Convert EOL to UNIX-style to render MD properly (#8925)

* Convert EOL to UNIX-style to render MD properly

* Update modules/markup/markdown/markdown.go

Co-Authored-By: zeripath <art27@cantab.net>

* Fix lint optimization

* Check for empty content before conversion

* Update modules/util/util.go

Co-Authored-By: zeripath <art27@cantab.net>

* Improved checks and tests

* Add paragraph render test

* Improve speed even more, improve tests

* Small improvement by @gary-kim

* Fix test for DOS

* More improvements

* Restart CI

1 files changed, 37 insertions, 0 deletions

diff --git a/modules/util/util.go b/modules/util/util.go
index 4203b5eb51..6d02b5f52f 100644
--- a/modules/util/util.go
+++ b/modules/util/util.go
@@ -5,6 +5,7 @@
 package util
 
 import (
+	"bytes"
 	"strings"
 )
 
@@ -63,3 +64,39 @@ func Min(a, b int) int {
 func IsEmptyString(s string) bool {
 	return len(strings.TrimSpace(s)) == 0
 }
+
+// NormalizeEOL will convert Windows (CRLF) and Mac (CR) EOLs to UNIX (LF)
+func NormalizeEOL(input []byte) []byte {
+	var right, left, pos int
+	if right = bytes.IndexByte(input, '\r'); right == -1 {
+		return input
+	}
+	length := len(input)
+	tmp := make([]byte, length)
+
+	// We know that left < length because otherwise right would be -1 from IndexByte.
+	copy(tmp[pos:pos+right], input[left:left+right])
+	pos += right
+	tmp[pos] = '\n'
+	left += right + 1
+	pos++
+
+	for left < length {
+		if input[left] == '\n' {
+			left++
+		}
+
+		right = bytes.IndexByte(input[left:], '\r')
+		if right == -1 {
+			copy(tmp[pos:], input[left:])
+			pos += length - left
+			break
+		}
+		copy(tmp[pos:pos+right], input[left:left+right])
+		pos += right
+		tmp[pos] = '\n'
+		left += right + 1
+		pos++
+	}
+	return tmp[:pos]
+}