deps: update undici to 7.15.0

[nodejs-github-bot]

This is an automated update of undici to 7.15.0.

[nodejs-github-bot]

Review requested:

• @nodejs/security-wg

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.93%. Comparing base (14c68e3) to head (b6559d5).
⚠️ Report is 38 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #59605      +/-   ##
==========================================
- Coverage   89.93%   89.93%   -0.01%     
==========================================
  Files         669      669              
  Lines      197473   197473              
  Branches    38588    38589       +1     
==========================================
- Hits       177604   177601       -3     
- Misses      12284    12287       +3     
  Partials     7585     7585              

see 24 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/68944/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/68984/

[targos]

@nodejs/undici The CI failures seem related.

[Uzlopak]

The thing is actually that we had some heavier changes since 7.15.0 in undici. 7.15.0 is more harmless,then what we have currently prepared for 7.16.0.

I am currently fighting with some stability or CI flakyness issues. And one test for shared-lib undici is failing on node 24 regularly. But this flakyness happened after release of 7.15.0.

[KhafraDev]

---
duration_ms: 175.13
exitcode: 1
severity: fail
stack: |-
  node:internal/deps/undici/undici:15523
        Error.captureStackTrace(err);
              ^

  TypeError: fetch failed
      at node:internal/deps/undici/undici:15523:13
      at async file:///home/iojs/build/workspace/node-test-commit-plinux/nodes/rhel9-ppc64le/test/parallel/test-fetch.mjs:28:18 {
    [cause]: HTTPParserError: Response does not match the HTTP/1.1 protocol (Invalid header value char)
        at Parser.execute (node:internal/deps/undici/undici:6456:21)
        at Parser.readMore (node:internal/deps/undici/undici:6411:16)
        at Socket.onHttpSocketReadable (node:internal/deps/undici/undici:6845:22)
        at Socket.emit (node:events:508:28)
        at emitReadable_ (node:internal/streams/readable:832:12)
        at process.processTicksAndRejections (node:internal/process/task_queues:89:21) {
      code: undefined,
      data: 'h: 11\r\n\r\nHello world'
    }
  }

  Node.js v25.0.0-pre

all similar errors from what I'm seeing

[Uzlopak]

Feels like llhttp 9.3.0 has issues. But it works on nodejs core?

[metcoder95]

Errors seems to be not always related with Undici, but under specific platforms; the only change that could be related might be lhttp, maybe there is some new lenient flag that might be causing that

[mcollina]

@nodejs/platform-aix The problem is actually specific to AIX and RedHat powerpc. I guess there is some problem there between our fetch WASM addons.

[mcollina]

My guess by looking at those errors points to the new SIMD-accelerated patches that went into llhttp: nodejs/llparse#73.

I think we could try loading https://github.com/nodejs/undici/blob/95d835cd9db685e4f7baaea8e10aee8ce77e7989/lib/dispatcher/client-h1.js#L66-L77 without SIMD in AIX to see if it fixes it.

[richardlau]

For Power, SIMD support requires Power 9+. Our OSUOSL hosted AIX machines are Power 8, as are our OSUOSL hosted RHEL 8 ppc64le machines.

https://ci.nodejs.org/job/node-test-commit-plinux/60592/ passed on RHEL 8 (provisioned as Power 8) but failed on rhel9, which are Power 9.

(nb. there's no link between RHEL version and Power architecture version -- it's just that the older VMs were created as Power 8 and the newer ones as Power 9.)

[richardlau]

https://ci.nodejs.org/job/node-test-commit-aix/58851/nodes=aix72-ppc64/ ran on an IBM Cloud hosted AIX machine, which I think is Power 9.

[targos]

So maybe V8's implementation of WASM SIMD is broken on Power 9?

[richardlau]

Spoke to @mcollina -- suggested for now to use non-SIMD on all of AIX and Linux ppc64le (os.type() == 'ppc64') for now. IBM will investigate separately whether there is an issue in V8.

[Uzlopak]

@richardlau

undici 7.16.0 got released, can you create a update undici 7.16.0 PR?

[richardlau]

Being superseded by undici 7.16.0.

[richardlau]

undici 7.16.0 got released, can you create a update undici 7.16.0 PR?

@Uzlopak I ran https://github.com/nodejs/node/actions/runs/17585683821/job/49952913962 and that picked up undici 7.15.0 again: #59827

https://github.com/nodejs/undici/releases is still showing 7.15.0 as the latest undici release.

[targos]

The release publication workflow failed: https://github.com/nodejs/undici/actions/runs/17582439038/job/49941839558