=================================================================== RCS file: /cvs/mandoc/mandocdb.c,v retrieving revision 1.136 retrieving revision 1.165 diff -u -p -r1.136 -r1.165 --- mandoc/mandocdb.c 2014/04/16 18:59:43 1.136 +++ mandoc/mandocdb.c 2014/09/09 19:35:25 1.165 @@ -1,4 +1,4 @@ -/* $Id: mandocdb.c,v 1.136 2014/04/16 18:59:43 schwarze Exp $ */ +/* $Id: mandocdb.c,v 1.165 2014/09/09 19:35:25 schwarze Exp $ */ /* * Copyright (c) 2011, 2012 Kristaps Dzonsons * Copyright (c) 2011, 2012, 2013, 2014 Ingo Schwarze @@ -15,10 +15,9 @@ * 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 #include #include @@ -26,7 +25,11 @@ #include #include #include +#if HAVE_FTS #include +#else +#include "compat_fts.h" +#endif #include #include #include @@ -36,7 +39,7 @@ #include #include -#ifdef HAVE_OHASH +#if HAVE_OHASH #include #else #include "compat_ohash.h" @@ -80,12 +83,6 @@ enum op { OP_TEST /* change no databases, report potential problems */ }; -enum form { - FORM_NONE, /* format is unknown */ - FORM_SRC, /* format is -man or -mdoc */ - FORM_CAT /* format is cat */ -}; - struct str { char *rendered; /* key in UTF-8 or ASCII form */ const struct mpage *mpage; /* if set, the owning parse */ @@ -100,25 +97,25 @@ struct inodev { struct mpage { struct inodev inodev; /* used for hashing routine */ - int64_t recno; /* id in mpages SQL table */ - enum form form; /* format from file content */ + int64_t pageid; /* pageid in mpages SQL table */ 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 mlink *mlinks; /* singly linked list */ + int form; /* format from file content */ }; struct mlink { char file[PATH_MAX]; /* filename rel. to manpath */ - enum form dform; /* format from directory */ - enum form fform; /* format from file name suffix */ 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 dform; /* format from directory */ + int fform; /* format from file name suffix */ int gzip; /* filename has a .gz suffix */ }; @@ -145,8 +142,8 @@ static int dbopen(int); static void dbprune(void); 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 hash_free(void *, void *); +static void *hash_calloc(size_t, size_t, void *); static void mlink_add(struct mlink *, const struct stat *); static void mlink_check(struct mpage *, struct mlink *); static void mlink_free(struct mlink *); @@ -172,7 +169,7 @@ static void putmdockey(const struct mpage *, const struct mdoc_node *, uint64_t); static void render_key(struct mchars *, struct str *); static void say(const char *, const char *, ...); -static int set_basedir(const char *); +static int set_basedir(const char *, int); static int treescan(void); static size_t utf8(unsigned int, char [7]); @@ -180,12 +177,12 @@ static char tempfilename[32]; static char *progname; static int nodb; /* no database changes */ 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 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 ohash mpages; /* table of distinct manual pages */ static struct ohash mlinks; /* table of directory entries */ @@ -320,6 +317,7 @@ static const struct mdoc_handler mdocs[MDOC_MAX] = { { NULL, 0 }, /* Ta */ }; + int main(int argc, char *argv[]) { @@ -335,8 +333,8 @@ main(int argc, char *argv[]) memset(&dirs, 0, sizeof(struct manpaths)); mpages_info.alloc = mlinks_info.alloc = hash_alloc; - mpages_info.halloc = mlinks_info.halloc = hash_halloc; - mpages_info.hfree = mlinks_info.hfree = hash_free; + mpages_info.calloc = mlinks_info.calloc = hash_calloc; + mpages_info.free = mlinks_info.free = hash_free; mpages_info.key_offset = offsetof(struct mpage, inodev); mlinks_info.key_offset = offsetof(struct mlink, file); @@ -348,13 +346,14 @@ main(int argc, char *argv[]) ++progname; /* - * 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)); \ + fprintf(stderr, "%s: -%c: Conflicting option\n", \ + progname, (_ch)); \ goto usage; \ } while (/*CONSTCOND*/0) @@ -363,51 +362,52 @@ main(int argc, char *argv[]) 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'): + case 'd': CHECKOP(op, ch); path_arg = optarg; op = OP_UPDATE; break; - case ('n'): + case 'n': nodb = 1; break; - case ('p'): + case 'p': warnings = 1; break; - case ('Q'): + case 'Q': mparse_options |= MPARSE_QUICK; break; - case ('T'): + case 'T': if (strcmp(optarg, "utf8")) { - fprintf(stderr, "-T%s: Unsupported " - "output format\n", optarg); + fprintf(stderr, "%s: -T%s: " + "Unsupported output format\n", + progname, optarg); goto usage; } write_utf8 = 1; break; - case ('t'): + 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'): + case 'v': /* Compatibility with espie@'s makewhatis. */ break; default: @@ -418,7 +418,8 @@ main(int argc, char *argv[]) argv += optind; if (OP_CONFFILE == op && argc > 0) { - fprintf(stderr, "-C: Too many arguments\n"); + fprintf(stderr, "%s: -C: Too many arguments\n", + progname); goto usage; } @@ -430,27 +431,39 @@ main(int argc, char *argv[]) ohash_init(&mlinks, 6, &mlinks_info); 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(); + + if (dbopen(1)) { + /* + * The existing database is usable. Process + * all files specified on the command-line. + */ + use_all = 1; + for (i = 0; i < argc; i++) + filescan(argv[i]); + if (OP_TEST != op) + dbprune(); + } else { + /* + * Database missing or corrupt. + * Recreate from scratch. + */ + exitcode = (int)MANDOCLEVEL_OK; + op = OP_DEFAULT; + if (0 == treescan()) + goto out; + if (0 == dbopen(0)) + goto out; + } if (OP_DELETE != op) mpages_merge(mc, mp); - dbclose(1); + dbclose(OP_DEFAULT == op ? 0 : 1); } else { /* * If we have arguments, use them as our manpaths. @@ -458,8 +471,8 @@ main(int argc, char *argv[]) * manpath_parse() wants to do it. */ if (argc > 0) { - dirs.paths = mandoc_calloc - (argc, sizeof(char *)); + dirs.paths = mandoc_reallocarray(NULL, + argc, sizeof(char *)); dirs.sz = (size_t)argc; for (i = 0; i < argc; i++) dirs.paths[i] = mandoc_strdup(argv[i]); @@ -489,17 +502,15 @@ main(int argc, char *argv[]) ohash_init(&mlinks, 6, &mlinks_info); } - if (0 == set_basedir(dirs.paths[j])) - goto out; + if (0 == set_basedir(dirs.paths[j], argc > 0)) + continue; if (0 == treescan()) - goto out; - if (0 == set_basedir(dirs.paths[j])) - goto out; + continue; if (0 == dbopen(0)) - goto out; + continue; mpages_merge(mc, mp); - if (warnings && + if (warnings && !nodb && ! (MPARSE_QUICK & mparse_options)) names_check(); dbclose(0); @@ -512,7 +523,6 @@ main(int argc, char *argv[]) } } out: - set_basedir(NULL); manpath_free(&dirs); mchars_free(mc); mparse_free(mp); @@ -526,7 +536,7 @@ usage: " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n" " %s [-Dnp] -u dir [file ...]\n" " %s [-Q] -t file ...\n", - progname, progname, progname, + progname, progname, progname, progname, progname); return((int)MANDOCLEVEL_BADARG); @@ -540,7 +550,7 @@ usage: * If use_all has been specified, grok all files. * If not, sanitise paths to the following: * - * [./]man*[/]/.
+ * [./]man*[/]/.
* or * [./]cat
[/]/.0 * @@ -549,6 +559,7 @@ usage: static int treescan(void) { + char buf[PATH_MAX]; FTS *f; FTSENT *ff; struct mlink *mlink; @@ -560,11 +571,8 @@ treescan(void) argv[0] = "."; argv[1] = (char *)NULL; - /* - * Walk through all components under the directory, using the - * logical descent of files. - */ - f = fts_open((char * const *)argv, FTS_LOGICAL, NULL); + f = fts_open((char * const *)argv, + FTS_PHYSICAL | FTS_NOCHDIR, NULL); if (NULL == f) { exitcode = (int)MANDOCLEVEL_SYSERR; say("", "&fts_open"); @@ -576,11 +584,36 @@ treescan(void) while (NULL != (ff = fts_read(f))) { path = ff->fts_path + 2; + switch (ff->fts_info) { + /* + * Symbolic links require various sanity checks, + * then get handled just like regular files. + */ + case FTS_SL: + if (NULL == realpath(path, buf)) { + if (warnings) + say(path, "&realpath"); + continue; + } + if (strstr(buf, basedir) != buf) { + if (warnings) say("", + "%s: outside base directory", buf); + continue; + } + /* Use logical inode to avoid mpages dupe. */ + if (-1 == stat(path, ff->fts_statp)) { + 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. */ - if (FTS_F == ff->fts_info) { + case FTS_F: if (0 == strcmp(path, MANDOC_DB)) continue; if ( ! use_all && ff->fts_level < 2) { @@ -627,7 +660,13 @@ treescan(void) fsec[-1] = '\0'; mlink = mandoc_calloc(1, sizeof(struct mlink)); - strlcpy(mlink->file, path, sizeof(mlink->file)); + 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; @@ -636,18 +675,23 @@ treescan(void) 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: + /* FALLTHROUGH */ + 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. @@ -668,14 +712,14 @@ treescan(void) dsec = NULL; } - if (NULL != dsec || use_all) + if (NULL != dsec || 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. @@ -707,7 +751,7 @@ treescan(void) * Try to infer the manual section, architecture, and page name from the * path, assuming it looks like * - * [./]man*[/]/.
+ * [./]man*[/]/.
* or * [./]cat
[/]/.0 * @@ -726,34 +770,74 @@ filescan(const char *file) if (0 == strncmp(file, "./", 2)) file += 2; + /* + * 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, "&lstat"); + return; + } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) { + exitcode = (int)MANDOCLEVEL_BADARG; + say(file, "Not a regular file"); + return; + } + + /* + * 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, "&realpath"); return; } - if (strstr(buf, basedir) == buf) - start = buf + strlen(basedir) + 1; - else if (OP_TEST == op) + if (OP_TEST == op) start = buf; + else if (strstr(buf, basedir) == buf) + start = buf + strlen(basedir); else { exitcode = (int)MANDOCLEVEL_BADARG; say("", "%s: outside base directory", buf); return; } - if (-1 == stat(buf, &st)) { - exitcode = (int)MANDOCLEVEL_BADARG; - say(file, "&stat"); - return; - } else if ( ! (S_IFREG & st.st_mode)) { - exitcode = (int)MANDOCLEVEL_BADARG; - say(file, "Not a regular file"); - return; + /* + * 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)); - strlcpy(mlink->file, start, sizeof(mlink->file)); + mlink->dform = FORM_NONE; + if (strlcpy(mlink->file, start, sizeof(mlink->file)) >= + sizeof(mlink->file)) { + say(start, "Filename too long"); + return; + } /* * First try to guess our directory structure. @@ -901,19 +985,15 @@ mlinks_undupe(struct mpage *mpage) mpage->form = FORM_NONE; goto nextlink; } - if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) { - if (warnings) - say(mlink->file, "Filename too long"); - 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'; - strlcat(buf, mlink->dsec, PATH_MAX); + (void)strlcat(buf, mlink->dsec, sizeof(buf)); if (NULL == ohash_find(&mlinks, - ohash_qlookup(&mlinks, buf))) + ohash_qlookup(&mlinks, buf))) goto nextlink; if (warnings) say(mlink->file, "Man source exists: %s", buf); @@ -996,7 +1076,6 @@ mpages_merge(struct mchars *mc, struct mparse *mp) { char any[] = "any"; struct ohash_info str_info; - int fd[2]; struct mpage *mpage, *mpage_dest; struct mlink *mlink, *mlink_dest; struct mdoc *mdoc; @@ -1004,13 +1083,13 @@ mpages_merge(struct mchars *mc, struct mparse *mp) char *sodest; char *cp; pid_t child_pid; - int status; + int fd; unsigned int pslot; enum mandoclevel lvl; str_info.alloc = hash_alloc; - str_info.halloc = hash_halloc; - str_info.hfree = hash_free; + str_info.calloc = hash_calloc; + str_info.free = hash_free; str_info.key_offset = offsetof(struct str, key); if (0 == nodb) @@ -1032,38 +1111,11 @@ mpages_merge(struct mchars *mc, struct mparse *mp) man = NULL; sodest = NULL; child_pid = 0; - fd[0] = -1; - fd[1] = -1; - if (mpage->mlinks->gzip) { - if (-1 == pipe(fd)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say(mpage->mlinks->file, "&pipe gunzip"); - goto nextpage; - } - switch (child_pid = fork()) { - case (-1): - exitcode = (int)MANDOCLEVEL_SYSERR; - say(mpage->mlinks->file, "&fork gunzip"); - child_pid = 0; - close(fd[1]); - close(fd[0]); - goto nextpage; - case (0): - close(fd[0]); - if (-1 == dup2(fd[1], STDOUT_FILENO)) { - say(mpage->mlinks->file, - "&dup gunzip"); - exit(1); - } - execlp("gunzip", "gunzip", "-c", - mpage->mlinks->file, NULL); - say(mpage->mlinks->file, "&exec gunzip"); - exit(1); - default: - close(fd[1]); - break; - } + mparse_open(mp, &fd, mpage->mlinks->file, &child_pid); + if (fd == -1) { + say(mpage->mlinks->file, "&open"); + goto nextpage; } /* @@ -1073,7 +1125,7 @@ mpages_merge(struct mchars *mc, struct mparse *mp) */ if (FORM_CAT != mpage->mlinks->dform || FORM_CAT != mpage->mlinks->fform) { - lvl = mparse_readfd(mp, fd[0], mpage->mlinks->file); + lvl = mparse_readfd(mp, fd, mpage->mlinks->file); if (lvl < MANDOCLEVEL_FATAL) mparse_result(mp, &mdoc, &man, &sodest); } @@ -1099,7 +1151,7 @@ mpages_merge(struct mchars *mc, struct mparse *mp) * to the target. */ - if (mpage_dest->recno) + if (mpage_dest->pageid) dbadd_mlink(mlink); if (NULL == mlink->next) @@ -1116,8 +1168,9 @@ mpages_merge(struct mchars *mc, struct mparse *mp) goto nextpage; } else if (NULL != mdoc) { mpage->form = FORM_SRC; - mpage->sec = - mandoc_strdup(mdoc_meta(mdoc)->msec); + mpage->sec = mdoc_meta(mdoc)->msec; + mpage->sec = mandoc_strdup( + NULL == mpage->sec ? "" : mpage->sec); mpage->arch = mdoc_meta(mdoc)->arch; mpage->arch = mandoc_strdup( NULL == mpage->arch ? "" : mpage->arch); @@ -1140,9 +1193,11 @@ mpages_merge(struct mchars *mc, struct mparse *mp) mpage->title = mandoc_strdup(mpage->mlinks->name); } + if (mpage->mlinks->gzip) + mpage->form |= FORM_GZ; putkey(mpage, mpage->sec, TYPE_sec); - putkey(mpage, '\0' == *mpage->arch ? - any : mpage->arch, TYPE_arch); + if (*mpage->arch != '\0') + putkey(mpage, mpage->arch, TYPE_arch); for (mlink = mpage->mlinks; mlink; mlink = mlink->next) { if ('\0' != *mlink->dsec) @@ -1162,7 +1217,7 @@ mpages_merge(struct mchars *mc, struct mparse *mp) } else if (NULL != man) parse_man(mpage, man_node(man)); else - parse_cat(mpage, fd[0]); + parse_cat(mpage, fd); if (NULL == mpage->desc) mpage->desc = mandoc_strdup(mpage->mlinks->name); @@ -1174,21 +1229,10 @@ mpages_merge(struct mchars *mc, struct mparse *mp) dbadd(mpage, mc); nextpage: - if (child_pid) { - if (-1 == waitpid(child_pid, &status, 0)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say(mpage->mlinks->file, "&wait gunzip"); - } else if (WIFSIGNALED(status)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say(mpage->mlinks->file, - "gunzip died from signal %d", - WTERMSIG(status)); - } else if (WEXITSTATUS(status)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say(mpage->mlinks->file, - "gunzip failed with code %d", - WEXITSTATUS(status)); - } + if (child_pid && + mparse_wait(mp, child_pid) != MANDOCLEVEL_OK) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say(mpage->mlinks->file, "&wait gunzip"); } ohash_delete(&strings); ohash_delete(&names); @@ -1215,7 +1259,8 @@ names_check(void) "AND mlinks.name == names.name" ")" ") JOIN (" - "SELECT * FROM mlinks GROUP BY pageid" + "SELECT sec, arch, name, pageid FROM mlinks " + "GROUP BY pageid" ") USING (pageid);", -1, &stmt, NULL); @@ -1223,10 +1268,10 @@ names_check(void) say("", "%s", sqlite3_errmsg(db)); while (SQLITE_ROW == (irc = sqlite3_step(stmt))) { - name = sqlite3_column_text(stmt, 0); - sec = sqlite3_column_text(stmt, 1); - arch = sqlite3_column_text(stmt, 2); - key = sqlite3_column_text(stmt, 3); + 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); @@ -1245,6 +1290,8 @@ parse_cat(struct mpage *mpage, int fd) fopen(mpage->mlinks->file, "r") : fdopen(fd, "r"); if (NULL == stream) { + if (-1 != fd) + close(fd); if (warnings) say(mpage->mlinks->file, "&fopen"); return; @@ -1264,7 +1311,7 @@ parse_cat(struct mpage *mpage, int fd) while (NULL != (line = fgetln(stream, &len))) if ('\n' != *line && ' ' != *line) break; - + /* * Read up until the next section into a buffer. * Strip the leading and trailing newline from each read line, @@ -1334,7 +1381,7 @@ parse_cat(struct mpage *mpage, int fd) if (0 == len) { memmove(line, line + 1, plen--); continue; - } + } memmove(line - 1, line + 1, plen - len); plen -= 2; } @@ -1398,11 +1445,11 @@ parse_man(struct mpage *mpage, const struct man_node * 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->child) { + 1 == head->nchild && + NULL != (head = (head->child)) && + MAN_TEXT == head->type && + 0 == strcmp(head->string, "NAME") && + NULL != body->child) { /* * Suck the entire NAME section into memory. @@ -1416,7 +1463,7 @@ parse_man(struct mpage *mpage, const struct man_node * if (NULL == title) return; - /* + /* * Go through a special heuristic dance here. * Conventionally, one or more manual names are * comma-specified prior to a whitespace, then a @@ -1498,15 +1545,15 @@ parse_mdoc(struct mpage *mpage, const struct mdoc_node assert(NULL != n); for (n = n->child; NULL != n; n = n->next) { switch (n->type) { - case (MDOC_ELEM): + case MDOC_ELEM: /* FALLTHROUGH */ - case (MDOC_BLOCK): + case MDOC_BLOCK: /* FALLTHROUGH */ - case (MDOC_HEAD): + case MDOC_HEAD: /* FALLTHROUGH */ - case (MDOC_BODY): + case MDOC_BODY: /* FALLTHROUGH */ - case (MDOC_TAIL): + case MDOC_TAIL: if (NULL != mdocs[n->tok].fp) if (0 == (*mdocs[n->tok].fp)(mpage, n)) break; @@ -1530,8 +1577,8 @@ parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_n size_t sz; if (SEC_SYNOPSIS != n->sec || - NULL == (n = n->child) || - MDOC_TEXT != n->type) + NULL == (n = n->child) || + MDOC_TEXT != n->type) return(0); /* @@ -1574,11 +1621,11 @@ parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_n if (NULL == (n = n->child) || MDOC_TEXT != n->type) return(0); - /* + /* * Parse: .Fn "struct type *name" "char *arg". - * First strip away pointer symbol. + * First strip away pointer symbol. * Then store the function name, then type. - * Finally, store the arguments. + * Finally, store the arguments. */ if (NULL == (cp = strrchr(n->string, ' '))) @@ -1688,7 +1735,7 @@ putkeys(const struct mpage *mpage, htab = &strings; if (debug > 1) for (i = 0; i < mansearch_keymax; i++) - if (1 << i & v) + if ((uint64_t)1 << i & v) say(mpage->mlinks->file, "Adding key %s=%*s", mansearch_keynames[i], sz, cp); @@ -1702,7 +1749,7 @@ putkeys(const struct mpage *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(htab, slot, s); } @@ -1795,7 +1842,7 @@ render_key(struct mchars *mc, struct str *key) if (strcspn(val, res) == bsz) { key->rendered = key->key; return; - } + } /* Pre-allocate by the length of the input */ @@ -1815,17 +1862,17 @@ render_key(struct mchars *mc, struct str *key) } switch (*val) { - case (ASCII_HYPH): + case ASCII_HYPH: buf[pos++] = '-'; val++; continue; - case ('\t'): + case '\t': /* FALLTHROUGH */ - case (ASCII_NBRSP): + case ASCII_NBRSP: buf[pos++] = ' '; val++; /* FALLTHROUGH */ - case (ASCII_BREAK): + case ASCII_BREAK: continue; default: break; @@ -1842,8 +1889,8 @@ render_key(struct mchars *mc, struct str *key) * predefined character or special character. */ - esc = mandoc_escape - ((const char **)&val, &seq, &len); + esc = mandoc_escape((const char **)&val, + &seq, &len); if (ESCAPE_ERROR == esc) break; if (ESCAPE_SPECIAL != esc) @@ -1892,9 +1939,16 @@ dbadd_mlink(const struct mlink *mlink) 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->recno); + SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid); SQL_STEP(stmts[STMT_INSERT_LINK]); sqlite3_reset(stmts[STMT_INSERT_LINK]); + + i = 1; + SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, NAME_FILE); + 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]); } /* @@ -1914,6 +1968,20 @@ dbadd(struct mpage *mpage, struct mchars *mc) mlink = mpage->mlinks; if (nodb) { + for (key = ohash_first(&names, &slot); NULL != key; + key = ohash_next(&names, &slot)) { + if (key->rendered != key->key) + free(key->rendered); + free(key); + } + for (key = ohash_first(&strings, &slot); NULL != key; + key = ohash_next(&strings, &slot)) { + if (key->rendered != key->key) + free(key->rendered); + free(key); + } + if (0 == debug) + return; while (NULL != mlink) { fputs(mlink->name, stdout); if (NULL == mlink->next || @@ -1940,13 +2008,22 @@ dbadd(struct mpage *mpage, struct mchars *mc) if (debug) say(mlink->file, "Adding to database"); + i = strlen(mpage->desc) + 1; + key = mandoc_calloc(1, sizeof(struct str) + i); + memcpy(key->key, mpage->desc, i); + render_key(mc, key); + i = 1; - SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->desc); - SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form); + SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, key->rendered); + SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form); SQL_STEP(stmts[STMT_INSERT_PAGE]); - mpage->recno = sqlite3_last_insert_rowid(db); + mpage->pageid = sqlite3_last_insert_rowid(db); sqlite3_reset(stmts[STMT_INSERT_PAGE]); + if (key->rendered != key->key) + free(key->rendered); + free(key); + while (NULL != mlink) { dbadd_mlink(mlink); mlink = mlink->next; @@ -1961,7 +2038,7 @@ dbadd(struct mpage *mpage, struct mchars *mc) i = 1; SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask); SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->rendered); - SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->recno); + SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid); SQL_STEP(stmts[STMT_INSERT_NAME]); sqlite3_reset(stmts[STMT_INSERT_NAME]); if (key->rendered != key->key) @@ -1976,7 +2053,7 @@ dbadd(struct mpage *mpage, struct mchars *mc) i = 1; SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask); SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered); - SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->recno); + SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid); SQL_STEP(stmts[STMT_INSERT_KEY]); sqlite3_reset(stmts[STMT_INSERT_KEY]); if (key->rendered != key->key) @@ -2054,11 +2131,11 @@ dbclose(int real) } switch (child = fork()) { - case (-1): + case -1: exitcode = (int)MANDOCLEVEL_SYSERR; say("", "&fork cmp"); return; - case (0): + case 0: execlp("cmp", "cmp", "-s", tempfilename, MANDOC_DB, NULL); say("", "&exec cmp"); @@ -2080,11 +2157,11 @@ dbclose(int real) *strrchr(tempfilename, '/') = '\0'; switch (child = fork()) { - case (-1): + case -1: exitcode = (int)MANDOCLEVEL_SYSERR; say("", "&fork rm"); return; - case (0): + case 0: execlp("rm", "rm", "-rf", tempfilename, NULL); say("", "&exec rm"); exit((int)MANDOCLEVEL_SYSERR); @@ -2115,7 +2192,7 @@ dbopen(int real) const char *sql; int rc, ofl; - if (nodb) + if (nodb) return(1); *tempfilename = '\0'; @@ -2125,7 +2202,8 @@ dbopen(int real) rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL); if (SQLITE_OK != rc) { exitcode = (int)MANDOCLEVEL_SYSERR; - say(MANDOC_DB, "%s", sqlite3_errmsg(db)); + if (SQLITE_CANTOPEN != rc) + say(MANDOC_DB, "%s", sqlite3_errstr(rc)); return(0); } goto prepare_statements; @@ -2135,36 +2213,27 @@ dbopen(int real) remove(MANDOC_DB "~"); rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL); - if (SQLITE_OK == rc) + if (SQLITE_OK == rc) goto create_tables; if (MPARSE_QUICK & mparse_options) { exitcode = (int)MANDOCLEVEL_SYSERR; - say(MANDOC_DB "~", "%s", sqlite3_errmsg(db)); + say(MANDOC_DB "~", "%s", sqlite3_errstr(rc)); return(0); } - if (strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX", - sizeof(tempfilename)) >= sizeof(tempfilename)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say("", "/tmp/mandocdb.XXXXXX: Filename too long"); - return(0); - } + (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX", + sizeof(tempfilename)); if (NULL == mkdtemp(tempfilename)) { exitcode = (int)MANDOCLEVEL_SYSERR; say("", "&%s", tempfilename); return(0); } - if (strlcat(tempfilename, "/" MANDOC_DB, - sizeof(tempfilename)) >= sizeof(tempfilename)) { - exitcode = (int)MANDOCLEVEL_SYSERR; - say("", "%s/" MANDOC_DB ": Filename too long", - tempfilename); - return(0); - } + (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_errmsg(db)); + say("", "%s: %s", tempfilename, sqlite3_errstr(rc)); return(0); } @@ -2172,14 +2241,14 @@ create_tables: sql = "CREATE TABLE \"mpages\" (\n" " \"desc\" TEXT NOT NULL,\n" " \"form\" INTEGER NOT NULL,\n" - " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT 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(id) " + " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) " "ON DELETE CASCADE\n" ");\n" "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n" @@ -2187,14 +2256,14 @@ create_tables: "CREATE TABLE \"names\" (\n" " \"bits\" INTEGER NOT NULL,\n" " \"name\" TEXT NOT NULL,\n" - " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) " + " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) " "ON DELETE CASCADE\n" ");\n" "\n" "CREATE TABLE \"keys\" (\n" " \"bits\" INTEGER NOT NULL,\n" " \"key\" TEXT NOT NULL,\n" - " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) " + " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) " "ON DELETE CASCADE\n" ");\n" "CREATE INDEX keys_pageid_idx ON keys (pageid);\n"; @@ -2202,12 +2271,21 @@ create_tables: 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); } prepare_statements: - SQL_EXEC("PRAGMA foreign_keys = ON"); - sql = "DELETE FROM mpages WHERE id IN " + 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); + } + + 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); @@ -2230,18 +2308,24 @@ prepare_statements: * synchronous mode for much better performance. */ - if (real) - SQL_EXEC("PRAGMA synchronous = OFF"); + 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); } static void * -hash_halloc(size_t sz, void *arg) +hash_calloc(size_t n, size_t sz, void *arg) { - return(mandoc_calloc(sz, 1)); + return(mandoc_calloc(n, sz)); } static void * @@ -2252,63 +2336,89 @@ hash_alloc(size_t sz, void *arg) } static void -hash_free(void *p, size_t sz, void *arg) +hash_free(void *p, void *arg) { free(p); } 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)) { + 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; - if (NULL != targetdir) - say("", "&getcwd"); + say("", "getcwd: %s", startdir); return(0); } - if (-1 == (fd = open(startdir, O_RDONLY, 0))) { + if (-1 == chdir(startdir)) { exitcode = (int)MANDOCLEVEL_SYSERR; - say("", "&open %s", startdir); - return(0); - } - if (NULL == targetdir) - targetdir = startdir; - } else { - if (-1 == fd) - return(0); - if (-1 == fchdir(fd)) { - close(fd); - basedir[0] = '\0'; - exitcode = (int)MANDOCLEVEL_SYSERR; say("", "&chdir %s", startdir); return(0); } - if (NULL == targetdir) { - close(fd); - return(1); - } } + + /* + * 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("", "&%s: realpath", targetdir); + 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("", "&chdir"); + if (report_baddir || errno != ENOENT) { + exitcode = (int)MANDOCLEVEL_BADARG; + say("", "&chdir"); + } return(0); } + 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); } @@ -2321,17 +2431,17 @@ say(const char *file, const char *format, ...) if ('\0' != *basedir) fprintf(stderr, "%s", basedir); if ('\0' != *basedir && '\0' != *file) - fputs("//", stderr); + fputc('/', stderr); if ('\0' != *file) fprintf(stderr, "%s", file); use_errno = 1; if (NULL != format) { switch (*format) { - case ('&'): + case '&': format++; break; - case ('\0'): + case '\0': format = NULL; break; default: