How do I stop my rspec capybara tests failing intermittently

Question

The intermittent failures are hard to pin down and they can be in many different tests.

Often you can see the element you are targeting in the browser but the program just doesn't seem to be in sync and can't find it. All attempts to wait don't work because of the disconnect with the actual page showing the element you want.

This has been the case in various ways for a couple of years and huge attempts by us to eliminate the failures with various capybara wait methods have helped but not eliminated the issue.

What else can we try?

Answer 1

This turned out to have a lot to do with the version of chromedriver I have been using. This was versions 2.7 which is 2-3 years old at this point. When you look at the version list at https://sites.google.com/a/chromium.org/chromedriver/downloads you can see a LOT of bug fixes since then.

At first I had to upgrade to at least 2.24 to avoid a "windows have 'data:.' in the URL and are blank pages. This seems to help address most of the intermittent failures. Further upgrading to 2.27 right now and is seeming even more reliable (no failures yet).

This general issue may reflect the move of selenium driver support to the vendors and given that this means Google in this case that might garner some good resources and explain some marked improvements and bug fixes.