Ask to clean cache if a corrupted workspace cache is detected #1011
Conversation
testforstephen
requested review from
CsCherrYY,
Eskibear and
jdneo
as code owners
Screen.Recording.2022-06-20.at.14.36.02.movcc:// @rgrunber |
src/daemon/index.ts
Outdated
| sendInfo("", { | ||
| name: "corrupted-cache", | ||
| }); | ||
| const ans = await vscode.window.showErrorMessage("A corrupted Java extension cache has been detected, please try to \"Clean Workspace\".", |
There was a problem hiding this comment.
Should the quoted Clean Workspace be considered as a noun?
There was a problem hiding this comment.
I just reword the notification message as follows:
|
Nice. Not sure if that exception is always present for a corrupted workspace but if it's what you're generally seeing, this should help. A log watching facility on the client/server logs with appropriate API for consumers to subscribe (eg. for telemetry) probably belongs in vscode-java though. Down the road, it would be good to look into moving it, if possible. |
This exception "org.eclipse.core.internal.dtree.ObjectNotFoundException" is a known exception when the workspace cache is not saved correctly. Also we only check this when the language server fails to start, so it's safe for me to pop up a message to clean the workspace. Here, the daemon simply treats the Java Language Server as a black box and uses a whitelist to check logs to monitor its state. I believe the language server itself could have a more systematic way to reflect its state. At the moment, when the language server fails to startup, the language status icon is always spinning and didn't show the correct state. |
4 participants
Fix #954