FYI, I was not sure which if any components to tag this with.

Brett, looks like maybe this was added by https://codereview.chromium.org/2479593006 - we don't actually use anything from //extensions, we're just checking if enable_extensions is set in our BUILD.gn file. Is even this dependency (on whether extensions are enabled or not) disallowed?

Maksim, maybe we should just nuke our dependency on ENABLE_EXTENSIONS? It's limited to schema_registry.cc, and I think we could safely call SetExtensionsDomainsReady() regardless of whether the platform enables extensions or not - it's just an unnecessary observer invocation, right?

Re schema_registry.cc - I don't think that the change to always call SetExtensionsDomainsReady is safe. It controls the initialization order of different parts (obtaining extensions' schemas, loading policies, exposing them). Normally SetExtensionsDomainsReady() is called from the extensions subsystem (from extensions::ManagedValueStoreCache::ExtensionTracker::Register), and that's why  schema_registry.cc has to fake this when under !ENABLE_EXTENSIONS.
But we can probably move this stub call of SetExtensionsDomainsReady() out from //components/policy - e.g. into //chrome/browser/policy/schema_registry_service_factory.cc where it's instantiated.

Makes sense - maybe you can eliminate this dependency at some point - definitely seems cleaner.