Different intercept behavior in Firefox 151?

[oliver]

Since last night some of our Cypress tests fail unexpectedly (both in CI and on developer systems). They even fail with older versions of our software; so this is probably not caused by changes in our software. And with Chrome the same tests suceed!

We usually run Cypress tests with Firefox. And there was a Firefox upgrade last night (from 150 to Firefox 151). I don't know yet how to install an older FF version in parallel though, so haven't tested with older FF versions.

I haven't debugged the problem in detail so far; but it looks like it's related to cy.intercept(). We have code like this:

  beforeEach(() => {
    cy.intercept({
      method: 'GET',
      url: /abc
    }).as('abcRequest');
...
    cy.wait('@abcRequest');
  });

This cy.wait() call works fine. But later in the test our software does another request to /abc, which triggers the intercept again; and it looks like Cypress then sends back the same response it intercepted in the initial cy.wait() call.

Does anyone else see similar problems since the Firefox 151 upgrade? Or do we use Cypress incorrectly here?

[MikeMcC399]

Which operating system are you using for your developer and CI systems?

Which is your CI provider?

You already found #33843 so perhaps Mozilla did something strange with this release?

https://developer.mozilla.org/en-US/docs/Mozilla/Firefox/Releases/151

[oliver]

We run Cypress under Windows 11 (on developer systems and via self-hosted Jenkins). We use Cypress 14.5.3 though; maybe Cypress 15 behaves differently.

[MikeMcC399]

I looked at the issue with cookies in more depth today and I confirmed that Firefox 151.0 sometimes returns cookies in a different order than previous versions. I believe that the problem is in the Cypress test which made an incorrect assumption that it could rely on the cookie order.

Firefox may have made similar internal changes that affect your tests. To get any sort of feedback or technical response, you'd need to provide a reproducible example, otherwise any comments are probably going to be guesswork.

[MikeMcC399]

I don't know yet how to install an older FF version in parallel though, so haven't tested with older FF versions.

You should be able to download specific Firefox versions / architectures / languages from:

https://download-installer.cdn.mozilla.net/pub/firefox/releases/

Additionally, if you are able to use Docker, you can use pre-built Cypress Docker images or you can build your own with cypress/factory

[oliver]

We have debugged our failing Cypress tests and found a bug in our code: in some of the Cypress tests we accidentally sent two cookies instead of just one. This didn't cause problems before; but I guess with Firefox 151 the two cookies are now sent in a different order, causing different behavior by the backend and leading to the test failure.

Thanks for describing changed cookie behavior in #33843; this was helpful when debugging the problem on our side.

[MikeMcC399]

That's great you found the issue based on cookies! 🎉 Glad to have been able to help in that case!

It looks like Firefox 151.0 may be doing some internal sorting when it stores cookies. I'm guessing that the order depends on the domain, as this was the difference when running on localhost and on the external test app website.

You might like to mark your last reply as Answer, if that is possible, so other readers can benefit.