Annotation of mandoc/mandocdb.c, Revision 1.115
1.115 ! schwarze 1: /* $Id: mandocdb.c,v 1.114 2014/01/22 20:58:39 schwarze Exp $ */
1.1 kristaps 2: /*
1.48 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.95 schwarze 4: * Copyright (c) 2011, 2012, 2013, 2014 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>
1.115 ! schwarze 23: #include <sys/wait.h>
1.1 kristaps 24:
25: #include <assert.h>
1.43 kristaps 26: #include <ctype.h>
1.63 schwarze 27: #include <errno.h>
1.1 kristaps 28: #include <fcntl.h>
1.50 kristaps 29: #include <fts.h>
1.1 kristaps 30: #include <getopt.h>
1.58 schwarze 31: #include <limits.h>
1.50 kristaps 32: #include <stddef.h>
1.59 schwarze 33: #include <stdio.h>
1.1 kristaps 34: #include <stdint.h>
35: #include <stdlib.h>
36: #include <string.h>
1.17 schwarze 37: #include <unistd.h>
1.1 kristaps 38:
1.53 kristaps 39: #ifdef HAVE_OHASH
1.50 kristaps 40: #include <ohash.h>
1.53 kristaps 41: #else
42: #include "compat_ohash.h"
43: #endif
1.50 kristaps 44: #include <sqlite3.h>
1.1 kristaps 45:
1.50 kristaps 46: #include "mdoc.h"
1.1 kristaps 47: #include "man.h"
48: #include "mandoc.h"
1.10 kristaps 49: #include "manpath.h"
1.55 kristaps 50: #include "mansearch.h"
1.1 kristaps 51:
1.111 schwarze 52: extern int mansearch_keymax;
53: extern const char *const mansearch_keynames[];
54:
1.52 kristaps 55: #define SQL_EXEC(_v) \
56: if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
57: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
58: #define SQL_BIND_TEXT(_s, _i, _v) \
59: if (SQLITE_OK != sqlite3_bind_text \
60: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
61: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
62: #define SQL_BIND_INT(_s, _i, _v) \
63: if (SQLITE_OK != sqlite3_bind_int \
64: ((_s), (_i)++, (_v))) \
65: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
66: #define SQL_BIND_INT64(_s, _i, _v) \
67: if (SQLITE_OK != sqlite3_bind_int64 \
68: ((_s), (_i)++, (_v))) \
69: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
70: #define SQL_STEP(_s) \
71: if (SQLITE_DONE != sqlite3_step((_s))) \
72: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
73:
1.50 kristaps 74: enum op {
75: OP_DEFAULT = 0, /* new dbs from dir list or default config */
76: OP_CONFFILE, /* new databases from custom config file */
77: OP_UPDATE, /* delete/add entries in existing database */
78: OP_DELETE, /* delete entries from existing database */
79: OP_TEST /* change no databases, report potential problems */
1.38 schwarze 80: };
81:
1.50 kristaps 82: enum form {
1.84 schwarze 83: FORM_NONE, /* format is unknown */
84: FORM_SRC, /* format is -man or -mdoc */
85: FORM_CAT /* format is cat */
1.50 kristaps 86: };
1.38 schwarze 87:
1.50 kristaps 88: struct str {
1.96 schwarze 89: char *rendered; /* key in UTF-8 or ASCII form */
1.78 schwarze 90: const struct mpage *mpage; /* if set, the owning parse */
1.50 kristaps 91: uint64_t mask; /* bitmask in sequence */
1.96 schwarze 92: char key[]; /* may contain escape sequences */
1.38 schwarze 93: };
94:
1.79 schwarze 95: struct inodev {
96: ino_t st_ino;
97: dev_t st_dev;
1.50 kristaps 98: };
1.5 kristaps 99:
1.78 schwarze 100: struct mpage {
1.79 schwarze 101: struct inodev inodev; /* used for hashing routine */
1.82 schwarze 102: enum form form; /* format from file content */
103: char *sec; /* section from file content */
104: char *arch; /* architecture from file content */
105: char *title; /* title from file content */
106: char *desc; /* description from file content */
107: struct mlink *mlinks; /* singly linked list */
108: };
109:
110: struct mlink {
1.58 schwarze 111: char file[PATH_MAX]; /* filename rel. to manpath */
1.82 schwarze 112: enum form dform; /* format from directory */
1.84 schwarze 113: enum form fform; /* format from file name suffix */
1.82 schwarze 114: char *dsec; /* section from directory */
115: char *arch; /* architecture from directory */
116: char *name; /* name from file name (not empty) */
117: char *fsec; /* section from file name suffix */
1.85 schwarze 118: struct mlink *next; /* singly linked list */
1.3 kristaps 119: };
120:
1.50 kristaps 121: enum stmt {
1.81 schwarze 122: STMT_DELETE_PAGE = 0, /* delete mpage */
123: STMT_INSERT_PAGE, /* insert mpage */
124: STMT_INSERT_LINK, /* insert mlink */
125: STMT_INSERT_KEY, /* insert parsed key */
1.50 kristaps 126: STMT__MAX
1.1 kristaps 127: };
128:
1.78 schwarze 129: typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
1.1 kristaps 130:
1.50 kristaps 131: struct mdoc_handler {
132: mdoc_fp fp; /* optional handler */
133: uint64_t mask; /* set unless handler returns 0 */
1.1 kristaps 134: };
135:
1.59 schwarze 136: static void dbclose(int);
1.105 schwarze 137: static void dbadd(const struct mpage *, struct mchars *);
1.59 schwarze 138: static int dbopen(int);
139: static void dbprune(void);
140: static void filescan(const char *);
1.50 kristaps 141: static void *hash_alloc(size_t, void *);
142: static void hash_free(void *, size_t, void *);
143: static void *hash_halloc(size_t, void *);
1.84 schwarze 144: static void mlink_add(struct mlink *, const struct stat *);
1.93 schwarze 145: static int mlink_check(struct mpage *, struct mlink *);
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);
1.101 schwarze 149: static void mpages_merge(struct mchars *, struct mparse *);
1.78 schwarze 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 *);
1.112 schwarze 161: static void putkey(const struct mpage *, char *, uint64_t);
1.78 schwarze 162: static void putkeys(const struct mpage *,
1.64 schwarze 163: const char *, size_t, uint64_t);
1.78 schwarze 164: static void putmdockey(const struct mpage *,
1.50 kristaps 165: const struct mdoc_node *, uint64_t);
1.96 schwarze 166: static void render_key(struct mchars *, struct str *);
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:
1.115 ! schwarze 172: static char tempfilename[32];
1.50 kristaps 173: static char *progname;
1.102 schwarze 174: static int nodb; /* no database changes */
175: static int quick; /* abort the parse early */
1.50 kristaps 176: static int use_all; /* use all found files */
177: static int verb; /* print what we're doing */
178: static int warnings; /* warn about crap */
1.95 schwarze 179: static int write_utf8; /* write UTF-8 output; else ASCII */
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 */
1.92 schwarze 219: { NULL, TYPE_In }, /* In */
1.68 schwarze 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 */
1.92 schwarze 227: { NULL, TYPE_St }, /* St */
1.68 schwarze 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:
1.102 schwarze 355: while (-1 != (ch = getopt(argc, argv, "aC:d:nQT:tu: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;
1.102 schwarze 373: case ('Q'):
374: quick = 1;
375: break;
1.95 schwarze 376: case ('T'):
377: if (strcmp(optarg, "utf8")) {
378: fprintf(stderr, "-T%s: Unsupported "
379: "output format\n", optarg);
380: goto usage;
381: }
382: write_utf8 = 1;
383: break;
1.38 schwarze 384: case ('t'):
1.50 kristaps 385: CHECKOP(op, ch);
1.38 schwarze 386: dup2(STDOUT_FILENO, STDERR_FILENO);
387: op = OP_TEST;
1.50 kristaps 388: nodb = warnings = 1;
1.38 schwarze 389: break;
1.5 kristaps 390: case ('u'):
1.50 kristaps 391: CHECKOP(op, ch);
1.59 schwarze 392: path_arg = optarg;
1.5 kristaps 393: op = OP_DELETE;
394: break;
395: case ('v'):
396: verb++;
397: break;
1.38 schwarze 398: case ('W'):
399: warnings = 1;
400: break;
1.1 kristaps 401: default:
1.38 schwarze 402: goto usage;
1.1 kristaps 403: }
404:
405: argc -= optind;
406: argv += optind;
407:
1.38 schwarze 408: if (OP_CONFFILE == op && argc > 0) {
1.50 kristaps 409: fprintf(stderr, "-C: Too many arguments\n");
1.38 schwarze 410: goto usage;
411: }
412:
1.59 schwarze 413: exitcode = (int)MANDOCLEVEL_OK;
1.50 kristaps 414: mp = mparse_alloc(MPARSE_AUTO,
1.102 schwarze 415: MANDOCLEVEL_FATAL, NULL, NULL, quick);
1.50 kristaps 416: mc = mchars_alloc();
417:
1.78 schwarze 418: ohash_init(&mpages, 6, &mpages_info);
1.83 schwarze 419: ohash_init(&mlinks, 6, &mlinks_info);
1.50 kristaps 420:
421: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
422: /*
423: * Force processing all files.
424: */
425: use_all = 1;
1.59 schwarze 426:
1.50 kristaps 427: /*
428: * All of these deal with a specific directory.
429: * Jump into that directory then collect files specified
430: * on the command-line.
431: */
1.59 schwarze 432: if (0 == set_basedir(path_arg))
1.50 kristaps 433: goto out;
434: for (i = 0; i < argc; i++)
1.59 schwarze 435: filescan(argv[i]);
436: if (0 == dbopen(1))
1.50 kristaps 437: goto out;
438: if (OP_TEST != op)
1.59 schwarze 439: dbprune();
1.50 kristaps 440: if (OP_DELETE != op)
1.101 schwarze 441: mpages_merge(mc, mp);
1.59 schwarze 442: dbclose(1);
1.50 kristaps 443: } else {
444: /*
445: * If we have arguments, use them as our manpaths.
446: * If we don't, grok from manpath(1) or however else
447: * manpath_parse() wants to do it.
448: */
449: if (argc > 0) {
450: dirs.paths = mandoc_calloc
451: (argc, sizeof(char *));
452: dirs.sz = (size_t)argc;
453: for (i = 0; i < argc; i++)
454: dirs.paths[i] = mandoc_strdup(argv[i]);
455: } else
1.59 schwarze 456: manpath_parse(&dirs, path_arg, NULL, NULL);
1.50 kristaps 457:
458: /*
1.69 schwarze 459: * First scan the tree rooted at a base directory, then
460: * build a new database and finally move it into place.
1.50 kristaps 461: * Ignore zero-length directories and strip trailing
462: * slashes.
463: */
464: for (j = 0; j < dirs.sz; j++) {
465: sz = strlen(dirs.paths[j]);
466: if (sz && '/' == dirs.paths[j][sz - 1])
467: dirs.paths[j][--sz] = '\0';
468: if (0 == sz)
469: continue;
1.66 schwarze 470:
471: if (j) {
1.78 schwarze 472: ohash_init(&mpages, 6, &mpages_info);
1.83 schwarze 473: ohash_init(&mlinks, 6, &mlinks_info);
1.66 schwarze 474: }
475:
1.59 schwarze 476: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 477: goto out;
1.59 schwarze 478: if (0 == treescan())
1.50 kristaps 479: goto out;
1.59 schwarze 480: if (0 == set_basedir(dirs.paths[j]))
1.50 kristaps 481: goto out;
1.59 schwarze 482: if (0 == dbopen(0))
1.50 kristaps 483: goto out;
1.52 kristaps 484:
1.101 schwarze 485: mpages_merge(mc, mp);
1.59 schwarze 486: dbclose(0);
1.66 schwarze 487:
488: if (j + 1 < dirs.sz) {
1.80 schwarze 489: mpages_free();
1.78 schwarze 490: ohash_delete(&mpages);
1.83 schwarze 491: ohash_delete(&mlinks);
1.66 schwarze 492: }
1.50 kristaps 493: }
494: }
495: out:
1.59 schwarze 496: set_basedir(NULL);
1.50 kristaps 497: manpath_free(&dirs);
498: mchars_free(mc);
499: mparse_free(mp);
1.80 schwarze 500: mpages_free();
1.78 schwarze 501: ohash_delete(&mpages);
1.83 schwarze 502: ohash_delete(&mlinks);
1.59 schwarze 503: return(exitcode);
1.50 kristaps 504: usage:
1.102 schwarze 505: fprintf(stderr, "usage: %s [-anQvW] [-C file] [-Tutf8]\n"
506: " %s [-anQvW] [-Tutf8] dir ...\n"
507: " %s [-nQvW] [-Tutf8] -d dir [file ...]\n"
1.50 kristaps 508: " %s [-nvW] -u dir [file ...]\n"
1.102 schwarze 509: " %s [-Q] -t file ...\n",
1.50 kristaps 510: progname, progname, progname,
511: progname, progname);
512:
1.59 schwarze 513: return((int)MANDOCLEVEL_BADARG);
1.50 kristaps 514: }
515:
516: /*
1.59 schwarze 517: * Scan a directory tree rooted at "basedir" for manpages.
1.50 kristaps 518: * We use fts(), scanning directory parts along the way for clues to our
519: * section and architecture.
520: *
521: * If use_all has been specified, grok all files.
522: * If not, sanitise paths to the following:
523: *
524: * [./]man*[/<arch>]/<name>.<section>
525: * or
526: * [./]cat<section>[/<arch>]/<name>.0
527: *
528: * TODO: accomodate for multi-language directories.
529: */
530: static int
1.59 schwarze 531: treescan(void)
1.50 kristaps 532: {
533: FTS *f;
534: FTSENT *ff;
1.84 schwarze 535: struct mlink *mlink;
1.50 kristaps 536: int dform;
1.94 schwarze 537: char *dsec, *arch, *fsec, *cp;
538: const char *path;
1.50 kristaps 539: const char *argv[2];
540:
541: argv[0] = ".";
542: argv[1] = (char *)NULL;
543:
544: /*
545: * Walk through all components under the directory, using the
546: * logical descent of files.
547: */
548: f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
549: if (NULL == f) {
1.59 schwarze 550: exitcode = (int)MANDOCLEVEL_SYSERR;
551: say("", NULL);
1.50 kristaps 552: return(0);
553: }
554:
555: dsec = arch = NULL;
556: dform = FORM_NONE;
557:
558: while (NULL != (ff = fts_read(f))) {
559: path = ff->fts_path + 2;
560: /*
1.83 schwarze 561: * If we're a regular file, add an mlink by using the
1.50 kristaps 562: * stored directory data and handling the filename.
563: */
564: if (FTS_F == ff->fts_info) {
1.60 schwarze 565: if (0 == strcmp(path, MANDOC_DB))
566: continue;
1.50 kristaps 567: if ( ! use_all && ff->fts_level < 2) {
1.56 schwarze 568: if (warnings)
1.59 schwarze 569: say(path, "Extraneous file");
1.50 kristaps 570: continue;
1.84 schwarze 571: } else if (NULL == (fsec =
1.60 schwarze 572: strrchr(ff->fts_name, '.'))) {
573: if ( ! use_all) {
1.56 schwarze 574: if (warnings)
1.60 schwarze 575: say(path,
576: "No filename suffix");
1.50 kristaps 577: continue;
578: }
1.84 schwarze 579: } else if (0 == strcmp(++fsec, "html")) {
1.60 schwarze 580: if (warnings)
581: say(path, "Skip html");
582: continue;
1.84 schwarze 583: } else if (0 == strcmp(fsec, "gz")) {
1.60 schwarze 584: if (warnings)
585: say(path, "Skip gz");
586: continue;
1.84 schwarze 587: } else if (0 == strcmp(fsec, "ps")) {
1.60 schwarze 588: if (warnings)
589: say(path, "Skip ps");
590: continue;
1.84 schwarze 591: } else if (0 == strcmp(fsec, "pdf")) {
1.60 schwarze 592: if (warnings)
593: say(path, "Skip pdf");
594: continue;
595: } else if ( ! use_all &&
1.84 schwarze 596: ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
597: (FORM_CAT == dform && strcmp(fsec, "0")))) {
1.60 schwarze 598: if (warnings)
599: say(path, "Wrong filename suffix");
600: continue;
1.66 schwarze 601: } else
1.84 schwarze 602: fsec[-1] = '\0';
1.94 schwarze 603:
1.84 schwarze 604: mlink = mandoc_calloc(1, sizeof(struct mlink));
605: strlcpy(mlink->file, path, sizeof(mlink->file));
606: mlink->dform = dform;
1.94 schwarze 607: mlink->dsec = dsec;
608: mlink->arch = arch;
609: mlink->name = ff->fts_name;
610: mlink->fsec = fsec;
1.84 schwarze 611: mlink_add(mlink, ff->fts_statp);
1.50 kristaps 612: continue;
1.84 schwarze 613: } else if (FTS_D != ff->fts_info &&
1.60 schwarze 614: FTS_DP != ff->fts_info) {
615: if (warnings)
616: say(path, "Not a regular file");
1.50 kristaps 617: continue;
1.60 schwarze 618: }
1.1 kristaps 619:
1.50 kristaps 620: switch (ff->fts_level) {
621: case (0):
622: /* Ignore the root directory. */
623: break;
624: case (1):
625: /*
626: * This might contain manX/ or catX/.
627: * Try to infer this from the name.
628: * If we're not in use_all, enforce it.
629: */
630: cp = ff->fts_name;
631: if (FTS_DP == ff->fts_info)
632: break;
1.1 kristaps 633:
1.50 kristaps 634: if (0 == strncmp(cp, "man", 3)) {
635: dform = FORM_SRC;
1.66 schwarze 636: dsec = cp + 3;
1.50 kristaps 637: } else if (0 == strncmp(cp, "cat", 3)) {
638: dform = FORM_CAT;
1.66 schwarze 639: dsec = cp + 3;
1.94 schwarze 640: } else {
641: dform = FORM_NONE;
642: dsec = NULL;
1.50 kristaps 643: }
1.1 kristaps 644:
1.50 kristaps 645: if (NULL != dsec || use_all)
646: break;
1.1 kristaps 647:
1.56 schwarze 648: if (warnings)
1.59 schwarze 649: say(path, "Unknown directory part");
1.50 kristaps 650: fts_set(f, ff, FTS_SKIP);
651: break;
652: case (2):
653: /*
654: * Possibly our architecture.
655: * If we're descending, keep tabs on it.
656: */
657: if (FTS_DP != ff->fts_info && NULL != dsec)
1.66 schwarze 658: arch = ff->fts_name;
1.94 schwarze 659: else
660: arch = NULL;
1.50 kristaps 661: break;
662: default:
663: if (FTS_DP == ff->fts_info || use_all)
664: break;
1.56 schwarze 665: if (warnings)
1.59 schwarze 666: say(path, "Extraneous directory part");
1.50 kristaps 667: fts_set(f, ff, FTS_SKIP);
668: break;
1.5 kristaps 669: }
1.50 kristaps 670: }
671:
672: fts_close(f);
673: return(1);
674: }
1.5 kristaps 675:
1.50 kristaps 676: /*
1.89 schwarze 677: * Add a file to the mlinks table.
1.50 kristaps 678: * Do not verify that it's a "valid" looking manpage (we'll do that
679: * later).
680: *
681: * Try to infer the manual section, architecture, and page name from the
682: * path, assuming it looks like
683: *
684: * [./]man*[/<arch>]/<name>.<section>
685: * or
686: * [./]cat<section>[/<arch>]/<name>.0
687: *
688: * See treescan() for the fts(3) version of this.
689: */
690: static void
1.59 schwarze 691: filescan(const char *file)
1.50 kristaps 692: {
1.59 schwarze 693: char buf[PATH_MAX];
1.84 schwarze 694: struct stat st;
695: struct mlink *mlink;
1.59 schwarze 696: char *p, *start;
1.5 kristaps 697:
1.50 kristaps 698: assert(use_all);
1.5 kristaps 699:
1.50 kristaps 700: if (0 == strncmp(file, "./", 2))
701: file += 2;
1.5 kristaps 702:
1.59 schwarze 703: if (NULL == realpath(file, buf)) {
704: exitcode = (int)MANDOCLEVEL_BADARG;
705: say(file, NULL);
706: return;
1.106 schwarze 707: }
708:
709: if (strstr(buf, basedir) == buf)
710: start = buf + strlen(basedir) + 1;
711: else if (OP_TEST == op)
712: start = buf;
713: else {
1.59 schwarze 714: exitcode = (int)MANDOCLEVEL_BADARG;
715: say("", "%s: outside base directory", buf);
716: return;
1.106 schwarze 717: }
718:
719: if (-1 == stat(buf, &st)) {
1.59 schwarze 720: exitcode = (int)MANDOCLEVEL_BADARG;
721: say(file, NULL);
1.50 kristaps 722: return;
723: } else if ( ! (S_IFREG & st.st_mode)) {
1.59 schwarze 724: exitcode = (int)MANDOCLEVEL_BADARG;
725: say(file, "Not a regular file");
1.50 kristaps 726: return;
727: }
1.106 schwarze 728:
1.84 schwarze 729: mlink = mandoc_calloc(1, sizeof(struct mlink));
730: strlcpy(mlink->file, start, sizeof(mlink->file));
1.17 schwarze 731:
1.50 kristaps 732: /*
733: * First try to guess our directory structure.
734: * If we find a separator, try to look for man* or cat*.
735: * If we find one of these and what's underneath is a directory,
736: * assume it's an architecture.
737: */
738: if (NULL != (p = strchr(start, '/'))) {
739: *p++ = '\0';
740: if (0 == strncmp(start, "man", 3)) {
1.84 schwarze 741: mlink->dform = FORM_SRC;
1.94 schwarze 742: mlink->dsec = start + 3;
1.50 kristaps 743: } else if (0 == strncmp(start, "cat", 3)) {
1.84 schwarze 744: mlink->dform = FORM_CAT;
1.94 schwarze 745: mlink->dsec = start + 3;
1.17 schwarze 746: }
1.5 kristaps 747:
1.50 kristaps 748: start = p;
1.84 schwarze 749: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
1.50 kristaps 750: *p++ = '\0';
1.94 schwarze 751: mlink->arch = start;
1.50 kristaps 752: start = p;
1.84 schwarze 753: }
1.50 kristaps 754: }
755:
756: /*
757: * Now check the file suffix.
758: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
759: */
760: p = strrchr(start, '\0');
761: while (p-- > start && '/' != *p && '.' != *p)
762: /* Loop. */ ;
763:
764: if ('.' == *p) {
765: *p++ = '\0';
1.94 schwarze 766: mlink->fsec = p;
1.5 kristaps 767: }
768:
1.10 kristaps 769: /*
1.50 kristaps 770: * Now try to parse the name.
771: * Use the filename portion of the path.
1.10 kristaps 772: */
1.84 schwarze 773: mlink->name = start;
1.50 kristaps 774: if (NULL != (p = strrchr(start, '/'))) {
1.84 schwarze 775: mlink->name = p + 1;
1.50 kristaps 776: *p = '\0';
1.84 schwarze 777: }
778: mlink_add(mlink, &st);
1.50 kristaps 779: }
1.5 kristaps 780:
1.50 kristaps 781: static void
1.84 schwarze 782: mlink_add(struct mlink *mlink, const struct stat *st)
1.50 kristaps 783: {
1.83 schwarze 784: struct inodev inodev;
1.78 schwarze 785: struct mpage *mpage;
1.83 schwarze 786: unsigned int slot;
1.50 kristaps 787:
1.84 schwarze 788: assert(NULL != mlink->file);
1.50 kristaps 789:
1.94 schwarze 790: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
791: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
792: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
793: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.84 schwarze 794:
795: if ('0' == *mlink->fsec) {
796: free(mlink->fsec);
797: mlink->fsec = mandoc_strdup(mlink->dsec);
798: mlink->fform = FORM_CAT;
799: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
800: mlink->fform = FORM_SRC;
1.74 schwarze 801: else
1.84 schwarze 802: mlink->fform = FORM_NONE;
1.82 schwarze 803:
1.83 schwarze 804: slot = ohash_qlookup(&mlinks, mlink->file);
805: assert(NULL == ohash_find(&mlinks, slot));
806: ohash_insert(&mlinks, slot, mlink);
807:
808: inodev.st_ino = st->st_ino;
809: inodev.st_dev = st->st_dev;
810: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
811: sizeof(struct inodev), inodev.st_ino);
812: mpage = ohash_find(&mpages, slot);
813: if (NULL == mpage) {
814: mpage = mandoc_calloc(1, sizeof(struct mpage));
815: mpage->inodev.st_ino = inodev.st_ino;
816: mpage->inodev.st_dev = inodev.st_dev;
817: ohash_insert(&mpages, slot, mpage);
818: } else
1.85 schwarze 819: mlink->next = mpage->mlinks;
1.82 schwarze 820: mpage->mlinks = mlink;
821: }
822:
823: static void
824: mlink_free(struct mlink *mlink)
825: {
826:
827: free(mlink->dsec);
828: free(mlink->arch);
829: free(mlink->name);
830: free(mlink->fsec);
831: free(mlink);
1.50 kristaps 832: }
1.3 kristaps 833:
1.50 kristaps 834: static void
1.78 schwarze 835: mpages_free(void)
1.50 kristaps 836: {
1.78 schwarze 837: struct mpage *mpage;
1.82 schwarze 838: struct mlink *mlink;
1.80 schwarze 839: unsigned int slot;
1.3 kristaps 840:
1.80 schwarze 841: mpage = ohash_first(&mpages, &slot);
842: while (NULL != mpage) {
1.82 schwarze 843: while (NULL != (mlink = mpage->mlinks)) {
1.85 schwarze 844: mpage->mlinks = mlink->next;
1.82 schwarze 845: mlink_free(mlink);
846: }
1.78 schwarze 847: free(mpage->sec);
848: free(mpage->arch);
1.82 schwarze 849: free(mpage->title);
850: free(mpage->desc);
1.78 schwarze 851: free(mpage);
1.80 schwarze 852: mpage = ohash_next(&mpages, &slot);
1.50 kristaps 853: }
854: }
1.38 schwarze 855:
1.50 kristaps 856: /*
1.89 schwarze 857: * For each mlink to the mpage, check whether the path looks like
858: * it is formatted, and if it does, check whether a source manual
859: * exists by the same name, ignoring the suffix.
860: * If both conditions hold, drop the mlink.
861: */
862: static void
863: mlinks_undupe(struct mpage *mpage)
864: {
865: char buf[PATH_MAX];
866: struct mlink **prev;
867: struct mlink *mlink;
868: char *bufp;
869:
870: mpage->form = FORM_CAT;
1.90 schwarze 871: prev = &mpage->mlinks;
872: while (NULL != (mlink = *prev)) {
1.89 schwarze 873: if (FORM_CAT != mlink->dform) {
874: mpage->form = FORM_NONE;
1.90 schwarze 875: goto nextlink;
1.89 schwarze 876: }
877: if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) {
878: if (warnings)
879: say(mlink->file, "Filename too long");
1.90 schwarze 880: goto nextlink;
1.89 schwarze 881: }
882: bufp = strstr(buf, "cat");
883: assert(NULL != bufp);
884: memcpy(bufp, "man", 3);
885: if (NULL != (bufp = strrchr(buf, '.')))
886: *++bufp = '\0';
887: strlcat(buf, mlink->dsec, PATH_MAX);
888: if (NULL == ohash_find(&mlinks,
889: ohash_qlookup(&mlinks, buf)))
1.90 schwarze 890: goto nextlink;
1.89 schwarze 891: if (warnings)
892: say(mlink->file, "Man source exists: %s", buf);
893: if (use_all)
1.90 schwarze 894: goto nextlink;
1.89 schwarze 895: *prev = mlink->next;
896: mlink_free(mlink);
1.90 schwarze 897: continue;
898: nextlink:
899: prev = &(*prev)->next;
1.89 schwarze 900: }
901: }
902:
1.93 schwarze 903: static int
904: mlink_check(struct mpage *mpage, struct mlink *mlink)
905: {
906: int match;
907:
908: match = 1;
909:
910: /*
911: * Check whether the manual section given in a file
912: * agrees with the directory where the file is located.
913: * Some manuals have suffixes like (3p) on their
914: * section number either inside the file or in the
915: * directory name, some are linked into more than one
916: * section, like encrypt(1) = makekey(8).
917: */
918:
919: if (FORM_SRC == mpage->form &&
920: strcasecmp(mpage->sec, mlink->dsec)) {
921: match = 0;
922: say(mlink->file, "Section \"%s\" manual in %s directory",
923: mpage->sec, mlink->dsec);
924: }
925:
926: /*
927: * Manual page directories exist for each kernel
928: * architecture as returned by machine(1).
929: * However, many manuals only depend on the
930: * application architecture as returned by arch(1).
931: * For example, some (2/ARM) manuals are shared
932: * across the "armish" and "zaurus" kernel
933: * architectures.
934: * A few manuals are even shared across completely
935: * different architectures, for example fdformat(1)
936: * on amd64, i386, sparc, and sparc64.
937: */
938:
939: if (strcasecmp(mpage->arch, mlink->arch)) {
940: match = 0;
941: say(mlink->file, "Architecture \"%s\" manual in "
942: "\"%s\" directory", mpage->arch, mlink->arch);
943: }
944:
945: if (strcasecmp(mpage->title, mlink->name))
946: match = 0;
947:
948: return(match);
949: }
950:
1.89 schwarze 951: /*
1.80 schwarze 952: * Run through the files in the global vector "mpages"
953: * and add them to the database specified in "basedir".
1.50 kristaps 954: *
955: * This handles the parsing scheme itself, using the cues of directory
956: * and filename to determine whether the file is parsable or not.
957: */
1.59 schwarze 958: static void
1.101 schwarze 959: mpages_merge(struct mchars *mc, struct mparse *mp)
1.50 kristaps 960: {
1.113 schwarze 961: char any[] = "any";
1.101 schwarze 962: struct ohash_info str_info;
1.78 schwarze 963: struct mpage *mpage;
1.93 schwarze 964: struct mlink *mlink;
1.69 schwarze 965: struct mdoc *mdoc;
966: struct man *man;
1.112 schwarze 967: char *cp;
1.69 schwarze 968: int match;
1.101 schwarze 969: unsigned int pslot;
1.69 schwarze 970: enum mandoclevel lvl;
971:
1.71 schwarze 972: str_info.alloc = hash_alloc;
973: str_info.halloc = hash_halloc;
974: str_info.hfree = hash_free;
975: str_info.key_offset = offsetof(struct str, key);
976:
1.107 schwarze 977: if (0 == nodb)
978: SQL_EXEC("BEGIN TRANSACTION");
979:
1.80 schwarze 980: mpage = ohash_first(&mpages, &pslot);
981: while (NULL != mpage) {
1.89 schwarze 982: mlinks_undupe(mpage);
983: if (NULL == mpage->mlinks) {
984: mpage = ohash_next(&mpages, &pslot);
985: continue;
1.50 kristaps 986: }
987:
1.71 schwarze 988: ohash_init(&strings, 6, &str_info);
1.50 kristaps 989: mparse_reset(mp);
990: mdoc = NULL;
991: man = NULL;
1.14 schwarze 992:
993: /*
1.33 schwarze 994: * Try interpreting the file as mdoc(7) or man(7)
995: * source code, unless it is already known to be
996: * formatted. Fall back to formatted mode.
1.14 schwarze 997: */
1.82 schwarze 998: if (FORM_CAT != mpage->mlinks->dform ||
1.84 schwarze 999: FORM_CAT != mpage->mlinks->fform) {
1.82 schwarze 1000: lvl = mparse_readfd(mp, -1, mpage->mlinks->file);
1.50 kristaps 1001: if (lvl < MANDOCLEVEL_FATAL)
1002: mparse_result(mp, &mdoc, &man);
1.76 schwarze 1003: }
1.14 schwarze 1004:
1005: if (NULL != mdoc) {
1.82 schwarze 1006: mpage->form = FORM_SRC;
1007: mpage->sec =
1008: mandoc_strdup(mdoc_meta(mdoc)->msec);
1009: mpage->arch = mdoc_meta(mdoc)->arch;
1.84 schwarze 1010: mpage->arch = mandoc_strdup(
1011: NULL == mpage->arch ? "" : mpage->arch);
1.82 schwarze 1012: mpage->title =
1013: mandoc_strdup(mdoc_meta(mdoc)->title);
1.14 schwarze 1014: } else if (NULL != man) {
1.82 schwarze 1015: mpage->form = FORM_SRC;
1016: mpage->sec =
1017: mandoc_strdup(man_meta(man)->msec);
1018: mpage->arch =
1019: mandoc_strdup(mpage->mlinks->arch);
1020: mpage->title =
1021: mandoc_strdup(man_meta(man)->title);
1.76 schwarze 1022: } else {
1.82 schwarze 1023: mpage->form = FORM_CAT;
1024: mpage->sec =
1025: mandoc_strdup(mpage->mlinks->dsec);
1026: mpage->arch =
1027: mandoc_strdup(mpage->mlinks->arch);
1028: mpage->title =
1029: mandoc_strdup(mpage->mlinks->name);
1.76 schwarze 1030: }
1.97 schwarze 1031: putkey(mpage, mpage->sec, TYPE_sec);
1.98 schwarze 1032: putkey(mpage, '\0' == *mpage->arch ?
1.113 schwarze 1033: any : mpage->arch, TYPE_arch);
1.50 kristaps 1034:
1.97 schwarze 1035: for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1036: if ('\0' != *mlink->dsec)
1037: putkey(mpage, mlink->dsec, TYPE_sec);
1038: if ('\0' != *mlink->fsec)
1039: putkey(mpage, mlink->fsec, TYPE_sec);
1.98 schwarze 1040: putkey(mpage, '\0' == *mlink->arch ?
1.113 schwarze 1041: any : mlink->arch, TYPE_arch);
1.93 schwarze 1042: putkey(mpage, mlink->name, TYPE_Nm);
1.97 schwarze 1043: }
1.12 schwarze 1044:
1.93 schwarze 1045: if (warnings && !use_all) {
1.69 schwarze 1046: match = 0;
1.93 schwarze 1047: for (mlink = mpage->mlinks; mlink;
1048: mlink = mlink->next)
1049: if (mlink_check(mpage, mlink))
1050: match = 1;
1051: } else
1052: match = 1;
1.12 schwarze 1053:
1.50 kristaps 1054: if (NULL != mdoc) {
1055: if (NULL != (cp = mdoc_meta(mdoc)->name))
1.78 schwarze 1056: putkey(mpage, cp, TYPE_Nm);
1057: assert(NULL == mpage->desc);
1058: parse_mdoc(mpage, mdoc_node(mdoc));
1059: putkey(mpage, NULL != mpage->desc ?
1.82 schwarze 1060: mpage->desc : mpage->mlinks->name, TYPE_Nd);
1.50 kristaps 1061: } else if (NULL != man)
1.78 schwarze 1062: parse_man(mpage, man_node(man));
1.50 kristaps 1063: else
1.78 schwarze 1064: parse_cat(mpage);
1.44 kristaps 1065:
1.105 schwarze 1066: dbadd(mpage, mc);
1.66 schwarze 1067: ohash_delete(&strings);
1.80 schwarze 1068: mpage = ohash_next(&mpages, &pslot);
1.50 kristaps 1069: }
1.107 schwarze 1070:
1071: if (0 == nodb)
1072: SQL_EXEC("END TRANSACTION");
1.50 kristaps 1073: }
1.12 schwarze 1074:
1.50 kristaps 1075: static void
1.78 schwarze 1076: parse_cat(struct mpage *mpage)
1.50 kristaps 1077: {
1078: FILE *stream;
1079: char *line, *p, *title;
1080: size_t len, plen, titlesz;
1.12 schwarze 1081:
1.82 schwarze 1082: if (NULL == (stream = fopen(mpage->mlinks->file, "r"))) {
1.56 schwarze 1083: if (warnings)
1.82 schwarze 1084: say(mpage->mlinks->file, NULL);
1.50 kristaps 1085: return;
1086: }
1.1 kristaps 1087:
1.50 kristaps 1088: /* Skip to first blank line. */
1.1 kristaps 1089:
1.50 kristaps 1090: while (NULL != (line = fgetln(stream, &len)))
1091: if ('\n' == *line)
1092: break;
1.1 kristaps 1093:
1.50 kristaps 1094: /*
1095: * Assume the first line that is not indented
1096: * is the first section header. Skip to it.
1097: */
1.1 kristaps 1098:
1.50 kristaps 1099: while (NULL != (line = fgetln(stream, &len)))
1100: if ('\n' != *line && ' ' != *line)
1101: break;
1102:
1103: /*
1104: * Read up until the next section into a buffer.
1105: * Strip the leading and trailing newline from each read line,
1106: * appending a trailing space.
1107: * Ignore empty (whitespace-only) lines.
1108: */
1.1 kristaps 1109:
1.50 kristaps 1110: titlesz = 0;
1111: title = NULL;
1.38 schwarze 1112:
1.50 kristaps 1113: while (NULL != (line = fgetln(stream, &len))) {
1114: if (' ' != *line || '\n' != line[len - 1])
1115: break;
1116: while (len > 0 && isspace((unsigned char)*line)) {
1117: line++;
1118: len--;
1119: }
1120: if (1 == len)
1.38 schwarze 1121: continue;
1.50 kristaps 1122: title = mandoc_realloc(title, titlesz + len);
1123: memcpy(title + titlesz, line, len);
1124: titlesz += len;
1125: title[titlesz - 1] = ' ';
1126: }
1.38 schwarze 1127:
1.50 kristaps 1128: /*
1129: * If no page content can be found, or the input line
1130: * is already the next section header, or there is no
1131: * trailing newline, reuse the page title as the page
1132: * description.
1133: */
1.44 kristaps 1134:
1.50 kristaps 1135: if (NULL == title || '\0' == *title) {
1.56 schwarze 1136: if (warnings)
1.82 schwarze 1137: say(mpage->mlinks->file,
1138: "Cannot find NAME section");
1.78 schwarze 1139: assert(NULL == mpage->desc);
1.82 schwarze 1140: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1141: putkey(mpage, mpage->mlinks->name, TYPE_Nd);
1.50 kristaps 1142: fclose(stream);
1143: free(title);
1144: return;
1145: }
1.1 kristaps 1146:
1.50 kristaps 1147: title = mandoc_realloc(title, titlesz + 1);
1148: title[titlesz] = '\0';
1.33 schwarze 1149:
1.50 kristaps 1150: /*
1151: * Skip to the first dash.
1152: * Use the remaining line as the description (no more than 70
1153: * bytes).
1154: */
1.33 schwarze 1155:
1.50 kristaps 1156: if (NULL != (p = strstr(title, "- "))) {
1157: for (p += 2; ' ' == *p || '\b' == *p; p++)
1158: /* Skip to next word. */ ;
1159: } else {
1.56 schwarze 1160: if (warnings)
1.82 schwarze 1161: say(mpage->mlinks->file,
1162: "No dash in title line");
1.50 kristaps 1163: p = title;
1164: }
1.38 schwarze 1165:
1.50 kristaps 1166: plen = strlen(p);
1.1 kristaps 1167:
1.50 kristaps 1168: /* Strip backspace-encoding from line. */
1.1 kristaps 1169:
1.50 kristaps 1170: while (NULL != (line = memchr(p, '\b', plen))) {
1171: len = line - p;
1172: if (0 == len) {
1173: memmove(line, line + 1, plen--);
1174: continue;
1175: }
1176: memmove(line - 1, line + 1, plen - len);
1177: plen -= 2;
1178: }
1.1 kristaps 1179:
1.78 schwarze 1180: assert(NULL == mpage->desc);
1181: mpage->desc = mandoc_strdup(p);
1182: putkey(mpage, mpage->desc, TYPE_Nd);
1.50 kristaps 1183: fclose(stream);
1184: free(title);
1185: }
1.1 kristaps 1186:
1.50 kristaps 1187: /*
1188: * Put a type/word pair into the word database for this particular file.
1189: */
1190: static void
1.112 schwarze 1191: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.50 kristaps 1192: {
1.112 schwarze 1193: char *cp;
1.18 kristaps 1194:
1.50 kristaps 1195: assert(NULL != value);
1.112 schwarze 1196: if (TYPE_arch == type)
1197: for (cp = value; *cp; cp++)
1198: if (isupper((unsigned char)*cp))
1199: *cp = _tolower((unsigned char)*cp);
1.78 schwarze 1200: putkeys(mpage, value, strlen(value), type);
1.3 kristaps 1201: }
1202:
1203: /*
1.50 kristaps 1204: * Grok all nodes at or below a certain mdoc node into putkey().
1.3 kristaps 1205: */
1206: static void
1.78 schwarze 1207: putmdockey(const struct mpage *mpage,
1208: const struct mdoc_node *n, uint64_t m)
1.3 kristaps 1209: {
1.18 kristaps 1210:
1.50 kristaps 1211: for ( ; NULL != n; n = n->next) {
1212: if (NULL != n->child)
1.78 schwarze 1213: putmdockey(mpage, n->child, m);
1.50 kristaps 1214: if (MDOC_TEXT == n->type)
1.78 schwarze 1215: putkey(mpage, n->string, m);
1.50 kristaps 1216: }
1217: }
1.18 kristaps 1218:
1.61 schwarze 1219: static void
1.78 schwarze 1220: parse_man(struct mpage *mpage, const struct man_node *n)
1.50 kristaps 1221: {
1222: const struct man_node *head, *body;
1223: char *start, *sv, *title;
1224: char byte;
1225: size_t sz, titlesz;
1.18 kristaps 1226:
1.50 kristaps 1227: if (NULL == n)
1.61 schwarze 1228: return;
1.18 kristaps 1229:
1.50 kristaps 1230: /*
1231: * We're only searching for one thing: the first text child in
1232: * the BODY of a NAME section. Since we don't keep track of
1233: * sections in -man, run some hoops to find out whether we're in
1234: * the correct section or not.
1235: */
1.18 kristaps 1236:
1.50 kristaps 1237: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1238: body = n;
1239: assert(body->parent);
1240: if (NULL != (head = body->parent->head) &&
1241: 1 == head->nchild &&
1242: NULL != (head = (head->child)) &&
1243: MAN_TEXT == head->type &&
1244: 0 == strcmp(head->string, "NAME") &&
1245: NULL != (body = body->child) &&
1246: MAN_TEXT == body->type) {
1.3 kristaps 1247:
1.50 kristaps 1248: title = NULL;
1249: titlesz = 0;
1.3 kristaps 1250:
1.50 kristaps 1251: /*
1252: * Suck the entire NAME section into memory.
1253: * Yes, we might run away.
1254: * But too many manuals have big, spread-out
1255: * NAME sections over many lines.
1256: */
1.3 kristaps 1257:
1.50 kristaps 1258: for ( ; NULL != body; body = body->next) {
1259: if (MAN_TEXT != body->type)
1260: break;
1261: if (0 == (sz = strlen(body->string)))
1262: continue;
1263: title = mandoc_realloc
1264: (title, titlesz + sz + 1);
1265: memcpy(title + titlesz, body->string, sz);
1266: titlesz += sz + 1;
1267: title[titlesz - 1] = ' ';
1268: }
1269: if (NULL == title)
1.61 schwarze 1270: return;
1.18 kristaps 1271:
1.50 kristaps 1272: title = mandoc_realloc(title, titlesz + 1);
1273: title[titlesz] = '\0';
1.18 kristaps 1274:
1.50 kristaps 1275: /* Skip leading space. */
1.18 kristaps 1276:
1.50 kristaps 1277: sv = title;
1278: while (isspace((unsigned char)*sv))
1279: sv++;
1.18 kristaps 1280:
1.50 kristaps 1281: if (0 == (sz = strlen(sv))) {
1282: free(title);
1.61 schwarze 1283: return;
1.50 kristaps 1284: }
1.1 kristaps 1285:
1.50 kristaps 1286: /* Erase trailing space. */
1.1 kristaps 1287:
1.50 kristaps 1288: start = &sv[sz - 1];
1289: while (start > sv && isspace((unsigned char)*start))
1290: *start-- = '\0';
1.1 kristaps 1291:
1.50 kristaps 1292: if (start == sv) {
1293: free(title);
1.61 schwarze 1294: return;
1.50 kristaps 1295: }
1.1 kristaps 1296:
1.50 kristaps 1297: start = sv;
1.18 kristaps 1298:
1.50 kristaps 1299: /*
1300: * Go through a special heuristic dance here.
1301: * Conventionally, one or more manual names are
1302: * comma-specified prior to a whitespace, then a
1303: * dash, then a description. Try to puzzle out
1304: * the name parts here.
1305: */
1.18 kristaps 1306:
1.50 kristaps 1307: for ( ;; ) {
1308: sz = strcspn(start, " ,");
1309: if ('\0' == start[sz])
1310: break;
1.1 kristaps 1311:
1.50 kristaps 1312: byte = start[sz];
1313: start[sz] = '\0';
1.110 schwarze 1314:
1315: /*
1316: * Assume a stray trailing comma in the
1317: * name list if a name begins with a dash.
1318: */
1319:
1320: if ('-' == start[0] ||
1321: ('\\' == start[0] && '-' == start[1]))
1322: break;
1.1 kristaps 1323:
1.78 schwarze 1324: putkey(mpage, start, TYPE_Nm);
1.1 kristaps 1325:
1.50 kristaps 1326: if (' ' == byte) {
1327: start += sz + 1;
1328: break;
1329: }
1.1 kristaps 1330:
1.50 kristaps 1331: assert(',' == byte);
1332: start += sz + 1;
1333: while (' ' == *start)
1334: start++;
1335: }
1.1 kristaps 1336:
1.50 kristaps 1337: if (sv == start) {
1.78 schwarze 1338: putkey(mpage, start, TYPE_Nm);
1.50 kristaps 1339: free(title);
1.61 schwarze 1340: return;
1.50 kristaps 1341: }
1.1 kristaps 1342:
1.50 kristaps 1343: while (isspace((unsigned char)*start))
1344: start++;
1.1 kristaps 1345:
1.50 kristaps 1346: if (0 == strncmp(start, "-", 1))
1347: start += 1;
1348: else if (0 == strncmp(start, "\\-\\-", 4))
1349: start += 4;
1350: else if (0 == strncmp(start, "\\-", 2))
1351: start += 2;
1352: else if (0 == strncmp(start, "\\(en", 4))
1353: start += 4;
1354: else if (0 == strncmp(start, "\\(em", 4))
1355: start += 4;
1.1 kristaps 1356:
1.50 kristaps 1357: while (' ' == *start)
1358: start++;
1.1 kristaps 1359:
1.78 schwarze 1360: assert(NULL == mpage->desc);
1361: mpage->desc = mandoc_strdup(start);
1362: putkey(mpage, mpage->desc, TYPE_Nd);
1.50 kristaps 1363: free(title);
1.61 schwarze 1364: return;
1.50 kristaps 1365: }
1366: }
1.1 kristaps 1367:
1.77 schwarze 1368: for (n = n->child; n; n = n->next) {
1.78 schwarze 1369: if (NULL != mpage->desc)
1.77 schwarze 1370: break;
1.78 schwarze 1371: parse_man(mpage, n);
1.77 schwarze 1372: }
1.1 kristaps 1373: }
1374:
1375: static void
1.78 schwarze 1376: parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1377: {
1378:
1.50 kristaps 1379: assert(NULL != n);
1380: for (n = n->child; NULL != n; n = n->next) {
1381: switch (n->type) {
1382: case (MDOC_ELEM):
1383: /* FALLTHROUGH */
1384: case (MDOC_BLOCK):
1385: /* FALLTHROUGH */
1386: case (MDOC_HEAD):
1387: /* FALLTHROUGH */
1388: case (MDOC_BODY):
1389: /* FALLTHROUGH */
1390: case (MDOC_TAIL):
1391: if (NULL != mdocs[n->tok].fp)
1.78 schwarze 1392: if (0 == (*mdocs[n->tok].fp)(mpage, n))
1.50 kristaps 1393: break;
1.68 schwarze 1394: if (mdocs[n->tok].mask)
1.78 schwarze 1395: putmdockey(mpage, n->child,
1396: mdocs[n->tok].mask);
1.50 kristaps 1397: break;
1398: default:
1399: assert(MDOC_ROOT != n->type);
1400: continue;
1401: }
1402: if (NULL != n->child)
1.78 schwarze 1403: parse_mdoc(mpage, n);
1.1 kristaps 1404: }
1405: }
1406:
1.25 schwarze 1407: static int
1.78 schwarze 1408: parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1409: {
1410: const char *start, *end;
1411: size_t sz;
1.25 schwarze 1412:
1.50 kristaps 1413: if (SEC_SYNOPSIS != n->sec ||
1414: NULL == (n = n->child) ||
1415: MDOC_TEXT != n->type)
1.25 schwarze 1416: return(0);
1.1 kristaps 1417:
1418: /*
1419: * Only consider those `Fd' macro fields that begin with an
1420: * "inclusion" token (versus, e.g., #define).
1421: */
1.50 kristaps 1422:
1.1 kristaps 1423: if (strcmp("#include", n->string))
1.25 schwarze 1424: return(0);
1.1 kristaps 1425:
1426: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.25 schwarze 1427: return(0);
1.1 kristaps 1428:
1429: /*
1430: * Strip away the enclosing angle brackets and make sure we're
1431: * not zero-length.
1432: */
1433:
1434: start = n->string;
1435: if ('<' == *start || '"' == *start)
1436: start++;
1437:
1438: if (0 == (sz = strlen(start)))
1.25 schwarze 1439: return(0);
1.1 kristaps 1440:
1441: end = &start[(int)sz - 1];
1442: if ('>' == *end || '"' == *end)
1443: end--;
1444:
1.50 kristaps 1445: if (end > start)
1.78 schwarze 1446: putkeys(mpage, start, end - start + 1, TYPE_In);
1.92 schwarze 1447: return(0);
1.1 kristaps 1448: }
1449:
1.25 schwarze 1450: static int
1.78 schwarze 1451: parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1452: {
1.112 schwarze 1453: char *cp;
1.1 kristaps 1454:
1.50 kristaps 1455: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.25 schwarze 1456: return(0);
1457:
1.50 kristaps 1458: /*
1459: * Parse: .Fn "struct type *name" "char *arg".
1460: * First strip away pointer symbol.
1461: * Then store the function name, then type.
1462: * Finally, store the arguments.
1463: */
1.1 kristaps 1464:
1.50 kristaps 1465: if (NULL == (cp = strrchr(n->string, ' ')))
1466: cp = n->string;
1.1 kristaps 1467:
1468: while ('*' == *cp)
1469: cp++;
1470:
1.78 schwarze 1471: putkey(mpage, cp, TYPE_Fn);
1.25 schwarze 1472:
1.50 kristaps 1473: if (n->string < cp)
1.78 schwarze 1474: putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1.25 schwarze 1475:
1.50 kristaps 1476: for (n = n->next; NULL != n; n = n->next)
1477: if (MDOC_TEXT == n->type)
1.78 schwarze 1478: putkey(mpage, n->string, TYPE_Fa);
1.25 schwarze 1479:
1480: return(0);
1.1 kristaps 1481: }
1482:
1.25 schwarze 1483: static int
1.78 schwarze 1484: parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1485: {
1.67 schwarze 1486: char *cp;
1.1 kristaps 1487:
1488: if (NULL == (n = n->child))
1.25 schwarze 1489: return(0);
1.1 kristaps 1490:
1.67 schwarze 1491: if (NULL == n->next) {
1.78 schwarze 1492: putkey(mpage, n->string, TYPE_Xr);
1.67 schwarze 1493: return(0);
1494: }
1495:
1496: if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) {
1497: perror(NULL);
1498: exit((int)MANDOCLEVEL_SYSERR);
1499: }
1.78 schwarze 1500: putkey(mpage, cp, TYPE_Xr);
1.67 schwarze 1501: free(cp);
1502: return(0);
1.1 kristaps 1503: }
1504:
1.25 schwarze 1505: static int
1.78 schwarze 1506: parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1507: {
1.50 kristaps 1508: size_t sz;
1.1 kristaps 1509:
1510: if (MDOC_BODY != n->type)
1.25 schwarze 1511: return(0);
1.1 kristaps 1512:
1.50 kristaps 1513: /*
1514: * Special-case the `Nd' because we need to put the description
1515: * into the document table.
1516: */
1517:
1518: for (n = n->child; NULL != n; n = n->next) {
1519: if (MDOC_TEXT == n->type) {
1.78 schwarze 1520: if (NULL != mpage->desc) {
1521: sz = strlen(mpage->desc) +
1.66 schwarze 1522: strlen(n->string) + 2;
1.78 schwarze 1523: mpage->desc = mandoc_realloc(
1524: mpage->desc, sz);
1525: strlcat(mpage->desc, " ", sz);
1526: strlcat(mpage->desc, n->string, sz);
1.66 schwarze 1527: } else
1.78 schwarze 1528: mpage->desc = mandoc_strdup(n->string);
1.50 kristaps 1529: }
1530: if (NULL != n->child)
1.78 schwarze 1531: parse_mdoc_Nd(mpage, n);
1.50 kristaps 1532: }
1.25 schwarze 1533: return(1);
1.1 kristaps 1534: }
1535:
1.25 schwarze 1536: static int
1.78 schwarze 1537: parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1538: {
1539:
1.92 schwarze 1540: return(SEC_NAME == n->sec ||
1541: (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type));
1.1 kristaps 1542: }
1543:
1.25 schwarze 1544: static int
1.78 schwarze 1545: parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1546: {
1547:
1.25 schwarze 1548: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 kristaps 1549: }
1550:
1.50 kristaps 1551: static int
1.78 schwarze 1552: parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1.1 kristaps 1553: {
1554:
1.50 kristaps 1555: return(MDOC_HEAD == n->type);
1556: }
1.1 kristaps 1557:
1.50 kristaps 1558: static int
1.78 schwarze 1559: parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1.50 kristaps 1560: {
1.1 kristaps 1561:
1.50 kristaps 1562: return(MDOC_BODY == n->type);
1.1 kristaps 1563: }
1564:
1.50 kristaps 1565: /*
1.66 schwarze 1566: * Add a string to the hash table for the current manual.
1567: * Each string has a bitmask telling which macros it belongs to.
1568: * When we finish the manual, we'll dump the table.
1.50 kristaps 1569: */
1570: static void
1.78 schwarze 1571: putkeys(const struct mpage *mpage,
1572: const char *cp, size_t sz, uint64_t v)
1.50 kristaps 1573: {
1574: struct str *s;
1.111 schwarze 1575: const char *end;
1576: uint64_t mask;
1.65 schwarze 1577: unsigned int slot;
1.111 schwarze 1578: int i;
1.25 schwarze 1579:
1.50 kristaps 1580: if (0 == sz)
1581: return;
1.111 schwarze 1582:
1583: if (verb > 1) {
1584: for (i = 0, mask = 1;
1585: i < mansearch_keymax;
1586: i++, mask <<= 1)
1587: if (mask & v)
1588: break;
1589: say(mpage->mlinks->file, "Adding key %s=%*s",
1590: mansearch_keynames[i], sz, cp);
1591: }
1.25 schwarze 1592:
1.65 schwarze 1593: end = cp + sz;
1594: slot = ohash_qlookupi(&strings, cp, &end);
1595: s = ohash_find(&strings, slot);
1.25 schwarze 1596:
1.78 schwarze 1597: if (NULL != s && mpage == s->mpage) {
1.50 kristaps 1598: s->mask |= v;
1599: return;
1600: } else if (NULL == s) {
1.51 kristaps 1601: s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1.50 kristaps 1602: memcpy(s->key, cp, sz);
1.65 schwarze 1603: ohash_insert(&strings, slot, s);
1.1 kristaps 1604: }
1.78 schwarze 1605: s->mpage = mpage;
1.50 kristaps 1606: s->mask = v;
1.1 kristaps 1607: }
1608:
1.50 kristaps 1609: /*
1610: * Take a Unicode codepoint and produce its UTF-8 encoding.
1611: * This isn't the best way to do this, but it works.
1612: * The magic numbers are from the UTF-8 packaging.
1613: * They're not as scary as they seem: read the UTF-8 spec for details.
1614: */
1615: static size_t
1616: utf8(unsigned int cp, char out[7])
1.1 kristaps 1617: {
1.50 kristaps 1618: size_t rc;
1.1 kristaps 1619:
1.50 kristaps 1620: rc = 0;
1621: if (cp <= 0x0000007F) {
1622: rc = 1;
1623: out[0] = (char)cp;
1624: } else if (cp <= 0x000007FF) {
1625: rc = 2;
1626: out[0] = (cp >> 6 & 31) | 192;
1627: out[1] = (cp & 63) | 128;
1628: } else if (cp <= 0x0000FFFF) {
1629: rc = 3;
1630: out[0] = (cp >> 12 & 15) | 224;
1631: out[1] = (cp >> 6 & 63) | 128;
1632: out[2] = (cp & 63) | 128;
1633: } else if (cp <= 0x001FFFFF) {
1634: rc = 4;
1635: out[0] = (cp >> 18 & 7) | 240;
1636: out[1] = (cp >> 12 & 63) | 128;
1637: out[2] = (cp >> 6 & 63) | 128;
1638: out[3] = (cp & 63) | 128;
1639: } else if (cp <= 0x03FFFFFF) {
1640: rc = 5;
1641: out[0] = (cp >> 24 & 3) | 248;
1642: out[1] = (cp >> 18 & 63) | 128;
1643: out[2] = (cp >> 12 & 63) | 128;
1644: out[3] = (cp >> 6 & 63) | 128;
1645: out[4] = (cp & 63) | 128;
1646: } else if (cp <= 0x7FFFFFFF) {
1647: rc = 6;
1648: out[0] = (cp >> 30 & 1) | 252;
1649: out[1] = (cp >> 24 & 63) | 128;
1650: out[2] = (cp >> 18 & 63) | 128;
1651: out[3] = (cp >> 12 & 63) | 128;
1652: out[4] = (cp >> 6 & 63) | 128;
1653: out[5] = (cp & 63) | 128;
1654: } else
1.1 kristaps 1655: return(0);
1656:
1.50 kristaps 1657: out[rc] = '\0';
1658: return(rc);
1659: }
1.1 kristaps 1660:
1.50 kristaps 1661: /*
1.96 schwarze 1662: * Store the rendered version of a key, or alias the pointer
1663: * if the key contains no escape sequences.
1.50 kristaps 1664: */
1665: static void
1.96 schwarze 1666: render_key(struct mchars *mc, struct str *key)
1.50 kristaps 1667: {
1668: size_t sz, bsz, pos;
1.114 schwarze 1669: char utfbuf[7], res[6];
1.50 kristaps 1670: char *buf;
1671: const char *seq, *cpp, *val;
1672: int len, u;
1673: enum mandoc_esc esc;
1674:
1.96 schwarze 1675: assert(NULL == key->rendered);
1.50 kristaps 1676:
1677: res[0] = '\\';
1678: res[1] = '\t';
1679: res[2] = ASCII_NBRSP;
1680: res[3] = ASCII_HYPH;
1.114 schwarze 1681: res[4] = ASCII_BREAK;
1682: res[5] = '\0';
1.1 kristaps 1683:
1.50 kristaps 1684: val = key->key;
1685: bsz = strlen(val);
1.46 kristaps 1686:
1.50 kristaps 1687: /*
1688: * Pre-check: if we have no stop-characters, then set the
1689: * pointer as ourselvse and get out of here.
1690: */
1691: if (strcspn(val, res) == bsz) {
1.96 schwarze 1692: key->rendered = key->key;
1.50 kristaps 1693: return;
1694: }
1.46 kristaps 1695:
1.50 kristaps 1696: /* Pre-allocate by the length of the input */
1.46 kristaps 1697:
1.50 kristaps 1698: buf = mandoc_malloc(++bsz);
1699: pos = 0;
1.46 kristaps 1700:
1.50 kristaps 1701: while ('\0' != *val) {
1702: /*
1703: * Halt on the first escape sequence.
1704: * This also halts on the end of string, in which case
1705: * we just copy, fallthrough, and exit the loop.
1706: */
1707: if ((sz = strcspn(val, res)) > 0) {
1708: memcpy(&buf[pos], val, sz);
1709: pos += sz;
1710: val += sz;
1711: }
1.46 kristaps 1712:
1.114 schwarze 1713: switch (*val) {
1714: case (ASCII_HYPH):
1.50 kristaps 1715: buf[pos++] = '-';
1716: val++;
1717: continue;
1.114 schwarze 1718: case ('\t'):
1719: /* FALLTHROUGH */
1720: case (ASCII_NBRSP):
1.50 kristaps 1721: buf[pos++] = ' ';
1722: val++;
1.114 schwarze 1723: /* FALLTHROUGH */
1724: case (ASCII_BREAK):
1.50 kristaps 1725: continue;
1.114 schwarze 1726: default:
1727: break;
1728: }
1729: if ('\\' != *val)
1.50 kristaps 1730: break;
1.46 kristaps 1731:
1.50 kristaps 1732: /* Read past the slash. */
1.46 kristaps 1733:
1.50 kristaps 1734: val++;
1.46 kristaps 1735:
1.50 kristaps 1736: /*
1737: * Parse the escape sequence and see if it's a
1738: * predefined character or special character.
1739: */
1.95 schwarze 1740:
1.50 kristaps 1741: esc = mandoc_escape
1742: ((const char **)&val, &seq, &len);
1743: if (ESCAPE_ERROR == esc)
1744: break;
1745: if (ESCAPE_SPECIAL != esc)
1746: continue;
1.1 kristaps 1747:
1.50 kristaps 1748: /*
1.95 schwarze 1749: * Render the special character
1750: * as either UTF-8 or ASCII.
1.50 kristaps 1751: */
1.95 schwarze 1752:
1753: if (write_utf8) {
1754: if (0 == (u = mchars_spec2cp(mc, seq, len)))
1755: continue;
1756: cpp = utfbuf;
1757: if (0 == (sz = utf8(u, utfbuf)))
1758: continue;
1759: sz = strlen(cpp);
1760: } else {
1761: cpp = mchars_spec2str(mc, seq, len, &sz);
1762: if (NULL == cpp)
1763: continue;
1764: if (ASCII_NBRSP == *cpp) {
1765: cpp = " ";
1766: sz = 1;
1767: }
1768: }
1.1 kristaps 1769:
1.50 kristaps 1770: /* Copy the rendered glyph into the stream. */
1.1 kristaps 1771:
1.50 kristaps 1772: bsz += sz;
1773: buf = mandoc_realloc(buf, bsz);
1774: memcpy(&buf[pos], cpp, sz);
1775: pos += sz;
1.1 kristaps 1776: }
1777:
1.50 kristaps 1778: buf[pos] = '\0';
1.96 schwarze 1779: key->rendered = buf;
1.1 kristaps 1780: }
1781:
1.14 schwarze 1782: /*
1.50 kristaps 1783: * Flush the current page's terms (and their bits) into the database.
1784: * Wrap the entire set of additions in a transaction to make sqlite be a
1785: * little faster.
1.96 schwarze 1786: * Also, handle escape sequences at the last possible moment.
1.14 schwarze 1787: */
1788: static void
1.105 schwarze 1789: dbadd(const struct mpage *mpage, struct mchars *mc)
1.14 schwarze 1790: {
1.87 schwarze 1791: struct mlink *mlink;
1.50 kristaps 1792: struct str *key;
1793: int64_t recno;
1.52 kristaps 1794: size_t i;
1.66 schwarze 1795: unsigned int slot;
1.14 schwarze 1796:
1.56 schwarze 1797: if (verb)
1.105 schwarze 1798: say(mpage->mlinks->file, "Adding to database");
1.43 kristaps 1799:
1.50 kristaps 1800: if (nodb)
1.14 schwarze 1801: return;
1.28 kristaps 1802:
1.52 kristaps 1803: i = 1;
1.82 schwarze 1804: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
1.81 schwarze 1805: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.50 kristaps 1806: recno = sqlite3_last_insert_rowid(db);
1.81 schwarze 1807: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1808:
1.87 schwarze 1809: for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1810: i = 1;
1811: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1812: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1.88 schwarze 1813: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.87 schwarze 1814: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, recno);
1815: SQL_STEP(stmts[STMT_INSERT_LINK]);
1816: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1817: }
1.50 kristaps 1818:
1.66 schwarze 1819: for (key = ohash_first(&strings, &slot); NULL != key;
1820: key = ohash_next(&strings, &slot)) {
1.78 schwarze 1821: assert(key->mpage == mpage);
1.96 schwarze 1822: if (NULL == key->rendered)
1823: render_key(mc, key);
1.52 kristaps 1824: i = 1;
1825: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.96 schwarze 1826: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
1.52 kristaps 1827: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
1828: SQL_STEP(stmts[STMT_INSERT_KEY]);
1.50 kristaps 1829: sqlite3_reset(stmts[STMT_INSERT_KEY]);
1.96 schwarze 1830: if (key->rendered != key->key)
1831: free(key->rendered);
1.66 schwarze 1832: free(key);
1.38 schwarze 1833: }
1.14 schwarze 1834: }
1835:
1.5 kristaps 1836: static void
1.59 schwarze 1837: dbprune(void)
1.5 kristaps 1838: {
1.78 schwarze 1839: struct mpage *mpage;
1.82 schwarze 1840: struct mlink *mlink;
1.52 kristaps 1841: size_t i;
1.80 schwarze 1842: unsigned int slot;
1.5 kristaps 1843:
1.106 schwarze 1844: if (0 == nodb)
1845: SQL_EXEC("BEGIN TRANSACTION");
1.12 schwarze 1846:
1.106 schwarze 1847: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
1848: mpage = ohash_next(&mpages, &slot)) {
1.82 schwarze 1849: mlink = mpage->mlinks;
1.56 schwarze 1850: if (verb)
1.106 schwarze 1851: say(mlink->file, "Deleting from database");
1852: if (nodb)
1853: continue;
1854: for ( ; NULL != mlink; mlink = mlink->next) {
1855: i = 1;
1856: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1857: i, mlink->dsec);
1858: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1859: i, mlink->arch);
1860: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1861: i, mlink->name);
1862: SQL_STEP(stmts[STMT_DELETE_PAGE]);
1863: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
1864: }
1.5 kristaps 1865: }
1.106 schwarze 1866:
1867: if (0 == nodb)
1868: SQL_EXEC("END TRANSACTION");
1.5 kristaps 1869: }
1870:
1.4 kristaps 1871: /*
1.50 kristaps 1872: * Close an existing database and its prepared statements.
1873: * If "real" is not set, rename the temporary file into the real one.
1.4 kristaps 1874: */
1.35 kristaps 1875: static void
1.59 schwarze 1876: dbclose(int real)
1.4 kristaps 1877: {
1.50 kristaps 1878: size_t i;
1.115 ! schwarze 1879: int status;
! 1880: pid_t child;
1.4 kristaps 1881:
1.50 kristaps 1882: if (nodb)
1.38 schwarze 1883: return;
1.50 kristaps 1884:
1885: for (i = 0; i < STMT__MAX; i++) {
1886: sqlite3_finalize(stmts[i]);
1887: stmts[i] = NULL;
1.4 kristaps 1888: }
1889:
1.50 kristaps 1890: sqlite3_close(db);
1891: db = NULL;
1.12 schwarze 1892:
1.50 kristaps 1893: if (real)
1894: return;
1.12 schwarze 1895:
1.115 ! schwarze 1896: if ('\0' == *tempfilename) {
! 1897: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
! 1898: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1899: say(MANDOC_DB, "%s", strerror(errno));
! 1900: }
! 1901: return;
! 1902: }
! 1903:
! 1904: switch (child = fork()) {
! 1905: case (-1):
! 1906: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1907: say("fork cmp", "%s", strerror(errno));
! 1908: return;
! 1909: case (0):
! 1910: execlp("cmp", "cmp", "-s",
! 1911: tempfilename, MANDOC_DB, NULL);
! 1912: say("exec cmp", "%s", strerror(errno));
! 1913: exit(0);
! 1914: default:
! 1915: break;
! 1916: }
! 1917: if (-1 == waitpid(child, &status, 0)) {
! 1918: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1919: say("wait cmp", "%s", strerror(errno));
! 1920: } else if (WIFSIGNALED(status)) {
! 1921: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1922: say("cmp", "Died from a signal");
! 1923: } else if (WEXITSTATUS(status)) {
1.59 schwarze 1924: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 ! schwarze 1925: say(MANDOC_DB,
! 1926: "Data changed, but cannot replace database");
! 1927: }
! 1928:
! 1929: *strrchr(tempfilename, '/') = '\0';
! 1930: switch (child = fork()) {
! 1931: case (-1):
! 1932: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1933: say("fork rm", "%s", strerror(errno));
! 1934: return;
! 1935: case (0):
! 1936: execlp("rm", "rm", "-rf", tempfilename, NULL);
! 1937: say("exec rm", "%s", strerror(errno));
! 1938: exit((int)MANDOCLEVEL_SYSERR);
! 1939: default:
! 1940: break;
! 1941: }
! 1942: if (-1 == waitpid(child, &status, 0)) {
! 1943: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1944: say("wait rm", "%s", strerror(errno));
! 1945: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
! 1946: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1947: say(tempfilename,
! 1948: "Cannot remove temporary directory");
1.59 schwarze 1949: }
1.50 kristaps 1950: }
1.14 schwarze 1951:
1.50 kristaps 1952: /*
1953: * This is straightforward stuff.
1954: * Open a database connection to a "temporary" database, then open a set
1955: * of prepared statements we'll use over and over again.
1956: * If "real" is set, we use the existing database; if not, we truncate a
1957: * temporary one.
1958: * Must be matched by dbclose().
1959: */
1960: static int
1.59 schwarze 1961: dbopen(int real)
1.50 kristaps 1962: {
1.115 ! schwarze 1963: const char *sql;
1.50 kristaps 1964: int rc, ofl;
1.12 schwarze 1965:
1.50 kristaps 1966: if (nodb)
1967: return(1);
1.12 schwarze 1968:
1.115 ! schwarze 1969: *tempfilename = '\0';
1.63 schwarze 1970: ofl = SQLITE_OPEN_READWRITE;
1.115 ! schwarze 1971:
! 1972: if (real) {
! 1973: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
! 1974: if (SQLITE_OK != rc) {
1.63 schwarze 1975: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 ! schwarze 1976: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.63 schwarze 1977: return(0);
1978: }
1.115 ! schwarze 1979: goto prepare_statements;
! 1980: }
! 1981:
! 1982: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.45 kristaps 1983:
1.115 ! schwarze 1984: remove(MANDOC_DB "~");
! 1985: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.50 kristaps 1986: if (SQLITE_OK == rc)
1.115 ! schwarze 1987: goto create_tables;
! 1988: if (quick) {
1.59 schwarze 1989: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 ! schwarze 1990: say(MANDOC_DB "~", "%s", sqlite3_errmsg(db));
1.50 kristaps 1991: return(0);
1992: }
1.12 schwarze 1993:
1.115 ! schwarze 1994: if (strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
! 1995: sizeof(tempfilename)) >= sizeof(tempfilename)) {
! 1996: exitcode = (int)MANDOCLEVEL_SYSERR;
! 1997: say("/tmp/mandocdb.XXXXXX", "Filename too long");
! 1998: return(0);
! 1999: }
! 2000: if (NULL == mkdtemp(tempfilename)) {
! 2001: exitcode = (int)MANDOCLEVEL_SYSERR;
! 2002: say(tempfilename, "%s", strerror(errno));
! 2003: return(0);
! 2004: }
! 2005: if (strlcat(tempfilename, "/" MANDOC_DB,
! 2006: sizeof(tempfilename)) >= sizeof(tempfilename)) {
! 2007: exitcode = (int)MANDOCLEVEL_SYSERR;
! 2008: say(tempfilename, "Filename too long");
! 2009: return(0);
! 2010: }
! 2011: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
! 2012: if (SQLITE_OK != rc) {
1.59 schwarze 2013: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 ! schwarze 2014: say(tempfilename, "%s", sqlite3_errmsg(db));
1.50 kristaps 2015: return(0);
2016: }
1.12 schwarze 2017:
1.115 ! schwarze 2018: create_tables:
1.81 schwarze 2019: sql = "CREATE TABLE \"mpages\" (\n"
1.50 kristaps 2020: " \"form\" INTEGER NOT NULL,\n"
2021: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2022: ");\n"
2023: "\n"
1.81 schwarze 2024: "CREATE TABLE \"mlinks\" (\n"
2025: " \"sec\" TEXT NOT NULL,\n"
2026: " \"arch\" TEXT NOT NULL,\n"
2027: " \"name\" TEXT NOT NULL,\n"
2028: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1.109 schwarze 2029: "ON DELETE CASCADE\n"
1.81 schwarze 2030: ");\n"
2031: "\n"
1.50 kristaps 2032: "CREATE TABLE \"keys\" (\n"
2033: " \"bits\" INTEGER NOT NULL,\n"
2034: " \"key\" TEXT NOT NULL,\n"
1.81 schwarze 2035: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1.109 schwarze 2036: "ON DELETE CASCADE\n"
1.108 schwarze 2037: ");\n";
1.14 schwarze 2038:
1.50 kristaps 2039: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1.59 schwarze 2040: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 ! schwarze 2041: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.50 kristaps 2042: return(0);
2043: }
1.4 kristaps 2044:
1.57 schwarze 2045: prepare_statements:
2046: SQL_EXEC("PRAGMA foreign_keys = ON");
1.106 schwarze 2047: sql = "DELETE FROM mpages WHERE id IN "
2048: "(SELECT pageid FROM mlinks WHERE "
2049: "sec=? AND arch=? AND name=?)";
1.81 schwarze 2050: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2051: sql = "INSERT INTO mpages "
1.103 schwarze 2052: "(form) VALUES (?)";
1.81 schwarze 2053: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2054: sql = "INSERT INTO mlinks "
1.104 schwarze 2055: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.81 schwarze 2056: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.50 kristaps 2057: sql = "INSERT INTO keys "
1.81 schwarze 2058: "(bits,key,pageid) VALUES (?,?,?)";
1.50 kristaps 2059: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.70 schwarze 2060:
2061: #ifndef __APPLE__
2062: /*
2063: * When opening a new database, we can turn off
2064: * synchronous mode for much better performance.
2065: */
2066:
2067: if (real)
2068: SQL_EXEC("PRAGMA synchronous = OFF");
2069: #endif
2070:
1.50 kristaps 2071: return(1);
2072: }
1.5 kristaps 2073:
1.50 kristaps 2074: static void *
2075: hash_halloc(size_t sz, void *arg)
2076: {
1.12 schwarze 2077:
1.50 kristaps 2078: return(mandoc_calloc(sz, 1));
2079: }
1.12 schwarze 2080:
1.50 kristaps 2081: static void *
2082: hash_alloc(size_t sz, void *arg)
2083: {
1.5 kristaps 2084:
1.50 kristaps 2085: return(mandoc_malloc(sz));
2086: }
1.41 kristaps 2087:
1.50 kristaps 2088: static void
2089: hash_free(void *p, size_t sz, void *arg)
2090: {
1.4 kristaps 2091:
1.50 kristaps 2092: free(p);
1.4 kristaps 2093: }
2094:
1.50 kristaps 2095: static int
1.59 schwarze 2096: set_basedir(const char *targetdir)
1.4 kristaps 2097: {
1.59 schwarze 2098: static char startdir[PATH_MAX];
2099: static int fd;
1.4 kristaps 2100:
1.59 schwarze 2101: /*
2102: * Remember where we started by keeping a fd open to the origin
2103: * path component: throughout this utility, we chdir() a lot to
2104: * handle relative paths, and by doing this, we can return to
2105: * the starting point.
2106: */
2107: if ('\0' == *startdir) {
2108: if (NULL == getcwd(startdir, PATH_MAX)) {
2109: exitcode = (int)MANDOCLEVEL_SYSERR;
2110: if (NULL != targetdir)
2111: say(".", NULL);
2112: return(0);
2113: }
2114: if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
2115: exitcode = (int)MANDOCLEVEL_SYSERR;
2116: say(startdir, NULL);
2117: return(0);
2118: }
2119: if (NULL == targetdir)
2120: targetdir = startdir;
2121: } else {
2122: if (-1 == fd)
2123: return(0);
2124: if (-1 == fchdir(fd)) {
2125: close(fd);
2126: basedir[0] = '\0';
2127: exitcode = (int)MANDOCLEVEL_SYSERR;
2128: say(startdir, NULL);
2129: return(0);
2130: }
2131: if (NULL == targetdir) {
2132: close(fd);
2133: return(1);
2134: }
2135: }
2136: if (NULL == realpath(targetdir, basedir)) {
2137: basedir[0] = '\0';
2138: exitcode = (int)MANDOCLEVEL_BADARG;
2139: say(targetdir, NULL);
1.50 kristaps 2140: return(0);
1.59 schwarze 2141: } else if (-1 == chdir(basedir)) {
2142: exitcode = (int)MANDOCLEVEL_BADARG;
2143: say("", NULL);
1.50 kristaps 2144: return(0);
1.4 kristaps 2145: }
1.50 kristaps 2146: return(1);
1.56 schwarze 2147: }
2148:
2149: static void
1.59 schwarze 2150: say(const char *file, const char *format, ...)
1.56 schwarze 2151: {
2152: va_list ap;
2153:
1.59 schwarze 2154: if ('\0' != *basedir)
2155: fprintf(stderr, "%s", basedir);
2156: if ('\0' != *basedir && '\0' != *file)
2157: fputs("//", stderr);
1.56 schwarze 2158: if ('\0' != *file)
1.59 schwarze 2159: fprintf(stderr, "%s", file);
1.56 schwarze 2160: fputs(": ", stderr);
1.59 schwarze 2161:
2162: if (NULL == format) {
2163: perror(NULL);
2164: return;
2165: }
1.56 schwarze 2166:
2167: va_start(ap, format);
2168: vfprintf(stderr, format, ap);
2169: va_end(ap);
2170:
2171: fputc('\n', stderr);
1.1 kristaps 2172: }
CVSweb