Upgrade `cargo-deny` to `0.19.0` on self-hosted runners.

[mukilan]

The versions older than 0.18.6 have a bug which causes them to error out prematurely. Servo's mach test-tidy currently doesn't catch this issue so we are silently allow patches that would otherwise fail cargo-deny to land.

After servo/servo#41853 lands, ./mach test-tidy will start failing if the self-hosted runner's cargo-deny version is older than 0.18.6. So, we should ensure the self-hosted runners move the 0.19.0 to match the github hosted runners.

[mukilan]

Looks like we need to do this before servo/servo#41853 as the PR itself ends up running on self-hosted runners.

[mukilan]

@delan is there a way to force a single PR to use github-hosted runners? I have a PR that upgrades cargo-deny version used by Servo, but it is blocked because self-hosted runners use the older cargo-deny version that fails the build.

If not, I guess the other options are to globally force GH runners until PR lands, or split the PR into two part - upgrade cargo-deny and then patch test-tidy to fail on unknown errors, after a new self-hosted image is generated with latest cargo-deny

[jschwe]

@mukilan We could install cargo-deny as part of the workflow (even when running on self-hosted runners), until we get the image updated.

[mukilan]

Good call, I've updated the PR to use the action directly. We can update lint.yml to be conditional again once this issue is completed.

[delan]

done! (wow we really need a simpler way to force image rebuilds, especially after #75…)

$ ( for i in {0..4}; do ./do run ci$i -- sh -c 'rm -Rv /var/lib/libvirt/images/base; rm -v /config/monitor/data/profiles/*/snapshot; virsh list --all --name | rg \^ci-template- | tr \\n \\0 | xargs -0n1 virsh undefine --'; ./do restart-monitor ci$i; done )

i see that the cargo-deny version is hardcoded in mach bootstrap, but servo/servo#41853 bumped it, so rebuilding all of the images against the same profile definitions should be enough.

[delan]

@mukilan if you want to make lint.yml conditional again, feel free to close this issue when you do :)

[mukilan]

Thanks @delan! I'll close this issue as servo/servo#41866 has landed.