Closed
Description
As I'm converting more of our test scripts to fail-if-no-tests to avoid the unfortunately-common pattern of us discovering that various tests weren't actually being run (e.g., because they were in the wrong location), .cirrus.yml
is getting more and more information embedded into it about what tests should and shouldn't run. This has several drawbacks:
- It clutters the CI config, which should describe at a high level what we are running and how, with details.
- It doesn't translate to other CI system (e.g., our Windows tests that are run on LUCI).
- It doesn't apply when running locally, so it's annoying to try to run a command on the whole repo locally.
We have the same issue with the "allow custom analysis" list, which has most of the same problems (and works against the goal of eliminating the use of shell scripts in the repo)