Integrate rust-analyzer and improve LSP Integration (required function, guards for empty replies, icons, deadlock fix)#9249
Conversation
…hDiagnostic Test with rust-analyzer yielded the following exception: java.lang.UnsupportedOperationException at org.eclipse.lsp4j.services.LanguageClient.refreshDiagnostics(LanguageClient.java:251) Caused: java.lang.reflect.InvocationTargetException at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:77) at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.base/java.lang.reflect.Method.invoke(Method.java:568) at org.eclipse.lsp4j.jsonrpc.services.GenericEndpoint.lambda$null$0(GenericEndpoint.java:65) Caused: java.lang.RuntimeException at org.eclipse.lsp4j.jsonrpc.services.GenericEndpoint.lambda$null$0(GenericEndpoint.java:67) at org.eclipse.lsp4j.jsonrpc.services.GenericEndpoint.request(GenericEndpoint.java:120) [catch] at org.eclipse.lsp4j.jsonrpc.RemoteEndpoint.handleRequest(RemoteEndpoint.java:261) at org.eclipse.lsp4j.jsonrpc.RemoteEndpoint.consume(RemoteEndpoint.java:190) at org.eclipse.lsp4j.jsonrpc.json.StreamMessageProducer.handleMessage(StreamMessageProducer.java:194) at org.eclipse.lsp4j.jsonrpc.json.StreamMessageProducer.listen(StreamMessageProducer.java:94) at org.eclipse.lsp4j.jsonrpc.json.ConcurrentMessageProcessor.run(ConcurrentMessageProcessor.java:113) at java.base/java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:539) at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264) at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136) at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635) at java.base/java.lang.Thread.run(Thread.java:840)
…cument It was observed, that with the rust-analyzer language server inserting a completion led to a dead-lock. Checking the stacks of running threads shows, that the LS tries to publish diagostics, while in parallel a documentFormat is called. The assumption is, that the rust side block execution of the `documentFormat` request until the `publishDiagnostics` call from the LS finishes. The latter though needs access to the document lock and that is already taken by the `documentFormat` task. "AWT-EventQueue-0" apache#35 prio=6 os_prio=0 cpu=8224,39ms elapsed=119,39s tid=0x00007f5384156f60 nid=0x2b0da7 waiting on condition [0x00007f537e7f9000] java.lang.Thread.State: WAITING (parking) at jdk.internal.misc.Unsafe.park(java.base@17.0.9/Native Method) - parking to wait for <0x00000006448493e8> (a java.util.concurrent.CompletableFuture$Signaller) at java.util.concurrent.locks.LockSupport.park(java.base@17.0.9/LockSupport.java:211) at java.util.concurrent.CompletableFuture$Signaller.block(java.base@17.0.9/CompletableFuture.java:1864) at java.util.concurrent.ForkJoinPool.unmanagedBlock(java.base@17.0.9/ForkJoinPool.java:3465) at java.util.concurrent.ForkJoinPool.managedBlock(java.base@17.0.9/ForkJoinPool.java:3436) at java.util.concurrent.CompletableFuture.waitingGet(java.base@17.0.9/CompletableFuture.java:1898) at java.util.concurrent.CompletableFuture.get(java.base@17.0.9/CompletableFuture.java:2072) at org.netbeans.modules.lsp.client.bindings.Formatter.documentFormat(Formatter.java:122) at org.netbeans.modules.lsp.client.bindings.Formatter.reformat(Formatter.java:87) at org.netbeans.modules.editor.indent.TaskHandler$MimeItem.runTask(TaskHandler.java:550) at org.netbeans.modules.editor.indent.TaskHandler.runTasks(TaskHandler.java:309) at org.netbeans.modules.editor.indent.IndentImpl.reformat(IndentImpl.java:349) at org.netbeans.modules.editor.indent.api.Reformat.reformat(Reformat.java:129) at org.netbeans.lib.editor.codetemplates.CodeTemplateInsertHandler.run(CodeTemplateInsertHandler.java:352) at org.netbeans.editor.GuardedDocument.runAtomicAsUser(GuardedDocument.java:333) at org.netbeans.lib.editor.codetemplates.CodeTemplateInsertHandler.insertTemplate(CodeTemplateInsertHandler.java:258) at org.netbeans.lib.editor.codetemplates.CodeTemplateInsertHandler.processTemplate(CodeTemplateInsertHandler.java:229) at org.netbeans.lib.editor.codetemplates.CodeTemplateManagerOperation.insert(CodeTemplateManagerOperation.java:273) at org.netbeans.lib.editor.codetemplates.api.CodeTemplate.insert(CodeTemplate.java:82) at org.netbeans.modules.lsp.client.bindings.CompletionProviderImpl$3.lambda$commit$0(CompletionProviderImpl.java:307) at org.netbeans.modules.lsp.client.bindings.CompletionProviderImpl$3$$Lambda$825/0x00007f53e4ae8498.run(Unknown Source) at org.netbeans.editor.GuardedDocument.runAtomic(GuardedDocument.java:296) at org.openide.text.NbDocument.runAtomic(NbDocument.java:411) at org.netbeans.modules.lsp.client.bindings.CompletionProviderImpl$3.commit(CompletionProviderImpl.java:254) at org.netbeans.modules.lsp.client.bindings.CompletionProviderImpl$3.defaultAction(CompletionProviderImpl.java:231) at org.netbeans.modules.editor.completion.CompletionImpl.dispatchKeyEvent(CompletionImpl.java:785) at org.netbeans.modules.editor.completion.CompletionImpl.keyPressed(CompletionImpl.java:386) at java.awt.AWTEventMulticaster.keyPressed(java.desktop@17.0.9/AWTEventMulticaster.java:258) [...] "pool-4-thread-1" apache#75 prio=5 os_prio=0 cpu=136,37ms elapsed=76,96s tid=0x00007f5380ae1080 nid=0x2b0e84 in Object.wait() [0x00007f52e99fc000] java.lang.Thread.State: WAITING (on object monitor) at java.lang.Object.wait(java.base@17.0.9/Native Method) - waiting on <no object reference available> at java.lang.Object.wait(java.base@17.0.9/Object.java:338) at javax.swing.text.AbstractDocument.readLock(java.desktop@17.0.9/AbstractDocument.java:1421) - locked <0x000000063c6af680> (a org.netbeans.modules.csl.core.GsfDocument) at org.netbeans.editor.BaseDocument.render(BaseDocument.java:1405) at org.openide.text.PositionRef$Manager$DocumentRenderer.render(PositionRef.java:927) at org.openide.text.PositionRef$Manager$DocumentRenderer.renderToObject(PositionRef.java:948) at org.openide.text.PositionRef$Manager.addPosition(PositionRef.java:389) at org.openide.text.PositionRef.init(PositionRef.java:105) at org.openide.text.PositionRef.<init>(PositionRef.java:90) at org.openide.text.PositionRef.<init>(PositionRef.java:68) at org.openide.text.CloneableEditorSupport.createPositionRef(CloneableEditorSupport.java:1264) at org.netbeans.modules.editor.hints.HintsControllerImpl.linePart(HintsControllerImpl.java:231) at org.netbeans.spi.editor.hints.ErrorDescriptionFactory.createErrorDescription(ErrorDescriptionFactory.java:245) at org.netbeans.spi.editor.hints.ErrorDescriptionFactory.createErrorDescription(ErrorDescriptionFactory.java:218) at org.netbeans.modules.lsp.client.bindings.LanguageClientImpl.lambda$publishDiagnostics$0(LanguageClientImpl.java:176) at org.netbeans.modules.lsp.client.bindings.LanguageClientImpl$$Lambda$705/0x00007f53e4ac73a8.apply(Unknown Source) at java.util.stream.ReferencePipeline$3$1.accept(java.base@17.0.9/ReferencePipeline.java:197) at java.util.ArrayList$ArrayListSpliterator.forEachRemaining(java.base@17.0.9/ArrayList.java:1625) at java.util.stream.AbstractPipeline.copyInto(java.base@17.0.9/AbstractPipeline.java:509) at java.util.stream.AbstractPipeline.wrapAndCopyInto(java.base@17.0.9/AbstractPipeline.java:499) at java.util.stream.ReduceOps$ReduceOp.evaluateSequential(java.base@17.0.9/ReduceOps.java:921) at java.util.stream.AbstractPipeline.evaluate(java.base@17.0.9/AbstractPipeline.java:234) at java.util.stream.ReferencePipeline.collect(java.base@17.0.9/ReferencePipeline.java:682) at org.netbeans.modules.lsp.client.bindings.LanguageClientImpl.publishDiagnostics(LanguageClientImpl.java:177) at jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(java.base@17.0.9/Native Method) at jdk.internal.reflect.NativeMethodAccessorImpl.invoke(java.base@17.0.9/NativeMethodAccessorImpl.java:77) at jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(java.base@17.0.9/DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(java.base@17.0.9/Method.java:568) at org.eclipse.lsp4j.jsonrpc.services.GenericEndpoint.lambda$recursiveFindRpcMethods$0(GenericEndpoint.java:65) at org.eclipse.lsp4j.jsonrpc.services.GenericEndpoint$$Lambda$648/0x00007f53e4a4dd50.apply(Unknown Source) at org.eclipse.lsp4j.jsonrpc.services.GenericEndpoint.notify(GenericEndpoint.java:160) at org.eclipse.lsp4j.jsonrpc.RemoteEndpoint.handleNotification(RemoteEndpoint.java:231) [...]
- add support for rust LSP implementation rust-analyzer - align rust icons more with existing NB icons - show hint if rust analyser is not configured
|
Hi @matthiasblaesing , Thanks for this effort! I wasn't able to get LSP server running properly with: The NetBeans UI froze, with the EDT trying to update nodes [1]. I haven't found time to look at it in detail, but from [2] I wonder if we should be using I think it's premature to add this to the IDE right now. Maybe we want to iterate a little bit more. Kind regards, [1] [2] |
|
The thread dump.txt I was able to extract when the IDE froze. |
| .start(); | ||
| return LanguageServerDescription.create(p.getInputStream(), p.getOutputStream(), p); | ||
| } catch (IOException ex) { | ||
| LOG.log(Level.FINE, null, ex); |
There was a problem hiding this comment.
I'd use Level.INFO for this, so one can see what may be failing when launching rust-analyzer as a background process. It's also worth investigating when the LSP server is restarted.
rust/rust.grammar/src/org/netbeans/modules/rust/grammar/lsp/RustLSP.java
Show resolved
Hide resolved
I'll look into the points you brought up. BUT I have to ask: for what do we want to wait? Rust Support in NetBeans has not moved an inch and while this is rough, not doing it will not yield improvements. |
|
Where does that version of rust-analyzer come from? I looked through https://github.com/rust-lang/rust-analyzer/releases and https://github.com/rust-lang/rust-analyzer/tags there is neither a tag, nor a release entry in september of last year. I fail to find the reference |
|
@vieiro could you please describe how you reached the deadlock and whether that is reproducible? |
Very true! If this is on master we'll have more chances to get contributions! |
I had a project opened (and a Rust file opened in the editor). Then the IDE got restarted (after changing the rust-analyzer path) and then it happened. I think there's a race condition between the LSP server being started and the opened files being sent to the LSP server, or something like that. IIRC there were two rust-analyzer servers running at the same time. I'll try to get a reproducer, though. |
| updateFocused(); | ||
| } | ||
| }); | ||
| updateFocused(); |
There was a problem hiding this comment.
just a guess regarding the reported freeze:
- wrap this line into SwingUtilities.invokeLater(line) or replace
@OnStartwith@OnShowing - remove synchronized from
updateFocused()(since no longer needed)
There was a problem hiding this comment.
It could be.
Maybe we should redesign the LSPClient in NetBeans.
This should be a state machine, that queues requests in order and does not execute them until the LSP server has been properly initialized.
I have a small draft that initializes servers somewhere (but all the rest is still missing). May want to recover it from backup...
General improvements (lsp.client)
To be able to sanely use rust-analyzer (the rust LSP implementation) the NetBeans lsp client implementation needs some adjustments:
refreshDiagnostics, failing to correctly work if that is not the casenullas a result of LSP operations. The existingnullguards were thus extendedpublishDiagnosticscall as part of the formatting.Rust specific
The cargo TOML support is updated to the current version of TOML to be able to parse project files using extended syntax.
While NetBeans can provide parts of the IDE services based on custom implementations, the rust-analyzer LSP server provides extended support of the box. Compared with the built-in solution rust-analyzer has the drawback, that it is huge and platform dependent, so it can't be bundled with NetBeans.
To provide both minimal rust support in all cases and improved editing support by using rust-analyzer the solution provided here will switch from internal support to rust-analyzer only if rust-analyzer is installed and configured.
When a user now opens a rust file and rust-analyzer is not configured a warning is attached to the file (as is done for cpplite):
But structure is still present:
The icons used in the rust module were aligned with the other NetBeans icons from lsp.client and PHP support.
The warning allows to directly jump into options to configure rust-analyzer:
With that done extended structure support is activated:
Code completion is provided:
and Find usages works: