Bugzilla – Full Text Bug Listing
| Summary: | Segmentation Fault When Widget looses focus | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | [Eclipse Project] Platform | Reporter: | Kris De Volder <kdevolder> | ||||||||||
| Component: | SWT | Assignee: | Felipe Heidrich <eclipse.felipe> | ||||||||||
| Status: | RESOLVED FIXED | QA Contact: | |||||||||||
| Severity: | critical | ||||||||||||
| Priority: | P3 | CC: | andrew.eisenberg, eclipse.felipe, jaroslav.resler, mseele, pwebster, remy.suen, Silenio_Quarti | ||||||||||
| Version: | 3.6 | Flags: | Silenio_Quarti:
review+
|
||||||||||
| Target Milestone: | 3.6.1 | ||||||||||||
| Hardware: | PC | ||||||||||||
| OS: | Linux-GTK | ||||||||||||
| Whiteboard: | |||||||||||||
| Attachments: |
|
||||||||||||
|
Description
Kris De Volder
Created attachment 172703 [details]
crahs log
Thought I already attached this, but seems I didn't.
I've got some more info which may help. I am still able to reproduce this bug 100% accurately. This sad in a sense, since I'm trying to find a workaround... But it also gives me a chance to try and track down where exactly it happens. I can put a breakpoint in an eventlistener just before the bug happens and then a few steps backing out of the stack and into an event dispatch brings me here: Thread [main] (Suspended (breakpoint at line 753 in Shell)) Shell.filterProc(int, int, int) line: 753 Display.filterProc(int, int, int) line: 1537 OS._g_main_context_iteration(int, boolean) line: not available [native method] OS.g_main_context_iteration(int, boolean) line: 2224 Display.readAndDispatch() line: 3169 Workbench.runEventLoop(Window$IExceptionHandler, Display) line: 2629 Workbench.runUI() line: 2593 Workbench.access$4(Workbench) line: 2427 Workbench$7.run() line: 670 Realm.runWithDefault(Realm, Runnable) line: 332 Workbench.createAndRunWorkbench(Display, WorkbenchAdvisor) line: 663 PlatformUI.createAndRunWorkbench(Display, WorkbenchAdvisor) line: 149 IDEApplication.start(IApplicationContext) line: 115 EclipseAppHandle.run(Object) line: 196 EclipseAppLauncher.runApplication(Object) line: 110 EclipseAppLauncher.start(Object) line: 79 EclipseStarter.run(Object) line: 369 EclipseStarter.run(String[], Runnable) line: 179 NativeMethodAccessorImpl.invoke0(Method, Object, Object[]) line: not available [native method] NativeMethodAccessorImpl.invoke(Object, Object[]) line: 39 DelegatingMethodAccessorImpl.invoke(Object, Object[]) line: 25 Method.invoke(Object, Object...) line: 597 Main.invokeFramework(String[], URL[]) line: 619 Main.basicRun(String[]) line: 574 Main.run(String[]) line: 1407 Main.main(String[]) line: 1383 The version of the Jar this stacktrace originates in is: /home/kdvolder/Applications/springsource/sts-2.3.3.201006180757/plugins/org.eclipse.swt.gtk.linux.x86_3.6.0.v3650b.jar An what I'm seeing there (looking at it right now) is this code: case OS.FocusOut: if (xFocusEvent.mode == OS.NotifyNormal || xFocusEvent.mode == OS.NotifyWhileGrabbed) { switch (xFocusEvent.detail) { case OS.NotifyNonlinear: case OS.NotifyNonlinearVirtual: case OS.NotifyVirtual: if (tooltipsHandle != 0) OS.gtk_tooltips_disable (tooltipsHandle); Display display = this.display; sendEvent (SWT.Deactivate); setActiveControl (null); if (display.activeShell == this) { display.activeShell = null; display.activePending = false; } if (isCustomResize ()) { === HERE ===> OS.gdk_window_invalidate_rect (OS.GTK_WIDGET_WINDOW (shellHandle), null, false); } break; } } break; And in the variables of the debugger, I can see that the "this" pointer of this Shell displays as Shell {*Disposed*} Then, when we step this line: OS.gdk_window_invalidate_rect (OS.GTK_WIDGET_WINDOW (shellHandle), null, false) => KABOOM! In case this helps as well, this the stack at the point where I put the breakpoint. It's inside a listener that wants to be notified of deactivation of a popup style window/widget (like content assist in Eclipse). Thread [main] (Suspended (breakpoint at line 1088 in GrailsInplaceDialog$11)) GrailsInplaceDialog$11.proposalPopupClosed(ContentProposalAdapter) line: 1088 ContentProposalAdapter.notifyPopupClosed() line: 2119 ContentProposalAdapter.access$8(ContentProposalAdapter) line: 2112 ContentProposalAdapter$ContentProposalPopup.close() line: 916 ContentProposalAdapter$ContentProposalPopup$PopupCloserListener.handleEvent(Event) line: 132 EventTable.sendEvent(Event) line: 84 Shell(Widget).sendEvent(Event) line: 1258 Shell(Widget).sendEvent(int, Event, boolean) line: 1282 Shell(Widget).sendEvent(int) line: 1263 Shell.filterProc(int, int, int) line: 746 Display.filterProc(int, int, int) line: 1537 OS._g_main_context_iteration(int, boolean) line: not available [native method] OS.g_main_context_iteration(int, boolean) line: 2224 Display.readAndDispatch() line: 3169 Workbench.runEventLoop(Window$IExceptionHandler, Display) line: 2629 Workbench.runUI() line: 2593 Workbench.access$4(Workbench) line: 2427 Workbench$7.run() line: 670 Realm.runWithDefault(Realm, Runnable) line: 332 Workbench.createAndRunWorkbench(Display, WorkbenchAdvisor) line: 663 PlatformUI.createAndRunWorkbench(Display, WorkbenchAdvisor) line: 149 IDEApplication.start(IApplicationContext) line: 115 EclipseAppHandle.run(Object) line: 196 EclipseAppLauncher.runApplication(Object) line: 110 EclipseAppLauncher.start(Object) line: 79 EclipseStarter.run(Object) line: 369 EclipseStarter.run(String[], Runnable) line: 179 NativeMethodAccessorImpl.invoke0(Method, Object, Object[]) line: not available [native method] NativeMethodAccessorImpl.invoke(Object, Object[]) line: 39 DelegatingMethodAccessorImpl.invoke(Object, Object[]) line: 25 Method.invoke(Object, Object...) line: 597 Main.invokeFramework(String[], URL[]) line: 619 Main.basicRun(String[]) line: 574 Main.run(String[]) line: 1407 Main.main(String[]) line: 1383 When I am at this breakpoint I can see that the shell is already disposed... Let me know if I can do some more digging (before I figure out a workaround and won't be able to reproduce this anymore :-) One more stacktrace, this one points to the actual call to Shell.dispose that causes the shell to be disposed. Thread [main] (Suspended (breakpoint at line 2236 in Shell)) Shell.dispose() line: 2236 ContentProposalAdapter$ContentProposalPopup(Window).close() line: 335 ContentProposalAdapter$ContentProposalPopup(PopupDialog).close() line: 1191 ContentProposalAdapter$ContentProposalPopup.close() line: 915 ContentProposalAdapter$ContentProposalPopup$PopupCloserListener.handleEvent(Event) line: 132 EventTable.sendEvent(Event) line: 84 Shell(Widget).sendEvent(Event) line: 1258 Shell(Widget).sendEvent(int, Event, boolean) line: 1282 Shell(Widget).sendEvent(int) line: 1263 Shell.filterProc(int, int, int) line: 746 Display.filterProc(int, int, int) line: 1537 OS._g_main_context_iteration(int, boolean) line: not available [native method] OS.g_main_context_iteration(int, boolean) line: 2224 Display.readAndDispatch() line: 3169 Workbench.runEventLoop(Window$IExceptionHandler, Display) line: 2629 Workbench.runUI() line: 2593 Workbench.access$4(Workbench) line: 2427 Workbench$7.run() line: 670 Realm.runWithDefault(Realm, Runnable) line: 332 Workbench.createAndRunWorkbench(Display, WorkbenchAdvisor) line: 663 PlatformUI.createAndRunWorkbench(Display, WorkbenchAdvisor) line: 149 IDEApplication.start(IApplicationContext) line: 115 EclipseAppHandle.run(Object) line: 196 EclipseAppLauncher.runApplication(Object) line: 110 EclipseAppLauncher.start(Object) line: 79 EclipseStarter.run(Object) line: 369 EclipseStarter.run(String[], Runnable) line: 179 NativeMethodAccessorImpl.invoke0(Method, Object, Object[]) line: not available [native method] NativeMethodAccessorImpl.invoke(Object, Object[]) line: 39 DelegatingMethodAccessorImpl.invoke(Object, Object[]) line: 25 Method.invoke(Object, Object...) line: 597 Main.invokeFramework(String[], URL[]) line: 619 Main.basicRun(String[]) line: 574 Main.run(String[]) line: 1407 Main.main(String[]) line: 1383 I have got a fix / workaround here.
We can wrap the call to close from handle event
At stacktrace line:
ContentProposalAdapter$ContentProposalPopup$PopupCloserListener.handleEvent(Event)
in a call to async exec like so
// For all other events, merely getting them dictates closure.
/* STS Change: new code: */
e.display.asyncExec(new Runnable() {
public void run() {
close();
}
});
/* STS Change: old code:
close();
*/
Sorry that got a bit messed up (why is this darned text layout contrained to so few colums?)
//Change at line: 132
//Class: org.eclipse.jface.fieldassist.ContentProposalAdapter
//Version of the jar: org.eclipse.jface_3.6.0.I20100601-0800.jar
// For all other events, merely getting them dictates closure.
/* STS Change: new code: */
e.display.asyncExec(new Runnable() {
public void run() {
close();
}
});
/* STS Change: old code: (line 132)
close();
*/
Created attachment 173751 [details]
Snippet that shows the bug
i can also reproduce this bug. I've wrote a snippet that shows the bug.
Please execute the snippet, set focus to text field. Press space key to show proposal adapter, then click into proposal adapter to get them the focus. If you now click into the shell, vm crashes.
Created attachment 173755 [details]
patch which fiexs the problem
I've investigated something around this bug:
- #23980 opens this bug
- the crash is, like kris said before, in Shell:753. gdk_window_invalidate_rect is made on a disposed shell
- i've added a patch which fixes the bug by checking the shell disposed state
*** Bug 319119 has been marked as a duplicate of this bug. *** These are the steps we believe should reproduce the problem in a regular Eclipse SDK without additional plug-ins. 1) From a java editor, I open CTRL+F. 2) check regular expressions 3) Invoke Ctrl+Space in the 'Find' text field. 4) Click on an entry in the popup, but don't double click 5) Click inside the editor in the "background". When will be Michaels patch available in eclipse updates? This is a critical bug causing eclipse to be unsuitable for regular work. Created attachment 176067 [details]
patch for HEAD
Michaels was right on the money, the shell was being disposed by the deactive event handler. thanks for the test case and the fix.
Fixed in HEAD > 20100806
Silenio, please review for 3.6.1 Note: the patch for HEAD is good for the R3_6_maintenance stream as well. Fixed in R3_6_maintenance > 2010-08-06 |