ci: run tests with --no-fail-fast

[ada4a]

When adding a lint that has a lot of hits in the repo itself, each run of CI only shows a couple of failures, as cargo test stops after the first error. This makes it so that all the failures are shown at once

changelog: none

[rustbot]

r? @flip1995

rustbot has assigned @flip1995.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[flip1995]

failing fast is kinda what you want in CI, so that the CI resources are freed as quickly as possible. What's the issue of seeing only the first error on CI? I can't imagine that someone works on Clippy by pushing to a PR in order to run tests.

I would definitely not do this for the merge queue, as this will slow down merging of PRs.

[ada4a]

I can't imagine that someone works on Clippy by pushing to a PR in order to run tests.

I must admit that I do... it takes ages on my machine to run both dogfood and argument-less cargo uitest (both of which are needed when implementing a new lint). Maybe it's time to look into dev-desktops as I was advised.

I would definitely not do this for the merge queue, as this will slow down merging of PRs.

Fair enough – I only did that because of the comment asking to keep the contents of the two workflows identical

[flip1995]

Maybe it's time to look into dev-desktops as I was advised.

I just wanted to mention that. CI is also really slow compared to a mid/high-end laptop. As a member of the clippy-contributors team, you shouldn't have problems getting access to one.

I don't really want to promote this workflow 😅

[ada4a]

Fine fine, I'll go request it:)