Annotation of src/usr.bin/mandoc/mandocdb.c, Revision 1.50
1.50 ! schwarze 1: /* $Id: mandocdb.c,v 1.49 2013/12/31 19:39:09 schwarze Exp $ */
1.1 schwarze 2: /*
1.47 schwarze 3: * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4: * Copyright (c) 2011, 2012, 2013 Ingo Schwarze <schwarze@openbsd.org>
1.1 schwarze 5: *
6: * Permission to use, copy, modify, and distribute this software for any
7: * purpose with or without fee is hereby granted, provided that the above
8: * copyright notice and this permission notice appear in all copies.
9: *
10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17: */
1.47 schwarze 18: #include <sys/stat.h>
1.1 schwarze 19:
20: #include <assert.h>
1.33 schwarze 21: #include <ctype.h>
1.34 schwarze 22: #include <errno.h>
1.1 schwarze 23: #include <fcntl.h>
1.47 schwarze 24: #include <fts.h>
1.1 schwarze 25: #include <getopt.h>
1.44 schwarze 26: #include <limits.h>
1.47 schwarze 27: #include <stddef.h>
1.1 schwarze 28: #include <stdio.h>
29: #include <stdint.h>
30: #include <stdlib.h>
31: #include <string.h>
1.14 schwarze 32: #include <unistd.h>
1.1 schwarze 33:
1.47 schwarze 34: #include <ohash.h>
35: #include <sqlite3.h>
36:
37: #include "mdoc.h"
1.1 schwarze 38: #include "man.h"
39: #include "mandoc.h"
1.10 schwarze 40: #include "manpath.h"
1.47 schwarze 41: #include "mansearch.h"
1.1 schwarze 42:
1.47 schwarze 43: #define SQL_EXEC(_v) \
44: if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
45: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
46: #define SQL_BIND_TEXT(_s, _i, _v) \
47: if (SQLITE_OK != sqlite3_bind_text \
48: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
49: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
50: #define SQL_BIND_INT(_s, _i, _v) \
51: if (SQLITE_OK != sqlite3_bind_int \
52: ((_s), (_i)++, (_v))) \
53: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
54: #define SQL_BIND_INT64(_s, _i, _v) \
55: if (SQLITE_OK != sqlite3_bind_int64 \
56: ((_s), (_i)++, (_v))) \
57: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
58: #define SQL_STEP(_s) \
59: if (SQLITE_DONE != sqlite3_step((_s))) \
60: fprintf(stderr, "%s\n", sqlite3_errmsg(db))
1.1 schwarze 61:
1.47 schwarze 62: enum op {
63: OP_DEFAULT = 0, /* new dbs from dir list or default config */
64: OP_CONFFILE, /* new databases from custom config file */
65: OP_UPDATE, /* delete/add entries in existing database */
66: OP_DELETE, /* delete entries from existing database */
67: OP_TEST /* change no databases, report potential problems */
68: };
1.11 schwarze 69:
1.47 schwarze 70: enum form {
71: FORM_NONE, /* format is unknown */
72: FORM_SRC, /* format is -man or -mdoc */
73: FORM_CAT /* format is cat */
74: };
1.28 schwarze 75:
1.47 schwarze 76: struct str {
77: char *utf8; /* key in UTF-8 form */
78: const struct mpage *mpage; /* if set, the owning parse */
79: uint64_t mask; /* bitmask in sequence */
80: char key[]; /* the string itself */
1.28 schwarze 81: };
82:
1.47 schwarze 83: struct inodev {
84: ino_t st_ino;
85: dev_t st_dev;
86: };
1.28 schwarze 87:
1.47 schwarze 88: struct mpage {
89: struct inodev inodev; /* used for hashing routine */
90: enum form form; /* format from file content */
91: char *sec; /* section from file content */
92: char *arch; /* architecture from file content */
93: char *title; /* title from file content */
94: char *desc; /* description from file content */
95: struct mlink *mlinks; /* singly linked list */
1.28 schwarze 96: };
97:
1.47 schwarze 98: struct mlink {
99: char file[PATH_MAX]; /* filename rel. to manpath */
100: enum form dform; /* format from directory */
101: enum form fform; /* format from file name suffix */
102: char *dsec; /* section from directory */
103: char *arch; /* architecture from directory */
104: char *name; /* name from file name (not empty) */
105: char *fsec; /* section from file name suffix */
106: struct mlink *next; /* singly linked list */
1.2 schwarze 107: };
108:
1.47 schwarze 109: struct title {
110: char *title; /* name(sec/arch) given inside the file */
111: char *file; /* file name in case of mismatch */
1.1 schwarze 112: };
113:
1.47 schwarze 114: enum stmt {
115: STMT_DELETE_PAGE = 0, /* delete mpage */
116: STMT_INSERT_PAGE, /* insert mpage */
117: STMT_INSERT_LINK, /* insert mlink */
118: STMT_INSERT_KEY, /* insert parsed key */
119: STMT__MAX
1.1 schwarze 120: };
121:
1.47 schwarze 122: typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
1.1 schwarze 123:
1.19 schwarze 124: struct mdoc_handler {
1.47 schwarze 125: mdoc_fp fp; /* optional handler */
126: uint64_t mask; /* set unless handler returns 0 */
1.19 schwarze 127: };
128:
1.47 schwarze 129: static void dbclose(int);
130: static void dbindex(const struct mpage *, struct mchars *);
131: static int dbopen(int);
132: static void dbprune(void);
133: static void filescan(const char *);
134: static void *hash_alloc(size_t, void *);
135: static void hash_free(void *, size_t, void *);
136: static void *hash_halloc(size_t, void *);
137: static void mlink_add(struct mlink *, const struct stat *);
1.50 ! schwarze 138: static int mlink_check(struct mpage *, struct mlink *);
1.47 schwarze 139: static void mlink_free(struct mlink *);
140: static void mlinks_undupe(struct mpage *);
141: static void mpages_free(void);
142: static void mpages_merge(struct mchars *, struct mparse *, int);
143: static void parse_cat(struct mpage *);
144: static void parse_man(struct mpage *, const struct man_node *);
145: static void parse_mdoc(struct mpage *, const struct mdoc_node *);
146: static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
147: static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
148: static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
149: static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
150: static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
151: static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
152: static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
153: static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
154: static void putkey(const struct mpage *,
155: const char *, uint64_t);
156: static void putkeys(const struct mpage *,
157: const char *, size_t, uint64_t);
158: static void putmdockey(const struct mpage *,
159: const struct mdoc_node *, uint64_t);
160: static void say(const char *, const char *, ...);
161: static int set_basedir(const char *);
162: static int treescan(void);
163: static size_t utf8(unsigned int, char [7]);
164: static void utf8key(struct mchars *, struct str *);
165:
166: static char *progname;
167: static int use_all; /* use all found files */
168: static int nodb; /* no database changes */
169: static int verb; /* print what we're doing */
170: static int warnings; /* warn about crap */
171: static int exitcode; /* to be returned by main */
172: static enum op op; /* operational mode */
173: static char basedir[PATH_MAX]; /* current base directory */
174: static struct ohash mpages; /* table of distinct manual pages */
175: static struct ohash mlinks; /* table of directory entries */
176: static struct ohash strings; /* table of all strings */
177: static sqlite3 *db = NULL; /* current database */
178: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
179:
1.19 schwarze 180: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.47 schwarze 181: { NULL, 0 }, /* Ap */
182: { NULL, 0 }, /* Dd */
183: { NULL, 0 }, /* Dt */
184: { NULL, 0 }, /* Os */
185: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
186: { parse_mdoc_head, TYPE_Ss }, /* Ss */
187: { NULL, 0 }, /* Pp */
188: { NULL, 0 }, /* D1 */
189: { NULL, 0 }, /* Dl */
190: { NULL, 0 }, /* Bd */
191: { NULL, 0 }, /* Ed */
192: { NULL, 0 }, /* Bl */
193: { NULL, 0 }, /* El */
194: { NULL, 0 }, /* It */
195: { NULL, 0 }, /* Ad */
196: { NULL, TYPE_An }, /* An */
197: { NULL, TYPE_Ar }, /* Ar */
198: { NULL, TYPE_Cd }, /* Cd */
199: { NULL, TYPE_Cm }, /* Cm */
200: { NULL, TYPE_Dv }, /* Dv */
201: { NULL, TYPE_Er }, /* Er */
202: { NULL, TYPE_Ev }, /* Ev */
203: { NULL, 0 }, /* Ex */
204: { NULL, TYPE_Fa }, /* Fa */
205: { parse_mdoc_Fd, 0 }, /* Fd */
206: { NULL, TYPE_Fl }, /* Fl */
207: { parse_mdoc_Fn, 0 }, /* Fn */
208: { NULL, TYPE_Ft }, /* Ft */
209: { NULL, TYPE_Ic }, /* Ic */
1.49 schwarze 210: { NULL, TYPE_In }, /* In */
1.47 schwarze 211: { NULL, TYPE_Li }, /* Li */
212: { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
213: { parse_mdoc_Nm, TYPE_Nm }, /* Nm */
214: { NULL, 0 }, /* Op */
215: { NULL, 0 }, /* Ot */
216: { NULL, TYPE_Pa }, /* Pa */
217: { NULL, 0 }, /* Rv */
1.49 schwarze 218: { NULL, TYPE_St }, /* St */
1.47 schwarze 219: { NULL, TYPE_Va }, /* Va */
220: { parse_mdoc_body, TYPE_Va }, /* Vt */
221: { parse_mdoc_Xr, 0 }, /* Xr */
222: { NULL, 0 }, /* %A */
223: { NULL, 0 }, /* %B */
224: { NULL, 0 }, /* %D */
225: { NULL, 0 }, /* %I */
226: { NULL, 0 }, /* %J */
227: { NULL, 0 }, /* %N */
228: { NULL, 0 }, /* %O */
229: { NULL, 0 }, /* %P */
230: { NULL, 0 }, /* %R */
231: { NULL, 0 }, /* %T */
232: { NULL, 0 }, /* %V */
233: { NULL, 0 }, /* Ac */
234: { NULL, 0 }, /* Ao */
235: { NULL, 0 }, /* Aq */
236: { NULL, TYPE_At }, /* At */
237: { NULL, 0 }, /* Bc */
238: { NULL, 0 }, /* Bf */
239: { NULL, 0 }, /* Bo */
240: { NULL, 0 }, /* Bq */
241: { NULL, TYPE_Bsx }, /* Bsx */
242: { NULL, TYPE_Bx }, /* Bx */
243: { NULL, 0 }, /* Db */
244: { NULL, 0 }, /* Dc */
245: { NULL, 0 }, /* Do */
246: { NULL, 0 }, /* Dq */
247: { NULL, 0 }, /* Ec */
248: { NULL, 0 }, /* Ef */
249: { NULL, TYPE_Em }, /* Em */
250: { NULL, 0 }, /* Eo */
251: { NULL, TYPE_Fx }, /* Fx */
252: { NULL, TYPE_Ms }, /* Ms */
253: { NULL, 0 }, /* No */
254: { NULL, 0 }, /* Ns */
255: { NULL, TYPE_Nx }, /* Nx */
256: { NULL, TYPE_Ox }, /* Ox */
257: { NULL, 0 }, /* Pc */
258: { NULL, 0 }, /* Pf */
259: { NULL, 0 }, /* Po */
260: { NULL, 0 }, /* Pq */
261: { NULL, 0 }, /* Qc */
262: { NULL, 0 }, /* Ql */
263: { NULL, 0 }, /* Qo */
264: { NULL, 0 }, /* Qq */
265: { NULL, 0 }, /* Re */
266: { NULL, 0 }, /* Rs */
267: { NULL, 0 }, /* Sc */
268: { NULL, 0 }, /* So */
269: { NULL, 0 }, /* Sq */
270: { NULL, 0 }, /* Sm */
271: { NULL, 0 }, /* Sx */
272: { NULL, TYPE_Sy }, /* Sy */
273: { NULL, TYPE_Tn }, /* Tn */
274: { NULL, 0 }, /* Ux */
275: { NULL, 0 }, /* Xc */
276: { NULL, 0 }, /* Xo */
277: { parse_mdoc_head, 0 }, /* Fo */
278: { NULL, 0 }, /* Fc */
279: { NULL, 0 }, /* Oo */
280: { NULL, 0 }, /* Oc */
281: { NULL, 0 }, /* Bk */
282: { NULL, 0 }, /* Ek */
283: { NULL, 0 }, /* Bt */
284: { NULL, 0 }, /* Hf */
285: { NULL, 0 }, /* Fr */
286: { NULL, 0 }, /* Ud */
287: { NULL, TYPE_Lb }, /* Lb */
288: { NULL, 0 }, /* Lp */
289: { NULL, TYPE_Lk }, /* Lk */
290: { NULL, TYPE_Mt }, /* Mt */
291: { NULL, 0 }, /* Brq */
292: { NULL, 0 }, /* Bro */
293: { NULL, 0 }, /* Brc */
294: { NULL, 0 }, /* %C */
295: { NULL, 0 }, /* Es */
296: { NULL, 0 }, /* En */
297: { NULL, TYPE_Dx }, /* Dx */
298: { NULL, 0 }, /* %Q */
299: { NULL, 0 }, /* br */
300: { NULL, 0 }, /* sp */
301: { NULL, 0 }, /* %U */
302: { NULL, 0 }, /* Ta */
1.1 schwarze 303: };
304:
305: int
1.3 schwarze 306: mandocdb(int argc, char *argv[])
1.1 schwarze 307: {
1.47 schwarze 308: int ch, i;
309: size_t j, sz;
310: const char *path_arg;
311: struct mchars *mc;
312: struct manpaths dirs;
313: struct mparse *mp;
314: struct ohash_info mpages_info, mlinks_info;
315:
316: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
317: memset(&dirs, 0, sizeof(struct manpaths));
318:
319: mpages_info.alloc = mlinks_info.alloc = hash_alloc;
320: mpages_info.halloc = mlinks_info.halloc = hash_halloc;
321: mpages_info.hfree = mlinks_info.hfree = hash_free;
322:
323: mpages_info.key_offset = offsetof(struct mpage, inodev);
324: mlinks_info.key_offset = offsetof(struct mlink, file);
1.1 schwarze 325:
326: progname = strrchr(argv[0], '/');
327: if (progname == NULL)
328: progname = argv[0];
329: else
330: ++progname;
331:
1.47 schwarze 332: /*
333: * We accept a few different invocations.
334: * The CHECKOP macro makes sure that invocation styles don't
335: * clobber each other.
336: */
337: #define CHECKOP(_op, _ch) do \
338: if (OP_DEFAULT != (_op)) { \
339: fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
340: goto usage; \
341: } while (/*CONSTCOND*/0)
1.10 schwarze 342:
1.47 schwarze 343: path_arg = NULL;
1.28 schwarze 344: op = OP_DEFAULT;
1.1 schwarze 345:
1.47 schwarze 346: while (-1 != (ch = getopt(argc, argv, "aC:d:ntu:vW")))
1.1 schwarze 347: switch (ch) {
1.6 schwarze 348: case ('a'):
349: use_all = 1;
350: break;
1.25 schwarze 351: case ('C'):
1.47 schwarze 352: CHECKOP(op, ch);
353: path_arg = optarg;
1.28 schwarze 354: op = OP_CONFFILE;
1.25 schwarze 355: break;
1.1 schwarze 356: case ('d'):
1.47 schwarze 357: CHECKOP(op, ch);
358: path_arg = optarg;
1.2 schwarze 359: op = OP_UPDATE;
1.1 schwarze 360: break;
1.47 schwarze 361: case ('n'):
362: nodb = 1;
363: break;
1.28 schwarze 364: case ('t'):
1.47 schwarze 365: CHECKOP(op, ch);
1.28 schwarze 366: dup2(STDOUT_FILENO, STDERR_FILENO);
367: op = OP_TEST;
1.47 schwarze 368: nodb = warnings = 1;
1.28 schwarze 369: break;
1.2 schwarze 370: case ('u'):
1.47 schwarze 371: CHECKOP(op, ch);
372: path_arg = optarg;
1.1 schwarze 373: op = OP_DELETE;
374: break;
375: case ('v'):
376: verb++;
377: break;
1.28 schwarze 378: case ('W'):
379: warnings = 1;
380: break;
1.1 schwarze 381: default:
1.28 schwarze 382: goto usage;
1.1 schwarze 383: }
384:
385: argc -= optind;
386: argv += optind;
387:
1.28 schwarze 388: if (OP_CONFFILE == op && argc > 0) {
1.47 schwarze 389: fprintf(stderr, "-C: Too many arguments\n");
1.28 schwarze 390: goto usage;
391: }
392:
1.47 schwarze 393: exitcode = (int)MANDOCLEVEL_OK;
394: mp = mparse_alloc(MPARSE_AUTO,
395: MANDOCLEVEL_FATAL, NULL, NULL, NULL);
396: mc = mchars_alloc();
1.2 schwarze 397:
1.47 schwarze 398: ohash_init(&mpages, 6, &mpages_info);
399: ohash_init(&mlinks, 6, &mlinks_info);
1.2 schwarze 400:
1.47 schwarze 401: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
402: /*
403: * Force processing all files.
404: */
405: use_all = 1;
1.2 schwarze 406:
1.47 schwarze 407: /*
408: * All of these deal with a specific directory.
409: * Jump into that directory then collect files specified
410: * on the command-line.
411: */
412: if (0 == set_basedir(path_arg))
413: goto out;
414: for (i = 0; i < argc; i++)
415: filescan(argv[i]);
416: if (0 == dbopen(1))
417: goto out;
418: if (OP_TEST != op)
419: dbprune();
420: if (OP_DELETE != op)
421: mpages_merge(mc, mp, 0);
422: dbclose(1);
423: } else {
424: /*
425: * If we have arguments, use them as our manpaths.
426: * If we don't, grok from manpath(1) or however else
427: * manpath_parse() wants to do it.
428: */
429: if (argc > 0) {
430: dirs.paths = mandoc_calloc
431: (argc, sizeof(char *));
432: dirs.sz = (size_t)argc;
433: for (i = 0; i < argc; i++)
434: dirs.paths[i] = mandoc_strdup(argv[i]);
435: } else
436: manpath_parse(&dirs, path_arg, NULL, NULL);
1.2 schwarze 437:
1.47 schwarze 438: /*
439: * First scan the tree rooted at a base directory, then
440: * build a new database and finally move it into place.
441: * Ignore zero-length directories and strip trailing
442: * slashes.
443: */
444: for (j = 0; j < dirs.sz; j++) {
445: sz = strlen(dirs.paths[j]);
446: if (sz && '/' == dirs.paths[j][sz - 1])
447: dirs.paths[j][--sz] = '\0';
448: if (0 == sz)
449: continue;
1.2 schwarze 450:
1.47 schwarze 451: if (j) {
452: ohash_init(&mpages, 6, &mpages_info);
453: ohash_init(&mlinks, 6, &mlinks_info);
454: }
1.1 schwarze 455:
1.47 schwarze 456: if (0 == set_basedir(dirs.paths[j]))
457: goto out;
458: if (0 == treescan())
459: goto out;
460: if (0 == set_basedir(dirs.paths[j]))
461: goto out;
462: if (0 == dbopen(0))
463: goto out;
1.41 deraadt 464:
1.47 schwarze 465: mpages_merge(mc, mp, warnings && !use_all);
466: dbclose(0);
1.28 schwarze 467:
1.47 schwarze 468: if (j + 1 < dirs.sz) {
469: mpages_free();
470: ohash_delete(&mpages);
471: ohash_delete(&mlinks);
472: }
1.2 schwarze 473: }
1.47 schwarze 474: }
475: out:
476: set_basedir(NULL);
477: manpath_free(&dirs);
478: mchars_free(mc);
479: mparse_free(mp);
480: mpages_free();
481: ohash_delete(&mpages);
482: ohash_delete(&mlinks);
483: return(exitcode);
484: usage:
485: fprintf(stderr, "usage: %s [-anvW] [-C file]\n"
486: " %s [-anvW] dir ...\n"
487: " %s [-nvW] -d dir [file ...]\n"
488: " %s [-nvW] -u dir [file ...]\n"
489: " %s -t file ...\n",
490: progname, progname, progname,
491: progname, progname);
1.1 schwarze 492:
1.47 schwarze 493: return((int)MANDOCLEVEL_BADARG);
494: }
1.1 schwarze 495:
1.47 schwarze 496: /*
497: * Scan a directory tree rooted at "basedir" for manpages.
498: * We use fts(), scanning directory parts along the way for clues to our
499: * section and architecture.
500: *
501: * If use_all has been specified, grok all files.
502: * If not, sanitise paths to the following:
503: *
504: * [./]man*[/<arch>]/<name>.<section>
505: * or
506: * [./]cat<section>[/<arch>]/<name>.0
507: *
508: * TODO: accomodate for multi-language directories.
509: */
510: static int
511: treescan(void)
512: {
513: FTS *f;
514: FTSENT *ff;
515: struct mlink *mlink;
516: int dform;
517: char *fsec;
518: const char *dsec, *arch, *cp, *path;
519: const char *argv[2];
1.1 schwarze 520:
1.47 schwarze 521: argv[0] = ".";
522: argv[1] = (char *)NULL;
1.1 schwarze 523:
1.47 schwarze 524: /*
525: * Walk through all components under the directory, using the
526: * logical descent of files.
527: */
528: f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
529: if (NULL == f) {
530: exitcode = (int)MANDOCLEVEL_SYSERR;
531: say("", NULL);
532: return(0);
533: }
1.2 schwarze 534:
1.47 schwarze 535: dsec = arch = NULL;
536: dform = FORM_NONE;
1.2 schwarze 537:
1.47 schwarze 538: while (NULL != (ff = fts_read(f))) {
539: path = ff->fts_path + 2;
1.15 schwarze 540: /*
1.47 schwarze 541: * If we're a regular file, add an mlink by using the
542: * stored directory data and handling the filename.
1.15 schwarze 543: */
1.47 schwarze 544: if (FTS_F == ff->fts_info) {
545: if (0 == strcmp(path, MANDOC_DB))
546: continue;
547: if ( ! use_all && ff->fts_level < 2) {
548: if (warnings)
549: say(path, "Extraneous file");
550: continue;
551: } else if (NULL == (fsec =
552: strrchr(ff->fts_name, '.'))) {
553: if ( ! use_all) {
554: if (warnings)
555: say(path,
556: "No filename suffix");
557: continue;
558: }
559: } else if (0 == strcmp(++fsec, "html")) {
560: if (warnings)
561: say(path, "Skip html");
562: continue;
563: } else if (0 == strcmp(fsec, "gz")) {
564: if (warnings)
565: say(path, "Skip gz");
566: continue;
567: } else if (0 == strcmp(fsec, "ps")) {
568: if (warnings)
569: say(path, "Skip ps");
570: continue;
571: } else if (0 == strcmp(fsec, "pdf")) {
572: if (warnings)
573: say(path, "Skip pdf");
574: continue;
575: } else if ( ! use_all &&
576: ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
577: (FORM_CAT == dform && strcmp(fsec, "0")))) {
578: if (warnings)
579: say(path, "Wrong filename suffix");
580: continue;
581: } else
582: fsec[-1] = '\0';
583: mlink = mandoc_calloc(1, sizeof(struct mlink));
584: strlcpy(mlink->file, path, sizeof(mlink->file));
585: mlink->dform = dform;
586: if (NULL != dsec)
587: mlink->dsec = mandoc_strdup(dsec);
588: if (NULL != arch)
589: mlink->arch = mandoc_strdup(arch);
590: mlink->name = mandoc_strdup(ff->fts_name);
591: if (NULL != fsec)
592: mlink->fsec = mandoc_strdup(fsec);
593: mlink_add(mlink, ff->fts_statp);
594: continue;
595: } else if (FTS_D != ff->fts_info &&
596: FTS_DP != ff->fts_info) {
597: if (warnings)
598: say(path, "Not a regular file");
599: continue;
600: }
601:
602: switch (ff->fts_level) {
603: case (0):
604: /* Ignore the root directory. */
605: break;
606: case (1):
607: /*
608: * This might contain manX/ or catX/.
609: * Try to infer this from the name.
610: * If we're not in use_all, enforce it.
611: */
612: dsec = NULL;
613: dform = FORM_NONE;
614: cp = ff->fts_name;
615: if (FTS_DP == ff->fts_info)
616: break;
1.15 schwarze 617:
1.47 schwarze 618: if (0 == strncmp(cp, "man", 3)) {
619: dform = FORM_SRC;
620: dsec = cp + 3;
621: } else if (0 == strncmp(cp, "cat", 3)) {
622: dform = FORM_CAT;
623: dsec = cp + 3;
1.26 schwarze 624: }
1.47 schwarze 625:
626: if (NULL != dsec || use_all)
627: break;
628:
629: if (warnings)
630: say(path, "Unknown directory part");
631: fts_set(f, ff, FTS_SKIP);
632: break;
633: case (2):
634: /*
635: * Possibly our architecture.
636: * If we're descending, keep tabs on it.
637: */
638: arch = NULL;
639: if (FTS_DP != ff->fts_info && NULL != dsec)
640: arch = ff->fts_name;
641: break;
642: default:
643: if (FTS_DP == ff->fts_info || use_all)
644: break;
645: if (warnings)
646: say(path, "Extraneous directory part");
647: fts_set(f, ff, FTS_SKIP);
648: break;
1.14 schwarze 649: }
1.47 schwarze 650: }
651:
652: fts_close(f);
653: return(1);
654: }
1.1 schwarze 655:
1.47 schwarze 656: /*
657: * Add a file to the mlinks table.
658: * Do not verify that it's a "valid" looking manpage (we'll do that
659: * later).
660: *
661: * Try to infer the manual section, architecture, and page name from the
662: * path, assuming it looks like
663: *
664: * [./]man*[/<arch>]/<name>.<section>
665: * or
666: * [./]cat<section>[/<arch>]/<name>.0
667: *
668: * See treescan() for the fts(3) version of this.
669: */
670: static void
671: filescan(const char *file)
672: {
673: char buf[PATH_MAX];
674: struct stat st;
675: struct mlink *mlink;
676: char *p, *start;
677:
678: assert(use_all);
679:
680: if (0 == strncmp(file, "./", 2))
681: file += 2;
682:
683: if (NULL == realpath(file, buf)) {
684: exitcode = (int)MANDOCLEVEL_BADARG;
685: say(file, NULL);
686: return;
687: } else if (OP_TEST != op && strstr(buf, basedir) != buf) {
688: exitcode = (int)MANDOCLEVEL_BADARG;
689: say("", "%s: outside base directory", buf);
690: return;
691: } else if (-1 == stat(buf, &st)) {
692: exitcode = (int)MANDOCLEVEL_BADARG;
693: say(file, NULL);
694: return;
695: } else if ( ! (S_IFREG & st.st_mode)) {
696: exitcode = (int)MANDOCLEVEL_BADARG;
697: say(file, "Not a regular file");
698: return;
1.1 schwarze 699: }
1.47 schwarze 700: start = buf + strlen(basedir);
701: mlink = mandoc_calloc(1, sizeof(struct mlink));
702: strlcpy(mlink->file, start, sizeof(mlink->file));
1.1 schwarze 703:
1.10 schwarze 704: /*
1.47 schwarze 705: * First try to guess our directory structure.
706: * If we find a separator, try to look for man* or cat*.
707: * If we find one of these and what's underneath is a directory,
708: * assume it's an architecture.
1.10 schwarze 709: */
1.47 schwarze 710: if (NULL != (p = strchr(start, '/'))) {
711: *p++ = '\0';
712: if (0 == strncmp(start, "man", 3)) {
713: mlink->dform = FORM_SRC;
714: mlink->dsec = mandoc_strdup(start + 3);
715: } else if (0 == strncmp(start, "cat", 3)) {
716: mlink->dform = FORM_CAT;
717: mlink->dsec = mandoc_strdup(start + 3);
718: }
1.10 schwarze 719:
1.47 schwarze 720: start = p;
721: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
722: *p++ = '\0';
723: mlink->arch = mandoc_strdup(start);
724: start = p;
1.41 deraadt 725: }
1.47 schwarze 726: }
1.7 schwarze 727:
1.47 schwarze 728: /*
729: * Now check the file suffix.
730: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
731: */
732: p = strrchr(start, '\0');
733: while (p-- > start && '/' != *p && '.' != *p)
734: /* Loop. */ ;
735:
736: if ('.' == *p) {
737: *p++ = '\0';
738: mlink->fsec = mandoc_strdup(p);
739: }
1.41 deraadt 740:
1.47 schwarze 741: /*
742: * Now try to parse the name.
743: * Use the filename portion of the path.
744: */
745: mlink->name = start;
746: if (NULL != (p = strrchr(start, '/'))) {
747: mlink->name = p + 1;
748: *p = '\0';
749: }
750: mlink->name = mandoc_strdup(mlink->name);
1.28 schwarze 751:
1.47 schwarze 752: mlink_add(mlink, &st);
753: }
1.2 schwarze 754:
1.47 schwarze 755: static void
756: mlink_add(struct mlink *mlink, const struct stat *st)
757: {
758: struct inodev inodev;
759: struct mpage *mpage;
760: unsigned int slot;
761:
762: assert(NULL != mlink->file);
763:
764: if (NULL == mlink->dsec)
765: mlink->dsec = mandoc_strdup("");
766: if (NULL == mlink->arch)
767: mlink->arch = mandoc_strdup("");
768: if (NULL == mlink->name)
769: mlink->name = mandoc_strdup("");
770: if (NULL == mlink->fsec)
771: mlink->fsec = mandoc_strdup("");
772:
773: if ('0' == *mlink->fsec) {
774: free(mlink->fsec);
775: mlink->fsec = mandoc_strdup(mlink->dsec);
776: mlink->fform = FORM_CAT;
777: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
778: mlink->fform = FORM_SRC;
779: else
780: mlink->fform = FORM_NONE;
1.1 schwarze 781:
1.47 schwarze 782: slot = ohash_qlookup(&mlinks, mlink->file);
783: assert(NULL == ohash_find(&mlinks, slot));
784: ohash_insert(&mlinks, slot, mlink);
785:
786: inodev.st_ino = st->st_ino;
787: inodev.st_dev = st->st_dev;
788: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
789: sizeof(struct inodev), inodev.st_ino);
790: mpage = ohash_find(&mpages, slot);
791: if (NULL == mpage) {
792: mpage = mandoc_calloc(1, sizeof(struct mpage));
793: mpage->inodev.st_ino = inodev.st_ino;
794: mpage->inodev.st_dev = inodev.st_dev;
795: ohash_insert(&mpages, slot, mpage);
796: } else
797: mlink->next = mpage->mlinks;
798: mpage->mlinks = mlink;
799: }
1.1 schwarze 800:
1.47 schwarze 801: static void
802: mlink_free(struct mlink *mlink)
803: {
1.1 schwarze 804:
1.47 schwarze 805: free(mlink->dsec);
806: free(mlink->arch);
807: free(mlink->name);
808: free(mlink->fsec);
809: free(mlink);
810: }
1.1 schwarze 811:
1.47 schwarze 812: static void
813: mpages_free(void)
814: {
815: struct mpage *mpage;
816: struct mlink *mlink;
817: unsigned int slot;
818:
819: mpage = ohash_first(&mpages, &slot);
820: while (NULL != mpage) {
821: while (NULL != (mlink = mpage->mlinks)) {
822: mpage->mlinks = mlink->next;
823: mlink_free(mlink);
824: }
825: free(mpage->sec);
826: free(mpage->arch);
827: free(mpage->title);
828: free(mpage->desc);
829: free(mpage);
830: mpage = ohash_next(&mpages, &slot);
831: }
832: }
1.1 schwarze 833:
1.47 schwarze 834: /*
835: * For each mlink to the mpage, check whether the path looks like
836: * it is formatted, and if it does, check whether a source manual
837: * exists by the same name, ignoring the suffix.
838: * If both conditions hold, drop the mlink.
839: */
840: static void
841: mlinks_undupe(struct mpage *mpage)
842: {
843: char buf[PATH_MAX];
844: struct mlink **prev;
845: struct mlink *mlink;
846: char *bufp;
847:
848: mpage->form = FORM_CAT;
849: prev = &mpage->mlinks;
850: while (NULL != (mlink = *prev)) {
851: if (FORM_CAT != mlink->dform) {
852: mpage->form = FORM_NONE;
853: goto nextlink;
854: }
855: if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) {
856: if (warnings)
857: say(mlink->file, "Filename too long");
858: goto nextlink;
1.26 schwarze 859: }
1.47 schwarze 860: bufp = strstr(buf, "cat");
861: assert(NULL != bufp);
862: memcpy(bufp, "man", 3);
863: if (NULL != (bufp = strrchr(buf, '.')))
864: *++bufp = '\0';
865: strlcat(buf, mlink->dsec, PATH_MAX);
866: if (NULL == ohash_find(&mlinks,
867: ohash_qlookup(&mlinks, buf)))
868: goto nextlink;
869: if (warnings)
870: say(mlink->file, "Man source exists: %s", buf);
871: if (use_all)
872: goto nextlink;
873: *prev = mlink->next;
874: mlink_free(mlink);
875: continue;
876: nextlink:
877: prev = &(*prev)->next;
1.1 schwarze 878: }
1.47 schwarze 879: }
1.1 schwarze 880:
1.50 ! schwarze 881: static int
! 882: mlink_check(struct mpage *mpage, struct mlink *mlink)
! 883: {
! 884: int match;
! 885:
! 886: match = 1;
! 887:
! 888: /*
! 889: * Check whether the manual section given in a file
! 890: * agrees with the directory where the file is located.
! 891: * Some manuals have suffixes like (3p) on their
! 892: * section number either inside the file or in the
! 893: * directory name, some are linked into more than one
! 894: * section, like encrypt(1) = makekey(8).
! 895: */
! 896:
! 897: if (FORM_SRC == mpage->form &&
! 898: strcasecmp(mpage->sec, mlink->dsec)) {
! 899: match = 0;
! 900: say(mlink->file, "Section \"%s\" manual in %s directory",
! 901: mpage->sec, mlink->dsec);
! 902: }
! 903:
! 904: /*
! 905: * Manual page directories exist for each kernel
! 906: * architecture as returned by machine(1).
! 907: * However, many manuals only depend on the
! 908: * application architecture as returned by arch(1).
! 909: * For example, some (2/ARM) manuals are shared
! 910: * across the "armish" and "zaurus" kernel
! 911: * architectures.
! 912: * A few manuals are even shared across completely
! 913: * different architectures, for example fdformat(1)
! 914: * on amd64, i386, sparc, and sparc64.
! 915: */
! 916:
! 917: if (strcasecmp(mpage->arch, mlink->arch)) {
! 918: match = 0;
! 919: say(mlink->file, "Architecture \"%s\" manual in "
! 920: "\"%s\" directory", mpage->arch, mlink->arch);
! 921: }
! 922:
! 923: if (strcasecmp(mpage->title, mlink->name))
! 924: match = 0;
! 925:
! 926: return(match);
! 927: }
! 928:
1.47 schwarze 929: /*
930: * Run through the files in the global vector "mpages"
931: * and add them to the database specified in "basedir".
932: *
933: * This handles the parsing scheme itself, using the cues of directory
934: * and filename to determine whether the file is parsable or not.
935: */
936: static void
937: mpages_merge(struct mchars *mc, struct mparse *mp, int check_reachable)
938: {
939: struct ohash title_table;
940: struct ohash_info title_info, str_info;
941: struct mpage *mpage;
1.50 ! schwarze 942: struct mlink *mlink;
1.47 schwarze 943: struct mdoc *mdoc;
944: struct man *man;
945: struct title *title_entry;
946: char *title_str;
947: const char *cp;
948: int match;
949: unsigned int pslot, tslot;
950: enum mandoclevel lvl;
951:
952: str_info.alloc = hash_alloc;
953: str_info.halloc = hash_halloc;
954: str_info.hfree = hash_free;
955: str_info.key_offset = offsetof(struct str, key);
956:
957: if (check_reachable) {
958: title_info.alloc = hash_alloc;
959: title_info.halloc = hash_halloc;
960: title_info.hfree = hash_free;
961: title_info.key_offset = offsetof(struct title, title);
962: ohash_init(&title_table, 6, &title_info);
963: }
964:
965: mpage = ohash_first(&mpages, &pslot);
966: while (NULL != mpage) {
967: mlinks_undupe(mpage);
968: if (NULL == mpage->mlinks) {
969: mpage = ohash_next(&mpages, &pslot);
970: continue;
971: }
1.1 schwarze 972:
1.47 schwarze 973: ohash_init(&strings, 6, &str_info);
974: mparse_reset(mp);
975: mdoc = NULL;
976: man = NULL;
1.11 schwarze 977:
978: /*
1.24 schwarze 979: * Try interpreting the file as mdoc(7) or man(7)
980: * source code, unless it is already known to be
981: * formatted. Fall back to formatted mode.
1.11 schwarze 982: */
1.47 schwarze 983: if (FORM_CAT != mpage->mlinks->dform ||
984: FORM_CAT != mpage->mlinks->fform) {
985: lvl = mparse_readfd(mp, -1, mpage->mlinks->file);
986: if (lvl < MANDOCLEVEL_FATAL)
987: mparse_result(mp, &mdoc, &man);
988: }
1.11 schwarze 989:
990: if (NULL != mdoc) {
1.47 schwarze 991: mpage->form = FORM_SRC;
992: mpage->sec =
993: mandoc_strdup(mdoc_meta(mdoc)->msec);
994: mpage->arch = mdoc_meta(mdoc)->arch;
995: mpage->arch = mandoc_strdup(
996: NULL == mpage->arch ? "" : mpage->arch);
997: mpage->title =
998: mandoc_strdup(mdoc_meta(mdoc)->title);
1.11 schwarze 999: } else if (NULL != man) {
1.47 schwarze 1000: mpage->form = FORM_SRC;
1001: mpage->sec =
1002: mandoc_strdup(man_meta(man)->msec);
1003: mpage->arch =
1004: mandoc_strdup(mpage->mlinks->arch);
1005: mpage->title =
1006: mandoc_strdup(man_meta(man)->title);
1.11 schwarze 1007: } else {
1.47 schwarze 1008: mpage->form = FORM_CAT;
1009: mpage->sec =
1010: mandoc_strdup(mpage->mlinks->dsec);
1011: mpage->arch =
1012: mandoc_strdup(mpage->mlinks->arch);
1013: mpage->title =
1014: mandoc_strdup(mpage->mlinks->name);
1.1 schwarze 1015: }
1016:
1.50 ! schwarze 1017: for (mlink = mpage->mlinks; mlink; mlink = mlink->next)
! 1018: putkey(mpage, mlink->name, TYPE_Nm);
1.41 deraadt 1019:
1.50 ! schwarze 1020: if (warnings && !use_all) {
1.47 schwarze 1021: match = 0;
1.50 ! schwarze 1022: for (mlink = mpage->mlinks; mlink;
! 1023: mlink = mlink->next)
! 1024: if (mlink_check(mpage, mlink))
! 1025: match = 1;
! 1026: } else
! 1027: match = 1;
1.6 schwarze 1028:
1.47 schwarze 1029: if (NULL != mdoc) {
1030: if (NULL != (cp = mdoc_meta(mdoc)->name))
1031: putkey(mpage, cp, TYPE_Nm);
1032: assert(NULL == mpage->desc);
1033: parse_mdoc(mpage, mdoc_node(mdoc));
1034: putkey(mpage, NULL != mpage->desc ?
1035: mpage->desc : mpage->mlinks->name, TYPE_Nd);
1036: } else if (NULL != man)
1037: parse_man(mpage, man_node(man));
1038: else
1039: parse_cat(mpage);
1.6 schwarze 1040:
1.37 schwarze 1041: /*
1042: * Build a title string for the file. If it matches
1043: * the location of the file, remember the title as
1044: * found; else, remember it as missing.
1045: */
1046:
1.47 schwarze 1047: if (check_reachable) {
1048: if (-1 == asprintf(&title_str, "%s(%s%s%s)",
1049: mpage->title, mpage->sec,
1050: '\0' == *mpage->arch ? "" : "/",
1051: mpage->arch)) {
1052: perror(NULL);
1053: exit((int)MANDOCLEVEL_SYSERR);
1.37 schwarze 1054: }
1.47 schwarze 1055: tslot = ohash_qlookup(&title_table, title_str);
1056: title_entry = ohash_find(&title_table, tslot);
1057: if (NULL == title_entry) {
1058: title_entry = mandoc_malloc(
1059: sizeof(struct title));
1060: title_entry->title = title_str;
1061: title_entry->file = mandoc_strdup(
1062: match ? "" : mpage->mlinks->file);
1063: ohash_insert(&title_table, tslot,
1064: title_entry);
1065: } else {
1066: if (match)
1067: *title_entry->file = '\0';
1068: free(title_str);
1.37 schwarze 1069: }
1070: }
1071:
1.47 schwarze 1072: dbindex(mpage, mc);
1073: ohash_delete(&strings);
1074: mpage = ohash_next(&mpages, &pslot);
1075: }
1076:
1077: if (check_reachable) {
1078: title_entry = ohash_first(&title_table, &tslot);
1079: while (NULL != title_entry) {
1080: if ('\0' != *title_entry->file)
1081: say(title_entry->file,
1082: "Probably unreachable, title is %s",
1083: title_entry->title);
1084: free(title_entry->title);
1085: free(title_entry->file);
1086: free(title_entry);
1087: title_entry = ohash_next(&title_table, &tslot);
1088: }
1089: ohash_delete(&title_table);
1090: }
1091: }
1.6 schwarze 1092:
1.47 schwarze 1093: static void
1094: parse_cat(struct mpage *mpage)
1095: {
1096: FILE *stream;
1097: char *line, *p, *title;
1098: size_t len, plen, titlesz;
1.1 schwarze 1099:
1.47 schwarze 1100: if (NULL == (stream = fopen(mpage->mlinks->file, "r"))) {
1101: if (warnings)
1102: say(mpage->mlinks->file, NULL);
1103: return;
1104: }
1.1 schwarze 1105:
1.47 schwarze 1106: /* Skip to first blank line. */
1.1 schwarze 1107:
1.47 schwarze 1108: while (NULL != (line = fgetln(stream, &len)))
1109: if ('\n' == *line)
1110: break;
1.1 schwarze 1111:
1.47 schwarze 1112: /*
1113: * Assume the first line that is not indented
1114: * is the first section header. Skip to it.
1115: */
1.1 schwarze 1116:
1.47 schwarze 1117: while (NULL != (line = fgetln(stream, &len)))
1118: if ('\n' != *line && ' ' != *line)
1119: break;
1120:
1121: /*
1122: * Read up until the next section into a buffer.
1123: * Strip the leading and trailing newline from each read line,
1124: * appending a trailing space.
1125: * Ignore empty (whitespace-only) lines.
1126: */
1.28 schwarze 1127:
1.47 schwarze 1128: titlesz = 0;
1129: title = NULL;
1.38 schwarze 1130:
1.47 schwarze 1131: while (NULL != (line = fgetln(stream, &len))) {
1132: if (' ' != *line || '\n' != line[len - 1])
1133: break;
1134: while (len > 0 && isspace((unsigned char)*line)) {
1135: line++;
1136: len--;
1137: }
1138: if (1 == len)
1139: continue;
1140: title = mandoc_realloc(title, titlesz + len);
1141: memcpy(title + titlesz, line, len);
1142: titlesz += len;
1143: title[titlesz - 1] = ' ';
1144: }
1.28 schwarze 1145:
1.47 schwarze 1146: /*
1147: * If no page content can be found, or the input line
1148: * is already the next section header, or there is no
1149: * trailing newline, reuse the page title as the page
1150: * description.
1151: */
1.1 schwarze 1152:
1.47 schwarze 1153: if (NULL == title || '\0' == *title) {
1154: if (warnings)
1155: say(mpage->mlinks->file,
1156: "Cannot find NAME section");
1157: assert(NULL == mpage->desc);
1158: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1159: putkey(mpage, mpage->mlinks->name, TYPE_Nd);
1160: fclose(stream);
1161: free(title);
1162: return;
1163: }
1.24 schwarze 1164:
1.47 schwarze 1165: title = mandoc_realloc(title, titlesz + 1);
1166: title[titlesz] = '\0';
1.24 schwarze 1167:
1.47 schwarze 1168: /*
1169: * Skip to the first dash.
1170: * Use the remaining line as the description (no more than 70
1171: * bytes).
1172: */
1.28 schwarze 1173:
1.47 schwarze 1174: if (NULL != (p = strstr(title, "- "))) {
1175: for (p += 2; ' ' == *p || '\b' == *p; p++)
1176: /* Skip to next word. */ ;
1177: } else {
1178: if (warnings)
1179: say(mpage->mlinks->file,
1180: "No dash in title line");
1181: p = title;
1182: }
1.1 schwarze 1183:
1.47 schwarze 1184: plen = strlen(p);
1.1 schwarze 1185:
1.47 schwarze 1186: /* Strip backspace-encoding from line. */
1.1 schwarze 1187:
1.47 schwarze 1188: while (NULL != (line = memchr(p, '\b', plen))) {
1189: len = line - p;
1190: if (0 == len) {
1191: memmove(line, line + 1, plen--);
1192: continue;
1193: }
1194: memmove(line - 1, line + 1, plen - len);
1195: plen -= 2;
1196: }
1.1 schwarze 1197:
1.47 schwarze 1198: assert(NULL == mpage->desc);
1199: mpage->desc = mandoc_strdup(p);
1200: putkey(mpage, mpage->desc, TYPE_Nd);
1201: fclose(stream);
1202: free(title);
1203: }
1.16 schwarze 1204:
1.47 schwarze 1205: /*
1206: * Put a type/word pair into the word database for this particular file.
1207: */
1208: static void
1209: putkey(const struct mpage *mpage, const char *value, uint64_t type)
1210: {
1.37 schwarze 1211:
1.47 schwarze 1212: assert(NULL != value);
1213: putkeys(mpage, value, strlen(value), type);
1.2 schwarze 1214: }
1215:
1216: /*
1.47 schwarze 1217: * Grok all nodes at or below a certain mdoc node into putkey().
1.2 schwarze 1218: */
1219: static void
1.47 schwarze 1220: putmdockey(const struct mpage *mpage,
1221: const struct mdoc_node *n, uint64_t m)
1.2 schwarze 1222: {
1.16 schwarze 1223:
1.47 schwarze 1224: for ( ; NULL != n; n = n->next) {
1225: if (NULL != n->child)
1226: putmdockey(mpage, n->child, m);
1227: if (MDOC_TEXT == n->type)
1228: putkey(mpage, n->string, m);
1229: }
1230: }
1.16 schwarze 1231:
1.47 schwarze 1232: static void
1233: parse_man(struct mpage *mpage, const struct man_node *n)
1234: {
1235: const struct man_node *head, *body;
1236: char *start, *sv, *title;
1237: char byte;
1238: size_t sz, titlesz;
1.16 schwarze 1239:
1.47 schwarze 1240: if (NULL == n)
1241: return;
1.16 schwarze 1242:
1.47 schwarze 1243: /*
1244: * We're only searching for one thing: the first text child in
1245: * the BODY of a NAME section. Since we don't keep track of
1246: * sections in -man, run some hoops to find out whether we're in
1247: * the correct section or not.
1248: */
1.16 schwarze 1249:
1.47 schwarze 1250: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1251: body = n;
1252: assert(body->parent);
1253: if (NULL != (head = body->parent->head) &&
1254: 1 == head->nchild &&
1255: NULL != (head = (head->child)) &&
1256: MAN_TEXT == head->type &&
1257: 0 == strcmp(head->string, "NAME") &&
1258: NULL != (body = body->child) &&
1259: MAN_TEXT == body->type) {
1.2 schwarze 1260:
1.47 schwarze 1261: title = NULL;
1262: titlesz = 0;
1.2 schwarze 1263:
1.47 schwarze 1264: /*
1265: * Suck the entire NAME section into memory.
1266: * Yes, we might run away.
1267: * But too many manuals have big, spread-out
1268: * NAME sections over many lines.
1269: */
1.2 schwarze 1270:
1.47 schwarze 1271: for ( ; NULL != body; body = body->next) {
1272: if (MAN_TEXT != body->type)
1273: break;
1274: if (0 == (sz = strlen(body->string)))
1275: continue;
1276: title = mandoc_realloc
1277: (title, titlesz + sz + 1);
1278: memcpy(title + titlesz, body->string, sz);
1279: titlesz += sz + 1;
1280: title[titlesz - 1] = ' ';
1281: }
1282: if (NULL == title)
1283: return;
1.16 schwarze 1284:
1.47 schwarze 1285: title = mandoc_realloc(title, titlesz + 1);
1286: title[titlesz] = '\0';
1.16 schwarze 1287:
1.47 schwarze 1288: /* Skip leading space. */
1.16 schwarze 1289:
1.47 schwarze 1290: sv = title;
1291: while (isspace((unsigned char)*sv))
1292: sv++;
1.16 schwarze 1293:
1.47 schwarze 1294: if (0 == (sz = strlen(sv))) {
1295: free(title);
1296: return;
1297: }
1.1 schwarze 1298:
1.47 schwarze 1299: /* Erase trailing space. */
1.1 schwarze 1300:
1.47 schwarze 1301: start = &sv[sz - 1];
1302: while (start > sv && isspace((unsigned char)*start))
1303: *start-- = '\0';
1.1 schwarze 1304:
1.47 schwarze 1305: if (start == sv) {
1306: free(title);
1307: return;
1308: }
1.1 schwarze 1309:
1.47 schwarze 1310: start = sv;
1.16 schwarze 1311:
1.47 schwarze 1312: /*
1313: * Go through a special heuristic dance here.
1314: * Conventionally, one or more manual names are
1315: * comma-specified prior to a whitespace, then a
1316: * dash, then a description. Try to puzzle out
1317: * the name parts here.
1318: */
1.16 schwarze 1319:
1.47 schwarze 1320: for ( ;; ) {
1321: sz = strcspn(start, " ,");
1322: if ('\0' == start[sz])
1323: break;
1.1 schwarze 1324:
1.47 schwarze 1325: byte = start[sz];
1326: start[sz] = '\0';
1.1 schwarze 1327:
1.47 schwarze 1328: putkey(mpage, start, TYPE_Nm);
1.1 schwarze 1329:
1.47 schwarze 1330: if (' ' == byte) {
1331: start += sz + 1;
1332: break;
1333: }
1.1 schwarze 1334:
1.47 schwarze 1335: assert(',' == byte);
1336: start += sz + 1;
1337: while (' ' == *start)
1338: start++;
1339: }
1.1 schwarze 1340:
1.47 schwarze 1341: if (sv == start) {
1342: putkey(mpage, start, TYPE_Nm);
1343: free(title);
1344: return;
1345: }
1.1 schwarze 1346:
1.47 schwarze 1347: while (isspace((unsigned char)*start))
1348: start++;
1.1 schwarze 1349:
1.47 schwarze 1350: if (0 == strncmp(start, "-", 1))
1351: start += 1;
1352: else if (0 == strncmp(start, "\\-\\-", 4))
1353: start += 4;
1354: else if (0 == strncmp(start, "\\-", 2))
1355: start += 2;
1356: else if (0 == strncmp(start, "\\(en", 4))
1357: start += 4;
1358: else if (0 == strncmp(start, "\\(em", 4))
1359: start += 4;
1.1 schwarze 1360:
1.47 schwarze 1361: while (' ' == *start)
1362: start++;
1.1 schwarze 1363:
1.47 schwarze 1364: assert(NULL == mpage->desc);
1365: mpage->desc = mandoc_strdup(start);
1366: putkey(mpage, mpage->desc, TYPE_Nd);
1367: free(title);
1368: return;
1369: }
1370: }
1.1 schwarze 1371:
1.47 schwarze 1372: for (n = n->child; n; n = n->next) {
1373: if (NULL != mpage->desc)
1374: break;
1375: parse_man(mpage, n);
1.1 schwarze 1376: }
1377: }
1378:
1379: static void
1.47 schwarze 1380: parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1381: {
1382:
1.47 schwarze 1383: assert(NULL != n);
1384: for (n = n->child; NULL != n; n = n->next) {
1385: switch (n->type) {
1386: case (MDOC_ELEM):
1387: /* FALLTHROUGH */
1388: case (MDOC_BLOCK):
1389: /* FALLTHROUGH */
1390: case (MDOC_HEAD):
1391: /* FALLTHROUGH */
1392: case (MDOC_BODY):
1393: /* FALLTHROUGH */
1394: case (MDOC_TAIL):
1395: if (NULL != mdocs[n->tok].fp)
1396: if (0 == (*mdocs[n->tok].fp)(mpage, n))
1397: break;
1398: if (mdocs[n->tok].mask)
1399: putmdockey(mpage, n->child,
1400: mdocs[n->tok].mask);
1401: break;
1402: default:
1403: assert(MDOC_ROOT != n->type);
1404: continue;
1405: }
1406: if (NULL != n->child)
1407: parse_mdoc(mpage, n);
1.1 schwarze 1408: }
1409: }
1410:
1.19 schwarze 1411: static int
1.47 schwarze 1412: parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1413: {
1414: const char *start, *end;
1415: size_t sz;
1.19 schwarze 1416:
1.47 schwarze 1417: if (SEC_SYNOPSIS != n->sec ||
1418: NULL == (n = n->child) ||
1419: MDOC_TEXT != n->type)
1.19 schwarze 1420: return(0);
1.1 schwarze 1421:
1422: /*
1423: * Only consider those `Fd' macro fields that begin with an
1424: * "inclusion" token (versus, e.g., #define).
1425: */
1.47 schwarze 1426:
1.1 schwarze 1427: if (strcmp("#include", n->string))
1.19 schwarze 1428: return(0);
1.1 schwarze 1429:
1430: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.19 schwarze 1431: return(0);
1.1 schwarze 1432:
1433: /*
1434: * Strip away the enclosing angle brackets and make sure we're
1435: * not zero-length.
1436: */
1437:
1438: start = n->string;
1439: if ('<' == *start || '"' == *start)
1440: start++;
1441:
1442: if (0 == (sz = strlen(start)))
1.19 schwarze 1443: return(0);
1.1 schwarze 1444:
1445: end = &start[(int)sz - 1];
1446: if ('>' == *end || '"' == *end)
1447: end--;
1448:
1.47 schwarze 1449: if (end > start)
1450: putkeys(mpage, start, end - start + 1, TYPE_In);
1.49 schwarze 1451: return(0);
1.1 schwarze 1452: }
1453:
1.19 schwarze 1454: static int
1.47 schwarze 1455: parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1456: {
1457: const char *cp;
1458:
1.47 schwarze 1459: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.19 schwarze 1460: return(0);
1461:
1.47 schwarze 1462: /*
1463: * Parse: .Fn "struct type *name" "char *arg".
1464: * First strip away pointer symbol.
1465: * Then store the function name, then type.
1466: * Finally, store the arguments.
1467: */
1.1 schwarze 1468:
1.47 schwarze 1469: if (NULL == (cp = strrchr(n->string, ' ')))
1470: cp = n->string;
1.1 schwarze 1471:
1472: while ('*' == *cp)
1473: cp++;
1474:
1.47 schwarze 1475: putkey(mpage, cp, TYPE_Fn);
1.19 schwarze 1476:
1.47 schwarze 1477: if (n->string < cp)
1478: putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1.19 schwarze 1479:
1.47 schwarze 1480: for (n = n->next; NULL != n; n = n->next)
1481: if (MDOC_TEXT == n->type)
1482: putkey(mpage, n->string, TYPE_Fa);
1.19 schwarze 1483:
1484: return(0);
1.1 schwarze 1485: }
1486:
1.19 schwarze 1487: static int
1.47 schwarze 1488: parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1489: {
1.47 schwarze 1490: char *cp;
1.1 schwarze 1491:
1492: if (NULL == (n = n->child))
1.19 schwarze 1493: return(0);
1.1 schwarze 1494:
1.47 schwarze 1495: if (NULL == n->next) {
1496: putkey(mpage, n->string, TYPE_Xr);
1497: return(0);
1498: }
1.1 schwarze 1499:
1.47 schwarze 1500: if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) {
1501: perror(NULL);
1502: exit((int)MANDOCLEVEL_SYSERR);
1503: }
1504: putkey(mpage, cp, TYPE_Xr);
1505: free(cp);
1506: return(0);
1.1 schwarze 1507: }
1508:
1.19 schwarze 1509: static int
1.47 schwarze 1510: parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1511: {
1.47 schwarze 1512: size_t sz;
1.1 schwarze 1513:
1514: if (MDOC_BODY != n->type)
1.19 schwarze 1515: return(0);
1.1 schwarze 1516:
1.47 schwarze 1517: /*
1518: * Special-case the `Nd' because we need to put the description
1519: * into the document table.
1520: */
1521:
1522: for (n = n->child; NULL != n; n = n->next) {
1523: if (MDOC_TEXT == n->type) {
1524: if (NULL != mpage->desc) {
1525: sz = strlen(mpage->desc) +
1526: strlen(n->string) + 2;
1527: mpage->desc = mandoc_realloc(
1528: mpage->desc, sz);
1529: strlcat(mpage->desc, " ", sz);
1530: strlcat(mpage->desc, n->string, sz);
1531: } else
1532: mpage->desc = mandoc_strdup(n->string);
1533: }
1534: if (NULL != n->child)
1535: parse_mdoc_Nd(mpage, n);
1536: }
1.19 schwarze 1537: return(1);
1.1 schwarze 1538: }
1539:
1.19 schwarze 1540: static int
1.47 schwarze 1541: parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1542: {
1543:
1.49 schwarze 1544: return(SEC_NAME == n->sec ||
1545: (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type));
1.1 schwarze 1546: }
1547:
1.19 schwarze 1548: static int
1.47 schwarze 1549: parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1550: {
1551:
1.19 schwarze 1552: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 schwarze 1553: }
1554:
1.47 schwarze 1555: static int
1556: parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1.1 schwarze 1557: {
1558:
1.47 schwarze 1559: return(MDOC_HEAD == n->type);
1560: }
1.1 schwarze 1561:
1.47 schwarze 1562: static int
1563: parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1564: {
1.1 schwarze 1565:
1.47 schwarze 1566: return(MDOC_BODY == n->type);
1.1 schwarze 1567: }
1568:
1.47 schwarze 1569: /*
1570: * Add a string to the hash table for the current manual.
1571: * Each string has a bitmask telling which macros it belongs to.
1572: * When we finish the manual, we'll dump the table.
1573: */
1.1 schwarze 1574: static void
1.47 schwarze 1575: putkeys(const struct mpage *mpage,
1576: const char *cp, size_t sz, uint64_t v)
1.1 schwarze 1577: {
1.47 schwarze 1578: struct str *s;
1579: unsigned int slot;
1580: const char *end;
1.1 schwarze 1581:
1.47 schwarze 1582: if (0 == sz)
1583: return;
1584:
1585: end = cp + sz;
1586: slot = ohash_qlookupi(&strings, cp, &end);
1587: s = ohash_find(&strings, slot);
1.1 schwarze 1588:
1.47 schwarze 1589: if (NULL != s && mpage == s->mpage) {
1590: s->mask |= v;
1.1 schwarze 1591: return;
1.47 schwarze 1592: } else if (NULL == s) {
1593: s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1594: memcpy(s->key, cp, sz);
1595: ohash_insert(&strings, slot, s);
1596: }
1597: s->mpage = mpage;
1598: s->mask = v;
1.1 schwarze 1599: }
1600:
1601: /*
1.47 schwarze 1602: * Take a Unicode codepoint and produce its UTF-8 encoding.
1603: * This isn't the best way to do this, but it works.
1604: * The magic numbers are from the UTF-8 packaging.
1605: * They're not as scary as they seem: read the UTF-8 spec for details.
1.1 schwarze 1606: */
1.47 schwarze 1607: static size_t
1608: utf8(unsigned int cp, char out[7])
1.1 schwarze 1609: {
1.47 schwarze 1610: size_t rc;
1.1 schwarze 1611:
1.47 schwarze 1612: rc = 0;
1613: if (cp <= 0x0000007F) {
1614: rc = 1;
1615: out[0] = (char)cp;
1616: } else if (cp <= 0x000007FF) {
1617: rc = 2;
1618: out[0] = (cp >> 6 & 31) | 192;
1619: out[1] = (cp & 63) | 128;
1620: } else if (cp <= 0x0000FFFF) {
1621: rc = 3;
1622: out[0] = (cp >> 12 & 15) | 224;
1623: out[1] = (cp >> 6 & 63) | 128;
1624: out[2] = (cp & 63) | 128;
1625: } else if (cp <= 0x001FFFFF) {
1626: rc = 4;
1627: out[0] = (cp >> 18 & 7) | 240;
1628: out[1] = (cp >> 12 & 63) | 128;
1629: out[2] = (cp >> 6 & 63) | 128;
1630: out[3] = (cp & 63) | 128;
1631: } else if (cp <= 0x03FFFFFF) {
1632: rc = 5;
1633: out[0] = (cp >> 24 & 3) | 248;
1634: out[1] = (cp >> 18 & 63) | 128;
1635: out[2] = (cp >> 12 & 63) | 128;
1636: out[3] = (cp >> 6 & 63) | 128;
1637: out[4] = (cp & 63) | 128;
1638: } else if (cp <= 0x7FFFFFFF) {
1639: rc = 6;
1640: out[0] = (cp >> 30 & 1) | 252;
1641: out[1] = (cp >> 24 & 63) | 128;
1642: out[2] = (cp >> 18 & 63) | 128;
1643: out[3] = (cp >> 12 & 63) | 128;
1644: out[4] = (cp >> 6 & 63) | 128;
1645: out[5] = (cp & 63) | 128;
1646: } else
1647: return(0);
1.19 schwarze 1648:
1.47 schwarze 1649: out[rc] = '\0';
1650: return(rc);
1.1 schwarze 1651: }
1652:
1.47 schwarze 1653: /*
1654: * Store the UTF-8 version of a key, or alias the pointer if the key has
1655: * no UTF-8 transcription marks in it.
1656: */
1657: static void
1658: utf8key(struct mchars *mc, struct str *key)
1.1 schwarze 1659: {
1.47 schwarze 1660: size_t sz, bsz, pos;
1661: char utfbuf[7], res[5];
1662: char *buf;
1663: const char *seq, *cpp, *val;
1664: int len, u;
1665: enum mandoc_esc esc;
1666:
1667: assert(NULL == key->utf8);
1668:
1669: res[0] = '\\';
1670: res[1] = '\t';
1671: res[2] = ASCII_NBRSP;
1672: res[3] = ASCII_HYPH;
1673: res[4] = '\0';
1.1 schwarze 1674:
1.47 schwarze 1675: val = key->key;
1676: bsz = strlen(val);
1.1 schwarze 1677:
1678: /*
1.47 schwarze 1679: * Pre-check: if we have no stop-characters, then set the
1680: * pointer as ourselvse and get out of here.
1.1 schwarze 1681: */
1.47 schwarze 1682: if (strcspn(val, res) == bsz) {
1683: key->utf8 = key->key;
1684: return;
1685: }
1.1 schwarze 1686:
1.47 schwarze 1687: /* Pre-allocate by the length of the input */
1.39 schwarze 1688:
1.47 schwarze 1689: buf = mandoc_malloc(++bsz);
1690: pos = 0;
1.39 schwarze 1691:
1.47 schwarze 1692: while ('\0' != *val) {
1693: /*
1694: * Halt on the first escape sequence.
1695: * This also halts on the end of string, in which case
1696: * we just copy, fallthrough, and exit the loop.
1697: */
1698: if ((sz = strcspn(val, res)) > 0) {
1699: memcpy(&buf[pos], val, sz);
1700: pos += sz;
1701: val += sz;
1702: }
1.39 schwarze 1703:
1.47 schwarze 1704: if (ASCII_HYPH == *val) {
1705: buf[pos++] = '-';
1706: val++;
1707: continue;
1708: } else if ('\t' == *val || ASCII_NBRSP == *val) {
1709: buf[pos++] = ' ';
1710: val++;
1711: continue;
1712: } else if ('\\' != *val)
1713: break;
1.39 schwarze 1714:
1.47 schwarze 1715: /* Read past the slash. */
1.39 schwarze 1716:
1.47 schwarze 1717: val++;
1.39 schwarze 1718:
1.47 schwarze 1719: /*
1720: * Parse the escape sequence and see if it's a
1721: * predefined character or special character.
1722: */
1723: esc = mandoc_escape
1724: ((const char **)&val, &seq, &len);
1725: if (ESCAPE_ERROR == esc)
1726: break;
1.39 schwarze 1727:
1.47 schwarze 1728: if (ESCAPE_SPECIAL != esc)
1729: continue;
1730: if (0 == (u = mchars_spec2cp(mc, seq, len)))
1731: continue;
1.39 schwarze 1732:
1.47 schwarze 1733: /*
1734: * If we have a Unicode codepoint, try to convert that
1735: * to a UTF-8 byte string.
1736: */
1737: cpp = utfbuf;
1738: if (0 == (sz = utf8(u, utfbuf)))
1739: continue;
1.1 schwarze 1740:
1.47 schwarze 1741: /* Copy the rendered glyph into the stream. */
1.1 schwarze 1742:
1.47 schwarze 1743: sz = strlen(cpp);
1744: bsz += sz;
1.1 schwarze 1745:
1.47 schwarze 1746: buf = mandoc_realloc(buf, bsz);
1.1 schwarze 1747:
1.47 schwarze 1748: memcpy(&buf[pos], cpp, sz);
1749: pos += sz;
1.1 schwarze 1750: }
1751:
1.47 schwarze 1752: buf[pos] = '\0';
1753: key->utf8 = buf;
1.1 schwarze 1754: }
1755:
1.11 schwarze 1756: /*
1.47 schwarze 1757: * Flush the current page's terms (and their bits) into the database.
1758: * Wrap the entire set of additions in a transaction to make sqlite be a
1759: * little faster.
1760: * Also, UTF-8-encode the description at the last possible moment.
1.11 schwarze 1761: */
1762: static void
1.47 schwarze 1763: dbindex(const struct mpage *mpage, struct mchars *mc)
1.11 schwarze 1764: {
1.47 schwarze 1765: struct mlink *mlink;
1766: struct str *key;
1767: const char *desc;
1768: int64_t recno;
1769: size_t i;
1770: unsigned int slot;
1771:
1772: if (verb)
1773: say(mpage->mlinks->file, "Adding to index");
1.11 schwarze 1774:
1.47 schwarze 1775: if (nodb)
1.11 schwarze 1776: return;
1.47 schwarze 1777:
1778: desc = "";
1779: if (NULL != mpage->desc && '\0' != *mpage->desc) {
1780: key = ohash_find(&strings,
1781: ohash_qlookup(&strings, mpage->desc));
1782: assert(NULL != key);
1783: if (NULL == key->utf8)
1784: utf8key(mc, key);
1785: desc = key->utf8;
1.11 schwarze 1786: }
1787:
1.47 schwarze 1788: SQL_EXEC("BEGIN TRANSACTION");
1.22 schwarze 1789:
1.47 schwarze 1790: i = 1;
1.22 schwarze 1791: /*
1.47 schwarze 1792: * XXX The following three lines are obsolete
1793: * and only kept for backward compatibility
1794: * until apropos(1) and friends have caught up.
1.22 schwarze 1795: */
1.47 schwarze 1796: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->file);
1797: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->dsec);
1798: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->arch);
1799: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, desc);
1800: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
1801: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1802: recno = sqlite3_last_insert_rowid(db);
1803: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1804:
1805: for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1806: i = 1;
1807: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->file);
1808: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1809: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1810: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1811: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, recno);
1812: SQL_STEP(stmts[STMT_INSERT_LINK]);
1813: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1814: }
1815:
1816: for (key = ohash_first(&strings, &slot); NULL != key;
1817: key = ohash_next(&strings, &slot)) {
1818: assert(key->mpage == mpage);
1819: if (NULL == key->utf8)
1820: utf8key(mc, key);
1821: i = 1;
1822: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1823: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->utf8);
1824: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
1825: SQL_STEP(stmts[STMT_INSERT_KEY]);
1826: sqlite3_reset(stmts[STMT_INSERT_KEY]);
1827: if (key->utf8 != key->key)
1828: free(key->utf8);
1829: free(key);
1.33 schwarze 1830: }
1831:
1.47 schwarze 1832: SQL_EXEC("END TRANSACTION");
1833: }
1.41 deraadt 1834:
1.47 schwarze 1835: static void
1836: dbprune(void)
1837: {
1838: struct mpage *mpage;
1839: struct mlink *mlink;
1840: size_t i;
1841: unsigned int slot;
1.11 schwarze 1842:
1.47 schwarze 1843: if (nodb)
1.11 schwarze 1844: return;
1.47 schwarze 1845:
1846: mpage = ohash_first(&mpages, &slot);
1847: while (NULL != mpage) {
1848: mlink = mpage->mlinks;
1849: i = 1;
1850: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE], i, mlink->file);
1851: SQL_STEP(stmts[STMT_DELETE_PAGE]);
1852: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
1853: if (verb)
1854: say(mlink->file, "Deleted from index");
1855: mpage = ohash_next(&mpages, &slot);
1.11 schwarze 1856: }
1.47 schwarze 1857: }
1.22 schwarze 1858:
1.47 schwarze 1859: /*
1860: * Close an existing database and its prepared statements.
1861: * If "real" is not set, rename the temporary file into the real one.
1862: */
1863: static void
1864: dbclose(int real)
1865: {
1866: size_t i;
1.11 schwarze 1867:
1.47 schwarze 1868: if (nodb)
1869: return;
1.11 schwarze 1870:
1.47 schwarze 1871: for (i = 0; i < STMT__MAX; i++) {
1872: sqlite3_finalize(stmts[i]);
1873: stmts[i] = NULL;
1.28 schwarze 1874: }
1.22 schwarze 1875:
1.47 schwarze 1876: sqlite3_close(db);
1877: db = NULL;
1.11 schwarze 1878:
1.47 schwarze 1879: if (real)
1880: return;
1.22 schwarze 1881:
1.47 schwarze 1882: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
1883: exitcode = (int)MANDOCLEVEL_SYSERR;
1884: say(MANDOC_DB, NULL);
1.22 schwarze 1885: }
1.11 schwarze 1886: }
1887:
1.47 schwarze 1888: /*
1889: * This is straightforward stuff.
1890: * Open a database connection to a "temporary" database, then open a set
1891: * of prepared statements we'll use over and over again.
1892: * If "real" is set, we use the existing database; if not, we truncate a
1893: * temporary one.
1894: * Must be matched by dbclose().
1895: */
1896: static int
1897: dbopen(int real)
1.2 schwarze 1898: {
1.47 schwarze 1899: const char *file, *sql;
1900: int rc, ofl;
1.6 schwarze 1901:
1.47 schwarze 1902: if (nodb)
1903: return(1);
1.6 schwarze 1904:
1.47 schwarze 1905: ofl = SQLITE_OPEN_READWRITE;
1906: if (0 == real) {
1907: file = MANDOC_DB "~";
1908: if (-1 == remove(file) && ENOENT != errno) {
1909: exitcode = (int)MANDOCLEVEL_SYSERR;
1910: say(file, NULL);
1911: return(0);
1.28 schwarze 1912: }
1.47 schwarze 1913: ofl |= SQLITE_OPEN_EXCLUSIVE;
1914: } else
1915: file = MANDOC_DB;
1.6 schwarze 1916:
1.47 schwarze 1917: rc = sqlite3_open_v2(file, &db, ofl, NULL);
1918: if (SQLITE_OK == rc)
1919: goto prepare_statements;
1920: if (SQLITE_CANTOPEN != rc) {
1921: exitcode = (int)MANDOCLEVEL_SYSERR;
1922: say(file, NULL);
1923: return(0);
1924: }
1.6 schwarze 1925:
1.47 schwarze 1926: sqlite3_close(db);
1927: db = NULL;
1.6 schwarze 1928:
1.47 schwarze 1929: if (SQLITE_OK != (rc = sqlite3_open(file, &db))) {
1930: exitcode = (int)MANDOCLEVEL_SYSERR;
1931: say(file, NULL);
1932: return(0);
1.2 schwarze 1933: }
1934:
1.47 schwarze 1935: /*
1936: * XXX The first three columns in table mpages are obsolete
1937: * and only kept for backward compatibility
1938: * until apropos(1) and friends have caught up.
1939: */
1940: sql = "CREATE TABLE \"mpages\" (\n"
1941: " \"file\" TEXT NOT NULL,\n"
1942: " \"sec\" TEXT NOT NULL,\n"
1943: " \"arch\" TEXT NOT NULL,\n"
1944: " \"desc\" TEXT NOT NULL,\n"
1945: " \"form\" INTEGER NOT NULL,\n"
1946: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1947: ");\n"
1948: "\n"
1949: "CREATE TABLE \"mlinks\" (\n"
1950: " \"file\" TEXT NOT NULL,\n"
1951: " \"sec\" TEXT NOT NULL,\n"
1952: " \"arch\" TEXT NOT NULL,\n"
1953: " \"name\" TEXT NOT NULL,\n"
1954: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1955: "ON DELETE CASCADE,\n"
1956: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1957: ");\n"
1958: "\n"
1959: "CREATE TABLE \"keys\" (\n"
1960: " \"bits\" INTEGER NOT NULL,\n"
1961: " \"key\" TEXT NOT NULL,\n"
1962: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
1963: "ON DELETE CASCADE,\n"
1964: " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1965: ");\n"
1966: "\n"
1967: "CREATE INDEX \"key_index\" ON keys (key);\n";
1968:
1969: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1970: exitcode = (int)MANDOCLEVEL_SYSERR;
1971: say(file, "%s", sqlite3_errmsg(db));
1972: return(0);
1.2 schwarze 1973: }
1974:
1.47 schwarze 1975: prepare_statements:
1976: SQL_EXEC("PRAGMA foreign_keys = ON");
1977: sql = "DELETE FROM mpages where file=?";
1978: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
1979: sql = "INSERT INTO mpages "
1980: "(file,sec,arch,desc,form) VALUES (?,?,?,?,?)";
1981: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
1982: sql = "INSERT INTO mlinks "
1983: "(file,sec,arch,name,pageid) VALUES (?,?,?,?,?)";
1984: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1985: sql = "INSERT INTO keys "
1986: "(bits,key,pageid) VALUES (?,?,?)";
1987: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.6 schwarze 1988:
1.47 schwarze 1989: /*
1990: * When opening a new database, we can turn off
1991: * synchronous mode for much better performance.
1992: */
1.6 schwarze 1993:
1.47 schwarze 1994: if (real)
1995: SQL_EXEC("PRAGMA synchronous = OFF");
1.11 schwarze 1996:
1.47 schwarze 1997: return(1);
1998: }
1.6 schwarze 1999:
1.47 schwarze 2000: static void *
2001: hash_halloc(size_t sz, void *arg)
2002: {
1.6 schwarze 2003:
1.47 schwarze 2004: return(mandoc_calloc(sz, 1));
2005: }
1.2 schwarze 2006:
1.47 schwarze 2007: static void *
2008: hash_alloc(size_t sz, void *arg)
2009: {
1.28 schwarze 2010:
1.47 schwarze 2011: return(mandoc_malloc(sz));
2012: }
1.6 schwarze 2013:
1.47 schwarze 2014: static void
2015: hash_free(void *p, size_t sz, void *arg)
2016: {
1.26 schwarze 2017:
1.47 schwarze 2018: free(p);
2019: }
1.6 schwarze 2020:
1.47 schwarze 2021: static int
2022: set_basedir(const char *targetdir)
2023: {
2024: static char startdir[PATH_MAX];
2025: static int fd;
1.6 schwarze 2026:
1.47 schwarze 2027: /*
2028: * Remember where we started by keeping a fd open to the origin
2029: * path component: throughout this utility, we chdir() a lot to
2030: * handle relative paths, and by doing this, we can return to
2031: * the starting point.
2032: */
2033: if ('\0' == *startdir) {
2034: if (NULL == getcwd(startdir, PATH_MAX)) {
2035: exitcode = (int)MANDOCLEVEL_SYSERR;
2036: if (NULL != targetdir)
2037: say(".", NULL);
2038: return(0);
2039: }
2040: if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
2041: exitcode = (int)MANDOCLEVEL_SYSERR;
2042: say(startdir, NULL);
2043: return(0);
1.11 schwarze 2044: }
1.47 schwarze 2045: if (NULL == targetdir)
2046: targetdir = startdir;
2047: } else {
2048: if (-1 == fd)
2049: return(0);
2050: if (-1 == fchdir(fd)) {
2051: close(fd);
2052: basedir[0] = '\0';
2053: exitcode = (int)MANDOCLEVEL_SYSERR;
2054: say(startdir, NULL);
2055: return(0);
1.2 schwarze 2056: }
1.47 schwarze 2057: if (NULL == targetdir) {
2058: close(fd);
2059: return(1);
1.2 schwarze 2060: }
2061: }
1.47 schwarze 2062: if (NULL == realpath(targetdir, basedir)) {
2063: basedir[0] = '\0';
2064: exitcode = (int)MANDOCLEVEL_BADARG;
2065: say(targetdir, NULL);
2066: return(0);
2067: } else if (-1 == chdir(basedir)) {
2068: exitcode = (int)MANDOCLEVEL_BADARG;
2069: say("", NULL);
2070: return(0);
2071: }
2072: return(1);
1.2 schwarze 2073: }
2074:
2075: static void
1.47 schwarze 2076: say(const char *file, const char *format, ...)
1.2 schwarze 2077: {
1.47 schwarze 2078: va_list ap;
1.2 schwarze 2079:
1.47 schwarze 2080: if ('\0' != *basedir)
2081: fprintf(stderr, "%s", basedir);
2082: if ('\0' != *basedir && '\0' != *file)
2083: fputs("//", stderr);
2084: if ('\0' != *file)
2085: fprintf(stderr, "%s", file);
2086: fputs(": ", stderr);
1.31 schwarze 2087:
1.47 schwarze 2088: if (NULL == format) {
2089: perror(NULL);
2090: return;
1.2 schwarze 2091: }
1.47 schwarze 2092:
2093: va_start(ap, format);
2094: vfprintf(stderr, format, ap);
2095: va_end(ap);
2096:
2097: fputc('\n', stderr);
1.1 schwarze 2098: }