Last Updated: 2024-04-18

I was getting sporadic failures on CircleCI due to a downloaded file not being found. Here was my code for configuring a download path and accessing downloaded files:

Capybara.save_path = Rails.root

def downloaded_file(filename)
  Capybara.save_path.join('downloads', filename)
end

def clear_test_downloads
  FileUtils.rm_rf(Capybara.save_path.join('/downloads'))
end

The issue was that the downloaded_file method was searching in the ./downloads directory whereas the clear_test_downloads method was operating on /downloads - i.e. the absolute path - way outside my project. This was incorrect and potentially dangerous.

This mistake occurred because I wrongly assumed that the PathName object, of which Rails.root is one, would interpret the /downloads as being a sub-directory off of the current one. Never assume.

The fix:

# Share the download path to prevent them going out of sync
def download_path
  Capybara.save_path.join('downloads')
end

def clear_test_downloads
  FileUtils.rm_rf(download_path)
end

def downloaded_file(filename)
  download_path.join(filename)
end

Lesson:

• Always be careful with folder slashes and explicitly test that the resulting folder path is as expected.
• When there is symmetry (e.g. between saving a file and removing one, between writing a cache and clearing it), test they match - or better yet use the same variable to refer to both.