Add option to disable ambiguous unicode characters detection (#28454)

* Close #24483
* Close #28123
* Close #23682
* Close #23149

(maybe more)

modules/charset/escape.go

@@ -8,11 +8,12 @@
 package charset
 
 import (
-	"bufio"
+	"html/template"
 	"io"
 	"strings"
 
 	"code.gitea.io/gitea/modules/log"
+	"code.gitea.io/gitea/modules/setting"
 	"code.gitea.io/gitea/modules/translation"
 )
 
@@ -20,20 +21,18 @@ import (
 const RuneNBSP = 0xa0
 
 // EscapeControlHTML escapes the unicode control sequences in a provided html document
-func EscapeControlHTML(text string, locale translation.Locale, allowed ...rune) (escaped *EscapeStatus, output string) {
+func EscapeControlHTML(html template.HTML, locale translation.Locale, allowed ...rune) (escaped *EscapeStatus, output template.HTML) {
 	sb := &strings.Builder{}
-	outputStream := &HTMLStreamerWriter{Writer: sb}
-	streamer := NewEscapeStreamer(locale, outputStream, allowed...).(*escapeStreamer)
-
-	if err := StreamHTML(strings.NewReader(text), streamer); err != nil {
-		streamer.escaped.HasError = true
-		log.Error("Error whilst escaping: %v", err)
-	}
-	return streamer.escaped, sb.String()
+	escaped, _ = EscapeControlReader(strings.NewReader(string(html)), sb, locale, allowed...) // err has been handled in EscapeControlReader
+	return escaped, template.HTML(sb.String())
 }
 
-// EscapeControlReaders escapes the unicode control sequences in a provided reader of HTML content and writer in a locale and returns the findings as an EscapeStatus and the escaped []byte
+// EscapeControlReader escapes the unicode control sequences in a provided reader of HTML content and writer in a locale and returns the findings as an EscapeStatus
 func EscapeControlReader(reader io.Reader, writer io.Writer, locale translation.Locale, allowed ...rune) (escaped *EscapeStatus, err error) {
+	if !setting.UI.AmbiguousUnicodeDetection {
+		_, err = io.Copy(writer, reader)
+		return &EscapeStatus{}, err
+	}
 	outputStream := &HTMLStreamerWriter{Writer: writer}
 	streamer := NewEscapeStreamer(locale, outputStream, allowed...).(*escapeStreamer)
 
@@ -43,41 +42,3 @@ func EscapeControlReader(reader io.Reader, writer io.Writer, locale translation.
 	}
 	return streamer.escaped, err
 }
-
-// EscapeControlStringReader escapes the unicode control sequences in a provided reader of string content and writer in a locale and returns the findings as an EscapeStatus and the escaped []byte. HTML line breaks are not inserted after every newline by this method.
-func EscapeControlStringReader(reader io.Reader, writer io.Writer, locale translation.Locale, allowed ...rune) (escaped *EscapeStatus, err error) {
-	bufRd := bufio.NewReader(reader)
-	outputStream := &HTMLStreamerWriter{Writer: writer}
-	streamer := NewEscapeStreamer(locale, outputStream, allowed...).(*escapeStreamer)
-
-	for {
-		line, rdErr := bufRd.ReadString('\n')
-		if len(line) > 0 {
-			if err := streamer.Text(line); err != nil {
-				streamer.escaped.HasError = true
-				log.Error("Error whilst escaping: %v", err)
-				return streamer.escaped, err
-			}
-		}
-		if rdErr != nil {
-			if rdErr != io.EOF {
-				err = rdErr
-			}
-			break
-		}
-	}
-	return streamer.escaped, err
-}
-
-// EscapeControlString escapes the unicode control sequences in a provided string and returns the findings as an EscapeStatus and the escaped string
-func EscapeControlString(text string, locale translation.Locale, allowed ...rune) (escaped *EscapeStatus, output string) {
-	sb := &strings.Builder{}
-	outputStream := &HTMLStreamerWriter{Writer: sb}
-	streamer := NewEscapeStreamer(locale, outputStream, allowed...).(*escapeStreamer)
-
-	if err := streamer.Text(text); err != nil {
-		streamer.escaped.HasError = true
-		log.Error("Error whilst escaping: %v", err)
-	}
-	return streamer.escaped, sb.String()
-}

modules/charset/escape_stream.go

@@ -64,7 +64,7 @@ func (e *escapeStreamer) Text(data string) error {
 			until, next = nextIdxs[0]+pos, nextIdxs[1]+pos
 		}
 
-		// from pos until until we know that the runes are not \r\t\n or even ' '
+		// from pos until we know that the runes are not \r\t\n or even ' '
 		runes := make([]rune, 0, next-until)
 		positions := make([]int, 0, next-until+1)
 

modules/charset/escape_test.go

@@ -4,11 +4,14 @@
 package charset
 
 import (
-	"reflect"
 	"strings"
 	"testing"
 
+	"code.gitea.io/gitea/modules/setting"
+	"code.gitea.io/gitea/modules/test"
 	"code.gitea.io/gitea/modules/translation"
+
+	"github.com/stretchr/testify/assert"
 )
 
 type escapeControlTest struct {
@@ -132,22 +135,8 @@ then resh (ר), and finally heh (ה) (which should appear leftmost).`,
 	},
 }
 
-func TestEscapeControlString(t *testing.T) {
-	for _, tt := range escapeControlTests {
-		t.Run(tt.name, func(t *testing.T) {
-			status, result := EscapeControlString(tt.text, &translation.MockLocale{})
-			if !reflect.DeepEqual(*status, tt.status) {
-				t.Errorf("EscapeControlString() status = %v, wanted= %v", status, tt.status)
-			}
-			if result != tt.result {
-				t.Errorf("EscapeControlString()\nresult= %v,\nwanted= %v", result, tt.result)
-			}
-		})
-	}
-}
-
 func TestEscapeControlReader(t *testing.T) {
-	// lets add some control characters to the tests
+	// add some control characters to the tests
 	tests := make([]escapeControlTest, 0, len(escapeControlTests)*3)
 	copy(tests, escapeControlTests)
 
@@ -169,29 +158,20 @@ func TestEscapeControlReader(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			input := strings.NewReader(tt.text)
 			output := &strings.Builder{}
-			status, err := EscapeControlReader(input, output, &translation.MockLocale{})
-			result := output.String()
-			if err != nil {
-				t.Errorf("EscapeControlReader(): err = %v", err)
-			}
-
-			if !reflect.DeepEqual(*status, tt.status) {
-				t.Errorf("EscapeControlReader() status = %v, wanted= %v", status, tt.status)
-			}
-			if result != tt.result {
-				t.Errorf("EscapeControlReader()\nresult= %v,\nwanted= %v", result, tt.result)
-			}
+			status, err := EscapeControlReader(strings.NewReader(tt.text), output, &translation.MockLocale{})
+			assert.NoError(t, err)
+			assert.Equal(t, tt.status, *status)
+			assert.Equal(t, tt.result, output.String())
 		})
 	}
 }
 
-func TestEscapeControlReader_panic(t *testing.T) {
-	bs := make([]byte, 0, 20479)
-	bs = append(bs, 'A')
-	for i := 0; i < 6826; i++ {
-		bs = append(bs, []byte("—")...)
-	}
-	_, _ = EscapeControlString(string(bs), &translation.MockLocale{})
+func TestSettingAmbiguousUnicodeDetection(t *testing.T) {
+	defer test.MockVariableValue(&setting.UI.AmbiguousUnicodeDetection, true)()
+	_, out := EscapeControlHTML("a test", &translation.MockLocale{})
+	assert.EqualValues(t, `a<span class="escaped-code-point" data-escaped="[U+00A0]"><span class="char"> </span></span>test`, out)
+	setting.UI.AmbiguousUnicodeDetection = false
+	_, out = EscapeControlHTML("a test", &translation.MockLocale{})
+	assert.EqualValues(t, `a test`, out)
 }

modules/git/command.go

@@ -14,7 +14,6 @@ import (
 	"os/exec"
 	"strings"
 	"time"
-	"unsafe"
 
 	"code.gitea.io/gitea/modules/git/internal" //nolint:depguard // only this file can use the internal type CmdArg, other files and packages should use AddXxx functions
 	"code.gitea.io/gitea/modules/log"
@@ -389,15 +388,11 @@ func (r *runStdError) IsExitCode(code int) bool {
 	return false
 }
 
-func bytesToString(b []byte) string {
-	return *(*string)(unsafe.Pointer(&b)) // that's what Golang's strings.Builder.String() does (go/src/strings/builder.go)
-}
-
 // RunStdString runs the command with options and returns stdout/stderr as string. and store stderr to returned error (err combined with stderr).
 func (c *Command) RunStdString(opts *RunOpts) (stdout, stderr string, runErr RunStdError) {
 	stdoutBytes, stderrBytes, err := c.RunStdBytes(opts)
-	stdout = bytesToString(stdoutBytes)
-	stderr = bytesToString(stderrBytes)
+	stdout = util.UnsafeBytesToString(stdoutBytes)
+	stderr = util.UnsafeBytesToString(stderrBytes)
 	if err != nil {
 		return stdout, stderr, &runStdError{err: err, stderr: stderr}
 	}
@@ -432,7 +427,7 @@ func (c *Command) RunStdBytes(opts *RunOpts) (stdout, stderr []byte, runErr RunS
 	err := c.Run(newOpts)
 	stderr = stderrBuf.Bytes()
 	if err != nil {
-		return nil, stderr, &runStdError{err: err, stderr: bytesToString(stderr)}
+		return nil, stderr, &runStdError{err: err, stderr: util.UnsafeBytesToString(stderr)}
 	}
 	// even if there is no err, there could still be some stderr output
 	return stdoutBuf.Bytes(), stderr, nil

modules/highlight/highlight.go

@@ -9,6 +9,7 @@ import (
 	"bytes"
 	"fmt"
 	gohtml "html"
+	"html/template"
 	"io"
 	"path/filepath"
 	"strings"
@@ -55,7 +56,7 @@ func NewContext() {
 }
 
 // Code returns a HTML version of code string with chroma syntax highlighting classes and the matched lexer name
-func Code(fileName, language, code string) (string, string) {
+func Code(fileName, language, code string) (output template.HTML, lexerName string) {
 	NewContext()
 
 	// diff view newline will be passed as empty, change to literal '\n' so it can be copied
@@ -65,7 +66,7 @@ func Code(fileName, language, code string) (string, string) {
 	}
 
 	if len(code) > sizeLimit {
-		return code, ""
+		return template.HTML(template.HTMLEscapeString(code)), ""
 	}
 
 	var lexer chroma.Lexer
@@ -102,13 +103,11 @@ func Code(fileName, language, code string) (string, string) {
 		cache.Add(fileName, lexer)
 	}
 
-	lexerName := formatLexerName(lexer.Config().Name)
-
-	return CodeFromLexer(lexer, code), lexerName
+	return CodeFromLexer(lexer, code), formatLexerName(lexer.Config().Name)
 }
 
 // CodeFromLexer returns a HTML version of code string with chroma syntax highlighting classes
-func CodeFromLexer(lexer chroma.Lexer, code string) string {
+func CodeFromLexer(lexer chroma.Lexer, code string) template.HTML {
 	formatter := html.New(html.WithClasses(true),
 		html.WithLineNumbers(false),
 		html.PreventSurroundingPre(true),
@@ -120,23 +119,23 @@ func CodeFromLexer(lexer chroma.Lexer, code string) string {
 	iterator, err := lexer.Tokenise(nil, code)
 	if err != nil {
 		log.Error("Can't tokenize code: %v", err)
-		return code
+		return template.HTML(template.HTMLEscapeString(code))
 	}
 	// style not used for live site but need to pass something
 	err = formatter.Format(htmlw, githubStyles, iterator)
 	if err != nil {
 		log.Error("Can't format code: %v", err)
-		return code
+		return template.HTML(template.HTMLEscapeString(code))
 	}
 
 	_ = htmlw.Flush()
 	// Chroma will add newlines for certain lexers in order to highlight them properly
 	// Once highlighted, strip them here, so they don't cause copy/paste trouble in HTML output
-	return strings.TrimSuffix(htmlbuf.String(), "\n")
+	return template.HTML(strings.TrimSuffix(htmlbuf.String(), "\n"))
 }
 
 // File returns a slice of chroma syntax highlighted HTML lines of code and the matched lexer name
-func File(fileName, language string, code []byte) ([]string, string, error) {
+func File(fileName, language string, code []byte) ([]template.HTML, string, error) {
 	NewContext()
 
 	if len(code) > sizeLimit {
@@ -183,14 +182,14 @@ func File(fileName, language string, code []byte) ([]string, string, error) {
 	tokensLines := chroma.SplitTokensIntoLines(iterator.Tokens())
 	htmlBuf := &bytes.Buffer{}
 
-	lines := make([]string, 0, len(tokensLines))
+	lines := make([]template.HTML, 0, len(tokensLines))
 	for _, tokens := range tokensLines {
 		iterator = chroma.Literator(tokens...)
 		err = formatter.Format(htmlBuf, githubStyles, iterator)
 		if err != nil {
 			return nil, "", fmt.Errorf("can't format code: %w", err)
 		}
-		lines = append(lines, htmlBuf.String())
+		lines = append(lines, template.HTML(htmlBuf.String()))
 		htmlBuf.Reset()
 	}
 
@@ -198,9 +197,9 @@ func File(fileName, language string, code []byte) ([]string, string, error) {
 }
 
 // PlainText returns non-highlighted HTML for code
-func PlainText(code []byte) []string {
+func PlainText(code []byte) []template.HTML {
 	r := bufio.NewReader(bytes.NewReader(code))
-	m := make([]string, 0, bytes.Count(code, []byte{'\n'})+1)
+	m := make([]template.HTML, 0, bytes.Count(code, []byte{'\n'})+1)
 	for {
 		content, err := r.ReadString('\n')
 		if err != nil && err != io.EOF {
@@ -210,7 +209,7 @@ func PlainText(code []byte) []string {
 		if content == "" && err == io.EOF {
 			break
 		}
-		s := gohtml.EscapeString(content)
+		s := template.HTML(gohtml.EscapeString(content))
 		m = append(m, s)
 	}
 	return m

modules/highlight/highlight_test.go

@@ -4,21 +4,36 @@
 package highlight
 
 import (
+	"html/template"
 	"strings"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
 )
 
-func lines(s string) []string {
-	return strings.Split(strings.ReplaceAll(strings.TrimSpace(s), `\n`, "\n"), "\n")
+func lines(s string) (out []template.HTML) {
+	// "" => [], "a" => ["a"], "a\n" => ["a\n"], "a\nb" => ["a\n", "b"] (each line always includes EOL "\n" if it exists)
+	out = make([]template.HTML, 0)
+	s = strings.ReplaceAll(strings.ReplaceAll(strings.TrimSpace(s), "\n", ""), `\n`, "\n")
+	for {
+		if p := strings.IndexByte(s, '\n'); p != -1 {
+			out = append(out, template.HTML(s[:p+1]))
+			s = s[p+1:]
+		} else {
+			break
+		}
+	}
+	if s != "" {
+		out = append(out, template.HTML(s))
+	}
+	return out
 }
 
 func TestFile(t *testing.T) {
 	tests := []struct {
 		name      string
 		code      string
-		want      []string
+		want      []template.HTML
 		lexerName string
 	}{
 		{
@@ -99,10 +114,7 @@ c=2
 		t.Run(tt.name, func(t *testing.T) {
 			out, lexerName, err := File(tt.name, "", []byte(tt.code))
 			assert.NoError(t, err)
-			expected := strings.Join(tt.want, "\n")
-			actual := strings.Join(out, "\n")
-			assert.Equal(t, strings.Count(actual, "<span"), strings.Count(actual, "</span>"))
-			assert.EqualValues(t, expected, actual)
+			assert.EqualValues(t, tt.want, out)
 			assert.Equal(t, tt.lexerName, lexerName)
 		})
 	}
@@ -112,7 +124,7 @@ func TestPlainText(t *testing.T) {
 	tests := []struct {
 		name string
 		code string
-		want []string
+		want []template.HTML
 	}{
 		{
 			name: "empty.py",
@@ -165,9 +177,7 @@ c=2`),
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
 			out := PlainText([]byte(tt.code))
-			expected := strings.Join(tt.want, "\n")
-			actual := strings.Join(out, "\n")
-			assert.EqualValues(t, expected, actual)
+			assert.EqualValues(t, tt.want, out)
 		})
 	}
 }

modules/indexer/code/search.go

@@ -6,6 +6,7 @@ package code
 import (
 	"bytes"
 	"context"
+	"html/template"
 	"strings"
 
 	"code.gitea.io/gitea/modules/highlight"
@@ -22,7 +23,7 @@ type Result struct {
 	Language       string
 	Color          string
 	LineNumbers    []int
-	FormattedLines string
+	FormattedLines template.HTML
 }
 
 type SearchResultLanguages = internal.SearchResultLanguages

modules/markup/orgmode/orgmode.go

@@ -87,7 +87,7 @@ func Render(ctx *markup.RenderContext, input io.Reader, output io.Writer) error
 			}
 			lexer = chroma.Coalesce(lexer)
 
-			if _, err := w.WriteString(highlight.CodeFromLexer(lexer, source)); err != nil {
+			if _, err := w.WriteString(string(highlight.CodeFromLexer(lexer, source))); err != nil {
 				return ""
 			}
 		}

modules/setting/ui.go

@@ -35,6 +35,8 @@ var UI = struct {
 	OnlyShowRelevantRepos bool
 	ExploreDefaultSort    string `ini:"EXPLORE_PAGING_DEFAULT_SORT"`
 
+	AmbiguousUnicodeDetection bool
+
 	Notification struct {
 		MinTimeout            time.Duration
 		TimeoutStep           time.Duration
@@ -82,6 +84,9 @@ var UI = struct {
 	Reactions:           []string{`+1`, `-1`, `laugh`, `hooray`, `confused`, `heart`, `rocket`, `eyes`},
 	CustomEmojis:        []string{`git`, `gitea`, `codeberg`, `gitlab`, `github`, `gogs`},
 	CustomEmojisMap:     map[string]string{"git": ":git:", "gitea": ":gitea:", "codeberg": ":codeberg:", "gitlab": ":gitlab:", "github": ":github:", "gogs": ":gogs:"},
+
+	AmbiguousUnicodeDetection: true,
+
 	Notification: struct {
 		MinTimeout            time.Duration
 		TimeoutStep           time.Duration

modules/util/string.go

@@ -3,7 +3,7 @@
 
 package util
 
-import "github.com/yuin/goldmark/util"
+import "unsafe"
 
 func isSnakeCaseUpper(c byte) bool {
 	return 'A' <= c && c <= 'Z'
@@ -83,5 +83,15 @@ func ToSnakeCase(input string) string {
 			}
 		}
 	}
-	return util.BytesToReadOnlyString(res)
+	return UnsafeBytesToString(res)
+}
+
+// UnsafeBytesToString uses Go's unsafe package to convert a byte slice to a string.
+// TODO: replace all "goldmark/util.BytesToReadOnlyString" with this official approach
+func UnsafeBytesToString(b []byte) string {
+	return unsafe.String(unsafe.SliceData(b), len(b))
+}
+
+func UnsafeStringToBytes(s string) []byte {
+	return unsafe.Slice(unsafe.StringData(s), len(s))
 }