[pmontra]

It's usually enough but not always. Sometimes it happened that my customers using MacOS or WSL were not able to pass some tests or reproduce some bugs. That was due to some differences between the userland of the Linux servers, which are our build and deployment targets, and what they have on their Macs. I work on a Debian laptop (it used to be Ubuntu) and I can always run on it whatever sw runs on the servers. The languages are Python and Ruby, some bash.

The developers on WSL (the Python project, Django) tend to have a simplified environment. For example they don't run Celery (I never investigated why) and run all the background jobs synchronously or they don't run those jobs at all.

The ones on Macs (the Ruby project, Rails) have the full environment but I remember that they skipped some integration tests because they always failed on their Macs (Capybara and Chromedriver, I don't remember the details.) I was the one running the full test suite. By the way: all the CI services I used in the last 10 years are particularly bad at running those kind of tests. Maybe it's the amount of memory or the timing of the operations and those CI VMs (or containers?) don't play well with the assumptions of the test frameworks. Any language, any framework.