Annotation of mandoc/mandocdb.c, Revision 1.92
1.92 ! schwarze 1: /* $Id: mandocdb.c,v 1.91 2013/12/31 18:07:42 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 *);
1.84 schwarze 145: static void mlink_add(struct mlink *, const struct stat *);
1.82 schwarze 146: static void mlink_free(struct mlink *);
1.89 schwarze 147: static void mlinks_undupe(struct mpage *);
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_Nd(struct mpage *, const struct mdoc_node *);
158: static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
159: static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
160: static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
161: static void putkey(const struct mpage *,
1.50 kristaps 162: const char *, uint64_t);
1.78 schwarze 163: static void putkeys(const struct mpage *,
1.64 schwarze 164: const char *, size_t, uint64_t);
1.78 schwarze 165: static void putmdockey(const struct mpage *,
1.50 kristaps 166: const struct mdoc_node *, uint64_t);
1.59 schwarze 167: static void say(const char *, const char *, ...);
1.80 schwarze 168: static int set_basedir(const char *);
1.59 schwarze 169: static int treescan(void);
1.50 kristaps 170: static size_t utf8(unsigned int, char [7]);
171: static void utf8key(struct mchars *, struct str *);
172:
173: static char *progname;
174: static int use_all; /* use all found files */
175: static int nodb; /* no database changes */
176: static int verb; /* print what we're doing */
177: static int warnings; /* warn about crap */
1.59 schwarze 178: static int exitcode; /* to be returned by main */
1.50 kristaps 179: static enum op op; /* operational mode */
1.59 schwarze 180: static char basedir[PATH_MAX]; /* current base directory */
1.78 schwarze 181: static struct ohash mpages; /* table of distinct manual pages */
1.83 schwarze 182: static struct ohash mlinks; /* table of directory entries */
1.50 kristaps 183: static struct ohash strings; /* table of all strings */
184: static sqlite3 *db = NULL; /* current database */
185: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.25 schwarze 186:
187: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.68 schwarze 188: { NULL, 0 }, /* Ap */
189: { NULL, 0 }, /* Dd */
190: { NULL, 0 }, /* Dt */
191: { NULL, 0 }, /* Os */
192: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
193: { parse_mdoc_head, TYPE_Ss }, /* Ss */
194: { NULL, 0 }, /* Pp */
195: { NULL, 0 }, /* D1 */
196: { NULL, 0 }, /* Dl */
197: { NULL, 0 }, /* Bd */
198: { NULL, 0 }, /* Ed */
199: { NULL, 0 }, /* Bl */
200: { NULL, 0 }, /* El */
201: { NULL, 0 }, /* It */
202: { NULL, 0 }, /* Ad */
203: { NULL, TYPE_An }, /* An */
204: { NULL, TYPE_Ar }, /* Ar */
205: { NULL, TYPE_Cd }, /* Cd */
206: { NULL, TYPE_Cm }, /* Cm */
207: { NULL, TYPE_Dv }, /* Dv */
208: { NULL, TYPE_Er }, /* Er */
209: { NULL, TYPE_Ev }, /* Ev */
210: { NULL, 0 }, /* Ex */
211: { NULL, TYPE_Fa }, /* Fa */
212: { parse_mdoc_Fd, 0 }, /* Fd */
213: { NULL, TYPE_Fl }, /* Fl */
214: { parse_mdoc_Fn, 0 }, /* Fn */
215: { NULL, TYPE_Ft }, /* Ft */
216: { NULL, TYPE_Ic }, /* Ic */
1.92 ! schwarze 217: { NULL, TYPE_In }, /* In */
1.68 schwarze 218: { NULL, TYPE_Li }, /* Li */
219: { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
220: { parse_mdoc_Nm, TYPE_Nm }, /* Nm */
221: { NULL, 0 }, /* Op */
222: { NULL, 0 }, /* Ot */
223: { NULL, TYPE_Pa }, /* Pa */
224: { NULL, 0 }, /* Rv */
1.92 ! schwarze 225: { NULL, TYPE_St }, /* St */
1.68 schwarze 226: { NULL, TYPE_Va }, /* Va */
227: { parse_mdoc_body, TYPE_Va }, /* Vt */
228: { parse_mdoc_Xr, 0 }, /* Xr */
229: { NULL, 0 }, /* %A */
230: { NULL, 0 }, /* %B */
231: { NULL, 0 }, /* %D */
232: { NULL, 0 }, /* %I */
233: { NULL, 0 }, /* %J */
234: { NULL, 0 }, /* %N */
235: { NULL, 0 }, /* %O */
236: { NULL, 0 }, /* %P */
237: { NULL, 0 }, /* %R */
238: { NULL, 0 }, /* %T */
239: { NULL, 0 }, /* %V */
240: { NULL, 0 }, /* Ac */
241: { NULL, 0 }, /* Ao */
242: { NULL, 0 }, /* Aq */
243: { NULL, TYPE_At }, /* At */
244: { NULL, 0 }, /* Bc */
245: { NULL, 0 }, /* Bf */
246: { NULL, 0 }, /* Bo */
247: { NULL, 0 }, /* Bq */
248: { NULL, TYPE_Bsx }, /* Bsx */
249: { NULL, TYPE_Bx }, /* Bx */
250: { NULL, 0 }, /* Db */
251: { NULL, 0 }, /* Dc */
252: { NULL, 0 }, /* Do */
253: { NULL, 0 }, /* Dq */
254: { NULL, 0 }, /* Ec */
255: { NULL, 0 }, /* Ef */
256: { NULL, TYPE_Em }, /* Em */
257: { NULL, 0 }, /* Eo */
258: { NULL, TYPE_Fx }, /* Fx */
259: { NULL, TYPE_Ms }, /* Ms */
260: { NULL, 0 }, /* No */
261: { NULL, 0 }, /* Ns */
262: { NULL, TYPE_Nx }, /* Nx */
263: { NULL, TYPE_Ox }, /* Ox */
264: { NULL, 0 }, /* Pc */
265: { NULL, 0 }, /* Pf */
266: { NULL, 0 }, /* Po */
267: { NULL, 0 }, /* Pq */
268: { NULL, 0 }, /* Qc */
269: { NULL, 0 }, /* Ql */
270: { NULL, 0 }, /* Qo */
271: { NULL, 0 }, /* Qq */
272: { NULL, 0 }, /* Re */
273: { NULL, 0 }, /* Rs */
274: { NULL, 0 }, /* Sc */
275: { NULL, 0 }, /* So */
276: { NULL, 0 }, /* Sq */
277: { NULL, 0 }, /* Sm */
278: { NULL, 0 }, /* Sx */
279: { NULL, TYPE_Sy }, /* Sy */
280: { NULL, TYPE_Tn }, /* Tn */
281: { NULL, 0 }, /* Ux */
282: { NULL, 0 }, /* Xc */
283: { NULL, 0 }, /* Xo */
284: { parse_mdoc_head, 0 }, /* Fo */
285: { NULL, 0 }, /* Fc */
286: { NULL, 0 }, /* Oo */
287: { NULL, 0 }, /* Oc */
288: { NULL, 0 }, /* Bk */
289: { NULL, 0 }, /* Ek */
290: { NULL, 0 }, /* Bt */
291: { NULL, 0 }, /* Hf */
292: { NULL, 0 }, /* Fr */
293: { NULL, 0 }, /* Ud */
294: { NULL, TYPE_Lb }, /* Lb */
295: { NULL, 0 }, /* Lp */
296: { NULL, TYPE_Lk }, /* Lk */
297: { NULL, TYPE_Mt }, /* Mt */
298: { NULL, 0 }, /* Brq */
299: { NULL, 0 }, /* Bro */
300: { NULL, 0 }, /* Brc */
301: { NULL, 0 }, /* %C */
302: { NULL, 0 }, /* Es */
303: { NULL, 0 }, /* En */
304: { NULL, TYPE_Dx }, /* Dx */
305: { NULL, 0 }, /* %Q */
306: { NULL, 0 }, /* br */
307: { NULL, 0 }, /* sp */
308: { NULL, 0 }, /* %U */
309: { NULL, 0 }, /* Ta */
1.1 kristaps 310: };
311:
312: int
313: main(int argc, char *argv[])
314: {
1.59 schwarze 315: int ch, i;
1.50 kristaps 316: size_t j, sz;
1.59 schwarze 317: const char *path_arg;
1.50 kristaps 318: struct mchars *mc;
319: struct manpaths dirs;
320: struct mparse *mp;
1.83 schwarze 321: struct ohash_info mpages_info, mlinks_info;
1.50 kristaps 322:
323: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
324: memset(&dirs, 0, sizeof(struct manpaths));
325:
1.83 schwarze 326: mpages_info.alloc = mlinks_info.alloc = hash_alloc;
327: mpages_info.halloc = mlinks_info.halloc = hash_halloc;
328: mpages_info.hfree = mlinks_info.hfree = hash_free;
1.50 kristaps 329:
1.79 schwarze 330: mpages_info.key_offset = offsetof(struct mpage, inodev);
1.83 schwarze 331: mlinks_info.key_offset = offsetof(struct mlink, file);
1.1 kristaps 332:
333: progname = strrchr(argv[0], '/');
334: if (progname == NULL)
335: progname = argv[0];
336: else
337: ++progname;
338:
1.50 kristaps 339: /*
340: * We accept a few different invocations.
341: * The CHECKOP macro makes sure that invocation styles don't
342: * clobber each other.
343: */
344: #define CHECKOP(_op, _ch) do \
345: if (OP_DEFAULT != (_op)) { \
346: fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
347: goto usage; \
348: } while (/*CONSTCOND*/0)
1.10 kristaps 349:
1.59 schwarze 350: path_arg = NULL;
1.38 schwarze 351: op = OP_DEFAULT;
1.1 kristaps 352:
1.50 kristaps 353: while (-1 != (ch = getopt(argc, argv, "aC:d:ntu:vW")))
1.1 kristaps 354: switch (ch) {
1.12 schwarze 355: case ('a'):
356: use_all = 1;
357: break;
1.34 schwarze 358: case ('C'):
1.50 kristaps 359: CHECKOP(op, ch);
1.59 schwarze 360: path_arg = optarg;
1.38 schwarze 361: op = OP_CONFFILE;
1.34 schwarze 362: break;
1.5 kristaps 363: case ('d'):
1.50 kristaps 364: CHECKOP(op, ch);
1.59 schwarze 365: path_arg = optarg;
1.5 kristaps 366: op = OP_UPDATE;
367: break;
1.50 kristaps 368: case ('n'):
369: nodb = 1;
370: break;
1.38 schwarze 371: case ('t'):
1.50 kristaps 372: CHECKOP(op, ch);
1.38 schwarze 373: dup2(STDOUT_FILENO, STDERR_FILENO);
374: op = OP_TEST;
1.50 kristaps 375: nodb = warnings = 1;
1.38 schwarze 376: break;
1.5 kristaps 377: case ('u'):
1.50 kristaps 378: CHECKOP(op, ch);
1.59 schwarze 379: path_arg = optarg;
1.5 kristaps 380: op = OP_DELETE;
381: break;
382: case ('v'):
383: verb++;
384: break;
1.38 schwarze 385: case ('W'):
386: warnings = 1;
387: break;
1.1 kristaps 388: default:
1.38 schwarze 389: goto usage;
1.1 kristaps 390: }
391:
392: argc -= optind;
393: argv += optind;
394:
1.38 schwarze 395: if (OP_CONFFILE == op && argc > 0) {
1.50 kristaps 396: fprintf(stderr, "-C: Too many arguments\n");
1.38 schwarze 397: goto usage;
398: }
399:
1.59 schwarze 400: exitcode = (int)MANDOCLEVEL_OK;
1.50 kristaps 401: mp = mparse_alloc(MPARSE_AUTO,
402: MANDOCLEVEL_FATAL, NULL, NULL, NULL);
403: mc = mchars_alloc();
404:
1.78 schwarze 405: ohash_init(&mpages, 6, &mpages_info);
1.83 schwarze 406: ohash_init(&mlinks, 6, &mlinks_info);
1.50 kristaps 407:
408: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
409: /*
410: * Force processing all files.
411: */
412: use_all = 1;
1.59 schwarze 413:
1.50 kristaps 414: /*
415: * All of these deal with a specific directory.
416: * Jump into that directory then collect files specified
417: * on the command-line.
418: */
1.59 schwarze 419: if (0 == set_basedir(path_arg))
1.50 kristaps 420: goto out;
421: for (i = 0; i < argc; i++)
1.59 schwarze 422: filescan(argv[i]);
423: if (0 == dbopen(1))
1.50 kristaps 424: goto out;
425: if (OP_TEST != op)
1.59 schwarze 426: dbprune();
1.50 kristaps 427: if (OP_DELETE != op)
1.78 schwarze 428: mpages_merge(mc, mp, 0);
1.59 schwarze 429: dbclose(1);
1.50 kristaps 430: } else {
431: /*
432: * If we have arguments, use them as our manpaths.
433: * If we don't, grok from manpath(1) or however else
434: * manpath_parse() wants to do it.
435: */
436: if (argc > 0) {
437: dirs.paths = mandoc_calloc
438: (argc, sizeof(char *));
439: dirs.sz = (size_t)argc;
440: for (i = 0; i < argc; i++)
441: dirs.paths[i] = mandoc_strdup(argv[i]);
442: } else
1.59 schwarze 443: manpath_parse(&dirs, path_arg, NULL, NULL);
1.50 kristaps 444:
445: /*
1.69 schwarze 446: * First scan the tree rooted at a base directory, then
447: * build a new database and finally move it into place.
1.50 kristaps 448: * Ignore zero-length directories and strip trailing
449: * slashes.
450: */
451: for (j = 0; j < dirs.sz; j++) {
452: sz = strlen(dirs.paths[j]);
453: if (sz && '/' == dirs.paths[j][sz - 1])
454: dirs.paths[j][--sz] = '\0';
455: if (0 == sz)
456: continue;
1.66 schwarze 457:
458: if (j) {
1.78 schwarze 459: ohash_init(&mpages, 6, &mpages_info);
1.83 schwarze 460: ohash_init(&mlinks, 6, &mlinks_info);
1.66 schwarze 461: }
462:
1.59 schwarze 463: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 464: goto out;
1.59 schwarze 465: if (0 == treescan())
1.50 kristaps 466: goto out;
1.59 schwarze 467: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 468: goto out;
1.59 schwarze 469: if (0 == dbopen(0))
1.50 kristaps 470: goto out;
1.52 kristaps 471:
1.78 schwarze 472: mpages_merge(mc, mp, warnings && !use_all);
1.59 schwarze 473: dbclose(0);
1.66 schwarze 474:
475: if (j + 1 < dirs.sz) {
1.80 schwarze 476: mpages_free();
1.78 schwarze 477: ohash_delete(&mpages);
1.83 schwarze 478: ohash_delete(&mlinks);
1.66 schwarze 479: }
1.50 kristaps 480: }
481: }
482: out:
1.59 schwarze 483: set_basedir(NULL);
1.50 kristaps 484: manpath_free(&dirs);
485: mchars_free(mc);
486: mparse_free(mp);
1.80 schwarze 487: mpages_free();
1.78 schwarze 488: ohash_delete(&mpages);
1.83 schwarze 489: ohash_delete(&mlinks);
1.59 schwarze 490: return(exitcode);
1.50 kristaps 491: usage:
492: fprintf(stderr, "usage: %s [-anvW] [-C file]\n"
493: " %s [-anvW] dir ...\n"
494: " %s [-nvW] -d dir [file ...]\n"
495: " %s [-nvW] -u dir [file ...]\n"
496: " %s -t file ...\n",
497: progname, progname, progname,
498: progname, progname);
499:
1.59 schwarze 500: return((int)MANDOCLEVEL_BADARG);
1.50 kristaps 501: }
502:
503: /*
1.59 schwarze 504: * Scan a directory tree rooted at "basedir" for manpages.
1.50 kristaps 505: * We use fts(), scanning directory parts along the way for clues to our
506: * section and architecture.
507: *
508: * If use_all has been specified, grok all files.
509: * If not, sanitise paths to the following:
510: *
511: * [./]man*[/<arch>]/<name>.<section>
512: * or
513: * [./]cat<section>[/<arch>]/<name>.0
514: *
515: * TODO: accomodate for multi-language directories.
516: */
517: static int
1.59 schwarze 518: treescan(void)
1.50 kristaps 519: {
520: FTS *f;
521: FTSENT *ff;
1.84 schwarze 522: struct mlink *mlink;
1.50 kristaps 523: int dform;
1.84 schwarze 524: char *fsec;
1.66 schwarze 525: const char *dsec, *arch, *cp, *path;
1.50 kristaps 526: const char *argv[2];
527:
528: argv[0] = ".";
529: argv[1] = (char *)NULL;
530:
531: /*
532: * Walk through all components under the directory, using the
533: * logical descent of files.
534: */
535: f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
536: if (NULL == f) {
1.59 schwarze 537: exitcode = (int)MANDOCLEVEL_SYSERR;
538: say("", NULL);
1.50 kristaps 539: return(0);
540: }
541:
542: dsec = arch = NULL;
543: dform = FORM_NONE;
544:
545: while (NULL != (ff = fts_read(f))) {
546: path = ff->fts_path + 2;
547: /*
1.83 schwarze 548: * If we're a regular file, add an mlink by using the
1.50 kristaps 549: * stored directory data and handling the filename.
550: */
551: if (FTS_F == ff->fts_info) {
1.60 schwarze 552: if (0 == strcmp(path, MANDOC_DB))
553: continue;
1.50 kristaps 554: if ( ! use_all && ff->fts_level < 2) {
1.56 schwarze 555: if (warnings)
1.59 schwarze 556: say(path, "Extraneous file");
1.50 kristaps 557: continue;
1.84 schwarze 558: } else if (NULL == (fsec =
1.60 schwarze 559: strrchr(ff->fts_name, '.'))) {
560: if ( ! use_all) {
1.56 schwarze 561: if (warnings)
1.60 schwarze 562: say(path,
563: "No filename suffix");
1.50 kristaps 564: continue;
565: }
1.84 schwarze 566: } else if (0 == strcmp(++fsec, "html")) {
1.60 schwarze 567: if (warnings)
568: say(path, "Skip html");
569: continue;
1.84 schwarze 570: } else if (0 == strcmp(fsec, "gz")) {
1.60 schwarze 571: if (warnings)
572: say(path, "Skip gz");
573: continue;
1.84 schwarze 574: } else if (0 == strcmp(fsec, "ps")) {
1.60 schwarze 575: if (warnings)
576: say(path, "Skip ps");
577: continue;
1.84 schwarze 578: } else if (0 == strcmp(fsec, "pdf")) {
1.60 schwarze 579: if (warnings)
580: say(path, "Skip pdf");
581: continue;
582: } else if ( ! use_all &&
1.84 schwarze 583: ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
584: (FORM_CAT == dform && strcmp(fsec, "0")))) {
1.60 schwarze 585: if (warnings)
586: say(path, "Wrong filename suffix");
587: continue;
1.66 schwarze 588: } else
1.84 schwarze 589: fsec[-1] = '\0';
590: mlink = mandoc_calloc(1, sizeof(struct mlink));
591: strlcpy(mlink->file, path, sizeof(mlink->file));
592: mlink->dform = dform;
593: if (NULL != dsec)
594: mlink->dsec = mandoc_strdup(dsec);
595: if (NULL != arch)
596: mlink->arch = mandoc_strdup(arch);
597: mlink->name = mandoc_strdup(ff->fts_name);
598: if (NULL != fsec)
599: mlink->fsec = mandoc_strdup(fsec);
600: mlink_add(mlink, ff->fts_statp);
1.50 kristaps 601: continue;
1.84 schwarze 602: } else if (FTS_D != ff->fts_info &&
1.60 schwarze 603: FTS_DP != ff->fts_info) {
604: if (warnings)
605: say(path, "Not a regular file");
1.50 kristaps 606: continue;
1.60 schwarze 607: }
1.1 kristaps 608:
1.50 kristaps 609: switch (ff->fts_level) {
610: case (0):
611: /* Ignore the root directory. */
612: break;
613: case (1):
614: /*
615: * This might contain manX/ or catX/.
616: * Try to infer this from the name.
617: * If we're not in use_all, enforce it.
618: */
619: dsec = NULL;
620: dform = FORM_NONE;
621: cp = ff->fts_name;
622: if (FTS_DP == ff->fts_info)
623: break;
1.1 kristaps 624:
1.50 kristaps 625: if (0 == strncmp(cp, "man", 3)) {
626: dform = FORM_SRC;
1.66 schwarze 627: dsec = cp + 3;
1.50 kristaps 628: } else if (0 == strncmp(cp, "cat", 3)) {
629: dform = FORM_CAT;
1.66 schwarze 630: dsec = cp + 3;
1.50 kristaps 631: }
1.1 kristaps 632:
1.50 kristaps 633: if (NULL != dsec || use_all)
634: break;
1.1 kristaps 635:
1.56 schwarze 636: if (warnings)
1.59 schwarze 637: say(path, "Unknown directory part");
1.50 kristaps 638: fts_set(f, ff, FTS_SKIP);
639: break;
640: case (2):
641: /*
642: * Possibly our architecture.
643: * If we're descending, keep tabs on it.
644: */
645: arch = NULL;
646: if (FTS_DP != ff->fts_info && NULL != dsec)
1.66 schwarze 647: arch = ff->fts_name;
1.50 kristaps 648: break;
649: default:
650: if (FTS_DP == ff->fts_info || use_all)
651: break;
1.56 schwarze 652: if (warnings)
1.59 schwarze 653: say(path, "Extraneous directory part");
1.50 kristaps 654: fts_set(f, ff, FTS_SKIP);
655: break;
1.5 kristaps 656: }
1.50 kristaps 657: }
658:
659: fts_close(f);
660: return(1);
661: }
1.5 kristaps 662:
1.50 kristaps 663: /*
1.89 schwarze 664: * Add a file to the mlinks table.
1.50 kristaps 665: * Do not verify that it's a "valid" looking manpage (we'll do that
666: * later).
667: *
668: * Try to infer the manual section, architecture, and page name from the
669: * path, assuming it looks like
670: *
671: * [./]man*[/<arch>]/<name>.<section>
672: * or
673: * [./]cat<section>[/<arch>]/<name>.0
674: *
675: * See treescan() for the fts(3) version of this.
676: */
677: static void
1.59 schwarze 678: filescan(const char *file)
1.50 kristaps 679: {
1.59 schwarze 680: char buf[PATH_MAX];
1.84 schwarze 681: struct stat st;
682: struct mlink *mlink;
1.59 schwarze 683: char *p, *start;
1.5 kristaps 684:
1.50 kristaps 685: assert(use_all);
1.5 kristaps 686:
1.50 kristaps 687: if (0 == strncmp(file, "./", 2))
688: file += 2;
1.5 kristaps 689:
1.59 schwarze 690: if (NULL == realpath(file, buf)) {
691: exitcode = (int)MANDOCLEVEL_BADARG;
692: say(file, NULL);
693: return;
1.72 schwarze 694: } else if (OP_TEST != op && strstr(buf, basedir) != buf) {
1.59 schwarze 695: exitcode = (int)MANDOCLEVEL_BADARG;
696: say("", "%s: outside base directory", buf);
697: return;
698: } else if (-1 == stat(buf, &st)) {
699: exitcode = (int)MANDOCLEVEL_BADARG;
700: say(file, NULL);
1.50 kristaps 701: return;
702: } else if ( ! (S_IFREG & st.st_mode)) {
1.59 schwarze 703: exitcode = (int)MANDOCLEVEL_BADARG;
704: say(file, "Not a regular file");
1.50 kristaps 705: return;
706: }
1.59 schwarze 707: start = buf + strlen(basedir);
1.84 schwarze 708: mlink = mandoc_calloc(1, sizeof(struct mlink));
709: strlcpy(mlink->file, start, sizeof(mlink->file));
1.17 schwarze 710:
1.50 kristaps 711: /*
712: * First try to guess our directory structure.
713: * If we find a separator, try to look for man* or cat*.
714: * If we find one of these and what's underneath is a directory,
715: * assume it's an architecture.
716: */
717: if (NULL != (p = strchr(start, '/'))) {
718: *p++ = '\0';
719: if (0 == strncmp(start, "man", 3)) {
1.84 schwarze 720: mlink->dform = FORM_SRC;
721: mlink->dsec = mandoc_strdup(start + 3);
1.50 kristaps 722: } else if (0 == strncmp(start, "cat", 3)) {
1.84 schwarze 723: mlink->dform = FORM_CAT;
724: mlink->dsec = mandoc_strdup(start + 3);
1.17 schwarze 725: }
1.5 kristaps 726:
1.50 kristaps 727: start = p;
1.84 schwarze 728: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
1.50 kristaps 729: *p++ = '\0';
1.84 schwarze 730: mlink->arch = mandoc_strdup(start);
1.50 kristaps 731: start = p;
1.84 schwarze 732: }
1.50 kristaps 733: }
734:
735: /*
736: * Now check the file suffix.
737: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
738: */
739: p = strrchr(start, '\0');
740: while (p-- > start && '/' != *p && '.' != *p)
741: /* Loop. */ ;
742:
743: if ('.' == *p) {
744: *p++ = '\0';
1.84 schwarze 745: mlink->fsec = mandoc_strdup(p);
1.5 kristaps 746: }
747:
1.10 kristaps 748: /*
1.50 kristaps 749: * Now try to parse the name.
750: * Use the filename portion of the path.
1.10 kristaps 751: */
1.84 schwarze 752: mlink->name = start;
1.50 kristaps 753: if (NULL != (p = strrchr(start, '/'))) {
1.84 schwarze 754: mlink->name = p + 1;
1.50 kristaps 755: *p = '\0';
1.84 schwarze 756: }
757: mlink->name = mandoc_strdup(mlink->name);
1.50 kristaps 758:
1.84 schwarze 759: mlink_add(mlink, &st);
1.50 kristaps 760: }
1.5 kristaps 761:
1.50 kristaps 762: static void
1.84 schwarze 763: mlink_add(struct mlink *mlink, const struct stat *st)
1.50 kristaps 764: {
1.83 schwarze 765: struct inodev inodev;
1.78 schwarze 766: struct mpage *mpage;
1.83 schwarze 767: unsigned int slot;
1.50 kristaps 768:
1.84 schwarze 769: assert(NULL != mlink->file);
1.50 kristaps 770:
1.84 schwarze 771: if (NULL == mlink->dsec)
772: mlink->dsec = mandoc_strdup("");
773: if (NULL == mlink->arch)
774: mlink->arch = mandoc_strdup("");
775: if (NULL == mlink->name)
776: mlink->name = mandoc_strdup("");
777: if (NULL == mlink->fsec)
778: mlink->fsec = mandoc_strdup("");
779:
780: if ('0' == *mlink->fsec) {
781: free(mlink->fsec);
782: mlink->fsec = mandoc_strdup(mlink->dsec);
783: mlink->fform = FORM_CAT;
784: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
785: mlink->fform = FORM_SRC;
1.74 schwarze 786: else
1.84 schwarze 787: mlink->fform = FORM_NONE;
1.82 schwarze 788:
1.83 schwarze 789: slot = ohash_qlookup(&mlinks, mlink->file);
790: assert(NULL == ohash_find(&mlinks, slot));
791: ohash_insert(&mlinks, slot, mlink);
792:
793: inodev.st_ino = st->st_ino;
794: inodev.st_dev = st->st_dev;
795: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
796: sizeof(struct inodev), inodev.st_ino);
797: mpage = ohash_find(&mpages, slot);
798: if (NULL == mpage) {
799: mpage = mandoc_calloc(1, sizeof(struct mpage));
800: mpage->inodev.st_ino = inodev.st_ino;
801: mpage->inodev.st_dev = inodev.st_dev;
802: ohash_insert(&mpages, slot, mpage);
803: } else
1.85 schwarze 804: mlink->next = mpage->mlinks;
1.82 schwarze 805: mpage->mlinks = mlink;
806: }
807:
808: static void
809: mlink_free(struct mlink *mlink)
810: {
811:
812: free(mlink->dsec);
813: free(mlink->arch);
814: free(mlink->name);
815: free(mlink->fsec);
816: free(mlink);
1.50 kristaps 817: }
1.3 kristaps 818:
1.50 kristaps 819: static void
1.78 schwarze 820: mpages_free(void)
1.50 kristaps 821: {
1.78 schwarze 822: struct mpage *mpage;
1.82 schwarze 823: struct mlink *mlink;
1.80 schwarze 824: unsigned int slot;
1.3 kristaps 825:
1.80 schwarze 826: mpage = ohash_first(&mpages, &slot);
827: while (NULL != mpage) {
1.82 schwarze 828: while (NULL != (mlink = mpage->mlinks)) {
1.85 schwarze 829: mpage->mlinks = mlink->next;
1.82 schwarze 830: mlink_free(mlink);
831: }
1.78 schwarze 832: free(mpage->sec);
833: free(mpage->arch);
1.82 schwarze 834: free(mpage->title);
835: free(mpage->desc);
1.78 schwarze 836: free(mpage);
1.80 schwarze 837: mpage = ohash_next(&mpages, &slot);
1.50 kristaps 838: }
839: }
1.38 schwarze 840:
1.50 kristaps 841: /*
1.89 schwarze 842: * For each mlink to the mpage, check whether the path looks like
843: * it is formatted, and if it does, check whether a source manual
844: * exists by the same name, ignoring the suffix.
845: * If both conditions hold, drop the mlink.
846: */
847: static void
848: mlinks_undupe(struct mpage *mpage)
849: {
850: char buf[PATH_MAX];
851: struct mlink **prev;
852: struct mlink *mlink;
853: char *bufp;
854:
855: mpage->form = FORM_CAT;
1.90 schwarze 856: prev = &mpage->mlinks;
857: while (NULL != (mlink = *prev)) {
1.89 schwarze 858: if (FORM_CAT != mlink->dform) {
859: mpage->form = FORM_NONE;
1.90 schwarze 860: goto nextlink;
1.89 schwarze 861: }
862: if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) {
863: if (warnings)
864: say(mlink->file, "Filename too long");
1.90 schwarze 865: goto nextlink;
1.89 schwarze 866: }
867: bufp = strstr(buf, "cat");
868: assert(NULL != bufp);
869: memcpy(bufp, "man", 3);
870: if (NULL != (bufp = strrchr(buf, '.')))
871: *++bufp = '\0';
872: strlcat(buf, mlink->dsec, PATH_MAX);
873: if (NULL == ohash_find(&mlinks,
874: ohash_qlookup(&mlinks, buf)))
1.90 schwarze 875: goto nextlink;
1.89 schwarze 876: if (warnings)
877: say(mlink->file, "Man source exists: %s", buf);
878: if (use_all)
1.90 schwarze 879: goto nextlink;
1.89 schwarze 880: *prev = mlink->next;
881: mlink_free(mlink);
1.90 schwarze 882: continue;
883: nextlink:
884: prev = &(*prev)->next;
1.89 schwarze 885: }
886: }
887:
888: /*
1.80 schwarze 889: * Run through the files in the global vector "mpages"
890: * and add them to the database specified in "basedir".
1.50 kristaps 891: *
892: * This handles the parsing scheme itself, using the cues of directory
893: * and filename to determine whether the file is parsable or not.
894: */
1.59 schwarze 895: static void
1.78 schwarze 896: mpages_merge(struct mchars *mc, struct mparse *mp, int check_reachable)
1.50 kristaps 897: {
1.69 schwarze 898: struct ohash title_table;
1.71 schwarze 899: struct ohash_info title_info, str_info;
1.78 schwarze 900: struct mpage *mpage;
1.69 schwarze 901: struct mdoc *mdoc;
902: struct man *man;
903: struct title *title_entry;
1.89 schwarze 904: char *title_str;
1.82 schwarze 905: const char *cp;
1.69 schwarze 906: int match;
1.80 schwarze 907: unsigned int pslot, tslot;
1.69 schwarze 908: enum mandoclevel lvl;
909:
1.71 schwarze 910: str_info.alloc = hash_alloc;
911: str_info.halloc = hash_halloc;
912: str_info.hfree = hash_free;
913: str_info.key_offset = offsetof(struct str, key);
914:
1.69 schwarze 915: if (check_reachable) {
916: title_info.alloc = hash_alloc;
917: title_info.halloc = hash_halloc;
918: title_info.hfree = hash_free;
919: title_info.key_offset = offsetof(struct title, title);
920: ohash_init(&title_table, 6, &title_info);
921: }
1.50 kristaps 922:
1.80 schwarze 923: mpage = ohash_first(&mpages, &pslot);
924: while (NULL != mpage) {
1.89 schwarze 925: mlinks_undupe(mpage);
926: if (NULL == mpage->mlinks) {
927: mpage = ohash_next(&mpages, &pslot);
928: continue;
1.50 kristaps 929: }
930:
1.71 schwarze 931: ohash_init(&strings, 6, &str_info);
1.50 kristaps 932: mparse_reset(mp);
933: mdoc = NULL;
934: man = NULL;
1.69 schwarze 935: match = 1;
1.14 schwarze 936:
937: /*
1.33 schwarze 938: * Try interpreting the file as mdoc(7) or man(7)
939: * source code, unless it is already known to be
940: * formatted. Fall back to formatted mode.
1.14 schwarze 941: */
1.82 schwarze 942: if (FORM_CAT != mpage->mlinks->dform ||
1.84 schwarze 943: FORM_CAT != mpage->mlinks->fform) {
1.82 schwarze 944: lvl = mparse_readfd(mp, -1, mpage->mlinks->file);
1.50 kristaps 945: if (lvl < MANDOCLEVEL_FATAL)
946: mparse_result(mp, &mdoc, &man);
1.76 schwarze 947: }
1.14 schwarze 948:
949: if (NULL != mdoc) {
1.82 schwarze 950: mpage->form = FORM_SRC;
951: mpage->sec =
952: mandoc_strdup(mdoc_meta(mdoc)->msec);
953: mpage->arch = mdoc_meta(mdoc)->arch;
1.84 schwarze 954: mpage->arch = mandoc_strdup(
955: NULL == mpage->arch ? "" : mpage->arch);
1.82 schwarze 956: mpage->title =
957: mandoc_strdup(mdoc_meta(mdoc)->title);
1.14 schwarze 958: } else if (NULL != man) {
1.82 schwarze 959: mpage->form = FORM_SRC;
960: mpage->sec =
961: mandoc_strdup(man_meta(man)->msec);
962: mpage->arch =
963: mandoc_strdup(mpage->mlinks->arch);
964: mpage->title =
965: mandoc_strdup(man_meta(man)->title);
1.76 schwarze 966: } else {
1.82 schwarze 967: mpage->form = FORM_CAT;
968: mpage->sec =
969: mandoc_strdup(mpage->mlinks->dsec);
970: mpage->arch =
971: mandoc_strdup(mpage->mlinks->arch);
972: mpage->title =
973: mandoc_strdup(mpage->mlinks->name);
1.76 schwarze 974: }
1.50 kristaps 975:
1.12 schwarze 976: /*
1.44 kristaps 977: * Check whether the manual section given in a file
978: * agrees with the directory where the file is located.
979: * Some manuals have suffixes like (3p) on their
980: * section number either inside the file or in the
981: * directory name, some are linked into more than one
982: * section, like encrypt(1) = makekey(8). Do not skip
983: * manuals for such reasons.
1.12 schwarze 984: */
1.82 schwarze 985: if (warnings && !use_all && FORM_SRC == mpage->form &&
986: strcasecmp(mpage->sec, mpage->mlinks->dsec)) {
1.69 schwarze 987: match = 0;
1.82 schwarze 988: say(mpage->mlinks->file, "Section \"%s\" "
1.78 schwarze 989: "manual in %s directory",
1.82 schwarze 990: mpage->sec, mpage->mlinks->dsec);
1.69 schwarze 991: }
1.12 schwarze 992:
1.42 schwarze 993: /*
994: * Manual page directories exist for each kernel
995: * architecture as returned by machine(1).
996: * However, many manuals only depend on the
997: * application architecture as returned by arch(1).
998: * For example, some (2/ARM) manuals are shared
999: * across the "armish" and "zaurus" kernel
1000: * architectures.
1001: * A few manuals are even shared across completely
1002: * different architectures, for example fdformat(1)
1003: * on amd64, i386, sparc, and sparc64.
1004: * Thus, warn about architecture mismatches,
1005: * but don't skip manuals for this reason.
1006: */
1.82 schwarze 1007: if (warnings && !use_all &&
1008: strcasecmp(mpage->arch, mpage->mlinks->arch)) {
1.69 schwarze 1009: match = 0;
1.82 schwarze 1010: say(mpage->mlinks->file, "Architecture \"%s\" "
1.47 schwarze 1011: "manual in \"%s\" directory",
1.82 schwarze 1012: mpage->arch, mpage->mlinks->arch);
1.69 schwarze 1013: }
1.82 schwarze 1014: if (warnings && !use_all &&
1015: strcasecmp(mpage->title, mpage->mlinks->name))
1.69 schwarze 1016: match = 0;
1.12 schwarze 1017:
1.82 schwarze 1018: putkey(mpage, mpage->mlinks->name, TYPE_Nm);
1.12 schwarze 1019:
1.50 kristaps 1020: if (NULL != mdoc) {
1021: if (NULL != (cp = mdoc_meta(mdoc)->name))
1.78 schwarze 1022: putkey(mpage, cp, TYPE_Nm);
1023: assert(NULL == mpage->desc);
1024: parse_mdoc(mpage, mdoc_node(mdoc));
1025: putkey(mpage, NULL != mpage->desc ?
1.82 schwarze 1026: mpage->desc : mpage->mlinks->name, TYPE_Nd);
1.50 kristaps 1027: } else if (NULL != man)
1.78 schwarze 1028: parse_man(mpage, man_node(man));
1.50 kristaps 1029: else
1.78 schwarze 1030: parse_cat(mpage);
1.44 kristaps 1031:
1.69 schwarze 1032: /*
1033: * Build a title string for the file. If it matches
1034: * the location of the file, remember the title as
1035: * found; else, remember it as missing.
1036: */
1037:
1038: if (check_reachable) {
1.82 schwarze 1039: if (-1 == asprintf(&title_str, "%s(%s%s%s)",
1040: mpage->title, mpage->sec,
1041: '\0' == *mpage->arch ? "" : "/",
1042: mpage->arch)) {
1.69 schwarze 1043: perror(NULL);
1044: exit((int)MANDOCLEVEL_SYSERR);
1045: }
1.80 schwarze 1046: tslot = ohash_qlookup(&title_table, title_str);
1047: title_entry = ohash_find(&title_table, tslot);
1.69 schwarze 1048: if (NULL == title_entry) {
1049: title_entry = mandoc_malloc(
1050: sizeof(struct title));
1051: title_entry->title = title_str;
1052: title_entry->file = mandoc_strdup(
1.82 schwarze 1053: match ? "" : mpage->mlinks->file);
1.80 schwarze 1054: ohash_insert(&title_table, tslot,
1.69 schwarze 1055: title_entry);
1056: } else {
1057: if (match)
1058: *title_entry->file = '\0';
1059: free(title_str);
1060: }
1061: }
1062:
1.82 schwarze 1063: dbindex(mpage, mc);
1.66 schwarze 1064: ohash_delete(&strings);
1.80 schwarze 1065: mpage = ohash_next(&mpages, &pslot);
1.69 schwarze 1066: }
1067:
1068: if (check_reachable) {
1.80 schwarze 1069: title_entry = ohash_first(&title_table, &tslot);
1.69 schwarze 1070: while (NULL != title_entry) {
1071: if ('\0' != *title_entry->file)
1072: say(title_entry->file,
1073: "Probably unreachable, title is %s",
1074: title_entry->title);
1075: free(title_entry->title);
1076: free(title_entry->file);
1077: free(title_entry);
1.80 schwarze 1078: title_entry = ohash_next(&title_table, &tslot);
1.69 schwarze 1079: }
1080: ohash_delete(&title_table);
1.50 kristaps 1081: }
1082: }
1.12 schwarze 1083:
1.50 kristaps 1084: static void
1.78 schwarze 1085: parse_cat(struct mpage *mpage)
1.50 kristaps 1086: {
1087: FILE *stream;
1088: char *line, *p, *title;
1089: size_t len, plen, titlesz;
1.12 schwarze 1090:
1.82 schwarze 1091: if (NULL == (stream = fopen(mpage->mlinks->file, "r"))) {
1.56 schwarze 1092: if (warnings)
1.82 schwarze 1093: say(mpage->mlinks->file, NULL);
1.50 kristaps 1094: return;
1095: }
1.1 kristaps 1096:
1.50 kristaps 1097: /* Skip to first blank line. */
1.1 kristaps 1098:
1.50 kristaps 1099: while (NULL != (line = fgetln(stream, &len)))
1100: if ('\n' == *line)
1101: break;
1.1 kristaps 1102:
1.50 kristaps 1103: /*
1104: * Assume the first line that is not indented
1105: * is the first section header. Skip to it.
1106: */
1.1 kristaps 1107:
1.50 kristaps 1108: while (NULL != (line = fgetln(stream, &len)))
1109: if ('\n' != *line && ' ' != *line)
1110: break;
1111:
1112: /*
1113: * Read up until the next section into a buffer.
1114: * Strip the leading and trailing newline from each read line,
1115: * appending a trailing space.
1116: * Ignore empty (whitespace-only) lines.
1117: */
1.1 kristaps 1118:
1.50 kristaps 1119: titlesz = 0;
1120: title = NULL;
1.38 schwarze 1121:
1.50 kristaps 1122: while (NULL != (line = fgetln(stream, &len))) {
1123: if (' ' != *line || '\n' != line[len - 1])
1124: break;
1125: while (len > 0 && isspace((unsigned char)*line)) {
1126: line++;
1127: len--;
1128: }
1129: if (1 == len)
1.38 schwarze 1130: continue;
1.50 kristaps 1131: title = mandoc_realloc(title, titlesz + len);
1132: memcpy(title + titlesz, line, len);
1133: titlesz += len;
1134: title[titlesz - 1] = ' ';
1135: }
1.38 schwarze 1136:
1.50 kristaps 1137: /*
1138: * If no page content can be found, or the input line
1139: * is already the next section header, or there is no
1140: * trailing newline, reuse the page title as the page
1141: * description.
1142: */
1.44 kristaps 1143:
1.50 kristaps 1144: if (NULL == title || '\0' == *title) {
1.56 schwarze 1145: if (warnings)
1.82 schwarze 1146: say(mpage->mlinks->file,
1147: "Cannot find NAME section");
1.78 schwarze 1148: assert(NULL == mpage->desc);
1.82 schwarze 1149: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1150: putkey(mpage, mpage->mlinks->name, TYPE_Nd);
1.50 kristaps 1151: fclose(stream);
1152: free(title);
1153: return;
1154: }
1.1 kristaps 1155:
1.50 kristaps 1156: title = mandoc_realloc(title, titlesz + 1);
1157: title[titlesz] = '\0';
1.33 schwarze 1158:
1.50 kristaps 1159: /*
1160: * Skip to the first dash.
1161: * Use the remaining line as the description (no more than 70
1162: * bytes).
1163: */
1.33 schwarze 1164:
1.50 kristaps 1165: if (NULL != (p = strstr(title, "- "))) {
1166: for (p += 2; ' ' == *p || '\b' == *p; p++)
1167: /* Skip to next word. */ ;
1168: } else {
1.56 schwarze 1169: if (warnings)
1.82 schwarze 1170: say(mpage->mlinks->file,
1171: "No dash in title line");
1.50 kristaps 1172: p = title;
1173: }
1.38 schwarze 1174:
1.50 kristaps 1175: plen = strlen(p);
1.1 kristaps 1176:
1.50 kristaps 1177: /* Strip backspace-encoding from line. */
1.1 kristaps 1178:
1.50 kristaps 1179: while (NULL != (line = memchr(p, '\b', plen))) {
1180: len = line - p;
1181: if (0 == len) {
1182: memmove(line, line + 1, plen--);
1183: continue;
1184: }
1185: memmove(line - 1, line + 1, plen - len);
1186: plen -= 2;
1187: }
1.1 kristaps 1188:
1.78 schwarze 1189: assert(NULL == mpage->desc);
1190: mpage->desc = mandoc_strdup(p);
1191: putkey(mpage, mpage->desc, TYPE_Nd);
1.50 kristaps 1192: fclose(stream);
1193: free(title);
1194: }
1.1 kristaps 1195:
1.50 kristaps 1196: /*
1197: * Put a type/word pair into the word database for this particular file.
1198: */
1199: static void
1.78 schwarze 1200: putkey(const struct mpage *mpage, const char *value, uint64_t type)
1.50 kristaps 1201: {
1.18 kristaps 1202:
1.50 kristaps 1203: assert(NULL != value);
1.78 schwarze 1204: putkeys(mpage, value, strlen(value), type);
1.3 kristaps 1205: }
1206:
1207: /*
1.50 kristaps 1208: * Grok all nodes at or below a certain mdoc node into putkey().
1.3 kristaps 1209: */
1210: static void
1.78 schwarze 1211: putmdockey(const struct mpage *mpage,
1212: const struct mdoc_node *n, uint64_t m)
1.3 kristaps 1213: {
1.18 kristaps 1214:
1.50 kristaps 1215: for ( ; NULL != n; n = n->next) {
1216: if (NULL != n->child)
1.78 schwarze 1217: putmdockey(mpage, n->child, m);
1.50 kristaps 1218: if (MDOC_TEXT == n->type)
1.78 schwarze 1219: putkey(mpage, n->string, m);
1.50 kristaps 1220: }
1221: }
1.18 kristaps 1222:
1.61 schwarze 1223: static void
1.78 schwarze 1224: parse_man(struct mpage *mpage, const struct man_node *n)
1.50 kristaps 1225: {
1226: const struct man_node *head, *body;
1227: char *start, *sv, *title;
1228: char byte;
1229: size_t sz, titlesz;
1.18 kristaps 1230:
1.50 kristaps 1231: if (NULL == n)
1.61 schwarze 1232: return;
1.18 kristaps 1233:
1.50 kristaps 1234: /*
1235: * We're only searching for one thing: the first text child in
1236: * the BODY of a NAME section. Since we don't keep track of
1237: * sections in -man, run some hoops to find out whether we're in
1238: * the correct section or not.
1239: */
1.18 kristaps 1240:
1.50 kristaps 1241: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1242: body = n;
1243: assert(body->parent);
1244: if (NULL != (head = body->parent->head) &&
1245: 1 == head->nchild &&
1246: NULL != (head = (head->child)) &&
1247: MAN_TEXT == head->type &&
1248: 0 == strcmp(head->string, "NAME") &&
1249: NULL != (body = body->child) &&
1250: MAN_TEXT == body->type) {
1.3 kristaps 1251:
1.50 kristaps 1252: title = NULL;
1253: titlesz = 0;
1.3 kristaps 1254:
1.50 kristaps 1255: /*
1256: * Suck the entire NAME section into memory.
1257: * Yes, we might run away.
1258: * But too many manuals have big, spread-out
1259: * NAME sections over many lines.
1260: */
1.3 kristaps 1261:
1.50 kristaps 1262: for ( ; NULL != body; body = body->next) {
1263: if (MAN_TEXT != body->type)
1264: break;
1265: if (0 == (sz = strlen(body->string)))
1266: continue;
1267: title = mandoc_realloc
1268: (title, titlesz + sz + 1);
1269: memcpy(title + titlesz, body->string, sz);
1270: titlesz += sz + 1;
1271: title[titlesz - 1] = ' ';
1272: }
1273: if (NULL == title)
1.61 schwarze 1274: return;
1.18 kristaps 1275:
1.50 kristaps 1276: title = mandoc_realloc(title, titlesz + 1);
1277: title[titlesz] = '\0';
1.18 kristaps 1278:
1.50 kristaps 1279: /* Skip leading space. */
1.18 kristaps 1280:
1.50 kristaps 1281: sv = title;
1282: while (isspace((unsigned char)*sv))
1283: sv++;
1.18 kristaps 1284:
1.50 kristaps 1285: if (0 == (sz = strlen(sv))) {
1286: free(title);
1.61 schwarze 1287: return;
1.50 kristaps 1288: }
1.1 kristaps 1289:
1.50 kristaps 1290: /* Erase trailing space. */
1.1 kristaps 1291:
1.50 kristaps 1292: start = &sv[sz - 1];
1293: while (start > sv && isspace((unsigned char)*start))
1294: *start-- = '\0';
1.1 kristaps 1295:
1.50 kristaps 1296: if (start == sv) {
1297: free(title);
1.61 schwarze 1298: return;
1.50 kristaps 1299: }
1.1 kristaps 1300:
1.50 kristaps 1301: start = sv;
1.18 kristaps 1302:
1.50 kristaps 1303: /*
1304: * Go through a special heuristic dance here.
1305: * Conventionally, one or more manual names are
1306: * comma-specified prior to a whitespace, then a
1307: * dash, then a description. Try to puzzle out
1308: * the name parts here.
1309: */
1.18 kristaps 1310:
1.50 kristaps 1311: for ( ;; ) {
1312: sz = strcspn(start, " ,");
1313: if ('\0' == start[sz])
1314: break;
1.1 kristaps 1315:
1.50 kristaps 1316: byte = start[sz];
1317: start[sz] = '\0';
1.1 kristaps 1318:
1.78 schwarze 1319: putkey(mpage, start, TYPE_Nm);
1.1 kristaps 1320:
1.50 kristaps 1321: if (' ' == byte) {
1322: start += sz + 1;
1323: break;
1324: }
1.1 kristaps 1325:
1.50 kristaps 1326: assert(',' == byte);
1327: start += sz + 1;
1328: while (' ' == *start)
1329: start++;
1330: }
1.1 kristaps 1331:
1.50 kristaps 1332: if (sv == start) {
1.78 schwarze 1333: putkey(mpage, start, TYPE_Nm);
1.50 kristaps 1334: free(title);
1.61 schwarze 1335: return;
1.50 kristaps 1336: }
1.1 kristaps 1337:
1.50 kristaps 1338: while (isspace((unsigned char)*start))
1339: start++;
1.1 kristaps 1340:
1.50 kristaps 1341: if (0 == strncmp(start, "-", 1))
1342: start += 1;
1343: else if (0 == strncmp(start, "\\-\\-", 4))
1344: start += 4;
1345: else if (0 == strncmp(start, "\\-", 2))
1346: start += 2;
1347: else if (0 == strncmp(start, "\\(en", 4))
1348: start += 4;
1349: else if (0 == strncmp(start, "\\(em", 4))
1350: start += 4;
1.1 kristaps 1351:
1.50 kristaps 1352: while (' ' == *start)
1353: start++;
1.1 kristaps 1354:
1.78 schwarze 1355: assert(NULL == mpage->desc);
1356: mpage->desc = mandoc_strdup(start);
1357: putkey(mpage, mpage->desc, TYPE_Nd);
1.50 kristaps 1358: free(title);
1.61 schwarze 1359: return;
1.50 kristaps 1360: }
1361: }
1.1 kristaps 1362:
1.77 schwarze 1363: for (n = n->child; n; n = n->next) {
1.78 schwarze 1364: if (NULL != mpage->desc)
1.77 schwarze 1365: break;
1.78 schwarze 1366: parse_man(mpage, n);
1.77 schwarze 1367: }
1.1 kristaps 1368: }
1369:
1370: static void
1.78 schwarze 1371: parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1372: {
1373:
1.50 kristaps 1374: assert(NULL != n);
1375: for (n = n->child; NULL != n; n = n->next) {
1376: switch (n->type) {
1377: case (MDOC_ELEM):
1378: /* FALLTHROUGH */
1379: case (MDOC_BLOCK):
1380: /* FALLTHROUGH */
1381: case (MDOC_HEAD):
1382: /* FALLTHROUGH */
1383: case (MDOC_BODY):
1384: /* FALLTHROUGH */
1385: case (MDOC_TAIL):
1386: if (NULL != mdocs[n->tok].fp)
1.78 schwarze 1387: if (0 == (*mdocs[n->tok].fp)(mpage, n))
1.50 kristaps 1388: break;
1.68 schwarze 1389: if (mdocs[n->tok].mask)
1.78 schwarze 1390: putmdockey(mpage, n->child,
1391: mdocs[n->tok].mask);
1.50 kristaps 1392: break;
1393: default:
1394: assert(MDOC_ROOT != n->type);
1395: continue;
1396: }
1397: if (NULL != n->child)
1.78 schwarze 1398: parse_mdoc(mpage, n);
1.1 kristaps 1399: }
1400: }
1401:
1.25 schwarze 1402: static int
1.78 schwarze 1403: parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1404: {
1405: const char *start, *end;
1406: size_t sz;
1.25 schwarze 1407:
1.50 kristaps 1408: if (SEC_SYNOPSIS != n->sec ||
1409: NULL == (n = n->child) ||
1410: MDOC_TEXT != n->type)
1.25 schwarze 1411: return(0);
1.1 kristaps 1412:
1413: /*
1414: * Only consider those `Fd' macro fields that begin with an
1415: * "inclusion" token (versus, e.g., #define).
1416: */
1.50 kristaps 1417:
1.1 kristaps 1418: if (strcmp("#include", n->string))
1.25 schwarze 1419: return(0);
1.1 kristaps 1420:
1421: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.25 schwarze 1422: return(0);
1.1 kristaps 1423:
1424: /*
1425: * Strip away the enclosing angle brackets and make sure we're
1426: * not zero-length.
1427: */
1428:
1429: start = n->string;
1430: if ('<' == *start || '"' == *start)
1431: start++;
1432:
1433: if (0 == (sz = strlen(start)))
1.25 schwarze 1434: return(0);
1.1 kristaps 1435:
1436: end = &start[(int)sz - 1];
1437: if ('>' == *end || '"' == *end)
1438: end--;
1439:
1.50 kristaps 1440: if (end > start)
1.78 schwarze 1441: putkeys(mpage, start, end - start + 1, TYPE_In);
1.92 ! schwarze 1442: return(0);
1.1 kristaps 1443: }
1444:
1.25 schwarze 1445: static int
1.78 schwarze 1446: parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1447: {
1448: const char *cp;
1449:
1.50 kristaps 1450: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.25 schwarze 1451: return(0);
1452:
1.50 kristaps 1453: /*
1454: * Parse: .Fn "struct type *name" "char *arg".
1455: * First strip away pointer symbol.
1456: * Then store the function name, then type.
1457: * Finally, store the arguments.
1458: */
1.1 kristaps 1459:
1.50 kristaps 1460: if (NULL == (cp = strrchr(n->string, ' ')))
1461: cp = n->string;
1.1 kristaps 1462:
1463: while ('*' == *cp)
1464: cp++;
1465:
1.78 schwarze 1466: putkey(mpage, cp, TYPE_Fn);
1.25 schwarze 1467:
1.50 kristaps 1468: if (n->string < cp)
1.78 schwarze 1469: putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1.25 schwarze 1470:
1.50 kristaps 1471: for (n = n->next; NULL != n; n = n->next)
1472: if (MDOC_TEXT == n->type)
1.78 schwarze 1473: putkey(mpage, n->string, TYPE_Fa);
1.25 schwarze 1474:
1475: return(0);
1.1 kristaps 1476: }
1477:
1.25 schwarze 1478: static int
1.78 schwarze 1479: parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1480: {
1.67 schwarze 1481: char *cp;
1.1 kristaps 1482:
1483: if (NULL == (n = n->child))
1.25 schwarze 1484: return(0);
1.1 kristaps 1485:
1.67 schwarze 1486: if (NULL == n->next) {
1.78 schwarze 1487: putkey(mpage, n->string, TYPE_Xr);
1.67 schwarze 1488: return(0);
1489: }
1490:
1491: if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) {
1492: perror(NULL);
1493: exit((int)MANDOCLEVEL_SYSERR);
1494: }
1.78 schwarze 1495: putkey(mpage, cp, TYPE_Xr);
1.67 schwarze 1496: free(cp);
1497: return(0);
1.1 kristaps 1498: }
1499:
1.25 schwarze 1500: static int
1.78 schwarze 1501: parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1502: {
1.50 kristaps 1503: size_t sz;
1.1 kristaps 1504:
1505: if (MDOC_BODY != n->type)
1.25 schwarze 1506: return(0);
1.1 kristaps 1507:
1.50 kristaps 1508: /*
1509: * Special-case the `Nd' because we need to put the description
1510: * into the document table.
1511: */
1512:
1513: for (n = n->child; NULL != n; n = n->next) {
1514: if (MDOC_TEXT == n->type) {
1.78 schwarze 1515: if (NULL != mpage->desc) {
1516: sz = strlen(mpage->desc) +
1.66 schwarze 1517: strlen(n->string) + 2;
1.78 schwarze 1518: mpage->desc = mandoc_realloc(
1519: mpage->desc, sz);
1520: strlcat(mpage->desc, " ", sz);
1521: strlcat(mpage->desc, n->string, sz);
1.66 schwarze 1522: } else
1.78 schwarze 1523: mpage->desc = mandoc_strdup(n->string);
1.50 kristaps 1524: }
1525: if (NULL != n->child)
1.78 schwarze 1526: parse_mdoc_Nd(mpage, n);
1.50 kristaps 1527: }
1.25 schwarze 1528: return(1);
1.1 kristaps 1529: }
1530:
1.25 schwarze 1531: static int
1.78 schwarze 1532: parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1533: {
1534:
1.92 ! schwarze 1535: return(SEC_NAME == n->sec ||
! 1536: (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type));
1.1 kristaps 1537: }
1538:
1.25 schwarze 1539: static int
1.78 schwarze 1540: parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1541: {
1542:
1.25 schwarze 1543: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 kristaps 1544: }
1545:
1.50 kristaps 1546: static int
1.78 schwarze 1547: parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1548: {
1549:
1.50 kristaps 1550: return(MDOC_HEAD == n->type);
1551: }
1.1 kristaps 1552:
1.50 kristaps 1553: static int
1.78 schwarze 1554: parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1.50 kristaps 1555: {
1.1 kristaps 1556:
1.50 kristaps 1557: return(MDOC_BODY == n->type);
1.1 kristaps 1558: }
1559:
1.50 kristaps 1560: /*
1.66 schwarze 1561: * Add a string to the hash table for the current manual.
1562: * Each string has a bitmask telling which macros it belongs to.
1563: * When we finish the manual, we'll dump the table.
1.50 kristaps 1564: */
1565: static void
1.78 schwarze 1566: putkeys(const struct mpage *mpage,
1567: const char *cp, size_t sz, uint64_t v)
1.50 kristaps 1568: {
1569: struct str *s;
1.65 schwarze 1570: unsigned int slot;
1.50 kristaps 1571: const char *end;
1.25 schwarze 1572:
1.50 kristaps 1573: if (0 == sz)
1574: return;
1.25 schwarze 1575:
1.65 schwarze 1576: end = cp + sz;
1577: slot = ohash_qlookupi(&strings, cp, &end);
1578: s = ohash_find(&strings, slot);
1.25 schwarze 1579:
1.78 schwarze 1580: if (NULL != s && mpage == s->mpage) {
1.50 kristaps 1581: s->mask |= v;
1582: return;
1583: } else if (NULL == s) {
1.51 kristaps 1584: s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1.50 kristaps 1585: memcpy(s->key, cp, sz);
1.65 schwarze 1586: ohash_insert(&strings, slot, s);
1.1 kristaps 1587: }
1.78 schwarze 1588: s->mpage = mpage;
1.50 kristaps 1589: s->mask = v;
1.1 kristaps 1590: }
1591:
1.50 kristaps 1592: /*
1593: * Take a Unicode codepoint and produce its UTF-8 encoding.
1594: * This isn't the best way to do this, but it works.
1595: * The magic numbers are from the UTF-8 packaging.
1596: * They're not as scary as they seem: read the UTF-8 spec for details.
1597: */
1598: static size_t
1599: utf8(unsigned int cp, char out[7])
1.1 kristaps 1600: {
1.50 kristaps 1601: size_t rc;
1.1 kristaps 1602:
1.50 kristaps 1603: rc = 0;
1604: if (cp <= 0x0000007F) {
1605: rc = 1;
1606: out[0] = (char)cp;
1607: } else if (cp <= 0x000007FF) {
1608: rc = 2;
1609: out[0] = (cp >> 6 & 31) | 192;
1610: out[1] = (cp & 63) | 128;
1611: } else if (cp <= 0x0000FFFF) {
1612: rc = 3;
1613: out[0] = (cp >> 12 & 15) | 224;
1614: out[1] = (cp >> 6 & 63) | 128;
1615: out[2] = (cp & 63) | 128;
1616: } else if (cp <= 0x001FFFFF) {
1617: rc = 4;
1618: out[0] = (cp >> 18 & 7) | 240;
1619: out[1] = (cp >> 12 & 63) | 128;
1620: out[2] = (cp >> 6 & 63) | 128;
1621: out[3] = (cp & 63) | 128;
1622: } else if (cp <= 0x03FFFFFF) {
1623: rc = 5;
1624: out[0] = (cp >> 24 & 3) | 248;
1625: out[1] = (cp >> 18 & 63) | 128;
1626: out[2] = (cp >> 12 & 63) | 128;
1627: out[3] = (cp >> 6 & 63) | 128;
1628: out[4] = (cp & 63) | 128;
1629: } else if (cp <= 0x7FFFFFFF) {
1630: rc = 6;
1631: out[0] = (cp >> 30 & 1) | 252;
1632: out[1] = (cp >> 24 & 63) | 128;
1633: out[2] = (cp >> 18 & 63) | 128;
1634: out[3] = (cp >> 12 & 63) | 128;
1635: out[4] = (cp >> 6 & 63) | 128;
1636: out[5] = (cp & 63) | 128;
1637: } else
1.1 kristaps 1638: return(0);
1639:
1.50 kristaps 1640: out[rc] = '\0';
1641: return(rc);
1642: }
1.1 kristaps 1643:
1.50 kristaps 1644: /*
1645: * Store the UTF-8 version of a key, or alias the pointer if the key has
1646: * no UTF-8 transcription marks in it.
1647: */
1648: static void
1649: utf8key(struct mchars *mc, struct str *key)
1650: {
1651: size_t sz, bsz, pos;
1652: char utfbuf[7], res[5];
1653: char *buf;
1654: const char *seq, *cpp, *val;
1655: int len, u;
1656: enum mandoc_esc esc;
1657:
1658: assert(NULL == key->utf8);
1659:
1660: res[0] = '\\';
1661: res[1] = '\t';
1662: res[2] = ASCII_NBRSP;
1663: res[3] = ASCII_HYPH;
1664: res[4] = '\0';
1.1 kristaps 1665:
1.50 kristaps 1666: val = key->key;
1667: bsz = strlen(val);
1.46 kristaps 1668:
1.50 kristaps 1669: /*
1670: * Pre-check: if we have no stop-characters, then set the
1671: * pointer as ourselvse and get out of here.
1672: */
1673: if (strcspn(val, res) == bsz) {
1674: key->utf8 = key->key;
1675: return;
1676: }
1.46 kristaps 1677:
1.50 kristaps 1678: /* Pre-allocate by the length of the input */
1.46 kristaps 1679:
1.50 kristaps 1680: buf = mandoc_malloc(++bsz);
1681: pos = 0;
1.46 kristaps 1682:
1.50 kristaps 1683: while ('\0' != *val) {
1684: /*
1685: * Halt on the first escape sequence.
1686: * This also halts on the end of string, in which case
1687: * we just copy, fallthrough, and exit the loop.
1688: */
1689: if ((sz = strcspn(val, res)) > 0) {
1690: memcpy(&buf[pos], val, sz);
1691: pos += sz;
1692: val += sz;
1693: }
1.46 kristaps 1694:
1.50 kristaps 1695: if (ASCII_HYPH == *val) {
1696: buf[pos++] = '-';
1697: val++;
1698: continue;
1699: } else if ('\t' == *val || ASCII_NBRSP == *val) {
1700: buf[pos++] = ' ';
1701: val++;
1702: continue;
1703: } else if ('\\' != *val)
1704: break;
1.46 kristaps 1705:
1.50 kristaps 1706: /* Read past the slash. */
1.46 kristaps 1707:
1.50 kristaps 1708: val++;
1.46 kristaps 1709:
1.50 kristaps 1710: /*
1711: * Parse the escape sequence and see if it's a
1712: * predefined character or special character.
1713: */
1714: esc = mandoc_escape
1715: ((const char **)&val, &seq, &len);
1716: if (ESCAPE_ERROR == esc)
1717: break;
1.1 kristaps 1718:
1.50 kristaps 1719: if (ESCAPE_SPECIAL != esc)
1720: continue;
1721: if (0 == (u = mchars_spec2cp(mc, seq, len)))
1722: continue;
1.1 kristaps 1723:
1.50 kristaps 1724: /*
1725: * If we have a Unicode codepoint, try to convert that
1726: * to a UTF-8 byte string.
1727: */
1728: cpp = utfbuf;
1729: if (0 == (sz = utf8(u, utfbuf)))
1730: continue;
1.1 kristaps 1731:
1.50 kristaps 1732: /* Copy the rendered glyph into the stream. */
1.1 kristaps 1733:
1.50 kristaps 1734: sz = strlen(cpp);
1735: bsz += sz;
1.1 kristaps 1736:
1.50 kristaps 1737: buf = mandoc_realloc(buf, bsz);
1.1 kristaps 1738:
1.50 kristaps 1739: memcpy(&buf[pos], cpp, sz);
1740: pos += sz;
1.1 kristaps 1741: }
1742:
1.50 kristaps 1743: buf[pos] = '\0';
1744: key->utf8 = buf;
1.1 kristaps 1745: }
1746:
1.14 schwarze 1747: /*
1.50 kristaps 1748: * Flush the current page's terms (and their bits) into the database.
1749: * Wrap the entire set of additions in a transaction to make sqlite be a
1750: * little faster.
1751: * Also, UTF-8-encode the description at the last possible moment.
1.14 schwarze 1752: */
1753: static void
1.82 schwarze 1754: dbindex(const struct mpage *mpage, struct mchars *mc)
1.14 schwarze 1755: {
1.87 schwarze 1756: struct mlink *mlink;
1.50 kristaps 1757: struct str *key;
1758: const char *desc;
1759: int64_t recno;
1.52 kristaps 1760: size_t i;
1.66 schwarze 1761: unsigned int slot;
1.14 schwarze 1762:
1.56 schwarze 1763: if (verb)
1.82 schwarze 1764: say(mpage->mlinks->file, "Adding to index");
1.43 kristaps 1765:
1.50 kristaps 1766: if (nodb)
1.14 schwarze 1767: return;
1.28 kristaps 1768:
1.50 kristaps 1769: desc = "";
1.78 schwarze 1770: if (NULL != mpage->desc && '\0' != *mpage->desc) {
1.65 schwarze 1771: key = ohash_find(&strings,
1.78 schwarze 1772: ohash_qlookup(&strings, mpage->desc));
1.50 kristaps 1773: assert(NULL != key);
1774: if (NULL == key->utf8)
1775: utf8key(mc, key);
1776: desc = key->utf8;
1777: }
1778:
1.52 kristaps 1779: SQL_EXEC("BEGIN TRANSACTION");
1.50 kristaps 1780:
1.52 kristaps 1781: i = 1;
1.81 schwarze 1782: /*
1783: * XXX The following three lines are obsolete
1784: * and only kept for backward compatibility
1785: * until apropos(1) and friends have caught up.
1786: */
1.82 schwarze 1787: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->file);
1788: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->dsec);
1789: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->arch);
1.81 schwarze 1790: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, desc);
1.82 schwarze 1791: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
1.81 schwarze 1792: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.50 kristaps 1793: recno = sqlite3_last_insert_rowid(db);
1.81 schwarze 1794: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1795:
1.87 schwarze 1796: for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1797: i = 1;
1.88 schwarze 1798: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->file);
1.87 schwarze 1799: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1800: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1.88 schwarze 1801: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.87 schwarze 1802: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, recno);
1803: SQL_STEP(stmts[STMT_INSERT_LINK]);
1804: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1805: }
1.50 kristaps 1806:
1.66 schwarze 1807: for (key = ohash_first(&strings, &slot); NULL != key;
1808: key = ohash_next(&strings, &slot)) {
1.78 schwarze 1809: assert(key->mpage == mpage);
1.50 kristaps 1810: if (NULL == key->utf8)
1811: utf8key(mc, key);
1.52 kristaps 1812: i = 1;
1813: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1814: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->utf8);
1815: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
1816: SQL_STEP(stmts[STMT_INSERT_KEY]);
1.50 kristaps 1817: sqlite3_reset(stmts[STMT_INSERT_KEY]);
1.66 schwarze 1818: if (key->utf8 != key->key)
1819: free(key->utf8);
1820: free(key);
1.38 schwarze 1821: }
1.28 kristaps 1822:
1.52 kristaps 1823: SQL_EXEC("END TRANSACTION");
1.14 schwarze 1824: }
1825:
1.5 kristaps 1826: static void
1.59 schwarze 1827: dbprune(void)
1.5 kristaps 1828: {
1.78 schwarze 1829: struct mpage *mpage;
1.82 schwarze 1830: struct mlink *mlink;
1.52 kristaps 1831: size_t i;
1.80 schwarze 1832: unsigned int slot;
1.5 kristaps 1833:
1.50 kristaps 1834: if (nodb)
1835: return;
1.12 schwarze 1836:
1.80 schwarze 1837: mpage = ohash_first(&mpages, &slot);
1838: while (NULL != mpage) {
1.82 schwarze 1839: mlink = mpage->mlinks;
1.52 kristaps 1840: i = 1;
1.82 schwarze 1841: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE], i, mlink->file);
1.81 schwarze 1842: SQL_STEP(stmts[STMT_DELETE_PAGE]);
1843: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
1.56 schwarze 1844: if (verb)
1.82 schwarze 1845: say(mlink->file, "Deleted from index");
1.80 schwarze 1846: mpage = ohash_next(&mpages, &slot);
1.5 kristaps 1847: }
1848: }
1849:
1.4 kristaps 1850: /*
1.50 kristaps 1851: * Close an existing database and its prepared statements.
1852: * If "real" is not set, rename the temporary file into the real one.
1.4 kristaps 1853: */
1.35 kristaps 1854: static void
1.59 schwarze 1855: dbclose(int real)
1.4 kristaps 1856: {
1.50 kristaps 1857: size_t i;
1.4 kristaps 1858:
1.50 kristaps 1859: if (nodb)
1.38 schwarze 1860: return;
1.50 kristaps 1861:
1862: for (i = 0; i < STMT__MAX; i++) {
1863: sqlite3_finalize(stmts[i]);
1864: stmts[i] = NULL;
1.4 kristaps 1865: }
1866:
1.50 kristaps 1867: sqlite3_close(db);
1868: db = NULL;
1.12 schwarze 1869:
1.50 kristaps 1870: if (real)
1871: return;
1.12 schwarze 1872:
1.63 schwarze 1873: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
1.59 schwarze 1874: exitcode = (int)MANDOCLEVEL_SYSERR;
1875: say(MANDOC_DB, NULL);
1876: }
1.50 kristaps 1877: }
1.14 schwarze 1878:
1.50 kristaps 1879: /*
1880: * This is straightforward stuff.
1881: * Open a database connection to a "temporary" database, then open a set
1882: * of prepared statements we'll use over and over again.
1883: * If "real" is set, we use the existing database; if not, we truncate a
1884: * temporary one.
1885: * Must be matched by dbclose().
1886: */
1887: static int
1.59 schwarze 1888: dbopen(int real)
1.50 kristaps 1889: {
1.63 schwarze 1890: const char *file, *sql;
1.50 kristaps 1891: int rc, ofl;
1.12 schwarze 1892:
1.50 kristaps 1893: if (nodb)
1894: return(1);
1.12 schwarze 1895:
1.63 schwarze 1896: ofl = SQLITE_OPEN_READWRITE;
1897: if (0 == real) {
1898: file = MANDOC_DB "~";
1899: if (-1 == remove(file) && ENOENT != errno) {
1900: exitcode = (int)MANDOCLEVEL_SYSERR;
1901: say(file, NULL);
1902: return(0);
1903: }
1904: ofl |= SQLITE_OPEN_EXCLUSIVE;
1905: } else
1906: file = MANDOC_DB;
1.45 kristaps 1907:
1.50 kristaps 1908: rc = sqlite3_open_v2(file, &db, ofl, NULL);
1909: if (SQLITE_OK == rc)
1.57 schwarze 1910: goto prepare_statements;
1.50 kristaps 1911: if (SQLITE_CANTOPEN != rc) {
1.59 schwarze 1912: exitcode = (int)MANDOCLEVEL_SYSERR;
1913: say(file, NULL);
1.50 kristaps 1914: return(0);
1915: }
1.12 schwarze 1916:
1.50 kristaps 1917: sqlite3_close(db);
1918: db = NULL;
1.12 schwarze 1919:
1.50 kristaps 1920: if (SQLITE_OK != (rc = sqlite3_open(file, &db))) {
1.59 schwarze 1921: exitcode = (int)MANDOCLEVEL_SYSERR;
1922: say(file, NULL);
1.50 kristaps 1923: return(0);
1924: }
1.12 schwarze 1925:
1.81 schwarze 1926: /*
1927: * XXX The first three columns in table mpages are obsolete
1928: * and only kept for backward compatibility
1929: * until apropos(1) and friends have caught up.
1930: */
1931: sql = "CREATE TABLE \"mpages\" (\n"
1.50 kristaps 1932: " \"file\" TEXT NOT NULL,\n"
1933: " \"sec\" TEXT NOT NULL,\n"
1934: " \"arch\" TEXT NOT NULL,\n"
1935: " \"desc\" TEXT NOT NULL,\n"
1936: " \"form\" INTEGER NOT NULL,\n"
1937: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1938: ");\n"
1939: "\n"
1.81 schwarze 1940: "CREATE TABLE \"mlinks\" (\n"
1.88 schwarze 1941: " \"file\" TEXT NOT NULL,\n"
1.81 schwarze 1942: " \"sec\" TEXT NOT NULL,\n"
1943: " \"arch\" TEXT NOT NULL,\n"
1944: " \"name\" TEXT NOT NULL,\n"
1945: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1946: "ON DELETE CASCADE,\n"
1947: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1948: ");\n"
1949: "\n"
1.50 kristaps 1950: "CREATE TABLE \"keys\" (\n"
1951: " \"bits\" INTEGER NOT NULL,\n"
1952: " \"key\" TEXT NOT NULL,\n"
1.81 schwarze 1953: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1954: "ON DELETE CASCADE,\n"
1.50 kristaps 1955: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1956: ");\n"
1957: "\n"
1958: "CREATE INDEX \"key_index\" ON keys (key);\n";
1.14 schwarze 1959:
1.50 kristaps 1960: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1.59 schwarze 1961: exitcode = (int)MANDOCLEVEL_SYSERR;
1962: say(file, "%s", sqlite3_errmsg(db));
1.50 kristaps 1963: return(0);
1964: }
1.4 kristaps 1965:
1.57 schwarze 1966: prepare_statements:
1967: SQL_EXEC("PRAGMA foreign_keys = ON");
1.81 schwarze 1968: sql = "DELETE FROM mpages where file=?";
1969: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
1970: sql = "INSERT INTO mpages "
1.50 kristaps 1971: "(file,sec,arch,desc,form) VALUES (?,?,?,?,?)";
1.81 schwarze 1972: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
1973: sql = "INSERT INTO mlinks "
1.88 schwarze 1974: "(file,sec,arch,name,pageid) VALUES (?,?,?,?,?)";
1.81 schwarze 1975: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.50 kristaps 1976: sql = "INSERT INTO keys "
1.81 schwarze 1977: "(bits,key,pageid) VALUES (?,?,?)";
1.50 kristaps 1978: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.70 schwarze 1979:
1980: #ifndef __APPLE__
1981: /*
1982: * When opening a new database, we can turn off
1983: * synchronous mode for much better performance.
1984: */
1985:
1986: if (real)
1987: SQL_EXEC("PRAGMA synchronous = OFF");
1988: #endif
1989:
1.50 kristaps 1990: return(1);
1991: }
1.5 kristaps 1992:
1.50 kristaps 1993: static void *
1994: hash_halloc(size_t sz, void *arg)
1995: {
1.12 schwarze 1996:
1.50 kristaps 1997: return(mandoc_calloc(sz, 1));
1998: }
1.12 schwarze 1999:
1.50 kristaps 2000: static void *
2001: hash_alloc(size_t sz, void *arg)
2002: {
1.5 kristaps 2003:
1.50 kristaps 2004: return(mandoc_malloc(sz));
2005: }
1.41 kristaps 2006:
1.50 kristaps 2007: static void
2008: hash_free(void *p, size_t sz, void *arg)
2009: {
1.4 kristaps 2010:
1.50 kristaps 2011: free(p);
1.4 kristaps 2012: }
2013:
1.50 kristaps 2014: static int
1.59 schwarze 2015: set_basedir(const char *targetdir)
1.4 kristaps 2016: {
1.59 schwarze 2017: static char startdir[PATH_MAX];
2018: static int fd;
1.4 kristaps 2019:
1.59 schwarze 2020: /*
2021: * Remember where we started by keeping a fd open to the origin
2022: * path component: throughout this utility, we chdir() a lot to
2023: * handle relative paths, and by doing this, we can return to
2024: * the starting point.
2025: */
2026: if ('\0' == *startdir) {
2027: if (NULL == getcwd(startdir, PATH_MAX)) {
2028: exitcode = (int)MANDOCLEVEL_SYSERR;
2029: if (NULL != targetdir)
2030: say(".", NULL);
2031: return(0);
2032: }
2033: if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
2034: exitcode = (int)MANDOCLEVEL_SYSERR;
2035: say(startdir, NULL);
2036: return(0);
2037: }
2038: if (NULL == targetdir)
2039: targetdir = startdir;
2040: } else {
2041: if (-1 == fd)
2042: return(0);
2043: if (-1 == fchdir(fd)) {
2044: close(fd);
2045: basedir[0] = '\0';
2046: exitcode = (int)MANDOCLEVEL_SYSERR;
2047: say(startdir, NULL);
2048: return(0);
2049: }
2050: if (NULL == targetdir) {
2051: close(fd);
2052: return(1);
2053: }
2054: }
2055: if (NULL == realpath(targetdir, basedir)) {
2056: basedir[0] = '\0';
2057: exitcode = (int)MANDOCLEVEL_BADARG;
2058: say(targetdir, NULL);
1.50 kristaps 2059: return(0);
1.59 schwarze 2060: } else if (-1 == chdir(basedir)) {
2061: exitcode = (int)MANDOCLEVEL_BADARG;
2062: say("", NULL);
1.50 kristaps 2063: return(0);
1.4 kristaps 2064: }
1.50 kristaps 2065: return(1);
1.56 schwarze 2066: }
2067:
2068: static void
1.59 schwarze 2069: say(const char *file, const char *format, ...)
1.56 schwarze 2070: {
2071: va_list ap;
2072:
1.59 schwarze 2073: if ('\0' != *basedir)
2074: fprintf(stderr, "%s", basedir);
2075: if ('\0' != *basedir && '\0' != *file)
2076: fputs("//", stderr);
1.56 schwarze 2077: if ('\0' != *file)
1.59 schwarze 2078: fprintf(stderr, "%s", file);
1.56 schwarze 2079: fputs(": ", stderr);
1.59 schwarze 2080:
2081: if (NULL == format) {
2082: perror(NULL);
2083: return;
2084: }
1.56 schwarze 2085:
2086: va_start(ap, format);
2087: vfprintf(stderr, format, ap);
2088: va_end(ap);
2089:
2090: fputc('\n', stderr);
1.1 kristaps 2091: }
CVSweb