Bugzilla – Full Text Bug Listing
| Summary: | [Compatibility] NPE opening editor | ||
|---|---|---|---|
| Product: | [Eclipse Project] Platform | Reporter: | Ed Willink <ed> |
| Component: | UI | Assignee: | Platform-UI-Inbox <Platform-UI-Inbox> |
| Status: | RESOLVED INVALID | QA Contact: | |
| Severity: | normal | ||
| Priority: | P3 | CC: | adolfosbh, david_williams, remy.suen, tom.schindl |
| Version: | 4.2 | ||
| Target Milestone: | 4.2 M6 | ||
| Hardware: | PC | ||
| OS: | Windows Vista | ||
| Whiteboard: | |||
| Bug Depends on: | |||
| Bug Blocks: | 366990, 367844 | ||
|
Description
Ed Willink
What happens on Windows on 4.2? PW (In reply to comment #0) > Caused by: java.lang.NullPointerException > at > org.eclipse.ui.internal.WorkbenchPage.setPerspective(WorkbenchPage.java:3312) > at org.eclipse.ui.internal.WorkbenchWindow.setup(WorkbenchWindow.java:530) > at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) > at > sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39) > at > sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25) > at java.lang.reflect.Method.invoke(Method.java:597) > at > org.eclipse.e4.core.internal.di.MethodRequestor.execute(MethodRequestor.java:56) > ... 59 more Odd place to fail. I'd have expected another error that precedes this one. Which build of Eclipse 4.x are you testing with? (In reply to comment #2) > Which build of Eclipse 4.x are you testing with? The Buckminster scripts look like their after <rm:uri format="{0}/eclipse/updates/4.2-I-builds"> The Hudson workspace contains https://hudson.eclipse.org/hudson/job/buckminster-mdt-ocl-branch-tests/ws/buildroot/target.platform/plugins/org.eclipse.e4.core.di_1.1.0.v20111221-1333.jar Does this NPE occur if you are using Eclipse regularly or only when the tests are being run? Just checked with 4.2M4 interactively; no apparent problem. Problem is for JUnit plugin tests on Hudson. (In reply to comment #5) > Problem is for JUnit plugin tests on Hudson. So they only fail when running on Hudson? Where can I get the source code for these tests? Also, you say "JUnit Plugin tests on Windows and Eclipse 3.8" was that on Hudson? If Hudson only (for 3.8, or 4.2) I'd be sure to remind you that a "display" is needed on Hudson ... I think most people specify this by checking "Run Xvnc during build" ... but suspect you, Ed, know this. Just reminding, just in case. The failing build was https://hudson.eclipse.org/hudson/job/buckminster-mdt-ocl-branch-tests/63/ The sources are all in: http://git.eclipse.org/c/mdt/org.eclipse.ocl.git. The tests are: tests\org.eclipse.ocl.examples.xtext.tests\src\org\eclipse\ocl\examples\test\xtext\EditorTests.java (In reply to comment #7) > If Hudson only (for 3.8, or 4.2) I'd be sure to remind you that a "display" is > needed on Hudson ... I think most people specify this by checking "Run Xvnc > during build" ... but suspect you, Ed, know this. Just reminding, just in case. I do see an error in the Hudson log complaining about a display. --------- Standard Error Xlib: extension "RANDR" missing on display ":12.0". (In reply to comment #8) > The failing build was > https://hudson.eclipse.org/hudson/job/buckminster-mdt-ocl-branch-tests/63/ The log is filled with errors complaining about EventAdmin. Can you ensure that the org.eclipse.equinox.event bundle is available to your build while it's running the tests? !ENTRY org.eclipse.e4.ui.workbench 4 0 2012-01-04 04:48:49.631 !MESSAGE Event Admin service is not available, unable to publish event org.osgi.service.event.Event [topic=org/eclipse/e4/ui/model/menu/ToolBarContributions/toolBarContributions/UNKNOWN]. (In reply to comment #7) > Also, you say > "JUnit Plugin tests on Windows > and Eclipse 3.8" > was that on Hudson? > > If Hudson only (for 3.8, or 4.2) I'd be sure to remind you that a "display" is > needed on Hudson ... I think most people specify this by checking "Run Xvnc > during build" ... but suspect you, Ed, know this. Just reminding, just in case. I suspect that there's something subtle here. The log file starts with 04/01/2012 04:34:49 Xvnc version X.org/xf4vnc custom version 04/01/2012 04:34:49 Copyright (C) 2001-2004 Alan Hourihane. ... and ends with Terminating xvnc. But after the problem tests there is another 04/01/2012 04:48:50 Xvnc version X.org/xf4vnc custom version 04/01/2012 04:48:50 Copyright (C) 2001-2004 Alan Hourihane. ... Perhaps this means that Xvnc is not running during the problem tests. Perhaps my releng can shed some light on why Xvnc gets triggered half way through. However if the problem is a bad device, a DI NPE doesn't seem like a good diagnostic. (In reply to comment #9) > I do see an error in the Hudson log complaining about a display. > Xlib: extension "RANDR" missing on display ":12.0". > > !ENTRY org.eclipse.e4.ui.workbench 4 0 2012-01-04 04:48:49.631 > !MESSAGE Event Admin service is not available, unable to publish event I cannot find any "RANDR" or "MESSAGE" in https://hudson.eclipse.org/hudson/job/buckminster-mdt-ocl-branch-tests/63/consoleFull If an Event Admin is needed shouldn't it be there already? How do we add it? There are many non-error errors in Hudson/Buckminster/Athena logs so it is difficult to know what matters.... Could not init font path element /usr/share/fonts/TTF/, removing from list! Could not init font path element /usr/share/fonts/OTF, removing from list! (In reply to comment #9) > > I do see an error in the Hudson log complaining about a display. > > --------- > > Standard Error > > Xlib: extension "RANDR" missing on display ":12.0". > That error is pretty normal ... more of a "warning" if you need RANR (which you don't, as far as I know ... "resize and rotate"). > But after the problem tests there is another > > 04/01/2012 04:48:50 Xvnc version X.org/xf4vnc custom version > 04/01/2012 04:48:50 Copyright (C) 2001-2004 Alan Hourihane. > ... > > Perhaps this means that Xvnc is not running during the problem tests. > > Perhaps my releng can shed some light on why Xvnc gets triggered half way > through. > I think I've raised a Red Herring ... if there is a "display" at all (and you have shown there is) you should be fine in that regard. Maybe some bundle not installed, as Remy suggests? (In reply to comment #11) > (In reply to comment #9) > > I do see an error in the Hudson log complaining about a display. > > > Xlib: extension "RANDR" missing on display ":12.0". > > > > !ENTRY org.eclipse.e4.ui.workbench 4 0 2012-01-04 04:48:49.631 > > !MESSAGE Event Admin service is not available, unable to publish event > > I cannot find any "RANDR" or "MESSAGE" in > https://hudson.eclipse.org/hudson/job/buckminster-mdt-ocl-branch-tests/63/consoleFull I was looking at the test failure's logs. > If an Event Admin is needed shouldn't it be there already? This is your standard API problem. The interface is available, it's that an implementation is not (the org.eclipse.equinox.event bundle). (In reply to comment #13) > I was looking at the test failure's logs. > > > If an Event Admin is needed shouldn't it be there already? > > This is your standard API problem. The interface is available, it's that an > implementation is not (the org.eclipse.equinox.event bundle). org.eclipse.pde.build is reported as missing and is indeed missing from the workspace. It looks like we're missing some Hudson/Buckminster configuration magic to enable the UI parts of the platform to load. (In reply to comment #7) > If Hudson only (for 3.8, or 4.2) I'd be sure to remind you that a "display" is > needed on Hudson ... I think most people specify this by checking "Run Xvnc > during build" ... but suspect you, Ed, know this. Just reminding, just in case. I'll just add that you also need to have a useful Window Manager running within your display for automated UI tests to do something useful (SWT depends on WM events). We use metacity on build.eclipse.org: metacity --replace --sm-disable But based on further comments it sounds like a configuration issue, so you might have that already running. PW (In reply to comment #15) > I'll just add that you also need to have a useful Window Manager running within > your display for automated UI tests to do something useful (SWT depends on WM > events). We use metacity on build.eclipse.org: > > metacity --replace --sm-disable > > But based on further comments it sounds like a configuration issue, so you > might have that already running. I don't think we have metacity so I think the onus is on us. I've added the MDT/OCL releng. (In reply to comment #14) > (In reply to comment #13) > > I was looking at the test failure's logs. > > > > > If an Event Admin is needed shouldn't it be there already? > > > > This is your standard API problem. The interface is available, it's that an > > implementation is not (the org.eclipse.equinox.event bundle). > > org.eclipse.pde.build is reported as missing and is indeed missing from the > workspace. > > It looks like we're missing some Hudson/Buckminster configuration magic to > enable the UI parts of the platform to load. If such a plugin with the interface implementation is required, can't the test plugin simply depend on that plugin, then buckminster would automatically install it in the target platform ?. (In reply to comment #15) > I'll just add that you also need to have a useful Window Manager running within > your display for automated UI tests to do something useful (SWT depends on WM > events). We use metacity on build.eclipse.org: > > metacity --replace --sm-disable > > But based on further comments it sounds like a configuration issue, so you > might have that already running. > > PW No idea about what metacity is ;P... I'll have to google it for info. Cheers, Adolfo. (In reply to comment #17) > > metacity --replace --sm-disable > > No idea about what metacity is ;P... I'll have to google it for info. One of the cross-project-dev observations on initial slave5 testing was that metacity wasn't installed; so I guess it is installed on fully functional Hudson. I suspect we just add the metacity line to a shell/startup script. Is this still an issue ? Is it an issue that we can help with or is it a config issue on Hudson ? Hi Eric, I think that the ball is in our roof. Yesterday, I wanted to try the metacity command in our tests job, however Eclipse servers are not helping to verify the solution. Let's see if we are luckier along the day. Regards, Adolfo. Ok, I've just confirmed that the option makes the offending tests (and the job) work [1]. Resolving as Invalid. BTW, I've noticed the following message when terminating xvnc: "Window manager warning: Fatal IO error 11 (Resource temporarily unavailable) on display ':10'." Paul, do you have any extra information/guide about this ? [1] https://hudson.eclipse.org/hudson/job/buckminster-mdt-ocl-branch-tests/81/ Cheers, Adolfo. (In reply to comment #21) > > "Window manager warning: Fatal IO error 11 (Resource temporarily unavailable) > on display ':10'." > > Paul, do you have any extra information/guide about this ? > I've never seen that one myself. Since it says "temp" then if it is not interfering with the tests I'd ignore it (but my experience here is lacking). PW (In reply to comment #22) > (In reply to comment #21) > > > > "Window manager warning: Fatal IO error 11 (Resource temporarily unavailable) > > on display ':10'." > > > > Paul, do you have any extra information/guide about this ? > > > > I've never seen that one myself. Since it says "temp" then if it is not > interfering with the tests I'd ignore it (but my experience here is lacking). > > PW I'll (temporarily) do it ;) Thanks, Adolfo. |