[Bug]: v1.59 breaks WebKit WASM

[peterhirn]

Version

1.59.1

Steps to reproduce

Repo: https://github.com/phi-ag/argon2

I'm running a very simple e2e test, see https://github.com/phi-ag/argon2/blob/main/e2e/index.e2e.ts

Everything works with v1.58.2, but after upgrading to v1.59.x WebKit always fails. Releated PR: phi-ag/argon2#704

Running 3 tests using 1 worker
[WebServer] Listening on port 3001
··×F

  1) [webkit] › e2e/index.e2e.ts:5:1 › hash and verify ─────────────────────────────────────────────

    Error: expect(locator).toHaveText(expected) failed

    Locator:  locator('#encodedWithSalt')
    Expected: "$argon2id$v=19$m=65536,t=3,p=4$YXNkZmFzZGZhc2RmYXNkZg$Z6jf3u1V2pXhzHdPMexmG6mG+i5486N1/fV/nlvUI60"
    Received: "$argon2id$v=19$m=65536,t=3,p=4$YXNkZmFzZGZhc2RmYXNkZg$gUlORCnWzMxGpicykVAPxO9zQhQDc56uZ+HFlRYl4pA"
    Timeout:  5000ms

    Call log:
      - Expect "toHaveText" with timeout 5000ms
      - waiting for locator('#encodedWithSalt')
        9 × locator resolved to <div id="encodedWithSalt">$argon2id$v=19$m=65536,t=3,p=4$YXNkZmFzZGZhc2RmYX…</div>
          - unexpected value "$argon2id$v=19$m=65536,t=3,p=4$YXNkZmFzZGZhc2RmYXNkZg$gUlORCnWzMxGpicykVAPxO9zQhQDc56uZ+HFlRYl4pA"


      10 |
      11 |   const encodedWithSalt = page.locator("#encodedWithSalt");
    > 12 |   await expect(encodedWithSalt).toHaveText(
         |                                 ^
      13 |     "$argon2id$v=19$m=65536,t=3,p=4$YXNkZmFzZGZhc2RmYXNkZg$Z6jf3u1V2pXhzHdPMexmG6mG+i5486N1/fV/nlvUI60"
      14 |   );
      15 |
        at /workdir/e2e/index.e2e.ts:12:33

While researching why this is broken I upgraded my iPad to the latest version (iPadOS 26.4.1) and opened the e2e test page in Safari (node e2e/server.js). Everything seems to work there, but I don't know which exact WebKit version is used.

Expected behavior

Actual behavior

Additional context

No response

Environment

Docker, see https://github.com/phi-ag/argon2/blob/main/e2e_docker.sh

[yury-s]

Could you provide a minimal repro? This likely points to an upstream WebKit regression, but we need a small, self-contained case—we don’t have the bandwidth to dig through a complex project.

[peterhirn]

@yury-s Can you take a quick look into the repo linked above? It is very small. Not a minimal repro though, maybe I can find some time in the next couple of weeks to create something even smaller.

[yury-s]

There are no build/test instructions there, I've tried the following and it fails.

npm i
npm run build
npm run test

Help us help you and provide a simple repro.

[peterhirn]

Right, sorry, will add full instructions to the readme when I'm back from holiday.

Off the top of my head:

fnm use
pnpm i
./build_docker.sh
pnpm build
./e2e_docker.sh