Commit Graph

4846 Commits

Author SHA1 Message Date
Norman Maurer
e71a521284 [#1414] Use gathering writes in AbstractMemoryHttpData if the buffer is backed by multiple ByteBuffers
* This fix the bug which caused an UnsupportedOperationException when renameTo(...) was called and the underlying ByteBuf was backed by multiple ByteBuffers
2013-06-10 09:25:03 +02:00
Norman Maurer
fa6999cd42 [#1425] Allow to access the EventLoopGroups via the Bootstraps 2013-06-10 09:24:57 +02:00
Wolfgang Profer
d9af92f354 [#1430] Fixed byte order problem on little endian platforms that support Unsafe but not unaligned memory access. 2013-06-10 09:24:22 +02:00
Derek Troy-West
f2d8a745b1 [#1422] ReadOnlyByteBuffer.isWritable() should return false 2013-06-10 09:23:57 +02:00
Trustin Lee
ccd43e0cda Fix compilation error 2013-06-10 16:18:22 +09:00
Trustin Lee
14158070bf Revamp the core API to reduce memory footprint and consumption
The API changes made so far turned out to increase the memory footprint
and consumption while our intention was actually decreasing them.

Memory consumption issue:

When there are many connections which does not exchange data frequently,
the old Netty 4 API spent a lot more memory than 3 because it always
allocates per-handler buffer for each connection unless otherwise
explicitly stated by a user.  In a usual real world load, a client
doesn't always send requests without pausing, so the idea of having a
buffer whose life cycle if bound to the life cycle of a connection
didn't work as expected.

Memory footprint issue:

The old Netty 4 API decreased overall memory footprint by a great deal
in many cases.  It was mainly because the old Netty 4 API did not
allocate a new buffer and event object for each read.  Instead, it
created a new buffer for each handler in a pipeline.  This works pretty
well as long as the number of handlers in a pipeline is only a few.
However, for a highly modular application with many handlers which
handles connections which lasts for relatively short period, it actually
makes the memory footprint issue much worse.

Changes:

All in all, this is about retaining all the good changes we made in 4 so
far such as better thread model and going back to the way how we dealt
with message events in 3.

To fix the memory consumption/footprint issue mentioned above, we made a
hard decision to break the backward compatibility again with the
following changes:

- Remove MessageBuf
- Merge Buf into ByteBuf
- Merge ChannelInboundByte/MessageHandler and ChannelStateHandler into ChannelInboundHandler
  - Similar changes were made to the adapter classes
- Merge ChannelOutboundByte/MessageHandler and ChannelOperationHandler into ChannelOutboundHandler
  - Similar changes were made to the adapter classes
- Introduce MessageList which is similar to `MessageEvent` in Netty 3
- Replace inboundBufferUpdated(ctx) with messageReceived(ctx, MessageList)
- Replace flush(ctx, promise) with write(ctx, MessageList, promise)
- Remove ByteToByteEncoder/Decoder/Codec
  - Replaced by MessageToByteEncoder<ByteBuf>, ByteToMessageDecoder<ByteBuf>, and ByteMessageCodec<ByteBuf>
- Merge EmbeddedByteChannel and EmbeddedMessageChannel into EmbeddedChannel
- Add SimpleChannelInboundHandler which is sometimes more useful than
  ChannelInboundHandlerAdapter
- Bring back Channel.isWritable() from Netty 3
- Add ChannelInboundHandler.channelWritabilityChanges() event
- Add RecvByteBufAllocator configuration property
  - Similar to ReceiveBufferSizePredictor in Netty 3
  - Some existing configuration properties such as
    DatagramChannelConfig.receivePacketSize is gone now.
- Remove suspend/resumeIntermediaryDeallocation() in ByteBuf

This change would have been impossible without @normanmaurer's help. He
fixed, ported, and improved many parts of the changes.
2013-06-10 16:10:39 +09:00
Jeff Pinner
fe711f65fb Merge pull request #1421 from jpinner/spdy_fix_stalled_streams_master
SPDY: always enqueue data frames if stream is stalled
2013-06-05 07:45:05 -07:00
Jeff Pinner
cf757f17f6 SPDY: always enqueue data frames if stream is stalled 2013-06-04 23:58:49 -07:00
Norman Maurer
bc20107b68 Use correct value to disable/enable direct arenas in PooledByteBufAllocator 2013-05-30 20:24:11 +02:00
Norman Maurer
89f1f3f4d1 [#1399] DefaultChannelHandlerPipeline.firstContext() should return null if no user handlers are in in the pipeline 2013-05-27 15:45:34 +02:00
Norman Maurer
f7931af704 Re-add Unsafe.voidPromise() which can be used for Unsafe operations for which no notification should be done [#1375] 2013-05-25 14:35:22 +02:00
Norman Maurer
d9c700e9fe [#1392] Fix bug where SslHandler returned wrong value for getCloseNotifyTimeoutMillis() 2013-05-24 23:18:14 +02:00
Norman Maurer
aa96b4b286 No need to create temporary arrays 2013-05-24 19:58:55 +02:00
Norman Maurer
f5dc482a59 No need to clear buffer as it is cleared later anyway and only update interestedOps if needed 2013-05-24 12:50:04 +02:00
Norman Maurer
d31ccebd62 Make sure that setAutoRead(false) has a direct effect and only update interestedOps if needed 2013-05-24 12:49:21 +02:00
Norman Maurer
252bd25855 Only call key.interestedOps() if needed 2013-05-24 12:46:30 +02:00
Norman Maurer
0e8fb21554 Some optimizations to the http codec 2013-05-24 09:07:17 +02:00
Trustin Lee
a3b4cdd614 Fix StackOverflowError in LocalEcho.doBeginRead() when the peer channel keeps writing data
- Fixes #1380
2013-05-24 11:55:21 +09:00
Norman Maurer
5398792ffa [#1388] Correctly break the loop on exceptions 2013-05-23 17:17:21 +02:00
Norman Maurer
50ac0cdfcb [#1388] Ensure AbstractNioMessageChannel based Channels will call fireInboundBufferUpdated() soon enough to release resources 2013-05-23 16:59:03 +02:00
Norman Maurer
83dcf829d6 [#1384] Cache HttpHeaderDateFormat in ThreadLocal 2013-05-23 11:39:50 +02:00
Norman Maurer
9c925b104a [#1385] Fix NPE which was triggered if a write was executed but the HeadHandler not init yet 2013-05-23 07:42:01 +02:00
shreyharia
1801ecfe30 Update HttpHeaderDateFormat.java
Output for "E, dd-MMM-y HH:mm:ss z"
on java 1.6 I get: Wed, 22-May-13 09:11:41 GM
on java 1.7 I get: Wed, 22-May-2013 09:11:14 GMT

To be uniform on both jdks.. modified it to
"E, dd-MMM-yy HH:mm:ss z"

learnt from - http://stackoverflow.com/questions/16687298/simpledateformat-different-behaviour-depending-on-java-version
2013-05-22 13:36:03 +02:00
Norman Maurer
9a7addbebe No need to use NioSocketChannel.class directly 2013-05-21 20:27:16 +02:00
Norman Maurer
548540bc2d Fix a which could cause data corruption when using AioSocketChannel.
This was because it was possible to have the JDK read into a wrong buffer region if the user called discardReadBytes() later. Fixes #1377
2013-05-21 20:19:00 +02:00
Luke Wood
c1cbe8bec3 Clarify SnappyFramedDecoder javadoc to make it as clear as possible regarding the default settings 2013-05-20 17:31:04 +01:00
Norman Maurer
81e3c1719a [maven-release-plugin] prepare for next development iteration 2013-05-18 09:59:13 +02:00
Norman Maurer
99caefdf39 [maven-release-plugin] prepare release netty-4.0.0.CR3 2013-05-18 09:57:11 +02:00
Trustin Lee
50433f7b9a Make sure successful SslHandshakeCompletionEvent is a singleton
- Related: #1373
- This commit is an amendment to 2a7bea2ad3
2013-05-18 05:50:58 +09:00
Norman Maurer
620c3e025a Just some tiny javadoc fixes 2013-05-17 22:16:29 +02:00
Norman Maurer
bd2230961d Fix checkstyle 2013-05-17 22:09:11 +02:00
Norman Maurer
aaca9df99a @Deprecate a method as it's not needed anymore 2013-05-17 22:00:55 +02:00
Norman Maurer
8a5e249d6f [#1374] Allow users to all all subprotocols by using * as wildcard 2013-05-17 21:57:47 +02:00
Norman Maurer
abb4e20d0b [#1369] Move ImmediateEventExecutor to common and let it access via a static
* Also fix a bug there to return a correct implementation of ProgressivPRomi
  ImmediateEventExecutor
2013-05-17 21:35:01 +02:00
Norman Maurer
a8830aee42 [#1369] Move ImmediateEventExecutor to common and let it access via a static public field
* Also fix a bug there to return a correct implementation of ProgressivPRomise to work with the
 ImmediateEventExecutor
2013-05-17 21:19:59 +02:00
Norman Maurer
6942dba855 [#1363] Make sure ChannnelConfig.setAutoRead(false) will stop read from socket directly 2013-05-17 20:56:09 +02:00
Norman Maurer
2a7bea2ad3 [#1373] Make SslHandshakeCompletionEvent constructor public 2013-05-17 19:39:52 +02:00
Norman Maurer
699ef0784e [#1317] Allow to use VoidPromise for flush(...), write(...) and sendFile(...)
* This also move rename Channel.Unsafe.voidFuture() to ChannelPropertyAccess.voidPromise()
2013-05-17 15:50:14 +02:00
Trustin Lee
fd1d31e7d8 Remove unnecessary inEventLoop() checks in Channel.Unsafe
.. because HeadHandler in the pipeline always ensures those methods are always invoked from the correct I/O thread
2013-05-17 19:20:46 +09:00
Trustin Lee
41f5d5650d Add Javadoc to tarball / Upgrade maven-compiler-plugin
- Fixes #1360
2013-05-17 18:32:27 +09:00
Trustin Lee
e58af1bc94 Readability / Fix inception year 2013-05-17 17:59:13 +09:00
kxbmap
3a5aa12000 Add LengthFieldPrepender.lengthAdjustment parameter 2013-05-17 17:58:53 +09:00
Trustin Lee
f92cfba388 Remove DefaultCompositeByteBuf.lastAccessed and use binary search instead
- Fixes #1364
- Even if a user creates a duplicate/slice, lastAccessed was shared between the derived buffers and it's updated even by a read operation, which made multithread access impossible
2013-05-17 16:54:47 +09:00
Trustin Lee
7140e4e63b Test if ChannelHandler.handlerRemoved() is called on closure / Reduced timeout 2013-05-17 11:07:53 +09:00
Trustin Lee
e1a378aa03 Clean up DefaultChannelPipelineTest
- Use the local transport in a correct way (i.e. no need to trigger channelActive et al by ourselves)
- Use Promise/Future instead of CountDownLatch where they simplifies
2013-05-17 10:54:20 +09:00
Luke Wood
f841056752 Fix SnappyFramedDecoder checksum validation and add a pair of suitable unit tests 2013-05-16 14:10:54 +01:00
Trustin Lee
2040b07849 Free the cleared buffer as early as possible / Better naming 2013-05-16 19:41:02 +09:00
Trustin Lee
dc13b68632 Make sure ChannelHandler.handlerRemoved() is always invoked
- Fixes #1366: No elegant way to free non-in/outbound buffers held by a handler
- handlerRemoved() is now also invoked when a channel is deregistered, as well as when a handler is removed from a pipeline.
- A little bit of clean-up for readability
- Fix a bug in forwardBufferContentAndRemove() where the handler buffers are not freed (mainly because we were relying on channel.isRegistered() to determine if the handler has been removed from inside the handler.
- ChunkedWriteHandler.handlerRemoved() is unnecessary anymore because ChannelPipeline now always forwards the content of the buffer.
2013-05-16 19:32:39 +09:00
Trustin Lee
670d3f53a8 Make uninitialization code in DefaultChannel easier to understand
.. by fixing bad / outdated method names
2013-05-16 16:18:01 +09:00
Norman Maurer
a331c87a7a [#1358] Fix Encoding of multipart attribute names and filenames for non ASCII chars 2013-05-14 15:24:01 +02:00