Fixes a 60 second hang that could occur in wallet-tool once the send request was processed. Fixing this revealed another bug - WalletTool was depending on this VM shutdown delay caused by the worker pool timeout, so clean up the code and replace with a blocking get of the future.
Also support the --offline and --password flags when using the payment protocol.
Run the wallet txConfidenceListener on the calling thread. Move tx broadcast that occurs inside setTransactionBroadcaster outside the wallet lock to avoid inverting on the mempool class (via PeerGroup, but that doesn't lock on the broadcast path).
This makes memory usage smoother by avoiding the creation of tons of closures and spamming the event thread with them.
- Remove needless recalculations of the Bloom filter on any wallet change, instead of just when keys/scripts are added. This may fix one of the privacy leak issues too.
- Run fast catchup/filter recalculations on the dedicated PeerGroup thread instead of abusing the user thread. Resolves a TODO.
- Replace the user thread SingleThreadedExecutor with a custom class that blocks the submitting thread and logs a warning if the queue is saturated, to avoid building up a backlog of closures.
Add lots more nullity annotations.
Clear lots of nullity static analysis warnings.
Delete some dead code.
Simplify a few expressions.
Resolves issue 317.
There was a nasty bug where a connection-failed interrupt was
run afer any interruptible calls were made, making the next
test-case fail due to interruption.
Remove Netty entirely, using the new Nio wrapper classes instead
* BitcoinSerializer now uses ByteBuffers directly instead of
InputStreams.
* TCPNetworkConnection and NetworkConnection interface deleted,
Peer now extends the abstract class PeerSocketHandler which
handles deserialization and interfaces with the Nio wrapper
classes.
* As a part of this, all version message handling has been moved
to Peer, instead of doing it in TCPNetworkConnection.
* Peer.setMinProtocolVersion() now returns a boolean instead of a
null/non-null future which holds the now-closing channel.
* Peer.sendMessage (now PeerSocketHandler.sendMessage()) now
returns void.
* PeerGroup has some significant API changes:
* removed constructors which take pipeline factories,
makePipelineFactory, createClientBootstrap
* Replaced with a setSocketTimeoutMillis method that sets a
timeout between openConnection() and version/verack exchange.
(Note that because Peer extends AbstractTimeoutHandler, it has
useful timeout setters public already).
* connectTo returns a Peer future, not a ChannelFuture
* removed peerFromChannelFuture and peerFromChannel
* Peer and PeerGroup Tests have semi-significant rewrites:
* They use actual TCP connections to localhost
* The "remote" side is a InboundMessageQueuer, which queues
inbound messages and allows for writing arbitrary messages.
* It ignores certain special pings which come from pingAndWait,
which is used to wait for message processing in the Peer.
* Removed a broken test in PeerGroupTest that should be reenabled
if we ever prefer a different version than our minimum version
again.
* Removed two duplicate tests in PeerTest (testRun_*Exception)
which are tested for in badMessage as well.
* Added a test for peer timeout and large message deserialization
Author: Matt Corallo <git@bluematt.me>
If there's insufficient balance (typical error) then InsufficientMoneyException is thrown (checked).
If the SendRequest is bad in some way, like asking to create a spend that would violate the protocol rules, IllegalArgumentException is thrown (unchecked).
Also make VerificationException (and thus ProtocolException+ScriptException) unchecked.
Resolves issue 425.
Disable signature checking in ChannelTest to make it run faster, especially in the debugger, and bump timeouts so client/server don't timeout when debugging.
Now connects all inputs of an overriding transaction, meaning balance is correct if a bit-tweaked but semantically identical transaction double spends its shadow (e.g. during key rotation on a cloned device).
Still does not recursively kill transactions however.
Resolves issue 439.
