[BUG] WebKit (Linux): Assertion error in WKPage._onTargetCreated on a cross-site redirect chain
Maintainers usually reply within 1 day
@yury-s is already working on this.
Since Oct 6, 2026.
Assessment
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Newbie friendliness
- 72/100
- Issue type
- Bug
- Clarity
- Clearly specified
- Activity status
- Active
- Tech stack
- javascript, typescript
- Domain
- testing-qa
Research direction
Start at WKPage._onTargetCreated and inspect the page-goto.spec.ts regression test named “should work with cross-process redirect chain through visited sites.” Reproduce the Linux WebKit redirect chain with repro.mjs, then verify the assertion is avoided and navigation reaches /done while the regression test passes.
Written by the indexing model from the issue text.
Description
Version
1.63.0 (WebKit r2359); also current main (WebKit r2369). 1.64.0-alpha-2026-10-02 is affected as well.
Environment
- Linux:
mcr.microsoft.com/playwright:v1.63.0-noble, GitHububuntu-24.04runners. Headless, and headed underxvfb-run. - macOS (WebKit r2369): not affected.
Steps to reproduce
Three loopback hosts act as three sites: localhost (C), 127.0.0.1 (B), [::1] (A).
The page visits C → B → A by clicking links. A form on A then posts to /go, which returns a 302 to B/hop, which returns a 302 to C/done.
// repro.mjs: node repro.mjs (env: VARIANT=get, PRE=c,b (default), N=3, HEADED=1)
import http from 'node:http';
import { webkit } from 'playwright';
const PORT = 8123;
const H = { a: '[::1]', b: '127.0.0.1', c: 'localhost' };
const method = process.env.VARIANT === 'get' ? 'get' : 'post';
const pre = (process.env.PRE ?? 'c,b').split(',').filter(Boolean);
const server = http.createServer((req, res) => {
const u = new URL(req.url, `http://${req.headers.host}`);
const next = u.searchParams.get('next');
if (u.pathname === '/page') {
res.writeHead(200, { 'content-type': 'text/html' });
const href = next ? `http://${H[next.split(',')[0]]}:${PORT}/page?next=${next.split(',').slice(1).join(',')}` : '/form';
res.end(next ? `<a href="${href}">next</a>` : `<form method="${method}" action="/go"><button>go</button></form>`);
} else if (u.pathname === '/go') {
req.resume(); res.writeHead(302, { location: `http://${H.b}:${PORT}/hop` }); res.end();
} else if (u.pathname === '/hop') {
res.writeHead(302, { location: `http://${H.c}:${PORT}/done` }); res.end();
} else {
res.writeHead(200, { 'content-type': 'text/html' }); res.end('<h1>done</h1>');
}
});
await new Promise(r => server.listen(PORT, '::', r));
const browser = await webkit.launch({ headless: !process.env.HEADED });
let fails = 0;
const N = Number(process.env.N || 3);
for (let i = 0; i < N; i++) {
const page = await browser.newPage();
try {
const hosts = [...pre, 'a'];
await page.goto(`http://${H[hosts[0]]}:${PORT}/page?next=${hosts.slice(1).join(',')}`);
for (let k = 1; k < hosts.length; k++) {
await page.getByRole('link', { name: 'next' }).click();
await page.waitForURL(url => url.host.startsWith(H[hosts[k]]));
}
await page.getByRole('button', { name: 'go' }).click();
await page.waitForURL(/\/done$/, { timeout: 10_000 });
console.log(`run ${i}: ok ${page.url()}`);
} catch (e) {
fails++; console.log(`run ${i}: FAIL ${String(e.message).split('\n')[0]}`);
}
await page.close().catch(() => {});
}
console.log(`variant=${method} pre=${pre.join('>')}>a fails=${fails}/${N}`);
await browser.close(); server.close();
docker run --rm --ipc=host -v "$PWD":/w -w /w mcr.microsoft.com/playwright:v1.63.0-noble \
bash -c 'npm i --silent [email protected] >/dev/null && node repro.mjs'
Expected behavior
The page lands on http://localhost:8123/done, as it does on macOS.
Actual behavior
On the first run, the driver process crashes with an unhandled assertion:
Error: Assertion error
at assert (playwright-core/lib/coreBundle.js:838)
at _WKPage._onTargetCreated (playwright-core/lib/coreBundle.js:47026)
at WKSession.emit ...
In the test runner, the assertion is swallowed and the navigation times out instead.
It reproduces with POST and with GET. It does not reproduce when only one of B and C was visited before, or neither (PRE=b, PRE=).
Analysis
Protocol sequence on Linux:
Target.targetCreated page-25 {isProvisional:true} // swap #1, after the 302 from A to B
[page-25] requestWillBeSent http://127.0.0.1/hop
[page-25] Page.willCheckNavigationPolicy
Target.targetCreated page-28 {isProvisional:true} // swap #2 while page-25 is still provisional -> assert(!this._provisionalPage)
Target.targetDestroyed page-25 // old provisional target is destroyed only after the new one is created
On macOS, provisional page-30 follows the /hop → /done redirect itself and commits, so there is no second swap.
On Linux, WebKit swaps processes again for the redirect target before the first provisional page commits, and it creates the new target before it destroys the old one.
WKPage._onTargetCreated asserts that no provisional page exists yet.
This is a different cause from #37766 (a WebKit SIGABRT on reload) and #42957 (COOP). No COOP/COEP headers are involved.
Proposed fix
In WKPage._onTargetCreated, when a provisional target arrives while _provisionalPage is set, dispose the stale provisional page instead of asserting.
The targetDestroyed that follows for the old id then matches nothing and is a no-op.
A PR with this fix and a regression test (page-goto.spec.ts › should work with cross-process redirect chain through visited sites) follows.
- Dominant language
- TypeScript
- Stars
- 100k
- Forks
- 6.9k
- Avg merge
- 16h 8m
- Merged PRs (30d)
- 281
Getting set up
- No Dockerfile or Docker Compose file
- No pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from microsoft/playwright
-
[MCP]: page snapshots in .playwright-mcp/ are not git-ignored (.playwright-cli/ is, since #42318)Open
Difficulty 2/5 1-3 hours Newbie friendliness 62/100
microsoft/playwright#44289 ·
Maintainers usually reply within 1 day
-
agent-triage-done
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
microsoft/playwright#44264 · 2 comments ·
Maintainers usually reply within 1 day
-
Difficulty 5/5 Over a week Newbie friendliness 12/100
microsoft/playwright#44290 ·
Maintainers usually reply within 1 day
-
Difficulty 4/5 3-5 days Newbie friendliness 22/100
microsoft/playwright#44288 ·
Maintainers usually reply within 1 day
-
[Bug]: page.goto never returns for a load started as a newer service worker takes over (Chromium)Open
Difficulty 5/5 Over a week Newbie friendliness 12/100
microsoft/playwright#44287 ·
Maintainers usually reply within 1 day
All issues in microsoft/playwright
Similar issues
-
clawsweeper:needs-maintainer-review clawsweeper:needs-product-decision clawsweeper:no-new-fix-pr clawsweeper:source-repro impact:ux-friction issue-rating: 🦞 diamond lobster P3
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
openclaw/openclaw#203385 · 1 comment · 1 reaction ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
nwverse/persona#12800 · 1 comment ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 80/100
Maintainers usually reply within 1 day
-
good first issue
Difficulty 2/5 Half a day Newbie friendliness 65/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
WISE-Community/WISE-Client#2404 ·
Maintainers usually reply within 1 day