Vite removes source code on MacOS

[sheremet-va]

Describe the bug

When building browser package, I expect code from client/main.ts to be bundled inside assets, but instead Vite omits it.

Reproduction

https://github.com/vitest-dev/vitest/tree/main/packages/browser

Steps to reproduce

Only reproducible on MacOS.

pnpm install
pnpm build
cd packages/browser

Inside dist/client folder you should see html file and assets folder. I expect client/main.ts code to be inside assets/index.[hash].js file, but instead it contains only polyfill for preload.

System Info

System:
    OS: macOS 13.0
    CPU: (8) arm64 Apple M1
    Memory: 67.31 MB / 8.00 GB
    Shell: 5.8.1 - /bin/zsh
  Binaries:
    Node: 16.17.1 - ~/.nvm/versions/node/v16.17.1/bin/node
    Yarn: 1.22.19 - ~/.nvm/versions/node/v16.17.1/bin/yarn
    npm: 8.15.0 - ~/.nvm/versions/node/v16.17.1/bin/npm
  Browsers:
    Chrome: 107.0.5304.110
    Firefox: 107.0
    Safari: 16.1

Used Package Manager

pnpm

Logs

No response

Validations

• Follow our Code of Conduct
• Read the Contributing Guidelines.
• Read the docs.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.
• Make sure this is a Vite issue and not a framework-specific issue. For example, if it's a Vue SFC related bug, it should likely be reported to vuejs/core instead.
• Check that this is a concrete bug. For Q&A open a GitHub Discussion or join our Discord Chat Server.
• The provided reproduction is a minimal reproducible example of the bug.

[sodatea]

Because of sideEffects: false: https://github.com/vitest-dev/vitest/blob/f789776546d3b08d40ffee3cec402c1beda65a04/packages/browser/package.json#L11

[userquin]

@sodatea why Windows is working with Node 16.17.0? I build the package and the code is there...

D:\work\projects\userquin\GitHub\antfu\vitest>node --version
v16.17.0

D:\work\projects\userquin\GitHub\antfu\vitest>pnpm vite --version
vite/3.2.3 win32-x64 node-v16.17.0

[sheremet-va]

Yes, why windows works, and why „sideEffects“ matter when bundling source code? There is no other way in Vite to import my code.

[sodatea]

I think it's because the actual entry file passed to Rollup is index.html and it tree-shakes its dependencies?
But I don't know

1. whether sideEffects should be taken into account in this case. My gut tells me it should not.
2. why it doesn't tree-shake on Windows.

[userquin]

@sodatea I've refactored browser package structure and now it seems to be working properly: maybe Vite with rollup.config and vite.config mixing builds? I mean, we've 2 builds: rollup.config is configured to build node stuff (the entry point is node/index.ts, in the PR moved to src/node folder), and vite.config to build client stuff (in the PR, I've moved all vite stuff inside src/client folder changing the root in the build script): vitest-dev/vitest#2367

[powerfulyang]

I think it's because the actual entry file passed to Rollup is index.html and it tree-shakes its dependencies? But I don't know

1. whether sideEffects should be taken into account in this case. My gut tells me it should not.
2. why it doesn't tree-shake on Windows.

when use relative path, <script type="module" src="./client/main.ts"></script> will reproduce the bug, but it can't be triggered by absolute path <script type="module" src="/client/main.ts"></script>. By the way, this bug only appear in unix or linux.

[fregante]

I'm pretty sure this is a duplicate of:

• v3.2.0 missing output #10735

[patak-dev]

@sheremet-va this seems to be fixed to me. Could you check on your side?
I thought that vite@4.3.0-beta.2 fixed this one after #12628, but I'm not able to reproduce it in 4.2.1 (maybe the fix in rollup did the trick here). Or I'm not testing it correctly 👀

[patak-dev]

I think this is indeed a duplicate of #10735. I just sent a PR for it, let's keep discussing on the other issue.