packages/vite/src/utils/logger.ts (1)

93-107: The suggestion timer is never cancelled on server shutdown.

If the Vite dev server is stopped (e.g. user restarts Nuxt or hits Ctrl+C) while the 2500 ms timer is pending, the callback fires after teardown and calls logger.info on a potentially stale context. This is a minor concern in practice (dev-only, process usually exits), but for correctness you could clear it via a Nuxt close hook or expose a cleanup function.

Additionally, consider whether the return on line 77 (if (hasShownSuggestion) { return }) should come before or after the dependency-parsing block. Currently, when hasShownSuggestion is true, the entire log message (including Vite's original "new dependencies optimized" info) is suppressed. If the intent is only to suppress the hint, you'd want to move the early return to just before the timer scheduling (line 93) and let the original Vite message pass through.

packages/vite/src/plugins/dev-server.ts (1)

151-156: Consider extracting the 3000 ms magic number and clearing the timer on server close.

The 3-second delay is meaningful (waiting for the initial page load to settle) but unexplained. A named constant or brief comment would help future readers understand the rationale. Also, if the server is torn down within those 3 seconds, the timer fires on a stale context.

Suggested improvement

+      const FIRST_LOAD_SETTLE_MS = 3000 // Wait for initial page load to complete before silencing dep suggestions
       let firstRequest = true
+      let firstLoadTimer: ReturnType<typeof setTimeout> | undefined
       const viteMiddleware = defineEventHandler(async (event: H3V1Event | H3V2Event) => {
         if (firstRequest) {
           firstRequest = false
-          setTimeout(onFirstPageLoad, 3000)
+          firstLoadTimer = setTimeout(onFirstPageLoad, FIRST_LOAD_SETTLE_MS)
         }

Then clear firstLoadTimer in a Nuxt close hook if one is available in this scope.

packages/vite/src/utils/logger.test.ts (2)

25-29: Add afterEach to restore real timers.

vi.useFakeTimers() is called in beforeEach but real timers are never restored. While vitest may handle this between test files, it's best practice to pair it with vi.useRealTimers() in afterEach to avoid leaking fake timers into other tests in the same suite if the suite grows.

Proposed fix

   beforeEach(() => {
     vi.clearAllMocks()
     vi.useFakeTimers()
     resetSuggestionState()
   })
+
+  afterEach(() => {
+    vi.useRealTimers()
+  })

Don't forget to add afterEach to the import on line 1.

---

68-68: Brittle exact call-count assertion.

toHaveBeenCalledTimes(4) depends on knowing that output() forwards each of the 3 info messages to logger.info plus the 1 suggestion. If the internal forwarding logic changes (e.g. one of those messages gets suppressed or an extra log is added), this test breaks for the wrong reason. Consider asserting only that the suggestion message appeared exactly once, which is the behaviour under test.

Proposed alternative

-    expect(logger.info).toHaveBeenCalledTimes(4) // 3 original info logs + 1 suggestion
+    // Verify exactly one suggestion was emitted
+    const suggestionCalls = vi.mocked(logger.info).mock.calls.filter(
+      call => (call[0] as string).includes('Vite has discovered new dependencies'),
+    )
+    expect(suggestionCalls).toHaveLength(1)

packages/vite/src/utils/logger.test.ts (1)

25-28: Restore real timers after each test to avoid leaking fake timers.

vi.useFakeTimers() is called in beforeEach but never restored. This can affect other test suites in the same process.

Proposed fix

   beforeEach(() => {
     vi.clearAllMocks()
     vi.useFakeTimers()
   })
+
+  afterEach(() => {
+    vi.useRealTimers()
+  })

You'll also need to add afterEach to the import on line 1:

-import { beforeEach, describe, expect, it, vi } from 'vitest'
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'