Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.155
1.155 ! schwarze 1: /* $OpenBSD: mandocdb.c,v 1.154 2015/10/12 21:16:32 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:
1.155 ! schwarze 341: if (pledge("stdio rpath wpath cpath fattr proc exec", NULL) == -1) {
! 342: perror("pledge");
! 343: return (int)MANDOCLEVEL_SYSERR;
! 344: }
! 345:
1.140 schwarze 346: memset(&conf, 0, sizeof(conf));
1.47 schwarze 347: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
348:
349: mpages_info.alloc = mlinks_info.alloc = hash_alloc;
1.106 espie 350: mpages_info.calloc = mlinks_info.calloc = hash_calloc;
1.135 schwarze 351: mpages_info.free = mlinks_info.free = hash_free;
352: mpages_info.data = mlinks_info.data = NULL;
1.47 schwarze 353:
354: mpages_info.key_offset = offsetof(struct mpage, inodev);
355: mlinks_info.key_offset = offsetof(struct mlink, file);
1.1 schwarze 356:
1.47 schwarze 357: /*
1.98 schwarze 358: * We accept a few different invocations.
1.47 schwarze 359: * The CHECKOP macro makes sure that invocation styles don't
360: * clobber each other.
361: */
362: #define CHECKOP(_op, _ch) do \
363: if (OP_DEFAULT != (_op)) { \
1.151 schwarze 364: warnx("-%c: Conflicting option", (_ch)); \
1.47 schwarze 365: goto usage; \
366: } while (/*CONSTCOND*/0)
1.10 schwarze 367:
1.47 schwarze 368: path_arg = NULL;
1.28 schwarze 369: op = OP_DEFAULT;
1.1 schwarze 370:
1.83 schwarze 371: while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
1.1 schwarze 372: switch (ch) {
1.98 schwarze 373: case 'a':
1.6 schwarze 374: use_all = 1;
375: break;
1.98 schwarze 376: case 'C':
1.47 schwarze 377: CHECKOP(op, ch);
378: path_arg = optarg;
1.28 schwarze 379: op = OP_CONFFILE;
1.25 schwarze 380: break;
1.98 schwarze 381: case 'D':
1.82 schwarze 382: debug++;
383: break;
1.98 schwarze 384: case 'd':
1.47 schwarze 385: CHECKOP(op, ch);
386: path_arg = optarg;
1.2 schwarze 387: op = OP_UPDATE;
1.1 schwarze 388: break;
1.98 schwarze 389: case 'n':
1.47 schwarze 390: nodb = 1;
391: break;
1.98 schwarze 392: case 'p':
1.83 schwarze 393: warnings = 1;
394: break;
1.98 schwarze 395: case 'Q':
1.73 schwarze 396: mparse_options |= MPARSE_QUICK;
1.59 schwarze 397: break;
1.98 schwarze 398: case 'T':
1.52 schwarze 399: if (strcmp(optarg, "utf8")) {
1.151 schwarze 400: warnx("-T%s: Unsupported output format",
401: optarg);
1.52 schwarze 402: goto usage;
403: }
404: write_utf8 = 1;
405: break;
1.98 schwarze 406: case 't':
1.47 schwarze 407: CHECKOP(op, ch);
1.28 schwarze 408: dup2(STDOUT_FILENO, STDERR_FILENO);
409: op = OP_TEST;
1.47 schwarze 410: nodb = warnings = 1;
1.28 schwarze 411: break;
1.98 schwarze 412: case 'u':
1.47 schwarze 413: CHECKOP(op, ch);
414: path_arg = optarg;
1.1 schwarze 415: op = OP_DELETE;
416: break;
1.98 schwarze 417: case 'v':
1.83 schwarze 418: /* Compatibility with espie@'s makewhatis. */
1.28 schwarze 419: break;
1.1 schwarze 420: default:
1.28 schwarze 421: goto usage;
1.1 schwarze 422: }
423:
424: argc -= optind;
425: argv += optind;
426:
1.155 ! schwarze 427: if (nodb && pledge("stdio rpath", NULL) == -1) {
! 428: perror("pledge");
! 429: return (int)MANDOCLEVEL_SYSERR;
! 430: }
! 431:
1.28 schwarze 432: if (OP_CONFFILE == op && argc > 0) {
1.151 schwarze 433: warnx("-C: Too many arguments");
1.28 schwarze 434: goto usage;
435: }
436:
1.47 schwarze 437: exitcode = (int)MANDOCLEVEL_OK;
1.131 schwarze 438: mchars = mchars_alloc();
1.137 schwarze 439: mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL,
1.131 schwarze 440: mchars, NULL);
1.47 schwarze 441: ohash_init(&mpages, 6, &mpages_info);
442: ohash_init(&mlinks, 6, &mlinks_info);
1.2 schwarze 443:
1.47 schwarze 444: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.2 schwarze 445:
1.47 schwarze 446: /*
1.109 schwarze 447: * Most of these deal with a specific directory.
1.97 schwarze 448: * Jump into that directory first.
1.47 schwarze 449: */
1.120 schwarze 450: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.47 schwarze 451: goto out;
1.97 schwarze 452:
1.95 schwarze 453: if (dbopen(1)) {
1.97 schwarze 454: /*
455: * The existing database is usable. Process
456: * all files specified on the command-line.
457: */
1.155 ! schwarze 458: if (!nodb && pledge("stdio rpath wpath cpath fattr",
! 459: NULL) == -1) {
! 460: perror("pledge");
! 461: exitcode = (int)MANDOCLEVEL_SYSERR;
! 462: goto out;
! 463: }
1.97 schwarze 464: use_all = 1;
1.95 schwarze 465: for (i = 0; i < argc; i++)
466: filescan(argv[i]);
467: if (OP_TEST != op)
468: dbprune();
469: } else {
470: /*
471: * Database missing or corrupt.
472: * Recreate from scratch.
473: */
1.97 schwarze 474: exitcode = (int)MANDOCLEVEL_OK;
1.95 schwarze 475: op = OP_DEFAULT;
476: if (0 == treescan())
477: goto out;
478: if (0 == dbopen(0))
479: goto out;
480: }
1.47 schwarze 481: if (OP_DELETE != op)
1.131 schwarze 482: mpages_merge(mp);
1.95 schwarze 483: dbclose(OP_DEFAULT == op ? 0 : 1);
1.47 schwarze 484: } else {
485: /*
486: * If we have arguments, use them as our manpaths.
487: * If we don't, grok from manpath(1) or however else
1.140 schwarze 488: * manconf_parse() wants to do it.
1.47 schwarze 489: */
490: if (argc > 0) {
1.140 schwarze 491: conf.manpath.paths = mandoc_reallocarray(NULL,
1.101 schwarze 492: argc, sizeof(char *));
1.140 schwarze 493: conf.manpath.sz = (size_t)argc;
1.47 schwarze 494: for (i = 0; i < argc; i++)
1.140 schwarze 495: conf.manpath.paths[i] = mandoc_strdup(argv[i]);
1.47 schwarze 496: } else
1.140 schwarze 497: manconf_parse(&conf, path_arg, NULL, NULL);
1.84 schwarze 498:
1.140 schwarze 499: if (conf.manpath.sz == 0) {
1.84 schwarze 500: exitcode = (int)MANDOCLEVEL_BADARG;
501: say("", "Empty manpath");
502: }
1.2 schwarze 503:
1.47 schwarze 504: /*
505: * First scan the tree rooted at a base directory, then
506: * build a new database and finally move it into place.
507: * Ignore zero-length directories and strip trailing
508: * slashes.
509: */
1.140 schwarze 510: for (j = 0; j < conf.manpath.sz; j++) {
511: sz = strlen(conf.manpath.paths[j]);
512: if (sz && conf.manpath.paths[j][sz - 1] == '/')
513: conf.manpath.paths[j][--sz] = '\0';
1.47 schwarze 514: if (0 == sz)
515: continue;
1.2 schwarze 516:
1.47 schwarze 517: if (j) {
518: ohash_init(&mpages, 6, &mpages_info);
519: ohash_init(&mlinks, 6, &mlinks_info);
520: }
1.1 schwarze 521:
1.140 schwarze 522: if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
1.120 schwarze 523: continue;
1.47 schwarze 524: if (0 == treescan())
1.120 schwarze 525: continue;
1.47 schwarze 526: if (0 == dbopen(0))
1.120 schwarze 527: continue;
1.41 deraadt 528:
1.131 schwarze 529: mpages_merge(mp);
1.104 schwarze 530: if (warnings && !nodb &&
1.87 schwarze 531: ! (MPARSE_QUICK & mparse_options))
532: names_check();
1.47 schwarze 533: dbclose(0);
1.28 schwarze 534:
1.140 schwarze 535: if (j + 1 < conf.manpath.sz) {
1.47 schwarze 536: mpages_free();
537: ohash_delete(&mpages);
538: ohash_delete(&mlinks);
539: }
1.2 schwarze 540: }
1.47 schwarze 541: }
542: out:
1.140 schwarze 543: manconf_free(&conf);
1.123 schwarze 544: mparse_free(mp);
1.131 schwarze 545: mchars_free(mchars);
1.47 schwarze 546: mpages_free();
547: ohash_delete(&mpages);
548: ohash_delete(&mlinks);
1.150 schwarze 549: return exitcode;
1.47 schwarze 550: usage:
1.83 schwarze 551: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
552: " %s [-aDnpQ] [-Tutf8] dir ...\n"
553: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
554: " %s [-Dnp] -u dir [file ...]\n"
1.59 schwarze 555: " %s [-Q] -t file ...\n",
1.151 schwarze 556: __progname, __progname, __progname,
557: __progname, __progname);
1.1 schwarze 558:
1.150 schwarze 559: return (int)MANDOCLEVEL_BADARG;
1.47 schwarze 560: }
1.1 schwarze 561:
1.47 schwarze 562: /*
563: * Scan a directory tree rooted at "basedir" for manpages.
564: * We use fts(), scanning directory parts along the way for clues to our
565: * section and architecture.
566: *
567: * If use_all has been specified, grok all files.
568: * If not, sanitise paths to the following:
569: *
1.98 schwarze 570: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 571: * or
572: * [./]cat<section>[/<arch>]/<name>.0
573: *
574: * TODO: accomodate for multi-language directories.
575: */
576: static int
577: treescan(void)
578: {
1.96 schwarze 579: char buf[PATH_MAX];
1.47 schwarze 580: FTS *f;
581: FTSENT *ff;
582: struct mlink *mlink;
1.81 schwarze 583: int dform, gzip;
1.51 schwarze 584: char *dsec, *arch, *fsec, *cp;
585: const char *path;
1.47 schwarze 586: const char *argv[2];
1.1 schwarze 587:
1.47 schwarze 588: argv[0] = ".";
589: argv[1] = (char *)NULL;
1.1 schwarze 590:
1.96 schwarze 591: f = fts_open((char * const *)argv,
592: FTS_PHYSICAL | FTS_NOCHDIR, NULL);
1.153 schwarze 593: if (f == NULL) {
1.47 schwarze 594: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 595: say("", "&fts_open");
1.150 schwarze 596: return 0;
1.47 schwarze 597: }
1.2 schwarze 598:
1.47 schwarze 599: dsec = arch = NULL;
600: dform = FORM_NONE;
1.2 schwarze 601:
1.153 schwarze 602: while ((ff = fts_read(f)) != NULL) {
1.47 schwarze 603: path = ff->fts_path + 2;
1.96 schwarze 604: switch (ff->fts_info) {
605:
606: /*
607: * Symbolic links require various sanity checks,
608: * then get handled just like regular files.
609: */
1.98 schwarze 610: case FTS_SL:
1.153 schwarze 611: if (realpath(path, buf) == NULL) {
1.96 schwarze 612: if (warnings)
613: say(path, "&realpath");
614: continue;
615: }
616: if (strstr(buf, basedir) != buf) {
617: if (warnings) say("",
618: "%s: outside base directory", buf);
619: continue;
620: }
621: /* Use logical inode to avoid mpages dupe. */
1.153 schwarze 622: if (stat(path, ff->fts_statp) == -1) {
1.96 schwarze 623: if (warnings)
624: say(path, "&stat");
625: continue;
626: }
627: /* FALLTHROUGH */
628:
1.15 schwarze 629: /*
1.47 schwarze 630: * If we're a regular file, add an mlink by using the
631: * stored directory data and handling the filename.
1.15 schwarze 632: */
1.98 schwarze 633: case FTS_F:
1.153 schwarze 634: if ( ! strcmp(path, MANDOC_DB))
1.47 schwarze 635: continue;
636: if ( ! use_all && ff->fts_level < 2) {
637: if (warnings)
638: say(path, "Extraneous file");
639: continue;
1.81 schwarze 640: }
641: gzip = 0;
642: fsec = NULL;
1.153 schwarze 643: while (fsec == NULL) {
1.81 schwarze 644: fsec = strrchr(ff->fts_name, '.');
1.153 schwarze 645: if (fsec == NULL || strcmp(fsec+1, "gz"))
1.81 schwarze 646: break;
647: gzip = 1;
648: *fsec = '\0';
649: fsec = NULL;
650: }
1.153 schwarze 651: if (fsec == NULL) {
1.47 schwarze 652: if ( ! use_all) {
653: if (warnings)
654: say(path,
655: "No filename suffix");
656: continue;
657: }
1.153 schwarze 658: } else if ( ! strcmp(++fsec, "html")) {
1.47 schwarze 659: if (warnings)
660: say(path, "Skip html");
661: continue;
1.153 schwarze 662: } else if ( ! strcmp(fsec, "ps")) {
1.47 schwarze 663: if (warnings)
664: say(path, "Skip ps");
665: continue;
1.153 schwarze 666: } else if ( ! strcmp(fsec, "pdf")) {
1.47 schwarze 667: if (warnings)
668: say(path, "Skip pdf");
669: continue;
670: } else if ( ! use_all &&
1.153 schwarze 671: ((dform == FORM_SRC &&
1.138 schwarze 672: strncmp(fsec, dsec, strlen(dsec))) ||
1.153 schwarze 673: (dform == FORM_CAT && strcmp(fsec, "0")))) {
1.47 schwarze 674: if (warnings)
675: say(path, "Wrong filename suffix");
676: continue;
677: } else
678: fsec[-1] = '\0';
1.51 schwarze 679:
1.47 schwarze 680: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.100 schwarze 681: if (strlcpy(mlink->file, path,
682: sizeof(mlink->file)) >=
683: sizeof(mlink->file)) {
684: say(path, "Filename too long");
685: free(mlink);
686: continue;
687: }
1.47 schwarze 688: mlink->dform = dform;
1.51 schwarze 689: mlink->dsec = dsec;
690: mlink->arch = arch;
691: mlink->name = ff->fts_name;
692: mlink->fsec = fsec;
1.81 schwarze 693: mlink->gzip = gzip;
1.47 schwarze 694: mlink_add(mlink, ff->fts_statp);
695: continue;
1.96 schwarze 696:
1.98 schwarze 697: case FTS_D:
698: case FTS_DP:
1.96 schwarze 699: break;
700:
701: default:
1.47 schwarze 702: if (warnings)
703: say(path, "Not a regular file");
704: continue;
705: }
706:
707: switch (ff->fts_level) {
1.98 schwarze 708: case 0:
1.47 schwarze 709: /* Ignore the root directory. */
710: break;
1.98 schwarze 711: case 1:
1.47 schwarze 712: /*
713: * This might contain manX/ or catX/.
714: * Try to infer this from the name.
715: * If we're not in use_all, enforce it.
716: */
717: cp = ff->fts_name;
1.153 schwarze 718: if (ff->fts_info == FTS_DP) {
719: dform = FORM_NONE;
720: dsec = NULL;
1.47 schwarze 721: break;
1.153 schwarze 722: }
1.15 schwarze 723:
1.153 schwarze 724: if ( ! strncmp(cp, "man", 3)) {
1.47 schwarze 725: dform = FORM_SRC;
726: dsec = cp + 3;
1.153 schwarze 727: } else if ( ! strncmp(cp, "cat", 3)) {
1.47 schwarze 728: dform = FORM_CAT;
729: dsec = cp + 3;
1.51 schwarze 730: } else {
731: dform = FORM_NONE;
732: dsec = NULL;
1.26 schwarze 733: }
1.47 schwarze 734:
1.153 schwarze 735: if (dsec != NULL || use_all)
1.47 schwarze 736: break;
737:
738: if (warnings)
739: say(path, "Unknown directory part");
740: fts_set(f, ff, FTS_SKIP);
741: break;
1.98 schwarze 742: case 2:
1.47 schwarze 743: /*
744: * Possibly our architecture.
745: * If we're descending, keep tabs on it.
746: */
1.153 schwarze 747: if (ff->fts_info != FTS_DP && dsec != NULL)
1.47 schwarze 748: arch = ff->fts_name;
1.51 schwarze 749: else
750: arch = NULL;
1.47 schwarze 751: break;
752: default:
1.153 schwarze 753: if (ff->fts_info == FTS_DP || use_all)
1.47 schwarze 754: break;
755: if (warnings)
756: say(path, "Extraneous directory part");
757: fts_set(f, ff, FTS_SKIP);
758: break;
1.14 schwarze 759: }
1.47 schwarze 760: }
761:
762: fts_close(f);
1.150 schwarze 763: return 1;
1.47 schwarze 764: }
1.1 schwarze 765:
1.47 schwarze 766: /*
767: * Add a file to the mlinks table.
768: * Do not verify that it's a "valid" looking manpage (we'll do that
769: * later).
770: *
771: * Try to infer the manual section, architecture, and page name from the
772: * path, assuming it looks like
773: *
1.98 schwarze 774: * [./]man*[/<arch>]/<name>.<section>
1.47 schwarze 775: * or
776: * [./]cat<section>[/<arch>]/<name>.0
777: *
778: * See treescan() for the fts(3) version of this.
779: */
780: static void
781: filescan(const char *file)
782: {
783: char buf[PATH_MAX];
784: struct stat st;
785: struct mlink *mlink;
786: char *p, *start;
787:
788: assert(use_all);
789:
790: if (0 == strncmp(file, "./", 2))
791: file += 2;
792:
1.96 schwarze 793: /*
794: * We have to do lstat(2) before realpath(3) loses
795: * the information whether this is a symbolic link.
796: * We need to know that because for symbolic links,
797: * we want to use the orginal file name, while for
798: * regular files, we want to use the real path.
799: */
800: if (-1 == lstat(file, &st)) {
801: exitcode = (int)MANDOCLEVEL_BADARG;
802: say(file, "&lstat");
803: return;
804: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
805: exitcode = (int)MANDOCLEVEL_BADARG;
806: say(file, "Not a regular file");
807: return;
808: }
809:
810: /*
811: * We have to resolve the file name to the real path
812: * in any case for the base directory check.
813: */
1.47 schwarze 814: if (NULL == realpath(file, buf)) {
815: exitcode = (int)MANDOCLEVEL_BADARG;
1.80 schwarze 816: say(file, "&realpath");
1.47 schwarze 817: return;
1.63 schwarze 818: }
819:
1.109 schwarze 820: if (OP_TEST == op)
1.63 schwarze 821: start = buf;
1.109 schwarze 822: else if (strstr(buf, basedir) == buf)
823: start = buf + strlen(basedir);
1.63 schwarze 824: else {
1.47 schwarze 825: exitcode = (int)MANDOCLEVEL_BADARG;
826: say("", "%s: outside base directory", buf);
827: return;
1.63 schwarze 828: }
829:
1.96 schwarze 830: /*
831: * Now we are sure the file is inside our tree.
832: * If it is a symbolic link, ignore the real path
833: * and use the original name.
834: * This implies passing stuff like "cat1/../man1/foo.1"
835: * on the command line won't work. So don't do that.
836: * Note the stat(2) can still fail if the link target
837: * doesn't exist.
838: */
839: if (S_IFLNK & st.st_mode) {
840: if (-1 == stat(buf, &st)) {
841: exitcode = (int)MANDOCLEVEL_BADARG;
842: say(file, "&stat");
843: return;
844: }
1.100 schwarze 845: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
846: say(file, "Filename too long");
847: return;
848: }
1.109 schwarze 849: start = buf;
850: if (OP_TEST != op && strstr(buf, basedir) == buf)
851: start += strlen(basedir);
1.1 schwarze 852: }
1.63 schwarze 853:
1.47 schwarze 854: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.116 schwarze 855: mlink->dform = FORM_NONE;
1.100 schwarze 856: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
857: sizeof(mlink->file)) {
858: say(start, "Filename too long");
1.134 schwarze 859: free(mlink);
1.100 schwarze 860: return;
861: }
1.1 schwarze 862:
1.10 schwarze 863: /*
1.47 schwarze 864: * First try to guess our directory structure.
865: * If we find a separator, try to look for man* or cat*.
866: * If we find one of these and what's underneath is a directory,
867: * assume it's an architecture.
1.10 schwarze 868: */
1.47 schwarze 869: if (NULL != (p = strchr(start, '/'))) {
870: *p++ = '\0';
871: if (0 == strncmp(start, "man", 3)) {
872: mlink->dform = FORM_SRC;
1.51 schwarze 873: mlink->dsec = start + 3;
1.47 schwarze 874: } else if (0 == strncmp(start, "cat", 3)) {
875: mlink->dform = FORM_CAT;
1.51 schwarze 876: mlink->dsec = start + 3;
1.47 schwarze 877: }
1.10 schwarze 878:
1.47 schwarze 879: start = p;
880: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
881: *p++ = '\0';
1.51 schwarze 882: mlink->arch = start;
1.47 schwarze 883: start = p;
1.41 deraadt 884: }
1.47 schwarze 885: }
1.7 schwarze 886:
1.47 schwarze 887: /*
888: * Now check the file suffix.
889: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
890: */
891: p = strrchr(start, '\0');
892: while (p-- > start && '/' != *p && '.' != *p)
893: /* Loop. */ ;
894:
895: if ('.' == *p) {
896: *p++ = '\0';
1.51 schwarze 897: mlink->fsec = p;
1.47 schwarze 898: }
1.41 deraadt 899:
1.47 schwarze 900: /*
901: * Now try to parse the name.
902: * Use the filename portion of the path.
903: */
904: mlink->name = start;
905: if (NULL != (p = strrchr(start, '/'))) {
906: mlink->name = p + 1;
907: *p = '\0';
908: }
909: mlink_add(mlink, &st);
910: }
1.2 schwarze 911:
1.47 schwarze 912: static void
913: mlink_add(struct mlink *mlink, const struct stat *st)
914: {
915: struct inodev inodev;
916: struct mpage *mpage;
917: unsigned int slot;
918:
919: assert(NULL != mlink->file);
920:
1.51 schwarze 921: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
922: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
923: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
924: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.47 schwarze 925:
926: if ('0' == *mlink->fsec) {
927: free(mlink->fsec);
928: mlink->fsec = mandoc_strdup(mlink->dsec);
929: mlink->fform = FORM_CAT;
930: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
931: mlink->fform = FORM_SRC;
932: else
933: mlink->fform = FORM_NONE;
1.1 schwarze 934:
1.47 schwarze 935: slot = ohash_qlookup(&mlinks, mlink->file);
936: assert(NULL == ohash_find(&mlinks, slot));
937: ohash_insert(&mlinks, slot, mlink);
938:
1.139 schwarze 939: memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1.47 schwarze 940: inodev.st_ino = st->st_ino;
941: inodev.st_dev = st->st_dev;
942: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
943: sizeof(struct inodev), inodev.st_ino);
944: mpage = ohash_find(&mpages, slot);
945: if (NULL == mpage) {
946: mpage = mandoc_calloc(1, sizeof(struct mpage));
947: mpage->inodev.st_ino = inodev.st_ino;
948: mpage->inodev.st_dev = inodev.st_dev;
949: ohash_insert(&mpages, slot, mpage);
950: } else
951: mlink->next = mpage->mlinks;
952: mpage->mlinks = mlink;
1.75 schwarze 953: mlink->mpage = mpage;
1.47 schwarze 954: }
1.1 schwarze 955:
1.47 schwarze 956: static void
957: mlink_free(struct mlink *mlink)
958: {
1.1 schwarze 959:
1.47 schwarze 960: free(mlink->dsec);
961: free(mlink->arch);
962: free(mlink->name);
963: free(mlink->fsec);
964: free(mlink);
965: }
1.1 schwarze 966:
1.47 schwarze 967: static void
968: mpages_free(void)
969: {
970: struct mpage *mpage;
971: struct mlink *mlink;
972: unsigned int slot;
973:
974: mpage = ohash_first(&mpages, &slot);
975: while (NULL != mpage) {
976: while (NULL != (mlink = mpage->mlinks)) {
977: mpage->mlinks = mlink->next;
978: mlink_free(mlink);
979: }
980: free(mpage->sec);
981: free(mpage->arch);
982: free(mpage->title);
983: free(mpage->desc);
984: free(mpage);
985: mpage = ohash_next(&mpages, &slot);
986: }
987: }
1.1 schwarze 988:
1.47 schwarze 989: /*
990: * For each mlink to the mpage, check whether the path looks like
991: * it is formatted, and if it does, check whether a source manual
992: * exists by the same name, ignoring the suffix.
993: * If both conditions hold, drop the mlink.
994: */
995: static void
996: mlinks_undupe(struct mpage *mpage)
997: {
998: char buf[PATH_MAX];
999: struct mlink **prev;
1000: struct mlink *mlink;
1001: char *bufp;
1002:
1003: mpage->form = FORM_CAT;
1004: prev = &mpage->mlinks;
1005: while (NULL != (mlink = *prev)) {
1006: if (FORM_CAT != mlink->dform) {
1007: mpage->form = FORM_NONE;
1008: goto nextlink;
1009: }
1.100 schwarze 1010: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.47 schwarze 1011: bufp = strstr(buf, "cat");
1012: assert(NULL != bufp);
1013: memcpy(bufp, "man", 3);
1014: if (NULL != (bufp = strrchr(buf, '.')))
1015: *++bufp = '\0';
1.100 schwarze 1016: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.47 schwarze 1017: if (NULL == ohash_find(&mlinks,
1.98 schwarze 1018: ohash_qlookup(&mlinks, buf)))
1.47 schwarze 1019: goto nextlink;
1020: if (warnings)
1021: say(mlink->file, "Man source exists: %s", buf);
1022: if (use_all)
1023: goto nextlink;
1024: *prev = mlink->next;
1025: mlink_free(mlink);
1026: continue;
1027: nextlink:
1028: prev = &(*prev)->next;
1.1 schwarze 1029: }
1.47 schwarze 1030: }
1.1 schwarze 1031:
1.88 schwarze 1032: static void
1.50 schwarze 1033: mlink_check(struct mpage *mpage, struct mlink *mlink)
1034: {
1.88 schwarze 1035: struct str *str;
1036: unsigned int slot;
1.50 schwarze 1037:
1038: /*
1039: * Check whether the manual section given in a file
1040: * agrees with the directory where the file is located.
1041: * Some manuals have suffixes like (3p) on their
1042: * section number either inside the file or in the
1043: * directory name, some are linked into more than one
1044: * section, like encrypt(1) = makekey(8).
1045: */
1046:
1047: if (FORM_SRC == mpage->form &&
1.88 schwarze 1048: strcasecmp(mpage->sec, mlink->dsec))
1.50 schwarze 1049: say(mlink->file, "Section \"%s\" manual in %s directory",
1050: mpage->sec, mlink->dsec);
1051:
1052: /*
1053: * Manual page directories exist for each kernel
1054: * architecture as returned by machine(1).
1055: * However, many manuals only depend on the
1056: * application architecture as returned by arch(1).
1057: * For example, some (2/ARM) manuals are shared
1058: * across the "armish" and "zaurus" kernel
1059: * architectures.
1060: * A few manuals are even shared across completely
1061: * different architectures, for example fdformat(1)
1062: * on amd64, i386, sparc, and sparc64.
1063: */
1064:
1.88 schwarze 1065: if (strcasecmp(mpage->arch, mlink->arch))
1.50 schwarze 1066: say(mlink->file, "Architecture \"%s\" manual in "
1067: "\"%s\" directory", mpage->arch, mlink->arch);
1068:
1.88 schwarze 1069: /*
1070: * XXX
1.90 schwarze 1071: * parse_cat() doesn't set NAME_TITLE yet.
1.88 schwarze 1072: */
1073:
1074: if (FORM_CAT == mpage->form)
1075: return;
1076:
1077: /*
1078: * Check whether this mlink
1079: * appears as a name in the NAME section.
1080: */
1.50 schwarze 1081:
1.90 schwarze 1082: slot = ohash_qlookup(&names, mlink->name);
1083: str = ohash_find(&names, slot);
1.88 schwarze 1084: assert(NULL != str);
1.90 schwarze 1085: if ( ! (NAME_TITLE & str->mask))
1.88 schwarze 1086: say(mlink->file, "Name missing in NAME section");
1.50 schwarze 1087: }
1088:
1.47 schwarze 1089: /*
1090: * Run through the files in the global vector "mpages"
1091: * and add them to the database specified in "basedir".
1092: *
1093: * This handles the parsing scheme itself, using the cues of directory
1094: * and filename to determine whether the file is parsable or not.
1095: */
1096: static void
1.131 schwarze 1097: mpages_merge(struct mparse *mp)
1.47 schwarze 1098: {
1.70 schwarze 1099: char any[] = "any";
1.58 schwarze 1100: struct ohash_info str_info;
1.75 schwarze 1101: struct mpage *mpage, *mpage_dest;
1102: struct mlink *mlink, *mlink_dest;
1.144 schwarze 1103: struct roff_man *man;
1.75 schwarze 1104: char *sodest;
1.69 schwarze 1105: char *cp;
1.117 schwarze 1106: int fd;
1.58 schwarze 1107: unsigned int pslot;
1.47 schwarze 1108:
1109: str_info.alloc = hash_alloc;
1.106 espie 1110: str_info.calloc = hash_calloc;
1111: str_info.free = hash_free;
1.135 schwarze 1112: str_info.data = NULL;
1.47 schwarze 1113: str_info.key_offset = offsetof(struct str, key);
1114:
1.126 schwarze 1115: if ( ! nodb)
1.64 schwarze 1116: SQL_EXEC("BEGIN TRANSACTION");
1117:
1.47 schwarze 1118: mpage = ohash_first(&mpages, &pslot);
1.126 schwarze 1119: while (mpage != NULL) {
1.47 schwarze 1120: mlinks_undupe(mpage);
1.136 schwarze 1121: if ((mlink = mpage->mlinks) == NULL) {
1.47 schwarze 1122: mpage = ohash_next(&mpages, &pslot);
1123: continue;
1124: }
1.1 schwarze 1125:
1.90 schwarze 1126: name_mask = NAME_MASK;
1127: ohash_init(&names, 4, &str_info);
1.47 schwarze 1128: ohash_init(&strings, 6, &str_info);
1129: mparse_reset(mp);
1130: man = NULL;
1.81 schwarze 1131: sodest = NULL;
1132:
1.136 schwarze 1133: mparse_open(mp, &fd, mlink->file);
1.117 schwarze 1134: if (fd == -1) {
1.136 schwarze 1135: say(mlink->file, "&open");
1.117 schwarze 1136: goto nextpage;
1.81 schwarze 1137: }
1.11 schwarze 1138:
1139: /*
1.137 schwarze 1140: * Interpret the file as mdoc(7) or man(7) source
1141: * code, unless it is known to be formatted.
1.11 schwarze 1142: */
1.136 schwarze 1143: if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1.137 schwarze 1144: mparse_readfd(mp, fd, mlink->file);
1.145 schwarze 1145: mparse_result(mp, &man, &sodest);
1.47 schwarze 1146: }
1.11 schwarze 1147:
1.126 schwarze 1148: if (sodest != NULL) {
1.75 schwarze 1149: mlink_dest = ohash_find(&mlinks,
1150: ohash_qlookup(&mlinks, sodest));
1.126 schwarze 1151: if (mlink_dest == NULL) {
1152: mandoc_asprintf(&cp, "%s.gz", sodest);
1153: mlink_dest = ohash_find(&mlinks,
1154: ohash_qlookup(&mlinks, cp));
1155: free(cp);
1156: }
1157: if (mlink_dest != NULL) {
1.75 schwarze 1158:
1159: /* The .so target exists. */
1160:
1161: mpage_dest = mlink_dest->mpage;
1162: while (1) {
1163: mlink->mpage = mpage_dest;
1164:
1165: /*
1166: * If the target was already
1167: * processed, add the links
1168: * to the database now.
1169: * Otherwise, this will
1170: * happen when we come
1171: * to the target.
1172: */
1173:
1.94 schwarze 1174: if (mpage_dest->pageid)
1.124 schwarze 1175: dbadd_mlink_name(mlink);
1.75 schwarze 1176:
1.126 schwarze 1177: if (mlink->next == NULL)
1.75 schwarze 1178: break;
1179: mlink = mlink->next;
1180: }
1181:
1182: /* Move all links to the target. */
1183:
1184: mlink->next = mlink_dest->next;
1185: mlink_dest->next = mpage->mlinks;
1186: mpage->mlinks = NULL;
1187: }
1.81 schwarze 1188: goto nextpage;
1.145 schwarze 1189: } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1.47 schwarze 1190: mpage->form = FORM_SRC;
1.146 schwarze 1191: mpage->sec = man->meta.msec;
1.112 schwarze 1192: mpage->sec = mandoc_strdup(
1.126 schwarze 1193: mpage->sec == NULL ? "" : mpage->sec);
1.146 schwarze 1194: mpage->arch = man->meta.arch;
1.47 schwarze 1195: mpage->arch = mandoc_strdup(
1.126 schwarze 1196: mpage->arch == NULL ? "" : mpage->arch);
1.146 schwarze 1197: mpage->title = mandoc_strdup(man->meta.title);
1.145 schwarze 1198: } else if (man != NULL && man->macroset == MACROSET_MAN) {
1.47 schwarze 1199: mpage->form = FORM_SRC;
1.146 schwarze 1200: mpage->sec = mandoc_strdup(man->meta.msec);
1.136 schwarze 1201: mpage->arch = mandoc_strdup(mlink->arch);
1.146 schwarze 1202: mpage->title = mandoc_strdup(man->meta.title);
1.11 schwarze 1203: } else {
1.47 schwarze 1204: mpage->form = FORM_CAT;
1.136 schwarze 1205: mpage->sec = mandoc_strdup(mlink->dsec);
1206: mpage->arch = mandoc_strdup(mlink->arch);
1207: mpage->title = mandoc_strdup(mlink->name);
1.1 schwarze 1208: }
1.54 schwarze 1209: putkey(mpage, mpage->sec, TYPE_sec);
1.119 schwarze 1210: if (*mpage->arch != '\0')
1211: putkey(mpage, mpage->arch, TYPE_arch);
1.1 schwarze 1212:
1.136 schwarze 1213: for ( ; mlink != NULL; mlink = mlink->next) {
1.54 schwarze 1214: if ('\0' != *mlink->dsec)
1215: putkey(mpage, mlink->dsec, TYPE_sec);
1216: if ('\0' != *mlink->fsec)
1217: putkey(mpage, mlink->fsec, TYPE_sec);
1.55 schwarze 1218: putkey(mpage, '\0' == *mlink->arch ?
1.70 schwarze 1219: any : mlink->arch, TYPE_arch);
1.90 schwarze 1220: putkey(mpage, mlink->name, NAME_FILE);
1.54 schwarze 1221: }
1.41 deraadt 1222:
1.129 schwarze 1223: assert(mpage->desc == NULL);
1.145 schwarze 1224: if (man != NULL && man->macroset == MACROSET_MDOC)
1.146 schwarze 1225: parse_mdoc(mpage, &man->meta, man->first);
1.129 schwarze 1226: else if (man != NULL)
1.146 schwarze 1227: parse_man(mpage, &man->meta, man->first);
1.47 schwarze 1228: else
1.117 schwarze 1229: parse_cat(mpage, fd);
1.129 schwarze 1230: if (mpage->desc == NULL)
1.92 schwarze 1231: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1.88 schwarze 1232:
1233: if (warnings && !use_all)
1234: for (mlink = mpage->mlinks; mlink;
1235: mlink = mlink->next)
1236: mlink_check(mpage, mlink);
1.6 schwarze 1237:
1.131 schwarze 1238: dbadd(mpage);
1.136 schwarze 1239: mlink = mpage->mlinks;
1.81 schwarze 1240:
1241: nextpage:
1.47 schwarze 1242: ohash_delete(&strings);
1.90 schwarze 1243: ohash_delete(&names);
1.47 schwarze 1244: mpage = ohash_next(&mpages, &pslot);
1245: }
1.64 schwarze 1246:
1247: if (0 == nodb)
1248: SQL_EXEC("END TRANSACTION");
1.87 schwarze 1249: }
1250:
1251: static void
1252: names_check(void)
1253: {
1254: sqlite3_stmt *stmt;
1255: const char *name, *sec, *arch, *key;
1256:
1257: sqlite3_prepare_v2(db,
1258: "SELECT name, sec, arch, key FROM ("
1.90 schwarze 1259: "SELECT name AS key, pageid FROM names "
1.87 schwarze 1260: "WHERE bits & ? AND NOT EXISTS ("
1261: "SELECT pageid FROM mlinks "
1.90 schwarze 1262: "WHERE mlinks.pageid == names.pageid "
1263: "AND mlinks.name == names.name"
1.87 schwarze 1264: ")"
1265: ") JOIN ("
1.99 schwarze 1266: "SELECT sec, arch, name, pageid FROM mlinks "
1267: "GROUP BY pageid"
1.87 schwarze 1268: ") USING (pageid);",
1269: -1, &stmt, NULL);
1270:
1.154 schwarze 1271: if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK)
1.91 schwarze 1272: say("", "%s", sqlite3_errmsg(db));
1.87 schwarze 1273:
1.154 schwarze 1274: while (sqlite3_step(stmt) == SQLITE_ROW) {
1.113 schwarze 1275: name = (const char *)sqlite3_column_text(stmt, 0);
1276: sec = (const char *)sqlite3_column_text(stmt, 1);
1277: arch = (const char *)sqlite3_column_text(stmt, 2);
1278: key = (const char *)sqlite3_column_text(stmt, 3);
1.87 schwarze 1279: say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1280: '\0' == *arch ? "" : "/",
1281: '\0' == *arch ? "" : arch, key);
1282: }
1283: sqlite3_finalize(stmt);
1.47 schwarze 1284: }
1.6 schwarze 1285:
1.47 schwarze 1286: static void
1.81 schwarze 1287: parse_cat(struct mpage *mpage, int fd)
1.47 schwarze 1288: {
1289: FILE *stream;
1290: char *line, *p, *title;
1291: size_t len, plen, titlesz;
1.1 schwarze 1292:
1.81 schwarze 1293: stream = (-1 == fd) ?
1294: fopen(mpage->mlinks->file, "r") :
1295: fdopen(fd, "r");
1296: if (NULL == stream) {
1.114 doug 1297: if (-1 != fd)
1298: close(fd);
1.47 schwarze 1299: if (warnings)
1.80 schwarze 1300: say(mpage->mlinks->file, "&fopen");
1.47 schwarze 1301: return;
1302: }
1.1 schwarze 1303:
1.47 schwarze 1304: /* Skip to first blank line. */
1.1 schwarze 1305:
1.47 schwarze 1306: while (NULL != (line = fgetln(stream, &len)))
1307: if ('\n' == *line)
1308: break;
1.1 schwarze 1309:
1.47 schwarze 1310: /*
1311: * Assume the first line that is not indented
1312: * is the first section header. Skip to it.
1313: */
1.1 schwarze 1314:
1.47 schwarze 1315: while (NULL != (line = fgetln(stream, &len)))
1316: if ('\n' != *line && ' ' != *line)
1317: break;
1.98 schwarze 1318:
1.47 schwarze 1319: /*
1320: * Read up until the next section into a buffer.
1321: * Strip the leading and trailing newline from each read line,
1322: * appending a trailing space.
1323: * Ignore empty (whitespace-only) lines.
1324: */
1.28 schwarze 1325:
1.47 schwarze 1326: titlesz = 0;
1327: title = NULL;
1.38 schwarze 1328:
1.47 schwarze 1329: while (NULL != (line = fgetln(stream, &len))) {
1330: if (' ' != *line || '\n' != line[len - 1])
1331: break;
1332: while (len > 0 && isspace((unsigned char)*line)) {
1333: line++;
1334: len--;
1335: }
1336: if (1 == len)
1337: continue;
1338: title = mandoc_realloc(title, titlesz + len);
1339: memcpy(title + titlesz, line, len);
1340: titlesz += len;
1341: title[titlesz - 1] = ' ';
1342: }
1.28 schwarze 1343:
1.47 schwarze 1344: /*
1345: * If no page content can be found, or the input line
1346: * is already the next section header, or there is no
1347: * trailing newline, reuse the page title as the page
1348: * description.
1349: */
1.1 schwarze 1350:
1.47 schwarze 1351: if (NULL == title || '\0' == *title) {
1352: if (warnings)
1353: say(mpage->mlinks->file,
1354: "Cannot find NAME section");
1355: fclose(stream);
1356: free(title);
1357: return;
1358: }
1.24 schwarze 1359:
1.47 schwarze 1360: title = mandoc_realloc(title, titlesz + 1);
1361: title[titlesz] = '\0';
1.24 schwarze 1362:
1.47 schwarze 1363: /*
1364: * Skip to the first dash.
1365: * Use the remaining line as the description (no more than 70
1366: * bytes).
1367: */
1.28 schwarze 1368:
1.47 schwarze 1369: if (NULL != (p = strstr(title, "- "))) {
1370: for (p += 2; ' ' == *p || '\b' == *p; p++)
1371: /* Skip to next word. */ ;
1372: } else {
1373: if (warnings)
1374: say(mpage->mlinks->file,
1375: "No dash in title line");
1376: p = title;
1377: }
1.1 schwarze 1378:
1.47 schwarze 1379: plen = strlen(p);
1.1 schwarze 1380:
1.47 schwarze 1381: /* Strip backspace-encoding from line. */
1.1 schwarze 1382:
1.47 schwarze 1383: while (NULL != (line = memchr(p, '\b', plen))) {
1384: len = line - p;
1385: if (0 == len) {
1386: memmove(line, line + 1, plen--);
1387: continue;
1.98 schwarze 1388: }
1.47 schwarze 1389: memmove(line - 1, line + 1, plen - len);
1390: plen -= 2;
1391: }
1.1 schwarze 1392:
1.47 schwarze 1393: mpage->desc = mandoc_strdup(p);
1394: fclose(stream);
1395: free(title);
1396: }
1.16 schwarze 1397:
1.47 schwarze 1398: /*
1399: * Put a type/word pair into the word database for this particular file.
1400: */
1401: static void
1.69 schwarze 1402: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.47 schwarze 1403: {
1.69 schwarze 1404: char *cp;
1.37 schwarze 1405:
1.47 schwarze 1406: assert(NULL != value);
1.69 schwarze 1407: if (TYPE_arch == type)
1408: for (cp = value; *cp; cp++)
1409: if (isupper((unsigned char)*cp))
1410: *cp = _tolower((unsigned char)*cp);
1.47 schwarze 1411: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1412: }
1413:
1414: /*
1.47 schwarze 1415: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1416: */
1417: static void
1.47 schwarze 1418: putmdockey(const struct mpage *mpage,
1.142 schwarze 1419: const struct roff_node *n, uint64_t m)
1.2 schwarze 1420: {
1.16 schwarze 1421:
1.47 schwarze 1422: for ( ; NULL != n; n = n->next) {
1423: if (NULL != n->child)
1424: putmdockey(mpage, n->child, m);
1.141 schwarze 1425: if (n->type == ROFFT_TEXT)
1.47 schwarze 1426: putkey(mpage, n->string, m);
1427: }
1428: }
1.16 schwarze 1429:
1.47 schwarze 1430: static void
1.143 schwarze 1431: parse_man(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1432: const struct roff_node *n)
1.47 schwarze 1433: {
1.142 schwarze 1434: const struct roff_node *head, *body;
1.78 schwarze 1435: char *start, *title;
1.47 schwarze 1436: char byte;
1.78 schwarze 1437: size_t sz;
1.16 schwarze 1438:
1.47 schwarze 1439: if (NULL == n)
1440: return;
1.16 schwarze 1441:
1.47 schwarze 1442: /*
1443: * We're only searching for one thing: the first text child in
1444: * the BODY of a NAME section. Since we don't keep track of
1445: * sections in -man, run some hoops to find out whether we're in
1446: * the correct section or not.
1447: */
1.16 schwarze 1448:
1.141 schwarze 1449: if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1.47 schwarze 1450: body = n;
1451: assert(body->parent);
1452: if (NULL != (head = body->parent->head) &&
1.98 schwarze 1453: 1 == head->nchild &&
1454: NULL != (head = (head->child)) &&
1.141 schwarze 1455: head->type == ROFFT_TEXT &&
1.98 schwarze 1456: 0 == strcmp(head->string, "NAME") &&
1457: NULL != body->child) {
1.2 schwarze 1458:
1.47 schwarze 1459: /*
1460: * Suck the entire NAME section into memory.
1461: * Yes, we might run away.
1462: * But too many manuals have big, spread-out
1463: * NAME sections over many lines.
1464: */
1.2 schwarze 1465:
1.78 schwarze 1466: title = NULL;
1.147 schwarze 1467: deroff(&title, body);
1.47 schwarze 1468: if (NULL == title)
1469: return;
1.16 schwarze 1470:
1.98 schwarze 1471: /*
1.47 schwarze 1472: * Go through a special heuristic dance here.
1473: * Conventionally, one or more manual names are
1474: * comma-specified prior to a whitespace, then a
1475: * dash, then a description. Try to puzzle out
1476: * the name parts here.
1477: */
1.16 schwarze 1478:
1.78 schwarze 1479: start = title;
1.47 schwarze 1480: for ( ;; ) {
1481: sz = strcspn(start, " ,");
1482: if ('\0' == start[sz])
1483: break;
1.1 schwarze 1484:
1.47 schwarze 1485: byte = start[sz];
1486: start[sz] = '\0';
1.67 schwarze 1487:
1488: /*
1489: * Assume a stray trailing comma in the
1490: * name list if a name begins with a dash.
1491: */
1492:
1493: if ('-' == start[0] ||
1494: ('\\' == start[0] && '-' == start[1]))
1495: break;
1.1 schwarze 1496:
1.90 schwarze 1497: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1498: if ( ! (mpage->name_head_done ||
1499: strcasecmp(start, meta->title))) {
1500: putkey(mpage, start, NAME_HEAD);
1501: mpage->name_head_done = 1;
1502: }
1.1 schwarze 1503:
1.47 schwarze 1504: if (' ' == byte) {
1505: start += sz + 1;
1506: break;
1507: }
1.1 schwarze 1508:
1.47 schwarze 1509: assert(',' == byte);
1510: start += sz + 1;
1511: while (' ' == *start)
1512: start++;
1513: }
1.1 schwarze 1514:
1.78 schwarze 1515: if (start == title) {
1.90 schwarze 1516: putkey(mpage, start, NAME_TITLE);
1.129 schwarze 1517: if ( ! (mpage->name_head_done ||
1518: strcasecmp(start, meta->title))) {
1519: putkey(mpage, start, NAME_HEAD);
1520: mpage->name_head_done = 1;
1521: }
1.47 schwarze 1522: free(title);
1523: return;
1524: }
1.1 schwarze 1525:
1.47 schwarze 1526: while (isspace((unsigned char)*start))
1527: start++;
1.1 schwarze 1528:
1.47 schwarze 1529: if (0 == strncmp(start, "-", 1))
1530: start += 1;
1531: else if (0 == strncmp(start, "\\-\\-", 4))
1532: start += 4;
1533: else if (0 == strncmp(start, "\\-", 2))
1534: start += 2;
1535: else if (0 == strncmp(start, "\\(en", 4))
1536: start += 4;
1537: else if (0 == strncmp(start, "\\(em", 4))
1538: start += 4;
1.1 schwarze 1539:
1.47 schwarze 1540: while (' ' == *start)
1541: start++;
1.1 schwarze 1542:
1.47 schwarze 1543: mpage->desc = mandoc_strdup(start);
1544: free(title);
1545: return;
1546: }
1547: }
1.1 schwarze 1548:
1.47 schwarze 1549: for (n = n->child; n; n = n->next) {
1550: if (NULL != mpage->desc)
1551: break;
1.129 schwarze 1552: parse_man(mpage, meta, n);
1.1 schwarze 1553: }
1554: }
1555:
1556: static void
1.143 schwarze 1557: parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1558: const struct roff_node *n)
1.1 schwarze 1559: {
1560:
1.47 schwarze 1561: assert(NULL != n);
1562: for (n = n->child; NULL != n; n = n->next) {
1563: switch (n->type) {
1.141 schwarze 1564: case ROFFT_ELEM:
1565: case ROFFT_BLOCK:
1566: case ROFFT_HEAD:
1567: case ROFFT_BODY:
1568: case ROFFT_TAIL:
1.47 schwarze 1569: if (NULL != mdocs[n->tok].fp)
1.127 schwarze 1570: if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1.47 schwarze 1571: break;
1572: if (mdocs[n->tok].mask)
1573: putmdockey(mpage, n->child,
1574: mdocs[n->tok].mask);
1575: break;
1576: default:
1.141 schwarze 1577: assert(n->type != ROFFT_ROOT);
1.47 schwarze 1578: continue;
1579: }
1580: if (NULL != n->child)
1.127 schwarze 1581: parse_mdoc(mpage, meta, n);
1.1 schwarze 1582: }
1583: }
1584:
1.19 schwarze 1585: static int
1.143 schwarze 1586: parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1587: const struct roff_node *n)
1.1 schwarze 1588: {
1.131 schwarze 1589: char *start, *end;
1.1 schwarze 1590: size_t sz;
1.19 schwarze 1591:
1.47 schwarze 1592: if (SEC_SYNOPSIS != n->sec ||
1.98 schwarze 1593: NULL == (n = n->child) ||
1.141 schwarze 1594: n->type != ROFFT_TEXT)
1.150 schwarze 1595: return 0;
1.1 schwarze 1596:
1597: /*
1598: * Only consider those `Fd' macro fields that begin with an
1599: * "inclusion" token (versus, e.g., #define).
1600: */
1.47 schwarze 1601:
1.1 schwarze 1602: if (strcmp("#include", n->string))
1.150 schwarze 1603: return 0;
1.1 schwarze 1604:
1.141 schwarze 1605: if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1.150 schwarze 1606: return 0;
1.1 schwarze 1607:
1608: /*
1609: * Strip away the enclosing angle brackets and make sure we're
1610: * not zero-length.
1611: */
1612:
1613: start = n->string;
1614: if ('<' == *start || '"' == *start)
1615: start++;
1616:
1617: if (0 == (sz = strlen(start)))
1.150 schwarze 1618: return 0;
1.1 schwarze 1619:
1620: end = &start[(int)sz - 1];
1621: if ('>' == *end || '"' == *end)
1622: end--;
1623:
1.47 schwarze 1624: if (end > start)
1625: putkeys(mpage, start, end - start + 1, TYPE_In);
1.150 schwarze 1626: return 0;
1.1 schwarze 1627: }
1628:
1.133 schwarze 1629: static void
1.142 schwarze 1630: parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1.1 schwarze 1631: {
1.69 schwarze 1632: char *cp;
1.133 schwarze 1633: size_t sz;
1.1 schwarze 1634:
1.141 schwarze 1635: if (n->type != ROFFT_TEXT)
1.133 schwarze 1636: return;
1.19 schwarze 1637:
1.133 schwarze 1638: /* Skip function pointer punctuation. */
1.1 schwarze 1639:
1.133 schwarze 1640: cp = n->string;
1641: while (*cp == '(' || *cp == '*')
1.1 schwarze 1642: cp++;
1.133 schwarze 1643: sz = strcspn(cp, "()");
1.1 schwarze 1644:
1.133 schwarze 1645: putkeys(mpage, cp, sz, TYPE_Fn);
1.128 schwarze 1646: if (n->sec == SEC_SYNOPSIS)
1.133 schwarze 1647: putkeys(mpage, cp, sz, NAME_SYN);
1648: }
1.19 schwarze 1649:
1.133 schwarze 1650: static int
1.143 schwarze 1651: parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1652: const struct roff_node *n)
1.133 schwarze 1653: {
1654:
1655: if (n->child == NULL)
1.150 schwarze 1656: return 0;
1.133 schwarze 1657:
1658: parse_mdoc_fname(mpage, n->child);
1.19 schwarze 1659:
1.133 schwarze 1660: for (n = n->child->next; n != NULL; n = n->next)
1.141 schwarze 1661: if (n->type == ROFFT_TEXT)
1.47 schwarze 1662: putkey(mpage, n->string, TYPE_Fa);
1.19 schwarze 1663:
1.150 schwarze 1664: return 0;
1.128 schwarze 1665: }
1666:
1667: static int
1.143 schwarze 1668: parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1669: const struct roff_node *n)
1.128 schwarze 1670: {
1.132 schwarze 1671:
1.141 schwarze 1672: if (n->type != ROFFT_HEAD)
1.150 schwarze 1673: return 1;
1.128 schwarze 1674:
1.133 schwarze 1675: if (n->child != NULL)
1676: parse_mdoc_fname(mpage, n->child);
1677:
1.150 schwarze 1678: return 0;
1.1 schwarze 1679: }
1680:
1.19 schwarze 1681: static int
1.143 schwarze 1682: parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1683: const struct roff_node *n)
1.1 schwarze 1684: {
1.47 schwarze 1685: char *cp;
1.1 schwarze 1686:
1687: if (NULL == (n = n->child))
1.150 schwarze 1688: return 0;
1.1 schwarze 1689:
1.47 schwarze 1690: if (NULL == n->next) {
1691: putkey(mpage, n->string, TYPE_Xr);
1.150 schwarze 1692: return 0;
1.47 schwarze 1693: }
1.1 schwarze 1694:
1.77 schwarze 1695: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.47 schwarze 1696: putkey(mpage, cp, TYPE_Xr);
1697: free(cp);
1.150 schwarze 1698: return 0;
1.1 schwarze 1699: }
1700:
1.19 schwarze 1701: static int
1.143 schwarze 1702: parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1703: const struct roff_node *n)
1.1 schwarze 1704: {
1705:
1.141 schwarze 1706: if (n->type == ROFFT_BODY)
1.147 schwarze 1707: deroff(&mpage->desc, n);
1.150 schwarze 1708: return 0;
1.1 schwarze 1709: }
1710:
1.19 schwarze 1711: static int
1.143 schwarze 1712: parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1713: const struct roff_node *n)
1.1 schwarze 1714: {
1715:
1.86 schwarze 1716: if (SEC_NAME == n->sec)
1.90 schwarze 1717: putmdockey(mpage, n->child, NAME_TITLE);
1.141 schwarze 1718: else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1.127 schwarze 1719: if (n->child == NULL)
1720: putkey(mpage, meta->name, NAME_SYN);
1721: else
1722: putmdockey(mpage, n->child, NAME_SYN);
1.129 schwarze 1723: }
1724: if ( ! (mpage->name_head_done ||
1725: n->child == NULL || n->child->string == NULL ||
1726: strcasecmp(n->child->string, meta->title))) {
1.141 schwarze 1727: putkey(mpage, n->child->string, ROFFT_HEAD);
1.129 schwarze 1728: mpage->name_head_done = 1;
1.127 schwarze 1729: }
1.150 schwarze 1730: return 0;
1.1 schwarze 1731: }
1732:
1.19 schwarze 1733: static int
1.143 schwarze 1734: parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1735: const struct roff_node *n)
1.1 schwarze 1736: {
1737:
1.150 schwarze 1738: return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1.1 schwarze 1739: }
1740:
1.47 schwarze 1741: static int
1.143 schwarze 1742: parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1743: const struct roff_node *n)
1.1 schwarze 1744: {
1745:
1.150 schwarze 1746: return n->type == ROFFT_HEAD;
1.47 schwarze 1747: }
1.1 schwarze 1748:
1.47 schwarze 1749: static int
1.143 schwarze 1750: parse_mdoc_body(struct mpage *mpage, const struct roff_meta *meta,
1.142 schwarze 1751: const struct roff_node *n)
1.47 schwarze 1752: {
1.1 schwarze 1753:
1.150 schwarze 1754: return n->type == ROFFT_BODY;
1.1 schwarze 1755: }
1756:
1.47 schwarze 1757: /*
1758: * Add a string to the hash table for the current manual.
1759: * Each string has a bitmask telling which macros it belongs to.
1760: * When we finish the manual, we'll dump the table.
1761: */
1.1 schwarze 1762: static void
1.131 schwarze 1763: putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1.1 schwarze 1764: {
1.90 schwarze 1765: struct ohash *htab;
1.47 schwarze 1766: struct str *s;
1.68 schwarze 1767: const char *end;
1.47 schwarze 1768: unsigned int slot;
1.131 schwarze 1769: int i, mustfree;
1.1 schwarze 1770:
1.47 schwarze 1771: if (0 == sz)
1772: return;
1.68 schwarze 1773:
1.131 schwarze 1774: mustfree = render_string(&cp, &sz);
1775:
1.90 schwarze 1776: if (TYPE_Nm & v) {
1777: htab = &names;
1778: v &= name_mask;
1.124 schwarze 1779: if (v & NAME_FIRST)
1780: name_mask &= ~NAME_FIRST;
1.90 schwarze 1781: if (debug > 1)
1782: say(mpage->mlinks->file,
1.130 schwarze 1783: "Adding name %*s, bits=%d", sz, cp, v);
1.90 schwarze 1784: } else {
1785: htab = &strings;
1786: if (debug > 1)
1787: for (i = 0; i < mansearch_keymax; i++)
1.118 schwarze 1788: if ((uint64_t)1 << i & v)
1.90 schwarze 1789: say(mpage->mlinks->file,
1790: "Adding key %s=%*s",
1791: mansearch_keynames[i], sz, cp);
1.68 schwarze 1792: }
1.47 schwarze 1793:
1794: end = cp + sz;
1.90 schwarze 1795: slot = ohash_qlookupi(htab, cp, &end);
1796: s = ohash_find(htab, slot);
1.1 schwarze 1797:
1.47 schwarze 1798: if (NULL != s && mpage == s->mpage) {
1799: s->mask |= v;
1.1 schwarze 1800: return;
1.47 schwarze 1801: } else if (NULL == s) {
1.101 schwarze 1802: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.47 schwarze 1803: memcpy(s->key, cp, sz);
1.90 schwarze 1804: ohash_insert(htab, slot, s);
1.47 schwarze 1805: }
1806: s->mpage = mpage;
1807: s->mask = v;
1.131 schwarze 1808:
1809: if (mustfree)
1810: free(cp);
1.1 schwarze 1811: }
1812:
1813: /*
1.47 schwarze 1814: * Take a Unicode codepoint and produce its UTF-8 encoding.
1815: * This isn't the best way to do this, but it works.
1816: * The magic numbers are from the UTF-8 packaging.
1817: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1818: */
1.47 schwarze 1819: static size_t
1820: utf8(unsigned int cp, char out[7])
1.1 schwarze 1821: {
1.47 schwarze 1822: size_t rc;
1.1 schwarze 1823:
1.47 schwarze 1824: rc = 0;
1825: if (cp <= 0x0000007F) {
1826: rc = 1;
1827: out[0] = (char)cp;
1828: } else if (cp <= 0x000007FF) {
1829: rc = 2;
1830: out[0] = (cp >> 6 & 31) | 192;
1831: out[1] = (cp & 63) | 128;
1832: } else if (cp <= 0x0000FFFF) {
1833: rc = 3;
1834: out[0] = (cp >> 12 & 15) | 224;
1835: out[1] = (cp >> 6 & 63) | 128;
1836: out[2] = (cp & 63) | 128;
1837: } else if (cp <= 0x001FFFFF) {
1838: rc = 4;
1839: out[0] = (cp >> 18 & 7) | 240;
1840: out[1] = (cp >> 12 & 63) | 128;
1841: out[2] = (cp >> 6 & 63) | 128;
1842: out[3] = (cp & 63) | 128;
1843: } else if (cp <= 0x03FFFFFF) {
1844: rc = 5;
1845: out[0] = (cp >> 24 & 3) | 248;
1846: out[1] = (cp >> 18 & 63) | 128;
1847: out[2] = (cp >> 12 & 63) | 128;
1848: out[3] = (cp >> 6 & 63) | 128;
1849: out[4] = (cp & 63) | 128;
1850: } else if (cp <= 0x7FFFFFFF) {
1851: rc = 6;
1852: out[0] = (cp >> 30 & 1) | 252;
1853: out[1] = (cp >> 24 & 63) | 128;
1854: out[2] = (cp >> 18 & 63) | 128;
1855: out[3] = (cp >> 12 & 63) | 128;
1856: out[4] = (cp >> 6 & 63) | 128;
1857: out[5] = (cp & 63) | 128;
1858: } else
1.150 schwarze 1859: return 0;
1.19 schwarze 1860:
1.47 schwarze 1861: out[rc] = '\0';
1.150 schwarze 1862: return rc;
1.1 schwarze 1863: }
1864:
1.47 schwarze 1865: /*
1.131 schwarze 1866: * If the string contains escape sequences,
1867: * replace it with an allocated rendering and return 1,
1868: * such that the caller can free it after use.
1869: * Otherwise, do nothing and return 0.
1.47 schwarze 1870: */
1.131 schwarze 1871: static int
1872: render_string(char **public, size_t *psz)
1.1 schwarze 1873: {
1.131 schwarze 1874: const char *src, *scp, *addcp, *seq;
1875: char *dst;
1876: size_t ssz, dsz, addsz;
1.71 schwarze 1877: char utfbuf[7], res[6];
1.131 schwarze 1878: int seqlen, unicode;
1.47 schwarze 1879:
1880: res[0] = '\\';
1881: res[1] = '\t';
1882: res[2] = ASCII_NBRSP;
1883: res[3] = ASCII_HYPH;
1.71 schwarze 1884: res[4] = ASCII_BREAK;
1885: res[5] = '\0';
1.1 schwarze 1886:
1.131 schwarze 1887: src = scp = *public;
1888: ssz = *psz;
1889: dst = NULL;
1890: dsz = 0;
1891:
1892: while (scp < src + *psz) {
1893:
1894: /* Leave normal characters unchanged. */
1895:
1896: if (strchr(res, *scp) == NULL) {
1897: if (dst != NULL)
1898: dst[dsz++] = *scp;
1899: scp++;
1900: continue;
1901: }
1.1 schwarze 1902:
1.47 schwarze 1903: /*
1.131 schwarze 1904: * Found something that requires replacing,
1905: * make sure we have a destination buffer.
1.47 schwarze 1906: */
1.131 schwarze 1907:
1908: if (dst == NULL) {
1909: dst = mandoc_malloc(ssz + 1);
1910: dsz = scp - src;
1911: memcpy(dst, src, dsz);
1.47 schwarze 1912: }
1.39 schwarze 1913:
1.131 schwarze 1914: /* Handle single-char special characters. */
1915:
1916: switch (*scp) {
1917: case '\\':
1918: break;
1.98 schwarze 1919: case '\t':
1920: case ASCII_NBRSP:
1.131 schwarze 1921: dst[dsz++] = ' ';
1922: scp++;
1923: continue;
1924: case ASCII_HYPH:
1925: dst[dsz++] = '-';
1.71 schwarze 1926: /* FALLTHROUGH */
1.98 schwarze 1927: case ASCII_BREAK:
1.131 schwarze 1928: scp++;
1.47 schwarze 1929: continue;
1.71 schwarze 1930: default:
1.131 schwarze 1931: abort();
1.71 schwarze 1932: }
1.39 schwarze 1933:
1.47 schwarze 1934: /*
1.131 schwarze 1935: * Found an escape sequence.
1936: * Read past the slash, then parse it.
1937: * Ignore everything except characters.
1.47 schwarze 1938: */
1.52 schwarze 1939:
1.131 schwarze 1940: scp++;
1941: if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1.47 schwarze 1942: continue;
1.39 schwarze 1943:
1.47 schwarze 1944: /*
1.52 schwarze 1945: * Render the special character
1946: * as either UTF-8 or ASCII.
1.47 schwarze 1947: */
1.52 schwarze 1948:
1949: if (write_utf8) {
1.131 schwarze 1950: unicode = mchars_spec2cp(mchars, seq, seqlen);
1951: if (unicode <= 0)
1.52 schwarze 1952: continue;
1.131 schwarze 1953: addsz = utf8(unicode, utfbuf);
1954: if (addsz == 0)
1.52 schwarze 1955: continue;
1.131 schwarze 1956: addcp = utfbuf;
1.52 schwarze 1957: } else {
1.131 schwarze 1958: addcp = mchars_spec2str(mchars, seq, seqlen, &addsz);
1959: if (addcp == NULL)
1.52 schwarze 1960: continue;
1.131 schwarze 1961: if (*addcp == ASCII_NBRSP) {
1962: addcp = " ";
1963: addsz = 1;
1.52 schwarze 1964: }
1965: }
1.1 schwarze 1966:
1.47 schwarze 1967: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1968:
1.131 schwarze 1969: ssz += addsz;
1970: dst = mandoc_realloc(dst, ssz + 1);
1971: memcpy(dst + dsz, addcp, addsz);
1972: dsz += addsz;
1973: }
1974: if (dst != NULL) {
1975: *public = dst;
1976: *psz = dsz;
1977: }
1978:
1979: /* Trim trailing whitespace and NUL-terminate. */
1980:
1981: while (*psz > 0 && (*public)[*psz - 1] == ' ')
1982: --*psz;
1983: if (dst != NULL) {
1984: (*public)[*psz] = '\0';
1.150 schwarze 1985: return 1;
1.131 schwarze 1986: } else
1.150 schwarze 1987: return 0;
1.1 schwarze 1988: }
1989:
1.75 schwarze 1990: static void
1991: dbadd_mlink(const struct mlink *mlink)
1992: {
1993: size_t i;
1994:
1995: i = 1;
1996: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1997: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1998: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.94 schwarze 1999: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1.75 schwarze 2000: SQL_STEP(stmts[STMT_INSERT_LINK]);
2001: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1.124 schwarze 2002: }
2003:
2004: static void
2005: dbadd_mlink_name(const struct mlink *mlink)
2006: {
1.130 schwarze 2007: uint64_t bits;
1.124 schwarze 2008: size_t i;
2009:
2010: dbadd_mlink(mlink);
1.115 schwarze 2011:
2012: i = 1;
1.130 schwarze 2013: SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
2014: bits = NAME_FILE & NAME_MASK;
2015: if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
2016: bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
2017: sqlite3_reset(stmts[STMT_SELECT_NAME]);
2018: }
2019:
2020: i = 1;
2021: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
1.115 schwarze 2022: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2023: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2024: SQL_STEP(stmts[STMT_INSERT_NAME]);
2025: sqlite3_reset(stmts[STMT_INSERT_NAME]);
1.75 schwarze 2026: }
2027:
1.11 schwarze 2028: /*
1.47 schwarze 2029: * Flush the current page's terms (and their bits) into the database.
2030: * Wrap the entire set of additions in a transaction to make sqlite be a
2031: * little faster.
1.53 schwarze 2032: * Also, handle escape sequences at the last possible moment.
1.11 schwarze 2033: */
2034: static void
1.131 schwarze 2035: dbadd(struct mpage *mpage)
1.11 schwarze 2036: {
1.47 schwarze 2037: struct mlink *mlink;
2038: struct str *key;
1.131 schwarze 2039: char *cp;
1.47 schwarze 2040: size_t i;
2041: unsigned int slot;
1.131 schwarze 2042: int mustfree;
1.47 schwarze 2043:
1.85 schwarze 2044: mlink = mpage->mlinks;
1.11 schwarze 2045:
1.85 schwarze 2046: if (nodb) {
1.104 schwarze 2047: for (key = ohash_first(&names, &slot); NULL != key;
1.131 schwarze 2048: key = ohash_next(&names, &slot))
1.104 schwarze 2049: free(key);
2050: for (key = ohash_first(&strings, &slot); NULL != key;
1.131 schwarze 2051: key = ohash_next(&strings, &slot))
1.104 schwarze 2052: free(key);
1.102 schwarze 2053: if (0 == debug)
2054: return;
1.85 schwarze 2055: while (NULL != mlink) {
2056: fputs(mlink->name, stdout);
2057: if (NULL == mlink->next ||
2058: strcmp(mlink->dsec, mlink->next->dsec) ||
2059: strcmp(mlink->fsec, mlink->next->fsec) ||
2060: strcmp(mlink->arch, mlink->next->arch)) {
2061: putchar('(');
2062: if ('\0' == *mlink->dsec)
2063: fputs(mlink->fsec, stdout);
2064: else
2065: fputs(mlink->dsec, stdout);
2066: if ('\0' != *mlink->arch)
2067: printf("/%s", mlink->arch);
2068: putchar(')');
2069: }
2070: mlink = mlink->next;
2071: if (NULL != mlink)
2072: fputs(", ", stdout);
2073: }
1.89 schwarze 2074: printf(" - %s\n", mpage->desc);
1.11 schwarze 2075: return;
1.85 schwarze 2076: }
2077:
2078: if (debug)
2079: say(mlink->file, "Adding to database");
1.47 schwarze 2080:
1.131 schwarze 2081: cp = mpage->desc;
2082: i = strlen(cp);
2083: mustfree = render_string(&cp, &i);
1.47 schwarze 2084: i = 1;
1.131 schwarze 2085: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
1.116 schwarze 2086: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
1.47 schwarze 2087: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.94 schwarze 2088: mpage->pageid = sqlite3_last_insert_rowid(db);
1.47 schwarze 2089: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1.131 schwarze 2090: if (mustfree)
2091: free(cp);
1.47 schwarze 2092:
1.85 schwarze 2093: while (NULL != mlink) {
1.75 schwarze 2094: dbadd_mlink(mlink);
1.85 schwarze 2095: mlink = mlink->next;
2096: }
1.91 schwarze 2097: mlink = mpage->mlinks;
1.47 schwarze 2098:
1.90 schwarze 2099: for (key = ohash_first(&names, &slot); NULL != key;
2100: key = ohash_next(&names, &slot)) {
2101: assert(key->mpage == mpage);
2102: i = 1;
2103: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
1.131 schwarze 2104: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
1.94 schwarze 2105: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
1.90 schwarze 2106: SQL_STEP(stmts[STMT_INSERT_NAME]);
2107: sqlite3_reset(stmts[STMT_INSERT_NAME]);
2108: free(key);
2109: }
1.47 schwarze 2110: for (key = ohash_first(&strings, &slot); NULL != key;
2111: key = ohash_next(&strings, &slot)) {
2112: assert(key->mpage == mpage);
2113: i = 1;
2114: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.131 schwarze 2115: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
1.94 schwarze 2116: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
1.47 schwarze 2117: SQL_STEP(stmts[STMT_INSERT_KEY]);
2118: sqlite3_reset(stmts[STMT_INSERT_KEY]);
2119: free(key);
1.33 schwarze 2120: }
1.47 schwarze 2121: }
1.41 deraadt 2122:
1.47 schwarze 2123: static void
2124: dbprune(void)
2125: {
2126: struct mpage *mpage;
2127: struct mlink *mlink;
2128: size_t i;
2129: unsigned int slot;
1.11 schwarze 2130:
1.63 schwarze 2131: if (0 == nodb)
2132: SQL_EXEC("BEGIN TRANSACTION");
1.47 schwarze 2133:
1.63 schwarze 2134: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2135: mpage = ohash_next(&mpages, &slot)) {
1.47 schwarze 2136: mlink = mpage->mlinks;
1.82 schwarze 2137: if (debug)
1.63 schwarze 2138: say(mlink->file, "Deleting from database");
2139: if (nodb)
2140: continue;
2141: for ( ; NULL != mlink; mlink = mlink->next) {
2142: i = 1;
2143: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2144: i, mlink->dsec);
2145: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2146: i, mlink->arch);
2147: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2148: i, mlink->name);
2149: SQL_STEP(stmts[STMT_DELETE_PAGE]);
2150: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2151: }
1.11 schwarze 2152: }
1.63 schwarze 2153:
2154: if (0 == nodb)
2155: SQL_EXEC("END TRANSACTION");
1.47 schwarze 2156: }
1.22 schwarze 2157:
1.47 schwarze 2158: /*
2159: * Close an existing database and its prepared statements.
2160: * If "real" is not set, rename the temporary file into the real one.
2161: */
2162: static void
2163: dbclose(int real)
2164: {
2165: size_t i;
1.72 schwarze 2166: int status;
2167: pid_t child;
1.11 schwarze 2168:
1.47 schwarze 2169: if (nodb)
2170: return;
1.11 schwarze 2171:
1.47 schwarze 2172: for (i = 0; i < STMT__MAX; i++) {
2173: sqlite3_finalize(stmts[i]);
2174: stmts[i] = NULL;
1.28 schwarze 2175: }
1.22 schwarze 2176:
1.47 schwarze 2177: sqlite3_close(db);
2178: db = NULL;
1.11 schwarze 2179:
1.47 schwarze 2180: if (real)
2181: return;
1.22 schwarze 2182:
1.72 schwarze 2183: if ('\0' == *tempfilename) {
2184: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2185: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2186: say(MANDOC_DB, "&rename");
1.72 schwarze 2187: }
2188: return;
2189: }
2190:
2191: switch (child = fork()) {
1.98 schwarze 2192: case -1:
1.72 schwarze 2193: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2194: say("", "&fork cmp");
1.72 schwarze 2195: return;
1.98 schwarze 2196: case 0:
1.72 schwarze 2197: execlp("cmp", "cmp", "-s",
1.149 schwarze 2198: tempfilename, MANDOC_DB, (char *)NULL);
1.80 schwarze 2199: say("", "&exec cmp");
1.72 schwarze 2200: exit(0);
2201: default:
2202: break;
2203: }
2204: if (-1 == waitpid(child, &status, 0)) {
2205: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2206: say("", "&wait cmp");
1.72 schwarze 2207: } else if (WIFSIGNALED(status)) {
2208: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2209: say("", "cmp died from signal %d", WTERMSIG(status));
1.72 schwarze 2210: } else if (WEXITSTATUS(status)) {
1.47 schwarze 2211: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2212: say(MANDOC_DB,
2213: "Data changed, but cannot replace database");
2214: }
2215:
2216: *strrchr(tempfilename, '/') = '\0';
2217: switch (child = fork()) {
1.98 schwarze 2218: case -1:
1.72 schwarze 2219: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2220: say("", "&fork rm");
1.72 schwarze 2221: return;
1.98 schwarze 2222: case 0:
1.149 schwarze 2223: execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
1.80 schwarze 2224: say("", "&exec rm");
1.72 schwarze 2225: exit((int)MANDOCLEVEL_SYSERR);
2226: default:
2227: break;
2228: }
2229: if (-1 == waitpid(child, &status, 0)) {
2230: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2231: say("", "&wait rm");
1.72 schwarze 2232: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2233: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2234: say("", "%s: Cannot remove temporary directory",
2235: tempfilename);
1.22 schwarze 2236: }
1.11 schwarze 2237: }
2238:
1.47 schwarze 2239: /*
2240: * This is straightforward stuff.
2241: * Open a database connection to a "temporary" database, then open a set
2242: * of prepared statements we'll use over and over again.
2243: * If "real" is set, we use the existing database; if not, we truncate a
2244: * temporary one.
2245: * Must be matched by dbclose().
2246: */
2247: static int
2248: dbopen(int real)
1.2 schwarze 2249: {
1.72 schwarze 2250: const char *sql;
1.47 schwarze 2251: int rc, ofl;
1.6 schwarze 2252:
1.98 schwarze 2253: if (nodb)
1.150 schwarze 2254: return 1;
1.6 schwarze 2255:
1.72 schwarze 2256: *tempfilename = '\0';
1.47 schwarze 2257: ofl = SQLITE_OPEN_READWRITE;
1.72 schwarze 2258:
2259: if (real) {
2260: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2261: if (SQLITE_OK != rc) {
1.47 schwarze 2262: exitcode = (int)MANDOCLEVEL_SYSERR;
1.108 sthen 2263: if (SQLITE_CANTOPEN != rc)
2264: say(MANDOC_DB, "%s", sqlite3_errstr(rc));
1.150 schwarze 2265: return 0;
1.28 schwarze 2266: }
1.72 schwarze 2267: goto prepare_statements;
2268: }
2269:
2270: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.6 schwarze 2271:
1.72 schwarze 2272: remove(MANDOC_DB "~");
2273: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.98 schwarze 2274: if (SQLITE_OK == rc)
1.72 schwarze 2275: goto create_tables;
1.73 schwarze 2276: if (MPARSE_QUICK & mparse_options) {
1.47 schwarze 2277: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2278: say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
1.150 schwarze 2279: return 0;
1.47 schwarze 2280: }
1.6 schwarze 2281:
1.100 schwarze 2282: (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2283: sizeof(tempfilename));
1.72 schwarze 2284: if (NULL == mkdtemp(tempfilename)) {
2285: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2286: say("", "&%s", tempfilename);
1.150 schwarze 2287: return 0;
1.72 schwarze 2288: }
1.100 schwarze 2289: (void)strlcat(tempfilename, "/" MANDOC_DB,
2290: sizeof(tempfilename));
1.72 schwarze 2291: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2292: if (SQLITE_OK != rc) {
1.47 schwarze 2293: exitcode = (int)MANDOCLEVEL_SYSERR;
1.103 schwarze 2294: say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
1.150 schwarze 2295: return 0;
1.2 schwarze 2296: }
2297:
1.72 schwarze 2298: create_tables:
1.47 schwarze 2299: sql = "CREATE TABLE \"mpages\" (\n"
1.89 schwarze 2300: " \"desc\" TEXT NOT NULL,\n"
1.47 schwarze 2301: " \"form\" INTEGER NOT NULL,\n"
1.94 schwarze 2302: " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1.47 schwarze 2303: ");\n"
2304: "\n"
2305: "CREATE TABLE \"mlinks\" (\n"
2306: " \"sec\" TEXT NOT NULL,\n"
2307: " \"arch\" TEXT NOT NULL,\n"
2308: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2309: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2310: "ON DELETE CASCADE\n"
1.47 schwarze 2311: ");\n"
1.93 schwarze 2312: "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
1.47 schwarze 2313: "\n"
1.90 schwarze 2314: "CREATE TABLE \"names\" (\n"
2315: " \"bits\" INTEGER NOT NULL,\n"
2316: " \"name\" TEXT NOT NULL,\n"
1.94 schwarze 2317: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.130 schwarze 2318: "ON DELETE CASCADE,\n"
2319: " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
1.90 schwarze 2320: ");\n"
2321: "\n"
1.47 schwarze 2322: "CREATE TABLE \"keys\" (\n"
2323: " \"bits\" INTEGER NOT NULL,\n"
2324: " \"key\" TEXT NOT NULL,\n"
1.94 schwarze 2325: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.66 schwarze 2326: "ON DELETE CASCADE\n"
1.93 schwarze 2327: ");\n"
2328: "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
1.47 schwarze 2329:
2330: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2331: exitcode = (int)MANDOCLEVEL_SYSERR;
1.72 schwarze 2332: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.103 schwarze 2333: sqlite3_close(db);
1.150 schwarze 2334: return 0;
1.2 schwarze 2335: }
2336:
1.47 schwarze 2337: prepare_statements:
1.103 schwarze 2338: if (SQLITE_OK != sqlite3_exec(db,
2339: "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2340: exitcode = (int)MANDOCLEVEL_SYSERR;
2341: say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2342: sqlite3_errmsg(db));
2343: sqlite3_close(db);
1.150 schwarze 2344: return 0;
1.103 schwarze 2345: }
2346:
1.94 schwarze 2347: sql = "DELETE FROM mpages WHERE pageid IN "
1.63 schwarze 2348: "(SELECT pageid FROM mlinks WHERE "
2349: "sec=? AND arch=? AND name=?)";
1.47 schwarze 2350: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2351: sql = "INSERT INTO mpages "
1.89 schwarze 2352: "(desc,form) VALUES (?,?)";
1.47 schwarze 2353: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2354: sql = "INSERT INTO mlinks "
1.61 schwarze 2355: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.47 schwarze 2356: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.130 schwarze 2357: sql = "SELECT bits FROM names where pageid = ?";
2358: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
1.90 schwarze 2359: sql = "INSERT INTO names "
2360: "(bits,name,pageid) VALUES (?,?,?)";
2361: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
1.47 schwarze 2362: sql = "INSERT INTO keys "
2363: "(bits,key,pageid) VALUES (?,?,?)";
2364: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.6 schwarze 2365:
1.47 schwarze 2366: /*
2367: * When opening a new database, we can turn off
2368: * synchronous mode for much better performance.
2369: */
1.6 schwarze 2370:
1.103 schwarze 2371: if (real && SQLITE_OK != sqlite3_exec(db,
2372: "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2373: exitcode = (int)MANDOCLEVEL_SYSERR;
2374: say(MANDOC_DB, "PRAGMA synchronous: %s",
2375: sqlite3_errmsg(db));
2376: sqlite3_close(db);
1.150 schwarze 2377: return 0;
1.103 schwarze 2378: }
1.11 schwarze 2379:
1.150 schwarze 2380: return 1;
1.47 schwarze 2381: }
1.6 schwarze 2382:
1.47 schwarze 2383: static void *
1.106 espie 2384: hash_calloc(size_t n, size_t sz, void *arg)
1.47 schwarze 2385: {
1.6 schwarze 2386:
1.150 schwarze 2387: return mandoc_calloc(n, sz);
1.47 schwarze 2388: }
1.2 schwarze 2389:
1.47 schwarze 2390: static void *
2391: hash_alloc(size_t sz, void *arg)
2392: {
1.28 schwarze 2393:
1.150 schwarze 2394: return mandoc_malloc(sz);
1.47 schwarze 2395: }
1.6 schwarze 2396:
1.47 schwarze 2397: static void
1.106 espie 2398: hash_free(void *p, void *arg)
1.47 schwarze 2399: {
1.26 schwarze 2400:
1.47 schwarze 2401: free(p);
2402: }
1.6 schwarze 2403:
1.47 schwarze 2404: static int
1.120 schwarze 2405: set_basedir(const char *targetdir, int report_baddir)
1.47 schwarze 2406: {
2407: static char startdir[PATH_MAX];
1.110 schwarze 2408: static int getcwd_status; /* 1 = ok, 2 = failure */
2409: static int chdir_status; /* 1 = changed directory */
1.109 schwarze 2410: char *cp;
1.6 schwarze 2411:
1.47 schwarze 2412: /*
1.110 schwarze 2413: * Remember the original working directory, if possible.
2414: * This will be needed if the second or a later directory
2415: * on the command line is given as a relative path.
2416: * Do not error out if the current directory is not
2417: * searchable: Maybe it won't be needed after all.
2418: */
2419: if (0 == getcwd_status) {
2420: if (NULL == getcwd(startdir, sizeof(startdir))) {
2421: getcwd_status = 2;
2422: (void)strlcpy(startdir, strerror(errno),
2423: sizeof(startdir));
2424: } else
2425: getcwd_status = 1;
2426: }
2427:
2428: /*
2429: * We are leaving the old base directory.
2430: * Do not use it any longer, not even for messages.
2431: */
2432: *basedir = '\0';
2433:
2434: /*
2435: * If and only if the directory was changed earlier and
2436: * the next directory to process is given as a relative path,
2437: * first go back, or bail out if that is impossible.
1.47 schwarze 2438: */
1.110 schwarze 2439: if (chdir_status && '/' != *targetdir) {
2440: if (2 == getcwd_status) {
1.47 schwarze 2441: exitcode = (int)MANDOCLEVEL_SYSERR;
1.110 schwarze 2442: say("", "getcwd: %s", startdir);
1.150 schwarze 2443: return 0;
1.47 schwarze 2444: }
1.110 schwarze 2445: if (-1 == chdir(startdir)) {
1.47 schwarze 2446: exitcode = (int)MANDOCLEVEL_SYSERR;
1.80 schwarze 2447: say("", "&chdir %s", startdir);
1.150 schwarze 2448: return 0;
1.2 schwarze 2449: }
2450: }
1.110 schwarze 2451:
2452: /*
2453: * Always resolve basedir to the canonicalized absolute
2454: * pathname and append a trailing slash, such that
2455: * we can reliably check whether files are inside.
2456: */
1.47 schwarze 2457: if (NULL == realpath(targetdir, basedir)) {
1.120 schwarze 2458: if (report_baddir || errno != ENOENT) {
2459: exitcode = (int)MANDOCLEVEL_BADARG;
2460: say("", "&%s: realpath", targetdir);
2461: }
1.150 schwarze 2462: return 0;
1.47 schwarze 2463: } else if (-1 == chdir(basedir)) {
1.120 schwarze 2464: if (report_baddir || errno != ENOENT) {
2465: exitcode = (int)MANDOCLEVEL_BADARG;
2466: say("", "&chdir");
2467: }
1.150 schwarze 2468: return 0;
1.109 schwarze 2469: }
1.110 schwarze 2470: chdir_status = 1;
1.109 schwarze 2471: cp = strchr(basedir, '\0');
2472: if ('/' != cp[-1]) {
2473: if (cp - basedir >= PATH_MAX - 1) {
2474: exitcode = (int)MANDOCLEVEL_SYSERR;
2475: say("", "Filename too long");
1.150 schwarze 2476: return 0;
1.109 schwarze 2477: }
2478: *cp++ = '/';
2479: *cp = '\0';
1.47 schwarze 2480: }
1.150 schwarze 2481: return 1;
1.2 schwarze 2482: }
2483:
2484: static void
1.47 schwarze 2485: say(const char *file, const char *format, ...)
1.2 schwarze 2486: {
1.47 schwarze 2487: va_list ap;
1.80 schwarze 2488: int use_errno;
1.2 schwarze 2489:
1.47 schwarze 2490: if ('\0' != *basedir)
2491: fprintf(stderr, "%s", basedir);
2492: if ('\0' != *basedir && '\0' != *file)
1.110 schwarze 2493: fputc('/', stderr);
1.47 schwarze 2494: if ('\0' != *file)
2495: fprintf(stderr, "%s", file);
1.31 schwarze 2496:
1.80 schwarze 2497: use_errno = 1;
2498: if (NULL != format) {
2499: switch (*format) {
1.98 schwarze 2500: case '&':
1.80 schwarze 2501: format++;
2502: break;
1.98 schwarze 2503: case '\0':
1.80 schwarze 2504: format = NULL;
2505: break;
2506: default:
2507: use_errno = 0;
2508: break;
2509: }
2510: }
2511: if (NULL != format) {
2512: if ('\0' != *basedir || '\0' != *file)
2513: fputs(": ", stderr);
2514: va_start(ap, format);
2515: vfprintf(stderr, format, ap);
2516: va_end(ap);
2517: }
2518: if (use_errno) {
2519: if ('\0' != *basedir || '\0' != *file || NULL != format)
2520: fputs(": ", stderr);
1.47 schwarze 2521: perror(NULL);
1.80 schwarze 2522: } else
2523: fputc('\n', stderr);
1.1 schwarze 2524: }