Annotation of mandoc/mandocdb.c, Revision 1.173
1.173 ! schwarze 1: /* $Id: mandocdb.c,v 1.172 2014/12/04 17:36:00 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: #include "config.h"
19:
1.162 schwarze 20: #include <sys/types.h>
1.50 kristaps 21: #include <sys/stat.h>
1.115 schwarze 22: #include <sys/wait.h>
1.1 kristaps 23:
24: #include <assert.h>
1.43 kristaps 25: #include <ctype.h>
1.63 schwarze 26: #include <errno.h>
1.1 kristaps 27: #include <fcntl.h>
1.158 schwarze 28: #if HAVE_FTS
1.50 kristaps 29: #include <fts.h>
1.157 schwarze 30: #else
31: #include "compat_fts.h"
32: #endif
1.1 kristaps 33: #include <getopt.h>
1.58 schwarze 34: #include <limits.h>
1.50 kristaps 35: #include <stddef.h>
1.59 schwarze 36: #include <stdio.h>
1.1 kristaps 37: #include <stdint.h>
38: #include <stdlib.h>
39: #include <string.h>
1.17 schwarze 40: #include <unistd.h>
1.1 kristaps 41:
1.158 schwarze 42: #if HAVE_OHASH
1.50 kristaps 43: #include <ohash.h>
1.53 kristaps 44: #else
45: #include "compat_ohash.h"
46: #endif
1.50 kristaps 47: #include <sqlite3.h>
1.1 kristaps 48:
1.50 kristaps 49: #include "mdoc.h"
1.1 kristaps 50: #include "man.h"
51: #include "mandoc.h"
1.119 schwarze 52: #include "mandoc_aux.h"
1.10 kristaps 53: #include "manpath.h"
1.55 kristaps 54: #include "mansearch.h"
1.1 kristaps 55:
1.111 schwarze 56: extern int mansearch_keymax;
57: extern const char *const mansearch_keynames[];
58:
1.52 kristaps 59: #define SQL_EXEC(_v) \
60: if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
1.134 schwarze 61: say("", "%s: %s", (_v), sqlite3_errmsg(db))
1.52 kristaps 62: #define SQL_BIND_TEXT(_s, _i, _v) \
63: if (SQLITE_OK != sqlite3_bind_text \
64: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.134 schwarze 65: say(mlink->file, "%s", sqlite3_errmsg(db))
1.52 kristaps 66: #define SQL_BIND_INT(_s, _i, _v) \
67: if (SQLITE_OK != sqlite3_bind_int \
68: ((_s), (_i)++, (_v))) \
1.134 schwarze 69: say(mlink->file, "%s", sqlite3_errmsg(db))
1.52 kristaps 70: #define SQL_BIND_INT64(_s, _i, _v) \
71: if (SQLITE_OK != sqlite3_bind_int64 \
72: ((_s), (_i)++, (_v))) \
1.134 schwarze 73: say(mlink->file, "%s", sqlite3_errmsg(db))
1.52 kristaps 74: #define SQL_STEP(_s) \
75: if (SQLITE_DONE != sqlite3_step((_s))) \
1.134 schwarze 76: say(mlink->file, "%s", sqlite3_errmsg(db))
1.52 kristaps 77:
1.50 kristaps 78: enum op {
79: OP_DEFAULT = 0, /* new dbs from dir list or default config */
80: OP_CONFFILE, /* new databases from custom config file */
81: OP_UPDATE, /* delete/add entries in existing database */
82: OP_DELETE, /* delete entries from existing database */
83: OP_TEST /* change no databases, report potential problems */
1.38 schwarze 84: };
85:
1.50 kristaps 86: struct str {
1.96 schwarze 87: char *rendered; /* key in UTF-8 or ASCII form */
1.78 schwarze 88: const struct mpage *mpage; /* if set, the owning parse */
1.50 kristaps 89: uint64_t mask; /* bitmask in sequence */
1.96 schwarze 90: char key[]; /* may contain escape sequences */
1.38 schwarze 91: };
92:
1.79 schwarze 93: struct inodev {
94: ino_t st_ino;
95: dev_t st_dev;
1.50 kristaps 96: };
1.5 kristaps 97:
1.78 schwarze 98: struct mpage {
1.79 schwarze 99: struct inodev inodev; /* used for hashing routine */
1.137 schwarze 100: int64_t pageid; /* pageid in mpages SQL table */
1.82 schwarze 101: char *sec; /* section from file content */
102: char *arch; /* architecture from file content */
103: char *title; /* title from file content */
104: char *desc; /* description from file content */
105: struct mlink *mlinks; /* singly linked list */
1.161 schwarze 106: int form; /* format from file content */
1.82 schwarze 107: };
108:
109: struct mlink {
1.58 schwarze 110: char file[PATH_MAX]; /* filename rel. to manpath */
1.82 schwarze 111: char *dsec; /* section from directory */
112: char *arch; /* architecture from directory */
113: char *name; /* name from file name (not empty) */
114: char *fsec; /* section from file name suffix */
1.85 schwarze 115: struct mlink *next; /* singly linked list */
1.118 schwarze 116: struct mpage *mpage; /* parent */
1.161 schwarze 117: int dform; /* format from directory */
118: int fform; /* format from file name suffix */
1.124 schwarze 119: int gzip; /* filename has a .gz suffix */
1.3 kristaps 120: };
121:
1.50 kristaps 122: enum stmt {
1.81 schwarze 123: STMT_DELETE_PAGE = 0, /* delete mpage */
124: STMT_INSERT_PAGE, /* insert mpage */
125: STMT_INSERT_LINK, /* insert mlink */
1.133 schwarze 126: STMT_INSERT_NAME, /* insert name */
1.81 schwarze 127: STMT_INSERT_KEY, /* insert parsed key */
1.50 kristaps 128: STMT__MAX
1.1 kristaps 129: };
130:
1.172 schwarze 131: typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_meta *,
132: const struct mdoc_node *);
1.1 kristaps 133:
1.50 kristaps 134: struct mdoc_handler {
135: mdoc_fp fp; /* optional handler */
136: uint64_t mask; /* set unless handler returns 0 */
1.1 kristaps 137: };
138:
1.59 schwarze 139: static void dbclose(int);
1.118 schwarze 140: static void dbadd(struct mpage *, struct mchars *);
141: static void dbadd_mlink(const struct mlink *mlink);
1.169 schwarze 142: static void dbadd_mlink_name(const struct mlink *mlink);
1.59 schwarze 143: static int dbopen(int);
144: static void dbprune(void);
145: static void filescan(const char *);
1.50 kristaps 146: static void *hash_alloc(size_t, void *);
1.152 schwarze 147: static void hash_free(void *, void *);
148: static void *hash_calloc(size_t, size_t, void *);
1.84 schwarze 149: static void mlink_add(struct mlink *, const struct stat *);
1.131 schwarze 150: static void mlink_check(struct mpage *, struct mlink *);
1.82 schwarze 151: static void mlink_free(struct mlink *);
1.89 schwarze 152: static void mlinks_undupe(struct mpage *);
1.78 schwarze 153: static void mpages_free(void);
1.101 schwarze 154: static void mpages_merge(struct mchars *, struct mparse *);
1.130 schwarze 155: static void names_check(void);
1.124 schwarze 156: static void parse_cat(struct mpage *, int);
1.78 schwarze 157: static void parse_man(struct mpage *, const struct man_node *);
1.172 schwarze 158: static void parse_mdoc(struct mpage *, const struct mdoc_meta *,
159: const struct mdoc_node *);
160: static int parse_mdoc_body(struct mpage *, const struct mdoc_meta *,
161: const struct mdoc_node *);
162: static int parse_mdoc_head(struct mpage *, const struct mdoc_meta *,
163: const struct mdoc_node *);
164: static int parse_mdoc_Fd(struct mpage *, const struct mdoc_meta *,
165: const struct mdoc_node *);
166: static int parse_mdoc_Fn(struct mpage *, const struct mdoc_meta *,
167: const struct mdoc_node *);
1.173 ! schwarze 168: static int parse_mdoc_Fo(struct mpage *, const struct mdoc_meta *,
! 169: const struct mdoc_node *);
1.172 schwarze 170: static int parse_mdoc_Nd(struct mpage *, const struct mdoc_meta *,
171: const struct mdoc_node *);
172: static int parse_mdoc_Nm(struct mpage *, const struct mdoc_meta *,
173: const struct mdoc_node *);
174: static int parse_mdoc_Sh(struct mpage *, const struct mdoc_meta *,
175: const struct mdoc_node *);
176: static int parse_mdoc_Xr(struct mpage *, const struct mdoc_meta *,
177: const struct mdoc_node *);
1.112 schwarze 178: static void putkey(const struct mpage *, char *, uint64_t);
1.78 schwarze 179: static void putkeys(const struct mpage *,
1.64 schwarze 180: const char *, size_t, uint64_t);
1.78 schwarze 181: static void putmdockey(const struct mpage *,
1.50 kristaps 182: const struct mdoc_node *, uint64_t);
1.96 schwarze 183: static void render_key(struct mchars *, struct str *);
1.59 schwarze 184: static void say(const char *, const char *, ...);
1.165 schwarze 185: static int set_basedir(const char *, int);
1.59 schwarze 186: static int treescan(void);
1.50 kristaps 187: static size_t utf8(unsigned int, char [7]);
188:
1.115 schwarze 189: static char tempfilename[32];
1.50 kristaps 190: static char *progname;
1.102 schwarze 191: static int nodb; /* no database changes */
1.116 schwarze 192: static int mparse_options; /* abort the parse early */
1.141 schwarze 193: static int use_all; /* use all found files */
194: static int debug; /* print what we're doing */
195: static int warnings; /* warn about crap */
1.95 schwarze 196: static int write_utf8; /* write UTF-8 output; else ASCII */
1.59 schwarze 197: static int exitcode; /* to be returned by main */
1.141 schwarze 198: static enum op op; /* operational mode */
1.59 schwarze 199: static char basedir[PATH_MAX]; /* current base directory */
1.78 schwarze 200: static struct ohash mpages; /* table of distinct manual pages */
1.83 schwarze 201: static struct ohash mlinks; /* table of directory entries */
1.133 schwarze 202: static struct ohash names; /* table of all names */
1.50 kristaps 203: static struct ohash strings; /* table of all strings */
204: static sqlite3 *db = NULL; /* current database */
205: static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
1.133 schwarze 206: static uint64_t name_mask;
1.25 schwarze 207:
208: static const struct mdoc_handler mdocs[MDOC_MAX] = {
1.68 schwarze 209: { NULL, 0 }, /* Ap */
210: { NULL, 0 }, /* Dd */
211: { NULL, 0 }, /* Dt */
212: { NULL, 0 }, /* Os */
213: { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
214: { parse_mdoc_head, TYPE_Ss }, /* Ss */
215: { NULL, 0 }, /* Pp */
216: { NULL, 0 }, /* D1 */
217: { NULL, 0 }, /* Dl */
218: { NULL, 0 }, /* Bd */
219: { NULL, 0 }, /* Ed */
220: { NULL, 0 }, /* Bl */
221: { NULL, 0 }, /* El */
222: { NULL, 0 }, /* It */
223: { NULL, 0 }, /* Ad */
224: { NULL, TYPE_An }, /* An */
225: { NULL, TYPE_Ar }, /* Ar */
226: { NULL, TYPE_Cd }, /* Cd */
227: { NULL, TYPE_Cm }, /* Cm */
228: { NULL, TYPE_Dv }, /* Dv */
229: { NULL, TYPE_Er }, /* Er */
230: { NULL, TYPE_Ev }, /* Ev */
231: { NULL, 0 }, /* Ex */
232: { NULL, TYPE_Fa }, /* Fa */
233: { parse_mdoc_Fd, 0 }, /* Fd */
234: { NULL, TYPE_Fl }, /* Fl */
235: { parse_mdoc_Fn, 0 }, /* Fn */
236: { NULL, TYPE_Ft }, /* Ft */
237: { NULL, TYPE_Ic }, /* Ic */
1.92 schwarze 238: { NULL, TYPE_In }, /* In */
1.68 schwarze 239: { NULL, TYPE_Li }, /* Li */
1.132 schwarze 240: { parse_mdoc_Nd, 0 }, /* Nd */
1.129 schwarze 241: { parse_mdoc_Nm, 0 }, /* Nm */
1.68 schwarze 242: { NULL, 0 }, /* Op */
243: { NULL, 0 }, /* Ot */
244: { NULL, TYPE_Pa }, /* Pa */
245: { NULL, 0 }, /* Rv */
1.92 schwarze 246: { NULL, TYPE_St }, /* St */
1.68 schwarze 247: { NULL, TYPE_Va }, /* Va */
248: { parse_mdoc_body, TYPE_Va }, /* Vt */
249: { parse_mdoc_Xr, 0 }, /* Xr */
250: { NULL, 0 }, /* %A */
251: { NULL, 0 }, /* %B */
252: { NULL, 0 }, /* %D */
253: { NULL, 0 }, /* %I */
254: { NULL, 0 }, /* %J */
255: { NULL, 0 }, /* %N */
256: { NULL, 0 }, /* %O */
257: { NULL, 0 }, /* %P */
258: { NULL, 0 }, /* %R */
259: { NULL, 0 }, /* %T */
260: { NULL, 0 }, /* %V */
261: { NULL, 0 }, /* Ac */
262: { NULL, 0 }, /* Ao */
263: { NULL, 0 }, /* Aq */
264: { NULL, TYPE_At }, /* At */
265: { NULL, 0 }, /* Bc */
266: { NULL, 0 }, /* Bf */
267: { NULL, 0 }, /* Bo */
268: { NULL, 0 }, /* Bq */
269: { NULL, TYPE_Bsx }, /* Bsx */
270: { NULL, TYPE_Bx }, /* Bx */
271: { NULL, 0 }, /* Db */
272: { NULL, 0 }, /* Dc */
273: { NULL, 0 }, /* Do */
274: { NULL, 0 }, /* Dq */
275: { NULL, 0 }, /* Ec */
276: { NULL, 0 }, /* Ef */
277: { NULL, TYPE_Em }, /* Em */
278: { NULL, 0 }, /* Eo */
279: { NULL, TYPE_Fx }, /* Fx */
280: { NULL, TYPE_Ms }, /* Ms */
281: { NULL, 0 }, /* No */
282: { NULL, 0 }, /* Ns */
283: { NULL, TYPE_Nx }, /* Nx */
284: { NULL, TYPE_Ox }, /* Ox */
285: { NULL, 0 }, /* Pc */
286: { NULL, 0 }, /* Pf */
287: { NULL, 0 }, /* Po */
288: { NULL, 0 }, /* Pq */
289: { NULL, 0 }, /* Qc */
290: { NULL, 0 }, /* Ql */
291: { NULL, 0 }, /* Qo */
292: { NULL, 0 }, /* Qq */
293: { NULL, 0 }, /* Re */
294: { NULL, 0 }, /* Rs */
295: { NULL, 0 }, /* Sc */
296: { NULL, 0 }, /* So */
297: { NULL, 0 }, /* Sq */
298: { NULL, 0 }, /* Sm */
299: { NULL, 0 }, /* Sx */
300: { NULL, TYPE_Sy }, /* Sy */
301: { NULL, TYPE_Tn }, /* Tn */
302: { NULL, 0 }, /* Ux */
303: { NULL, 0 }, /* Xc */
304: { NULL, 0 }, /* Xo */
1.173 ! schwarze 305: { parse_mdoc_Fo, 0 }, /* Fo */
1.68 schwarze 306: { NULL, 0 }, /* Fc */
307: { NULL, 0 }, /* Oo */
308: { NULL, 0 }, /* Oc */
309: { NULL, 0 }, /* Bk */
310: { NULL, 0 }, /* Ek */
311: { NULL, 0 }, /* Bt */
312: { NULL, 0 }, /* Hf */
313: { NULL, 0 }, /* Fr */
314: { NULL, 0 }, /* Ud */
315: { NULL, TYPE_Lb }, /* Lb */
316: { NULL, 0 }, /* Lp */
317: { NULL, TYPE_Lk }, /* Lk */
318: { NULL, TYPE_Mt }, /* Mt */
319: { NULL, 0 }, /* Brq */
320: { NULL, 0 }, /* Bro */
321: { NULL, 0 }, /* Brc */
322: { NULL, 0 }, /* %C */
323: { NULL, 0 }, /* Es */
324: { NULL, 0 }, /* En */
325: { NULL, TYPE_Dx }, /* Dx */
326: { NULL, 0 }, /* %Q */
327: { NULL, 0 }, /* br */
328: { NULL, 0 }, /* sp */
329: { NULL, 0 }, /* %U */
330: { NULL, 0 }, /* Ta */
1.166 schwarze 331: { NULL, 0 }, /* ll */
1.1 kristaps 332: };
333:
1.141 schwarze 334:
1.1 kristaps 335: int
336: main(int argc, char *argv[])
337: {
1.59 schwarze 338: int ch, i;
1.50 kristaps 339: size_t j, sz;
1.59 schwarze 340: const char *path_arg;
1.50 kristaps 341: struct mchars *mc;
342: struct manpaths dirs;
343: struct mparse *mp;
1.83 schwarze 344: struct ohash_info mpages_info, mlinks_info;
1.50 kristaps 345:
346: memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
347: memset(&dirs, 0, sizeof(struct manpaths));
348:
1.83 schwarze 349: mpages_info.alloc = mlinks_info.alloc = hash_alloc;
1.152 schwarze 350: mpages_info.calloc = mlinks_info.calloc = hash_calloc;
351: mpages_info.free = mlinks_info.free = hash_free;
1.50 kristaps 352:
1.79 schwarze 353: mpages_info.key_offset = offsetof(struct mpage, inodev);
1.83 schwarze 354: mlinks_info.key_offset = offsetof(struct mlink, file);
1.1 kristaps 355:
356: progname = strrchr(argv[0], '/');
357: if (progname == NULL)
358: progname = argv[0];
359: else
360: ++progname;
361:
1.50 kristaps 362: /*
1.141 schwarze 363: * We accept a few different invocations.
1.50 kristaps 364: * The CHECKOP macro makes sure that invocation styles don't
365: * clobber each other.
366: */
367: #define CHECKOP(_op, _ch) do \
368: if (OP_DEFAULT != (_op)) { \
1.153 schwarze 369: fprintf(stderr, "%s: -%c: Conflicting option\n", \
370: progname, (_ch)); \
1.50 kristaps 371: goto usage; \
372: } while (/*CONSTCOND*/0)
1.10 kristaps 373:
1.59 schwarze 374: path_arg = NULL;
1.38 schwarze 375: op = OP_DEFAULT;
1.1 kristaps 376:
1.126 schwarze 377: while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
1.1 kristaps 378: switch (ch) {
1.141 schwarze 379: case 'a':
1.12 schwarze 380: use_all = 1;
381: break;
1.141 schwarze 382: case 'C':
1.50 kristaps 383: CHECKOP(op, ch);
1.59 schwarze 384: path_arg = optarg;
1.38 schwarze 385: op = OP_CONFFILE;
1.34 schwarze 386: break;
1.141 schwarze 387: case 'D':
1.125 schwarze 388: debug++;
389: break;
1.141 schwarze 390: case 'd':
1.50 kristaps 391: CHECKOP(op, ch);
1.59 schwarze 392: path_arg = optarg;
1.5 kristaps 393: op = OP_UPDATE;
394: break;
1.141 schwarze 395: case 'n':
1.50 kristaps 396: nodb = 1;
397: break;
1.141 schwarze 398: case 'p':
1.126 schwarze 399: warnings = 1;
400: break;
1.141 schwarze 401: case 'Q':
1.116 schwarze 402: mparse_options |= MPARSE_QUICK;
1.102 schwarze 403: break;
1.141 schwarze 404: case 'T':
1.95 schwarze 405: if (strcmp(optarg, "utf8")) {
1.153 schwarze 406: fprintf(stderr, "%s: -T%s: "
407: "Unsupported output format\n",
408: progname, optarg);
1.95 schwarze 409: goto usage;
410: }
411: write_utf8 = 1;
412: break;
1.141 schwarze 413: case 't':
1.50 kristaps 414: CHECKOP(op, ch);
1.38 schwarze 415: dup2(STDOUT_FILENO, STDERR_FILENO);
416: op = OP_TEST;
1.50 kristaps 417: nodb = warnings = 1;
1.38 schwarze 418: break;
1.141 schwarze 419: case 'u':
1.50 kristaps 420: CHECKOP(op, ch);
1.59 schwarze 421: path_arg = optarg;
1.5 kristaps 422: op = OP_DELETE;
423: break;
1.141 schwarze 424: case 'v':
1.126 schwarze 425: /* Compatibility with espie@'s makewhatis. */
1.38 schwarze 426: break;
1.1 kristaps 427: default:
1.38 schwarze 428: goto usage;
1.1 kristaps 429: }
430:
431: argc -= optind;
432: argv += optind;
433:
1.38 schwarze 434: if (OP_CONFFILE == op && argc > 0) {
1.153 schwarze 435: fprintf(stderr, "%s: -C: Too many arguments\n",
436: progname);
1.38 schwarze 437: goto usage;
438: }
439:
1.59 schwarze 440: exitcode = (int)MANDOCLEVEL_OK;
1.50 kristaps 441: mc = mchars_alloc();
1.168 schwarze 442: mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL,
443: mc, NULL);
1.78 schwarze 444: ohash_init(&mpages, 6, &mpages_info);
1.83 schwarze 445: ohash_init(&mlinks, 6, &mlinks_info);
1.50 kristaps 446:
447: if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
1.59 schwarze 448:
1.50 kristaps 449: /*
1.150 schwarze 450: * Most of these deal with a specific directory.
1.140 schwarze 451: * Jump into that directory first.
1.50 kristaps 452: */
1.165 schwarze 453: if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
1.50 kristaps 454: goto out;
1.140 schwarze 455:
1.138 schwarze 456: if (dbopen(1)) {
1.140 schwarze 457: /*
458: * The existing database is usable. Process
459: * all files specified on the command-line.
460: */
461: use_all = 1;
1.138 schwarze 462: for (i = 0; i < argc; i++)
463: filescan(argv[i]);
464: if (OP_TEST != op)
465: dbprune();
466: } else {
467: /*
468: * Database missing or corrupt.
469: * Recreate from scratch.
470: */
1.140 schwarze 471: exitcode = (int)MANDOCLEVEL_OK;
1.138 schwarze 472: op = OP_DEFAULT;
473: if (0 == treescan())
474: goto out;
475: if (0 == dbopen(0))
476: goto out;
477: }
1.50 kristaps 478: if (OP_DELETE != op)
1.101 schwarze 479: mpages_merge(mc, mp);
1.138 schwarze 480: dbclose(OP_DEFAULT == op ? 0 : 1);
1.50 kristaps 481: } else {
482: /*
483: * If we have arguments, use them as our manpaths.
484: * If we don't, grok from manpath(1) or however else
485: * manpath_parse() wants to do it.
486: */
487: if (argc > 0) {
1.144 schwarze 488: dirs.paths = mandoc_reallocarray(NULL,
489: argc, sizeof(char *));
1.50 kristaps 490: dirs.sz = (size_t)argc;
491: for (i = 0; i < argc; i++)
492: dirs.paths[i] = mandoc_strdup(argv[i]);
493: } else
1.59 schwarze 494: manpath_parse(&dirs, path_arg, NULL, NULL);
1.127 schwarze 495:
496: if (0 == dirs.sz) {
497: exitcode = (int)MANDOCLEVEL_BADARG;
498: say("", "Empty manpath");
499: }
1.50 kristaps 500:
501: /*
1.69 schwarze 502: * First scan the tree rooted at a base directory, then
503: * build a new database and finally move it into place.
1.50 kristaps 504: * Ignore zero-length directories and strip trailing
505: * slashes.
506: */
507: for (j = 0; j < dirs.sz; j++) {
508: sz = strlen(dirs.paths[j]);
509: if (sz && '/' == dirs.paths[j][sz - 1])
510: dirs.paths[j][--sz] = '\0';
511: if (0 == sz)
512: continue;
1.66 schwarze 513:
514: if (j) {
1.78 schwarze 515: ohash_init(&mpages, 6, &mpages_info);
1.83 schwarze 516: ohash_init(&mlinks, 6, &mlinks_info);
1.66 schwarze 517: }
518:
1.165 schwarze 519: if (0 == set_basedir(dirs.paths[j], argc > 0))
520: continue;
1.59 schwarze 521: if (0 == treescan())
1.165 schwarze 522: continue;
1.59 schwarze 523: if (0 == dbopen(0))
1.165 schwarze 524: continue;
1.52 kristaps 525:
1.101 schwarze 526: mpages_merge(mc, mp);
1.147 schwarze 527: if (warnings && !nodb &&
1.130 schwarze 528: ! (MPARSE_QUICK & mparse_options))
529: names_check();
1.59 schwarze 530: dbclose(0);
1.66 schwarze 531:
532: if (j + 1 < dirs.sz) {
1.80 schwarze 533: mpages_free();
1.78 schwarze 534: ohash_delete(&mpages);
1.83 schwarze 535: ohash_delete(&mlinks);
1.66 schwarze 536: }
1.50 kristaps 537: }
538: }
539: out:
540: manpath_free(&dirs);
1.168 schwarze 541: mparse_free(mp);
1.50 kristaps 542: mchars_free(mc);
1.80 schwarze 543: mpages_free();
1.78 schwarze 544: ohash_delete(&mpages);
1.83 schwarze 545: ohash_delete(&mlinks);
1.59 schwarze 546: return(exitcode);
1.50 kristaps 547: usage:
1.126 schwarze 548: fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
549: " %s [-aDnpQ] [-Tutf8] dir ...\n"
550: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
551: " %s [-Dnp] -u dir [file ...]\n"
1.102 schwarze 552: " %s [-Q] -t file ...\n",
1.141 schwarze 553: progname, progname, progname,
1.50 kristaps 554: progname, progname);
555:
1.59 schwarze 556: return((int)MANDOCLEVEL_BADARG);
1.50 kristaps 557: }
558:
559: /*
1.59 schwarze 560: * Scan a directory tree rooted at "basedir" for manpages.
1.50 kristaps 561: * We use fts(), scanning directory parts along the way for clues to our
562: * section and architecture.
563: *
564: * If use_all has been specified, grok all files.
565: * If not, sanitise paths to the following:
566: *
1.141 schwarze 567: * [./]man*[/<arch>]/<name>.<section>
1.50 kristaps 568: * or
569: * [./]cat<section>[/<arch>]/<name>.0
570: *
571: * TODO: accomodate for multi-language directories.
572: */
573: static int
1.59 schwarze 574: treescan(void)
1.50 kristaps 575: {
1.139 schwarze 576: char buf[PATH_MAX];
1.50 kristaps 577: FTS *f;
578: FTSENT *ff;
1.84 schwarze 579: struct mlink *mlink;
1.124 schwarze 580: int dform, gzip;
1.94 schwarze 581: char *dsec, *arch, *fsec, *cp;
582: const char *path;
1.50 kristaps 583: const char *argv[2];
584:
585: argv[0] = ".";
586: argv[1] = (char *)NULL;
587:
1.139 schwarze 588: f = fts_open((char * const *)argv,
589: FTS_PHYSICAL | FTS_NOCHDIR, NULL);
1.50 kristaps 590: if (NULL == f) {
1.59 schwarze 591: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 592: say("", "&fts_open");
1.50 kristaps 593: return(0);
594: }
595:
596: dsec = arch = NULL;
597: dform = FORM_NONE;
598:
599: while (NULL != (ff = fts_read(f))) {
600: path = ff->fts_path + 2;
1.139 schwarze 601: switch (ff->fts_info) {
602:
603: /*
604: * Symbolic links require various sanity checks,
605: * then get handled just like regular files.
606: */
1.141 schwarze 607: case FTS_SL:
1.139 schwarze 608: if (NULL == realpath(path, buf)) {
609: if (warnings)
610: say(path, "&realpath");
611: continue;
612: }
613: if (strstr(buf, basedir) != buf) {
614: if (warnings) say("",
615: "%s: outside base directory", buf);
616: continue;
617: }
618: /* Use logical inode to avoid mpages dupe. */
619: if (-1 == stat(path, ff->fts_statp)) {
620: if (warnings)
621: say(path, "&stat");
622: continue;
623: }
624: /* FALLTHROUGH */
625:
1.50 kristaps 626: /*
1.83 schwarze 627: * If we're a regular file, add an mlink by using the
1.50 kristaps 628: * stored directory data and handling the filename.
629: */
1.141 schwarze 630: case FTS_F:
1.60 schwarze 631: if (0 == strcmp(path, MANDOC_DB))
632: continue;
1.50 kristaps 633: if ( ! use_all && ff->fts_level < 2) {
1.56 schwarze 634: if (warnings)
1.59 schwarze 635: say(path, "Extraneous file");
1.50 kristaps 636: continue;
1.124 schwarze 637: }
638: gzip = 0;
639: fsec = NULL;
640: while (NULL == fsec) {
641: fsec = strrchr(ff->fts_name, '.');
642: if (NULL == fsec || strcmp(fsec+1, "gz"))
643: break;
644: gzip = 1;
645: *fsec = '\0';
646: fsec = NULL;
647: }
648: if (NULL == fsec) {
1.60 schwarze 649: if ( ! use_all) {
1.56 schwarze 650: if (warnings)
1.60 schwarze 651: say(path,
652: "No filename suffix");
1.50 kristaps 653: continue;
654: }
1.84 schwarze 655: } else if (0 == strcmp(++fsec, "html")) {
1.60 schwarze 656: if (warnings)
657: say(path, "Skip html");
658: continue;
1.84 schwarze 659: } else if (0 == strcmp(fsec, "ps")) {
1.60 schwarze 660: if (warnings)
661: say(path, "Skip ps");
662: continue;
1.84 schwarze 663: } else if (0 == strcmp(fsec, "pdf")) {
1.60 schwarze 664: if (warnings)
665: say(path, "Skip pdf");
666: continue;
667: } else if ( ! use_all &&
1.84 schwarze 668: ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
669: (FORM_CAT == dform && strcmp(fsec, "0")))) {
1.60 schwarze 670: if (warnings)
671: say(path, "Wrong filename suffix");
672: continue;
1.66 schwarze 673: } else
1.84 schwarze 674: fsec[-1] = '\0';
1.94 schwarze 675:
1.84 schwarze 676: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.143 schwarze 677: if (strlcpy(mlink->file, path,
678: sizeof(mlink->file)) >=
679: sizeof(mlink->file)) {
680: say(path, "Filename too long");
681: free(mlink);
682: continue;
683: }
1.84 schwarze 684: mlink->dform = dform;
1.94 schwarze 685: mlink->dsec = dsec;
686: mlink->arch = arch;
687: mlink->name = ff->fts_name;
688: mlink->fsec = fsec;
1.124 schwarze 689: mlink->gzip = gzip;
1.84 schwarze 690: mlink_add(mlink, ff->fts_statp);
1.50 kristaps 691: continue;
1.139 schwarze 692:
1.141 schwarze 693: case FTS_D:
1.139 schwarze 694: /* FALLTHROUGH */
1.141 schwarze 695: case FTS_DP:
1.139 schwarze 696: break;
697:
698: default:
1.60 schwarze 699: if (warnings)
700: say(path, "Not a regular file");
1.50 kristaps 701: continue;
1.60 schwarze 702: }
1.1 kristaps 703:
1.50 kristaps 704: switch (ff->fts_level) {
1.141 schwarze 705: case 0:
1.50 kristaps 706: /* Ignore the root directory. */
707: break;
1.141 schwarze 708: case 1:
1.50 kristaps 709: /*
710: * This might contain manX/ or catX/.
711: * Try to infer this from the name.
712: * If we're not in use_all, enforce it.
713: */
714: cp = ff->fts_name;
715: if (FTS_DP == ff->fts_info)
716: break;
1.1 kristaps 717:
1.50 kristaps 718: if (0 == strncmp(cp, "man", 3)) {
719: dform = FORM_SRC;
1.66 schwarze 720: dsec = cp + 3;
1.50 kristaps 721: } else if (0 == strncmp(cp, "cat", 3)) {
722: dform = FORM_CAT;
1.66 schwarze 723: dsec = cp + 3;
1.94 schwarze 724: } else {
725: dform = FORM_NONE;
726: dsec = NULL;
1.50 kristaps 727: }
1.1 kristaps 728:
1.141 schwarze 729: if (NULL != dsec || use_all)
1.50 kristaps 730: break;
1.1 kristaps 731:
1.56 schwarze 732: if (warnings)
1.59 schwarze 733: say(path, "Unknown directory part");
1.50 kristaps 734: fts_set(f, ff, FTS_SKIP);
735: break;
1.141 schwarze 736: case 2:
1.50 kristaps 737: /*
738: * Possibly our architecture.
739: * If we're descending, keep tabs on it.
740: */
741: if (FTS_DP != ff->fts_info && NULL != dsec)
1.66 schwarze 742: arch = ff->fts_name;
1.94 schwarze 743: else
744: arch = NULL;
1.50 kristaps 745: break;
746: default:
747: if (FTS_DP == ff->fts_info || use_all)
748: break;
1.56 schwarze 749: if (warnings)
1.59 schwarze 750: say(path, "Extraneous directory part");
1.50 kristaps 751: fts_set(f, ff, FTS_SKIP);
752: break;
1.5 kristaps 753: }
1.50 kristaps 754: }
755:
756: fts_close(f);
757: return(1);
758: }
1.5 kristaps 759:
1.50 kristaps 760: /*
1.89 schwarze 761: * Add a file to the mlinks table.
1.50 kristaps 762: * Do not verify that it's a "valid" looking manpage (we'll do that
763: * later).
764: *
765: * Try to infer the manual section, architecture, and page name from the
766: * path, assuming it looks like
767: *
1.141 schwarze 768: * [./]man*[/<arch>]/<name>.<section>
1.50 kristaps 769: * or
770: * [./]cat<section>[/<arch>]/<name>.0
771: *
772: * See treescan() for the fts(3) version of this.
773: */
774: static void
1.59 schwarze 775: filescan(const char *file)
1.50 kristaps 776: {
1.59 schwarze 777: char buf[PATH_MAX];
1.84 schwarze 778: struct stat st;
779: struct mlink *mlink;
1.59 schwarze 780: char *p, *start;
1.5 kristaps 781:
1.50 kristaps 782: assert(use_all);
1.5 kristaps 783:
1.50 kristaps 784: if (0 == strncmp(file, "./", 2))
785: file += 2;
1.5 kristaps 786:
1.139 schwarze 787: /*
788: * We have to do lstat(2) before realpath(3) loses
789: * the information whether this is a symbolic link.
790: * We need to know that because for symbolic links,
791: * we want to use the orginal file name, while for
792: * regular files, we want to use the real path.
793: */
794: if (-1 == lstat(file, &st)) {
795: exitcode = (int)MANDOCLEVEL_BADARG;
796: say(file, "&lstat");
797: return;
798: } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
799: exitcode = (int)MANDOCLEVEL_BADARG;
800: say(file, "Not a regular file");
801: return;
802: }
803:
804: /*
805: * We have to resolve the file name to the real path
806: * in any case for the base directory check.
807: */
1.59 schwarze 808: if (NULL == realpath(file, buf)) {
809: exitcode = (int)MANDOCLEVEL_BADARG;
1.123 schwarze 810: say(file, "&realpath");
1.59 schwarze 811: return;
1.106 schwarze 812: }
813:
1.150 schwarze 814: if (OP_TEST == op)
1.106 schwarze 815: start = buf;
1.150 schwarze 816: else if (strstr(buf, basedir) == buf)
817: start = buf + strlen(basedir);
1.106 schwarze 818: else {
1.59 schwarze 819: exitcode = (int)MANDOCLEVEL_BADARG;
820: say("", "%s: outside base directory", buf);
821: return;
1.106 schwarze 822: }
823:
1.139 schwarze 824: /*
825: * Now we are sure the file is inside our tree.
826: * If it is a symbolic link, ignore the real path
827: * and use the original name.
828: * This implies passing stuff like "cat1/../man1/foo.1"
829: * on the command line won't work. So don't do that.
830: * Note the stat(2) can still fail if the link target
831: * doesn't exist.
832: */
833: if (S_IFLNK & st.st_mode) {
834: if (-1 == stat(buf, &st)) {
835: exitcode = (int)MANDOCLEVEL_BADARG;
836: say(file, "&stat");
837: return;
838: }
1.143 schwarze 839: if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
840: say(file, "Filename too long");
841: return;
842: }
1.150 schwarze 843: start = buf;
844: if (OP_TEST != op && strstr(buf, basedir) == buf)
845: start += strlen(basedir);
1.50 kristaps 846: }
1.106 schwarze 847:
1.84 schwarze 848: mlink = mandoc_calloc(1, sizeof(struct mlink));
1.161 schwarze 849: mlink->dform = FORM_NONE;
1.143 schwarze 850: if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
851: sizeof(mlink->file)) {
852: say(start, "Filename too long");
853: return;
854: }
1.17 schwarze 855:
1.50 kristaps 856: /*
857: * First try to guess our directory structure.
858: * If we find a separator, try to look for man* or cat*.
859: * If we find one of these and what's underneath is a directory,
860: * assume it's an architecture.
861: */
862: if (NULL != (p = strchr(start, '/'))) {
863: *p++ = '\0';
864: if (0 == strncmp(start, "man", 3)) {
1.84 schwarze 865: mlink->dform = FORM_SRC;
1.94 schwarze 866: mlink->dsec = start + 3;
1.50 kristaps 867: } else if (0 == strncmp(start, "cat", 3)) {
1.84 schwarze 868: mlink->dform = FORM_CAT;
1.94 schwarze 869: mlink->dsec = start + 3;
1.17 schwarze 870: }
1.5 kristaps 871:
1.50 kristaps 872: start = p;
1.84 schwarze 873: if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
1.50 kristaps 874: *p++ = '\0';
1.94 schwarze 875: mlink->arch = start;
1.50 kristaps 876: start = p;
1.84 schwarze 877: }
1.50 kristaps 878: }
879:
880: /*
881: * Now check the file suffix.
882: * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
883: */
884: p = strrchr(start, '\0');
885: while (p-- > start && '/' != *p && '.' != *p)
886: /* Loop. */ ;
887:
888: if ('.' == *p) {
889: *p++ = '\0';
1.94 schwarze 890: mlink->fsec = p;
1.5 kristaps 891: }
892:
1.10 kristaps 893: /*
1.50 kristaps 894: * Now try to parse the name.
895: * Use the filename portion of the path.
1.10 kristaps 896: */
1.84 schwarze 897: mlink->name = start;
1.50 kristaps 898: if (NULL != (p = strrchr(start, '/'))) {
1.84 schwarze 899: mlink->name = p + 1;
1.50 kristaps 900: *p = '\0';
1.84 schwarze 901: }
902: mlink_add(mlink, &st);
1.50 kristaps 903: }
1.5 kristaps 904:
1.50 kristaps 905: static void
1.84 schwarze 906: mlink_add(struct mlink *mlink, const struct stat *st)
1.50 kristaps 907: {
1.83 schwarze 908: struct inodev inodev;
1.78 schwarze 909: struct mpage *mpage;
1.83 schwarze 910: unsigned int slot;
1.50 kristaps 911:
1.84 schwarze 912: assert(NULL != mlink->file);
1.50 kristaps 913:
1.94 schwarze 914: mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
915: mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
916: mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
917: mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1.84 schwarze 918:
919: if ('0' == *mlink->fsec) {
920: free(mlink->fsec);
921: mlink->fsec = mandoc_strdup(mlink->dsec);
922: mlink->fform = FORM_CAT;
923: } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
924: mlink->fform = FORM_SRC;
1.74 schwarze 925: else
1.84 schwarze 926: mlink->fform = FORM_NONE;
1.82 schwarze 927:
1.83 schwarze 928: slot = ohash_qlookup(&mlinks, mlink->file);
929: assert(NULL == ohash_find(&mlinks, slot));
930: ohash_insert(&mlinks, slot, mlink);
931:
932: inodev.st_ino = st->st_ino;
933: inodev.st_dev = st->st_dev;
934: slot = ohash_lookup_memory(&mpages, (char *)&inodev,
935: sizeof(struct inodev), inodev.st_ino);
936: mpage = ohash_find(&mpages, slot);
937: if (NULL == mpage) {
938: mpage = mandoc_calloc(1, sizeof(struct mpage));
939: mpage->inodev.st_ino = inodev.st_ino;
940: mpage->inodev.st_dev = inodev.st_dev;
941: ohash_insert(&mpages, slot, mpage);
942: } else
1.85 schwarze 943: mlink->next = mpage->mlinks;
1.82 schwarze 944: mpage->mlinks = mlink;
1.118 schwarze 945: mlink->mpage = mpage;
1.82 schwarze 946: }
947:
948: static void
949: mlink_free(struct mlink *mlink)
950: {
951:
952: free(mlink->dsec);
953: free(mlink->arch);
954: free(mlink->name);
955: free(mlink->fsec);
956: free(mlink);
1.50 kristaps 957: }
1.3 kristaps 958:
1.50 kristaps 959: static void
1.78 schwarze 960: mpages_free(void)
1.50 kristaps 961: {
1.78 schwarze 962: struct mpage *mpage;
1.82 schwarze 963: struct mlink *mlink;
1.80 schwarze 964: unsigned int slot;
1.3 kristaps 965:
1.80 schwarze 966: mpage = ohash_first(&mpages, &slot);
967: while (NULL != mpage) {
1.82 schwarze 968: while (NULL != (mlink = mpage->mlinks)) {
1.85 schwarze 969: mpage->mlinks = mlink->next;
1.82 schwarze 970: mlink_free(mlink);
971: }
1.78 schwarze 972: free(mpage->sec);
973: free(mpage->arch);
1.82 schwarze 974: free(mpage->title);
975: free(mpage->desc);
1.78 schwarze 976: free(mpage);
1.80 schwarze 977: mpage = ohash_next(&mpages, &slot);
1.50 kristaps 978: }
979: }
1.38 schwarze 980:
1.50 kristaps 981: /*
1.89 schwarze 982: * For each mlink to the mpage, check whether the path looks like
983: * it is formatted, and if it does, check whether a source manual
984: * exists by the same name, ignoring the suffix.
985: * If both conditions hold, drop the mlink.
986: */
987: static void
988: mlinks_undupe(struct mpage *mpage)
989: {
990: char buf[PATH_MAX];
991: struct mlink **prev;
992: struct mlink *mlink;
993: char *bufp;
994:
995: mpage->form = FORM_CAT;
1.90 schwarze 996: prev = &mpage->mlinks;
997: while (NULL != (mlink = *prev)) {
1.89 schwarze 998: if (FORM_CAT != mlink->dform) {
999: mpage->form = FORM_NONE;
1.90 schwarze 1000: goto nextlink;
1.89 schwarze 1001: }
1.143 schwarze 1002: (void)strlcpy(buf, mlink->file, sizeof(buf));
1.89 schwarze 1003: bufp = strstr(buf, "cat");
1004: assert(NULL != bufp);
1005: memcpy(bufp, "man", 3);
1006: if (NULL != (bufp = strrchr(buf, '.')))
1007: *++bufp = '\0';
1.143 schwarze 1008: (void)strlcat(buf, mlink->dsec, sizeof(buf));
1.89 schwarze 1009: if (NULL == ohash_find(&mlinks,
1.141 schwarze 1010: ohash_qlookup(&mlinks, buf)))
1.90 schwarze 1011: goto nextlink;
1.89 schwarze 1012: if (warnings)
1013: say(mlink->file, "Man source exists: %s", buf);
1014: if (use_all)
1.90 schwarze 1015: goto nextlink;
1.89 schwarze 1016: *prev = mlink->next;
1017: mlink_free(mlink);
1.90 schwarze 1018: continue;
1019: nextlink:
1020: prev = &(*prev)->next;
1.89 schwarze 1021: }
1022: }
1023:
1.131 schwarze 1024: static void
1.93 schwarze 1025: mlink_check(struct mpage *mpage, struct mlink *mlink)
1026: {
1.131 schwarze 1027: struct str *str;
1028: unsigned int slot;
1.93 schwarze 1029:
1030: /*
1031: * Check whether the manual section given in a file
1032: * agrees with the directory where the file is located.
1033: * Some manuals have suffixes like (3p) on their
1034: * section number either inside the file or in the
1035: * directory name, some are linked into more than one
1036: * section, like encrypt(1) = makekey(8).
1037: */
1038:
1039: if (FORM_SRC == mpage->form &&
1.131 schwarze 1040: strcasecmp(mpage->sec, mlink->dsec))
1.93 schwarze 1041: say(mlink->file, "Section \"%s\" manual in %s directory",
1042: mpage->sec, mlink->dsec);
1043:
1044: /*
1045: * Manual page directories exist for each kernel
1046: * architecture as returned by machine(1).
1047: * However, many manuals only depend on the
1048: * application architecture as returned by arch(1).
1049: * For example, some (2/ARM) manuals are shared
1050: * across the "armish" and "zaurus" kernel
1051: * architectures.
1052: * A few manuals are even shared across completely
1053: * different architectures, for example fdformat(1)
1054: * on amd64, i386, sparc, and sparc64.
1055: */
1056:
1.131 schwarze 1057: if (strcasecmp(mpage->arch, mlink->arch))
1.93 schwarze 1058: say(mlink->file, "Architecture \"%s\" manual in "
1059: "\"%s\" directory", mpage->arch, mlink->arch);
1060:
1.131 schwarze 1061: /*
1062: * XXX
1.133 schwarze 1063: * parse_cat() doesn't set NAME_TITLE yet.
1.131 schwarze 1064: */
1065:
1066: if (FORM_CAT == mpage->form)
1067: return;
1068:
1069: /*
1070: * Check whether this mlink
1071: * appears as a name in the NAME section.
1072: */
1.93 schwarze 1073:
1.133 schwarze 1074: slot = ohash_qlookup(&names, mlink->name);
1075: str = ohash_find(&names, slot);
1.131 schwarze 1076: assert(NULL != str);
1.133 schwarze 1077: if ( ! (NAME_TITLE & str->mask))
1.131 schwarze 1078: say(mlink->file, "Name missing in NAME section");
1.93 schwarze 1079: }
1080:
1.89 schwarze 1081: /*
1.80 schwarze 1082: * Run through the files in the global vector "mpages"
1083: * and add them to the database specified in "basedir".
1.50 kristaps 1084: *
1085: * This handles the parsing scheme itself, using the cues of directory
1086: * and filename to determine whether the file is parsable or not.
1087: */
1.59 schwarze 1088: static void
1.101 schwarze 1089: mpages_merge(struct mchars *mc, struct mparse *mp)
1.50 kristaps 1090: {
1.113 schwarze 1091: char any[] = "any";
1.101 schwarze 1092: struct ohash_info str_info;
1.118 schwarze 1093: struct mpage *mpage, *mpage_dest;
1094: struct mlink *mlink, *mlink_dest;
1.69 schwarze 1095: struct mdoc *mdoc;
1096: struct man *man;
1.118 schwarze 1097: char *sodest;
1.112 schwarze 1098: char *cp;
1.162 schwarze 1099: int fd;
1.101 schwarze 1100: unsigned int pslot;
1.69 schwarze 1101: enum mandoclevel lvl;
1102:
1.71 schwarze 1103: str_info.alloc = hash_alloc;
1.152 schwarze 1104: str_info.calloc = hash_calloc;
1105: str_info.free = hash_free;
1.71 schwarze 1106: str_info.key_offset = offsetof(struct str, key);
1107:
1.171 schwarze 1108: if ( ! nodb)
1.107 schwarze 1109: SQL_EXEC("BEGIN TRANSACTION");
1110:
1.80 schwarze 1111: mpage = ohash_first(&mpages, &pslot);
1.171 schwarze 1112: while (mpage != NULL) {
1.89 schwarze 1113: mlinks_undupe(mpage);
1.171 schwarze 1114: if (mpage->mlinks == NULL) {
1.89 schwarze 1115: mpage = ohash_next(&mpages, &pslot);
1116: continue;
1.50 kristaps 1117: }
1118:
1.133 schwarze 1119: name_mask = NAME_MASK;
1120: ohash_init(&names, 4, &str_info);
1.71 schwarze 1121: ohash_init(&strings, 6, &str_info);
1.50 kristaps 1122: mparse_reset(mp);
1123: mdoc = NULL;
1124: man = NULL;
1.124 schwarze 1125: sodest = NULL;
1126:
1.170 schwarze 1127: mparse_open(mp, &fd, mpage->mlinks->file);
1.162 schwarze 1128: if (fd == -1) {
1129: say(mpage->mlinks->file, "&open");
1130: goto nextpage;
1.124 schwarze 1131: }
1.14 schwarze 1132:
1133: /*
1.33 schwarze 1134: * Try interpreting the file as mdoc(7) or man(7)
1135: * source code, unless it is already known to be
1136: * formatted. Fall back to formatted mode.
1.14 schwarze 1137: */
1.171 schwarze 1138: if (mpage->mlinks->dform != FORM_CAT ||
1139: mpage->mlinks->fform != FORM_CAT) {
1.162 schwarze 1140: lvl = mparse_readfd(mp, fd, mpage->mlinks->file);
1.50 kristaps 1141: if (lvl < MANDOCLEVEL_FATAL)
1.118 schwarze 1142: mparse_result(mp, &mdoc, &man, &sodest);
1.76 schwarze 1143: }
1.14 schwarze 1144:
1.171 schwarze 1145: if (sodest != NULL) {
1.118 schwarze 1146: mlink_dest = ohash_find(&mlinks,
1147: ohash_qlookup(&mlinks, sodest));
1.171 schwarze 1148: if (mlink_dest == NULL) {
1149: mandoc_asprintf(&cp, "%s.gz", sodest);
1150: mlink_dest = ohash_find(&mlinks,
1151: ohash_qlookup(&mlinks, cp));
1152: free(cp);
1153: }
1154: if (mlink_dest != NULL) {
1.118 schwarze 1155:
1156: /* The .so target exists. */
1157:
1158: mpage_dest = mlink_dest->mpage;
1159: mlink = mpage->mlinks;
1160: while (1) {
1161: mlink->mpage = mpage_dest;
1162:
1163: /*
1164: * If the target was already
1165: * processed, add the links
1166: * to the database now.
1167: * Otherwise, this will
1168: * happen when we come
1169: * to the target.
1170: */
1171:
1.137 schwarze 1172: if (mpage_dest->pageid)
1.169 schwarze 1173: dbadd_mlink_name(mlink);
1.118 schwarze 1174:
1.171 schwarze 1175: if (mlink->next == NULL)
1.118 schwarze 1176: break;
1177: mlink = mlink->next;
1178: }
1179:
1180: /* Move all links to the target. */
1181:
1182: mlink->next = mlink_dest->next;
1183: mlink_dest->next = mpage->mlinks;
1184: mpage->mlinks = NULL;
1185: }
1.124 schwarze 1186: goto nextpage;
1.171 schwarze 1187: } else if (mdoc != NULL) {
1.82 schwarze 1188: mpage->form = FORM_SRC;
1.155 schwarze 1189: mpage->sec = mdoc_meta(mdoc)->msec;
1190: mpage->sec = mandoc_strdup(
1.171 schwarze 1191: mpage->sec == NULL ? "" : mpage->sec);
1.82 schwarze 1192: mpage->arch = mdoc_meta(mdoc)->arch;
1.84 schwarze 1193: mpage->arch = mandoc_strdup(
1.171 schwarze 1194: mpage->arch == NULL ? "" : mpage->arch);
1.82 schwarze 1195: mpage->title =
1196: mandoc_strdup(mdoc_meta(mdoc)->title);
1.171 schwarze 1197: } else if (man != NULL) {
1.82 schwarze 1198: mpage->form = FORM_SRC;
1199: mpage->sec =
1200: mandoc_strdup(man_meta(man)->msec);
1201: mpage->arch =
1202: mandoc_strdup(mpage->mlinks->arch);
1203: mpage->title =
1204: mandoc_strdup(man_meta(man)->title);
1.76 schwarze 1205: } else {
1.82 schwarze 1206: mpage->form = FORM_CAT;
1207: mpage->sec =
1208: mandoc_strdup(mpage->mlinks->dsec);
1209: mpage->arch =
1210: mandoc_strdup(mpage->mlinks->arch);
1211: mpage->title =
1212: mandoc_strdup(mpage->mlinks->name);
1.76 schwarze 1213: }
1.97 schwarze 1214: putkey(mpage, mpage->sec, TYPE_sec);
1.164 schwarze 1215: if (*mpage->arch != '\0')
1216: putkey(mpage, mpage->arch, TYPE_arch);
1.50 kristaps 1217:
1.97 schwarze 1218: for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1219: if ('\0' != *mlink->dsec)
1220: putkey(mpage, mlink->dsec, TYPE_sec);
1221: if ('\0' != *mlink->fsec)
1222: putkey(mpage, mlink->fsec, TYPE_sec);
1.98 schwarze 1223: putkey(mpage, '\0' == *mlink->arch ?
1.113 schwarze 1224: any : mlink->arch, TYPE_arch);
1.133 schwarze 1225: putkey(mpage, mlink->name, NAME_FILE);
1.97 schwarze 1226: }
1.12 schwarze 1227:
1.135 schwarze 1228: assert(NULL == mpage->desc);
1.50 kristaps 1229: if (NULL != mdoc) {
1230: if (NULL != (cp = mdoc_meta(mdoc)->name))
1.133 schwarze 1231: putkey(mpage, cp, NAME_HEAD);
1.172 schwarze 1232: parse_mdoc(mpage, mdoc_meta(mdoc), mdoc_node(mdoc));
1.50 kristaps 1233: } else if (NULL != man)
1.78 schwarze 1234: parse_man(mpage, man_node(man));
1.50 kristaps 1235: else
1.162 schwarze 1236: parse_cat(mpage, fd);
1.135 schwarze 1237: if (NULL == mpage->desc)
1238: mpage->desc = mandoc_strdup(mpage->mlinks->name);
1.131 schwarze 1239:
1240: if (warnings && !use_all)
1241: for (mlink = mpage->mlinks; mlink;
1242: mlink = mlink->next)
1243: mlink_check(mpage, mlink);
1.44 kristaps 1244:
1.105 schwarze 1245: dbadd(mpage, mc);
1.124 schwarze 1246:
1247: nextpage:
1.170 schwarze 1248: if (mparse_wait(mp) != MANDOCLEVEL_OK) {
1.162 schwarze 1249: exitcode = (int)MANDOCLEVEL_SYSERR;
1250: say(mpage->mlinks->file, "&wait gunzip");
1.124 schwarze 1251: }
1.66 schwarze 1252: ohash_delete(&strings);
1.133 schwarze 1253: ohash_delete(&names);
1.80 schwarze 1254: mpage = ohash_next(&mpages, &pslot);
1.50 kristaps 1255: }
1.107 schwarze 1256:
1257: if (0 == nodb)
1258: SQL_EXEC("END TRANSACTION");
1.130 schwarze 1259: }
1260:
1261: static void
1262: names_check(void)
1263: {
1264: sqlite3_stmt *stmt;
1265: const char *name, *sec, *arch, *key;
1266: int irc;
1267:
1268: sqlite3_prepare_v2(db,
1269: "SELECT name, sec, arch, key FROM ("
1.133 schwarze 1270: "SELECT name AS key, pageid FROM names "
1.130 schwarze 1271: "WHERE bits & ? AND NOT EXISTS ("
1272: "SELECT pageid FROM mlinks "
1.133 schwarze 1273: "WHERE mlinks.pageid == names.pageid "
1274: "AND mlinks.name == names.name"
1.130 schwarze 1275: ")"
1276: ") JOIN ("
1.142 schwarze 1277: "SELECT sec, arch, name, pageid FROM mlinks "
1278: "GROUP BY pageid"
1.130 schwarze 1279: ") USING (pageid);",
1280: -1, &stmt, NULL);
1281:
1.134 schwarze 1282: if (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE))
1283: say("", "%s", sqlite3_errmsg(db));
1.130 schwarze 1284:
1285: while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1.154 schwarze 1286: name = (const char *)sqlite3_column_text(stmt, 0);
1287: sec = (const char *)sqlite3_column_text(stmt, 1);
1288: arch = (const char *)sqlite3_column_text(stmt, 2);
1289: key = (const char *)sqlite3_column_text(stmt, 3);
1.130 schwarze 1290: say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1291: '\0' == *arch ? "" : "/",
1292: '\0' == *arch ? "" : arch, key);
1293: }
1294: sqlite3_finalize(stmt);
1.50 kristaps 1295: }
1.12 schwarze 1296:
1.50 kristaps 1297: static void
1.124 schwarze 1298: parse_cat(struct mpage *mpage, int fd)
1.50 kristaps 1299: {
1300: FILE *stream;
1301: char *line, *p, *title;
1302: size_t len, plen, titlesz;
1.12 schwarze 1303:
1.124 schwarze 1304: stream = (-1 == fd) ?
1305: fopen(mpage->mlinks->file, "r") :
1306: fdopen(fd, "r");
1307: if (NULL == stream) {
1.159 schwarze 1308: if (-1 != fd)
1309: close(fd);
1.56 schwarze 1310: if (warnings)
1.123 schwarze 1311: say(mpage->mlinks->file, "&fopen");
1.50 kristaps 1312: return;
1313: }
1.1 kristaps 1314:
1.50 kristaps 1315: /* Skip to first blank line. */
1.1 kristaps 1316:
1.50 kristaps 1317: while (NULL != (line = fgetln(stream, &len)))
1318: if ('\n' == *line)
1319: break;
1.1 kristaps 1320:
1.50 kristaps 1321: /*
1322: * Assume the first line that is not indented
1323: * is the first section header. Skip to it.
1324: */
1.1 kristaps 1325:
1.50 kristaps 1326: while (NULL != (line = fgetln(stream, &len)))
1327: if ('\n' != *line && ' ' != *line)
1328: break;
1.141 schwarze 1329:
1.50 kristaps 1330: /*
1331: * Read up until the next section into a buffer.
1332: * Strip the leading and trailing newline from each read line,
1333: * appending a trailing space.
1334: * Ignore empty (whitespace-only) lines.
1335: */
1.1 kristaps 1336:
1.50 kristaps 1337: titlesz = 0;
1338: title = NULL;
1.38 schwarze 1339:
1.50 kristaps 1340: while (NULL != (line = fgetln(stream, &len))) {
1341: if (' ' != *line || '\n' != line[len - 1])
1342: break;
1343: while (len > 0 && isspace((unsigned char)*line)) {
1344: line++;
1345: len--;
1346: }
1347: if (1 == len)
1.38 schwarze 1348: continue;
1.50 kristaps 1349: title = mandoc_realloc(title, titlesz + len);
1350: memcpy(title + titlesz, line, len);
1351: titlesz += len;
1352: title[titlesz - 1] = ' ';
1353: }
1.38 schwarze 1354:
1.50 kristaps 1355: /*
1356: * If no page content can be found, or the input line
1357: * is already the next section header, or there is no
1358: * trailing newline, reuse the page title as the page
1359: * description.
1360: */
1.44 kristaps 1361:
1.50 kristaps 1362: if (NULL == title || '\0' == *title) {
1.56 schwarze 1363: if (warnings)
1.82 schwarze 1364: say(mpage->mlinks->file,
1365: "Cannot find NAME section");
1.50 kristaps 1366: fclose(stream);
1367: free(title);
1368: return;
1369: }
1.1 kristaps 1370:
1.50 kristaps 1371: title = mandoc_realloc(title, titlesz + 1);
1372: title[titlesz] = '\0';
1.33 schwarze 1373:
1.50 kristaps 1374: /*
1375: * Skip to the first dash.
1376: * Use the remaining line as the description (no more than 70
1377: * bytes).
1378: */
1.33 schwarze 1379:
1.50 kristaps 1380: if (NULL != (p = strstr(title, "- "))) {
1381: for (p += 2; ' ' == *p || '\b' == *p; p++)
1382: /* Skip to next word. */ ;
1383: } else {
1.56 schwarze 1384: if (warnings)
1.82 schwarze 1385: say(mpage->mlinks->file,
1386: "No dash in title line");
1.50 kristaps 1387: p = title;
1388: }
1.38 schwarze 1389:
1.50 kristaps 1390: plen = strlen(p);
1.1 kristaps 1391:
1.50 kristaps 1392: /* Strip backspace-encoding from line. */
1.1 kristaps 1393:
1.50 kristaps 1394: while (NULL != (line = memchr(p, '\b', plen))) {
1395: len = line - p;
1396: if (0 == len) {
1397: memmove(line, line + 1, plen--);
1398: continue;
1.141 schwarze 1399: }
1.50 kristaps 1400: memmove(line - 1, line + 1, plen - len);
1401: plen -= 2;
1402: }
1.1 kristaps 1403:
1.78 schwarze 1404: mpage->desc = mandoc_strdup(p);
1.50 kristaps 1405: fclose(stream);
1406: free(title);
1407: }
1.1 kristaps 1408:
1.50 kristaps 1409: /*
1410: * Put a type/word pair into the word database for this particular file.
1411: */
1412: static void
1.112 schwarze 1413: putkey(const struct mpage *mpage, char *value, uint64_t type)
1.50 kristaps 1414: {
1.112 schwarze 1415: char *cp;
1.18 kristaps 1416:
1.50 kristaps 1417: assert(NULL != value);
1.112 schwarze 1418: if (TYPE_arch == type)
1419: for (cp = value; *cp; cp++)
1420: if (isupper((unsigned char)*cp))
1421: *cp = _tolower((unsigned char)*cp);
1.78 schwarze 1422: putkeys(mpage, value, strlen(value), type);
1.3 kristaps 1423: }
1424:
1425: /*
1.50 kristaps 1426: * Grok all nodes at or below a certain mdoc node into putkey().
1.3 kristaps 1427: */
1428: static void
1.78 schwarze 1429: putmdockey(const struct mpage *mpage,
1430: const struct mdoc_node *n, uint64_t m)
1.3 kristaps 1431: {
1.18 kristaps 1432:
1.50 kristaps 1433: for ( ; NULL != n; n = n->next) {
1434: if (NULL != n->child)
1.78 schwarze 1435: putmdockey(mpage, n->child, m);
1.50 kristaps 1436: if (MDOC_TEXT == n->type)
1.78 schwarze 1437: putkey(mpage, n->string, m);
1.50 kristaps 1438: }
1439: }
1.18 kristaps 1440:
1.61 schwarze 1441: static void
1.78 schwarze 1442: parse_man(struct mpage *mpage, const struct man_node *n)
1.50 kristaps 1443: {
1444: const struct man_node *head, *body;
1.121 schwarze 1445: char *start, *title;
1.50 kristaps 1446: char byte;
1.121 schwarze 1447: size_t sz;
1.18 kristaps 1448:
1.50 kristaps 1449: if (NULL == n)
1.61 schwarze 1450: return;
1.18 kristaps 1451:
1.50 kristaps 1452: /*
1453: * We're only searching for one thing: the first text child in
1454: * the BODY of a NAME section. Since we don't keep track of
1455: * sections in -man, run some hoops to find out whether we're in
1456: * the correct section or not.
1457: */
1.18 kristaps 1458:
1.50 kristaps 1459: if (MAN_BODY == n->type && MAN_SH == n->tok) {
1460: body = n;
1461: assert(body->parent);
1462: if (NULL != (head = body->parent->head) &&
1.141 schwarze 1463: 1 == head->nchild &&
1464: NULL != (head = (head->child)) &&
1465: MAN_TEXT == head->type &&
1466: 0 == strcmp(head->string, "NAME") &&
1467: NULL != body->child) {
1.3 kristaps 1468:
1.50 kristaps 1469: /*
1470: * Suck the entire NAME section into memory.
1471: * Yes, we might run away.
1472: * But too many manuals have big, spread-out
1473: * NAME sections over many lines.
1474: */
1.3 kristaps 1475:
1.121 schwarze 1476: title = NULL;
1477: man_deroff(&title, body);
1.50 kristaps 1478: if (NULL == title)
1.61 schwarze 1479: return;
1.18 kristaps 1480:
1.141 schwarze 1481: /*
1.50 kristaps 1482: * Go through a special heuristic dance here.
1483: * Conventionally, one or more manual names are
1484: * comma-specified prior to a whitespace, then a
1485: * dash, then a description. Try to puzzle out
1486: * the name parts here.
1487: */
1.18 kristaps 1488:
1.121 schwarze 1489: start = title;
1.50 kristaps 1490: for ( ;; ) {
1491: sz = strcspn(start, " ,");
1492: if ('\0' == start[sz])
1493: break;
1.1 kristaps 1494:
1.50 kristaps 1495: byte = start[sz];
1496: start[sz] = '\0';
1.110 schwarze 1497:
1498: /*
1499: * Assume a stray trailing comma in the
1500: * name list if a name begins with a dash.
1501: */
1502:
1503: if ('-' == start[0] ||
1504: ('\\' == start[0] && '-' == start[1]))
1505: break;
1.1 kristaps 1506:
1.133 schwarze 1507: putkey(mpage, start, NAME_TITLE);
1.1 kristaps 1508:
1.50 kristaps 1509: if (' ' == byte) {
1510: start += sz + 1;
1511: break;
1512: }
1.1 kristaps 1513:
1.50 kristaps 1514: assert(',' == byte);
1515: start += sz + 1;
1516: while (' ' == *start)
1517: start++;
1518: }
1.1 kristaps 1519:
1.121 schwarze 1520: if (start == title) {
1.133 schwarze 1521: putkey(mpage, start, NAME_TITLE);
1.50 kristaps 1522: free(title);
1.61 schwarze 1523: return;
1.50 kristaps 1524: }
1.1 kristaps 1525:
1.50 kristaps 1526: while (isspace((unsigned char)*start))
1527: start++;
1.1 kristaps 1528:
1.50 kristaps 1529: if (0 == strncmp(start, "-", 1))
1530: start += 1;
1531: else if (0 == strncmp(start, "\\-\\-", 4))
1532: start += 4;
1533: else if (0 == strncmp(start, "\\-", 2))
1534: start += 2;
1535: else if (0 == strncmp(start, "\\(en", 4))
1536: start += 4;
1537: else if (0 == strncmp(start, "\\(em", 4))
1538: start += 4;
1.1 kristaps 1539:
1.50 kristaps 1540: while (' ' == *start)
1541: start++;
1.1 kristaps 1542:
1.78 schwarze 1543: mpage->desc = mandoc_strdup(start);
1.50 kristaps 1544: free(title);
1.61 schwarze 1545: return;
1.50 kristaps 1546: }
1547: }
1.1 kristaps 1548:
1.77 schwarze 1549: for (n = n->child; n; n = n->next) {
1.78 schwarze 1550: if (NULL != mpage->desc)
1.77 schwarze 1551: break;
1.78 schwarze 1552: parse_man(mpage, n);
1.77 schwarze 1553: }
1.1 kristaps 1554: }
1555:
1556: static void
1.172 schwarze 1557: parse_mdoc(struct mpage *mpage, const struct mdoc_meta *meta,
1558: const struct mdoc_node *n)
1.1 kristaps 1559: {
1560:
1.50 kristaps 1561: assert(NULL != n);
1562: for (n = n->child; NULL != n; n = n->next) {
1563: switch (n->type) {
1.141 schwarze 1564: case MDOC_ELEM:
1.50 kristaps 1565: /* FALLTHROUGH */
1.141 schwarze 1566: case MDOC_BLOCK:
1.50 kristaps 1567: /* FALLTHROUGH */
1.141 schwarze 1568: case MDOC_HEAD:
1.50 kristaps 1569: /* FALLTHROUGH */
1.141 schwarze 1570: case MDOC_BODY:
1.50 kristaps 1571: /* FALLTHROUGH */
1.141 schwarze 1572: case MDOC_TAIL:
1.50 kristaps 1573: if (NULL != mdocs[n->tok].fp)
1.172 schwarze 1574: if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1.50 kristaps 1575: break;
1.68 schwarze 1576: if (mdocs[n->tok].mask)
1.78 schwarze 1577: putmdockey(mpage, n->child,
1578: mdocs[n->tok].mask);
1.50 kristaps 1579: break;
1580: default:
1581: assert(MDOC_ROOT != n->type);
1582: continue;
1583: }
1584: if (NULL != n->child)
1.172 schwarze 1585: parse_mdoc(mpage, meta, n);
1.1 kristaps 1586: }
1587: }
1588:
1.25 schwarze 1589: static int
1.172 schwarze 1590: parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_meta *meta,
1591: const struct mdoc_node *n)
1.1 kristaps 1592: {
1593: const char *start, *end;
1594: size_t sz;
1.25 schwarze 1595:
1.50 kristaps 1596: if (SEC_SYNOPSIS != n->sec ||
1.141 schwarze 1597: NULL == (n = n->child) ||
1598: MDOC_TEXT != n->type)
1.25 schwarze 1599: return(0);
1.1 kristaps 1600:
1601: /*
1602: * Only consider those `Fd' macro fields that begin with an
1603: * "inclusion" token (versus, e.g., #define).
1604: */
1.50 kristaps 1605:
1.1 kristaps 1606: if (strcmp("#include", n->string))
1.25 schwarze 1607: return(0);
1.1 kristaps 1608:
1609: if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1.25 schwarze 1610: return(0);
1.1 kristaps 1611:
1612: /*
1613: * Strip away the enclosing angle brackets and make sure we're
1614: * not zero-length.
1615: */
1616:
1617: start = n->string;
1618: if ('<' == *start || '"' == *start)
1619: start++;
1620:
1621: if (0 == (sz = strlen(start)))
1.25 schwarze 1622: return(0);
1.1 kristaps 1623:
1624: end = &start[(int)sz - 1];
1625: if ('>' == *end || '"' == *end)
1626: end--;
1627:
1.50 kristaps 1628: if (end > start)
1.78 schwarze 1629: putkeys(mpage, start, end - start + 1, TYPE_In);
1.92 schwarze 1630: return(0);
1.1 kristaps 1631: }
1632:
1.25 schwarze 1633: static int
1.172 schwarze 1634: parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_meta *meta,
1635: const struct mdoc_node *n)
1.1 kristaps 1636: {
1.112 schwarze 1637: char *cp;
1.1 kristaps 1638:
1.50 kristaps 1639: if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1.25 schwarze 1640: return(0);
1641:
1.141 schwarze 1642: /*
1.50 kristaps 1643: * Parse: .Fn "struct type *name" "char *arg".
1.141 schwarze 1644: * First strip away pointer symbol.
1.50 kristaps 1645: * Then store the function name, then type.
1.141 schwarze 1646: * Finally, store the arguments.
1.50 kristaps 1647: */
1.1 kristaps 1648:
1.50 kristaps 1649: if (NULL == (cp = strrchr(n->string, ' ')))
1650: cp = n->string;
1.1 kristaps 1651:
1652: while ('*' == *cp)
1653: cp++;
1654:
1.78 schwarze 1655: putkey(mpage, cp, TYPE_Fn);
1.173 ! schwarze 1656: if (n->sec == SEC_SYNOPSIS)
! 1657: putkey(mpage, cp, NAME_SYN);
1.25 schwarze 1658:
1.50 kristaps 1659: if (n->string < cp)
1.78 schwarze 1660: putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1.25 schwarze 1661:
1.50 kristaps 1662: for (n = n->next; NULL != n; n = n->next)
1663: if (MDOC_TEXT == n->type)
1.78 schwarze 1664: putkey(mpage, n->string, TYPE_Fa);
1.25 schwarze 1665:
1.173 ! schwarze 1666: return(0);
! 1667: }
! 1668:
! 1669: static int
! 1670: parse_mdoc_Fo(struct mpage *mpage, const struct mdoc_meta *meta,
! 1671: const struct mdoc_node *n)
! 1672: {
! 1673:
! 1674: putmdockey(mpage, n->child, TYPE_Fn);
! 1675: if (n->sec == SEC_SYNOPSIS)
! 1676: putmdockey(mpage, n->child, NAME_SYN);
1.25 schwarze 1677: return(0);
1.1 kristaps 1678: }
1679:
1.25 schwarze 1680: static int
1.172 schwarze 1681: parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_meta *meta,
1682: const struct mdoc_node *n)
1.1 kristaps 1683: {
1.67 schwarze 1684: char *cp;
1.1 kristaps 1685:
1686: if (NULL == (n = n->child))
1.25 schwarze 1687: return(0);
1.1 kristaps 1688:
1.67 schwarze 1689: if (NULL == n->next) {
1.78 schwarze 1690: putkey(mpage, n->string, TYPE_Xr);
1.67 schwarze 1691: return(0);
1692: }
1693:
1.120 schwarze 1694: mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1.78 schwarze 1695: putkey(mpage, cp, TYPE_Xr);
1.67 schwarze 1696: free(cp);
1697: return(0);
1.1 kristaps 1698: }
1699:
1.25 schwarze 1700: static int
1.172 schwarze 1701: parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_meta *meta,
1702: const struct mdoc_node *n)
1.1 kristaps 1703: {
1704:
1.122 schwarze 1705: if (MDOC_BODY == n->type)
1706: mdoc_deroff(&mpage->desc, n);
1707: return(0);
1.1 kristaps 1708: }
1709:
1.25 schwarze 1710: static int
1.172 schwarze 1711: parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_meta *meta,
1712: const struct mdoc_node *n)
1.1 kristaps 1713: {
1714:
1.129 schwarze 1715: if (SEC_NAME == n->sec)
1.133 schwarze 1716: putmdockey(mpage, n->child, NAME_TITLE);
1.172 schwarze 1717: else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type) {
1718: if (n->child == NULL)
1719: putkey(mpage, meta->name, NAME_SYN);
1720: else
1721: putmdockey(mpage, n->child, NAME_SYN);
1722: }
1.129 schwarze 1723: return(0);
1.1 kristaps 1724: }
1725:
1.25 schwarze 1726: static int
1.172 schwarze 1727: parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_meta *meta,
1728: const struct mdoc_node *n)
1.1 kristaps 1729: {
1730:
1.25 schwarze 1731: return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1.1 kristaps 1732: }
1733:
1.50 kristaps 1734: static int
1.172 schwarze 1735: parse_mdoc_head(struct mpage *mpage, const struct mdoc_meta *meta,
1736: const struct mdoc_node *n)
1.1 kristaps 1737: {
1738:
1.50 kristaps 1739: return(MDOC_HEAD == n->type);
1740: }
1.1 kristaps 1741:
1.50 kristaps 1742: static int
1.172 schwarze 1743: parse_mdoc_body(struct mpage *mpage, const struct mdoc_meta *meta,
1744: const struct mdoc_node *n)
1.50 kristaps 1745: {
1.1 kristaps 1746:
1.50 kristaps 1747: return(MDOC_BODY == n->type);
1.1 kristaps 1748: }
1749:
1.50 kristaps 1750: /*
1.66 schwarze 1751: * Add a string to the hash table for the current manual.
1752: * Each string has a bitmask telling which macros it belongs to.
1753: * When we finish the manual, we'll dump the table.
1.50 kristaps 1754: */
1755: static void
1.78 schwarze 1756: putkeys(const struct mpage *mpage,
1757: const char *cp, size_t sz, uint64_t v)
1.50 kristaps 1758: {
1.133 schwarze 1759: struct ohash *htab;
1.50 kristaps 1760: struct str *s;
1.111 schwarze 1761: const char *end;
1.65 schwarze 1762: unsigned int slot;
1.111 schwarze 1763: int i;
1.25 schwarze 1764:
1.50 kristaps 1765: if (0 == sz)
1766: return;
1.111 schwarze 1767:
1.133 schwarze 1768: if (TYPE_Nm & v) {
1769: htab = &names;
1770: v &= name_mask;
1.169 schwarze 1771: if (v & NAME_FIRST)
1772: name_mask &= ~NAME_FIRST;
1.133 schwarze 1773: if (debug > 1)
1774: say(mpage->mlinks->file,
1775: "Adding name %*s", sz, cp);
1776: } else {
1777: htab = &strings;
1778: if (debug > 1)
1779: for (i = 0; i < mansearch_keymax; i++)
1.163 schwarze 1780: if ((uint64_t)1 << i & v)
1.133 schwarze 1781: say(mpage->mlinks->file,
1782: "Adding key %s=%*s",
1783: mansearch_keynames[i], sz, cp);
1.111 schwarze 1784: }
1.25 schwarze 1785:
1.65 schwarze 1786: end = cp + sz;
1.133 schwarze 1787: slot = ohash_qlookupi(htab, cp, &end);
1788: s = ohash_find(htab, slot);
1.25 schwarze 1789:
1.78 schwarze 1790: if (NULL != s && mpage == s->mpage) {
1.50 kristaps 1791: s->mask |= v;
1792: return;
1793: } else if (NULL == s) {
1.144 schwarze 1794: s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1.50 kristaps 1795: memcpy(s->key, cp, sz);
1.133 schwarze 1796: ohash_insert(htab, slot, s);
1.1 kristaps 1797: }
1.78 schwarze 1798: s->mpage = mpage;
1.50 kristaps 1799: s->mask = v;
1.1 kristaps 1800: }
1801:
1.50 kristaps 1802: /*
1803: * Take a Unicode codepoint and produce its UTF-8 encoding.
1804: * This isn't the best way to do this, but it works.
1805: * The magic numbers are from the UTF-8 packaging.
1806: * They're not as scary as they seem: read the UTF-8 spec for details.
1807: */
1808: static size_t
1809: utf8(unsigned int cp, char out[7])
1.1 kristaps 1810: {
1.50 kristaps 1811: size_t rc;
1.1 kristaps 1812:
1.50 kristaps 1813: rc = 0;
1814: if (cp <= 0x0000007F) {
1815: rc = 1;
1816: out[0] = (char)cp;
1817: } else if (cp <= 0x000007FF) {
1818: rc = 2;
1819: out[0] = (cp >> 6 & 31) | 192;
1820: out[1] = (cp & 63) | 128;
1821: } else if (cp <= 0x0000FFFF) {
1822: rc = 3;
1823: out[0] = (cp >> 12 & 15) | 224;
1824: out[1] = (cp >> 6 & 63) | 128;
1825: out[2] = (cp & 63) | 128;
1826: } else if (cp <= 0x001FFFFF) {
1827: rc = 4;
1828: out[0] = (cp >> 18 & 7) | 240;
1829: out[1] = (cp >> 12 & 63) | 128;
1830: out[2] = (cp >> 6 & 63) | 128;
1831: out[3] = (cp & 63) | 128;
1832: } else if (cp <= 0x03FFFFFF) {
1833: rc = 5;
1834: out[0] = (cp >> 24 & 3) | 248;
1835: out[1] = (cp >> 18 & 63) | 128;
1836: out[2] = (cp >> 12 & 63) | 128;
1837: out[3] = (cp >> 6 & 63) | 128;
1838: out[4] = (cp & 63) | 128;
1839: } else if (cp <= 0x7FFFFFFF) {
1840: rc = 6;
1841: out[0] = (cp >> 30 & 1) | 252;
1842: out[1] = (cp >> 24 & 63) | 128;
1843: out[2] = (cp >> 18 & 63) | 128;
1844: out[3] = (cp >> 12 & 63) | 128;
1845: out[4] = (cp >> 6 & 63) | 128;
1846: out[5] = (cp & 63) | 128;
1847: } else
1.1 kristaps 1848: return(0);
1849:
1.50 kristaps 1850: out[rc] = '\0';
1851: return(rc);
1852: }
1.1 kristaps 1853:
1.50 kristaps 1854: /*
1.96 schwarze 1855: * Store the rendered version of a key, or alias the pointer
1856: * if the key contains no escape sequences.
1.50 kristaps 1857: */
1858: static void
1.96 schwarze 1859: render_key(struct mchars *mc, struct str *key)
1.50 kristaps 1860: {
1861: size_t sz, bsz, pos;
1.114 schwarze 1862: char utfbuf[7], res[6];
1.50 kristaps 1863: char *buf;
1864: const char *seq, *cpp, *val;
1865: int len, u;
1866: enum mandoc_esc esc;
1867:
1.96 schwarze 1868: assert(NULL == key->rendered);
1.50 kristaps 1869:
1870: res[0] = '\\';
1871: res[1] = '\t';
1872: res[2] = ASCII_NBRSP;
1873: res[3] = ASCII_HYPH;
1.114 schwarze 1874: res[4] = ASCII_BREAK;
1875: res[5] = '\0';
1.1 kristaps 1876:
1.50 kristaps 1877: val = key->key;
1878: bsz = strlen(val);
1.46 kristaps 1879:
1.50 kristaps 1880: /*
1881: * Pre-check: if we have no stop-characters, then set the
1882: * pointer as ourselvse and get out of here.
1883: */
1884: if (strcspn(val, res) == bsz) {
1.96 schwarze 1885: key->rendered = key->key;
1.50 kristaps 1886: return;
1.141 schwarze 1887: }
1.46 kristaps 1888:
1.50 kristaps 1889: /* Pre-allocate by the length of the input */
1.46 kristaps 1890:
1.50 kristaps 1891: buf = mandoc_malloc(++bsz);
1892: pos = 0;
1.46 kristaps 1893:
1.50 kristaps 1894: while ('\0' != *val) {
1895: /*
1896: * Halt on the first escape sequence.
1897: * This also halts on the end of string, in which case
1898: * we just copy, fallthrough, and exit the loop.
1899: */
1900: if ((sz = strcspn(val, res)) > 0) {
1901: memcpy(&buf[pos], val, sz);
1902: pos += sz;
1903: val += sz;
1904: }
1.46 kristaps 1905:
1.114 schwarze 1906: switch (*val) {
1.141 schwarze 1907: case ASCII_HYPH:
1.50 kristaps 1908: buf[pos++] = '-';
1909: val++;
1910: continue;
1.141 schwarze 1911: case '\t':
1.114 schwarze 1912: /* FALLTHROUGH */
1.141 schwarze 1913: case ASCII_NBRSP:
1.50 kristaps 1914: buf[pos++] = ' ';
1915: val++;
1.114 schwarze 1916: /* FALLTHROUGH */
1.141 schwarze 1917: case ASCII_BREAK:
1.50 kristaps 1918: continue;
1.114 schwarze 1919: default:
1920: break;
1921: }
1922: if ('\\' != *val)
1.50 kristaps 1923: break;
1.46 kristaps 1924:
1.50 kristaps 1925: /* Read past the slash. */
1.46 kristaps 1926:
1.50 kristaps 1927: val++;
1.46 kristaps 1928:
1.50 kristaps 1929: /*
1930: * Parse the escape sequence and see if it's a
1931: * predefined character or special character.
1932: */
1.95 schwarze 1933:
1.141 schwarze 1934: esc = mandoc_escape((const char **)&val,
1935: &seq, &len);
1.50 kristaps 1936: if (ESCAPE_ERROR == esc)
1937: break;
1938: if (ESCAPE_SPECIAL != esc)
1939: continue;
1.1 kristaps 1940:
1.50 kristaps 1941: /*
1.95 schwarze 1942: * Render the special character
1943: * as either UTF-8 or ASCII.
1.50 kristaps 1944: */
1.95 schwarze 1945:
1946: if (write_utf8) {
1.167 schwarze 1947: if ((u = mchars_spec2cp(mc, seq, len)) <= 0)
1.95 schwarze 1948: continue;
1949: cpp = utfbuf;
1950: if (0 == (sz = utf8(u, utfbuf)))
1951: continue;
1952: sz = strlen(cpp);
1953: } else {
1954: cpp = mchars_spec2str(mc, seq, len, &sz);
1955: if (NULL == cpp)
1956: continue;
1957: if (ASCII_NBRSP == *cpp) {
1958: cpp = " ";
1959: sz = 1;
1960: }
1961: }
1.1 kristaps 1962:
1.50 kristaps 1963: /* Copy the rendered glyph into the stream. */
1.1 kristaps 1964:
1.50 kristaps 1965: bsz += sz;
1966: buf = mandoc_realloc(buf, bsz);
1967: memcpy(&buf[pos], cpp, sz);
1968: pos += sz;
1.1 kristaps 1969: }
1970:
1.50 kristaps 1971: buf[pos] = '\0';
1.96 schwarze 1972: key->rendered = buf;
1.1 kristaps 1973: }
1974:
1.118 schwarze 1975: static void
1976: dbadd_mlink(const struct mlink *mlink)
1977: {
1978: size_t i;
1979:
1980: i = 1;
1981: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1982: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1983: SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1.137 schwarze 1984: SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1.118 schwarze 1985: SQL_STEP(stmts[STMT_INSERT_LINK]);
1986: sqlite3_reset(stmts[STMT_INSERT_LINK]);
1.169 schwarze 1987: }
1988:
1989: static void
1990: dbadd_mlink_name(const struct mlink *mlink)
1991: {
1992: size_t i;
1993:
1994: dbadd_mlink(mlink);
1.160 schwarze 1995:
1996: i = 1;
1.169 schwarze 1997: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, NAME_FILE & NAME_MASK);
1.160 schwarze 1998: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
1999: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2000: SQL_STEP(stmts[STMT_INSERT_NAME]);
2001: sqlite3_reset(stmts[STMT_INSERT_NAME]);
1.118 schwarze 2002: }
2003:
1.14 schwarze 2004: /*
1.50 kristaps 2005: * Flush the current page's terms (and their bits) into the database.
2006: * Wrap the entire set of additions in a transaction to make sqlite be a
2007: * little faster.
1.96 schwarze 2008: * Also, handle escape sequences at the last possible moment.
1.14 schwarze 2009: */
2010: static void
1.118 schwarze 2011: dbadd(struct mpage *mpage, struct mchars *mc)
1.14 schwarze 2012: {
1.87 schwarze 2013: struct mlink *mlink;
1.50 kristaps 2014: struct str *key;
1.52 kristaps 2015: size_t i;
1.66 schwarze 2016: unsigned int slot;
1.14 schwarze 2017:
1.128 schwarze 2018: mlink = mpage->mlinks;
1.43 kristaps 2019:
1.128 schwarze 2020: if (nodb) {
1.147 schwarze 2021: for (key = ohash_first(&names, &slot); NULL != key;
2022: key = ohash_next(&names, &slot)) {
2023: if (key->rendered != key->key)
2024: free(key->rendered);
2025: free(key);
2026: }
2027: for (key = ohash_first(&strings, &slot); NULL != key;
2028: key = ohash_next(&strings, &slot)) {
2029: if (key->rendered != key->key)
2030: free(key->rendered);
2031: free(key);
2032: }
1.145 schwarze 2033: if (0 == debug)
2034: return;
1.128 schwarze 2035: while (NULL != mlink) {
2036: fputs(mlink->name, stdout);
2037: if (NULL == mlink->next ||
2038: strcmp(mlink->dsec, mlink->next->dsec) ||
2039: strcmp(mlink->fsec, mlink->next->fsec) ||
2040: strcmp(mlink->arch, mlink->next->arch)) {
2041: putchar('(');
2042: if ('\0' == *mlink->dsec)
2043: fputs(mlink->fsec, stdout);
2044: else
2045: fputs(mlink->dsec, stdout);
2046: if ('\0' != *mlink->arch)
2047: printf("/%s", mlink->arch);
2048: putchar(')');
2049: }
2050: mlink = mlink->next;
2051: if (NULL != mlink)
2052: fputs(", ", stdout);
2053: }
1.132 schwarze 2054: printf(" - %s\n", mpage->desc);
1.14 schwarze 2055: return;
1.128 schwarze 2056: }
2057:
2058: if (debug)
2059: say(mlink->file, "Adding to database");
1.28 kristaps 2060:
1.148 schwarze 2061: i = strlen(mpage->desc) + 1;
2062: key = mandoc_calloc(1, sizeof(struct str) + i);
2063: memcpy(key->key, mpage->desc, i);
2064: render_key(mc, key);
2065:
1.52 kristaps 2066: i = 1;
1.148 schwarze 2067: SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, key->rendered);
1.161 schwarze 2068: SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
1.81 schwarze 2069: SQL_STEP(stmts[STMT_INSERT_PAGE]);
1.137 schwarze 2070: mpage->pageid = sqlite3_last_insert_rowid(db);
1.81 schwarze 2071: sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1.148 schwarze 2072:
2073: if (key->rendered != key->key)
2074: free(key->rendered);
2075: free(key);
1.81 schwarze 2076:
1.128 schwarze 2077: while (NULL != mlink) {
1.118 schwarze 2078: dbadd_mlink(mlink);
1.128 schwarze 2079: mlink = mlink->next;
2080: }
1.134 schwarze 2081: mlink = mpage->mlinks;
1.50 kristaps 2082:
1.133 schwarze 2083: for (key = ohash_first(&names, &slot); NULL != key;
2084: key = ohash_next(&names, &slot)) {
2085: assert(key->mpage == mpage);
2086: if (NULL == key->rendered)
2087: render_key(mc, key);
2088: i = 1;
2089: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2090: SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->rendered);
1.137 schwarze 2091: SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
1.133 schwarze 2092: SQL_STEP(stmts[STMT_INSERT_NAME]);
2093: sqlite3_reset(stmts[STMT_INSERT_NAME]);
2094: if (key->rendered != key->key)
2095: free(key->rendered);
2096: free(key);
2097: }
1.66 schwarze 2098: for (key = ohash_first(&strings, &slot); NULL != key;
2099: key = ohash_next(&strings, &slot)) {
1.78 schwarze 2100: assert(key->mpage == mpage);
1.96 schwarze 2101: if (NULL == key->rendered)
2102: render_key(mc, key);
1.52 kristaps 2103: i = 1;
2104: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1.96 schwarze 2105: SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
1.137 schwarze 2106: SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
1.52 kristaps 2107: SQL_STEP(stmts[STMT_INSERT_KEY]);
1.50 kristaps 2108: sqlite3_reset(stmts[STMT_INSERT_KEY]);
1.96 schwarze 2109: if (key->rendered != key->key)
2110: free(key->rendered);
1.66 schwarze 2111: free(key);
1.38 schwarze 2112: }
1.14 schwarze 2113: }
2114:
1.5 kristaps 2115: static void
1.59 schwarze 2116: dbprune(void)
1.5 kristaps 2117: {
1.78 schwarze 2118: struct mpage *mpage;
1.82 schwarze 2119: struct mlink *mlink;
1.52 kristaps 2120: size_t i;
1.80 schwarze 2121: unsigned int slot;
1.5 kristaps 2122:
1.106 schwarze 2123: if (0 == nodb)
2124: SQL_EXEC("BEGIN TRANSACTION");
1.12 schwarze 2125:
1.106 schwarze 2126: for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2127: mpage = ohash_next(&mpages, &slot)) {
1.82 schwarze 2128: mlink = mpage->mlinks;
1.125 schwarze 2129: if (debug)
1.106 schwarze 2130: say(mlink->file, "Deleting from database");
2131: if (nodb)
2132: continue;
2133: for ( ; NULL != mlink; mlink = mlink->next) {
2134: i = 1;
2135: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2136: i, mlink->dsec);
2137: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2138: i, mlink->arch);
2139: SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2140: i, mlink->name);
2141: SQL_STEP(stmts[STMT_DELETE_PAGE]);
2142: sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2143: }
1.5 kristaps 2144: }
1.106 schwarze 2145:
2146: if (0 == nodb)
2147: SQL_EXEC("END TRANSACTION");
1.5 kristaps 2148: }
2149:
1.4 kristaps 2150: /*
1.50 kristaps 2151: * Close an existing database and its prepared statements.
2152: * If "real" is not set, rename the temporary file into the real one.
1.4 kristaps 2153: */
1.35 kristaps 2154: static void
1.59 schwarze 2155: dbclose(int real)
1.4 kristaps 2156: {
1.50 kristaps 2157: size_t i;
1.115 schwarze 2158: int status;
2159: pid_t child;
1.4 kristaps 2160:
1.50 kristaps 2161: if (nodb)
1.38 schwarze 2162: return;
1.50 kristaps 2163:
2164: for (i = 0; i < STMT__MAX; i++) {
2165: sqlite3_finalize(stmts[i]);
2166: stmts[i] = NULL;
1.4 kristaps 2167: }
2168:
1.50 kristaps 2169: sqlite3_close(db);
2170: db = NULL;
1.12 schwarze 2171:
1.50 kristaps 2172: if (real)
2173: return;
1.12 schwarze 2174:
1.115 schwarze 2175: if ('\0' == *tempfilename) {
2176: if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2177: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2178: say(MANDOC_DB, "&rename");
1.115 schwarze 2179: }
2180: return;
2181: }
2182:
2183: switch (child = fork()) {
1.141 schwarze 2184: case -1:
1.115 schwarze 2185: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2186: say("", "&fork cmp");
1.115 schwarze 2187: return;
1.141 schwarze 2188: case 0:
1.115 schwarze 2189: execlp("cmp", "cmp", "-s",
2190: tempfilename, MANDOC_DB, NULL);
1.123 schwarze 2191: say("", "&exec cmp");
1.115 schwarze 2192: exit(0);
2193: default:
2194: break;
2195: }
2196: if (-1 == waitpid(child, &status, 0)) {
2197: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2198: say("", "&wait cmp");
1.115 schwarze 2199: } else if (WIFSIGNALED(status)) {
2200: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2201: say("", "cmp died from signal %d", WTERMSIG(status));
1.115 schwarze 2202: } else if (WEXITSTATUS(status)) {
1.59 schwarze 2203: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 schwarze 2204: say(MANDOC_DB,
2205: "Data changed, but cannot replace database");
2206: }
2207:
2208: *strrchr(tempfilename, '/') = '\0';
2209: switch (child = fork()) {
1.141 schwarze 2210: case -1:
1.115 schwarze 2211: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2212: say("", "&fork rm");
1.115 schwarze 2213: return;
1.141 schwarze 2214: case 0:
1.115 schwarze 2215: execlp("rm", "rm", "-rf", tempfilename, NULL);
1.123 schwarze 2216: say("", "&exec rm");
1.115 schwarze 2217: exit((int)MANDOCLEVEL_SYSERR);
2218: default:
2219: break;
2220: }
2221: if (-1 == waitpid(child, &status, 0)) {
2222: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2223: say("", "&wait rm");
1.115 schwarze 2224: } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2225: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2226: say("", "%s: Cannot remove temporary directory",
2227: tempfilename);
1.59 schwarze 2228: }
1.50 kristaps 2229: }
1.14 schwarze 2230:
1.50 kristaps 2231: /*
2232: * This is straightforward stuff.
2233: * Open a database connection to a "temporary" database, then open a set
2234: * of prepared statements we'll use over and over again.
2235: * If "real" is set, we use the existing database; if not, we truncate a
2236: * temporary one.
2237: * Must be matched by dbclose().
2238: */
2239: static int
1.59 schwarze 2240: dbopen(int real)
1.50 kristaps 2241: {
1.115 schwarze 2242: const char *sql;
1.50 kristaps 2243: int rc, ofl;
1.12 schwarze 2244:
1.141 schwarze 2245: if (nodb)
1.50 kristaps 2246: return(1);
1.12 schwarze 2247:
1.115 schwarze 2248: *tempfilename = '\0';
1.63 schwarze 2249: ofl = SQLITE_OPEN_READWRITE;
1.115 schwarze 2250:
2251: if (real) {
2252: rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2253: if (SQLITE_OK != rc) {
1.63 schwarze 2254: exitcode = (int)MANDOCLEVEL_SYSERR;
1.149 schwarze 2255: if (SQLITE_CANTOPEN != rc)
2256: say(MANDOC_DB, "%s", sqlite3_errstr(rc));
1.63 schwarze 2257: return(0);
2258: }
1.115 schwarze 2259: goto prepare_statements;
2260: }
2261:
2262: ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
1.45 kristaps 2263:
1.115 schwarze 2264: remove(MANDOC_DB "~");
2265: rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
1.141 schwarze 2266: if (SQLITE_OK == rc)
1.115 schwarze 2267: goto create_tables;
1.116 schwarze 2268: if (MPARSE_QUICK & mparse_options) {
1.59 schwarze 2269: exitcode = (int)MANDOCLEVEL_SYSERR;
1.146 schwarze 2270: say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
1.50 kristaps 2271: return(0);
2272: }
1.12 schwarze 2273:
1.143 schwarze 2274: (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2275: sizeof(tempfilename));
1.115 schwarze 2276: if (NULL == mkdtemp(tempfilename)) {
2277: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2278: say("", "&%s", tempfilename);
1.115 schwarze 2279: return(0);
2280: }
1.143 schwarze 2281: (void)strlcat(tempfilename, "/" MANDOC_DB,
2282: sizeof(tempfilename));
1.115 schwarze 2283: rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2284: if (SQLITE_OK != rc) {
1.59 schwarze 2285: exitcode = (int)MANDOCLEVEL_SYSERR;
1.146 schwarze 2286: say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
1.50 kristaps 2287: return(0);
2288: }
1.12 schwarze 2289:
1.115 schwarze 2290: create_tables:
1.81 schwarze 2291: sql = "CREATE TABLE \"mpages\" (\n"
1.132 schwarze 2292: " \"desc\" TEXT NOT NULL,\n"
1.50 kristaps 2293: " \"form\" INTEGER NOT NULL,\n"
1.137 schwarze 2294: " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
1.50 kristaps 2295: ");\n"
2296: "\n"
1.81 schwarze 2297: "CREATE TABLE \"mlinks\" (\n"
2298: " \"sec\" TEXT NOT NULL,\n"
2299: " \"arch\" TEXT NOT NULL,\n"
2300: " \"name\" TEXT NOT NULL,\n"
1.137 schwarze 2301: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.109 schwarze 2302: "ON DELETE CASCADE\n"
1.81 schwarze 2303: ");\n"
1.136 schwarze 2304: "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
1.81 schwarze 2305: "\n"
1.133 schwarze 2306: "CREATE TABLE \"names\" (\n"
2307: " \"bits\" INTEGER NOT NULL,\n"
2308: " \"name\" TEXT NOT NULL,\n"
1.137 schwarze 2309: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.133 schwarze 2310: "ON DELETE CASCADE\n"
2311: ");\n"
2312: "\n"
1.50 kristaps 2313: "CREATE TABLE \"keys\" (\n"
2314: " \"bits\" INTEGER NOT NULL,\n"
2315: " \"key\" TEXT NOT NULL,\n"
1.137 schwarze 2316: " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
1.109 schwarze 2317: "ON DELETE CASCADE\n"
1.136 schwarze 2318: ");\n"
2319: "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
1.14 schwarze 2320:
1.50 kristaps 2321: if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
1.59 schwarze 2322: exitcode = (int)MANDOCLEVEL_SYSERR;
1.115 schwarze 2323: say(MANDOC_DB, "%s", sqlite3_errmsg(db));
1.146 schwarze 2324: sqlite3_close(db);
1.50 kristaps 2325: return(0);
2326: }
1.4 kristaps 2327:
1.57 schwarze 2328: prepare_statements:
1.146 schwarze 2329: if (SQLITE_OK != sqlite3_exec(db,
2330: "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2331: exitcode = (int)MANDOCLEVEL_SYSERR;
2332: say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2333: sqlite3_errmsg(db));
2334: sqlite3_close(db);
2335: return(0);
2336: }
2337:
1.137 schwarze 2338: sql = "DELETE FROM mpages WHERE pageid IN "
1.106 schwarze 2339: "(SELECT pageid FROM mlinks WHERE "
2340: "sec=? AND arch=? AND name=?)";
1.81 schwarze 2341: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2342: sql = "INSERT INTO mpages "
1.132 schwarze 2343: "(desc,form) VALUES (?,?)";
1.81 schwarze 2344: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2345: sql = "INSERT INTO mlinks "
1.104 schwarze 2346: "(sec,arch,name,pageid) VALUES (?,?,?,?)";
1.81 schwarze 2347: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
1.133 schwarze 2348: sql = "INSERT INTO names "
2349: "(bits,name,pageid) VALUES (?,?,?)";
2350: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
1.50 kristaps 2351: sql = "INSERT INTO keys "
1.81 schwarze 2352: "(bits,key,pageid) VALUES (?,?,?)";
1.50 kristaps 2353: sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
1.70 schwarze 2354:
2355: #ifndef __APPLE__
2356: /*
2357: * When opening a new database, we can turn off
2358: * synchronous mode for much better performance.
2359: */
2360:
1.146 schwarze 2361: if (real && SQLITE_OK != sqlite3_exec(db,
2362: "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2363: exitcode = (int)MANDOCLEVEL_SYSERR;
2364: say(MANDOC_DB, "PRAGMA synchronous: %s",
1.162 schwarze 2365: sqlite3_errmsg(db));
1.146 schwarze 2366: sqlite3_close(db);
2367: return(0);
2368: }
1.70 schwarze 2369: #endif
2370:
1.50 kristaps 2371: return(1);
2372: }
1.5 kristaps 2373:
1.50 kristaps 2374: static void *
1.152 schwarze 2375: hash_calloc(size_t n, size_t sz, void *arg)
1.50 kristaps 2376: {
1.12 schwarze 2377:
1.152 schwarze 2378: return(mandoc_calloc(n, sz));
1.50 kristaps 2379: }
1.12 schwarze 2380:
1.50 kristaps 2381: static void *
2382: hash_alloc(size_t sz, void *arg)
2383: {
1.5 kristaps 2384:
1.50 kristaps 2385: return(mandoc_malloc(sz));
2386: }
1.41 kristaps 2387:
1.50 kristaps 2388: static void
1.152 schwarze 2389: hash_free(void *p, void *arg)
1.50 kristaps 2390: {
1.4 kristaps 2391:
1.50 kristaps 2392: free(p);
1.4 kristaps 2393: }
2394:
1.50 kristaps 2395: static int
1.165 schwarze 2396: set_basedir(const char *targetdir, int report_baddir)
1.4 kristaps 2397: {
1.59 schwarze 2398: static char startdir[PATH_MAX];
1.151 schwarze 2399: static int getcwd_status; /* 1 = ok, 2 = failure */
2400: static int chdir_status; /* 1 = changed directory */
1.150 schwarze 2401: char *cp;
1.4 kristaps 2402:
1.59 schwarze 2403: /*
1.151 schwarze 2404: * Remember the original working directory, if possible.
2405: * This will be needed if the second or a later directory
2406: * on the command line is given as a relative path.
2407: * Do not error out if the current directory is not
2408: * searchable: Maybe it won't be needed after all.
2409: */
2410: if (0 == getcwd_status) {
2411: if (NULL == getcwd(startdir, sizeof(startdir))) {
2412: getcwd_status = 2;
2413: (void)strlcpy(startdir, strerror(errno),
2414: sizeof(startdir));
2415: } else
2416: getcwd_status = 1;
2417: }
2418:
2419: /*
2420: * We are leaving the old base directory.
2421: * Do not use it any longer, not even for messages.
2422: */
2423: *basedir = '\0';
2424:
2425: /*
2426: * If and only if the directory was changed earlier and
2427: * the next directory to process is given as a relative path,
2428: * first go back, or bail out if that is impossible.
1.59 schwarze 2429: */
1.151 schwarze 2430: if (chdir_status && '/' != *targetdir) {
2431: if (2 == getcwd_status) {
1.59 schwarze 2432: exitcode = (int)MANDOCLEVEL_SYSERR;
1.151 schwarze 2433: say("", "getcwd: %s", startdir);
1.59 schwarze 2434: return(0);
2435: }
1.151 schwarze 2436: if (-1 == chdir(startdir)) {
1.59 schwarze 2437: exitcode = (int)MANDOCLEVEL_SYSERR;
1.123 schwarze 2438: say("", "&chdir %s", startdir);
1.59 schwarze 2439: return(0);
2440: }
2441: }
1.151 schwarze 2442:
2443: /*
2444: * Always resolve basedir to the canonicalized absolute
2445: * pathname and append a trailing slash, such that
2446: * we can reliably check whether files are inside.
2447: */
1.59 schwarze 2448: if (NULL == realpath(targetdir, basedir)) {
1.165 schwarze 2449: if (report_baddir || errno != ENOENT) {
2450: exitcode = (int)MANDOCLEVEL_BADARG;
2451: say("", "&%s: realpath", targetdir);
2452: }
1.50 kristaps 2453: return(0);
1.59 schwarze 2454: } else if (-1 == chdir(basedir)) {
1.165 schwarze 2455: if (report_baddir || errno != ENOENT) {
2456: exitcode = (int)MANDOCLEVEL_BADARG;
2457: say("", "&chdir");
2458: }
1.50 kristaps 2459: return(0);
1.150 schwarze 2460: }
1.151 schwarze 2461: chdir_status = 1;
1.150 schwarze 2462: cp = strchr(basedir, '\0');
2463: if ('/' != cp[-1]) {
2464: if (cp - basedir >= PATH_MAX - 1) {
2465: exitcode = (int)MANDOCLEVEL_SYSERR;
2466: say("", "Filename too long");
2467: return(0);
2468: }
2469: *cp++ = '/';
2470: *cp = '\0';
1.4 kristaps 2471: }
1.50 kristaps 2472: return(1);
1.56 schwarze 2473: }
2474:
2475: static void
1.59 schwarze 2476: say(const char *file, const char *format, ...)
1.56 schwarze 2477: {
2478: va_list ap;
1.123 schwarze 2479: int use_errno;
1.56 schwarze 2480:
1.59 schwarze 2481: if ('\0' != *basedir)
2482: fprintf(stderr, "%s", basedir);
2483: if ('\0' != *basedir && '\0' != *file)
1.151 schwarze 2484: fputc('/', stderr);
1.56 schwarze 2485: if ('\0' != *file)
1.59 schwarze 2486: fprintf(stderr, "%s", file);
2487:
1.123 schwarze 2488: use_errno = 1;
2489: if (NULL != format) {
2490: switch (*format) {
1.141 schwarze 2491: case '&':
1.123 schwarze 2492: format++;
2493: break;
1.141 schwarze 2494: case '\0':
1.123 schwarze 2495: format = NULL;
2496: break;
2497: default:
2498: use_errno = 0;
2499: break;
2500: }
2501: }
2502: if (NULL != format) {
2503: if ('\0' != *basedir || '\0' != *file)
2504: fputs(": ", stderr);
2505: va_start(ap, format);
2506: vfprintf(stderr, format, ap);
2507: va_end(ap);
2508: }
2509: if (use_errno) {
2510: if ('\0' != *basedir || '\0' != *file || NULL != format)
2511: fputs(": ", stderr);
1.59 schwarze 2512: perror(NULL);
1.123 schwarze 2513: } else
2514: fputc('\n', stderr);
1.1 kristaps 2515: }
CVSweb