test: fix test "test_prepare_uncles"; it could failed with a very small chance

[yangby-cryptape]

Issue

• ChainController::process_block(..) is synchronized, it will wait for the response.

  ckb/chain/src/chain.rs

  Lines 200 to 208 in 6e34d89

  	recv(process_block_receiver) -> msg => match msg {
  	Ok(Request { responder, arguments: (block, verify) }) => {
  	let _ = responder.send(self.process_block(block, verify));
  	},
  	_ => {
  	error!("process_block_receiver closed");
  	break;
  	},
  	},

• But when ChainService::process_block(..), it will send NewUncle(..) to TxPoolController without waiting for the response.

  ckb/chain/src/chain.rs

  Lines 491 to 497 in 6e34d89

  	if let Err(e) = self
  	.shared
  	.tx_pool_controller()
  	.notify_new_uncle(block_ref.as_uncle())
  	{
  	error!("notify new_uncle error {}", e);
  	}

• So, with a very small probability, the uncle block may not have been processed by TxPoolService.

  So the follow line will throw the error: panicked at 'index out of bounds: the len is 0 but the index is 0.

  ckb/chain/src/tests/block_assembler.rs

  Line 210 in 6e34d89

  assert_eq!(block_template.uncles[0].hash, block0_0.hash().unpack());

Reproduce

Append ::std::thread::sleep(::std::time::Duration::from_millis(500)); to line 629 in the follow code section to delay TxPool update NewUncle can reproduce this issue.

ckb/tx-pool/src/service.rs

Lines 629 to 637 in 6e34d89

	Message::NewUncle(Notify { arguments: uncle }) => {
	if service.block_assembler.is_some() {
	let block_assembler = service.block_assembler.clone().unwrap();
	block_assembler.candidate_uncles.lock().await.insert(uncle);
	block_assembler
	.last_uncles_updated_at
	.store(unix_time_as_millis(), Ordering::SeqCst);
	}
	}

[keroro520]

Cool

[yangby-cryptape]

bors r=zhangsoledad,keroro520

[bors]

Build succeeded:

• continuous-integration/travis-ci/push