=================================================================== RCS file: /cvs/mandoc/mandocdb.c,v retrieving revision 1.56 retrieving revision 1.67 diff -u -p -r1.56 -r1.67 --- mandoc/mandocdb.c 2013/06/03 19:01:31 1.56 +++ mandoc/mandocdb.c 2013/06/07 04:51:26 1.67 @@ -1,4 +1,4 @@ -/* $Id: mandocdb.c,v 1.56 2013/06/03 19:01:31 schwarze Exp $ */ +/* $Id: mandocdb.c,v 1.67 2013/06/07 04:51:26 schwarze Exp $ */ /* * Copyright (c) 2011, 2012 Kristaps Dzonsons * Copyright (c) 2011, 2012, 2013 Ingo Schwarze @@ -19,7 +19,6 @@ #include "config.h" #endif -#include #include #include @@ -28,7 +27,9 @@ #include #include #include +#include #include +#include #include #include #include @@ -83,7 +84,6 @@ enum form { struct str { char *utf8; /* key in UTF-8 form */ const struct of *of; /* if set, the owning parse */ - struct str *next; /* next in owning parse sequence */ uint64_t mask; /* bitmask in sequence */ char key[]; /* the string itself */ }; @@ -98,12 +98,12 @@ struct of { struct of *next; /* next in ofs */ enum form dform; /* path-cued form */ enum form sform; /* suffix-cued form */ - char file[MAXPATHLEN]; /* filename rel. to manpath */ - const char *desc; /* parsed description */ - const char *sec; /* suffix-cued section (or empty) */ - const char *dsec; /* path-cued section (or empty) */ - const char *arch; /* path-cued arch. (or empty) */ - const char *name; /* name (from filename) (not empty) */ + 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) */ }; enum stmt { @@ -122,28 +122,25 @@ struct mdoc_handler { #define MDOCF_CHILD 0x01 /* automatically index child nodes */ }; -static void dbclose(const char *, int); -static void dbindex(struct mchars *, int, - const struct of *, const char *); -static int dbopen(const char *, int); -static void dbprune(const char *); +static void dbclose(int); +static void dbindex(struct mchars *, int, const struct of *); +static int dbopen(int); +static void dbprune(void); static void fileadd(struct of *); static int filecheck(const char *); -static void filescan(const char *, const char *); -static struct str *hashget(const char *, size_t); +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(const char *, int, const char *, - const char *, const char *, const char *, - const char *, 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 int ofmerge(struct mchars *, struct mparse *, const char *); -static void parse_catpage(struct of *, const char *); -static int parse_man(struct of *, - const struct man_node *); +static void ofmerge(struct mchars *, struct mparse *, + struct ohash_info*); +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 *); @@ -155,33 +152,30 @@ static int parse_mdoc_Nm(struct of *, const struct md 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 path_reset(const char *, int, const char *); +static int set_basedir(const char *); static void putkey(const struct of *, const char *, uint64_t); static void putkeys(const struct of *, - const char *, int, uint64_t); + const char *, size_t, uint64_t); static void putmdockey(const struct of *, const struct mdoc_node *, uint64_t); -static void say(const char *, const char *, const char *, ...); -static char *stradd(const char *); -static char *straddbuf(const char *, size_t); -static int treescan(const char *); +static void say(const char *, const char *, ...); +static int treescan(void); static size_t utf8(unsigned int, char [7]); static void utf8key(struct mchars *, struct str *); -static void wordaddbuf(const struct of *, - const char *, size_t, uint64_t); 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 exitcode; /* to be returned by main */ 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 strings; /* table of all strings */ static struct of *ofs = NULL; /* vector of files to parse */ -static struct str *words = NULL; /* word list in current parse */ static sqlite3 *db = NULL; /* current database */ static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */ @@ -313,12 +307,9 @@ static const struct mdoc_handler mdocs[MDOC_MAX] = { int main(int argc, char *argv[]) { - char cwd[MAXPATHLEN]; - int ch, rc, fd, i; - unsigned int index; + int ch, i; size_t j, sz; - const char *dir; - struct str *s; + const char *path_arg; struct mchars *mc; struct manpaths dirs; struct mparse *mp; @@ -345,20 +336,6 @@ main(int argc, char *argv[]) ++progname; /* - * 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. - */ - if (NULL == getcwd(cwd, MAXPATHLEN)) { - perror(NULL); - return(EXIT_FAILURE); - } else if (-1 == (fd = open(cwd, O_RDONLY, 0))) { - perror(cwd); - return(EXIT_FAILURE); - } - - /* * We accept a few different invocations. * The CHECKOP macro makes sure that invocation styles don't * clobber each other. @@ -369,7 +346,7 @@ main(int argc, char *argv[]) goto usage; \ } while (/*CONSTCOND*/0) - dir = NULL; + path_arg = NULL; op = OP_DEFAULT; while (-1 != (ch = getopt(argc, argv, "aC:d:ntu:vW"))) @@ -379,12 +356,12 @@ main(int argc, char *argv[]) break; case ('C'): CHECKOP(op, ch); - dir = optarg; + path_arg = optarg; op = OP_CONFFILE; break; case ('d'): CHECKOP(op, ch); - dir = optarg; + path_arg = optarg; op = OP_UPDATE; break; case ('n'): @@ -398,7 +375,7 @@ main(int argc, char *argv[]) break; case ('u'): CHECKOP(op, ch); - dir = optarg; + path_arg = optarg; op = OP_DELETE; break; case ('v'): @@ -419,12 +396,11 @@ main(int argc, char *argv[]) goto usage; } - rc = 1; + exitcode = (int)MANDOCLEVEL_OK; mp = mparse_alloc(MPARSE_AUTO, MANDOCLEVEL_FATAL, NULL, NULL, NULL); mc = mchars_alloc(); - ohash_init(&strings, 6, &str_info); ohash_init(&inos, 6, &ino_info); ohash_init(&filenames, 6, &filename_info); @@ -433,24 +409,23 @@ main(int argc, char *argv[]) * Force processing all files. */ use_all = 1; - if (NULL == dir) - dir = cwd; + /* * All of these deal with a specific directory. * Jump into that directory then collect files specified * on the command-line. */ - if (0 == path_reset(cwd, fd, dir)) + if (0 == set_basedir(path_arg)) goto out; for (i = 0; i < argc; i++) - filescan(argv[i], dir); - if (0 == dbopen(dir, 1)) + filescan(argv[i]); + if (0 == dbopen(1)) goto out; if (OP_TEST != op) - dbprune(dir); + dbprune(); if (OP_DELETE != op) - rc = ofmerge(mc, mp, dir); - dbclose(dir, 1); + ofmerge(mc, mp, &str_info); + dbclose(1); } else { /* * If we have arguments, use them as our manpaths. @@ -464,7 +439,7 @@ main(int argc, char *argv[]) for (i = 0; i < argc; i++) dirs.paths[i] = mandoc_strdup(argv[i]); } else - manpath_parse(&dirs, dir, NULL, NULL); + manpath_parse(&dirs, path_arg, NULL, NULL); /* * First scan the tree rooted at a base directory. @@ -479,13 +454,19 @@ main(int argc, char *argv[]) dirs.paths[j][--sz] = '\0'; if (0 == sz) continue; - if (0 == path_reset(cwd, fd, dirs.paths[j])) + + if (j) { + ohash_init(&inos, 6, &ino_info); + ohash_init(&filenames, 6, &filename_info); + } + + if (0 == set_basedir(dirs.paths[j])) goto out; - if (0 == treescan(dirs.paths[j])) + if (0 == treescan()) goto out; - if (0 == path_reset(cwd, fd, dirs.paths[j])) + if (0 == set_basedir(dirs.paths[j])) goto out; - if (0 == dbopen(dirs.paths[j], 0)) + if (0 == dbopen(0)) goto out; /* @@ -497,32 +478,25 @@ main(int argc, char *argv[]) SQL_EXEC("PRAGMA synchronous = OFF"); #endif - if (0 == ofmerge(mc, mp, dirs.paths[j])) - goto out; - dbclose(dirs.paths[j], 0); - offree(); - ohash_delete(&inos); - ohash_init(&inos, 6, &ino_info); - ohash_delete(&filenames); - ohash_init(&filenames, 6, &filename_info); + ofmerge(mc, mp, &str_info); + dbclose(0); + + if (j + 1 < dirs.sz) { + ohash_delete(&inos); + ohash_delete(&filenames); + offree(); + } } } out: - close(fd); + set_basedir(NULL); manpath_free(&dirs); mchars_free(mc); mparse_free(mp); - for (s = ohash_first(&strings, &index); - NULL != s; s = ohash_next(&strings, &index)) { - if (s->utf8 != s->key) - free(s->utf8); - free(s); - } - ohash_delete(&strings); ohash_delete(&inos); ohash_delete(&filenames); offree(); - return(rc ? EXIT_SUCCESS : EXIT_FAILURE); + return(exitcode); usage: fprintf(stderr, "usage: %s [-anvW] [-C file]\n" " %s [-anvW] dir ...\n" @@ -532,11 +506,11 @@ usage: progname, progname, progname, progname, progname); - return(EXIT_FAILURE); + return((int)MANDOCLEVEL_BADARG); } /* - * Scan a directory tree rooted at "base" for manpages. + * Scan a directory tree rooted at "basedir" for manpages. * We use fts(), scanning directory parts along the way for clues to our * section and architecture. * @@ -550,13 +524,13 @@ usage: * TODO: accomodate for multi-language directories. */ static int -treescan(const char *base) +treescan(void) { FTS *f; FTSENT *ff; int dform; char *sec; - const char *dsec, *arch, *cp, *name, *path; + const char *dsec, *arch, *cp, *path; const char *argv[2]; argv[0] = "."; @@ -568,7 +542,8 @@ treescan(const char *base) */ f = fts_open((char * const *)argv, FTS_LOGICAL, NULL); if (NULL == f) { - perror(base); + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", NULL); return(0); } @@ -583,53 +558,57 @@ treescan(const char *base) * Disallow duplicate (hard-linked) files. */ if (FTS_F == ff->fts_info) { + if (0 == strcmp(path, MANDOC_DB)) + continue; if ( ! use_all && ff->fts_level < 2) { if (warnings) - say(base, path, "Extraneous file"); + say(path, "Extraneous file"); continue; } else if (inocheck(ff->fts_statp)) { if (warnings) - say(base, path, "Duplicate file"); + say(path, "Duplicate file"); continue; - } - - cp = ff->fts_name; - - if (0 == strcmp(cp, "mandocdb.db")) { - if (warnings) - say(base, path, "Skip database"); - continue; - } else if (NULL != (cp = strrchr(cp, '.'))) { - if (0 == strcmp(cp + 1, "html")) { + } else if (NULL == (sec = + strrchr(ff->fts_name, '.'))) { + if ( ! use_all) { if (warnings) - say(base, path, "Skip html"); + say(path, + "No filename suffix"); continue; - } else if (0 == strcmp(cp + 1, "gz")) { - if (warnings) - say(base, path, "Skip gz"); - continue; - } else if (0 == strcmp(cp + 1, "ps")) { - if (warnings) - say(base, path, "Skip ps"); - continue; - } else if (0 == strcmp(cp + 1, "pdf")) { - if (warnings) - say(base, path, "Skip pdf"); - continue; } - } - - if (NULL != (sec = strrchr(ff->fts_name, '.'))) { - *sec = '\0'; - sec = stradd(sec + 1); - } - name = stradd(ff->fts_name); - ofadd(base, dform, path, - name, dsec, sec, arch, ff->fts_statp); + } else if (0 == strcmp(++sec, "html")) { + if (warnings) + say(path, "Skip html"); + continue; + } else if (0 == strcmp(sec, "gz")) { + if (warnings) + say(path, "Skip gz"); + continue; + } else if (0 == strcmp(sec, "ps")) { + if (warnings) + say(path, "Skip ps"); + continue; + } else if (0 == strcmp(sec, "pdf")) { + if (warnings) + say(path, "Skip pdf"); + continue; + } else if ( ! use_all && + ((FORM_SRC == dform && strcmp(sec, dsec)) || + (FORM_CAT == dform && strcmp(sec, "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); continue; } else if (FTS_D != ff->fts_info && - FTS_DP != ff->fts_info) + FTS_DP != ff->fts_info) { + if (warnings) + say(path, "Not a regular file"); continue; + } switch (ff->fts_level) { case (0): @@ -649,17 +628,17 @@ treescan(const char *base) if (0 == strncmp(cp, "man", 3)) { dform = FORM_SRC; - dsec = stradd(cp + 3); + dsec = cp + 3; } else if (0 == strncmp(cp, "cat", 3)) { dform = FORM_CAT; - dsec = stradd(cp + 3); + dsec = cp + 3; } if (NULL != dsec || use_all) break; if (warnings) - say(base, path, "Unknown directory part"); + say(path, "Unknown directory part"); fts_set(f, ff, FTS_SKIP); break; case (2): @@ -669,13 +648,13 @@ treescan(const char *base) */ arch = NULL; if (FTS_DP != ff->fts_info && NULL != dsec) - arch = stradd(ff->fts_name); + arch = ff->fts_name; break; default: if (FTS_DP == ff->fts_info || use_all) break; if (warnings) - say(base, path, "Extraneous directory part"); + say(path, "Extraneous directory part"); fts_set(f, ff, FTS_SKIP); break; } @@ -701,10 +680,11 @@ treescan(const char *base) * See treescan() for the fts(3) version of this. */ static void -filescan(const char *file, const char *base) +filescan(const char *file) { + char buf[PATH_MAX]; const char *sec, *arch, *name, *dsec; - char *p, *start, *buf; + char *p, *start; int dform; struct stat st; @@ -713,22 +693,28 @@ filescan(const char *file, const char *base) if (0 == strncmp(file, "./", 2)) file += 2; - if (-1 == stat(file, &st)) { - if (warnings) - say(base, file, "%s", strerror(errno)); + if (NULL == realpath(file, buf)) { + exitcode = (int)MANDOCLEVEL_BADARG; + say(file, NULL); return; + } else if (strstr(buf, basedir) != buf) { + exitcode = (int)MANDOCLEVEL_BADARG; + say("", "%s: outside base directory", buf); + return; + } else if (-1 == stat(buf, &st)) { + exitcode = (int)MANDOCLEVEL_BADARG; + say(file, NULL); + return; } else if ( ! (S_IFREG & st.st_mode)) { - if (warnings) - say(base, file, "Not a regular file"); + exitcode = (int)MANDOCLEVEL_BADARG; + say(file, "Not a regular file"); return; } else if (inocheck(&st)) { if (warnings) - say(base, file, "Duplicate file"); + say(file, "Duplicate file"); return; } - - buf = mandoc_strdup(file); - start = buf; + start = buf + strlen(basedir); sec = arch = name = dsec = NULL; dform = FORM_NONE; @@ -779,8 +765,7 @@ filescan(const char *file, const char *base) *p = '\0'; } - ofadd(base, dform, file, name, dsec, sec, arch, &st); - free(buf); + ofadd(dform, file, name, dsec, sec, arch, &st); } /* @@ -789,10 +774,9 @@ filescan(const char *file, const char *base) static int filecheck(const char *name) { - unsigned int index; - index = ohash_qlookup(&filenames, name); - return(NULL != ohash_find(&filenames, index)); + return(NULL != ohash_find(&filenames, + ohash_qlookup(&filenames, name))); } /* @@ -802,11 +786,11 @@ filecheck(const char *name) static void fileadd(struct of *of) { - unsigned int index; + unsigned int slot; - index = ohash_qlookup(&filenames, of->file); - assert(NULL == ohash_find(&filenames, index)); - ohash_insert(&filenames, index, of); + slot = ohash_qlookup(&filenames, of->file); + assert(NULL == ohash_find(&filenames, slot)); + ohash_insert(&filenames, slot, of); } /* @@ -817,15 +801,13 @@ inocheck(const struct stat *st) { struct id id; uint32_t hash; - unsigned int index; memset(&id, 0, sizeof(id)); id.ino = hash = st->st_ino; id.dev = st->st_dev; - index = ohash_lookup_memory - (&inos, (char *)&id, sizeof(id), hash); - return(NULL != ohash_find(&inos, index)); + return(NULL != ohash_find(&inos, ohash_lookup_memory( + &inos, (char *)&id, sizeof(id), hash))); } /* @@ -837,21 +819,20 @@ static void inoadd(const struct stat *st, struct of *of) { uint32_t hash; - unsigned int index; + unsigned int slot; of->id.ino = hash = st->st_ino; of->id.dev = st->st_dev; - index = ohash_lookup_memory + slot = ohash_lookup_memory (&inos, (char *)&of->id, sizeof(of->id), hash); - assert(NULL == ohash_find(&inos, index)); - ohash_insert(&inos, index, of); + assert(NULL == ohash_find(&inos, slot)); + ohash_insert(&inos, slot, of); } static void -ofadd(const char *base, int dform, const char *file, - const char *name, const char *dsec, const char *sec, - const char *arch, const struct stat *st) +ofadd(int dform, const char *file, const char *name, const char *dsec, + const char *sec, const char *arch, const struct stat *st) { struct of *of; int sform; @@ -876,11 +857,11 @@ ofadd(const char *base, int dform, const char *file, } of = mandoc_calloc(1, sizeof(struct of)); - strlcpy(of->file, file, MAXPATHLEN); - of->name = name; - of->sec = sec; - of->dsec = dsec; - of->arch = arch; + 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; @@ -902,25 +883,30 @@ offree(void) while (NULL != (of = ofs)) { ofs = of->next; + free(of->name); + free(of->sec); + free(of->dsec); + free(of->arch); free(of); } } /* * Run through the files in the global vector "ofs" and add them to the - * database specified in "base". + * 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 int -ofmerge(struct mchars *mc, struct mparse *mp, const char *base) +static void +ofmerge(struct mchars *mc, struct mparse *mp, + struct ohash_info *infop) { int form; size_t sz; struct mdoc *mdoc; struct man *man; - char buf[MAXPATHLEN]; + char buf[PATH_MAX]; char *bufp; const char *msec, *march, *mtitle, *cp; struct of *of; @@ -935,11 +921,10 @@ ofmerge(struct mchars *mc, struct mparse *mp, const ch * own. */ if ( ! use_all && FORM_CAT == of->dform) { - sz = strlcpy(buf, of->file, MAXPATHLEN); - if (sz >= MAXPATHLEN) { + sz = strlcpy(buf, of->file, PATH_MAX); + if (sz >= PATH_MAX) { if (warnings) - say(base, of->file, - "Filename too long"); + say(of->file, "Filename too long"); continue; } bufp = strstr(buf, "cat"); @@ -947,16 +932,16 @@ ofmerge(struct mchars *mc, struct mparse *mp, const ch memcpy(bufp, "man", 3); if (NULL != (bufp = strrchr(buf, '.'))) *++bufp = '\0'; - strlcat(buf, of->dsec, MAXPATHLEN); + strlcat(buf, of->dsec, PATH_MAX); if (filecheck(buf)) { if (warnings) - say(base, of->file, "Man " + say(of->file, "Man " "source exists: %s", buf); continue; } } - words = NULL; + ohash_init(&strings, 6, infop); mparse_reset(mp); mdoc = NULL; man = NULL; @@ -1006,7 +991,7 @@ ofmerge(struct mchars *mc, struct mparse *mp, const ch */ if (warnings && !use_all && form && strcasecmp(msec, of->dsec)) - say(base, of->file, "Section \"%s\" " + say(of->file, "Section \"%s\" " "manual in %s directory", msec, of->dsec); @@ -1025,7 +1010,7 @@ ofmerge(struct mchars *mc, struct mparse *mp, const ch * but don't skip manuals for this reason. */ if (warnings && !use_all && strcasecmp(march, of->arch)) - say(base, of->file, "Architecture \"%s\" " + say(of->file, "Architecture \"%s\" " "manual in \"%s\" directory", march, of->arch); @@ -1034,20 +1019,22 @@ ofmerge(struct mchars *mc, struct mparse *mp, const ch 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); } else if (NULL != man) parse_man(of, man_node(man)); else - parse_catpage(of, base); + parse_catpage(of); - dbindex(mc, form, of, base); + dbindex(mc, form, of); + ohash_delete(&strings); } - - return(1); } static void -parse_catpage(struct of *of, const char *base) +parse_catpage(struct of *of) { FILE *stream; char *line, *p, *title; @@ -1055,7 +1042,7 @@ parse_catpage(struct of *of, const char *base) if (NULL == (stream = fopen(of->file, "r"))) { if (warnings) - say(base, of->file, "%s", strerror(errno)); + say(of->file, NULL); return; } @@ -1108,7 +1095,10 @@ parse_catpage(struct of *of, const char *base) if (NULL == title || '\0' == *title) { if (warnings) - say(base, of->file, "Cannot find NAME section"); + say(of->file, "Cannot find NAME section"); + assert(NULL == of->desc); + of->desc = mandoc_strdup(of->name); + putkey(of, of->name, TYPE_Nd); fclose(stream); free(title); return; @@ -1128,7 +1118,7 @@ parse_catpage(struct of *of, const char *base) /* Skip to next word. */ ; } else { if (warnings) - say(base, of->file, "No dash in title line"); + say(of->file, "No dash in title line"); p = title; } @@ -1146,8 +1136,9 @@ parse_catpage(struct of *of, const char *base) plen -= 2; } - of->desc = stradd(p); - putkey(of, p, TYPE_Nd); + assert(NULL == of->desc); + of->desc = mandoc_strdup(p); + putkey(of, of->desc, TYPE_Nd); fclose(stream); free(title); } @@ -1160,20 +1151,10 @@ putkey(const struct of *of, const char *value, uint64_ { assert(NULL != value); - wordaddbuf(of, value, strlen(value), type); + putkeys(of, value, strlen(value), type); } /* - * Like putkey() but for unterminated strings. - */ -static void -putkeys(const struct of *of, const char *value, int sz, uint64_t type) -{ - - wordaddbuf(of, value, sz, type); -} - -/* * Grok all nodes at or below a certain mdoc node into putkey(). */ static void @@ -1188,7 +1169,7 @@ putmdockey(const struct of *of, const struct mdoc_node } } -static int +static void parse_man(struct of *of, const struct man_node *n) { const struct man_node *head, *body; @@ -1197,7 +1178,7 @@ parse_man(struct of *of, const struct man_node *n) size_t sz, titlesz; if (NULL == n) - return(0); + return; /* * We're only searching for one thing: the first text child in @@ -1239,7 +1220,7 @@ parse_man(struct of *of, const struct man_node *n) title[titlesz - 1] = ' '; } if (NULL == title) - return(1); + return; title = mandoc_realloc(title, titlesz + 1); title[titlesz] = '\0'; @@ -1252,7 +1233,7 @@ parse_man(struct of *of, const struct man_node *n) if (0 == (sz = strlen(sv))) { free(title); - return(1); + return; } /* Erase trailing space. */ @@ -1263,7 +1244,7 @@ parse_man(struct of *of, const struct man_node *n) if (start == sv) { free(title); - return(1); + return; } start = sv; @@ -1300,7 +1281,7 @@ parse_man(struct of *of, const struct man_node *n) if (sv == start) { putkey(of, start, TYPE_Nm); free(title); - return(1); + return; } while (isspace((unsigned char)*start)) @@ -1321,18 +1302,15 @@ parse_man(struct of *of, const struct man_node *n) start++; assert(NULL == of->desc); - of->desc = stradd(start); - putkey(of, start, TYPE_Nd); + of->desc = mandoc_strdup(start); + putkey(of, of->desc, TYPE_Nd); free(title); - return(1); + return; } } for (n = n->child; n; n = n->next) - if (parse_man(of, n)) - return(1); - - return(0); + parse_man(of, n); } static void @@ -1468,19 +1446,29 @@ parse_mdoc_St(struct of *of, const struct mdoc_node *n static int parse_mdoc_Xr(struct of *of, const struct mdoc_node *n) { + char *cp; if (NULL == (n = n->child)) return(0); - putkey(of, n->string, TYPE_Xr); - return(1); + if (NULL == n->next) { + putkey(of, 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); + free(cp); + return(0); } static int parse_mdoc_Nd(struct of *of, const struct mdoc_node *n) { size_t sz; - char *sv, *desc; if (MDOC_BODY != n->type) return(0); @@ -1490,25 +1478,20 @@ parse_mdoc_Nd(struct of *of, const struct mdoc_node *n * into the document table. */ - desc = NULL; for (n = n->child; NULL != n; n = n->next) { if (MDOC_TEXT == n->type) { - sz = strlen(n->string) + 1; - if (NULL != (sv = desc)) - sz += strlen(desc) + 1; - desc = mandoc_realloc(desc, sz); - if (NULL != sv) - strlcat(desc, " ", sz); - else - *desc = '\0'; - strlcat(desc, n->string, sz); + 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); } - - of->desc = NULL != desc ? stradd(desc) : NULL; - free(desc); return(1); } @@ -1546,76 +1529,23 @@ parse_mdoc_body(struct of *of, const struct mdoc_node } /* - * See straddbuf(). + * Add a string to the hash table for the current manual. + * Each string has a bitmask telling which macros it belongs to. + * When we finish the manual, we'll dump the table. */ -static char * -stradd(const char *cp) -{ - - return(straddbuf(cp, strlen(cp))); -} - -/* - * This looks up or adds a string to the string table. - * The string table is a table of all strings encountered during parse - * or file scan. - * In using it, we avoid having thousands of (e.g.) "cat1" string - * allocations for the "of" table. - * We also have a layer atop the string table for keeping track of words - * in a parse sequence (see wordaddbuf()). - */ -static char * -straddbuf(const char *cp, size_t sz) -{ - struct str *s; - unsigned int index; - const char *end; - - if (NULL != (s = hashget(cp, sz))) - return(s->key); - - s = mandoc_calloc(sizeof(struct str) + sz + 1, 1); - memcpy(s->key, cp, sz); - - end = cp + sz; - index = ohash_qlookupi(&strings, cp, &end); - assert(NULL == ohash_find(&strings, index)); - ohash_insert(&strings, index, s); - return(s->key); -} - -static struct str * -hashget(const char *cp, size_t sz) -{ - unsigned int index; - const char *end; - - end = cp + sz; - index = ohash_qlookupi(&strings, cp, &end); - return(ohash_find(&strings, index)); -} - -/* - * Add a word to the current parse sequence. - * Within the hashtable of strings, we maintain a list of strings that - * are currently indexed. - * Each of these ("words") has a bitmask modified within the parse. - * When we finish a parse, we'll dump the list, then remove the head - * entry -- since the next parse will have a new "of", it can keep track - * of its entries without conflict. - */ static void -wordaddbuf(const struct of *of, - const char *cp, size_t sz, uint64_t v) +putkeys(const struct of *of, const char *cp, size_t sz, uint64_t v) { struct str *s; - unsigned int index; + unsigned int slot; const char *end; if (0 == sz) return; - s = hashget(cp, sz); + end = cp + sz; + slot = ohash_qlookupi(&strings, cp, &end); + s = ohash_find(&strings, slot); if (NULL != s && of == s->of) { s->mask |= v; @@ -1623,16 +1553,10 @@ wordaddbuf(const struct of *of, } else if (NULL == s) { s = mandoc_calloc(sizeof(struct str) + sz + 1, 1); memcpy(s->key, cp, sz); - end = cp + sz; - index = ohash_qlookupi(&strings, cp, &end); - assert(NULL == ohash_find(&strings, index)); - ohash_insert(&strings, index, s); + ohash_insert(&strings, slot, s); } - - s->next = words; s->of = of; s->mask = v; - words = s; } /* @@ -1798,23 +1722,24 @@ utf8key(struct mchars *mc, struct str *key) * Also, UTF-8-encode the description at the last possible moment. */ static void -dbindex(struct mchars *mc, int form, - const struct of *of, const char *base) +dbindex(struct mchars *mc, int form, const struct of *of) { struct str *key; const char *desc; int64_t recno; size_t i; + unsigned int slot; if (verb) - say(base, of->file, "Adding to index"); + say(of->file, "Adding to index"); if (nodb) return; desc = ""; if (NULL != of->desc) { - key = hashget(of->desc, strlen(of->desc)); + key = ohash_find(&strings, + ohash_qlookup(&strings, of->desc)); assert(NULL != key); if (NULL == key->utf8) utf8key(mc, key); @@ -1833,7 +1758,8 @@ dbindex(struct mchars *mc, int form, recno = sqlite3_last_insert_rowid(db); sqlite3_reset(stmts[STMT_INSERT_DOC]); - for (key = words; NULL != key; key = key->next) { + for (key = ohash_first(&strings, &slot); NULL != key; + key = ohash_next(&strings, &slot)) { assert(key->of == of); if (NULL == key->utf8) utf8key(mc, key); @@ -1843,13 +1769,16 @@ dbindex(struct mchars *mc, int form, SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, recno); SQL_STEP(stmts[STMT_INSERT_KEY]); sqlite3_reset(stmts[STMT_INSERT_KEY]); + if (key->utf8 != key->key) + free(key->utf8); + free(key); } SQL_EXEC("END TRANSACTION"); } static void -dbprune(const char *base) +dbprune(void) { struct of *of; size_t i; @@ -1863,7 +1792,7 @@ dbprune(const char *base) SQL_STEP(stmts[STMT_DELETE]); sqlite3_reset(stmts[STMT_DELETE]); if (verb) - say(base, of->file, "Deleted from index"); + say(of->file, "Deleted from index"); } } @@ -1872,10 +1801,9 @@ dbprune(const char *base) * If "real" is not set, rename the temporary file into the real one. */ static void -dbclose(const char *base, int real) +dbclose(int real) { size_t i; - char file[MAXPATHLEN]; if (nodb) return; @@ -1891,10 +1819,10 @@ dbclose(const char *base, int real) if (real) return; - strlcpy(file, MANDOC_DB, MAXPATHLEN); - strlcat(file, "~", MAXPATHLEN); - if (-1 == rename(file, MANDOC_DB)) - perror(MANDOC_DB); + if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say(MANDOC_DB, NULL); + } } /* @@ -1906,36 +1834,32 @@ dbclose(const char *base, int real) * Must be matched by dbclose(). */ static int -dbopen(const char *base, int real) +dbopen(int real) { - char file[MAXPATHLEN]; - const char *sql; + const char *file, *sql; int rc, ofl; - size_t sz; if (nodb) return(1); - sz = strlcpy(file, MANDOC_DB, MAXPATHLEN); - if ( ! real) - sz = strlcat(file, "~", MAXPATHLEN); + ofl = SQLITE_OPEN_READWRITE; + if (0 == real) { + file = MANDOC_DB "~"; + if (-1 == remove(file) && ENOENT != errno) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say(file, NULL); + return(0); + } + ofl |= SQLITE_OPEN_EXCLUSIVE; + } else + file = MANDOC_DB; - if (sz >= MAXPATHLEN) { - fprintf(stderr, "%s: Path too long\n", file); - return(0); - } - - if ( ! real) - remove(file); - - ofl = SQLITE_OPEN_READWRITE | - (0 == real ? SQLITE_OPEN_EXCLUSIVE : 0); - rc = sqlite3_open_v2(file, &db, ofl, NULL); if (SQLITE_OK == rc) - return(1); + goto prepare_statements; if (SQLITE_CANTOPEN != rc) { - perror(file); + exitcode = (int)MANDOCLEVEL_SYSERR; + say(file, NULL); return(0); } @@ -1943,7 +1867,8 @@ dbopen(const char *base, int real) db = NULL; if (SQLITE_OK != (rc = sqlite3_open(file, &db))) { - perror(file); + exitcode = (int)MANDOCLEVEL_SYSERR; + say(file, NULL); return(0); } @@ -1967,10 +1892,13 @@ dbopen(const char *base, int real) "CREATE INDEX \"key_index\" ON keys (key);\n"; if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) { - perror(sqlite3_errmsg(db)); + exitcode = (int)MANDOCLEVEL_SYSERR; + say(file, "%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 " @@ -2004,28 +1932,76 @@ hash_free(void *p, size_t sz, void *arg) } static int -path_reset(const char *cwd, int fd, const char *base) +set_basedir(const char *targetdir) { + static char startdir[PATH_MAX]; + static int fd; - if (-1 == fchdir(fd)) { - perror(cwd); + /* + * 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. + */ + if ('\0' == *startdir) { + if (NULL == getcwd(startdir, PATH_MAX)) { + exitcode = (int)MANDOCLEVEL_SYSERR; + if (NULL != targetdir) + say(".", NULL); + return(0); + } + if (-1 == (fd = open(startdir, O_RDONLY, 0))) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say(startdir, NULL); + 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(startdir, NULL); + return(0); + } + if (NULL == targetdir) { + close(fd); + return(1); + } + } + if (NULL == realpath(targetdir, basedir)) { + basedir[0] = '\0'; + exitcode = (int)MANDOCLEVEL_BADARG; + say(targetdir, NULL); return(0); - } else if (-1 == chdir(base)) { - perror(base); + } else if (-1 == chdir(basedir)) { + exitcode = (int)MANDOCLEVEL_BADARG; + say("", NULL); return(0); } return(1); } static void -say(const char *dir, const char *file, const char *format, ...) +say(const char *file, const char *format, ...) { va_list ap; - fprintf(stderr, "%s", dir); + if ('\0' != *basedir) + fprintf(stderr, "%s", basedir); + if ('\0' != *basedir && '\0' != *file) + fputs("//", stderr); if ('\0' != *file) - fprintf(stderr, "//%s", file); + fprintf(stderr, "%s", file); fputs(": ", stderr); + + if (NULL == format) { + perror(NULL); + return; + } va_start(ap, format); vfprintf(stderr, format, ap);