===================================================================
RCS file: /cvs/mandoc/term.c,v
retrieving revision 1.211
retrieving revision 1.220
diff -u -p -r1.211 -r1.220
--- mandoc/term.c	2013/12/22 23:34:13	1.211
+++ mandoc/term.c	2014/04/05 21:18:19	1.220
@@ -1,7 +1,7 @@
-/*	$Id: term.c,v 1.211 2013/12/22 23:34:13 schwarze Exp $ */
+/*	$Id: term.c,v 1.220 2014/04/05 21:18:19 schwarze Exp $ */
 /*
  * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
- * Copyright (c) 2010, 2011, 2012, 2013 Ingo Schwarze <schwarze@openbsd.org>
+ * Copyright (c) 2010-2014 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
@@ -23,12 +23,12 @@
 
 #include <assert.h>
 #include <ctype.h>
-#include <stdint.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
 
 #include "mandoc.h"
+#include "mandoc_aux.h"
 #include "out.h"
 #include "term.h"
 #include "main.h"
@@ -120,7 +120,12 @@ term_flushln(struct termp *p)
 	 * First, establish the maximum columns of "visible" content.
 	 * This is usually the difference between the right-margin and
 	 * an indentation, but can be, for tagged lists or columns, a
-	 * small set of values. 
+	 * small set of values.
+	 *
+	 * The following unsigned-signed subtractions look strange,
+	 * but they are actually correct.  If the int p->overstep
+	 * is negative, it gets sign extended.  Subtracting that
+	 * very large size_t effectively adds a small number to dv.
 	 */
 	assert  (p->rmargin >= p->offset);
 	dv     = p->rmargin - p->offset;
@@ -174,9 +179,17 @@ term_flushln(struct termp *p)
 			/* Regular word. */
 			/* Break at the hyphen point if we overrun. */
 			if (vend > vis && vend < bp && 
-					ASCII_HYPH == p->buf[j])
+			    (ASCII_HYPH == p->buf[j] ||
+			     ASCII_BREAK == p->buf[j]))
 				jhy = j;
 
+			/*
+			 * Hyphenation now decided, put back a real
+			 * hyphen such that we get the correct width.
+			 */
+			if (ASCII_HYPH == p->buf[j])
+				p->buf[j] = '-';
+
 			vend += (*p->width)(p, p->buf[j]);
 		}
 
@@ -199,7 +212,11 @@ term_flushln(struct termp *p)
 			if (0 < ntab)
 				vbl += ntab * p->tabwidth;
 
-			/* Remove the p->overstep width. */
+			/*
+			 * Remove the p->overstep width.
+			 * Again, if p->overstep is negative,
+			 * sign extension does the right thing.
+			 */
 
 			bp += (size_t)p->overstep;
 			p->overstep = 0;
@@ -224,6 +241,8 @@ term_flushln(struct termp *p)
 				vbl += (*p->width)(p, ' ');
 				continue;
 			}
+			if (ASCII_BREAK == p->buf[i])
+				continue;
 
 			/*
 			 * Now we definitely know there will be
@@ -236,12 +255,6 @@ term_flushln(struct termp *p)
 				vbl = 0;
 			}
 
-			if (ASCII_HYPH == p->buf[i]) {
-				(*p->letter)(p, '-');
-				p->viscol += (*p->width)(p, '-');
-				continue;
-			}
-
 			(*p->letter)(p, p->buf[i]);
 			if (8 == p->buf[i])
 				p->viscol -= (*p->width)(p, p->buf[i-1]);
@@ -275,8 +288,10 @@ term_flushln(struct termp *p)
 		 * If we have overstepped the margin, temporarily move
 		 * it to the right and flag the rest of the line to be
 		 * shorter.
+		 * If there is a request to keep the columns together,
+		 * allow negative overstep when the column is not full.
 		 */
-		if (p->overstep < 0)
+		if (p->trailspace && p->overstep < 0)
 			p->overstep = 0;
 		return;
 
@@ -396,6 +411,7 @@ term_fontpop(struct termp *p)
 void
 term_word(struct termp *p, const char *word)
 {
+	const char	 nbrsp[2] = { ASCII_NBRSP, 0 };
 	const char	*seq, *cp;
 	char		 c;
 	int		 sz, uc;
@@ -418,7 +434,7 @@ term_word(struct termp *p, const char *word)
 	else
 		p->flags |= TERMP_NOSPACE;
 
-	p->flags &= ~(TERMP_SENTENCE | TERMP_IGNDELIM);
+	p->flags &= ~TERMP_SENTENCE;
 
 	while ('\0' != *word) {
 		if ('\\' != *word) {
@@ -427,7 +443,15 @@ term_word(struct termp *p, const char *word)
 				word++;
 				continue;
 			}
-			ssz = strcspn(word, "\\");
+			if (TERMP_NBRWORD & p->flags) {
+				if (' ' == *word) {
+					encode(p, nbrsp, 1);
+					word++;
+					continue;
+				}
+				ssz = strcspn(word, "\\ ");
+			} else
+				ssz = strcspn(word, "\\");
 			encode(p, word, ssz);
 			word += (int)ssz;
 			continue;
@@ -502,6 +526,7 @@ term_word(struct termp *p, const char *word)
 			break;
 		}
 	}
+	p->flags &= ~TERMP_NBRWORD;
 }
 
 static void
@@ -598,6 +623,36 @@ encode(struct termp *p, const char *word, size_t sz)
 	}
 }
 
+void
+term_setwidth(struct termp *p, const char *wstr)
+{
+	struct roffsu	 su;
+	size_t		 width;
+	int		 iop;
+
+	iop = 0;
+	width = 0;
+	if (NULL != wstr) {
+		switch (*wstr) {
+		case ('+'):
+			iop = 1;
+			wstr++;
+			break;
+		case ('-'):
+			iop = -1;
+			wstr++;
+			break;
+		default:
+			break;
+		}
+		if (a2roffsu(wstr, &su, SCALE_MAX))
+			width = term_hspan(p, &su);
+		else
+			iop = 0;
+	}
+	(*p->setwidth)(p, iop, width);
+}
+
 size_t
 term_len(const struct termp *p, size_t sz)
 {
@@ -623,7 +678,8 @@ term_strlen(const struct termp *p, const char *cp)
 	int		 ssz, skip, c;
 	const char	*seq, *rhs;
 	enum mandoc_esc	 esc;
-	static const char rej[] = { '\\', ASCII_HYPH, ASCII_NBRSP, '\0' };
+	static const char rej[] = { '\\', ASCII_NBRSP, ASCII_HYPH,
+			ASCII_BREAK, '\0' };
 
 	/*
 	 * Account for escaped sequences within string length
@@ -638,7 +694,6 @@ term_strlen(const struct termp *p, const char *cp)
 		for (i = 0; i < rsz; i++)
 			sz += cond_width(p, *cp++, &skip);
 
-		c = 0;
 		switch (*cp) {
 		case ('\\'):
 			cp++;
@@ -712,6 +767,8 @@ term_strlen(const struct termp *p, const char *cp)
 		case (ASCII_HYPH):
 			sz += cond_width(p, '-', &skip);
 			cp++;
+			/* FALLTHROUGH */
+		case (ASCII_BREAK):
 			break;
 		default:
 			break;