1 /* $Id: mandocdb.c,v 1.269 2021/08/19 16:55:31 schwarze Exp $ */
3 * Copyright (c) 2011-2020 Ingo Schwarze <schwarze@openbsd.org>
4 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
5 * Copyright (c) 2016 Ed Maste <emaste@freebsd.org>
7 * Permission to use, copy, modify, and distribute this software for any
8 * purpose with or without fee is hereby granted, provided that the above
9 * copyright notice and this permission notice appear in all copies.
11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
19 * Implementation of the makewhatis(8) program.
23 #include <sys/types.h>
37 #include "compat_fts.h"
51 #include "mandoc_aux.h"
52 #include "mandoc_ohash.h"
57 #include "mandoc_parse.h"
59 #include "mansearch.h"
60 #include "dba_array.h"
63 extern const char *const mansearch_keynames[];
66 OP_DEFAULT = 0, /* new dbs from dir list or default config */
67 OP_CONFFILE, /* new databases from custom config file */
68 OP_UPDATE, /* delete/add entries in existing database */
69 OP_DELETE, /* delete entries from existing database */
70 OP_TEST /* change no databases, report potential problems */
74 const struct mpage *mpage; /* if set, the owning parse */
75 uint64_t mask; /* bitmask in sequence */
76 char key[]; /* rendered text */
85 struct inodev inodev; /* used for hashing routine */
86 struct dba_array *dba;
87 char *sec; /* section from file content */
88 char *arch; /* architecture from file content */
89 char *title; /* title from file content */
90 char *desc; /* description from file content */
91 struct mpage *next; /* singly linked list */
92 struct mlink *mlinks; /* singly linked list */
94 enum form form; /* format from file content */
98 char file[PATH_MAX]; /* filename rel. to manpath */
99 char *dsec; /* section from directory */
100 char *arch; /* architecture from directory */
101 char *name; /* name from file name (not empty) */
102 char *fsec; /* section from file name suffix */
103 struct mlink *next; /* singly linked list */
104 struct mpage *mpage; /* parent */
105 int gzip; /* filename has a .gz suffix */
106 enum form dform; /* format from directory */
107 enum form fform; /* format from file name suffix */
110 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
111 const struct roff_node *);
113 struct mdoc_handler {
114 mdoc_fp fp; /* optional handler */
115 uint64_t mask; /* set unless handler returns 0 */
116 int taboo; /* node flags that must not be set */
120 int mandocdb(int, char *[]);
122 static void dbadd(struct dba *, struct mpage *);
123 static void dbadd_mlink(const struct mlink *);
124 static void dbprune(struct dba *);
125 static void dbwrite(struct dba *);
126 static void filescan(const char *);
127 #if HAVE_FTS_COMPARE_CONST
128 static int fts_compare(const FTSENT *const *, const FTSENT *const *);
130 static int fts_compare(const FTSENT **, const FTSENT **);
132 static void mlink_add(struct mlink *, const struct stat *);
133 static void mlink_check(struct mpage *, struct mlink *);
134 static void mlink_free(struct mlink *);
135 static void mlinks_undupe(struct mpage *);
136 static void mpages_free(void);
137 static void mpages_merge(struct dba *, struct mparse *);
138 static void parse_cat(struct mpage *, int);
139 static void parse_man(struct mpage *, const struct roff_meta *,
140 const struct roff_node *);
141 static void parse_mdoc(struct mpage *, const struct roff_meta *,
142 const struct roff_node *);
143 static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
144 const struct roff_node *);
145 static int parse_mdoc_Fa(struct mpage *, const struct roff_meta *,
146 const struct roff_node *);
147 static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
148 const struct roff_node *);
149 static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
150 static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
151 const struct roff_node *);
152 static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
153 const struct roff_node *);
154 static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
155 const struct roff_node *);
156 static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
157 const struct roff_node *);
158 static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
159 const struct roff_node *);
160 static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
161 const struct roff_node *);
162 static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
163 const struct roff_node *);
164 static void putkey(const struct mpage *, char *, uint64_t);
165 static void putkeys(const struct mpage *, char *, size_t, uint64_t);
166 static void putmdockey(const struct mpage *,
167 const struct roff_node *, uint64_t, int);
168 #ifdef READ_ALLOWED_PATH
169 static int read_allowed(const char *);
171 static int render_string(char **, size_t *);
172 static void say(const char *, const char *, ...)
173 __attribute__((__format__ (__printf__, 2, 3)));
174 static int set_basedir(const char *, int);
175 static int treescan(void);
176 static size_t utf8(unsigned int, char [7]);
178 static int nodb; /* no database changes */
179 static int mparse_options; /* abort the parse early */
180 static int use_all; /* use all found files */
181 static int debug; /* print what we're doing */
182 static int warnings; /* warn about crap */
183 static int write_utf8; /* write UTF-8 output; else ASCII */
184 static int exitcode; /* to be returned by main */
185 static enum op op; /* operational mode */
186 static char basedir[PATH_MAX]; /* current base directory */
187 static size_t basedir_len; /* strlen(basedir) */
188 static struct mpage *mpage_head; /* list of distinct manual pages */
189 static struct ohash mpages; /* table of distinct manual pages */
190 static struct ohash mlinks; /* table of directory entries */
191 static struct ohash names; /* table of all names */
192 static struct ohash strings; /* table of all strings */
193 static uint64_t name_mask;
195 static const struct mdoc_handler mdoc_handlers[MDOC_MAX - MDOC_Dd] = {
196 { NULL, 0, NODE_NOPRT }, /* Dd */
197 { NULL, 0, NODE_NOPRT }, /* Dt */
198 { NULL, 0, NODE_NOPRT }, /* Os */
199 { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
200 { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
201 { NULL, 0, 0 }, /* Pp */
202 { NULL, 0, 0 }, /* D1 */
203 { NULL, 0, 0 }, /* Dl */
204 { NULL, 0, 0 }, /* Bd */
205 { NULL, 0, 0 }, /* Ed */
206 { NULL, 0, 0 }, /* Bl */
207 { NULL, 0, 0 }, /* El */
208 { NULL, 0, 0 }, /* It */
209 { NULL, 0, 0 }, /* Ad */
210 { NULL, TYPE_An, 0 }, /* An */
211 { NULL, 0, 0 }, /* Ap */
212 { NULL, TYPE_Ar, 0 }, /* Ar */
213 { NULL, TYPE_Cd, 0 }, /* Cd */
214 { NULL, TYPE_Cm, 0 }, /* Cm */
215 { NULL, TYPE_Dv, 0 }, /* Dv */
216 { NULL, TYPE_Er, 0 }, /* Er */
217 { NULL, TYPE_Ev, 0 }, /* Ev */
218 { NULL, 0, 0 }, /* Ex */
219 { parse_mdoc_Fa, 0, 0 }, /* Fa */
220 { parse_mdoc_Fd, 0, 0 }, /* Fd */
221 { NULL, TYPE_Fl, 0 }, /* Fl */
222 { parse_mdoc_Fn, 0, 0 }, /* Fn */
223 { NULL, TYPE_Ft | TYPE_Vt, 0 }, /* Ft */
224 { NULL, TYPE_Ic, 0 }, /* Ic */
225 { NULL, TYPE_In, 0 }, /* In */
226 { NULL, TYPE_Li, 0 }, /* Li */
227 { parse_mdoc_Nd, 0, 0 }, /* Nd */
228 { parse_mdoc_Nm, 0, 0 }, /* Nm */
229 { NULL, 0, 0 }, /* Op */
230 { NULL, 0, 0 }, /* Ot */
231 { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
232 { NULL, 0, 0 }, /* Rv */
233 { NULL, TYPE_St, 0 }, /* St */
234 { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
235 { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
236 { parse_mdoc_Xr, 0, 0 }, /* Xr */
237 { NULL, 0, 0 }, /* %A */
238 { NULL, 0, 0 }, /* %B */
239 { NULL, 0, 0 }, /* %D */
240 { NULL, 0, 0 }, /* %I */
241 { NULL, 0, 0 }, /* %J */
242 { NULL, 0, 0 }, /* %N */
243 { NULL, 0, 0 }, /* %O */
244 { NULL, 0, 0 }, /* %P */
245 { NULL, 0, 0 }, /* %R */
246 { NULL, 0, 0 }, /* %T */
247 { NULL, 0, 0 }, /* %V */
248 { NULL, 0, 0 }, /* Ac */
249 { NULL, 0, 0 }, /* Ao */
250 { NULL, 0, 0 }, /* Aq */
251 { NULL, TYPE_At, 0 }, /* At */
252 { NULL, 0, 0 }, /* Bc */
253 { NULL, 0, 0 }, /* Bf */
254 { NULL, 0, 0 }, /* Bo */
255 { NULL, 0, 0 }, /* Bq */
256 { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
257 { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
258 { NULL, 0, 0 }, /* Db */
259 { NULL, 0, 0 }, /* Dc */
260 { NULL, 0, 0 }, /* Do */
261 { NULL, 0, 0 }, /* Dq */
262 { NULL, 0, 0 }, /* Ec */
263 { NULL, 0, 0 }, /* Ef */
264 { NULL, TYPE_Em, 0 }, /* Em */
265 { NULL, 0, 0 }, /* Eo */
266 { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
267 { NULL, TYPE_Ms, 0 }, /* Ms */
268 { NULL, 0, 0 }, /* No */
269 { NULL, 0, 0 }, /* Ns */
270 { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
271 { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
272 { NULL, 0, 0 }, /* Pc */
273 { NULL, 0, 0 }, /* Pf */
274 { NULL, 0, 0 }, /* Po */
275 { NULL, 0, 0 }, /* Pq */
276 { NULL, 0, 0 }, /* Qc */
277 { NULL, 0, 0 }, /* Ql */
278 { NULL, 0, 0 }, /* Qo */
279 { NULL, 0, 0 }, /* Qq */
280 { NULL, 0, 0 }, /* Re */
281 { NULL, 0, 0 }, /* Rs */
282 { NULL, 0, 0 }, /* Sc */
283 { NULL, 0, 0 }, /* So */
284 { NULL, 0, 0 }, /* Sq */
285 { NULL, 0, 0 }, /* Sm */
286 { NULL, 0, 0 }, /* Sx */
287 { NULL, TYPE_Sy, 0 }, /* Sy */
288 { NULL, TYPE_Tn, 0 }, /* Tn */
289 { NULL, 0, NODE_NOSRC }, /* Ux */
290 { NULL, 0, 0 }, /* Xc */
291 { NULL, 0, 0 }, /* Xo */
292 { parse_mdoc_Fo, 0, 0 }, /* Fo */
293 { NULL, 0, 0 }, /* Fc */
294 { NULL, 0, 0 }, /* Oo */
295 { NULL, 0, 0 }, /* Oc */
296 { NULL, 0, 0 }, /* Bk */
297 { NULL, 0, 0 }, /* Ek */
298 { NULL, 0, 0 }, /* Bt */
299 { NULL, 0, 0 }, /* Hf */
300 { NULL, 0, 0 }, /* Fr */
301 { NULL, 0, 0 }, /* Ud */
302 { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */
303 { NULL, 0, 0 }, /* Lp */
304 { NULL, TYPE_Lk, 0 }, /* Lk */
305 { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
306 { NULL, 0, 0 }, /* Brq */
307 { NULL, 0, 0 }, /* Bro */
308 { NULL, 0, 0 }, /* Brc */
309 { NULL, 0, 0 }, /* %C */
310 { NULL, 0, 0 }, /* Es */
311 { NULL, 0, 0 }, /* En */
312 { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
313 { NULL, 0, 0 }, /* %Q */
314 { NULL, 0, 0 }, /* %U */
315 { NULL, 0, 0 }, /* Ta */
320 mandocdb(int argc, char *argv[])
325 const char *path_arg, *progname;
330 if (pledge("stdio rpath wpath cpath", NULL) == -1) {
332 return (int)MANDOCLEVEL_SYSERR;
336 #if HAVE_SANDBOX_INIT
337 if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) {
338 warnx("sandbox_init");
339 return (int)MANDOCLEVEL_SYSERR;
343 memset(&conf, 0, sizeof(conf));
346 * We accept a few different invocations.
347 * The CHECKOP macro makes sure that invocation styles don't
348 * clobber each other.
350 #define CHECKOP(_op, _ch) do \
351 if ((_op) != OP_DEFAULT) { \
352 warnx("-%c: Conflicting option", (_ch)); \
354 } while (/*CONSTCOND*/0)
356 mparse_options = MPARSE_VALIDATE;
360 while ((ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")) != -1)
385 mparse_options |= MPARSE_QUICK;
388 if (strcmp(optarg, "utf8") != 0) {
389 warnx("-T%s: Unsupported output format",
397 dup2(STDOUT_FILENO, STDERR_FILENO);
407 /* Compatibility with espie@'s makewhatis. */
418 if (pledge("stdio rpath", NULL) == -1) {
420 return (int)MANDOCLEVEL_SYSERR;
425 if (op == OP_CONFFILE && argc > 0) {
426 warnx("-C: Too many arguments");
430 exitcode = (int)MANDOCLEVEL_OK;
432 mp = mparse_alloc(mparse_options, MANDOC_OS_OTHER, NULL);
433 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
434 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
436 if (op == OP_UPDATE || op == OP_DELETE || op == OP_TEST) {
439 * Most of these deal with a specific directory.
440 * Jump into that directory first.
442 if (op != OP_TEST && set_basedir(path_arg, 1) == 0)
445 dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
448 * The existing database is usable. Process
449 * all files specified on the command-line.
452 for (i = 0; i < argc; i++)
457 /* Database missing or corrupt. */
458 if (op != OP_UPDATE || errno != ENOENT)
459 say(MANDOC_DB, "%s: Automatically recreating"
460 " from scratch", strerror(errno));
461 exitcode = (int)MANDOCLEVEL_OK;
468 mpages_merge(dba, mp);
474 * If we have arguments, use them as our manpaths.
475 * If we don't, use man.conf(5).
478 conf.manpath.paths = mandoc_reallocarray(NULL,
479 argc, sizeof(char *));
480 conf.manpath.sz = (size_t)argc;
481 for (i = 0; i < argc; i++)
482 conf.manpath.paths[i] = mandoc_strdup(argv[i]);
484 manconf_parse(&conf, path_arg, NULL, NULL);
486 if (conf.manpath.sz == 0) {
487 exitcode = (int)MANDOCLEVEL_BADARG;
488 say("", "Empty manpath");
492 * First scan the tree rooted at a base directory, then
493 * build a new database and finally move it into place.
494 * Ignore zero-length directories and strip trailing
497 for (j = 0; j < conf.manpath.sz; j++) {
498 sz = strlen(conf.manpath.paths[j]);
499 if (sz && conf.manpath.paths[j][sz - 1] == '/')
500 conf.manpath.paths[j][--sz] = '\0';
505 mandoc_ohash_init(&mpages, 6,
506 offsetof(struct mpage, inodev));
507 mandoc_ohash_init(&mlinks, 6,
508 offsetof(struct mlink, file));
511 if (set_basedir(conf.manpath.paths[j], argc > 0) == 0)
516 mpages_merge(dba, mp);
521 if (j + 1 < conf.manpath.sz) {
523 ohash_delete(&mpages);
524 ohash_delete(&mlinks);
533 ohash_delete(&mpages);
534 ohash_delete(&mlinks);
537 progname = getprogname();
538 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
539 " %s [-aDnpQ] [-Tutf8] dir ...\n"
540 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
541 " %s [-Dnp] -u dir [file ...]\n"
542 " %s [-Q] -t file ...\n",
543 progname, progname, progname, progname, progname);
545 return (int)MANDOCLEVEL_BADARG;
549 * To get a singly linked list in alpha order while inserting entries
550 * at the beginning, process directory entries in reverse alpha order.
553 #if HAVE_FTS_COMPARE_CONST
554 fts_compare(const FTSENT *const *a, const FTSENT *const *b)
556 fts_compare(const FTSENT **a, const FTSENT **b)
559 return -strcmp((*a)->fts_name, (*b)->fts_name);
563 * Scan a directory tree rooted at "basedir" for manpages.
564 * We use fts(), scanning directory parts along the way for clues to our
565 * section and architecture.
567 * If use_all has been specified, grok all files.
568 * If not, sanitise paths to the following:
570 * [./]man*[/<arch>]/<name>.<section>
572 * [./]cat<section>[/<arch>]/<name>.0
574 * TODO: accommodate for multi-language directories.
585 char *dsec, *arch, *fsec, *cp;
592 f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR,
595 exitcode = (int)MANDOCLEVEL_SYSERR;
596 say("", "&fts_open");
603 while ((ff = fts_read(f)) != NULL) {
604 path = ff->fts_path + 2;
605 switch (ff->fts_info) {
608 * Symbolic links require various sanity checks,
609 * then get handled just like regular files.
612 if (realpath(path, buf) == NULL) {
614 say(path, "&realpath");
617 if (strncmp(buf, basedir, basedir_len) != 0
618 #ifdef READ_ALLOWED_PATH
619 && !read_allowed(buf)
622 if (warnings) say("",
623 "%s: outside base directory", buf);
626 /* Use logical inode to avoid mpages dupe. */
627 if (stat(path, ff->fts_statp) == -1) {
632 if ((ff->fts_statp->st_mode & S_IFMT) != S_IFREG)
637 * If we're a regular file, add an mlink by using the
638 * stored directory data and handling the filename.
641 if ( ! strcmp(path, MANDOC_DB))
643 if ( ! use_all && ff->fts_level < 2) {
645 say(path, "Extraneous file");
650 while (fsec == NULL) {
651 fsec = strrchr(ff->fts_name, '.');
652 if (fsec == NULL || strcmp(fsec+1, "gz"))
662 "No filename suffix");
665 } else if ( ! strcmp(++fsec, "html")) {
667 say(path, "Skip html");
669 } else if ( ! strcmp(fsec, "ps")) {
671 say(path, "Skip ps");
673 } else if ( ! strcmp(fsec, "pdf")) {
675 say(path, "Skip pdf");
677 } else if ( ! use_all &&
678 ((dform == FORM_SRC &&
679 strncmp(fsec, dsec, strlen(dsec))) ||
680 (dform == FORM_CAT && strcmp(fsec, "0")))) {
682 say(path, "Wrong filename suffix");
687 mlink = mandoc_calloc(1, sizeof(struct mlink));
688 if (strlcpy(mlink->file, path,
689 sizeof(mlink->file)) >=
690 sizeof(mlink->file)) {
691 say(path, "Filename too long");
695 mlink->dform = dform;
698 mlink->name = ff->fts_name;
701 mlink_add(mlink, ff->fts_statp);
710 say(path, "Not a regular file");
714 switch (ff->fts_level) {
716 /* Ignore the root directory. */
720 * This might contain manX/ or catX/.
721 * Try to infer this from the name.
722 * If we're not in use_all, enforce it.
725 if (ff->fts_info == FTS_DP) {
731 if ( ! strncmp(cp, "man", 3)) {
734 } else if ( ! strncmp(cp, "cat", 3)) {
742 if (dsec != NULL || use_all)
746 say(path, "Unknown directory part");
747 fts_set(f, ff, FTS_SKIP);
751 * Possibly our architecture.
752 * If we're descending, keep tabs on it.
754 if (ff->fts_info != FTS_DP && dsec != NULL)
760 if (ff->fts_info == FTS_DP || use_all)
763 say(path, "Extraneous directory part");
764 fts_set(f, ff, FTS_SKIP);
774 * Add a file to the mlinks table.
775 * Do not verify that it's a "valid" looking manpage (we'll do that
778 * Try to infer the manual section, architecture, and page name from the
779 * path, assuming it looks like
781 * [./]man*[/<arch>]/<name>.<section>
783 * [./]cat<section>[/<arch>]/<name>.0
785 * See treescan() for the fts(3) version of this.
788 filescan(const char *infile)
792 char *linkfile, *p, *realdir, *start, *usefile;
797 if (strncmp(infile, "./", 2) == 0)
801 * We have to do lstat(2) before realpath(3) loses
802 * the information whether this is a symbolic link.
803 * We need to know that because for symbolic links,
804 * we want to use the orginal file name, while for
805 * regular files, we want to use the real path.
807 if (lstat(infile, &st) == -1) {
808 exitcode = (int)MANDOCLEVEL_BADARG;
809 say(infile, "&lstat");
811 } else if (S_ISREG(st.st_mode) == 0 && S_ISLNK(st.st_mode) == 0) {
812 exitcode = (int)MANDOCLEVEL_BADARG;
813 say(infile, "Not a regular file");
818 * We have to resolve the file name to the real path
819 * in any case for the base directory check.
821 if ((usefile = realpath(infile, NULL)) == NULL) {
822 exitcode = (int)MANDOCLEVEL_BADARG;
823 say(infile, "&realpath");
829 else if (strncmp(usefile, basedir, basedir_len) == 0)
830 start = usefile + basedir_len;
831 #ifdef READ_ALLOWED_PATH
832 else if (read_allowed(usefile))
836 exitcode = (int)MANDOCLEVEL_BADARG;
837 say("", "%s: outside base directory", infile);
843 * Now we are sure the file is inside our tree.
844 * If it is a symbolic link, ignore the real path
845 * and use the original name.
848 if (S_ISLNK(st.st_mode) == 0)
852 * Some implementations of realpath(3) may succeed
853 * even if the target of the link does not exist,
854 * so check again for extra safety.
856 if (stat(usefile, &st) == -1) {
857 exitcode = (int)MANDOCLEVEL_BADARG;
858 say(infile, "&stat");
862 linkfile = mandoc_strdup(infile);
865 start = usefile = linkfile;
868 if (strncmp(infile, basedir, basedir_len) == 0) {
871 start = usefile + basedir_len;
876 * This symbolic link points into the basedir
877 * from the outside. Let's see whether any of
878 * the parent directories resolve to the basedir.
880 p = strchr(linkfile, '\0');
885 if ((realdir = realpath(linkfile, NULL)) == NULL) {
886 exitcode = (int)MANDOCLEVEL_BADARG;
887 say(infile, "&realpath");
892 realdir_len = strlen(realdir) + 1;
895 } while (realdir_len > basedir_len);
898 * If one of the directories resolves to the basedir,
899 * use the rest of the original name.
900 * Otherwise, the best we can do
901 * is to use the filename pointed to.
903 if (realdir_len == basedir_len) {
909 start = usefile + basedir_len;
911 } while (/* CONSTCOND */ 0);
913 mlink = mandoc_calloc(1, sizeof(struct mlink));
914 mlink->dform = FORM_NONE;
915 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
916 sizeof(mlink->file)) {
917 say(start, "Filename too long");
924 * In test mode or when the original name is absolute
925 * but outside our tree, guess the base directory.
928 if (op == OP_TEST || (start == usefile && *start == '/')) {
929 if (strncmp(usefile, "man/", 4) == 0)
931 else if ((start = strstr(usefile, "/man/")) != NULL)
938 * First try to guess our directory structure.
939 * If we find a separator, try to look for man* or cat*.
940 * If we find one of these and what's underneath is a directory,
941 * assume it's an architecture.
943 if ((p = strchr(start, '/')) != NULL) {
945 if (strncmp(start, "man", 3) == 0) {
946 mlink->dform = FORM_SRC;
947 mlink->dsec = start + 3;
948 } else if (strncmp(start, "cat", 3) == 0) {
949 mlink->dform = FORM_CAT;
950 mlink->dsec = start + 3;
954 if (mlink->dsec != NULL && (p = strchr(start, '/')) != NULL) {
962 * Now check the file suffix.
963 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
965 p = strrchr(start, '\0');
966 while (p-- > start && *p != '/' && *p != '.')
975 * Now try to parse the name.
976 * Use the filename portion of the path.
979 if ((p = strrchr(start, '/')) != NULL) {
983 mlink_add(mlink, &st);
988 mlink_add(struct mlink *mlink, const struct stat *st)
990 struct inodev inodev;
994 assert(NULL != mlink->file);
996 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
997 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
998 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
999 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1001 if ('0' == *mlink->fsec) {
1003 mlink->fsec = mandoc_strdup(mlink->dsec);
1004 mlink->fform = FORM_CAT;
1005 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
1006 mlink->fform = FORM_SRC;
1008 mlink->fform = FORM_NONE;
1010 slot = ohash_qlookup(&mlinks, mlink->file);
1011 assert(NULL == ohash_find(&mlinks, slot));
1012 ohash_insert(&mlinks, slot, mlink);
1014 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1015 inodev.st_ino = st->st_ino;
1016 inodev.st_dev = st->st_dev;
1017 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
1018 sizeof(struct inodev), inodev.st_ino);
1019 mpage = ohash_find(&mpages, slot);
1020 if (NULL == mpage) {
1021 mpage = mandoc_calloc(1, sizeof(struct mpage));
1022 mpage->inodev.st_ino = inodev.st_ino;
1023 mpage->inodev.st_dev = inodev.st_dev;
1024 mpage->form = FORM_NONE;
1025 mpage->next = mpage_head;
1027 ohash_insert(&mpages, slot, mpage);
1029 mlink->next = mpage->mlinks;
1030 mpage->mlinks = mlink;
1031 mlink->mpage = mpage;
1035 mlink_free(struct mlink *mlink)
1048 struct mpage *mpage;
1049 struct mlink *mlink;
1051 while ((mpage = mpage_head) != NULL) {
1052 while ((mlink = mpage->mlinks) != NULL) {
1053 mpage->mlinks = mlink->next;
1056 mpage_head = mpage->next;
1066 * For each mlink to the mpage, check whether the path looks like
1067 * it is formatted, and if it does, check whether a source manual
1068 * exists by the same name, ignoring the suffix.
1069 * If both conditions hold, drop the mlink.
1072 mlinks_undupe(struct mpage *mpage)
1075 struct mlink **prev;
1076 struct mlink *mlink;
1079 mpage->form = FORM_CAT;
1080 prev = &mpage->mlinks;
1081 while (NULL != (mlink = *prev)) {
1082 if (FORM_CAT != mlink->dform) {
1083 mpage->form = FORM_NONE;
1086 (void)strlcpy(buf, mlink->file, sizeof(buf));
1087 bufp = strstr(buf, "cat");
1088 assert(NULL != bufp);
1089 memcpy(bufp, "man", 3);
1090 if (NULL != (bufp = strrchr(buf, '.')))
1092 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1093 if (NULL == ohash_find(&mlinks,
1094 ohash_qlookup(&mlinks, buf)))
1097 say(mlink->file, "Man source exists: %s", buf);
1100 *prev = mlink->next;
1104 prev = &(*prev)->next;
1109 mlink_check(struct mpage *mpage, struct mlink *mlink)
1115 * Check whether the manual section given in a file
1116 * agrees with the directory where the file is located.
1117 * Some manuals have suffixes like (3p) on their
1118 * section number either inside the file or in the
1119 * directory name, some are linked into more than one
1120 * section, like encrypt(1) = makekey(8).
1123 if (FORM_SRC == mpage->form &&
1124 strcasecmp(mpage->sec, mlink->dsec))
1125 say(mlink->file, "Section \"%s\" manual in %s directory",
1126 mpage->sec, mlink->dsec);
1129 * Manual page directories exist for each kernel
1130 * architecture as returned by machine(1).
1131 * However, many manuals only depend on the
1132 * application architecture as returned by arch(1).
1133 * For example, some (2/ARM) manuals are shared
1134 * across the "armish" and "zaurus" kernel
1136 * A few manuals are even shared across completely
1137 * different architectures, for example fdformat(1)
1138 * on amd64, i386, and sparc64.
1141 if (strcasecmp(mpage->arch, mlink->arch))
1142 say(mlink->file, "Architecture \"%s\" manual in "
1143 "\"%s\" directory", mpage->arch, mlink->arch);
1147 * parse_cat() doesn't set NAME_TITLE yet.
1150 if (FORM_CAT == mpage->form)
1154 * Check whether this mlink
1155 * appears as a name in the NAME section.
1158 slot = ohash_qlookup(&names, mlink->name);
1159 str = ohash_find(&names, slot);
1160 assert(NULL != str);
1161 if ( ! (NAME_TITLE & str->mask))
1162 say(mlink->file, "Name missing in NAME section");
1166 * Run through the files in the global vector "mpages"
1167 * and add them to the database specified in "basedir".
1169 * This handles the parsing scheme itself, using the cues of directory
1170 * and filename to determine whether the file is parsable or not.
1173 mpages_merge(struct dba *dba, struct mparse *mp)
1175 struct mpage *mpage, *mpage_dest;
1176 struct mlink *mlink, *mlink_dest;
1177 struct roff_meta *meta;
1181 for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
1182 mlinks_undupe(mpage);
1183 if ((mlink = mpage->mlinks) == NULL)
1186 name_mask = NAME_MASK;
1187 mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1188 mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1192 if ((fd = mparse_open(mp, mlink->file)) == -1) {
1193 say(mlink->file, "&open");
1198 * Interpret the file as mdoc(7) or man(7) source
1199 * code, unless it is known to be formatted.
1201 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1202 mparse_readfd(mp, fd, mlink->file);
1205 meta = mparse_result(mp);
1208 if (meta != NULL && meta->sodest != NULL) {
1209 mlink_dest = ohash_find(&mlinks,
1210 ohash_qlookup(&mlinks, meta->sodest));
1211 if (mlink_dest == NULL) {
1212 mandoc_asprintf(&cp, "%s.gz", meta->sodest);
1213 mlink_dest = ohash_find(&mlinks,
1214 ohash_qlookup(&mlinks, cp));
1217 if (mlink_dest != NULL) {
1219 /* The .so target exists. */
1221 mpage_dest = mlink_dest->mpage;
1223 mlink->mpage = mpage_dest;
1226 * If the target was already
1227 * processed, add the links
1228 * to the database now.
1229 * Otherwise, this will
1230 * happen when we come
1234 if (mpage_dest->dba != NULL)
1237 if (mlink->next == NULL)
1239 mlink = mlink->next;
1242 /* Move all links to the target. */
1244 mlink->next = mlink_dest->next;
1245 mlink_dest->next = mpage->mlinks;
1246 mpage->mlinks = NULL;
1249 meta->macroset = MACROSET_NONE;
1251 if (meta != NULL && meta->macroset == MACROSET_MDOC) {
1252 mpage->form = FORM_SRC;
1253 mpage->sec = meta->msec;
1254 mpage->sec = mandoc_strdup(
1255 mpage->sec == NULL ? "" : mpage->sec);
1256 mpage->arch = meta->arch;
1257 mpage->arch = mandoc_strdup(
1258 mpage->arch == NULL ? "" : mpage->arch);
1259 mpage->title = mandoc_strdup(meta->title);
1260 } else if (meta != NULL && meta->macroset == MACROSET_MAN) {
1261 if (*meta->msec != '\0' || *meta->title != '\0') {
1262 mpage->form = FORM_SRC;
1263 mpage->sec = mandoc_strdup(meta->msec);
1264 mpage->arch = mandoc_strdup(mlink->arch);
1265 mpage->title = mandoc_strdup(meta->title);
1270 assert(mpage->desc == NULL);
1271 if (meta == NULL || meta->sodest != NULL) {
1272 mpage->sec = mandoc_strdup(mlink->dsec);
1273 mpage->arch = mandoc_strdup(mlink->arch);
1274 mpage->title = mandoc_strdup(mlink->name);
1276 mpage->form = FORM_CAT;
1277 parse_cat(mpage, fd);
1279 mpage->form = FORM_SRC;
1280 } else if (meta->macroset == MACROSET_MDOC)
1281 parse_mdoc(mpage, meta, meta->first);
1283 parse_man(mpage, meta, meta->first);
1284 if (mpage->desc == NULL) {
1285 mpage->desc = mandoc_strdup(mlink->name);
1287 say(mlink->file, "No one-line description, "
1288 "using filename \"%s\"", mlink->name);
1291 for (mlink = mpage->mlinks;
1293 mlink = mlink->next) {
1294 putkey(mpage, mlink->name, NAME_FILE);
1295 if (warnings && !use_all)
1296 mlink_check(mpage, mlink);
1302 ohash_delete(&strings);
1303 ohash_delete(&names);
1308 parse_cat(struct mpage *mpage, int fd)
1311 struct mlink *mlink;
1312 char *line, *p, *title, *sec;
1313 size_t linesz, plen, titlesz;
1317 mlink = mpage->mlinks;
1318 stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1319 if (stream == NULL) {
1323 say(mlink->file, "&fopen");
1330 /* Parse the section number from the header line. */
1332 while (getline(&line, &linesz, stream) != -1) {
1335 if ((sec = strchr(line, '(')) == NULL)
1337 if ((p = strchr(++sec, ')')) == NULL)
1340 mpage->sec = mandoc_strndup(sec, p - sec);
1341 if (warnings && *mlink->dsec != '\0' &&
1342 strcasecmp(mpage->sec, mlink->dsec))
1344 "Section \"%s\" manual in %s directory",
1345 mpage->sec, mlink->dsec);
1349 /* Skip to first blank line. */
1351 while (line == NULL || *line != '\n')
1352 if (getline(&line, &linesz, stream) == -1)
1356 * Assume the first line that is not indented
1357 * is the first section header. Skip to it.
1360 while (getline(&line, &linesz, stream) != -1)
1361 if (*line != '\n' && *line != ' ')
1365 * Read up until the next section into a buffer.
1366 * Strip the leading and trailing newline from each read line,
1367 * appending a trailing space.
1368 * Ignore empty (whitespace-only) lines.
1374 while ((len = getline(&line, &linesz, stream)) != -1) {
1378 while (isspace((unsigned char)line[offs]))
1380 if (line[offs] == '\0')
1382 title = mandoc_realloc(title, titlesz + len - offs);
1383 memcpy(title + titlesz, line + offs, len - offs);
1384 titlesz += len - offs;
1385 title[titlesz - 1] = ' ';
1390 * If no page content can be found, or the input line
1391 * is already the next section header, or there is no
1392 * trailing newline, reuse the page title as the page
1396 if (NULL == title || '\0' == *title) {
1398 say(mlink->file, "Cannot find NAME section");
1404 title[titlesz - 1] = '\0';
1407 * Skip to the first dash.
1408 * Use the remaining line as the description (no more than 70
1412 if (NULL != (p = strstr(title, "- "))) {
1413 for (p += 2; ' ' == *p || '\b' == *p; p++)
1414 /* Skip to next word. */ ;
1417 say(mlink->file, "No dash in title line, "
1418 "reusing \"%s\" as one-line description", title);
1424 /* Strip backspace-encoding from line. */
1426 while (NULL != (line = memchr(p, '\b', plen))) {
1429 memmove(line, line + 1, plen--);
1432 memmove(line - 1, line + 1, plen - len);
1437 * Cut off excessive one-line descriptions.
1438 * Bad pages are not worth better heuristics.
1441 mpage->desc = mandoc_strndup(p, 150);
1447 * Put a type/word pair into the word database for this particular file.
1450 putkey(const struct mpage *mpage, char *value, uint64_t type)
1452 putkeys(mpage, value, strlen(value), type);
1456 * Grok all nodes at or below a certain mdoc node into putkey().
1459 putmdockey(const struct mpage *mpage,
1460 const struct roff_node *n, uint64_t m, int taboo)
1463 for ( ; NULL != n; n = n->next) {
1464 if (n->flags & taboo)
1466 if (NULL != n->child)
1467 putmdockey(mpage, n->child, m, taboo);
1468 if (n->type == ROFFT_TEXT)
1469 putkey(mpage, n->string, m);
1474 parse_man(struct mpage *mpage, const struct roff_meta *meta,
1475 const struct roff_node *n)
1477 const struct roff_node *head, *body;
1478 char *start, *title;
1486 * We're only searching for one thing: the first text child in
1487 * the BODY of a NAME section. Since we don't keep track of
1488 * sections in -man, run some hoops to find out whether we're in
1489 * the correct section or not.
1492 if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1494 if ((head = body->parent->head) != NULL &&
1495 (head = head->child) != NULL &&
1496 head->next == NULL &&
1497 head->type == ROFFT_TEXT &&
1498 strcmp(head->string, "NAME") == 0 &&
1499 body->child != NULL) {
1502 * Suck the entire NAME section into memory.
1503 * Yes, we might run away.
1504 * But too many manuals have big, spread-out
1505 * NAME sections over many lines.
1509 deroff(&title, body);
1514 * Go through a special heuristic dance here.
1515 * Conventionally, one or more manual names are
1516 * comma-specified prior to a whitespace, then a
1517 * dash, then a description. Try to puzzle out
1518 * the name parts here.
1523 sz = strcspn(start, " ,");
1524 if ('\0' == start[sz])
1531 * Assume a stray trailing comma in the
1532 * name list if a name begins with a dash.
1535 if ('-' == start[0] ||
1536 ('\\' == start[0] && '-' == start[1]))
1539 putkey(mpage, start, NAME_TITLE);
1540 if ( ! (mpage->name_head_done ||
1541 strcasecmp(start, meta->title))) {
1542 putkey(mpage, start, NAME_HEAD);
1543 mpage->name_head_done = 1;
1551 assert(',' == byte);
1553 while (' ' == *start)
1557 if (start == title) {
1558 putkey(mpage, start, NAME_TITLE);
1559 if ( ! (mpage->name_head_done ||
1560 strcasecmp(start, meta->title))) {
1561 putkey(mpage, start, NAME_HEAD);
1562 mpage->name_head_done = 1;
1568 while (isspace((unsigned char)*start))
1571 if (0 == strncmp(start, "-", 1))
1573 else if (0 == strncmp(start, "\\-\\-", 4))
1575 else if (0 == strncmp(start, "\\-", 2))
1577 else if (0 == strncmp(start, "\\(en", 4))
1579 else if (0 == strncmp(start, "\\(em", 4))
1582 while (' ' == *start)
1586 * Cut off excessive one-line descriptions.
1587 * Bad pages are not worth better heuristics.
1590 mpage->desc = mandoc_strndup(start, 150);
1596 for (n = n->child; n; n = n->next) {
1597 if (NULL != mpage->desc)
1599 parse_man(mpage, meta, n);
1604 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1605 const struct roff_node *n)
1607 const struct mdoc_handler *handler;
1609 for (n = n->child; n != NULL; n = n->next) {
1610 if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX)
1612 assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1613 handler = mdoc_handlers + (n->tok - MDOC_Dd);
1614 if (n->flags & handler->taboo)
1623 if (handler->fp != NULL &&
1624 (*handler->fp)(mpage, meta, n) == 0)
1627 putmdockey(mpage, n->child,
1628 handler->mask, handler->taboo);
1633 if (NULL != n->child)
1634 parse_mdoc(mpage, meta, n);
1639 parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta,
1640 const struct roff_node *n)
1645 if (n->sec == SEC_SYNOPSIS)
1648 putmdockey(mpage, n->child, mask, 0);
1653 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1654 const struct roff_node *n)
1659 if (SEC_SYNOPSIS != n->sec ||
1660 NULL == (n = n->child) ||
1661 n->type != ROFFT_TEXT)
1665 * Only consider those `Fd' macro fields that begin with an
1666 * "inclusion" token (versus, e.g., #define).
1669 if (strcmp("#include", n->string))
1672 if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1676 * Strip away the enclosing angle brackets and make sure we're
1681 if ('<' == *start || '"' == *start)
1684 if (0 == (sz = strlen(start)))
1687 end = &start[(int)sz - 1];
1688 if ('>' == *end || '"' == *end)
1692 putkeys(mpage, start, end - start + 1, TYPE_In);
1697 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1702 if (n->type != ROFFT_TEXT)
1705 /* Skip function pointer punctuation. */
1708 while (*cp == '(' || *cp == '*')
1710 sz = strcspn(cp, "()");
1712 putkeys(mpage, cp, sz, TYPE_Fn);
1713 if (n->sec == SEC_SYNOPSIS)
1714 putkeys(mpage, cp, sz, NAME_SYN);
1718 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1719 const struct roff_node *n)
1723 if (n->child == NULL)
1726 parse_mdoc_fname(mpage, n->child);
1729 if (n != NULL && n->type == ROFFT_TEXT) {
1731 if (n->sec == SEC_SYNOPSIS)
1733 putmdockey(mpage, n, mask, 0);
1740 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1741 const struct roff_node *n)
1744 if (n->type != ROFFT_HEAD)
1747 if (n->child != NULL)
1748 parse_mdoc_fname(mpage, n->child);
1754 parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1755 const struct roff_node *n)
1759 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1762 if (n->child != NULL &&
1763 n->child->next == NULL &&
1764 n->child->type == ROFFT_TEXT)
1770 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1771 n->type == ROFFT_BODY ? TYPE_Va : 0));
1779 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1780 const struct roff_node *n)
1784 if (NULL == (n = n->child))
1787 if (NULL == n->next) {
1788 putkey(mpage, n->string, TYPE_Xr);
1792 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1793 putkey(mpage, cp, TYPE_Xr);
1799 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1800 const struct roff_node *n)
1803 if (n->type == ROFFT_BODY)
1804 deroff(&mpage->desc, n);
1809 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1810 const struct roff_node *n)
1813 if (SEC_NAME == n->sec)
1814 putmdockey(mpage, n->child, NAME_TITLE, 0);
1815 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1816 if (n->child == NULL)
1817 putkey(mpage, meta->name, NAME_SYN);
1819 putmdockey(mpage, n->child, NAME_SYN, 0);
1821 if ( ! (mpage->name_head_done ||
1822 n->child == NULL || n->child->string == NULL ||
1823 strcasecmp(n->child->string, meta->title))) {
1824 putkey(mpage, n->child->string, NAME_HEAD);
1825 mpage->name_head_done = 1;
1831 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1832 const struct roff_node *n)
1835 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1839 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1840 const struct roff_node *n)
1843 return n->type == ROFFT_HEAD;
1847 * Add a string to the hash table for the current manual.
1848 * Each string has a bitmask telling which macros it belongs to.
1849 * When we finish the manual, we'll dump the table.
1852 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1863 mustfree = render_string(&cp, &sz);
1869 name_mask &= ~NAME_FIRST;
1871 say(mpage->mlinks->file,
1872 "Adding name %*s, bits=0x%llx", (int)sz, cp,
1873 (unsigned long long)v);
1877 for (i = 0; i < KEY_MAX; i++)
1878 if ((uint64_t)1 << i & v)
1879 say(mpage->mlinks->file,
1880 "Adding key %s=%*s",
1881 mansearch_keynames[i], (int)sz, cp);
1885 slot = ohash_qlookupi(htab, cp, &end);
1886 s = ohash_find(htab, slot);
1888 if (NULL != s && mpage == s->mpage) {
1891 } else if (NULL == s) {
1892 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1893 memcpy(s->key, cp, sz);
1894 ohash_insert(htab, slot, s);
1904 * Take a Unicode codepoint and produce its UTF-8 encoding.
1905 * This isn't the best way to do this, but it works.
1906 * The magic numbers are from the UTF-8 packaging.
1907 * They're not as scary as they seem: read the UTF-8 spec for details.
1910 utf8(unsigned int cp, char out[7])
1915 if (cp <= 0x0000007F) {
1918 } else if (cp <= 0x000007FF) {
1920 out[0] = (cp >> 6 & 31) | 192;
1921 out[1] = (cp & 63) | 128;
1922 } else if (cp <= 0x0000FFFF) {
1924 out[0] = (cp >> 12 & 15) | 224;
1925 out[1] = (cp >> 6 & 63) | 128;
1926 out[2] = (cp & 63) | 128;
1927 } else if (cp <= 0x001FFFFF) {
1929 out[0] = (cp >> 18 & 7) | 240;
1930 out[1] = (cp >> 12 & 63) | 128;
1931 out[2] = (cp >> 6 & 63) | 128;
1932 out[3] = (cp & 63) | 128;
1933 } else if (cp <= 0x03FFFFFF) {
1935 out[0] = (cp >> 24 & 3) | 248;
1936 out[1] = (cp >> 18 & 63) | 128;
1937 out[2] = (cp >> 12 & 63) | 128;
1938 out[3] = (cp >> 6 & 63) | 128;
1939 out[4] = (cp & 63) | 128;
1940 } else if (cp <= 0x7FFFFFFF) {
1942 out[0] = (cp >> 30 & 1) | 252;
1943 out[1] = (cp >> 24 & 63) | 128;
1944 out[2] = (cp >> 18 & 63) | 128;
1945 out[3] = (cp >> 12 & 63) | 128;
1946 out[4] = (cp >> 6 & 63) | 128;
1947 out[5] = (cp & 63) | 128;
1956 * If the string contains escape sequences,
1957 * replace it with an allocated rendering and return 1,
1958 * such that the caller can free it after use.
1959 * Otherwise, do nothing and return 0.
1962 render_string(char **public, size_t *psz)
1964 const char *src, *scp, *addcp, *seq;
1966 size_t ssz, dsz, addsz;
1967 char utfbuf[7], res[6];
1968 int seqlen, unicode;
1972 res[2] = ASCII_NBRSP;
1973 res[3] = ASCII_HYPH;
1974 res[4] = ASCII_BREAK;
1977 src = scp = *public;
1982 while (scp < src + *psz) {
1984 /* Leave normal characters unchanged. */
1986 if (strchr(res, *scp) == NULL) {
1994 * Found something that requires replacing,
1995 * make sure we have a destination buffer.
1999 dst = mandoc_malloc(ssz + 1);
2001 memcpy(dst, src, dsz);
2004 /* Handle single-char special characters. */
2025 * Found an escape sequence.
2026 * Read past the slash, then parse it.
2027 * Ignore everything except characters.
2031 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
2035 * Render the special character
2036 * as either UTF-8 or ASCII.
2040 unicode = mchars_spec2cp(seq, seqlen);
2043 addsz = utf8(unicode, utfbuf);
2048 addcp = mchars_spec2str(seq, seqlen, &addsz);
2051 if (*addcp == ASCII_NBRSP) {
2057 /* Copy the rendered glyph into the stream. */
2060 dst = mandoc_realloc(dst, ssz + 1);
2061 memcpy(dst + dsz, addcp, addsz);
2069 /* Trim trailing whitespace and NUL-terminate. */
2071 while (*psz > 0 && (*public)[*psz - 1] == ' ')
2074 (*public)[*psz] = '\0';
2081 dbadd_mlink(const struct mlink *mlink)
2083 dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
2084 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
2085 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
2086 dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
2087 dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
2091 * Flush the current page's terms (and their bits) into the database.
2092 * Also, handle escape sequences at the last possible moment.
2095 dbadd(struct dba *dba, struct mpage *mpage)
2097 struct mlink *mlink;
2105 mlink = mpage->mlinks;
2108 for (key = ohash_first(&names, &slot); NULL != key;
2109 key = ohash_next(&names, &slot))
2111 for (key = ohash_first(&strings, &slot); NULL != key;
2112 key = ohash_next(&strings, &slot))
2116 while (NULL != mlink) {
2117 fputs(mlink->name, stdout);
2118 if (NULL == mlink->next ||
2119 strcmp(mlink->dsec, mlink->next->dsec) ||
2120 strcmp(mlink->fsec, mlink->next->fsec) ||
2121 strcmp(mlink->arch, mlink->next->arch)) {
2123 if ('\0' == *mlink->dsec)
2124 fputs(mlink->fsec, stdout);
2126 fputs(mlink->dsec, stdout);
2127 if ('\0' != *mlink->arch)
2128 printf("/%s", mlink->arch);
2131 mlink = mlink->next;
2133 fputs(", ", stdout);
2135 printf(" - %s\n", mpage->desc);
2140 say(mlink->file, "Adding to database");
2144 mustfree = render_string(&cp, &i);
2145 mpage->dba = dba_page_new(dba->pages,
2146 *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2147 cp, mlink->file, mpage->form);
2150 dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
2152 while (mlink != NULL) {
2154 mlink = mlink->next;
2157 for (key = ohash_first(&names, &slot); NULL != key;
2158 key = ohash_next(&names, &slot)) {
2159 assert(key->mpage == mpage);
2160 dba_page_alias(mpage->dba, key->key, key->mask);
2163 for (key = ohash_first(&strings, &slot); NULL != key;
2164 key = ohash_next(&strings, &slot)) {
2165 assert(key->mpage == mpage);
2167 for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2168 if (key->mask & mask)
2169 dba_macro_add(dba->macros, i,
2170 key->key, mpage->dba);
2178 dbprune(struct dba *dba)
2180 struct dba_array *page, *files;
2183 dba_array_FOREACH(dba->pages, page) {
2184 files = dba_array_get(page, DBP_FILE);
2185 dba_array_FOREACH(files, file) {
2188 if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2191 say(file, "Deleting from database");
2192 dba_array_del(dba->pages);
2200 * Write the database from memory to disk.
2203 dbwrite(struct dba *dba)
2205 struct stat sb1, sb2;
2206 char tfn[33], *cp1, *cp2;
2211 * Do not write empty databases, and delete existing ones
2212 * when makewhatis -u causes them to become empty.
2215 dba_array_start(dba->pages);
2216 if (dba_array_next(dba->pages) == NULL) {
2217 if (unlink(MANDOC_DB) == -1 && errno != ENOENT)
2218 say(MANDOC_DB, "&unlink");
2223 * Build the database in a temporary file,
2224 * then atomically move it into place.
2227 if (dba_write(MANDOC_DB "~", dba) != -1) {
2228 if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2229 exitcode = (int)MANDOCLEVEL_SYSERR;
2230 say(MANDOC_DB, "&rename");
2231 unlink(MANDOC_DB "~");
2237 * We lack write permission and cannot replace the database
2238 * file, but let's at least check whether the data changed.
2241 (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2242 if (mkdtemp(tfn) == NULL) {
2243 exitcode = (int)MANDOCLEVEL_SYSERR;
2244 say("", "&%s", tfn);
2247 cp1 = cp2 = MAP_FAILED;
2249 (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2250 if (dba_write(tfn, dba) == -1) {
2251 say(tfn, "&dba_write");
2254 if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) {
2255 say(MANDOC_DB, "&open");
2258 if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) {
2262 if (fstat(fd1, &sb1) == -1) {
2263 say(MANDOC_DB, "&fstat");
2266 if (fstat(fd2, &sb2) == -1) {
2270 if (sb1.st_size != sb2.st_size)
2272 if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE,
2273 fd1, 0)) == MAP_FAILED) {
2274 say(MANDOC_DB, "&mmap");
2277 if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE,
2278 fd2, 0)) == MAP_FAILED) {
2282 for (i = 0; i < sb1.st_size; i++)
2283 if (cp1[i] != cp2[i])
2288 exitcode = (int)MANDOCLEVEL_SYSERR;
2289 say(MANDOC_DB, "Data changed, but cannot replace database");
2292 if (cp1 != MAP_FAILED)
2293 munmap(cp1, sb1.st_size);
2294 if (cp2 != MAP_FAILED)
2295 munmap(cp2, sb2.st_size);
2301 *strrchr(tfn, '/') = '\0';
2306 set_basedir(const char *targetdir, int report_baddir)
2308 static char startdir[PATH_MAX];
2309 static int getcwd_status; /* 1 = ok, 2 = failure */
2310 static int chdir_status; /* 1 = changed directory */
2313 * Remember the original working directory, if possible.
2314 * This will be needed if the second or a later directory
2315 * on the command line is given as a relative path.
2316 * Do not error out if the current directory is not
2317 * searchable: Maybe it won't be needed after all.
2319 if (getcwd_status == 0) {
2320 if (getcwd(startdir, sizeof(startdir)) == NULL) {
2322 (void)strlcpy(startdir, strerror(errno),
2329 * We are leaving the old base directory.
2330 * Do not use it any longer, not even for messages.
2336 * If and only if the directory was changed earlier and
2337 * the next directory to process is given as a relative path,
2338 * first go back, or bail out if that is impossible.
2340 if (chdir_status && *targetdir != '/') {
2341 if (getcwd_status == 2) {
2342 exitcode = (int)MANDOCLEVEL_SYSERR;
2343 say("", "getcwd: %s", startdir);
2346 if (chdir(startdir) == -1) {
2347 exitcode = (int)MANDOCLEVEL_SYSERR;
2348 say("", "&chdir %s", startdir);
2354 * Always resolve basedir to the canonicalized absolute
2355 * pathname and append a trailing slash, such that
2356 * we can reliably check whether files are inside.
2358 if (realpath(targetdir, basedir) == NULL) {
2359 if (report_baddir || errno != ENOENT) {
2360 exitcode = (int)MANDOCLEVEL_BADARG;
2361 say("", "&%s: realpath", targetdir);
2365 } else if (chdir(basedir) == -1) {
2366 if (report_baddir || errno != ENOENT) {
2367 exitcode = (int)MANDOCLEVEL_BADARG;
2374 basedir_len = strlen(basedir);
2375 if (basedir[basedir_len - 1] != '/') {
2376 if (basedir_len >= PATH_MAX - 1) {
2377 exitcode = (int)MANDOCLEVEL_SYSERR;
2378 say("", "Filename too long");
2383 basedir[basedir_len++] = '/';
2384 basedir[basedir_len] = '\0';
2389 #ifdef READ_ALLOWED_PATH
2391 read_allowed(const char *candidate)
2396 for (cp = READ_ALLOWED_PATH;; cp += len) {
2401 len = strcspn(cp, ":");
2402 if (strncmp(candidate, cp, len) == 0)
2409 say(const char *file, const char *format, ...)
2414 if (*basedir != '\0')
2415 fprintf(stderr, "%s", basedir);
2416 if (*basedir != '\0' && *file != '\0')
2419 fprintf(stderr, "%s", file);
2422 if (format != NULL) {
2435 if (format != NULL) {
2436 if (*basedir != '\0' || *file != '\0')
2437 fputs(": ", stderr);
2438 va_start(ap, format);
2439 vfprintf(stderr, format, ap);
2443 if (*basedir != '\0' || *file != '\0' || format != NULL)
2444 fputs(": ", stderr);
2447 fputc('\n', stderr);