[Bug]: eval- based source maps generate broken build on some systems

[cgarwood]

System Info

vscode ➜ /workspaces/frontend (dev) $ npx envinfo --system --npmPackages '@rspack/*' --binaries --browser

  System:
    OS: Linux 6.8 Debian GNU/Linux 12 (bookworm) 12 (bookworm)
    CPU: (4) x64 Common KVM processor
    Memory: 3.13 GB / 11.68 GB
    Container: Yes
    Shell: 5.2.15 - /bin/bash
  Binaries:
    Node: 20.18.1 - /usr/local/share/nvm/versions/node/v20.18.1/bin/node
    Yarn: 4.5.3 - /usr/bin/yarn
    npm: 10.8.2 - /usr/local/share/nvm/versions/node/v20.18.1/bin/npm
  npmPackages:
    @rspack/cli: 1.1.6 => 1.1.6 
    @rspack/core: 1.1.6 => 1.1.6 

Details

Developing on an Ubuntu system, I'm getting broken builds when using eval- based options for the devtool source maps. Same issue on Ubuntu 20, 22, and on a VSCode devcontainer hosted on both systems. Tested across two VMs (one running Ubuntu 20.04, another VM running Ubuntu 24.04)

Interestingly enough, the exact same code, run in a VSCode devcontainer on a Windows system builds OK with the eval- options.

For example:
devtool: "cheap-module-source-map" works fine
devtool: "eval-cheap-module-source-map" generates a broken build.

Broken builds are typically failing with an error about const variables being redeclared.

A diff for one of the main files in the build matches between the Windows and Ubuntu machines with non-eval source maps.
A diff for the same file between both machines using eval- source maps generates a pretty significant diff, but I'm not well enough versed to be able to parse it enough to tell what's going on.

home-assistant/frontend#23231 has more information on the issue I'm seeing and contains the diff files.

Reproduce link

https://github.com/home-assistant/frontend

Reproduce Steps

Looks like something that may be difficult to reproduce as nobody else working on the Home Assistant project has been able to reproduce it.

https://developers.home-assistant.io/docs/frontend/development has details for the Home Assistant project and setting up a full development environment. It's a little tricky as you need both the Python backend as well as the frontend repository.