You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Every message runs the workspace folder scan 3 times in a row, even when t3.json ignores every repository #40659
What happens
On a project whose workspace root has many entries (ours: about 660 top-level entries and about 2,700 second-level entries), each message waits 7–9 s before the provider starts. The server trace shows three full WorkspaceFolders.scan runs, one after another, for every turn:
ProjectService.syncFolders also runs the same scan every 30 s.
The project's t3.json is {"ignoreRepositories": ["."]}. That ignores every discovered repository, so discoverFolders throws away the whole scan result and returns [] every time. The walk does nothing useful, but it still costs about 80% of the time before the turn starts.
Expected
When ignoreRepositories covers the workspace root ("."), skip the child walk. Folders listed in .code-workspace and tracked folders would keep working as they do now.
Scan at most once per turn: share one discover result between RuntimePolicyV2.resolve, CheckpointService.repositoryScopes and WorkspaceFolders.submodules instead of rescanning three times.
Workaround we used
We moved about 380 old directories under a dot-named folder, which the scan skips. That brought one scan from 4.2 s down to 2.3 s. It only helps for a while, because the scan still runs three times per message.
Activity