Don't log invalid token in sentry

Sentry should be for exceptional errors. Having this reported to sentry was creating a lot of noise (~20k events a day) and using up a large proportion of our Sentry limits. If in the future we want to track this, we should instead log it or track unauthorized responses as a proxy.

Before doing this I did some testing to see if the tokens were long lasting and are automatically refresh - as far as I can see they are and wasn't able to reproduce any issues by leaving tabs open or by offloading tabs. This would point to most tokens being invalid due to expiry. This could be due to the refresh token not working in inactive tabs.

Author: zetter-rpf

lib/hydra_public_api_client.rb

@@ -18,8 +18,7 @@ def fetch_oauth_user(token:)
 
     response = get('userinfo', {}, { Authorization: "Bearer #{token}" })
     response.body.to_h
-  rescue Faraday::UnauthorizedError => e
-    Sentry.capture_exception(e)
+  rescue Faraday::UnauthorizedError
     nil
   end
 

spec/models/user_spec.rb

@@ -120,20 +120,9 @@
       end
     end
 
-    context 'when the access token is invalid' do
-      before do
-        allow(Sentry).to receive(:capture_exception)
-        stub_request(:get, "#{HydraPublicApiClient::API_URL}/userinfo").to_return(status: 401)
-      end
-
-      it 'returns nil' do
-        expect(user).to be_nil
-      end
-
-      it 'reports the Faraday::UnauthorizedError exception to Sentry' do
-        user
-        expect(Sentry).to have_received(:capture_exception).with(instance_of(Faraday::UnauthorizedError))
-      end
+    it 'returns nil when the access token is invalid' do
+      stub_request(:get, "#{HydraPublicApiClient::API_URL}/userinfo").to_return(status: 401)
+      expect(user).to be_nil
     end
   end