From c12bf4054c37b88eb37f7c59631d55ae512bcd29 Mon Sep 17 00:00:00 2001
From: Matthew Holt <mholt@users.noreply.github.com>
Date: Sat, 28 Sep 2019 21:18:36 -0600
Subject: caddyfile: Fix lexer behavior with regards to escaped newlines

Newlines (\n) can be escaped outside of quoted areas and the newline
will be treated as whitespace but not as an actual line break. Escaping
newlines inside a quoted area is not necessary, and because quotes
trigger literal interpretation of the contents, the escaping backslash
will be parsed as a literal backslash, and the newline will not be
escaped.

Caveat: When a newline is escaped, tokens after it until an unescaped
newline will appear to the parser be on the same line as the initial
token after the last unescaped newline. This may technically lead to
some false line numbers if errors are given, but escaped newlines are
counted so that the next token after an unescaped newline is correct.

See #2766
---
 caddyconfig/caddyfile/lexer.go      | 45 ++++++++++++++++++++++---------------
 caddyconfig/caddyfile/lexer_test.go | 42 +++++++++++++++++++++++++++++++---
 2 files changed, 66 insertions(+), 21 deletions(-)

(limited to 'caddyconfig')

diff --git a/caddyconfig/caddyfile/lexer.go b/caddyconfig/caddyfile/lexer.go
index efe648d..c0b6e1d 100755
--- a/caddyconfig/caddyfile/lexer.go
+++ b/caddyconfig/caddyfile/lexer.go
@@ -26,9 +26,10 @@ type (
 	// are separated by whitespace. A word can be enclosed
 	// in quotes if it contains whitespace.
 	lexer struct {
-		reader *bufio.Reader
-		token  Token
-		line   int
+		reader       *bufio.Reader
+		token        Token
+		line         int
+		skippedLines int
 	}
 
 	// Token represents a single parsable unit.
@@ -91,27 +92,30 @@ func (l *lexer) next() bool {
 			panic(err)
 		}
 
+		if !escaped && ch == '\\' {
+			escaped = true
+			continue
+		}
+
 		if quoted {
-			if !escaped {
-				if ch == '\\' {
-					escaped = true
-					continue
-				} else if ch == '"' {
+			if escaped {
+				// all is literal in quoted area,
+				// so only escape quotes
+				if ch != '"' {
+					val = append(val, '\\')
+				}
+				escaped = false
+			} else {
+				if ch == '"' {
 					quoted = false
 					return makeToken()
 				}
 			}
 			if ch == '\n' {
-				l.line++
-			}
-			if escaped {
-				// only escape quotes and newlines
-				if ch != '"' && ch != '\n' {
-					val = append(val, '\\')
-				}
+				l.line += 1 + l.skippedLines
+				l.skippedLines = 0
 			}
 			val = append(val, ch)
-			escaped = false
 			continue
 		}
 
@@ -120,7 +124,13 @@ func (l *lexer) next() bool {
 				continue
 			}
 			if ch == '\n' {
-				l.line++
+				if escaped {
+					l.skippedLines++
+					escaped = false
+				} else {
+					l.line += 1 + l.skippedLines
+					l.skippedLines = 0
+				}
 				comment = false
 			}
 			if len(val) > 0 {
@@ -132,7 +142,6 @@ func (l *lexer) next() bool {
 		if ch == '#' {
 			comment = true
 		}
-
 		if comment {
 			continue
 		}
diff --git a/caddyconfig/caddyfile/lexer_test.go b/caddyconfig/caddyfile/lexer_test.go
index f9a843c..ce2e147 100755
--- a/caddyconfig/caddyfile/lexer_test.go
+++ b/caddyconfig/caddyfile/lexer_test.go
@@ -96,13 +96,49 @@ func TestLexer(t *testing.T) {
 			},
 		},
 		{
-			input: "A \"newline \\\ninside\" quotes",
+			input: "An escaped \"newline\\\ninside\" quotes",
 			expected: []Token{
-				{Line: 1, Text: "A"},
-				{Line: 1, Text: "newline \ninside"},
+				{Line: 1, Text: "An"},
+				{Line: 1, Text: "escaped"},
+				{Line: 1, Text: "newline\\\ninside"},
 				{Line: 2, Text: "quotes"},
 			},
 		},
+		{
+			input: "An escaped newline\\\noutside quotes",
+			expected: []Token{
+				{Line: 1, Text: "An"},
+				{Line: 1, Text: "escaped"},
+				{Line: 1, Text: "newline"},
+				{Line: 1, Text: "outside"},
+				{Line: 1, Text: "quotes"},
+			},
+		},
+		{
+			input: "line1\\\nescaped\nline2\nline3",
+			expected: []Token{
+				{Line: 1, Text: "line1"},
+				{Line: 1, Text: "escaped"},
+				{Line: 3, Text: "line2"},
+				{Line: 4, Text: "line3"},
+			},
+		},
+		{
+			input: "line1\\\nescaped1\\\nescaped2\nline4\nline5",
+			expected: []Token{
+				{Line: 1, Text: "line1"},
+				{Line: 1, Text: "escaped1"},
+				{Line: 1, Text: "escaped2"},
+				{Line: 4, Text: "line4"},
+				{Line: 5, Text: "line5"},
+			},
+		},
+		{
+			input: `"unescapable\ in quotes"`,
+			expected: []Token{
+				{Line: 1, Text: `unescapable\ in quotes`},
+			},
+		},
 		{
 			input: `"don't\escape"`,
 			expected: []Token{
-- 
cgit v1.2.3