Tests are passing locally. Does anyone know of a way to locally reproduce the Travis environment?

read-ACK.
The Travis bug seems to be random.

Very much agree with this, but won't ACK unless it at least passes travis. I'll look into it.

The failure in https://travis-ci.org/bitcoin/bitcoin/jobs/51080203 is triggered by the comparison tool detecting that the node disconnected

02:07:41 14 BitcoindComparisonTool$1.onPeerDisconnected: bitcoind node disconnected!

The subsequent log however shows that the tool is the one that disconnects, and then a normal shutdown procedure follows.

2015-02-17 14:07:42 socket closed
2015-02-17 14:07:42 disconnecting peer=1
2015-02-17 14:07:42 Shutdown: In progress...

It's hard to filter out what is happening with the amount of lock-related low-level noise in that log. It may be. It may be that the interesting part of the log precedes that part of the tail that is printed ....

The travis respin made it pass. There's nothing in this change that could believably introduce intermittent errors, so ACK.

ACK