Repository navigation
Description
Describe the bug
My organization's managed settings include forceRemoteSettingsRefresh. Since then, opening a new session or switching sessions inside a running CLI often shows:
Initializing chat using settings required by your organization admin…
Plugins required by your organization admin could not be checked. Continuing with the current setup. Error: Managed settings could not be confirmed
The logs show that the cache-only startup path correctly refuses the cached policy (source=none), but no live fetch is started at that point. Managed plugin preparation waits about 12.5 s and gives up, and permission seeding times out after 5 s. The live fetch only runs at the next policy-refresh trigger, here the next session switch, and then succeeds immediately (source=server, serverFetchFailed=false).
In one CLI process today: 27 session registrations, 10 of these warnings and 13 Permission seeding did not complete within 5000ms warnings. The CLI does not hang; it continues without the policy for that init. The network is not the cause: api.github.com answers in about 0.3 s, and the live fetch, when it runs, completes immediately.
Possibly related to #5341 (same forceRemoteSettingsRefresh setup, but there the CLI hangs; here it only degrades with a warning).
Affected version
GitHub Copilot CLI 1.0.111 (first seen on 1.0.109, when the org started returning forceRemoteSettingsRefresh)
Steps to reproduce the behavior
- Sign in with an account whose organization's managed settings include
forceRemoteSettingsRefresh. - Start
copilotand use a session normally. - Open a new session (or switch to another one and back) a few times.
- Some of the time, see the "Managed settings could not be confirmed" message. In
~/.copilot/logs/process-*.log, see the sequence below.
Sanitized log excerpt (one occurrence):
18:50:31.510Z [INFO] Registering foreground session: <new-session>
18:50:31.615Z [INFO] [managedSettings] device MDM: no policy present on this device
18:50:31.615Z [INFO] [managedSettings] forceRemoteSettingsRefresh requires a live fetch for https://github.com/<org>/; cache-only startup will continue without server policy
18:50:31.615Z [INFO] [managedSettings] effective policy resolved: source=none, bypassDisabled=false, serverFetchFailed=false, ...
18:50:31.816Z [INFO] Signed in to https://github.com/<org>/ as <user>.
18:50:41.586Z [WARNING] Permission seeding did not complete within 5000ms for session <new-session>; releasing the extension load gate without authorizing interrupted work.
18:50:44.076Z [WARNING] Managed plugin preparation did not complete: Plugins required by your organization admin could not be checked. Continuing with the current setup. Error: Managed settings could not be confirmed
18:50:44.275Z [INFO] Registering foreground session: <previous-session>
18:50:44.531Z [INFO] [managedSettings] forceRemoteSettingsRefresh set — forcing a live fetch for https://github.com/<org>/ (a cached policy exists but will not be served or used as a fallback)
18:50:44.531Z [INFO] [managedSettings] server policy received from https://github.com/<org>/: bypassDisabled=false, keys=[features,extraKnownMarketplaces,model,permissions,forceRemoteSettingsRefresh]
18:50:44.531Z [INFO] [managedSettings] effective policy resolved: source=server, bypassDisabled=false, serverFetchFailed=false, ...
Between 18:50:31.6 and 18:50:44.5 no live fetch is logged for the new session.
Expected behavior
When forceRemoteSettingsRefresh makes the cache-only startup unusable, session init should start (or await) a live policy fetch. Managed plugin preparation and permission seeding would then receive the server policy instead of timing out and continuing without it.
Additional context
- OS: Windows 11 Enterprise 10.0.26200 (build 26200)
- CPU architecture: x86_64 (AMD64)
- Terminal: Windows Terminal
- Shell: PowerShell 7.7.0
- Account: Enterprise Managed User; org policy fetched from github.com; no device MDM policy
- Full logs available on request
Activity