`nx serve --prod` not working with cache hit

Current Behavior

When the build is retrieved from cache, nx serve --prod does not work

Expected Behavior

nx serve --prod should always work, even when the cache is hit for retrieving the build

Steps to Reproduce

npx create-nx-workspace@latest repo --preset=next
cd repo
nx generate @nrwl/next:application webapp

// Commands sequence

nx serve
nx serve --prod
// Error (expected): "Could not start production server. Try building your app with `nx build webapp`."
nx build
nx serve --prod

// Now, let's re-do the exact same flow

nx serve
nx build
// Build is instant as expected because it is retrieved from cache
nx serve --prod
// Error (unexpected): "Could not start production server. Try building your app with `nx build webapp`."

Environment

Node : 16.15.0 OS : darwin arm64 pnpm : 7.1.0

nx : 14.1.5 @nrwl/angular : Not Found @nrwl/cypress : 14.1.5 @nrwl/detox : Not Found @nrwl/devkit : 14.1.5 @nrwl/eslint-plugin-nx : 14.1.5 @nrwl/express : Not Found @nrwl/jest : 14.1.5 @nrwl/js : 14.1.5 @nrwl/linter : 14.1.5 @nrwl/nest : Not Found @nrwl/next : 14.1.5 @nrwl/node : Not Found @nrwl/nx-cloud : Not Found @nrwl/nx-plugin : Not Found @nrwl/react : 14.1.5 @nrwl/react-native : Not Found @nrwl/schematics : Not Found @nrwl/storybook : 14.1.5 @nrwl/web : 14.1.5 @nrwl/workspace : 14.1.5 typescript : 4.6.4 rxjs : 6.6.7

[jaysoo]

This will be fixed in the next release. For the time being, set outputPath to something else as the workaround.

{
  // ...
  "targets": {
    "build": {
      "executor": "@nrwl/next:build",
      "outputs": [
        "{options.outputPath}"
      ],
      "defaultConfiguration": "production",
      "options": {
        "root": "apps/demo",
        "outputPath": "dist/apps/demo"
      },
      "configurations": {
        "development": {
          // NOTE: Point to a different path so there aren't any collisions.
          "outputPath": "tmp/apps/demo"
        },
        "production": {}
      }
    },
    // ...
}

[Phault]

Pretty big issue I’d say, as this can lead to shipping broken production builds. In our case the CI built production builds, then ran e2e tests which mistakenly was testing with serve:development, if tests succeeded it would deploy. That’s obviously not great, but the deploy step was dependent on build:production, so it should’ve been all good. Nx said there was a cache hit but the development build would end up deployed.

Nothing is actually in production yet, so no real harm done.