[teamleaderleo]

Ready: an independent Codex review says SHIP at the head, and the fleet tests pass at the head and fail at the test-only commit (see the receipt comment).

Leo-ask: ask-76f83a7020

"this watch is kind aass also id like to be able to watch .mp4s direclty in cmux next somehow (chromium?)"

What

This PR handles the ".mp4 in cmux" half of the ask. The link to the What's New reel (Watch what changed in nightly <sha> ↗) points at raw.githubusercontent.com/.../reel.mp4. GitHub sends every .mp4 there as application/octet-stream with nosniff, so a cmux browser tab downloaded the file instead of playing it. Any other web .mp4 served the same way did too.

Now a WebKit tab plays such a video at the video's own address, in a plain <video controls autoplay> page (black background, fitted to the tab), when a main-frame response meets all of these:

• WebKit cannot show its type;
• it is not Content-Disposition: attachment;
• its http(s) path ends in mp4, m4v, mov, webm or ogv;
• the navigation was a GET and not Back or Forward.

The page loads with loadSimulatedRequest, so it is a normal back-list entry: Back returns to the page the video was opened from.

These cases still download, as before:

• attachments;
• every other file type;
• a POST, because the player could not fetch the video again;
• a Back or Forward to an entry WebKit had evicted, because that must not add a new back-list entry.

The link's look ("this watch is kind aass") is the What's New page, which the whats-new lane owns. I've sent them that half of the ask, with the screenshot.

Known limitation

A page that calls location.replace() on such a video gets a new back-list entry instead of replacing its own, because WebKit doesn't tell the app that a navigation is a replace. This is rare, and the player still plays.

Chromium tabs don't take this path. CEF downloads octet-stream responses in the shim, and changing that means a shim rebuild. This PR keeps to the WebKit change.

Tests

Tests/CmuxNextBrowserTests/RemoteVideoPageTests.swift runs a real WebKit tab against a local server. The server sends a generated H.264 .mp4 the way GitHub does: application/octet-stream, nosniff, with byte ranges. The test:

• goes from page a to page b, then follows b's link to the video;
• waits for the video's currentTime to pass 0;
• checks there was no download and that the back list is exactly a, b, reel.mp4 after Back and Forward;
• checks that a form POST for the same video still downloads.

Observed / Not run

• Red first: on the fleet, the test-only commit cf8e824 fails. The tab downloads the video, so it never plays.
• Green: on the fleet, RemoteVideoPageTests passes at the head (5ff010c).
• The prepush gate passed with no new violations.
• Not run: a click on the real What's New link in a built app.

Changelog

• Fixed: a web .mp4 that its server sends as a generic file, such as the nightly What's New reel, plays in a cmux browser tab instead of downloading.

No capture: the player is the system <video> control on black. The test proves playback.

Agent session: cosmic-gray

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

---

Summary by cubic

Plays web videos whose servers send them as generic files in a cmux WebKit tab instead of downloading them — the nightly What's New reel among them, since raw.githubusercontent.com serves every .mp4 as application/octet-stream with nosniff. The video now opens in a plain player page (system controls, black background) at its own address, and takes a normal spot in the back list: Back returns to the page the video was opened from.

Only a main-frame GET for an http(s) URL ending in mp4, m4v, mov, webm, or ogv becomes the player. The tab records the video URL when its newest own main-frame navigation asks for it, and only a response for that URL plays — a target=_blank click, an older held GET, or an in-flight reload cannot turn a POST's response into it. Attachments, other file types, POSTs, and Back or Forward navigations still download as before, as do Chromium tabs.

Known limitation

• A page calling location.replace() on such a video adds a new back-list entry instead of replacing its own; the player still plays.

Tests

• RemoteVideoPageTests plays a generated H.264 .mp4 served like GitHub does (octet-stream, nosniff, byte ranges) and checks playback, the back-list order, that a form POST still downloads, and that only the matching response becomes the player.

^{Written for commit 5ff010c. Summary will update on new commits.}

[teamleaderleo]

Owned by browser-import-2 on cosmic-gray · session

[github-actions]

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

Review 5ff010c: 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=pr24338 -f ref=browser-import-2/remote-video-plays -f platform=macos -f backend_mode=local

red: cmux-next checks (god files, concurrency, crash safety, l10n); NEW: red check: Checks job structure (+1 more) is not red on the base 2d13b66; do not merge until green; see lane-rules

[blacksmith-sh]

Found 9 test failures on Blacksmith runners:

Failures

Test	View Logs
/tmp/tmpig4s6our/summary-missing.json (live):/summary-missing.json (live):	View Logs
/tmp/tmpr5bbl1x6/summary-fail.json (live):/summary-fail.json (live):	View Logs
[new_workspaces] debug.new_workspace timed 0 of 50 new workspace rows/ [new_workspaces] debug.new_workspace timed 0 of 50 new workspace rows	View Logs
[new_workspaces] the window draws no frames/[new_workspaces] the window draws no frames	View Logs
[tabs] debug.tab_switch timed 0 of 40 switches/ [tabs] debug.tab_switch timed 0 of 40 switches	View Logs
[tabs] the window draws no frames/[tabs] the window draws no frames	View Logs
test_every_part_of_bun_run_check_runs_in_the_warn_branch (main.WebviewsWarnSplitCov ersCheck.test_every_part_of_bun_run_check_runs_in_the_warn_branch)/ test_every_part_of_bun_run_check_runs_in_the_warn_branch (main.WebviewsWarnSplitCov ersCheck.test_every_part_of_bun_run_check_runs_in_the_warn_branch)	View Logs
test_every_submodule_free_checkout_on_an_owned_runner_resets_submodules (main.Reuse dWorkspaceSubmodules.test_every_submodule_free_checkout_on_an_owned_runner_resets_submo dules)/ test_every_submodule_free_checkout_on_an_owned_runner_resets_submodules (main.Reuse dWorkspaceSubmodules.test_every_submodule_free_checkout_on_an_owned_runner_resets_submo dules)	View Logs
test_every_submodule_free_checkout_on_an_owned_runner_resets_submodules (main.Reuse dWorkspaceSubmodules.test_every_submodule_free_checkout_on_an_owned_runner_resets_submo dules) (job='web-bundles')/ test_every_submodule_free_checkout_on_an_owned_runner_resets_submodules (main.Reuse dWorkspaceSubmodules.test_every_submodule_free_checkout_on_an_owned_runner_resets_submo dules) (job='web-bundles')	View Logs

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[teamleaderleo]

Bead cx-945kp created for this PR by the cmux coordinator.

[cursor]

Bugbot needs on-demand usage enabled

Bugbot uses usage-based billing for this team and requires on-demand usage to be enabled.

A team admin can enable on-demand usage in the Cursor dashboard.

[teamleaderleo]

Review (independent subagent): SHIP at 5ff010c

Reviewer: Codex gpt-6.1-sol on cmux16 (a different model from this lane), adversarial, fifth round. Rounds 1-4 blocked and each finding was fixed in a following commit: hidden history entry, POST responses, new-window actions, a late certificate-admission callback.

Verdict

SHIP at 5ff010c

No blocking findings in manaflow-ai/cmux, browser-import-2/remote-video-plays against feat-cmux-next.

• Non-blocking limitation — Packages/macOS/CmuxNext/Sources/CmuxNextBrowser/WebKit/WebKitTab+Delegates.swift:60: A GET link to /reel.mp4 that redirects to an extensionless /media?id=123 clears the candidate at the redirect and downloads. Redirects ending in a supported video extension remain eligible. This preserves base behavior for the extensionless destination; it does not break the direct raw.githubusercontent.com reel case.
• Non-blocking test coverage — Packages/macOS/CmuxNext/Tests/CmuxNextBrowserTests/RemoteVideoPageTests.swift:35: The integration test covers the direct octet-stream/nosniff GET, advancing playback, visible URL, Back/Forward history, and POST download. It does not exercise a redirect to another .mp4 or an attachment response; a regression in either scenario could escape this test. The current policy retains the attachment download branch and reevaluates redirected requests.

The main-frame, method, traversal, and response-URL guards do not reveal a remaining wrong-response conversion. Explicit downloads and synchronous cancellation paths precede player eligibility; organization filtering and certificate admission still precede the original load. The generated title and double-quoted source escape HTML delimiters, and the source is restricted to HTTP(S). Bypassing automatic-download admission here renders media rather than creating a download.

Reviewed the complete patch and complete changed files fetched with curl at the exact head above. Static review only: no checkout, clone, build, tests, or gh operations; no runtime or CI success claimed.

Tests (fleet, cmux-ci)

• Head 5ff010c: RemoteVideoPageTests passed, 2 tests (the video plays, the back list is exactly a, b, reel.mp4 through Back and Forward, a POST still downloads).
• Test-only commit cf8e824 (red first): failed. The video never played, because the tab downloaded it.

Reds at this head

• generated files: test compile error in PaneMotionBench.swift, a base file. cmux NEXT tests: PaneMotionBench's argument list is nonisolated, so CmuxNextAppTests build again #24377 has since fixed it on base.
• checks: "Checks job structure" and "No fleet host names in gallery fixtures", both base reds that are also red on cmux-next tests: the Extensions menu suite scales its budgets by the runner's reference workload #24157 and cmux NEXT: a window closed during the launch restore stays closed #23695.
• fast tier: cancelled at its 25-minute limit on a cmux-austin-mini runner, the same hang as on cmux NEXT: a window closed during the launch restore stays closed #23695.

[github-actions]

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.3991367032401 (b9c3d5aabc)
• cmux NEXT nightly 1.0.0-nightly.3991575404801 (a62351daaa)

[github-actions]

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.3991503232101 (cd84bc62d1)