I’ve tried all 3 methods of async gRPC calls and they all give a seg fault. When trying the callback example here I get a seg fault when calling the RPC. When trying the AsyncX method here I get a seg fault at cq_.Next once call->response_reader->Finish has been called. If I use PrepareAsyncX, StartCall it’s the same thing. Is there something missing in these examples?

My code looks like:

proto:

service FooService
{
  rpc Foo(FooRequest) returns (FooResponse);
}

AsyncX method

ServerBuilder builder_;
std::unique_ptr<ServerComplestionQueue> cq = builder_.AddCompletionQueue();
ServerComplestionQueue *cq_ = cq.get();
//...
ClientContext cc;
FooRequest req;
TagWrapper* tag = new TagWrapper;

tag->reader = stub_.AsyncFoo(&cc, req, cq_);
tag->reader->Finish(&tag->resp_, &tag->status_, (void*)tag);

//In another thread
void* tag = NULL;
bool ok = false;
cq_->Next(&tag, &ok);

cq_.Next is called in its own thread before Finish and once call->response_reader->Finish is called, the call to cq_.Next seg faults. PrepareAsyncFoo is the same, it just has a call to call->response_reader->StartCall() before Finish.

Callback method, which would be preferable since it’s far more straight forward:

ClientContext cc;
FooRequest req;
FooResponse resp;
stub_->async()->Foo(&cc, &req, &resp, [](grpc::Status status) { printf("%dn", status.ok()); });

Here the call to stub_->async()->Foo seg faults.