[eps1lon]

Backports #97931
Part of https://linear.app/vercel/issue/VOC-34654/

Since 15.x supports a Node.js version not supported by a secure sharp version, we have to conditionally block AVIF optimization based on the used libheif version. I added a test that runs for each support major sharp version to validate only the fixed version optimizes AVIF.

[github-actions]

Failing test suites

Commit: ac65ec4 | About building and testing Next.js

pnpm test-start test/e2e/app-dir/ppr-partial-hydration/ppr-partial-hydration.test.ts (job)

• PPR - partial hydration > Static shell, no streaming metadata > should hydrate the shell without waiting for slow suspense boundaries (DD)
• PPR - partial hydration > Static shell, streaming metadata > should hydrate the shell without waiting for slow suspense boundaries (DD)

Expand output

● PPR - partial hydration › Static shell, no streaming metadata › should hydrate the shell without waiting for slow suspense boundaries

page.waitForSelector: Timeout 10ms exceeded.
Call log:
  - waiting for locator('#shell-hydrated') to be visible
  -   locator resolved to visible <div id="shell-hydrated" data-is-hydrated="true">🟢 Hydrated</div>

  491 |
  492 |     return this.startChain(async () => {
> 493 |       const el = await page.waitForSelector(selector, {
      |                             ^
  494 |         timeout,
  495 |         state,
  496 |       })

  at waitForSelector (lib/browsers/playwright.ts:493:29)
  at Playwright._chain (lib/browsers/playwright.ts:616:23)
  at Playwright._chain [as startChain] (lib/browsers/playwright.ts:597:17)
  at Playwright.startChain [as waitForElementByCss] (lib/browsers/playwright.ts:492:17)
  at Playwright.waitForElementByCss (lib/browsers/playwright.ts:398:17)
  at elementByCssInstant (e2e/app-dir/ppr-partial-hydration/ppr-partial-hydration.test.ts:75:14)
  at fn (lib/next-test-utils.ts:845:20)
  at Object.<anonymous> (e2e/app-dir/ppr-partial-hydration/ppr-partial-hydration.test.ts:71:7)
  at Proxy._chain (lib/browsers/playwright.ts:616:23)
  at Proxy._chain (lib/browsers/playwright.ts:592:17)
  at Proxy.continueChain (lib/browsers/playwright.ts:440:17)
  at getAttribute (e2e/app-dir/ppr-partial-hydration/ppr-partial-hydration.test.ts:76:14)
  at fn (lib/next-test-utils.ts:845:20)
  at Object.<anonymous> (e2e/app-dir/ppr-partial-hydration/ppr-partial-hydration.test.ts:71:7)

● PPR - partial hydration › Static shell, streaming metadata › should hydrate the shell without waiting for slow suspense boundaries

expect(received).toHaveLength(expected)

Expected length: 0
Received length: 1
Received array:  [{"_guid": "handle@93518a7147a7c3401e24ab07931186d8", "_type": "ElementHandle"}]

  57 |
  58 |           // The dynamic content hasn't streamed in yet, we should only see the fallback
> 59 |           expect(await browser.elementsByCss('main #dynamic')).toHaveLength(0)
     |                                                                ^
  60 |           expect(
  61 |             await browser
  62 |               .elementByCssInstant('#dynamic-fallback', { state: 'visible' })

  at toHaveLength (e2e/app-dir/ppr-partial-hydration/ppr-partial-hydration.test.ts:59:64)
  at retry (lib/next-test-utils.ts:845:14)
  at Object.<anonymous> (e2e/app-dir/ppr-partial-hydration/ppr-partial-hydration.test.ts:48:7)

not meant to be reviewed

[styfle]

Do we need to hardcode q here?

[eps1lon]

Removed the hardcoding which got reintroduced by reverting. It was hardcoded before we disabled AVIF.

[styfle]

Same question - Do we need to hardcode q here?

[eps1lon]

Removed the hardcoding which got reintroduced by reverting. It was hardcoded before we disabled AVIF.

[eps1lon]

Flaky test. Force-merging.