From ede644560890e63f58ed159bab1a28ed7f3f888b Mon Sep 17 00:00:00 2001
From: Thomas Pelletier <pelletier.thomas@gmail.com>
Date: Tue, 30 Nov 2021 10:44:48 -0500
Subject: [PATCH] Decoder: flag bad \r in literal multiline strings (#687)

Fixes #685
---
 scanner.go          | 48 ++++++++++++++++++++++++++++-----------------
 unmarshaler_test.go | 12 ++++++++++--
 2 files changed, 40 insertions(+), 20 deletions(-)

diff --git a/scanner.go b/scanner.go
index 4884713..bb445fa 100644
--- a/scanner.go
+++ b/scanner.go
@@ -76,30 +76,42 @@ func scanMultilineLiteralString(b []byte) ([]byte, []byte, error) {
 	// mll-char = %x09 / %x20-26 / %x28-7E / non-ascii
 	// mll-quotes = 1*2apostrophe
 	for i := 3; i < len(b); {
-		if scanFollowsMultilineLiteralStringDelimiter(b[i:]) {
-			i += 3
+		switch b[i] {
+		case '\'':
+			if scanFollowsMultilineLiteralStringDelimiter(b[i:]) {
+				i += 3
 
-			// At that point we found 3 apostrophe, and i is the
-			// index of the byte after the third one. The scanner
-			// needs to be eager, because there can be an extra 2
-			// apostrophe that can be accepted at the end of the
-			// string.
+				// At that point we found 3 apostrophe, and i is the
+				// index of the byte after the third one. The scanner
+				// needs to be eager, because there can be an extra 2
+				// apostrophe that can be accepted at the end of the
+				// string.
+
+				if i >= len(b) || b[i] != '\'' {
+					return b[:i], b[i:], nil
+				}
+				i++
+
+				if i >= len(b) || b[i] != '\'' {
+					return b[:i], b[i:], nil
+				}
+				i++
+
+				if i < len(b) && b[i] == '\'' {
+					return nil, nil, newDecodeError(b[i-3:i+1], "''' not allowed in multiline literal string")
+				}
 
-			if i >= len(b) || b[i] != '\'' {
 				return b[:i], b[i:], nil
 			}
-			i++
-
-			if i >= len(b) || b[i] != '\'' {
-				return b[:i], b[i:], nil
+		case '\r':
+			if len(b) < i+2 {
+				return nil, nil, newDecodeError(b[len(b):], `need a \n after \r`)
 			}
-			i++
-
-			if i < len(b) && b[i] == '\'' {
-				return nil, nil, newDecodeError(b[i-3:i+1], "''' not allowed in multiline literal string")
+			if b[i+1] != '\n' {
+				return nil, nil, newDecodeError(b[i:i+2], `need a \n after \r`)
 			}
-
-			return b[:i], b[i:], nil
+			i += 2 // skip the \n
+			continue
 		}
 		size := utf8ValidNext(b[i:])
 		if size == 0 {
diff --git a/unmarshaler_test.go b/unmarshaler_test.go
index b10d703..6f0b910 100644
--- a/unmarshaler_test.go
+++ b/unmarshaler_test.go
@@ -571,7 +571,7 @@ func TestUnmarshal(t *testing.T) {
 		},
 		{
 			desc:  "multiline basic string with windows newline",
-			input: "A = \"\"\"\r\nTest\"\"\"",
+			input: "A = \"\"\"\r\nTe\r\nst\"\"\"",
 			gen: func() test {
 				type doc struct {
 					A string
@@ -579,7 +579,7 @@ func TestUnmarshal(t *testing.T) {
 
 				return test{
 					target:   &doc{},
-					expected: &doc{A: "Test"},
+					expected: &doc{A: "Te\r\nst"},
 				}
 			},
 		},
@@ -2652,6 +2652,14 @@ world'`,
 			desc: `carriage return inside literal string`,
 			data: "A = '\r'",
 		},
+		{
+			desc: `carriage return inside multiline literal string`,
+			data: "a='''\r'''",
+		},
+		{
+			desc: `carriage return at trail of multiline literal string`,
+			data: "a='''\r",
+		},
 		{
 			desc: `carriage return in comment`,
 			data: "# this is a test\ra=1",