We're migrating from an old BuildBot CI system to GitHub CI and we've run into some problems with tests. We run a bunch of unit tests for low level socket communication, ssdp discovery and http stress tests. We have tests failing on macOS, Linux and Windows when run on GitHub CI. Is there some special considerations to take when running those kinds of tests? Are certain ports blocked? Is certain types of network communication blocked? IPv4 vs IPv6? Connecting to `localhost`? Is there any other reason these kinds of connections and tests would fail on GitHub CI but work when run locally and on our own BuildBot instance?
Nothing is explicity blocked that should affect local networking that I am aware of. A runner runs on a standard vm, although as a user called "runner". It may be necessary to elevate to do certain operations, but you can `sudo ...` in scripts
Can you share more information about what exactly the tests are doing and how they are failing? Any error messages or logs?
Thank you for your reply. I have been able to make some progress by tweaking socket timeout values and the number of concurrent socket connections being created in the tests. I haven't managed to get all tests to pass though but your answer at least tells me that there is nothing in particular that it is blocked and that is more a matter of how the connections are set up and how the tests are designed.