gitlab-org--gitlab-foss/spec/services/projects/destroy_service_spec.rb
Lin Jen-Shin 5747b0d3ed Make deleting with optimistic locking respect NULL
For now deleting with optimistic locking is broken when
lock_version is still NULL, because Rails would try to
delete with `lock_version = 0` while in the database
the column is still `NULL`.

The monkey patches would force Rails just pass whatever
in the column, and stop Rails from casting `NULL` into `0`
when the value is read from database.

Closes #24766
2016-12-01 16:17:20 +08:00

94 lines
2.8 KiB
Ruby

require 'spec_helper'
describe Projects::DestroyService, services: true do
let!(:user) { create(:user) }
let!(:project) { create(:project, namespace: user.namespace) }
let!(:path) { project.repository.path_to_repo }
let!(:remove_path) { path.sub(/\.git\Z/, "+#{project.id}+deleted.git") }
let!(:async) { false } # execute or async_execute
shared_examples 'deleting the project' do
it 'deletes the project' do
expect(Project.all).not_to include(project)
expect(Dir.exist?(path)).to be_falsey
expect(Dir.exist?(remove_path)).to be_falsey
end
end
context 'Sidekiq inline' do
before do
# Run sidekiq immediatly to check that renamed repository will be removed
Sidekiq::Testing.inline! { destroy_project(project, user, {}) }
end
it_behaves_like 'deleting the project'
end
context 'Sidekiq fake' do
before do
# Dont run sidekiq to check if renamed repository exists
Sidekiq::Testing.fake! { destroy_project(project, user, {}) }
end
it { expect(Project.all).not_to include(project) }
it { expect(Dir.exist?(path)).to be_falsey }
it { expect(Dir.exist?(remove_path)).to be_truthy }
end
context 'async delete of project with private issue visibility' do
let!(:async) { true }
before do
project.project_feature.update_attribute("issues_access_level", ProjectFeature::PRIVATE)
# Run sidekiq immediately to check that renamed repository will be removed
Sidekiq::Testing.inline! { destroy_project(project, user, {}) }
end
it_behaves_like 'deleting the project'
end
context 'delete with pipeline' do # which has optimistic locking
let!(:pipeline) { create(:ci_pipeline, project: project) }
before do
expect(project).to receive(:destroy!).and_call_original
perform_enqueued_jobs do
destroy_project(project, user, {})
end
end
it_behaves_like 'deleting the project'
end
context 'container registry' do
before do
stub_container_registry_config(enabled: true)
stub_container_registry_tags('tag')
end
context 'tags deletion succeeds' do
it do
expect_any_instance_of(ContainerRegistry::Tag).to receive(:delete).and_return(true)
destroy_project(project, user, {})
end
end
context 'tags deletion fails' do
before { expect_any_instance_of(ContainerRegistry::Tag).to receive(:delete).and_return(false) }
subject { destroy_project(project, user, {}) }
it { expect{subject}.to raise_error(Projects::DestroyService::DestroyError) }
end
end
def destroy_project(project, user, params)
if async
Projects::DestroyService.new(project, user, params).async_execute
else
Projects::DestroyService.new(project, user, params).execute
end
end
end