[teamleaderleo]

The browser shard can outlive its 40-minute job: agent-pane-attach-huge-files.test.ts declares a 3,000-second file allowance, other sweeps declare 600–900 seconds, and up to 60 failed files are retried serially. On #25304, run 40368616367 was cancelled without writing its head-results artifact. That shard reported 23 first-pass failures; its 900-second gallery sweep took about 11 minutes, and silent serial retries then ran for 17 minutes before cancellation. The 3,000-second stress-file allowance is a separate ceiling violation, not identified as the active file in this cancelled shard.

The runner now caps each file at 900 seconds, reports its name before launch and failure log as soon as it finishes, keeps TIMEOUT failures without another serial attempt, and limits the whole retry phase, including an in-flight attempt, to a 120-second shared budget. Every selected file still runs; failures left after the retry budget remain failures. Intentional standalone stress runs can explicitly raise CMUX_WEB_TEST_MAX_FILE_TIMEOUT. An incomplete retry retains the first named failure. Bun's default per-test timeout remains 30 seconds, enclosed by the hard file deadline even if a suite overrides it. The existing results artifact also carries the exact runner and result parser for base replay, so the head and base use the same deadline policy with their own tests and Bun version. PR path routing is unchanged.

claude-sr2 requested: “give every test a hard timeout so a hang turns into a named failure instead of a 37-min cancel.” This addresses the base blocker seen on #25304 and #25325. claimPaint is a separate false-positive: it finishes in 1.76s and counts two correctly disabled ordered-list boundary buttons. Existing #24469 owns that repair; its Codex owner has resumed it.

Regression command: python3 tests/test_webviews_tests_per_file.py PerFileRunner.test_a_declared_long_file_cannot_exceed_the_runner_hard_limit PerFileRunner.test_a_timed_out_file_reports_once_without_a_serial_retry PerFileRunner.test_serial_retries_have_a_shared_time_budget. The new tests fail at regression commit dd84051fdbf915 over base c80354516c120e04a3adba32258268611a5e1a73; all pass with the fix.

Additional red/green regressions cover an in-flight retry exceeding the remaining budget (7781af1dac4340), keeping its original named failure when that retry is incomplete (29826b3de600f4), and executing head/base workflow scripts with a stale base runner (af0430b3b59c7a). python3 tests/test_webviews_tests_per_file.py and python3 tests/test_webviews_tests_new_reds.py pass on the final head.

Changelog

none

Observed / Not run

• 19 per-file runner tests: passed, 26.086s.
• 20 new-red comparison wrapper/workflow tests: passed, 20.441s.
• bash scripts/ci/tests/run-webviews-tests-retry.test.sh: passed.
• bash -n scripts/ci/run-webviews-tests.sh, diff check and prepush gate: passed.
• actionlint .github/workflows/ci-web.yml: passed.
• No native builds. Hosted shard validation on base and two webview PRs is still pending; no completion claim from a skipped check.
• Independent Codex xhigh review held the initial head for mixed head/base deadline policies. That finding is repaired and the independent review returned Review (Codex xhigh): SHIP at 1736adf93024567e0c596b9df56360c48f30a1ab after rerunning all 39 focused tests, retry shell, actionlint and prepush gate.

Agent session: agent session

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review 0f27fcc: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr25548 -f ref=nx-ci-browser-shard-timeouts -f platform=macos -f backend_mode=local

Yours to fix: next-docs.test.ts (this PR changes the web test runner (.github/workflows/ci-web.yml)). Also failing here, not on main: composer-paste-images.test.tsx, composer-paste-images.test.tsx, chat-minimap.test.tsx and 8 more. Seen on other PRs too (likely flaky): chat-environment-scroll-stability.test.tsx, chat-environment-scroll-stability.test.tsx, composer-paste-images.test.tsx and 18 more. This PR merged before its CI finished: fix forward on main.

CI failed on 1736adf930 (run 40384050518 attempt 1): 5 code, 1 unknown.

Job	Verdict	Why
web / Web tests (2/4)	code	a web test failed
web / react-apps-base-check (3)	code	a web test failed
web / react-apps-base-check (4)	code	a web test failed
web / react-apps-base-check (2)	code	a web test failed
web / react-apps-base-check (1)	code	a web test failed
web / react-apps-check	unknown	no known signature; failed step: Check webview test shards

Failures:

• yours next-docs.test.ts cmux-next docs > the CLI and MCP reference is fresh against the action catalog: failed (job)
• seen on other PRs chat-environment-scroll-stability.test.tsx long-chat scroll leaves the Left chip mounted and stable, while own cwd changes update it: failed (job)
• seen on other PRs chat-environment-scroll-stability.test.tsx long-chat scroll leaves the Right chip mounted and stable, while own cwd changes update it: failed (job)
• seen on other PRs composer-paste-images.test.tsx pasting two copied images makes two chips and puts nothing in the prompt: failed (job)
• seen on other PRs composer-paste-images.test.tsx an image copied as HTML alone (no file on the clipboard) still comes in as a chip, its text as text: failed (job)
• seen on other PRs composer-paste-images.test.tsx a pasted image goes to the agent as an image block; the prompt's text has no data: URI: failed (job)
• seen on other PRs composer-paste-images.test.tsx an image copied as HTML into a plain text field comes in as a chip and the text still goes in: failed (job)
• new in this PR composer-paste-images.test.tsx a prompt copied from a bubble with '![…](data:…)' text pastes as a chip and its text, nothing inline: failed (job)
• new in this PR composer-paste-images.test.tsx the same text pasted into a plain text field leaves its text there without the image: failed (job)
• seen on other PRs claimPaint.test.tsx a pooled Settings page claimed by the host reports its paint once its first screen is in: failed (job)
• ... 23 more in the job logs

Matched log lines

web / Web tests (2/4): (fail) cmux-next docs > the CLI and MCP reference is fresh against the action catalog [3.02ms]
web / react-apps-base-check (3): (fail) long-chat scroll leaves the Left chip mounted and stable, while own cwd changes update it [21019.65ms]
web / react-apps-base-check (4): (fail) a pooled Settings page claimed by the host reports its paint once its first screen is in [326.16ms]
web / react-apps-base-check (2): (fail) agent pane page > rounds its controls as small-radius rects, not capsules [6.42ms]
web / react-apps-base-check (1): (fail) chromium > the gutter beside every part of the stack scrolls the transcript, and the part keeps the pointer [2971.75ms]

Not re-run automatically: web / Web tests (2/4), web / react-apps-base-check (3), web / react-apps-base-check (4), web / react-apps-base-check (2), web / react-apps-base-check (1) are not machine failures.

Written by scripts/ci/classify_failures.py (ci-failure-attribution.yml); signatures are its SIGNATURES table. A machine verdict is the runner's fault, not this PR's; yours means the failing file is one this PR changes, also red on main that main's latest full suite fails the same way, seen on other PRs that it failed on another pull request's run lately.

red: web / Web status; base red: the same failure on the base c803545; does not block
red: web / Web tests (2/4); no finished base run on c803545 to compare; do not merge until green; see lane-rules
red: web / react-apps-base-check (1); no finished base run on c803545 to compare; do not merge until green; see lane-rules
red: web / react-apps-base-check (2); no finished base run on c803545 to compare; do not merge until green; see lane-rules
red: web / react-apps-base-check (3); no finished base run on c803545 to compare; do not merge until green; see lane-rules
red: web / react-apps-base-check (4); no finished base run on c803545 to compare; do not merge until green; see lane-rules
red: web / react-apps-check; base red: the same failure on the base c803545; does not block

[teamleaderleo]

Owned by an agent session · session

[teamleaderleo]

Bead cx-vte03 created for this PR by the cmux coordinator.

[teamleaderleo]

cmux-next fast gates / post cmux-next/fast-gates is red on 8 of 12 fresh feat-cmux-next PR heads in the last 2:00:00, so it is most likely a base red, not this PR's. The base-reds lane has it; a base merge picks up the fix once it lands.

[teamleaderleo]

Review (Codex xhigh): SHIP at 1736adf

Independent session nx-ci-review-browser-deadlines (01a17981-c863-7580-aa4b-2b77f68b5779, gpt-6.1-sol xhigh through acpmux) inspected the integrated four-file diff and repaired deltas. Its initial P1 about mixed head/base policies is fixed. At the final head it reran 19 runner tests (25.533s), 20 wrapper/workflow tests (20.151s), retry shell test, syntax, diff, actionlint and prepush gate: all passed, worktree clean.

Primary verdict: SHIP at the same exact head. I independently inspected the final workflow/script diff, checked test failure evidence and the clean worktree. All selected files still execute; timeouts are named failures, budget-incomplete retries retain their first real named failures, and base replay uses identical runner/parser policy with base tests and Bun. No new job, step, trigger or credential privilege. #24438 overlaps only the lint-step file and was notified; its diff is disjoint.

No full browser or native execution is claimed by this review. The remaining stop condition is browser-shard completion on merged base and two webview PRs, plus #24469's separately owned claimPaint repair.

[teamleaderleo]

Exact-head non-green classification at 1736adf93024567e0c596b9df56360c48f30a1ab: both completed failures are infrastructure status writers, with no test/source failure.

• high-signal review link: PATCH .../issues/comments/... failed with API rate limit exceeded for installation (HTTP403).
• post cmux-next/fast-gates: posting the status failed with the same installation quota error. The root fast gates job in that run passed.

Risk evidence and merge freeze are green; remaining checks were pending at the bounded snapshot. No rerun/dispatch was started to work around quota. Independent xhigh SHIP and primary review are recorded above at this exact head.

[teamleaderleo]

high-signal review link / link is red on 6 of 10 fresh feat-cmux-next PR heads in the last 2:00:00, so it is most likely a base red, not this PR's. The base-reds lane has it; a base merge picks up the fix once it lands.

[teamleaderleo]

Review (subagent, correctness first): SHIP at 1736adf

Independent Codex xhigh review also returned SHIP at this exact head. The completed Web tests (2/4) red is a base/infrastructure finding (cmux-next docs > the CLI and MCP reference is fresh against the action catalog) unrelated to the four changed paths; its census could not run because DNS failed while downloading historical artifacts. Status-writer failures are separately documented as installation-quota infrastructure reds.

[github-actions]

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4044337729102 (b195c241a9)
• cmux NEXT nightly 1.0.0-nightly.4045248345901 (ab883bf956)
• cmux NEXT nightly 1.0.0-nightly.4045474536502 (a6e613294e)
• cmux NEXT nightly 1.0.0-nightly.4048947256901 (273534101b)
• cmux NEXT nightly 1.0.0-nightly.4049353932801 (e7cbb89fab)