I’m configured to use node with browsers. There’s a org.openqa.selenium.WebDriverException that causes my build to fail where it wouldn’t before. To test, that it’s not my build, I pushed up my last good build and then pushed a README change that fails.
First line of failure:
firefox 47 Tests failed: {"value":{"stacktrace":"org.openqa.selenium.WebDriverException: address not available\n...
My config:
docker:
- image: circleci/node:8-browsers
working_directory: ~/repo
steps:
...
- run:
name: install JS dependencies
command: |
cd frontend/
npm install
npm run bower install
...
- run:
name: build and run JS tests
command: |
cd frontend/
npm test
That’s unlikely to be enough information to help someone help you. In general, browser test failure on CI should be debugged in the same way as the same happening locally.
Find out what “address not available” means. Does it mean you’re using a URL that is never available even when Selenium is up? Has the Selenium server gone down? Have you waited for it to start? Have you looked at your Selenium logs?
I would expect so. I would do it by creating a custom Docker image containing the browsers and drivers required, and then using that instead of a prebuilt one from Circle. I do this with PhantomJS, and it works very well. You can then bump up the browsers in use on your own schedule, rather than suffer build breakage because something in the CircleCI image was upgraded.