content.js: types/messages.d.ts describes replies that do not exist, dead plumbing that still costs round trips, and handleCaptionsScrape answers early on a synchronous request failure #233

Severity: low

1. types/messages.d.ts: interface FSVideoSize { width; height } while content.js answers GET_VIDEO_SIZE with sendResponse(video ? video.size : 0) (a number, the visible area) and the background sorts by b.videoSize - a.videoSize. FSOpenPlayer omits attempt (content.js uses it) and its noRedirect comment ("True only for the top frame, which navigates rather than redirects") is backwards: the top frame never navigates, it overlays. The file's own header says an inaccurate type is worse than none. Fix: type FSVideoSizeReply = number, add attempt, reword.

2. Dead plumbing (content.js, main.mjs): Config.customVideoQuery/hasCustomPlaylist/hasCustomLinkHandler/customIframeId and the 'config' message: nothing under chrome/ sends SEND_TO_CONTENT or a config message. sendToOtherContents({type: 'active-state', ...}) goes background -> same frame with destination: 'custom', which no script handles: two wasted messages per player open and close. request.force and request.softReplace in the OPEN_PLAYER handler are never sent by openPlayer, so the "overlay without a video" branch and the per-request soft/hard choice are unreachable. httpRequest's POST and bust branches have no caller. main.mjs's 'key_down'/'seek_to' MESSAGE_FROM_PAGE handlers: nothing sends them. Fix: delete, or wire up the custom-script config path if the site scripts are meant to use it.

3. handleCaptionsScrape answers early when a request fails synchronously: if (done === pending) sendResponse(tracks) inside the callback, with pending++ per track in the same loop, and httpRequest's catch (e) { callback(e); } runs synchronously. Two <track> elements where xhr.open throws for the first (a malformed src): done becomes 1 while pending is still 1, sendResponse([]) goes out before the second track is even requested, and the player opens with no captions although one was loadable. Fix: count pending for all tracks before issuing any request, or defer the sync failure with queueMicrotask.

4. Overlay guard and windowed fullscreen ignore shadow trees (overlay-guard.js and fillScreenIframe use document.querySelectorAll('*') while the rest of content.js walks shadow roots): a web-component player keeps its control bar inside the host's shadow root; elementsFromPoint retargets the iframe to the host, host.contains(iframe) is false across the boundary, nothing is examined, so the site's bar stays over the player; in windowed fullscreen the host is expanded but its shadow siblings are never hidden. Fix: walk openOrClosedShadowRoot (already used in content.js) in both scans and use getRootNode().host for containment.

Confidence: high (1, 2, 3), medium (4).

From a line-by-line code review of the tree at 41363fd (2026-10-02). Each claim above was checked against the source before filing; the reproduction steps and line references are as of that commit.

Generated by Claude Code