-
-
Notifications
You must be signed in to change notification settings - Fork 1.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
chore: reduce windows coverage #8632
Conversation
|
15 minutes, still! Why is Windows so terrible? |
I did a review of recent PRs where there was some Windows involvement, which has made me more comfortable with this change. We should make sure that we're running the tests in For reference, here are my notes: |
Well I guess it would have been 30 without this change, so it's probably still a good change. I closed #8630 since I didn't see a difference with it, but folks were reporting that |
Would it be worth separating e2e vs everything else into separate runs, so they can happen concurrently? Or would that just add more complexity for little gain? |
From that last run, the unit tests took 6s and the browser tests 10m, so I don't know there's much win in separating them Unit tests:
Integration tests:
There were two tests that flaked a total of three times with 45s timeouts, so we might be able to make things faster by addressing some of that flakiness:
|
Hmm. Actually I'm wondering if it ran the browser tests twice. Take a look at these log messages: 2023-01-20T17:03:28.0580491Z @sveltejs/kit:test:cross-browser: > pnpm test:unit && pnpm -r --workspace-concurrency 1 --filter="./test/**" test:cross-browser |
Once in dev mode (the 211 tests), once in prod (the 202 tests) |
Co-authored-by: Ben McCann <322311+benmccann@users.noreply.github.com>
We've observed that when Windows-specific test failures happen (other than those caused by vitejs/vite#9986), it's either specific unit tests or all end-to-end tests. I can't recall a time when a Windows bug has caused an end-to-end test to fail by itself.
A corollary is that we don't need to run the entire test suite in Windows, since doing so is slow as balls (#8630 notwithstanding). We can just run the unit tests and a subset of prerendering/e2e tests. That's the theory anyway.