Re: [ve-dev] Initialization error with VisualComponentAdapter#componentV

[Rich Kulp <richkulp@xxxxxxxxxx>]

Hi,

I  don't know what the problem
is. You should be able to call back to the remote vm, but socketClosed
is a weird message. That should be a log entry indicating from the remote
vm why the socket was closed. A socket should not close while in a transaction
unless there was a severe error on the remote vm.

Rich

Gerald Rosenberg <gerald@xxxxxxxxxx>
Sent by: ve-dev-bounces@xxxxxxxxxxx

08/31/2006 01:50 AM

Please respond to
Discussions people developing code for the Visual Editor project <ve-dev@xxxxxxxxxxx>

To	Discussions people developing code for the Visual Editor project <ve-dev@xxxxxxxxxxx>
cc
Subject	[ve-dev] Initialization error with VisualComponentAdapter#componentValidated()

Rich -- thanks very much for getting the "fix"
in for the TreeVisualContainerEditpolicy.activate()
problem.

As you may remember, I have been getting a rather strange initialization
message (
http://dev.eclipse.org/mhonarc/lists/ve-dev/msg01342.html).
 Think I have finally localized the source.

I am using VisualComponentAdapter#componentValidated() as a trigger to
reconcile layout configuration changes in the remote VM with the model
held by my helper.  To implement the reconcile, I am doing a number
of calls out to my remoteVMAdapter to dig out the informaiton I need.  (These
calls are being made within the callback thread -- which appears to already
be an async thread.)

During initialization of VE on an existing source file, VisualComponentAdapter#componentValidated()
is being called, starting a reconcile.  Invariably, one of the calls
to the remoteVMAdapter will fail, provoking the error message.  The
specific call that fails seems to vary, as does the content of the error
message, *apparently* depending on the number of formlayout/JPanels are
in the design.  The listing in the prior message comes from a design
with three formlayout/JPanels.  For a single formlayout/JPanel, I
get only this as the error:

!ENTRY org.eclipse.jem.proxy 4 0 2006-08-30 21:08:35.625
!MESSAGE In REMCallbackThread
!STACK 0
java.net.SocketException: Socket closed
at java.net.SocketOutputStream.socketWrite(Unknown Source)
at java.net.SocketOutputStream.write(Unknown Source)
at java.io.BufferedOutputStream.flushBuffer(Unknown Source)
at java.io.BufferedOutputStream.flush(Unknown Source)
at java.io.DataOutputStream.flush(Unknown Source)
at org.eclipse.jem.internal.proxy.common.remote.Commands.sendErrorCommand(Commands.java:1372)
at org.eclipse.jem.internal.proxy.common.remote.Commands.sendCallbackDoneCommand(Commands.java:1134)
at org.eclipse.jem.internal.proxy.remote.REMCallbackThread.run(REMCallbackThread.java:246)

Currently, I am flagging and ignoring the first  #componentValidated()
callback, thereby skipping the reconcile during VE initialization.  Subsequent
#componentValidated()/reconciles work fine.

Question is, is this just an initialization problem internal to VE -- and
have I identified a valid workaround -- or is my doing the reconcile on
the callback thread not proper?  If not, what is the correct way to
use #componentValidated() to initiate the reconcile process.

Appreciate any help and advice you can give,
Gerald
_______________________________________________
ve-dev mailing list
ve-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ve-dev