-/* $Id: mandocdb.c,v 1.79 2013/12/26 22:30:41 schwarze Exp $ */
+/* $Id: mandocdb.c,v 1.250 2017/05/17 22:27:12 schwarze Exp $ */
/*
* Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
- * Copyright (c) 2011, 2012, 2013 Ingo Schwarze <schwarze@openbsd.org>
+ * Copyright (c) 2011-2017 Ingo Schwarze <schwarze@openbsd.org>
+ * Copyright (c) 2016 Ed Maste <emaste@freebsd.org>
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
- * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
- * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
-#ifdef HAVE_CONFIG_H
#include "config.h"
-#endif
+#include <sys/types.h>
#include <sys/stat.h>
+#include <sys/wait.h>
#include <assert.h>
#include <ctype.h>
+#if HAVE_ERR
+#include <err.h>
+#endif
#include <errno.h>
#include <fcntl.h>
+#if HAVE_FTS
#include <fts.h>
-#include <getopt.h>
+#else
+#include "compat_fts.h"
+#endif
#include <limits.h>
+#if HAVE_SANDBOX_INIT
+#include <sandbox.h>
+#endif
+#include <stdarg.h>
#include <stddef.h>
#include <stdio.h>
#include <stdint.h>
#include <string.h>
#include <unistd.h>
-#ifdef HAVE_OHASH
-#include <ohash.h>
-#else
-#include "compat_ohash.h"
-#endif
-#include <sqlite3.h>
-
+#include "mandoc_aux.h"
+#include "mandoc_ohash.h"
+#include "mandoc.h"
+#include "roff.h"
#include "mdoc.h"
#include "man.h"
-#include "mandoc.h"
-#include "manpath.h"
+#include "manconf.h"
#include "mansearch.h"
+#include "dba_array.h"
+#include "dba.h"
-#define SQL_EXEC(_v) \
- if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
- fprintf(stderr, "%s\n", sqlite3_errmsg(db))
-#define SQL_BIND_TEXT(_s, _i, _v) \
- if (SQLITE_OK != sqlite3_bind_text \
- ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
- fprintf(stderr, "%s\n", sqlite3_errmsg(db))
-#define SQL_BIND_INT(_s, _i, _v) \
- if (SQLITE_OK != sqlite3_bind_int \
- ((_s), (_i)++, (_v))) \
- fprintf(stderr, "%s\n", sqlite3_errmsg(db))
-#define SQL_BIND_INT64(_s, _i, _v) \
- if (SQLITE_OK != sqlite3_bind_int64 \
- ((_s), (_i)++, (_v))) \
- fprintf(stderr, "%s\n", sqlite3_errmsg(db))
-#define SQL_STEP(_s) \
- if (SQLITE_DONE != sqlite3_step((_s))) \
- fprintf(stderr, "%s\n", sqlite3_errmsg(db))
+extern const char *const mansearch_keynames[];
enum op {
OP_DEFAULT = 0, /* new dbs from dir list or default config */
OP_TEST /* change no databases, report potential problems */
};
-enum form {
- FORM_SRC, /* format is -man or -mdoc */
- FORM_CAT, /* format is cat */
- FORM_NONE /* format is unknown */
-};
-
struct str {
- char *utf8; /* key in UTF-8 form */
const struct mpage *mpage; /* if set, the owning parse */
uint64_t mask; /* bitmask in sequence */
- char key[]; /* the string itself */
+ char key[]; /* rendered text */
};
struct inodev {
struct mpage {
struct inodev inodev; /* used for hashing routine */
- struct mpage *next; /* next in mpages_list */
- enum form dform; /* path-cued form */
- enum form sform; /* suffix-cued form */
- char file[PATH_MAX]; /* filename rel. to manpath */
- char *desc; /* parsed description */
- char *name; /* name (from filename) (not empty) */
- char *sec; /* suffix-cued section (or empty) */
- char *dsec; /* path-cued section (or empty) */
- char *arch; /* path-cued arch. (or empty) */
-};
-
-struct title {
- char *title; /* name(sec/arch) given inside the file */
- char *file; /* file name in case of mismatch */
+ struct dba_array *dba;
+ char *sec; /* section from file content */
+ char *arch; /* architecture from file content */
+ char *title; /* title from file content */
+ char *desc; /* description from file content */
+ struct mpage *next; /* singly linked list */
+ struct mlink *mlinks; /* singly linked list */
+ int name_head_done;
+ enum form form; /* format from file content */
};
-enum stmt {
- STMT_DELETE = 0, /* delete manpage */
- STMT_INSERT_DOC, /* insert manpage */
- STMT_INSERT_KEY, /* insert parsed key */
- STMT__MAX
+struct mlink {
+ char file[PATH_MAX]; /* filename rel. to manpath */
+ char *dsec; /* section from directory */
+ char *arch; /* architecture from directory */
+ char *name; /* name from file name (not empty) */
+ char *fsec; /* section from file name suffix */
+ struct mlink *next; /* singly linked list */
+ struct mpage *mpage; /* parent */
+ int gzip; /* filename has a .gz suffix */
+ enum form dform; /* format from directory */
+ enum form fform; /* format from file name suffix */
};
-typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
+typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
struct mdoc_handler {
mdoc_fp fp; /* optional handler */
uint64_t mask; /* set unless handler returns 0 */
+ int taboo; /* node flags that must not be set */
};
-static void dbclose(int);
-static void dbindex(struct mchars *, int, const struct mpage *);
-static int dbopen(int);
-static void dbprune(void);
-static void fileadd(struct mpage *);
-static int filecheck(const char *);
+
+int mandocdb(int, char *[]);
+
+static void dbadd(struct dba *, struct mpage *);
+static void dbadd_mlink(const struct mlink *mlink);
+static void dbprune(struct dba *);
+static void dbwrite(struct dba *);
static void filescan(const char *);
-static void *hash_alloc(size_t, void *);
-static void hash_free(void *, size_t, void *);
-static void *hash_halloc(size_t, void *);
-static void inoadd(const struct stat *, struct mpage *);
-static int inocheck(const struct stat *);
-static void ofadd(int, const char *, const char *, const char *,
- const char *, const char *, const struct stat *);
+#if HAVE_FTS_COMPARE_CONST
+static int fts_compare(const FTSENT *const *, const FTSENT *const *);
+#else
+static int fts_compare(const FTSENT **, const FTSENT **);
+#endif
+static void mlink_add(struct mlink *, const struct stat *);
+static void mlink_check(struct mpage *, struct mlink *);
+static void mlink_free(struct mlink *);
+static void mlinks_undupe(struct mpage *);
static void mpages_free(void);
-static void mpages_merge(struct mchars *, struct mparse *, int);
-static void parse_cat(struct mpage *);
-static void parse_man(struct mpage *, const struct man_node *);
-static void parse_mdoc(struct mpage *, const struct mdoc_node *);
-static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
-static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
-static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
-static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
-static int parse_mdoc_In(struct mpage *, const struct mdoc_node *);
-static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
-static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
-static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
-static int parse_mdoc_St(struct mpage *, const struct mdoc_node *);
-static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
-static int set_basedir(const char *);
-static void putkey(const struct mpage *,
- const char *, uint64_t);
-static void putkeys(const struct mpage *,
- const char *, size_t, uint64_t);
+static void mpages_merge(struct dba *, struct mparse *);
+static void parse_cat(struct mpage *, int);
+static void parse_man(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static void parse_mdoc(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
+static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
+ const struct roff_node *);
+static void putkey(const struct mpage *, char *, uint64_t);
+static void putkeys(const struct mpage *, char *, size_t, uint64_t);
static void putmdockey(const struct mpage *,
- const struct mdoc_node *, uint64_t);
-static void say(const char *, const char *, ...);
+ const struct roff_node *, uint64_t, int);
+static int render_string(char **, size_t *);
+static void say(const char *, const char *, ...)
+ __attribute__((__format__ (__printf__, 2, 3)));
+static int set_basedir(const char *, int);
static int treescan(void);
static size_t utf8(unsigned int, char [7]);
-static void utf8key(struct mchars *, struct str *);
-static char *progname;
-static int use_all; /* use all found files */
static int nodb; /* no database changes */
-static int verb; /* print what we're doing */
-static int warnings; /* warn about crap */
+static int mparse_options; /* abort the parse early */
+static int use_all; /* use all found files */
+static int debug; /* print what we're doing */
+static int warnings; /* warn about crap */
+static int write_utf8; /* write UTF-8 output; else ASCII */
static int exitcode; /* to be returned by main */
-static enum op op; /* operational mode */
+static enum op op; /* operational mode */
static char basedir[PATH_MAX]; /* current base directory */
+static struct mpage *mpage_head; /* list of distinct manual pages */
static struct ohash mpages; /* table of distinct manual pages */
-static struct ohash filenames; /* table of filenames */
+static struct ohash mlinks; /* table of directory entries */
+static struct ohash names; /* table of all names */
static struct ohash strings; /* table of all strings */
-static struct mpage *mpages_list = NULL; /* vector of files to parse */
-static sqlite3 *db = NULL; /* current database */
-static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
-
-static const struct mdoc_handler mdocs[MDOC_MAX] = {
- { NULL, 0 }, /* Ap */
- { NULL, 0 }, /* Dd */
- { NULL, 0 }, /* Dt */
- { NULL, 0 }, /* Os */
- { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
- { parse_mdoc_head, TYPE_Ss }, /* Ss */
- { NULL, 0 }, /* Pp */
- { NULL, 0 }, /* D1 */
- { NULL, 0 }, /* Dl */
- { NULL, 0 }, /* Bd */
- { NULL, 0 }, /* Ed */
- { NULL, 0 }, /* Bl */
- { NULL, 0 }, /* El */
- { NULL, 0 }, /* It */
- { NULL, 0 }, /* Ad */
- { NULL, TYPE_An }, /* An */
- { NULL, TYPE_Ar }, /* Ar */
- { NULL, TYPE_Cd }, /* Cd */
- { NULL, TYPE_Cm }, /* Cm */
- { NULL, TYPE_Dv }, /* Dv */
- { NULL, TYPE_Er }, /* Er */
- { NULL, TYPE_Ev }, /* Ev */
- { NULL, 0 }, /* Ex */
- { NULL, TYPE_Fa }, /* Fa */
- { parse_mdoc_Fd, 0 }, /* Fd */
- { NULL, TYPE_Fl }, /* Fl */
- { parse_mdoc_Fn, 0 }, /* Fn */
- { NULL, TYPE_Ft }, /* Ft */
- { NULL, TYPE_Ic }, /* Ic */
- { parse_mdoc_In, TYPE_In }, /* In */
- { NULL, TYPE_Li }, /* Li */
- { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
- { parse_mdoc_Nm, TYPE_Nm }, /* Nm */
- { NULL, 0 }, /* Op */
- { NULL, 0 }, /* Ot */
- { NULL, TYPE_Pa }, /* Pa */
- { NULL, 0 }, /* Rv */
- { parse_mdoc_St, 0 }, /* St */
- { NULL, TYPE_Va }, /* Va */
- { parse_mdoc_body, TYPE_Va }, /* Vt */
- { parse_mdoc_Xr, 0 }, /* Xr */
- { NULL, 0 }, /* %A */
- { NULL, 0 }, /* %B */
- { NULL, 0 }, /* %D */
- { NULL, 0 }, /* %I */
- { NULL, 0 }, /* %J */
- { NULL, 0 }, /* %N */
- { NULL, 0 }, /* %O */
- { NULL, 0 }, /* %P */
- { NULL, 0 }, /* %R */
- { NULL, 0 }, /* %T */
- { NULL, 0 }, /* %V */
- { NULL, 0 }, /* Ac */
- { NULL, 0 }, /* Ao */
- { NULL, 0 }, /* Aq */
- { NULL, TYPE_At }, /* At */
- { NULL, 0 }, /* Bc */
- { NULL, 0 }, /* Bf */
- { NULL, 0 }, /* Bo */
- { NULL, 0 }, /* Bq */
- { NULL, TYPE_Bsx }, /* Bsx */
- { NULL, TYPE_Bx }, /* Bx */
- { NULL, 0 }, /* Db */
- { NULL, 0 }, /* Dc */
- { NULL, 0 }, /* Do */
- { NULL, 0 }, /* Dq */
- { NULL, 0 }, /* Ec */
- { NULL, 0 }, /* Ef */
- { NULL, TYPE_Em }, /* Em */
- { NULL, 0 }, /* Eo */
- { NULL, TYPE_Fx }, /* Fx */
- { NULL, TYPE_Ms }, /* Ms */
- { NULL, 0 }, /* No */
- { NULL, 0 }, /* Ns */
- { NULL, TYPE_Nx }, /* Nx */
- { NULL, TYPE_Ox }, /* Ox */
- { NULL, 0 }, /* Pc */
- { NULL, 0 }, /* Pf */
- { NULL, 0 }, /* Po */
- { NULL, 0 }, /* Pq */
- { NULL, 0 }, /* Qc */
- { NULL, 0 }, /* Ql */
- { NULL, 0 }, /* Qo */
- { NULL, 0 }, /* Qq */
- { NULL, 0 }, /* Re */
- { NULL, 0 }, /* Rs */
- { NULL, 0 }, /* Sc */
- { NULL, 0 }, /* So */
- { NULL, 0 }, /* Sq */
- { NULL, 0 }, /* Sm */
- { NULL, 0 }, /* Sx */
- { NULL, TYPE_Sy }, /* Sy */
- { NULL, TYPE_Tn }, /* Tn */
- { NULL, 0 }, /* Ux */
- { NULL, 0 }, /* Xc */
- { NULL, 0 }, /* Xo */
- { parse_mdoc_head, 0 }, /* Fo */
- { NULL, 0 }, /* Fc */
- { NULL, 0 }, /* Oo */
- { NULL, 0 }, /* Oc */
- { NULL, 0 }, /* Bk */
- { NULL, 0 }, /* Ek */
- { NULL, 0 }, /* Bt */
- { NULL, 0 }, /* Hf */
- { NULL, 0 }, /* Fr */
- { NULL, 0 }, /* Ud */
- { NULL, TYPE_Lb }, /* Lb */
- { NULL, 0 }, /* Lp */
- { NULL, TYPE_Lk }, /* Lk */
- { NULL, TYPE_Mt }, /* Mt */
- { NULL, 0 }, /* Brq */
- { NULL, 0 }, /* Bro */
- { NULL, 0 }, /* Brc */
- { NULL, 0 }, /* %C */
- { NULL, 0 }, /* Es */
- { NULL, 0 }, /* En */
- { NULL, TYPE_Dx }, /* Dx */
- { NULL, 0 }, /* %Q */
- { NULL, 0 }, /* br */
- { NULL, 0 }, /* sp */
- { NULL, 0 }, /* %U */
- { NULL, 0 }, /* Ta */
+static uint64_t name_mask;
+
+static const struct mdoc_handler __mdocs[MDOC_MAX - MDOC_Dd] = {
+ { NULL, 0, NODE_NOPRT }, /* Dd */
+ { NULL, 0, NODE_NOPRT }, /* Dt */
+ { NULL, 0, NODE_NOPRT }, /* Os */
+ { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
+ { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
+ { NULL, 0, 0 }, /* Pp */
+ { NULL, 0, 0 }, /* D1 */
+ { NULL, 0, 0 }, /* Dl */
+ { NULL, 0, 0 }, /* Bd */
+ { NULL, 0, 0 }, /* Ed */
+ { NULL, 0, 0 }, /* Bl */
+ { NULL, 0, 0 }, /* El */
+ { NULL, 0, 0 }, /* It */
+ { NULL, 0, 0 }, /* Ad */
+ { NULL, TYPE_An, 0 }, /* An */
+ { NULL, 0, 0 }, /* Ap */
+ { NULL, TYPE_Ar, 0 }, /* Ar */
+ { NULL, TYPE_Cd, 0 }, /* Cd */
+ { NULL, TYPE_Cm, 0 }, /* Cm */
+ { NULL, TYPE_Dv, 0 }, /* Dv */
+ { NULL, TYPE_Er, 0 }, /* Er */
+ { NULL, TYPE_Ev, 0 }, /* Ev */
+ { NULL, 0, 0 }, /* Ex */
+ { NULL, TYPE_Fa, 0 }, /* Fa */
+ { parse_mdoc_Fd, 0, 0 }, /* Fd */
+ { NULL, TYPE_Fl, 0 }, /* Fl */
+ { parse_mdoc_Fn, 0, 0 }, /* Fn */
+ { NULL, TYPE_Ft, 0 }, /* Ft */
+ { NULL, TYPE_Ic, 0 }, /* Ic */
+ { NULL, TYPE_In, 0 }, /* In */
+ { NULL, TYPE_Li, 0 }, /* Li */
+ { parse_mdoc_Nd, 0, 0 }, /* Nd */
+ { parse_mdoc_Nm, 0, 0 }, /* Nm */
+ { NULL, 0, 0 }, /* Op */
+ { NULL, 0, 0 }, /* Ot */
+ { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
+ { NULL, 0, 0 }, /* Rv */
+ { NULL, TYPE_St, 0 }, /* St */
+ { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
+ { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
+ { parse_mdoc_Xr, 0, 0 }, /* Xr */
+ { NULL, 0, 0 }, /* %A */
+ { NULL, 0, 0 }, /* %B */
+ { NULL, 0, 0 }, /* %D */
+ { NULL, 0, 0 }, /* %I */
+ { NULL, 0, 0 }, /* %J */
+ { NULL, 0, 0 }, /* %N */
+ { NULL, 0, 0 }, /* %O */
+ { NULL, 0, 0 }, /* %P */
+ { NULL, 0, 0 }, /* %R */
+ { NULL, 0, 0 }, /* %T */
+ { NULL, 0, 0 }, /* %V */
+ { NULL, 0, 0 }, /* Ac */
+ { NULL, 0, 0 }, /* Ao */
+ { NULL, 0, 0 }, /* Aq */
+ { NULL, TYPE_At, 0 }, /* At */
+ { NULL, 0, 0 }, /* Bc */
+ { NULL, 0, 0 }, /* Bf */
+ { NULL, 0, 0 }, /* Bo */
+ { NULL, 0, 0 }, /* Bq */
+ { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
+ { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
+ { NULL, 0, 0 }, /* Db */
+ { NULL, 0, 0 }, /* Dc */
+ { NULL, 0, 0 }, /* Do */
+ { NULL, 0, 0 }, /* Dq */
+ { NULL, 0, 0 }, /* Ec */
+ { NULL, 0, 0 }, /* Ef */
+ { NULL, TYPE_Em, 0 }, /* Em */
+ { NULL, 0, 0 }, /* Eo */
+ { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
+ { NULL, TYPE_Ms, 0 }, /* Ms */
+ { NULL, 0, 0 }, /* No */
+ { NULL, 0, 0 }, /* Ns */
+ { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
+ { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
+ { NULL, 0, 0 }, /* Pc */
+ { NULL, 0, 0 }, /* Pf */
+ { NULL, 0, 0 }, /* Po */
+ { NULL, 0, 0 }, /* Pq */
+ { NULL, 0, 0 }, /* Qc */
+ { NULL, 0, 0 }, /* Ql */
+ { NULL, 0, 0 }, /* Qo */
+ { NULL, 0, 0 }, /* Qq */
+ { NULL, 0, 0 }, /* Re */
+ { NULL, 0, 0 }, /* Rs */
+ { NULL, 0, 0 }, /* Sc */
+ { NULL, 0, 0 }, /* So */
+ { NULL, 0, 0 }, /* Sq */
+ { NULL, 0, 0 }, /* Sm */
+ { NULL, 0, 0 }, /* Sx */
+ { NULL, TYPE_Sy, 0 }, /* Sy */
+ { NULL, TYPE_Tn, 0 }, /* Tn */
+ { NULL, 0, NODE_NOSRC }, /* Ux */
+ { NULL, 0, 0 }, /* Xc */
+ { NULL, 0, 0 }, /* Xo */
+ { parse_mdoc_Fo, 0, 0 }, /* Fo */
+ { NULL, 0, 0 }, /* Fc */
+ { NULL, 0, 0 }, /* Oo */
+ { NULL, 0, 0 }, /* Oc */
+ { NULL, 0, 0 }, /* Bk */
+ { NULL, 0, 0 }, /* Ek */
+ { NULL, 0, 0 }, /* Bt */
+ { NULL, 0, 0 }, /* Hf */
+ { NULL, 0, 0 }, /* Fr */
+ { NULL, 0, 0 }, /* Ud */
+ { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */
+ { NULL, 0, 0 }, /* Lp */
+ { NULL, TYPE_Lk, 0 }, /* Lk */
+ { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
+ { NULL, 0, 0 }, /* Brq */
+ { NULL, 0, 0 }, /* Bro */
+ { NULL, 0, 0 }, /* Brc */
+ { NULL, 0, 0 }, /* %C */
+ { NULL, 0, 0 }, /* Es */
+ { NULL, 0, 0 }, /* En */
+ { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
+ { NULL, 0, 0 }, /* %Q */
+ { NULL, 0, 0 }, /* %U */
+ { NULL, 0, 0 }, /* Ta */
};
+static const struct mdoc_handler *const mdocs = __mdocs - MDOC_Dd;
+
int
-main(int argc, char *argv[])
+mandocdb(int argc, char *argv[])
{
- int ch, i;
- size_t j, sz;
- const char *path_arg;
- struct mchars *mc;
- struct manpaths dirs;
+ struct manconf conf;
struct mparse *mp;
- struct ohash_info mpages_info, filename_info;
-
- memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
- memset(&dirs, 0, sizeof(struct manpaths));
+ struct dba *dba;
+ const char *path_arg, *progname;
+ size_t j, sz;
+ int ch, i;
- mpages_info.alloc = filename_info.alloc = hash_alloc;
- mpages_info.halloc = filename_info.halloc = hash_halloc;
- mpages_info.hfree = filename_info.hfree = hash_free;
+#if HAVE_PLEDGE
+ if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) {
+ warn("pledge");
+ return (int)MANDOCLEVEL_SYSERR;
+ }
+#endif
- mpages_info.key_offset = offsetof(struct mpage, inodev);
- filename_info.key_offset = offsetof(struct mpage, file);
+#if HAVE_SANDBOX_INIT
+ if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) {
+ warnx("sandbox_init");
+ return (int)MANDOCLEVEL_SYSERR;
+ }
+#endif
- progname = strrchr(argv[0], '/');
- if (progname == NULL)
- progname = argv[0];
- else
- ++progname;
+ memset(&conf, 0, sizeof(conf));
/*
- * We accept a few different invocations.
+ * We accept a few different invocations.
* The CHECKOP macro makes sure that invocation styles don't
* clobber each other.
*/
#define CHECKOP(_op, _ch) do \
if (OP_DEFAULT != (_op)) { \
- fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
+ warnx("-%c: Conflicting option", (_ch)); \
goto usage; \
} while (/*CONSTCOND*/0)
path_arg = NULL;
op = OP_DEFAULT;
- while (-1 != (ch = getopt(argc, argv, "aC:d:ntu:vW")))
+ while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
switch (ch) {
- case ('a'):
+ case 'a':
use_all = 1;
break;
- case ('C'):
+ case 'C':
CHECKOP(op, ch);
path_arg = optarg;
op = OP_CONFFILE;
break;
- case ('d'):
+ case 'D':
+ debug++;
+ break;
+ case 'd':
CHECKOP(op, ch);
path_arg = optarg;
op = OP_UPDATE;
break;
- case ('n'):
+ case 'n':
nodb = 1;
break;
- case ('t'):
+ case 'p':
+ warnings = 1;
+ break;
+ case 'Q':
+ mparse_options |= MPARSE_QUICK;
+ break;
+ case 'T':
+ if (strcmp(optarg, "utf8")) {
+ warnx("-T%s: Unsupported output format",
+ optarg);
+ goto usage;
+ }
+ write_utf8 = 1;
+ break;
+ case 't':
CHECKOP(op, ch);
dup2(STDOUT_FILENO, STDERR_FILENO);
op = OP_TEST;
nodb = warnings = 1;
break;
- case ('u'):
+ case 'u':
CHECKOP(op, ch);
path_arg = optarg;
op = OP_DELETE;
break;
- case ('v'):
- verb++;
- break;
- case ('W'):
- warnings = 1;
+ case 'v':
+ /* Compatibility with espie@'s makewhatis. */
break;
default:
goto usage;
argc -= optind;
argv += optind;
+#if HAVE_PLEDGE
+ if (nodb) {
+ if (pledge("stdio rpath", NULL) == -1) {
+ warn("pledge");
+ return (int)MANDOCLEVEL_SYSERR;
+ }
+ }
+#endif
+
if (OP_CONFFILE == op && argc > 0) {
- fprintf(stderr, "-C: Too many arguments\n");
+ warnx("-C: Too many arguments");
goto usage;
}
exitcode = (int)MANDOCLEVEL_OK;
- mp = mparse_alloc(MPARSE_AUTO,
- MANDOCLEVEL_FATAL, NULL, NULL, NULL);
- mc = mchars_alloc();
-
- ohash_init(&mpages, 6, &mpages_info);
- ohash_init(&filenames, 6, &filename_info);
+ mchars_alloc();
+ mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
+ mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
+ mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
- /*
- * Force processing all files.
- */
- use_all = 1;
/*
- * All of these deal with a specific directory.
- * Jump into that directory then collect files specified
- * on the command-line.
+ * Most of these deal with a specific directory.
+ * Jump into that directory first.
*/
- if (0 == set_basedir(path_arg))
+ if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
goto out;
- for (i = 0; i < argc; i++)
- filescan(argv[i]);
- if (0 == dbopen(1))
- goto out;
- if (OP_TEST != op)
- dbprune();
+
+ dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
+ if (dba != NULL) {
+ /*
+ * The existing database is usable. Process
+ * all files specified on the command-line.
+ */
+#if HAVE_PLEDGE
+ if (!nodb) {
+ if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) {
+ warn("pledge");
+ exitcode = (int)MANDOCLEVEL_SYSERR;
+ goto out;
+ }
+ }
+#endif
+ use_all = 1;
+ for (i = 0; i < argc; i++)
+ filescan(argv[i]);
+ if (nodb == 0)
+ dbprune(dba);
+ } else {
+ /* Database missing or corrupt. */
+ if (op != OP_UPDATE || errno != ENOENT)
+ say(MANDOC_DB, "%s: Automatically recreating"
+ " from scratch", strerror(errno));
+ exitcode = (int)MANDOCLEVEL_OK;
+ op = OP_DEFAULT;
+ if (0 == treescan())
+ goto out;
+ dba = dba_new(128);
+ }
if (OP_DELETE != op)
- mpages_merge(mc, mp, 0);
- dbclose(1);
+ mpages_merge(dba, mp);
+ if (nodb == 0)
+ dbwrite(dba);
+ dba_free(dba);
} else {
/*
* If we have arguments, use them as our manpaths.
- * If we don't, grok from manpath(1) or however else
- * manpath_parse() wants to do it.
+ * If we don't, use man.conf(5).
*/
if (argc > 0) {
- dirs.paths = mandoc_calloc
- (argc, sizeof(char *));
- dirs.sz = (size_t)argc;
+ conf.manpath.paths = mandoc_reallocarray(NULL,
+ argc, sizeof(char *));
+ conf.manpath.sz = (size_t)argc;
for (i = 0; i < argc; i++)
- dirs.paths[i] = mandoc_strdup(argv[i]);
+ conf.manpath.paths[i] = mandoc_strdup(argv[i]);
} else
- manpath_parse(&dirs, path_arg, NULL, NULL);
+ manconf_parse(&conf, path_arg, NULL, NULL);
+
+ if (conf.manpath.sz == 0) {
+ exitcode = (int)MANDOCLEVEL_BADARG;
+ say("", "Empty manpath");
+ }
/*
* First scan the tree rooted at a base directory, then
* Ignore zero-length directories and strip trailing
* slashes.
*/
- for (j = 0; j < dirs.sz; j++) {
- sz = strlen(dirs.paths[j]);
- if (sz && '/' == dirs.paths[j][sz - 1])
- dirs.paths[j][--sz] = '\0';
+ for (j = 0; j < conf.manpath.sz; j++) {
+ sz = strlen(conf.manpath.paths[j]);
+ if (sz && conf.manpath.paths[j][sz - 1] == '/')
+ conf.manpath.paths[j][--sz] = '\0';
if (0 == sz)
continue;
if (j) {
- ohash_init(&mpages, 6, &mpages_info);
- ohash_init(&filenames, 6, &filename_info);
+ mandoc_ohash_init(&mpages, 6,
+ offsetof(struct mpage, inodev));
+ mandoc_ohash_init(&mlinks, 6,
+ offsetof(struct mlink, file));
}
- if (0 == set_basedir(dirs.paths[j]))
- goto out;
+ if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
+ continue;
if (0 == treescan())
- goto out;
- if (0 == set_basedir(dirs.paths[j]))
- goto out;
- if (0 == dbopen(0))
- goto out;
-
- mpages_merge(mc, mp, warnings && !use_all);
- dbclose(0);
+ continue;
+ dba = dba_new(128);
+ mpages_merge(dba, mp);
+ if (nodb == 0)
+ dbwrite(dba);
+ dba_free(dba);
- if (j + 1 < dirs.sz) {
- ohash_delete(&mpages);
- ohash_delete(&filenames);
+ if (j + 1 < conf.manpath.sz) {
mpages_free();
+ ohash_delete(&mpages);
+ ohash_delete(&mlinks);
}
}
}
out:
- set_basedir(NULL);
- manpath_free(&dirs);
- mchars_free(mc);
+ manconf_free(&conf);
mparse_free(mp);
- ohash_delete(&mpages);
- ohash_delete(&filenames);
+ mchars_free();
mpages_free();
- return(exitcode);
+ ohash_delete(&mpages);
+ ohash_delete(&mlinks);
+ return exitcode;
usage:
- fprintf(stderr, "usage: %s [-anvW] [-C file]\n"
- " %s [-anvW] dir ...\n"
- " %s [-nvW] -d dir [file ...]\n"
- " %s [-nvW] -u dir [file ...]\n"
- " %s -t file ...\n",
- progname, progname, progname,
- progname, progname);
-
- return((int)MANDOCLEVEL_BADARG);
+ progname = getprogname();
+ fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
+ " %s [-aDnpQ] [-Tutf8] dir ...\n"
+ " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
+ " %s [-Dnp] -u dir [file ...]\n"
+ " %s [-Q] -t file ...\n",
+ progname, progname, progname, progname, progname);
+
+ return (int)MANDOCLEVEL_BADARG;
+}
+
+/*
+ * To get a singly linked list in alpha order while inserting entries
+ * at the beginning, process directory entries in reverse alpha order.
+ */
+static int
+#if HAVE_FTS_COMPARE_CONST
+fts_compare(const FTSENT *const *a, const FTSENT *const *b)
+#else
+fts_compare(const FTSENT **a, const FTSENT **b)
+#endif
+{
+ return -strcmp((*a)->fts_name, (*b)->fts_name);
}
/*
* If use_all has been specified, grok all files.
* If not, sanitise paths to the following:
*
- * [./]man*[/<arch>]/<name>.<section>
+ * [./]man*[/<arch>]/<name>.<section>
* or
* [./]cat<section>[/<arch>]/<name>.0
*
- * TODO: accomodate for multi-language directories.
+ * TODO: accommodate for multi-language directories.
*/
static int
treescan(void)
{
+ char buf[PATH_MAX];
FTS *f;
FTSENT *ff;
- int dform;
- char *sec;
- const char *dsec, *arch, *cp, *path;
+ struct mlink *mlink;
+ int gzip;
+ enum form dform;
+ char *dsec, *arch, *fsec, *cp;
+ const char *path;
const char *argv[2];
argv[0] = ".";
- argv[1] = (char *)NULL;
+ argv[1] = NULL;
- /*
- * Walk through all components under the directory, using the
- * logical descent of files.
- */
- f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
- if (NULL == f) {
+ f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR,
+ fts_compare);
+ if (f == NULL) {
exitcode = (int)MANDOCLEVEL_SYSERR;
- say("", NULL);
- return(0);
+ say("", "&fts_open");
+ return 0;
}
dsec = arch = NULL;
dform = FORM_NONE;
- while (NULL != (ff = fts_read(f))) {
+ while ((ff = fts_read(f)) != NULL) {
path = ff->fts_path + 2;
+ switch (ff->fts_info) {
+
/*
- * If we're a regular file, add an mpage by using the
+ * Symbolic links require various sanity checks,
+ * then get handled just like regular files.
+ */
+ case FTS_SL:
+ if (realpath(path, buf) == NULL) {
+ if (warnings)
+ say(path, "&realpath");
+ continue;
+ }
+ if (strstr(buf, basedir) != buf
+#ifdef HOMEBREWDIR
+ && strstr(buf, HOMEBREWDIR) != buf
+#endif
+ ) {
+ if (warnings) say("",
+ "%s: outside base directory", buf);
+ continue;
+ }
+ /* Use logical inode to avoid mpages dupe. */
+ if (stat(path, ff->fts_statp) == -1) {
+ if (warnings)
+ say(path, "&stat");
+ continue;
+ }
+ /* FALLTHROUGH */
+
+ /*
+ * If we're a regular file, add an mlink by using the
* stored directory data and handling the filename.
- * Disallow duplicate (hard-linked) files.
*/
- if (FTS_F == ff->fts_info) {
- if (0 == strcmp(path, MANDOC_DB))
+ case FTS_F:
+ if ( ! strcmp(path, MANDOC_DB))
continue;
if ( ! use_all && ff->fts_level < 2) {
if (warnings)
say(path, "Extraneous file");
continue;
- } else if (inocheck(ff->fts_statp)) {
- if (warnings)
- say(path, "Duplicate file");
- continue;
- } else if (NULL == (sec =
- strrchr(ff->fts_name, '.'))) {
+ }
+ gzip = 0;
+ fsec = NULL;
+ while (fsec == NULL) {
+ fsec = strrchr(ff->fts_name, '.');
+ if (fsec == NULL || strcmp(fsec+1, "gz"))
+ break;
+ gzip = 1;
+ *fsec = '\0';
+ fsec = NULL;
+ }
+ if (fsec == NULL) {
if ( ! use_all) {
if (warnings)
say(path,
"No filename suffix");
continue;
}
- } else if (0 == strcmp(++sec, "html")) {
+ } else if ( ! strcmp(++fsec, "html")) {
if (warnings)
say(path, "Skip html");
continue;
- } else if (0 == strcmp(sec, "gz")) {
- if (warnings)
- say(path, "Skip gz");
- continue;
- } else if (0 == strcmp(sec, "ps")) {
+ } else if ( ! strcmp(fsec, "ps")) {
if (warnings)
say(path, "Skip ps");
continue;
- } else if (0 == strcmp(sec, "pdf")) {
+ } else if ( ! strcmp(fsec, "pdf")) {
if (warnings)
say(path, "Skip pdf");
continue;
} else if ( ! use_all &&
- ((FORM_SRC == dform && strcmp(sec, dsec)) ||
- (FORM_CAT == dform && strcmp(sec, "0")))) {
+ ((dform == FORM_SRC &&
+ strncmp(fsec, dsec, strlen(dsec))) ||
+ (dform == FORM_CAT && strcmp(fsec, "0")))) {
if (warnings)
say(path, "Wrong filename suffix");
continue;
} else
- sec[-1] = '\0';
- ofadd(dform, path, ff->fts_name, dsec, sec,
- arch, ff->fts_statp);
+ fsec[-1] = '\0';
+
+ mlink = mandoc_calloc(1, sizeof(struct mlink));
+ if (strlcpy(mlink->file, path,
+ sizeof(mlink->file)) >=
+ sizeof(mlink->file)) {
+ say(path, "Filename too long");
+ free(mlink);
+ continue;
+ }
+ mlink->dform = dform;
+ mlink->dsec = dsec;
+ mlink->arch = arch;
+ mlink->name = ff->fts_name;
+ mlink->fsec = fsec;
+ mlink->gzip = gzip;
+ mlink_add(mlink, ff->fts_statp);
continue;
- } else if (FTS_D != ff->fts_info &&
- FTS_DP != ff->fts_info) {
+
+ case FTS_D:
+ case FTS_DP:
+ break;
+
+ default:
if (warnings)
say(path, "Not a regular file");
continue;
}
switch (ff->fts_level) {
- case (0):
+ case 0:
/* Ignore the root directory. */
break;
- case (1):
+ case 1:
/*
* This might contain manX/ or catX/.
* Try to infer this from the name.
* If we're not in use_all, enforce it.
*/
- dsec = NULL;
- dform = FORM_NONE;
cp = ff->fts_name;
- if (FTS_DP == ff->fts_info)
+ if (ff->fts_info == FTS_DP) {
+ dform = FORM_NONE;
+ dsec = NULL;
break;
+ }
- if (0 == strncmp(cp, "man", 3)) {
+ if ( ! strncmp(cp, "man", 3)) {
dform = FORM_SRC;
dsec = cp + 3;
- } else if (0 == strncmp(cp, "cat", 3)) {
+ } else if ( ! strncmp(cp, "cat", 3)) {
dform = FORM_CAT;
dsec = cp + 3;
+ } else {
+ dform = FORM_NONE;
+ dsec = NULL;
}
- if (NULL != dsec || use_all)
+ if (dsec != NULL || use_all)
break;
if (warnings)
say(path, "Unknown directory part");
fts_set(f, ff, FTS_SKIP);
break;
- case (2):
+ case 2:
/*
* Possibly our architecture.
* If we're descending, keep tabs on it.
*/
- arch = NULL;
- if (FTS_DP != ff->fts_info && NULL != dsec)
+ if (ff->fts_info != FTS_DP && dsec != NULL)
arch = ff->fts_name;
+ else
+ arch = NULL;
break;
default:
- if (FTS_DP == ff->fts_info || use_all)
+ if (ff->fts_info == FTS_DP || use_all)
break;
if (warnings)
say(path, "Extraneous directory part");
}
fts_close(f);
- return(1);
+ return 1;
}
/*
- * Add a file to the file vector.
+ * Add a file to the mlinks table.
* Do not verify that it's a "valid" looking manpage (we'll do that
* later).
*
* Try to infer the manual section, architecture, and page name from the
* path, assuming it looks like
*
- * [./]man*[/<arch>]/<name>.<section>
+ * [./]man*[/<arch>]/<name>.<section>
* or
* [./]cat<section>[/<arch>]/<name>.0
*
- * Stuff this information directly into the mpage vector.
* See treescan() for the fts(3) version of this.
*/
static void
filescan(const char *file)
{
char buf[PATH_MAX];
- const char *sec, *arch, *name, *dsec;
- char *p, *start;
- int dform;
struct stat st;
+ struct mlink *mlink;
+ char *p, *start;
assert(use_all);
if (0 == strncmp(file, "./", 2))
file += 2;
- if (NULL == realpath(file, buf)) {
+ /*
+ * We have to do lstat(2) before realpath(3) loses
+ * the information whether this is a symbolic link.
+ * We need to know that because for symbolic links,
+ * we want to use the orginal file name, while for
+ * regular files, we want to use the real path.
+ */
+ if (-1 == lstat(file, &st)) {
exitcode = (int)MANDOCLEVEL_BADARG;
- say(file, NULL);
+ say(file, "&lstat");
return;
- } else if (OP_TEST != op && strstr(buf, basedir) != buf) {
+ } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
exitcode = (int)MANDOCLEVEL_BADARG;
- say("", "%s: outside base directory", buf);
+ say(file, "Not a regular file");
return;
- } else if (-1 == stat(buf, &st)) {
+ }
+
+ /*
+ * We have to resolve the file name to the real path
+ * in any case for the base directory check.
+ */
+ if (NULL == realpath(file, buf)) {
exitcode = (int)MANDOCLEVEL_BADARG;
- say(file, NULL);
+ say(file, "&realpath");
return;
- } else if ( ! (S_IFREG & st.st_mode)) {
+ }
+
+ if (OP_TEST == op)
+ start = buf;
+ else if (strstr(buf, basedir) == buf)
+ start = buf + strlen(basedir);
+#ifdef HOMEBREWDIR
+ else if (strstr(buf, HOMEBREWDIR) == buf)
+ start = buf;
+#endif
+ else {
exitcode = (int)MANDOCLEVEL_BADARG;
- say(file, "Not a regular file");
+ say("", "%s: outside base directory", buf);
return;
- } else if (inocheck(&st)) {
- if (warnings)
- say(file, "Duplicate file");
+ }
+
+ /*
+ * Now we are sure the file is inside our tree.
+ * If it is a symbolic link, ignore the real path
+ * and use the original name.
+ * This implies passing stuff like "cat1/../man1/foo.1"
+ * on the command line won't work. So don't do that.
+ * Note the stat(2) can still fail if the link target
+ * doesn't exist.
+ */
+ if (S_IFLNK & st.st_mode) {
+ if (-1 == stat(buf, &st)) {
+ exitcode = (int)MANDOCLEVEL_BADARG;
+ say(file, "&stat");
+ return;
+ }
+ if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
+ say(file, "Filename too long");
+ return;
+ }
+ start = buf;
+ if (OP_TEST != op && strstr(buf, basedir) == buf)
+ start += strlen(basedir);
+ }
+
+ mlink = mandoc_calloc(1, sizeof(struct mlink));
+ mlink->dform = FORM_NONE;
+ if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
+ sizeof(mlink->file)) {
+ say(start, "Filename too long");
+ free(mlink);
return;
}
- start = buf + strlen(basedir);
- sec = arch = name = dsec = NULL;
- dform = FORM_NONE;
+
+ /*
+ * In test mode or when the original name is absolute
+ * but outside our tree, guess the base directory.
+ */
+
+ if (op == OP_TEST || (start == buf && *start == '/')) {
+ if (strncmp(buf, "man/", 4) == 0)
+ start = buf + 4;
+ else if ((start = strstr(buf, "/man/")) != NULL)
+ start += 5;
+ else
+ start = buf;
+ }
/*
* First try to guess our directory structure.
if (NULL != (p = strchr(start, '/'))) {
*p++ = '\0';
if (0 == strncmp(start, "man", 3)) {
- dform = FORM_SRC;
- dsec = start + 3;
+ mlink->dform = FORM_SRC;
+ mlink->dsec = start + 3;
} else if (0 == strncmp(start, "cat", 3)) {
- dform = FORM_CAT;
- dsec = start + 3;
+ mlink->dform = FORM_CAT;
+ mlink->dsec = start + 3;
}
start = p;
- if (NULL != dsec && NULL != (p = strchr(start, '/'))) {
+ if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
*p++ = '\0';
- arch = start;
+ mlink->arch = start;
start = p;
- }
+ }
}
/*
if ('.' == *p) {
*p++ = '\0';
- sec = p;
+ mlink->fsec = p;
}
/*
* Now try to parse the name.
* Use the filename portion of the path.
*/
- name = start;
+ mlink->name = start;
if (NULL != (p = strrchr(start, '/'))) {
- name = p + 1;
+ mlink->name = p + 1;
*p = '\0';
- }
-
- ofadd(dform, file, name, dsec, sec, arch, &st);
+ }
+ mlink_add(mlink, &st);
}
-/*
- * See fileadd().
- */
-static int
-filecheck(const char *name)
+static void
+mlink_add(struct mlink *mlink, const struct stat *st)
{
+ struct inodev inodev;
+ struct mpage *mpage;
+ unsigned int slot;
+
+ assert(NULL != mlink->file);
+
+ mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
+ mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
+ mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
+ mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
- return(NULL != ohash_find(&filenames,
- ohash_qlookup(&filenames, name)));
+ if ('0' == *mlink->fsec) {
+ free(mlink->fsec);
+ mlink->fsec = mandoc_strdup(mlink->dsec);
+ mlink->fform = FORM_CAT;
+ } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
+ mlink->fform = FORM_SRC;
+ else
+ mlink->fform = FORM_NONE;
+
+ slot = ohash_qlookup(&mlinks, mlink->file);
+ assert(NULL == ohash_find(&mlinks, slot));
+ ohash_insert(&mlinks, slot, mlink);
+
+ memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
+ inodev.st_ino = st->st_ino;
+ inodev.st_dev = st->st_dev;
+ slot = ohash_lookup_memory(&mpages, (char *)&inodev,
+ sizeof(struct inodev), inodev.st_ino);
+ mpage = ohash_find(&mpages, slot);
+ if (NULL == mpage) {
+ mpage = mandoc_calloc(1, sizeof(struct mpage));
+ mpage->inodev.st_ino = inodev.st_ino;
+ mpage->inodev.st_dev = inodev.st_dev;
+ mpage->form = FORM_NONE;
+ mpage->next = mpage_head;
+ mpage_head = mpage;
+ ohash_insert(&mpages, slot, mpage);
+ } else
+ mlink->next = mpage->mlinks;
+ mpage->mlinks = mlink;
+ mlink->mpage = mpage;
}
-/*
- * Use the standard hashing mechanism (K&R) to see if the given filename
- * already exists.
- */
static void
-fileadd(struct mpage *mpage)
+mlink_free(struct mlink *mlink)
{
- unsigned int slot;
- slot = ohash_qlookup(&filenames, mpage->file);
- assert(NULL == ohash_find(&filenames, slot));
- ohash_insert(&filenames, slot, mpage);
+ free(mlink->dsec);
+ free(mlink->arch);
+ free(mlink->name);
+ free(mlink->fsec);
+ free(mlink);
}
-/*
- * See inoadd().
- */
-static int
-inocheck(const struct stat *st)
+static void
+mpages_free(void)
{
- struct inodev inodev;
- uint32_t hash;
-
- memset(&inodev, 0, sizeof(inodev));
- inodev.st_ino = hash = st->st_ino;
- inodev.st_dev = st->st_dev;
+ struct mpage *mpage;
+ struct mlink *mlink;
- return(NULL != ohash_find(&mpages, ohash_lookup_memory(
- &mpages, (char *)&inodev, sizeof(inodev), hash)));
+ while ((mpage = mpage_head) != NULL) {
+ while ((mlink = mpage->mlinks) != NULL) {
+ mpage->mlinks = mlink->next;
+ mlink_free(mlink);
+ }
+ mpage_head = mpage->next;
+ free(mpage->sec);
+ free(mpage->arch);
+ free(mpage->title);
+ free(mpage->desc);
+ free(mpage);
+ }
}
/*
- * The hashing function used here is quite simple: simply take the inode
- * and use uint32_t of its bits.
- * Then when we do the lookup, use both the inode and device identifier.
+ * For each mlink to the mpage, check whether the path looks like
+ * it is formatted, and if it does, check whether a source manual
+ * exists by the same name, ignoring the suffix.
+ * If both conditions hold, drop the mlink.
*/
static void
-inoadd(const struct stat *st, struct mpage *mpage)
+mlinks_undupe(struct mpage *mpage)
{
- uint32_t hash;
- unsigned int slot;
-
- mpage->inodev.st_ino = hash = st->st_ino;
- mpage->inodev.st_dev = st->st_dev;
- slot = ohash_lookup_memory(&mpages,
- (char *)&mpage->inodev, sizeof(mpage->inodev), hash);
-
- assert(NULL == ohash_find(&mpages, slot));
- ohash_insert(&mpages, slot, mpage);
+ char buf[PATH_MAX];
+ struct mlink **prev;
+ struct mlink *mlink;
+ char *bufp;
+
+ mpage->form = FORM_CAT;
+ prev = &mpage->mlinks;
+ while (NULL != (mlink = *prev)) {
+ if (FORM_CAT != mlink->dform) {
+ mpage->form = FORM_NONE;
+ goto nextlink;
+ }
+ (void)strlcpy(buf, mlink->file, sizeof(buf));
+ bufp = strstr(buf, "cat");
+ assert(NULL != bufp);
+ memcpy(bufp, "man", 3);
+ if (NULL != (bufp = strrchr(buf, '.')))
+ *++bufp = '\0';
+ (void)strlcat(buf, mlink->dsec, sizeof(buf));
+ if (NULL == ohash_find(&mlinks,
+ ohash_qlookup(&mlinks, buf)))
+ goto nextlink;
+ if (warnings)
+ say(mlink->file, "Man source exists: %s", buf);
+ if (use_all)
+ goto nextlink;
+ *prev = mlink->next;
+ mlink_free(mlink);
+ continue;
+nextlink:
+ prev = &(*prev)->next;
+ }
}
static void
-ofadd(int dform, const char *file, const char *name, const char *dsec,
- const char *sec, const char *arch, const struct stat *st)
+mlink_check(struct mpage *mpage, struct mlink *mlink)
{
- struct mpage *mpage;
- int sform;
-
- assert(NULL != file);
-
- if (NULL == name)
- name = "";
- if (NULL == sec)
- sec = "";
- if (NULL == dsec)
- dsec = "";
- if (NULL == arch)
- arch = "";
-
- if ('0' == *sec) {
- sec = dsec;
- sform = FORM_CAT;
- } else if ('1' <= *sec && '9' >= *sec)
- sform = FORM_SRC;
- else
- sform = FORM_NONE;
-
- mpage = mandoc_calloc(1, sizeof(struct mpage));
- strlcpy(mpage->file, file, PATH_MAX);
- mpage->name = mandoc_strdup(name);
- mpage->sec = mandoc_strdup(sec);
- mpage->dsec = mandoc_strdup(dsec);
- mpage->arch = mandoc_strdup(arch);
- mpage->sform = sform;
- mpage->dform = dform;
- mpage->next = mpages_list;
- mpages_list = mpage;
+ struct str *str;
+ unsigned int slot;
/*
- * Add to unique identifier hash.
- * Then if it's a source manual and we're going to use source in
- * favour of catpages, add it to that hash.
+ * Check whether the manual section given in a file
+ * agrees with the directory where the file is located.
+ * Some manuals have suffixes like (3p) on their
+ * section number either inside the file or in the
+ * directory name, some are linked into more than one
+ * section, like encrypt(1) = makekey(8).
*/
- inoadd(st, mpage);
- fileadd(mpage);
-}
-static void
-mpages_free(void)
-{
- struct mpage *mpage;
+ if (FORM_SRC == mpage->form &&
+ strcasecmp(mpage->sec, mlink->dsec))
+ say(mlink->file, "Section \"%s\" manual in %s directory",
+ mpage->sec, mlink->dsec);
- while (NULL != (mpage = mpages_list)) {
- mpages_list = mpage->next;
- free(mpage->name);
- free(mpage->sec);
- free(mpage->dsec);
- free(mpage->arch);
- free(mpage);
- }
+ /*
+ * Manual page directories exist for each kernel
+ * architecture as returned by machine(1).
+ * However, many manuals only depend on the
+ * application architecture as returned by arch(1).
+ * For example, some (2/ARM) manuals are shared
+ * across the "armish" and "zaurus" kernel
+ * architectures.
+ * A few manuals are even shared across completely
+ * different architectures, for example fdformat(1)
+ * on amd64, i386, and sparc64.
+ */
+
+ if (strcasecmp(mpage->arch, mlink->arch))
+ say(mlink->file, "Architecture \"%s\" manual in "
+ "\"%s\" directory", mpage->arch, mlink->arch);
+
+ /*
+ * XXX
+ * parse_cat() doesn't set NAME_TITLE yet.
+ */
+
+ if (FORM_CAT == mpage->form)
+ return;
+
+ /*
+ * Check whether this mlink
+ * appears as a name in the NAME section.
+ */
+
+ slot = ohash_qlookup(&names, mlink->name);
+ str = ohash_find(&names, slot);
+ assert(NULL != str);
+ if ( ! (NAME_TITLE & str->mask))
+ say(mlink->file, "Name missing in NAME section");
}
/*
- * Run through the files in the global vector "mpages_list" and add them to the
- * database specified in "basedir".
+ * Run through the files in the global vector "mpages"
+ * and add them to the database specified in "basedir".
*
* This handles the parsing scheme itself, using the cues of directory
* and filename to determine whether the file is parsable or not.
*/
static void
-mpages_merge(struct mchars *mc, struct mparse *mp, int check_reachable)
+mpages_merge(struct dba *dba, struct mparse *mp)
{
- struct ohash title_table;
- struct ohash_info title_info, str_info;
- char buf[PATH_MAX];
- struct mpage *mpage;
- struct mdoc *mdoc;
- struct man *man;
- struct title *title_entry;
- char *bufp, *title_str;
- const char *msec, *march, *mtitle, *cp;
- size_t sz;
- int form;
- int match;
- unsigned int slot;
- enum mandoclevel lvl;
-
- str_info.alloc = hash_alloc;
- str_info.halloc = hash_halloc;
- str_info.hfree = hash_free;
- str_info.key_offset = offsetof(struct str, key);
-
- if (check_reachable) {
- title_info.alloc = hash_alloc;
- title_info.halloc = hash_halloc;
- title_info.hfree = hash_free;
- title_info.key_offset = offsetof(struct title, title);
- ohash_init(&title_table, 6, &title_info);
- }
-
- for (mpage = mpages_list; NULL != mpage; mpage = mpage->next) {
- /*
- * If we're a catpage (as defined by our path), then see
- * if a manpage exists by the same name (ignoring the
- * suffix).
- * If it does, then we want to use it instead of our
- * own.
- */
- if ( ! use_all && FORM_CAT == mpage->dform) {
- sz = strlcpy(buf, mpage->file, PATH_MAX);
- if (sz >= PATH_MAX) {
- if (warnings)
- say(mpage->file,
- "Filename too long");
- continue;
- }
- bufp = strstr(buf, "cat");
- assert(NULL != bufp);
- memcpy(bufp, "man", 3);
- if (NULL != (bufp = strrchr(buf, '.')))
- *++bufp = '\0';
- strlcat(buf, mpage->dsec, PATH_MAX);
- if (filecheck(buf)) {
- if (warnings)
- say(mpage->file, "Man "
- "source exists: %s", buf);
- continue;
- }
- }
+ struct mpage *mpage, *mpage_dest;
+ struct mlink *mlink, *mlink_dest;
+ struct roff_man *man;
+ char *sodest;
+ char *cp;
+ int fd;
+
+ for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
+ mlinks_undupe(mpage);
+ if ((mlink = mpage->mlinks) == NULL)
+ continue;
- ohash_init(&strings, 6, &str_info);
+ name_mask = NAME_MASK;
+ mandoc_ohash_init(&names, 4, offsetof(struct str, key));
+ mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
mparse_reset(mp);
- mdoc = NULL;
man = NULL;
- match = 1;
+ sodest = NULL;
+
+ if ((fd = mparse_open(mp, mlink->file)) == -1) {
+ say(mlink->file, "&open");
+ goto nextpage;
+ }
/*
- * Try interpreting the file as mdoc(7) or man(7)
- * source code, unless it is already known to be
- * formatted. Fall back to formatted mode.
+ * Interpret the file as mdoc(7) or man(7) source
+ * code, unless it is known to be formatted.
*/
- if (FORM_CAT != mpage->dform || FORM_CAT != mpage->sform) {
- lvl = mparse_readfd(mp, -1, mpage->file);
- if (lvl < MANDOCLEVEL_FATAL)
- mparse_result(mp, &mdoc, &man);
+ if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
+ mparse_readfd(mp, fd, mlink->file);
+ close(fd);
+ fd = -1;
+ mparse_result(mp, &man, &sodest);
}
- if (NULL != mdoc) {
- form = 1;
- msec = mdoc_meta(mdoc)->msec;
- march = mdoc_meta(mdoc)->arch;
- mtitle = mdoc_meta(mdoc)->title;
- } else if (NULL != man) {
- form = 1;
- msec = man_meta(man)->msec;
- march = mpage->arch;
- mtitle = man_meta(man)->title;
- } else {
- form = 0;
- msec = mpage->dsec;
- march = mpage->arch;
- mtitle = mpage->name;
- }
+ if (sodest != NULL) {
+ mlink_dest = ohash_find(&mlinks,
+ ohash_qlookup(&mlinks, sodest));
+ if (mlink_dest == NULL) {
+ mandoc_asprintf(&cp, "%s.gz", sodest);
+ mlink_dest = ohash_find(&mlinks,
+ ohash_qlookup(&mlinks, cp));
+ free(cp);
+ }
+ if (mlink_dest != NULL) {
- if (NULL == msec)
- msec = "";
- if (NULL == march)
- march = "";
- if (NULL == mtitle)
- mtitle = "";
+ /* The .so target exists. */
- /*
- * Check whether the manual section given in a file
- * agrees with the directory where the file is located.
- * Some manuals have suffixes like (3p) on their
- * section number either inside the file or in the
- * directory name, some are linked into more than one
- * section, like encrypt(1) = makekey(8). Do not skip
- * manuals for such reasons.
- */
- if (warnings && !use_all && form &&
- strcasecmp(msec, mpage->dsec)) {
- match = 0;
- say(mpage->file, "Section \"%s\" "
- "manual in %s directory",
- msec, mpage->dsec);
- }
+ mpage_dest = mlink_dest->mpage;
+ while (1) {
+ mlink->mpage = mpage_dest;
- /*
- * Manual page directories exist for each kernel
- * architecture as returned by machine(1).
- * However, many manuals only depend on the
- * application architecture as returned by arch(1).
- * For example, some (2/ARM) manuals are shared
- * across the "armish" and "zaurus" kernel
- * architectures.
- * A few manuals are even shared across completely
- * different architectures, for example fdformat(1)
- * on amd64, i386, sparc, and sparc64.
- * Thus, warn about architecture mismatches,
- * but don't skip manuals for this reason.
- */
- if (warnings && !use_all && strcasecmp(march, mpage->arch)) {
- match = 0;
- say(mpage->file, "Architecture \"%s\" "
- "manual in \"%s\" directory",
- march, mpage->arch);
- }
- if (warnings && !use_all && strcasecmp(mtitle, mpage->name))
- match = 0;
-
- putkey(mpage, mpage->name, TYPE_Nm);
-
- if (NULL != mdoc) {
- if (NULL != (cp = mdoc_meta(mdoc)->name))
- putkey(mpage, cp, TYPE_Nm);
- assert(NULL == mpage->desc);
- parse_mdoc(mpage, mdoc_node(mdoc));
- putkey(mpage, NULL != mpage->desc ?
- mpage->desc : mpage->name, TYPE_Nd);
- } else if (NULL != man)
- parse_man(mpage, man_node(man));
- else
- parse_cat(mpage);
+ /*
+ * If the target was already
+ * processed, add the links
+ * to the database now.
+ * Otherwise, this will
+ * happen when we come
+ * to the target.
+ */
- /*
- * Build a title string for the file. If it matches
- * the location of the file, remember the title as
- * found; else, remember it as missing.
- */
+ if (mpage_dest->dba != NULL)
+ dbadd_mlink(mlink);
- if (check_reachable) {
- if (-1 == asprintf(&title_str, "%s(%s%s%s)", mtitle,
- msec, '\0' == *march ? "" : "/", march)) {
- perror(NULL);
- exit((int)MANDOCLEVEL_SYSERR);
- }
- slot = ohash_qlookup(&title_table, title_str);
- title_entry = ohash_find(&title_table, slot);
- if (NULL == title_entry) {
- title_entry = mandoc_malloc(
- sizeof(struct title));
- title_entry->title = title_str;
- title_entry->file = mandoc_strdup(
- match ? "" : mpage->file);
- ohash_insert(&title_table, slot,
- title_entry);
- } else {
- if (match)
- *title_entry->file = '\0';
- free(title_str);
+ if (mlink->next == NULL)
+ break;
+ mlink = mlink->next;
+ }
+
+ /* Move all links to the target. */
+
+ mlink->next = mlink_dest->next;
+ mlink_dest->next = mpage->mlinks;
+ mpage->mlinks = NULL;
}
+ goto nextpage;
+ } else if (man != NULL && man->macroset == MACROSET_MDOC) {
+ mdoc_validate(man);
+ mpage->form = FORM_SRC;
+ mpage->sec = man->meta.msec;
+ mpage->sec = mandoc_strdup(
+ mpage->sec == NULL ? "" : mpage->sec);
+ mpage->arch = man->meta.arch;
+ mpage->arch = mandoc_strdup(
+ mpage->arch == NULL ? "" : mpage->arch);
+ mpage->title = mandoc_strdup(man->meta.title);
+ } else if (man != NULL && man->macroset == MACROSET_MAN) {
+ man_validate(man);
+ if (*man->meta.msec != '\0' ||
+ *man->meta.title != '\0') {
+ mpage->form = FORM_SRC;
+ mpage->sec = mandoc_strdup(man->meta.msec);
+ mpage->arch = mandoc_strdup(mlink->arch);
+ mpage->title = mandoc_strdup(man->meta.title);
+ } else
+ man = NULL;
}
- dbindex(mc, form, mpage);
- ohash_delete(&strings);
- }
+ assert(mpage->desc == NULL);
+ if (man == NULL) {
+ mpage->form = FORM_CAT;
+ mpage->sec = mandoc_strdup(mlink->dsec);
+ mpage->arch = mandoc_strdup(mlink->arch);
+ mpage->title = mandoc_strdup(mlink->name);
+ parse_cat(mpage, fd);
+ } else if (man->macroset == MACROSET_MDOC)
+ parse_mdoc(mpage, &man->meta, man->first);
+ else
+ parse_man(mpage, &man->meta, man->first);
+ if (mpage->desc == NULL) {
+ mpage->desc = mandoc_strdup(mlink->name);
+ if (warnings)
+ say(mlink->file, "No one-line description, "
+ "using filename \"%s\"", mlink->name);
+ }
- if (check_reachable) {
- title_entry = ohash_first(&title_table, &slot);
- while (NULL != title_entry) {
- if ('\0' != *title_entry->file)
- say(title_entry->file,
- "Probably unreachable, title is %s",
- title_entry->title);
- free(title_entry->title);
- free(title_entry->file);
- free(title_entry);
- title_entry = ohash_next(&title_table, &slot);
+ for (mlink = mpage->mlinks;
+ mlink != NULL;
+ mlink = mlink->next) {
+ putkey(mpage, mlink->name, NAME_FILE);
+ if (warnings && !use_all)
+ mlink_check(mpage, mlink);
}
- ohash_delete(&title_table);
+
+ dbadd(dba, mpage);
+
+nextpage:
+ ohash_delete(&strings);
+ ohash_delete(&names);
}
}
static void
-parse_cat(struct mpage *mpage)
+parse_cat(struct mpage *mpage, int fd)
{
FILE *stream;
- char *line, *p, *title;
- size_t len, plen, titlesz;
-
- if (NULL == (stream = fopen(mpage->file, "r"))) {
+ struct mlink *mlink;
+ char *line, *p, *title, *sec;
+ size_t linesz, plen, titlesz;
+ ssize_t len;
+ int offs;
+
+ mlink = mpage->mlinks;
+ stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
+ if (stream == NULL) {
+ if (fd != -1)
+ close(fd);
if (warnings)
- say(mpage->file, NULL);
+ say(mlink->file, "&fopen");
return;
}
+ line = NULL;
+ linesz = 0;
+
+ /* Parse the section number from the header line. */
+
+ while (getline(&line, &linesz, stream) != -1) {
+ if (*line == '\n')
+ continue;
+ if ((sec = strchr(line, '(')) == NULL)
+ break;
+ if ((p = strchr(++sec, ')')) == NULL)
+ break;
+ free(mpage->sec);
+ mpage->sec = mandoc_strndup(sec, p - sec);
+ if (warnings && *mlink->dsec != '\0' &&
+ strcasecmp(mpage->sec, mlink->dsec))
+ say(mlink->file,
+ "Section \"%s\" manual in %s directory",
+ mpage->sec, mlink->dsec);
+ break;
+ }
+
/* Skip to first blank line. */
- while (NULL != (line = fgetln(stream, &len)))
- if ('\n' == *line)
+ while (line == NULL || *line != '\n')
+ if (getline(&line, &linesz, stream) == -1)
break;
/*
* is the first section header. Skip to it.
*/
- while (NULL != (line = fgetln(stream, &len)))
- if ('\n' != *line && ' ' != *line)
+ while (getline(&line, &linesz, stream) != -1)
+ if (*line != '\n' && *line != ' ')
break;
-
+
/*
* Read up until the next section into a buffer.
* Strip the leading and trailing newline from each read line,
titlesz = 0;
title = NULL;
- while (NULL != (line = fgetln(stream, &len))) {
- if (' ' != *line || '\n' != line[len - 1])
+ while ((len = getline(&line, &linesz, stream)) != -1) {
+ if (*line != ' ')
break;
- while (len > 0 && isspace((unsigned char)*line)) {
- line++;
- len--;
- }
- if (1 == len)
+ offs = 0;
+ while (isspace((unsigned char)line[offs]))
+ offs++;
+ if (line[offs] == '\0')
continue;
- title = mandoc_realloc(title, titlesz + len);
- memcpy(title + titlesz, line, len);
- titlesz += len;
+ title = mandoc_realloc(title, titlesz + len - offs);
+ memcpy(title + titlesz, line + offs, len - offs);
+ titlesz += len - offs;
title[titlesz - 1] = ' ';
}
+ free(line);
/*
* If no page content can be found, or the input line
if (NULL == title || '\0' == *title) {
if (warnings)
- say(mpage->file, "Cannot find NAME section");
- assert(NULL == mpage->desc);
- mpage->desc = mandoc_strdup(mpage->name);
- putkey(mpage, mpage->name, TYPE_Nd);
+ say(mlink->file, "Cannot find NAME section");
fclose(stream);
free(title);
return;
}
- title = mandoc_realloc(title, titlesz + 1);
- title[titlesz] = '\0';
+ title[titlesz - 1] = '\0';
/*
* Skip to the first dash.
/* Skip to next word. */ ;
} else {
if (warnings)
- say(mpage->file, "No dash in title line");
+ say(mlink->file, "No dash in title line, "
+ "reusing \"%s\" as one-line description", title);
p = title;
}
if (0 == len) {
memmove(line, line + 1, plen--);
continue;
- }
+ }
memmove(line - 1, line + 1, plen - len);
plen -= 2;
}
- assert(NULL == mpage->desc);
mpage->desc = mandoc_strdup(p);
- putkey(mpage, mpage->desc, TYPE_Nd);
fclose(stream);
free(title);
}
* Put a type/word pair into the word database for this particular file.
*/
static void
-putkey(const struct mpage *mpage, const char *value, uint64_t type)
+putkey(const struct mpage *mpage, char *value, uint64_t type)
{
-
- assert(NULL != value);
putkeys(mpage, value, strlen(value), type);
}
*/
static void
putmdockey(const struct mpage *mpage,
- const struct mdoc_node *n, uint64_t m)
+ const struct roff_node *n, uint64_t m, int taboo)
{
for ( ; NULL != n; n = n->next) {
+ if (n->flags & taboo)
+ continue;
if (NULL != n->child)
- putmdockey(mpage, n->child, m);
- if (MDOC_TEXT == n->type)
+ putmdockey(mpage, n->child, m, taboo);
+ if (n->type == ROFFT_TEXT)
putkey(mpage, n->string, m);
}
}
static void
-parse_man(struct mpage *mpage, const struct man_node *n)
+parse_man(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
{
- const struct man_node *head, *body;
- char *start, *sv, *title;
+ const struct roff_node *head, *body;
+ char *start, *title;
char byte;
- size_t sz, titlesz;
+ size_t sz;
- if (NULL == n)
+ if (n == NULL)
return;
/*
* the correct section or not.
*/
- if (MAN_BODY == n->type && MAN_SH == n->tok) {
+ if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
body = n;
- assert(body->parent);
- if (NULL != (head = body->parent->head) &&
- 1 == head->nchild &&
- NULL != (head = (head->child)) &&
- MAN_TEXT == head->type &&
- 0 == strcmp(head->string, "NAME") &&
- NULL != (body = body->child) &&
- MAN_TEXT == body->type) {
-
- title = NULL;
- titlesz = 0;
+ if ((head = body->parent->head) != NULL &&
+ (head = head->child) != NULL &&
+ head->next == NULL &&
+ head->type == ROFFT_TEXT &&
+ strcmp(head->string, "NAME") == 0 &&
+ body->child != NULL) {
/*
* Suck the entire NAME section into memory.
* NAME sections over many lines.
*/
- for ( ; NULL != body; body = body->next) {
- if (MAN_TEXT != body->type)
- break;
- if (0 == (sz = strlen(body->string)))
- continue;
- title = mandoc_realloc
- (title, titlesz + sz + 1);
- memcpy(title + titlesz, body->string, sz);
- titlesz += sz + 1;
- title[titlesz - 1] = ' ';
- }
+ title = NULL;
+ deroff(&title, body);
if (NULL == title)
return;
- title = mandoc_realloc(title, titlesz + 1);
- title[titlesz] = '\0';
-
- /* Skip leading space. */
-
- sv = title;
- while (isspace((unsigned char)*sv))
- sv++;
-
- if (0 == (sz = strlen(sv))) {
- free(title);
- return;
- }
-
- /* Erase trailing space. */
-
- start = &sv[sz - 1];
- while (start > sv && isspace((unsigned char)*start))
- *start-- = '\0';
-
- if (start == sv) {
- free(title);
- return;
- }
-
- start = sv;
-
- /*
+ /*
* Go through a special heuristic dance here.
* Conventionally, one or more manual names are
* comma-specified prior to a whitespace, then a
* the name parts here.
*/
+ start = title;
for ( ;; ) {
sz = strcspn(start, " ,");
if ('\0' == start[sz])
byte = start[sz];
start[sz] = '\0';
- putkey(mpage, start, TYPE_Nm);
+ /*
+ * Assume a stray trailing comma in the
+ * name list if a name begins with a dash.
+ */
+
+ if ('-' == start[0] ||
+ ('\\' == start[0] && '-' == start[1]))
+ break;
+
+ putkey(mpage, start, NAME_TITLE);
+ if ( ! (mpage->name_head_done ||
+ strcasecmp(start, meta->title))) {
+ putkey(mpage, start, NAME_HEAD);
+ mpage->name_head_done = 1;
+ }
if (' ' == byte) {
start += sz + 1;
start++;
}
- if (sv == start) {
- putkey(mpage, start, TYPE_Nm);
+ if (start == title) {
+ putkey(mpage, start, NAME_TITLE);
+ if ( ! (mpage->name_head_done ||
+ strcasecmp(start, meta->title))) {
+ putkey(mpage, start, NAME_HEAD);
+ mpage->name_head_done = 1;
+ }
free(title);
return;
}
while (' ' == *start)
start++;
- assert(NULL == mpage->desc);
mpage->desc = mandoc_strdup(start);
- putkey(mpage, mpage->desc, TYPE_Nd);
free(title);
return;
}
for (n = n->child; n; n = n->next) {
if (NULL != mpage->desc)
break;
- parse_man(mpage, n);
+ parse_man(mpage, meta, n);
}
}
static void
-parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
+parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
{
- assert(NULL != n);
- for (n = n->child; NULL != n; n = n->next) {
+ for (n = n->child; n != NULL; n = n->next) {
+ if (n->tok == TOKEN_NONE ||
+ n->tok < ROFF_MAX ||
+ n->flags & mdocs[n->tok].taboo)
+ continue;
+ assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
switch (n->type) {
- case (MDOC_ELEM):
- /* FALLTHROUGH */
- case (MDOC_BLOCK):
- /* FALLTHROUGH */
- case (MDOC_HEAD):
- /* FALLTHROUGH */
- case (MDOC_BODY):
- /* FALLTHROUGH */
- case (MDOC_TAIL):
- if (NULL != mdocs[n->tok].fp)
- if (0 == (*mdocs[n->tok].fp)(mpage, n))
- break;
+ case ROFFT_ELEM:
+ case ROFFT_BLOCK:
+ case ROFFT_HEAD:
+ case ROFFT_BODY:
+ case ROFFT_TAIL:
+ if (mdocs[n->tok].fp != NULL &&
+ (*mdocs[n->tok].fp)(mpage, meta, n) == 0)
+ break;
if (mdocs[n->tok].mask)
putmdockey(mpage, n->child,
- mdocs[n->tok].mask);
+ mdocs[n->tok].mask, mdocs[n->tok].taboo);
break;
default:
- assert(MDOC_ROOT != n->type);
continue;
}
if (NULL != n->child)
- parse_mdoc(mpage, n);
+ parse_mdoc(mpage, meta, n);
}
}
static int
-parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
+parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
{
- const char *start, *end;
+ char *start, *end;
size_t sz;
if (SEC_SYNOPSIS != n->sec ||
- NULL == (n = n->child) ||
- MDOC_TEXT != n->type)
- return(0);
+ NULL == (n = n->child) ||
+ n->type != ROFFT_TEXT)
+ return 0;
/*
* Only consider those `Fd' macro fields that begin with an
*/
if (strcmp("#include", n->string))
- return(0);
+ return 0;
- if (NULL == (n = n->next) || MDOC_TEXT != n->type)
- return(0);
+ if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
+ return 0;
/*
* Strip away the enclosing angle brackets and make sure we're
start++;
if (0 == (sz = strlen(start)))
- return(0);
+ return 0;
end = &start[(int)sz - 1];
if ('>' == *end || '"' == *end)
if (end > start)
putkeys(mpage, start, end - start + 1, TYPE_In);
- return(1);
+ return 0;
}
-static int
-parse_mdoc_In(struct mpage *mpage, const struct mdoc_node *n)
+static void
+parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
{
+ char *cp;
+ size_t sz;
- if (NULL != n->child && MDOC_TEXT == n->child->type)
- return(0);
+ if (n->type != ROFFT_TEXT)
+ return;
+
+ /* Skip function pointer punctuation. */
+
+ cp = n->string;
+ while (*cp == '(' || *cp == '*')
+ cp++;
+ sz = strcspn(cp, "()");
- putkey(mpage, n->child->string, TYPE_In);
- return(1);
+ putkeys(mpage, cp, sz, TYPE_Fn);
+ if (n->sec == SEC_SYNOPSIS)
+ putkeys(mpage, cp, sz, NAME_SYN);
}
static int
-parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
+parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
{
- const char *cp;
- if (NULL == (n = n->child) || MDOC_TEXT != n->type)
- return(0);
+ if (n->child == NULL)
+ return 0;
- /*
- * Parse: .Fn "struct type *name" "char *arg".
- * First strip away pointer symbol.
- * Then store the function name, then type.
- * Finally, store the arguments.
- */
+ parse_mdoc_fname(mpage, n->child);
- if (NULL == (cp = strrchr(n->string, ' ')))
- cp = n->string;
+ for (n = n->child->next; n != NULL; n = n->next)
+ if (n->type == ROFFT_TEXT)
+ putkey(mpage, n->string, TYPE_Fa);
- while ('*' == *cp)
- cp++;
+ return 0;
+}
- putkey(mpage, cp, TYPE_Fn);
+static int
+parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
+{
- if (n->string < cp)
- putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
+ if (n->type != ROFFT_HEAD)
+ return 1;
- for (n = n->next; NULL != n; n = n->next)
- if (MDOC_TEXT == n->type)
- putkey(mpage, n->string, TYPE_Fa);
+ if (n->child != NULL)
+ parse_mdoc_fname(mpage, n->child);
- return(0);
+ return 0;
}
static int
-parse_mdoc_St(struct mpage *mpage, const struct mdoc_node *n)
+parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
{
+ char *cp;
+
+ if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
+ return 0;
+
+ if (n->child != NULL &&
+ n->child->next == NULL &&
+ n->child->type == ROFFT_TEXT)
+ return 1;
+
+ cp = NULL;
+ deroff(&cp, n);
+ if (cp != NULL) {
+ putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
+ n->type == ROFFT_BODY ? TYPE_Va : 0));
+ free(cp);
+ }
- if (NULL == n->child || MDOC_TEXT != n->child->type)
- return(0);
-
- putkey(mpage, n->child->string, TYPE_St);
- return(1);
+ return 0;
}
static int
-parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
+parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
{
char *cp;
if (NULL == (n = n->child))
- return(0);
+ return 0;
if (NULL == n->next) {
putkey(mpage, n->string, TYPE_Xr);
- return(0);
+ return 0;
}
- if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) {
- perror(NULL);
- exit((int)MANDOCLEVEL_SYSERR);
- }
+ mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
putkey(mpage, cp, TYPE_Xr);
free(cp);
- return(0);
+ return 0;
}
static int
-parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
+parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
{
- size_t sz;
-
- if (MDOC_BODY != n->type)
- return(0);
- /*
- * Special-case the `Nd' because we need to put the description
- * into the document table.
- */
-
- for (n = n->child; NULL != n; n = n->next) {
- if (MDOC_TEXT == n->type) {
- if (NULL != mpage->desc) {
- sz = strlen(mpage->desc) +
- strlen(n->string) + 2;
- mpage->desc = mandoc_realloc(
- mpage->desc, sz);
- strlcat(mpage->desc, " ", sz);
- strlcat(mpage->desc, n->string, sz);
- } else
- mpage->desc = mandoc_strdup(n->string);
- }
- if (NULL != n->child)
- parse_mdoc_Nd(mpage, n);
- }
- return(1);
+ if (n->type == ROFFT_BODY)
+ deroff(&mpage->desc, n);
+ return 0;
}
static int
-parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
+parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
{
if (SEC_NAME == n->sec)
- return(1);
- else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type)
- return(0);
-
- return(1);
-}
-
-static int
-parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
-{
-
- return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
+ putmdockey(mpage, n->child, NAME_TITLE, 0);
+ else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
+ if (n->child == NULL)
+ putkey(mpage, meta->name, NAME_SYN);
+ else
+ putmdockey(mpage, n->child, NAME_SYN, 0);
+ }
+ if ( ! (mpage->name_head_done ||
+ n->child == NULL || n->child->string == NULL ||
+ strcasecmp(n->child->string, meta->title))) {
+ putkey(mpage, n->child->string, NAME_HEAD);
+ mpage->name_head_done = 1;
+ }
+ return 0;
}
static int
-parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
+parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
{
- return(MDOC_HEAD == n->type);
+ return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
}
static int
-parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
+parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
+ const struct roff_node *n)
{
- return(MDOC_BODY == n->type);
+ return n->type == ROFFT_HEAD;
}
/*
* When we finish the manual, we'll dump the table.
*/
static void
-putkeys(const struct mpage *mpage,
- const char *cp, size_t sz, uint64_t v)
+putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
{
+ struct ohash *htab;
struct str *s;
- unsigned int slot;
const char *end;
+ unsigned int slot;
+ int i, mustfree;
if (0 == sz)
return;
+ mustfree = render_string(&cp, &sz);
+
+ if (TYPE_Nm & v) {
+ htab = &names;
+ v &= name_mask;
+ if (v & NAME_FIRST)
+ name_mask &= ~NAME_FIRST;
+ if (debug > 1)
+ say(mpage->mlinks->file,
+ "Adding name %*s, bits=0x%llx", (int)sz, cp,
+ (unsigned long long)v);
+ } else {
+ htab = &strings;
+ if (debug > 1)
+ for (i = 0; i < KEY_MAX; i++)
+ if ((uint64_t)1 << i & v)
+ say(mpage->mlinks->file,
+ "Adding key %s=%*s",
+ mansearch_keynames[i], (int)sz, cp);
+ }
+
end = cp + sz;
- slot = ohash_qlookupi(&strings, cp, &end);
- s = ohash_find(&strings, slot);
+ slot = ohash_qlookupi(htab, cp, &end);
+ s = ohash_find(htab, slot);
if (NULL != s && mpage == s->mpage) {
s->mask |= v;
return;
} else if (NULL == s) {
- s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
+ s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
memcpy(s->key, cp, sz);
- ohash_insert(&strings, slot, s);
+ ohash_insert(htab, slot, s);
}
s->mpage = mpage;
s->mask = v;
+
+ if (mustfree)
+ free(cp);
}
/*
out[4] = (cp >> 6 & 63) | 128;
out[5] = (cp & 63) | 128;
} else
- return(0);
+ return 0;
out[rc] = '\0';
- return(rc);
+ return rc;
}
/*
- * Store the UTF-8 version of a key, or alias the pointer if the key has
- * no UTF-8 transcription marks in it.
+ * If the string contains escape sequences,
+ * replace it with an allocated rendering and return 1,
+ * such that the caller can free it after use.
+ * Otherwise, do nothing and return 0.
*/
-static void
-utf8key(struct mchars *mc, struct str *key)
+static int
+render_string(char **public, size_t *psz)
{
- size_t sz, bsz, pos;
- char utfbuf[7], res[5];
- char *buf;
- const char *seq, *cpp, *val;
- int len, u;
- enum mandoc_esc esc;
-
- assert(NULL == key->utf8);
+ const char *src, *scp, *addcp, *seq;
+ char *dst;
+ size_t ssz, dsz, addsz;
+ char utfbuf[7], res[6];
+ int seqlen, unicode;
res[0] = '\\';
res[1] = '\t';
res[2] = ASCII_NBRSP;
res[3] = ASCII_HYPH;
- res[4] = '\0';
+ res[4] = ASCII_BREAK;
+ res[5] = '\0';
- val = key->key;
- bsz = strlen(val);
+ src = scp = *public;
+ ssz = *psz;
+ dst = NULL;
+ dsz = 0;
- /*
- * Pre-check: if we have no stop-characters, then set the
- * pointer as ourselvse and get out of here.
- */
- if (strcspn(val, res) == bsz) {
- key->utf8 = key->key;
- return;
- }
+ while (scp < src + *psz) {
- /* Pre-allocate by the length of the input */
+ /* Leave normal characters unchanged. */
- buf = mandoc_malloc(++bsz);
- pos = 0;
+ if (strchr(res, *scp) == NULL) {
+ if (dst != NULL)
+ dst[dsz++] = *scp;
+ scp++;
+ continue;
+ }
- while ('\0' != *val) {
/*
- * Halt on the first escape sequence.
- * This also halts on the end of string, in which case
- * we just copy, fallthrough, and exit the loop.
+ * Found something that requires replacing,
+ * make sure we have a destination buffer.
*/
- if ((sz = strcspn(val, res)) > 0) {
- memcpy(&buf[pos], val, sz);
- pos += sz;
- val += sz;
+
+ if (dst == NULL) {
+ dst = mandoc_malloc(ssz + 1);
+ dsz = scp - src;
+ memcpy(dst, src, dsz);
}
- if (ASCII_HYPH == *val) {
- buf[pos++] = '-';
- val++;
+ /* Handle single-char special characters. */
+
+ switch (*scp) {
+ case '\\':
+ break;
+ case '\t':
+ case ASCII_NBRSP:
+ dst[dsz++] = ' ';
+ scp++;
continue;
- } else if ('\t' == *val || ASCII_NBRSP == *val) {
- buf[pos++] = ' ';
- val++;
+ case ASCII_HYPH:
+ dst[dsz++] = '-';
+ /* FALLTHROUGH */
+ case ASCII_BREAK:
+ scp++;
continue;
- } else if ('\\' != *val)
- break;
-
- /* Read past the slash. */
-
- val++;
- u = 0;
+ default:
+ abort();
+ }
/*
- * Parse the escape sequence and see if it's a
- * predefined character or special character.
+ * Found an escape sequence.
+ * Read past the slash, then parse it.
+ * Ignore everything except characters.
*/
- esc = mandoc_escape
- ((const char **)&val, &seq, &len);
- if (ESCAPE_ERROR == esc)
- break;
- if (ESCAPE_SPECIAL != esc)
- continue;
- if (0 == (u = mchars_spec2cp(mc, seq, len)))
+ scp++;
+ if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
continue;
/*
- * If we have a Unicode codepoint, try to convert that
- * to a UTF-8 byte string.
+ * Render the special character
+ * as either UTF-8 or ASCII.
*/
- cpp = utfbuf;
- if (0 == (sz = utf8(u, utfbuf)))
- continue;
+
+ if (write_utf8) {
+ unicode = mchars_spec2cp(seq, seqlen);
+ if (unicode <= 0)
+ continue;
+ addsz = utf8(unicode, utfbuf);
+ if (addsz == 0)
+ continue;
+ addcp = utfbuf;
+ } else {
+ addcp = mchars_spec2str(seq, seqlen, &addsz);
+ if (addcp == NULL)
+ continue;
+ if (*addcp == ASCII_NBRSP) {
+ addcp = " ";
+ addsz = 1;
+ }
+ }
/* Copy the rendered glyph into the stream. */
- sz = strlen(cpp);
- bsz += sz;
+ ssz += addsz;
+ dst = mandoc_realloc(dst, ssz + 1);
+ memcpy(dst + dsz, addcp, addsz);
+ dsz += addsz;
+ }
+ if (dst != NULL) {
+ *public = dst;
+ *psz = dsz;
+ }
- buf = mandoc_realloc(buf, bsz);
+ /* Trim trailing whitespace and NUL-terminate. */
- memcpy(&buf[pos], cpp, sz);
- pos += sz;
- }
+ while (*psz > 0 && (*public)[*psz - 1] == ' ')
+ --*psz;
+ if (dst != NULL) {
+ (*public)[*psz] = '\0';
+ return 1;
+ } else
+ return 0;
+}
- buf[pos] = '\0';
- key->utf8 = buf;
+static void
+dbadd_mlink(const struct mlink *mlink)
+{
+ dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
+ dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
+ dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
+ dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
+ dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
}
/*
* Flush the current page's terms (and their bits) into the database.
- * Wrap the entire set of additions in a transaction to make sqlite be a
- * little faster.
- * Also, UTF-8-encode the description at the last possible moment.
+ * Also, handle escape sequences at the last possible moment.
*/
static void
-dbindex(struct mchars *mc, int form, const struct mpage *mpage)
+dbadd(struct dba *dba, struct mpage *mpage)
{
+ struct mlink *mlink;
struct str *key;
- const char *desc;
- int64_t recno;
+ char *cp;
+ uint64_t mask;
size_t i;
unsigned int slot;
-
- if (verb)
- say(mpage->file, "Adding to index");
-
- if (nodb)
+ int mustfree;
+
+ mlink = mpage->mlinks;
+
+ if (nodb) {
+ for (key = ohash_first(&names, &slot); NULL != key;
+ key = ohash_next(&names, &slot))
+ free(key);
+ for (key = ohash_first(&strings, &slot); NULL != key;
+ key = ohash_next(&strings, &slot))
+ free(key);
+ if (0 == debug)
+ return;
+ while (NULL != mlink) {
+ fputs(mlink->name, stdout);
+ if (NULL == mlink->next ||
+ strcmp(mlink->dsec, mlink->next->dsec) ||
+ strcmp(mlink->fsec, mlink->next->fsec) ||
+ strcmp(mlink->arch, mlink->next->arch)) {
+ putchar('(');
+ if ('\0' == *mlink->dsec)
+ fputs(mlink->fsec, stdout);
+ else
+ fputs(mlink->dsec, stdout);
+ if ('\0' != *mlink->arch)
+ printf("/%s", mlink->arch);
+ putchar(')');
+ }
+ mlink = mlink->next;
+ if (NULL != mlink)
+ fputs(", ", stdout);
+ }
+ printf(" - %s\n", mpage->desc);
return;
-
- desc = "";
- if (NULL != mpage->desc && '\0' != *mpage->desc) {
- key = ohash_find(&strings,
- ohash_qlookup(&strings, mpage->desc));
- assert(NULL != key);
- if (NULL == key->utf8)
- utf8key(mc, key);
- desc = key->utf8;
}
- SQL_EXEC("BEGIN TRANSACTION");
-
- i = 1;
- SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, mpage->file);
- SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, mpage->sec);
- SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, mpage->arch);
- SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, desc);
- SQL_BIND_INT(stmts[STMT_INSERT_DOC], i, form);
- SQL_STEP(stmts[STMT_INSERT_DOC]);
- recno = sqlite3_last_insert_rowid(db);
- sqlite3_reset(stmts[STMT_INSERT_DOC]);
+ if (debug)
+ say(mlink->file, "Adding to database");
+
+ cp = mpage->desc;
+ i = strlen(cp);
+ mustfree = render_string(&cp, &i);
+ mpage->dba = dba_page_new(dba->pages,
+ *mpage->arch == '\0' ? mlink->arch : mpage->arch,
+ cp, mlink->file, mpage->form);
+ if (mustfree)
+ free(cp);
+ dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
+
+ while (mlink != NULL) {
+ dbadd_mlink(mlink);
+ mlink = mlink->next;
+ }
+ for (key = ohash_first(&names, &slot); NULL != key;
+ key = ohash_next(&names, &slot)) {
+ assert(key->mpage == mpage);
+ dba_page_alias(mpage->dba, key->key, key->mask);
+ free(key);
+ }
for (key = ohash_first(&strings, &slot); NULL != key;
key = ohash_next(&strings, &slot)) {
assert(key->mpage == mpage);
- if (NULL == key->utf8)
- utf8key(mc, key);
- i = 1;
- SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
- SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->utf8);
- SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno);
- SQL_STEP(stmts[STMT_INSERT_KEY]);
- sqlite3_reset(stmts[STMT_INSERT_KEY]);
- if (key->utf8 != key->key)
- free(key->utf8);
+ i = 0;
+ for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
+ if (key->mask & mask)
+ dba_macro_add(dba->macros, i,
+ key->key, mpage->dba);
+ i++;
+ }
free(key);
}
-
- SQL_EXEC("END TRANSACTION");
}
static void
-dbprune(void)
+dbprune(struct dba *dba)
{
- struct mpage *mpage;
- size_t i;
-
- if (nodb)
- return;
-
- for (mpage = mpages_list; NULL != mpage; mpage = mpage->next) {
- i = 1;
- SQL_BIND_TEXT(stmts[STMT_DELETE], i, mpage->file);
- SQL_STEP(stmts[STMT_DELETE]);
- sqlite3_reset(stmts[STMT_DELETE]);
- if (verb)
- say(mpage->file, "Deleted from index");
+ struct dba_array *page, *files;
+ char *file;
+
+ dba_array_FOREACH(dba->pages, page) {
+ files = dba_array_get(page, DBP_FILE);
+ dba_array_FOREACH(files, file) {
+ if (*file < ' ')
+ file++;
+ if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
+ file)) != NULL) {
+ if (debug)
+ say(file, "Deleting from database");
+ dba_array_del(dba->pages);
+ break;
+ }
+ }
}
}
/*
- * Close an existing database and its prepared statements.
- * If "real" is not set, rename the temporary file into the real one.
+ * Write the database from memory to disk.
*/
static void
-dbclose(int real)
+dbwrite(struct dba *dba)
{
- size_t i;
+ char tfn[32];
+ int status;
+ pid_t child;
- if (nodb)
- return;
-
- for (i = 0; i < STMT__MAX; i++) {
- sqlite3_finalize(stmts[i]);
- stmts[i] = NULL;
- }
-
- sqlite3_close(db);
- db = NULL;
+ /*
+ * Do not write empty databases, and delete existing ones
+ * when makewhatis -u causes them to become empty.
+ */
- if (real)
+ dba_array_start(dba->pages);
+ if (dba_array_next(dba->pages) == NULL) {
+ if (unlink(MANDOC_DB) == -1)
+ say(MANDOC_DB, "&unlink");
return;
-
- if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
- exitcode = (int)MANDOCLEVEL_SYSERR;
- say(MANDOC_DB, NULL);
}
-}
-/*
- * This is straightforward stuff.
- * Open a database connection to a "temporary" database, then open a set
- * of prepared statements we'll use over and over again.
- * If "real" is set, we use the existing database; if not, we truncate a
- * temporary one.
- * Must be matched by dbclose().
- */
-static int
-dbopen(int real)
-{
- const char *file, *sql;
- int rc, ofl;
-
- if (nodb)
- return(1);
+ /*
+ * Build the database in a temporary file,
+ * then atomically move it into place.
+ */
- ofl = SQLITE_OPEN_READWRITE;
- if (0 == real) {
- file = MANDOC_DB "~";
- if (-1 == remove(file) && ENOENT != errno) {
+ if (dba_write(MANDOC_DB "~", dba) != -1) {
+ if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
exitcode = (int)MANDOCLEVEL_SYSERR;
- say(file, NULL);
- return(0);
+ say(MANDOC_DB, "&rename");
+ unlink(MANDOC_DB "~");
}
- ofl |= SQLITE_OPEN_EXCLUSIVE;
- } else
- file = MANDOC_DB;
-
- rc = sqlite3_open_v2(file, &db, ofl, NULL);
- if (SQLITE_OK == rc)
- goto prepare_statements;
- if (SQLITE_CANTOPEN != rc) {
- exitcode = (int)MANDOCLEVEL_SYSERR;
- say(file, NULL);
- return(0);
+ return;
}
- sqlite3_close(db);
- db = NULL;
+ /*
+ * We lack write permission and cannot replace the database
+ * file, but let's at least check whether the data changed.
+ */
- if (SQLITE_OK != (rc = sqlite3_open(file, &db))) {
+ (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
+ if (mkdtemp(tfn) == NULL) {
exitcode = (int)MANDOCLEVEL_SYSERR;
- say(file, NULL);
- return(0);
+ say("", "&%s", tfn);
+ return;
}
- sql = "CREATE TABLE \"docs\" (\n"
- " \"file\" TEXT NOT NULL,\n"
- " \"sec\" TEXT NOT NULL,\n"
- " \"arch\" TEXT NOT NULL,\n"
- " \"desc\" TEXT NOT NULL,\n"
- " \"form\" INTEGER NOT NULL,\n"
- " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
- ");\n"
- "\n"
- "CREATE TABLE \"keys\" (\n"
- " \"bits\" INTEGER NOT NULL,\n"
- " \"key\" TEXT NOT NULL,\n"
- " \"docid\" INTEGER NOT NULL REFERENCES docs(id) "
- "ON DELETE CASCADE,\n"
- " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
- ");\n"
- "\n"
- "CREATE INDEX \"key_index\" ON keys (key);\n";
-
- if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
+ (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
+ if (dba_write(tfn, dba) == -1) {
exitcode = (int)MANDOCLEVEL_SYSERR;
- say(file, "%s", sqlite3_errmsg(db));
- return(0);
+ say(tfn, "&dba_write");
+ goto out;
}
-prepare_statements:
- SQL_EXEC("PRAGMA foreign_keys = ON");
- sql = "DELETE FROM docs where file=?";
- sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE], NULL);
- sql = "INSERT INTO docs "
- "(file,sec,arch,desc,form) VALUES (?,?,?,?,?)";
- sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_DOC], NULL);
- sql = "INSERT INTO keys "
- "(bits,key,docid) VALUES (?,?,?)";
- sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
-
-#ifndef __APPLE__
- /*
- * When opening a new database, we can turn off
- * synchronous mode for much better performance.
- */
-
- if (real)
- SQL_EXEC("PRAGMA synchronous = OFF");
-#endif
-
- return(1);
-}
-
-static void *
-hash_halloc(size_t sz, void *arg)
-{
-
- return(mandoc_calloc(sz, 1));
-}
-
-static void *
-hash_alloc(size_t sz, void *arg)
-{
-
- return(mandoc_malloc(sz));
-}
-
-static void
-hash_free(void *p, size_t sz, void *arg)
-{
+ switch (child = fork()) {
+ case -1:
+ exitcode = (int)MANDOCLEVEL_SYSERR;
+ say("", "&fork cmp");
+ return;
+ case 0:
+ execlp("cmp", "cmp", "-s", tfn, MANDOC_DB, (char *)NULL);
+ say("", "&exec cmp");
+ exit(0);
+ default:
+ break;
+ }
+ if (waitpid(child, &status, 0) == -1) {
+ exitcode = (int)MANDOCLEVEL_SYSERR;
+ say("", "&wait cmp");
+ } else if (WIFSIGNALED(status)) {
+ exitcode = (int)MANDOCLEVEL_SYSERR;
+ say("", "cmp died from signal %d", WTERMSIG(status));
+ } else if (WEXITSTATUS(status)) {
+ exitcode = (int)MANDOCLEVEL_SYSERR;
+ say(MANDOC_DB,
+ "Data changed, but cannot replace database");
+ }
- free(p);
+out:
+ *strrchr(tfn, '/') = '\0';
+ switch (child = fork()) {
+ case -1:
+ exitcode = (int)MANDOCLEVEL_SYSERR;
+ say("", "&fork rm");
+ return;
+ case 0:
+ execlp("rm", "rm", "-rf", tfn, (char *)NULL);
+ say("", "&exec rm");
+ exit((int)MANDOCLEVEL_SYSERR);
+ default:
+ break;
+ }
+ if (waitpid(child, &status, 0) == -1) {
+ exitcode = (int)MANDOCLEVEL_SYSERR;
+ say("", "&wait rm");
+ } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
+ exitcode = (int)MANDOCLEVEL_SYSERR;
+ say("", "%s: Cannot remove temporary directory", tfn);
+ }
}
static int
-set_basedir(const char *targetdir)
+set_basedir(const char *targetdir, int report_baddir)
{
static char startdir[PATH_MAX];
- static int fd;
+ static int getcwd_status; /* 1 = ok, 2 = failure */
+ static int chdir_status; /* 1 = changed directory */
+ char *cp;
/*
- * Remember where we started by keeping a fd open to the origin
- * path component: throughout this utility, we chdir() a lot to
- * handle relative paths, and by doing this, we can return to
- * the starting point.
+ * Remember the original working directory, if possible.
+ * This will be needed if the second or a later directory
+ * on the command line is given as a relative path.
+ * Do not error out if the current directory is not
+ * searchable: Maybe it won't be needed after all.
*/
- if ('\0' == *startdir) {
- if (NULL == getcwd(startdir, PATH_MAX)) {
- exitcode = (int)MANDOCLEVEL_SYSERR;
- if (NULL != targetdir)
- say(".", NULL);
- return(0);
- }
- if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
+ if (0 == getcwd_status) {
+ if (NULL == getcwd(startdir, sizeof(startdir))) {
+ getcwd_status = 2;
+ (void)strlcpy(startdir, strerror(errno),
+ sizeof(startdir));
+ } else
+ getcwd_status = 1;
+ }
+
+ /*
+ * We are leaving the old base directory.
+ * Do not use it any longer, not even for messages.
+ */
+ *basedir = '\0';
+
+ /*
+ * If and only if the directory was changed earlier and
+ * the next directory to process is given as a relative path,
+ * first go back, or bail out if that is impossible.
+ */
+ if (chdir_status && '/' != *targetdir) {
+ if (2 == getcwd_status) {
exitcode = (int)MANDOCLEVEL_SYSERR;
- say(startdir, NULL);
- return(0);
+ say("", "getcwd: %s", startdir);
+ return 0;
}
- if (NULL == targetdir)
- targetdir = startdir;
- } else {
- if (-1 == fd)
- return(0);
- if (-1 == fchdir(fd)) {
- close(fd);
- basedir[0] = '\0';
+ if (-1 == chdir(startdir)) {
exitcode = (int)MANDOCLEVEL_SYSERR;
- say(startdir, NULL);
- return(0);
- }
- if (NULL == targetdir) {
- close(fd);
- return(1);
+ say("", "&chdir %s", startdir);
+ return 0;
}
}
+
+ /*
+ * Always resolve basedir to the canonicalized absolute
+ * pathname and append a trailing slash, such that
+ * we can reliably check whether files are inside.
+ */
if (NULL == realpath(targetdir, basedir)) {
- basedir[0] = '\0';
- exitcode = (int)MANDOCLEVEL_BADARG;
- say(targetdir, NULL);
- return(0);
+ if (report_baddir || errno != ENOENT) {
+ exitcode = (int)MANDOCLEVEL_BADARG;
+ say("", "&%s: realpath", targetdir);
+ }
+ return 0;
} else if (-1 == chdir(basedir)) {
- exitcode = (int)MANDOCLEVEL_BADARG;
- say("", NULL);
- return(0);
+ if (report_baddir || errno != ENOENT) {
+ exitcode = (int)MANDOCLEVEL_BADARG;
+ say("", "&chdir");
+ }
+ return 0;
}
- return(1);
+ chdir_status = 1;
+ cp = strchr(basedir, '\0');
+ if ('/' != cp[-1]) {
+ if (cp - basedir >= PATH_MAX - 1) {
+ exitcode = (int)MANDOCLEVEL_SYSERR;
+ say("", "Filename too long");
+ return 0;
+ }
+ *cp++ = '/';
+ *cp = '\0';
+ }
+ return 1;
}
static void
say(const char *file, const char *format, ...)
{
va_list ap;
+ int use_errno;
if ('\0' != *basedir)
fprintf(stderr, "%s", basedir);
if ('\0' != *basedir && '\0' != *file)
- fputs("//", stderr);
+ fputc('/', stderr);
if ('\0' != *file)
fprintf(stderr, "%s", file);
- fputs(": ", stderr);
- if (NULL == format) {
- perror(NULL);
- return;
+ use_errno = 1;
+ if (NULL != format) {
+ switch (*format) {
+ case '&':
+ format++;
+ break;
+ case '\0':
+ format = NULL;
+ break;
+ default:
+ use_errno = 0;
+ break;
+ }
}
-
- va_start(ap, format);
- vfprintf(stderr, format, ap);
- va_end(ap);
-
- fputc('\n', stderr);
+ if (NULL != format) {
+ if ('\0' != *basedir || '\0' != *file)
+ fputs(": ", stderr);
+ va_start(ap, format);
+ vfprintf(stderr, format, ap);
+ va_end(ap);
+ }
+ if (use_errno) {
+ if ('\0' != *basedir || '\0' != *file || NULL != format)
+ fputs(": ", stderr);
+ perror(NULL);
+ } else
+ fputc('\n', stderr);
}
git.ckatri.com / mandoc.git / blobdiff