* HttpListenerManager.cs : remove consumed HttpListenerContext from
the pending queue. Compare "wsdl" request in case insensitive
manner. Patch by Matt Dargavel.
svn path=/trunk/mcs/; revision=154120
* HttpRequestContext.cs : apply the patch by Matt Dargavel, which is
supposed to provide detailed 400 error message.
svn path=/trunk/mcs/; revision=154112
* ChannelListenerBase.cs, TransactionFlowBindingElement.cs: fix
GetProperty<T>() bug that fails to retrieve properties. (This may
be rewritten later if an ongoing change proposal is to take place.)
svn path=/trunk/mcs/; revision=154107
* HttpListenerManager.cs : For ASP.NET, use correct GenericIdentity
type, and use null password unless we find a chance to get password
in ASP.NET HttpContext.
svn path=/trunk/mcs/; revision=153310
* HttpReplyChannel.cs, AspNetReplyChannel.cs, HttpChannelListener.cs,
HttpListenerManager.cs : ongoing authentication support. It returns
401 as expected, but somehow fails to retrieve user info.
svn path=/trunk/mcs/; revision=152941
* MessageHeader.cs, MessageHeaders.cs: MessageHeader does not treat
EndpointAddress as a special case. Instead, each MessageHeaders
property should convert it to appropriate serializable objects.
(Our DataContractSerializer fails to reject EndpointAddress as
serializable type, so it's rather a preparation step.)
svn path=/trunk/mcs/; revision=151196
* MessageHeader.cs, Message.cs : use dictionary string whenever
applicable. The binary output on #577139 is getting closer (no
idea if it is really about binary output though).
svn path=/trunk/mcs/; revision=151194
* BinaryMessageEncoder.cs : moved soap_dictionary to Constants.
* Constants.cs : moved SoapDictionary from BinaryMessageEncoder.
svn path=/trunk/mcs/; revision=151193
* SvcHttpHandler.cs : a wait handle could be used for more than one
requests and hence it sometimes blocked one reply channel. Now
it creates a wait handle for each ASP.NET request so that such
multiple use should not occur. This should fix blocking on
concurrent requests (it indeed fixes as long as I can see).
* HTTP_listener_notes.txt: some notes updated.
svn path=/trunk/mcs/; revision=150850
* TcpDuplexSessionChannel.cs : added a workaround receiver delay
that somehow makes callback client to not cause infinite loop.
* CallbackBehaviorAttributeTest.cs : enable the test again.
svn path=/trunk/mcs/; revision=150052
* TcpDuplexSessionChannel.cs, PeerDuplexChannel.cs,
DuplexChannelBase.cs : Receive() should rather use TryReceive().
It should not be in reverse order.
svn path=/trunk/mcs/; revision=150040
* HttpReplyChannel.cs, HttpListenerManager.cs :
Implement async cancellation in the expected manner (though with
some hack). This mostly removed nunit blockers.
* NetTcpBindingTest.cs : enabled connection tests again, as they
got working again (and more importantly, they don't block).
* CallbackBehaviorAttributeTest.cs : on the other hand, disabled
a test so far. It is the only active test && the blocker now.
svn path=/trunk/mcs/; revision=150037
* CommunicationObject.cs : when process state changes, lock the
object to make sure the state transition is valid.
Change OnClosed/OnClosing to do the way OnOpened/OnOpening does.
* ChannelDispatcher.cs : another error audit.
svn path=/trunk/mcs/; revision=149793
* TcpBinaryFrameManager.cs : treat EOF as interrupted stream too (it
does not happen when communicating with .NET client but happens with
mono). Seealso FIXME comment.
* TcpDuplexSessionChannel.cs : do not try to close session when it
failed to read sized message. The connection is already dead.
Fixed bug #567949.
svn path=/trunk/mcs/; revision=149166
* TcpDuplexSessionChannel.cs :
Write EndRecord only when the connection is available.
Remove hack - do let ChannelDispatcher close the channel.
* TcpBinaryFrameManager.cs :
The socket may be disconnected when trying to read a sized message,
so check some socket exceptions.
In case the client silently shuts down without EndRecord and
reconnects to the server with preamble while expecting a sized
message (which is so lame, but that's what .NET often does), refresh
preamble and continue.
This fixes some cases of bug #567949 but not all.
svn path=/trunk/mcs/; revision=149163
* BinaryMessageEncoder.cs : enable workaround buffering only when
the argument stream does *not* support seek. This fixes some
net.tcp connection problem.
* BinaryMessageEncodingBindingElementTest.cs : added test for
connection-based tests for both TCP and HTTP (they give difference:
see BinaryMessageEncoder.cs change too).
svn path=/trunk/mcs/; revision=149103
* TcpBinaryFrameManager.cs : If preamble byte is not available from
the client stream, detect it to close the reply or input channel.
* TcpRequestChannel.cs : send preamble at each request time.
It differentiates communication behavior on such cases that involve
single open and multiple requests.
* TcpReplyChannel.cs : close the channel only when the channel is
oeprated to close i.e. do not close the socket every time.
Process preambles on each request (the same as request channel).
svn path=/trunk/mcs/; revision=148901
* TcpBinaryFrameManager.cs : unsized envelope record may have more
than one data block, so consume them accordingly. This should fix
large buffer consumption.
svn path=/trunk/mcs/; revision=148900
* MessageFault.cs : support EnvelopeVersion.None as well, as almost
the same as SOAP12.
* MessageFaultTest.cs : add CreateFault test with MessageVersion.None.
svn path=/trunk/mcs/; revision=148797