returned by main */
1.98 schwarze 175: static enum op op; /* operational mode */
1.47 schwarze 176: static char basedir[PATH_MAX]; /* current base directory */
177: static struct ohash mpages; /* table of distinct manual pages */
178: static struct ohash mlinks; /* table of directory entries */
1.90 schwarze 179: static struct ohash names; /* table of all names */
1.47 schwarze 180: static struct ohash strings; /* table of all strings */
181: static sqlite3 *db = NULL; /* current database */
182: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.90 schwarze 183: static uint64_t name_mask;
1.47 schwarze 184:
1.19 schwarze 185: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.47 schwarze 186: { NULL, 0 }, /* Ap */
187: { NULL, 0 }, /* Dd */
188: { NULL, 0 }, /* Dt */
189: { NULL, 0 }, /* Os */
190: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
191: { parse_mdoc_head, TYPE_Ss }, /* Ss */
192: { NULL, 0 }, /* Pp */
193: { NULL, 0 }, /* D1 */
194: { NULL, 0 }, /* Dl */
195: { NULL, 0 }, /* Bd */
196: { NULL, 0 }, /* Ed */
197: { NULL, 0 }, /* Bl */
198: { NULL, 0 }, /* El */
199: { NULL, 0 }, /* It */
200: { NULL, 0 }, /* Ad */
201: { NULL, TYPE_An }, /* An */
202: { NULL, TYPE_Ar }, /* Ar */
203: { NULL, TYPE_Cd }, /* Cd */
204: { NULL, TYPE_Cm }, /* Cm */
205: { NULL, TYPE_Dv }, /* Dv */
206: { NULL, TYPE_Er }, /* Er */
207: { NULL, TYPE_Ev }, /* Ev */
208: { NULL, 0 }, /* Ex */
209: { NULL, TYPE_Fa }, /* Fa */
210: { parse_mdoc_Fd, 0 }, /* Fd */
211: { NULL, TYPE_Fl }, /* Fl */
212: { parse_mdoc_Fn, 0 }, /* Fn */
213: { NULL, TYPE_Ft }, /* Ft */
214: { NULL, TYPE_Ic }, /* Ic */
1.49 schwarze 215: { NULL, TYPE_In }, /* In */
1.47 schwarze 216: { NULL, TYPE_Li }, /* Li */
1.89 schwarze 217: { parse_mdoc_Nd, 0 }, /* Nd */
1.86 schwarze 218: { parse_mdoc_Nm, 0 }, /* Nm */
1.47 schwarze 219: { NULL, 0 }, /* Op */
220: { NULL, 0 }, /* Ot */
221: { NULL, TYPE_Pa }, /* Pa */
222: { NULL, 0 }, /* Rv */
1.49 schwarze 223: { NULL, TYPE_St }, /* St */
1.47 schwarze 224: { NULL, TYPE_Va }, /* Va */
225: { parse_mdoc_body, TYPE_Va }, /* Vt */
226: { parse_mdoc_Xr, 0 }, /* Xr */
227: { NULL, 0 }, /* %A */
228: { NULL, 0 }, /* %B */
229: { NULL, 0 }, /* %D */
230: { NULL, 0 }, /* %I */
231: { NULL, 0 }, /* %J */
232: { NULL, 0 }, /* %N */
233: { NULL, 0 }, /* %O */
234: { NULL, 0 }, /* %P */
235: { NULL, 0 }, /* %R */
236: { NULL, 0 }, /* %T */
237: { NULL, 0 }, /* %V */
238: { NULL, 0 }, /* Ac */
239: { NULL, 0 }, /* Ao */
240: { NULL, 0 }, /* Aq */
241: { NULL, TYPE_At }, /* At */
242: { NULL, 0 }, /* Bc */
243: { NULL, 0 }, /* Bf */
244: { NULL, 0 }, /* Bo */
245: { NULL, 0 }, /* Bq */
246: { NULL, TYPE_Bsx }, /* Bsx */
247: { NULL, TYPE_Bx }, /* Bx */
248: { NULL, 0 }, /* Db */
249: { NULL, 0 }, /* Dc */
250: { NULL, 0 }, /* Do */
251: { NULL, 0 }, /* Dq */
252: { NULL, 0 }, /* Ec */
253: { NULL, 0 }, /* Ef */
254: { NULL, TYPE_Em }, /* Em */
255: { NULL, 0 }, /* Eo */
256: { NULL, TYPE_Fx }, /* Fx */
257: { NULL, TYPE_Ms }, /* Ms */
258: { NULL, 0 }, /* No */
259: { NULL, 0 }, /* Ns */
260: { NULL, TYPE_Nx }, /* Nx */
261: { NULL, TYPE_Ox }, /* Ox */
262: { NULL, 0 }, /* Pc */
263: { NULL, 0 }, /* Pf */
264: { NULL, 0 }, /* Po */
265: { NULL, 0 }, /* Pq */
266: { NULL, 0 }, /* Qc */
267: { NULL, 0 }, /* Ql */
268: { NULL, 0 }, /* Qo */
269: { NULL, 0 }, /* Qq */
270: { NULL, 0 }, /* Re */
271: { NULL, 0 }, /* Rs */
272: { NULL, 0 }, /* Sc */
273: { NULL, 0 }, /* So */
274: { NULL, 0 }, /* Sq */
275: { NULL, 0 }, /* Sm */
276: { NULL, 0 }, /* Sx */
277: { NULL, TYPE_Sy }, /* Sy */
278: { NULL, TYPE_Tn }, /* Tn */
279: { NULL, 0 }, /* Ux */
280: { NULL, 0 }, /* Xc */
281: { NULL, 0 }, /* Xo */
282: { parse_mdoc_head, 0 }, /* Fo */
283: { NULL, 0 }, /* Fc */
284: { NULL, 0 }, /* Oo */
285: { NULL, 0 }, /* Oc */
286: { NULL, 0 }, /* Bk */
287: { NULL, 0 }, /* Ek */
288: { NULL, 0 }, /* Bt */
289: { NULL, 0 }, /* Hf */
290: { NULL, 0 }, /* Fr */
291: { NULL, 0 }, /* Ud */
292: { NULL, TYPE_Lb }, /* Lb */
293: { NULL, 0 }, /* Lp */
294: { NULL, TYPE_Lk }, /* Lk */
295: { NULL, TYPE_Mt }, /* Mt */
296: { NULL, 0 }, /* Brq */
297: { NULL, 0 }, /* Bro */
298: { NULL, 0 }, /* Brc */
299: { NULL, 0 }, /* %C */
300: { NULL, 0 }, /* Es */
301: { NULL, 0 }, /* En */
302: { NULL, TYPE_Dx }, /* Dx */
303: { NULL, 0 }, /* %Q */
304: { NULL, 0 }, /* br */
305: { NULL, 0 }, /* sp */
306: { NULL, 0 }, /* %U */
307: { NULL, 0 }, /* Ta */
1.1 schwarze 308: };
309:
1.98 schwarze 310:
1.1 schwarze 311: int
1.3 schwarze 312: mandocdb(int argc, char *argv[])
1.1 schwarze 313: {
1.47 schwarze 314: int ch, i;
315: size_t j, sz;
316: const char *path_arg;
317: struct mchars *mc;
318: struct manpaths dirs;
319: struct mparse *mp;
320: struct ohash_info mpages_info, mlinks_info;
321:
322: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
323: memset(&dirs, 0, sizeof(struct manpaths));
324:
325: mpages_info.alloc = mlinks_info.alloc = hash_alloc;
1.106 espie 326: mpages_info.calloc = mlinks_info.calloc = hash_calloc;
327: mpages_info.free = mlinks_info.free = hash_free;
1.47 schwarze 328:
329: mpages_info.key_offset = offsetof(struct mpage, inodev);
330: mlinks_info.key_offset = offsetof(struct mlink, file);
1.1 schwarze 331:
332: progname = strrchr(argv[0], '/');
333: if (progname == NULL)
334: progname = argv[0];
335: else
336: ++progname;
337:
1.47 schwarze 338: /*
1.98 schwarze 339: * We accept a few different invocations.
1.47 schwarze 340: * The CHECKOP macro makes sure that invocation styles don't
341: * clobber each other.
342: */
343: #define CHECKOP(_op, _ch) do \
344: if (OP_DEFAULT != (_op)) { \
1.111 schwarze 345: fprintf(stderr, "%s: -%c: Conflicting option\n", \
346: progname, (_ch)); \
1.47 schwarze 347: goto usage; \
348: } while (/*CONSTCOND*/0)
1.10 schwarze 349:
1.47 schwarze 350: path_arg = NULL;
1.28 schwarze 351: op = OP_DEFAULT;
1.1 schwarze 352:
1.83 schwarze 353: while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
1.1 schwarze 354: switch (ch) {
1.98 schwarze 355: case 'a':
1.6 schwarze 356: use_all = 1;
357: break;
1.98 schwarze 358: case 'C':
1.47 schwarze 359: CHECKOP(op, ch);
360: path_arg = optarg;
1.28 schwarze 361: op = OP_CONFFILE;
1.25 schwarze 362: break;
1.98 schwarze 363: case 'D':
1.82 schwarze 364: debug++;
365: break;
1.98 schwarze 366: case 'd':
1.47 schwarze 367: CHECKOP(op, ch);
368: path_arg = optarg;
1.2 schwarze 369: op = OP_UPDATE;
1.1 schwarze 370: break;
1.98 schwarze 371: case 'n':
1.47 schwarze 372: nodb = 1;
373: break;
1.98 schwarze 374: case 'p':
1.83 schwarze 375: warnings = 1;
376: break;
1.98 schwarze 377: case 'Q':
1.73 schwarze 378: mparse_options |= MPARSE_QUICK;
1.59 schwarze 379: break;
1.98 schwarze 380: case 'T':
1.52 schwarze 381: if (strcmp(optarg, "utf8")) {
1.111 schwarze 382: fprintf(stderr, "%s: -T%s: "
383: "Unsupported output format\n",
384: progname, optarg);
1.52 schwarze 385: goto usage;
386: }
387: write_utf8 = 1;
388: break;
1.98 schwarze 389: case 't':
1.47 schwarze 390: CHECKOP(op, ch);
1.28 schwarze 391: dup2(STDOUT_FILENO, STDERR_FILENO);
392: op = OP_TEST;
1.47 schwarze 393: nodb = warnings = 1;
1.28 schwarze 394: break;
1.98 schwarze 395: case 'u':
1.47 schwarze 396: CHECKOP(op, ch);
397: path_arg = optarg;
1.1 schwarze 398: op = OP_DELETE;
399: break;
1.98 schwarze 400: case 'v':
1.83 schwarze 401: /* Compatibility with espie@'s makewhatis. */
1.28 schwarze 402: break;
1.1 schwarze 403: default:
1.28 schwarze 404: goto usage;
1.1 schwarze 405: }
406:
407: argc -= optind;
408: argv += optind;
409:
1.28 schwarze 410: if (OP_CONFFILE == op && argc > 0) {
1.111 schwarze 411: fprintf(stderr, "%s: -C: Too many arguments\n",
412: progname);
1.28 schwarze 413: goto usage;
414: }
415:
1.47 schwarze 416: exitcode = (int)MANDOCLEVEL_OK;
1.73 schwarze 417: mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL, NULL);
1.47 schwarze 418: mc = mchars_alloc();
1.2 schwarze 419:
1.47 schwarze 420: ohash_init(&mpages, 6, &mpages_info);
421: ohash_init(&mlinks, 6, &mlinks_info);
1.2 schwarze 422:
1.47 schwarze 423: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.2 schwarze 424:
1.47 schwarze 425: /*
1.109 schwarze 426: * Most of these deal with a specific directory.
1.97 schwarze 427: * Jump into that directory first.
1.47 schwarze 428: */
1.109 schwarze 429: if (OP_TEST != op && 0 == set_basedir(path_arg))
1.47 schwarze 430: goto out;
1.97 schwarze 431:
1.95 schwarze 432: if (dbopen(1)) {
1.97 schwarze 433: /*
434: * The existing database is usable. Process
435: * all files specified on the command-line.
436: */
437: use_all = 1;
1.95 schwarze 438: for (i = 0; i < argc; i++)
439: filescan(argv[i]);
440: if (OP_TEST != op)
441: dbprune();
442: } else {
443: /*
444: * Database missing or corrupt.
445: * Recreate from scratch.
446: */
1.97 schwarze 447: exitcode = (int)MANDOCLEVEL_OK;
1.95 schwarze 448: op = OP_DEFAULT;
449: if (0 == treescan())
450: goto out;
451: if (0 == dbopen(0))
452: goto out;
453: }
1.47 schwarze 454: if (OP_DELETE != op)
1.58 schwarze 455: mpages_merge(mc, mp);
1.95 schwarze 456: dbclose(OP_DEFAULT == op ? 0 : 1);
1.47 schwarze 457: } else {
458: /*
459: * If we have arguments, use them as our manpaths.
460: * If we don't, grok from manpath(1) or however else
461: * manpath_parse() wants to do it.
462: */
463: if (argc > 0) {
1.101 schwarze 464: dirs.paths = mandoc_reallocarray(NULL,
465: argc, sizeof(char *));
1.47 schwarze 466: dirs.sz = (size_t)argc;
467: for (i = 0; i < argc; i++)
468: dirs.paths[i] = mandoc_strdup(argv[i]);
469: } else
470: manpath_parse(&dirs, path_arg, NULL, NULL);
1.84 schwarze 471:
472: if (0 == dirs.sz) {
473: exitcode = (int)MANDOCLEVEL_BADARG;
474: say("", "Empty manpath");
475: }
1.2 schwarze 476:
1.47 schwarze 477: /*
478: * First scan the tree rooted at a base directory, then
479: * build a new database and finally move it into place.
480: * Ignore zero-length directories and strip trailing
481: * slashes.
482: */
483: for (j = 0; j < dirs.sz; j++) {
484: sz = strlen(dirs.paths[j]);
485: if (sz && '/' == dirs.paths[j][sz - 1])
486: dirs.paths[j][--sz] = '\0';
487: if (0 == sz)
488: continue;
1.2 schwarze 489:
1.47 schwarze 490: if (j) {
491: ohash_init(&mpages, 6, &mpages_info);
492: ohash_init(&mlinks, 6, &mlinks_info);
493: }
1.1 schwarze 494:
1.47 schwarze 495: if (0 == set_basedir(dirs.paths[j]))
496: goto out;
497: if (0 == treescan())
498: goto out;
499: if (0 == dbopen(0))
500: goto out;
1.41 deraadt 501:
1.58 schwarze 502: mpages_merge(mc, mp);
1.104 schwarze 503: if (warnings && !nodb &&
1.87 schwarze 504: ! (MPARSE_QUICK & mparse_options))
505: names_check();
1.47 schwarze 506: dbclose(0);
1.28 schwarze 507:
1.47 schwarze 508: if (j + 1 < dirs.sz) {
509: mpages_free();
510: ohash_delete(&mpages);
511: ohash_delete(&mlinks);
512: }
1.2 schwarze 513: }
1.47 schwarze 514: }
515: out:
516: manpath_free(&dirs);
517: mchars_free(mc);
518: mparse_free(mp);
519: mpages_free();
520: ohash_delete(&mpages);
521: ohash_delete(&mlinks);
522: return(exitcode);
523: usage:
1.83 schwarze 524: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
525: " %s [-aDnpQ] [-Tutf8] dir ...\n"
526: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
527: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 528: " %s [-Q] -t file ...\n",
1.98 schwarze 529: progname, progname, progname,
1.47 schwarze 530: progname, progname);
1.1 schwarze 531:
1.47 schwarze 532: return((int)MANDOCLEVEL_BADARG);
533: }
1.1 schwarze 534:
1.47 schwarze 535: /*
536: * Scan a directory tree rooted at "basedir" for manpages.
537: * We use fts(), scanning directory parts along the way for clues to our
538: * section and architecture.
539: *
540: * If use_all has been specified, grok all files.
541: * If not, sanitise paths to the following:
542: *
1.98 schwarze 543: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 544: * or
545: * [./]cat<section>[/<arch>]/<name>.0
546: *
547: * TODO: accomodate for multi-language directories.
548: */
549: static int
550: treescan(void)
551: {
1.96 schwarze 552: char buf[PATH_MAX];
1.47 schwarze 553: FTS *f;
554: FTSENT *ff;
555: struct mlink *mlink;
1.81 schwarze 556: int dform, gzip;
1.51 schwarze 557: char *dsec, *arch, *fsec, *cp;
558: const char *path;
1.47 schwarze 559: const char *argv[2];
1.1 schwarze 560:
1.47 schwarze 561: argv[0] = ".";
562: argv[1] = (char *)NULL;
1.1 schwarze 563:
1.96 schwarze 564: f = fts_open((char * const *)argv,
565: FTS_PHYSICAL | FTS_NOCHDIR, NULL);
1.47 schwarze 566: if (NULL == f) {
567: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 568: say("", "&fts_open");
1.47 schwarze 569: return(0);
570: }
1.2 schwarze 571:
1.47 schwarze 572: dsec = arch = NULL;
573: dform = FORM_NONE;
1.2 schwarze 574:
1.47 schwarze 575: while (NULL != (ff = fts_read(f))) {
576: path = ff->fts_path + 2;
1.96 schwarze 577: switch (ff->fts_info) {
578:
579: /*
580: * Symbolic links require various sanity checks,
581: * then get handled just like regular files.
582: */
1.98 schwarze 583: case FTS_SL:
1.96 schwarze 584: if (NULL == realpath(path, buf)) {
585: if (warnings)
586: say(path, "&realpath");
587: continue;
588: }
589: if (strstr(buf, basedir) != buf) {
590: if (warnings) say("",
591: "%s: outside base directory", buf);
592: continue;
593: }
594: /* Use logical inode to avoid mpages dupe. */
595: if (-1 == stat(path, ff->fts_statp)) {
596: if (warnings)
597: say(path, "&stat");
598: continue;
599: }
600: /* FALLTHROUGH */
601:
1.15 schwarze 602: /*
1.47 schwarze 603: * If we're a regular file, add an mlink by using the
604: * stored directory data and handling the filename.
1.15 schwarze 605: */
1.98 schwarze 606: case FTS_F:
1.47 schwarze 607: if (0 == strcmp(path, MANDOC_DB))
608: continue;
609: if ( ! use_all && ff->fts_level < 2) {
610: if (warnings)
611: say(path, "Extraneous file");
612: continue;
1.81 schwarze 613: }
614: gzip = 0;
615: fsec = NULL;
616: while (NULL == fsec) {
617: fsec = strrchr(ff->fts_name, '.');
618: if (NULL == fsec || strcmp(fsec+1, "gz"))
619: break;
620: gzip = 1;
621: *fsec = '\0';
622: fsec = NULL;
623: }
624: if (NULL == fsec) {
1.47 schwarze 625: if ( ! use_all) {
626: if (warnings)
627: say(path,
628: "No filename suffix");
629: continue;
630: }
631: } else if (0 == strcmp(++fsec, "html")) {
632: if (warnings)
633: say(path, "Skip html");
634: continue;
635: } else if (0 == strcmp(fsec, "ps")) {
636: if (warnings)
637: say(path, "Skip ps");
638: continue;
639: } else if (0 == strcmp(fsec, "pdf")) {
640: if (warnings)
641: say(path, "Skip pdf");
642: continue;
643: } else if ( ! use_all &&
644: ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
645: (FORM_CAT == dform && strcmp(fsec, "0")))) {
646: if (warnings)
647: say(path, "Wrong filename suffix");
648: continue;
649: } else
650: fsec[-1] = '\0';
1.51 schwarze 651:
1.47 schwarze 652: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.100 schwarze 653: if (strlcpy(mlink->file, path,
654: sizeof(mlink->file)) >=
655: sizeof(mlink->file)) {
656: say(path, "Filename too long");
657: free(mlink);
658: continue;
659: }
1.47 schwarze 660: mlink->dform = dform;
1.51 schwarze 661: mlink->dsec = dsec;
662: mlink->arch = arch;
663: mlink->name = ff->fts_name;
664: mlink->fsec = fsec;
1.81 schwarze 665: mlink->gzip = gzip;
1.47 schwarze 666: mlink_add(mlink, ff->fts_statp);
667: continue;
1.96 schwarze 668:
1.98 schwarze 669: case FTS_D:
1.96 schwarze 670: /* FALLTHROUGH */
1.98 schwarze 671: case FTS_DP:
1.96 schwarze 672: break;
673:
674: default:
1.47 schwarze 675: if (warnings)
676: say(path, "Not a regular file");
677: continue;
678: }
679:
680: switch (ff->fts_level) {
1.98 schwarze 681: case 0:
1.47 schwarze 682: /* Ignore the root directory. */
683: break;
1.98 schwarze 684: case 1:
1.47 schwarze 685: /*
686: * This might contain manX/ or catX/.
687: * Try to infer this from the name.
688: * If we're not in use_all, enforce it.
689: */
690: cp = ff->fts_name;
691: if (FTS_DP == ff->fts_info)
692: break;
1.15 schwarze 693:
1.47 schwarze 694: if (0 == strncmp(cp, "man", 3)) {
695: dform = FORM_SRC;
696: dsec = cp + 3;
697: } else if (0 == strncmp(cp, "cat", 3)) {
698: dform = FORM_CAT;
699: dsec = cp + 3;
1.51 schwarze 700: } else {
701: dform = FORM_NONE;
702: dsec = NULL;
1.26 schwarze 703: }
1.47 schwarze 704:
1.98 schwarze 705: if (NULL != dsec || use_all)
1.47 schwarze 706: break;
707:
708: if (warnings)
709: say(path, "Unknown directory part");
710: fts_set(f, ff, FTS_SKIP);
711: break;
1.98 schwarze 712: case 2:
1.47 schwarze 713: /*
714: * Possibly our architecture.
715: * If we're descending, keep tabs on it.
716: */
717: if (FTS_DP != ff->fts_info && NULL != dsec)
718: arch = ff->fts_name;
1.51 schwarze 719: else
720: arch = NULL;
1.47 schwarze 721: break;
722: default:
723: if (FTS_DP == ff->fts_info || use_all)
724: break;
725: if (warnings)
726: say(path, "Extraneous directory part");
727: fts_set(f, ff, FTS_SKIP);
728: break;
1.14 schwarze 729: }
1.47 schwarze 730: }
731:
732: fts_close(f);
733: return(1);
734: }
1.1 schwarze 735:
1.47 schwarze 736: /*
737: * Add a file to the mlinks table.
738: * Do not verify that it's a "valid" looking manpage (we'll do that
739: * later).
740: *
741: * Try to infer the manual section, architecture, and page name from the
742: * path, assuming it looks like
743: *
1.98 schwarze 744: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 745: * or
746: * [./]cat<section>[/<arch>]/<name>.0
747: *
748: * See treescan() for the fts(3) version of this.
749: */
750: static void
751: filescan(const char *file)
752: {
753: char buf[PATH_MAX];
754: struct stat st;
755: struct mlink *mlink;
756: char *p, *start;
757:
758: assert(use_all);
759:
760: if (0 == strncmp(file, "./", 2))
761: file += 2;
762:
1.96 schwarze 763: /*
764: * We have to do lstat(2) before realpath(3) loses
765: * the information whether this is a symbolic link.
766: * We need to know that because for symbolic links,
767: * we want to use the orginal file name, while for
768: * regular files, we want to use the real path.
769: */
770: if (-1 == lstat(file, &st)) {
771: exitcode = (int)MANDOCLEVEL_BADARG;
772: say(file, "&lstat");
773: return;
774: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
775: exitcode = (int)MANDOCLEVEL_BADARG;
776: say(file, "Not a regular file");
777: return;
778: }
779:
780: /*
781: * We have to resolve the file name to the real path
782: * in any case for the base directory check.
783: */
1.47 schwarze 784: if (NULL == realpath(file, buf)) {
785: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 786: say(file, "&realpath");
1.47 schwarze 787: return;
1.63 schwarze 788: }
789:
1.109 schwarze 790: if (OP_TEST == op)
1.63 schwarze 791: start = buf;
1.109 schwarze 792: else if (strstr(buf, basedir) == buf)
793: start = buf + strlen(basedir);
1.63 schwarze 794: else {
1.47 schwarze 795: exitcode = (int)MANDOCLEVEL_BADARG;
796: say("", "%s: outside base directory", buf);
797: return;
1.63 schwarze 798: }
799:
1.96 schwarze 800: /*
801: * Now we are sure the file is inside our tree.
802: * If it is a symbolic link, ignore the real path
803: * and use the original name.
804: * This implies passing stuff like "cat1/../man1/foo.1"
805: * on the command line won't work. So don't do that.
806: * Note the stat(2) can still fail if the link target
807: * doesn't exist.
808: */
809: if (S_IFLNK & st.st_mode) {
810: if (-1 == stat(buf, &st)) {
811: exitcode = (int)MANDOCLEVEL_BADARG;
812: say(file, "&stat");
813: return;
814: }
1.100 schwarze 815: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
816: say(file, "Filename too long");
817: return;
818: }
1.109 schwarze 819: start = buf;
820: if (OP_TEST != op && strstr(buf, basedir) == buf)
821: start += strlen(basedir);
1.1 schwarze 822: }
1.63 schwarze 823:
1.47 schwarze 824: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 825: mlink->dform = FORM_NONE;
1.100 schwarze 826: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
827: sizeof(mlink->file)) {
828: say(start, "Filename too long");
829: return;
830: }
1.1 schwarze 831:
1.10 schwarze 832: /*
1.47 schwarze 833: * First try to guess our directory structure.
834: * If we find a separator, try to look for man* or cat*.
835: * If we find one of these and what's underneath is a directory,
836: * assume it's an architecture.
1.10 schwarze 837: */
1.47 schwarze 838: if (NULL != (p = strchr(start, '/'))) {
839: *p++ = '\0';
840: if (0 == strncmp(start, "man", 3)) {
841: mlink->dform = FORM_SRC;
1.51 schwarze 842: mlink->dsec = start + 3;
1.47 schwarze 843: } else if (0 == strncmp(start, "cat", 3)) {
844: mlink->dform = FORM_CAT;
1.51 schwarze 845: mlink->dsec = start + 3;
1.47 schwarze 846: }
1.10 schwarze 847:
1.47 schwarze 848: start = p;
849: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
850: *p++ = '\0';
1.51 schwarze 851: mlink->arch = start;
1.47 schwarze 852: start = p;
1.41 deraadt 853: }
1.47 schwarze 854: }
1.7 schwarze 855:
1.47 schwarze 856: /*
857: * Now check the file suffix.
858: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
859: */
860: p = strrchr(start, '\0');
861: while (p-- > start && '/' != *p && '.' != *p)
862: /* Loop. */ ;
863:
864: if ('.' == *p) {
865: *p++ = '\0';
1.51 schwarze 866: mlink->fsec = p;
1.47 schwarze 867: }
1.41 deraadt 868:
1.47 schwarze 869: /*
870: * Now try to parse the name.
871: * Use the filename portion of the path.
872: */
873: mlink->name = start;
874: if (NULL != (p = strrchr(start, '/'))) {
875: mlink->name = p + 1;
876: *p = '\0';
877: }
878: mlink_add(mlink, &st);
879: }
1.2 schwarze 880:
1.47 schwarze 881: static void
882: mlink_add(struct mlink *mlink, const struct stat *st)
883: {
884: struct inodev inodev;
885: struct mpage *mpage;
886: unsigned int slot;
887:
888: assert(NULL != mlink->file);
889:
1.51 schwarze 890: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
891: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
892: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
893: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 894:
895: if ('0' == *mlink->fsec) {
896: free(mlink->fsec);
897: mlink->fsec = mandoc_strdup(mlink->dsec);
898: mlink->fform = FORM_CAT;
899: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
900: mlink->fform = FORM_SRC;
901: else
902: mlink->fform = FORM_NONE;
1.1 schwarze 903:
1.47 schwarze 904: slot = ohash_qlookup(&mlinks, mlink->file);
905: assert(NULL == ohash_find(&mlinks, slot));
906: ohash_insert(&mlinks, slot, mlink);
907:
908: inodev.st_ino = st->st_ino;
909: inodev.st_dev = st->st_dev;
910: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
911: sizeof(struct inodev), inodev.st_ino);
912: mpage = ohash_find(&mpages, slot);
913: if (NULL == mpage) {
914: mpage = mandoc_calloc(1, sizeof(struct mpage));
915: mpage->inodev.st_ino = inodev.st_ino;
916: mpage->inodev.st_dev = inodev.st_dev;
917: ohash_insert(&mpages, slot, mpage);
918: } else
919: mlink->next = mpage->mlinks;
920: mpage->mlinks = mlink;
1.75 schwarze 921: mlink->mpage = mpage;
1.47 schwarze 922: }
1.1 schwarze 923:
1.47 schwarze 924: static void
925: mlink_free(struct mlink *mlink)
926: {
1.1 schwarze 927:
1.47 schwarze 928: free(mlink->dsec);
929: free(mlink->arch);
930: free(mlink->name);
931: free(mlink->fsec);
932: free(mlink);
933: }
1.1 schwarze 934:
1.47 schwarze 935: static void
936: mpages_free(void)
937: {
938: struct mpage *mpage;
939: struct mlink *mlink;
940: unsigned int slot;
941:
942: mpage = ohash_first(&mpages, &slot);
943: while (NULL != mpage) {
944: while (NULL != (mlink = mpage->mlinks)) {
945: mpage->mlinks = mlink->next;
946: mlink_free(mlink);
947: }
948: free(mpage->sec);
949: free(mpage->arch);
950: free(mpage->title);
951: free(mpage->desc);
952: free(mpage);
953: mpage = ohash_next(&mpages, &slot);
954: }
955: }
1.1 schwarze 956:
1.47 schwarze 957: /*
958: * For each mlink to the mpage, check whether the path looks like
959: * it is formatted, and if it does, check whether a source manual
960: * exists by the same name, ignoring the suffix.
961: * If both conditions hold, drop the mlink.
962: */
963: static void
964: mlinks_undupe(struct mpage *mpage)
965: {
966: char buf[PATH_MAX];
967: struct mlink **prev;
968: struct mlink *mlink;
969: char *bufp;
970:
971: mpage->form = FORM_CAT;
972: prev = &mpage->mlinks;
973: while (NULL != (mlink = *prev)) {
974: if (FORM_CAT != mlink->dform) {
975: mpage->form = FORM_NONE;
976: goto nextlink;
977: }
1.100 schwarze 978: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 979: bufp = strstr(buf, "cat");
980: assert(NULL != bufp);
981: memcpy(bufp, "man", 3);
982: if (NULL != (bufp = strrchr(buf, '.')))
983: *++bufp = '\0';
1.100 schwarze 984: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 985: if (NULL == ohash_find(&mlinks,
1.98 schwarze 986: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 987: goto nextlink;
988: if (warnings)
989: say(mlink->file, "Man source exists: %s", buf);
990: if (use_all)
991: goto nextlink;
992: *prev = mlink->next;
993: mlink_free(mlink);
994: continue;
995: nextlink:
996: prev = &(*prev)->next;
1.1 schwarze 997: }
1.47 schwarze 998: }
1.1 schwarze 999:
1.88 schwarze 1000: static void
1.50 schwarze 1001: mlink_check(struct mpage *mpage, struct mlink *mlink)
1002: {
1.88 schwarze 1003: struct str *str;
1004: unsigned int slot;
1.50 schwarze 1005:
1006: /*
1007: * Check whether the manual section given in a file
1008: * agrees with the directory where the file is located.
1009: * Some manuals have suffixes like (3p) on their
1010: * section number either inside the file or in the
1011: * directory name, some are linked into more than one
1012: * section, like encrypt(1) = makekey(8).
1013: */
1014:
1015: if (FORM_SRC == mpage->form &&
1.88 schwarze 1016: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1017: say(mlink->file, "Section \"%s\" manual in %s directory",
1018: mpage->sec, mlink->dsec);
1019:
1020: /*
1021: * Manual page directories exist for each kernel
1022: * architecture as returned by machine(1).
1023: * However, many manuals only depend on the
1024: * application architecture as returned by arch(1).
1025: * For example, some (2/ARM) manuals are shared
1026: * across the "armish" and "zaurus" kernel
1027: * architectures.
1028: * A few manuals are even shared across completely
1029: * different architectures, for example fdformat(1)
1030: * on amd64, i386, sparc, and sparc64.
1031: */
1032:
1.88 schwarze 1033: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1034: say(mlink->file, "Architecture \"%s\" manual in "
1035: "\"%s\" directory", mpage->arch, mlink->arch);
1036:
1.88 schwarze 1037: /*
1038: * XXX
1.90 schwarze 1039: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1040: */
1041:
1042: if (FORM_CAT == mpage->form)
1043: return;
1044:
1045: /*
1046: * Check whether this mlink
1047: * appears as a name in the NAME section.
1048: */
1.50 schwarze 1049:
1.90 schwarze 1050: slot = ohash_qlookup(&names, mlink->name);
1051: str = ohash_find(&names, slot);
1.88 schwarze 1052: assert(NULL != str);
1.90 schwarze 1053: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1054: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1055: }
1056:
1.47 schwarze 1057: /*
1058: * Run through the files in the global vector "mpages"
1059: * and add them to the database specified in "basedir".
1060: *
1061: * This handles the parsing scheme itself, using the cues of directory
1062: * and filename to determine whether the file is parsable or not.
1063: */
1064: static void
1.58 schwarze 1065: mpages_merge(struct mchars *mc, struct mparse *mp)
1.47 schwarze 1066: {
1.70 schwarze 1067: char any[] = "any";
1.58 schwarze 1068: struct ohash_info str_info;
1.75 schwarze 1069: struct mpage *mpage, *mpage_dest;
1070: struct mlink *mlink, *mlink_dest;
1.47 schwarze 1071: struct mdoc *mdoc;
1072: struct man *man;
1.75 schwarze 1073: char *sodest;
1.69 schwarze 1074: char *cp;
1.81 schwarze 1075: pid_t child_pid;
1.117 ! schwarze 1076: int fd;
1.58 schwarze 1077: unsigned int pslot;
1.47 schwarze 1078: enum mandoclevel lvl;
1079:
1080: str_info.alloc = hash_alloc;
1.106 espie 1081: str_info.calloc = hash_calloc;
1082: str_info.free = hash_free;
1.47 schwarze 1083: str_info.key_offset = offsetof(struct str, key);
1084:
1.64 schwarze 1085: if (0 == nodb)
1086: SQL_EXEC("BEGIN TRANSACTION");
1087:
1.47 schwarze 1088: mpage = ohash_first(&mpages, &pslot);
1089: while (NULL != mpage) {
1090: mlinks_undupe(mpage);
1091: if (NULL == mpage->mlinks) {
1092: mpage = ohash_next(&mpages, &pslot);
1093: continue;
1094: }
1.1 schwarze 1095:
1.90 schwarze 1096: name_mask = NAME_MASK;
1097: ohash_init(&names, 4, &str_info);
1.47 schwarze 1098: ohash_init(&strings, 6, &str_info);
1099: mparse_reset(mp);
1100: mdoc = NULL;
1101: man = NULL;
1.81 schwarze 1102: sodest = NULL;
1103: child_pid = 0;
1104:
1.117 ! schwarze 1105: mparse_open(mp, &fd, mpage->mlinks->file, &child_pid);
! 1106: if (fd == -1) {
! 1107: say(mpage->mlinks->file, "&open");
! 1108: goto nextpage;
1.81 schwarze 1109: }
1.11 schwarze 1110:
1111: /*
1.24 schwarze 1112: * Try interpreting the file as mdoc(7) or man(7)
1113: * source code, unless it is already known to be
1114: * formatted. Fall back to formatted mode.
1.11 schwarze 1115: */
1.47 schwarze 1116: if (FORM_CAT != mpage->mlinks->dform ||
1117: FORM_CAT != mpage->mlinks->fform) {
1.117 ! schwarze 1118: lvl = mparse_readfd(mp, fd, mpage->mlinks->file);
1.47 schwarze 1119: if (lvl < MANDOCLEVEL_FATAL)
1.75 schwarze 1120: mparse_result(mp, &mdoc, &man, &sodest);
1.47 schwarze 1121: }
1.11 schwarze 1122:
1.75 schwarze 1123: if (NULL != sodest) {
1124: mlink_dest = ohash_find(&mlinks,
1125: ohash_qlookup(&mlinks, sodest));
1126: if (NULL != mlink_dest) {
1127:
1128: /* The .so target exists. */
1129:
1130: mpage_dest = mlink_dest->mpage;
1131: mlink = mpage->mlinks;
1132: while (1) {
1133: mlink->mpage = mpage_dest;
1134:
1135: /*
1136: * If the target was already
1137: * processed, add the links
1138: * to the database now.
1139: * Otherwise, this will
1140: * happen when we come
1141: * to the target.
1142: */
1143:
1.94 schwarze 1144: if (mpage_dest->pageid)
1.75 schwarze 1145: dbadd_mlink(mlink);
1146:
1147: if (NULL == mlink->next)
1148: break;
1149: mlink = mlink->next;
1150: }
1151:
1152: /* Move all links to the target. */
1153:
1154: mlink->next = mlink_dest->next;
1155: mlink_dest->next = mpage->mlinks;
1156: mpage->mlinks = NULL;
1157: }
1.81 schwarze 1158: goto nextpage;
1.75 schwarze 1159: } else if (NULL != mdoc) {
1.47 schwarze 1160: mpage->form = FORM_SRC;
1.112 schwarze 1161: mpage->sec = mdoc_meta(mdoc)->msec;
1162: mpage->sec = mandoc_strdup(
1163: NULL == mpage->sec ? "" : mpage->sec);
1.47 schwarze 1164: mpage->arch = mdoc_meta(mdoc)->arch;
1165: mpage->arch = mandoc_strdup(
1166: NULL == mpage->arch ? "" : mpage->arch);
1167: mpage->title =
1168: mandoc_strdup(mdoc_meta(mdoc)->title);
1.11 schwarze 1169: } else if (NULL != man) {
1.47 schwarze 1170: mpage->form = FORM_SRC;
1171: mpage->sec =
1172: mandoc_strdup(man_meta(man)->msec);
1173: mpage->arch =
1174: mandoc_strdup(mpage->mlinks->arch);
1175: mpage->title =
1176: mandoc_strdup(man_meta(man)->title);
1.11 schwarze 1177: } else {
1.47 schwarze 1178: mpage->form = FORM_CAT;
1179: mpage->sec =
1180: mandoc_strdup(mpage->mlinks->dsec);
1181: mpage->arch =
1182: mandoc_strdup(mpage->mlinks->arch);
1183: mpage->title =
1184: mandoc_strdup(mpage->mlinks->name);
1.1 schwarze 1185: }
1.116 schwarze 1186: if (mpage->mlinks->gzip)
1187: mpage->form |= FORM_GZ;
1.54 schwarze 1188: putkey(mpage, mpage->sec, TYPE_sec);
1.55 schwarze 1189: putkey(mpage, '\0' == *mpage->arch ?
1.70 schwarze 1190: any : mpage->arch, TYPE_arch);
1.1 schwarze 1191:
1.54 schwarze 1192: for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1193: if ('\0' != *mlink->dsec)
1194: putkey(mpage, mlink->dsec, TYPE_sec);
1195: if ('\0' != *mlink->fsec)
1196: putkey(mpage, mlink->fsec, TYPE_sec);
1.55 schwarze 1197: putkey(mpage, '\0' == *mlink->arch ?
1.70 schwarze 1198: any : mlink->arch, TYPE_arch);
1.90 schwarze 1199: putkey(mpage, mlink->name, NAME_FILE);
1.54 schwarze 1200: }
1.41 deraadt 1201:
1.92 schwarze 1202: assert(NULL == mpage->desc);
1.47 schwarze 1203: if (NULL != mdoc) {
1204: if (NULL != (cp = mdoc_meta(mdoc)->name))
1.90 schwarze 1205: putkey(mpage, cp, NAME_HEAD);
1.47 schwarze 1206: parse_mdoc(mpage, mdoc_node(mdoc));
1207: } else if (NULL != man)
1208: parse_man(mpage, man_node(man));
1209: else
1.117 ! schwarze 1210: parse_cat(mpage, fd);
1.92 schwarze 1211: if (NULL == mpage->desc)
1212: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1.88 schwarze 1213:
1214: if (warnings && !use_all)
1215: for (mlink = mpage->mlinks; mlink;
1216: mlink = mlink->next)
1217: mlink_check(mpage, mlink);
1.6 schwarze 1218:
1.62 schwarze 1219: dbadd(mpage, mc);
1.81 schwarze 1220:
1221: nextpage:
1.117 ! schwarze 1222: if (child_pid &&
! 1223: mparse_wait(mp, child_pid) != MANDOCLEVEL_OK) {
! 1224: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1225: say(mpage->mlinks->file, "&wait gunzip");
1.81 schwarze 1226: }
1.47 schwarze 1227: ohash_delete(&strings);
1.90 schwarze 1228: ohash_delete(&names);
1.47 schwarze 1229: mpage = ohash_next(&mpages, &pslot);
1230: }
1.64 schwarze 1231:
1232: if (0 == nodb)
1233: SQL_EXEC("END TRANSACTION");
1.87 schwarze 1234: }
1235:
1236: static void
1237: names_check(void)
1238: {
1239: sqlite3_stmt *stmt;
1240: const char *name, *sec, *arch, *key;
1241: int irc;
1242:
1243: sqlite3_prepare_v2(db,
1244: "SELECT name, sec, arch, key FROM ("
1.90 schwarze 1245: "SELECT name AS key, pageid FROM names "
1.87 schwarze 1246: "WHERE bits & ? AND NOT EXISTS ("
1247: "SELECT pageid FROM mlinks "
1.90 schwarze 1248: "WHERE mlinks.pageid == names.pageid "
1249: "AND mlinks.name == names.name"
1.87 schwarze 1250: ")"
1251: ") JOIN ("
1.99 schwarze 1252: "SELECT sec, arch, name, pageid FROM mlinks "
1253: "GROUP BY pageid"
1.87 schwarze 1254: ") USING (pageid);",
1255: -1, &stmt, NULL);
1256:
1.91 schwarze 1257: if (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE))
1258: say("", "%s", sqlite3_errmsg(db));
1.87 schwarze 1259:
1260: while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1.113 schwarze 1261: name = (const char *)sqlite3_column_text(stmt, 0);
1262: sec = (const char *)sqlite3_column_text(stmt, 1);
1263: arch = (const char *)sqlite3_column_text(stmt, 2);
1264: key = (const char *)sqlite3_column_text(stmt, 3);
1.87 schwarze 1265: say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1266: '\0' == *arch ? "" : "/",
1267: '\0' == *arch ? "" : arch, key);
1268: }
1269: sqlite3_finalize(stmt);
1.47 schwarze 1270: }
1.6 schwarze 1271:
1.47 schwarze 1272: static void
1.81 schwarze 1273: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1274: {
1275: FILE *stream;
1276: char *line, *p, *title;
1277: size_t len, plen, titlesz;
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:
1.47 schwarze 1290: /* Skip to first blank line. */
1.1 schwarze 1291:
1.47 schwarze 1292: while (NULL != (line = fgetln(stream, &len)))
1293: if ('\n' == *line)
1294: break;
1.1 schwarze 1295:
1.47 schwarze 1296: /*
1297: * Assume the first line that is not indented
1298: * is the first section header. Skip to it.
1299: */
1.1 schwarze 1300:
1.47 schwarze 1301: while (NULL != (line = fgetln(stream, &len)))
1302: if ('\n' != *line && ' ' != *line)
1303: break;
1.98 schwarze 1304:
1.47 schwarze 1305: /*
1306: * Read up until the next section into a buffer.
1307: * Strip the leading and trailing newline from each read line,
1308: * appending a trailing space.
1309: * Ignore empty (whitespace-only) lines.
1310: */
1.28 schwarze 1311:
1.47 schwarze 1312: titlesz = 0;
1313: title = NULL;
1.38 schwarze 1314:
1.47 schwarze 1315: while (NULL != (line = fgetln(stream, &len))) {
1316: if (' ' != *line || '\n' != line[len - 1])
1317: break;
1318: while (len > 0 && isspace((unsigned char)*line)) {
1319: line++;
1320: len--;
1321: }
1322: if (1 == len)
1323: continue;
1324: title = mandoc_realloc(title, titlesz + len);
1325: memcpy(title + titlesz, line, len);
1326: titlesz += len;
1327: title[titlesz - 1] = ' ';
1328: }
1.28 schwarze 1329:
1.47 schwarze 1330: /*
1331: * If no page content can be found, or the input line
1332: * is already the next section header, or there is no
1333: * trailing newline, reuse the page title as the page
1334: * description.
1335: */
1.1 schwarze 1336:
1.47 schwarze 1337: if (NULL == title || '\0' == *title) {
1338: if (warnings)
1339: say(mpage->mlinks->file,
1340: "Cannot find NAME section");
1341: fclose(stream);
1342: free(title);
1343: return;
1344: }
1.24 schwarze 1345:
1.47 schwarze 1346: title = mandoc_realloc(title, titlesz + 1);
1347: title[titlesz] = '\0';
1.24 schwarze 1348:
1.47 schwarze 1349: /*
1350: * Skip to the first dash.
1351: * Use the remaining line as the description (no more than 70
1352: * bytes).
1353: */
1.28 schwarze 1354:
1.47 schwarze 1355: if (NULL != (p = strstr(title, "- "))) {
1356: for (p += 2; ' ' == *p || '\b' == *p; p++)
1357: /* Skip to next word. */ ;
1358: } else {
1359: if (warnings)
1360: say(mpage->mlinks->file,
1361: "No dash in title line");
1362: p = title;
1363: }
1.1 schwarze 1364:
1.47 schwarze 1365: plen = strlen(p);
1.1 schwarze 1366:
1.47 schwarze 1367: /* Strip backspace-encoding from line. */
1.1 schwarze 1368:
1.47 schwarze 1369: while (NULL != (line = memchr(p, '\b', plen))) {
1370: len = line - p;
1371: if (0 == len) {
1372: memmove(line, line + 1, plen--);
1373: continue;
1.98 schwarze 1374: }
1.47 schwarze 1375: memmove(line - 1, line + 1, plen - len);
1376: plen -= 2;
1377: }
1.1 schwarze 1378:
1.47 schwarze 1379: mpage->desc = mandoc_strdup(p);
1380: fclose(stream);
1381: free(title);
1382: }
1.16 schwarze 1383:
1.47 schwarze 1384: /*
1385: * Put a type/word pair into the word database for this particular file.
1386: */
1387: static void
1.69 schwarze 1388: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1389: {
1.69 schwarze 1390: char *cp;
1.37 schwarze 1391:
1.47 schwarze 1392: assert(NULL != value);
1.69 schwarze 1393: if (TYPE_arch == type)
1394: for (cp = value; *cp; cp++)
1395: if (isupper((unsigned char)*cp))
1396: *cp = _tolower((unsigned char)*cp);
1.47 schwarze 1397: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1398: }
1399:
1400: /*
1.47 schwarze 1401: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1402: */
1403: static void
1.47 schwarze 1404: putmdockey(const struct mpage *mpage,
1405: const struct mdoc_node *n, uint64_t m)
1.2 schwarze 1406: {
1.16 schwarze 1407:
1.47 schwarze 1408: for ( ; NULL != n; n = n->next) {
1409: if (NULL != n->child)
1410: putmdockey(mpage, n->child, m);
1411: if (MDOC_TEXT == n->type)
1412: putkey(mpage, n->string, m);
1413: }
1414: }
1.16 schwarze 1415:
1.47 schwarze 1416: static void
1417: parse_man(struct mpage *mpage, const struct man_node *n)
1418: {
1419: const struct man_node *head, *body;
1.78 schwarze 1420: char *start, *title;
1.47 schwarze 1421: char byte;
1.78 schwarze 1422: size_t sz;
1.16 schwarze 1423:
1.47 schwarze 1424: if (NULL == n)
1425: return;
1.16 schwarze 1426:
1.47 schwarze 1427: /*
1428: * We're only searching for one thing: the first text child in
1429: * the BODY of a NAME section. Since we don't keep track of
1430: * sections in -man, run some hoops to find out whether we're in
1431: * the correct section or not.
1432: */
1.16 schwarze 1433:
1.47 schwarze 1434: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1435: body = n;
1436: assert(body->parent);
1437: if (NULL != (head = body->parent->head) &&
1.98 schwarze 1438: 1 == head->nchild &&
1439: NULL != (head = (head->child)) &&
1440: MAN_TEXT == head->type &&
1441: 0 == strcmp(head->string, "NAME") &&
1442: NULL != body->child) {
1.2 schwarze 1443:
1.47 schwarze 1444: /*
1445: * Suck the entire NAME section into memory.
1446: * Yes, we might run away.
1447: * But too many manuals have big, spread-out
1448: * NAME sections over many lines.
1449: */
1.2 schwarze 1450:
1.78 schwarze 1451: title = NULL;
1452: man_deroff(&title, body);
1.47 schwarze 1453: if (NULL == title)
1454: return;
1.16 schwarze 1455:
1.98 schwarze 1456: /*
1.47 schwarze 1457: * Go through a special heuristic dance here.
1458: * Conventionally, one or more manual names are
1459: * comma-specified prior to a whitespace, then a
1460: * dash, then a description. Try to puzzle out
1461: * the name parts here.
1462: */
1.16 schwarze 1463:
1.78 schwarze 1464: start = title;
1.47 schwarze 1465: for ( ;; ) {
1466: sz = strcspn(start, " ,");
1467: if ('\0' == start[sz])
1468: break;
1.1 schwarze 1469:
1.47 schwarze 1470: byte = start[sz];
1471: start[sz] = '\0';
1.67 schwarze 1472:
1473: /*
1474: * Assume a stray trailing comma in the
1475: * name list if a name begins with a dash.
1476: */
1477:
1478: if ('-' == start[0] ||
1479: ('\\' == start[0] && '-' == start[1]))
1480: break;
1.1 schwarze 1481:
1.90 schwarze 1482: putkey(mpage, start, NAME_TITLE);
1.1 schwarze 1483:
1.47 schwarze 1484: if (' ' == byte) {
1485: start += sz + 1;
1486: break;
1487: }
1.1 schwarze 1488:
1.47 schwarze 1489: assert(',' == byte);
1490: start += sz + 1;
1491: while (' ' == *start)
1492: start++;
1493: }
1.1 schwarze 1494:
1.78 schwarze 1495: if (start == title) {
1.90 schwarze 1496: putkey(mpage, start, NAME_TITLE);
1.47 schwarze 1497: free(title);
1498: return;
1499: }
1.1 schwarze 1500:
1.47 schwarze 1501: while (isspace((unsigned char)*start))
1502: start++;
1.1 schwarze 1503:
1.47 schwarze 1504: if (0 == strncmp(start, "-", 1))
1505: start += 1;
1506: else if (0 == strncmp(start, "\\-\\-", 4))
1507: start += 4;
1508: else if (0 == strncmp(start, "\\-", 2))
1509: start += 2;
1510: else if (0 == strncmp(start, "\\(en", 4))
1511: start += 4;
1512: else if (0 == strncmp(start, "\\(em", 4))
1513: start += 4;
1.1 schwarze 1514:
1.47 schwarze 1515: while (' ' == *start)
1516: start++;
1.1 schwarze 1517:
1.47 schwarze 1518: mpage->desc = mandoc_strdup(start);
1519: free(title);
1520: return;
1521: }
1522: }
1.1 schwarze 1523:
1.47 schwarze 1524: for (n = n->child; n; n = n->next) {
1525: if (NULL != mpage->desc)
1526: break;
1527: parse_man(mpage, n);
1.1 schwarze 1528: }
1529: }
1530:
1531: static void
1.47 schwarze 1532: parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1533: {
1534:
1.47 schwarze 1535: assert(NULL != n);
1536: for (n = n->child; NULL != n; n = n->next) {
1537: switch (n->type) {
1.98 schwarze 1538: case MDOC_ELEM:
1.47 schwarze 1539: /* FALLTHROUGH */
1.98 schwarze 1540: case MDOC_BLOCK:
1.47 schwarze 1541: /* FALLTHROUGH */
1.98 schwarze 1542: case MDOC_HEAD:
1.47 schwarze 1543: /* FALLTHROUGH */
1.98 schwarze 1544: case MDOC_BODY:
1.47 schwarze 1545: /* FALLTHROUGH */
1.98 schwarze 1546: case MDOC_TAIL:
1.47 schwarze 1547: if (NULL != mdocs[n->tok].fp)
1548: if (0 == (*mdocs[n->tok].fp)(mpage, n))
1549: break;
1550: if (mdocs[n->tok].mask)
1551: putmdockey(mpage, n->child,
1552: mdocs[n->tok].mask);
1553: break;
1554: default:
1555: assert(MDOC_ROOT != n->type);
1556: continue;
1557: }
1558: if (NULL != n->child)
1559: parse_mdoc(mpage, n);
1.1 schwarze 1560: }
1561: }
1562:
1.19 schwarze 1563: static int
1.47 schwarze 1564: parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1565: {
1566: const char *start, *end;
1567: size_t sz;
1.19 schwarze 1568:
1.47 schwarze 1569: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1570: NULL == (n = n->child) ||
1571: MDOC_TEXT != n->type)
1.19 schwarze 1572: return(0);
1.1 schwarze 1573:
1574: /*
1575: * Only consider those `Fd' macro fields that begin with an
1576: * "inclusion" token (versus, e.g., #define).
1577: */
1.47 schwarze 1578:
1.1 schwarze 1579: if (strcmp("#include", n->string))
1.19 schwarze 1580: return(0);
1.1 schwarze 1581:
1582: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.19 schwarze 1583: return(0);
1.1 schwarze 1584:
1585: /*
1586: * Strip away the enclosing angle brackets and make sure we're
1587: * not zero-length.
1588: */
1589:
1590: start = n->string;
1591: if ('<' == *start || '"' == *start)
1592: start++;
1593:
1594: if (0 == (sz = strlen(start)))
1.19 schwarze 1595: return(0);
1.1 schwarze 1596:
1597: end = &start[(int)sz - 1];
1598: if ('>' == *end || '"' == *end)
1599: end--;
1600:
1.47 schwarze 1601: if (end > start)
1602: putkeys(mpage, start, end - start + 1, TYPE_In);
1.49 schwarze 1603: return(0);
1.1 schwarze 1604: }
1605:
1.19 schwarze 1606: static int
1.47 schwarze 1607: parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1608: {
1.69 schwarze 1609: char *cp;
1.1 schwarze 1610:
1.47 schwarze 1611: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.19 schwarze 1612: return(0);
1613:
1.98 schwarze 1614: /*
1.47 schwarze 1615: * Parse: .Fn "struct type *name" "char *arg".
1.98 schwarze 1616: * First strip away pointer symbol.
1.47 schwarze 1617: * Then store the function name, then type.
1.98 schwarze 1618: * Finally, store the arguments.
1.47 schwarze 1619: */
1.1 schwarze 1620:
1.47 schwarze 1621: if (NULL == (cp = strrchr(n->string, ' ')))
1622: cp = n->string;
1.1 schwarze 1623:
1624: while ('*' == *cp)
1625: cp++;
1626:
1.47 schwarze 1627: putkey(mpage, cp, TYPE_Fn);
1.19 schwarze 1628:
1.47 schwarze 1629: if (n->string < cp)
1630: putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1.19 schwarze 1631:
1.47 schwarze 1632: for (n = n->next; NULL != n; n = n->next)
1633: if (MDOC_TEXT == n->type)
1634: putkey(mpage, n->string, TYPE_Fa);
1.19 schwarze 1635:
1636: return(0);
1.1 schwarze 1637: }
1638:
1.19 schwarze 1639: static int
1.47 schwarze 1640: parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1641: {
1.47 schwarze 1642: char *cp;
1.1 schwarze 1643:
1644: if (NULL == (n = n->child))
1.19 schwarze 1645: return(0);
1.1 schwarze 1646:
1.47 schwarze 1647: if (NULL == n->next) {
1648: putkey(mpage, n->string, TYPE_Xr);
1649: return(0);
1650: }
1.1 schwarze 1651:
1.77 schwarze 1652: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1653: putkey(mpage, cp, TYPE_Xr);
1654: free(cp);
1655: return(0);
1.1 schwarze 1656: }
1657:
1.19 schwarze 1658: static int
1.47 schwarze 1659: parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1660: {
1661:
1.79 schwarze 1662: if (MDOC_BODY == n->type)
1663: mdoc_deroff(&mpage->desc, n);
1664: return(0);
1.1 schwarze 1665: }
1666:
1.19 schwarze 1667: static int
1.47 schwarze 1668: parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1669: {
1670:
1.86 schwarze 1671: if (SEC_NAME == n->sec)
1.90 schwarze 1672: putmdockey(mpage, n->child, NAME_TITLE);
1.86 schwarze 1673: else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type)
1.90 schwarze 1674: putmdockey(mpage, n->child, NAME_SYN);
1.86 schwarze 1675: return(0);
1.1 schwarze 1676: }
1677:
1.19 schwarze 1678: static int
1.47 schwarze 1679: parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1680: {
1681:
1.19 schwarze 1682: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 schwarze 1683: }
1684:
1.47 schwarze 1685: static int
1686: parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1687: {
1688:
1.47 schwarze 1689: return(MDOC_HEAD == n->type);
1690: }
1.1 schwarze 1691:
1.47 schwarze 1692: static int
1693: parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1694: {
1.1 schwarze 1695:
1.47 schwarze 1696: return(MDOC_BODY == n->type);
1.1 schwarze 1697: }
1698:
1.47 schwarze 1699: /*
1700: * Add a string to the hash table for the current manual.
1701: * Each string has a bitmask telling which macros it belongs to.
1702: * When we finish the manual, we'll dump the table.
1703: */
1.1 schwarze 1704: static void
1.47 schwarze 1705: putkeys(const struct mpage *mpage,
1706: const char *cp, size_t sz, uint64_t v)
1.1 schwarze 1707: {
1.90 schwarze 1708: struct ohash *htab;
1.47 schwarze 1709: struct str *s;
1.68 schwarze 1710: const char *end;
1.47 schwarze 1711: unsigned int slot;
1.68 schwarze 1712: int i;
1.1 schwarze 1713:
1.47 schwarze 1714: if (0 == sz)
1715: return;
1.68 schwarze 1716:
1.90 schwarze 1717: if (TYPE_Nm & v) {
1718: htab = &names;
1719: v &= name_mask;
1720: name_mask &= ~NAME_FIRST;
1721: if (debug > 1)
1722: say(mpage->mlinks->file,
1723: "Adding name %*s", sz, cp);
1724: } else {
1725: htab = &strings;
1726: if (debug > 1)
1727: for (i = 0; i < mansearch_keymax; i++)
1728: if (1 << i & v)
1729: say(mpage->mlinks->file,
1730: "Adding key %s=%*s",
1731: mansearch_keynames[i], sz, cp);
1.68 schwarze 1732: }
1.47 schwarze 1733:
1734: end = cp + sz;
1.90 schwarze 1735: slot = ohash_qlookupi(htab, cp, &end);
1736: s = ohash_find(htab, slot);
1.1 schwarze 1737:
1.47 schwarze 1738: if (NULL != s && mpage == s->mpage) {
1739: s->mask |= v;
1.1 schwarze 1740: return;
1.47 schwarze 1741: } else if (NULL == s) {
1.101 schwarze 1742: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1743: memcpy(s->key, cp, sz);
1.90 schwarze 1744: ohash_insert(htab, slot, s);
1.47 schwarze 1745: }
1746: s->mpage = mpage;
1747: s->mask = v;
1.1 schwarze 1748: }
1749:
1750: /*
1.47 schwarze 1751: * Take a Unicode codepoint and produce its UTF-8 encoding.
1752: * This isn't the best way to do this, but it works.
1753: * The magic numbers are from the UTF-8 packaging.
1754: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1755: */
1.47 schwarze 1756: static size_t
1757: utf8(unsigned int cp, char out[7])
1.1 schwarze 1758: {
1.47 schwarze 1759: size_t rc;
1.1 schwarze 1760:
1.47 schwarze 1761: rc = 0;
1762: if (cp <= 0x0000007F) {
1763: rc = 1;
1764: out[0] = (char)cp;
1765: } else if (cp <= 0x000007FF) {
1766: rc = 2;
1767: out[0] = (cp >> 6 & 31) | 192;
1768: out[1] = (cp & 63) | 128;
1769: } else if (cp <= 0x0000FFFF) {
1770: rc = 3;
1771: out[0] = (cp >> 12 & 15) | 224;
1772: out[1] = (cp >> 6 & 63) | 128;
1773: out[2] = (cp & 63) | 128;
1774: } else if (cp <= 0x001FFFFF) {
1775: rc = 4;
1776: out[0] = (cp >> 18 & 7) | 240;
1777: out[1] = (cp >> 12 & 63) | 128;
1778: out[2] = (cp >> 6 & 63) | 128;
1779: out[3] = (cp & 63) | 128;
1780: } else if (cp <= 0x03FFFFFF) {
1781: rc = 5;
1782: out[0] = (cp >> 24 & 3) | 248;
1783: out[1] = (cp >> 18 & 63) | 128;
1784: out[2] = (cp >> 12 & 63) | 128;
1785: out[3] = (cp >> 6 & 63) | 128;
1786: out[4] = (cp & 63) | 128;
1787: } else if (cp <= 0x7FFFFFFF) {
1788: rc = 6;
1789: out[0] = (cp >> 30 & 1) | 252;
1790: out[1] = (cp >> 24 & 63) | 128;
1791: out[2] = (cp >> 18 & 63) | 128;
1792: out[3] = (cp >> 12 & 63) | 128;
1793: out[4] = (cp >> 6 & 63) | 128;
1794: out[5] = (cp & 63) | 128;
1795: } else
1796: return(0);
1.19 schwarze 1797:
1.47 schwarze 1798: out[rc] = '\0';
1799: return(rc);
1.1 schwarze 1800: }
1801:
1.47 schwarze 1802: /*
1.53 schwarze 1803: * Store the rendered version of a key, or alias the pointer
1804: * if the key contains no escape sequences.
1.47 schwarze 1805: */
1806: static void
1.53 schwarze 1807: render_key(struct mchars *mc, struct str *key)
1.1 schwarze 1808: {
1.47 schwarze 1809: size_t sz, bsz, pos;
1.71 schwarze 1810: char utfbuf[7], res[6];
1.47 schwarze 1811: char *buf;
1812: const char *seq, *cpp, *val;
1813: int len, u;
1814: enum mandoc_esc esc;
1815:
1.53 schwarze 1816: assert(NULL == key->rendered);
1.47 schwarze 1817:
1818: res[0] = '\\';
1819: res[1] = '\t';
1820: res[2] = ASCII_NBRSP;
1821: res[3] = ASCII_HYPH;
1.71 schwarze 1822: res[4] = ASCII_BREAK;
1823: res[5] = '\0';
1.1 schwarze 1824:
1.47 schwarze 1825: val = key->key;
1826: bsz = strlen(val);
1.1 schwarze 1827:
1828: /*
1.47 schwarze 1829: * Pre-check: if we have no stop-characters, then set the
1830: * pointer as ourselvse and get out of here.
1.1 schwarze 1831: */
1.47 schwarze 1832: if (strcspn(val, res) == bsz) {
1.53 schwarze 1833: key->rendered = key->key;
1.47 schwarze 1834: return;
1.98 schwarze 1835: }
1.1 schwarze 1836:
1.47 schwarze 1837: /* Pre-allocate by the length of the input */
1.39 schwarze 1838:
1.47 schwarze 1839: buf = mandoc_malloc(++bsz);
1840: pos = 0;
1.39 schwarze 1841:
1.47 schwarze 1842: while ('\0' != *val) {
1843: /*
1844: * Halt on the first escape sequence.
1845: * This also halts on the end of string, in which case
1846: * we just copy, fallthrough, and exit the loop.
1847: */
1848: if ((sz = strcspn(val, res)) > 0) {
1849: memcpy(&buf[pos], val, sz);
1850: pos += sz;
1851: val += sz;
1852: }
1.39 schwarze 1853:
1.71 schwarze 1854: switch (*val) {
1.98 schwarze 1855: case ASCII_HYPH:
1.47 schwarze 1856: buf[pos++] = '-';
1857: val++;
1858: continue;
1.98 schwarze 1859: case '\t':
1.71 schwarze 1860: /* FALLTHROUGH */
1.98 schwarze 1861: case ASCII_NBRSP:
1.47 schwarze 1862: buf[pos++] = ' ';
1863: val++;
1.71 schwarze 1864: /* FALLTHROUGH */
1.98 schwarze 1865: case ASCII_BREAK:
1.47 schwarze 1866: continue;
1.71 schwarze 1867: default:
1868: break;
1869: }
1870: if ('\\' != *val)
1.47 schwarze 1871: break;
1.39 schwarze 1872:
1.47 schwarze 1873: /* Read past the slash. */
1.39 schwarze 1874:
1.47 schwarze 1875: val++;
1.39 schwarze 1876:
1.47 schwarze 1877: /*
1878: * Parse the escape sequence and see if it's a
1879: * predefined character or special character.
1880: */
1.52 schwarze 1881:
1.98 schwarze 1882: esc = mandoc_escape((const char **)&val,
1883: &seq, &len);
1.47 schwarze 1884: if (ESCAPE_ERROR == esc)
1885: break;
1886: if (ESCAPE_SPECIAL != esc)
1887: continue;
1.39 schwarze 1888:
1.47 schwarze 1889: /*
1.52 schwarze 1890: * Render the special character
1891: * as either UTF-8 or ASCII.
1.47 schwarze 1892: */
1.52 schwarze 1893:
1894: if (write_utf8) {
1895: if (0 == (u = mchars_spec2cp(mc, seq, len)))
1896: continue;
1897: cpp = utfbuf;
1898: if (0 == (sz = utf8(u, utfbuf)))
1899: continue;
1900: sz = strlen(cpp);
1901: } else {
1902: cpp = mchars_spec2str(mc, seq, len, &sz);
1903: if (NULL == cpp)
1904: continue;
1905: if (ASCII_NBRSP == *cpp) {
1906: cpp = " ";
1907: sz = 1;
1908: }
1909: }
1.1 schwarze 1910:
1.47 schwarze 1911: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1912:
1.47 schwarze 1913: bsz += sz;
1914: buf = mandoc_realloc(buf, bsz);
1915: memcpy(&buf[pos], cpp, sz);
1916: pos += sz;
1.1 schwarze 1917: }
1918:
1.47 schwarze 1919: buf[pos] = '\0';
1.53 schwarze 1920: key->rendered = buf;
1.1 schwarze 1921: }
1922:
1.75 schwarze 1923: static void
1924: dbadd_mlink(const struct mlink *mlink)
1925: {
1926: size_t i;
1927:
1928: i = 1;
1929: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1930: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1931: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.94 schwarze 1932: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1.75 schwarze 1933: SQL_STEP(stmts[STMT_INSERT_LINK]);
1934: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1.115 schwarze 1935:
1936: i = 1;
1937: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, NAME_FILE);
1938: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
1939: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
1940: SQL_STEP(stmts[STMT_INSERT_NAME]);
1941: sqlite3_reset(stmts[STMT_INSERT_NAME]);
1.75 schwarze 1942: }
1943:
1.11 schwarze 1944: /*
1.47 schwarze 1945: * Flush the current page's terms (and their bits) into the database.
1946: * Wrap the entire set of additions in a transaction to make sqlite be a
1947: * little faster.
1.53 schwarze 1948: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 1949: */
1950: static void
1.75 schwarze 1951: dbadd(struct mpage *mpage, struct mchars *mc)
1.11 schwarze 1952: {
1.47 schwarze 1953: struct mlink *mlink;
1954: struct str *key;
1955: size_t i;
1956: unsigned int slot;
1957:
1.85 schwarze 1958: mlink = mpage->mlinks;
1.11 schwarze 1959:
1.85 schwarze 1960: if (nodb) {
1.104 schwarze 1961: for (key = ohash_first(&names, &slot); NULL != key;
1962: key = ohash_next(&names, &slot)) {
1963: if (key->rendered != key->key)
1964: free(key->rendered);
1965: free(key);
1966: }
1967: for (key = ohash_first(&strings, &slot); NULL != key;
1968: key = ohash_next(&strings, &slot)) {
1969: if (key->rendered != key->key)
1970: free(key->rendered);
1971: free(key);
1972: }
1.102 schwarze 1973: if (0 == debug)
1974: return;
1.85 schwarze 1975: while (NULL != mlink) {
1976: fputs(mlink->name, stdout);
1977: if (NULL == mlink->next ||
1978: strcmp(mlink->dsec, mlink->next->dsec) ||
1979: strcmp(mlink->fsec, mlink->next->fsec) ||
1980: strcmp(mlink->arch, mlink->next->arch)) {
1981: putchar('(');
1982: if ('\0' == *mlink->dsec)
1983: fputs(mlink->fsec, stdout);
1984: else
1985: fputs(mlink->dsec, stdout);
1986: if ('\0' != *mlink->arch)
1987: printf("/%s", mlink->arch);
1988: putchar(')');
1989: }
1990: mlink = mlink->next;
1991: if (NULL != mlink)
1992: fputs(", ", stdout);
1993: }
1.89 schwarze 1994: printf(" - %s\n", mpage->desc);
1.11 schwarze 1995: return;
1.85 schwarze 1996: }
1997:
1998: if (debug)
1999: say(mlink->file, "Adding to database");
1.47 schwarze 2000:
1.105 schwarze 2001: i = strlen(mpage->desc) + 1;
2002: key = mandoc_calloc(1, sizeof(struct str) + i);
2003: memcpy(key->key, mpage->desc, i);
2004: render_key(mc, key);
2005:
1.47 schwarze 2006: i = 1;
1.105 schwarze 2007: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, key->rendered);
1.116 schwarze 2008: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
1.47 schwarze 2009: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.94 schwarze 2010: mpage->pageid = sqlite3_last_insert_rowid(db);
1.47 schwarze 2011: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1.105 schwarze 2012:
2013: if (key->rendered != key->key)
2014: free(key->rendered);
2015: free(key);
1.47 schwarze 2016:
1.85 schwarze 2017: while (NULL != mlink) {
1.75 schwarze 2018: dbadd_mlink(mlink);
1.85 schwarze 2019: mlink = mlink->next;
2020: }
1.91 schwarze 2021: mlink = mpage->mlinks;
1.47 schwarze 2022:
1.90 schwarze 2023: for (key = ohash_first(&names, &slot); NULL != key;
2024: key = ohash_next(&names, &slot)) {
2025: assert(key->mpage == mpage);
2026: if (NULL == key->rendered)
2027: render_key(mc, key);
2028: i = 1;
2029: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2030: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->rendered);
1.94 schwarze 2031: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
1.90 schwarze 2032: SQL_STEP(stmts[STMT_INSERT_NAME]);
2033: sqlite3_reset(stmts[STMT_INSERT_NAME]);
2034: if (key->rendered != key->key)
2035: free(key->rendered);
2036: free(key);
2037: }
1.47 schwarze 2038: for (key = ohash_first(&strings, &slot); NULL != key;
2039: key = ohash_next(&strings, &slot)) {
2040: assert(key->mpage == mpage);
1.53 schwarze 2041: if (NULL == key->rendered)
2042: render_key(mc, key);
1.47 schwarze 2043: i = 1;
2044: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.53 schwarze 2045: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
1.94 schwarze 2046: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
1.47 schwarze 2047: SQL_STEP(stmts[STMT_INSERT_KEY]);
2048: sqlite3_reset(stmts[STMT_INSERT_KEY]);
1.53 schwarze 2049: if (key->rendered != key->key)
2050: free(key->rendered);
1.47 schwarze 2051: free(key);
1.33 schwarze 2052: }
1.47 schwarze 2053: }
1.41 deraadt 2054:
1.47 schwarze 2055: static void
2056: dbprune(void)
2057: {
2058: struct mpage *mpage;
2059: struct mlink *mlink;
2060: size_t i;
2061: unsigned int slot;
1.11 schwarze 2062:
1.63 schwarze 2063: if (0 == nodb)
2064: SQL_EXEC("BEGIN TRANSACTION");
1.47 schwarze 2065:
1.63 schwarze 2066: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2067: mpage = ohash_next(&mpages, &slot)) {
1.47 schwarze 2068: mlink = mpage->mlinks;
1.82 schwarze 2069: if (debug)
1.63 schwarze 2070: say(mlink->file, "Deleting from database");
2071: if (nodb)
2072: continue;
2073: for ( ; NULL != mlink; mlink = mlink->next) {
2074: i = 1;
2075: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2076: i, mlink->dsec);
2077: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2078: i, mlink->arch);
2079: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2080: i, mlink->name);
2081: SQL_STEP(stmts[STMT_DELETE_PAGE]);
2082: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2083: }
1.11 schwarze 2084: }
1.63 schwarze 2085:
2086: if (0 == nodb)
2087: SQL_EXEC("END TRANSACTION");
1.47 schwarze 2088: }
1.22 schwarze 2089:
1.47 schwarze 2090: /*
2091: * Close an existing database and its prepared statements.
2092: * If "real" is not set, rename the temporary file into the real one.
2093: */
2094: static void
2095: dbclose(int real)
2096: {
2097: size_t i;
1.72 schwarze 2098: int status;
2099: pid_t child;
1.11 schwarze 2100:
1.47 schwarze 2101: if (nodb)
2102: return;
1.11 schwarze 2103:
1.47 schwarze 2104: for (i = 0; i < STMT__MAX; i++) {
2105: sqlite3_finalize(stmts[i]);
2106: stmts[i] = NULL;
1.28 schwarze 2107: }
1.22 schwarze 2108:
1.47 schwarze 2109: sqlite3_close(db);
2110: db = NULL;
1.11 schwarze 2111:
1.47 schwarze 2112: if (real)
2113: return;
1.22 schwarze 2114:
1.72 schwarze 2115: if ('\0' == *tempfilename) {
2116: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2117: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2118: say(MANDOC_DB, "&rename");
1.72 schwarze 2119: }
2120: return;
2121: }
2122:
2123: switch (child = fork()) {
1.98 schwarze 2124: case -1:
1.72 schwarze 2125: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2126: say("", "&fork cmp");
1.72 schwarze 2127: return;
1.98 schwarze 2128: case 0:
1.72 schwarze 2129: execlp("cmp", "cmp", "-s",
2130: tempfilename, MANDOC_DB, NULL);
1.80 schwarze 2131: say("", "&exec cmp");
1.72 schwarze 2132: exit(0);
2133: default:
2134: break;
2135: }
2136: if (-1 == waitpid(child, &status, 0)) {
2137: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2138: say("", "&wait cmp");
1.72 schwarze 2139: } else if (WIFSIGNALED(status)) {
2140: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2141: say("", "cmp died from signal %d", WTERMSIG(status));
1.72 schwarze 2142: } else if (WEXITSTATUS(status)) {
1.47 schwarze 2143: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2144: say(MANDOC_DB,
2145: "Data changed, but cannot replace database");
2146: }
2147:
2148: *strrchr(tempfilename, '/') = '\0';
2149: switch (child = fork()) {
1.98 schwarze 2150: case -1:
1.72 schwarze 2151: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2152: say("", "&fork rm");
1.72 schwarze 2153: return;
1.98 schwarze 2154: case 0:
1.72 schwarze 2155: execlp("rm", "rm", "-rf", tempfilename, NULL);
1.80 schwarze 2156: say("", "&exec rm");
1.72 schwarze 2157: exit((int)MANDOCLEVEL_SYSERR);
2158: default:
2159: break;
2160: }
2161: if (-1 == waitpid(child, &status, 0)) {
2162: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2163: say("", "&wait rm");
1.72 schwarze 2164: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2165: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2166: say("", "%s: Cannot remove temporary directory",
2167: tempfilename);
1.22 schwarze 2168: }
1.11 schwarze 2169: }
2170:
1.47 schwarze 2171: /*
2172: * This is straightforward stuff.
2173: * Open a database connection to a "temporary" database, then open a set
2174: * of prepared statements we'll use over and over again.
2175: * If "real" is set, we use the existing database; if not, we truncate a
2176: * temporary one.
2177: * Must be matched by dbclose().
2178: */
2179: static int
2180: dbopen(int real)
1.2 schwarze 2181: {
1.72 schwarze 2182: const char *sql;
1.47 schwarze 2183: int rc, ofl;
1.6 schwarze 2184:
1.98 schwarze 2185: if (nodb)
1.47 schwarze 2186: return(1);
1.6 schwarze 2187:
1.72 schwarze 2188: *tempfilename = '\0';
1.47 schwarze 2189: ofl = SQLITE_OPEN_READWRITE;
1.72 schwarze 2190:
2191: if (real) {
2192: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2193: if (SQLITE_OK != rc) {
1.47 schwarze 2194: exitcode = (int)MANDOCLEVEL_SYSERR;
1.108 sthen 2195: if (SQLITE_CANTOPEN != rc)
2196: say(MANDOC_DB, "%s", sqlite3_errstr(rc));
1.47 schwarze 2197: return(0);
1.28 schwarze 2198: }
1.72 schwarze 2199: goto prepare_statements;
2200: }
2201:
2202: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.6 schwarze 2203:
1.72 schwarze 2204: remove(MANDOC_DB "~");
2205: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.98 schwarze 2206: if (SQLITE_OK == rc)
1.72 schwarze 2207: goto create_tables;
1.73 schwarze 2208: if (MPARSE_QUICK & mparse_options) {
1.47 schwarze 2209: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2210: say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
1.47 schwarze 2211: return(0);
2212: }
1.6 schwarze 2213:
1.100 schwarze 2214: (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2215: sizeof(tempfilename));
1.72 schwarze 2216: if (NULL == mkdtemp(tempfilename)) {
2217: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2218: say("", "&%s", tempfilename);
1.72 schwarze 2219: return(0);
2220: }
1.100 schwarze 2221: (void)strlcat(tempfilename, "/" MANDOC_DB,
2222: sizeof(tempfilename));
1.72 schwarze 2223: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2224: if (SQLITE_OK != rc) {
1.47 schwarze 2225: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2226: say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
1.47 schwarze 2227: return(0);
1.2 schwarze 2228: }
2229:
1.72 schwarze 2230: create_tables:
1.47 schwarze 2231: sql = "CREATE TABLE \"mpages\" (\n"
1.89 schwarze 2232: " \"desc\" TEXT NOT NULL,\n"
1.47 schwarze 2233: " \"form\" INTEGER NOT NULL,\n"
1.94 schwarze 2234: " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1.47 schwarze 2235: ");\n"
2236: "\n"
2237: "CREATE TABLE \"mlinks\" (\n"
2238: " \"sec\" TEXT NOT NULL,\n"
2239: " \"arch\" TEXT NOT NULL,\n"
2240: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2241: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2242: "ON DELETE CASCADE\n"
1.47 schwarze 2243: ");\n"
1.93 schwarze 2244: "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
1.47 schwarze 2245: "\n"
1.90 schwarze 2246: "CREATE TABLE \"names\" (\n"
2247: " \"bits\" INTEGER NOT NULL,\n"
2248: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2249: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.90 schwarze 2250: "ON DELETE CASCADE\n"
2251: ");\n"
2252: "\n"
1.47 schwarze 2253: "CREATE TABLE \"keys\" (\n"
2254: " \"bits\" INTEGER NOT NULL,\n"
2255: " \"key\" TEXT NOT NULL,\n"
1.94 schwarze 2256: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2257: "ON DELETE CASCADE\n"
1.93 schwarze 2258: ");\n"
2259: "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
1.47 schwarze 2260:
2261: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2262: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2263: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.103 schwarze 2264: sqlite3_close(db);
1.47 schwarze 2265: return(0);
1.2 schwarze 2266: }
2267:
1.47 schwarze 2268: prepare_statements:
1.103 schwarze 2269: if (SQLITE_OK != sqlite3_exec(db,
2270: "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2271: exitcode = (int)MANDOCLEVEL_SYSERR;
2272: say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2273: sqlite3_errmsg(db));
2274: sqlite3_close(db);
2275: return(0);
2276: }
2277:
1.94 schwarze 2278: sql = "DELETE FROM mpages WHERE pageid IN "
1.63 schwarze 2279: "(SELECT pageid FROM mlinks WHERE "
2280: "sec=? AND arch=? AND name=?)";
1.47 schwarze 2281: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2282: sql = "INSERT INTO mpages "
1.89 schwarze 2283: "(desc,form) VALUES (?,?)";
1.47 schwarze 2284: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2285: sql = "INSERT INTO mlinks "
1.61 schwarze 2286: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.47 schwarze 2287: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.90 schwarze 2288: sql = "INSERT INTO names "
2289: "(bits,name,pageid) VALUES (?,?,?)";
2290: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
1.47 schwarze 2291: sql = "INSERT INTO keys "
2292: "(bits,key,pageid) VALUES (?,?,?)";
2293: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.6 schwarze 2294:
1.47 schwarze 2295: /*
2296: * When opening a new database, we can turn off
2297: * synchronous mode for much better performance.
2298: */
1.6 schwarze 2299:
1.103 schwarze 2300: if (real && SQLITE_OK != sqlite3_exec(db,
2301: "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2302: exitcode = (int)MANDOCLEVEL_SYSERR;
2303: say(MANDOC_DB, "PRAGMA synchronous: %s",
2304: sqlite3_errmsg(db));
2305: sqlite3_close(db);
2306: return(0);
2307: }
1.11 schwarze 2308:
1.47 schwarze 2309: return(1);
2310: }
1.6 schwarze 2311:
1.47 schwarze 2312: static void *
1.106 espie 2313: hash_calloc(size_t n, size_t sz, void *arg)
1.47 schwarze 2314: {
1.6 schwarze 2315:
1.106 espie 2316: return(mandoc_calloc(n, sz));
1.47 schwarze 2317: }
1.2 schwarze 2318:
1.47 schwarze 2319: static void *
2320: hash_alloc(size_t sz, void *arg)
2321: {
1.28 schwarze 2322:
1.47 schwarze 2323: return(mandoc_malloc(sz));
2324: }
1.6 schwarze 2325:
1.47 schwarze 2326: static void
1.106 espie 2327: hash_free(void *p, void *arg)
1.47 schwarze 2328: {
1.26 schwarze 2329:
1.47 schwarze 2330: free(p);
2331: }
1.6 schwarze 2332:
1.47 schwarze 2333: static int
2334: set_basedir(const char *targetdir)
2335: {
2336: static char startdir[PATH_MAX];
1.110 schwarze 2337: static int getcwd_status; /* 1 = ok, 2 = failure */
2338: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2339: char *cp;
1.6 schwarze 2340:
1.47 schwarze 2341: /*
1.110 schwarze 2342: * Remember the original working directory, if possible.
2343: * This will be needed if the second or a later directory
2344: * on the command line is given as a relative path.
2345: * Do not error out if the current directory is not
2346: * searchable: Maybe it won't be needed after all.
2347: */
2348: if (0 == getcwd_status) {
2349: if (NULL == getcwd(startdir, sizeof(startdir))) {
2350: getcwd_status = 2;
2351: (void)strlcpy(startdir, strerror(errno),
2352: sizeof(startdir));
2353: } else
2354: getcwd_status = 1;
2355: }
2356:
2357: /*
2358: * We are leaving the old base directory.
2359: * Do not use it any longer, not even for messages.
2360: */
2361: *basedir = '\0';
2362:
2363: /*
2364: * If and only if the directory was changed earlier and
2365: * the next directory to process is given as a relative path,
2366: * first go back, or bail out if that is impossible.
1.47 schwarze 2367: */
1.110 schwarze 2368: if (chdir_status && '/' != *targetdir) {
2369: if (2 == getcwd_status) {
1.47 schwarze 2370: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2371: say("", "getcwd: %s", startdir);
1.47 schwarze 2372: return(0);
2373: }
1.110 schwarze 2374: if (-1 == chdir(startdir)) {
1.47 schwarze 2375: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2376: say("", "&chdir %s", startdir);
1.47 schwarze 2377: return(0);
1.2 schwarze 2378: }
2379: }
1.110 schwarze 2380:
2381: /*
2382: * Always resolve basedir to the canonicalized absolute
2383: * pathname and append a trailing slash, such that
2384: * we can reliably check whether files are inside.
2385: */
1.47 schwarze 2386: if (NULL == realpath(targetdir, basedir)) {
2387: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 2388: say("", "&%s: realpath", targetdir);
1.47 schwarze 2389: return(0);
2390: } else if (-1 == chdir(basedir)) {
2391: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 2392: say("", "&chdir");
1.47 schwarze 2393: return(0);
1.109 schwarze 2394: }
1.110 schwarze 2395: chdir_status = 1;
1.109 schwarze 2396: cp = strchr(basedir, '\0');
2397: if ('/' != cp[-1]) {
2398: if (cp - basedir >= PATH_MAX - 1) {
2399: exitcode = (int)MANDOCLEVEL_SYSERR;
2400: say("", "Filename too long");
2401: return(0);
2402: }
2403: *cp++ = '/';
2404: *cp = '\0';
1.47 schwarze 2405: }
2406: return(1);
1.2 schwarze 2407: }
2408:
2409: static void
1.47 schwarze 2410: say(const char *file, const char *format, ...)
1.2 schwarze 2411: {
1.47 schwarze 2412: va_list ap;
1.80 schwarze 2413: int use_errno;
1.2 schwarze 2414:
1.47 schwarze 2415: if ('\0' != *basedir)
2416: fprintf(stderr, "%s", basedir);
2417: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2418: fputc('/', stderr);
1.47 schwarze 2419: if ('\0' != *file)
2420: fprintf(stderr, "%s", file);
1.31 schwarze 2421:
1.80 schwarze 2422: use_errno = 1;
2423: if (NULL != format) {
2424: switch (*format) {
1.98 schwarze 2425: case '&':
1.80 schwarze 2426: format++;
2427: break;
1.98 schwarze 2428: case '\0':
1.80 schwarze 2429: format = NULL;
2430: break;
2431: default:
2432: use_errno = 0;
2433: break;
2434: }
2435: }
2436: if (NULL != format) {
2437: if ('\0' != *basedir || '\0' != *file)
2438: fputs(": ", stderr);
2439: va_start(ap, format);
2440: vfprintf(stderr, format, ap);
2441: va_end(ap);
2442: }
2443: if (use_errno) {
2444: if ('\0' != *basedir || '\0' != *file || NULL != format)
2445: fputs(": ", stderr);
1.47 schwarze 2446: perror(NULL);
1.80 schwarze 2447: } else
2448: fputc('\n', stderr);
1.1 schwarze 2449: }