andreacavalli/netty5
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
3,131 Commits 27 Branches 234 Tags
Commit Graph

2771 Commits

Author SHA1 Message Date
Norman Maurer
0522c4ffc4 Merge pull request #572 from fredericBregier/3 
3 - fix for issue  #569
 2012-08-28 09:08:22 -07:00
Frédéric Brégier
148fc841ac Second part of fix for issue #569 
When moving to take into account arrayOffset, it should have been taken into account also in setReadPosition and other places. Fix it now...
Also fix mismatch algorithm between SeekAheadOptimize and not SeekAheadOptimize (standard) versions.
 2012-08-28 16:39:47 +03:00
Frédéric Brégier
a3cedc8b47 First part of fix for issue #569 
When moving to take into account arrayOffset, it should have been taken into account also in setReadPosition and other places. Fix it now...
 2012-08-28 16:23:35 +03:00
Norman Maurer
dc56905630 Merge pull request #545 from jaens/3-fix-spdy-compress 
Do not write compressed SPDY frames out-of-band in another thread
 2012-08-28 05:38:58 -07:00
Norman Maurer
e685e535f9 Introduce a new abstract class called OneToOneStrictEncoder which helps to ensure strict ordering. This should be used if that is needed like in the case of ZIP. See ##546 2012-08-28 13:32:38 +02:00
Norman Maurer
dc8ff7e173 Correctly convert to nanos 2012-08-28 13:04:59 +02:00
norman
039c7563cf Cleanup 2012-08-27 11:21:05 +02:00
norman
292a186d16 Add workaround for epoll bug that cause 100% cpu usage also in the NioClientSocketPipelineSink 2012-08-27 11:12:15 +02:00
norman
db1a72c02e Handle ClosedChannelException on re-create of the selector 2012-08-27 11:11:34 +02:00
norman
f8a99a0108 Use nanos to detect the jdk epoll bug. Also use 80% of the select timeout to detect it to be more save. Thanks to @kimchy for spot this 2012-08-27 10:20:40 +02:00
Norman Maurer
798390fc4d Switch to System.nanoTime() to calculate block time of Selector.select(..) 2012-08-26 18:48:50 +02:00
Norman Maurer
8b4f593397 Allow to adjust timeout of Selector.select(timeout) via org.jboss.netty.selectTimeout property. See #568 2012-08-26 09:02:59 +02:00
Norman Maurer
5d07dea3b7 Use Selecor.select() to accept new Sockets to not need to schedule a timeout if not needed anyway. See #567 2012-08-26 08:46:48 +02:00
Norman Maurer
ff3f2b6361 Re-create Selector if we hit the epoll(..) jdk bug which leads to 100% cpu load. This is just a workaround but helps to recover. See #327 2012-08-25 21:51:52 +02:00
Norman Maurer
1a6e7b4be1 Merge pull request #556 from netty/jdk_bug_workaround 
Don't rely on the return value fo Selector.select(..) as it is buggy and...
 2012-08-25 11:11:05 -07:00
Cruz Julian Bishop
f34eb62104 Remove AbstractChannel.getRandom() 
Signed-off-by: Cruz Julian Bishop <cruzjbishop@gmail.com>
 2012-08-25 16:49:38 +10:00
Cruz Julian Bishop
93c990dd7a Generate Channel IDs in a pseudorandom fashion 
Requested by @psweeny in #547

Signed-off-by: Cruz Julian Bishop <cruzjbishop@gmail.com>
 2012-08-25 13:43:02 +10:00
norman
5c6e3fe6ba Don't rely on the return value fo Selector.select(..) as it is buggy and so can cause 100% cpu usage. See #535 2012-08-24 08:59:39 +02:00
Norman Maurer
14d01f4a8f Move the catch block out of the connect method, which makes more sense as we need to catch it anyway in a more general scope. See See #535 2012-08-23 21:04:34 +02:00
Norman Maurer
85e55eebd8 Add a workaround for the 100% cpu usage that can happen because of an endless loop. See #535 2012-08-23 20:41:22 +02:00
norman
03c4283020 Backport fix for make SslHandler close the connection on SSLException or NotSslRecordException. This needs to get enabled via setter to keep backward compatibility. See #160 2012-08-22 07:46:49 +02:00
Trustin Lee
01ba0ff075 [#160] Revert the fix for #160 for the branch '3' 
- Fix will be part of 4.x
 2012-08-22 13:44:58 +09:00
Trustin Lee
81305d87b4 [#160] No response to write if server is using SslHandler and client is not 
- Make SslHandler close the connection immediately on SSLException or
  or NotSslRecordException
 2012-08-22 12:16:40 +09:00
Trustin Lee
31a51b4937 [#239] IdleStateHandler and ReadTimeoutHandler starts two timers 
- Ensure initialize does not start timer twice
 2012-08-21 20:13:44 +09:00
Jaen Saul
4e351f7399 Do not write compressed SPDY frames out-of-band in another thread 2012-08-20 21:39:28 +03:00
Trustin Lee
a93ada2031 [#539] Potential direct memory leak in HttpContentEn/Decoder 2012-08-20 13:40:58 +09:00
Trustin Lee
bf74b16774 [#539] Potential direct memory leak in HttpContentEn/Decoder 2012-08-20 13:35:12 +09:00
Trustin Lee
88c3fd306b Typo 2012-08-20 12:18:59 +09:00
Trustin Lee
513fc4f78b [#539] Fix potential direct memory leak in HttpContentEn/Decoder 2012-08-20 12:14:38 +09:00
Trustin Lee
ebbcfbc185 Fix test failure 2012-08-20 12:04:16 +09:00
Norman Maurer
f7f0511193 Set maxAge to Integer.MIN_VALUE by default. See #534 2012-08-19 10:53:13 +02:00
Norman Maurer
af672039d7 Allow to set a negative value for maxAge of DefaultCookie. See #533 2012-08-19 10:45:55 +02:00
Trustin Lee
963b7c20ac Merge pull request #496 from CruzBishop/static-fixes 
Some static analysis fixes
 2012-08-17 00:12:53 -07:00
Trustin Lee
195a7bb953 Use UTF-8 to encode URI - see #521 2012-08-17 11:44:21 +09:00
Trustin Lee
c509a278c3 Rename variables to reduce confusion 2012-08-17 11:33:54 +09:00
James Tyrrell
e5a7fa2c96 [#494] Automatically adding chunked encoding header breaks streaming 2012-08-16 18:23:22 +10:00
Norman Maurer
3f709efcea Merge pull request #522 from jpinner/utf8_encode_uri_3 
Fix #521: Encode URI using UTF-8 charset
 2012-08-16 00:22:19 -07:00
norman
24f1b54c86 Only cache the localAddress if its a non wildcard address, so its possible to retrieve the 'real' adress later once the channel is fully bound/connected. See #524 2012-08-16 07:38:38 +02:00
Jeff Pinner
4f001f3246 Fix #521: Encode URI using UTF-8 charset 2012-08-15 14:04:30 -07:00
Norman Maurer
b6264c02d9 Make sure that it continue to try to read from the socket even if the SocketTimeoutException was triggered because of the SO_TIMEOUT. See #520 2012-08-15 22:39:40 +02:00
Norman Maurer
d3d5a931d0 Set the SO_TIMEOUT on the underlying Socket so we will be able to run submitted tasks in the IO-Thread even if the read operation would block because of nothing to read. See #520 2012-08-15 22:29:03 +02:00
norman
88124d88ce Remove synchronized blocks to optimize BufferedWriteHandler. See #519 2012-08-15 15:02:53 +02:00
Cruz Julian Bishop
8af95f0897 Fixed a typo in ObjectEchoClientHandler 
This fixes #510 in branch 3

Signed-off-by: Cruz Julian Bishop <cruzjbishop@gmail.com>
 2012-08-14 19:54:52 +10:00
Norman Maurer
3a3c5de76b Merge branch '3' of github.com:netty/netty into 3 2012-08-12 10:59:04 +02:00
Norman Maurer
6d87b7629e Fix NPE in DefaultChannelPipeline if toString() is called with an empty pipeline. See #505 2012-08-12 10:58:43 +02:00
Cruz Julian Bishop
0aebfb762b Removes unnecessary parentheses for @trustin 
Part of #496

Signed-off-by: Cruz Julian Bishop <cruzjbishop@gmail.com>
 2012-08-09 19:10:17 +10:00
Cruz Julian Bishop
abc2f87f2f Merge remote-tracking branch 'netty/3' into static-fixes 2012-08-09 19:01:05 +10:00
Trustin Lee
6d5a332667 Fix a buf in CompositeChannelBuffer.setBytes() where -1 is not returned 2012-08-09 17:50:27 +09:00
norman
eb8757f1cb Fix bug which leads to return HttpMessageEncoder implementations null if encoding a HttpChunk which is the last one. This vilates the OneToOneEncoder contract and so leads to a missing notification of the write future. See #493 2012-08-09 08:02:27 +02:00
Cruz Julian Bishop
618ddb42e3 Replaces two manual array copies 
This is a potential performance boost, but there shouldn't
be too much of a difference.

Signed-off-by: Cruz Julian Bishop <cruzjbishop@gmail.com>
 2012-08-09 09:51:43 +10:00