Extra tests inside shared_examples:

Feedback from:
https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/4093#note_12321421

The advantage of this is that tests are less about logic and more
straightforward, thus could be easier to reason about each individual
tests.

The disadvantage of this is that we write more duplicated codes and
once something changed we might need to change all places and it's
harder to reason all tests as a whole. Because now we need to look at
more places to figure out how it should work under another option!

spec/models/ci/runner_spec.rb

@@ -154,7 +154,7 @@ describe Ci::Runner, models: true do
         runner.locked = true
       end
 
-      shared_examples 'locked build picker' do |serve_matching_tags|
+      shared_examples 'locked build picker' do
         context 'when runner cannot pick untagged jobs' do
           before do
             runner.run_untagged = false
@@ -170,16 +170,6 @@ describe Ci::Runner, models: true do
             runner.tag_list = ['bb', 'cc']
           end
 
-          it "#{serve_matching_tags} handle it for matching tags" do
-            build.tag_list = ['bb']
-            expected = if serve_matching_tags
-                         be_truthy
-                       else
-                         be_falsey
-                       end
-            expect(runner.can_pick?(build)).to expected
-          end
-
           it 'cannot handle it for builds without matching tags' do
             build.tag_list = ['aa']
             expect(runner.can_pick?(build)).to be_falsey
@@ -192,7 +182,18 @@ describe Ci::Runner, models: true do
           expect(runner.can_pick?(build)).to be_truthy
         end
 
-        it_behaves_like 'locked build picker', true
+        it_behaves_like 'locked build picker'
+
+        context 'when having runner tags' do
+          before do
+            runner.tag_list = ['bb', 'cc']
+            build.tag_list = ['bb']
+          end
+
+          it 'can handle it for matching tags' do
+            expect(runner.can_pick?(build)).to be_truthy
+          end
+        end
       end
 
       context 'serving a different project' do
@@ -204,7 +205,18 @@ describe Ci::Runner, models: true do
           expect(runner.can_pick?(build)).to be_falsey
         end
 
-        it_behaves_like 'locked build picker', false
+        it_behaves_like 'locked build picker'
+
+        context 'when having runner tags' do
+          before do
+            runner.tag_list = ['bb', 'cc']
+            build.tag_list = ['bb']
+          end
+
+          it 'cannot handle it for matching tags' do
+            expect(runner.can_pick?(build)).to be_falsey
+          end
+        end
       end
     end
   end