Vite dev server crashes Remix / React when reloading after finding new dependencies to optimize

[astonfuture]

Reproduction

https://github.com/astonfuture/remix-vite-optimize-dependency-crash

1. Run the dev server and force clear vite's cache:
   npm run dev -- --force or delete node_module/.vite then npm run dev

2. Navigate to the index http://localhost:5173

3. Directly navigate to page-two by typing http://localhost:5173/page-two in the address bar and hitting enter

4. Vite detects the new dependency (lodash-es), runs the optimization and reloads the page which causes it to crash with hydration errors.

System Info

System:
    OS: macOS 14.6.1
    CPU: (12) arm64 Apple M3 Pro
    Memory: 68.77 MB / 18.00 GB
    Shell: 5.9 - /bin/zsh
  Binaries:
    Node: 21.7.1 - /opt/homebrew/bin/node
    npm: 10.7.0 - /opt/homebrew/bin/npm
    Watchman: 2024.04.08.00 - /opt/homebrew/bin/watchman
  Browsers:
    Brave Browser: 130.1.71.114
    Chrome: 130.0.6723.60
    Safari: 17.6
  npmPackages:
    @remix-run/cloudflare: ^2.13.1 => 2.13.1 
    @remix-run/dev: ^2.13.1 => 2.13.1 
    @remix-run/react: ^2.13.1 => 2.13.1 
    @remix-run/server-runtime: ^2.13.1 => 2.13.1 
    vite: ^5.4.10 => 5.4.10

Used Package Manager

npm

Expected Behavior

There should not be hydration errors which crash the page as you navigate

Actual Behavior

The reproduction repo above is the official Remix Cloudflare workers template with one extra route added. The new route imports a package (lodash-es) and uses it in the component. When you navigate to that page, vite detects the new package and runs its optimizer which causes a page reload. This causes all manner of errors and the page crashes.

Starting the dev server

> remix vite:dev --force

Forced re-optimization of dependencies
  ➜  Local:   http://localhost:5173/
  ➜  Network: use --host to expose
  ➜  press h + enter to show help

When navigating to a page with new dependencies:

7:09:40 AM [vite] ✨ new dependencies optimized: lodash-es
7:09:40 AM [vite] ✨ optimized dependencies changed. reloading

Page crashes with a white screen with console errors in the browser such as:

Uncaught TypeError: Cannot read properties of null (reading 'useState')

Warning: An error occurred during hydration. The server HTML was replaced with client content in <#document>.

Warning: Invalid hook call. Hooks can only be called inside of the body of a function component. This could happen for one of the following reasons:

Reloading the page after it crashes works fine, leading me to believe this has something to do with the HMR that happens when vite reloads.

[clgeoio]

I have seen a similar thing before, but it was hard to reproduce.
I was able to reproduce the behaviour with the above repo using both Safari and Brave browsers 👍

[rorcores]

A fix for this would be great please, thank you

[brampono]

I'm also experiencing this issue! Would love to see a fix 🙏

[joshbuddy]

From what I understand, its causing a hook error due to two versions of React. But this is sort of beyond my React paygrade to solve.

[clgeoio]

@joshbuddy, I found a similar issue on Github that suggested the same thing about different react versions.
I tried adding dedupe: ['react', 'react-dom'] to the vite.config, as suggested in that issue, however that did not fix.

I'll keep poking around to see what can I do to find where this other version of React is coming from 👀

[joshbuddy]

My guess at what is happening under the hood is that the two versions of react are coming from two different builds, and both builds are being loaded simultaneously. Conspicuously, if I do a hard refresh, Safari behaves normally again.

[astonfuture]

I think the hard refresh fix is due to the fact that the new optimized dependency is already loaded in node_modules/.vite and so it's available when you refresh. The error is something to do with the fact that the optimization and reload happens in a HMR-ey way while the page is still loaded.

[joshbode]

Adding the warmup option to vite.config.ts appears to work around the issue:

  server: {
    warmup: {
      clientFiles: ['app/**/*.tsx'],
    },
  },

[alexfoxy]

I'm running into this issue. Seems to be specific to safari in my case, on iOS and mac. @joshbode your config has somewhat helped but it still seemed to happen on occasion. Very strange - and annoying!

[f5io]

i'm also running into this issue, thought i was going mad, unfortunately @joshbode's config didnt seem to help. i'm migrating an older remix app to the new vite approach and attempting to get it running on cloudflare workers. a simple page refresh after the error is hit and everything is hunky-dory, but its not a great dev-experience.

[astonfuture]

i'm also running into this issue, thought i was going mad, unfortunately @joshbode's config didnt seem to help. i'm migrating an older remix app to the new vite approach and attempting to get it running on cloudflare workers. a simple page refresh after the error is hit and everything is hunky-dory, but its not a great dev-experience.

This is exactly the behaviour we're seeing and it's also a migration from an older remix on CF workers to vite.

[halljus]

I was hoping the combination of #9921 and enabling unstable_optimizeDeps would fix this, but unfortunately that was not the case. 🙁