cmd/devp2p: ethtest suite runnable as unit test

[renaynay]

This PR adds a test file to allow the eth test suite to be runnable as independent unit tests for easier testing.

TODO:

• allow logging from test suite
• when running all tests, reuse suite to avoid multiple geth setups

[fjl]

ok  	github.com/ethereum/go-ethereum/cmd/devp2p/internal/ethtest	164.936s

So it passes, but also takes a long time. We need to make it run faster.

[fjl]

--- PASS: TestEthSuite (169.65s)
    --- PASS: TestEthSuite/Status (0.00s)
    --- PASS: TestEthSuite/Status_66 (0.00s)
    --- PASS: TestEthSuite/GetBlockHeaders (0.00s)
    --- PASS: TestEthSuite/GetBlockHeaders_66 (0.00s)
    --- PASS: TestEthSuite/SimultaneousRequests_66 (0.00s)
    --- PASS: TestEthSuite/SameRequestID_66 (0.00s)
    --- PASS: TestEthSuite/ZeroRequestID_66 (0.00s)
    --- PASS: TestEthSuite/GetBlockBodies (0.00s)
    --- PASS: TestEthSuite/GetBlockBodies_66 (0.00s)
    --- PASS: TestEthSuite/Broadcast (0.11s)
    --- PASS: TestEthSuite/Broadcast_66 (0.11s)
    --- PASS: TestEthSuite/LargeAnnounce (20.23s)
    --- PASS: TestEthSuite/LargeAnnounce_66 (20.26s)
    --- PASS: TestEthSuite/OldAnnounce (10.01s)
    --- PASS: TestEthSuite/OldAnnounce_66 (10.01s)
    --- PASS: TestEthSuite/MaliciousHandshake (0.12s)
    --- PASS: TestEthSuite/MaliciousStatus (0.02s)
    --- PASS: TestEthSuite/MaliciousHandshake_66 (0.09s)
    --- PASS: TestEthSuite/MaliciousStatus_66 (0.02s)
    --- PASS: TestEthSuite/Transactions (0.21s)
    --- PASS: TestEthSuite/Transactions_66 (0.22s)
    --- PASS: TestEthSuite/MaliciousTransactions (50.16s)
    --- PASS: TestEthSuite/MaliciousTransactions_66 (50.18s)

The long tests are

    --- PASS: TestEthSuite/MaliciousTransactions (50.16s)
    --- PASS: TestEthSuite/MaliciousTransactions_66 (50.18s)
    --- PASS: TestEthSuite/LargeAnnounce (20.23s)
    --- PASS: TestEthSuite/LargeAnnounce_66 (20.26s)
    --- PASS: TestEthSuite/OldAnnounce (10.01s)
    --- PASS: TestEthSuite/OldAnnounce_66 (10.01s)

We need to see why they take so much time. Is it because of large data transfer, or because the test has a large timeout?

[fjl]

I tried setting global timeout to 10s, but no difference.

[fjl]

In MaliciousTransactions_66, the issue seems to be that it hits the timeout multiple times:

# Testing malicious tx propagation: 0
# unexpected message, logging: (*ethtest.Error)(could not read from connection: read tcp 127.0.0.1:55753->127.0.0.1:55751: i/o timeout)
# 
# Testing malicious tx propagation: 1
# unexpected message, logging: (*ethtest.Error)(could not read from connection: read tcp 127.0.0.1:55763->127.0.0.1:55751: i/o timeout)
# 
# Testing malicious tx propagation: 2
# unexpected message, logging: (*ethtest.Error)(could not read from connection: read tcp 127.0.0.1:55765->127.0.0.1:55751: i/o timeout)
# 
# Testing malicious tx propagation: 3
# unexpected message, logging: (*ethtest.Error)(could not read from connection: read tcp 127.0.0.1:55817->127.0.0.1:55751: i/o timeout)
# 
# Testing malicious tx propagation: 4
# unexpected message, logging: (*ethtest.Error)(could not read from connection: read tcp 127.0.0.1:55819->127.0.0.1:55751: i/o timeout)
# 

[fjl]

I think we should announce the txs individually here. Some of them can be large, and the node might reject the entire announcement when the first item is too big.

[fjl]

What I meant is, you can just send them in a loop like

for _, tx := range txs {
    if err := sendConn.Write(&Transactions{tx}); err != nil { ... }
}