[news.eclipse.modeling.gmf] Re: StackOverflowError with shared EditingDo

[Rene Ladan <r.c.ladan@xxxxxx>]

Rene Ladan wrote:
> Dawid wrote:
> > Hi,
> >
> > I am having the same kind of problem. Could anybody help us out ?
> Maybe the GMF build is somehow bad?  No updates received for GMF since 
> 20070621
>
> > Rene Ladan schrieb:
> > > Andrew Montalenti wrote:
> > > > Hi everyone,
> > > >
> > > > I am trying to get the generated GMF diagram editor to share its 
> > > > TransactionalEditingDomain with my custom views and the EMF editor.
> > > >  
> > > [..snip overrides..]
> > >
> > > > Now, what ends up happening is the EditingDomain sharing seems to be 
> > > > working for certain operations.  Notably, renames, property changes, 
> > > > copy/
> > > > paste, even drag and drop.  All of these operations take effect 
> > > > without error.
> > > >  
> > > > However, any _pure create_ events -- creating a new Node, or 
> > > > creating a child of a Node inside a compartment, or creating a 
> > > > connection among nodes -- in short, anything that utilizes a 
> > > > CreateElementCommand, and is accessible via the Palette Creation 
> > > > Tools, causes a huge stack trace.
> > > >  
> > > > The cause of failure is a StackOverflowError.
> > >
> > > I'm seeing the same kind of error, although my stack trace (attached) 
> > > doesn't display anything useful at the bottom (maybe the message 
> > > stack was full?):
> > >
> > > !SESSION 2007-08-13 10:17:07.067 
> > > -----------------------------------------------
> > > eclipse.buildId=I20070601-1539
> > > java.version=1.6.0_01
> > > java.vendor=Sun Microsystems Inc.
> > > BootLoader constants: OS=win32, ARCH=x86, WS=win32, NL=nl_NL
> > > Framework arguments:  -product org.eclipse.sdk.ide -pdelaunch
> > > Command-line arguments:  -product org.eclipse.sdk.ide -data 
> > > C:\Documents and Settings\rladan\workspace\mt-test -dev 
> > > file:C:/Documents and 
> > > Settings/rladan/workspace/.metadata/.plugins/org.eclipse.pde.core/metatest/dev.properties 
> > > -pdelaunch -debug C:\Documents and 
> > > Settings\rladan\workspace\.metadata\.plugins\org.eclipse.pde.core\metatest/.options 
> > > -os win32 -ws win32 -arch x86
> > >
> > > <whole bunch of "MESSAGE NLS unused" messages>
> > >
> > > !ENTRY org.eclipse.ui 4 0 2007-08-13 10:18:13.509
> > > !MESSAGE Unhandled event loop exception
> > > !STACK 0
> > > java.lang.StackOverflowError
> > >     at 
> > > org.eclipse.core.runtime.ProgressMonitorWrapper.internalWorked(ProgressMonitorWrapper.java:94) 
> > >
> > >      at 
> > > org.eclipse.core.runtime.SubProgressMonitor.internalWorked(SubProgressMonitor.java:155) 
> > >
> > >      <ad infinitum>
> > >
> > > > This is especially puzzling for me because certain operations -- 
> > > > like a node Copy/Paste operation -- actually create a new element in 
> > > > the semantic model, and this occurs without error.  Only "pure 
> > > > creation" commands will fail.
> > > Here too.
> > >
> > > > The following notes may also be helpful:
> > > >
> > > >    1. although this stack trace occurs, the semantic element is 
> > > > actually still created, but the notational view for it (the 
> > > > NodeEditPart in the case of a Node) is not created.
> > > Here too, but if I close and re-open the resource, GMF shows the 
> > > newly created element.
> > >
> > > >    2. a knock-on effect is that after this stack overflow occurs, 
> > > > the diagram editor no longer functions properly.  Even if I close 
> > > > all resources and re-open them, the diagram editor will not allow 
> > > > any changes to the View -- you can't move a NodeEditPart, change its 
> > > > color, or anything like that.
> > > Here too, but if I close and re-open the resource, movements work again.
> > >
> > > [..snip renaming..]
I did a fresh installation of Eclipse in a new directory to see if the above
problems disappeared, but they didn't :(.  Versions :
* Eclipse 3.4 Integration 20070921919
* EMF 2.3.1 20070925
* OCL 2.0.* 20070921
* GEF 3.4.0 20070919
* GMF SDK 2.0 20070903 (Integration)
* Apache Batik 1.6.0 200706111724

I created a new test project (the coffee tutorial from OnJava.com) which
actually works fine.  Steps:
1. Create a new GMF project
2a. Import the ecore, gmfmap, gmftool, gmfgraph files
2b. Set the nsURI/nsPrefix fiels in the ecore file
2c. Migrate the gmfgraph and gmfmap files to GMF 2.0
3. Generate the genmodel, gmfgen files using the defaults
4. Generate the code

So maybe the stack problems are gmf-model related?  In both cases all models
validate fine.

> > > > p.s. I'm using GMF 2.0RC1
> > >
> > > Eclipse 3.3.0 I20070625-1500, EMF 2.3.0v20070626, GEF 3.3.0v20070620, 
> > > GMF 1.0.100.v20070601, JDK 1.6.0 update 2
> > >
> > > The strange thing is that it used to work.  A working setup seemed to 
> > > be   one before July 10th.  After that I _did_ change the models:
> > >
> > > 1. change & validate the ecore model because of some new requirements
> > > 2. update the genmodel file
> > > 3. regenerate the EMF editor (ok)
> > > Now the GMF editor is broken as described.
> > > 4. update & validate the gmfgraph file to match some renamings in the 
> > > ecore file.
> > > 5. check if the gmftool file needs updating (no)
> > > 6. update & validate the gmfmap file
> > > 7. regenerate & validate the gmfgen file
> > > 8. regenerate the GMF editor
> > > Now it is still broken.
> > >
> > > Regards,
> > > Rene