fix: Invalid UTF-8 commit messages in JSON API responses (#37542) (#37585)

Backport #37542

Co-authored-by: Nicolas <bircni@icloud.com>

---------

Co-authored-by: Nicolas <bircni@icloud.com>

modules/git/commit.go

@@ -37,11 +37,7 @@ type CommitSignature struct {
 
 // Message returns the commit message. Same as retrieving CommitMessage directly.
 func (c *Commit) Message() string {
-	// FIXME: GIT-COMMIT-MESSAGE-ENCODING: this logic is not right
-	// * When need to use commit message in templates/database, it should be valid UTF-8
-	// * When need to get the original commit message, it should just use "c.CommitMessage"
-	// It's not easy to refactor at the moment, many templates need to be updated and tested
-	return c.CommitMessage
+	return strings.ToValidUTF8(c.CommitMessage, "?")
 }
 
 // Summary returns first line of commit message.

modules/git/commit_test.go

@@ -159,6 +159,14 @@ ISO-8859-1`, commitFromReader.Signature.Payload)
 	assert.Equal(t, commitFromReader, commitFromReader2)
 }
 
+func TestCommitMessageSanitizesInvalidUTF8(t *testing.T) {
+	commit := &Commit{
+		CommitMessage: "title \xff\n\n\nbody \xff\n\n\n",
+	}
+	assert.Equal(t, "title ?\n\n\nbody ?\n\n\n", commit.Message())
+	assert.Equal(t, "title ?", commit.Summary())
+}
+
 func TestHasPreviousCommit(t *testing.T) {
 	bareRepo1Path := filepath.Join(testReposDir, "repo1_bare")
 

modules/git/pipeline/lfs_nogogit.go

@@ -11,7 +11,6 @@ import (
 	"encoding/hex"
 	"io"
 	"sort"
-	"strings"
 
 	"code.gitea.io/gitea/modules/git"
 	"code.gitea.io/gitea/modules/git/gitcmd"
@@ -102,7 +101,7 @@ func findLFSFileFunc(repo *git.Repository, objectID git.ObjectID, revListReader
 						result := LFSResult{
 							Name:         curPath + string(fname),
 							SHA:          curCommit.ID.String(),
-							Summary:      strings.Split(strings.TrimSpace(curCommit.CommitMessage), "\n")[0],
+							Summary:      curCommit.Summary(),
 							When:         curCommit.Author.When,
 							ParentHashes: curCommit.Parents,
 						}