Repository navigation
Exceptions activating vscode-java extension #2405
Description
Exceptions activating vscode-java extension. The spinning wheel just keeps going in the bottom-right corner.
Environment
- Operating System: OSX 12.3.1
- JDK version: embedded into vscode-java
- Visual Studio Code version: 1.66.1
- Java extension version: 1.4.0
Steps To Reproduce
- Open a Java file
- Java spinning wheel spinning in the bottom right corner. Exception in debug output:
rejected promise not handled within 1 second: Error: write EPIPE
/Applications/Visual Studio Code.app/Contents/Resources/app/out/vs/workbench/api/node/extensionHostProcess.js:104
stack trace: Error: write EPIPE
at WriteWrap.onWriteComplete [as oncomplete] (node:internal[/stream_base_commons:98:16]())
at WriteWrap.callbackTrampoline (node:internal[/async_hooks:130:17]())
/Applications/Visual Studio Code.app/Contents/Resources/app/out/vs/workbench/api/node/extensionHostProcess.js:104
rejected promise not handled within 1 second: Error [ERR_STREAM_DESTROYED]: Cannot call write after a stream was destroyed
/Applications/Visual Studio Code.app/Contents/Resources/app/out/vs/workbench/api/node/extensionHostProcess.js:104
stack trace: Error: Cannot call write after a stream was destroyed
at new NodeError (node:internal[/errors:371:5]())
at _write (node:internal[/streams/writable:321:11]())
at Socket.write (node:internal[/streams/writable:334:10]())
at [/Users/aboyko/.vscode/extensions/redhat.java-1.4.0-darwin-x64/dist/extension.js:2:1014845]()
at new Promise (<anonymous>)
at c.write ([/Users/aboyko/.vscode/extensions/redhat.java-1.4.0-darwin-x64/dist/extension.js:2:1014763]())
at _.doWrite ([/Users/aboyko/.vscode/extensions/redhat.java-1.4.0-darwin-x64/dist/extension.js:2:1004314]())
at [/Users/aboyko/.vscode/extensions/redhat.java-1.4.0-darwin-x64/dist/extension.js:2:1004209]()
/Applications/Visual Studio Code.app/Contents/Resources/app/out/vs/workbench/api/node/extensionHostProcess.js:104
Test Project: https://github.com/spring-projects/spring-petclinic
Metadata
Metadata
Assignees
Type
Projects
- Status
Activity
snjeza commentedon Apr 13, 2022
I can't reproduce the issue.
@BoykoAlex could you try to add
and start VS Code in the following way:
or
BoykoAlex commentedon Apr 13, 2022
Actually i think the project should have been: https://github.com/spring-projects/spring-petclinic (main or 1.5.x branch). I have updated the project in the description as well.
snjeza commentedon Apr 13, 2022
@BoykoAlex I still can't reproduce the issue.
Could you set
and send your .project, .classpath files and the .settings folder from the project's root (spring-petclinic)?
Eskibear commentedon Apr 14, 2022
I glanced at the location and that should be something inside langugeclient lib. Just a thought that if you have SERVER_PORT or JDTLS_CLIENT_PORT environment variable set? That would leads to wrong connection between client and server.
BTW, you can open extension log file via "Java: Open Extension Log File" command. Search "Starting Java server with:" for full command that Java LS was launched by, see if anything wrong there.
BoykoAlex commentedon May 26, 2022
Circled back to this issue.
Opened
spring-petclinicproject main branch (https://github.com/spring-projects/spring-petclinic).Spinning wheel in the bottom-right corner. Java extension doesn't seem to start properly.
Here is the Java LS log:
Eskibear commentedon May 27, 2022
@BoykoAlex The callstack looks similar to a known issue due to broken workspace state. microsoft/vscode-java-pack#954 (comment)
Workaround
Can you try clone spring-petclinic into another location and open it. Due to a brand new location, the workspace state should be totally clean then.
/cc @testforstephen @rgrunber Alex now has a concrete case about the broken cache issue... Do we want to collect any files related to the broken cache of workspace state, in order to find out the root cause?
testforstephen commentedon May 27, 2022
The workspace state is saved to a *.snap file, which is large and might not be collected by logs.
BoykoAlex commentedon May 27, 2022
Yes, having the project in a different folder fixed the issue. I'd be happy to attach
.snapfile but could not find it niether in$HOME/.vscodefolder nor in the spring-petclinic project root folder :-\rgrunber commentedon May 27, 2022
As @testforstephen mentioned in #2471 (comment), this might be related to the shutdown bug. I filed a bug in vscode-pde where it seemed to corrupt my workspace and prevent a successful startup.
If you're able to reproduce, it would be interesting if you could watch the contents of the server logs on shutdown before the problem happens consistently on startup. This is assuming this is caused by an abrupt shutdown.
So for spring-petclinic, the server logs would be at
/Users/aboyko/Library/Application Support/Code/User/workspaceStorage/29a927f04cb6b1515e547291e230a24c/redhat.java/jdt_ws/.metadata/.log. On MacOS, the commandline should be able to dotail -fon the above file.BoykoAlex commentedon May 30, 2022
jdt-log.zip
rgrunber commentedon May 30, 2022
I think the above logs just match your post from a few days ago.
!MESSAGE The workspace exited with unsaved changes in the previous session; refreshing workspace to recover changes.I think we need to try petclinic or some project and find a way to reproduce the corruption. My best guess is performing some kind of build/rebuild operation and issuing a SIGKILL to the JDT-LS process. If we can reproduce the stacktrace, then that's likely the issue.I don't know how reasonable it is to recover from this, so maybe the best we can do is detect it.
testforstephen commentedon May 31, 2022
I also reproduce this issue when running spring-petclinic on macOS.
My case is when I update macOS and restart it, reopening spring-petclinic failed due to the corrupt workspace state. From the server log, the last session failed because of ObjectNotFoundException. And the previous session didn't contain any exit and shutdown trace.
testforstephen commentedon May 31, 2022
In another VS Code window, when I click
xto close VS Code, the ParentProcessWatcher mechanism actually didn't work either. This is another issue we need to fix. I tracked it in a new issue #2488shihabuddin commentedon Mar 28, 2023
I just hit by this issue on Windows. Instead of cloning the project to a new location, is there a way to clear the workspace state?
fbricon commentedon Mar 28, 2023
fabiobsantosprogrow commentedon Jan 23, 2024
For me this is the real fix. Visual studio extension starts working after some hours struggling with this problem!
rgrunber commentedon Jan 23, 2024
If this is still a problem, I think we could come up with some kind of detection for it. The error message :
The workspace exited with unsaved changes in the previous session; refreshing workspace to recover changes.is emitted by :setCrashedfromrestoreSnapshots. See also :https://github.com/eclipse-platform/eclipse.platform/blob/3ec9f496b1a31bec24f54f701243078e7d5c0323/resources/bundles/org.eclipse.core.resources/src/org/eclipse/core/internal/resources/LocalMetaArea.java#L199-L207
https://github.com/eclipse-platform/eclipse.platform/blob/3ec9f496b1a31bec24f54f701243078e7d5c0323/resources/bundles/org.eclipse.core.resources/src/org/eclipse/core/internal/resources/LocalMetaArea.java#L58
I think we could pretty easily check whether
${storagePath}/jdt_ws/.metadata/.plugins/org.eclipse.core.resources/contains some file of the form[0-9]+.snap. If so, we can prompt the user to clean the workspace, or potentially do it on their behalf.