=================================================================== RCS file: /cvs/mandoc/mandocdb.c,v retrieving revision 1.139 retrieving revision 1.179 diff -u -p -r1.139 -r1.179 --- mandoc/mandocdb.c 2014/04/19 02:30:19 1.139 +++ mandoc/mandocdb.c 2014/12/09 07:29:42 1.179 @@ -1,4 +1,4 @@ -/* $Id: mandocdb.c,v 1.139 2014/04/19 02:30:19 schwarze Exp $ */ +/* $Id: mandocdb.c,v 1.179 2014/12/09 07:29:42 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" @@ -80,17 +83,10 @@ 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 */ uint64_t mask; /* bitmask in sequence */ - char key[]; /* may contain escape sequences */ + char key[]; /* rendered text */ }; struct inodev { @@ -101,24 +97,25 @@ struct inodev { struct mpage { struct inodev inodev; /* used for hashing routine */ int64_t pageid; /* pageid in mpages SQL table */ - enum form form; /* format from file content */ 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 */ + int name_head_done; }; 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 */ }; @@ -127,11 +124,13 @@ enum stmt { 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 mdoc_node *); +typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); struct mdoc_handler { mdoc_fp fp; /* optional handler */ @@ -139,40 +138,53 @@ struct mdoc_handler { }; static void dbclose(int); -static void dbadd(struct mpage *, struct mchars *); +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 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 *); static void mlinks_undupe(struct mpage *); static void mpages_free(void); -static void mpages_merge(struct mchars *, struct mparse *); +static void mpages_merge(struct mparse *); static void names_check(void); static void parse_cat(struct mpage *, int); -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_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_Xr(struct mpage *, const struct mdoc_node *); +static void parse_man(struct mpage *, const struct man_meta *, + const struct man_node *); +static void parse_mdoc(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); +static int parse_mdoc_body(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); +static int parse_mdoc_head(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); +static int parse_mdoc_Fd(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); +static void parse_mdoc_fname(struct mpage *, const struct mdoc_node *); +static int parse_mdoc_Fn(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); +static int parse_mdoc_Fo(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); +static int parse_mdoc_Nd(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); +static int parse_mdoc_Nm(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); +static int parse_mdoc_Sh(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); +static int parse_mdoc_Xr(struct mpage *, const struct mdoc_meta *, + const struct mdoc_node *); static void putkey(const struct mpage *, char *, uint64_t); -static void putkeys(const struct mpage *, - const char *, size_t, 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 render_key(struct mchars *, struct str *); +static int render_string(char **, size_t *); 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]); @@ -180,13 +192,14 @@ 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 mchars *mchars; /* table of named characters */ 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 */ @@ -292,7 +305,7 @@ static const struct mdoc_handler mdocs[MDOC_MAX] = { { NULL, 0 }, /* Ux */ { NULL, 0 }, /* Xc */ { NULL, 0 }, /* Xo */ - { parse_mdoc_head, 0 }, /* Fo */ + { parse_mdoc_Fo, 0 }, /* Fo */ { NULL, 0 }, /* Fc */ { NULL, 0 }, /* Oo */ { NULL, 0 }, /* Oc */ @@ -318,15 +331,16 @@ static const struct mdoc_handler mdocs[MDOC_MAX] = { { NULL, 0 }, /* sp */ { NULL, 0 }, /* %U */ { NULL, 0 }, /* Ta */ + { NULL, 0 }, /* ll */ }; + 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 mparse *mp; struct ohash_info mpages_info, mlinks_info; @@ -335,8 +349,8 @@ 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 +362,14 @@ 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 +378,52 @@ 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: @@ -418,31 +434,33 @@ 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; } exitcode = (int)MANDOCLEVEL_OK; - mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL, NULL); - mc = mchars_alloc(); - + mchars = mchars_alloc(); + mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL, + mchars, NULL); ohash_init(&mpages, 6, &mpages_info); 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; + 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) @@ -452,6 +470,7 @@ main(int argc, char *argv[]) * Database missing or corrupt. * Recreate from scratch. */ + exitcode = (int)MANDOCLEVEL_OK; op = OP_DEFAULT; if (0 == treescan()) goto out; @@ -459,7 +478,7 @@ main(int argc, char *argv[]) goto out; } if (OP_DELETE != op) - mpages_merge(mc, mp); + mpages_merge(mp); dbclose(OP_DEFAULT == op ? 0 : 1); } else { /* @@ -468,8 +487,8 @@ 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]); @@ -499,17 +518,15 @@ 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 && + mpages_merge(mp); + if (warnings && !nodb && ! (MPARSE_QUICK & mparse_options)) names_check(); dbclose(0); @@ -522,10 +539,9 @@ main(int argc, char *argv[]) } } out: - set_basedir(NULL); manpath_free(&dirs); - mchars_free(mc); mparse_free(mp); + mchars_free(mchars); mpages_free(); ohash_delete(&mpages); ohash_delete(&mlinks); @@ -536,7 +552,7 @@ 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); @@ -550,7 +566,7 @@ usage: * If use_all has been specified, grok all files. * If not, sanitise paths to the following: * - * [./]man*[/]/.
+ * [./]man*[/]/.
* or * [./]cat
[/]/.0 * @@ -590,7 +606,7 @@ treescan(void) * Symbolic links require various sanity checks, * then get handled just like regular files. */ - case (FTS_SL): + case FTS_SL: if (NULL == realpath(path, buf)) { if (warnings) say(path, "&realpath"); @@ -613,7 +629,7 @@ treescan(void) * If we're a regular file, add an mlink by using the * stored directory data and handling the filename. */ - case (FTS_F): + case FTS_F: if (0 == strcmp(path, MANDOC_DB)) continue; if ( ! use_all && ff->fts_level < 2) { @@ -660,7 +676,13 @@ 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; @@ -670,9 +692,9 @@ treescan(void) mlink_add(mlink, ff->fts_statp); continue; - case (FTS_D): + case FTS_D: /* FALLTHROUGH */ - case (FTS_DP): + case FTS_DP: break; default: @@ -682,10 +704,10 @@ treescan(void) } 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. @@ -706,14 +728,14 @@ 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. @@ -745,7 +767,7 @@ treescan(void) * Try to infer the manual section, architecture, and page name from the * path, assuming it looks like * - * [./]man*[/]/.
+ * [./]man*[/]/.
* or * [./]cat
[/]/.0 * @@ -791,10 +813,10 @@ filescan(const char *file) 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); @@ -816,13 +838,22 @@ filescan(const char *file) say(file, "&stat"); return; } - strlcpy(buf, file, sizeof(buf)); - start = strstr(buf, basedir) == buf ? - buf + strlen(basedir) + 1 : buf; + 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. @@ -970,19 +1001,15 @@ 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); @@ -1061,34 +1088,32 @@ mlink_check(struct mpage *mpage, struct mlink *mlink) * and filename to determine whether the file is parsable or not. */ static void -mpages_merge(struct mchars *mc, struct mparse *mp) +mpages_merge(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; struct man *man; 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) + if ( ! nodb) SQL_EXEC("BEGIN TRANSACTION"); mpage = ohash_first(&mpages, &pslot); - while (NULL != mpage) { + while (mpage != NULL) { mlinks_undupe(mpage); - if (NULL == mpage->mlinks) { + if (mpage->mlinks == NULL) { mpage = ohash_next(&mpages, &pslot); continue; } @@ -1100,39 +1125,11 @@ mpages_merge(struct mchars *mc, struct mparse *mp) mdoc = NULL; 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); + if (fd == -1) { + say(mpage->mlinks->file, "&open"); + goto nextpage; } /* @@ -1140,17 +1137,23 @@ mpages_merge(struct mchars *mc, struct mparse *mp) * source code, unless it is already known to be * formatted. Fall back to formatted mode. */ - if (FORM_CAT != mpage->mlinks->dform || - FORM_CAT != mpage->mlinks->fform) { - lvl = mparse_readfd(mp, fd[0], mpage->mlinks->file); + if (mpage->mlinks->dform != FORM_CAT || + mpage->mlinks->fform != FORM_CAT) { + lvl = mparse_readfd(mp, fd, mpage->mlinks->file); if (lvl < MANDOCLEVEL_FATAL) mparse_result(mp, &mdoc, &man, &sodest); } - if (NULL != sodest) { + if (sodest != NULL) { mlink_dest = ohash_find(&mlinks, ohash_qlookup(&mlinks, sodest)); - if (NULL != mlink_dest) { + 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) { /* The .so target exists. */ @@ -1169,9 +1172,9 @@ mpages_merge(struct mchars *mc, struct mparse *mp) */ if (mpage_dest->pageid) - dbadd_mlink(mlink); + dbadd_mlink_name(mlink); - if (NULL == mlink->next) + if (mlink->next == NULL) break; mlink = mlink->next; } @@ -1183,16 +1186,17 @@ mpages_merge(struct mchars *mc, struct mparse *mp) mpage->mlinks = NULL; } goto nextpage; - } else if (NULL != mdoc) { + } else if (mdoc != NULL) { mpage->form = FORM_SRC; - mpage->sec = - mandoc_strdup(mdoc_meta(mdoc)->msec); + mpage->sec = mdoc_meta(mdoc)->msec; + mpage->sec = mandoc_strdup( + mpage->sec == NULL ? "" : mpage->sec); mpage->arch = mdoc_meta(mdoc)->arch; mpage->arch = mandoc_strdup( - NULL == mpage->arch ? "" : mpage->arch); + mpage->arch == NULL ? "" : mpage->arch); mpage->title = mandoc_strdup(mdoc_meta(mdoc)->title); - } else if (NULL != man) { + } else if (man != NULL) { mpage->form = FORM_SRC; mpage->sec = mandoc_strdup(man_meta(man)->msec); @@ -1210,8 +1214,8 @@ mpages_merge(struct mchars *mc, struct mparse *mp) mandoc_strdup(mpage->mlinks->name); } 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) @@ -1223,16 +1227,14 @@ mpages_merge(struct mchars *mc, struct mparse *mp) putkey(mpage, mlink->name, NAME_FILE); } - assert(NULL == mpage->desc); - if (NULL != mdoc) { - if (NULL != (cp = mdoc_meta(mdoc)->name)) - putkey(mpage, cp, NAME_HEAD); - parse_mdoc(mpage, mdoc_node(mdoc)); - } else if (NULL != man) - parse_man(mpage, man_node(man)); + assert(mpage->desc == NULL); + if (mdoc != NULL) + parse_mdoc(mpage, mdoc_meta(mdoc), mdoc_node(mdoc)); + else if (man != NULL) + parse_man(mpage, man_meta(man), man_node(man)); else - parse_cat(mpage, fd[0]); - if (NULL == mpage->desc) + parse_cat(mpage, fd); + if (mpage->desc == NULL) mpage->desc = mandoc_strdup(mpage->mlinks->name); if (warnings && !use_all) @@ -1240,24 +1242,12 @@ mpages_merge(struct mchars *mc, struct mparse *mp) mlink = mlink->next) mlink_check(mpage, mlink); - dbadd(mpage, mc); + dbadd(mpage); 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 (mparse_wait(mp) != MANDOCLEVEL_OK) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say(mpage->mlinks->file, "&wait gunzip"); } ohash_delete(&strings); ohash_delete(&names); @@ -1284,7 +1274,8 @@ 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); @@ -1292,10 +1283,10 @@ 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); @@ -1314,6 +1305,8 @@ 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; @@ -1333,7 +1326,7 @@ 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, @@ -1403,7 +1396,7 @@ 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; } @@ -1446,7 +1439,8 @@ putmdockey(const struct mpage *mpage, } static void -parse_man(struct mpage *mpage, const struct man_node *n) +parse_man(struct mpage *mpage, const struct man_meta *meta, + const struct man_node *n) { const struct man_node *head, *body; char *start, *title; @@ -1467,11 +1461,11 @@ 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. @@ -1485,7 +1479,7 @@ 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 @@ -1512,6 +1506,11 @@ parse_man(struct mpage *mpage, const struct man_node * 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; @@ -1526,6 +1525,11 @@ parse_man(struct mpage *mpage, const struct man_node * 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; } @@ -1556,28 +1560,29 @@ parse_man(struct mpage *mpage, const struct man_node * 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 mdoc_meta *meta, + const struct mdoc_node *n) { 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)) + if (0 == (*mdocs[n->tok].fp)(mpage, meta, n)) break; if (mdocs[n->tok].mask) putmdockey(mpage, n->child, @@ -1588,19 +1593,20 @@ parse_mdoc(struct mpage *mpage, const struct mdoc_node 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 mdoc_meta *meta, + const struct mdoc_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) + NULL == (n = n->child) || + MDOC_TEXT != n->type) return(0); /* @@ -1635,42 +1641,62 @@ parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_n return(0); } -static int -parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n) +static void +parse_mdoc_fname(struct mpage *mpage, const struct mdoc_node *n) { char *cp; + size_t sz; - if (NULL == (n = n->child) || MDOC_TEXT != n->type) - return(0); + if (n->type != MDOC_TEXT) + return; - /* - * Parse: .Fn "struct type *name" "char *arg". - * First strip away pointer symbol. - * Then store the function name, then type. - * Finally, store the arguments. - */ + /* Skip function pointer punctuation. */ - if (NULL == (cp = strrchr(n->string, ' '))) - cp = n->string; - - while ('*' == *cp) + cp = n->string; + while (*cp == '(' || *cp == '*') cp++; + sz = strcspn(cp, "()"); - putkey(mpage, cp, TYPE_Fn); + putkeys(mpage, cp, sz, TYPE_Fn); + if (n->sec == SEC_SYNOPSIS) + putkeys(mpage, cp, sz, NAME_SYN); +} - if (n->string < cp) - putkeys(mpage, n->string, cp - n->string, TYPE_Ft); +static int +parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_meta *meta, + const struct mdoc_node *n) +{ - for (n = n->next; NULL != n; n = n->next) - if (MDOC_TEXT == n->type) + 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 == MDOC_TEXT) putkey(mpage, n->string, TYPE_Fa); return(0); } static int -parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n) +parse_mdoc_Fo(struct mpage *mpage, const struct mdoc_meta *meta, + const struct mdoc_node *n) { + + if (n->type != MDOC_HEAD) + return(1); + + if (n->child != NULL) + parse_mdoc_fname(mpage, n->child); + + return(0); +} + +static int +parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_meta *meta, + const struct mdoc_node *n) +{ char *cp; if (NULL == (n = n->child)) @@ -1688,7 +1714,8 @@ parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_n } static int -parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n) +parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_meta *meta, + const struct mdoc_node *n) { if (MDOC_BODY == n->type) @@ -1697,32 +1724,46 @@ parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_n } static int -parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n) +parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_meta *meta, + const struct mdoc_node *n) { if (SEC_NAME == n->sec) putmdockey(mpage, n->child, NAME_TITLE); - else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type) - putmdockey(mpage, n->child, NAME_SYN); + else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type) { + if (n->child == NULL) + putkey(mpage, meta->name, NAME_SYN); + else + putmdockey(mpage, n->child, NAME_SYN); + } + 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_Sh(struct mpage *mpage, const struct mdoc_node *n) +parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_meta *meta, + const struct mdoc_node *n) { return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type); } static int -parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n) +parse_mdoc_head(struct mpage *mpage, const struct mdoc_meta *meta, + const struct mdoc_node *n) { return(MDOC_HEAD == n->type); } static int -parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n) +parse_mdoc_body(struct mpage *mpage, const struct mdoc_meta *meta, + const struct mdoc_node *n) { return(MDOC_BODY == n->type); @@ -1734,30 +1775,32 @@ parse_mdoc_body(struct mpage *mpage, const struct mdoc * 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; const char *end; unsigned int slot; - int i; + int i, mustfree; if (0 == sz) return; + mustfree = render_string(&cp, &sz); + if (TYPE_Nm & v) { htab = &names; v &= name_mask; - name_mask &= ~NAME_FIRST; + if (v & NAME_FIRST) + name_mask &= ~NAME_FIRST; if (debug > 1) say(mpage->mlinks->file, - "Adding name %*s", sz, cp); + "Adding name %*s, bits=%d", sz, cp, v); } else { 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); @@ -1771,12 +1814,15 @@ 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); } s->mpage = mpage; s->mask = v; + + if (mustfree) + free(cp); } /* @@ -1832,21 +1878,20 @@ utf8(unsigned int cp, char out[7]) } /* - * Store the rendered version of a key, or alias the pointer - * if the key contains no escape sequences. + * 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 -render_key(struct mchars *mc, struct str *key) +static int +render_string(char **public, size_t *psz) { - size_t sz, bsz, pos; + const char *src, *scp, *addcp, *seq; + char *dst; + size_t ssz, dsz, addsz; char utfbuf[7], res[6]; - char *buf; - const char *seq, *cpp, *val; - int len, u; - enum mandoc_esc esc; + int seqlen, unicode; - assert(NULL == key->rendered); - res[0] = '\\'; res[1] = '\t'; res[2] = ASCII_NBRSP; @@ -1854,68 +1899,62 @@ render_key(struct mchars *mc, struct str *key) 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->rendered = 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); } - switch (*val) { - case (ASCII_HYPH): - buf[pos++] = '-'; - val++; + /* Handle single-char special characters. */ + + switch (*scp) { + case '\\': + break; + case '\t': + /* FALLTHROUGH */ + case ASCII_NBRSP: + dst[dsz++] = ' '; + scp++; continue; - case ('\t'): + case ASCII_HYPH: + dst[dsz++] = '-'; /* FALLTHROUGH */ - case (ASCII_NBRSP): - buf[pos++] = ' '; - val++; - /* FALLTHROUGH */ - case (ASCII_BREAK): + case ASCII_BREAK: + scp++; continue; default: - break; + abort(); } - if ('\\' != *val) - break; - /* Read past the slash. */ - - val++; - /* - * 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) + scp++; + if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL) continue; /* @@ -1924,32 +1963,44 @@ render_key(struct mchars *mc, struct str *key) */ if (write_utf8) { - if (0 == (u = mchars_spec2cp(mc, seq, len))) + unicode = mchars_spec2cp(mchars, seq, seqlen); + if (unicode <= 0) continue; - cpp = utfbuf; - if (0 == (sz = utf8(u, utfbuf))) + addsz = utf8(unicode, utfbuf); + if (addsz == 0) continue; - sz = strlen(cpp); + addcp = utfbuf; } else { - cpp = mchars_spec2str(mc, seq, len, &sz); - if (NULL == cpp) + addcp = mchars_spec2str(mchars, seq, seqlen, &addsz); + if (addcp == NULL) continue; - if (ASCII_NBRSP == *cpp) { - cpp = " "; - sz = 1; + if (*addcp == ASCII_NBRSP) { + addcp = " "; + addsz = 1; } } /* Copy the rendered glyph into the stream. */ - bsz += sz; - buf = mandoc_realloc(buf, bsz); - memcpy(&buf[pos], cpp, sz); - pos += sz; + ssz += addsz; + dst = mandoc_realloc(dst, ssz + 1); + memcpy(dst + dsz, addcp, addsz); + dsz += addsz; } + if (dst != NULL) { + *public = dst; + *psz = dsz; + } - buf[pos] = '\0'; - key->rendered = buf; + /* Trim trailing whitespace and NUL-terminate. */ + + while (*psz > 0 && (*public)[*psz - 1] == ' ') + --*psz; + if (dst != NULL) { + (*public)[*psz] = '\0'; + return(1); + } else + return(0); } static void @@ -1966,6 +2017,30 @@ dbadd_mlink(const struct mlink *mlink) sqlite3_reset(stmts[STMT_INSERT_LINK]); } +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 @@ -1973,16 +2048,26 @@ dbadd_mlink(const struct mlink *mlink) * Also, handle escape sequences at the last possible moment. */ static void -dbadd(struct mpage *mpage, struct mchars *mc) +dbadd(struct mpage *mpage) { struct mlink *mlink; struct str *key; + char *cp; size_t i; unsigned int slot; + 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 || @@ -2009,12 +2094,17 @@ dbadd(struct mpage *mpage, struct mchars *mc) if (debug) say(mlink->file, "Adding to database"); + cp = mpage->desc; + i = strlen(cp); + mustfree = render_string(&cp, &i); 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, 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]); + if (mustfree) + free(cp); while (NULL != mlink) { dbadd_mlink(mlink); @@ -2025,31 +2115,23 @@ dbadd(struct mpage *mpage, struct mchars *mc) for (key = ohash_first(&names, &slot); NULL != key; key = ohash_next(&names, &slot)) { assert(key->mpage == mpage); - if (NULL == key->rendered) - render_key(mc, key); i = 1; SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask); - SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->rendered); + 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]); - if (key->rendered != key->key) - free(key->rendered); free(key); } for (key = ohash_first(&strings, &slot); NULL != key; key = ohash_next(&strings, &slot)) { assert(key->mpage == mpage); - if (NULL == key->rendered) - render_key(mc, key); i = 1; SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask); - SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered); + 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]); - if (key->rendered != key->key) - free(key->rendered); free(key); } } @@ -2123,11 +2205,11 @@ 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"); @@ -2149,11 +2231,11 @@ 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); @@ -2184,7 +2266,7 @@ dbopen(int real) const char *sql; int rc, ofl; - if (nodb) + if (nodb) return(1); *tempfilename = '\0'; @@ -2194,7 +2276,8 @@ 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; @@ -2204,36 +2287,27 @@ 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); } @@ -2257,7 +2331,8 @@ create_tables: " \"bits\" INTEGER NOT NULL,\n" " \"name\" TEXT NOT NULL,\n" " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) " - "ON DELETE CASCADE\n" + "ON DELETE CASCADE,\n" + " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n" ");\n" "\n" "CREATE TABLE \"keys\" (\n" @@ -2271,11 +2346,20 @@ 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"); + 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=?)"; @@ -2286,6 +2370,8 @@ prepare_statements: 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); @@ -2299,18 +2385,24 @@ 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 * @@ -2321,63 +2413,89 @@ 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); } @@ -2390,17 +2508,17 @@ 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: