build: use Ninja in action workflows

[targos]

Until we have a fix for the makefiles generated by GYP.

Refs: #37429

[targos]

Haha I would have thought that Ninja was available by default in GitHub action runners

[targos]

actions/runner-images#741 (comment)

[gengjiawen]

actions/virtual-environments#741 (comment)

haha, same here. In the end, I build this in docker to avoid such issues (https://github.com/gengjiawen/node-build/blob/7bbcd29d624a79cccf3aa7bafd94a5747ae12c48/Dockerfile#L23).
Also suggest add it to gitpods gitpod-io/workspace-images#357.

[gengjiawen]

Until we have a fix for the makefiles generated by GYP.

If we have fix some makefiles. I still hope we use ninja by default since it's much fast.

[aduh95]

It seems it doesn't improve the situation – maybe you need to specify CONFIG_FLAGS=--ninja to the make test command as well?

[targos]

It seems it doesn't improve the situation – maybe you need to specify CONFIG_FLAGS=--ninja to the make test command as well?

Looks like you're right. I wonder what's special about github actions here. On my computer, it's enough to just ./configure --ninja and then make test will use Ninja

[aduh95]

Tests completed in less than 10m 🎉

[aduh95]

coverage-linux tests are still using GYP and are taking very long to complete.

[targos]

@aduh95 right, I missed coverage-linux, thanks!

[richardlau]

FWIW I think I've found the source of the regression with makefiles. PR: #37502

[aduh95]

fast-track?

[aduh95]

#37505 fixes the issue with GYP – although we probably should still land this and use Ninja on GH Actions considering the improvement for build time.

[richardlau]

#37505 fixes the issue with GYP – although we probably should still land this and use Ninja on GH Actions considering the improvement for build time.

The build times in #37505 on GitHub Actions are comparable to the times for this PR. I would not expect them to be wildly different as GitHub actions start with a clean workspace/checkout.