=================================================================== RCS file: /cvs/mandoc/mandocdb.c,v retrieving revision 1.3 retrieving revision 1.249 diff -u -p -r1.3 -r1.249 --- mandoc/mandocdb.c 2011/07/15 09:06:23 1.3 +++ mandoc/mandocdb.c 2017/05/05 15:17:32 1.249 @@ -1,740 +1,1597 @@ -/* $Id: mandocdb.c,v 1.3 2011/07/15 09:06:23 kristaps Exp $ */ +/* $Id: mandocdb.c,v 1.249 2017/05/05 15:17:32 schwarze Exp $ */ /* - * Copyright (c) 2011 Kristaps Dzonsons + * Copyright (c) 2011, 2012 Kristaps Dzonsons + * Copyright (c) 2011-2017 Ingo Schwarze + * Copyright (c) 2016 Ed Maste * * 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 +#if HAVE_ERR +#include +#endif +#include #include -#include +#if HAVE_FTS +#include +#else +#include "compat_fts.h" +#endif +#include +#if HAVE_SANDBOX_INIT +#include +#endif +#include +#include #include #include #include #include +#include -#ifdef __linux__ -# include -#else -# include -#endif - -#include "man.h" -#include "mdoc.h" +#include "mandoc_aux.h" +#include "mandoc_ohash.h" #include "mandoc.h" +#include "roff.h" +#include "mdoc.h" +#include "man.h" +#include "manconf.h" +#include "mansearch.h" +#include "dba_array.h" +#include "dba.h" -#define MANDOC_DB "mandoc.db" -#define MANDOC_IDX "mandoc.index" -#define MANDOC_BUFSZ BUFSIZ -#define MANDOC_FLAGS O_CREAT|O_TRUNC|O_RDWR -#define MANDOC_SLOP 1024 +extern const char *const mansearch_keynames[]; -/* Bit-fields. See mandocdb.8. */ +enum op { + OP_DEFAULT = 0, /* new dbs from dir list or default config */ + OP_CONFFILE, /* new databases from custom config file */ + OP_UPDATE, /* delete/add entries in existing database */ + OP_DELETE, /* delete entries from existing database */ + OP_TEST /* change no databases, report potential problems */ +}; -#define TYPE_NAME 0x01 -#define TYPE_FUNCTION 0x02 -#define TYPE_UTILITY 0x04 -#define TYPE_INCLUDES 0x08 -#define TYPE_VARIABLE 0x10 -#define TYPE_STANDARD 0x20 -#define TYPE_AUTHOR 0x40 -#define TYPE_CONFIG 0x80 -#define TYPE_DESC 0x100 -#define TYPE_XREF 0x200 -#define TYPE_PATH 0x400 -#define TYPE_ENV 0x800 -#define TYPE_ERR 0x1000 +struct str { + const struct mpage *mpage; /* if set, the owning parse */ + uint64_t mask; /* bitmask in sequence */ + char key[]; /* rendered text */ +}; -struct of { - char *fname; - struct of *next; +struct inodev { + ino_t st_ino; + dev_t st_dev; }; -/* Buffer for storing growable data. */ +struct mpage { + struct inodev inodev; /* used for hashing routine */ + struct dba_array *dba; + 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 mpage *next; /* singly linked list */ + struct mlink *mlinks; /* singly linked list */ + int name_head_done; + enum form form; /* format from file content */ +}; -struct buf { - char *cp; - size_t len; - size_t size; +struct mlink { + char file[PATH_MAX]; /* filename rel. to manpath */ + char *dsec; /* section from directory */ + char *arch; /* architecture from directory */ + char *name; /* name from file name (not empty) */ + char *fsec; /* section from file name suffix */ + struct mlink *next; /* singly linked list */ + struct mpage *mpage; /* parent */ + int gzip; /* filename has a .gz suffix */ + enum form dform; /* format from directory */ + enum form fform; /* format from file name suffix */ }; -/* Operation we're going to perform. */ +typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *, + const struct roff_node *); -enum op { - OP_NEW = 0, /* new database */ - OP_UPDATE, /* update entries in existing database */ - OP_DELETE /* delete entries from existing database */ +struct mdoc_handler { + mdoc_fp fp; /* optional handler */ + uint64_t mask; /* set unless handler returns 0 */ + int taboo; /* node flags that must not be set */ }; -#define MAN_ARGS DB *hash, \ - struct buf *buf, \ - struct buf *dbuf, \ - const struct man_node *n -#define MDOC_ARGS DB *hash, \ - struct buf *buf, \ - struct buf *dbuf, \ - const struct mdoc_node *n, \ - const struct mdoc_meta *m -static void buf_appendmdoc(struct buf *, - const struct mdoc_node *, int); -static void buf_append(struct buf *, const char *); -static void buf_appendb(struct buf *, - const void *, size_t); -static void dbt_put(DB *, const char *, DBT *, DBT *); -static void hash_put(DB *, const struct buf *, int); -static void hash_reset(DB **); -static void index_merge(const struct of *, struct mparse *, - struct buf *, struct buf *, - DB *, DB *, const char *, - DB *, const char *, - recno_t, const recno_t *, size_t); -static void index_prune(const struct of *, DB *, - const char *, DB *, const char *, - recno_t *, recno_t **, size_t *); -static int pman_node(MAN_ARGS); -static void pmdoc_node(MDOC_ARGS); -static void pmdoc_An(MDOC_ARGS); -static void pmdoc_Cd(MDOC_ARGS); -static void pmdoc_Er(MDOC_ARGS); -static void pmdoc_Ev(MDOC_ARGS); -static void pmdoc_Fd(MDOC_ARGS); -static void pmdoc_In(MDOC_ARGS); -static void pmdoc_Fn(MDOC_ARGS); -static void pmdoc_Fo(MDOC_ARGS); -static void pmdoc_Nd(MDOC_ARGS); -static void pmdoc_Nm(MDOC_ARGS); -static void pmdoc_Pa(MDOC_ARGS); -static void pmdoc_St(MDOC_ARGS); -static void pmdoc_Vt(MDOC_ARGS); -static void pmdoc_Xr(MDOC_ARGS); -static void usage(void); +int mandocdb(int, char *[]); -typedef void (*pmdoc_nf)(MDOC_ARGS); +static void dbadd(struct dba *, struct mpage *); +static void dbadd_mlink(const struct mlink *mlink); +static void dbprune(struct dba *); +static void dbwrite(struct dba *); +static void filescan(const char *); +#if HAVE_FTS_COMPARE_CONST +static int fts_compare(const FTSENT *const *, const FTSENT *const *); +#else +static int fts_compare(const FTSENT **, const FTSENT **); +#endif +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 dba *, struct mparse *); +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_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_Va(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, int); +static int render_string(char **, size_t *); +static void say(const char *, const char *, ...) + __attribute__((__format__ (__printf__, 2, 3))); +static int set_basedir(const char *, int); +static int treescan(void); +static size_t utf8(unsigned int, char [7]); -static const pmdoc_nf mdocs[MDOC_MAX] = { - NULL, /* Ap */ - NULL, /* Dd */ - NULL, /* Dt */ - NULL, /* Os */ - NULL, /* Sh */ - NULL, /* Ss */ - NULL, /* Pp */ - NULL, /* D1 */ - NULL, /* Dl */ - NULL, /* Bd */ - NULL, /* Ed */ - NULL, /* Bl */ - NULL, /* El */ - NULL, /* It */ - NULL, /* Ad */ - pmdoc_An, /* An */ - NULL, /* Ar */ - pmdoc_Cd, /* Cd */ - NULL, /* Cm */ - NULL, /* Dv */ - pmdoc_Er, /* Er */ - pmdoc_Ev, /* Ev */ - NULL, /* Ex */ - NULL, /* Fa */ - pmdoc_Fd, /* Fd */ - NULL, /* Fl */ - pmdoc_Fn, /* Fn */ - NULL, /* Ft */ - NULL, /* Ic */ - pmdoc_In, /* In */ - NULL, /* Li */ - pmdoc_Nd, /* Nd */ - pmdoc_Nm, /* Nm */ - NULL, /* Op */ - NULL, /* Ot */ - pmdoc_Pa, /* Pa */ - NULL, /* Rv */ - pmdoc_St, /* St */ - pmdoc_Vt, /* Va */ - pmdoc_Vt, /* Vt */ - pmdoc_Xr, /* Xr */ - NULL, /* %A */ - NULL, /* %B */ - NULL, /* %D */ - NULL, /* %I */ - NULL, /* %J */ - NULL, /* %N */ - NULL, /* %O */ - NULL, /* %P */ - NULL, /* %R */ - NULL, /* %T */ - NULL, /* %V */ - NULL, /* Ac */ - NULL, /* Ao */ - NULL, /* Aq */ - NULL, /* At */ - NULL, /* Bc */ - NULL, /* Bf */ - NULL, /* Bo */ - NULL, /* Bq */ - NULL, /* Bsx */ - NULL, /* Bx */ - NULL, /* Db */ - NULL, /* Dc */ - NULL, /* Do */ - NULL, /* Dq */ - NULL, /* Ec */ - NULL, /* Ef */ - NULL, /* Em */ - NULL, /* Eo */ - NULL, /* Fx */ - NULL, /* Ms */ - NULL, /* No */ - NULL, /* Ns */ - NULL, /* Nx */ - NULL, /* Ox */ - NULL, /* Pc */ - NULL, /* Pf */ - NULL, /* Po */ - NULL, /* Pq */ - NULL, /* Qc */ - NULL, /* Ql */ - NULL, /* Qo */ - NULL, /* Qq */ - NULL, /* Re */ - NULL, /* Rs */ - NULL, /* Sc */ - NULL, /* So */ - NULL, /* Sq */ - NULL, /* Sm */ - NULL, /* Sx */ - NULL, /* Sy */ - NULL, /* Tn */ - NULL, /* Ux */ - NULL, /* Xc */ - NULL, /* Xo */ - pmdoc_Fo, /* Fo */ - NULL, /* Fc */ - NULL, /* Oo */ - NULL, /* Oc */ - NULL, /* Bk */ - NULL, /* Ek */ - NULL, /* Bt */ - NULL, /* Hf */ - NULL, /* Fr */ - NULL, /* Ud */ - NULL, /* Lb */ - NULL, /* Lp */ - NULL, /* Lk */ - NULL, /* Mt */ - NULL, /* Brq */ - NULL, /* Bro */ - NULL, /* Brc */ - NULL, /* %C */ - NULL, /* Es */ - NULL, /* En */ - NULL, /* Dx */ - NULL, /* %Q */ - NULL, /* br */ - NULL, /* sp */ - NULL, /* %U */ - NULL, /* Ta */ +static int nodb; /* no database changes */ +static int mparse_options; /* abort the parse early */ +static int use_all; /* use all found files */ +static int debug; /* print what we're doing */ +static int warnings; /* warn about crap */ +static int 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 mpage *mpage_head; /* list of distinct manual pages */ +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 uint64_t name_mask; + +static const struct mdoc_handler __mdocs[MDOC_MAX - MDOC_Dd] = { + { NULL, 0, NODE_NOPRT }, /* Dd */ + { NULL, 0, NODE_NOPRT }, /* Dt */ + { NULL, 0, NODE_NOPRT }, /* Os */ + { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */ + { parse_mdoc_head, TYPE_Ss, 0 }, /* 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, 0 }, /* An */ + { NULL, 0, 0 }, /* Ap */ + { NULL, TYPE_Ar, 0 }, /* Ar */ + { NULL, TYPE_Cd, 0 }, /* Cd */ + { NULL, TYPE_Cm, 0 }, /* Cm */ + { NULL, TYPE_Dv, 0 }, /* Dv */ + { NULL, TYPE_Er, 0 }, /* Er */ + { NULL, TYPE_Ev, 0 }, /* Ev */ + { NULL, 0, 0 }, /* Ex */ + { NULL, TYPE_Fa, 0 }, /* Fa */ + { parse_mdoc_Fd, 0, 0 }, /* Fd */ + { NULL, TYPE_Fl, 0 }, /* Fl */ + { parse_mdoc_Fn, 0, 0 }, /* Fn */ + { NULL, TYPE_Ft, 0 }, /* Ft */ + { NULL, TYPE_Ic, 0 }, /* Ic */ + { NULL, TYPE_In, 0 }, /* In */ + { NULL, TYPE_Li, 0 }, /* Li */ + { parse_mdoc_Nd, 0, 0 }, /* Nd */ + { parse_mdoc_Nm, 0, 0 }, /* Nm */ + { NULL, 0, 0 }, /* Op */ + { NULL, 0, 0 }, /* Ot */ + { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */ + { NULL, 0, 0 }, /* Rv */ + { NULL, TYPE_St, 0 }, /* St */ + { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */ + { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */ + { parse_mdoc_Xr, 0, 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, 0 }, /* At */ + { NULL, 0, 0 }, /* Bc */ + { NULL, 0, 0 }, /* Bf */ + { NULL, 0, 0 }, /* Bo */ + { NULL, 0, 0 }, /* Bq */ + { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */ + { NULL, TYPE_Bx, NODE_NOSRC }, /* 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, 0 }, /* Em */ + { NULL, 0, 0 }, /* Eo */ + { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */ + { NULL, TYPE_Ms, 0 }, /* Ms */ + { NULL, 0, 0 }, /* No */ + { NULL, 0, 0 }, /* Ns */ + { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */ + { NULL, TYPE_Ox, NODE_NOSRC }, /* 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, 0 }, /* Sy */ + { NULL, TYPE_Tn, 0 }, /* Tn */ + { NULL, 0, NODE_NOSRC }, /* Ux */ + { NULL, 0, 0 }, /* Xc */ + { NULL, 0, 0 }, /* Xo */ + { parse_mdoc_Fo, 0, 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, NODE_NOSRC }, /* Lb */ + { NULL, 0, 0 }, /* Lp */ + { NULL, TYPE_Lk, 0 }, /* Lk */ + { NULL, TYPE_Mt, NODE_NOSRC }, /* 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, NODE_NOSRC }, /* Dx */ + { NULL, 0, 0 }, /* %Q */ + { NULL, 0, 0 }, /* %U */ + { NULL, 0, 0 }, /* Ta */ }; +static const struct mdoc_handler *const mdocs = __mdocs - MDOC_Dd; -static const char *progname; int -main(int argc, char *argv[]) +mandocdb(int argc, char *argv[]) { - struct mparse *mp; /* parse sequence */ - enum op op; /* current operation */ - const char *dir; /* result dir (default: cwd) */ - char ibuf[MAXPATHLEN], /* index fname */ - fbuf[MAXPATHLEN]; /* btree fname */ - int ch, verb, i; - DB *idx, /* index database */ - *db, /* keyword database */ - *hash; /* temporary keyword hashtable */ - enum mandoclevel ec; /* exit status */ - BTREEINFO info; /* btree configuration */ - recno_t maxrec; /* supremum of all records */ - recno_t *recs; /* buffer of empty records */ - size_t recsz, /* buffer size of recs */ - reccur; /* valid number of recs */ - struct buf buf, /* keyword buffer */ - dbuf; /* description buffer */ - struct of *ofile; - extern int optind; - extern char *optarg; + struct manconf conf; + struct mparse *mp; + struct dba *dba; + const char *path_arg, *progname; + size_t j, sz; + int ch, i; - progname = strrchr(argv[0], '/'); - if (progname == NULL) - progname = argv[0]; - else - ++progname; +#if HAVE_PLEDGE + if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) { + warn("pledge"); + return (int)MANDOCLEVEL_SYSERR; + } +#endif - ofile = NULL; - dir = ""; - verb = 0; - db = idx = NULL; - mp = NULL; - hash = NULL; - recs = NULL; - recsz = reccur = 0; - maxrec = 0; - op = OP_NEW; - ec = MANDOCLEVEL_SYSERR; +#if HAVE_SANDBOX_INIT + if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) { + warnx("sandbox_init"); + return (int)MANDOCLEVEL_SYSERR; + } +#endif - memset(&buf, 0, sizeof(struct buf)); - memset(&dbuf, 0, sizeof(struct buf)); + memset(&conf, 0, sizeof(conf)); - while (-1 != (ch = getopt(argc, argv, "d:ruv"))) + /* + * 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)) { \ + warnx("-%c: Conflicting option", (_ch)); \ + goto usage; \ + } while (/*CONSTCOND*/0) + + path_arg = NULL; + op = OP_DEFAULT; + + while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v"))) switch (ch) { - case ('d'): - dir = optarg; + case 'a': + use_all = 1; break; - case ('r'): - op = OP_DELETE; + case 'C': + CHECKOP(op, ch); + path_arg = optarg; + op = OP_CONFFILE; break; - case ('u'): + case 'D': + debug++; + break; + case 'd': + CHECKOP(op, ch); + path_arg = optarg; op = OP_UPDATE; break; - case ('v'): - verb++; + case 'n': + nodb = 1; break; + case 'p': + warnings = 1; + break; + case 'Q': + mparse_options |= MPARSE_QUICK; + break; + case 'T': + if (strcmp(optarg, "utf8")) { + warnx("-T%s: Unsupported output format", + 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': + CHECKOP(op, ch); + path_arg = optarg; + op = OP_DELETE; + break; + case 'v': + /* Compatibility with espie@'s makewhatis. */ + break; default: - usage(); - return((int)MANDOCLEVEL_BADARG); + goto usage; } argc -= optind; argv += optind; - ibuf[0] = ibuf[MAXPATHLEN - 2] = - fbuf[0] = fbuf[MAXPATHLEN - 2] = '\0'; +#if HAVE_PLEDGE + if (nodb) { + if (pledge("stdio rpath", NULL) == -1) { + warn("pledge"); + return (int)MANDOCLEVEL_SYSERR; + } + } +#endif - strlcat(fbuf, dir, MAXPATHLEN); - strlcat(fbuf, MANDOC_DB, MAXPATHLEN); + if (OP_CONFFILE == op && argc > 0) { + warnx("-C: Too many arguments"); + goto usage; + } - strlcat(ibuf, dir, MAXPATHLEN); - strlcat(ibuf, MANDOC_IDX, MAXPATHLEN); + exitcode = (int)MANDOCLEVEL_OK; + mchars_alloc(); + mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL); + mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev)); + mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file)); - if ('\0' != fbuf[MAXPATHLEN - 2] || - '\0' != ibuf[MAXPATHLEN - 2]) { - fprintf(stderr, "%s: Path too long\n", dir); - goto out; + if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) { + + /* + * Most of these deal with a specific directory. + * Jump into that directory first. + */ + if (OP_TEST != op && 0 == set_basedir(path_arg, 1)) + goto out; + + dba = nodb ? dba_new(128) : dba_read(MANDOC_DB); + if (dba != NULL) { + /* + * The existing database is usable. Process + * all files specified on the command-line. + */ +#if HAVE_PLEDGE + if (!nodb) { + if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) { + warn("pledge"); + exitcode = (int)MANDOCLEVEL_SYSERR; + goto out; + } + } +#endif + use_all = 1; + for (i = 0; i < argc; i++) + filescan(argv[i]); + if (nodb == 0) + dbprune(dba); + } else { + /* Database missing or corrupt. */ + if (op != OP_UPDATE || errno != ENOENT) + say(MANDOC_DB, "%s: Automatically recreating" + " from scratch", strerror(errno)); + exitcode = (int)MANDOCLEVEL_OK; + op = OP_DEFAULT; + if (0 == treescan()) + goto out; + dba = dba_new(128); + } + if (OP_DELETE != op) + mpages_merge(dba, mp); + if (nodb == 0) + dbwrite(dba); + dba_free(dba); + } else { + /* + * If we have arguments, use them as our manpaths. + * If we don't, use man.conf(5). + */ + if (argc > 0) { + conf.manpath.paths = mandoc_reallocarray(NULL, + argc, sizeof(char *)); + conf.manpath.sz = (size_t)argc; + for (i = 0; i < argc; i++) + conf.manpath.paths[i] = mandoc_strdup(argv[i]); + } else + 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 + * build a new database and finally move it into place. + * Ignore zero-length directories and strip trailing + * slashes. + */ + 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 (j) { + mandoc_ohash_init(&mpages, 6, + offsetof(struct mpage, inodev)); + mandoc_ohash_init(&mlinks, 6, + offsetof(struct mlink, file)); + } + + if ( ! set_basedir(conf.manpath.paths[j], argc > 0)) + continue; + if (0 == treescan()) + continue; + dba = dba_new(128); + mpages_merge(dba, mp); + if (nodb == 0) + dbwrite(dba); + dba_free(dba); + + if (j + 1 < conf.manpath.sz) { + mpages_free(); + ohash_delete(&mpages); + ohash_delete(&mlinks); + } + } } +out: + manconf_free(&conf); + mparse_free(mp); + mchars_free(); + mpages_free(); + ohash_delete(&mpages); + ohash_delete(&mlinks); + return exitcode; +usage: + progname = getprogname(); + fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n" + " %s [-aDnpQ] [-Tutf8] dir ...\n" + " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n" + " %s [-Dnp] -u dir [file ...]\n" + " %s [-Q] -t file ...\n", + progname, progname, progname, progname, progname); + return (int)MANDOCLEVEL_BADARG; +} + +/* + * To get a singly linked list in alpha order while inserting entries + * at the beginning, process directory entries in reverse alpha order. + */ +static int +#if HAVE_FTS_COMPARE_CONST +fts_compare(const FTSENT *const *a, const FTSENT *const *b) +#else +fts_compare(const FTSENT **a, const FTSENT **b) +#endif +{ + return -strcmp((*a)->fts_name, (*b)->fts_name); +} + +/* + * 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*[/]/.
+ * or + * [./]cat
[/]/.0 + * + * TODO: accommodate for multi-language directories. + */ +static int +treescan(void) +{ + char buf[PATH_MAX]; + FTS *f; + FTSENT *ff; + struct mlink *mlink; + int gzip; + enum form dform; + char *dsec, *arch, *fsec, *cp; + const char *path; + const char *argv[2]; + + argv[0] = "."; + argv[1] = NULL; + + f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR, + fts_compare); + if (f == NULL) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "&fts_open"); + return 0; + } + + dsec = arch = NULL; + dform = FORM_NONE; + + while ((ff = fts_read(f)) != NULL) { + path = ff->fts_path + 2; + switch (ff->fts_info) { + + /* + * Symbolic links require various sanity checks, + * then get handled just like regular files. + */ + case FTS_SL: + if (realpath(path, buf) == NULL) { + if (warnings) + say(path, "&realpath"); + continue; + } + 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 (stat(path, ff->fts_statp) == -1) { + if (warnings) + say(path, "&stat"); + continue; + } + /* FALLTHROUGH */ + + /* + * If we're a regular file, add an mlink by using the + * stored directory data and handling the filename. + */ + case FTS_F: + if ( ! strcmp(path, MANDOC_DB)) + continue; + if ( ! use_all && ff->fts_level < 2) { + if (warnings) + say(path, "Extraneous file"); + continue; + } + gzip = 0; + fsec = NULL; + while (fsec == NULL) { + fsec = strrchr(ff->fts_name, '.'); + if (fsec == NULL || strcmp(fsec+1, "gz")) + break; + gzip = 1; + *fsec = '\0'; + fsec = NULL; + } + if (fsec == NULL) { + if ( ! use_all) { + if (warnings) + say(path, + "No filename suffix"); + continue; + } + } else if ( ! strcmp(++fsec, "html")) { + if (warnings) + say(path, "Skip html"); + continue; + } else if ( ! strcmp(fsec, "ps")) { + if (warnings) + say(path, "Skip ps"); + continue; + } else if ( ! strcmp(fsec, "pdf")) { + if (warnings) + say(path, "Skip pdf"); + continue; + } else if ( ! use_all && + ((dform == FORM_SRC && + strncmp(fsec, dsec, strlen(dsec))) || + (dform == FORM_CAT && strcmp(fsec, "0")))) { + if (warnings) + say(path, "Wrong filename suffix"); + continue; + } else + fsec[-1] = '\0'; + + mlink = mandoc_calloc(1, sizeof(struct mlink)); + if (strlcpy(mlink->file, path, + sizeof(mlink->file)) >= + sizeof(mlink->file)) { + say(path, "Filename too long"); + free(mlink); + continue; + } + mlink->dform = dform; + mlink->dsec = dsec; + mlink->arch = arch; + mlink->name = ff->fts_name; + mlink->fsec = fsec; + mlink->gzip = gzip; + mlink_add(mlink, ff->fts_statp); + continue; + + case FTS_D: + case FTS_DP: + break; + + default: + if (warnings) + say(path, "Not a regular file"); + continue; + } + + switch (ff->fts_level) { + case 0: + /* Ignore the root directory. */ + break; + case 1: + /* + * This might contain manX/ or catX/. + * Try to infer this from the name. + * If we're not in use_all, enforce it. + */ + cp = ff->fts_name; + if (ff->fts_info == FTS_DP) { + dform = FORM_NONE; + dsec = NULL; + break; + } + + if ( ! strncmp(cp, "man", 3)) { + dform = FORM_SRC; + dsec = cp + 3; + } else if ( ! strncmp(cp, "cat", 3)) { + dform = FORM_CAT; + dsec = cp + 3; + } else { + dform = FORM_NONE; + dsec = NULL; + } + + if (dsec != NULL || use_all) + break; + + if (warnings) + say(path, "Unknown directory part"); + fts_set(f, ff, FTS_SKIP); + break; + case 2: + /* + * Possibly our architecture. + * If we're descending, keep tabs on it. + */ + if (ff->fts_info != FTS_DP && dsec != NULL) + arch = ff->fts_name; + else + arch = NULL; + break; + default: + if (ff->fts_info == FTS_DP || use_all) + break; + if (warnings) + say(path, "Extraneous directory part"); + fts_set(f, ff, FTS_SKIP); + break; + } + } + + fts_close(f); + return 1; +} + +/* + * 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*[/]/.
+ * or + * [./]cat
[/]/.0 + * + * See treescan() for the fts(3) version of this. + */ +static void +filescan(const char *file) +{ + char buf[PATH_MAX]; + struct stat st; + struct mlink *mlink; + char *p, *start; + + assert(use_all); + + if (0 == strncmp(file, "./", 2)) + file += 2; + /* - * For the keyword database, open a BTREE database that allows - * duplicates. - * For the index database, use a standard RECNO database type. - * Truncate the database if we're creating a new one. + * 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 (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) { + exitcode = (int)MANDOCLEVEL_BADARG; + say(file, "Not a regular file"); + return; + } - memset(&info, 0, sizeof(BTREEINFO)); - info.flags = R_DUP; + /* + * 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; + } - if (OP_NEW == op) { - db = dbopen(fbuf, MANDOC_FLAGS, 0644, DB_BTREE, &info); - idx = dbopen(ibuf, MANDOC_FLAGS, 0644, DB_RECNO, NULL); - } else { - db = dbopen(fbuf, O_CREAT|O_RDWR, 0644, DB_BTREE, &info); - idx = dbopen(ibuf, O_CREAT|O_RDWR, 0644, DB_RECNO, NULL); + 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; } - if (NULL == db) { - perror(fbuf); - goto out; - } else if (NULL == db) { - perror(ibuf); - goto out; + /* + * 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); } - ofile = mandoc_calloc(argc, sizeof(struct of)); - for (i = 0; i < argc; i++) { - ofile[i].next = &ofile[i + 1]; - ofile[i].fname = argv[i]; + 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; } - ofile[argc - 1].next = NULL; + /* + * In test mode or when the original name is absolute + * but outside our tree, guess the base directory. + */ + if (op == OP_TEST || (start == buf && *start == '/')) { + if (strncmp(buf, "man/", 4) == 0) + start = buf + 4; + else if ((start = strstr(buf, "/man/")) != NULL) + start += 5; + else + start = buf; + } + /* - * If we're going to delete or update a database, remove the - * entries now (both the index and all keywords pointing to it). - * This doesn't actually remove them: it only sets their record - * value lengths to zero. - * While doing so, add the empty records to a list we'll access - * later in re-adding entries to the database. + * 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, + * assume it's an architecture. */ + if (NULL != (p = strchr(start, '/'))) { + *p++ = '\0'; + if (0 == strncmp(start, "man", 3)) { + mlink->dform = FORM_SRC; + mlink->dsec = start + 3; + } else if (0 == strncmp(start, "cat", 3)) { + mlink->dform = FORM_CAT; + mlink->dsec = start + 3; + } - if (OP_DELETE == op || OP_UPDATE == op) - index_prune(ofile, db, fbuf, idx, ibuf, - &maxrec, &recs, &recsz); + start = p; + if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) { + *p++ = '\0'; + mlink->arch = start; + start = p; + } + } - if (OP_DELETE == op) { - ec = MANDOCLEVEL_OK; - goto out; + /* + * Now check the file suffix. + * Suffix of `.0' indicates a catpage, `.1-9' is a manpage. + */ + p = strrchr(start, '\0'); + while (p-- > start && '/' != *p && '.' != *p) + /* Loop. */ ; + + if ('.' == *p) { + *p++ = '\0'; + mlink->fsec = p; } /* - * Add records to the database. - * Try parsing each manual given on the command line. - * If we fail, then emit an error and keep on going. - * Take resulting trees and push them down into the database code. - * Use the auto-parser and don't report any errors. + * Now try to parse the name. + * Use the filename portion of the path. */ + mlink->name = start; + if (NULL != (p = strrchr(start, '/'))) { + mlink->name = p + 1; + *p = '\0'; + } + mlink_add(mlink, &st); +} - mp = mparse_alloc(MPARSE_AUTO, MANDOCLEVEL_FATAL, NULL, NULL); +static void +mlink_add(struct mlink *mlink, const struct stat *st) +{ + struct inodev inodev; + struct mpage *mpage; + unsigned int slot; - buf.size = dbuf.size = MANDOC_BUFSZ; - buf.cp = mandoc_malloc(buf.size); - dbuf.cp = mandoc_malloc(dbuf.size); + assert(NULL != mlink->file); - index_merge(ofile, mp, &dbuf, &buf, hash, db, - fbuf, idx, ibuf, maxrec, recs, reccur); + 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 : ""); - ec = MANDOCLEVEL_OK; -out: - if (db) - (*db->close)(db); - if (idx) - (*idx->close)(idx); - if (hash) - (*hash->close)(hash); - if (mp) - mparse_free(mp); + 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; - free(ofile); - free(buf.cp); - free(dbuf.cp); - free(recs); + slot = ohash_qlookup(&mlinks, mlink->file); + assert(NULL == ohash_find(&mlinks, slot)); + ohash_insert(&mlinks, slot, mlink); - return((int)ec); + 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; + mpage->form = FORM_NONE; + mpage->next = mpage_head; + mpage_head = mpage; + ohash_insert(&mpages, slot, mpage); + } else + mlink->next = mpage->mlinks; + mpage->mlinks = mlink; + mlink->mpage = mpage; } -void -index_merge(const struct of *of, struct mparse *mp, - struct buf *dbuf, struct buf *buf, - DB *hash, DB *db, const char *dbf, - DB *idx, const char *idxf, - recno_t maxrec, const recno_t *recs, size_t reccur) +static void +mlink_free(struct mlink *mlink) { - recno_t rec; - int ch; - DBT key, val; - struct mdoc *mdoc; - struct man *man; - const char *fn, *msec, *mtitle, *arch; - size_t sv; - unsigned seq; - char vbuf[8]; - for (rec = 0; of; of = of->next) { - fn = of->fname; - if (reccur > 0) { - --reccur; - rec = recs[(int)reccur]; - } else if (maxrec > 0) { - rec = maxrec; - maxrec = 0; - } else - rec++; + free(mlink->dsec); + free(mlink->arch); + free(mlink->name); + free(mlink->fsec); + free(mlink); +} - mparse_reset(mp); - hash_reset(&hash); +static void +mpages_free(void) +{ + struct mpage *mpage; + struct mlink *mlink; - if (mparse_readfd(mp, -1, fn) >= MANDOCLEVEL_FATAL) { - fprintf(stderr, "%s: Parse failure\n", fn); - continue; + while ((mpage = mpage_head) != NULL) { + while ((mlink = mpage->mlinks) != NULL) { + mpage->mlinks = mlink->next; + mlink_free(mlink); } + mpage_head = mpage->next; + free(mpage->sec); + free(mpage->arch); + free(mpage->title); + free(mpage->desc); + free(mpage); + } +} - mparse_result(mp, &mdoc, &man); - if (NULL == mdoc && NULL == man) - continue; +/* + * 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 +mlinks_undupe(struct mpage *mpage) +{ + char buf[PATH_MAX]; + struct mlink **prev; + struct mlink *mlink; + char *bufp; - msec = NULL != mdoc ? - mdoc_meta(mdoc)->msec : man_meta(man)->msec; - mtitle = NULL != mdoc ? - mdoc_meta(mdoc)->title : man_meta(man)->title; - arch = NULL != mdoc ? - mdoc_meta(mdoc)->arch : NULL; + 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; + } +} - if (NULL == arch) - arch = ""; +static void +mlink_check(struct mpage *mpage, struct mlink *mlink) +{ + struct str *str; + unsigned int slot; - /* - * The index record value consists of a nil-terminated - * filename, a nil-terminated manual section, and a - * nil-terminated description. Since the description - * may not be set, we set a sentinel to see if we're - * going to write a nil byte in its place. - */ + /* + * 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). + */ - dbuf->len = 0; - buf_appendb(dbuf, fn, strlen(fn) + 1); - buf_appendb(dbuf, msec, strlen(msec) + 1); - buf_appendb(dbuf, mtitle, strlen(mtitle) + 1); - buf_appendb(dbuf, arch, strlen(arch) + 1); + if (FORM_SRC == mpage->form && + strcasecmp(mpage->sec, mlink->dsec)) + say(mlink->file, "Section \"%s\" manual in %s directory", + mpage->sec, mlink->dsec); - sv = dbuf->len; + /* + * 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, and sparc64. + */ - /* Fix the record number in the btree value. */ + if (strcasecmp(mpage->arch, mlink->arch)) + say(mlink->file, "Architecture \"%s\" manual in " + "\"%s\" directory", mpage->arch, mlink->arch); - if (mdoc) - pmdoc_node(hash, buf, dbuf, - mdoc_node(mdoc), mdoc_meta(mdoc)); - else - pman_node(hash, buf, dbuf, man_node(man)); + /* + * XXX + * parse_cat() doesn't set NAME_TITLE yet. + */ - /* - * Copy from the in-memory hashtable of pending keywords - * into the database. - */ - - memset(vbuf, 0, sizeof(uint32_t)); - memcpy(vbuf + 4, &rec, sizeof(uint32_t)); + if (FORM_CAT == mpage->form) + return; - seq = R_FIRST; - while (0 == (ch = (*hash->seq)(hash, &key, &val, seq))) { - seq = R_NEXT; + /* + * Check whether this mlink + * appears as a name in the NAME section. + */ - memcpy(vbuf, val.data, sizeof(uint32_t)); - val.size = sizeof(vbuf); - val.data = vbuf; + 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"); +} - printf("%s: Added keyword: %s\n", - fn, (char *)key.data); - dbt_put(db, dbf, &key, &val); +/* + * 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 +mpages_merge(struct dba *dba, struct mparse *mp) +{ + struct mpage *mpage, *mpage_dest; + struct mlink *mlink, *mlink_dest; + struct roff_man *man; + char *sodest; + char *cp; + int fd; + + for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) { + mlinks_undupe(mpage); + if ((mlink = mpage->mlinks) == NULL) + continue; + + name_mask = NAME_MASK; + mandoc_ohash_init(&names, 4, offsetof(struct str, key)); + mandoc_ohash_init(&strings, 6, offsetof(struct str, key)); + mparse_reset(mp); + man = NULL; + sodest = NULL; + + if ((fd = mparse_open(mp, mlink->file)) == -1) { + say(mlink->file, "&open"); + goto nextpage; } - if (ch < 0) { - perror("hash"); - exit((int)MANDOCLEVEL_SYSERR); - } - + /* - * Apply to the index. If we haven't had a description - * set, put an empty one in now. + * Interpret the file as mdoc(7) or man(7) source + * code, unless it is known to be formatted. */ + if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) { + mparse_readfd(mp, fd, mlink->file); + close(fd); + fd = -1; + mparse_result(mp, &man, &sodest); + } - if (dbuf->len == sv) - buf_appendb(dbuf, "", 1); + 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) { - key.data = &rec; - key.size = sizeof(recno_t); + /* The .so target exists. */ - val.data = dbuf->cp; - val.size = dbuf->len; + mpage_dest = mlink_dest->mpage; + while (1) { + mlink->mpage = mpage_dest; - printf("%s: Added index\n", fn); - dbt_put(idx, idxf, &key, &val); + /* + * If the target was already + * processed, add the links + * to the database now. + * Otherwise, this will + * happen when we come + * to the target. + */ + + if (mpage_dest->dba != NULL) + dbadd_mlink(mlink); + + 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) { + mdoc_validate(man); + 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) { + man_validate(man); + if (*man->meta.msec != '\0' || + *man->meta.title != '\0') { + 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 + man = NULL; + } + + assert(mpage->desc == NULL); + if (man == NULL) { + mpage->form = FORM_CAT; + mpage->sec = mandoc_strdup(mlink->dsec); + mpage->arch = mandoc_strdup(mlink->arch); + mpage->title = mandoc_strdup(mlink->name); + parse_cat(mpage, fd); + } else if (man->macroset == MACROSET_MDOC) + parse_mdoc(mpage, &man->meta, man->first); + else + parse_man(mpage, &man->meta, man->first); + if (mpage->desc == NULL) { + mpage->desc = mandoc_strdup(mlink->name); + if (warnings) + say(mlink->file, "No one-line description, " + "using filename \"%s\"", mlink->name); + } + + for (mlink = mpage->mlinks; + mlink != NULL; + mlink = mlink->next) { + putkey(mpage, mlink->name, NAME_FILE); + if (warnings && !use_all) + mlink_check(mpage, mlink); + } + + dbadd(dba, mpage); + +nextpage: + ohash_delete(&strings); + ohash_delete(&names); } } -/* - * Scan through all entries in the index file `idx' and prune those - * entries in `ofile'. - * Pruning consists of removing from `db', then invalidating the entry - * in `idx' (zeroing its value size). - */ static void -index_prune(const struct of *ofile, DB *db, const char *dbf, - DB *idx, const char *idxf, - recno_t *maxrec, recno_t **recs, size_t *recsz) +parse_cat(struct mpage *mpage, int fd) { - const struct of *of; - const char *fn; - unsigned seq, sseq; - DBT key, val; - size_t reccur; - int ch; + FILE *stream; + struct mlink *mlink; + char *line, *p, *title, *sec; + size_t linesz, plen, titlesz; + ssize_t len; + int offs; - reccur = 0; - seq = R_FIRST; - while (0 == (ch = (*idx->seq)(idx, &key, &val, seq))) { - seq = R_NEXT; - *maxrec = *(recno_t *)key.data; - if (0 == val.size) { - if (reccur >= *recsz) { - *recsz += MANDOC_SLOP; - *recs = mandoc_realloc(*recs, - *recsz * sizeof(recno_t)); - } - (*recs)[(int)reccur] = *maxrec; - reccur++; + mlink = mpage->mlinks; + stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r"); + if (stream == NULL) { + if (fd != -1) + close(fd); + if (warnings) + say(mlink->file, "&fopen"); + return; + } + + line = NULL; + linesz = 0; + + /* Parse the section number from the header line. */ + + while (getline(&line, &linesz, stream) != -1) { + if (*line == '\n') continue; - } + if ((sec = strchr(line, '(')) == NULL) + break; + if ((p = strchr(++sec, ')')) == NULL) + break; + free(mpage->sec); + mpage->sec = mandoc_strndup(sec, p - sec); + if (warnings && *mlink->dsec != '\0' && + strcasecmp(mpage->sec, mlink->dsec)) + say(mlink->file, + "Section \"%s\" manual in %s directory", + mpage->sec, mlink->dsec); + break; + } - fn = (char *)val.data; - for (of = ofile; of; of = of->next) - if (0 == strcmp(fn, of->fname)) - break; + /* Skip to first blank line. */ - if (NULL == of) + while (line == NULL || *line != '\n') + if (getline(&line, &linesz, stream) == -1) + break; + + /* + * Assume the first line that is not indented + * is the first section header. Skip to it. + */ + + while (getline(&line, &linesz, stream) != -1) + if (*line != '\n' && *line != ' ') + break; + + /* + * Read up until the next section into a buffer. + * Strip the leading and trailing newline from each read line, + * appending a trailing space. + * Ignore empty (whitespace-only) lines. + */ + + titlesz = 0; + title = NULL; + + while ((len = getline(&line, &linesz, stream)) != -1) { + if (*line != ' ') + break; + offs = 0; + while (isspace((unsigned char)line[offs])) + offs++; + if (line[offs] == '\0') continue; + title = mandoc_realloc(title, titlesz + len - offs); + memcpy(title + titlesz, line + offs, len - offs); + titlesz += len - offs; + title[titlesz - 1] = ' '; + } + free(line); - sseq = R_FIRST; - while (0 == (ch = (*db->seq)(db, &key, &val, sseq))) { - sseq = R_NEXT; - assert(8 == val.size); - if (*maxrec != *(recno_t *)(val.data + 4)) - continue; - printf("%s: Deleted keyword: %s\n", - fn, (char *)key.data); - ch = (*db->del)(db, &key, R_CURSOR); - if (ch < 0) - break; - } - if (ch < 0) { - perror(dbf); - exit((int)MANDOCLEVEL_SYSERR); - } + /* + * If no page content can be found, or the input line + * is already the next section header, or there is no + * trailing newline, reuse the page title as the page + * description. + */ - printf("%s: Deleted index\n", fn); + if (NULL == title || '\0' == *title) { + if (warnings) + say(mlink->file, "Cannot find NAME section"); + fclose(stream); + free(title); + return; + } - val.size = 0; - ch = (*idx->put)(idx, &key, &val, R_CURSOR); - if (ch < 0) { - perror(idxf); - exit((int)MANDOCLEVEL_SYSERR); - } + title[titlesz - 1] = '\0'; - if (reccur >= *recsz) { - *recsz += MANDOC_SLOP; - *recs = mandoc_realloc - (*recs, *recsz * sizeof(recno_t)); - } + /* + * Skip to the first dash. + * Use the remaining line as the description (no more than 70 + * bytes). + */ - (*recs)[(int)reccur] = *maxrec; - reccur++; + if (NULL != (p = strstr(title, "- "))) { + for (p += 2; ' ' == *p || '\b' == *p; p++) + /* Skip to next word. */ ; + } else { + if (warnings) + say(mlink->file, "No dash in title line, " + "reusing \"%s\" as one-line description", title); + p = title; } - (*maxrec)++; -} -/* - * Grow the buffer (if necessary) and copy in a binary string. - */ -static void -buf_appendb(struct buf *buf, const void *cp, size_t sz) -{ + plen = strlen(p); - /* Overshoot by MANDOC_BUFSZ. */ + /* Strip backspace-encoding from line. */ - while (buf->len + sz >= buf->size) { - buf->size = buf->len + sz + MANDOC_BUFSZ; - buf->cp = mandoc_realloc(buf->cp, buf->size); + while (NULL != (line = memchr(p, '\b', plen))) { + len = line - p; + if (0 == len) { + memmove(line, line + 1, plen--); + continue; + } + memmove(line - 1, line + 1, plen - len); + plen -= 2; } - memcpy(buf->cp + (int)buf->len, cp, sz); - buf->len += sz; + mpage->desc = mandoc_strdup(p); + fclose(stream); + free(title); } /* - * Append a nil-terminated string to the buffer. - * This can be invoked multiple times. - * The buffer string will be nil-terminated. - * If invoked multiple times, a space is put between strings. + * Put a type/word pair into the word database for this particular file. */ static void -buf_append(struct buf *buf, const char *cp) +putkey(const struct mpage *mpage, char *value, uint64_t type) { - size_t sz; - - if (0 == (sz = strlen(cp))) - return; - - if (buf->len) - buf->cp[(int)buf->len - 1] = ' '; - - buf_appendb(buf, cp, sz + 1); + putkeys(mpage, value, strlen(value), type); } /* - * Recursively add all text from a given node. - * This is optimised for general mdoc nodes in this context, which do - * not consist of subexpressions and having a recursive call for n->next - * would be wasteful. - * The "f" variable should be 0 unless called from pmdoc_Nd for the - * description buffer, which does not start at the beginning of the - * buffer. + * Grok all nodes at or below a certain mdoc node into putkey(). */ static void -buf_appendmdoc(struct buf *buf, const struct mdoc_node *n, int f) +putmdockey(const struct mpage *mpage, + const struct roff_node *n, uint64_t m, int taboo) { - for ( ; n; n = n->next) { - if (n->child) - buf_appendmdoc(buf, n->child, f); - - if (MDOC_TEXT == n->type && f) { - f = 0; - buf_appendb(buf, n->string, - strlen(n->string) + 1); - } else if (MDOC_TEXT == n->type) - buf_append(buf, n->string); - + for ( ; NULL != n; n = n->next) { + if (n->flags & taboo) + continue; + if (NULL != n->child) + putmdockey(mpage, n->child, m, taboo); + if (n->type == ROFFT_TEXT) + putkey(mpage, n->string, m); } } -/* ARGSUSED */ static void -pmdoc_An(MDOC_ARGS) +parse_man(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - - if (SEC_AUTHORS != n->sec) + const struct roff_node *head, *body; + char *start, *title; + char byte; + size_t sz; + + if (n == NULL) return; - buf_appendmdoc(buf, n->child, 0); - hash_put(hash, buf, TYPE_AUTHOR); + /* + * We're only searching for one thing: the first text child in + * the BODY of a NAME section. Since we don't keep track of + * sections in -man, run some hoops to find out whether we're in + * the correct section or not. + */ + + if (n->type == ROFFT_BODY && n->tok == MAN_SH) { + body = n; + if ((head = body->parent->head) != NULL && + (head = head->child) != NULL && + head->next == NULL && + head->type == ROFFT_TEXT && + strcmp(head->string, "NAME") == 0 && + body->child != NULL) { + + /* + * Suck the entire NAME section into memory. + * Yes, we might run away. + * But too many manuals have big, spread-out + * NAME sections over many lines. + */ + + title = NULL; + deroff(&title, body); + if (NULL == title) + return; + + /* + * Go through a special heuristic dance here. + * Conventionally, one or more manual names are + * comma-specified prior to a whitespace, then a + * dash, then a description. Try to puzzle out + * the name parts here. + */ + + start = title; + for ( ;; ) { + sz = strcspn(start, " ,"); + if ('\0' == start[sz]) + break; + + byte = start[sz]; + start[sz] = '\0'; + + /* + * 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; + } + + assert(',' == byte); + start += sz + 1; + while (' ' == *start) + start++; + } + + 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; + } + + while (isspace((unsigned char)*start)) + start++; + + if (0 == strncmp(start, "-", 1)) + start += 1; + else if (0 == strncmp(start, "\\-\\-", 4)) + start += 4; + else if (0 == strncmp(start, "\\-", 2)) + start += 2; + else if (0 == strncmp(start, "\\(en", 4)) + start += 4; + else if (0 == strncmp(start, "\\(em", 4)) + start += 4; + + while (' ' == *start) + start++; + + mpage->desc = mandoc_strdup(start); + free(title); + return; + } + } + + for (n = n->child; n; n = n->next) { + if (NULL != mpage->desc) + break; + parse_man(mpage, meta, n); + } } static void -hash_reset(DB **db) +parse_mdoc(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - DB *hash; - if (NULL != (hash = *db)) - (*hash->close)(hash); - - *db = dbopen(NULL, MANDOC_FLAGS, 0644, DB_HASH, NULL); - if (NULL == *db) { - perror("hash"); - exit((int)MANDOCLEVEL_SYSERR); + for (n = n->child; n != NULL; n = n->next) { + if (n->tok == TOKEN_NONE || + n->tok < ROFF_MAX || + n->flags & mdocs[n->tok].taboo) + continue; + assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX); + switch (n->type) { + case ROFFT_ELEM: + case ROFFT_BLOCK: + case ROFFT_HEAD: + case ROFFT_BODY: + case ROFFT_TAIL: + if (mdocs[n->tok].fp != NULL && + (*mdocs[n->tok].fp)(mpage, meta, n) == 0) + break; + if (mdocs[n->tok].mask) + putmdockey(mpage, n->child, + mdocs[n->tok].mask, mdocs[n->tok].taboo); + break; + default: + continue; + } + if (NULL != n->child) + parse_mdoc(mpage, meta, n); } } -/* ARGSUSED */ -static void -pmdoc_Fd(MDOC_ARGS) +static int +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) - return; - if (NULL == (n = n->child) || MDOC_TEXT != n->type) - return; + if (SEC_SYNOPSIS != n->sec || + NULL == (n = n->child) || + n->type != ROFFT_TEXT) + return 0; + /* * Only consider those `Fd' macro fields that begin with an * "inclusion" token (versus, e.g., #define). */ + if (strcmp("#include", n->string)) - return; + return 0; - if (NULL == (n = n->next) || MDOC_TEXT != n->type) - return; + if ((n = n->next) == NULL || n->type != ROFFT_TEXT) + return 0; /* * Strip away the enclosing angle brackets and make sure we're @@ -746,408 +1603,713 @@ pmdoc_Fd(MDOC_ARGS) start++; if (0 == (sz = strlen(start))) - return; + return 0; end = &start[(int)sz - 1]; if ('>' == *end || '"' == *end) end--; - assert(end >= start); - - buf_appendb(buf, start, (size_t)(end - start + 1)); - buf_appendb(buf, "", 1); - - hash_put(hash, buf, TYPE_INCLUDES); + if (end > start) + putkeys(mpage, start, end - start + 1, TYPE_In); + return 0; } -/* ARGSUSED */ static void -pmdoc_Cd(MDOC_ARGS) +parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n) { - - if (SEC_SYNOPSIS != n->sec) + char *cp; + size_t sz; + + if (n->type != ROFFT_TEXT) return; - buf_appendmdoc(buf, n->child, 0); - hash_put(hash, buf, TYPE_CONFIG); -} + /* Skip function pointer punctuation. */ -/* ARGSUSED */ -static void -pmdoc_In(MDOC_ARGS) -{ - - if (SEC_SYNOPSIS != n->sec) - return; - if (NULL == n->child || MDOC_TEXT != n->child->type) - return; + cp = n->string; + while (*cp == '(' || *cp == '*') + cp++; + sz = strcspn(cp, "()"); - buf_append(buf, n->child->string); - hash_put(hash, buf, TYPE_INCLUDES); + putkeys(mpage, cp, sz, TYPE_Fn); + if (n->sec == SEC_SYNOPSIS) + putkeys(mpage, cp, sz, NAME_SYN); } -/* ARGSUSED */ -static void -pmdoc_Fn(MDOC_ARGS) +static int +parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - const char *cp; - - if (SEC_SYNOPSIS != n->sec) - return; - if (NULL == n->child || MDOC_TEXT != n->child->type) - return; - /* .Fn "struct type *arg" "foo" */ + if (n->child == NULL) + return 0; - cp = strrchr(n->child->string, ' '); - if (NULL == cp) - cp = n->child->string; + parse_mdoc_fname(mpage, n->child); - /* Strip away pointer symbol. */ + for (n = n->child->next; n != NULL; n = n->next) + if (n->type == ROFFT_TEXT) + putkey(mpage, n->string, TYPE_Fa); - while ('*' == *cp) - cp++; - - buf_append(buf, cp); - hash_put(hash, buf, TYPE_FUNCTION); + return 0; } -/* ARGSUSED */ -static void -pmdoc_St(MDOC_ARGS) +static int +parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - - if (SEC_STANDARDS != n->sec) - return; - if (NULL == n->child || MDOC_TEXT != n->child->type) - return; - buf_append(buf, n->child->string); - hash_put(hash, buf, TYPE_STANDARD); + if (n->type != ROFFT_HEAD) + return 1; + + if (n->child != NULL) + parse_mdoc_fname(mpage, n->child); + + return 0; } -/* ARGSUSED */ -static void -pmdoc_Xr(MDOC_ARGS) +static int +parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { + char *cp; - if (NULL == (n = n->child)) - return; + if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY) + return 0; - buf_appendb(buf, n->string, strlen(n->string)); + if (n->child != NULL && + n->child->next == NULL && + n->child->type == ROFFT_TEXT) + return 1; - if (NULL != (n = n->next)) { - buf_appendb(buf, ".", 1); - buf_appendb(buf, n->string, strlen(n->string) + 1); - } else - buf_appendb(buf, ".", 2); + cp = NULL; + deroff(&cp, n); + if (cp != NULL) { + putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va || + n->type == ROFFT_BODY ? TYPE_Va : 0)); + free(cp); + } - hash_put(hash, buf, TYPE_XREF); + return 0; } -/* ARGSUSED */ -static void -pmdoc_Vt(MDOC_ARGS) +static int +parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - const char *start; - size_t sz; - - if (SEC_SYNOPSIS != n->sec) - return; - if (MDOC_Vt == n->tok && MDOC_BODY != n->type) - return; - if (NULL == n->last || MDOC_TEXT != n->last->type) - return; + char *cp; - /* - * Strip away leading pointer symbol '*' and trailing ';'. - */ + if (NULL == (n = n->child)) + return 0; - start = n->last->string; + if (NULL == n->next) { + putkey(mpage, n->string, TYPE_Xr); + return 0; + } - while ('*' == *start) - start++; + mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string); + putkey(mpage, cp, TYPE_Xr); + free(cp); + return 0; +} - if (0 == (sz = strlen(start))) - return; +static int +parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) +{ - if (';' == start[(int)sz - 1]) - sz--; + if (n->type == ROFFT_BODY) + deroff(&mpage->desc, n); + return 0; +} - if (0 == sz) - return; +static int +parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) +{ - buf_appendb(buf, start, sz); - buf_appendb(buf, "", 1); - hash_put(hash, buf, TYPE_VARIABLE); + if (SEC_NAME == n->sec) + putmdockey(mpage, n->child, NAME_TITLE, 0); + 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, 0); + } + if ( ! (mpage->name_head_done || + n->child == NULL || n->child->string == NULL || + strcasecmp(n->child->string, meta->title))) { + putkey(mpage, n->child->string, NAME_HEAD); + mpage->name_head_done = 1; + } + return 0; } -/* ARGSUSED */ -static void -pmdoc_Fo(MDOC_ARGS) +static int +parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) { - - if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type) - return; - if (NULL == n->child || MDOC_TEXT != n->child->type) - return; - buf_append(buf, n->child->string); - hash_put(hash, buf, TYPE_FUNCTION); + return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD; } +static int +parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta, + const struct roff_node *n) +{ -/* ARGSUSED */ + return n->type == ROFFT_HEAD; +} + +/* + * 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 void -pmdoc_Nd(MDOC_ARGS) +putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v) { + struct ohash *htab; + struct str *s; + const char *end; + unsigned int slot; + int i, mustfree; - if (MDOC_BODY != n->type) + if (0 == sz) return; - buf_appendmdoc(dbuf, n->child, 1); - buf_appendmdoc(buf, n->child, 0); + mustfree = render_string(&cp, &sz); - hash_put(hash, buf, TYPE_DESC); -} + 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=0x%llx", (int)sz, cp, + (unsigned long long)v); + } else { + htab = &strings; + if (debug > 1) + for (i = 0; i < KEY_MAX; i++) + if ((uint64_t)1 << i & v) + say(mpage->mlinks->file, + "Adding key %s=%*s", + mansearch_keynames[i], (int)sz, cp); + } -/* ARGSUSED */ -static void -pmdoc_Er(MDOC_ARGS) -{ + end = cp + sz; + slot = ohash_qlookupi(htab, cp, &end); + s = ohash_find(htab, slot); - if (SEC_ERRORS != n->sec) + if (NULL != s && mpage == s->mpage) { + s->mask |= v; return; - - buf_appendmdoc(buf, n->child, 0); - hash_put(hash, buf, TYPE_ERR); + } else if (NULL == s) { + s = mandoc_calloc(1, sizeof(struct str) + sz + 1); + memcpy(s->key, cp, sz); + ohash_insert(htab, slot, s); + } + s->mpage = mpage; + s->mask = v; + + if (mustfree) + free(cp); } -/* ARGSUSED */ -static void -pmdoc_Ev(MDOC_ARGS) +/* + * Take a Unicode codepoint and produce its UTF-8 encoding. + * This isn't the best way to do this, but it works. + * The magic numbers are from the UTF-8 packaging. + * They're not as scary as they seem: read the UTF-8 spec for details. + */ +static size_t +utf8(unsigned int cp, char out[7]) { + size_t rc; - if (SEC_ENVIRONMENT != n->sec) - return; - - buf_appendmdoc(buf, n->child, 0); - hash_put(hash, buf, TYPE_ENV); + rc = 0; + if (cp <= 0x0000007F) { + rc = 1; + out[0] = (char)cp; + } else if (cp <= 0x000007FF) { + rc = 2; + out[0] = (cp >> 6 & 31) | 192; + out[1] = (cp & 63) | 128; + } else if (cp <= 0x0000FFFF) { + rc = 3; + out[0] = (cp >> 12 & 15) | 224; + out[1] = (cp >> 6 & 63) | 128; + out[2] = (cp & 63) | 128; + } else if (cp <= 0x001FFFFF) { + rc = 4; + out[0] = (cp >> 18 & 7) | 240; + out[1] = (cp >> 12 & 63) | 128; + out[2] = (cp >> 6 & 63) | 128; + out[3] = (cp & 63) | 128; + } else if (cp <= 0x03FFFFFF) { + rc = 5; + out[0] = (cp >> 24 & 3) | 248; + out[1] = (cp >> 18 & 63) | 128; + out[2] = (cp >> 12 & 63) | 128; + out[3] = (cp >> 6 & 63) | 128; + out[4] = (cp & 63) | 128; + } else if (cp <= 0x7FFFFFFF) { + rc = 6; + out[0] = (cp >> 30 & 1) | 252; + out[1] = (cp >> 24 & 63) | 128; + out[2] = (cp >> 18 & 63) | 128; + out[3] = (cp >> 12 & 63) | 128; + out[4] = (cp >> 6 & 63) | 128; + out[5] = (cp & 63) | 128; + } else + return 0; + + out[rc] = '\0'; + return rc; } -/* ARGSUSED */ -static void -pmdoc_Pa(MDOC_ARGS) +/* + * 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 int +render_string(char **public, size_t *psz) { + const char *src, *scp, *addcp, *seq; + char *dst; + size_t ssz, dsz, addsz; + char utfbuf[7], res[6]; + int seqlen, unicode; - if (SEC_FILES != n->sec) - return; - - buf_appendmdoc(buf, n->child, 0); - hash_put(hash, buf, TYPE_PATH); + res[0] = '\\'; + res[1] = '\t'; + res[2] = ASCII_NBRSP; + res[3] = ASCII_HYPH; + res[4] = ASCII_BREAK; + res[5] = '\0'; + + src = scp = *public; + ssz = *psz; + dst = NULL; + dsz = 0; + + while (scp < src + *psz) { + + /* Leave normal characters unchanged. */ + + if (strchr(res, *scp) == NULL) { + if (dst != NULL) + dst[dsz++] = *scp; + scp++; + continue; + } + + /* + * Found something that requires replacing, + * make sure we have a destination buffer. + */ + + if (dst == NULL) { + dst = mandoc_malloc(ssz + 1); + dsz = scp - src; + memcpy(dst, src, dsz); + } + + /* Handle single-char special characters. */ + + switch (*scp) { + case '\\': + break; + case '\t': + case ASCII_NBRSP: + dst[dsz++] = ' '; + scp++; + continue; + case ASCII_HYPH: + dst[dsz++] = '-'; + /* FALLTHROUGH */ + case ASCII_BREAK: + scp++; + continue; + default: + abort(); + } + + /* + * Found an escape sequence. + * Read past the slash, then parse it. + * Ignore everything except characters. + */ + + scp++; + if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL) + continue; + + /* + * Render the special character + * as either UTF-8 or ASCII. + */ + + if (write_utf8) { + unicode = mchars_spec2cp(seq, seqlen); + if (unicode <= 0) + continue; + addsz = utf8(unicode, utfbuf); + if (addsz == 0) + continue; + addcp = utfbuf; + } else { + addcp = mchars_spec2str(seq, seqlen, &addsz); + if (addcp == NULL) + continue; + if (*addcp == ASCII_NBRSP) { + addcp = " "; + addsz = 1; + } + } + + /* Copy the rendered glyph into the stream. */ + + ssz += addsz; + dst = mandoc_realloc(dst, ssz + 1); + memcpy(dst + dsz, addcp, addsz); + dsz += addsz; + } + if (dst != NULL) { + *public = dst; + *psz = dsz; + } + + /* Trim trailing whitespace and NUL-terminate. */ + + while (*psz > 0 && (*public)[*psz - 1] == ' ') + --*psz; + if (dst != NULL) { + (*public)[*psz] = '\0'; + return 1; + } else + return 0; } -/* ARGSUSED */ static void -pmdoc_Nm(MDOC_ARGS) +dbadd_mlink(const struct mlink *mlink) { - - if (SEC_NAME == n->sec) { - buf_appendmdoc(buf, n->child, 0); - hash_put(hash, buf, TYPE_NAME); - return; - } else if (SEC_SYNOPSIS != n->sec || MDOC_HEAD != n->type) - return; - - if (NULL == n->child) - buf_append(buf, m->name); - - buf_appendmdoc(buf, n->child, 0); - hash_put(hash, buf, TYPE_UTILITY); + dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE); + dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec); + dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec); + dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch); + dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file); } +/* + * Flush the current page's terms (and their bits) into the database. + * Also, handle escape sequences at the last possible moment. + */ static void -hash_put(DB *db, const struct buf *buf, int mask) +dbadd(struct dba *dba, struct mpage *mpage) { - DBT key, val; - int rc; + struct mlink *mlink; + struct str *key; + char *cp; + uint64_t mask; + size_t i; + unsigned int slot; + int mustfree; - if (buf->len < 2) + mlink = mpage->mlinks; + + if (nodb) { + for (key = ohash_first(&names, &slot); NULL != key; + key = ohash_next(&names, &slot)) + free(key); + for (key = ohash_first(&strings, &slot); NULL != key; + key = ohash_next(&strings, &slot)) + free(key); + if (0 == debug) + return; + while (NULL != mlink) { + fputs(mlink->name, stdout); + if (NULL == mlink->next || + 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; + } - key.data = buf->cp; - key.size = buf->len; + if (debug) + say(mlink->file, "Adding to database"); - if ((rc = (*db->get)(db, &key, &val, 0)) < 0) { - perror("hash"); - exit((int)MANDOCLEVEL_SYSERR); - } else if (0 == rc) - mask |= *(int *)val.data; + cp = mpage->desc; + i = strlen(cp); + mustfree = render_string(&cp, &i); + mpage->dba = dba_page_new(dba->pages, + *mpage->arch == '\0' ? mlink->arch : mpage->arch, + cp, mlink->file, mpage->form); + if (mustfree) + free(cp); + dba_page_add(mpage->dba, DBP_SECT, mpage->sec); - val.data = &mask; - val.size = sizeof(int); + while (mlink != NULL) { + dbadd_mlink(mlink); + mlink = mlink->next; + } - if ((rc = (*db->put)(db, &key, &val, 0)) < 0) { - perror("hash"); - exit((int)MANDOCLEVEL_SYSERR); - } + for (key = ohash_first(&names, &slot); NULL != key; + key = ohash_next(&names, &slot)) { + assert(key->mpage == mpage); + dba_page_alias(mpage->dba, key->key, key->mask); + free(key); + } + for (key = ohash_first(&strings, &slot); NULL != key; + key = ohash_next(&strings, &slot)) { + assert(key->mpage == mpage); + i = 0; + for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) { + if (key->mask & mask) + dba_macro_add(dba->macros, i, + key->key, mpage->dba); + i++; + } + free(key); + } } static void -dbt_put(DB *db, const char *dbn, DBT *key, DBT *val) +dbprune(struct dba *dba) { + struct dba_array *page, *files; + char *file; - assert(key->size); - assert(val->size); - - if (0 == (*db->put)(db, key, val, 0)) - return; - - perror(dbn); - exit((int)MANDOCLEVEL_SYSERR); - /* NOTREACHED */ + dba_array_FOREACH(dba->pages, page) { + files = dba_array_get(page, DBP_FILE); + dba_array_FOREACH(files, file) { + if (*file < ' ') + file++; + if (ohash_find(&mlinks, ohash_qlookup(&mlinks, + file)) != NULL) { + if (debug) + say(file, "Deleting from database"); + dba_array_del(dba->pages); + break; + } + } + } } /* - * Call out to per-macro handlers after clearing the persistent database - * key. If the macro sets the database key, flush it to the database. + * Write the database from memory to disk. */ static void -pmdoc_node(MDOC_ARGS) +dbwrite(struct dba *dba) { + char tfn[32]; + int status; + pid_t child; - if (NULL == n) + if (dba_write(MANDOC_DB "~", dba) != -1) { + if (rename(MANDOC_DB "~", MANDOC_DB) == -1) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say(MANDOC_DB, "&rename"); + unlink(MANDOC_DB "~"); + } return; + } - switch (n->type) { - case (MDOC_HEAD): - /* FALLTHROUGH */ - case (MDOC_BODY): - /* FALLTHROUGH */ - case (MDOC_TAIL): - /* FALLTHROUGH */ - case (MDOC_BLOCK): - /* FALLTHROUGH */ - case (MDOC_ELEM): - if (NULL == mdocs[n->tok]) - break; + (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn)); + if (mkdtemp(tfn) == NULL) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "&%s", tfn); + return; + } - buf->len = 0; - (*mdocs[n->tok])(hash, buf, dbuf, n, m); - break; + (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn)); + if (dba_write(tfn, dba) == -1) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say(tfn, "&dba_write"); + goto out; + } + + switch (child = fork()) { + case -1: + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "&fork cmp"); + return; + case 0: + execlp("cmp", "cmp", "-s", tfn, MANDOC_DB, (char *)NULL); + say("", "&exec cmp"); + exit(0); default: break; } + if (waitpid(child, &status, 0) == -1) { + 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"); + } - pmdoc_node(hash, buf, dbuf, n->child, m); - pmdoc_node(hash, buf, dbuf, n->next, m); +out: + *strrchr(tfn, '/') = '\0'; + switch (child = fork()) { + case -1: + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "&fork rm"); + return; + case 0: + execlp("rm", "rm", "-rf", tfn, (char *)NULL); + say("", "&exec rm"); + exit((int)MANDOCLEVEL_SYSERR); + default: + break; + } + if (waitpid(child, &status, 0) == -1) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "&wait rm"); + } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) { + exitcode = (int)MANDOCLEVEL_SYSERR; + say("", "%s: Cannot remove temporary directory", tfn); + } } static int -pman_node(MAN_ARGS) +set_basedir(const char *targetdir, int report_baddir) { - const struct man_node *head, *body; - const char *start, *sv; - size_t sz; + static char startdir[PATH_MAX]; + static int getcwd_status; /* 1 = ok, 2 = failure */ + static int chdir_status; /* 1 = changed directory */ + char *cp; - if (NULL == n) - 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; + } /* - * We're only searching for one thing: the first text child in - * the BODY of a NAME section. Since we don't keep track of - * sections in -man, run some hoops to find out whether we're in - * the correct section or not. + * We are leaving the old base directory. + * Do not use it any longer, not even for messages. */ + *basedir = '\0'; - if (MAN_BODY == n->type && MAN_SH == n->tok) { - 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) { - - assert(body->string); - start = sv = body->string; - - /* - * Go through a special heuristic dance here. - * This is why -man manuals are great! - * (I'm being sarcastic: my eyes are bleeding.) - * Conventionally, one or more manual names are - * comma-specified prior to a whitespace, then a - * dash, then a description. Try to puzzle out - * the name parts here. - */ - - for ( ;; ) { - sz = strcspn(start, " ,"); - if ('\0' == start[(int)sz]) - break; - - buf->len = 0; - buf_appendb(buf, start, sz); - buf_appendb(buf, "", 1); - - hash_put(hash, buf, TYPE_NAME); - - if (' ' == start[(int)sz]) { - start += (int)sz + 1; - break; - } - - assert(',' == start[(int)sz]); - start += (int)sz + 1; - while (' ' == *start) - start++; - } - - buf->len = 0; - - if (sv == start) { - buf_append(buf, start); - return(1); - } - - while (' ' == *start) - start++; - - if (0 == strncmp(start, "-", 1)) - start += 1; - else if (0 == strncmp(start, "\\-", 2)) - start += 2; - else if (0 == strncmp(start, "\\(en", 4)) - start += 4; - else if (0 == strncmp(start, "\\(em", 4)) - start += 4; - - while (' ' == *start) - start++; - - sz = strlen(start) + 1; - buf_appendb(dbuf, start, sz); - buf_appendb(buf, start, sz); - - hash_put(hash, buf, TYPE_DESC); + /* + * 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; + } } - if (pman_node(hash, buf, dbuf, n->child)) - return(1); - if (pman_node(hash, buf, dbuf, n->next)) - return(1); - - 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 -usage(void) +say(const char *file, const char *format, ...) { + va_list ap; + int use_errno; - fprintf(stderr, "usage: %s [-ruv] [-d path] [file...]\n", - progname); + if ('\0' != *basedir) + fprintf(stderr, "%s", basedir); + if ('\0' != *basedir && '\0' != *file) + fputc('/', stderr); + if ('\0' != *file) + fprintf(stderr, "%s", file); + + 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); }