reenable more of the CI matrix

[Conduitry]

Now that this is free, we can afford it. Rather than simply uncommenting the old lines, I've also enabled tests on Node 15. When 16 is released and 15 goes EOL, we can update this again.

[Conduitry]

I don't know why the heck Node 12 on Ubuntu is failing. That's one of the ones that was already running.

[GrygrFlzr]

Possibly #404 at play. Let me force re-run all jobs.

[Conduitry]

That's possible. I had tried running them twice already, and Node 12 Ubuntu failed both times, although I didn't look carefully to see whether it was the same failure message both times.

[GrygrFlzr]

Okay, it seems to be at least consistently erroring on:

dev:basics
2:54:59 AM [vite] Error when evaluating SSR module /home/runner/work/kit/kit/packages/kit/test/apps/basics/src/routes/errors/module-scope-server.svelte:
Error: Crashing now
    at /home/runner/work/kit/kit/packages/kit/test/apps/basics/src/routes/errors/module-scope-server.svelte:5:12
    at instantiateModule (/home/runner/work/kit/kit/node_modules/.pnpm/[email protected]/node_modules/vite/dist/node/chunks/dep-e0f09032.js:68041:166)
2:54:59 AM [vite] Error when evaluating SSR module /home/runner/work/kit/kit/packages/kit/test/apps/basics/src/routes/errors/module-scope-server.svelte:
Error: Crashing now
    at /home/runner/work/kit/kit/packages/kit/test/apps/basics/src/routes/errors/module-scope-server.svelte:5:12
    at instantiateModule (/home/runner/work/kit/kit/node_modules/.pnpm/[email protected]/node_modules/vite/dist/node/chunks/dep-e0f09032.js:68041:166) (x2)

I can also replicate that specific error locally, so I think there's an actual issue involved that was previously skipped in CI somehow.

[GrygrFlzr]

I debugged it a bit and it seems that the error page in that test case is correctly rendering, but vite is also re-throwing the SSR error and triggering uvu. Not sure why this wasn't an issue before, though.

[GrygrFlzr]

Nevermind - while the vite SSR error does appear, it doesn't actually fail the uvu test - the set of failing tests keep changing between runs, usually something related to clicking a link, regexp routes, or redirect loops.

I think it's just a case of more test runs = more chances of failure due to the tests themselves being flaky and timing-sensitive.

[GrygrFlzr]

Okay, I merged #501 into master which should have solved most of the test stability issues.

@Conduitry go ahead and pull master to re-test the full matrix.

[Conduitry]

Success! Just the Windows ones failing now. Thanks @GrygrFlzr !

[GrygrFlzr]

Huh, the failing Windows tests cancelled the still-running macOS test for Node 15. Should macOS be shifted earlier in the OS matrix so that it has a chance to complete the run before the Windows CIs fail?

[Conduitry]

That's probably a good idea. Updated.