Merge OpenBSD's `so' handling (plus some documentation). Great work to

[mandoc.git] / main.c

diff --git a/main.c b/main.c
index 188a3732276d8106009c0bde35b5a93fead1c9dc..16b54def251c26cfa738942ef75d5e06b3d1127f 100644 (file)
--- a/main.c
+++ b/main.c
@@ -1,109 +1,256 @@
-/* $Id: main.c,v 1.3 2009/03/19 16:40:49 kristaps Exp $ */
+/*     $Id: main.c,v 1.113 2010/12/01 16:54:25 kristaps Exp $ */

 /*
 /*

- * Copyright (c) 2008, 2009 Kristaps Dzonsons <kristaps@openbsd.org>
+ * Copyright (c) 2008, 2009, 2010 Kristaps Dzonsons <kristaps@bsd.lv>
+ * Copyright (c) 2010 Ingo Schwarze <schwarze@openbsd.org>

  *
  * Permission to use, copy, modify, and distribute this software for any
  *
  * 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.
+ * 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 WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED
- * WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE
- * AUTHOR 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.
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR 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 <sys/mman.h>

 #include <sys/stat.h>
 
 #include <assert.h>
 #include <sys/stat.h>
 
 #include <assert.h>

-#include <err.h>
+#include <ctype.h>

 #include <fcntl.h>
 #include <stdio.h>
 #include <fcntl.h>
 #include <stdio.h>

+#include <stdint.h>

 #include <stdlib.h>
 #include <string.h>
 #include <unistd.h>
 
 #include <stdlib.h>
 #include <string.h>
 #include <unistd.h>
 

+#include "mandoc.h"
+#include "main.h"

 #include "mdoc.h"
 #include "mdoc.h"

+#include "man.h"
+#include "roff.h"

 
 

-#ifdef __linux__
-extern int               getsubopt(char **, char * const *, char **);
-# ifndef __dead
-#  define __dead __attribute__((__noreturn__))
-# endif
+#ifndef MAP_FILE
+#define        MAP_FILE        0

 #endif
 
 #endif
 

-#define        WARN_WALL         0x03          /* All-warnings mask. */
-#define        WARN_WCOMPAT     (1 << 0)       /* Compatibility warnings. */
-#define        WARN_WSYNTAX     (1 << 1)       /* Syntax warnings. */
-#define        WARN_WERR        (1 << 2)       /* Warnings->errors. */
+#define        UNCONST(a)      ((void *)(uintptr_t)(const void *)(a))
+
+/* FIXME: Intel's compiler?  LLVM?  pcc?  */
+
+#if !defined(__GNUC__) || (__GNUC__ < 2)
+# if !defined(lint)
+#  define __attribute__(x)
+# endif
+#endif /* !defined(__GNUC__) || (__GNUC__ < 2) */
+
+typedef        void            (*out_mdoc)(void *, const struct mdoc *);
+typedef        void            (*out_man)(void *, const struct man *);
+typedef        void            (*out_free)(void *);

 
 

-enum outt {
-       OUTT_ASCII,
-       OUTT_LATIN1,
-       OUTT_UTF8,
+struct buf {
+       char             *buf;
+       size_t            sz;
+};
+
+enum   intt {
+       INTT_AUTO,
+       INTT_MDOC,
+       INTT_MAN
+};
+
+enum   outt {
+       OUTT_ASCII = 0,

        OUTT_TREE,
        OUTT_TREE,

-       OUTT_LINT
+       OUTT_HTML,
+       OUTT_XHTML,
+       OUTT_LINT,
+       OUTT_PS,
+       OUTT_PDF

 };
 
 };
 

-typedef        int             (*out_run)(void *, const struct mdoc *);
-typedef        void            (*out_free)(void *);
+struct curparse {
+       const char       *file;         /* Current parse. */
+       int               fd;           /* Current parse. */
+       int               line;         /* Line number in the file. */
+       enum mandoclevel  wlevel;       /* Ignore messages below this. */
+       int               wstop;        /* Stop after a file with a warning. */
+       enum intt         inttype;      /* which parser to use */
+       struct man       *pman;         /* persistent man parser */
+       struct mdoc      *pmdoc;        /* persistent mdoc parser */
+       struct man       *man;          /* man parser */
+       struct mdoc      *mdoc;         /* mdoc parser */
+       struct roff      *roff;         /* roff parser (!NULL) */
+       struct regset     regs;         /* roff registers */
+       enum outt         outtype;      /* which output to use */
+       out_mdoc          outmdoc;      /* mdoc output ptr */
+       out_man           outman;       /* man output ptr */
+       out_free          outfree;      /* free output ptr */
+       void             *outdata;      /* data for output */
+       char              outopts[BUFSIZ]; /* buf of output opts */
+};
+
+static const char * const      mandoclevels[MANDOCLEVEL_MAX] = {
+       "SUCCESS",
+       "RESERVED",
+       "WARNING",
+       "ERROR",
+       "FATAL",
+       "BADARG",
+       "SYSERR"
+};
+
+static const enum mandocerr    mandoclimits[MANDOCLEVEL_MAX] = {
+       MANDOCERR_OK,
+       MANDOCERR_WARNING,
+       MANDOCERR_WARNING,
+       MANDOCERR_ERROR,
+       MANDOCERR_FATAL,
+       MANDOCERR_MAX,
+       MANDOCERR_MAX
+};

 
 

-extern char             *__progname;
-
-extern void             *ascii_alloc(void);
-extern void             *latin1_alloc(void);
-extern void             *utf8_alloc(void);
-extern int               terminal_run(void *, const struct mdoc *);
-extern int               tree_run(void *, const struct mdoc *);
-extern void              terminal_free(void *);
-
-__dead static void       version(void);
-__dead static void       usage(void);
-static int               foptions(int *, char *);
-static int               toptions(enum outt *, char *);
-static int               woptions(int *, char *);
-static int               merr(void *, int, int, const char *);
-static int               mwarn(void *, int, int, 
-                               enum mdoc_warn, const char *);
-static int               file(char **, size_t *, char **, size_t *, 
-                               const char *, struct mdoc *);
-static int               fdesc(char **, size_t *, char **, size_t *, 
-                               const char *, int, struct mdoc *);
+static const char * const      mandocerrs[MANDOCERR_MAX] = {
+       "ok",
+
+       "generic warning",
+
+       ".so is fragile, better use ln(1)",
+       "text should be uppercase",
+       "sections out of conventional order",
+       "section name repeats",
+       "out of order prologue",
+       "repeated prologue entry",
+       "list type must come first",
+       "tab in non-literal context",
+       "bad escape sequence",
+       "unterminated quoted string",
+       "argument requires the width argument",
+       "superfluous width argument",
+       "bad date argument",
+       "bad width argument",
+       "unknown manual section",
+       "section not in conventional manual section",
+       "end of line whitespace",
+       "blocks badly nested",
+
+       "generic error",
+
+       "NAME section must come first",
+       "bad Boolean value",
+       "child violates parent syntax",
+       "bad AT&T symbol",
+       "bad standard",
+       "list type repeated",
+       "display type repeated",
+       "argument repeated",
+       "ignoring argument",
+       "manual name not yet set",
+       "obsolete macro ignored",
+       "empty macro ignored",
+       "macro not allowed in body",
+       "macro not allowed in prologue",
+       "bad character",
+       "bad NAME section contents",
+       "no blank lines",
+       "no text in this context",
+       "bad comment style",
+       "unknown macro will be lost",
+       "NOT IMPLEMENTED: skipping request",
+       "line scope broken",
+       "argument count wrong",
+       "request scope close w/none open",
+       "scope already open",
+       "scope open on exit",
+       "macro requires line argument(s)",
+       "macro requires body argument(s)",
+       "macro requires argument(s)",
+       "no title in document",
+       "missing list type",
+       "missing display type",
+       "missing font type",
+       "line argument(s) will be lost",
+       "body argument(s) will be lost",
+       "paragraph macro ignored",
+
+       "generic fatal error",
+
+       "column syntax is inconsistent",
+       "displays may not be nested",
+       "unsupported display type",
+       "blocks badly nested",
+       "no such block is open",
+       "line scope broken, syntax violated",
+       "argument count wrong, violates syntax",
+       "child violates parent syntax",
+       "argument count wrong, violates syntax",
+       "NOT IMPLEMENTED: .so with absolute path or \"..\"",
+       "no document body",
+       "no document prologue",
+       "utsname system call failed",
+       "static buffer exhausted",
+};

 
 

+static void              parsebuf(struct curparse *, struct buf, int);
+static void              pdesc(struct curparse *);
+static void              fdesc(struct curparse *);
+static void              ffile(const char *, struct curparse *);
+static int               pfile(const char *, struct curparse *);
+static int               moptions(enum intt *, char *);
+static int               mmsg(enum mandocerr, void *, 
+                               int, int, const char *);
+static void              pset(const char *, int, struct curparse *);
+static int               toptions(struct curparse *, char *);
+static void              usage(void) __attribute__((noreturn));
+static void              version(void) __attribute__((noreturn));
+static int               woptions(struct curparse *, char *);
+
+static const char       *progname;
+static enum mandoclevel  exit_status = MANDOCLEVEL_OK;

 
 int
 main(int argc, char *argv[])
 {
 
 int
 main(int argc, char *argv[])
 {

-       int              c, rc, fflags, wflags;
-       struct mdoc_cb   cb;
-       struct mdoc     *mdoc;
-       char            *buf, *line;
-       size_t           bufsz, linesz;
-       void            *outdata;
-       enum outt        outtype;
-       out_run          outrun;
-       out_free         outfree;
-
-       fflags = wflags = 0;
-       outtype = OUTT_ASCII;
+       int              c;
+       struct curparse  curp;
+
+       progname = strrchr(argv[0], '/');
+       if (progname == NULL)
+               progname = argv[0];
+       else
+               ++progname;
+
+       memset(&curp, 0, sizeof(struct curparse));
+
+       curp.inttype = INTT_AUTO;
+       curp.outtype = OUTT_ASCII;
+       curp.wlevel  = MANDOCLEVEL_FATAL;

 
        /* LINTED */
 
        /* LINTED */

-       while (-1 != (c = getopt(argc, argv, "f:VW:T:")))
+       while (-1 != (c = getopt(argc, argv, "m:O:T:VW:")))

                switch (c) {
                switch (c) {

-               case ('f'):
-                       if ( ! foptions(&fflags, optarg))
-                               return(0);
+               case ('m'):
+                       if ( ! moptions(&curp.inttype, optarg))
+                               return((int)MANDOCLEVEL_BADARG);
+                       break;
+               case ('O'):
+                       (void)strlcat(curp.outopts, optarg, BUFSIZ);
+                       (void)strlcat(curp.outopts, ",", BUFSIZ);

                        break;
                case ('T'):
                        break;
                case ('T'):

-                       if ( ! toptions(&outtype, optarg))
-                               return(0);
+                       if ( ! toptions(&curp, optarg))
+                               return((int)MANDOCLEVEL_BADARG);

                        break;
                case ('W'):
                        break;
                case ('W'):

-                       if ( ! woptions(&wflags, optarg))
-                               return(0);
+                       if ( ! woptions(&curp, optarg))
+                               return((int)MANDOCLEVEL_BADARG);

                        break;
                case ('V'):
                        version();
                        break;
                case ('V'):
                        version();


@@ -116,345 +263,724 @@ main(int argc, char *argv[])
        argc -= optind;
        argv += optind;
 
        argc -= optind;
        argv += optind;
 

-       /*
-        * Allocate the appropriate front-end.  Note that utf8, ascii
-        * and latin1 all resolve to the terminal front-end with
-        * different encodings (see terminal.c).  Not all frontends have
-        * cleanup or alloc routines.
-        */
+       if (NULL == *argv) {
+               curp.file = "<stdin>";
+               curp.fd = STDIN_FILENO;

 
 

-       switch (outtype) {
-       case (OUTT_LATIN1):
-               outdata = latin1_alloc();
-               outrun = terminal_run;
-               outfree = terminal_free;
-               break;
-       case (OUTT_UTF8):
-               outdata = utf8_alloc();
-               outrun = terminal_run;
-               outfree = terminal_free;
-               break;
-       case (OUTT_TREE):
-               outdata = NULL;
-               outrun = tree_run;
-               outfree = NULL;
-               break;
-       case (OUTT_LINT):
-               outdata = NULL;
-               outrun = NULL;
-               outfree = NULL;
-               break;
-       default:
-               outdata = ascii_alloc();
-               outrun = terminal_run;
-               outfree = terminal_free;
-               break;
+               fdesc(&curp);

        }
 
        }
 

-       /*
-        * All callbacks route into here, where we print them onto the
-        * screen.  XXX - for now, no path for debugging messages.
-        */
-
-       cb.mdoc_msg = NULL;
-       cb.mdoc_err = merr;
-       cb.mdoc_warn = mwarn;
-
-       buf = line = NULL;
-       bufsz = linesz = 0;
-
-       mdoc = mdoc_alloc(&wflags, fflags, &cb);
-

        while (*argv) {
        while (*argv) {

-               if ( ! file(&line, &linesz, &buf, &bufsz, *argv, mdoc))
-                       break;
-               if (outrun && ! (*outrun)(outdata, mdoc))
+               ffile(*argv, &curp);
+               if (MANDOCLEVEL_OK != exit_status && curp.wstop)

                        break;
                        break;

-
-               /* Reset the parser for another file. */
-               mdoc_reset(mdoc);
-               argv++;
+               ++argv;

        }
 
        }
 

-       rc = NULL == *argv;
-
-       if (buf)
-               free(buf);
-       if (line)
-               free(line);
-       if (outfree)
-               (*outfree)(outdata);
+       if (curp.outfree)
+               (*curp.outfree)(curp.outdata);
+       if (curp.pmdoc)
+               mdoc_free(curp.pmdoc);
+       if (curp.pman)
+               man_free(curp.pman);
+       if (curp.roff)
+               roff_free(curp.roff);

 
 

-       mdoc_free(mdoc);
-
-       return(rc ? EXIT_SUCCESS : EXIT_FAILURE);
+       return((int)exit_status);

 }
 
 
 }
 
 

-__dead static void
+static void

 version(void)
 {
 
 version(void)
 {
 

-       (void)printf("%s %s\n", __progname, VERSION);
-       exit(0);
-       /* NOTREACHED */
+       (void)printf("%s %s\n", progname, VERSION);
+       exit((int)MANDOCLEVEL_OK);

 }
 
 
 }
 
 

-__dead static void
+static void

 usage(void)
 {
 
 usage(void)
 {
 

-       (void)fprintf(stderr, "usage: %s\n", __progname);
-       exit(1);
-       /* NOTREACHED */
+       (void)fprintf(stderr, "usage: %s "
+                       "[-V] "
+                       "[-foption] "
+                       "[-mformat] "
+                       "[-Ooption] "
+                       "[-Toutput] "
+                       "[-Werr] "
+                       "[file...]\n", 
+                       progname);
+
+       exit((int)MANDOCLEVEL_BADARG);

 }
 
 }
 

+static void
+ffile(const char *file, struct curparse *curp)
+{
+
+       /*
+        * Called once per input file.  Get the file ready for reading,
+        * pass it through to the parser-driver, then close it out.
+        * XXX: don't do anything special as this is only called for
+        * files; stdin goes directly to fdesc().
+        */
+
+       curp->file = file;
+
+       if (-1 == (curp->fd = open(curp->file, O_RDONLY, 0))) {
+               perror(curp->file);
+               exit_status = MANDOCLEVEL_SYSERR;
+               return;
+       }
+
+       fdesc(curp);
+
+       if (-1 == close(curp->fd))
+               perror(curp->file);
+}

 
 static int
 
 static int

-file(char **ln, size_t *lnsz, char **buf, size_t *bufsz, 
-               const char *file, struct mdoc *mdoc)
+pfile(const char *file, struct curparse *curp)

 {
 {

-       int              fd, c;
+       const char      *savefile;
+       int              fd, savefd;

 
        if (-1 == (fd = open(file, O_RDONLY, 0))) {
 
        if (-1 == (fd = open(file, O_RDONLY, 0))) {

-               warn("%s", file);
+               perror(file);
+               exit_status = MANDOCLEVEL_SYSERR;

                return(0);
        }
 
                return(0);
        }
 

-       c = fdesc(ln, lnsz, buf, bufsz, file, fd, mdoc);
+       savefile = curp->file;
+       savefd = curp->fd;
+
+       curp->file = file;
+       curp->fd = fd;
+
+       pdesc(curp);
+
+       curp->file = savefile;
+       curp->fd = savefd;

 
        if (-1 == close(fd))
 
        if (-1 == close(fd))

-               warn("%s", file);
+               perror(file);
+
+       return(MANDOCLEVEL_FATAL > exit_status ? 1 : 0);
+}
+

 
 

-       return(c);
+static void
+resize_buf(struct buf *buf, size_t initial)
+{
+
+       buf->sz = buf->sz ? 2 * buf->sz : initial;
+       buf->buf = realloc(buf->buf, buf->sz);
+       if (NULL == buf->buf) {
+               perror(NULL);
+               exit((int)MANDOCLEVEL_SYSERR);
+       }

 }
 
 
 static int
 }
 
 
 static int

-fdesc(char **lnp, size_t *lnsz, char **bufp, size_t *bufsz, 
-               const char *f, int fd, struct mdoc *mdoc)
+read_whole_file(struct curparse *curp, struct buf *fb, int *with_mmap)

 {
 {

-       size_t           sz;
-       ssize_t          ssz;

        struct stat      st;
        struct stat      st;

-       int              j, i, pos, lnn;
-       char            *ln, *buf;
+       size_t           off;
+       ssize_t          ssz;

 
 

-       buf = *bufp;
-       ln = *lnp;
+       if (-1 == fstat(curp->fd, &st)) {
+               perror(curp->file);
+               return(0);
+       }

 
        /*
 
        /*

-        * Two buffers: ln and buf.  buf is the input buffer, optimised
-        * for each file's block size.  ln is a line buffer.  Both
-        * growable, hence passed in by ptr-ptr.
+        * If we're a regular file, try just reading in the whole entry
+        * via mmap().  This is faster than reading it into blocks, and
+        * since each file is only a few bytes to begin with, I'm not
+        * concerned that this is going to tank any machines.

         */
 
         */
 

-       if (-1 == fstat(fd, &st)) {
-               warnx("%s", f);
-               sz = BUFSIZ;
-       } else 
-               sz = (unsigned)BUFSIZ > st.st_blksize ?
-                       (size_t)BUFSIZ : st.st_blksize;
-
-       if (sz > *bufsz) {
-               if (NULL == (buf = realloc(buf, sz)))
-                       err(1, "realloc");
-               *bufp = buf;
-               *bufsz = sz;
+       if (S_ISREG(st.st_mode)) {
+               if (st.st_size >= (1U << 31)) {
+                       fprintf(stderr, "%s: input too large\n", 
+                                       curp->file);
+                       return(0);
+               }
+               *with_mmap = 1;
+               fb->sz = (size_t)st.st_size;
+               fb->buf = mmap(NULL, fb->sz, PROT_READ, 
+                               MAP_FILE|MAP_SHARED, curp->fd, 0);
+               if (fb->buf != MAP_FAILED)
+                       return(1);

        }
 
        /*
        }
 
        /*

-        * Fill buf with file blocksize and parse newlines into ln.
+        * If this isn't a regular file (like, say, stdin), then we must
+        * go the old way and just read things in bit by bit.

         */
 
         */
 

-       for (lnn = 1, pos = 0; ; ) {
-               if (-1 == (ssz = read(fd, buf, sz))) {
-                       warn("%s", f);
-                       return(0);
-               } else if (0 == ssz) 
+       *with_mmap = 0;
+       off = 0;
+       fb->sz = 0;
+       fb->buf = NULL;
+       for (;;) {
+               if (off == fb->sz) {
+                       if (fb->sz == (1U << 31)) {
+                               fprintf(stderr, "%s: input too large\n", 
+                                               curp->file);
+                               break;
+                       }
+                       resize_buf(fb, 65536);
+               }
+               ssz = read(curp->fd, fb->buf + (int)off, fb->sz - off);
+               if (ssz == 0) {
+                       fb->sz = off;
+                       return(1);
+               }
+               if (ssz == -1) {
+                       perror(curp->file);
+                       break;
+               }
+               off += (size_t)ssz;
+       }
+
+       free(fb->buf);
+       fb->buf = NULL;
+       return(0);
+}
+
+
+static void
+fdesc(struct curparse *curp)
+{
+
+       /*
+        * Called once per file with an opened file descriptor.  All
+        * pre-file-parse operations (whether stdin or a file) should go
+        * here.
+        *
+        * This calls down into the nested parser, which drills down and
+        * fully parses a file and all its dependences (i.e., `so').  It
+        * then runs the cleanup validators and pushes to output.
+        */
+
+       /* Zero the parse type. */
+
+       curp->mdoc = NULL;
+       curp->man = NULL;
+
+       /* Make sure the mandotory roff parser is initialised. */
+
+       if (NULL == curp->roff) {
+               curp->roff = roff_alloc(&curp->regs, curp, mmsg);
+               assert(curp->roff);
+       }
+
+       /* Fully parse the file. */
+
+       pdesc(curp);
+
+       if (MANDOCLEVEL_FATAL <= exit_status)
+               goto cleanup;
+
+       /* NOTE a parser may not have been assigned, yet. */
+
+       if ( ! (curp->man || curp->mdoc)) {
+               fprintf(stderr, "%s: Not a manual\n", curp->file);
+               exit_status = MANDOCLEVEL_FATAL;
+               goto cleanup;
+       }
+
+       /* Clean up the parse routine ASTs. */
+
+       if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
+               assert(MANDOCLEVEL_FATAL <= exit_status);
+               goto cleanup;
+       }
+
+       if (curp->man && ! man_endparse(curp->man)) {
+               assert(MANDOCLEVEL_FATAL <= exit_status);
+               goto cleanup;
+       }
+
+       assert(curp->roff);
+       if ( ! roff_endparse(curp->roff)) {
+               assert(MANDOCLEVEL_FATAL <= exit_status);
+               goto cleanup;
+       }
+
+       /*
+        * With -Wstop and warnings or errors of at least
+        * the requested level, do not produce output.
+        */
+
+       if (MANDOCLEVEL_OK != exit_status && curp->wstop)
+               goto cleanup;
+
+       /* If unset, allocate output dev now (if applicable). */
+
+       if ( ! (curp->outman && curp->outmdoc)) {
+               switch (curp->outtype) {
+               case (OUTT_XHTML):
+                       curp->outdata = xhtml_alloc(curp->outopts);
+                       break;
+               case (OUTT_HTML):
+                       curp->outdata = html_alloc(curp->outopts);
+                       break;
+               case (OUTT_ASCII):
+                       curp->outdata = ascii_alloc(curp->outopts);
+                       curp->outfree = ascii_free;
+                       break;
+               case (OUTT_PDF):
+                       curp->outdata = pdf_alloc(curp->outopts);
+                       curp->outfree = pspdf_free;
+                       break;
+               case (OUTT_PS):
+                       curp->outdata = ps_alloc(curp->outopts);
+                       curp->outfree = pspdf_free;
+                       break;
+               default:
+                       break;
+               }
+
+               switch (curp->outtype) {
+               case (OUTT_HTML):
+                       /* FALLTHROUGH */
+               case (OUTT_XHTML):
+                       curp->outman = html_man;
+                       curp->outmdoc = html_mdoc;
+                       curp->outfree = html_free;
+                       break;
+               case (OUTT_TREE):
+                       curp->outman = tree_man;
+                       curp->outmdoc = tree_mdoc;
+                       break;
+               case (OUTT_PDF):
+                       /* FALLTHROUGH */
+               case (OUTT_ASCII):
+                       /* FALLTHROUGH */
+               case (OUTT_PS):
+                       curp->outman = terminal_man;
+                       curp->outmdoc = terminal_mdoc;
+                       break;
+               default:

                        break;
                        break;

+               }
+       }
+
+       /* Execute the out device, if it exists. */
+
+       if (curp->man && curp->outman)
+               (*curp->outman)(curp->outdata, curp->man);
+       if (curp->mdoc && curp->outmdoc)
+               (*curp->outmdoc)(curp->outdata, curp->mdoc);
+
+ cleanup:
+
+       memset(&curp->regs, 0, sizeof(struct regset));
+
+       /* Reset the current-parse compilers. */
+
+       if (curp->mdoc)
+               mdoc_reset(curp->mdoc);
+       if (curp->man)
+               man_reset(curp->man);
+
+       assert(curp->roff);
+       roff_reset(curp->roff);
+
+       return;
+}
+
+static void
+pdesc(struct curparse *curp)
+{
+       struct buf       blk;
+       int              with_mmap;
+
+       /*
+        * Run for each opened file; may be called more than once for
+        * each full parse sequence if the opened file is nested (i.e.,
+        * from `so').  Simply sucks in the whole file and moves into
+        * the parse phase for the file.
+        */
+
+       if ( ! read_whole_file(curp, &blk, &with_mmap)) {
+               exit_status = MANDOCLEVEL_SYSERR;
+               return;
+       }
+
+       /* Line number is per-file. */
+
+       curp->line = 1;
+
+       parsebuf(curp, blk, 1);
+
+       if (with_mmap)
+               munmap(blk.buf, blk.sz);
+       else
+               free(blk.buf);
+}
+
+static void
+parsebuf(struct curparse *curp, struct buf blk, int start)
+{
+       struct buf       ln;
+       int              i, of, rc;
+       int              pos; /* byte number in the ln buffer */
+       int              lnn; /* line number in the real file */
+       unsigned char    c;
+
+       /*
+        * Main parse routine for an opened file.  This is called for
+        * each opened file and simply loops around the full input file,
+        * possibly nesting (i.e., with `so').
+        */
+
+       memset(&ln, 0, sizeof(struct buf));
+
+       lnn = curp->line; 
+       pos = 0; 

 
 

-               for (i = 0; i < (int)ssz; i++) {
-                       if (pos >= (int)*lnsz) {
-                               *lnsz += 256; /* Step-size. */
-                               ln = realloc(ln, *lnsz);
-                               if (NULL == ln)
-                                       err(1, "realloc");
-                               *lnp = ln;
+       for (i = 0; i < (int)blk.sz; ) {
+               if (0 == pos && '\0' == blk.buf[i])
+                       break;
+
+               if (start)
+                       curp->line = lnn;
+
+               while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
+                       if ('\n' == blk.buf[i]) {
+                               ++i;
+                               ++lnn;
+                               break;
+                       }
+
+                       /* 
+                        * Warn about bogus characters.  If you're using
+                        * non-ASCII encoding, you're screwing your
+                        * readers.  Since I'd rather this not happen,
+                        * I'll be helpful and drop these characters so
+                        * we don't display gibberish.  Note to manual
+                        * writers: use special characters.
+                        */
+
+                       c = (unsigned char) blk.buf[i];
+
+                       if ( ! (isascii(c) && 
+                                       (isgraph(c) || isblank(c)))) {
+                               mmsg(MANDOCERR_BADCHAR, curp, 
+                                   curp->line, pos, "ignoring byte");
+                               i++;
+                               continue;

                        }
 
                        }
 

-                       if ('\n' != buf[i]) {
-                               ln[pos++] = buf[i];
+                       /* Trailing backslash = a plain char. */
+
+                       if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
+                               if (pos >= (int)ln.sz)
+                                       resize_buf(&ln, 256);
+                               ln.buf[pos++] = blk.buf[i++];

                                continue;
                        }
 
                                continue;
                        }
 

-                       /* Check for CPP-escaped newline.  */
+                       /* Found escape & at least one other char. */

 
 

-                       if (pos > 0 && '\\' == ln[pos - 1]) {
-                               for (j = pos - 1; j >= 0; j--)
-                                       if ('\\' != ln[j])
+                       if ('\n' == blk.buf[i + 1]) {
+                               i += 2;
+                               /* Escaped newlines are skipped over */
+                               ++lnn;
+                               continue;
+                       }
+
+                       if ('"' == blk.buf[i + 1]) {
+                               i += 2;
+                               /* Comment, skip to end of line */
+                               for (; i < (int)blk.sz; ++i) {
+                                       if ('\n' == blk.buf[i]) {
+                                               ++i;
+                                               ++lnn;

                                                break;
                                                break;

+                                       }
+                               }

 
 

-                               if ( ! ((pos - j) % 2)) {
-                                       pos--;
-                                       lnn++;
-                                       continue;
+                               /* Backout trailing whitespaces */
+                               for (; pos > 0; --pos) {
+                                       if (ln.buf[pos - 1] != ' ')
+                                               break;
+                                       if (pos > 2 && ln.buf[pos - 2] == '\\')
+                                               break;

                                }
                                }

+                               break;

                        }
 
                        }
 

-                       ln[pos] = 0;
-                       if ( ! mdoc_parseln(mdoc, lnn, ln))
-                               return(0);
-                       lnn++;
+                       /* Some other escape sequence, copy & cont. */
+
+                       if (pos + 1 >= (int)ln.sz)
+                               resize_buf(&ln, 256);
+
+                       ln.buf[pos++] = blk.buf[i++];
+                       ln.buf[pos++] = blk.buf[i++];
+               }
+
+               if (pos >= (int)ln.sz)
+                       resize_buf(&ln, 256);
+
+               ln.buf[pos] = '\0';
+
+               /*
+                * A significant amount of complexity is contained by
+                * the roff preprocessor.  It's line-oriented but can be
+                * expressed on one line, so we need at times to
+                * readjust our starting point and re-run it.  The roff
+                * preprocessor can also readjust the buffers with new
+                * data, so we pass them in wholesale.
+                */
+
+               of = 0;
+
+rerun:
+               rc = roff_parseln
+                       (curp->roff, curp->line, 
+                        &ln.buf, &ln.sz, of, &of);
+
+               switch (rc) {
+               case (ROFF_REPARSE):
+                       parsebuf(curp, ln, 0);
+                       pos = 0;
+                       continue;
+               case (ROFF_APPEND):
+                       pos = strlen(ln.buf);
+                       continue;
+               case (ROFF_RERUN):
+                       goto rerun;
+               case (ROFF_IGN):

                        pos = 0;
                        pos = 0;

+                       continue;
+               case (ROFF_ERR):
+                       assert(MANDOCLEVEL_FATAL <= exit_status);
+                       break;
+               case (ROFF_SO):
+                       if (pfile(ln.buf + of, curp)) {
+                               pos = 0;
+                               continue;
+                       } else
+                               break;
+               case (ROFF_CONT):
+                       break;

                }
                }

+
+               /*
+                * If input parsers have not been allocated, do so now.
+                * We keep these instanced betwen parsers, but set them
+                * locally per parse routine since we can use different
+                * parsers with each one.
+                */
+
+               if ( ! (curp->man || curp->mdoc))
+                       pset(ln.buf + of, pos - of, curp);
+
+               /* 
+                * Lastly, push down into the parsers themselves.  One
+                * of these will have already been set in the pset()
+                * routine.
+                */
+
+               if (curp->man || curp->mdoc) {
+                       rc = curp->man ?
+                               man_parseln(curp->man, 
+                                       curp->line, ln.buf, of) :
+                               mdoc_parseln(curp->mdoc, 
+                                       curp->line, ln.buf, of);
+
+                       if ( ! rc) {
+                               assert(MANDOCLEVEL_FATAL <= exit_status);
+                               break;
+                       }
+               }
+
+               /* Temporary buffers typically are not full. */
+
+               if (0 == start && '\0' == blk.buf[i])
+                       break;
+
+               /* Start the next input line. */
+
+               pos = 0;

        }
 
        }
 

-       return(mdoc_endparse(mdoc));
+       free(ln.buf);

 }
 
 }
 

+static void
+pset(const char *buf, int pos, struct curparse *curp)
+{
+       int              i;
+
+       /*
+        * Try to intuit which kind of manual parser should be used.  If
+        * passed in by command-line (-man, -mdoc), then use that
+        * explicitly.  If passed as -mandoc, then try to guess from the
+        * line: either skip dot-lines, use -mdoc when finding `.Dt', or
+        * default to -man, which is more lenient.
+        *
+        * Separate out pmdoc/pman from mdoc/man: the first persists
+        * through all parsers, while the latter is used per-parse.
+        */
+
+       if ('.' == buf[0] || '\'' == buf[0]) {
+               for (i = 1; buf[i]; i++)
+                       if (' ' != buf[i] && '\t' != buf[i])
+                               break;
+               if ('\0' == buf[i])
+                       return;
+       }
+
+       switch (curp->inttype) {
+       case (INTT_MDOC):
+               if (NULL == curp->pmdoc) 
+                       curp->pmdoc = mdoc_alloc
+                               (&curp->regs, curp, mmsg);
+               assert(curp->pmdoc);
+               curp->mdoc = curp->pmdoc;
+               return;
+       case (INTT_MAN):
+               if (NULL == curp->pman) 
+                       curp->pman = man_alloc
+                               (&curp->regs, curp, mmsg);
+               assert(curp->pman);
+               curp->man = curp->pman;
+               return;
+       default:
+               break;
+       }
+
+       if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3))  {
+               if (NULL == curp->pmdoc) 
+                       curp->pmdoc = mdoc_alloc
+                               (&curp->regs, curp, mmsg);
+               assert(curp->pmdoc);
+               curp->mdoc = curp->pmdoc;
+               return;
+       } 
+
+       if (NULL == curp->pman) 
+               curp->pman = man_alloc(&curp->regs, curp, mmsg);
+       assert(curp->pman);
+       curp->man = curp->pman;
+}

 
 static int
 
 static int

-toptions(enum outt *tflags, char *arg)
+moptions(enum intt *tflags, char *arg)

 {
 
 {
 

-       if (0 == strcmp(arg, "ascii"))
-               *tflags = OUTT_ASCII;
-       else if (0 == strcmp(arg, "latin1"))
-               *tflags = OUTT_LATIN1;
-       else if (0 == strcmp(arg, "utf8"))
-               *tflags = OUTT_UTF8;
-       else if (0 == strcmp(arg, "lint"))
-               *tflags = OUTT_LINT;
-       else if (0 == strcmp(arg, "tree"))
-               *tflags = OUTT_TREE;
+       if (0 == strcmp(arg, "doc"))
+               *tflags = INTT_MDOC;
+       else if (0 == strcmp(arg, "andoc"))
+               *tflags = INTT_AUTO;
+       else if (0 == strcmp(arg, "an"))
+               *tflags = INTT_MAN;

        else {
        else {

-               warnx("bad argument: -T%s", arg);
+               fprintf(stderr, "%s: Bad argument\n", arg);

                return(0);
        }
 
        return(1);
 }
 
                return(0);
        }
 
        return(1);
 }
 

-
-/*
- * Parse out the options for [-fopt...] setting compiler options.  These
- * can be comma-delimited or called again.
- */

 static int
 static int

-foptions(int *fflags, char *arg)
+toptions(struct curparse *curp, char *arg)

 {
 {

-       char            *v;
-       char            *toks[4];

 
 

-       toks[0] = "ign-scope";
-       toks[1] = "ign-escape";
-       toks[2] = "ign-macro";
-       toks[3] = NULL;
-
-       while (*arg) 
-               switch (getsubopt(&arg, toks, &v)) {
-               case (0):
-                       *fflags |= MDOC_IGN_SCOPE;
-                       break;
-               case (1):
-                       *fflags |= MDOC_IGN_ESCAPE;
-                       break;
-               case (2):
-                       *fflags |= MDOC_IGN_MACRO;
-                       break;
-               default:
-                       warnx("bad argument: -f%s", arg);
-                       return(0);
-               }
+       if (0 == strcmp(arg, "ascii"))
+               curp->outtype = OUTT_ASCII;
+       else if (0 == strcmp(arg, "lint")) {
+               curp->outtype = OUTT_LINT;
+               curp->wlevel  = MANDOCLEVEL_WARNING;
+       }
+       else if (0 == strcmp(arg, "tree"))
+               curp->outtype = OUTT_TREE;
+       else if (0 == strcmp(arg, "html"))
+               curp->outtype = OUTT_HTML;
+       else if (0 == strcmp(arg, "xhtml"))
+               curp->outtype = OUTT_XHTML;
+       else if (0 == strcmp(arg, "ps"))
+               curp->outtype = OUTT_PS;
+       else if (0 == strcmp(arg, "pdf"))
+               curp->outtype = OUTT_PDF;
+       else {
+               fprintf(stderr, "%s: Bad argument\n", arg);
+               return(0);
+       }

 
        return(1);
 }
 
 
        return(1);
 }
 

-
-/* 
- * Parse out the options for [-Werr...], which sets warning modes.
- * These can be comma-delimited or called again.  
- */

 static int
 static int

-woptions(int *wflags, char *arg)
+woptions(struct curparse *curp, char *arg)

 {
 {

-       char            *v;
-       char            *toks[5]; 
+       char            *v, *o;
+       const char      *toks[6]; 

 
 

-       toks[0] = "all";
-       toks[1] = "compat";
-       toks[2] = "syntax";
+       toks[0] = "stop";
+       toks[1] = "all";
+       toks[2] = "warning";

        toks[3] = "error";
        toks[3] = "error";

-       toks[4] = NULL;
+       toks[4] = "fatal";
+       toks[5] = NULL;

 
 

-       while (*arg) 
-               switch (getsubopt(&arg, toks, &v)) {
+       while (*arg) {
+               o = arg;
+               switch (getsubopt(&arg, UNCONST(toks), &v)) {

                case (0):
                case (0):

-                       *wflags |= WARN_WALL;
+                       curp->wstop = 1;

                        break;
                case (1):
                        break;
                case (1):

-                       *wflags |= WARN_WCOMPAT;
-                       break;
+                       /* FALLTHROUGH */

                case (2):
                case (2):

-                       *wflags |= WARN_WSYNTAX;
+                       curp->wlevel = MANDOCLEVEL_WARNING;

                        break;
                case (3):
                        break;
                case (3):

-                       *wflags |= WARN_WERR;
+                       curp->wlevel = MANDOCLEVEL_ERROR;
+                       break;
+               case (4):
+                       curp->wlevel = MANDOCLEVEL_FATAL;

                        break;
                default:
                        break;
                default:

-                       warnx("bad argument: -W%s", arg);
+                       fprintf(stderr, "-W%s: Bad argument\n", o);

                        return(0);
                }
                        return(0);
                }

+       }

 
        return(1);
 }
 
 
        return(1);
 }
 

-
-/* ARGSUSED */

 static int
 static int

-merr(void *arg, int line, int col, const char *msg)
+mmsg(enum mandocerr t, void *arg, int ln, int col, const char *msg)

 {
 {

+       struct curparse *cp;
+       enum mandoclevel level;

 
 

-       warnx("error: %s (line %d, column %d)", msg, line, col);
-       return(0);
-}
+       level = MANDOCLEVEL_FATAL;
+       while (t < mandoclimits[level])
+               /* LINTED */
+               level--;

 
 

-
-static int
-mwarn(void *arg, int line, int col, 
-               enum mdoc_warn type, const char *msg)
-{
-       int              flags;
-       char            *wtype;
-
-       flags = *(int *)arg;
-       wtype = NULL;
-
-       switch (type) {
-       case (WARN_COMPAT):
-               wtype = "compat";
-               if (flags & WARN_WCOMPAT)
-                       break;
-               return(1);
-       case (WARN_SYNTAX):
-               wtype = "syntax";
-               if (flags & WARN_WSYNTAX)
-                       break;
+       cp = (struct curparse *)arg;
+       if (level < cp->wlevel)

                return(1);
                return(1);

-       }

 
 

-       assert(wtype);
-       warnx("%s warning: %s (line %d, column %d)", 
-                       wtype, msg, line, col);
-
-       if ( ! (flags & WARN_WERR))
-               return(1);
+       fprintf(stderr, "%s:%d:%d: %s: %s",
+           cp->file, ln, col + 1, mandoclevels[level], mandocerrs[t]);
+       if (msg)
+               fprintf(stderr, ": %s", msg);
+       fputc('\n', stderr);

 
 

-       warnx("%s: considering warnings as errors", 
-                       __progname);
-       return(0);
+       if (exit_status < level)
+               exit_status = level;
+       
+       return(level < MANDOCLEVEL_FATAL);

 }
 }

-
-