perf(vite): remove vite warm up

[antfu]

🔗 Linked issue

❓ Type of change

• 📖 Documentation (updates to the documentation or readme)
• 🐞 Bug fix (a non-breaking change that fixes an issue)
• 👌 Enhancement (improving an existing functionality like performance)
• ✨ New feature (a non-breaking change that adds functionality)
• ⚠️ Breaking change (fix or feature that would cause existing functionality to change)

📚 Description

This reduces the start-up time of nuxt.com repo with viteNode: true (nuxi dev to first render on the browser)
from 60s to 40s on my M1 Max Macbook Pro.

Reasoning:

• Vite will now do warm-up for sync deps automatically.
• Our warm-up uses server.transformRequest(url) for both client/server. Where the server bundler actually needs server.transformRequest(url, { ssr: true }), which is actually using a different plugin pipeline and caches.

📝 Checklist

• I have linked an issue or discussion.
• I have updated the documentation accordingly.

[netlify]

✅ Deploy Preview for nuxt3-docs canceled.

Name	Link
🔨 Latest commit	f0a7311
🔍 Latest deploy log	https://app.netlify.com/sites/nuxt3-docs/deploys/62fa622147e8880008a4a2e5

[pi0]

Vite will now do warm-up for sync deps automatically.

Does Vite also warmups the entrypoint in background? (Same for vite-node with ssr condition)

Our warm-up uses server.transformRequest(url) for both client/server. Where the server bundler actually needs server.transformRequest(url, { ssr: true }), which is actually using a different plugin pipeline and caches.

Nice find and probably to late to fix if warmup is not relevant anymore 😅

[pi0]

If this is not relevant anymore I think we can remove util as well. Only I want to make sure vite and vite-node does the same because in past, without entrywarmup we were simply delaying it until first request.

[antfu]

I also tried to only warm up the entry (and let Vite crawl the deps), it ends up around 50s start-up time. I guess it might be because it occupied the execution time of other parts. I could do a more detailed benchmarking, and probably try delaying the warm-up with a timeout.

[pi0]

I suggest this method to benchmark:

• Cleanup cache with nuxi cleanup each time
• In parallel initiate a fetch to http://localhost:300 and start nuxi dev
  • Measure time to first byte for SSR perf
• Open a new browser window ensuring no other windows are open
  • Measure time to load for entrypoint scripts

[pi0]

Let's experiment this as an option that we can disable. (sorry there were merge conflicts otherwise would push to same PR) ~> #6647.

We should also track this:

Our warm-up uses server.transformRequest(url) for both client/server. Where the server bundler actually needs server.transformRequest(url, { ssr: true }), which is actually using a different plugin pipeline and caches.

~> #6649