
Hibernate exceptions not marshalled properly.
	I did some work on our project's exception handling today, and discovered something strange when a Hibernate exception was sent up to the browser.
	The exception thrown out of the called method looked like this:
	    org.hibernate.exception.SQLGrammarException
	        message: "invalid bla bla..."
	        cause: ->
	            java.sql.SQLException
	                message: "ORA..."
	                cause: null
	but what came out on the other end was:
	    {
	        javaClassName: "org.hibernate.exception.SQLGrammarException",
	        cause: {
	            javaClassname: "java.sql.SQLException",
	            message: "ORA...",
	            cause: null
	        }
	    }
	No message property on the SQLGrammarException!
	I did a short debug session into DWR's BeanConverter and I can see that it gets the
	    PropertyDescriptor[] descriptors = info.getPropertyDescriptors();
	for SQLGrammarException, then in a loop subsequently getting to the descriptor for "message", which it skips out from because of
	    if (readRequired && descriptor.getReadMethod() == null)
	    {
	        continue;
	    }
	where getReadMethod() returns null. 
	Strange, because as far as I can see there is a getMessage() defined on the superclass org.hibernate.exception.NestableRuntimeException, so there *is* a read method...
	----
	I found the reason for the missing "message" property, and I think maybe you need to take a look at the ExceptionConverter code.
	When you ask about the properties through
	    Introspector.getBeanInfo(SQLGrammarException.class)
	and the introspector scans the methods of Hibernate's exception class hierarchy, it will find
	    org.hibernate.exception.NestableRuntimeException
	        getMessage();
	        getMessage(int);
	(Ie, Hibernate has added an overloaded version with an int parameter) 
	The Introspector now creates an IndexedPropertyDescriptor (used for "array" access) instead of a normal PropertyDescriptor, and the descriptor will have no readMethod but instead an indexedReadMethod.
	I can't say if it's Hibernate or Introspector that is at fault here, but it does cause problems in DWR when marshalling Hibernate exceptions so I think you need to provide a workaround. Maybe add code that assures that Throwable's properties "message" and "cause" (that are guranteed to always exist on any exception) are always marshalled, whatever Introspector says?

It seems that when DWR needs to convert a two-dimensional int array (as a property), it hangs/never returns.....
	When explicitely adding <convert type="array" class="[[I"/>, it works again.
	I guess you want to handle multi-dimensional arrays for primitives or convertable types.
	But even if not, then DWR should not hang I guess ;-)

Ensure script tag remoting calls _clearUp()

IE slow on opening new windows with RA turned on.

dwr.util.setOptions gets the wrong order in Opera

Turning RA off
	Say I've got a page with reverse ajax turned on. I want to refresh the page
	using a js call from the reverse ajax script conduit.  What seems to happen
	is that the page will sometimes get stuck and won't refresh (of course this
	an IE only issue).  Is there a way to turn off the reverse ajax on the the
	onunload event handler to allow the refresh to happen?  Or even explicitly
	cancel it before attempting to refresh the page?  I'm pretty sure the issue
	is the open connection, but I didn't see an obvious 'cancelReverseAjax'
	method.  Looking at engine.js, it appears that setting reverseAjax to false
	will only set the variable, but won't turn off the _poll() method.
	...
	What's the proper way to disconnect a client with a reverse ajax connection
	(either from javascript or from java) in an immediate fashion. I'm currently
	grabbing a ref to the script session, then writing a script to it via the
	DwrUtil.addScript. If the script is for a data refresh, I do the following
	to force a reverse ajax disconnect (from my java thread):
		scriptSession.flushConduits();
		Object lock = scriptSession.getScriptLock();
		synchronized(lock) {
		    lock.notify();
		}
		scriptSession.invalidate();
		scriptSession = null;
	Is there a better way?

Multiple call issue
	When calling the same function 4 times consecutively passing a different
	paramter the same result is always returned.  While debugging on the server side
	(jboss), I have noted that the result passed back (a collection) is different.
	However, DWR appears to be getting confused and always returns the same result
	for each call.
	I have noticed that if I put each call between a begin/endbatch that the correct
	result is returned.
	This is only for 1.1.  1.0 does not have this issue.
	LabBean.getMaterials(displayMaterialJewels, 1);
	LabBean.getMaterials(displayMaterialReversedItems, 2);
	LabBean.getMaterials(displayMaterialReversedJewels, 1);
	LabBean.getMaterials(displayMaterialItems, 2);

-- Below the line for DWR 2.0 --------------------------------------------------

Beans extending hashmap - Kristjn Bjarni Gumundsson <kristjg@hugvit.is>
	I have a class that extends HashMap, and I define that it should be
	converted using the bean converter, however it always get converted
	using the map converter. Is there any way to specify that it should be
	converted as a bean?
	Joe: This ought to work because DWR should pick the more specific
	declaration. I'll add it to the list of issues.

IFrame GET is broken with Safari
	iframe is only really useful with IE6 (with activex turned off) so we can
	ignore this breakage

Broken signatures
	We are users of DWR and have been trying to pass a javascript array of
	strings as a value in a Map from our client to our server and haven't been
	able to achieve the desired outcome however we have discovered the
	following.
	The client javascript looks something like :-
	      var refIds = new Array();
	      var count = 10;
	      for (var i = 0; i < count; i++) {
	            refIds[i] = i;
	      }
	    var params = new Object();
	    params.refIds = this.refIds;
	    QueryServiceGetLocationsDWRMethod.execute(dwrResponse, this.name, "handleResponse", requestId, params);
	The server side code looks something like :-
	    public QueryServiceGetLocationsDWRMethod.execute(String objectRef, String methodRef, String requestId, Map args)
	We are trying to pass the refIds array from the client to the server,
	expecting on the server side for DWR to build a map containing key/value
	entry of <String,String[]>
	We have configure the signatures section of dwr.xml to :-	
      <signatures>
		<![CDATA[
		import java.util.Map;
		import au.gov.vic.doi.xfc.web.service.query.QueryServiceGetLocationsDWRMethod;
		QueryServiceGetLocationsDWRMethod.execute(String, String, String, Map<String,[Ljava.lang.String;>);
		]]>
      </signatures>	
	The POST to the server contains :-
		callCount=1
		c0-scriptName=QueryServiceGetLocations
		c0-methodName=execute
		c0-id=3063_1144372351140
		c0-param0=string:Object1
		c0-param1=string:handleResponse
		c0-param2=string:reqId46
		c0-e2=number:0
		c0-e3=number:1
		c0-e4=number:2
		c0-e5=number:3
		c0-e6=number:4
		c0-e7=number:5
		c0-e8=number:6
		c0-e9=number:7
		c0-e10=number:8
		c0-e11=number:9
		c0-e1=Array:[reference:c0-e2,reference:c0-e3,reference:c0-e4,reference:c0-e5,reference:c0-e6,reference:c0-e7,reference:c0-e8,reference:c0-e9,reference:c0-e10,reference:c0-e11]
		c0-param3=Object:{refIds:reference:c0-e1}
		xml=true	
	Which looks correct to us.
	As you can see we are getting a map that contains a key/value pair where
	the value is a single string representing an array of object references
	rather than an array of string values (10 array elements in this example).
	In getting to the point we have, we have discovered the following :-
	   The SignatureParser class needs to be altered to take into account the
	   ClassLoader requires the string [Ljava.lang.String; to load the class as
	   an array of Strings. This string includes the ";" character which the
	   parser interprets as a line delimiter which in this case causes a
	   failure as thetokenized incorrectly tokenizes part lines.
	   We coded around this and the next parsing issue was that the "."
	   [Ljava.lang.String; string where being replaced with "$", the parser was
	   interpreting the string as representing an inner class, clearly this is
	   not the case. We coded around this and have now got to a point where the
	   [dwr-invoker]: Initialization successful is completing however we are
	   back to where we started in that whilst our code changes have taken us
	   through the parser successfully, the ArrrayConverter does not appear to
	   be doing its job as we are getting the same result as above where the
	   array is being constructed on the server as a single string of array
	   references.
	We're not able to spend more time on tracking down what's happening with
	ArrayConverter. It appears the ArrayConverter is working for the first
	level of indirection but not subsequent levels.
	Our fix for the first parser problem was to remove the ";" from the
	signatures section in the dwr.xml and if the class spec starts with an "["
	the append a ";" before loading the class, we simlely commented out the
	section of code that manages inner classes to get around the "$"
	substitution problem.
	Thanks Dan.
