Copilot commented on code in PR #3466:
URL: https://github.com/apache/dubbo-go/pull/3466#discussion_r3502724723
##########
protocol/triple/triple_protocol/duplex_http_call_test.go:
##########
@@ -182,3 +184,91 @@ func TestDuplexHTTPCallConcurrentWriteCloseRace(t
*testing.T) {
_ = call.CloseRead()
}
}
+
+// TestDuplexHTTPCallCloseReadDoesNotDrainResponseBody is a regression guard
+// for the stream-close hang fixed upstream in connect-go v1.18.0
+// (connectrpc/connect-go#791).
+//
+// CloseRead is the low-level operation behind CloseResponse and
+// ServerStreamForClient.Close. It is a cleanup path, so it should close the
+// response body without trying to read the rest of the stream. If it drains
the
+// body first, a peer that keeps the response stream open can make close block.
+func TestDuplexHTTPCallCloseReadDoesNotDrainResponseBody(t *testing.T) {
+ t.Parallel()
+ body := newBlockingReadCloser()
+ call := &duplexHTTPCall{
+ ctx: context.Background(),
+ responseReady: make(chan struct{}),
+ response: &http.Response{
+ Body: body,
+ Trailer: make(http.Header),
+ },
+ }
+ close(call.responseReady)
+
+ done := make(chan error, 1)
+ go func() {
+ done <- call.CloseRead()
+ }()
+
+ select {
+ case err := <-done:
+ assert.Nil(t, err)
+ case <-body.readStarted:
+ body.unblock()
+ assert.Nil(t, <-done)
+ t.Fatal("CloseRead attempted to drain the response body")
+ case <-time.After(200 * time.Millisecond):
Review Comment:
This regression test uses a 200ms timeout to detect blocking. That’s very
tight and can be flaky on slower/loaded CI runners; other timeouts in this
package typically use ~1s. Consider using a larger timeout to keep the failure
signal while reducing flakiness.
##########
protocol/triple/triple_protocol/stream_close_ext_test.go:
##########
@@ -0,0 +1,202 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package triple_protocol_test
+
+import (
+ "context"
+ "errors"
+ "io"
+ "net/http"
+ "net/http/httptest"
+ "sync"
+ "testing"
+ "time"
+)
+
+import (
+ triple "dubbo.apache.org/dubbo-go/v3/protocol/triple/triple_protocol"
+
"dubbo.apache.org/dubbo-go/v3/protocol/triple/triple_protocol/internal/assert"
+ pingv1
"dubbo.apache.org/dubbo-go/v3/protocol/triple/triple_protocol/internal/gen/proto/connect/ping/v1"
+
"dubbo.apache.org/dubbo-go/v3/protocol/triple/triple_protocol/internal/gen/proto/connect/ping/v1/pingv1connect"
+)
+
+// TestServerStreamCloseDoesNotDrainResponse covers the public
+// ServerStreamForClient.Close path for connect-go#791-style lifecycle
+// behavior. The server sends one response and then intentionally keeps the
+// stream open; client-side Close should still return promptly because it is a
+// cleanup operation, not a request to read the stream to EOF.
+func TestServerStreamCloseDoesNotDrainResponse(t *testing.T) {
+ release, unblock := newCloseRelease()
+ t.Cleanup(unblock)
+ client := newCloseLifecyclePingClient(t, &pluggablePingServer{
+ countUp: func(ctx context.Context, req *triple.Request, stream
*triple.ServerStream) error {
+ if err := stream.Send(&pingv1.CountUpResponse{Number:
1}); err != nil {
+ return err
+ }
+ <-release
+ return nil
+ },
+ })
+
+ stream, err := client.CountUp(context.Background(),
triple.NewRequest(&pingv1.CountUpRequest{}))
+ assert.Nil(t, err)
+ assert.True(t, stream.Receive(&pingv1.CountUpResponse{}))
+ msg := stream.Msg().(*pingv1.CountUpResponse)
+ assert.Equal(t, msg.Number, int64(1))
+
+ done := make(chan error, 1)
+ go func() {
+ done <- stream.Close()
+ }()
+ assert.Nil(t, assertCloseReturnsPromptly(t, done, unblock,
"ServerStreamForClient.Close blocked while draining the response"))
+}
+
+// TestBidiStreamCloseResponseDoesNotDrainResponse covers the public
+// BidiStreamForClient.CloseResponse path. After one successful receive, the
+// server keeps the response side open; closing the receive side should close
+// the HTTP response body instead of draining it.
+func TestBidiStreamCloseResponseDoesNotDrainResponse(t *testing.T) {
+ release, unblock := newCloseRelease()
+ t.Cleanup(unblock)
+ client := newCloseLifecyclePingClient(t, &pluggablePingServer{
+ cumSum: func(ctx context.Context, stream *triple.BidiStream)
error {
+ req := &pingv1.CumSumRequest{}
+ if err := stream.Receive(req); err != nil {
+ return err
+ }
+ if err := stream.Send(&pingv1.CumSumResponse{Sum:
req.Number}); err != nil {
+ return err
+ }
+ <-release
+ return nil
+ },
+ })
+
+ stream, err := client.CumSum(context.Background())
+ assert.Nil(t, err)
+ assert.Nil(t, stream.Send(&pingv1.CumSumRequest{Number: 2}))
+ res := &pingv1.CumSumResponse{}
+ assert.Nil(t, stream.Receive(res))
+ assert.Equal(t, res.Sum, int64(2))
+
+ done := make(chan error, 1)
+ go func() {
+ done <- stream.CloseResponse()
+ }()
+ assert.Nil(t, assertCloseReturnsPromptly(t, done, unblock,
"BidiStreamForClient.CloseResponse blocked while draining the response"))
+ assert.Nil(t, stream.CloseRequest())
+}
+
+// TestBidiStreamCloseResponseAfterServerStopsReading covers the case where the
+// server returns before consuming the rest of the request stream. Once the
+// client has received the server-side error, closing the receive side should
be
+// a local cleanup step and must not wait for additional response bytes.
+func TestBidiStreamCloseResponseAfterServerStopsReading(t *testing.T) {
+ client := newCloseLifecyclePingClient(t, &pluggablePingServer{
+ cumSum: func(ctx context.Context, stream *triple.BidiStream)
error {
+ req := &pingv1.CumSumRequest{}
+ if err := stream.Receive(req); err != nil {
+ return err
+ }
+ return triple.NewError(triple.CodeUnavailable,
errors.New("server stopped reading"))
+ },
+ })
+
+ stream, err := client.CumSum(context.Background())
+ assert.Nil(t, err)
+ assert.Nil(t, stream.Send(&pingv1.CumSumRequest{Number: 1}))
+ err = stream.Receive(&pingv1.CumSumResponse{})
+ assert.NotNil(t, err)
+ assert.Equal(t, triple.CodeOf(err), triple.CodeUnavailable)
+
+ done := make(chan error, 1)
+ go func() {
+ done <- stream.CloseResponse()
+ }()
+ assert.Nil(t, assertCloseReturnsPromptly(t, done, nil,
"BidiStreamForClient.CloseResponse blocked after server stopped reading"))
+ assert.Nil(t, stream.CloseRequest())
+}
+
+// TestClientStreamCloseAndReceiveAfterServerReturnsError covers a
client-stream
+// RPC where the server returns an error before the client finishes sending all
+// messages. CloseAndReceive should surface the server error and finish cleanup
+// without hanging in CloseResponse.
+func TestClientStreamCloseAndReceiveAfterServerReturnsError(t *testing.T) {
+ client := newCloseLifecyclePingClient(t, &pluggablePingServer{
+ sum: func(ctx context.Context, stream *triple.ClientStream)
(*triple.Response, error) {
+ return nil, triple.NewError(triple.CodeUnavailable,
errors.New("server returned early"))
+ },
+ })
+
+ stream, err := client.Sum(context.Background())
+ assert.Nil(t, err)
+ if err = stream.Send(&pingv1.SumRequest{Number: 1}); err != nil {
+ assert.ErrorIs(t, err, io.EOF)
+ }
+
+ done := make(chan error, 1)
+ go func() {
+ done <-
stream.CloseAndReceive(triple.NewResponse(&pingv1.SumResponse{}))
+ }()
+ err = assertCloseReturnsPromptly(t, done, nil,
"ClientStreamForClient.CloseAndReceive blocked after server returned early")
+ assert.NotNil(t, err)
+ assert.Equal(t, triple.CodeOf(err), triple.CodeUnavailable)
+}
+
+func newCloseLifecyclePingClient(t *testing.T, pingServer
pingv1connect.PingServiceHandler) pingv1connect.PingServiceClient {
+ t.Helper()
+ mux := http.NewServeMux()
+ mux.Handle(pingv1connect.NewPingServiceHandler(pingServer))
+ server := httptest.NewUnstartedServer(mux)
+ server.EnableHTTP2 = true
+ server.StartTLS()
+ t.Cleanup(server.Close)
+ return pingv1connect.NewPingServiceClient(server.Client(), server.URL)
+}
+
+// newCloseRelease returns a channel that lets tests keep the server handler
+// alive until the client-side close call has completed. The unblock function
is
+// idempotent so cleanup can safely release the handler after failures.
+func newCloseRelease() (<-chan struct{}, func()) {
+ release := make(chan struct{})
+ var once sync.Once
+ return release, func() {
+ once.Do(func() {
+ close(release)
+ })
+ }
+}
+
+// assertCloseReturnsPromptly treats a timeout as evidence that close attempted
+// to drain the response stream. On failure it releases the server first so the
+// goroutine running the close call can finish before the test exits.
+func assertCloseReturnsPromptly(t *testing.T, done <-chan error, unblock
func(), failure string) error {
+ t.Helper()
+ select {
+ case err := <-done:
+ return err
+ case <-time.After(200 * time.Millisecond):
Review Comment:
The prompt-close timeout is set to 200ms, which is unusually tight compared
to other tests in this package (most timeouts here use ~1s). This can make the
regression test flaky on slower/loaded CI runners even when Close* is
non-blocking.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
