_{tachometer-reporter-action v2 for CI}

Size Change: -59 B (-0.08%)

Total Size: 78.4 kB

_{compressed-size-action}

coverage: 99.523% (-0.09%) from 99.61%
when pulling a2dcfc6 on switch-to-vitest
into bb804cd on main.

It seems slower across the board and It's an extra 100MB+ of node_modules & an 800MB in ~/.cache/ms-playright (tried using local Chrome install but I think that'd require different config/setup, it doesn't seem to function properly out-of-the-box)

I agree, merging this feels like a regression, all though I am convinced that we should eventually move to a new tool because of karma being deprecated and it blocking upgrading our dependencies. I am going to explore the landscape a bit more, maybe there's a middle ground here.

Ideally webdriver would be as fast as playwright but... I am unsure where this is headed in terms of performance.

We've had to switch to some Vitest-specific APIs. While the "lock-in" doesn't look to be to rough, we are switching from runner agnostic (or meant to be runner agnostic) tools to something that's specific & will need to be changed if we need to migrate again in the future

This is more because sinon in more recent versions is very broken, when we use sinon.spyOn it will break the prototype of the underlying object 😅

We've encountered a fair number of Vitest bugs in the process of this PR

This was part of my intention, I wanted to uncover bugs so we can help the vitest people with testing browser mode as I am really looking for a community project to step into the shoes of karma.

The --skipLibCheck flags can probably be removed now that it was added to js-lint.json, right?

Well, sadly, the setting doesn't work 😅

I agree with you, let's hold off on merging this, I will surface this with the vitest people so there's information sharing.

+1 on waiting on some vitest improvements before merging this. I don't mind the switch away from sinon as it hasn't been well maintained either.

This is more because sinon in more recent versions is very broken, when we use sinon.spyOn it will break the prototype of the underlying object 😅

Oh fair enough, wasn't quite sure if it was sinon or vitest that was the problem there.

Well, sadly, the setting doesn't work 😅

Huh. I thought the flag would act the exact same there. Admittedly I can't even run test:ts:core at the moment, some Node error... test:ts:compat is a-ok though. Fair enough on that too then.