[GH-ISSUE #2916] [Bug] Real browser fixture teardown waits on speculative HTTP connections #1027

New issue

Closed

opened 2026-10-15 18:30:28 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-15 18:30:28 +00:00

Owner

Originally created by @N0zoM1z0 on GitHub (Oct 13, 2026).
Original GitHub issue: https://github.com/morluto/rea/issues/2916

The shared real-browser fixture can wait indefinitely during HTTP-server teardown when a connection opens without sending request headers. The complete browser verifier awaits this cleanup in finally, so any earlier assertion/error remains hidden while cleanup waits.

Confirmed reproduction

Pinned main d4f0c5e664e26fe7ed1566dcc4b85ea9b57d52da, actual Windows x64 Node24.18.0:

Start startBrowserVerifierSite() from scripts/fixtures/browser-verifier-site.mjs.
Control: fetch its HTTP origin and consume the body; site.close() settles within3s.
Start a fresh fixture; connect a retained node:net socket to its origin, without sending HTTP headers.
Call site.close(): it is still pending after3s.
Destroy that same recorded client socket: the pending close settles.

Fresh diagnostic result:

{"ordinary_http_closed":true,"speculative_tcp_closed_before_destroy":false,"closed_after_destroy":true,"root_cause_of_full_browser_run_established":false,"node":"v24.18.0","platform":"win32"}

The diagnostic uses real local sockets and no browser/mocked provider. Its enclosing owned Job settled with zero processes:8.92s elapsed,33.1MiB peak Job commit, two shared logical CPUs and BelowNormal priority.

Producing boundary

The fixture terminates its tracked WebSocket clients and awaits WebSocket-server closure, then awaits server.close() without force-closing its accepted HTTP connections. Other fixture producers already handle this boundary: commit e5bc4e85ec5669399ea506233632b118dead0692 fixed speculative browser connections in the scenario-storage fixture.

Expected: fixture teardown must close the fixture's own accepted connections and listener, independently of whether a client sends headers. Preserve WebSocket teardown and handle ownership; do not increase deadlines or hide earlier verifier failures.

Fresh actual Edge scripts/verify-real-browser.mjs reached this audit's300s enclosing budget; the last marker was closing fixture site, then the outer Job terminated with zero descendants. That composite run used729MiB Job commit; memory did not reach the2GiB bound.

The socket probe proves a fixture cleanup gap, not the root cause of that full Edge timeout. The ten-minute CI composite lane and an earlier functional failure must be assessed separately; source shows a pending finally can mask that earlier error. Add a small real-socket regression and rerun the actual browser workflow with useful phase/error diagnostics.

Local raw diagnostic/source/receipts: /tmp/rea-windows-e2e-20261013/results/browser-close-probe.*. This issue will be handled serially after the preceding Windows audit fixes.

Originally created by @N0zoM1z0 on GitHub (Oct 13, 2026). Original GitHub issue: https://github.com/morluto/rea/issues/2916 The shared real-browser fixture can wait indefinitely during HTTP-server teardown when a connection opens without sending request headers. The complete browser verifier awaits this cleanup in `finally`, so any earlier assertion/error remains hidden while cleanup waits. ## Confirmed reproduction Pinned main `d4f0c5e664e26fe7ed1566dcc4b85ea9b57d52da`, actual Windows x64 Node24.18.0: 1. Start `startBrowserVerifierSite()` from `scripts/fixtures/browser-verifier-site.mjs`. 2. Control: fetch its HTTP origin and consume the body; `site.close()` settles within3s. 3. Start a fresh fixture; connect a retained `node:net` socket to its origin, without sending HTTP headers. 4. Call `site.close()`: it is still pending after3s. 5. Destroy that same recorded client socket: the pending close settles. Fresh diagnostic result: ```json {"ordinary_http_closed":true,"speculative_tcp_closed_before_destroy":false,"closed_after_destroy":true,"root_cause_of_full_browser_run_established":false,"node":"v24.18.0","platform":"win32"} ``` The diagnostic uses real local sockets and no browser/mocked provider. Its enclosing owned Job settled with zero processes:8.92s elapsed,33.1MiB peak Job commit, two shared logical CPUs and BelowNormal priority. ## Producing boundary The fixture terminates its tracked WebSocket clients and awaits WebSocket-server closure, then awaits `server.close()` without force-closing its accepted HTTP connections. Other fixture producers already handle this boundary: commit `e5bc4e85ec5669399ea506233632b118dead0692` fixed speculative browser connections in the scenario-storage fixture. Expected: fixture teardown must close the fixture's own accepted connections and listener, independently of whether a client sends headers. Preserve WebSocket teardown and handle ownership; do not increase deadlines or hide earlier verifier failures. ## Related full-workflow observation and limits Fresh actual Edge `scripts/verify-real-browser.mjs` reached this audit's300s enclosing budget; the last marker was `closing fixture site`, then the outer Job terminated with zero descendants. That composite run used729MiB Job commit; memory did not reach the2GiB bound. The socket probe proves a fixture cleanup gap, **not the root cause of that full Edge timeout**. The ten-minute CI composite lane and an earlier functional failure must be assessed separately; source shows a pending `finally` can mask that earlier error. Add a small real-socket regression and rerun the actual browser workflow with useful phase/error diagnostics. Local raw diagnostic/source/receipts: `/tmp/rea-windows-e2e-20261013/results/browser-close-probe.*`. This issue will be handled serially after the preceding Windows audit fixes.

bsakaguchi

2026-10-15 18:30:28 +00:00

closed this issue
added the
area: runtime

bug
labels

bsakaguchi commented

2026-10-15 18:30:29 +00:00

Author

Owner

@morluto commented on GitHub (Oct 14, 2026):

Addressed on main in eca18e08ac80ac838eef78788bf82429bbb30673.

The browser fixture closes speculative idle HTTP connections and tracked WebSockets during teardown; a real idle TCP regression covers this owner.

Validation: Affected suites passed in the merged 65-file run (1169 passed, 2 explicit skips), with full static/architecture checks and documentation checks passing. Final installed-package verification is recorded separately in the audit. Verifier trust-boundary controls cover unavailable lineage, terminal modes, idle TCP teardown, and shared fixture cleanup; no real-provider verifier success is inferred.

Verification scope: Controlled source and production-boundary verification on macOS with pinned Node 24.18.0; real external engines and Windows hosts were not independently rerun. Closure covers the described source defect, not broader provider conformance or performance.

@morluto commented on GitHub (Oct 14, 2026): Addressed on main in [eca18e08ac80ac838eef78788bf82429bbb30673](https://github.com/morluto/rea/commit/eca18e08ac80ac838eef78788bf82429bbb30673). The browser fixture closes speculative idle HTTP connections and tracked WebSockets during teardown; a real idle TCP regression covers this owner. Validation: Affected suites passed in the merged 65-file run (1169 passed, 2 explicit skips), with full static/architecture checks and documentation checks passing. Final installed-package verification is recorded separately in the audit. Verifier trust-boundary controls cover unavailable lineage, terminal modes, idle TCP teardown, and shared fixture cleanup; no real-provider verifier success is inferred. Verification scope: Controlled source and production-boundary verification on macOS with pinned Node 24.18.0; real external engines and Windows hosts were not independently rerun. Closure covers the described source defect, not broader provider conformance or performance.

bsakaguchi referenced this issue

2026-10-16 21:02:34 +00:00

[GH-ISSUE #2383] [Tracker] Audit analysis and debugging coverage for the TIOBE top 100 languages (October 2026) #794

bsakaguchi referenced this issue

2026-10-17 21:07:57 +00:00

[PR #1027] [MERGED] fix(windows): allow Java to read retained target snapshots #2591

bsakaguchi referenced this issue

2026-10-17 21:07:58 +00:00

[PR #1030] [MERGED] feat(ghidra): support native x86 PE applications on Windows #2594