Cookies bleed between scenarios if last step is on a different domain

[j4m3s]

Mink v1.5.0, tests run using Selenium

We have acceptance tests that follow the user journey from our site (e.g. localhost) through to paypal, and have found that if the scenario ends on a PayPal page, the cookies from localhost are not cleared and (for example) a user remains logged-in at the start of the next scenario.

Adding a step calling getSession()->restart() at the end resolves the issue.

I haven't verified whether or not the issue occurs with Goutte too (as PayPal depends on javascript) - if it would help let me know and I'll create a test project.

[stof]

When using Selenium, we can indeed only remove cookies for the current domain when resetting the session. reset does not guarantee a full isolation.
GoutteDriver is able to reset all cookies when resetting the session.

Btw, if you use Mink in Behat through MinkExtension, you can mark a scenario as needing a full restart of the session by tagging it as @insulated (though it currently applies to the beginning of the scenario, not the end)

[j4m3s]

I realise I'm pushing the boundaries of this ticket but is the issue limited to Selenium? (i.e. would it work if we used zombie?). GoutteDriver isn't an option in this case as the Paypal sandbox depends so heavily on Javascript.

I only exposed the problem by accident - would it be possible to output a warning of some form to tell users that this unexpected behaviour is taking place thanks to Selenium? It wasn't too serious in our case but I imagine it could cause a lot of head scratching for others wondering what was going on. I had convinced myself scenarios were insulated by default (though I'm not sure where I got that from :).

Thanks for the info about @insulated. Unfortunately it's really the end of the scenario that's the problem - we'd need to insulate every other test too.

[stof]

ZombieDriver is also able to reset the session fully. Only Selenium and Sahi are limited to the current domain (because they are limited by what browsers are allowing from JS).

We are currently in the process of rewriting the Mink documentation. Can you open an issue on the docs repo to ask to document this for session resetting ?

Regarding stopping the session fully at the end of the scenario, I'm opening an issue on MinkExtension to add this feature

[j4m3s]

Thank you :) I'll open the docs issue now.
James.