get it to work with chrome

[catmando]

latest chrome driver fixes previous issue that was preventing chrome driver from working with Opal.

Here is some info on Stack Overflow regarding getting chrome driver settings sent over from rspec:

https://stackoverflow.com/questions/36407099/sending-capabilities-switches-to-chromedriver-with-capybara/46261462#46261462

[sfcgeorge]

I don't know what switches you need to set on Chrome, but Chrome in headless mode ignores some of them. Command line flags work, but prefs / profile don't.

This is only an issue when used with the --headless flag, so we can still use Chrome instead of Firefox but without the advantages of being headless.

https://bugs.chromium.org/p/chromedriver/issues/detail?id=1925
SeleniumHQ/selenium#4437

[catmando]

We should make sure it works head and headless

[janbiedermann]

done in sachsenring, yet these headless drivers all suck, sometimes tests run, sometimes not, can depend on subtle version differences.

[catmando]

then it ain't working :-)

[janbiedermann]

The fix would then be documentation:

1. Use the latest driver available
2. if the last installed version doesnt work, try the next earlier one
3. repeat two until you got it working or no more version is available

Thing is, because the drivers are external programs following their manufacturers release policies, we cant lock a version in the Gemfile. For chrome headless users usually get the latest, which is not always the greatest, same for geckodriver.

Very important here to add for users:
Do not reference circular objects!

Meaning: Do not test for "MyModel", but instead call a method that returns a value, any will do then, for example: "MyModel.respond_to?('find')"

[sfcgeorge]

Chrome Headless works for me, issue can be closed?

[janbiedermann]

almost, still issues with the hyper-gems tests

[janbiedermann]

currently no more issues with tests with the gems, though using any has_xpath or has_css is completely unreliable. workaround expect(page.body).to match(/<div>/) or .to include('<div>')
Leaving this open for documentation.