X-Git-Url: https://git.cameronkatri.com/mandoc.git/blobdiff_plain/88d20eead0ae7a10800a000fd389e1b49ed6431e..2dd7cbdf0bf0c97f0926fa2015fb1566cf7e233b:/main.c

diff --git a/main.c b/main.c
index 71836e3a..3324f8a8 100644
--- a/main.c
+++ b/main.c
@@ -1,120 +1,239 @@
-/* $Id: main.c,v 1.9 2009/03/22 19:10:48 kristaps Exp $ */
+/*	$Id: main.c,v 1.98 2010/07/07 15:04:54 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
- * 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 <err.h>
 #include <fcntl.h>
 #include <stdio.h>
+#include <stdint.h>
 #include <stdlib.h>
 #include <string.h>
 #include <unistd.h>
 
+#include "mandoc.h"
+#include "main.h"
 #include "mdoc.h"
+#include "man.h"
+#include "roff.h"
+
+#define	UNCONST(a)	((void *)(uintptr_t)(const void *)(a))
+
+/* FIXME: Intel's compiler?  LLVM?  pcc?  */
 
-#ifdef __linux__
-extern	int		  getsubopt(char **, char * const *, char **);
-# ifndef __dead
-#  define __dead __attribute__((__noreturn__))
+#if !defined(__GNUC__) || (__GNUC__ < 2)
+# if !defined(lint)
+#  define __attribute__(x)
 # endif
-#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 *);
 
 struct	buf {
 	char	 	 *buf;
 	size_t		  sz;
 };
 
-struct	curparse {
-	const char	 *file;
-	int		  wflags;
-#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. */
+enum	intt {
+	INTT_AUTO,
+	INTT_MDOC,
+	INTT_MAN
 };
 
-enum outt {
-	OUTT_ASCII,
-	OUTT_LATIN1,
-	OUTT_UTF8,
+enum	outt {
+	OUTT_ASCII = 0,
 	OUTT_TREE,
-	OUTT_LINT
+	OUTT_HTML,
+	OUTT_XHTML,
+	OUTT_LINT,
+	OUTT_PS
 };
 
-typedef	int		(*out_run)(void *, const struct mdoc *);
-typedef	void		(*out_free)(void *);
-
-extern	char		 *__progname;
+struct	curparse {
+	const char	 *file;		/* Current parse. */
+	int		  fd;		/* Current parse. */
+	int		  wflags;
+	/* FIXME: set by max error */
+#define	WARN_WALL	 (1 << 0)	/* All-warnings mask. */
+#define	WARN_WERR	 (1 << 2)	/* Warnings->errors. */
+	int		  fflags;
+#define	FL_IGN_SCOPE	 (1 << 0) 	/* Ignore scope errors. */
+#define	FL_NIGN_ESCAPE	 (1 << 1) 	/* Don't ignore bad escapes. */
+#define	FL_NIGN_MACRO	 (1 << 2) 	/* Don't ignore bad macros. */
+#define	FL_IGN_ERRORS	 (1 << 4)	/* Ignore failed parse. */
+#define	FL_STRICT	  FL_NIGN_ESCAPE | \
+			  FL_NIGN_MACRO /* ignore nothing */
+	enum intt	  inttype;	/* which parser to use */
+	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 */
+};
 
-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 *);
+static	const char * const	mandocerrs[MANDOCERR_MAX] = {
+	"ok",
+
+	"generic warning",
+
+	"text should be uppercase",
+	"sections out of conventional order",
+	"section name repeats",
+	"out of order prologue",
+	"repeated prologue entry",
+	"list type must come first",
+	"bad standard",
+	"bad library",
+	"bad escape sequence",
+	"unterminated quoted string",
+	"argument requires the width argument",
+	"superfluous width argument",
+	"ignoring argument",
+	"bad date argument",
+	"bad width argument",
+	"unknown manual section",
+	"section not in conventional manual section",
+	"end of line whitespace",
+	"blocks badly nested",
+	"scope open on exit",
+
+	"generic error",
+
+	"NAME section must come first",
+	"bad Boolean value",
+	"child violates parent syntax",
+	"bad AT&T symbol",
+	"list type repeated",
+	"display type repeated",
+	"argument repeated",
+	"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",
+	"line scope broken",
+	"argument count wrong",
+	"request scope close w/none open",
+	"scope already open",
+	"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",
+
+	"generic fatal error",
+
+	"column syntax is inconsistent",
+	"displays may not be nested",
+	"unsupported display type",
+	"blocks badly nested",
+	"no such block is open",
+	"scope broken, syntax violated",
+	"line scope broken, syntax violated",
+	"argument count wrong, violates syntax",
+	"child violates parent syntax",
+	"argument count wrong, violates syntax",
+	"no document body",
+	"no document prologue",
+	"utsname system call failed",
+	"memory exhausted",
+};
 
-__dead	static void	  version(void);
-__dead	static void	  usage(void);
+static	void		  fdesc(struct curparse *);
+static	void		  ffile(const char *, struct curparse *);
 static	int		  foptions(int *, char *);
-static	int		  toptions(enum outt *, char *);
+static	struct man	 *man_init(struct curparse *);
+static	struct mdoc	 *mdoc_init(struct curparse *);
+static	struct roff	 *roff_init(struct curparse *);
+static	int		  moptions(enum intt *, char *);
+static	int		  mmsg(enum mandocerr, void *, 
+				int, int, const char *);
+static	int		  pset(const char *, int, struct curparse *,
+				struct man **, struct mdoc **);
+static	int		  toptions(struct curparse *, char *);
+static	void		  usage(void) __attribute__((noreturn));
+static	void		  version(void) __attribute__((noreturn));
 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(struct buf *, struct buf *,
-				const char *, struct mdoc *);
-static	int		  fdesc(struct buf *, struct buf *,
-				const char *, int, struct mdoc *);
 
+static	const char	 *progname;
+static	int		  with_fatal;
+static	int		  with_error;
 
 int
 main(int argc, char *argv[])
 {
-	int		 c, rc, fflags;
-	struct mdoc_cb	 cb;
-	struct mdoc	*mdoc;
-	void		*outdata;
-	enum outt	 outtype;
-	struct buf	 ln, blk;
-	out_run		 outrun;
-	out_free	 outfree;
+	int		 c;
 	struct curparse	 curp;
 
-	fflags = 0;
-	outtype = OUTT_ASCII;
+	progname = strrchr(argv[0], '/');
+	if (progname == NULL)
+		progname = argv[0];
+	else
+		++progname;
 
-	bzero(&curp, sizeof(struct curparse));
+	memset(&curp, 0, sizeof(struct curparse));
+
+	curp.inttype = INTT_AUTO;
+	curp.outtype = OUTT_ASCII;
 
 	/* LINTED */
-	while (-1 != (c = getopt(argc, argv, "f:VW:T:")))
+	while (-1 != (c = getopt(argc, argv, "f:m:O:T:VW:")))
 		switch (c) {
 		case ('f'):
-			if ( ! foptions(&fflags, optarg))
-				return(0);
+			if ( ! foptions(&curp.fflags, optarg))
+				return(EXIT_FAILURE);
+			break;
+		case ('m'):
+			if ( ! moptions(&curp.inttype, optarg))
+				return(EXIT_FAILURE);
+			break;
+		case ('O'):
+			(void)strlcat(curp.outopts, optarg, BUFSIZ);
+			(void)strlcat(curp.outopts, ",", BUFSIZ);
 			break;
 		case ('T'):
-			if ( ! toptions(&outtype, optarg))
-				return(0);
+			if ( ! toptions(&curp, optarg))
+				return(EXIT_FAILURE);
 			break;
 		case ('W'):
 			if ( ! woptions(&curp.wflags, optarg))
-				return(0);
+				return(EXIT_FAILURE);
 			break;
 		case ('V'):
 			version();
@@ -127,311 +246,524 @@ main(int argc, char *argv[])
 	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.
-	 */
+	while (*argv) {
+		ffile(*argv, &curp);
 
-	cb.mdoc_msg = NULL;
-	cb.mdoc_err = merr;
-	cb.mdoc_warn = mwarn;
+		if (with_fatal && !(curp.fflags & FL_IGN_ERRORS))
+			break;
+		++argv;
+	}
 
-	bzero(&ln, sizeof(struct buf));
-	bzero(&blk, sizeof(struct buf));
+	if (curp.outfree)
+		(*curp.outfree)(curp.outdata);
+	if (curp.mdoc)
+		mdoc_free(curp.mdoc);
+	if (curp.man)
+		man_free(curp.man);
+	if (curp.roff)
+		roff_free(curp.roff);
+
+	return((with_fatal || with_error) ? 
+			EXIT_FAILURE :  EXIT_SUCCESS);
+}
 
-	mdoc = mdoc_alloc(&curp, fflags, &cb);
 
-	/*
-	 * Loop around available files.
-	 */
+static void
+version(void)
+{
 
-	if (NULL == *argv) {
-		curp.file = "<stdin>";
-		c = fdesc(&blk, &ln, "stdin", STDIN_FILENO, mdoc);
-		rc = 0;
-		if (c && NULL == outrun)
-			rc = 1;
-		else if (c && outrun && (*outrun)(outdata, mdoc))
-			rc = 1;
-	} else {
-		while (*argv) {
-			curp.file = *argv;
-			c = file(&blk, &ln, *argv, mdoc);
-			if ( ! c)
-				break;
-			if (outrun && ! (*outrun)(outdata, mdoc))
-				break;
-			/* Reset the parser for another file. */
-			mdoc_reset(mdoc);
-			argv++;
-		}
-		rc = NULL == *argv;
-	}
+	(void)printf("%s %s\n", progname, VERSION);
+	exit(EXIT_SUCCESS);
+}
 
-	if (blk.buf)
-		free(blk.buf);
-	if (ln.buf)
-		free(ln.buf);
-	if (outfree)
-		(*outfree)(outdata);
 
-	mdoc_free(mdoc);
+static void
+usage(void)
+{
 
-	return(rc ? EXIT_SUCCESS : EXIT_FAILURE);
+	(void)fprintf(stderr, "usage: %s [-V] [-foption] "
+			"[-mformat] [-Ooption] [-Toutput] "
+			"[-Werr] [file...]\n", progname);
+	exit(EXIT_FAILURE);
 }
 
 
-__dead static void
-version(void)
+static struct man *
+man_init(struct curparse *curp)
 {
+	int		 pflags;
+
+	/* Defaults from mandoc.1. */
 
-	(void)printf("%s %s\n", __progname, VERSION);
-	exit(0);
-	/* NOTREACHED */
+	pflags = MAN_IGN_MACRO | MAN_IGN_ESCAPE;
+
+	if (curp->fflags & FL_NIGN_MACRO)
+		pflags &= ~MAN_IGN_MACRO;
+	if (curp->fflags & FL_NIGN_ESCAPE)
+		pflags &= ~MAN_IGN_ESCAPE;
+
+	return(man_alloc(&curp->regs, curp, pflags, mmsg));
 }
 
 
-__dead static void
-usage(void)
+static struct roff *
+roff_init(struct curparse *curp)
 {
 
-	(void)fprintf(stderr, "usage: %s\n", __progname);
-	exit(1);
-	/* NOTREACHED */
+	return(roff_alloc(&curp->regs, mmsg, curp));
 }
 
 
-static int
-file(struct buf *blk, struct buf *ln,
-		const char *file, struct mdoc *mdoc)
+static struct mdoc *
+mdoc_init(struct curparse *curp)
 {
-	int		 fd, c;
+	int		 pflags;
 
-	if (-1 == (fd = open(file, O_RDONLY, 0))) {
-		warn("%s", file);
-		return(0);
+	/* Defaults from mandoc.1. */
+
+	pflags = MDOC_IGN_MACRO | MDOC_IGN_ESCAPE;
+
+	if (curp->fflags & FL_IGN_SCOPE)
+		pflags |= MDOC_IGN_SCOPE;
+	if (curp->fflags & FL_NIGN_ESCAPE)
+		pflags &= ~MDOC_IGN_ESCAPE;
+	if (curp->fflags & FL_NIGN_MACRO)
+		pflags &= ~MDOC_IGN_MACRO;
+
+	return(mdoc_alloc(&curp->regs, curp, pflags, mmsg));
+}
+
+
+static void
+ffile(const char *file, struct curparse *curp)
+{
+
+	curp->file = file;
+	if (-1 == (curp->fd = open(curp->file, O_RDONLY, 0))) {
+		perror(curp->file);
+		with_fatal = 1;
+		return;
 	}
 
-	c = fdesc(blk, ln, file, fd, mdoc);
+	fdesc(curp);
+
+	if (-1 == close(curp->fd))
+		perror(curp->file);
+}
 
-	if (-1 == close(fd))
-		warn("%s", file);
 
-	return(c);
+static int
+resize_buf(struct buf *buf, size_t initial)
+{
+	void *tmp;
+	size_t sz;
+
+	if (buf->sz == 0)
+		sz = initial;
+	else
+		sz = 2 * buf->sz;
+	tmp = realloc(buf->buf, sz);
+	if (NULL == tmp) {
+		perror(NULL);
+		return(0);
+	}
+	buf->buf = tmp;
+	buf->sz = sz;
+	return(1);
 }
 
 
 static int
-fdesc(struct buf *blk, struct buf *ln,
-		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;
-	int		 j, i, pos, lnn;
-#ifdef	STRIP_XO
-	int		 macro, xo, xeoln;
-#endif
+	size_t		 off;
+	ssize_t		 ssz;
+
+	if (-1 == fstat(curp->fd, &st)) {
+		perror(curp->file);
+		with_fatal = 1;
+		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.
 	 */
 
-	sz = BUFSIZ;
-
-	if (-1 == fstat(fd, &st))
-		warnx("%s", f);
-	else if ((size_t)st.st_blksize > sz)
-		sz = st.st_blksize;
-
-	if (sz > blk->sz) {
-		blk->buf = realloc(blk->buf, sz);
-		if (NULL == blk->buf)
-			err(1, "realloc");
-		blk->sz = sz;
+	if (S_ISREG(st.st_mode)) {
+		if (st.st_size >= (1U << 31)) {
+			fprintf(stderr, "%s: input too large\n", 
+					curp->file);
+			with_fatal = 1;
+			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.
 	 */
-#ifdef	STRIP_XO
-	macro = xo = xeoln = 0;
-#endif
 
-	for (lnn = 1, pos = 0; ; ) {
-		if (-1 == (ssz = read(fd, blk->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;
+			}
+			if (! resize_buf(fb, 65536))
+				break;
+		}
+		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;
+	}
 
-		for (i = 0; i < (int)ssz; i++) {
-			if (pos >= (int)ln->sz) {
-				ln->sz += 256; /* Step-size. */
-				ln->buf = realloc(ln->buf, ln->sz);
-				if (NULL == ln->buf)
-					err(1, "realloc");
-			}
+	free(fb->buf);
+	fb->buf = NULL;
+	with_fatal = 1;
+	return(0);
+}
 
-			if ('\n' != blk->buf[i]) {
-				/*
-				 * Ugly of uglies.  Here we handle the
-				 * dreaded `Xo/Xc' scoping.  Cover the
-				 * eyes of any nearby children.  This
-				 * makes `Xo/Xc' enclosures look like
-				 * one huge line.
-				 */
-#ifdef	STRIP_XO
-				/*
-				 * First, note whether we're in a macro
-				 * line.
-				 */
-				if (0 == pos && '.' == blk->buf[i])
-					macro = 1;
-
-				/*
-				 * If we're in an `Xo' context and just
-				 * nixed a newline, remove the control
-				 * character for new macro lines:
-				 * they're going to show up as all part
-				 * of the same line.
-				 */
-				if (xo && xeoln && '.' == blk->buf[i]) {
-					xeoln = 0;
-					continue;
-				}
-				xeoln = 0;
-
-				/*
-				 * If we've parsed `Xo', enter an xo
-				 * context.  `Xo' must be in a parsable
-				 * state.  This is the ugly part.  IT IS
-				 * NOT SMART ENOUGH TO HANDLE ESCAPED
-				 * WHITESPACE.
-				 */
-				if (macro && pos && 'o' == blk->buf[i]) {
-					if (xo && 'X' == ln->buf[pos - 1])  {
-						if (' ' == ln->buf[pos - 2])
-							xo++;
-					} else if ('X' == ln->buf[pos - 1]) {
-						if (2 == pos && '.' == ln->buf[pos - 2])
-							xo++;
-						else if (' ' == ln->buf[pos - 2])
-							xo++;
-					}
-				}
 
-				/*
-				 * If we're parsed `Xc', leave an xo
-				 * context if one's already pending.
-				 * `Xc' must be in a parsable state.
-				 * THIS IS NOT SMART ENOUGH TO HANDLE
-				 * ESCAPED WHITESPACE.
-				 */
-				if (macro && pos && 'c' == blk->buf[i])
-					if (xo && 'X' == ln->buf[pos - 1])
-						if (' ' == ln->buf[pos - 2])
-							xo--;
-#endif	/* STRIP_XO */
-
-				ln->buf[pos++] = blk->buf[i];
-				continue;
-			}
+static void
+fdesc(struct curparse *curp)
+{
+	struct buf	 ln, blk;
+	int		 i, pos, lnn, lnn_start, with_mmap, of;
+	enum rofferr	 re;
+	struct man	*man;
+	struct mdoc	*mdoc;
+	struct roff	*roff;
 
-			/* Check for CPP-escaped newline.  */
+	man = NULL;
+	mdoc = NULL;
+	roff = NULL;
 
-			if (pos > 0 && '\\' == ln->buf[pos - 1]) {
-				for (j = pos - 1; j >= 0; j--)
-					if ('\\' != ln->buf[j])
-						break;
+	memset(&ln, 0, sizeof(struct buf));
 
-				if ( ! ((pos - j) % 2)) {
-					pos--;
-					lnn++;
-					continue;
-				}
-			}
+	/*
+	 * Two buffers: ln and buf.  buf is the input file and may be
+	 * memory mapped.  ln is a line buffer and grows on-demand.
+	 */
 
-#ifdef	STRIP_XO
-			/*
-			 * If we're in an xo context, put a space in
-			 * place of the newline and continue parsing.
-			 * Mark that we just did a newline.
-			 */
-			if (xo) {
-				xeoln = 1;
-				ln->buf[pos++] = ' ';
-				lnn++;
+	if ( ! read_whole_file(curp, &blk, &with_mmap))
+		return;
+
+	if (NULL == curp->roff) 
+		curp->roff = roff_init(curp);
+	if (NULL == (roff = curp->roff))
+		goto bailout;
+
+	for (i = 0, lnn = 1; i < (int)blk.sz;) {
+		pos = 0;
+		lnn_start = lnn;
+		while (i < (int)blk.sz) {
+			if ('\n' == blk.buf[i]) {
+				++i;
+				++lnn;
+				break;
+			}
+			/* Trailing backslash is like a plain character. */
+			if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
+				if (pos >= (int)ln.sz)
+					if (! resize_buf(&ln, 256))
+						goto bailout;
+				ln.buf[pos++] = blk.buf[i++];
+				continue;
+			}
+			/* Found an escape and at least one other character. */
+			if ('\n' == blk.buf[i + 1]) {
+				/* Escaped newlines are skipped over */
+				i += 2;
+				++lnn;
 				continue;
 			}
-			macro = 0;
-#endif	/* STRIP_XO */
-
-			ln->buf[pos] = 0;
-			if ( ! mdoc_parseln(mdoc, lnn, ln->buf))
-				return(0);
-			lnn++;
-			pos = 0;
+			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;
+					}
+				}
+				/* Backout trailing whitespaces */
+				for (; pos > 0; --pos) {
+					if (ln.buf[pos - 1] != ' ')
+						break;
+					if (pos > 2 && ln.buf[pos - 2] == '\\')
+						break;
+				}
+				break;
+			}
+			/* Some other escape sequence, copy and continue. */
+			if (pos + 1 >= (int)ln.sz)
+				if (! resize_buf(&ln, 256))
+					goto bailout;
+
+			ln.buf[pos++] = blk.buf[i++];
+			ln.buf[pos++] = blk.buf[i++];
+		}
+
+ 		if (pos >= (int)ln.sz)
+			if (! resize_buf(&ln, 256))
+				goto bailout;
+		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;
+		do {
+			re = roff_parseln(roff, lnn_start, 
+					&ln.buf, &ln.sz, of, &of);
+		} while (ROFF_RERUN == re);
+
+		if (ROFF_IGN == re)
+			continue;
+		else if (ROFF_ERR == re)
+			goto bailout;
+
+		/*
+		 * 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 ( ! (man || mdoc))
+			if ( ! pset(ln.buf + of, pos - of, curp, &man, &mdoc))
+				goto bailout;
+
+		/* Lastly, push down into the parsers themselves. */
+
+		if (man && ! man_parseln(man, lnn_start, ln.buf, of))
+			goto bailout;
+		if (mdoc && ! mdoc_parseln(mdoc, lnn_start, ln.buf, of))
+			goto bailout;
+	}
+
+	/* NOTE a parser may not have been assigned, yet. */
+
+	if ( ! (man || mdoc)) {
+		fprintf(stderr, "%s: Not a manual\n", curp->file);
+		goto bailout;
+	}
+
+	/* Clean up the parse routine ASTs. */
+
+	if (mdoc && ! mdoc_endparse(mdoc))
+		goto bailout;
+	if (man && ! man_endparse(man))
+		goto bailout;
+	if (roff && ! roff_endparse(roff))
+		goto bailout;
+
+	/* 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_PS):
+			curp->outdata = ps_alloc(curp->outopts);
+			curp->outfree = ps_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_ASCII):
+			/* FALLTHROUGH */
+		case (OUTT_PS):
+			curp->outman = terminal_man;
+			curp->outmdoc = terminal_mdoc;
+			break;
+		default:
+			break;
+		}
+	}
+
+	/* Execute the out device, if it exists. */
+
+	if (man && curp->outman)
+		(*curp->outman)(curp->outdata, man);
+	if (mdoc && curp->outmdoc)
+		(*curp->outmdoc)(curp->outdata, mdoc);
+
+ cleanup:
+	memset(&curp->regs, 0, sizeof(struct regset));
+	if (mdoc)
+		mdoc_reset(mdoc);
+	if (man)
+		man_reset(man);
+	if (roff)
+		roff_reset(roff);
+	if (ln.buf)
+		free(ln.buf);
+	if (with_mmap)
+		munmap(blk.buf, blk.sz);
+	else
+		free(blk.buf);
+
+	return;
+
+ bailout:
+	with_fatal = 1;
+	goto cleanup;
+}
+
+
+static int
+pset(const char *buf, int pos, struct curparse *curp,
+		struct man **man, struct mdoc **mdoc)
+{
+	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.
+	 */
+
+	if ('.' == buf[0] || '\'' == buf[0]) {
+		for (i = 1; buf[i]; i++)
+			if (' ' != buf[i] && '\t' != buf[i])
+				break;
+		if (0 == buf[i])
+			return(1);
+	}
+
+	switch (curp->inttype) {
+	case (INTT_MDOC):
+		if (NULL == curp->mdoc) 
+			curp->mdoc = mdoc_init(curp);
+		if (NULL == (*mdoc = curp->mdoc))
+			return(0);
+		return(1);
+	case (INTT_MAN):
+		if (NULL == curp->man) 
+			curp->man = man_init(curp);
+		if (NULL == (*man = curp->man))
+			return(0);
+		return(1);
+	default:
+		break;
+	}
+
+	if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3))  {
+		if (NULL == curp->mdoc) 
+			curp->mdoc = mdoc_init(curp);
+		if (NULL == (*mdoc = curp->mdoc))
+			return(0);
+		return(1);
+	} 
+
+	if (NULL == curp->man) 
+		curp->man = man_init(curp);
+	if (NULL == (*man = curp->man))
+		return(0);
+	return(1);
+}
+
+
+static int
+moptions(enum intt *tflags, char *arg)
+{
+
+	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 {
+		fprintf(stderr, "%s: Bad argument\n", arg);
+		return(0);
 	}
 
-	return(mdoc_endparse(mdoc));
+	return(1);
 }
 
 
 static int
-toptions(enum outt *tflags, char *arg)
+toptions(struct curparse *curp, 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;
+		curp->outtype = OUTT_ASCII;
+	else if (0 == strcmp(arg, "lint")) {
+		curp->outtype = OUTT_LINT;
+		curp->wflags |= WARN_WALL;
+		curp->fflags |= FL_STRICT;
+	}
 	else if (0 == strcmp(arg, "tree"))
-		*tflags = OUTT_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 {
-		warnx("bad argument: -T%s", arg);
+		fprintf(stderr, "%s: Bad argument\n", arg);
 		return(0);
 	}
 
@@ -439,127 +771,115 @@ toptions(enum outt *tflags, char *arg)
 }
 
 
-/*
- * Parse out the options for [-fopt...] setting compiler options.  These
- * can be comma-delimited or called again.
- */
 static int
 foptions(int *fflags, char *arg)
 {
-	char		*v;
-	char		*toks[4];
+	char		*v, *o;
+	const char	*toks[8];
 
 	toks[0] = "ign-scope";
-	toks[1] = "ign-escape";
-	toks[2] = "ign-macro";
-	toks[3] = NULL;
-
-	while (*arg) 
-		switch (getsubopt(&arg, toks, &v)) {
+	toks[1] = "no-ign-escape";
+	toks[2] = "no-ign-macro";
+	toks[3] = "ign-errors";
+	toks[4] = "strict";
+	toks[5] = "ign-escape";
+	toks[6] = NULL;
+
+	while (*arg) {
+		o = arg;
+		switch (getsubopt(&arg, UNCONST(toks), &v)) {
 		case (0):
-			*fflags |= MDOC_IGN_SCOPE;
+			*fflags |= FL_IGN_SCOPE;
 			break;
 		case (1):
-			*fflags |= MDOC_IGN_ESCAPE;
+			*fflags |= FL_NIGN_ESCAPE;
 			break;
 		case (2):
-			*fflags |= MDOC_IGN_MACRO;
+			*fflags |= FL_NIGN_MACRO;
+			break;
+		case (3):
+			*fflags |= FL_IGN_ERRORS;
+			break;
+		case (4):
+			*fflags |= FL_STRICT;
+			break;
+		case (5):
+			*fflags &= ~FL_NIGN_ESCAPE;
 			break;
 		default:
-			warnx("bad argument: -f%s", arg);
+			fprintf(stderr, "%s: Bad argument\n", o);
 			return(0);
 		}
+	}
 
 	return(1);
 }
 
 
-/* 
- * Parse out the options for [-Werr...], which sets warning modes.
- * These can be comma-delimited or called again.  
- */
 static int
 woptions(int *wflags, char *arg)
 {
-	char		*v;
-	char		*toks[5]; 
+	char		*v, *o;
+	const char	*toks[3]; 
 
 	toks[0] = "all";
-	toks[1] = "compat";
-	toks[2] = "syntax";
-	toks[3] = "error";
-	toks[4] = NULL;
+	toks[1] = "error";
+	toks[2] = NULL;
 
-	while (*arg) 
-		switch (getsubopt(&arg, toks, &v)) {
+	while (*arg) {
+		o = arg;
+		switch (getsubopt(&arg, UNCONST(toks), &v)) {
 		case (0):
 			*wflags |= WARN_WALL;
 			break;
 		case (1):
-			*wflags |= WARN_WCOMPAT;
-			break;
-		case (2):
-			*wflags |= WARN_WSYNTAX;
-			break;
-		case (3):
 			*wflags |= WARN_WERR;
 			break;
 		default:
-			warnx("bad argument: -W%s", arg);
+			fprintf(stderr, "%s: Bad argument\n", o);
 			return(0);
 		}
+	}
 
 	return(1);
 }
 
 
-/* ARGSUSED */
-static int
-merr(void *arg, int line, int col, const char *msg)
-{
-	struct curparse *curp;
-
-	curp = (struct curparse *)arg;
-
-	warnx("%s:%d: error: %s (column %d)", 
-			curp->file, line, msg, col);
-	return(0);
-}
-
-
 static int
-mwarn(void *arg, int line, int col, 
-		enum mdoc_warn type, const char *msg)
+mmsg(enum mandocerr t, void *arg, int ln, int col, const char *msg)
 {
-	struct curparse *curp;
-	char		*wtype;
+	struct curparse *cp;
+	const char *level;
+	int rc;
 
-	curp = (struct curparse *)arg;
-	wtype = NULL;
+	cp = (struct curparse *)arg;
+	level = NULL;
+	rc = 1;
 
-	switch (type) {
-	case (WARN_COMPAT):
-		wtype = "compat";
-		if (curp->wflags & WARN_WCOMPAT)
-			break;
-		return(1);
-	case (WARN_SYNTAX):
-		wtype = "syntax";
-		if (curp->wflags & WARN_WSYNTAX)
-			break;
-		return(1);
+	if (t >= MANDOCERR_FATAL) {
+		with_fatal = 1;
+		level = "FATAL";
+		rc = 0;
+	} else {
+		if ( ! (WARN_WALL & cp->wflags))
+			return(1);
+		if (t >= MANDOCERR_ERROR) {
+			with_error = 1;
+			level = "ERROR";
+		}
+		if (WARN_WERR & cp->wflags) {
+			with_fatal = 1;
+			rc = 0;
+		}
 	}
 
-	assert(wtype);
-	warnx("%s:%d: %s warning: %s (column %d)", 
-			curp->file, line, wtype, msg, col);
-
-	if ( ! (curp->wflags & WARN_WERR))
-		return(1);
+	fprintf(stderr, "%s:%d:%d:", cp->file, ln, col + 1);
+	if (level)
+		fprintf(stderr, " %s:", level);
+	fprintf(stderr, " %s", mandocerrs[t]);
+	if (msg)
+		fprintf(stderr, ": %s", msg);
+	fputc('\n', stderr);
 
-	warnx("%s: considering warnings as errors", 
-			__progname);
-	return(0);
+	return(rc);
 }
-
-