caddyfile: Fix heredoc fuzz crasher, drop trailing newline (#5404)

Co-authored-by: Mohammed Al Sahaf <msaa1990@gmail.com>

2 files changed, 76 insertions, 16 deletions

diff --git a/caddyconfig/caddyfile/lexer.go b/caddyconfig/caddyfile/lexer.go
index ba8b879..1f531f4 100644
--- a/caddyconfig/caddyfile/lexer.go
+++ b/caddyconfig/caddyfile/lexer.go
@@ -284,15 +284,17 @@ func (l *lexer) next() (bool, error) {
 // and processes the text to strip leading whitespace, returning the final
 // value without the leading whitespace.
 func (l *lexer) finalizeHeredoc(val []rune, marker string) ([]rune, error) {
+	stringVal := string(val)
+
 	// find the last newline of the heredoc, which is where the contents end
-	lastNewline := strings.LastIndex(string(val), "\n")
+	lastNewline := strings.LastIndex(stringVal, "\n")
 
 	// collapse the content, then split into separate lines
-	lines := strings.Split(string(val[:lastNewline+1]), "\n")
+	lines := strings.Split(stringVal[:lastNewline+1], "\n")
 
 	// figure out how much whitespace we need to strip from the front of every line
 	// by getting the string that precedes the marker, on the last line
-	paddingToStrip := string(val[lastNewline+1 : len(val)-len(marker)])
+	paddingToStrip := stringVal[lastNewline+1 : len(stringVal)-len(marker)]
 
 	// iterate over each line and strip the whitespace from the front
 	var out string
@@ -310,6 +312,11 @@ func (l *lexer) finalizeHeredoc(val []rune, marker string) ([]rune, error) {
 		out += strings.ReplaceAll(lineText[len(paddingToStrip):]+"\n", "\r", "")
 	}
 
+	// Remove the trailing newline from the loop
+	if len(out) > 0 && out[len(out)-1] == '\n' {
+		out = out[:len(out)-1]
+	}
+
 	// return the final value
 	return []rune(out), nil
 }
@@ -340,9 +347,9 @@ func (t Token) NumLineBreaks() int {
 	lineBreaks := strings.Count(t.Text, "\n")
 	if t.wasQuoted == '<' {
 		// heredocs have an extra linebreak because the opening
-		// delimiter is on its own line and is not included in
-		// the token Text itself
-		lineBreaks++
+		// delimiter is on its own line and is not included in the
+		// token Text itself, and the trailing newline is removed.
+		lineBreaks += 2
 	}
 	return lineBreaks
 }
diff --git a/caddyconfig/caddyfile/lexer_test.go b/caddyconfig/caddyfile/lexer_test.go
index 3c7e157..801d81e 100644
--- a/caddyconfig/caddyfile/lexer_test.go
+++ b/caddyconfig/caddyfile/lexer_test.go
@@ -256,7 +256,7 @@ EOF same-line-arg
 	`),
 			expected: []Token{
 				{Line: 1, Text: `heredoc`},
-				{Line: 1, Text: "content\n"},
+				{Line: 1, Text: "content"},
 				{Line: 3, Text: `same-line-arg`},
 			},
 		},
@@ -267,18 +267,40 @@ VERY-LONG-MARKER same-line-arg
 	`),
 			expected: []Token{
 				{Line: 1, Text: `heredoc`},
-				{Line: 1, Text: "content\n"},
+				{Line: 1, Text: "content"},
 				{Line: 3, Text: `same-line-arg`},
 			},
 		},
 		{
 			input: []byte(`heredoc <<EOF
+extra-newline
+
+EOF same-line-arg
+	`),
+			expected: []Token{
+				{Line: 1, Text: `heredoc`},
+				{Line: 1, Text: "extra-newline\n"},
+				{Line: 4, Text: `same-line-arg`},
+			},
+		},
+		{
+			input: []byte(`heredoc <<EOF
+		EOF same-line-arg
+	`),
+			expected: []Token{
+				{Line: 1, Text: `heredoc`},
+				{Line: 1, Text: ""},
+				{Line: 2, Text: `same-line-arg`},
+			},
+		},
+		{
+			input: []byte(`heredoc <<EOF
 	content
 	EOF same-line-arg
 	`),
 			expected: []Token{
 				{Line: 1, Text: `heredoc`},
-				{Line: 1, Text: "content\n"},
+				{Line: 1, Text: "content"},
 				{Line: 3, Text: `same-line-arg`},
 			},
 		},
@@ -294,7 +316,7 @@ VERY-LONG-MARKER same-line-arg
 			expected: []Token{
 				{Line: 1, Text: `prev-line`},
 				{Line: 2, Text: `heredoc`},
-				{Line: 2, Text: "\tmulti\n\tline\n\tcontent\n"},
+				{Line: 2, Text: "\tmulti\n\tline\n\tcontent"},
 				{Line: 6, Text: `same-line-arg`},
 				{Line: 7, Text: `next-line`},
 			},
@@ -313,6 +335,37 @@ VERY-LONG-MARKER same-line-arg
 			},
 		},
 		{
+			input: []byte(`heredoc <<s
+			�
+			s
+	`),
+			expected: []Token{
+				{Line: 1, Text: `heredoc`},
+				{Line: 1, Text: "�"},
+			},
+		},
+		{
+			input: []byte("\u000Aheredoc \u003C\u003C\u0073\u0073\u000A\u00BF\u0057\u0001\u0000\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u003D\u001F\u000A\u0073\u0073\u000A\u00BF\u0057\u0001\u0000\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u003D\u001F\u000A\u00BF\u00BF\u0057\u0001\u0000\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u003D\u001F"),
+			expected: []Token{
+				{
+					Line: 2,
+					Text: "heredoc",
+				},
+				{
+					Line: 2,
+					Text: "\u00BF\u0057\u0001\u0000\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u003D\u001F",
+				},
+				{
+					Line: 5,
+					Text: "\u00BF\u0057\u0001\u0000\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u003D\u001F",
+				},
+				{
+					Line: 6,
+					Text: "\u00BF\u00BF\u0057\u0001\u0000\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u00FF\u003D\u001F",
+				},
+			},
+		},
+		{
 			input: []byte(`heredoc <<HERE SAME LINE
 	content
 	HERE same-line-arg
@@ -357,17 +410,17 @@ VERY-LONG-MARKER same-line-arg
 		actual, err := Tokenize(testCase.input, "")
 		if testCase.expectErr {
 			if err == nil {
-				t.Errorf("expected error, got actual: %v", actual)
+				t.Fatalf("expected error, got actual: %v", actual)
 				continue
 			}
 			if err.Error() != testCase.errorMessage {
-				t.Errorf("expected error '%v', got: %v", testCase.errorMessage, err)
+				t.Fatalf("expected error '%v', got: %v", testCase.errorMessage, err)
 			}
 			continue
 		}
 
 		if err != nil {
-			t.Errorf("%v", err)
+			t.Fatalf("%v", err)
 		}
 		lexerCompare(t, i, testCase.expected, actual)
 	}
@@ -375,17 +428,17 @@ VERY-LONG-MARKER same-line-arg
 
 func lexerCompare(t *testing.T, n int, expected, actual []Token) {
 	if len(expected) != len(actual) {
-		t.Errorf("Test case %d: expected %d token(s) but got %d", n, len(expected), len(actual))
+		t.Fatalf("Test case %d: expected %d token(s) but got %d", n, len(expected), len(actual))
 	}
 
 	for i := 0; i < len(actual) && i < len(expected); i++ {
 		if actual[i].Line != expected[i].Line {
-			t.Errorf("Test case %d token %d ('%s'): expected line %d but was line %d",
+			t.Fatalf("Test case %d token %d ('%s'): expected line %d but was line %d",
 				n, i, expected[i].Text, expected[i].Line, actual[i].Line)
 			break
 		}
 		if actual[i].Text != expected[i].Text {
-			t.Errorf("Test case %d token %d: expected text '%s' but was '%s'",
+			t.Fatalf("Test case %d token %d: expected text '%s' but was '%s'",
 				n, i, expected[i].Text, actual[i].Text)
 			break
 		}