Jun 10, 2024 · Jun 3, 2024 · Jun 7, 2024
diff --git a/codersdk/agentsdk/convert.go b/codersdk/agentsdk/convert.go
 }
 return &proto.Log{
 CreatedAt: timestamppb.New(log.CreatedAt),
 Output:    log.Output,
 Output:strings.ToValidUTF8(log.Output, "❌"),
 Level:     proto.Log_Level(lvl),
 }, nil
 }
diff --git a/codersdk/agentsdk/logs_internal_test.go b/codersdk/agentsdk/logs_internal_test.go
 require.ErrorIs(t, err, context.Canceled)
 }

 func TestLogSender_InvalidUTF8(t *testing.T) {
 t.Parallel()
 testCtx := testutil.Context(t, testutil.WaitShort)
 ctx, cancel := context.WithCancel(testCtx)
 logger := slogtest.Make(t, nil).Leveled(slog.LevelDebug)
 fDest := newFakeLogDest()
 uut := NewLogSender(logger)

 t0 := dbtime.Now()
 ls1 := uuid.UUID{0x11}

 uut.Enqueue(ls1,
 Log{
 CreatedAt: t0,
 Output:    "test log 0, src 1\xc3\x28",
 Level:     codersdk.LogLevelInfo,
 },
 Log{
 CreatedAt: t0,
 Output:    "test log 1, src 1",
 Level:     codersdk.LogLevelInfo,
 })

 loopErr := make(chan error, 1)
 go func() {
 err := uut.SendLoop(ctx, fDest)
 loopErr <- err
 }()

 req := testutil.RequireRecvCtx(ctx, t, fDest.reqs)
 require.NotNil(t, req)
 require.Len(t, req.Logs, 2, "it should sanitize invalid UTF-8, but still send")
 // the 0xc3, 0x28 is an invalid 2-byte sequence in UTF-8.  The sanitizer replaces 0xc3 with ❌, and then
 // interprets 0x28 as a 1-byte sequence "("
 require.Equal(t, "test log 0, src 1❌(", req.Logs[0].GetOutput())
 require.Equal(t, proto.Log_INFO, req.Logs[0].GetLevel())
 require.Equal(t, "test log 1, src 1", req.Logs[1].GetOutput())
 require.Equal(t, proto.Log_INFO, req.Logs[1].GetLevel())
 testutil.RequireSendCtx(ctx, t, fDest.resps, &proto.BatchCreateLogsResponse{})

 cancel()
 err := testutil.RequireRecvCtx(testCtx, t, loopErr)
 require.ErrorIs(t, err, context.Canceled)
 }

 func TestLogSender_Batch(t *testing.T) {
 t.Parallel()
 testCtx := testutil.Context(t, testutil.WaitShort)
Original file line number	Diff line number	Diff line change
Expand Up		@@ -348,7 +348,7 @@ func ProtoFromLog(log Log) (*proto.Log, error) {
		}
		return &proto.Log{
		CreatedAt: timestamppb.New(log.CreatedAt),
		Output: log.Output,
		Output:strings.ToValidUTF8(log.Output, "❌"),
		Level: proto.Log_Level(lvl),
		}, nil
		}
Expand Down
Original file line number	Diff line number	Diff line change
Expand Up		@@ -231,6 +231,51 @@ func TestLogSender_SkipHugeLog(t *testing.T) {
		require.ErrorIs(t, err, context.Canceled)
		}

		func TestLogSender_InvalidUTF8(t *testing.T) {
		t.Parallel()
		testCtx := testutil.Context(t, testutil.WaitShort)
		ctx, cancel := context.WithCancel(testCtx)
		logger := slogtest.Make(t, nil).Leveled(slog.LevelDebug)
		fDest := newFakeLogDest()
		uut := NewLogSender(logger)

		t0 := dbtime.Now()
		ls1 := uuid.UUID{0x11}

		uut.Enqueue(ls1,
		Log{
		CreatedAt: t0,
		Output: "test log 0, src 1\xc3\x28",
		Level: codersdk.LogLevelInfo,
		},
		Log{
		CreatedAt: t0,
		Output: "test log 1, src 1",
		Level: codersdk.LogLevelInfo,
		})

		loopErr := make(chan error, 1)
		go func() {
		err := uut.SendLoop(ctx, fDest)
		loopErr <- err
		}()

		req := testutil.RequireRecvCtx(ctx, t, fDest.reqs)
		require.NotNil(t, req)
		require.Len(t, req.Logs, 2, "it should sanitize invalid UTF-8, but still send")
		// the 0xc3, 0x28 is an invalid 2-byte sequence in UTF-8. The sanitizer replaces 0xc3 with ❌, and then
		// interprets 0x28 as a 1-byte sequence "("
		require.Equal(t, "test log 0, src 1❌(", req.Logs[0].GetOutput())
		require.Equal(t, proto.Log_INFO, req.Logs[0].GetLevel())
		require.Equal(t, "test log 1, src 1", req.Logs[1].GetOutput())
		require.Equal(t, proto.Log_INFO, req.Logs[1].GetLevel())
		testutil.RequireSendCtx(ctx, t, fDest.resps, &proto.BatchCreateLogsResponse{})

		cancel()
		err := testutil.RequireRecvCtx(testCtx, t, loopErr)
		require.ErrorIs(t, err, context.Canceled)
		}

		func TestLogSender_Batch(t *testing.T) {
		t.Parallel()
		testCtx := testutil.Context(t, testutil.WaitShort)
Expand Down