Annotation of mandoc/mandocdb.c, Revision 1.85
1.85 ! schwarze 1: /* $Id: mandocdb.c,v 1.84 2013/12/27 15:39:03 schwarze Exp $ */
1.1 kristaps 2: /*
1.48 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.56 schwarze 4: * Copyright (c) 2011, 2012, 2013 Ingo Schwarze <schwarze@openbsd.org>
1.1 kristaps 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: */
18: #ifdef HAVE_CONFIG_H
19: #include "config.h"
20: #endif
21:
1.50 kristaps 22: #include <sys/stat.h>
1.1 kristaps 23:
24: #include <assert.h>
1.43 kristaps 25: #include <ctype.h>
1.63 schwarze 26: #include <errno.h>
1.1 kristaps 27: #include <fcntl.h>
1.50 kristaps 28: #include <fts.h>
1.1 kristaps 29: #include <getopt.h>
1.58 schwarze 30: #include <limits.h>
1.50 kristaps 31: #include <stddef.h>
1.59 schwarze 32: #include <stdio.h>
1.1 kristaps 33: #include <stdint.h>
34: #include <stdlib.h>
35: #include <string.h>
1.17 schwarze 36: #include <unistd.h>
1.1 kristaps 37:
1.53 kristaps 38: #ifdef HAVE_OHASH
1.50 kristaps 39: #include <ohash.h>
1.53 kristaps 40: #else
41: #include "compat_ohash.h"
42: #endif
1.50 kristaps 43: #include <sqlite3.h>
1.1 kristaps 44:
1.50 kristaps 45: #include "mdoc.h"
1.1 kristaps 46: #include "man.h"
47: #include "mandoc.h"
1.10 kristaps 48: #include "manpath.h"
1.55 kristaps 49: #include "mansearch.h"
1.1 kristaps 50:
1.52 kristaps 51: #define SQL_EXEC(_v) \
52: if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
53: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
54: #define SQL_BIND_TEXT(_s, _i, _v) \
55: if (SQLITE_OK != sqlite3_bind_text \
56: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
57: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
58: #define SQL_BIND_INT(_s, _i, _v) \
59: if (SQLITE_OK != sqlite3_bind_int \
60: ((_s), (_i)++, (_v))) \
61: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
62: #define SQL_BIND_INT64(_s, _i, _v) \
63: if (SQLITE_OK != sqlite3_bind_int64 \
64: ((_s), (_i)++, (_v))) \
65: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
66: #define SQL_STEP(_s) \
67: if (SQLITE_DONE != sqlite3_step((_s))) \
68: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
69:
1.50 kristaps 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 */
1.38 schwarze 76: };
77:
1.50 kristaps 78: enum form {
1.84 schwarze 79: FORM_NONE, /* format is unknown */
80: FORM_SRC, /* format is -man or -mdoc */
81: FORM_CAT /* format is cat */
1.50 kristaps 82: };
1.38 schwarze 83:
1.50 kristaps 84: struct str {
85: char *utf8; /* key in UTF-8 form */
1.78 schwarze 86: const struct mpage *mpage; /* if set, the owning parse */
1.50 kristaps 87: uint64_t mask; /* bitmask in sequence */
1.51 kristaps 88: char key[]; /* the string itself */
1.38 schwarze 89: };
90:
1.79 schwarze 91: struct inodev {
92: ino_t st_ino;
93: dev_t st_dev;
1.50 kristaps 94: };
1.5 kristaps 95:
1.78 schwarze 96: struct mpage {
1.79 schwarze 97: struct inodev inodev; /* used for hashing routine */
1.82 schwarze 98: enum form form; /* format from file content */
99: char *sec; /* section from file content */
100: char *arch; /* architecture from file content */
101: char *title; /* title from file content */
102: char *desc; /* description from file content */
103: struct mlink *mlinks; /* singly linked list */
104: };
105:
106: struct mlink {
1.58 schwarze 107: char file[PATH_MAX]; /* filename rel. to manpath */
1.82 schwarze 108: enum form dform; /* format from directory */
1.84 schwarze 109: enum form fform; /* format from file name suffix */
1.82 schwarze 110: char *dsec; /* section from directory */
111: char *arch; /* architecture from directory */
112: char *name; /* name from file name (not empty) */
113: char *fsec; /* section from file name suffix */
1.85 ! schwarze 114: struct mlink *next; /* singly linked list */
1.3 kristaps 115: };
116:
1.69 schwarze 117: struct title {
118: char *title; /* name(sec/arch) given inside the file */
119: char *file; /* file name in case of mismatch */
120: };
121:
1.50 kristaps 122: enum stmt {
1.81 schwarze 123: STMT_DELETE_PAGE = 0, /* delete mpage */
124: STMT_INSERT_PAGE, /* insert mpage */
125: STMT_INSERT_LINK, /* insert mlink */
126: STMT_INSERT_KEY, /* insert parsed key */
1.50 kristaps 127: STMT__MAX
1.1 kristaps 128: };
129:
1.78 schwarze 130: typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
1.1 kristaps 131:
1.50 kristaps 132: struct mdoc_handler {
133: mdoc_fp fp; /* optional handler */
134: uint64_t mask; /* set unless handler returns 0 */
1.1 kristaps 135: };
136:
1.59 schwarze 137: static void dbclose(int);
1.82 schwarze 138: static void dbindex(const struct mpage *, struct mchars *);
1.59 schwarze 139: static int dbopen(int);
140: static void dbprune(void);
141: static void filescan(const char *);
1.50 kristaps 142: static void *hash_alloc(size_t, void *);
143: static void hash_free(void *, size_t, void *);
144: static void *hash_halloc(size_t, void *);
145: static int inocheck(const struct stat *);
1.84 schwarze 146: static void mlink_add(struct mlink *, const struct stat *);
1.82 schwarze 147: static void mlink_free(struct mlink *);
1.78 schwarze 148: static void mpages_free(void);
149: static void mpages_merge(struct mchars *, struct mparse *, int);
150: static void parse_cat(struct mpage *);
151: static void parse_man(struct mpage *, const struct man_node *);
152: static void parse_mdoc(struct mpage *, const struct mdoc_node *);
153: static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
154: static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
155: static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
156: static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
157: static int parse_mdoc_In(struct mpage *, const struct mdoc_node *);
158: static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
159: static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
160: static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
161: static int parse_mdoc_St(struct mpage *, const struct mdoc_node *);
162: static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
163: static void putkey(const struct mpage *,
1.50 kristaps 164: const char *, uint64_t);
1.78 schwarze 165: static void putkeys(const struct mpage *,
1.64 schwarze 166: const char *, size_t, uint64_t);
1.78 schwarze 167: static void putmdockey(const struct mpage *,
1.50 kristaps 168: const struct mdoc_node *, uint64_t);
1.59 schwarze 169: static void say(const char *, const char *, ...);
1.80 schwarze 170: static int set_basedir(const char *);
1.59 schwarze 171: static int treescan(void);
1.50 kristaps 172: static size_t utf8(unsigned int, char [7]);
173: static void utf8key(struct mchars *, struct str *);
174:
175: static char *progname;
176: static int use_all; /* use all found files */
177: static int nodb; /* no database changes */
178: static int verb; /* print what we're doing */
179: static int warnings; /* warn about crap */
1.59 schwarze 180: static int exitcode; /* to be returned by main */
1.50 kristaps 181: static enum op op; /* operational mode */
1.59 schwarze 182: static char basedir[PATH_MAX]; /* current base directory */
1.78 schwarze 183: static struct ohash mpages; /* table of distinct manual pages */
1.83 schwarze 184: static struct ohash mlinks; /* table of directory entries */
1.50 kristaps 185: static struct ohash strings; /* table of all strings */
186: static sqlite3 *db = NULL; /* current database */
187: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.25 schwarze 188:
189: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.68 schwarze 190: { NULL, 0 }, /* Ap */
191: { NULL, 0 }, /* Dd */
192: { NULL, 0 }, /* Dt */
193: { NULL, 0 }, /* Os */
194: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
195: { parse_mdoc_head, TYPE_Ss }, /* Ss */
196: { NULL, 0 }, /* Pp */
197: { NULL, 0 }, /* D1 */
198: { NULL, 0 }, /* Dl */
199: { NULL, 0 }, /* Bd */
200: { NULL, 0 }, /* Ed */
201: { NULL, 0 }, /* Bl */
202: { NULL, 0 }, /* El */
203: { NULL, 0 }, /* It */
204: { NULL, 0 }, /* Ad */
205: { NULL, TYPE_An }, /* An */
206: { NULL, TYPE_Ar }, /* Ar */
207: { NULL, TYPE_Cd }, /* Cd */
208: { NULL, TYPE_Cm }, /* Cm */
209: { NULL, TYPE_Dv }, /* Dv */
210: { NULL, TYPE_Er }, /* Er */
211: { NULL, TYPE_Ev }, /* Ev */
212: { NULL, 0 }, /* Ex */
213: { NULL, TYPE_Fa }, /* Fa */
214: { parse_mdoc_Fd, 0 }, /* Fd */
215: { NULL, TYPE_Fl }, /* Fl */
216: { parse_mdoc_Fn, 0 }, /* Fn */
217: { NULL, TYPE_Ft }, /* Ft */
218: { NULL, TYPE_Ic }, /* Ic */
219: { parse_mdoc_In, TYPE_In }, /* In */
220: { NULL, TYPE_Li }, /* Li */
221: { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
222: { parse_mdoc_Nm, TYPE_Nm }, /* Nm */
223: { NULL, 0 }, /* Op */
224: { NULL, 0 }, /* Ot */
225: { NULL, TYPE_Pa }, /* Pa */
226: { NULL, 0 }, /* Rv */
227: { parse_mdoc_St, 0 }, /* St */
228: { NULL, TYPE_Va }, /* Va */
229: { parse_mdoc_body, TYPE_Va }, /* Vt */
230: { parse_mdoc_Xr, 0 }, /* Xr */
231: { NULL, 0 }, /* %A */
232: { NULL, 0 }, /* %B */
233: { NULL, 0 }, /* %D */
234: { NULL, 0 }, /* %I */
235: { NULL, 0 }, /* %J */
236: { NULL, 0 }, /* %N */
237: { NULL, 0 }, /* %O */
238: { NULL, 0 }, /* %P */
239: { NULL, 0 }, /* %R */
240: { NULL, 0 }, /* %T */
241: { NULL, 0 }, /* %V */
242: { NULL, 0 }, /* Ac */
243: { NULL, 0 }, /* Ao */
244: { NULL, 0 }, /* Aq */
245: { NULL, TYPE_At }, /* At */
246: { NULL, 0 }, /* Bc */
247: { NULL, 0 }, /* Bf */
248: { NULL, 0 }, /* Bo */
249: { NULL, 0 }, /* Bq */
250: { NULL, TYPE_Bsx }, /* Bsx */
251: { NULL, TYPE_Bx }, /* Bx */
252: { NULL, 0 }, /* Db */
253: { NULL, 0 }, /* Dc */
254: { NULL, 0 }, /* Do */
255: { NULL, 0 }, /* Dq */
256: { NULL, 0 }, /* Ec */
257: { NULL, 0 }, /* Ef */
258: { NULL, TYPE_Em }, /* Em */
259: { NULL, 0 }, /* Eo */
260: { NULL, TYPE_Fx }, /* Fx */
261: { NULL, TYPE_Ms }, /* Ms */
262: { NULL, 0 }, /* No */
263: { NULL, 0 }, /* Ns */
264: { NULL, TYPE_Nx }, /* Nx */
265: { NULL, TYPE_Ox }, /* Ox */
266: { NULL, 0 }, /* Pc */
267: { NULL, 0 }, /* Pf */
268: { NULL, 0 }, /* Po */
269: { NULL, 0 }, /* Pq */
270: { NULL, 0 }, /* Qc */
271: { NULL, 0 }, /* Ql */
272: { NULL, 0 }, /* Qo */
273: { NULL, 0 }, /* Qq */
274: { NULL, 0 }, /* Re */
275: { NULL, 0 }, /* Rs */
276: { NULL, 0 }, /* Sc */
277: { NULL, 0 }, /* So */
278: { NULL, 0 }, /* Sq */
279: { NULL, 0 }, /* Sm */
280: { NULL, 0 }, /* Sx */
281: { NULL, TYPE_Sy }, /* Sy */
282: { NULL, TYPE_Tn }, /* Tn */
283: { NULL, 0 }, /* Ux */
284: { NULL, 0 }, /* Xc */
285: { NULL, 0 }, /* Xo */
286: { parse_mdoc_head, 0 }, /* Fo */
287: { NULL, 0 }, /* Fc */
288: { NULL, 0 }, /* Oo */
289: { NULL, 0 }, /* Oc */
290: { NULL, 0 }, /* Bk */
291: { NULL, 0 }, /* Ek */
292: { NULL, 0 }, /* Bt */
293: { NULL, 0 }, /* Hf */
294: { NULL, 0 }, /* Fr */
295: { NULL, 0 }, /* Ud */
296: { NULL, TYPE_Lb }, /* Lb */
297: { NULL, 0 }, /* Lp */
298: { NULL, TYPE_Lk }, /* Lk */
299: { NULL, TYPE_Mt }, /* Mt */
300: { NULL, 0 }, /* Brq */
301: { NULL, 0 }, /* Bro */
302: { NULL, 0 }, /* Brc */
303: { NULL, 0 }, /* %C */
304: { NULL, 0 }, /* Es */
305: { NULL, 0 }, /* En */
306: { NULL, TYPE_Dx }, /* Dx */
307: { NULL, 0 }, /* %Q */
308: { NULL, 0 }, /* br */
309: { NULL, 0 }, /* sp */
310: { NULL, 0 }, /* %U */
311: { NULL, 0 }, /* Ta */
1.1 kristaps 312: };
313:
314: int
315: main(int argc, char *argv[])
316: {
1.59 schwarze 317: int ch, i;
1.50 kristaps 318: size_t j, sz;
1.59 schwarze 319: const char *path_arg;
1.50 kristaps 320: struct mchars *mc;
321: struct manpaths dirs;
322: struct mparse *mp;
1.83 schwarze 323: struct ohash_info mpages_info, mlinks_info;
1.50 kristaps 324:
325: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
326: memset(&dirs, 0, sizeof(struct manpaths));
327:
1.83 schwarze 328: mpages_info.alloc = mlinks_info.alloc = hash_alloc;
329: mpages_info.halloc = mlinks_info.halloc = hash_halloc;
330: mpages_info.hfree = mlinks_info.hfree = hash_free;
1.50 kristaps 331:
1.79 schwarze 332: mpages_info.key_offset = offsetof(struct mpage, inodev);
1.83 schwarze 333: mlinks_info.key_offset = offsetof(struct mlink, file);
1.1 kristaps 334:
335: progname = strrchr(argv[0], '/');
336: if (progname == NULL)
337: progname = argv[0];
338: else
339: ++progname;
340:
1.50 kristaps 341: /*
342: * We accept a few different invocations.
343: * The CHECKOP macro makes sure that invocation styles don't
344: * clobber each other.
345: */
346: #define CHECKOP(_op, _ch) do \
347: if (OP_DEFAULT != (_op)) { \
348: fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
349: goto usage; \
350: } while (/*CONSTCOND*/0)
1.10 kristaps 351:
1.59 schwarze 352: path_arg = NULL;
1.38 schwarze 353: op = OP_DEFAULT;
1.1 kristaps 354:
1.50 kristaps 355: while (-1 != (ch = getopt(argc, argv, "aC:d:ntu:vW")))
1.1 kristaps 356: switch (ch) {
1.12 schwarze 357: case ('a'):
358: use_all = 1;
359: break;
1.34 schwarze 360: case ('C'):
1.50 kristaps 361: CHECKOP(op, ch);
1.59 schwarze 362: path_arg = optarg;
1.38 schwarze 363: op = OP_CONFFILE;
1.34 schwarze 364: break;
1.5 kristaps 365: case ('d'):
1.50 kristaps 366: CHECKOP(op, ch);
1.59 schwarze 367: path_arg = optarg;
1.5 kristaps 368: op = OP_UPDATE;
369: break;
1.50 kristaps 370: case ('n'):
371: nodb = 1;
372: break;
1.38 schwarze 373: case ('t'):
1.50 kristaps 374: CHECKOP(op, ch);
1.38 schwarze 375: dup2(STDOUT_FILENO, STDERR_FILENO);
376: op = OP_TEST;
1.50 kristaps 377: nodb = warnings = 1;
1.38 schwarze 378: break;
1.5 kristaps 379: case ('u'):
1.50 kristaps 380: CHECKOP(op, ch);
1.59 schwarze 381: path_arg = optarg;
1.5 kristaps 382: op = OP_DELETE;
383: break;
384: case ('v'):
385: verb++;
386: break;
1.38 schwarze 387: case ('W'):
388: warnings = 1;
389: break;
1.1 kristaps 390: default:
1.38 schwarze 391: goto usage;
1.1 kristaps 392: }
393:
394: argc -= optind;
395: argv += optind;
396:
1.38 schwarze 397: if (OP_CONFFILE == op && argc > 0) {
1.50 kristaps 398: fprintf(stderr, "-C: Too many arguments\n");
1.38 schwarze 399: goto usage;
400: }
401:
1.59 schwarze 402: exitcode = (int)MANDOCLEVEL_OK;
1.50 kristaps 403: mp = mparse_alloc(MPARSE_AUTO,
404: MANDOCLEVEL_FATAL, NULL, NULL, NULL);
405: mc = mchars_alloc();
406:
1.78 schwarze 407: ohash_init(&mpages, 6, &mpages_info);
1.83 schwarze 408: ohash_init(&mlinks, 6, &mlinks_info);
1.50 kristaps 409:
410: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
411: /*
412: * Force processing all files.
413: */
414: use_all = 1;
1.59 schwarze 415:
1.50 kristaps 416: /*
417: * All of these deal with a specific directory.
418: * Jump into that directory then collect files specified
419: * on the command-line.
420: */
1.59 schwarze 421: if (0 == set_basedir(path_arg))
1.50 kristaps 422: goto out;
423: for (i = 0; i < argc; i++)
1.59 schwarze 424: filescan(argv[i]);
425: if (0 == dbopen(1))
1.50 kristaps 426: goto out;
427: if (OP_TEST != op)
1.59 schwarze 428: dbprune();
1.50 kristaps 429: if (OP_DELETE != op)
1.78 schwarze 430: mpages_merge(mc, mp, 0);
1.59 schwarze 431: dbclose(1);
1.50 kristaps 432: } else {
433: /*
434: * If we have arguments, use them as our manpaths.
435: * If we don't, grok from manpath(1) or however else
436: * manpath_parse() wants to do it.
437: */
438: if (argc > 0) {
439: dirs.paths = mandoc_calloc
440: (argc, sizeof(char *));
441: dirs.sz = (size_t)argc;
442: for (i = 0; i < argc; i++)
443: dirs.paths[i] = mandoc_strdup(argv[i]);
444: } else
1.59 schwarze 445: manpath_parse(&dirs, path_arg, NULL, NULL);
1.50 kristaps 446:
447: /*
1.69 schwarze 448: * First scan the tree rooted at a base directory, then
449: * build a new database and finally move it into place.
1.50 kristaps 450: * Ignore zero-length directories and strip trailing
451: * slashes.
452: */
453: for (j = 0; j < dirs.sz; j++) {
454: sz = strlen(dirs.paths[j]);
455: if (sz && '/' == dirs.paths[j][sz - 1])
456: dirs.paths[j][--sz] = '\0';
457: if (0 == sz)
458: continue;
1.66 schwarze 459:
460: if (j) {
1.78 schwarze 461: ohash_init(&mpages, 6, &mpages_info);
1.83 schwarze 462: ohash_init(&mlinks, 6, &mlinks_info);
1.66 schwarze 463: }
464:
1.59 schwarze 465: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 466: goto out;
1.59 schwarze 467: if (0 == treescan())
1.50 kristaps 468: goto out;
1.59 schwarze 469: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 470: goto out;
1.59 schwarze 471: if (0 == dbopen(0))
1.50 kristaps 472: goto out;
1.52 kristaps 473:
1.78 schwarze 474: mpages_merge(mc, mp, warnings && !use_all);
1.59 schwarze 475: dbclose(0);
1.66 schwarze 476:
477: if (j + 1 < dirs.sz) {
1.80 schwarze 478: mpages_free();
1.78 schwarze 479: ohash_delete(&mpages);
1.83 schwarze 480: ohash_delete(&mlinks);
1.66 schwarze 481: }
1.50 kristaps 482: }
483: }
484: out:
1.59 schwarze 485: set_basedir(NULL);
1.50 kristaps 486: manpath_free(&dirs);
487: mchars_free(mc);
488: mparse_free(mp);
1.80 schwarze 489: mpages_free();
1.78 schwarze 490: ohash_delete(&mpages);
1.83 schwarze 491: ohash_delete(&mlinks);
1.59 schwarze 492: return(exitcode);
1.50 kristaps 493: usage:
494: fprintf(stderr, "usage: %s [-anvW] [-C file]\n"
495: " %s [-anvW] dir ...\n"
496: " %s [-nvW] -d dir [file ...]\n"
497: " %s [-nvW] -u dir [file ...]\n"
498: " %s -t file ...\n",
499: progname, progname, progname,
500: progname, progname);
501:
1.59 schwarze 502: return((int)MANDOCLEVEL_BADARG);
1.50 kristaps 503: }
504:
505: /*
1.59 schwarze 506: * Scan a directory tree rooted at "basedir" for manpages.
1.50 kristaps 507: * We use fts(), scanning directory parts along the way for clues to our
508: * section and architecture.
509: *
510: * If use_all has been specified, grok all files.
511: * If not, sanitise paths to the following:
512: *
513: * [./]man*[/<arch>]/<name>.<section>
514: * or
515: * [./]cat<section>[/<arch>]/<name>.0
516: *
517: * TODO: accomodate for multi-language directories.
518: */
519: static int
1.59 schwarze 520: treescan(void)
1.50 kristaps 521: {
522: FTS *f;
523: FTSENT *ff;
1.84 schwarze 524: struct mlink *mlink;
1.50 kristaps 525: int dform;
1.84 schwarze 526: char *fsec;
1.66 schwarze 527: const char *dsec, *arch, *cp, *path;
1.50 kristaps 528: const char *argv[2];
529:
530: argv[0] = ".";
531: argv[1] = (char *)NULL;
532:
533: /*
534: * Walk through all components under the directory, using the
535: * logical descent of files.
536: */
537: f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
538: if (NULL == f) {
1.59 schwarze 539: exitcode = (int)MANDOCLEVEL_SYSERR;
540: say("", NULL);
1.50 kristaps 541: return(0);
542: }
543:
544: dsec = arch = NULL;
545: dform = FORM_NONE;
546:
547: while (NULL != (ff = fts_read(f))) {
548: path = ff->fts_path + 2;
549: /*
1.83 schwarze 550: * If we're a regular file, add an mlink by using the
1.50 kristaps 551: * stored directory data and handling the filename.
552: * Disallow duplicate (hard-linked) files.
553: */
554: if (FTS_F == ff->fts_info) {
1.60 schwarze 555: if (0 == strcmp(path, MANDOC_DB))
556: continue;
1.50 kristaps 557: if ( ! use_all && ff->fts_level < 2) {
1.56 schwarze 558: if (warnings)
1.59 schwarze 559: say(path, "Extraneous file");
1.50 kristaps 560: continue;
561: } else if (inocheck(ff->fts_statp)) {
1.56 schwarze 562: if (warnings)
1.59 schwarze 563: say(path, "Duplicate file");
1.50 kristaps 564: continue;
1.84 schwarze 565: } else if (NULL == (fsec =
1.60 schwarze 566: strrchr(ff->fts_name, '.'))) {
567: if ( ! use_all) {
1.56 schwarze 568: if (warnings)
1.60 schwarze 569: say(path,
570: "No filename suffix");
1.50 kristaps 571: continue;
572: }
1.84 schwarze 573: } else if (0 == strcmp(++fsec, "html")) {
1.60 schwarze 574: if (warnings)
575: say(path, "Skip html");
576: continue;
1.84 schwarze 577: } else if (0 == strcmp(fsec, "gz")) {
1.60 schwarze 578: if (warnings)
579: say(path, "Skip gz");
580: continue;
1.84 schwarze 581: } else if (0 == strcmp(fsec, "ps")) {
1.60 schwarze 582: if (warnings)
583: say(path, "Skip ps");
584: continue;
1.84 schwarze 585: } else if (0 == strcmp(fsec, "pdf")) {
1.60 schwarze 586: if (warnings)
587: say(path, "Skip pdf");
588: continue;
589: } else if ( ! use_all &&
1.84 schwarze 590: ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
591: (FORM_CAT == dform && strcmp(fsec, "0")))) {
1.60 schwarze 592: if (warnings)
593: say(path, "Wrong filename suffix");
594: continue;
1.66 schwarze 595: } else
1.84 schwarze 596: fsec[-1] = '\0';
597: mlink = mandoc_calloc(1, sizeof(struct mlink));
598: strlcpy(mlink->file, path, sizeof(mlink->file));
599: mlink->dform = dform;
600: if (NULL != dsec)
601: mlink->dsec = mandoc_strdup(dsec);
602: if (NULL != arch)
603: mlink->arch = mandoc_strdup(arch);
604: mlink->name = mandoc_strdup(ff->fts_name);
605: if (NULL != fsec)
606: mlink->fsec = mandoc_strdup(fsec);
607: mlink_add(mlink, ff->fts_statp);
1.50 kristaps 608: continue;
1.84 schwarze 609: } else if (FTS_D != ff->fts_info &&
1.60 schwarze 610: FTS_DP != ff->fts_info) {
611: if (warnings)
612: say(path, "Not a regular file");
1.50 kristaps 613: continue;
1.60 schwarze 614: }
1.1 kristaps 615:
1.50 kristaps 616: switch (ff->fts_level) {
617: case (0):
618: /* Ignore the root directory. */
619: break;
620: case (1):
621: /*
622: * This might contain manX/ or catX/.
623: * Try to infer this from the name.
624: * If we're not in use_all, enforce it.
625: */
626: dsec = NULL;
627: dform = FORM_NONE;
628: cp = ff->fts_name;
629: if (FTS_DP == ff->fts_info)
630: break;
1.1 kristaps 631:
1.50 kristaps 632: if (0 == strncmp(cp, "man", 3)) {
633: dform = FORM_SRC;
1.66 schwarze 634: dsec = cp + 3;
1.50 kristaps 635: } else if (0 == strncmp(cp, "cat", 3)) {
636: dform = FORM_CAT;
1.66 schwarze 637: dsec = cp + 3;
1.50 kristaps 638: }
1.1 kristaps 639:
1.50 kristaps 640: if (NULL != dsec || use_all)
641: break;
1.1 kristaps 642:
1.56 schwarze 643: if (warnings)
1.59 schwarze 644: say(path, "Unknown directory part");
1.50 kristaps 645: fts_set(f, ff, FTS_SKIP);
646: break;
647: case (2):
648: /*
649: * Possibly our architecture.
650: * If we're descending, keep tabs on it.
651: */
652: arch = NULL;
653: if (FTS_DP != ff->fts_info && NULL != dsec)
1.66 schwarze 654: arch = ff->fts_name;
1.50 kristaps 655: break;
656: default:
657: if (FTS_DP == ff->fts_info || use_all)
658: break;
1.56 schwarze 659: if (warnings)
1.59 schwarze 660: say(path, "Extraneous directory part");
1.50 kristaps 661: fts_set(f, ff, FTS_SKIP);
662: break;
1.5 kristaps 663: }
1.50 kristaps 664: }
665:
666: fts_close(f);
667: return(1);
668: }
1.5 kristaps 669:
1.50 kristaps 670: /*
671: * Add a file to the file vector.
672: * Do not verify that it's a "valid" looking manpage (we'll do that
673: * later).
674: *
675: * Try to infer the manual section, architecture, and page name from the
676: * path, assuming it looks like
677: *
678: * [./]man*[/<arch>]/<name>.<section>
679: * or
680: * [./]cat<section>[/<arch>]/<name>.0
681: *
1.83 schwarze 682: * Stuff this information directly into the mlink vector.
1.50 kristaps 683: * See treescan() for the fts(3) version of this.
684: */
685: static void
1.59 schwarze 686: filescan(const char *file)
1.50 kristaps 687: {
1.59 schwarze 688: char buf[PATH_MAX];
1.84 schwarze 689: struct stat st;
690: struct mlink *mlink;
1.59 schwarze 691: char *p, *start;
1.5 kristaps 692:
1.50 kristaps 693: assert(use_all);
1.5 kristaps 694:
1.50 kristaps 695: if (0 == strncmp(file, "./", 2))
696: file += 2;
1.5 kristaps 697:
1.59 schwarze 698: if (NULL == realpath(file, buf)) {
699: exitcode = (int)MANDOCLEVEL_BADARG;
700: say(file, NULL);
701: return;
1.72 schwarze 702: } else if (OP_TEST != op && strstr(buf, basedir) != buf) {
1.59 schwarze 703: exitcode = (int)MANDOCLEVEL_BADARG;
704: say("", "%s: outside base directory", buf);
705: return;
706: } else if (-1 == stat(buf, &st)) {
707: exitcode = (int)MANDOCLEVEL_BADARG;
708: say(file, NULL);
1.50 kristaps 709: return;
710: } else if ( ! (S_IFREG & st.st_mode)) {
1.59 schwarze 711: exitcode = (int)MANDOCLEVEL_BADARG;
712: say(file, "Not a regular file");
1.50 kristaps 713: return;
714: } else if (inocheck(&st)) {
1.56 schwarze 715: if (warnings)
1.59 schwarze 716: say(file, "Duplicate file");
1.50 kristaps 717: return;
718: }
1.59 schwarze 719: start = buf + strlen(basedir);
1.84 schwarze 720: mlink = mandoc_calloc(1, sizeof(struct mlink));
721: strlcpy(mlink->file, start, sizeof(mlink->file));
1.17 schwarze 722:
1.50 kristaps 723: /*
724: * First try to guess our directory structure.
725: * If we find a separator, try to look for man* or cat*.
726: * If we find one of these and what's underneath is a directory,
727: * assume it's an architecture.
728: */
729: if (NULL != (p = strchr(start, '/'))) {
730: *p++ = '\0';
731: if (0 == strncmp(start, "man", 3)) {
1.84 schwarze 732: mlink->dform = FORM_SRC;
733: mlink->dsec = mandoc_strdup(start + 3);
1.50 kristaps 734: } else if (0 == strncmp(start, "cat", 3)) {
1.84 schwarze 735: mlink->dform = FORM_CAT;
736: mlink->dsec = mandoc_strdup(start + 3);
1.17 schwarze 737: }
1.5 kristaps 738:
1.50 kristaps 739: start = p;
1.84 schwarze 740: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
1.50 kristaps 741: *p++ = '\0';
1.84 schwarze 742: mlink->arch = mandoc_strdup(start);
1.50 kristaps 743: start = p;
1.84 schwarze 744: }
1.50 kristaps 745: }
746:
747: /*
748: * Now check the file suffix.
749: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
750: */
751: p = strrchr(start, '\0');
752: while (p-- > start && '/' != *p && '.' != *p)
753: /* Loop. */ ;
754:
755: if ('.' == *p) {
756: *p++ = '\0';
1.84 schwarze 757: mlink->fsec = mandoc_strdup(p);
1.5 kristaps 758: }
759:
1.10 kristaps 760: /*
1.50 kristaps 761: * Now try to parse the name.
762: * Use the filename portion of the path.
1.10 kristaps 763: */
1.84 schwarze 764: mlink->name = start;
1.50 kristaps 765: if (NULL != (p = strrchr(start, '/'))) {
1.84 schwarze 766: mlink->name = p + 1;
1.50 kristaps 767: *p = '\0';
1.84 schwarze 768: }
769: mlink->name = mandoc_strdup(mlink->name);
1.50 kristaps 770:
1.84 schwarze 771: mlink_add(mlink, &st);
1.50 kristaps 772: }
773:
774: static int
775: inocheck(const struct stat *st)
776: {
1.79 schwarze 777: struct inodev inodev;
1.50 kristaps 778: uint32_t hash;
779:
1.79 schwarze 780: memset(&inodev, 0, sizeof(inodev));
781: inodev.st_ino = hash = st->st_ino;
782: inodev.st_dev = st->st_dev;
1.13 schwarze 783:
1.78 schwarze 784: return(NULL != ohash_find(&mpages, ohash_lookup_memory(
1.79 schwarze 785: &mpages, (char *)&inodev, sizeof(inodev), hash)));
1.50 kristaps 786: }
1.5 kristaps 787:
1.50 kristaps 788: static void
1.84 schwarze 789: mlink_add(struct mlink *mlink, const struct stat *st)
1.50 kristaps 790: {
1.83 schwarze 791: struct inodev inodev;
1.78 schwarze 792: struct mpage *mpage;
1.83 schwarze 793: unsigned int slot;
1.50 kristaps 794:
1.84 schwarze 795: assert(NULL != mlink->file);
1.50 kristaps 796:
1.84 schwarze 797: if (NULL == mlink->dsec)
798: mlink->dsec = mandoc_strdup("");
799: if (NULL == mlink->arch)
800: mlink->arch = mandoc_strdup("");
801: if (NULL == mlink->name)
802: mlink->name = mandoc_strdup("");
803: if (NULL == mlink->fsec)
804: mlink->fsec = mandoc_strdup("");
805:
806: if ('0' == *mlink->fsec) {
807: free(mlink->fsec);
808: mlink->fsec = mandoc_strdup(mlink->dsec);
809: mlink->fform = FORM_CAT;
810: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
811: mlink->fform = FORM_SRC;
1.74 schwarze 812: else
1.84 schwarze 813: mlink->fform = FORM_NONE;
1.82 schwarze 814:
1.83 schwarze 815: slot = ohash_qlookup(&mlinks, mlink->file);
816: assert(NULL == ohash_find(&mlinks, slot));
817: ohash_insert(&mlinks, slot, mlink);
818:
819: inodev.st_ino = st->st_ino;
820: inodev.st_dev = st->st_dev;
821: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
822: sizeof(struct inodev), inodev.st_ino);
823: mpage = ohash_find(&mpages, slot);
824: if (NULL == mpage) {
825: mpage = mandoc_calloc(1, sizeof(struct mpage));
826: mpage->inodev.st_ino = inodev.st_ino;
827: mpage->inodev.st_dev = inodev.st_dev;
828: ohash_insert(&mpages, slot, mpage);
829: } else
1.85 ! schwarze 830: mlink->next = mpage->mlinks;
1.82 schwarze 831: mpage->mlinks = mlink;
832: }
833:
834: static void
835: mlink_free(struct mlink *mlink)
836: {
837:
838: free(mlink->dsec);
839: free(mlink->arch);
840: free(mlink->name);
841: free(mlink->fsec);
842: free(mlink);
1.50 kristaps 843: }
1.3 kristaps 844:
1.50 kristaps 845: static void
1.78 schwarze 846: mpages_free(void)
1.50 kristaps 847: {
1.78 schwarze 848: struct mpage *mpage;
1.82 schwarze 849: struct mlink *mlink;
1.80 schwarze 850: unsigned int slot;
1.3 kristaps 851:
1.80 schwarze 852: mpage = ohash_first(&mpages, &slot);
853: while (NULL != mpage) {
1.82 schwarze 854: while (NULL != (mlink = mpage->mlinks)) {
1.85 ! schwarze 855: mpage->mlinks = mlink->next;
1.82 schwarze 856: mlink_free(mlink);
857: }
1.78 schwarze 858: free(mpage->sec);
859: free(mpage->arch);
1.82 schwarze 860: free(mpage->title);
861: free(mpage->desc);
1.78 schwarze 862: free(mpage);
1.80 schwarze 863: mpage = ohash_next(&mpages, &slot);
1.50 kristaps 864: }
865: }
1.38 schwarze 866:
1.50 kristaps 867: /*
1.80 schwarze 868: * Run through the files in the global vector "mpages"
869: * and add them to the database specified in "basedir".
1.50 kristaps 870: *
871: * This handles the parsing scheme itself, using the cues of directory
872: * and filename to determine whether the file is parsable or not.
873: */
1.59 schwarze 874: static void
1.78 schwarze 875: mpages_merge(struct mchars *mc, struct mparse *mp, int check_reachable)
1.50 kristaps 876: {
1.69 schwarze 877: struct ohash title_table;
1.71 schwarze 878: struct ohash_info title_info, str_info;
1.69 schwarze 879: char buf[PATH_MAX];
1.78 schwarze 880: struct mpage *mpage;
1.69 schwarze 881: struct mdoc *mdoc;
882: struct man *man;
883: struct title *title_entry;
884: char *bufp, *title_str;
1.82 schwarze 885: const char *cp;
1.69 schwarze 886: size_t sz;
887: int match;
1.80 schwarze 888: unsigned int pslot, tslot;
1.69 schwarze 889: enum mandoclevel lvl;
890:
1.71 schwarze 891: str_info.alloc = hash_alloc;
892: str_info.halloc = hash_halloc;
893: str_info.hfree = hash_free;
894: str_info.key_offset = offsetof(struct str, key);
895:
1.69 schwarze 896: if (check_reachable) {
897: title_info.alloc = hash_alloc;
898: title_info.halloc = hash_halloc;
899: title_info.hfree = hash_free;
900: title_info.key_offset = offsetof(struct title, title);
901: ohash_init(&title_table, 6, &title_info);
902: }
1.50 kristaps 903:
1.80 schwarze 904: mpage = ohash_first(&mpages, &pslot);
905: while (NULL != mpage) {
1.50 kristaps 906: /*
907: * If we're a catpage (as defined by our path), then see
908: * if a manpage exists by the same name (ignoring the
909: * suffix).
910: * If it does, then we want to use it instead of our
911: * own.
912: */
1.82 schwarze 913: if ( ! use_all && FORM_CAT == mpage->mlinks->dform) {
914: sz = strlcpy(buf, mpage->mlinks->file, PATH_MAX);
1.58 schwarze 915: if (sz >= PATH_MAX) {
1.56 schwarze 916: if (warnings)
1.82 schwarze 917: say(mpage->mlinks->file,
1.78 schwarze 918: "Filename too long");
1.80 schwarze 919: mpage = ohash_next(&mpages, &pslot);
1.50 kristaps 920: continue;
921: }
922: bufp = strstr(buf, "cat");
923: assert(NULL != bufp);
924: memcpy(bufp, "man", 3);
925: if (NULL != (bufp = strrchr(buf, '.')))
926: *++bufp = '\0';
1.82 schwarze 927: strlcat(buf, mpage->mlinks->dsec, PATH_MAX);
1.83 schwarze 928: if (NULL != ohash_find(&mlinks,
929: ohash_qlookup(&mlinks, buf))) {
1.56 schwarze 930: if (warnings)
1.82 schwarze 931: say(mpage->mlinks->file, "Man "
1.56 schwarze 932: "source exists: %s", buf);
1.80 schwarze 933: mpage = ohash_next(&mpages, &pslot);
1.50 kristaps 934: continue;
935: }
936: }
937:
1.71 schwarze 938: ohash_init(&strings, 6, &str_info);
1.50 kristaps 939: mparse_reset(mp);
940: mdoc = NULL;
941: man = NULL;
1.69 schwarze 942: match = 1;
1.14 schwarze 943:
944: /*
1.33 schwarze 945: * Try interpreting the file as mdoc(7) or man(7)
946: * source code, unless it is already known to be
947: * formatted. Fall back to formatted mode.
1.14 schwarze 948: */
1.82 schwarze 949: if (FORM_CAT != mpage->mlinks->dform ||
1.84 schwarze 950: FORM_CAT != mpage->mlinks->fform) {
1.82 schwarze 951: lvl = mparse_readfd(mp, -1, mpage->mlinks->file);
1.50 kristaps 952: if (lvl < MANDOCLEVEL_FATAL)
953: mparse_result(mp, &mdoc, &man);
1.76 schwarze 954: }
1.14 schwarze 955:
956: if (NULL != mdoc) {
1.82 schwarze 957: mpage->form = FORM_SRC;
958: mpage->sec =
959: mandoc_strdup(mdoc_meta(mdoc)->msec);
960: mpage->arch = mdoc_meta(mdoc)->arch;
1.84 schwarze 961: mpage->arch = mandoc_strdup(
962: NULL == mpage->arch ? "" : mpage->arch);
1.82 schwarze 963: mpage->title =
964: mandoc_strdup(mdoc_meta(mdoc)->title);
1.14 schwarze 965: } else if (NULL != man) {
1.82 schwarze 966: mpage->form = FORM_SRC;
967: mpage->sec =
968: mandoc_strdup(man_meta(man)->msec);
969: mpage->arch =
970: mandoc_strdup(mpage->mlinks->arch);
971: mpage->title =
972: mandoc_strdup(man_meta(man)->title);
1.76 schwarze 973: } else {
1.82 schwarze 974: mpage->form = FORM_CAT;
975: mpage->sec =
976: mandoc_strdup(mpage->mlinks->dsec);
977: mpage->arch =
978: mandoc_strdup(mpage->mlinks->arch);
979: mpage->title =
980: mandoc_strdup(mpage->mlinks->name);
1.76 schwarze 981: }
1.50 kristaps 982:
1.12 schwarze 983: /*
1.44 kristaps 984: * Check whether the manual section given in a file
985: * agrees with the directory where the file is located.
986: * Some manuals have suffixes like (3p) on their
987: * section number either inside the file or in the
988: * directory name, some are linked into more than one
989: * section, like encrypt(1) = makekey(8). Do not skip
990: * manuals for such reasons.
1.12 schwarze 991: */
1.82 schwarze 992: if (warnings && !use_all && FORM_SRC == mpage->form &&
993: strcasecmp(mpage->sec, mpage->mlinks->dsec)) {
1.69 schwarze 994: match = 0;
1.82 schwarze 995: say(mpage->mlinks->file, "Section \"%s\" "
1.78 schwarze 996: "manual in %s directory",
1.82 schwarze 997: mpage->sec, mpage->mlinks->dsec);
1.69 schwarze 998: }
1.12 schwarze 999:
1.42 schwarze 1000: /*
1001: * Manual page directories exist for each kernel
1002: * architecture as returned by machine(1).
1003: * However, many manuals only depend on the
1004: * application architecture as returned by arch(1).
1005: * For example, some (2/ARM) manuals are shared
1006: * across the "armish" and "zaurus" kernel
1007: * architectures.
1008: * A few manuals are even shared across completely
1009: * different architectures, for example fdformat(1)
1010: * on amd64, i386, sparc, and sparc64.
1011: * Thus, warn about architecture mismatches,
1012: * but don't skip manuals for this reason.
1013: */
1.82 schwarze 1014: if (warnings && !use_all &&
1015: strcasecmp(mpage->arch, mpage->mlinks->arch)) {
1.69 schwarze 1016: match = 0;
1.82 schwarze 1017: say(mpage->mlinks->file, "Architecture \"%s\" "
1.47 schwarze 1018: "manual in \"%s\" directory",
1.82 schwarze 1019: mpage->arch, mpage->mlinks->arch);
1.69 schwarze 1020: }
1.82 schwarze 1021: if (warnings && !use_all &&
1022: strcasecmp(mpage->title, mpage->mlinks->name))
1.69 schwarze 1023: match = 0;
1.12 schwarze 1024:
1.82 schwarze 1025: putkey(mpage, mpage->mlinks->name, TYPE_Nm);
1.12 schwarze 1026:
1.50 kristaps 1027: if (NULL != mdoc) {
1028: if (NULL != (cp = mdoc_meta(mdoc)->name))
1.78 schwarze 1029: putkey(mpage, cp, TYPE_Nm);
1030: assert(NULL == mpage->desc);
1031: parse_mdoc(mpage, mdoc_node(mdoc));
1032: putkey(mpage, NULL != mpage->desc ?
1.82 schwarze 1033: mpage->desc : mpage->mlinks->name, TYPE_Nd);
1.50 kristaps 1034: } else if (NULL != man)
1.78 schwarze 1035: parse_man(mpage, man_node(man));
1.50 kristaps 1036: else
1.78 schwarze 1037: parse_cat(mpage);
1.44 kristaps 1038:
1.69 schwarze 1039: /*
1040: * Build a title string for the file. If it matches
1041: * the location of the file, remember the title as
1042: * found; else, remember it as missing.
1043: */
1044:
1045: if (check_reachable) {
1.82 schwarze 1046: if (-1 == asprintf(&title_str, "%s(%s%s%s)",
1047: mpage->title, mpage->sec,
1048: '\0' == *mpage->arch ? "" : "/",
1049: mpage->arch)) {
1.69 schwarze 1050: perror(NULL);
1051: exit((int)MANDOCLEVEL_SYSERR);
1052: }
1.80 schwarze 1053: tslot = ohash_qlookup(&title_table, title_str);
1054: title_entry = ohash_find(&title_table, tslot);
1.69 schwarze 1055: if (NULL == title_entry) {
1056: title_entry = mandoc_malloc(
1057: sizeof(struct title));
1058: title_entry->title = title_str;
1059: title_entry->file = mandoc_strdup(
1.82 schwarze 1060: match ? "" : mpage->mlinks->file);
1.80 schwarze 1061: ohash_insert(&title_table, tslot,
1.69 schwarze 1062: title_entry);
1063: } else {
1064: if (match)
1065: *title_entry->file = '\0';
1066: free(title_str);
1067: }
1068: }
1069:
1.82 schwarze 1070: dbindex(mpage, mc);
1.66 schwarze 1071: ohash_delete(&strings);
1.80 schwarze 1072: mpage = ohash_next(&mpages, &pslot);
1.69 schwarze 1073: }
1074:
1075: if (check_reachable) {
1.80 schwarze 1076: title_entry = ohash_first(&title_table, &tslot);
1.69 schwarze 1077: while (NULL != title_entry) {
1078: if ('\0' != *title_entry->file)
1079: say(title_entry->file,
1080: "Probably unreachable, title is %s",
1081: title_entry->title);
1082: free(title_entry->title);
1083: free(title_entry->file);
1084: free(title_entry);
1.80 schwarze 1085: title_entry = ohash_next(&title_table, &tslot);
1.69 schwarze 1086: }
1087: ohash_delete(&title_table);
1.50 kristaps 1088: }
1089: }
1.12 schwarze 1090:
1.50 kristaps 1091: static void
1.78 schwarze 1092: parse_cat(struct mpage *mpage)
1.50 kristaps 1093: {
1094: FILE *stream;
1095: char *line, *p, *title;
1096: size_t len, plen, titlesz;
1.12 schwarze 1097:
1.82 schwarze 1098: if (NULL == (stream = fopen(mpage->mlinks->file, "r"))) {
1.56 schwarze 1099: if (warnings)
1.82 schwarze 1100: say(mpage->mlinks->file, NULL);
1.50 kristaps 1101: return;
1102: }
1.1 kristaps 1103:
1.50 kristaps 1104: /* Skip to first blank line. */
1.1 kristaps 1105:
1.50 kristaps 1106: while (NULL != (line = fgetln(stream, &len)))
1107: if ('\n' == *line)
1108: break;
1.1 kristaps 1109:
1.50 kristaps 1110: /*
1111: * Assume the first line that is not indented
1112: * is the first section header. Skip to it.
1113: */
1.1 kristaps 1114:
1.50 kristaps 1115: while (NULL != (line = fgetln(stream, &len)))
1116: if ('\n' != *line && ' ' != *line)
1117: break;
1118:
1119: /*
1120: * Read up until the next section into a buffer.
1121: * Strip the leading and trailing newline from each read line,
1122: * appending a trailing space.
1123: * Ignore empty (whitespace-only) lines.
1124: */
1.1 kristaps 1125:
1.50 kristaps 1126: titlesz = 0;
1127: title = NULL;
1.38 schwarze 1128:
1.50 kristaps 1129: while (NULL != (line = fgetln(stream, &len))) {
1130: if (' ' != *line || '\n' != line[len - 1])
1131: break;
1132: while (len > 0 && isspace((unsigned char)*line)) {
1133: line++;
1134: len--;
1135: }
1136: if (1 == len)
1.38 schwarze 1137: continue;
1.50 kristaps 1138: title = mandoc_realloc(title, titlesz + len);
1139: memcpy(title + titlesz, line, len);
1140: titlesz += len;
1141: title[titlesz - 1] = ' ';
1142: }
1.38 schwarze 1143:
1.50 kristaps 1144: /*
1145: * If no page content can be found, or the input line
1146: * is already the next section header, or there is no
1147: * trailing newline, reuse the page title as the page
1148: * description.
1149: */
1.44 kristaps 1150:
1.50 kristaps 1151: if (NULL == title || '\0' == *title) {
1.56 schwarze 1152: if (warnings)
1.82 schwarze 1153: say(mpage->mlinks->file,
1154: "Cannot find NAME section");
1.78 schwarze 1155: assert(NULL == mpage->desc);
1.82 schwarze 1156: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1157: putkey(mpage, mpage->mlinks->name, TYPE_Nd);
1.50 kristaps 1158: fclose(stream);
1159: free(title);
1160: return;
1161: }
1.1 kristaps 1162:
1.50 kristaps 1163: title = mandoc_realloc(title, titlesz + 1);
1164: title[titlesz] = '\0';
1.33 schwarze 1165:
1.50 kristaps 1166: /*
1167: * Skip to the first dash.
1168: * Use the remaining line as the description (no more than 70
1169: * bytes).
1170: */
1.33 schwarze 1171:
1.50 kristaps 1172: if (NULL != (p = strstr(title, "- "))) {
1173: for (p += 2; ' ' == *p || '\b' == *p; p++)
1174: /* Skip to next word. */ ;
1175: } else {
1.56 schwarze 1176: if (warnings)
1.82 schwarze 1177: say(mpage->mlinks->file,
1178: "No dash in title line");
1.50 kristaps 1179: p = title;
1180: }
1.38 schwarze 1181:
1.50 kristaps 1182: plen = strlen(p);
1.1 kristaps 1183:
1.50 kristaps 1184: /* Strip backspace-encoding from line. */
1.1 kristaps 1185:
1.50 kristaps 1186: while (NULL != (line = memchr(p, '\b', plen))) {
1187: len = line - p;
1188: if (0 == len) {
1189: memmove(line, line + 1, plen--);
1190: continue;
1191: }
1192: memmove(line - 1, line + 1, plen - len);
1193: plen -= 2;
1194: }
1.1 kristaps 1195:
1.78 schwarze 1196: assert(NULL == mpage->desc);
1197: mpage->desc = mandoc_strdup(p);
1198: putkey(mpage, mpage->desc, TYPE_Nd);
1.50 kristaps 1199: fclose(stream);
1200: free(title);
1201: }
1.1 kristaps 1202:
1.50 kristaps 1203: /*
1204: * Put a type/word pair into the word database for this particular file.
1205: */
1206: static void
1.78 schwarze 1207: putkey(const struct mpage *mpage, const char *value, uint64_t type)
1.50 kristaps 1208: {
1.18 kristaps 1209:
1.50 kristaps 1210: assert(NULL != value);
1.78 schwarze 1211: putkeys(mpage, value, strlen(value), type);
1.3 kristaps 1212: }
1213:
1214: /*
1.50 kristaps 1215: * Grok all nodes at or below a certain mdoc node into putkey().
1.3 kristaps 1216: */
1217: static void
1.78 schwarze 1218: putmdockey(const struct mpage *mpage,
1219: const struct mdoc_node *n, uint64_t m)
1.3 kristaps 1220: {
1.18 kristaps 1221:
1.50 kristaps 1222: for ( ; NULL != n; n = n->next) {
1223: if (NULL != n->child)
1.78 schwarze 1224: putmdockey(mpage, n->child, m);
1.50 kristaps 1225: if (MDOC_TEXT == n->type)
1.78 schwarze 1226: putkey(mpage, n->string, m);
1.50 kristaps 1227: }
1228: }
1.18 kristaps 1229:
1.61 schwarze 1230: static void
1.78 schwarze 1231: parse_man(struct mpage *mpage, const struct man_node *n)
1.50 kristaps 1232: {
1233: const struct man_node *head, *body;
1234: char *start, *sv, *title;
1235: char byte;
1236: size_t sz, titlesz;
1.18 kristaps 1237:
1.50 kristaps 1238: if (NULL == n)
1.61 schwarze 1239: return;
1.18 kristaps 1240:
1.50 kristaps 1241: /*
1242: * We're only searching for one thing: the first text child in
1243: * the BODY of a NAME section. Since we don't keep track of
1244: * sections in -man, run some hoops to find out whether we're in
1245: * the correct section or not.
1246: */
1.18 kristaps 1247:
1.50 kristaps 1248: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1249: body = n;
1250: assert(body->parent);
1251: if (NULL != (head = body->parent->head) &&
1252: 1 == head->nchild &&
1253: NULL != (head = (head->child)) &&
1254: MAN_TEXT == head->type &&
1255: 0 == strcmp(head->string, "NAME") &&
1256: NULL != (body = body->child) &&
1257: MAN_TEXT == body->type) {
1.3 kristaps 1258:
1.50 kristaps 1259: title = NULL;
1260: titlesz = 0;
1.3 kristaps 1261:
1.50 kristaps 1262: /*
1263: * Suck the entire NAME section into memory.
1264: * Yes, we might run away.
1265: * But too many manuals have big, spread-out
1266: * NAME sections over many lines.
1267: */
1.3 kristaps 1268:
1.50 kristaps 1269: for ( ; NULL != body; body = body->next) {
1270: if (MAN_TEXT != body->type)
1271: break;
1272: if (0 == (sz = strlen(body->string)))
1273: continue;
1274: title = mandoc_realloc
1275: (title, titlesz + sz + 1);
1276: memcpy(title + titlesz, body->string, sz);
1277: titlesz += sz + 1;
1278: title[titlesz - 1] = ' ';
1279: }
1280: if (NULL == title)
1.61 schwarze 1281: return;
1.18 kristaps 1282:
1.50 kristaps 1283: title = mandoc_realloc(title, titlesz + 1);
1284: title[titlesz] = '\0';
1.18 kristaps 1285:
1.50 kristaps 1286: /* Skip leading space. */
1.18 kristaps 1287:
1.50 kristaps 1288: sv = title;
1289: while (isspace((unsigned char)*sv))
1290: sv++;
1.18 kristaps 1291:
1.50 kristaps 1292: if (0 == (sz = strlen(sv))) {
1293: free(title);
1.61 schwarze 1294: return;
1.50 kristaps 1295: }
1.1 kristaps 1296:
1.50 kristaps 1297: /* Erase trailing space. */
1.1 kristaps 1298:
1.50 kristaps 1299: start = &sv[sz - 1];
1300: while (start > sv && isspace((unsigned char)*start))
1301: *start-- = '\0';
1.1 kristaps 1302:
1.50 kristaps 1303: if (start == sv) {
1304: free(title);
1.61 schwarze 1305: return;
1.50 kristaps 1306: }
1.1 kristaps 1307:
1.50 kristaps 1308: start = sv;
1.18 kristaps 1309:
1.50 kristaps 1310: /*
1311: * Go through a special heuristic dance here.
1312: * Conventionally, one or more manual names are
1313: * comma-specified prior to a whitespace, then a
1314: * dash, then a description. Try to puzzle out
1315: * the name parts here.
1316: */
1.18 kristaps 1317:
1.50 kristaps 1318: for ( ;; ) {
1319: sz = strcspn(start, " ,");
1320: if ('\0' == start[sz])
1321: break;
1.1 kristaps 1322:
1.50 kristaps 1323: byte = start[sz];
1324: start[sz] = '\0';
1.1 kristaps 1325:
1.78 schwarze 1326: putkey(mpage, start, TYPE_Nm);
1.1 kristaps 1327:
1.50 kristaps 1328: if (' ' == byte) {
1329: start += sz + 1;
1330: break;
1331: }
1.1 kristaps 1332:
1.50 kristaps 1333: assert(',' == byte);
1334: start += sz + 1;
1335: while (' ' == *start)
1336: start++;
1337: }
1.1 kristaps 1338:
1.50 kristaps 1339: if (sv == start) {
1.78 schwarze 1340: putkey(mpage, start, TYPE_Nm);
1.50 kristaps 1341: free(title);
1.61 schwarze 1342: return;
1.50 kristaps 1343: }
1.1 kristaps 1344:
1.50 kristaps 1345: while (isspace((unsigned char)*start))
1346: start++;
1.1 kristaps 1347:
1.50 kristaps 1348: if (0 == strncmp(start, "-", 1))
1349: start += 1;
1350: else if (0 == strncmp(start, "\\-\\-", 4))
1351: start += 4;
1352: else if (0 == strncmp(start, "\\-", 2))
1353: start += 2;
1354: else if (0 == strncmp(start, "\\(en", 4))
1355: start += 4;
1356: else if (0 == strncmp(start, "\\(em", 4))
1357: start += 4;
1.1 kristaps 1358:
1.50 kristaps 1359: while (' ' == *start)
1360: start++;
1.1 kristaps 1361:
1.78 schwarze 1362: assert(NULL == mpage->desc);
1363: mpage->desc = mandoc_strdup(start);
1364: putkey(mpage, mpage->desc, TYPE_Nd);
1.50 kristaps 1365: free(title);
1.61 schwarze 1366: return;
1.50 kristaps 1367: }
1368: }
1.1 kristaps 1369:
1.77 schwarze 1370: for (n = n->child; n; n = n->next) {
1.78 schwarze 1371: if (NULL != mpage->desc)
1.77 schwarze 1372: break;
1.78 schwarze 1373: parse_man(mpage, n);
1.77 schwarze 1374: }
1.1 kristaps 1375: }
1376:
1377: static void
1.78 schwarze 1378: parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1379: {
1380:
1.50 kristaps 1381: assert(NULL != n);
1382: for (n = n->child; NULL != n; n = n->next) {
1383: switch (n->type) {
1384: case (MDOC_ELEM):
1385: /* FALLTHROUGH */
1386: case (MDOC_BLOCK):
1387: /* FALLTHROUGH */
1388: case (MDOC_HEAD):
1389: /* FALLTHROUGH */
1390: case (MDOC_BODY):
1391: /* FALLTHROUGH */
1392: case (MDOC_TAIL):
1393: if (NULL != mdocs[n->tok].fp)
1.78 schwarze 1394: if (0 == (*mdocs[n->tok].fp)(mpage, n))
1.50 kristaps 1395: break;
1.68 schwarze 1396: if (mdocs[n->tok].mask)
1.78 schwarze 1397: putmdockey(mpage, n->child,
1398: mdocs[n->tok].mask);
1.50 kristaps 1399: break;
1400: default:
1401: assert(MDOC_ROOT != n->type);
1402: continue;
1403: }
1404: if (NULL != n->child)
1.78 schwarze 1405: parse_mdoc(mpage, n);
1.1 kristaps 1406: }
1407: }
1408:
1.25 schwarze 1409: static int
1.78 schwarze 1410: parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1411: {
1412: const char *start, *end;
1413: size_t sz;
1.25 schwarze 1414:
1.50 kristaps 1415: if (SEC_SYNOPSIS != n->sec ||
1416: NULL == (n = n->child) ||
1417: MDOC_TEXT != n->type)
1.25 schwarze 1418: return(0);
1.1 kristaps 1419:
1420: /*
1421: * Only consider those `Fd' macro fields that begin with an
1422: * "inclusion" token (versus, e.g., #define).
1423: */
1.50 kristaps 1424:
1.1 kristaps 1425: if (strcmp("#include", n->string))
1.25 schwarze 1426: return(0);
1.1 kristaps 1427:
1428: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.25 schwarze 1429: return(0);
1.1 kristaps 1430:
1431: /*
1432: * Strip away the enclosing angle brackets and make sure we're
1433: * not zero-length.
1434: */
1435:
1436: start = n->string;
1437: if ('<' == *start || '"' == *start)
1438: start++;
1439:
1440: if (0 == (sz = strlen(start)))
1.25 schwarze 1441: return(0);
1.1 kristaps 1442:
1443: end = &start[(int)sz - 1];
1444: if ('>' == *end || '"' == *end)
1445: end--;
1446:
1.50 kristaps 1447: if (end > start)
1.78 schwarze 1448: putkeys(mpage, start, end - start + 1, TYPE_In);
1.25 schwarze 1449: return(1);
1.1 kristaps 1450: }
1451:
1.25 schwarze 1452: static int
1.78 schwarze 1453: parse_mdoc_In(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1454: {
1455:
1.50 kristaps 1456: if (NULL != n->child && MDOC_TEXT == n->child->type)
1.25 schwarze 1457: return(0);
1.1 kristaps 1458:
1.78 schwarze 1459: putkey(mpage, n->child->string, TYPE_In);
1.25 schwarze 1460: return(1);
1.1 kristaps 1461: }
1462:
1.25 schwarze 1463: static int
1.78 schwarze 1464: parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1465: {
1466: const char *cp;
1467:
1.50 kristaps 1468: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.25 schwarze 1469: return(0);
1470:
1.50 kristaps 1471: /*
1472: * Parse: .Fn "struct type *name" "char *arg".
1473: * First strip away pointer symbol.
1474: * Then store the function name, then type.
1475: * Finally, store the arguments.
1476: */
1.1 kristaps 1477:
1.50 kristaps 1478: if (NULL == (cp = strrchr(n->string, ' ')))
1479: cp = n->string;
1.1 kristaps 1480:
1481: while ('*' == *cp)
1482: cp++;
1483:
1.78 schwarze 1484: putkey(mpage, cp, TYPE_Fn);
1.25 schwarze 1485:
1.50 kristaps 1486: if (n->string < cp)
1.78 schwarze 1487: putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1.25 schwarze 1488:
1.50 kristaps 1489: for (n = n->next; NULL != n; n = n->next)
1490: if (MDOC_TEXT == n->type)
1.78 schwarze 1491: putkey(mpage, n->string, TYPE_Fa);
1.25 schwarze 1492:
1493: return(0);
1.1 kristaps 1494: }
1495:
1.25 schwarze 1496: static int
1.78 schwarze 1497: parse_mdoc_St(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1498: {
1.25 schwarze 1499:
1.1 kristaps 1500: if (NULL == n->child || MDOC_TEXT != n->child->type)
1.25 schwarze 1501: return(0);
1.1 kristaps 1502:
1.78 schwarze 1503: putkey(mpage, n->child->string, TYPE_St);
1.25 schwarze 1504: return(1);
1.1 kristaps 1505: }
1506:
1.25 schwarze 1507: static int
1.78 schwarze 1508: parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1509: {
1.67 schwarze 1510: char *cp;
1.1 kristaps 1511:
1512: if (NULL == (n = n->child))
1.25 schwarze 1513: return(0);
1.1 kristaps 1514:
1.67 schwarze 1515: if (NULL == n->next) {
1.78 schwarze 1516: putkey(mpage, n->string, TYPE_Xr);
1.67 schwarze 1517: return(0);
1518: }
1519:
1520: if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) {
1521: perror(NULL);
1522: exit((int)MANDOCLEVEL_SYSERR);
1523: }
1.78 schwarze 1524: putkey(mpage, cp, TYPE_Xr);
1.67 schwarze 1525: free(cp);
1526: return(0);
1.1 kristaps 1527: }
1528:
1.25 schwarze 1529: static int
1.78 schwarze 1530: parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1531: {
1.50 kristaps 1532: size_t sz;
1.1 kristaps 1533:
1534: if (MDOC_BODY != n->type)
1.25 schwarze 1535: return(0);
1.1 kristaps 1536:
1.50 kristaps 1537: /*
1538: * Special-case the `Nd' because we need to put the description
1539: * into the document table.
1540: */
1541:
1542: for (n = n->child; NULL != n; n = n->next) {
1543: if (MDOC_TEXT == n->type) {
1.78 schwarze 1544: if (NULL != mpage->desc) {
1545: sz = strlen(mpage->desc) +
1.66 schwarze 1546: strlen(n->string) + 2;
1.78 schwarze 1547: mpage->desc = mandoc_realloc(
1548: mpage->desc, sz);
1549: strlcat(mpage->desc, " ", sz);
1550: strlcat(mpage->desc, n->string, sz);
1.66 schwarze 1551: } else
1.78 schwarze 1552: mpage->desc = mandoc_strdup(n->string);
1.50 kristaps 1553: }
1554: if (NULL != n->child)
1.78 schwarze 1555: parse_mdoc_Nd(mpage, n);
1.50 kristaps 1556: }
1.25 schwarze 1557: return(1);
1.1 kristaps 1558: }
1559:
1.25 schwarze 1560: static int
1.78 schwarze 1561: parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1562: {
1563:
1.25 schwarze 1564: if (SEC_NAME == n->sec)
1565: return(1);
1566: else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
1567: return(0);
1.1 kristaps 1568:
1.25 schwarze 1569: return(1);
1.1 kristaps 1570: }
1571:
1.25 schwarze 1572: static int
1.78 schwarze 1573: parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1574: {
1575:
1.25 schwarze 1576: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 kristaps 1577: }
1578:
1.50 kristaps 1579: static int
1.78 schwarze 1580: parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1581: {
1582:
1.50 kristaps 1583: return(MDOC_HEAD == n->type);
1584: }
1.1 kristaps 1585:
1.50 kristaps 1586: static int
1.78 schwarze 1587: parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1.50 kristaps 1588: {
1.1 kristaps 1589:
1.50 kristaps 1590: return(MDOC_BODY == n->type);
1.1 kristaps 1591: }
1592:
1.50 kristaps 1593: /*
1.66 schwarze 1594: * Add a string to the hash table for the current manual.
1595: * Each string has a bitmask telling which macros it belongs to.
1596: * When we finish the manual, we'll dump the table.
1.50 kristaps 1597: */
1598: static void
1.78 schwarze 1599: putkeys(const struct mpage *mpage,
1600: const char *cp, size_t sz, uint64_t v)
1.50 kristaps 1601: {
1602: struct str *s;
1.65 schwarze 1603: unsigned int slot;
1.50 kristaps 1604: const char *end;
1.25 schwarze 1605:
1.50 kristaps 1606: if (0 == sz)
1607: return;
1.25 schwarze 1608:
1.65 schwarze 1609: end = cp + sz;
1610: slot = ohash_qlookupi(&strings, cp, &end);
1611: s = ohash_find(&strings, slot);
1.25 schwarze 1612:
1.78 schwarze 1613: if (NULL != s && mpage == s->mpage) {
1.50 kristaps 1614: s->mask |= v;
1615: return;
1616: } else if (NULL == s) {
1.51 kristaps 1617: s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1.50 kristaps 1618: memcpy(s->key, cp, sz);
1.65 schwarze 1619: ohash_insert(&strings, slot, s);
1.1 kristaps 1620: }
1.78 schwarze 1621: s->mpage = mpage;
1.50 kristaps 1622: s->mask = v;
1.1 kristaps 1623: }
1624:
1.50 kristaps 1625: /*
1626: * Take a Unicode codepoint and produce its UTF-8 encoding.
1627: * This isn't the best way to do this, but it works.
1628: * The magic numbers are from the UTF-8 packaging.
1629: * They're not as scary as they seem: read the UTF-8 spec for details.
1630: */
1631: static size_t
1632: utf8(unsigned int cp, char out[7])
1.1 kristaps 1633: {
1.50 kristaps 1634: size_t rc;
1.1 kristaps 1635:
1.50 kristaps 1636: rc = 0;
1637: if (cp <= 0x0000007F) {
1638: rc = 1;
1639: out[0] = (char)cp;
1640: } else if (cp <= 0x000007FF) {
1641: rc = 2;
1642: out[0] = (cp >> 6 & 31) | 192;
1643: out[1] = (cp & 63) | 128;
1644: } else if (cp <= 0x0000FFFF) {
1645: rc = 3;
1646: out[0] = (cp >> 12 & 15) | 224;
1647: out[1] = (cp >> 6 & 63) | 128;
1648: out[2] = (cp & 63) | 128;
1649: } else if (cp <= 0x001FFFFF) {
1650: rc = 4;
1651: out[0] = (cp >> 18 & 7) | 240;
1652: out[1] = (cp >> 12 & 63) | 128;
1653: out[2] = (cp >> 6 & 63) | 128;
1654: out[3] = (cp & 63) | 128;
1655: } else if (cp <= 0x03FFFFFF) {
1656: rc = 5;
1657: out[0] = (cp >> 24 & 3) | 248;
1658: out[1] = (cp >> 18 & 63) | 128;
1659: out[2] = (cp >> 12 & 63) | 128;
1660: out[3] = (cp >> 6 & 63) | 128;
1661: out[4] = (cp & 63) | 128;
1662: } else if (cp <= 0x7FFFFFFF) {
1663: rc = 6;
1664: out[0] = (cp >> 30 & 1) | 252;
1665: out[1] = (cp >> 24 & 63) | 128;
1666: out[2] = (cp >> 18 & 63) | 128;
1667: out[3] = (cp >> 12 & 63) | 128;
1668: out[4] = (cp >> 6 & 63) | 128;
1669: out[5] = (cp & 63) | 128;
1670: } else
1.1 kristaps 1671: return(0);
1672:
1.50 kristaps 1673: out[rc] = '\0';
1674: return(rc);
1675: }
1.1 kristaps 1676:
1.50 kristaps 1677: /*
1678: * Store the UTF-8 version of a key, or alias the pointer if the key has
1679: * no UTF-8 transcription marks in it.
1680: */
1681: static void
1682: utf8key(struct mchars *mc, struct str *key)
1683: {
1684: size_t sz, bsz, pos;
1685: char utfbuf[7], res[5];
1686: char *buf;
1687: const char *seq, *cpp, *val;
1688: int len, u;
1689: enum mandoc_esc esc;
1690:
1691: assert(NULL == key->utf8);
1692:
1693: res[0] = '\\';
1694: res[1] = '\t';
1695: res[2] = ASCII_NBRSP;
1696: res[3] = ASCII_HYPH;
1697: res[4] = '\0';
1.1 kristaps 1698:
1.50 kristaps 1699: val = key->key;
1700: bsz = strlen(val);
1.46 kristaps 1701:
1.50 kristaps 1702: /*
1703: * Pre-check: if we have no stop-characters, then set the
1704: * pointer as ourselvse and get out of here.
1705: */
1706: if (strcspn(val, res) == bsz) {
1707: key->utf8 = key->key;
1708: return;
1709: }
1.46 kristaps 1710:
1.50 kristaps 1711: /* Pre-allocate by the length of the input */
1.46 kristaps 1712:
1.50 kristaps 1713: buf = mandoc_malloc(++bsz);
1714: pos = 0;
1.46 kristaps 1715:
1.50 kristaps 1716: while ('\0' != *val) {
1717: /*
1718: * Halt on the first escape sequence.
1719: * This also halts on the end of string, in which case
1720: * we just copy, fallthrough, and exit the loop.
1721: */
1722: if ((sz = strcspn(val, res)) > 0) {
1723: memcpy(&buf[pos], val, sz);
1724: pos += sz;
1725: val += sz;
1726: }
1.46 kristaps 1727:
1.50 kristaps 1728: if (ASCII_HYPH == *val) {
1729: buf[pos++] = '-';
1730: val++;
1731: continue;
1732: } else if ('\t' == *val || ASCII_NBRSP == *val) {
1733: buf[pos++] = ' ';
1734: val++;
1735: continue;
1736: } else if ('\\' != *val)
1737: break;
1.46 kristaps 1738:
1.50 kristaps 1739: /* Read past the slash. */
1.46 kristaps 1740:
1.50 kristaps 1741: val++;
1742: u = 0;
1.46 kristaps 1743:
1.50 kristaps 1744: /*
1745: * Parse the escape sequence and see if it's a
1746: * predefined character or special character.
1747: */
1748: esc = mandoc_escape
1749: ((const char **)&val, &seq, &len);
1750: if (ESCAPE_ERROR == esc)
1751: break;
1.1 kristaps 1752:
1.50 kristaps 1753: if (ESCAPE_SPECIAL != esc)
1754: continue;
1755: if (0 == (u = mchars_spec2cp(mc, seq, len)))
1756: continue;
1.1 kristaps 1757:
1.50 kristaps 1758: /*
1759: * If we have a Unicode codepoint, try to convert that
1760: * to a UTF-8 byte string.
1761: */
1762: cpp = utfbuf;
1763: if (0 == (sz = utf8(u, utfbuf)))
1764: continue;
1.1 kristaps 1765:
1.50 kristaps 1766: /* Copy the rendered glyph into the stream. */
1.1 kristaps 1767:
1.50 kristaps 1768: sz = strlen(cpp);
1769: bsz += sz;
1.1 kristaps 1770:
1.50 kristaps 1771: buf = mandoc_realloc(buf, bsz);
1.1 kristaps 1772:
1.50 kristaps 1773: memcpy(&buf[pos], cpp, sz);
1774: pos += sz;
1.1 kristaps 1775: }
1776:
1.50 kristaps 1777: buf[pos] = '\0';
1778: key->utf8 = buf;
1.1 kristaps 1779: }
1780:
1.14 schwarze 1781: /*
1.50 kristaps 1782: * Flush the current page's terms (and their bits) into the database.
1783: * Wrap the entire set of additions in a transaction to make sqlite be a
1784: * little faster.
1785: * Also, UTF-8-encode the description at the last possible moment.
1.14 schwarze 1786: */
1787: static void
1.82 schwarze 1788: dbindex(const struct mpage *mpage, struct mchars *mc)
1.14 schwarze 1789: {
1.50 kristaps 1790: struct str *key;
1791: const char *desc;
1792: int64_t recno;
1.52 kristaps 1793: size_t i;
1.66 schwarze 1794: unsigned int slot;
1.14 schwarze 1795:
1.56 schwarze 1796: if (verb)
1.82 schwarze 1797: say(mpage->mlinks->file, "Adding to index");
1.43 kristaps 1798:
1.50 kristaps 1799: if (nodb)
1.14 schwarze 1800: return;
1.28 kristaps 1801:
1.50 kristaps 1802: desc = "";
1.78 schwarze 1803: if (NULL != mpage->desc && '\0' != *mpage->desc) {
1.65 schwarze 1804: key = ohash_find(&strings,
1.78 schwarze 1805: ohash_qlookup(&strings, mpage->desc));
1.50 kristaps 1806: assert(NULL != key);
1807: if (NULL == key->utf8)
1808: utf8key(mc, key);
1809: desc = key->utf8;
1810: }
1811:
1.52 kristaps 1812: SQL_EXEC("BEGIN TRANSACTION");
1.50 kristaps 1813:
1.52 kristaps 1814: i = 1;
1.81 schwarze 1815: /*
1816: * XXX The following three lines are obsolete
1817: * and only kept for backward compatibility
1818: * until apropos(1) and friends have caught up.
1819: */
1.82 schwarze 1820: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->file);
1821: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->dsec);
1822: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->arch);
1.81 schwarze 1823: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, desc);
1.82 schwarze 1824: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
1.81 schwarze 1825: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.50 kristaps 1826: recno = sqlite3_last_insert_rowid(db);
1.81 schwarze 1827: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1828:
1829: i = 1;
1.82 schwarze 1830: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mpage->mlinks->dsec);
1831: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mpage->mlinks->arch);
1832: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mpage->mlinks->file);
1.81 schwarze 1833: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, recno);
1834: SQL_STEP(stmts[STMT_INSERT_LINK]);
1835: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1.50 kristaps 1836:
1.66 schwarze 1837: for (key = ohash_first(&strings, &slot); NULL != key;
1838: key = ohash_next(&strings, &slot)) {
1.78 schwarze 1839: assert(key->mpage == mpage);
1.50 kristaps 1840: if (NULL == key->utf8)
1841: utf8key(mc, key);
1.52 kristaps 1842: i = 1;
1843: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1844: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->utf8);
1845: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
1846: SQL_STEP(stmts[STMT_INSERT_KEY]);
1.50 kristaps 1847: sqlite3_reset(stmts[STMT_INSERT_KEY]);
1.66 schwarze 1848: if (key->utf8 != key->key)
1849: free(key->utf8);
1850: free(key);
1.38 schwarze 1851: }
1.28 kristaps 1852:
1.52 kristaps 1853: SQL_EXEC("END TRANSACTION");
1.14 schwarze 1854: }
1855:
1.5 kristaps 1856: static void
1.59 schwarze 1857: dbprune(void)
1.5 kristaps 1858: {
1.78 schwarze 1859: struct mpage *mpage;
1.82 schwarze 1860: struct mlink *mlink;
1.52 kristaps 1861: size_t i;
1.80 schwarze 1862: unsigned int slot;
1.5 kristaps 1863:
1.50 kristaps 1864: if (nodb)
1865: return;
1.12 schwarze 1866:
1.80 schwarze 1867: mpage = ohash_first(&mpages, &slot);
1868: while (NULL != mpage) {
1.82 schwarze 1869: mlink = mpage->mlinks;
1.52 kristaps 1870: i = 1;
1.82 schwarze 1871: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE], i, mlink->file);
1.81 schwarze 1872: SQL_STEP(stmts[STMT_DELETE_PAGE]);
1873: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
1.56 schwarze 1874: if (verb)
1.82 schwarze 1875: say(mlink->file, "Deleted from index");
1.80 schwarze 1876: mpage = ohash_next(&mpages, &slot);
1.5 kristaps 1877: }
1878: }
1879:
1.4 kristaps 1880: /*
1.50 kristaps 1881: * Close an existing database and its prepared statements.
1882: * If "real" is not set, rename the temporary file into the real one.
1.4 kristaps 1883: */
1.35 kristaps 1884: static void
1.59 schwarze 1885: dbclose(int real)
1.4 kristaps 1886: {
1.50 kristaps 1887: size_t i;
1.4 kristaps 1888:
1.50 kristaps 1889: if (nodb)
1.38 schwarze 1890: return;
1.50 kristaps 1891:
1892: for (i = 0; i < STMT__MAX; i++) {
1893: sqlite3_finalize(stmts[i]);
1894: stmts[i] = NULL;
1.4 kristaps 1895: }
1896:
1.50 kristaps 1897: sqlite3_close(db);
1898: db = NULL;
1.12 schwarze 1899:
1.50 kristaps 1900: if (real)
1901: return;
1.12 schwarze 1902:
1.63 schwarze 1903: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
1.59 schwarze 1904: exitcode = (int)MANDOCLEVEL_SYSERR;
1905: say(MANDOC_DB, NULL);
1906: }
1.50 kristaps 1907: }
1.14 schwarze 1908:
1.50 kristaps 1909: /*
1910: * This is straightforward stuff.
1911: * Open a database connection to a "temporary" database, then open a set
1912: * of prepared statements we'll use over and over again.
1913: * If "real" is set, we use the existing database; if not, we truncate a
1914: * temporary one.
1915: * Must be matched by dbclose().
1916: */
1917: static int
1.59 schwarze 1918: dbopen(int real)
1.50 kristaps 1919: {
1.63 schwarze 1920: const char *file, *sql;
1.50 kristaps 1921: int rc, ofl;
1.12 schwarze 1922:
1.50 kristaps 1923: if (nodb)
1924: return(1);
1.12 schwarze 1925:
1.63 schwarze 1926: ofl = SQLITE_OPEN_READWRITE;
1927: if (0 == real) {
1928: file = MANDOC_DB "~";
1929: if (-1 == remove(file) && ENOENT != errno) {
1930: exitcode = (int)MANDOCLEVEL_SYSERR;
1931: say(file, NULL);
1932: return(0);
1933: }
1934: ofl |= SQLITE_OPEN_EXCLUSIVE;
1935: } else
1936: file = MANDOC_DB;
1.45 kristaps 1937:
1.50 kristaps 1938: rc = sqlite3_open_v2(file, &db, ofl, NULL);
1939: if (SQLITE_OK == rc)
1.57 schwarze 1940: goto prepare_statements;
1.50 kristaps 1941: if (SQLITE_CANTOPEN != rc) {
1.59 schwarze 1942: exitcode = (int)MANDOCLEVEL_SYSERR;
1943: say(file, NULL);
1.50 kristaps 1944: return(0);
1945: }
1.12 schwarze 1946:
1.50 kristaps 1947: sqlite3_close(db);
1948: db = NULL;
1.12 schwarze 1949:
1.50 kristaps 1950: if (SQLITE_OK != (rc = sqlite3_open(file, &db))) {
1.59 schwarze 1951: exitcode = (int)MANDOCLEVEL_SYSERR;
1952: say(file, NULL);
1.50 kristaps 1953: return(0);
1954: }
1.12 schwarze 1955:
1.81 schwarze 1956: /*
1957: * XXX The first three columns in table mpages are obsolete
1958: * and only kept for backward compatibility
1959: * until apropos(1) and friends have caught up.
1960: */
1961: sql = "CREATE TABLE \"mpages\" (\n"
1.50 kristaps 1962: " \"file\" TEXT NOT NULL,\n"
1963: " \"sec\" TEXT NOT NULL,\n"
1964: " \"arch\" TEXT NOT NULL,\n"
1965: " \"desc\" TEXT NOT NULL,\n"
1966: " \"form\" INTEGER NOT NULL,\n"
1967: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1968: ");\n"
1969: "\n"
1.81 schwarze 1970: "CREATE TABLE \"mlinks\" (\n"
1971: " \"sec\" TEXT NOT NULL,\n"
1972: " \"arch\" TEXT NOT NULL,\n"
1973: " \"name\" TEXT NOT NULL,\n"
1974: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1975: "ON DELETE CASCADE,\n"
1976: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1977: ");\n"
1978: "\n"
1.50 kristaps 1979: "CREATE TABLE \"keys\" (\n"
1980: " \"bits\" INTEGER NOT NULL,\n"
1981: " \"key\" TEXT NOT NULL,\n"
1.81 schwarze 1982: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1983: "ON DELETE CASCADE,\n"
1.50 kristaps 1984: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1985: ");\n"
1986: "\n"
1987: "CREATE INDEX \"key_index\" ON keys (key);\n";
1.14 schwarze 1988:
1.50 kristaps 1989: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1.59 schwarze 1990: exitcode = (int)MANDOCLEVEL_SYSERR;
1991: say(file, "%s", sqlite3_errmsg(db));
1.50 kristaps 1992: return(0);
1993: }
1.4 kristaps 1994:
1.57 schwarze 1995: prepare_statements:
1996: SQL_EXEC("PRAGMA foreign_keys = ON");
1.81 schwarze 1997: sql = "DELETE FROM mpages where file=?";
1998: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
1999: sql = "INSERT INTO mpages "
1.50 kristaps 2000: "(file,sec,arch,desc,form) VALUES (?,?,?,?,?)";
1.81 schwarze 2001: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2002: sql = "INSERT INTO mlinks "
2003: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2004: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.50 kristaps 2005: sql = "INSERT INTO keys "
1.81 schwarze 2006: "(bits,key,pageid) VALUES (?,?,?)";
1.50 kristaps 2007: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.70 schwarze 2008:
2009: #ifndef __APPLE__
2010: /*
2011: * When opening a new database, we can turn off
2012: * synchronous mode for much better performance.
2013: */
2014:
2015: if (real)
2016: SQL_EXEC("PRAGMA synchronous = OFF");
2017: #endif
2018:
1.50 kristaps 2019: return(1);
2020: }
1.5 kristaps 2021:
1.50 kristaps 2022: static void *
2023: hash_halloc(size_t sz, void *arg)
2024: {
1.12 schwarze 2025:
1.50 kristaps 2026: return(mandoc_calloc(sz, 1));
2027: }
1.12 schwarze 2028:
1.50 kristaps 2029: static void *
2030: hash_alloc(size_t sz, void *arg)
2031: {
1.5 kristaps 2032:
1.50 kristaps 2033: return(mandoc_malloc(sz));
2034: }
1.41 kristaps 2035:
1.50 kristaps 2036: static void
2037: hash_free(void *p, size_t sz, void *arg)
2038: {
1.4 kristaps 2039:
1.50 kristaps 2040: free(p);
1.4 kristaps 2041: }
2042:
1.50 kristaps 2043: static int
1.59 schwarze 2044: set_basedir(const char *targetdir)
1.4 kristaps 2045: {
1.59 schwarze 2046: static char startdir[PATH_MAX];
2047: static int fd;
1.4 kristaps 2048:
1.59 schwarze 2049: /*
2050: * Remember where we started by keeping a fd open to the origin
2051: * path component: throughout this utility, we chdir() a lot to
2052: * handle relative paths, and by doing this, we can return to
2053: * the starting point.
2054: */
2055: if ('\0' == *startdir) {
2056: if (NULL == getcwd(startdir, PATH_MAX)) {
2057: exitcode = (int)MANDOCLEVEL_SYSERR;
2058: if (NULL != targetdir)
2059: say(".", NULL);
2060: return(0);
2061: }
2062: if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
2063: exitcode = (int)MANDOCLEVEL_SYSERR;
2064: say(startdir, NULL);
2065: return(0);
2066: }
2067: if (NULL == targetdir)
2068: targetdir = startdir;
2069: } else {
2070: if (-1 == fd)
2071: return(0);
2072: if (-1 == fchdir(fd)) {
2073: close(fd);
2074: basedir[0] = '\0';
2075: exitcode = (int)MANDOCLEVEL_SYSERR;
2076: say(startdir, NULL);
2077: return(0);
2078: }
2079: if (NULL == targetdir) {
2080: close(fd);
2081: return(1);
2082: }
2083: }
2084: if (NULL == realpath(targetdir, basedir)) {
2085: basedir[0] = '\0';
2086: exitcode = (int)MANDOCLEVEL_BADARG;
2087: say(targetdir, NULL);
1.50 kristaps 2088: return(0);
1.59 schwarze 2089: } else if (-1 == chdir(basedir)) {
2090: exitcode = (int)MANDOCLEVEL_BADARG;
2091: say("", NULL);
1.50 kristaps 2092: return(0);
1.4 kristaps 2093: }
1.50 kristaps 2094: return(1);
1.56 schwarze 2095: }
2096:
2097: static void
1.59 schwarze 2098: say(const char *file, const char *format, ...)
1.56 schwarze 2099: {
2100: va_list ap;
2101:
1.59 schwarze 2102: if ('\0' != *basedir)
2103: fprintf(stderr, "%s", basedir);
2104: if ('\0' != *basedir && '\0' != *file)
2105: fputs("//", stderr);
1.56 schwarze 2106: if ('\0' != *file)
1.59 schwarze 2107: fprintf(stderr, "%s", file);
1.56 schwarze 2108: fputs(": ", stderr);
1.59 schwarze 2109:
2110: if (NULL == format) {
2111: perror(NULL);
2112: return;
2113: }
1.56 schwarze 2114:
2115: va_start(ap, format);
2116: vfprintf(stderr, format, ap);
2117: va_end(ap);
2118:
2119: fputc('\n', stderr);
1.1 kristaps 2120: }
CVSweb