=================================================================== RCS file: /cvs/mandoc/mandocdb.c,v retrieving revision 1.220.2.12 retrieving revision 1.271 diff -u -p -r1.220.2.12 -r1.271 --- mandoc/mandocdb.c 2017/02/17 14:50:13 1.220.2.12 +++ mandoc/mandocdb.c 2022/04/14 16:43:44 1.271 @@ -1,7 +1,7 @@ -/* $Id: mandocdb.c,v 1.220.2.12 2017/02/17 14:50:13 schwarze Exp $ */ +/* $Id: mandocdb.c,v 1.271 2022/04/14 16:43:44 schwarze Exp $ */ /* + * Copyright (c) 2011-2021 Ingo Schwarze * Copyright (c) 2011, 2012 Kristaps Dzonsons - * Copyright (c) 2011-2017 Ingo Schwarze * Copyright (c) 2016 Ed Maste * * Permission to use, copy, modify, and distribute this software for any @@ -15,12 +15,14 @@ * 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. + * + * Implementation of the makewhatis(8) program. */ #include "config.h" #include +#include #include -#include #include #include @@ -38,6 +40,7 @@ #if HAVE_SANDBOX_INIT #include #endif +#include #include #include #include @@ -45,39 +48,20 @@ #include #include -#include - #include "mandoc_aux.h" #include "mandoc_ohash.h" #include "mandoc.h" #include "roff.h" #include "mdoc.h" #include "man.h" +#include "mandoc_parse.h" #include "manconf.h" #include "mansearch.h" +#include "dba_array.h" +#include "dba.h" -extern int mansearch_keymax; extern const char *const mansearch_keynames[]; -#define SQL_EXEC(_v) \ - if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \ - say("", "%s: %s", (_v), sqlite3_errmsg(db)) -#define SQL_BIND_TEXT(_s, _i, _v) \ - if (SQLITE_OK != sqlite3_bind_text \ - ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \ - say(mlink->file, "%s", sqlite3_errmsg(db)) -#define SQL_BIND_INT(_s, _i, _v) \ - if (SQLITE_OK != sqlite3_bind_int \ - ((_s), (_i)++, (_v))) \ - say(mlink->file, "%s", sqlite3_errmsg(db)) -#define SQL_BIND_INT64(_s, _i, _v) \ - if (SQLITE_OK != sqlite3_bind_int64 \ - ((_s), (_i)++, (_v))) \ - say(mlink->file, "%s", sqlite3_errmsg(db)) -#define SQL_STEP(_s) \ - if (SQLITE_DONE != sqlite3_step((_s))) \ - say(mlink->file, "%s", sqlite3_errmsg(db)) - enum op { OP_DEFAULT = 0, /* new dbs from dir list or default config */ OP_CONFFILE, /* new databases from custom config file */ @@ -99,15 +83,15 @@ struct inodev { struct mpage { struct inodev inodev; /* used for hashing routine */ - int64_t pageid; /* pageid in mpages SQL table */ + 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 form; /* format from file content */ int name_head_done; + enum form form; /* format from file content */ }; struct mlink { @@ -118,21 +102,11 @@ struct mlink { char *fsec; /* section from file name suffix */ struct mlink *next; /* singly linked list */ struct mpage *mpage; /* parent */ - int dform; /* format from directory */ - int fform; /* format from file name suffix */ int gzip; /* filename has a .gz suffix */ + enum form dform; /* format from directory */ + enum form fform; /* format from file name suffix */ }; -enum stmt { - STMT_DELETE_PAGE = 0, /* delete mpage */ - STMT_INSERT_PAGE, /* insert mpage */ - STMT_INSERT_LINK, /* insert mlink */ - STMT_INSERT_NAME, /* insert name */ - STMT_SELECT_NAME, /* retrieve existing name flags */ - STMT_INSERT_KEY, /* insert parsed key */ - STMT__MAX -}; - typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *, const struct roff_node *); @@ -145,12 +119,10 @@ struct mdoc_handler { int mandocdb(int, char *[]); -static void dbclose(int); -static void dbadd(struct mpage *); -static void dbadd_mlink(const struct mlink *mlink); -static void dbadd_mlink_name(const struct mlink *mlink); -static int dbopen(int); -static void dbprune(void); +static void dbadd(struct dba *, struct mpage *); +static void dbadd_mlink(const struct mlink *); +static void dbprune(struct dba *); +static void dbwrite(struct dba *); static void filescan(const char *); #if HAVE_FTS_COMPARE_CONST static int fts_compare(const FTSENT *const *, const FTSENT *const *); @@ -162,8 +134,7 @@ 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 mparse *); -static void names_check(void); +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 *); @@ -171,6 +142,8 @@ static void parse_mdoc(struct mpage *, const struct r const struct roff_node *); static int parse_mdoc_head(struct mpage *, const struct roff_meta *, const struct roff_node *); +static int parse_mdoc_Fa(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 *); @@ -192,6 +165,9 @@ static void putkey(const struct mpage *, char *, uint static void putkeys(const struct mpage *, char *, size_t, uint64_t); static void putmdockey(const struct mpage *, const struct roff_node *, uint64_t, int); +#ifdef READ_ALLOWED_PATH +static int read_allowed(const char *); +#endif static int render_string(char **, size_t *); static void say(const char *, const char *, ...) __attribute__((__format__ (__printf__, 2, 3))); @@ -199,7 +175,6 @@ static int set_basedir(const char *, int); static int treescan(void); static size_t utf8(unsigned int, char [7]); -static char tempfilename[32]; static int nodb; /* no database changes */ static int mparse_options; /* abort the parse early */ static int use_all; /* use all found files */ @@ -209,17 +184,15 @@ static int write_utf8; /* write UTF-8 output; else A static int exitcode; /* to be returned by main */ static enum op op; /* operational mode */ static char basedir[PATH_MAX]; /* current base directory */ +static size_t basedir_len; /* strlen(basedir) */ static struct mpage *mpage_head; /* list of distinct manual pages */ static struct ohash mpages; /* table of distinct manual pages */ 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 sqlite3 *db = NULL; /* current database */ -static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */ static uint64_t name_mask; -static const struct mdoc_handler mdocs[MDOC_MAX] = { - { NULL, 0, 0 }, /* Ap */ +static const struct mdoc_handler mdoc_handlers[MDOC_MAX - MDOC_Dd] = { { NULL, 0, NODE_NOPRT }, /* Dd */ { NULL, 0, NODE_NOPRT }, /* Dt */ { NULL, 0, NODE_NOPRT }, /* Os */ @@ -235,6 +208,7 @@ static const struct mdoc_handler mdocs[MDOC_MAX] = { { 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 */ @@ -242,11 +216,11 @@ static const struct mdoc_handler mdocs[MDOC_MAX] = { { NULL, TYPE_Er, 0 }, /* Er */ { NULL, TYPE_Ev, 0 }, /* Ev */ { NULL, 0, 0 }, /* Ex */ - { NULL, TYPE_Fa, 0 }, /* Fa */ + { parse_mdoc_Fa, 0, 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_Ft | TYPE_Vt, 0 }, /* Ft */ { NULL, TYPE_Ic, 0 }, /* Ic */ { NULL, TYPE_In, 0 }, /* In */ { NULL, TYPE_Li, 0 }, /* Li */ @@ -337,11 +311,8 @@ static const struct mdoc_handler mdocs[MDOC_MAX] = { { NULL, 0, 0 }, /* En */ { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */ { NULL, 0, 0 }, /* %Q */ - { NULL, 0, 0 }, /* br */ - { NULL, 0, 0 }, /* sp */ { NULL, 0, 0 }, /* %U */ { NULL, 0, 0 }, /* Ta */ - { NULL, 0, 0 }, /* ll */ }; @@ -350,12 +321,13 @@ mandocdb(int argc, char *argv[]) { struct manconf conf; struct mparse *mp; + struct dba *dba; const char *path_arg, *progname; size_t j, sz; int ch, i; #if HAVE_PLEDGE - if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) { + if (pledge("stdio rpath wpath cpath", NULL) == -1) { warn("pledge"); return (int)MANDOCLEVEL_SYSERR; } @@ -369,7 +341,6 @@ mandocdb(int argc, char *argv[]) #endif memset(&conf, 0, sizeof(conf)); - memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *)); /* * We accept a few different invocations. @@ -377,15 +348,16 @@ mandocdb(int argc, char *argv[]) * clobber each other. */ #define CHECKOP(_op, _ch) do \ - if (OP_DEFAULT != (_op)) { \ + if ((_op) != OP_DEFAULT) { \ warnx("-%c: Conflicting option", (_ch)); \ goto usage; \ } while (/*CONSTCOND*/0) + mparse_options = MPARSE_VALIDATE; path_arg = NULL; op = OP_DEFAULT; - while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v"))) + while ((ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")) != -1) switch (ch) { case 'a': use_all = 1; @@ -413,7 +385,7 @@ mandocdb(int argc, char *argv[]) mparse_options |= MPARSE_QUICK; break; case 'T': - if (strcmp(optarg, "utf8")) { + if (strcmp(optarg, "utf8") != 0) { warnx("-T%s: Unsupported output format", optarg); goto usage; @@ -450,65 +422,57 @@ mandocdb(int argc, char *argv[]) } #endif - if (OP_CONFFILE == op && argc > 0) { + if (op == OP_CONFFILE && argc > 0) { warnx("-C: Too many arguments"); goto usage; } exitcode = (int)MANDOCLEVEL_OK; mchars_alloc(); - mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL); + mp = mparse_alloc(mparse_options, MANDOC_OS_OTHER, 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) { + if (op == OP_UPDATE || op == OP_DELETE || op == OP_TEST) { /* * Most of these deal with a specific directory. * Jump into that directory first. */ - if (OP_TEST != op && 0 == set_basedir(path_arg, 1)) + if (op != OP_TEST && set_basedir(path_arg, 1) == 0) goto out; - if (dbopen(1)) { + 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 (OP_TEST != op) - dbprune(); + if (nodb == 0) + dbprune(dba); } else { - /* - * Database missing or corrupt. - * Recreate from scratch. - */ + /* 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()) + if (treescan() == 0) goto out; - if (0 == dbopen(0)) - goto out; + dba = dba_new(128); } - if (OP_DELETE != op) - mpages_merge(mp); - dbclose(OP_DEFAULT == op ? 0 : 1); + if (op != OP_DELETE) + 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 - * manconf_parse() wants to do it. + * If we don't, use man.conf(5). */ if (argc > 0) { conf.manpath.paths = mandoc_reallocarray(NULL, @@ -534,7 +498,7 @@ mandocdb(int argc, char *argv[]) sz = strlen(conf.manpath.paths[j]); if (sz && conf.manpath.paths[j][sz - 1] == '/') conf.manpath.paths[j][--sz] = '\0'; - if (0 == sz) + if (sz == 0) continue; if (j) { @@ -544,19 +508,16 @@ mandocdb(int argc, char *argv[]) offsetof(struct mlink, file)); } - if ( ! set_basedir(conf.manpath.paths[j], argc > 0)) + if (set_basedir(conf.manpath.paths[j], argc > 0) == 0) continue; - if (0 == treescan()) + if (treescan() == 0) continue; - if (0 == dbopen(0)) - continue; + dba = dba_new(128); + mpages_merge(dba, mp); + if (nodb == 0) + dbwrite(dba); + dba_free(dba); - mpages_merge(mp); - if (warnings && !nodb && - ! (MPARSE_QUICK & mparse_options)) - names_check(); - dbclose(0); - if (j + 1 < conf.manpath.sz) { mpages_free(); ohash_delete(&mpages); @@ -571,6 +532,9 @@ out: mpages_free(); ohash_delete(&mpages); ohash_delete(&mlinks); +#if DEBUG_MEMORY + mandoc_dbg_finish(); +#endif return exitcode; usage: progname = getprogname(); @@ -619,7 +583,8 @@ treescan(void) FTS *f; FTSENT *ff; struct mlink *mlink; - int dform, gzip; + int gzip; + enum form dform; char *dsec, *arch, *fsec, *cp; const char *path; const char *argv[2]; @@ -652,9 +617,9 @@ treescan(void) say(path, "&realpath"); continue; } - if (strstr(buf, basedir) != buf -#ifdef HOMEBREWDIR - && strstr(buf, HOMEBREWDIR) != buf + if (strncmp(buf, basedir, basedir_len) != 0 +#ifdef READ_ALLOWED_PATH + && !read_allowed(buf) #endif ) { if (warnings) say("", @@ -667,6 +632,8 @@ treescan(void) say(path, "&stat"); continue; } + if ((ff->fts_statp->st_mode & S_IFMT) != S_IFREG) + continue; /* FALLTHROUGH */ /* @@ -821,17 +788,17 @@ treescan(void) * See treescan() for the fts(3) version of this. */ static void -filescan(const char *file) +filescan(const char *infile) { - char buf[PATH_MAX]; struct stat st; struct mlink *mlink; - char *p, *start; + char *linkfile, *p, *realdir, *start, *usefile; + size_t realdir_len; assert(use_all); - if (0 == strncmp(file, "./", 2)) - file += 2; + if (strncmp(infile, "./", 2) == 0) + infile += 2; /* * We have to do lstat(2) before realpath(3) loses @@ -840,13 +807,13 @@ filescan(const char *file) * we want to use the orginal file name, while for * regular files, we want to use the real path. */ - if (-1 == lstat(file, &st)) { + if (lstat(infile, &st) == -1) { exitcode = (int)MANDOCLEVEL_BADARG; - say(file, "&lstat"); + say(infile, "&lstat"); return; - } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) { + } else if (S_ISREG(st.st_mode) == 0 && S_ISLNK(st.st_mode) == 0) { exitcode = (int)MANDOCLEVEL_BADARG; - say(file, "Not a regular file"); + say(infile, "Not a regular file"); return; } @@ -854,23 +821,24 @@ filescan(const char *file) * We have to resolve the file name to the real path * in any case for the base directory check. */ - if (NULL == realpath(file, buf)) { + if ((usefile = realpath(infile, NULL)) == NULL) { exitcode = (int)MANDOCLEVEL_BADARG; - say(file, "&realpath"); + say(infile, "&realpath"); return; } - 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; + if (op == OP_TEST) + start = usefile; + else if (strncmp(usefile, basedir, basedir_len) == 0) + start = usefile + basedir_len; +#ifdef READ_ALLOWED_PATH + else if (read_allowed(usefile)) + start = usefile; #endif else { exitcode = (int)MANDOCLEVEL_BADARG; - say("", "%s: outside base directory", buf); + say("", "%s: outside base directory", infile); + free(usefile); return; } @@ -878,32 +846,80 @@ filescan(const char *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)) { + do { + if (S_ISLNK(st.st_mode) == 0) + break; + + /* + * Some implementations of realpath(3) may succeed + * even if the target of the link does not exist, + * so check again for extra safety. + */ + if (stat(usefile, &st) == -1) { exitcode = (int)MANDOCLEVEL_BADARG; - say(file, "&stat"); + say(infile, "&stat"); + free(usefile); return; } - if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) { - say(file, "Filename too long"); - return; + linkfile = mandoc_strdup(infile); + if (op == OP_TEST) { + free(usefile); + start = usefile = linkfile; + break; } - start = buf; - if (OP_TEST != op && strstr(buf, basedir) == buf) - start += strlen(basedir); - } + if (strncmp(infile, basedir, basedir_len) == 0) { + free(usefile); + usefile = linkfile; + start = usefile + basedir_len; + break; + } + /* + * This symbolic link points into the basedir + * from the outside. Let's see whether any of + * the parent directories resolve to the basedir. + */ + p = strchr(linkfile, '\0'); + do { + while (*--p != '/') + continue; + *p = '\0'; + if ((realdir = realpath(linkfile, NULL)) == NULL) { + exitcode = (int)MANDOCLEVEL_BADARG; + say(infile, "&realpath"); + free(linkfile); + free(usefile); + return; + } + realdir_len = strlen(realdir) + 1; + free(realdir); + *p = '/'; + } while (realdir_len > basedir_len); + + /* + * If one of the directories resolves to the basedir, + * use the rest of the original name. + * Otherwise, the best we can do + * is to use the filename pointed to. + */ + if (realdir_len == basedir_len) { + free(usefile); + usefile = linkfile; + start = p + 1; + } else { + free(linkfile); + start = usefile + basedir_len; + } + } while (/* CONSTCOND */ 0); + 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); + free(usefile); return; } @@ -912,13 +928,13 @@ filescan(const char *file) * 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) + if (op == OP_TEST || (start == usefile && *start == '/')) { + if (strncmp(usefile, "man/", 4) == 0) + start = usefile + 4; + else if ((start = strstr(usefile, "/man/")) != NULL) start += 5; else - start = buf; + start = usefile; } /* @@ -927,18 +943,18 @@ filescan(const char *file) * If we find one of these and what's underneath is a directory, * assume it's an architecture. */ - if (NULL != (p = strchr(start, '/'))) { + if ((p = strchr(start, '/')) != NULL) { *p++ = '\0'; - if (0 == strncmp(start, "man", 3)) { + if (strncmp(start, "man", 3) == 0) { mlink->dform = FORM_SRC; mlink->dsec = start + 3; - } else if (0 == strncmp(start, "cat", 3)) { + } else if (strncmp(start, "cat", 3) == 0) { mlink->dform = FORM_CAT; mlink->dsec = start + 3; } start = p; - if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) { + if (mlink->dsec != NULL && (p = strchr(start, '/')) != NULL) { *p++ = '\0'; mlink->arch = start; start = p; @@ -950,10 +966,10 @@ filescan(const char *file) * Suffix of `.0' indicates a catpage, `.1-9' is a manpage. */ p = strrchr(start, '\0'); - while (p-- > start && '/' != *p && '.' != *p) - /* Loop. */ ; + while (p-- > start && *p != '/' && *p != '.') + continue; - if ('.' == *p) { + if (*p == '.') { *p++ = '\0'; mlink->fsec = p; } @@ -963,11 +979,12 @@ filescan(const char *file) * Use the filename portion of the path. */ mlink->name = start; - if (NULL != (p = strrchr(start, '/'))) { + if ((p = strrchr(start, '/')) != NULL) { mlink->name = p + 1; *p = '\0'; } mlink_add(mlink, &st); + free(usefile); } static void @@ -1007,6 +1024,7 @@ mlink_add(struct mlink *mlink, const struct stat *st) 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); @@ -1120,7 +1138,7 @@ mlink_check(struct mpage *mpage, struct mlink *mlink) * architectures. * A few manuals are even shared across completely * different architectures, for example fdformat(1) - * on amd64, i386, sparc, and sparc64. + * on amd64, i386, and sparc64. */ if (strcasecmp(mpage->arch, mlink->arch)) @@ -1155,19 +1173,14 @@ mlink_check(struct mpage *mpage, struct mlink *mlink) * and filename to determine whether the file is parsable or not. */ static void -mpages_merge(struct mparse *mp) +mpages_merge(struct dba *dba, struct mparse *mp) { - char any[] = "any"; struct mpage *mpage, *mpage_dest; struct mlink *mlink, *mlink_dest; - struct roff_man *man; - char *sodest; + struct roff_meta *meta; char *cp; int fd; - if ( ! nodb) - SQL_EXEC("BEGIN TRANSACTION"); - for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) { mlinks_undupe(mpage); if ((mlink = mpage->mlinks) == NULL) @@ -1177,8 +1190,7 @@ mpages_merge(struct mparse *mp) mandoc_ohash_init(&names, 4, offsetof(struct str, key)); mandoc_ohash_init(&strings, 6, offsetof(struct str, key)); mparse_reset(mp); - man = NULL; - sodest = NULL; + meta = NULL; if ((fd = mparse_open(mp, mlink->file)) == -1) { say(mlink->file, "&open"); @@ -1193,14 +1205,14 @@ mpages_merge(struct mparse *mp) mparse_readfd(mp, fd, mlink->file); close(fd); fd = -1; - mparse_result(mp, &man, &sodest); + meta = mparse_result(mp); } - if (sodest != NULL) { + if (meta != NULL && meta->sodest != NULL) { mlink_dest = ohash_find(&mlinks, - ohash_qlookup(&mlinks, sodest)); + ohash_qlookup(&mlinks, meta->sodest)); if (mlink_dest == NULL) { - mandoc_asprintf(&cp, "%s.gz", sodest); + mandoc_asprintf(&cp, "%s.gz", meta->sodest); mlink_dest = ohash_find(&mlinks, ohash_qlookup(&mlinks, cp)); free(cp); @@ -1222,8 +1234,8 @@ mpages_merge(struct mparse *mp) * to the target. */ - if (mpage_dest->pageid) - dbadd_mlink_name(mlink); + if (mpage_dest->dba != NULL) + dbadd_mlink(mlink); if (mlink->next == NULL) break; @@ -1235,41 +1247,43 @@ mpages_merge(struct mparse *mp) mlink->next = mlink_dest->next; mlink_dest->next = mpage->mlinks; mpage->mlinks = NULL; + goto nextpage; } - goto nextpage; - } else if (man != NULL && man->macroset == MACROSET_MDOC) { - mdoc_validate(man); + meta->macroset = MACROSET_NONE; + } + if (meta != NULL && meta->macroset == MACROSET_MDOC) { mpage->form = FORM_SRC; - mpage->sec = man->meta.msec; + mpage->sec = meta->msec; mpage->sec = mandoc_strdup( mpage->sec == NULL ? "" : mpage->sec); - mpage->arch = man->meta.arch; + mpage->arch = 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.msec != '\0') { + mpage->title = mandoc_strdup(meta->title); + } else if (meta != NULL && meta->macroset == MACROSET_MAN) { + if (*meta->msec != '\0' || *meta->title != '\0') { mpage->form = FORM_SRC; - mpage->sec = mandoc_strdup(man->meta.msec); + mpage->sec = mandoc_strdup(meta->msec); mpage->arch = mandoc_strdup(mlink->arch); - mpage->title = mandoc_strdup(man->meta.title); + mpage->title = mandoc_strdup(meta->title); } else - man = NULL; + meta = NULL; } assert(mpage->desc == NULL); - if (man == NULL) { - mpage->form = FORM_CAT; + if (meta == NULL || meta->sodest != NULL) { 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); + if (meta == NULL) { + mpage->form = FORM_CAT; + parse_cat(mpage, fd); + } else + mpage->form = FORM_SRC; + } else if (meta->macroset == MACROSET_MDOC) + parse_mdoc(mpage, meta, meta->first); else - parse_man(mpage, &man->meta, man->first); + parse_man(mpage, meta, meta->first); if (mpage->desc == NULL) { mpage->desc = mandoc_strdup(mlink->name); if (warnings) @@ -1277,72 +1291,23 @@ mpages_merge(struct mparse *mp) "using filename \"%s\"", mlink->name); } - putkey(mpage, mpage->sec, TYPE_sec); - if (*mpage->arch != '\0') - putkey(mpage, mpage->arch, TYPE_arch); - - for ( ; mlink != NULL; mlink = mlink->next) { - if ('\0' != *mlink->dsec) - putkey(mpage, mlink->dsec, TYPE_sec); - if ('\0' != *mlink->fsec) - putkey(mpage, mlink->fsec, TYPE_sec); - putkey(mpage, '\0' == *mlink->arch ? - any : mlink->arch, TYPE_arch); + for (mlink = mpage->mlinks; + mlink != NULL; + mlink = mlink->next) { putkey(mpage, mlink->name, NAME_FILE); + if (warnings && !use_all) + mlink_check(mpage, mlink); } - if (warnings && !use_all) - for (mlink = mpage->mlinks; mlink; - mlink = mlink->next) - mlink_check(mpage, mlink); + dbadd(dba, mpage); - dbadd(mpage); - nextpage: ohash_delete(&strings); ohash_delete(&names); } - - if (0 == nodb) - SQL_EXEC("END TRANSACTION"); } static void -names_check(void) -{ - sqlite3_stmt *stmt; - const char *name, *sec, *arch, *key; - - sqlite3_prepare_v2(db, - "SELECT name, sec, arch, key FROM (" - "SELECT name AS key, pageid FROM names " - "WHERE bits & ? AND NOT EXISTS (" - "SELECT pageid FROM mlinks " - "WHERE mlinks.pageid == names.pageid " - "AND mlinks.name == names.name" - ")" - ") JOIN (" - "SELECT sec, arch, name, pageid FROM mlinks " - "GROUP BY pageid" - ") USING (pageid);", - -1, &stmt, NULL); - - if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK) - say("", "%s", sqlite3_errmsg(db)); - - while (sqlite3_step(stmt) == SQLITE_ROW) { - name = (const char *)sqlite3_column_text(stmt, 0); - sec = (const char *)sqlite3_column_text(stmt, 1); - arch = (const char *)sqlite3_column_text(stmt, 2); - key = (const char *)sqlite3_column_text(stmt, 3); - say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec, - '\0' == *arch ? "" : "/", - '\0' == *arch ? "" : arch, key); - } - sqlite3_finalize(stmt); -} - -static void parse_cat(struct mpage *mpage, int fd) { FILE *stream; @@ -1471,7 +1436,12 @@ parse_cat(struct mpage *mpage, int fd) plen -= 2; } - mpage->desc = mandoc_strdup(p); + /* + * Cut off excessive one-line descriptions. + * Bad pages are not worth better heuristics. + */ + + mpage->desc = mandoc_strndup(p, 150); fclose(stream); free(title); } @@ -1482,13 +1452,6 @@ parse_cat(struct mpage *mpage, int fd) static void putkey(const struct mpage *mpage, char *value, uint64_t type) { - char *cp; - - assert(NULL != value); - if (TYPE_arch == type) - for (cp = value; *cp; cp++) - if (isupper((unsigned char)*cp)) - *cp = _tolower((unsigned char)*cp); putkeys(mpage, value, strlen(value), type); } @@ -1622,7 +1585,12 @@ parse_man(struct mpage *mpage, const struct roff_meta while (' ' == *start) start++; - mpage->desc = mandoc_strdup(start); + /* + * Cut off excessive one-line descriptions. + * Bad pages are not worth better heuristics. + */ + + mpage->desc = mandoc_strndup(start, 150); free(title); return; } @@ -1639,26 +1607,30 @@ static void parse_mdoc(struct mpage *mpage, const struct roff_meta *meta, const struct roff_node *n) { + const struct mdoc_handler *handler; - assert(NULL != n); - for (n = n->child; NULL != n; n = n->next) { - if (n->flags & mdocs[n->tok].taboo) + for (n = n->child; n != NULL; n = n->next) { + if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX) continue; + assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX); + handler = mdoc_handlers + (n->tok - MDOC_Dd); + if (n->flags & handler->taboo) + continue; + switch (n->type) { case ROFFT_ELEM: case ROFFT_BLOCK: case ROFFT_HEAD: case ROFFT_BODY: case ROFFT_TAIL: - if (NULL != mdocs[n->tok].fp) - if (0 == (*mdocs[n->tok].fp)(mpage, meta, n)) - break; - if (mdocs[n->tok].mask) + if (handler->fp != NULL && + (*handler->fp)(mpage, meta, n) == 0) + break; + if (handler->mask) putmdockey(mpage, n->child, - mdocs[n->tok].mask, mdocs[n->tok].taboo); + handler->mask, handler->taboo); break; default: - assert(n->type != ROFFT_ROOT); continue; } if (NULL != n->child) @@ -1667,6 +1639,20 @@ parse_mdoc(struct mpage *mpage, const struct roff_meta } static int +parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) +{ + uint64_t mask; + + mask = TYPE_Fa; + if (n->sec == SEC_SYNOPSIS) + mask |= TYPE_Vt; + + putmdockey(mpage, n->child, mask, 0); + return 0; +} + +static int parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta, const struct roff_node *n) { @@ -1735,15 +1721,20 @@ static int parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta, const struct roff_node *n) { + uint64_t mask; if (n->child == NULL) return 0; parse_mdoc_fname(mpage, n->child); - for (n = n->child->next; n != NULL; n = n->next) - if (n->type == ROFFT_TEXT) - putkey(mpage, n->string, TYPE_Fa); + n = n->child->next; + if (n != NULL && n->type == ROFFT_TEXT) { + mask = TYPE_Fa; + if (n->sec == SEC_SYNOPSIS) + mask |= TYPE_Vt; + putmdockey(mpage, n, mask, 0); + } return 0; } @@ -1886,7 +1877,7 @@ putkeys(const struct mpage *mpage, char *cp, size_t sz } else { htab = &strings; if (debug > 1) - for (i = 0; i < mansearch_keymax; i++) + for (i = 0; i < KEY_MAX; i++) if ((uint64_t)1 << i & v) say(mpage->mlinks->file, "Adding key %s=%*s", @@ -2092,53 +2083,24 @@ render_string(char **public, size_t *psz) static void dbadd_mlink(const struct mlink *mlink) { - size_t i; - - i = 1; - SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec); - SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch); - SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name); - SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid); - SQL_STEP(stmts[STMT_INSERT_LINK]); - sqlite3_reset(stmts[STMT_INSERT_LINK]); + 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); } -static void -dbadd_mlink_name(const struct mlink *mlink) -{ - uint64_t bits; - size_t i; - - dbadd_mlink(mlink); - - i = 1; - SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid); - bits = NAME_FILE & NAME_MASK; - if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) { - bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0); - sqlite3_reset(stmts[STMT_SELECT_NAME]); - } - - i = 1; - SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits); - SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name); - SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid); - SQL_STEP(stmts[STMT_INSERT_NAME]); - sqlite3_reset(stmts[STMT_INSERT_NAME]); -} - /* * 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, handle escape sequences at the last possible moment. */ static void -dbadd(struct mpage *mpage) +dbadd(struct dba *dba, struct mpage *mpage) { struct mlink *mlink; struct str *key; char *cp; + uint64_t mask; size_t i; unsigned int slot; int mustfree; @@ -2183,305 +2145,164 @@ dbadd(struct mpage *mpage) cp = mpage->desc; i = strlen(cp); mustfree = render_string(&cp, &i); - i = 1; - SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp); - SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form); - SQL_STEP(stmts[STMT_INSERT_PAGE]); - mpage->pageid = sqlite3_last_insert_rowid(db); - sqlite3_reset(stmts[STMT_INSERT_PAGE]); + 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 (NULL != mlink) { + while (mlink != NULL) { dbadd_mlink(mlink); mlink = mlink->next; } - mlink = mpage->mlinks; for (key = ohash_first(&names, &slot); NULL != key; key = ohash_next(&names, &slot)) { assert(key->mpage == mpage); - i = 1; - SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask); - SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key); - SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid); - SQL_STEP(stmts[STMT_INSERT_NAME]); - sqlite3_reset(stmts[STMT_INSERT_NAME]); + 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); - i = 1; - SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask); - SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key); - SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid); - SQL_STEP(stmts[STMT_INSERT_KEY]); - sqlite3_reset(stmts[STMT_INSERT_KEY]); + 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); } } static void -dbprune(void) +dbprune(struct dba *dba) { - struct mpage *mpage; - struct mlink *mlink; - size_t i; - unsigned int slot; + struct dba_array *page, *files; + char *file; - if (0 == nodb) - SQL_EXEC("BEGIN TRANSACTION"); - - for (mpage = ohash_first(&mpages, &slot); NULL != mpage; - mpage = ohash_next(&mpages, &slot)) { - mlink = mpage->mlinks; - if (debug) - say(mlink->file, "Deleting from database"); - if (nodb) - continue; - for ( ; NULL != mlink; mlink = mlink->next) { - i = 1; - SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE], - i, mlink->dsec); - SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE], - i, mlink->arch); - SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE], - i, mlink->name); - SQL_STEP(stmts[STMT_DELETE_PAGE]); - sqlite3_reset(stmts[STMT_DELETE_PAGE]); + 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; + } } } - - if (0 == nodb) - SQL_EXEC("END TRANSACTION"); } /* - * 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; - int status; - pid_t child; + struct stat sb1, sb2; + char tfn[33], *cp1, *cp2; + off_t i; + int fd1, fd2; - if (nodb) - return; + /* + * Do not write empty databases, and delete existing ones + * when makewhatis -u causes them to become empty. + */ - for (i = 0; i < STMT__MAX; i++) { - sqlite3_finalize(stmts[i]); - stmts[i] = NULL; + dba_array_start(dba->pages); + if (dba_array_next(dba->pages) == NULL) { + if (unlink(MANDOC_DB) == -1 && errno != ENOENT) + say(MANDOC_DB, "&unlink"); + return; } - sqlite3_close(db); - db = NULL; + /* + * Build the database in a temporary file, + * then atomically move it into place. + */ - if (real) - return; - - if ('\0' == *tempfilename) { - if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) { + if (dba_write(MANDOC_DB "~", dba) != -1) { + if (rename(MANDOC_DB "~", MANDOC_DB) == -1) { exitcode = (int)MANDOCLEVEL_SYSERR; say(MANDOC_DB, "&rename"); + unlink(MANDOC_DB "~"); } return; } - switch (child = fork()) { - case -1: - exitcode = (int)MANDOCLEVEL_SYSERR; - say("", "&fork cmp"); - return; - case 0: - execlp("cmp", "cmp", "-s", - tempfilename, MANDOC_DB, (char *)NULL); - say("", "&exec cmp"); - exit(0); - default: - break; - } - if (-1 == waitpid(child, &status, 0)) { - 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"); - } + /* + * We lack write permission and cannot replace the database + * file, but let's at least check whether the data changed. + */ - *strrchr(tempfilename, '/') = '\0'; - switch (child = fork()) { - case -1: + (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn)); + if (mkdtemp(tfn) == NULL) { exitcode = (int)MANDOCLEVEL_SYSERR; - say("", "&fork rm"); + say("", "&%s", tfn); return; - case 0: - execlp("rm", "rm", "-rf", tempfilename, (char *)NULL); - say("", "&exec rm"); - exit((int)MANDOCLEVEL_SYSERR); - default: - break; } - if (-1 == waitpid(child, &status, 0)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say("", "&wait rm"); - } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say("", "%s: Cannot remove temporary directory", - tempfilename); + cp1 = cp2 = MAP_FAILED; + fd1 = fd2 = -1; + (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn)); + if (dba_write(tfn, dba) == -1) { + say(tfn, "&dba_write"); + goto err; } -} - -/* - * 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 *sql; - int rc, ofl; - - if (nodb) - return 1; - - *tempfilename = '\0'; - ofl = SQLITE_OPEN_READWRITE; - - if (real) { - rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL); - if (SQLITE_OK != rc) { - exitcode = (int)MANDOCLEVEL_SYSERR; - if (SQLITE_CANTOPEN != rc) - say(MANDOC_DB, "%s", sqlite3_errstr(rc)); - return 0; - } - goto prepare_statements; + if ((fd1 = open(MANDOC_DB, O_RDONLY)) == -1) { + say(MANDOC_DB, "&open"); + goto err; } - - ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE; - - remove(MANDOC_DB "~"); - rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL); - if (SQLITE_OK == rc) - goto create_tables; - if (MPARSE_QUICK & mparse_options) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say(MANDOC_DB "~", "%s", sqlite3_errstr(rc)); - return 0; + if ((fd2 = open(tfn, O_RDONLY)) == -1) { + say(tfn, "&open"); + goto err; } - - (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX", - sizeof(tempfilename)); - if (NULL == mkdtemp(tempfilename)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say("", "&%s", tempfilename); - return 0; + if (fstat(fd1, &sb1) == -1) { + say(MANDOC_DB, "&fstat"); + goto err; } - (void)strlcat(tempfilename, "/" MANDOC_DB, - sizeof(tempfilename)); - rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL); - if (SQLITE_OK != rc) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say("", "%s: %s", tempfilename, sqlite3_errstr(rc)); - return 0; + if (fstat(fd2, &sb2) == -1) { + say(tfn, "&fstat"); + goto err; } - -create_tables: - sql = "CREATE TABLE \"mpages\" (\n" - " \"desc\" TEXT NOT NULL,\n" - " \"form\" INTEGER NOT NULL,\n" - " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n" - ");\n" - "\n" - "CREATE TABLE \"mlinks\" (\n" - " \"sec\" TEXT NOT NULL,\n" - " \"arch\" TEXT NOT NULL,\n" - " \"name\" TEXT NOT NULL,\n" - " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) " - "ON DELETE CASCADE\n" - ");\n" - "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n" - "\n" - "CREATE TABLE \"names\" (\n" - " \"bits\" INTEGER NOT NULL,\n" - " \"name\" TEXT NOT NULL,\n" - " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) " - "ON DELETE CASCADE,\n" - " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n" - ");\n" - "\n" - "CREATE TABLE \"keys\" (\n" - " \"bits\" INTEGER NOT NULL,\n" - " \"key\" TEXT NOT NULL,\n" - " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) " - "ON DELETE CASCADE\n" - ");\n" - "CREATE INDEX keys_pageid_idx ON keys (pageid);\n"; - - if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say(MANDOC_DB, "%s", sqlite3_errmsg(db)); - sqlite3_close(db); - return 0; + if (sb1.st_size != sb2.st_size) + goto err; + if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE, + fd1, 0)) == MAP_FAILED) { + say(MANDOC_DB, "&mmap"); + goto err; } - -prepare_statements: - if (SQLITE_OK != sqlite3_exec(db, - "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say(MANDOC_DB, "PRAGMA foreign_keys: %s", - sqlite3_errmsg(db)); - sqlite3_close(db); - return 0; + if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE, + fd2, 0)) == MAP_FAILED) { + say(tfn, "&mmap"); + goto err; } + for (i = 0; i < sb1.st_size; i++) + if (cp1[i] != cp2[i]) + goto err; + goto out; - sql = "DELETE FROM mpages WHERE pageid IN " - "(SELECT pageid FROM mlinks WHERE " - "sec=? AND arch=? AND name=?)"; - sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL); - sql = "INSERT INTO mpages " - "(desc,form) VALUES (?,?)"; - sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL); - sql = "INSERT INTO mlinks " - "(sec,arch,name,pageid) VALUES (?,?,?,?)"; - sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL); - sql = "SELECT bits FROM names where pageid = ?"; - sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL); - sql = "INSERT INTO names " - "(bits,name,pageid) VALUES (?,?,?)"; - sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL); - sql = "INSERT INTO keys " - "(bits,key,pageid) VALUES (?,?,?)"; - sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL); +err: + exitcode = (int)MANDOCLEVEL_SYSERR; + say(MANDOC_DB, "Data changed, but cannot replace database"); -#ifndef __APPLE__ - /* - * When opening a new database, we can turn off - * synchronous mode for much better performance. - */ - - if (real && SQLITE_OK != sqlite3_exec(db, - "PRAGMA synchronous = OFF", NULL, NULL, NULL)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say(MANDOC_DB, "PRAGMA synchronous: %s", - sqlite3_errmsg(db)); - sqlite3_close(db); - return 0; - } -#endif - - return 1; +out: + if (cp1 != MAP_FAILED) + munmap(cp1, sb1.st_size); + if (cp2 != MAP_FAILED) + munmap(cp2, sb2.st_size); + if (fd1 != -1) + close(fd1); + if (fd2 != -1) + close(fd2); + unlink(tfn); + *strrchr(tfn, '/') = '\0'; + rmdir(tfn); } static int @@ -2490,7 +2311,6 @@ set_basedir(const char *targetdir, int report_baddir) static char startdir[PATH_MAX]; static int getcwd_status; /* 1 = ok, 2 = failure */ static int chdir_status; /* 1 = changed directory */ - char *cp; /* * Remember the original working directory, if possible. @@ -2499,8 +2319,8 @@ set_basedir(const char *targetdir, int report_baddir) * Do not error out if the current directory is not * searchable: Maybe it won't be needed after all. */ - if (0 == getcwd_status) { - if (NULL == getcwd(startdir, sizeof(startdir))) { + if (getcwd_status == 0) { + if (getcwd(startdir, sizeof(startdir)) == NULL) { getcwd_status = 2; (void)strlcpy(startdir, strerror(errno), sizeof(startdir)); @@ -2513,19 +2333,20 @@ set_basedir(const char *targetdir, int report_baddir) * Do not use it any longer, not even for messages. */ *basedir = '\0'; + basedir_len = 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) { + if (chdir_status && *targetdir != '/') { + if (getcwd_status == 2) { exitcode = (int)MANDOCLEVEL_SYSERR; say("", "getcwd: %s", startdir); return 0; } - if (-1 == chdir(startdir)) { + if (chdir(startdir) == -1) { exitcode = (int)MANDOCLEVEL_SYSERR; say("", "&chdir %s", startdir); return 0; @@ -2537,48 +2358,71 @@ set_basedir(const char *targetdir, int report_baddir) * pathname and append a trailing slash, such that * we can reliably check whether files are inside. */ - if (NULL == realpath(targetdir, basedir)) { + if (realpath(targetdir, basedir) == NULL) { if (report_baddir || errno != ENOENT) { exitcode = (int)MANDOCLEVEL_BADARG; say("", "&%s: realpath", targetdir); } + *basedir = '\0'; return 0; - } else if (-1 == chdir(basedir)) { + } else if (chdir(basedir) == -1) { if (report_baddir || errno != ENOENT) { exitcode = (int)MANDOCLEVEL_BADARG; say("", "&chdir"); } + *basedir = '\0'; return 0; } chdir_status = 1; - cp = strchr(basedir, '\0'); - if ('/' != cp[-1]) { - if (cp - basedir >= PATH_MAX - 1) { + basedir_len = strlen(basedir); + if (basedir[basedir_len - 1] != '/') { + if (basedir_len >= PATH_MAX - 1) { exitcode = (int)MANDOCLEVEL_SYSERR; say("", "Filename too long"); + *basedir = '\0'; + basedir_len = 0; return 0; } - *cp++ = '/'; - *cp = '\0'; + basedir[basedir_len++] = '/'; + basedir[basedir_len] = '\0'; } return 1; } +#ifdef READ_ALLOWED_PATH +static int +read_allowed(const char *candidate) +{ + const char *cp; + size_t len; + + for (cp = READ_ALLOWED_PATH;; cp += len) { + while (*cp == ':') + cp++; + if (*cp == '\0') + return 0; + len = strcspn(cp, ":"); + if (strncmp(candidate, cp, len) == 0) + return 1; + } +} +#endif + static void say(const char *file, const char *format, ...) { va_list ap; int use_errno; - if ('\0' != *basedir) + if (*basedir != '\0') fprintf(stderr, "%s", basedir); - if ('\0' != *basedir && '\0' != *file) + if (*basedir != '\0' && *file != '\0') fputc('/', stderr); - if ('\0' != *file) + if (*file != '\0') fprintf(stderr, "%s", file); use_errno = 1; - if (NULL != format) { + if (format != NULL) { switch (*format) { case '&': format++; @@ -2591,15 +2435,15 @@ say(const char *file, const char *format, ...) break; } } - if (NULL != format) { - if ('\0' != *basedir || '\0' != *file) + if (format != NULL) { + if (*basedir != '\0' || *file != '\0') fputs(": ", stderr); va_start(ap, format); vfprintf(stderr, format, ap); va_end(ap); } if (use_errno) { - if ('\0' != *basedir || '\0' != *file || NULL != format) + if (*basedir != '\0' || *file != '\0' || format != NULL) fputs(": ", stderr); perror(NULL); } else