Community
Participate
Working Groups
I am running into a deadlock situation. I seems like Highlighting and Outline are deadlocking. I have let the IDE run for a bit and then paused, when repeating it is stuck in the same two places... (See stack tracess below) The problem can be replicated - I always get the problem when in a particular position in the source text. If I remove my semantic highlighting, it does not hang, and I get an error marker instead (because of a syntax error - which is expected). Do I need to protect the highlighting in a UnitOfWork some way, or should it work without this? I read this: "Please refer to DefaultContentOutlineNodeFactory.getEObjectHandle(EObject, IStateAccess<XtextResource>) to learn more about IEObjectHandles. ", and I have seen that I may not work directly with the model. And I looked at the mentioned code, but I just don't get it... What I am doing now is like this... public void provideHighlightingFor2(XtextResource resource, IHighlightedPositionAcceptor acceptor) { if(resource == null) return; EList<EObject> contents = resource.getContents(); if(contents == null || contents.size() == 0) return; // nothing there at all - probably an empty file BeeModel model = (BeeModel) contents.get(0); // BuildUnit bu = model.getBody(); TreeIterator<EObject> all = model.eAllContents(); while(all.hasNext()) { EObject o = all.next(); // TODO: OPTIMIZE the if-then-else below using an EMF switch if(o instanceof BuildUnit) highlightBuildUnit((BuildUnit) o, acceptor); // handle literals else if(o instanceof BLiteralExpression) { Object val = ((BLiteralExpression) o).getValue(); // handle real numbers (INT and HEX are handled by terminal rules) if(val instanceof Float || val instanceof Double) highlightObject(o, BeeLangHighlightConfiguration.REAL_ID, acceptor); } else if(o instanceof PathVector) highlightPaths((PathVector) o, acceptor); // etc. etc... Where the the various highlightObject, highlightPaths, use a variant of finding "firstFeature" as I seen used by others... Am I doing something wrong? If I should use "readOnly" and "IUnitOfWork", how to I get to the model from the XtextResource passed in in the call to provideHighlightingFor ? (The version on b3's HEAD has the semantic highlighting disabled since I am working on a proposal provider for the position where this issue occurs...) // provideHighlightingFor2(resource, acceptor); Regards - henrik Daemon Thread [Thread-0] (Suspended) Unsafe.park(boolean, long) line: not available [native method] LockSupport.park(Object) line: 158 ReentrantReadWriteLock$NonfairSync(AbstractQueuedSynchronizer).parkAndCheckInterrupt() line: 747 ReentrantReadWriteLock$NonfairSync(AbstractQueuedSynchronizer).doAcquireShared(int) line: 877 ReentrantReadWriteLock$NonfairSync(AbstractQueuedSynchronizer).acquireShared(int) line: 1197 ReentrantReadWriteLock$ReadLock.lock() line: 594 XtextDocument$XtextDocumentLocker(IStateAccess$AbstractImpl<P>).readOnly(IUnitOfWork<T,P>) line: 36 XtextDocument.readOnly(IUnitOfWork<T,XtextResource>) line: 58 XtextContentOutlinePage.synchronizeOutlinePage() line: 312 EditorSelectionChangedListener.selectionChanged(SelectionChangedEvent) line: 31 XtextSourceViewer(TextViewer).firePostSelectionChanged(SelectionChangedEvent) line: 2745 XtextSourceViewer(TextViewer).firePostSelectionChanged(int, int) line: 2693 TextViewer$5.run() line: 2672 Display.runTimers() line: 3742 Display.readAndDispatch() line: 3256 Workbench.runEventLoop(Window$IExceptionHandler, Display) line: 2624 Workbench.runUI() line: 2588 Workbench.access$4(Workbench) line: 2422 Workbench$7.run() line: 670 Realm.runWithDefault(Realm, Runnable) line: 332 Workbench.createAndRunWorkbench(Display, WorkbenchAdvisor) line: 663 PlatformUI.createAndRunWorkbench(Display, WorkbenchAdvisor) line: 149 IDEApplication.start(IApplicationContext) line: 115 EclipseAppHandle.run(Object) line: 196 EclipseAppLauncher.runApplication(Object) line: 110 EclipseAppLauncher.start(Object) line: 79 EclipseStarter.run(Object) line: 369 EclipseStarter.run(String[], Runnable) line: 179 NativeMethodAccessorImpl.invoke0(Method, Object, Object[]) line: not available [native method] NativeMethodAccessorImpl.invoke(Object, Object[]) line: 39 DelegatingMethodAccessorImpl.invoke(Object, Object[]) line: 25 Method.invoke(Object, Object...) line: 597 Main.invokeFramework(String[], URL[]) line: 619 Main.basicRun(String[]) line: 574 Main.run(String[]) line: 1407 Main.main(String[]) line: 1383 AND Thread [Worker-3] (Suspended) MergingHighlightedPositionAcceptor.mergePositions(int, int, int, LightweightPosition$IntToStringArray[]) line: 88 MergingHighlightedPositionAcceptor.mergePositions() line: 71 MergingHighlightedPositionAcceptor.provideHighlightingFor(XtextResource, IHighlightedPositionAcceptor) line: 47 HighlightingReconciler.reconcilePositions(XtextResource) line: 87 HighlightingReconciler.modelChanged(XtextResource) line: 275 XtextDocument.notifyModelListeners(XtextResource) line: 90 XtextDocument$XtextDocumentLocker.afterModify(XtextResource, Object, IUnitOfWork<?,XtextResource>) line: 142 XtextDocument$XtextDocumentLocker.afterModify(Object, Object, IUnitOfWork) line: 1 XtextDocument$XtextDocumentLocker(IStateAccess$AbstractImpl<P>).modify(IUnitOfWork<T,P>) line: 58 XtextDocument$XtextDocumentLocker.modify(IUnitOfWork<T,XtextResource>) line: 149 XtextDocument.modify(IUnitOfWork<T,XtextResource>) line: 62 XtextDocumentReconcileStrategy.reconcile(IRegion) line: 27 XtextReconciler.run(IProgressMonitor) line: 179 Worker.run() line: 54
Copied from a forum message by Henrik (news://news.eclipse.org:119/i0jmvv$f4f$1@build.eclipse.org)
Sebastian wrote >can you confirm that it's an endless loop in MergedHighlightingPositionAcceptor.mergePositions? So far, I have only paused the execution and looked at where the different threads were executing - did not look at exact positions, but "mergePositions" was always on the call stack (afaict). How can I provide more info? To replicate: If you update your b3 workspace remove the o.e.b3.backend.[edit, editor and test] projects (obsolete generated stuff). Then uncomment line 145 in /org.eclipse.b3.beelang.ui/src/org/eclipse/b3/coloring/BeeLangSemanticHighlightingCalculator.java which says: // provideHighlightingFor2(resource, acceptor); And with a b3 file like this: unit HelloDeadlock version | { } Place the marker at | (if you get that far after typing 'version'), try typing a version '1.0.0', or invoke proposals.
I'ld guess that the problem does not disappear if you close the outline before you perform the necessary steps to reproduce the locked ui. A list of offset / lengths / ids that are passed to the HighlightedPositionAcceptor would be great, e.g something like 1, 5, 'myId' 3, 5, 'myIdOverlay', 'anotherId' 4, 6, ..
(In reply to comment #3) > I'ld guess that the problem does not disappear if you close the outline before > you perform the necessary steps to reproduce the locked ui. > > A list of offset / lengths / ids that are passed to the > HighlightedPositionAcceptor would be great, e.g something like > > 1, 5, 'myId' > 3, 5, 'myIdOverlay', 'anotherId' > 4, 6, .. Does not help to close the outline (tried that). Will see what I can generate in terms of data...
I wrote a wrapper to record what was called. For this input: unit a version { } unit b { } This is the positions and styles 5, 1, default 25, 1, default 14, 0, version And the last output gave me a clue to the cause. So, I filtered out requests to highlight something of zero length - and guess what... Now it does not hang anymore.
Thanks Henrik. I changed the subject of this bugzilla and scheduled it for SR1.
Fixed in HEAD.
I've run into this bug again in XText v2.2.0 Filtering the nodes of length 0, suggested here, did not help. Debugging shows that mergePositions() of MergingHighlightedPositionAcceptor is stuck into an endless while() loop, but only for certain files, but I've yet to understand the pattern. I'll post more information as soon as I have significant test cases.
Closing all bugs that were set to RESOLVED before Neon.0
