feat(vitest-plugin): add vite 8 support

[adriencaccia]

Add Vite 8 to the supported peer dependency range for @codspeed/vitest-plugin, and test against it.

Vite 8 replaces esbuild and Rollup with Rolldown and Oxc. The plugin only uses name, apply, enforce and config, none of which are affected by the breaking changes. Two build-level adjustments were needed:

• Vite 8 publishes its types through exports only, which the legacy moduleResolution: "node" cannot read, so the package now uses moduleResolution: "bundler".
• rollup-plugin-dts forces preserveSymlinks: true, which stops TypeScript from reaching rolldown (a transitive dep of Vite 8) in a pnpm store and leaves Plugin without its base members. It is disabled for the declaration build, with vite kept external.

The vitest dev dependency moves to ^4.1.11: vitest depends on vite directly, and 4.1 is the first release accepting Vite 8, so without it the dev environment resolves two different vite copies and the plugin never actually runs on Vite 8.

Verified on Node 24.19.0:

• pnpm turbo run lint typecheck test green (29/29) with --frozen-lockfile
• dist/index.d.ts byte-identical to main, build emits no warnings
• plugin benches pass on vitest 4.1.11 + vite 8.2.2 in all three runner modes (disabled, walltime, analysis)
• packed the plugin into a standalone project on vite 8.2.2 + vitest 4.1.11: benches run, walltime results written, and the published .d.ts typechecks
• backward compat: same tarball typechecks and benches on vite 7.3.6, and examples/with-vitest-v3 (vitest 3.2.4) and examples/with-typescript-esm (vitest 4.0.18) still pass on vite 7

Vitest 4.1 prints deprecation warnings for the vitest/runners and vitest/suite subpath imports the plugin uses. That is fixed by #75, stacked on this branch.

[codspeed-hq]

Merging this PR will regress 15 benchmarks

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 5 improved benchmarks
❌ 15 regressed benchmarks
✅ 218 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	recursive fibo 10	299.9 µs	1,372.5 µs	-78.15%
❌	Simulation	switch 1	202.1 µs	534.3 µs	-62.17%
❌	WallTime	fibo darwin	8.8 ms	11.9 ms	-25.59%
❌	Memory	end	2 KB	2.6 KB	-24.55%
❌	Memory	short body	656 B	784 B	-16.33%
❌	Memory	short body	656 B	784 B	-16.33%
❌	Memory	short body	656 B	784 B	-16.33%
❌	WallTime	short body	1.9 µs	2.2 µs	-14.29%
❌	WallTime	switch 1	72 ns	84 ns	-14.29%
❌	WallTime	switch 1	72 ns	84 ns	-14.29%
❌	WallTime	switch 2	84 ns	96 ns	-12.5%
❌	Memory	two	1.8 KB	2.1 KB	-11.94%
❌	Memory	one	2 KB	2.2 KB	-11.27%
❌	WallTime	short body	1.9 µs	2.2 µs	-10.56%
❌	WallTime	test sync baz 10	108 ns	120 ns	-10%
⚡	Simulation	test_recursive_fibo_10	11,699.9 µs	274.3 µs	×43
⚡	Simulation	recursive fibo 10	1,371.7 µs	300.5 µs	×4.6
⚡	Memory	wait 1sec	64.2 KB	20.3 KB	×3.2
⚡	WallTime	test_recursive_cached_fibo_10	2.3 µs	2 µs	+15.34%
⚡	WallTime	switch 2	96 ns	84 ns	+14.29%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing feat/vite-8-support (facdae6) with main (9338d9a)}

[tats-u]

This branch has conflicts that must be resolved

[adriencaccia]

This branch has conflicts that must be resolved

@tats-u as stated in the PR description, to push this pr we need to first bump the node version of the repo. More info in #76

[tats-u]

I don't think this is necessary. Either of the following options should be sufficient:

• Implement a dedicated workflow and project (package.json) for older Node versions, limiting testing scope (install using the latest Node until just before testing, then switch back)
• Reduce support level for older Node versions to "as-is" or "may address bug reports only if they come in"

It's important to note that the 20.19 release was over a year ago.

[tats-u]

If you want to detect whether Vite is v8, you can use this.meta.rolldownVersion.

[ematipico]

@adriencaccia any chance to have this PR resurrected? If not, I am happy to take over the works. We need this (plus some other fixes) to make the plugin work at Starlight.

At the moment, our biggest blockers are:

• Vite 8 support
• setup and teardown never called feat(vitest): call setup and teardown when running suite #46

[adriencaccia]

@adriencaccia any chance to have this PR resurrected? If not, I am happy to take over the works. We need this (plus some other fixes) to make the plugin work at Starlight.

It seems the biggest blocker for vite 8 was bumping the node version of the repo as explained in #76.
So since it is fixed, we can take up this PR I think yes!

Regarding the setup/teardown, @GuillaumeLagrange do you remember why we closed #46?

[GuillaumeLagrange]

Regarding the setup/teardown, @GuillaumeLagrange do you remember why we closed #46?

No particular reason comes to mind, I think it's just a branch that was forgotten and never merged

[ematipico]

Thank you. It would be great to have those hooks fixed. For some reason, the benchmarks stale https://github.com/withastro/starlight/actions/runs/33494510505/job/99813536944?pr=4157 and we believe the cause is those hooks