=================================================================== RCS file: /cvs/mandoc/mandocdb.c,v retrieving revision 1.61 retrieving revision 1.90 diff -u -p -r1.61 -r1.90 --- mandoc/mandocdb.c 2013/06/05 21:21:08 1.61 +++ mandoc/mandocdb.c 2013/12/27 23:41:55 1.90 @@ -1,4 +1,4 @@ -/* $Id: mandocdb.c,v 1.61 2013/06/05 21:21:08 schwarze Exp $ */ +/* $Id: mandocdb.c,v 1.90 2013/12/27 23:41:55 schwarze Exp $ */ /* * Copyright (c) 2011, 2012 Kristaps Dzonsons * Copyright (c) 2011, 2012, 2013 Ingo Schwarze @@ -23,6 +23,7 @@ #include #include +#include #include #include #include @@ -75,98 +76,101 @@ 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 */ - struct str *next; /* next in owning parse sequence */ + const struct mpage *mpage; /* if set, the owning parse */ uint64_t mask; /* bitmask in sequence */ char key[]; /* the string itself */ }; -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 */ +struct mpage { + struct inodev inodev; /* used for hashing routine */ + 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 mlink { char file[PATH_MAX]; /* 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) */ + 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 title { + char *title; /* name(sec/arch) given inside the file */ + char *file; /* file name in case of mismatch */ +}; + 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_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 */ uint64_t mask; /* set unless handler returns 0 */ - int flags; /* for use by pmdoc_node */ -#define MDOCF_CHILD 0x01 /* automatically index child nodes */ }; static void dbclose(int); -static void dbindex(struct mchars *, int, const struct of *); +static void dbindex(const struct mpage *, struct mchars *); static int dbopen(int); static void dbprune(void); -static void fileadd(struct of *); -static int filecheck(const char *); static void filescan(const char *); -static struct str *hashget(const char *, size_t); 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 *); -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 *, +static void mlink_add(struct mlink *, const struct stat *); +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 *, int); +static void parse_cat(struct mpage *); +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_In(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_St(struct mpage *, const struct mdoc_node *); +static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *); +static void putkey(const struct mpage *, const char *, uint64_t); -static void putkeys(const struct of *, - const char *, int, uint64_t); -static void putmdockey(const struct of *, +static void putkeys(const struct mpage *, + const char *, size_t, uint64_t); +static void putmdockey(const struct mpage *, const struct mdoc_node *, uint64_t); static void say(const char *, const char *, ...); -static char *stradd(const char *); -static char *straddbuf(const char *, size_t); +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 void wordaddbuf(const struct of *, - const char *, size_t, uint64_t); static char *progname; static int use_all; /* use all found files */ @@ -176,165 +180,157 @@ 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 mpages; /* table of distinct manual pages */ +static struct ohash mlinks; /* table of directory entries */ 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 */ static const struct mdoc_handler mdocs[MDOC_MAX] = { - { NULL, 0, 0 }, /* Ap */ - { NULL, 0, 0 }, /* Dd */ - { NULL, 0, 0 }, /* Dt */ - { NULL, 0, 0 }, /* Os */ - { parse_mdoc_Sh, TYPE_Sh, MDOCF_CHILD }, /* Sh */ - { parse_mdoc_head, TYPE_Ss, MDOCF_CHILD }, /* Ss */ - { NULL, 0, 0 }, /* Pp */ - { NULL, 0, 0 }, /* D1 */ - { NULL, 0, 0 }, /* Dl */ - { NULL, 0, 0 }, /* Bd */ - { NULL, 0, 0 }, /* Ed */ - { NULL, 0, 0 }, /* Bl */ - { NULL, 0, 0 }, /* El */ - { NULL, 0, 0 }, /* It */ - { NULL, 0, 0 }, /* Ad */ - { NULL, TYPE_An, MDOCF_CHILD }, /* An */ - { NULL, TYPE_Ar, MDOCF_CHILD }, /* Ar */ - { NULL, TYPE_Cd, MDOCF_CHILD }, /* Cd */ - { NULL, TYPE_Cm, MDOCF_CHILD }, /* Cm */ - { NULL, TYPE_Dv, MDOCF_CHILD }, /* Dv */ - { NULL, TYPE_Er, MDOCF_CHILD }, /* Er */ - { NULL, TYPE_Ev, MDOCF_CHILD }, /* Ev */ - { NULL, 0, 0 }, /* Ex */ - { NULL, TYPE_Fa, MDOCF_CHILD }, /* Fa */ - { parse_mdoc_Fd, TYPE_In, 0 }, /* Fd */ - { NULL, TYPE_Fl, MDOCF_CHILD }, /* Fl */ - { parse_mdoc_Fn, 0, 0 }, /* Fn */ - { NULL, TYPE_Ft, MDOCF_CHILD }, /* Ft */ - { NULL, TYPE_Ic, MDOCF_CHILD }, /* Ic */ - { parse_mdoc_In, TYPE_In, MDOCF_CHILD }, /* In */ - { NULL, TYPE_Li, MDOCF_CHILD }, /* Li */ - { parse_mdoc_Nd, TYPE_Nd, MDOCF_CHILD }, /* Nd */ - { parse_mdoc_Nm, TYPE_Nm, MDOCF_CHILD }, /* Nm */ - { NULL, 0, 0 }, /* Op */ - { NULL, 0, 0 }, /* Ot */ - { NULL, TYPE_Pa, MDOCF_CHILD }, /* Pa */ - { NULL, 0, 0 }, /* Rv */ - { parse_mdoc_St, TYPE_St, 0 }, /* St */ - { NULL, TYPE_Va, MDOCF_CHILD }, /* Va */ - { parse_mdoc_body, TYPE_Va, MDOCF_CHILD }, /* Vt */ - { parse_mdoc_Xr, TYPE_Xr, 0 }, /* Xr */ - { NULL, 0, 0 }, /* %A */ - { NULL, 0, 0 }, /* %B */ - { NULL, 0, 0 }, /* %D */ - { NULL, 0, 0 }, /* %I */ - { NULL, 0, 0 }, /* %J */ - { NULL, 0, 0 }, /* %N */ - { NULL, 0, 0 }, /* %O */ - { NULL, 0, 0 }, /* %P */ - { NULL, 0, 0 }, /* %R */ - { NULL, 0, 0 }, /* %T */ - { NULL, 0, 0 }, /* %V */ - { NULL, 0, 0 }, /* Ac */ - { NULL, 0, 0 }, /* Ao */ - { NULL, 0, 0 }, /* Aq */ - { NULL, TYPE_At, MDOCF_CHILD }, /* At */ - { NULL, 0, 0 }, /* Bc */ - { NULL, 0, 0 }, /* Bf */ - { NULL, 0, 0 }, /* Bo */ - { NULL, 0, 0 }, /* Bq */ - { NULL, TYPE_Bsx, MDOCF_CHILD }, /* Bsx */ - { NULL, TYPE_Bx, MDOCF_CHILD }, /* Bx */ - { NULL, 0, 0 }, /* Db */ - { NULL, 0, 0 }, /* Dc */ - { NULL, 0, 0 }, /* Do */ - { NULL, 0, 0 }, /* Dq */ - { NULL, 0, 0 }, /* Ec */ - { NULL, 0, 0 }, /* Ef */ - { NULL, TYPE_Em, MDOCF_CHILD }, /* Em */ - { NULL, 0, 0 }, /* Eo */ - { NULL, TYPE_Fx, MDOCF_CHILD }, /* Fx */ - { NULL, TYPE_Ms, MDOCF_CHILD }, /* Ms */ - { NULL, 0, 0 }, /* No */ - { NULL, 0, 0 }, /* Ns */ - { NULL, TYPE_Nx, MDOCF_CHILD }, /* Nx */ - { NULL, TYPE_Ox, MDOCF_CHILD }, /* Ox */ - { NULL, 0, 0 }, /* Pc */ - { NULL, 0, 0 }, /* Pf */ - { NULL, 0, 0 }, /* Po */ - { NULL, 0, 0 }, /* Pq */ - { NULL, 0, 0 }, /* Qc */ - { NULL, 0, 0 }, /* Ql */ - { NULL, 0, 0 }, /* Qo */ - { NULL, 0, 0 }, /* Qq */ - { NULL, 0, 0 }, /* Re */ - { NULL, 0, 0 }, /* Rs */ - { NULL, 0, 0 }, /* Sc */ - { NULL, 0, 0 }, /* So */ - { NULL, 0, 0 }, /* Sq */ - { NULL, 0, 0 }, /* Sm */ - { NULL, 0, 0 }, /* Sx */ - { NULL, TYPE_Sy, MDOCF_CHILD }, /* Sy */ - { NULL, TYPE_Tn, MDOCF_CHILD }, /* Tn */ - { NULL, 0, 0 }, /* Ux */ - { NULL, 0, 0 }, /* Xc */ - { NULL, 0, 0 }, /* Xo */ - { parse_mdoc_head, TYPE_Fn, 0 }, /* Fo */ - { NULL, 0, 0 }, /* Fc */ - { NULL, 0, 0 }, /* Oo */ - { NULL, 0, 0 }, /* Oc */ - { NULL, 0, 0 }, /* Bk */ - { NULL, 0, 0 }, /* Ek */ - { NULL, 0, 0 }, /* Bt */ - { NULL, 0, 0 }, /* Hf */ - { NULL, 0, 0 }, /* Fr */ - { NULL, 0, 0 }, /* Ud */ - { NULL, TYPE_Lb, MDOCF_CHILD }, /* Lb */ - { NULL, 0, 0 }, /* Lp */ - { NULL, TYPE_Lk, MDOCF_CHILD }, /* Lk */ - { NULL, TYPE_Mt, MDOCF_CHILD }, /* Mt */ - { NULL, 0, 0 }, /* Brq */ - { NULL, 0, 0 }, /* Bro */ - { NULL, 0, 0 }, /* Brc */ - { NULL, 0, 0 }, /* %C */ - { NULL, 0, 0 }, /* Es */ - { NULL, 0, 0 }, /* En */ - { NULL, TYPE_Dx, MDOCF_CHILD }, /* Dx */ - { NULL, 0, 0 }, /* %Q */ - { NULL, 0, 0 }, /* br */ - { NULL, 0, 0 }, /* sp */ - { NULL, 0, 0 }, /* %U */ - { NULL, 0, 0 }, /* Ta */ + { NULL, 0 }, /* Ap */ + { NULL, 0 }, /* Dd */ + { NULL, 0 }, /* Dt */ + { NULL, 0 }, /* Os */ + { parse_mdoc_Sh, TYPE_Sh }, /* Sh */ + { parse_mdoc_head, TYPE_Ss }, /* Ss */ + { NULL, 0 }, /* Pp */ + { NULL, 0 }, /* D1 */ + { NULL, 0 }, /* Dl */ + { NULL, 0 }, /* Bd */ + { NULL, 0 }, /* Ed */ + { NULL, 0 }, /* Bl */ + { NULL, 0 }, /* El */ + { NULL, 0 }, /* It */ + { NULL, 0 }, /* Ad */ + { NULL, TYPE_An }, /* An */ + { NULL, TYPE_Ar }, /* Ar */ + { NULL, TYPE_Cd }, /* Cd */ + { NULL, TYPE_Cm }, /* Cm */ + { NULL, TYPE_Dv }, /* Dv */ + { NULL, TYPE_Er }, /* Er */ + { NULL, TYPE_Ev }, /* Ev */ + { NULL, 0 }, /* Ex */ + { NULL, TYPE_Fa }, /* Fa */ + { parse_mdoc_Fd, 0 }, /* Fd */ + { NULL, TYPE_Fl }, /* Fl */ + { parse_mdoc_Fn, 0 }, /* Fn */ + { NULL, TYPE_Ft }, /* Ft */ + { NULL, TYPE_Ic }, /* Ic */ + { parse_mdoc_In, TYPE_In }, /* In */ + { NULL, TYPE_Li }, /* Li */ + { parse_mdoc_Nd, TYPE_Nd }, /* Nd */ + { parse_mdoc_Nm, TYPE_Nm }, /* Nm */ + { NULL, 0 }, /* Op */ + { NULL, 0 }, /* Ot */ + { NULL, TYPE_Pa }, /* Pa */ + { NULL, 0 }, /* Rv */ + { parse_mdoc_St, 0 }, /* St */ + { NULL, TYPE_Va }, /* Va */ + { parse_mdoc_body, TYPE_Va }, /* Vt */ + { parse_mdoc_Xr, 0 }, /* Xr */ + { NULL, 0 }, /* %A */ + { NULL, 0 }, /* %B */ + { NULL, 0 }, /* %D */ + { NULL, 0 }, /* %I */ + { NULL, 0 }, /* %J */ + { NULL, 0 }, /* %N */ + { NULL, 0 }, /* %O */ + { NULL, 0 }, /* %P */ + { NULL, 0 }, /* %R */ + { NULL, 0 }, /* %T */ + { NULL, 0 }, /* %V */ + { NULL, 0 }, /* Ac */ + { NULL, 0 }, /* Ao */ + { NULL, 0 }, /* Aq */ + { NULL, TYPE_At }, /* At */ + { NULL, 0 }, /* Bc */ + { NULL, 0 }, /* Bf */ + { NULL, 0 }, /* Bo */ + { NULL, 0 }, /* Bq */ + { NULL, TYPE_Bsx }, /* Bsx */ + { NULL, TYPE_Bx }, /* Bx */ + { NULL, 0 }, /* Db */ + { NULL, 0 }, /* Dc */ + { NULL, 0 }, /* Do */ + { NULL, 0 }, /* Dq */ + { NULL, 0 }, /* Ec */ + { NULL, 0 }, /* Ef */ + { NULL, TYPE_Em }, /* Em */ + { NULL, 0 }, /* Eo */ + { NULL, TYPE_Fx }, /* Fx */ + { NULL, TYPE_Ms }, /* Ms */ + { NULL, 0 }, /* No */ + { NULL, 0 }, /* Ns */ + { NULL, TYPE_Nx }, /* Nx */ + { NULL, TYPE_Ox }, /* Ox */ + { NULL, 0 }, /* Pc */ + { NULL, 0 }, /* Pf */ + { NULL, 0 }, /* Po */ + { NULL, 0 }, /* Pq */ + { NULL, 0 }, /* Qc */ + { NULL, 0 }, /* Ql */ + { NULL, 0 }, /* Qo */ + { NULL, 0 }, /* Qq */ + { NULL, 0 }, /* Re */ + { NULL, 0 }, /* Rs */ + { NULL, 0 }, /* Sc */ + { NULL, 0 }, /* So */ + { NULL, 0 }, /* Sq */ + { NULL, 0 }, /* Sm */ + { NULL, 0 }, /* Sx */ + { NULL, TYPE_Sy }, /* Sy */ + { NULL, TYPE_Tn }, /* Tn */ + { NULL, 0 }, /* Ux */ + { NULL, 0 }, /* Xc */ + { NULL, 0 }, /* Xo */ + { parse_mdoc_head, 0 }, /* Fo */ + { NULL, 0 }, /* Fc */ + { NULL, 0 }, /* Oo */ + { NULL, 0 }, /* Oc */ + { NULL, 0 }, /* Bk */ + { NULL, 0 }, /* Ek */ + { NULL, 0 }, /* Bt */ + { NULL, 0 }, /* Hf */ + { NULL, 0 }, /* Fr */ + { NULL, 0 }, /* Ud */ + { NULL, TYPE_Lb }, /* Lb */ + { NULL, 0 }, /* Lp */ + { NULL, TYPE_Lk }, /* Lk */ + { NULL, TYPE_Mt }, /* Mt */ + { NULL, 0 }, /* Brq */ + { NULL, 0 }, /* Bro */ + { NULL, 0 }, /* Brc */ + { NULL, 0 }, /* %C */ + { NULL, 0 }, /* Es */ + { NULL, 0 }, /* En */ + { NULL, TYPE_Dx }, /* Dx */ + { NULL, 0 }, /* %Q */ + { NULL, 0 }, /* br */ + { NULL, 0 }, /* sp */ + { NULL, 0 }, /* %U */ + { NULL, 0 }, /* Ta */ }; int main(int argc, char *argv[]) { int ch, i; - unsigned int index; size_t j, sz; const char *path_arg; - struct str *s; struct mchars *mc; struct manpaths dirs; struct mparse *mp; - struct ohash_info ino_info, filename_info, str_info; + struct ohash_info mpages_info, mlinks_info; memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *)); memset(&dirs, 0, sizeof(struct manpaths)); - ino_info.halloc = filename_info.halloc = - str_info.halloc = hash_halloc; - ino_info.hfree = filename_info.hfree = - str_info.hfree = hash_free; - ino_info.alloc = filename_info.alloc = - str_info.alloc = hash_alloc; + 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); - str_info.key_offset = offsetof(struct str, key); + mpages_info.key_offset = offsetof(struct mpage, inodev); + mlinks_info.key_offset = offsetof(struct mlink, file); progname = strrchr(argv[0], '/'); if (progname == NULL) @@ -408,9 +404,8 @@ main(int argc, char *argv[]) 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); + ohash_init(&mpages, 6, &mpages_info); + ohash_init(&mlinks, 6, &mlinks_info); if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) { /* @@ -432,7 +427,7 @@ main(int argc, char *argv[]) if (OP_TEST != op) dbprune(); if (OP_DELETE != op) - ofmerge(mc, mp); + mpages_merge(mc, mp, 0); dbclose(1); } else { /* @@ -450,9 +445,8 @@ main(int argc, char *argv[]) manpath_parse(&dirs, path_arg, NULL, NULL); /* - * First scan the tree rooted at a base directory. - * Then whak its database (if one exists), parse, and - * build up the database. + * First scan the tree rooted at a base directory, then + * build a new database and finally move it into place. * Ignore zero-length directories and strip trailing * slashes. */ @@ -462,6 +456,12 @@ main(int argc, char *argv[]) dirs.paths[j][--sz] = '\0'; if (0 == sz) continue; + + if (j) { + ohash_init(&mpages, 6, &mpages_info); + ohash_init(&mlinks, 6, &mlinks_info); + } + if (0 == set_basedir(dirs.paths[j])) goto out; if (0 == treescan()) @@ -471,22 +471,14 @@ main(int argc, char *argv[]) if (0 == dbopen(0)) goto out; - /* - * Since we're opening up a new database, we can - * turn off synchronous mode for much better - * performance. - */ -#ifndef __APPLE__ - SQL_EXEC("PRAGMA synchronous = OFF"); -#endif - - ofmerge(mc, mp); + mpages_merge(mc, mp, warnings && !use_all); dbclose(0); - offree(); - ohash_delete(&inos); - ohash_init(&inos, 6, &ino_info); - ohash_delete(&filenames); - ohash_init(&filenames, 6, &filename_info); + + if (j + 1 < dirs.sz) { + mpages_free(); + ohash_delete(&mpages); + ohash_delete(&mlinks); + } } } out: @@ -494,16 +486,9 @@ out: 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(); + mpages_free(); + ohash_delete(&mpages); + ohash_delete(&mlinks); return(exitcode); usage: fprintf(stderr, "usage: %s [-anvW] [-C file]\n" @@ -536,9 +521,10 @@ treescan(void) { FTS *f; FTSENT *ff; + struct mlink *mlink; int dform; - char *sec; - const char *dsec, *arch, *cp, *name, *path; + char *fsec; + const char *dsec, *arch, *cp, *path; const char *argv[2]; argv[0] = "."; @@ -561,9 +547,8 @@ treescan(void) while (NULL != (ff = fts_read(f))) { path = ff->fts_path + 2; /* - * If we're a regular file, add an "of" by using the + * 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) { if (0 == strcmp(path, MANDOC_DB)) @@ -572,11 +557,7 @@ treescan(void) if (warnings) say(path, "Extraneous file"); continue; - } else if (inocheck(ff->fts_statp)) { - if (warnings) - say(path, "Duplicate file"); - continue; - } else if (NULL == (sec = + } else if (NULL == (fsec = strrchr(ff->fts_name, '.'))) { if ( ! use_all) { if (warnings) @@ -584,37 +565,43 @@ treescan(void) "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, "gz")) { if (warnings) say(path, "Skip gz"); continue; - } else if (0 == strcmp(sec, "ps")) { + } else if (0 == strcmp(fsec, "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'; - sec = stradd(sec); - } - name = stradd(ff->fts_name); - ofadd(dform, path, - name, dsec, sec, arch, ff->fts_statp); + } else + fsec[-1] = '\0'; + mlink = mandoc_calloc(1, sizeof(struct mlink)); + strlcpy(mlink->file, path, sizeof(mlink->file)); + mlink->dform = dform; + if (NULL != dsec) + mlink->dsec = mandoc_strdup(dsec); + if (NULL != arch) + mlink->arch = mandoc_strdup(arch); + mlink->name = mandoc_strdup(ff->fts_name); + if (NULL != fsec) + mlink->fsec = mandoc_strdup(fsec); + mlink_add(mlink, ff->fts_statp); continue; - } else if (FTS_D != ff->fts_info && + } else if (FTS_D != ff->fts_info && FTS_DP != ff->fts_info) { if (warnings) say(path, "Not a regular file"); @@ -639,10 +626,10 @@ treescan(void) 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) @@ -659,7 +646,7 @@ treescan(void) */ 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) @@ -676,7 +663,7 @@ 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). * @@ -687,17 +674,15 @@ treescan(void) * 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); @@ -708,7 +693,7 @@ filescan(const char *file) exitcode = (int)MANDOCLEVEL_BADARG; say(file, NULL); return; - } else if (strstr(buf, basedir) != buf) { + } else if (OP_TEST != op && strstr(buf, basedir) != buf) { exitcode = (int)MANDOCLEVEL_BADARG; say("", "%s: outside base directory", buf); return; @@ -720,14 +705,10 @@ filescan(const char *file) exitcode = (int)MANDOCLEVEL_BADARG; say(file, "Not a regular file"); return; - } else if (inocheck(&st)) { - if (warnings) - say(file, "Duplicate file"); - return; } start = buf + strlen(basedir); - sec = arch = name = dsec = NULL; - dform = FORM_NONE; + mlink = mandoc_calloc(1, sizeof(struct mlink)); + strlcpy(mlink->file, start, sizeof(mlink->file)); /* * First try to guess our directory structure. @@ -738,19 +719,19 @@ 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 = mandoc_strdup(start + 3); } else if (0 == strncmp(start, "cat", 3)) { - dform = FORM_CAT; - dsec = start + 3; + mlink->dform = FORM_CAT; + mlink->dsec = mandoc_strdup(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 = mandoc_strdup(start); start = p; - } + } } /* @@ -763,232 +744,237 @@ filescan(const char *file) if ('.' == *p) { *p++ = '\0'; - sec = p; + mlink->fsec = mandoc_strdup(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'; - } + } + mlink->name = mandoc_strdup(mlink->name); - ofadd(dform, file, name, dsec, sec, arch, &st); + mlink_add(mlink, &st); } -/* - * See fileadd(). - */ -static int -filecheck(const char *name) -{ - unsigned int index; - - index = ohash_qlookup(&filenames, name); - return(NULL != ohash_find(&filenames, index)); -} - -/* - * Use the standard hashing mechanism (K&R) to see if the given filename - * already exists. - */ static void -fileadd(struct of *of) +mlink_add(struct mlink *mlink, const struct stat *st) { - unsigned int index; + struct inodev inodev; + struct mpage *mpage; + unsigned int slot; - index = ohash_qlookup(&filenames, of->file); - assert(NULL == ohash_find(&filenames, index)); - ohash_insert(&filenames, index, of); -} + assert(NULL != mlink->file); -/* - * See inoadd(). - */ -static int -inocheck(const struct stat *st) -{ - struct id id; - uint32_t hash; - unsigned int index; + if (NULL == mlink->dsec) + mlink->dsec = mandoc_strdup(""); + if (NULL == mlink->arch) + mlink->arch = mandoc_strdup(""); + if (NULL == mlink->name) + mlink->name = mandoc_strdup(""); + if (NULL == mlink->fsec) + mlink->fsec = mandoc_strdup(""); - 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); + 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; - return(NULL != ohash_find(&inos, index)); + 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; } -/* - * 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. - */ static void -inoadd(const struct stat *st, struct of *of) +mlink_free(struct mlink *mlink) { - uint32_t hash; - unsigned int index; - of->id.ino = hash = st->st_ino; - of->id.dev = st->st_dev; - index = ohash_lookup_memory - (&inos, (char *)&of->id, sizeof(of->id), hash); - - assert(NULL == ohash_find(&inos, index)); - ohash_insert(&inos, index, of); + free(mlink->dsec); + free(mlink->arch); + free(mlink->name); + free(mlink->fsec); + free(mlink); } static void -ofadd(int dform, const char *file, const char *name, const char *dsec, - const char *sec, const char *arch, const struct stat *st) +mpages_free(void) { - struct of *of; - int sform; + struct mpage *mpage; + struct mlink *mlink; + unsigned int slot; - assert(NULL != file); - - if (NULL == name) - name = ""; - if (NULL == sec) - sec = ""; - if (NULL == dsec) - dsec = ""; - if (NULL == arch) - arch = ""; - - sform = FORM_NONE; - if (NULL != sec && *sec <= '9' && *sec >= '1') - sform = FORM_SRC; - else if (NULL != sec && *sec == '0') { - sec = dsec; - sform = FORM_CAT; + 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); } - - of = mandoc_calloc(1, sizeof(struct of)); - strlcpy(of->file, file, PATH_MAX); - of->name = name; - of->sec = sec; - of->dsec = dsec; - of->arch = arch; - of->sform = sform; - of->dform = dform; - of->next = ofs; - ofs = of; - - /* - * 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. - */ - inoadd(st, of); - fileadd(of); } +/* + * 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 -offree(void) +mlinks_undupe(struct mpage *mpage) { - struct of *of; + char buf[PATH_MAX]; + struct mlink **prev; + struct mlink *mlink; + char *bufp; - while (NULL != (of = ofs)) { - ofs = of->next; - free(of); + mpage->form = FORM_CAT; + prev = &mpage->mlinks; + while (NULL != (mlink = *prev)) { + if (FORM_CAT != mlink->dform) { + mpage->form = FORM_NONE; + goto nextlink; + } + if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) { + if (warnings) + say(mlink->file, "Filename too long"); + goto nextlink; + } + bufp = strstr(buf, "cat"); + assert(NULL != bufp); + memcpy(bufp, "man", 3); + if (NULL != (bufp = strrchr(buf, '.'))) + *++bufp = '\0'; + strlcat(buf, mlink->dsec, PATH_MAX); + 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; } } /* - * 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) +mpages_merge(struct mchars *mc, struct mparse *mp, int check_reachable) { - int form; - size_t sz; - struct mdoc *mdoc; - struct man *man; - char buf[PATH_MAX]; - char *bufp; - const char *msec, *march, *mtitle, *cp; - struct of *of; - enum mandoclevel lvl; + struct ohash title_table; + struct ohash_info title_info, str_info; + struct mpage *mpage; + struct mdoc *mdoc; + struct man *man; + struct title *title_entry; + char *title_str; + const char *cp; + int match; + unsigned int pslot, tslot; + enum mandoclevel lvl; - 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; - } + str_info.alloc = hash_alloc; + str_info.halloc = hash_halloc; + 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); + } + + mpage = ohash_first(&mpages, &pslot); + while (NULL != mpage) { + mlinks_undupe(mpage); + if (NULL == mpage->mlinks) { + mpage = ohash_next(&mpages, &pslot); + continue; } - words = NULL; + 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; /* * 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, -1, mpage->mlinks->file); if (lvl < MANDOCLEVEL_FATAL) mparse_result(mp, &mdoc, &man); - } + } if (NULL != mdoc) { - form = 1; - msec = mdoc_meta(mdoc)->msec; - march = mdoc_meta(mdoc)->arch; - mtitle = mdoc_meta(mdoc)->title; + 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) { - form = 1; - msec = man_meta(man)->msec; - march = ""; - mtitle = man_meta(man)->title; - } + 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); + } - if (NULL == msec) - msec = ""; - if (NULL == march) - march = ""; - if (NULL == mtitle) - mtitle = ""; - /* * Check whether the manual section given in a file * agrees with the directory where the file is located. @@ -998,11 +984,13 @@ ofmerge(struct mchars *mc, struct mparse *mp) * section, like encrypt(1) = makekey(8). Do not skip * manuals for such reasons. */ - if (warnings && !use_all && form && - strcasecmp(msec, of->dsec)) - say(of->file, "Section \"%s\" " - "manual in %s directory", - msec, of->dsec); + if (warnings && !use_all && FORM_SRC == mpage->form && + strcasecmp(mpage->sec, mpage->mlinks->dsec)) { + match = 0; + say(mpage->mlinks->file, "Section \"%s\" " + "manual in %s directory", + mpage->sec, mpage->mlinks->dsec); + } /* * Manual page directories exist for each kernel @@ -1018,36 +1006,93 @@ ofmerge(struct mchars *mc, struct mparse *mp) * Thus, warn about architecture mismatches, * but don't skip manuals for this reason. */ - if (warnings && !use_all && strcasecmp(march, of->arch)) - say(of->file, "Architecture \"%s\" " + if (warnings && !use_all && + strcasecmp(mpage->arch, mpage->mlinks->arch)) { + match = 0; + say(mpage->mlinks->file, "Architecture \"%s\" " "manual in \"%s\" directory", - march, of->arch); + mpage->arch, mpage->mlinks->arch); + } + if (warnings && !use_all && + strcasecmp(mpage->title, mpage->mlinks->name)) + match = 0; - putkey(of, of->name, TYPE_Nm); + putkey(mpage, mpage->mlinks->name, TYPE_Nm); if (NULL != mdoc) { if (NULL != (cp = mdoc_meta(mdoc)->name)) - putkey(of, cp, TYPE_Nm); - parse_mdoc(of, mdoc_node(mdoc)); + putkey(mpage, cp, TYPE_Nm); + assert(NULL == mpage->desc); + parse_mdoc(mpage, mdoc_node(mdoc)); + putkey(mpage, NULL != mpage->desc ? + mpage->desc : mpage->mlinks->name, TYPE_Nd); } else if (NULL != man) - parse_man(of, man_node(man)); + parse_man(mpage, man_node(man)); else - parse_catpage(of); + parse_cat(mpage); - dbindex(mc, form, of); + /* + * 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 (check_reachable) { + if (-1 == asprintf(&title_str, "%s(%s%s%s)", + mpage->title, mpage->sec, + '\0' == *mpage->arch ? "" : "/", + mpage->arch)) { + perror(NULL); + exit((int)MANDOCLEVEL_SYSERR); + } + tslot = ohash_qlookup(&title_table, title_str); + title_entry = ohash_find(&title_table, tslot); + if (NULL == title_entry) { + title_entry = mandoc_malloc( + sizeof(struct title)); + title_entry->title = title_str; + title_entry->file = mandoc_strdup( + match ? "" : mpage->mlinks->file); + ohash_insert(&title_table, tslot, + title_entry); + } else { + if (match) + *title_entry->file = '\0'; + free(title_str); + } + } + + dbindex(mpage, mc); + ohash_delete(&strings); + mpage = ohash_next(&mpages, &pslot); } + + if (check_reachable) { + title_entry = ohash_first(&title_table, &tslot); + 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, &tslot); + } + ohash_delete(&title_table); + } } static void -parse_catpage(struct of *of) +parse_cat(struct mpage *mpage) { FILE *stream; char *line, *p, *title; size_t len, plen, titlesz; - if (NULL == (stream = fopen(of->file, "r"))) { + if (NULL == (stream = fopen(mpage->mlinks->file, "r"))) { if (warnings) - say(of->file, NULL); + say(mpage->mlinks->file, NULL); return; } @@ -1100,7 +1145,11 @@ parse_catpage(struct of *of) if (NULL == title || '\0' == *title) { if (warnings) - say(of->file, "Cannot find NAME section"); + say(mpage->mlinks->file, + "Cannot find NAME section"); + assert(NULL == mpage->desc); + mpage->desc = mandoc_strdup(mpage->mlinks->name); + putkey(mpage, mpage->mlinks->name, TYPE_Nd); fclose(stream); free(title); return; @@ -1120,7 +1169,8 @@ 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; } @@ -1138,8 +1188,9 @@ parse_catpage(struct of *of) plen -= 2; } - of->desc = stradd(p); - putkey(of, p, TYPE_Nd); + assert(NULL == mpage->desc); + mpage->desc = mandoc_strdup(p); + putkey(mpage, mpage->desc, TYPE_Nd); fclose(stream); free(title); } @@ -1148,40 +1199,31 @@ 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, const char *value, uint64_t type) { assert(NULL != value); - wordaddbuf(of, value, strlen(value), type); + putkeys(mpage, 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 -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; @@ -1276,7 +1318,7 @@ parse_man(struct of *of, const struct man_node *n) byte = start[sz]; start[sz] = '\0'; - putkey(of, start, TYPE_Nm); + putkey(mpage, start, TYPE_Nm); if (' ' == byte) { start += sz + 1; @@ -1290,7 +1332,7 @@ parse_man(struct of *of, const struct man_node *n) } if (sv == start) { - putkey(of, start, TYPE_Nm); + putkey(mpage, start, TYPE_Nm); free(title); return; } @@ -1312,20 +1354,23 @@ parse_man(struct of *of, const struct man_node *n) while (' ' == *start) start++; - assert(NULL == of->desc); - of->desc = stradd(start); - putkey(of, start, TYPE_Nd); + assert(NULL == mpage->desc); + mpage->desc = mandoc_strdup(start); + putkey(mpage, mpage->desc, TYPE_Nd); 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); @@ -1341,23 +1386,23 @@ parse_mdoc(struct of *of, const struct mdoc_node *n) /* FALLTHROUGH */ 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 (MDOCF_CHILD & mdocs[n->tok].flags) - putmdockey(of, n->child, mdocs[n->tok].mask); + if (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; @@ -1395,23 +1440,23 @@ parse_mdoc_Fd(struct of *of, const struct mdoc_node *n end--; if (end > start) - putkeys(of, start, end - start + 1, TYPE_In); + putkeys(mpage, start, end - start + 1, TYPE_In); return(1); } static int -parse_mdoc_In(struct of *of, const struct mdoc_node *n) +parse_mdoc_In(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_In); + putkey(mpage, n->child->string, TYPE_In); return(1); } static int -parse_mdoc_Fn(struct of *of, const struct mdoc_node *n) +parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n) { const char *cp; @@ -1431,45 +1476,55 @@ 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_St(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); + putkey(mpage, n->child->string, TYPE_St); return(1); } static int -parse_mdoc_Xr(struct of *of, const struct mdoc_node *n) +parse_mdoc_Xr(struct mpage *mpage, 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(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(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; - char *sv, *desc; if (MDOC_BODY != n->type) return(0); @@ -1479,30 +1534,26 @@ 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 != mpage->desc) { + sz = strlen(mpage->desc) + + strlen(n->string) + 2; + mpage->desc = mandoc_realloc( + mpage->desc, sz); + strlcat(mpage->desc, " ", sz); + strlcat(mpage->desc, n->string, sz); + } else + mpage->desc = mandoc_strdup(n->string); } if (NULL != n->child) - parse_mdoc_Nd(of, n); + parse_mdoc_Nd(mpage, n); } - - of->desc = NULL != desc ? stradd(desc) : NULL; - free(desc); return(1); } 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) @@ -1514,114 +1565,56 @@ parse_mdoc_Nm(struct of *of, const struct mdoc_node *n } 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); } /* - * 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 mpage *mpage, + 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) { + if (NULL != s && mpage == s->mpage) { s->mask |= v; return; } 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->mpage = mpage; s->mask = v; - words = s; } /* @@ -1787,22 +1780,25 @@ 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) +dbindex(const 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"); + say(mpage->mlinks->file, "Adding to index"); if (nodb) return; desc = ""; - if (NULL != of->desc) { - key = hashget(of->desc, strlen(of->desc)); + if (NULL != mpage->desc && '\0' != *mpage->desc) { + key = ohash_find(&strings, + ohash_qlookup(&strings, mpage->desc)); assert(NULL != key); if (NULL == key->utf8) utf8key(mc, key); @@ -1812,17 +1808,34 @@ dbindex(struct mchars *mc, int form, const struct of * SQL_EXEC("BEGIN TRANSACTION"); 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]); + /* + * XXX The following three lines are obsolete + * and only kept for backward compatibility + * until apropos(1) and friends have caught up. + */ + SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->file); + SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->dsec); + SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->mlinks->arch); + SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, desc); + SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form); + SQL_STEP(stmts[STMT_INSERT_PAGE]); recno = sqlite3_last_insert_rowid(db); - sqlite3_reset(stmts[STMT_INSERT_DOC]); + sqlite3_reset(stmts[STMT_INSERT_PAGE]); - for (key = words; NULL != key; key = key->next) { - assert(key->of == of); + for (mlink = mpage->mlinks; mlink; mlink = mlink->next) { + i = 1; + SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->file); + 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, recno); + SQL_STEP(stmts[STMT_INSERT_LINK]); + sqlite3_reset(stmts[STMT_INSERT_LINK]); + } + + for (key = ohash_first(&strings, &slot); NULL != key; + key = ohash_next(&strings, &slot)) { + assert(key->mpage == mpage); if (NULL == key->utf8) utf8key(mc, key); i = 1; @@ -1831,6 +1844,9 @@ dbindex(struct mchars *mc, int form, const struct of * 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"); @@ -1839,19 +1855,24 @@ dbindex(struct mchars *mc, int form, const struct of * static void dbprune(void) { - struct of *of; + struct mpage *mpage; + struct mlink *mlink; size_t i; + unsigned int slot; if (nodb) return; - for (of = ofs; NULL != of; of = of->next) { + mpage = ohash_first(&mpages, &slot); + while (NULL != mpage) { + mlink = mpage->mlinks; i = 1; - SQL_BIND_TEXT(stmts[STMT_DELETE], i, of->file); - SQL_STEP(stmts[STMT_DELETE]); - sqlite3_reset(stmts[STMT_DELETE]); + SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE], i, mlink->file); + SQL_STEP(stmts[STMT_DELETE_PAGE]); + sqlite3_reset(stmts[STMT_DELETE_PAGE]); if (verb) - say(of->file, "Deleted from index"); + say(mlink->file, "Deleted from index"); + mpage = ohash_next(&mpages, &slot); } } @@ -1863,7 +1884,6 @@ static void dbclose(int real) { size_t i; - char file[PATH_MAX]; if (nodb) return; @@ -1879,9 +1899,7 @@ dbclose(int real) if (real) return; - strlcpy(file, MANDOC_DB, PATH_MAX); - strlcat(file, "~", PATH_MAX); - if (-1 == rename(file, MANDOC_DB)) { + if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) { exitcode = (int)MANDOCLEVEL_SYSERR; say(MANDOC_DB, NULL); } @@ -1898,29 +1916,24 @@ dbclose(int real) static int dbopen(int real) { - char file[PATH_MAX]; - const char *sql; + const char *file, *sql; int rc, ofl; - size_t sz; if (nodb) return(1); - sz = strlcpy(file, MANDOC_DB, PATH_MAX); - if ( ! real) - sz = strlcat(file, "~", PATH_MAX); + 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 >= PATH_MAX) { - 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) goto prepare_statements; @@ -1939,7 +1952,12 @@ dbopen(int real) return(0); } - sql = "CREATE TABLE \"docs\" (\n" + /* + * XXX The first three columns in table mpages are obsolete + * and only kept for backward compatibility + * until apropos(1) and friends have caught up. + */ + sql = "CREATE TABLE \"mpages\" (\n" " \"file\" TEXT NOT NULL,\n" " \"sec\" TEXT NOT NULL,\n" " \"arch\" TEXT NOT NULL,\n" @@ -1948,11 +1966,21 @@ dbopen(int real) " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n" ");\n" "\n" + "CREATE TABLE \"mlinks\" (\n" + " \"file\" TEXT NOT NULL,\n" + " \"sec\" TEXT NOT NULL,\n" + " \"arch\" TEXT NOT NULL,\n" + " \"name\" TEXT NOT NULL,\n" + " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) " + "ON DELETE CASCADE,\n" + " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\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" + " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) " + "ON DELETE CASCADE,\n" " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n" ");\n" "\n" @@ -1966,14 +1994,28 @@ dbopen(int real) 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 " + sql = "DELETE FROM mpages where file=?"; + sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL); + sql = "INSERT INTO mpages " "(file,sec,arch,desc,form) VALUES (?,?,?,?,?)"; - sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_DOC], NULL); + sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL); + sql = "INSERT INTO mlinks " + "(file,sec,arch,name,pageid) VALUES (?,?,?,?,?)"; + sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], 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__ + /* + * When opening a new database, we can turn off + * synchronous mode for much better performance. + */ + + if (real) + SQL_EXEC("PRAGMA synchronous = OFF"); +#endif + return(1); }