Remove N+1 SQL query loading project feature in dashboard

Projects that have a pipeline may need to check whether the user has
permission to read the build (`can?(current_user, :read_build,
project)`), which requires checking the `project_features` table.
This would cause an N+1 SQL query for each project.

This change also has a beneficial side effect that may avoid a race
condition. When a user deletes a project, the project is queued for
deletion and the user is redirected back to the dashboard page. However,
the following may happen:

1. The dashboard page may load this deleted project in the list of
20 projects.
2. The view will load the project pipeline status from the cache and
attempt to show each project.
3. When the view encounters the deleted project, it calls
`can?(current_user, :read_build, project)` to determine whether to
display the pipeline status.
4. Sidekiq deletes the project from the database.
5. However, since the deleted project is still loaded in memory, it will
attempt to call `project.project_feature.access_level`.
6. Since `project_feature` was not eager loaded, a lazy `SELECT` call is
made to the database.
7. This `SELECT` call returns nothing, and the user sees a 500 error.

By eager loading `project_feature`, we can ensure that we have a
consistent view and avoid records from being deleted later.

Closes https://gitlab.com/gitlab-org/gitlab-ce/issues/66482

1 files changed, 22 insertions, 0 deletions

diff --git a/spec/features/dashboard/projects_spec.rb b/spec/features/dashboard/projects_spec.rb
index e2100c8562b..f362172e712 100644
--- a/spec/features/dashboard/projects_spec.rb
+++ b/spec/features/dashboard/projects_spec.rb
@@ -220,4 +220,26 @@ describe 'Dashboard Projects' do
       expect(find('input#merge_request_target_branch', visible: false).value).to eq 'master'
     end
   end
+
+  it 'avoids an N+1 query in dashboard index' do
+    create(:ci_pipeline, :with_job, status: :success, project: project, ref: project.default_branch, sha: project.commit.sha)
+    visit dashboard_projects_path
+
+    control_count = ActiveRecord::QueryRecorder.new { visit dashboard_projects_path }.count
+
+    new_project = create(:project, :repository, name: 'new project')
+    create(:ci_pipeline, :with_job, status: :success, project: new_project, ref: new_project.commit.sha)
+    new_project.add_developer(user)
+
+    ActiveRecord::QueryRecorder.new { visit dashboard_projects_path }.count
+
+    # There are three known N+1 queries:
+    # 1. Project#open_issues_count
+    # 2. Project#open_merge_requests_count
+    # 3. Project#forks_count
+    #
+    # In addition, ProjectsHelper#load_pipeline_status also adds an
+    # additional query.
+    expect { visit dashboard_projects_path }.not_to exceed_query_limit(control_count + 4)
+  end
 end