=================================================================== RCS file: /cvs/mandoc/mandocdb.c,v retrieving revision 1.58 retrieving revision 1.197 diff -u -p -r1.58 -r1.197 --- mandoc/mandocdb.c 2013/06/05 02:00:26 1.58 +++ mandoc/mandocdb.c 2015/10/06 18:32:19 1.197 @@ -1,70 +1,80 @@ -/* $Id: mandocdb.c,v 1.58 2013/06/05 02:00:26 schwarze Exp $ */ +/* $Id: mandocdb.c,v 1.197 2015/10/06 18:32:19 schwarze Exp $ */ /* * Copyright (c) 2011, 2012 Kristaps Dzonsons - * Copyright (c) 2011, 2012, 2013 Ingo Schwarze + * Copyright (c) 2011-2015 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 * copyright notice and this permission notice appear in all copies. * - * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF - * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ -#ifdef HAVE_CONFIG_H #include "config.h" -#endif +#include #include +#include #include #include #include #include +#if HAVE_FTS #include +#else +#include "compat_fts.h" +#endif #include #include #include +#include #include #include #include #include -#ifdef HAVE_OHASH +#if HAVE_OHASH #include #else #include "compat_ohash.h" #endif #include +#include "mandoc_aux.h" +#include "mandoc.h" +#include "roff.h" #include "mdoc.h" #include "man.h" -#include "mandoc.h" -#include "manpath.h" +#include "manconf.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 */ @@ -74,269 +84,278 @@ enum op { OP_TEST /* change no databases, report potential problems */ }; -enum form { - FORM_SRC, /* format is -man or -mdoc */ - FORM_CAT, /* format is cat */ - FORM_NONE /* format is unknown */ -}; - 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 */ + char key[]; /* rendered text */ }; -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 */ + int64_t pageid; /* pageid in mpages SQL table */ + char *sec; /* section from file content */ + char *arch; /* architecture from file content */ + char *title; /* title from file content */ + char *desc; /* description from file content */ + struct mlink *mlinks; /* singly linked list */ + int form; /* format from file content */ + int name_head_done; +}; + +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) */ + char *dsec; /* section from directory */ + char *arch; /* architecture from directory */ + char *name; /* name from file name (not empty) */ + char *fsec; /* section from file name suffix */ + struct mlink *next; /* singly linked list */ + struct mpage *mpage; /* parent */ + int dform; /* format from directory */ + int fform; /* format from file name suffix */ + int gzip; /* filename has a .gz suffix */ }; 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_SELECT_NAME, /* retrieve existing name flags */ + 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 roff_meta *, + const struct roff_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(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 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 dbclose(int); +static void dbadd(struct mpage *); +static void dbadd_mlink(const struct mlink *mlink); +static void dbadd_mlink_name(const struct mlink *mlink); +static int dbopen(int); +static void dbprune(void); +static void filescan(const char *); static void *hash_alloc(size_t, void *); -static void hash_free(void *, size_t, void *); -static void *hash_halloc(size_t, void *); -static void 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 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 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 path_reset(const char *, int, const char *); -static void putkey(const struct of *, - const char *, uint64_t); -static void putkeys(const struct of *, - const char *, int, 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 hash_free(void *, void *); +static void *hash_calloc(size_t, size_t, void *); +static void mlink_add(struct mlink *, const struct stat *); +static void mlink_check(struct mpage *, struct mlink *); +static void mlink_free(struct mlink *); +static void mlinks_undupe(struct mpage *); +static void mpages_free(void); +static void mpages_merge(struct mparse *); +static void names_check(void); +static void parse_cat(struct mpage *, int); +static void parse_man(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static void parse_mdoc(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static int parse_mdoc_body(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static int parse_mdoc_head(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static void parse_mdoc_fname(struct mpage *, const struct roff_node *); +static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *, + const struct roff_node *); +static void putkey(const struct mpage *, char *, uint64_t); +static void putkeys(const struct mpage *, char *, size_t, uint64_t); +static void putmdockey(const struct mpage *, + const struct roff_node *, uint64_t); +static int render_string(char **, size_t *); +static void say(const char *, const char *, ...); +static int set_basedir(const char *, int); +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 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 enum op op; /* operational mode */ -static struct ohash inos; /* table of inodes/devices */ -static struct ohash filenames; /* table of filenames */ +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 char basedir[PATH_MAX]; /* current base directory */ +static struct mchars *mchars; /* table of named characters */ +static struct ohash mpages; /* table of distinct manual pages */ +static struct ohash mlinks; /* table of directory entries */ +static struct ohash names; /* table of all names */ 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 uint64_t name_mask; 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 */ + { NULL, TYPE_In }, /* In */ + { NULL, TYPE_Li }, /* Li */ + { parse_mdoc_Nd, 0 }, /* Nd */ + { parse_mdoc_Nm, 0 }, /* Nm */ + { NULL, 0 }, /* Op */ + { NULL, 0 }, /* Ot */ + { NULL, TYPE_Pa }, /* Pa */ + { NULL, 0 }, /* Rv */ + { NULL, TYPE_St }, /* 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_Fo, 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 */ + { NULL, 0 }, /* ll */ }; + int -main(int argc, char *argv[]) +mandocdb(int argc, char *argv[]) { - char cwd[PATH_MAX]; - int ch, rc, fd, i; - unsigned int index; - size_t j, sz; - const char *dir; - struct str *s; - struct mchars *mc; - struct manpaths dirs; + struct manconf conf; + struct ohash_info mpages_info, mlinks_info; struct mparse *mp; - struct ohash_info ino_info, filename_info, str_info; + const char *path_arg; + size_t j, sz; + int ch, i; + memset(&conf, 0, sizeof(conf)); 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.calloc = mlinks_info.calloc = hash_calloc; + mpages_info.free = mlinks_info.free = hash_free; + mpages_info.data = mlinks_info.data = NULL; - 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) @@ -345,68 +364,70 @@ 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, PATH_MAX)) { - 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. + * We accept a few different invocations. * The CHECKOP macro makes sure that invocation styles don't * clobber each other. */ #define CHECKOP(_op, _ch) do \ if (OP_DEFAULT != (_op)) { \ - fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \ + fprintf(stderr, "%s: -%c: Conflicting option\n", \ + progname, (_ch)); \ goto usage; \ } while (/*CONSTCOND*/0) - dir = NULL; + 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); - dir = optarg; + path_arg = optarg; op = OP_CONFFILE; break; - case ('d'): + case 'D': + debug++; + break; + case 'd': CHECKOP(op, ch); - dir = optarg; + 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, "%s: -T%s: " + "Unsupported output format\n", + progname, 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); - dir = optarg; + 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; } @@ -415,161 +436,165 @@ main(int argc, char *argv[]) argv += optind; if (OP_CONFFILE == op && argc > 0) { - fprintf(stderr, "-C: Too many arguments\n"); + fprintf(stderr, "%s: -C: Too many arguments\n", + progname); goto usage; } - rc = 1; - mp = mparse_alloc(MPARSE_AUTO, - MANDOCLEVEL_FATAL, NULL, NULL, NULL); - mc = mchars_alloc(); + exitcode = (int)MANDOCLEVEL_OK; + mchars = mchars_alloc(); + mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, + mchars, NULL); + ohash_init(&mpages, 6, &mpages_info); + ohash_init(&mlinks, 6, &mlinks_info); - ohash_init(&strings, 6, &str_info); - ohash_init(&inos, 6, &ino_info); - ohash_init(&filenames, 6, &filename_info); - if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) { - /* - * 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. + * Most of these deal with a specific directory. + * Jump into that directory first. */ - if (0 == path_reset(cwd, fd, dir)) + if (OP_TEST != op && 0 == set_basedir(path_arg, 1)) goto out; - for (i = 0; i < argc; i++) - filescan(argv[i], dir); - if (0 == dbopen(dir, 1)) - goto out; - if (OP_TEST != op) - dbprune(dir); + + 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) - rc = ofmerge(mc, mp, dir); - dbclose(dir, 1); + mpages_merge(mp); + dbclose(OP_DEFAULT == op ? 0 : 1); } else { /* * If we have arguments, use them as our manpaths. * If we don't, grok from manpath(1) or however else - * manpath_parse() wants to do it. + * manconf_parse() wants to do it. */ if (argc > 0) { - dirs.paths = mandoc_calloc - (argc, sizeof(char *)); - dirs.sz = (size_t)argc; + conf.manpath.paths = mandoc_reallocarray(NULL, + argc, sizeof(char *)); + conf.manpath.sz = (size_t)argc; for (i = 0; i < argc; i++) - dirs.paths[i] = mandoc_strdup(argv[i]); + conf.manpath.paths[i] = mandoc_strdup(argv[i]); } else - manpath_parse(&dirs, dir, NULL, NULL); + manconf_parse(&conf, path_arg, NULL, NULL); + if (conf.manpath.sz == 0) { + exitcode = (int)MANDOCLEVEL_BADARG; + say("", "Empty manpath"); + } + /* - * 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. */ - for (j = 0; j < dirs.sz; j++) { - sz = strlen(dirs.paths[j]); - if (sz && '/' == dirs.paths[j][sz - 1]) - dirs.paths[j][--sz] = '\0'; + for (j = 0; j < conf.manpath.sz; j++) { + sz = strlen(conf.manpath.paths[j]); + if (sz && conf.manpath.paths[j][sz - 1] == '/') + conf.manpath.paths[j][--sz] = '\0'; if (0 == sz) continue; - if (0 == path_reset(cwd, fd, dirs.paths[j])) - goto out; - if (0 == treescan(dirs.paths[j])) - goto out; - if (0 == path_reset(cwd, fd, dirs.paths[j])) - goto out; - if (0 == dbopen(dirs.paths[j], 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 + if (j) { + ohash_init(&mpages, 6, &mpages_info); + ohash_init(&mlinks, 6, &mlinks_info); + } - 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); + if ( ! set_basedir(conf.manpath.paths[j], argc > 0)) + continue; + if (0 == treescan()) + continue; + if (0 == dbopen(0)) + continue; + + mpages_merge(mp); + if (warnings && !nodb && + ! (MPARSE_QUICK & mparse_options)) + names_check(); + dbclose(0); + + if (j + 1 < conf.manpath.sz) { + mpages_free(); + ohash_delete(&mpages); + ohash_delete(&mlinks); + } } } out: - close(fd); - manpath_free(&dirs); - mchars_free(mc); + manconf_free(&conf); 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); + mchars_free(mchars); + 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(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. * * If use_all has been specified, grok all files. * If not, sanitise paths to the following: * - * [./]man*[/]/.
+ * [./]man*[/]/.
* or * [./]cat
[/]/.0 * * TODO: accomodate for multi-language directories. */ static int -treescan(const char *base) +treescan(void) { + char buf[PATH_MAX]; FTS *f; FTSENT *ff; - int dform; - char *sec; - const char *dsec, *arch, *cp, *name, *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) { - perror(base); - return(0); + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "&fts_open"); + return 0; } dsec = arch = NULL; @@ -577,162 +602,272 @@ treescan(const char *base) 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 - * stored directory data and handling the filename. - * Disallow duplicate (hard-linked) files. + * Symbolic links require various sanity checks, + * then get handled just like regular files. */ - if (FTS_F == ff->fts_info) { - if ( ! use_all && ff->fts_level < 2) { + case FTS_SL: + if (NULL == realpath(path, buf)) { if (warnings) - say(base, path, "Extraneous file"); + say(path, "&realpath"); continue; - } else if (inocheck(ff->fts_statp)) { + } + if (strstr(buf, basedir) != buf +#ifdef HOMEBREWDIR + && strstr(buf, HOMEBREWDIR) != buf +#endif + ) { + 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(base, path, "Duplicate file"); + say(path, "&stat"); continue; - } + } + /* FALLTHROUGH */ - cp = ff->fts_name; - - if (0 == strcmp(cp, "mandocdb.db")) { + /* + * If we're a regular file, add an mlink by using the + * stored directory data and handling the filename. + */ + case FTS_F: + if (0 == strcmp(path, MANDOC_DB)) + continue; + if ( ! use_all && ff->fts_level < 2) { if (warnings) - say(base, path, "Skip database"); + say(path, "Extraneous file"); continue; - } else if (NULL != (cp = strrchr(cp, '.'))) { - if (0 == strcmp(cp + 1, "html")) { + } + 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(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; } - } + } else if (0 == strcmp(++fsec, "html")) { + if (warnings) + say(path, "Skip html"); + continue; + } else if (0 == strcmp(fsec, "ps")) { + if (warnings) + say(path, "Skip ps"); + continue; + } else if (0 == strcmp(fsec, "pdf")) { + if (warnings) + say(path, "Skip pdf"); + continue; + } else if ( ! use_all && + ((FORM_SRC == dform && + strncmp(fsec, dsec, strlen(dsec))) || + (FORM_CAT == dform && strcmp(fsec, "0")))) { + if (warnings) + say(path, "Wrong filename suffix"); + continue; + } else + fsec[-1] = '\0'; - if (NULL != (sec = strrchr(ff->fts_name, '.'))) { - *sec = '\0'; - sec = stradd(sec + 1); + 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; } - name = stradd(ff->fts_name); - ofadd(base, dform, path, - name, dsec, sec, arch, ff->fts_statp); + 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; 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; + } else { + dform = FORM_NONE; + dsec = NULL; } - if (NULL != dsec || use_all) + 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): + case 2: /* * Possibly our architecture. * If we're descending, keep tabs on it. */ - arch = NULL; if (FTS_DP != ff->fts_info && NULL != dsec) - arch = stradd(ff->fts_name); + arch = ff->fts_name; + else + arch = NULL; 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; } } fts_close(f); - return(1); + return 1; } /* - * 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, const char *base) +filescan(const char *file) { - const char *sec, *arch, *name, *dsec; - char *p, *start, *buf; - int dform; + char buf[PATH_MAX]; struct stat st; + struct mlink *mlink; + char *p, *start; assert(use_all); if (0 == strncmp(file, "./", 2)) file += 2; - if (-1 == stat(file, &st)) { - if (warnings) - say(base, file, "%s", strerror(errno)); + /* + * We have to do lstat(2) before realpath(3) loses + * the information whether this is a symbolic link. + * We need to know that because for symbolic links, + * we want to use the orginal file name, while for + * regular files, we want to use the real path. + */ + if (-1 == lstat(file, &st)) { + exitcode = (int)MANDOCLEVEL_BADARG; + say(file, "&lstat"); return; - } else if ( ! (S_IFREG & st.st_mode)) { - if (warnings) - say(base, file, "Not a regular file"); + } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) { + exitcode = (int)MANDOCLEVEL_BADARG; + say(file, "Not a regular file"); return; - } else if (inocheck(&st)) { - if (warnings) - say(base, file, "Duplicate file"); + } + + /* + * We have to resolve the file name to the real path + * in any case for the base directory check. + */ + if (NULL == realpath(file, buf)) { + exitcode = (int)MANDOCLEVEL_BADARG; + say(file, "&realpath"); return; } - buf = mandoc_strdup(file); - start = buf; - sec = arch = name = dsec = NULL; - dform = FORM_NONE; + if (OP_TEST == op) + start = buf; + else if (strstr(buf, basedir) == buf) + start = buf + strlen(basedir); +#ifdef HOMEBREWDIR + else if (strstr(buf, HOMEBREWDIR) == buf) + start = buf; +#endif + else { + exitcode = (int)MANDOCLEVEL_BADARG; + say("", "%s: outside base directory", buf); + return; + } /* + * Now we are sure the file is inside our tree. + * If it is a symbolic link, ignore the real path + * and use the original name. + * This implies passing stuff like "cat1/../man1/foo.1" + * on the command line won't work. So don't do that. + * Note the stat(2) can still fail if the link target + * doesn't exist. + */ + if (S_IFLNK & st.st_mode) { + if (-1 == stat(buf, &st)) { + exitcode = (int)MANDOCLEVEL_BADARG; + say(file, "&stat"); + return; + } + if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) { + say(file, "Filename too long"); + return; + } + start = buf; + if (OP_TEST != op && strstr(buf, basedir) == buf) + start += strlen(basedir); + } + + mlink = mandoc_calloc(1, sizeof(struct mlink)); + mlink->dform = FORM_NONE; + if (strlcpy(mlink->file, start, sizeof(mlink->file)) >= + sizeof(mlink->file)) { + say(start, "Filename too long"); + free(mlink); + return; + } + + /* * First try to guess our directory structure. * If we find a separator, try to look for man* or cat*. * If we find one of these and what's underneath is a directory, @@ -741,19 +876,19 @@ filescan(const char *file, const char *base) 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; - } + } } /* @@ -766,296 +901,411 @@ filescan(const char *file, const char *base) 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(base, dform, file, name, dsec, sec, arch, &st); - free(buf); + } + mlink_add(mlink, &st); } -/* - * See fileadd(). - */ -static int -filecheck(const char *name) +static void +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, name); - return(NULL != ohash_find(&filenames, index)); + 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); + + memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */ + 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 index; - index = ohash_qlookup(&filenames, of->file); - assert(NULL == ohash_find(&filenames, index)); - ohash_insert(&filenames, index, 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; - unsigned int index; + 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; - index = ohash_lookup_memory - (&inos, (char *)&id, sizeof(id), hash); - - return(NULL != ohash_find(&inos, index)); + 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 index; + 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; - index = ohash_lookup_memory - (&inos, (char *)&of->id, sizeof(of->id), hash); - - assert(NULL == ohash_find(&inos, index)); - ohash_insert(&inos, index, 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(const char *base, 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 = name; - of->sec = sec; - of->dsec = dsec; - of->arch = 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); - } + /* + * 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 "base". + * 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 int -ofmerge(struct mchars *mc, struct mparse *mp, const char *base) +static void +mpages_merge(struct mparse *mp) { - 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; + char any[] = "any"; + struct ohash_info str_info; + struct mpage *mpage, *mpage_dest; + struct mlink *mlink, *mlink_dest; + struct roff_man *man; + char *sodest; + char *cp; + int fd; + unsigned int pslot; - 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(base, 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(base, of->file, "Man " - "source exists: %s", buf); - continue; - } + str_info.alloc = hash_alloc; + str_info.calloc = hash_calloc; + str_info.free = hash_free; + str_info.data = NULL; + str_info.key_offset = offsetof(struct str, key); + + if ( ! nodb) + SQL_EXEC("BEGIN TRANSACTION"); + + mpage = ohash_first(&mpages, &pslot); + while (mpage != NULL) { + mlinks_undupe(mpage); + if ((mlink = mpage->mlinks) == NULL) { + mpage = ohash_next(&mpages, &pslot); + continue; } - words = NULL; + 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; + sodest = NULL; + mparse_open(mp, &fd, mlink->file); + if (fd == -1) { + say(mlink->file, "&open"); + goto nextpage; + } + /* - * 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. + * Interpret the file as mdoc(7) or man(7) source + * code, unless it is known to be formatted. */ - if (FORM_SRC == of->dform || FORM_SRC == of->sform) { - lvl = mparse_readfd(mp, -1, of->file); - if (lvl < MANDOCLEVEL_FATAL) - mparse_result(mp, &mdoc, &man); - } + if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) { + mparse_readfd(mp, fd, mlink->file); + mparse_result(mp, &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 (sodest != NULL) { + mlink_dest = ohash_find(&mlinks, + ohash_qlookup(&mlinks, sodest)); + if (mlink_dest == NULL) { + mandoc_asprintf(&cp, "%s.gz", sodest); + mlink_dest = ohash_find(&mlinks, + ohash_qlookup(&mlinks, cp)); + free(cp); + } + if (mlink_dest != NULL) { - 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)) - say(base, of->file, "Section \"%s\" " - "manual in %s directory", - msec, of->dsec); + mpage_dest = mlink_dest->mpage; + while (1) { + mlink->mpage = mpage_dest; - /* - * 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)) - say(base, of->file, "Architecture \"%s\" " - "manual in \"%s\" directory", - march, of->arch); + /* + * If the target was already + * processed, add the links + * to the database now. + * Otherwise, this will + * happen when we come + * to the target. + */ - putkey(of, of->name, TYPE_Nm); + if (mpage_dest->pageid) + dbadd_mlink_name(mlink); - if (NULL != mdoc) { - if (NULL != (cp = mdoc_meta(mdoc)->name)) - putkey(of, cp, TYPE_Nm); - parse_mdoc(of, mdoc_node(mdoc)); - } else if (NULL != man) - parse_man(of, man_node(man)); + if (mlink->next == NULL) + 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 (man != NULL && man->macroset == MACROSET_MDOC) { + mpage->form = FORM_SRC; + mpage->sec = man->meta.msec; + mpage->sec = mandoc_strdup( + mpage->sec == NULL ? "" : mpage->sec); + mpage->arch = man->meta.arch; + mpage->arch = mandoc_strdup( + mpage->arch == NULL ? "" : mpage->arch); + mpage->title = mandoc_strdup(man->meta.title); + } else if (man != NULL && man->macroset == MACROSET_MAN) { + mpage->form = FORM_SRC; + mpage->sec = mandoc_strdup(man->meta.msec); + mpage->arch = mandoc_strdup(mlink->arch); + mpage->title = mandoc_strdup(man->meta.title); + } else { + mpage->form = FORM_CAT; + mpage->sec = mandoc_strdup(mlink->dsec); + mpage->arch = mandoc_strdup(mlink->arch); + mpage->title = mandoc_strdup(mlink->name); + } + putkey(mpage, mpage->sec, TYPE_sec); + if (*mpage->arch != '\0') + putkey(mpage, mpage->arch, TYPE_arch); + + for ( ; mlink != NULL; 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); + } + + assert(mpage->desc == NULL); + if (man != NULL && man->macroset == MACROSET_MDOC) + parse_mdoc(mpage, &man->meta, man->first); + else if (man != NULL) + parse_man(mpage, &man->meta, man->first); else - parse_catpage(of, base); + parse_cat(mpage, fd); + if (mpage->desc == NULL) + mpage->desc = mandoc_strdup(mpage->mlinks->name); - dbindex(mc, form, of, base); + if (warnings && !use_all) + for (mlink = mpage->mlinks; mlink; + mlink = mlink->next) + mlink_check(mpage, mlink); + + dbadd(mpage); + mlink = mpage->mlinks; + +nextpage: + ohash_delete(&strings); + ohash_delete(&names); + mpage = ohash_next(&mpages, &pslot); } - return(1); + if (0 == nodb) + SQL_EXEC("END TRANSACTION"); } static void -parse_catpage(struct of *of, const char *base) +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 = (const char *)sqlite3_column_text(stmt, 0); + sec = (const char *)sqlite3_column_text(stmt, 1); + arch = (const char *)sqlite3_column_text(stmt, 2); + key = (const char *)sqlite3_column_text(stmt, 3); + say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec, + '\0' == *arch ? "" : "/", + '\0' == *arch ? "" : arch, key); + } + sqlite3_finalize(stmt); +} + +static void +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 (-1 != fd) + close(fd); if (warnings) - say(base, of->file, "%s", strerror(errno)); + say(mpage->mlinks->file, "&fopen"); return; } @@ -1073,7 +1323,7 @@ parse_catpage(struct of *of, const char *base) 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, @@ -1108,7 +1358,8 @@ parse_catpage(struct of *of, const char *base) if (NULL == title || '\0' == *title) { if (warnings) - say(base, of->file, "Cannot find NAME section"); + say(mpage->mlinks->file, + "Cannot find NAME section"); fclose(stream); free(title); return; @@ -1128,7 +1379,8 @@ 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(mpage->mlinks->file, + "No dash in title line"); p = title; } @@ -1141,13 +1393,12 @@ parse_catpage(struct of *of, const char *base) if (0 == len) { memmove(line, line + 1, plen--); continue; - } + } memmove(line - 1, line + 1, plen - len); plen -= 2; } - of->desc = stradd(p); - putkey(of, p, TYPE_Nd); + mpage->desc = mandoc_strdup(p); fclose(stream); free(title); } @@ -1156,48 +1407,45 @@ parse_catpage(struct of *of, const char *base) * 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); - wordaddbuf(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); } /* - * 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 roff_node *n, uint64_t m) { for ( ; NULL != n; n = n->next) { if (NULL != n->child) - putmdockey(of, n->child, m); - if (MDOC_TEXT == n->type) - putkey(of, n->string, m); + putmdockey(mpage, n->child, m); + if (n->type == ROFFT_TEXT) + putkey(mpage, n->string, m); } } -static int -parse_man(struct of *of, const struct man_node *n) +static void +parse_man(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - const struct man_node *head, *body; - char *start, *sv, *title; + const struct roff_node *head, *body; + char *start, *title; char byte; - size_t sz, titlesz; + size_t sz; if (NULL == n) - return(0); + return; /* * We're only searching for one thing: the first text child in @@ -1206,20 +1454,16 @@ parse_man(struct of *of, const struct man_node *n) * the correct section or not. */ - if (MAN_BODY == n->type && MAN_SH == n->tok) { + if (n->type == ROFFT_BODY && n->tok == MAN_SH) { 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)) && + head->type == ROFFT_TEXT && + 0 == strcmp(head->string, "NAME") && + NULL != body->child) { - title = NULL; - titlesz = 0; - /* * Suck the entire NAME section into memory. * Yes, we might run away. @@ -1227,48 +1471,12 @@ 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; + deroff(&title, body); if (NULL == title) - return(1); + 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(1); - } - - /* Erase trailing space. */ - - start = &sv[sz - 1]; - while (start > sv && isspace((unsigned char)*start)) - *start-- = '\0'; - - if (start == sv) { - free(title); - return(1); - } - - start = sv; - - /* + /* * Go through a special heuristic dance here. * Conventionally, one or more manual names are * comma-specified prior to a whitespace, then a @@ -1276,6 +1484,7 @@ parse_man(struct of *of, const struct man_node *n) * the name parts here. */ + start = title; for ( ;; ) { sz = strcspn(start, " ,"); if ('\0' == start[sz]) @@ -1284,8 +1493,22 @@ 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 ( ! (mpage->name_head_done || + strcasecmp(start, meta->title))) { + putkey(mpage, start, NAME_HEAD); + mpage->name_head_done = 1; + } + if (' ' == byte) { start += sz + 1; break; @@ -1297,10 +1520,15 @@ 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); + if ( ! (mpage->name_head_done || + strcasecmp(start, meta->title))) { + putkey(mpage, start, NAME_HEAD); + mpage->name_head_done = 1; + } free(title); - return(1); + return; } while (isspace((unsigned char)*start)) @@ -1320,63 +1548,63 @@ 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); + mpage->desc = mandoc_strdup(start); free(title); - return(1); + return; } } - for (n = n->child; n; n = n->next) - if (parse_man(of, n)) - return(1); - - return(0); + for (n = n->child; n; n = n->next) { + if (NULL != mpage->desc) + break; + parse_man(mpage, meta, n); + } } static void -parse_mdoc(struct of *of, const struct mdoc_node *n) +parse_mdoc(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { assert(NULL != n); for (n = n->child; NULL != n; n = n->next) { switch (n->type) { - case (MDOC_ELEM): + case ROFFT_ELEM: /* FALLTHROUGH */ - case (MDOC_BLOCK): + case ROFFT_BLOCK: /* FALLTHROUGH */ - case (MDOC_HEAD): + case ROFFT_HEAD: /* FALLTHROUGH */ - case (MDOC_BODY): + case ROFFT_BODY: /* FALLTHROUGH */ - case (MDOC_TAIL): + case ROFFT_TAIL: if (NULL != mdocs[n->tok].fp) - if (0 == (*mdocs[n->tok].fp)(of, n)) + if (0 == (*mdocs[n->tok].fp)(mpage, meta, 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); + assert(n->type != ROFFT_ROOT); continue; } if (NULL != n->child) - parse_mdoc(of, n); + parse_mdoc(mpage, meta, n); } } static int -parse_mdoc_Fd(struct of *of, const struct mdoc_node *n) +parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - const char *start, *end; + char *start, *end; size_t sz; if (SEC_SYNOPSIS != n->sec || - NULL == (n = n->child) || - MDOC_TEXT != n->type) - return(0); + NULL == (n = n->child) || + n->type != ROFFT_TEXT) + return 0; /* * Only consider those `Fd' macro fields that begin with an @@ -1384,10 +1612,10 @@ parse_mdoc_Fd(struct of *of, const struct mdoc_node *n */ if (strcmp("#include", n->string)) - return(0); + return 0; - if (NULL == (n = n->next) || MDOC_TEXT != n->type) - return(0); + if ((n = n->next) == NULL || n->type != ROFFT_TEXT) + return 0; /* * Strip away the enclosing angle brackets and make sure we're @@ -1399,240 +1627,199 @@ parse_mdoc_Fd(struct of *of, const struct mdoc_node *n start++; if (0 == (sz = strlen(start))) - return(0); + return 0; end = &start[(int)sz - 1]; if ('>' == *end || '"' == *end) 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) +static void +parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n) { + char *cp; + size_t sz; - if (NULL != n->child && MDOC_TEXT == n->child->type) - return(0); + if (n->type != ROFFT_TEXT) + return; - putkey(of, n->child->string, TYPE_In); - return(1); + /* Skip function pointer punctuation. */ + + cp = n->string; + while (*cp == '(' || *cp == '*') + cp++; + sz = strcspn(cp, "()"); + + putkeys(mpage, cp, sz, TYPE_Fn); + if (n->sec == SEC_SYNOPSIS) + putkeys(mpage, cp, sz, NAME_SYN); } static int -parse_mdoc_Fn(struct of *of, const struct mdoc_node *n) +parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - const char *cp; - if (NULL == (n = n->child) || MDOC_TEXT != n->type) - return(0); + if (n->child == NULL) + return 0; - /* - * Parse: .Fn "struct type *name" "char *arg". - * First strip away pointer symbol. - * Then store the function name, then type. - * Finally, store the arguments. - */ + parse_mdoc_fname(mpage, n->child); - if (NULL == (cp = strrchr(n->string, ' '))) - cp = n->string; + for (n = n->child->next; n != NULL; n = n->next) + if (n->type == ROFFT_TEXT) + putkey(mpage, n->string, TYPE_Fa); - while ('*' == *cp) - cp++; - - putkey(of, cp, TYPE_Fn); - - if (n->string < cp) - putkeys(of, 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); - - return(0); + return 0; } static int -parse_mdoc_St(struct of *of, const struct mdoc_node *n) +parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - if (NULL == n->child || MDOC_TEXT != n->child->type) - return(0); + if (n->type != ROFFT_HEAD) + return 1; - putkey(of, n->child->string, TYPE_St); - return(1); + if (n->child != NULL) + parse_mdoc_fname(mpage, n->child); + + return 0; } static int -parse_mdoc_Xr(struct of *of, const struct mdoc_node *n) +parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { + char *cp; if (NULL == (n = n->child)) - return(0); + return 0; - putkey(of, n->string, TYPE_Xr); - return(1); + if (NULL == n->next) { + putkey(mpage, n->string, TYPE_Xr); + return 0; + } + + 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 roff_meta *meta, + const struct roff_node *n) { - size_t sz; - char *sv, *desc; - if (MDOC_BODY != n->type) - return(0); - - /* - * Special-case the `Nd' because we need to put the description - * 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 != n->child) - parse_mdoc_Nd(of, n); - } - - of->desc = NULL != desc ? stradd(desc) : NULL; - free(desc); - return(1); + if (n->type == ROFFT_BODY) + 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 roff_meta *meta, + const struct roff_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 (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) { + if (n->child == NULL) + putkey(mpage, meta->name, NAME_SYN); + else + putmdockey(mpage, n->child, NAME_SYN); + } + if ( ! (mpage->name_head_done || + n->child == NULL || n->child->string == NULL || + strcasecmp(n->child->string, meta->title))) { + putkey(mpage, n->child->string, ROFFT_HEAD); + mpage->name_head_done = 1; + } + return 0; } static int -parse_mdoc_Sh(struct of *of, const struct mdoc_node *n) +parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type); + return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD; } static int -parse_mdoc_head(struct of *of, const struct mdoc_node *n) +parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - return(MDOC_HEAD == n->type); + return n->type == ROFFT_HEAD; } static int -parse_mdoc_body(struct of *of, const struct mdoc_node *n) +parse_mdoc_body(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - return(MDOC_BODY == n->type); + return n->type == ROFFT_BODY; } /* - * 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, char *cp, size_t sz, uint64_t v) { + struct ohash *htab; struct str *s; - unsigned int index; const char *end; + unsigned int slot; + int i, mustfree; if (0 == sz) return; - s = hashget(cp, sz); + mustfree = render_string(&cp, &sz); - if (NULL != s && of == s->of) { + if (TYPE_Nm & v) { + htab = &names; + v &= name_mask; + if (v & NAME_FIRST) + name_mask &= ~NAME_FIRST; + if (debug > 1) + say(mpage->mlinks->file, + "Adding name %*s, bits=%d", sz, cp, v); + } else { + htab = &strings; + if (debug > 1) + for (i = 0; i < mansearch_keymax; i++) + if ((uint64_t)1 << i & v) + say(mpage->mlinks->file, + "Adding key %s=%*s", + mansearch_keynames[i], sz, cp); + } + + end = cp + sz; + slot = ohash_qlookupi(htab, cp, &end); + s = ohash_find(htab, slot); + + 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); - end = cp + sz; - index = ohash_qlookupi(&strings, cp, &end); - assert(NULL == ohash_find(&strings, index)); - ohash_insert(&strings, index, s); + ohash_insert(htab, slot, s); } - - s->next = words; - s->of = of; + s->mpage = mpage; s->mask = v; - words = s; + + if (mustfree) + free(cp); } /* @@ -1681,190 +1868,304 @@ utf8(unsigned int cp, char out[7]) out[4] = (cp >> 6 & 63) | 128; out[5] = (cp & 63) | 128; } else - return(0); + return 0; out[rc] = '\0'; - return(rc); + return rc; } /* - * Store the UTF-8 version of a key, or alias the pointer if the key has - * no UTF-8 transcription marks in it. + * If the string contains escape sequences, + * replace it with an allocated rendering and return 1, + * such that the caller can free it after use. + * Otherwise, do nothing and return 0. */ -static void -utf8key(struct mchars *mc, struct str *key) +static int +render_string(char **public, size_t *psz) { - size_t sz, bsz, pos; - char utfbuf[7], res[5]; - char *buf; - const char *seq, *cpp, *val; - int len, u; - enum mandoc_esc esc; + const char *src, *scp, *addcp, *seq; + char *dst; + size_t ssz, dsz, addsz; + char utfbuf[7], res[6]; + int seqlen, unicode; - assert(NULL == key->utf8); - 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); + src = scp = *public; + ssz = *psz; + dst = NULL; + dsz = 0; - /* - * Pre-check: if we have no stop-characters, then set the - * pointer as ourselvse and get out of here. - */ - if (strcspn(val, res) == bsz) { - key->utf8 = key->key; - return; - } + while (scp < src + *psz) { - /* Pre-allocate by the length of the input */ + /* Leave normal characters unchanged. */ - buf = mandoc_malloc(++bsz); - pos = 0; + if (strchr(res, *scp) == NULL) { + if (dst != NULL) + dst[dsz++] = *scp; + scp++; + continue; + } - while ('\0' != *val) { /* - * Halt on the first escape sequence. - * This also halts on the end of string, in which case - * we just copy, fallthrough, and exit the loop. + * Found something that requires replacing, + * make sure we have a destination buffer. */ - if ((sz = strcspn(val, res)) > 0) { - memcpy(&buf[pos], val, sz); - pos += sz; - val += sz; + + if (dst == NULL) { + dst = mandoc_malloc(ssz + 1); + dsz = scp - src; + memcpy(dst, src, dsz); } - if (ASCII_HYPH == *val) { - buf[pos++] = '-'; - val++; + /* Handle single-char special characters. */ + + switch (*scp) { + case '\\': + break; + case '\t': + /* FALLTHROUGH */ + case ASCII_NBRSP: + dst[dsz++] = ' '; + scp++; continue; - } else if ('\t' == *val || ASCII_NBRSP == *val) { - buf[pos++] = ' '; - val++; + case ASCII_HYPH: + dst[dsz++] = '-'; + /* FALLTHROUGH */ + case ASCII_BREAK: + scp++; continue; - } else if ('\\' != *val) - break; + default: + abort(); + } - /* Read past the slash. */ - - val++; - u = 0; - /* - * Parse the escape sequence and see if it's a - * predefined character or special character. + * Found an escape sequence. + * Read past the slash, then parse it. + * Ignore everything except characters. */ - esc = mandoc_escape - ((const char **)&val, &seq, &len); - if (ESCAPE_ERROR == esc) - break; - if (ESCAPE_SPECIAL != esc) + scp++; + if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL) continue; - 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) { + unicode = mchars_spec2cp(mchars, seq, seqlen); + if (unicode <= 0) + continue; + addsz = utf8(unicode, utfbuf); + if (addsz == 0) + continue; + addcp = utfbuf; + } else { + addcp = mchars_spec2str(mchars, seq, seqlen, &addsz); + if (addcp == NULL) + continue; + if (*addcp == ASCII_NBRSP) { + addcp = " "; + addsz = 1; + } + } + /* Copy the rendered glyph into the stream. */ - sz = strlen(cpp); - bsz += sz; + ssz += addsz; + dst = mandoc_realloc(dst, ssz + 1); + memcpy(dst + dsz, addcp, addsz); + dsz += addsz; + } + if (dst != NULL) { + *public = dst; + *psz = dsz; + } - buf = mandoc_realloc(buf, bsz); + /* Trim trailing whitespace and NUL-terminate. */ - memcpy(&buf[pos], cpp, sz); - pos += sz; + while (*psz > 0 && (*public)[*psz - 1] == ' ') + --*psz; + if (dst != NULL) { + (*public)[*psz] = '\0'; + return 1; + } else + return 0; +} + +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]); +} + +static void +dbadd_mlink_name(const struct mlink *mlink) +{ + uint64_t bits; + size_t i; + + dbadd_mlink(mlink); + + i = 1; + SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid); + bits = NAME_FILE & NAME_MASK; + if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) { + bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0); + sqlite3_reset(stmts[STMT_SELECT_NAME]); } - buf[pos] = '\0'; - key->utf8 = buf; + i = 1; + SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits); + SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name); + SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid); + SQL_STEP(stmts[STMT_INSERT_NAME]); + sqlite3_reset(stmts[STMT_INSERT_NAME]); } /* * Flush the current page's terms (and their bits) into the database. * Wrap the entire set of additions in a transaction to make sqlite be a * 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, const char *base) +dbadd(struct mpage *mpage) { + struct mlink *mlink; struct str *key; - const char *desc; - int64_t recno; + char *cp; size_t i; + unsigned int slot; + int mustfree; - if (verb) - say(base, of->file, "Adding to index"); + mlink = mpage->mlinks; - if (nodb) + if (nodb) { + for (key = ohash_first(&names, &slot); NULL != key; + key = ohash_next(&names, &slot)) + free(key); + for (key = ohash_first(&strings, &slot); NULL != key; + key = ohash_next(&strings, &slot)) + free(key); + if (0 == debug) + return; + while (NULL != mlink) { + fputs(mlink->name, stdout); + if (NULL == mlink->next || + 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) { - key = hashget(of->desc, strlen(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"); + cp = mpage->desc; + i = strlen(cp); + mustfree = render_string(&cp, &i); 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, cp); + SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form); + SQL_STEP(stmts[STMT_INSERT_PAGE]); + mpage->pageid = sqlite3_last_insert_rowid(db); + sqlite3_reset(stmts[STMT_INSERT_PAGE]); + if (mustfree) + free(cp); - for (key = words; NULL != key; key = key->next) { - assert(key->of == of); - if (NULL == key->utf8) - utf8key(mc, key); + 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); i = 1; + SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask); + SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key); + SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid); + SQL_STEP(stmts[STMT_INSERT_NAME]); + sqlite3_reset(stmts[STMT_INSERT_NAME]); + free(key); + } + for (key = ohash_first(&strings, &slot); NULL != key; + key = ohash_next(&strings, &slot)) { + assert(key->mpage == mpage); + 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->key); + SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid); SQL_STEP(stmts[STMT_INSERT_KEY]); sqlite3_reset(stmts[STMT_INSERT_KEY]); + free(key); } - - SQL_EXEC("END TRANSACTION"); } static void -dbprune(const char *base) +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(base, 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"); } /* @@ -1872,10 +2173,11 @@ 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[PATH_MAX]; + int status; + pid_t child; if (nodb) return; @@ -1891,10 +2193,60 @@ dbclose(const char *base, int real) if (real) return; - strlcpy(file, MANDOC_DB, PATH_MAX); - strlcat(file, "~", PATH_MAX); - if (-1 == rename(file, MANDOC_DB)) - perror(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("", "&fork cmp"); + return; + case 0: + execlp("cmp", "cmp", "-s", + tempfilename, MANDOC_DB, (char *)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, (char *)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); + } } /* @@ -1906,132 +2258,282 @@ dbclose(const char *base, int real) * Must be matched by dbclose(). */ static int -dbopen(const char *base, int real) +dbopen(int real) { - char file[PATH_MAX]; const char *sql; int rc, ofl; - size_t sz; - if (nodb) - return(1); + if (nodb) + return 1; - sz = strlcpy(file, MANDOC_DB, PATH_MAX); - if ( ! real) - sz = strlcat(file, "~", PATH_MAX); + *tempfilename = '\0'; + ofl = SQLITE_OPEN_READWRITE; - if (sz >= PATH_MAX) { - fprintf(stderr, "%s: Path too long\n", file); - return(0); + if (real) { + rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL); + if (SQLITE_OK != rc) { + exitcode = (int)MANDOCLEVEL_SYSERR; + if (SQLITE_CANTOPEN != rc) + say(MANDOC_DB, "%s", sqlite3_errstr(rc)); + return 0; + } + goto prepare_statements; } - if ( ! real) - remove(file); + ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE; - 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; - if (SQLITE_CANTOPEN != rc) { - perror(file); - return(0); + 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(MANDOC_DB "~", "%s", sqlite3_errstr(rc)); + return 0; } - sqlite3_close(db); - db = NULL; - - if (SQLITE_OK != (rc = sqlite3_open(file, &db))) { - perror(file); - return(0); + (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX", + sizeof(tempfilename)); + if (NULL == mkdtemp(tempfilename)) { + exitcode = (int)MANDOCLEVEL_SYSERR; + 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_errstr(rc)); + 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" + " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\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)) { - perror(sqlite3_errmsg(db)); - return(0); + exitcode = (int)MANDOCLEVEL_SYSERR; + say(MANDOC_DB, "%s", sqlite3_errmsg(db)); + sqlite3_close(db); + return 0; } prepare_statements: - SQL_EXEC("PRAGMA foreign_keys = ON"); - sql = "DELETE FROM 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); + if (SQLITE_OK != sqlite3_exec(db, + "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say(MANDOC_DB, "PRAGMA foreign_keys: %s", + sqlite3_errmsg(db)); + sqlite3_close(db); + return 0; + } + + sql = "DELETE FROM mpages WHERE pageid IN " + "(SELECT pageid FROM mlinks WHERE " + "sec=? AND arch=? AND name=?)"; + sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL); + 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 = "SELECT bits FROM names where pageid = ?"; + sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL); + sql = "INSERT INTO names " + "(bits,name,pageid) VALUES (?,?,?)"; + sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL); sql = "INSERT INTO keys " - "(bits,key,docid) VALUES (?,?,?)"; + "(bits,key,pageid) VALUES (?,?,?)"; sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL); - return(1); + +#ifndef __APPLE__ + /* + * When opening a new database, we can turn off + * synchronous mode for much better performance. + */ + + if (real && SQLITE_OK != sqlite3_exec(db, + "PRAGMA synchronous = OFF", NULL, NULL, NULL)) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say(MANDOC_DB, "PRAGMA synchronous: %s", + sqlite3_errmsg(db)); + sqlite3_close(db); + return 0; + } +#endif + + return 1; } static void * -hash_halloc(size_t sz, void *arg) +hash_calloc(size_t n, size_t sz, void *arg) { - return(mandoc_calloc(sz, 1)); + return mandoc_calloc(n, sz); } static void * hash_alloc(size_t sz, void *arg) { - return(mandoc_malloc(sz)); + return mandoc_malloc(sz); } static void -hash_free(void *p, size_t sz, void *arg) +hash_free(void *p, void *arg) { free(p); } static int -path_reset(const char *cwd, int fd, const char *base) +set_basedir(const char *targetdir, int report_baddir) { + static char startdir[PATH_MAX]; + static int getcwd_status; /* 1 = ok, 2 = failure */ + static int chdir_status; /* 1 = changed directory */ + char *cp; - if (-1 == fchdir(fd)) { - perror(cwd); - return(0); - } else if (-1 == chdir(base)) { - perror(base); - return(0); + /* + * Remember the original working directory, if possible. + * This will be needed if the second or a later directory + * on the command line is given as a relative path. + * Do not error out if the current directory is not + * searchable: Maybe it won't be needed after all. + */ + if (0 == getcwd_status) { + if (NULL == getcwd(startdir, sizeof(startdir))) { + getcwd_status = 2; + (void)strlcpy(startdir, strerror(errno), + sizeof(startdir)); + } else + getcwd_status = 1; } - return(1); + + /* + * We are leaving the old base directory. + * Do not use it any longer, not even for messages. + */ + *basedir = '\0'; + + /* + * If and only if the directory was changed earlier and + * the next directory to process is given as a relative path, + * first go back, or bail out if that is impossible. + */ + if (chdir_status && '/' != *targetdir) { + if (2 == getcwd_status) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "getcwd: %s", startdir); + return 0; + } + if (-1 == chdir(startdir)) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "&chdir %s", startdir); + return 0; + } + } + + /* + * Always resolve basedir to the canonicalized absolute + * pathname and append a trailing slash, such that + * we can reliably check whether files are inside. + */ + if (NULL == realpath(targetdir, basedir)) { + if (report_baddir || errno != ENOENT) { + exitcode = (int)MANDOCLEVEL_BADARG; + say("", "&%s: realpath", targetdir); + } + return 0; + } else if (-1 == chdir(basedir)) { + if (report_baddir || errno != ENOENT) { + exitcode = (int)MANDOCLEVEL_BADARG; + say("", "&chdir"); + } + return 0; + } + chdir_status = 1; + cp = strchr(basedir, '\0'); + if ('/' != cp[-1]) { + if (cp - basedir >= PATH_MAX - 1) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "Filename too long"); + return 0; + } + *cp++ = '/'; + *cp = '\0'; + } + return 1; } static void -say(const char *dir, const char *file, const char *format, ...) +say(const char *file, const char *format, ...) { va_list ap; + int use_errno; - fprintf(stderr, "%s", dir); + if ('\0' != *basedir) + fprintf(stderr, "%s", basedir); + if ('\0' != *basedir && '\0' != *file) + fputc('/', stderr); if ('\0' != *file) - fprintf(stderr, "//%s", file); - fputs(": ", stderr); + fprintf(stderr, "%s", file); - va_start(ap, format); - vfprintf(stderr, format, ap); - va_end(ap); - - fputc('\n', stderr); + use_errno = 1; + if (NULL != format) { + switch (*format) { + case '&': + format++; + break; + case '\0': + format = NULL; + break; + default: + use_errno = 0; + break; + } + } + 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); }