Community
Participate
Working Groups
Since the update of the Eclipse Platform on SWT and auto-scaling, gmf-runtime based diagrams have issues to be rendered on some Windows platform. This can be for example reproduced on the geoshape example shipped with the modeling tools distribution Usage of the argument -Dswt.enable.autoScale=false removes the graphical glitches. Version of Eclipse: Eclipse Modeling Tools Version: Neon Milestone 7 (4.6.0M7) Build id: 20160505-1310
Created attachment 261863 [details] Comparison of 2 debug sessions on the same project with autoscaling disabled on the left and enabled on the right
I am trying to escalate this critical issue within our own IBM internal Eclipse support team to connect with the SWT team and determine what to do here. If anyone on the CC list is aware of a code fix we can apply to the GMF Runtime please let us know.
(In reply to Remi Schnekenburger from comment #0) > Since the update of the Eclipse Platform on SWT and auto-scaling, > gmf-runtime based diagrams have issues to be rendered on some Windows > platform. This can be for example reproduced on the geoshape example shipped > with the modeling tools distribution > > Usage of the argument -Dswt.enable.autoScale=false removes the graphical > glitches. > > Version of Eclipse: > Eclipse Modeling Tools > > Version: Neon Milestone 7 (4.6.0M7) > Build id: 20160505-1310 Are you using a high resolution monitor? or using zoomed resolution, what is the zoom factor? Can you try with a more recent build like 4.6RC1 or 4.6RC2?
(In reply to Lakshmi Shanmugam from comment #3) > (In reply to Remi Schnekenburger from comment #0) > > Since the update of the Eclipse Platform on SWT and auto-scaling, > > gmf-runtime based diagrams have issues to be rendered on some Windows > > platform. This can be for example reproduced on the geoshape example shipped > > with the modeling tools distribution > > > > Usage of the argument -Dswt.enable.autoScale=false removes the graphical > > glitches. > > > > Version of Eclipse: > > Eclipse Modeling Tools > > > > Version: Neon Milestone 7 (4.6.0M7) > > Build id: 20160505-1310 > > Are you using a high resolution monitor? or using zoomed resolution, what is > the zoom factor? Can you try with a more recent build like 4.6RC1 or 4.6RC2? By default windows runs at 125%. In M7 we supported 25% intervals. Due to this they are seeing this problem. In RC1 we have reverted to 100% multiples. They may not see this problem now. We definitely need to test it at 200% to confirm whether this problem exists or not.
As explained in bug 494145, I reproduce the problem with this configuration: * Windows 10 * Resolution : 3840x2160 * Font size modification : 250% (recommanded) * org.eclipse.swt.internal.deviceZoom=200 * java.runtime.version=1.8.0_77-b03 * org.eclipse.draw2d.sdk (3.10.100.201605020204) "Draw2d" * org.eclipse.gmf.runtime.sdk (1.10.0.201605032009) "Graphical Modeling Framework Runtime SDK"
Created attachment 261933 [details] Example of background not correctly drawn With the above configuration, it seems that there is a link between the backgound shift and the org.eclipse.swt.internal.deviceZoom (200%).
Couple of queries: 1. Do you have any simplified SWT snippet to reproduce this issue at our end. Please share detailed steps to reproduce this issue, suggest you to give it a try with latest Eclipse RC1 J2EE build(it has may important fixes w.r.t. HighDPI) https://www.eclipse.org/downloads/packages/eclipse-ide-java-ee-developers/neonrc1 2. Can you see this issue on Linux/GTK at 200% zoom settings ?
(In reply to Niraj Modi from comment #7) > Couple of queries: > 1. Do you have any simplified SWT snippet to reproduce this issue at our end. > Please share detailed steps to reproduce this issue, suggest you to give it > a try with latest Eclipse RC1 J2EE build(it has may important fixes w.r.t. > HighDPI) > https://www.eclipse.org/downloads/packages/eclipse-ide-java-ee-developers/ > neonrc1 > I have not really a simple snippet. But I have scenario with a fresh NeonRc1 modeling install: 1- Download Eclipse Neon RC1 Modeling: http://www.eclipse.org/downloads/packages/eclipse-modeling-tools/neonrc1 2- Create a new "Modeling Project" 3- Create a new "Ecore Model" in this project 4- Activate the Viewpoint "Design" on this project: use contextual menu "Viewpoints Selection" on the project 5- Create a diagram of the root package of the ecore file: use contextual menu "New Representation/xxx class diagram" on the root package of ecore file 6- Add a Note in this diagram: KO, the background of the Node is not correctly drawn. 6bis- Step 2 to 6 can be replaced by importing the project from SampleProject.zip and open diagram "root class diagram" > 2. Can you see this issue on Linux/GTK at 200% zoom settings ? I try it on a virtual box with the above scenario and the problem is here.
Created attachment 261937 [details] SampleProject.zip for steps to reproduce of comment 8
Thank you for providing the steps. We found the root cause this problem is there in the SWT layer. In GMF drawing the background is done in the following steps 1. Get the points list (corners/vertices) 2. create SWT Path (at this point the SWT scales up the points and addtem to the path) 3. Get the points list from the SWT Path (extract the points again from the created path) 4. Scale up or down based on the zoom selected in the diagram(please note this is zoom at GMF level) 5. recreate the path 6. Use it to fill the background here at step 3 SWT in required to scaledown the points list. But this is missed. Due to this the background dimensions are not calculated correctly. Actually scaling is getting applied twice at step 2 and step 5. This the root cause. the fix is very simple and I will attach gerrit patch very shortly
New Gerrit change created: https://git.eclipse.org/r/73418
(In reply to Eclipse Genie from comment #11) > New Gerrit change created: https://git.eclipse.org/r/73418 +1 Latest patch looks good to me, verified tested on Win7 as per steps in comment 8 on the bug.
Created attachment 261974 [details] snippet to reproduce the problem
Pushed patch set 4 that removes a redundant "if (isDisposed()) ..." on Windows. +1 for RC3. Verified on GTK3 and Win 7.
Gerrit change https://git.eclipse.org/r/73418 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=916064708a285dbc8d069868f2e6a9128285c28e
Verified on I20160525-2000 in linux 64 bit.
Verified the fix in Build id: I20160525-2000 on Win7.
