Installing CumulusCI v 4 installs the latest version of urllib3 which seems to be incompatable with running ROBOT tests

[liamcollerton-giveclarity]

Describe the bug

I am running a cci flow from a Github action and I started seeing the following error for every Robot test when running the Robot task.

Parent suite setup failed:

ValueError: Timeout value connect was <object object at 0x7f9212d244f0>, but it must be an int, float or None.

After some investigation I tracked it down to the urllib3 library. The latest version is 2.2.3 which is what is being installed with cumulusCI

Collecting urllib3<3,>=1.21.1 (from requests->cumulusci)
Downloading urllib3-2.2.3-py3-none-any.whl.metadata (6.5 kB)

If I add the step

  - name: Set urlib3 version
    run: |
      pip install urllib3==1.26.20
    shell: bash

after installing cumulus ci and before running the flow then it all works fine

Note that I am not using the standard cumulus Github actions. I had created my own which install the sf cli and cumulusci and then runs a version of the ci_feature flow.

I expect that this would be an issue in any installation that doesn't already have a version of urlib between 1.21.1 and 1.26.20

Reproduction steps

n. ...

Your CumulusCI and Python versions

CumulusCI version: 4.0.1
Python version: 3.13.0

Operating System

Ubuntu latest on a Github Action

Windows environment

No response

CumulusCI installation method

pip

Error Gist

No response

Additional information

No response

[jstvz]

Thanks for the report. Looks like we pinned this in Snowfakery until recently. The issue is that we're pinning robotframework-seleniumlibrary<6. We'll see whether we can remove that pin, and we'll restore the urllib3 pin if we can't.

[Szandor72]

We're also currently blocked due to this issue which hit us by surprise.

[liamcollerton-giveclarity]

@Szandor72 The workaround is to make sure that you are using the correct version of that library, after you have installed cumulusci.

How are you experiencing this issue and are you still blocked?

[Szandor72]

We downgraded to cumulusci 3.87 for now because we don't want to be responsible for managing CCI's internal package dependencies. We're consumers not Pythonians.

…

On Mon, Nov 25, 2024, 3:11 PM liamcollerton-giveclarity < ***@***.***> wrote: @Szandor72 <https://github.com/Szandor72> The workaround is to make sure that you are using the correct version of that library, after you have installed cumulusci. How are you experiencing this issue and are you still blocked? — Reply to this email directly, view it on GitHub <#3849 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEAGS6TUDU3I4UDS4FIJOUL2CMVX3AVCNFSM6AAAAABSJ3NOTCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIOJYGEZDMMBVHA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[jstvz]

@Szandor72 Sorry for the trouble caused by the recent changes. We made these updates to help with compliance on our end, but we didn't expect them to affect users as much as they have. It wasn't our intention to put you in a position where you have to manage our internal package dependencies—making things easier for ourselves while making it harder on our users is the opposite of what we want.

[Szandor72]

@jstvz I am deeply grateful for the work you are doing here. In years of usage, there were only very few issues.

Back to topic: Does the WI include regression tests for such a case as well? The biggest surprise was that it was not detected before release.

[jstvz]

@Szandor72 Thank you for the kind words. We’ve been working on improving our test coverage and recently fixed some issues in our integration tests, which helped uncover these errors. That said, we should have caught it before release.