=================================================================== RCS file: /cvs/mandoc/mandocdb.c,v retrieving revision 1.73 retrieving revision 1.144 diff -u -p -r1.73 -r1.144 --- mandoc/mandocdb.c 2013/10/18 23:07:23 1.73 +++ mandoc/mandocdb.c 2014/04/23 21:06:41 1.144 @@ -1,7 +1,7 @@ -/* $Id: mandocdb.c,v 1.73 2013/10/18 23:07:23 schwarze Exp $ */ +/* $Id: mandocdb.c,v 1.144 2014/04/23 21:06:41 schwarze Exp $ */ /* * Copyright (c) 2011, 2012 Kristaps Dzonsons - * Copyright (c) 2011, 2012, 2013 Ingo Schwarze + * Copyright (c) 2011, 2012, 2013, 2014 Ingo Schwarze * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above @@ -20,6 +20,7 @@ #endif #include +#include #include #include @@ -45,27 +46,31 @@ #include "mdoc.h" #include "man.h" #include "mandoc.h" +#include "mandoc_aux.h" #include "manpath.h" #include "mansearch.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)) \ - fprintf(stderr, "%s\n", sqlite3_errmsg(db)) + 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)) \ - fprintf(stderr, "%s\n", sqlite3_errmsg(db)) + say(mlink->file, "%s", sqlite3_errmsg(db)) #define SQL_BIND_INT(_s, _i, _v) \ if (SQLITE_OK != sqlite3_bind_int \ ((_s), (_i)++, (_v))) \ - fprintf(stderr, "%s\n", sqlite3_errmsg(db)) + say(mlink->file, "%s", sqlite3_errmsg(db)) #define SQL_BIND_INT64(_s, _i, _v) \ if (SQLITE_OK != sqlite3_bind_int64 \ ((_s), (_i)++, (_v))) \ - fprintf(stderr, "%s\n", sqlite3_errmsg(db)) + say(mlink->file, "%s", sqlite3_errmsg(db)) #define SQL_STEP(_s) \ if (SQLITE_DONE != sqlite3_step((_s))) \ - fprintf(stderr, "%s\n", sqlite3_errmsg(db)) + say(mlink->file, "%s", sqlite3_errmsg(db)) enum op { OP_DEFAULT = 0, /* new dbs from dir list or default config */ @@ -76,49 +81,57 @@ enum op { }; enum form { - FORM_SRC, /* format is -man or -mdoc */ - FORM_CAT, /* format is cat */ - FORM_NONE /* format is unknown */ + FORM_NONE, /* format is unknown */ + FORM_SRC, /* format is -man or -mdoc */ + FORM_CAT /* format is cat */ }; struct str { - char *utf8; /* key in UTF-8 form */ - const struct of *of; /* if set, the owning parse */ + 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[]; /* the string itself */ + char key[]; /* may contain escape sequences */ }; -struct id { - ino_t ino; - dev_t dev; +struct inodev { + ino_t st_ino; + dev_t st_dev; }; -struct of { - struct id id; /* used for hashing routine */ - struct of *next; /* next in ofs */ - enum form dform; /* path-cued form */ - enum form sform; /* suffix-cued form */ - char file[PATH_MAX]; /* filename rel. to manpath */ - char *desc; /* parsed description */ - char *name; /* name (from filename) (not empty) */ - char *sec; /* suffix-cued section (or empty) */ - char *dsec; /* path-cued section (or empty) */ - char *arch; /* path-cued arch. (or empty) */ +struct 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 */ }; -struct title { - char *title; /* name(sec/arch) given inside the file */ - char *file; /* file name in case of mismatch */ +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 gzip; /* filename has a .gz suffix */ }; enum stmt { - STMT_DELETE = 0, /* delete manpage */ - STMT_INSERT_DOC, /* insert manpage */ - STMT_INSERT_KEY, /* insert parsed key */ + STMT_DELETE_PAGE = 0, /* delete mpage */ + STMT_INSERT_PAGE, /* insert mpage */ + STMT_INSERT_LINK, /* insert mlink */ + STMT_INSERT_NAME, /* insert name */ + STMT_INSERT_KEY, /* insert parsed key */ STMT__MAX }; -typedef int (*mdoc_fp)(struct of *, const struct mdoc_node *); +typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *); struct mdoc_handler { mdoc_fp fp; /* optional handler */ @@ -126,60 +139,61 @@ struct mdoc_handler { }; static void dbclose(int); -static void dbindex(struct mchars *, int, const struct of *); +static void dbadd(struct mpage *, struct mchars *); +static void dbadd_mlink(const struct mlink *mlink); static int dbopen(int); static void dbprune(void); -static void fileadd(struct of *); -static int filecheck(const char *); static void filescan(const char *); static void *hash_alloc(size_t, void *); static void hash_free(void *, size_t, void *); static void *hash_halloc(size_t, void *); -static void inoadd(const struct stat *, struct of *); -static int inocheck(const struct stat *); -static void ofadd(int, const char *, const char *, const char *, - const char *, const char *, const struct stat *); -static void offree(void); -static void ofmerge(struct mchars *, struct mparse *, int); -static void parse_catpage(struct of *); -static void parse_man(struct of *, const struct man_node *); -static void parse_mdoc(struct of *, const struct mdoc_node *); -static int parse_mdoc_body(struct of *, const struct mdoc_node *); -static int parse_mdoc_head(struct of *, const struct mdoc_node *); -static int parse_mdoc_Fd(struct of *, const struct mdoc_node *); -static int parse_mdoc_Fn(struct of *, const struct mdoc_node *); -static int parse_mdoc_In(struct of *, const struct mdoc_node *); -static int parse_mdoc_Nd(struct of *, const struct mdoc_node *); -static int parse_mdoc_Nm(struct of *, const struct mdoc_node *); -static int parse_mdoc_Sh(struct of *, const struct mdoc_node *); -static int parse_mdoc_St(struct of *, const struct mdoc_node *); -static int parse_mdoc_Xr(struct of *, const struct mdoc_node *); -static int set_basedir(const char *); -static void putkey(const struct of *, - const char *, uint64_t); -static void putkeys(const struct of *, +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 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 putkey(const struct mpage *, char *, uint64_t); +static void putkeys(const struct mpage *, const char *, size_t, uint64_t); -static void putmdockey(const struct of *, +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 treescan(void); static size_t utf8(unsigned int, char [7]); -static void utf8key(struct mchars *, struct str *); +static char tempfilename[32]; static char *progname; -static int use_all; /* use all found files */ static int nodb; /* no database changes */ -static int verb; /* print what we're doing */ -static int warnings; /* warn about crap */ +static int mparse_options; /* abort the parse early */ +static int use_all; /* use all found files */ +static int debug; /* print what we're doing */ +static int warnings; /* warn about crap */ +static int write_utf8; /* write UTF-8 output; else ASCII */ static int exitcode; /* to be returned by main */ -static enum op op; /* operational mode */ +static enum op op; /* operational mode */ static char basedir[PATH_MAX]; /* current base directory */ -static struct ohash inos; /* table of inodes/devices */ -static struct ohash filenames; /* table of filenames */ +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 struct of *ofs = NULL; /* vector of files to parse */ 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 }, /* Ap */ @@ -211,15 +225,15 @@ static const struct mdoc_handler mdocs[MDOC_MAX] = { { parse_mdoc_Fn, 0 }, /* Fn */ { NULL, TYPE_Ft }, /* Ft */ { NULL, TYPE_Ic }, /* Ic */ - { parse_mdoc_In, TYPE_In }, /* In */ + { NULL, TYPE_In }, /* In */ { NULL, TYPE_Li }, /* Li */ - { parse_mdoc_Nd, TYPE_Nd }, /* Nd */ - { parse_mdoc_Nm, TYPE_Nm }, /* Nm */ + { parse_mdoc_Nd, 0 }, /* Nd */ + { parse_mdoc_Nm, 0 }, /* Nm */ { NULL, 0 }, /* Op */ { NULL, 0 }, /* Ot */ { NULL, TYPE_Pa }, /* Pa */ { NULL, 0 }, /* Rv */ - { parse_mdoc_St, 0 }, /* St */ + { NULL, TYPE_St }, /* St */ { NULL, TYPE_Va }, /* Va */ { parse_mdoc_body, TYPE_Va }, /* Vt */ { parse_mdoc_Xr, 0 }, /* Xr */ @@ -306,6 +320,7 @@ static const struct mdoc_handler mdocs[MDOC_MAX] = { { NULL, 0 }, /* Ta */ }; + int main(int argc, char *argv[]) { @@ -315,17 +330,17 @@ main(int argc, char *argv[]) struct mchars *mc; struct manpaths dirs; struct mparse *mp; - struct ohash_info ino_info, filename_info; + struct ohash_info mpages_info, mlinks_info; memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *)); memset(&dirs, 0, sizeof(struct manpaths)); - ino_info.alloc = filename_info.alloc = hash_alloc; - ino_info.halloc = filename_info.halloc = hash_halloc; - ino_info.hfree = filename_info.hfree = hash_free; + mpages_info.alloc = mlinks_info.alloc = hash_alloc; + mpages_info.halloc = mlinks_info.halloc = hash_halloc; + mpages_info.hfree = mlinks_info.hfree = hash_free; - ino_info.key_offset = offsetof(struct of, id); - filename_info.key_offset = offsetof(struct of, file); + mpages_info.key_offset = offsetof(struct mpage, inodev); + mlinks_info.key_offset = offsetof(struct mlink, file); progname = strrchr(argv[0], '/'); if (progname == NULL) @@ -334,7 +349,7 @@ 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. */ @@ -347,41 +362,55 @@ main(int argc, char *argv[]) path_arg = NULL; op = OP_DEFAULT; - while (-1 != (ch = getopt(argc, argv, "aC:d:ntu:vW"))) + while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v"))) switch (ch) { - case ('a'): + case 'a': use_all = 1; break; - case ('C'): + case 'C': CHECKOP(op, ch); path_arg = optarg; op = OP_CONFFILE; break; - case ('d'): + case 'D': + debug++; + break; + case 'd': CHECKOP(op, ch); path_arg = optarg; op = OP_UPDATE; break; - case ('n'): + case 'n': nodb = 1; break; - case ('t'): + case 'p': + warnings = 1; + break; + case 'Q': + mparse_options |= MPARSE_QUICK; + break; + case 'T': + if (strcmp(optarg, "utf8")) { + fprintf(stderr, "-T%s: Unsupported " + "output format\n", optarg); + goto usage; + } + write_utf8 = 1; + break; + case 't': CHECKOP(op, ch); dup2(STDOUT_FILENO, STDERR_FILENO); op = OP_TEST; nodb = warnings = 1; break; - case ('u'): + case 'u': CHECKOP(op, ch); path_arg = optarg; op = OP_DELETE; break; - case ('v'): - verb++; + case 'v': + /* Compatibility with espie@'s makewhatis. */ break; - case ('W'): - warnings = 1; - break; default: goto usage; } @@ -395,35 +424,46 @@ main(int argc, char *argv[]) } exitcode = (int)MANDOCLEVEL_OK; - mp = mparse_alloc(MPARSE_AUTO, - MANDOCLEVEL_FATAL, NULL, NULL, NULL); + mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL, NULL); mc = mchars_alloc(); - ohash_init(&inos, 6, &ino_info); - ohash_init(&filenames, 6, &filename_info); + 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. + * Jump into that directory first. */ if (0 == set_basedir(path_arg)) 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) - ofmerge(mc, mp, 0); - dbclose(1); + mpages_merge(mc, mp); + dbclose(OP_DEFAULT == op ? 0 : 1); } else { /* * If we have arguments, use them as our manpaths. @@ -431,14 +471,19 @@ 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]); } else manpath_parse(&dirs, path_arg, NULL, NULL); + if (0 == dirs.sz) { + exitcode = (int)MANDOCLEVEL_BADARG; + say("", "Empty manpath"); + } + /* * First scan the tree rooted at a base directory, then * build a new database and finally move it into place. @@ -453,8 +498,8 @@ main(int argc, char *argv[]) continue; if (j) { - ohash_init(&inos, 6, &ino_info); - ohash_init(&filenames, 6, &filename_info); + ohash_init(&mpages, 6, &mpages_info); + ohash_init(&mlinks, 6, &mlinks_info); } if (0 == set_basedir(dirs.paths[j])) @@ -466,13 +511,16 @@ main(int argc, char *argv[]) if (0 == dbopen(0)) goto out; - ofmerge(mc, mp, warnings && !use_all); + mpages_merge(mc, mp); + if (warnings && + ! (MPARSE_QUICK & mparse_options)) + names_check(); dbclose(0); if (j + 1 < dirs.sz) { - ohash_delete(&inos); - ohash_delete(&filenames); - offree(); + mpages_free(); + ohash_delete(&mpages); + ohash_delete(&mlinks); } } } @@ -481,17 +529,17 @@ out: manpath_free(&dirs); mchars_free(mc); mparse_free(mp); - ohash_delete(&inos); - ohash_delete(&filenames); - offree(); + mpages_free(); + ohash_delete(&mpages); + ohash_delete(&mlinks); return(exitcode); usage: - fprintf(stderr, "usage: %s [-anvW] [-C file]\n" - " %s [-anvW] dir ...\n" - " %s [-nvW] -d dir [file ...]\n" - " %s [-nvW] -u dir [file ...]\n" - " %s -t file ...\n", - progname, progname, progname, + fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n" + " %s [-aDnpQ] [-Tutf8] dir ...\n" + " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n" + " %s [-Dnp] -u dir [file ...]\n" + " %s [-Q] -t file ...\n", + progname, progname, progname, progname, progname); return((int)MANDOCLEVEL_BADARG); @@ -505,7 +553,7 @@ usage: * If use_all has been specified, grok all files. * If not, sanitise paths to the following: * - * [./]man*[/]/.
+ * [./]man*[/]/.
* or * [./]cat
[/]/.0 * @@ -514,24 +562,23 @@ usage: static int treescan(void) { + char buf[PATH_MAX]; FTS *f; FTSENT *ff; - int dform; - char *sec; - const char *dsec, *arch, *cp, *path; + struct mlink *mlink; + int dform, gzip; + char *dsec, *arch, *fsec, *cp; + const char *path; const char *argv[2]; 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("", NULL); + say("", "&fts_open"); return(0); } @@ -540,76 +587,119 @@ treescan(void) while (NULL != (ff = fts_read(f))) { path = ff->fts_path + 2; + switch (ff->fts_info) { + /* - * If we're a regular file, add an "of" by using the + * 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. - * Disallow duplicate (hard-linked) files. */ - if (FTS_F == ff->fts_info) { + case FTS_F: if (0 == strcmp(path, MANDOC_DB)) continue; if ( ! use_all && ff->fts_level < 2) { if (warnings) say(path, "Extraneous file"); continue; - } else if (inocheck(ff->fts_statp)) { - if (warnings) - say(path, "Duplicate file"); - continue; - } else if (NULL == (sec = - strrchr(ff->fts_name, '.'))) { + } + gzip = 0; + fsec = NULL; + while (NULL == fsec) { + fsec = strrchr(ff->fts_name, '.'); + if (NULL == fsec || strcmp(fsec+1, "gz")) + break; + gzip = 1; + *fsec = '\0'; + fsec = NULL; + } + if (NULL == fsec) { if ( ! use_all) { if (warnings) say(path, "No filename suffix"); continue; } - } else if (0 == strcmp(++sec, "html")) { + } else if (0 == strcmp(++fsec, "html")) { if (warnings) say(path, "Skip html"); continue; - } else if (0 == strcmp(sec, "gz")) { + } else if (0 == strcmp(fsec, "ps")) { if (warnings) - say(path, "Skip gz"); - continue; - } else if (0 == strcmp(sec, "ps")) { - if (warnings) say(path, "Skip ps"); continue; - } else if (0 == strcmp(sec, "pdf")) { + } else if (0 == strcmp(fsec, "pdf")) { if (warnings) say(path, "Skip pdf"); continue; } else if ( ! use_all && - ((FORM_SRC == dform && strcmp(sec, dsec)) || - (FORM_CAT == dform && strcmp(sec, "0")))) { + ((FORM_SRC == dform && strcmp(fsec, dsec)) || + (FORM_CAT == dform && strcmp(fsec, "0")))) { if (warnings) say(path, "Wrong filename suffix"); continue; } else - sec[-1] = '\0'; - ofadd(dform, path, ff->fts_name, dsec, sec, - arch, ff->fts_statp); + fsec[-1] = '\0'; + + mlink = mandoc_calloc(1, sizeof(struct mlink)); + if (strlcpy(mlink->file, path, + sizeof(mlink->file)) >= + sizeof(mlink->file)) { + say(path, "Filename too long"); + free(mlink); + continue; + } + mlink->dform = dform; + mlink->dsec = dsec; + mlink->arch = arch; + mlink->name = ff->fts_name; + mlink->fsec = fsec; + mlink->gzip = gzip; + mlink_add(mlink, ff->fts_statp); continue; - } else if (FTS_D != ff->fts_info && - FTS_DP != ff->fts_info) { + + case FTS_D: + /* 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. * If we're not in use_all, enforce it. */ - dsec = NULL; - dform = FORM_NONE; cp = ff->fts_name; if (FTS_DP == ff->fts_info) break; @@ -620,23 +710,27 @@ treescan(void) } else if (0 == strncmp(cp, "cat", 3)) { dform = FORM_CAT; dsec = cp + 3; + } else { + dform = FORM_NONE; + 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. */ - arch = NULL; if (FTS_DP != ff->fts_info && NULL != dsec) arch = ff->fts_name; + else + arch = NULL; break; default: if (FTS_DP == ff->fts_info || use_all) @@ -653,58 +747,98 @@ treescan(void) } /* - * Add a file to the file vector. + * Add a file to the mlinks table. * Do not verify that it's a "valid" looking manpage (we'll do that * later). * * Try to infer the manual section, architecture, and page name from the * path, assuming it looks like * - * [./]man*[/]/.
+ * [./]man*[/]/.
* or * [./]cat
[/]/.0 * - * Stuff this information directly into the "of" vector. * See treescan() for the fts(3) version of this. */ static void filescan(const char *file) { char buf[PATH_MAX]; - const char *sec, *arch, *name, *dsec; - char *p, *start; - int dform; struct stat st; + struct mlink *mlink; + char *p, *start; assert(use_all); if (0 == strncmp(file, "./", 2)) file += 2; - if (NULL == realpath(file, buf)) { + /* + * We have to do lstat(2) before realpath(3) loses + * the information whether this is a symbolic link. + * We need to know that because for symbolic links, + * we want to use the orginal file name, while for + * regular files, we want to use the real path. + */ + if (-1 == lstat(file, &st)) { exitcode = (int)MANDOCLEVEL_BADARG; - say(file, NULL); + say(file, "&lstat"); return; - } else if (OP_TEST != op && strstr(buf, basedir) != buf) { + } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) { exitcode = (int)MANDOCLEVEL_BADARG; - say("", "%s: outside base directory", buf); + say(file, "Not a regular file"); return; - } else if (-1 == stat(buf, &st)) { + } + + /* + * We have to resolve the file name to the real path + * in any case for the base directory check. + */ + if (NULL == realpath(file, buf)) { exitcode = (int)MANDOCLEVEL_BADARG; - say(file, NULL); + say(file, "&realpath"); return; - } else if ( ! (S_IFREG & st.st_mode)) { + } + + if (strstr(buf, basedir) == buf) + start = buf + strlen(basedir) + 1; + else if (OP_TEST == op) + start = buf; + else { exitcode = (int)MANDOCLEVEL_BADARG; - say(file, "Not a regular file"); + say("", "%s: outside base directory", buf); return; - } else if (inocheck(&st)) { - if (warnings) - say(file, "Duplicate file"); + } + + /* + * Now we are sure the file is inside our tree. + * If it is a symbolic link, ignore the real path + * and use the original name. + * This implies passing stuff like "cat1/../man1/foo.1" + * on the command line won't work. So don't do that. + * Note the stat(2) can still fail if the link target + * doesn't exist. + */ + if (S_IFLNK & st.st_mode) { + if (-1 == stat(buf, &st)) { + exitcode = (int)MANDOCLEVEL_BADARG; + say(file, "&stat"); + return; + } + if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) { + say(file, "Filename too long"); + return; + } + start = strstr(buf, basedir) == buf ? + buf + strlen(basedir) + 1 : buf; + } + + mlink = mandoc_calloc(1, sizeof(struct mlink)); + if (strlcpy(mlink->file, start, sizeof(mlink->file)) >= + sizeof(mlink->file)) { + say(start, "Filename too long"); return; } - start = buf + strlen(basedir); - sec = arch = name = dsec = NULL; - dform = FORM_NONE; /* * First try to guess our directory structure. @@ -715,19 +849,19 @@ filescan(const char *file) if (NULL != (p = strchr(start, '/'))) { *p++ = '\0'; if (0 == strncmp(start, "man", 3)) { - dform = FORM_SRC; - dsec = start + 3; + mlink->dform = FORM_SRC; + mlink->dsec = start + 3; } else if (0 == strncmp(start, "cat", 3)) { - dform = FORM_CAT; - dsec = start + 3; + mlink->dform = FORM_CAT; + mlink->dsec = start + 3; } start = p; - if (NULL != dsec && NULL != (p = strchr(start, '/'))) { + if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) { *p++ = '\0'; - arch = start; + mlink->arch = start; start = p; - } + } } /* @@ -740,168 +874,219 @@ filescan(const char *file) if ('.' == *p) { *p++ = '\0'; - sec = p; + mlink->fsec = p; } /* * Now try to parse the name. * Use the filename portion of the path. */ - name = start; + mlink->name = start; if (NULL != (p = strrchr(start, '/'))) { - name = p + 1; + mlink->name = p + 1; *p = '\0'; - } - - ofadd(dform, file, name, dsec, sec, arch, &st); + } + mlink_add(mlink, &st); } -/* - * See fileadd(). - */ -static int -filecheck(const char *name) +static void +mlink_add(struct mlink *mlink, const struct stat *st) { + struct inodev inodev; + struct mpage *mpage; + unsigned int slot; - return(NULL != ohash_find(&filenames, - ohash_qlookup(&filenames, name))); + assert(NULL != mlink->file); + + mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : ""); + mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : ""); + mlink->name = mandoc_strdup(mlink->name ? mlink->name : ""); + mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : ""); + + if ('0' == *mlink->fsec) { + free(mlink->fsec); + mlink->fsec = mandoc_strdup(mlink->dsec); + mlink->fform = FORM_CAT; + } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec) + mlink->fform = FORM_SRC; + else + mlink->fform = FORM_NONE; + + slot = ohash_qlookup(&mlinks, mlink->file); + assert(NULL == ohash_find(&mlinks, slot)); + ohash_insert(&mlinks, slot, mlink); + + inodev.st_ino = st->st_ino; + inodev.st_dev = st->st_dev; + slot = ohash_lookup_memory(&mpages, (char *)&inodev, + sizeof(struct inodev), inodev.st_ino); + mpage = ohash_find(&mpages, slot); + if (NULL == mpage) { + mpage = mandoc_calloc(1, sizeof(struct mpage)); + mpage->inodev.st_ino = inodev.st_ino; + mpage->inodev.st_dev = inodev.st_dev; + ohash_insert(&mpages, slot, mpage); + } else + mlink->next = mpage->mlinks; + mpage->mlinks = mlink; + mlink->mpage = mpage; } -/* - * Use the standard hashing mechanism (K&R) to see if the given filename - * already exists. - */ static void -fileadd(struct of *of) +mlink_free(struct mlink *mlink) { - unsigned int slot; - slot = ohash_qlookup(&filenames, of->file); - assert(NULL == ohash_find(&filenames, slot)); - ohash_insert(&filenames, slot, of); + free(mlink->dsec); + free(mlink->arch); + free(mlink->name); + free(mlink->fsec); + free(mlink); } -/* - * See inoadd(). - */ -static int -inocheck(const struct stat *st) +static void +mpages_free(void) { - struct id id; - uint32_t hash; + struct mpage *mpage; + struct mlink *mlink; + unsigned int slot; - memset(&id, 0, sizeof(id)); - id.ino = hash = st->st_ino; - id.dev = st->st_dev; - - return(NULL != ohash_find(&inos, ohash_lookup_memory( - &inos, (char *)&id, sizeof(id), hash))); + mpage = ohash_first(&mpages, &slot); + while (NULL != mpage) { + while (NULL != (mlink = mpage->mlinks)) { + mpage->mlinks = mlink->next; + mlink_free(mlink); + } + free(mpage->sec); + free(mpage->arch); + free(mpage->title); + free(mpage->desc); + free(mpage); + mpage = ohash_next(&mpages, &slot); + } } /* - * The hashing function used here is quite simple: simply take the inode - * and use uint32_t of its bits. - * Then when we do the lookup, use both the inode and device identifier. + * For each mlink to the mpage, check whether the path looks like + * it is formatted, and if it does, check whether a source manual + * exists by the same name, ignoring the suffix. + * If both conditions hold, drop the mlink. */ static void -inoadd(const struct stat *st, struct of *of) +mlinks_undupe(struct mpage *mpage) { - uint32_t hash; - unsigned int slot; + char buf[PATH_MAX]; + struct mlink **prev; + struct mlink *mlink; + char *bufp; - of->id.ino = hash = st->st_ino; - of->id.dev = st->st_dev; - slot = ohash_lookup_memory - (&inos, (char *)&of->id, sizeof(of->id), hash); - - assert(NULL == ohash_find(&inos, slot)); - ohash_insert(&inos, slot, of); + mpage->form = FORM_CAT; + prev = &mpage->mlinks; + while (NULL != (mlink = *prev)) { + if (FORM_CAT != mlink->dform) { + mpage->form = FORM_NONE; + goto nextlink; + } + (void)strlcpy(buf, mlink->file, sizeof(buf)); + bufp = strstr(buf, "cat"); + assert(NULL != bufp); + memcpy(bufp, "man", 3); + if (NULL != (bufp = strrchr(buf, '.'))) + *++bufp = '\0'; + (void)strlcat(buf, mlink->dsec, sizeof(buf)); + if (NULL == ohash_find(&mlinks, + ohash_qlookup(&mlinks, buf))) + goto nextlink; + if (warnings) + say(mlink->file, "Man source exists: %s", buf); + if (use_all) + goto nextlink; + *prev = mlink->next; + mlink_free(mlink); + continue; +nextlink: + prev = &(*prev)->next; + } } static void -ofadd(int dform, const char *file, const char *name, const char *dsec, - const char *sec, const char *arch, const struct stat *st) +mlink_check(struct mpage *mpage, struct mlink *mlink) { - struct of *of; - int sform; + struct str *str; + unsigned int slot; - assert(NULL != file); + /* + * Check whether the manual section given in a file + * agrees with the directory where the file is located. + * Some manuals have suffixes like (3p) on their + * section number either inside the file or in the + * directory name, some are linked into more than one + * section, like encrypt(1) = makekey(8). + */ - if (NULL == name) - name = ""; - if (NULL == sec) - sec = ""; - if (NULL == dsec) - dsec = ""; - if (NULL == arch) - arch = ""; + if (FORM_SRC == mpage->form && + strcasecmp(mpage->sec, mlink->dsec)) + say(mlink->file, "Section \"%s\" manual in %s directory", + mpage->sec, mlink->dsec); - sform = FORM_NONE; - if (NULL != sec && *sec <= '9' && *sec >= '1') - sform = FORM_SRC; - else if (NULL != sec && *sec == '0') { - sec = dsec; - sform = FORM_CAT; - } + /* + * Manual page directories exist for each kernel + * architecture as returned by machine(1). + * However, many manuals only depend on the + * application architecture as returned by arch(1). + * For example, some (2/ARM) manuals are shared + * across the "armish" and "zaurus" kernel + * architectures. + * A few manuals are even shared across completely + * different architectures, for example fdformat(1) + * on amd64, i386, sparc, and sparc64. + */ - of = mandoc_calloc(1, sizeof(struct of)); - strlcpy(of->file, file, PATH_MAX); - of->name = mandoc_strdup(name); - of->sec = mandoc_strdup(sec); - of->dsec = mandoc_strdup(dsec); - of->arch = mandoc_strdup(arch); - of->sform = sform; - of->dform = dform; - of->next = ofs; - ofs = of; + if (strcasecmp(mpage->arch, mlink->arch)) + say(mlink->file, "Architecture \"%s\" manual in " + "\"%s\" directory", mpage->arch, mlink->arch); /* - * Add to unique identifier hash. - * Then if it's a source manual and we're going to use source in - * favour of catpages, add it to that hash. + * XXX + * parse_cat() doesn't set NAME_TITLE yet. */ - inoadd(st, of); - fileadd(of); -} -static void -offree(void) -{ - struct of *of; + if (FORM_CAT == mpage->form) + return; - while (NULL != (of = ofs)) { - ofs = of->next; - free(of->name); - free(of->sec); - free(of->dsec); - free(of->arch); - free(of); - } + /* + * Check whether this mlink + * appears as a name in the NAME section. + */ + + slot = ohash_qlookup(&names, mlink->name); + str = ohash_find(&names, slot); + assert(NULL != str); + if ( ! (NAME_TITLE & str->mask)) + say(mlink->file, "Name missing in NAME section"); } /* - * Run through the files in the global vector "ofs" and add them to the - * database specified in "basedir". + * Run through the files in the global vector "mpages" + * and add them to the database specified in "basedir". * * This handles the parsing scheme itself, using the cues of directory * and filename to determine whether the file is parsable or not. */ static void -ofmerge(struct mchars *mc, struct mparse *mp, int check_reachable) +mpages_merge(struct mchars *mc, struct mparse *mp) { - struct ohash title_table; - struct ohash_info title_info, str_info; - char buf[PATH_MAX]; - struct of *of; + 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; - struct title *title_entry; - char *bufp, *title_str; - const char *msec, *march, *mtitle, *cp; - size_t sz; - int form; - int match; - unsigned int slot; + char *sodest; + char *cp; + pid_t child_pid; + int status; + unsigned int pslot; enum mandoclevel lvl; str_info.alloc = hash_alloc; @@ -909,196 +1094,241 @@ ofmerge(struct mchars *mc, struct mparse *mp, int chec str_info.hfree = hash_free; str_info.key_offset = offsetof(struct str, key); - if (check_reachable) { - title_info.alloc = hash_alloc; - title_info.halloc = hash_halloc; - title_info.hfree = hash_free; - title_info.key_offset = offsetof(struct title, title); - ohash_init(&title_table, 6, &title_info); - } + if (0 == nodb) + SQL_EXEC("BEGIN TRANSACTION"); - for (of = ofs; NULL != of; of = of->next) { - /* - * If we're a catpage (as defined by our path), then see - * if a manpage exists by the same name (ignoring the - * suffix). - * If it does, then we want to use it instead of our - * own. - */ - if ( ! use_all && FORM_CAT == of->dform) { - sz = strlcpy(buf, of->file, PATH_MAX); - if (sz >= PATH_MAX) { - if (warnings) - say(of->file, "Filename too long"); - continue; - } - bufp = strstr(buf, "cat"); - assert(NULL != bufp); - memcpy(bufp, "man", 3); - if (NULL != (bufp = strrchr(buf, '.'))) - *++bufp = '\0'; - strlcat(buf, of->dsec, PATH_MAX); - if (filecheck(buf)) { - if (warnings) - say(of->file, "Man " - "source exists: %s", buf); - continue; - } + mpage = ohash_first(&mpages, &pslot); + while (NULL != mpage) { + mlinks_undupe(mpage); + if (NULL == mpage->mlinks) { + mpage = ohash_next(&mpages, &pslot); + continue; } + name_mask = NAME_MASK; + ohash_init(&names, 4, &str_info); ohash_init(&strings, 6, &str_info); mparse_reset(mp); mdoc = NULL; man = NULL; - form = 0; - msec = of->dsec; - march = of->arch; - mtitle = of->name; - match = 1; + 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; + } + } + /* * Try interpreting the file as mdoc(7) or man(7) * source code, unless it is already known to be * formatted. Fall back to formatted mode. */ - if (FORM_SRC == of->dform || FORM_SRC == of->sform) { - lvl = mparse_readfd(mp, -1, of->file); + if (FORM_CAT != mpage->mlinks->dform || + FORM_CAT != mpage->mlinks->fform) { + lvl = mparse_readfd(mp, fd[0], mpage->mlinks->file); if (lvl < MANDOCLEVEL_FATAL) - mparse_result(mp, &mdoc, &man); - } + mparse_result(mp, &mdoc, &man, &sodest); + } - if (NULL != mdoc) { - form = 1; - msec = mdoc_meta(mdoc)->msec; - march = mdoc_meta(mdoc)->arch; - mtitle = mdoc_meta(mdoc)->title; - } else if (NULL != man) { - form = 1; - msec = man_meta(man)->msec; - march = ""; - mtitle = man_meta(man)->title; - } + if (NULL != sodest) { + mlink_dest = ohash_find(&mlinks, + ohash_qlookup(&mlinks, sodest)); + if (NULL != mlink_dest) { - if (NULL == msec) - msec = ""; - if (NULL == march) - march = ""; - if (NULL == mtitle) - mtitle = ""; + /* The .so target exists. */ - /* - * Check whether the manual section given in a file - * agrees with the directory where the file is located. - * Some manuals have suffixes like (3p) on their - * section number either inside the file or in the - * directory name, some are linked into more than one - * section, like encrypt(1) = makekey(8). Do not skip - * manuals for such reasons. - */ - if (warnings && !use_all && form && - strcasecmp(msec, of->dsec)) { - match = 0; - say(of->file, "Section \"%s\" " - "manual in %s directory", - msec, of->dsec); + mpage_dest = mlink_dest->mpage; + mlink = mpage->mlinks; + while (1) { + mlink->mpage = mpage_dest; + + /* + * If the target was already + * processed, add the links + * to the database now. + * Otherwise, this will + * happen when we come + * to the target. + */ + + if (mpage_dest->pageid) + dbadd_mlink(mlink); + + if (NULL == mlink->next) + break; + mlink = mlink->next; + } + + /* Move all links to the target. */ + + mlink->next = mlink_dest->next; + mlink_dest->next = mpage->mlinks; + mpage->mlinks = NULL; + } + goto nextpage; + } else if (NULL != mdoc) { + mpage->form = FORM_SRC; + mpage->sec = + mandoc_strdup(mdoc_meta(mdoc)->msec); + mpage->arch = mdoc_meta(mdoc)->arch; + mpage->arch = mandoc_strdup( + NULL == mpage->arch ? "" : mpage->arch); + mpage->title = + mandoc_strdup(mdoc_meta(mdoc)->title); + } else if (NULL != man) { + mpage->form = FORM_SRC; + mpage->sec = + mandoc_strdup(man_meta(man)->msec); + mpage->arch = + mandoc_strdup(mpage->mlinks->arch); + mpage->title = + mandoc_strdup(man_meta(man)->title); + } else { + mpage->form = FORM_CAT; + mpage->sec = + mandoc_strdup(mpage->mlinks->dsec); + mpage->arch = + mandoc_strdup(mpage->mlinks->arch); + mpage->title = + mandoc_strdup(mpage->mlinks->name); } + putkey(mpage, mpage->sec, TYPE_sec); + putkey(mpage, '\0' == *mpage->arch ? + any : mpage->arch, TYPE_arch); - /* - * Manual page directories exist for each kernel - * architecture as returned by machine(1). - * However, many manuals only depend on the - * application architecture as returned by arch(1). - * For example, some (2/ARM) manuals are shared - * across the "armish" and "zaurus" kernel - * architectures. - * A few manuals are even shared across completely - * different architectures, for example fdformat(1) - * on amd64, i386, sparc, and sparc64. - * Thus, warn about architecture mismatches, - * but don't skip manuals for this reason. - */ - if (warnings && !use_all && strcasecmp(march, of->arch)) { - match = 0; - say(of->file, "Architecture \"%s\" " - "manual in \"%s\" directory", - march, of->arch); + for (mlink = mpage->mlinks; mlink; 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); + putkey(mpage, mlink->name, NAME_FILE); } - if (warnings && !use_all && strcasecmp(mtitle, of->name)) - match = 0; - putkey(of, of->name, TYPE_Nm); - + assert(NULL == mpage->desc); if (NULL != mdoc) { if (NULL != (cp = mdoc_meta(mdoc)->name)) - putkey(of, cp, TYPE_Nm); - assert(NULL == of->desc); - parse_mdoc(of, mdoc_node(mdoc)); - putkey(of, NULL != of->desc ? - of->desc : of->name, TYPE_Nd); + putkey(mpage, cp, NAME_HEAD); + parse_mdoc(mpage, mdoc_node(mdoc)); } else if (NULL != man) - parse_man(of, man_node(man)); + parse_man(mpage, man_node(man)); else - parse_catpage(of); + parse_cat(mpage, fd[0]); + if (NULL == mpage->desc) + mpage->desc = mandoc_strdup(mpage->mlinks->name); - /* - * Build a title string for the file. If it matches - * the location of the file, remember the title as - * found; else, remember it as missing. - */ + if (warnings && !use_all) + for (mlink = mpage->mlinks; mlink; + mlink = mlink->next) + mlink_check(mpage, mlink); - if (check_reachable) { - if (-1 == asprintf(&title_str, "%s(%s%s%s)", mtitle, - msec, '\0' == *march ? "" : "/", march)) { - perror(NULL); - exit((int)MANDOCLEVEL_SYSERR); + 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)); } - slot = ohash_qlookup(&title_table, title_str); - title_entry = ohash_find(&title_table, slot); - if (NULL == title_entry) { - title_entry = mandoc_malloc( - sizeof(struct title)); - title_entry->title = title_str; - title_entry->file = mandoc_strdup( - match ? "" : of->file); - ohash_insert(&title_table, slot, - title_entry); - } else { - if (match) - *title_entry->file = '\0'; - free(title_str); - } } - - dbindex(mc, form, of); ohash_delete(&strings); + ohash_delete(&names); + mpage = ohash_next(&mpages, &pslot); } - if (check_reachable) { - title_entry = ohash_first(&title_table, &slot); - while (NULL != title_entry) { - if ('\0' != *title_entry->file) - say(title_entry->file, - "Probably unreachable, title is %s", - title_entry->title); - free(title_entry->title); - free(title_entry->file); - free(title_entry); - title_entry = ohash_next(&title_table, &slot); - } - ohash_delete(&title_table); + if (0 == nodb) + SQL_EXEC("END TRANSACTION"); +} + +static void +names_check(void) +{ + sqlite3_stmt *stmt; + const char *name, *sec, *arch, *key; + int irc; + + 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 (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE)) + 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); + say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec, + '\0' == *arch ? "" : "/", + '\0' == *arch ? "" : arch, key); } + sqlite3_finalize(stmt); } static void -parse_catpage(struct of *of) +parse_cat(struct mpage *mpage, int fd) { FILE *stream; char *line, *p, *title; size_t len, plen, titlesz; - if (NULL == (stream = fopen(of->file, "r"))) { + stream = (-1 == fd) ? + fopen(mpage->mlinks->file, "r") : + fdopen(fd, "r"); + if (NULL == stream) { if (warnings) - say(of->file, NULL); + say(mpage->mlinks->file, "&fopen"); return; } @@ -1116,7 +1346,7 @@ parse_catpage(struct of *of) 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, @@ -1151,10 +1381,8 @@ parse_catpage(struct of *of) if (NULL == title || '\0' == *title) { if (warnings) - say(of->file, "Cannot find NAME section"); - assert(NULL == of->desc); - of->desc = mandoc_strdup(of->name); - putkey(of, of->name, TYPE_Nd); + say(mpage->mlinks->file, + "Cannot find NAME section"); fclose(stream); free(title); return; @@ -1174,7 +1402,8 @@ parse_catpage(struct of *of) /* Skip to next word. */ ; } else { if (warnings) - say(of->file, "No dash in title line"); + say(mpage->mlinks->file, + "No dash in title line"); p = title; } @@ -1187,14 +1416,12 @@ parse_catpage(struct of *of) if (0 == len) { memmove(line, line + 1, plen--); continue; - } + } memmove(line - 1, line + 1, plen - len); plen -= 2; } - assert(NULL == of->desc); - of->desc = mandoc_strdup(p); - putkey(of, of->desc, TYPE_Nd); + mpage->desc = mandoc_strdup(p); fclose(stream); free(title); } @@ -1203,35 +1430,41 @@ parse_catpage(struct of *of) * Put a type/word pair into the word database for this particular file. */ static void -putkey(const struct of *of, const char *value, uint64_t type) +putkey(const struct mpage *mpage, char *value, uint64_t type) { + char *cp; assert(NULL != value); - putkeys(of, value, strlen(value), type); + 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); } /* * Grok all nodes at or below a certain mdoc node into putkey(). */ static void -putmdockey(const struct of *of, const struct mdoc_node *n, uint64_t m) +putmdockey(const struct mpage *mpage, + const struct mdoc_node *n, uint64_t m) { for ( ; NULL != n; n = n->next) { if (NULL != n->child) - putmdockey(of, n->child, m); + putmdockey(mpage, n->child, m); if (MDOC_TEXT == n->type) - putkey(of, n->string, m); + putkey(mpage, n->string, m); } } static void -parse_man(struct of *of, const struct man_node *n) +parse_man(struct mpage *mpage, const struct man_node *n) { const struct man_node *head, *body; - char *start, *sv, *title; + char *start, *title; char byte; - size_t sz, titlesz; + size_t sz; if (NULL == n) return; @@ -1247,16 +1480,12 @@ parse_man(struct of *of, const struct man_node *n) body = n; assert(body->parent); if (NULL != (head = body->parent->head) && - 1 == head->nchild && - NULL != (head = (head->child)) && - MAN_TEXT == head->type && - 0 == strcmp(head->string, "NAME") && - NULL != (body = body->child) && - MAN_TEXT == body->type) { + 1 == head->nchild && + NULL != (head = (head->child)) && + MAN_TEXT == head->type && + 0 == strcmp(head->string, "NAME") && + NULL != body->child) { - title = NULL; - titlesz = 0; - /* * Suck the entire NAME section into memory. * Yes, we might run away. @@ -1264,48 +1493,12 @@ parse_man(struct of *of, const struct man_node *n) * NAME sections over many lines. */ - for ( ; NULL != body; body = body->next) { - if (MAN_TEXT != body->type) - break; - if (0 == (sz = strlen(body->string))) - continue; - title = mandoc_realloc - (title, titlesz + sz + 1); - memcpy(title + titlesz, body->string, sz); - titlesz += sz + 1; - title[titlesz - 1] = ' '; - } + title = NULL; + man_deroff(&title, body); if (NULL == title) return; - title = mandoc_realloc(title, titlesz + 1); - title[titlesz] = '\0'; - - /* Skip leading space. */ - - sv = title; - while (isspace((unsigned char)*sv)) - sv++; - - if (0 == (sz = strlen(sv))) { - free(title); - return; - } - - /* Erase trailing space. */ - - start = &sv[sz - 1]; - while (start > sv && isspace((unsigned char)*start)) - *start-- = '\0'; - - if (start == sv) { - free(title); - return; - } - - start = sv; - - /* + /* * Go through a special heuristic dance here. * Conventionally, one or more manual names are * comma-specified prior to a whitespace, then a @@ -1313,6 +1506,7 @@ parse_man(struct of *of, const struct man_node *n) * the name parts here. */ + start = title; for ( ;; ) { sz = strcspn(start, " ,"); if ('\0' == start[sz]) @@ -1321,8 +1515,17 @@ parse_man(struct of *of, const struct man_node *n) byte = start[sz]; start[sz] = '\0'; - putkey(of, start, TYPE_Nm); + /* + * Assume a stray trailing comma in the + * name list if a name begins with a dash. + */ + if ('-' == start[0] || + ('\\' == start[0] && '-' == start[1])) + break; + + putkey(mpage, start, NAME_TITLE); + if (' ' == byte) { start += sz + 1; break; @@ -1334,8 +1537,8 @@ parse_man(struct of *of, const struct man_node *n) start++; } - if (sv == start) { - putkey(of, start, TYPE_Nm); + if (start == title) { + putkey(mpage, start, NAME_TITLE); free(title); return; } @@ -1357,58 +1560,60 @@ parse_man(struct of *of, const struct man_node *n) while (' ' == *start) start++; - assert(NULL == of->desc); - of->desc = mandoc_strdup(start); - putkey(of, of->desc, TYPE_Nd); + mpage->desc = mandoc_strdup(start); free(title); return; } } - for (n = n->child; n; n = n->next) - parse_man(of, n); + for (n = n->child; n; n = n->next) { + if (NULL != mpage->desc) + break; + parse_man(mpage, n); + } } static void -parse_mdoc(struct of *of, const struct mdoc_node *n) +parse_mdoc(struct mpage *mpage, 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)(of, n)) + if (0 == (*mdocs[n->tok].fp)(mpage, n)) break; if (mdocs[n->tok].mask) - putmdockey(of, n->child, mdocs[n->tok].mask); + putmdockey(mpage, n->child, + mdocs[n->tok].mask); break; default: assert(MDOC_ROOT != n->type); continue; } if (NULL != n->child) - parse_mdoc(of, n); + parse_mdoc(mpage, n); } } static int -parse_mdoc_Fd(struct of *of, const struct mdoc_node *n) +parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n) { const 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); /* @@ -1439,34 +1644,23 @@ parse_mdoc_Fd(struct of *of, const struct mdoc_node *n end--; if (end > start) - putkeys(of, start, end - start + 1, TYPE_In); - return(1); + putkeys(mpage, start, end - start + 1, TYPE_In); + return(0); } static int -parse_mdoc_In(struct of *of, const struct mdoc_node *n) +parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n) { + char *cp; - if (NULL != n->child && MDOC_TEXT == n->child->type) - return(0); - - putkey(of, n->child->string, TYPE_In); - return(1); -} - -static int -parse_mdoc_Fn(struct of *of, const struct mdoc_node *n) -{ - const char *cp; - 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, ' '))) @@ -1475,109 +1669,73 @@ parse_mdoc_Fn(struct of *of, const struct mdoc_node *n while ('*' == *cp) cp++; - putkey(of, cp, TYPE_Fn); + putkey(mpage, cp, TYPE_Fn); if (n->string < cp) - putkeys(of, n->string, cp - n->string, TYPE_Ft); + putkeys(mpage, n->string, cp - n->string, TYPE_Ft); for (n = n->next; NULL != n; n = n->next) if (MDOC_TEXT == n->type) - putkey(of, n->string, TYPE_Fa); + putkey(mpage, n->string, TYPE_Fa); return(0); } static int -parse_mdoc_St(struct of *of, const struct mdoc_node *n) +parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n) { - - if (NULL == n->child || MDOC_TEXT != n->child->type) - return(0); - - putkey(of, n->child->string, TYPE_St); - return(1); -} - -static int -parse_mdoc_Xr(struct of *of, const struct mdoc_node *n) -{ char *cp; if (NULL == (n = n->child)) return(0); if (NULL == n->next) { - putkey(of, n->string, TYPE_Xr); + putkey(mpage, n->string, TYPE_Xr); return(0); } - if (-1 == asprintf(&cp, "%s(%s)", n->string, n->next->string)) { - perror(NULL); - exit((int)MANDOCLEVEL_SYSERR); - } - putkey(of, cp, TYPE_Xr); + mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string); + putkey(mpage, cp, TYPE_Xr); free(cp); return(0); } static int -parse_mdoc_Nd(struct of *of, const struct mdoc_node *n) +parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n) { - size_t sz; - if (MDOC_BODY != n->type) - return(0); - - /* - * Special-case the `Nd' because we need to put the description - * into the document table. - */ - - for (n = n->child; NULL != n; n = n->next) { - if (MDOC_TEXT == n->type) { - if (NULL != of->desc) { - sz = strlen(of->desc) + - strlen(n->string) + 2; - of->desc = mandoc_realloc(of->desc, sz); - strlcat(of->desc, " ", sz); - strlcat(of->desc, n->string, sz); - } else - of->desc = mandoc_strdup(n->string); - } - if (NULL != n->child) - parse_mdoc_Nd(of, n); - } - return(1); + if (MDOC_BODY == n->type) + mdoc_deroff(&mpage->desc, n); + return(0); } static int -parse_mdoc_Nm(struct of *of, const struct mdoc_node *n) +parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n) { if (SEC_NAME == n->sec) - return(1); - else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type) - return(0); - - return(1); + putmdockey(mpage, n->child, NAME_TITLE); + else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type) + putmdockey(mpage, n->child, NAME_SYN); + return(0); } static int -parse_mdoc_Sh(struct of *of, const struct mdoc_node *n) +parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n) { return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type); } static int -parse_mdoc_head(struct of *of, const struct mdoc_node *n) +parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n) { return(MDOC_HEAD == n->type); } static int -parse_mdoc_body(struct of *of, const struct mdoc_node *n) +parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n) { return(MDOC_BODY == n->type); @@ -1589,28 +1747,48 @@ parse_mdoc_body(struct of *of, const struct mdoc_node * When we finish the manual, we'll dump the table. */ static void -putkeys(const struct of *of, const char *cp, size_t sz, uint64_t v) +putkeys(const struct mpage *mpage, + const char *cp, size_t sz, uint64_t v) { + struct ohash *htab; struct str *s; - unsigned int slot; const char *end; + unsigned int slot; + int i; if (0 == sz) return; + if (TYPE_Nm & v) { + htab = &names; + v &= name_mask; + name_mask &= ~NAME_FIRST; + if (debug > 1) + say(mpage->mlinks->file, + "Adding name %*s", sz, cp); + } else { + htab = &strings; + if (debug > 1) + for (i = 0; i < mansearch_keymax; i++) + if (1 << i & v) + say(mpage->mlinks->file, + "Adding key %s=%*s", + mansearch_keynames[i], sz, cp); + } + end = cp + sz; - slot = ohash_qlookupi(&strings, cp, &end); - s = ohash_find(&strings, slot); + slot = ohash_qlookupi(htab, cp, &end); + s = ohash_find(htab, slot); - if (NULL != s && of == s->of) { + if (NULL != s && mpage == s->mpage) { s->mask |= v; return; } else if (NULL == s) { - s = mandoc_calloc(sizeof(struct str) + sz + 1, 1); + s = mandoc_calloc(1, sizeof(struct str) + sz + 1); memcpy(s->key, cp, sz); - ohash_insert(&strings, slot, s); + ohash_insert(htab, slot, s); } - s->of = of; + s->mpage = mpage; s->mask = v; } @@ -1667,26 +1845,27 @@ utf8(unsigned int cp, char out[7]) } /* - * Store the UTF-8 version of a key, or alias the pointer if the key has - * no UTF-8 transcription marks in it. + * Store the rendered version of a key, or alias the pointer + * if the key contains no escape sequences. */ static void -utf8key(struct mchars *mc, struct str *key) +render_key(struct mchars *mc, struct str *key) { size_t sz, bsz, pos; - char utfbuf[7], res[5]; + char utfbuf[7], res[6]; char *buf; const char *seq, *cpp, *val; int len, u; enum mandoc_esc esc; - assert(NULL == key->utf8); + assert(NULL == key->rendered); res[0] = '\\'; res[1] = '\t'; res[2] = ASCII_NBRSP; res[3] = ASCII_HYPH; - res[4] = '\0'; + res[4] = ASCII_BREAK; + res[5] = '\0'; val = key->key; bsz = strlen(val); @@ -1696,9 +1875,9 @@ utf8key(struct mchars *mc, struct str *key) * pointer as ourselvse and get out of here. */ if (strcspn(val, res) == bsz) { - key->utf8 = key->key; + key->rendered = key->key; return; - } + } /* Pre-allocate by the length of the input */ @@ -1717,138 +1896,210 @@ utf8key(struct mchars *mc, struct str *key) val += sz; } - if (ASCII_HYPH == *val) { + switch (*val) { + case ASCII_HYPH: buf[pos++] = '-'; val++; continue; - } else if ('\t' == *val || ASCII_NBRSP == *val) { + case '\t': + /* FALLTHROUGH */ + case ASCII_NBRSP: buf[pos++] = ' '; val++; + /* FALLTHROUGH */ + case ASCII_BREAK: continue; - } else if ('\\' != *val) + default: break; + } + if ('\\' != *val) + break; /* Read past the slash. */ val++; - u = 0; /* * Parse the escape sequence and see if it's a * 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) continue; - if (0 == (u = mchars_spec2cp(mc, seq, len))) - continue; /* - * If we have a Unicode codepoint, try to convert that - * to a UTF-8 byte string. + * Render the special character + * as either UTF-8 or ASCII. */ - cpp = utfbuf; - if (0 == (sz = utf8(u, utfbuf))) - continue; + if (write_utf8) { + if (0 == (u = mchars_spec2cp(mc, seq, len))) + continue; + cpp = utfbuf; + if (0 == (sz = utf8(u, utfbuf))) + continue; + sz = strlen(cpp); + } else { + cpp = mchars_spec2str(mc, seq, len, &sz); + if (NULL == cpp) + continue; + if (ASCII_NBRSP == *cpp) { + cpp = " "; + sz = 1; + } + } + /* Copy the rendered glyph into the stream. */ - sz = strlen(cpp); bsz += sz; - buf = mandoc_realloc(buf, bsz); - memcpy(&buf[pos], cpp, sz); pos += sz; } buf[pos] = '\0'; - key->utf8 = buf; + key->rendered = buf; } +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]); +} + /* * Flush the current page's terms (and their bits) into the database. * Wrap the entire set of additions in a transaction to make sqlite be a * little faster. - * Also, UTF-8-encode the description at the last possible moment. + * Also, handle escape sequences at the last possible moment. */ static void -dbindex(struct mchars *mc, int form, const struct of *of) +dbadd(struct mpage *mpage, struct mchars *mc) { + struct mlink *mlink; struct str *key; - const char *desc; - int64_t recno; size_t i; unsigned int slot; - if (verb) - say(of->file, "Adding to index"); + mlink = mpage->mlinks; - if (nodb) + if (nodb) { + while (NULL != mlink) { + fputs(mlink->name, stdout); + if (NULL == mlink->next || + strcmp(mlink->dsec, mlink->next->dsec) || + strcmp(mlink->fsec, mlink->next->fsec) || + strcmp(mlink->arch, mlink->next->arch)) { + putchar('('); + if ('\0' == *mlink->dsec) + fputs(mlink->fsec, stdout); + else + fputs(mlink->dsec, stdout); + if ('\0' != *mlink->arch) + printf("/%s", mlink->arch); + putchar(')'); + } + mlink = mlink->next; + if (NULL != mlink) + fputs(", ", stdout); + } + printf(" - %s\n", mpage->desc); return; - - desc = ""; - if (NULL != of->desc && '\0' != *of->desc) { - key = ohash_find(&strings, - ohash_qlookup(&strings, of->desc)); - assert(NULL != key); - if (NULL == key->utf8) - utf8key(mc, key); - desc = key->utf8; } - SQL_EXEC("BEGIN TRANSACTION"); + if (debug) + say(mlink->file, "Adding to database"); i = 1; - SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->file); - SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->sec); - SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, of->arch); - SQL_BIND_TEXT(stmts[STMT_INSERT_DOC], i, desc); - SQL_BIND_INT(stmts[STMT_INSERT_DOC], i, form); - SQL_STEP(stmts[STMT_INSERT_DOC]); - recno = sqlite3_last_insert_rowid(db); - sqlite3_reset(stmts[STMT_INSERT_DOC]); + SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->desc); + SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form); + SQL_STEP(stmts[STMT_INSERT_PAGE]); + mpage->pageid = sqlite3_last_insert_rowid(db); + sqlite3_reset(stmts[STMT_INSERT_PAGE]); + while (NULL != mlink) { + 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); + 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_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->of == of); - if (NULL == key->utf8) - utf8key(mc, key); + 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->utf8); - SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno); + SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered); + SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid); SQL_STEP(stmts[STMT_INSERT_KEY]); sqlite3_reset(stmts[STMT_INSERT_KEY]); - if (key->utf8 != key->key) - free(key->utf8); + if (key->rendered != key->key) + free(key->rendered); free(key); } - - SQL_EXEC("END TRANSACTION"); } static void dbprune(void) { - struct of *of; + struct mpage *mpage; + struct mlink *mlink; size_t i; + unsigned int slot; - if (nodb) - return; + if (0 == nodb) + SQL_EXEC("BEGIN TRANSACTION"); - for (of = ofs; NULL != of; of = of->next) { - i = 1; - SQL_BIND_TEXT(stmts[STMT_DELETE], i, of->file); - SQL_STEP(stmts[STMT_DELETE]); - sqlite3_reset(stmts[STMT_DELETE]); - if (verb) - say(of->file, "Deleted from index"); + 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]); + } } + + if (0 == nodb) + SQL_EXEC("END TRANSACTION"); } /* @@ -1859,6 +2110,8 @@ static void dbclose(int real) { size_t i; + int status; + pid_t child; if (nodb) return; @@ -1874,10 +2127,60 @@ dbclose(int real) if (real) return; - if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) { + if ('\0' == *tempfilename) { + if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say(MANDOC_DB, "&rename"); + } + return; + } + + switch (child = fork()) { + case -1: exitcode = (int)MANDOCLEVEL_SYSERR; - say(MANDOC_DB, NULL); + say("", "&fork cmp"); + return; + case 0: + execlp("cmp", "cmp", "-s", + tempfilename, MANDOC_DB, 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"); + } + + *strrchr(tempfilename, '/') = '\0'; + switch (child = fork()) { + case -1: + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "&fork rm"); + return; + case 0: + execlp("rm", "rm", "-rf", tempfilename, 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); + } } /* @@ -1891,76 +2194,107 @@ dbclose(int real) static int dbopen(int real) { - const char *file, *sql; + const char *sql; int rc, ofl; - if (nodb) + if (nodb) return(1); + *tempfilename = '\0'; ofl = SQLITE_OPEN_READWRITE; - if (0 == real) { - file = MANDOC_DB "~"; - if (-1 == remove(file) && ENOENT != errno) { + + if (real) { + rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL); + if (SQLITE_OK != rc) { exitcode = (int)MANDOCLEVEL_SYSERR; - say(file, NULL); + say(MANDOC_DB, "%s", sqlite3_errmsg(db)); return(0); } - ofl |= SQLITE_OPEN_EXCLUSIVE; - } else - file = MANDOC_DB; - - rc = sqlite3_open_v2(file, &db, ofl, NULL); - if (SQLITE_OK == rc) goto prepare_statements; - if (SQLITE_CANTOPEN != rc) { + } + + 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(file, NULL); + say(MANDOC_DB "~", "%s", sqlite3_errmsg(db)); return(0); } - sqlite3_close(db); - db = NULL; - - if (SQLITE_OK != (rc = sqlite3_open(file, &db))) { + (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX", + sizeof(tempfilename)); + if (NULL == mkdtemp(tempfilename)) { exitcode = (int)MANDOCLEVEL_SYSERR; - say(file, NULL); + say("", "&%s", 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)); + return(0); + } - sql = "CREATE TABLE \"docs\" (\n" - " \"file\" TEXT NOT NULL,\n" - " \"sec\" TEXT NOT NULL,\n" - " \"arch\" TEXT NOT NULL,\n" +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(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" + ");\n" + "\n" "CREATE TABLE \"keys\" (\n" " \"bits\" INTEGER NOT NULL,\n" " \"key\" TEXT NOT NULL,\n" - " \"docid\" INTEGER NOT NULL REFERENCES docs(id) " - "ON DELETE CASCADE,\n" - " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n" + " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) " + "ON DELETE CASCADE\n" ");\n" - "\n" - "CREATE INDEX \"key_index\" ON keys (key);\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(file, "%s", sqlite3_errmsg(db)); + say(MANDOC_DB, "%s", sqlite3_errmsg(db)); return(0); } prepare_statements: SQL_EXEC("PRAGMA foreign_keys = ON"); - sql = "DELETE FROM docs where file=?"; - sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE], NULL); - sql = "INSERT INTO docs " - "(file,sec,arch,desc,form) VALUES (?,?,?,?,?)"; - sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_DOC], NULL); + sql = "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 = "INSERT INTO names " + "(bits,name,pageid) VALUES (?,?,?)"; + sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL); sql = "INSERT INTO keys " - "(bits,key,docid) VALUES (?,?,?)"; + "(bits,key,pageid) VALUES (?,?,?)"; sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL); #ifndef __APPLE__ @@ -1980,7 +2314,7 @@ static void * hash_halloc(size_t sz, void *arg) { - return(mandoc_calloc(sz, 1)); + return(mandoc_calloc(1, sz)); } static void * @@ -2013,12 +2347,12 @@ set_basedir(const char *targetdir) if (NULL == getcwd(startdir, PATH_MAX)) { exitcode = (int)MANDOCLEVEL_SYSERR; if (NULL != targetdir) - say(".", NULL); + say("", "&getcwd"); return(0); } if (-1 == (fd = open(startdir, O_RDONLY, 0))) { exitcode = (int)MANDOCLEVEL_SYSERR; - say(startdir, NULL); + say("", "&open %s", startdir); return(0); } if (NULL == targetdir) @@ -2030,7 +2364,7 @@ set_basedir(const char *targetdir) close(fd); basedir[0] = '\0'; exitcode = (int)MANDOCLEVEL_SYSERR; - say(startdir, NULL); + say("", "&chdir %s", startdir); return(0); } if (NULL == targetdir) { @@ -2041,11 +2375,11 @@ set_basedir(const char *targetdir) if (NULL == realpath(targetdir, basedir)) { basedir[0] = '\0'; exitcode = (int)MANDOCLEVEL_BADARG; - say(targetdir, NULL); + say("", "&%s: realpath", targetdir); return(0); } else if (-1 == chdir(basedir)) { exitcode = (int)MANDOCLEVEL_BADARG; - say("", NULL); + say("", "&chdir"); return(0); } return(1); @@ -2055,6 +2389,7 @@ static void say(const char *file, const char *format, ...) { va_list ap; + int use_errno; if ('\0' != *basedir) fprintf(stderr, "%s", basedir); @@ -2062,16 +2397,32 @@ say(const char *file, const char *format, ...) fputs("//", stderr); if ('\0' != *file) fprintf(stderr, "%s", file); - fputs(": ", stderr); - if (NULL == format) { - perror(NULL); - return; + use_errno = 1; + if (NULL != format) { + switch (*format) { + case '&': + format++; + break; + case '\0': + format = NULL; + break; + default: + use_errno = 0; + break; + } } - - va_start(ap, format); - vfprintf(stderr, format, ap); - va_end(ap); - - fputc('\n', stderr); + if (NULL != format) { + if ('\0' != *basedir || '\0' != *file) + fputs(": ", stderr); + va_start(ap, format); + vfprintf(stderr, format, ap); + va_end(ap); + } + if (use_errno) { + if ('\0' != *basedir || '\0' != *file || NULL != format) + fputs(": ", stderr); + perror(NULL); + } else + fputc('\n', stderr); }