Context

 
To run UI-tests, projects depend on core-Tf, which comes with artifacts from selenium, selenide, etc.. So - managing the version of the artifacts at one single place is done in core-TF.
 
But, additionally, we need docker-images from Selenium to run the tests. We typically have a docker-compose and and a .env file for each project. These two files are typically project-specific and cannot be shared among various projects.
The .env file contains this assignment:
SELENIUM_VERSION=4.13.0
The value for SELENIUM_VERSION is the same in all .env files of all projects.
(at least it should be)

Approach

 
Inspiration from here:
https://stackoverflow.com/questions/62121588/how-to-evaluate-a-dynamic-variable-in-a-docker-compose-yml-file
 
On the individual projects have this:
a file named env_raw or similar. It is identical to the final ray, but keeps a placeholder for the value of SELENIUM_VERSION
 
at aws-build-scripts repository we have this:
a downloadable script which contains the value for SELENIUM_VERSION and a function to transform env_raw into .env

In the Jenkins pipeline- we need to download the script and create "dynamically" the .env before calling docker-compose.

What about running tests locally?

The above mentioned approach should work fine on Jenkins.
BUT locally we would lack the .env file.

To workaround this - each project could still contain a .env file with a literal value for the SELENIUM_VERSION. On jenkins we would just override the file.