From e429c1164ec68b154b7f06db2c2e04dd25bad094 Mon Sep 17 00:00:00 2001
From: zeripath <art27@cantab.net>
Date: Sun, 23 Aug 2020 14:15:29 +0100
Subject: [PATCH] Ensure that the detected charset order is set in chardet test
 (#12574)

TestToUTF8WithFallback is the cause of recurrent spurious test failures
even despite code to set the detected charset order.

The reason why this happens is because the preferred detected charset order
is not being initialised for these tests.

This PR simply ensures that this is set at the start of each test and would
allow different tests to be written to allow differing orders.

Replaces #12571
Close #12571

Signed-off-by: Andrew Thornton <art27@cantab.net>
---
 modules/charset/charset_test.go | 28 +++++++++++++++++++++++-----
 1 file changed, 23 insertions(+), 5 deletions(-)

diff --git a/modules/charset/charset_test.go b/modules/charset/charset_test.go
index 394a42c71..33f0c10a7 100644
--- a/modules/charset/charset_test.go
+++ b/modules/charset/charset_test.go
@@ -5,6 +5,7 @@
 package charset
 
 import (
+	"strings"
 	"testing"
 
 	"code.gitea.io/gitea/modules/setting"
@@ -12,6 +13,22 @@ import (
 	"github.com/stretchr/testify/assert"
 )
 
+func resetDefaultCharsetsOrder() {
+	defaultDetectedCharsetsOrder := make([]string, 0, len(setting.Repository.DetectedCharsetsOrder))
+	for _, charset := range setting.Repository.DetectedCharsetsOrder {
+		defaultDetectedCharsetsOrder = append(defaultDetectedCharsetsOrder, strings.ToLower(strings.TrimSpace(charset)))
+	}
+	setting.Repository.DetectedCharsetScore = map[string]int{}
+	i := 0
+	for _, charset := range defaultDetectedCharsetsOrder {
+		canonicalCharset := strings.ToLower(strings.TrimSpace(charset))
+		if _, has := setting.Repository.DetectedCharsetScore[canonicalCharset]; !has {
+			setting.Repository.DetectedCharsetScore[canonicalCharset] = i
+			i++
+		}
+	}
+}
+
 func TestRemoveBOMIfPresent(t *testing.T) {
 	res := RemoveBOMIfPresent([]byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba})
 	assert.Equal(t, []byte{0xc3, 0xa1, 0xc3, 0xa9, 0xc3, 0xad, 0xc3, 0xb3, 0xc3, 0xba}, res)
@@ -21,6 +38,7 @@ func TestRemoveBOMIfPresent(t *testing.T) {
 }
 
 func TestToUTF8WithErr(t *testing.T) {
+	resetDefaultCharsetsOrder()
 	var res string
 	var err error
 
@@ -76,6 +94,7 @@ func TestToUTF8WithErr(t *testing.T) {
 }
 
 func TestToUTF8WithFallback(t *testing.T) {
+	resetDefaultCharsetsOrder()
 	// "ABC"
 	res := ToUTF8WithFallback([]byte{0x41, 0x42, 0x43})
 	assert.Equal(t, []byte{0x41, 0x42, 0x43}, res)
@@ -116,7 +135,7 @@ func TestToUTF8WithFallback(t *testing.T) {
 }
 
 func TestToUTF8(t *testing.T) {
-
+	resetDefaultCharsetsOrder()
 	// Note: golang compiler seems so behave differently depending on the current
 	// locale, so some conversions might behave differently. For that reason, we don't
 	// depend on particular conversions but in expected behaviors.
@@ -165,6 +184,7 @@ func TestToUTF8(t *testing.T) {
 }
 
 func TestToUTF8DropErrors(t *testing.T) {
+	resetDefaultCharsetsOrder()
 	// "ABC"
 	res := ToUTF8DropErrors([]byte{0x41, 0x42, 0x43})
 	assert.Equal(t, []byte{0x41, 0x42, 0x43}, res)
@@ -204,6 +224,7 @@ func TestToUTF8DropErrors(t *testing.T) {
 }
 
 func TestDetectEncoding(t *testing.T) {
+	resetDefaultCharsetsOrder()
 	testSuccess := func(b []byte, expected string) {
 		encoding, err := DetectEncoding(b)
 		assert.NoError(t, err)
@@ -225,10 +246,7 @@ func TestDetectEncoding(t *testing.T) {
 	b = []byte{0x44, 0xe9, 0x63, 0x6f, 0x72, 0x0a}
 	encoding, err := DetectEncoding(b)
 	assert.NoError(t, err)
-	// due to a race condition in `chardet` library, it could either detect
-	// "ISO-8859-1" or "IS0-8859-2" here. Technically either is correct, so
-	// we accept either.
-	assert.Contains(t, encoding, "ISO-8859")
+	assert.Contains(t, encoding, "ISO-8859-1")
 
 	old := setting.Repository.AnsiCharset
 	setting.Repository.AnsiCharset = "placeholder"