A series of RPC calls can lead to a crash #1961

I have found the following series of RPC calls that can make bitcoind crash. You need to know the rpc username and password, as well as the wallet password.

#!/bin/bash

RPCUSER=1
RPCPASSWORD=2
WALLETPASSWORD=1

set -e

for n in {1..999};
do
    bitcoind -rpcuser=$RPCUSER -rpcpassword=$RPCPASSWORD walletpassphrase $WALLETPASSWORD 99999
    bitcoind -rpcuser=$RPCUSER -rpcpassword=$RPCPASSWORD walletlock
    echo -n "$n "
done

This is due to the re-locking not killing the previous re-locking thread before starting itself. This has a negative impact on my application, which does frequent locks and unlocks.

@wowus Is this related to the problem described in #1960 and the mentioned fix? Can you check if current master fixes this for you.

@Diapolo It's unrelated.

I think we should have a look at this before 0.8.2.

walletpassphrase starts 2 threads, ThreadTopUpKeyPool() and ThreadCleanWalletPassphrase() I guess there is something wrong here when calling walletlock, while the ThreadCleanWalletPassphrase() is still running?

The wallet passphrase logic should be rewritten, preferably using a boost::asio timer (we're using boost::asio for RPC anyway).

Should be fixed by #2625

@wowus : I cannot reproduce this either, even on my wimpy 32-bit OSX 10.6 laptop. What OS/machine configuration are you running on?

Arch Linux Intel x86_64

Closing as "we can't reproduce"