===================================================================
RCS file: /cvs/mandoc/main.c,v
retrieving revision 1.69
retrieving revision 1.85
diff -u -p -r1.69 -r1.85
--- mandoc/main.c	2010/05/15 16:18:23	1.69
+++ mandoc/main.c	2010/06/07 20:57:09	1.85
@@ -1,4 +1,4 @@
-/*	$Id: main.c,v 1.69 2010/05/15 16:18:23 joerg Exp $ */
+/*	$Id: main.c,v 1.85 2010/06/07 20:57:09 kristaps Exp $ */
 /*
  * Copyright (c) 2008, 2009 Kristaps Dzonsons <kristaps@kth.se>
  *
@@ -29,8 +29,10 @@
 #include <string.h>
 #include <unistd.h>
 
+#include "mandoc.h"
 #include "mdoc.h"
 #include "man.h"
+#include "roff.h"
 #include "main.h"
 
 #define	UNCONST(a)	((void *)(uintptr_t)(const void *)(a))
@@ -63,13 +65,15 @@ enum	outt {
 	OUTT_TREE,
 	OUTT_HTML,
 	OUTT_XHTML,
-	OUTT_LINT
+	OUTT_LINT,
+	OUTT_PS
 };
 
 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;
@@ -78,26 +82,94 @@ struct	curparse {
 #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
-	enum intt	  inttype;	/* Input parsers... */
-	struct man	 *man;
-	struct mdoc	 *mdoc;
-	enum outt	  outtype;	/* Output devices... */
-	out_mdoc	  outmdoc;
-	out_man	  	  outman;
-	out_free	  outfree;
-	void		 *outdata;
-	char		  outopts[BUFSIZ];
+			  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) */
+	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	mandocerrs[MANDOCERR_MAX] = {
+	"ok",
+	"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",
+	"bad date argument",
+	"bad width argument",
+	"unknown manual section",
+	"section not in conventional manual section",
+	"end of line whitespace",
+	"scope open on exit",
+	"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",
+	"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",
+	"line argument(s) will be lost",
+	"body argument(s) will be lost",
+	"column syntax is inconsistent",
+	"missing font type",
+	"missing display type",
+	"displays may not be nested",
+	"no scope to rewind: syntax violated",
+	"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",
+};
+
 static	void		  fdesc(struct curparse *);
 static	void		  ffile(const char *, struct curparse *);
 static	int		  foptions(int *, char *);
 static	struct man	 *man_init(struct curparse *);
 static	struct mdoc	 *mdoc_init(struct curparse *);
-static	int		  merr(void *, int, int, const char *);
+static	struct roff	 *roff_init(struct curparse *);
 static	int		  moptions(enum intt *, char *);
-static	int		  mwarn(void *, int, int, const 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 *);
@@ -177,6 +249,12 @@ main(int argc, char *argv[])
 
 	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_warning || with_error) ? 
 			EXIT_FAILURE :  EXIT_SUCCESS);
@@ -207,11 +285,7 @@ static struct man *
 man_init(struct curparse *curp)
 {
 	int		 pflags;
-	struct man_cb	 mancb;
 
-	mancb.man_err = merr;
-	mancb.man_warn = mwarn;
-
 	/* Defaults from mandoc.1. */
 
 	pflags = MAN_IGN_MACRO | MAN_IGN_ESCAPE;
@@ -221,19 +295,23 @@ man_init(struct curparse *curp)
 	if (curp->fflags & FL_NIGN_ESCAPE)
 		pflags &= ~MAN_IGN_ESCAPE;
 
-	return(man_alloc(curp, pflags, &mancb));
+	return(man_alloc(curp, pflags, mmsg));
 }
 
 
+static struct roff *
+roff_init(struct curparse *curp)
+{
+
+	return(roff_alloc(mmsg, curp));
+}
+
+
 static struct mdoc *
 mdoc_init(struct curparse *curp)
 {
 	int		 pflags;
-	struct mdoc_cb	 mdoccb;
 
-	mdoccb.mdoc_err = merr;
-	mdoccb.mdoc_warn = mwarn;
-
 	/* Defaults from mandoc.1. */
 
 	pflags = MDOC_IGN_MACRO | MDOC_IGN_ESCAPE;
@@ -245,7 +323,7 @@ mdoc_init(struct curparse *curp)
 	if (curp->fflags & FL_NIGN_MACRO)
 		pflags &= ~MDOC_IGN_MACRO;
 
-	return(mdoc_alloc(curp, pflags, &mdoccb));
+	return(mdoc_alloc(curp, pflags, mmsg));
 }
 
 
@@ -316,9 +394,9 @@ read_whole_file(struct curparse *curp, struct buf *fb,
 			return(0);
 		}
 		*with_mmap = 1;
-		fb->sz = st.st_size;
+		fb->sz = (size_t)st.st_size;
 		fb->buf = mmap(NULL, fb->sz, PROT_READ, 
-				MAP_FILE, curp->fd, 0);
+				MAP_FILE|MAP_SHARED, curp->fd, 0);
 		if (fb->buf != MAP_FAILED)
 			return(1);
 	}
@@ -342,7 +420,7 @@ read_whole_file(struct curparse *curp, struct buf *fb,
 			if (! resize_buf(fb, 65536))
 				break;
 		}
-		ssz = read(curp->fd, fb->buf + off, fb->sz - off);
+		ssz = read(curp->fd, fb->buf + (int)off, fb->sz - off);
 		if (ssz == 0) {
 			fb->sz = off;
 			return(1);
@@ -351,7 +429,7 @@ read_whole_file(struct curparse *curp, struct buf *fb,
 			perror(curp->file);
 			break;
 		}
-		off += ssz;
+		off += (size_t)ssz;
 	}
 
 	free(fb->buf);
@@ -365,87 +443,123 @@ static void
 fdesc(struct curparse *curp)
 {
 	struct buf	 ln, blk;
-	int		 j, i, pos, lnn, comment, with_mmap;
+	int		 i, pos, lnn, lnn_start, with_mmap, of;
+	enum rofferr	 re;
 	struct man	*man;
 	struct mdoc	*mdoc;
+	struct roff	*roff;
 
 	man = NULL;
 	mdoc = NULL;
+	roff = NULL;
 	memset(&ln, 0, sizeof(struct buf));
 
 	/*
-	 * Two buffers: ln and buf.  buf is the input buffer optimised
-	 * here for each file's block size.  ln is a line buffer.  Both
-	 * growable, hence passed in by ptr-ptr.
+	 * Two buffers: ln and buf.  buf is the input file and may be
+	 * memory mapped.  ln is a line buffer and grows on-demand.
 	 */
 
-	if (!read_whole_file(curp, &blk, &with_mmap))
+	if ( ! read_whole_file(curp, &blk, &with_mmap))
 		return;
 
-	/* Fill buf with file blocksize. */
+	if (NULL == curp->roff) 
+		curp->roff = roff_init(curp);
+	if (NULL == (roff = curp->roff))
+		goto bailout;
 
-	for (i = lnn = pos = comment = 0; i < (int)blk.sz; ++i) {
-		if (pos >= (int)ln.sz) {
-			if (! resize_buf(&ln, 256))
-				goto bailout;
-		}
-
-		if ('\n' != blk.buf[i]) {
-			if (comment)
+	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;
-			ln.buf[pos++] = blk.buf[i];
-
-			/* Handle in-line `\"' comments. */
-
-			if (1 == pos || '\"' != ln.buf[pos - 1])
-				continue;
-
-			for (j = pos - 2; j >= 0; j--)
-				if ('\\' != ln.buf[j])
-					break;
-
-			if ( ! ((pos - 2 - j) % 2))
-				continue;
-
-			comment = 1;
-			pos -= 2;
-			for (; pos > 0; --pos) {
-				if (ln.buf[pos - 1] != ' ')
-					break;
-				if (pos > 2 && ln.buf[pos - 2] == '\\')
-					break;
 			}
-			continue;
-		} 
-
-		/* Handle escaped `\\n' newlines. */
-
-		if (pos > 0 && 0 == comment && '\\' == ln.buf[pos - 1]) {
-			for (j = pos - 1; j >= 0; j--)
-				if ('\\' != ln.buf[j])
-					break;
-			if ( ! ((pos - j) % 2)) {
-				pos--;
-				lnn++;
+			/* Found an escape and at least one other character. */
+			if ('\n' == blk.buf[i + 1]) {
+				/* Escaped newlines are skipped over */
+				i += 2;
+				++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;
+					}
+				}
+				/* 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++];
 		}
 
-		ln.buf[pos] = 0;
-		lnn++;
+ 		if (pos >= (int)ln.sz)
+			if (! resize_buf(&ln, 256))
+				goto bailout;
+		ln.buf[pos] = '\0';
 
-		/* If unset, assign parser in pset(). */
+		/*
+		 * 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.
+		 */
 
-		if ( ! (man || mdoc) && ! pset(ln.buf, pos, curp, &man, &mdoc))
+		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;
 
-		pos = comment = 0;
+		/*
+		 * 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.
+		 */
 
-		/* Pass down into parsers. */
+		if ( ! (man || mdoc))
+			if ( ! pset(ln.buf + of, pos - of, curp, &man, &mdoc))
+				goto bailout;
 
-		if (man && ! man_parseln(man, lnn, ln.buf))
+		/* 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, ln.buf))
+		if (mdoc && ! mdoc_parseln(mdoc, lnn_start, ln.buf, of))
 			goto bailout;
 	}
 
@@ -456,10 +570,14 @@ fdesc(struct curparse *curp)
 		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). */
 
@@ -467,12 +585,24 @@ fdesc(struct curparse *curp)
 		switch (curp->outtype) {
 		case (OUTT_XHTML):
 			curp->outdata = xhtml_alloc(curp->outopts);
-			curp->outman = html_man;
-			curp->outmdoc = html_mdoc;
-			curp->outfree = html_free;
 			break;
 		case (OUTT_HTML):
 			curp->outdata = html_alloc(curp->outopts);
+			break;
+		case (OUTT_ASCII):
+			curp->outdata = ascii_alloc(curp->outopts);
+			break;
+		case (OUTT_PS):
+			curp->outdata = ps_alloc();
+			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;
@@ -481,14 +611,15 @@ fdesc(struct curparse *curp)
 			curp->outman = tree_man;
 			curp->outmdoc = tree_mdoc;
 			break;
-		case (OUTT_LINT):
-			break;
-		default:
-			curp->outdata = ascii_alloc(80);
+		case (OUTT_ASCII):
+			/* FALLTHROUGH */
+		case (OUTT_PS):
 			curp->outman = terminal_man;
 			curp->outmdoc = terminal_mdoc;
 			curp->outfree = terminal_free;
 			break;
+		default:
+			break;
 		}
 	}
 
@@ -500,20 +631,19 @@ fdesc(struct curparse *curp)
 		(*curp->outmdoc)(curp->outdata, mdoc);
 
  cleanup:
-	if (curp->mdoc) {
-		mdoc_free(curp->mdoc);
-		curp->mdoc = NULL;
-	}
-	if (curp->man) {
-		man_free(curp->man);
-		curp->man = NULL;
-	}
+	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:
@@ -536,7 +666,7 @@ pset(const char *buf, int pos, struct curparse *curp,
 	 * default to -man, which is more lenient.
 	 */
 
-	if (buf[0] == '.') {
+	if ('.' == buf[0] || '\'' == buf[0]) {
 		for (i = 1; buf[i]; i++)
 			if (' ' != buf[i] && '\t' != buf[i])
 				break;
@@ -613,6 +743,8 @@ toptions(struct curparse *curp, char *arg)
 		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 {
 		fprintf(stderr, "%s: Bad argument\n", arg);
 		return(0);
@@ -696,42 +828,34 @@ woptions(int *wflags, char *arg)
 }
 
 
-/* ARGSUSED */
 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 *curp;
+	struct curparse *cp;
 
-	curp = (struct curparse *)arg;
+	cp = (struct curparse *)arg;
 
-	(void)fprintf(stderr, "%s:%d:%d: error: %s\n", 
-			curp->file, line, col + 1, msg);
+	if (t <= MANDOCERR_ERROR) {
+		if ( ! (cp->wflags & WARN_WALL))
+			return(1);
+		with_warning = 1;
+	} else
+		with_error = 1;
 
-	with_error = 1;
+	fprintf(stderr, "%s:%d:%d: %s", cp->file, 
+			ln, col + 1, mandocerrs[t]);
 
-	return(0);
-}
+	if (msg)
+		fprintf(stderr, ": %s", msg);
 
+	fputc('\n', stderr);
 
-static int
-mwarn(void *arg, int line, int col, const char *msg)
-{
-	struct curparse *curp;
-
-	curp = (struct curparse *)arg;
-
-	if ( ! (curp->wflags & WARN_WALL))
-		return(1);
-
-	(void)fprintf(stderr, "%s:%d:%d: warning: %s\n", 
-			curp->file, line, col + 1, msg);
-
-	with_warning = 1;
-	if (curp->wflags & WARN_WERR) {
+	/* This is superfluous, but whatever. */
+	if (t > MANDOCERR_ERROR)
+		return(0);
+	if (cp->wflags & WARN_WERR) {
 		with_error = 1;
 		return(0);
 	}
-
 	return(1);
 }
-