summaryrefslogtreecommitdiff
path: root/spec/migrations/rename_reserved_project_names_spec.rb
diff options
context:
space:
mode:
Diffstat (limited to 'spec/migrations/rename_reserved_project_names_spec.rb')
-rw-r--r--spec/migrations/rename_reserved_project_names_spec.rb12
1 files changed, 8 insertions, 4 deletions
diff --git a/spec/migrations/rename_reserved_project_names_spec.rb b/spec/migrations/rename_reserved_project_names_spec.rb
index e6555b1fe6b..34336d705b1 100644
--- a/spec/migrations/rename_reserved_project_names_spec.rb
+++ b/spec/migrations/rename_reserved_project_names_spec.rb
@@ -3,10 +3,14 @@
require 'spec_helper'
require Rails.root.join('db', 'post_migrate', '20161221153951_rename_reserved_project_names.rb')
-# This migration uses multiple threads, and thus different transactions. This
-# means data created in this spec may not be visible to some threads. To work
-# around this we use the DELETE cleaning strategy.
-describe RenameReservedProjectNames, :delete do
+# This migration is using factories, which set fields that don't actually
+# exist in the DB schema previous to 20161221153951. Thus we just use the
+# latest schema when testing this migration.
+# This is ok-ish because:
+# 1. This migration is a data migration
+# 2. It only relies on very stable DB fields: routes.id, routes.path, namespaces.id, projects.namespace_id
+# Ideally, the test should not use factories and rely on the `table` helper instead.
+describe RenameReservedProjectNames, :migration, schema: :latest do
let(:migration) { described_class.new }
let!(:project) { create(:project) }