Strip trailing, unescaped whitespace from free-form, non-literal lines (like groff).

4 files changed, 45 insertions, 24 deletions

diff --git a/man.7 b/man.7
index 5a3b22dc..a9733a11 100644
--- a/man.7
+++ b/man.7
@@ -126,10 +126,13 @@ and
 .Sq \ef
 attributes are forgotten when entering or exiting a macro block.
 .Ss Whitespace
-Unless specifically escaped, consecutive blocks of whitespace are pruned
-from input.  These are later re-added, if applicable, by a front-end
-utility such as
-.Xr mandoc 1 .
+In free-form lines, whitespace is preserved within a line; un-escaped
+trailing spaces are stripped from input (unless in a literal context).
+Blank free-form lines, which may include spaces, are permitted and
+rendered as an empty line.
+.Pp
+In macro lines, whitespace delimits arguments and is discarded.  If
+arguments are quoted, whitespace within the quotes is retained.
 .Ss Dates
 The
 .Sx \&TH
diff --git a/man.c b/man.c
index 828698d7..28c72df6 100644
--- a/man.c
+++ b/man.c
@@ -395,16 +395,30 @@ man_ptext(struct man *m, int line, char *buf)
 		goto descope;
 	}
 
-	/* Warn if the last un-escaped character is whitespace. */
+	/* 
+	 * Warn if the last un-escaped character is whitespace. Then
+	 * strip away the remaining spaces (tabs stay!).   
+	 */
 
 	i = (int)strlen(buf);
 	assert(i);
 
-	if (' ' == buf[i - 1] || '\t' == buf[i - 1])
-		if (1 == i || ('\\' != buf[i - 2]))
+	if (' ' == buf[i - 1] || '\t' == buf[i - 1]) {
+		assert(i > 1);
+		if ('\\' != buf[i - 2])
 			if ( ! man_pwarn(m, line, i - 1, WTSPACE))
 				return(0);
 
+		for (--i; i && ' ' == buf[i]; i--)
+			/* Spin back to non-space. */ ;
+
+		/* Jump ahead of escaped whitespace. */
+		assert(i);
+		i += '\\' == buf[i] ? 2 : 1;
+
+		buf[i] = '\0';
+	}
+
 	if ( ! man_word_alloc(m, line, 0, buf))
 		return(0);
 
diff --git a/mdoc.7 b/mdoc.7
index 1b807efa..d9c6b757 100644
--- a/mdoc.7
+++ b/mdoc.7
@@ -181,23 +181,13 @@ and
 .Sq \e*(Ba
 .Pq vertical bar .
 .Ss Whitespace
-In non-literal free-form lines, consecutive blocks of whitespace are
-pruned from input and added later in the output filter, if applicable:
-.Bd -literal -offset indent
-These     spaces   are    pruned       from    input.
-\&.Bd \-literal
-These         are              not.
-\&.Ed
-.Ed
+In free-form lines, whitespace is preserved within a line; un-escaped
+trailing spaces are stripped from input (unless in a literal context).
+Blank free-form lines, which may include spaces, are only permitted
+within literal contexts.
 .Pp
 In macro lines, whitespace delimits arguments and is discarded.  If
 arguments are quoted, whitespace within the quotes is retained.
-.Pp
-Blank lines are only permitted within literal contexts, as are lines
-containing only whitespace.  Tab characters are only acceptable when
-delimiting
-.Sq \&Bl \-column
-or when in a literal context.
 .Ss Quotation
 Macro arguments may be quoted with a double-quote to group
 space-delimited terms or to retain blocks of whitespace.  A quoted
diff --git a/mdoc.c b/mdoc.c
index 9bb391a2..1a6d23f8 100644
--- a/mdoc.c
+++ b/mdoc.c
@@ -665,16 +665,30 @@ mdoc_ptext(struct mdoc *m, int line, char *buf)
 		return(1);
 	}
 
-	/* Warn if the last un-escaped character is whitespace. */
+	/* 
+	 * Warn if the last un-escaped character is whitespace. Then
+	 * strip away the remaining spaces (tabs stay!).   
+	 */
 
 	i = (int)strlen(buf);
 	assert(i);
 
-	if (' ' == buf[i - 1] || '\t' == buf[i - 1])
-		if (1 == i || ('\\' != buf[i - 2]))
+	if (' ' == buf[i - 1] || '\t' == buf[i - 1]) {
+		assert(i > 1);
+		if ('\\' != buf[i - 2])
 			if ( ! mdoc_pwarn(m, line, i - 1, ETAILWS))
 				return(0);
 
+		for (--i; i && ' ' == buf[i]; i--)
+			/* Spin back to non-space. */ ;
+
+		/* Jump ahead of escaped whitespace. */
+		assert(i);
+		i += '\\' == buf[i] ? 2 : 1;
+
+		buf[i] = '\0';
+	}
+
 	/* Allocate the whole word. */
 
 	return(mdoc_word_alloc(m, line, 0, buf));