Commit Graph

5 Commits

Author SHA1 Message Date
khannurien
fac0e2db28 v4: an audio file's waveform is made once by the server, and the player seeks on one accessible scrubber — every player downloaded the whole file a second time and decoded all of it in the browser to draw 150 bars, so a long FLAC cost tens of megabytes of memory per page, and the waveform could only be clicked: no drag, no keyboard, no slider for a screen reader. GET /api/dumps/:id/waveform now answers an audio dump's peaks, which lib/ffmpeg.ts decodes with ffmpeg on first request, mono at 8 kHz and streamed, keeps beside the thumbnails as the response itself, and serves with an ETag (a ten-minute MP3 takes about a second, then a few hundred bytes). The peaks are taken by shared/waveform.ts's peaksOf, scaled to the loudest so a quiet recording fills the track, and the dump form's preview of a file not yet uploaded decodes it with the same function, at 8 kHz in an OfflineAudioContext. Scrubber replaces Waveform and SeekBar: a role="slider" that seeks where a pointer lets go, follows a drag (the clock shows where it is held), takes arrows, Home and End, and draws the bars as one path with the played part clipped by a nested svg, so it moves smoothly and a timeupdate redraws two attributes instead of 150 rects; audio without a waveform, and video, get a plain bar. MediaPlayer reads whether it plays from the element's events alone, so a click no longer reports the change twice, its handles for PlayerProvider are stable callbacks instead of refs to refs, the drag no longer reattaches every listener, and the <source> children that src made dead are gone, with the mime prop. The file, thumbnail and waveform routes resolve their dump, token and caching through lib/dump-media.ts, and what is generated from an upload is named and removed in lib/ffmpeg.ts, the custom thumbnail by dump-service's customThumbnailPath, instead of paths spelled out in each. The test setup's play() and pause() fire their events as a browser does. 277 backend tests and 684 frontend tests pass, and deno check, tsc, eslint, doc:lint and i18n:compile --strict are clean; checked in the app with a real FLAC on its dump page and in the global player at 1280 and 390px, seeking by keyboard.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 13:34:22 +00:00
khannurien
7960d74c55 v4: weekly digests and monthly recaps go out by email, every email is written once as Markdown in the app's one catalog, and a review of the last two commits removes what they stated twice — users get a weekly digest (what happened to them, what the people and playlists they follow posted, the week's top dumps, the community) and a monthly recap (the month's top dumps and the community), both on by default, in their own style and scheme. digest-service sends what is due every quarter of an hour from digest_schedule (migration 0020), which holds each user's rolling 7- or 30-day period; existing users' first periods are spread over the coming week or month, a digest more than a period late covers the last full one, an empty one isn't sent, and an address like admin@localhost is skipped. Each digest carries a one-click unsubscribe (List-Unsubscribe and List-Unsubscribe-Post, POST /api/users/unsubscribe) and a link to the new /unsubscribe page, which asks before turning that cadence off, since mail scanners open every link; the token (lib/jwt) names the user and the cadence and never expires. The welcome email now says the digests are on. The emails' words moved out of a hand-written api/mail/strings.ts into the Lingui catalogs the interface uses: lingui extracts api/mail/ too, the catalogs moved to a top-level locales/ that both halves read, compile emits ES modules the server imports (api/mail/i18n.ts), messages take ICU plurals, and the labels the settings tab, the unsubscribe page and the emails share are one message each (their French had already drifted apart). Each template writes its email once as Markdown, which is the plain-text part, and EmailKit.render draws its blocks in the style (a lone link is a button, a list item starting with an image a thumbnail row, a numbered list a ranking, --- the footer), so no string is declared for an HTML and a text version; user text is escaped for Markdown, which the welcome body didn't do for usernames, whose _ read as emphasis. paint() set colors before a border-top shorthand that reset them, so every divider was black. The review also gave one home to what was written twice: the locale list (shared/locale.ts, read by lingui.config, the client, the server and its catalog loader), the base style and default scheme (shared/theme.ts), an email's look (lookOf, which digest-service, welcome and reset share), whether email is on (EMAIL_ENABLED, read by email-service and index.html), a setting's column and its default in the digests' SQL (SETTINGS_COLUMNS, DEFAULT_SETTINGS), the unsubscribe token (made once, not parsed back from a URL), AuthProvider's compare-and-store, ThemeProvider's twin style and scheme code (useThemeChoice), the test JWT encoder, and duplicate digest messages. The email palettes stay a flattened copy of the themes' tokens, now under a test that caught Times' secondary and link colors drifting. The Dockerfile copies api/mail and locales so the build extracts the emails' messages and the server finds its catalogs, and deno.json maps @lingui/core for the runtime image. 236 backend tests (new: digest building, sending, catch-up, scheduling and unsubscribing; the Markdown renderer; French from the catalog; palettes against themes; the stagger migration; unsubscribe tokens) and 661 frontend tests (new: the unsubscribe page) pass, and tsc, eslint, doc:lint, deno check, deno lint and deno fmt are clean.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-04 22:14:46 +00:00
khannurien
364c498a24 v3: dependencies upgraded and deno audit clean, every API response typed end to end, no explicit any left — and the site search stops handing out password hashes — deno audit reported 19 advisories (11 high); none remain. nodemailer 9 → 10.0.10 closes four of them (a quadratic address parser, two recipient-domain allow-list bypasses, resolveContent ignoring disableFileAccess/disableUrlAccess); 10's one break is Node ≥ 20, and 10.0.11 is a day too young for the minimum release age. react-router 8.0.1 → 8.4.0 closes the RSC-mode CSRF bypass. Refreshing the lockfile clears the transitive ones that came in through Vite, ESLint and Lingui: postcss, nanoid, picomatch, browserslist, brace-expansion and baseline-browser-mapping. Everything else moves to its latest too — Lingui 6.8 (its SWC plugin 6.7), React 19.3, Vite 8.3.1, react-hook-form 7.89, marked, jose, ESLint 10.11, typescript-eslint 8.70, the @types — and frimousse 0.3 → 0.4, whose declarations only add (custom emoji data resolvers, a cache). TypeScript stays on 6.0: typescript-eslint supports < 6.1, so 7.0 would stop the lint. allowScripts names @swc/core 1.16.2 and drops esbuild, which Vite 8 no longer installs.
All checks were successful
Build and Publish Docker Image / build-and-push (push) Successful in 51s
Search: GET /api/search sent each matching user's full row — password hash and email included — to anyone, signed in or not, because searchUsersGlobal returned User[] and the route passed it through. It returns profiles only now, stripped where the rows are read, and the search test fails on the old code. Types alone wouldn't have caught it (a User is structurally a PublicUser), so the strip is explicit and tested.

Typed routes. Every JSON response is declared as APIResponse<T>, the way the typed routes already did, with the missing shapes named in api/model/interfaces.ts: PublicUser (no hash, no email) and SelfUser (your own profile, with email) plus toPublicUser/toSelfUser in place of ten hand-written destructurings, AuthTokens, AuthResponse, MediaTokenResponse, InviteResponse, AttachmentResponse, SearchResults, LinkPreview, UserSuggestion and Invitee. Request bodies are `unknown` and narrowed by the existing guards, or by a new fieldOf(obj, key) where a route reads one field — no more reading off Oak's `any` or casting to Record<string, unknown>. The password-reset routes answer `{ success: true, data: null }` rather than a bare `{ success: true }`, so every success fits the envelope; the frontend only reads `success`.

Socket protocol: routes/ws.ts had its own send() taking Record<string, unknown>, so nothing checked what it sent. It takes ServerToClientMessage now, which turned up two mismatches: `pong` was missing from the union although the server answers every client ping with one, and vote_ack carried a `success: true` no type declares and no client reads. The union has PongMessage and vote_ack lost the field.

No explicit any. The YouTube playlist parse reads ytInitialData as `unknown` through a small at(node, …path) that checks each step, instead of `type Json = any`. The backend test helpers type the response body as the server's envelope: request<T>() and get/post/put/patch/del<T>() name what `data` holds, res.data throws with the server's message when the call didn't succeed and res.error when it didn't fail, and each test says which API type it expects, so a test reading a field the route doesn't send no longer compiles. Assertions that a field is absent are `assertFalse("passwordHash" in …)` rather than reading a field the type says isn't there. The socket tests' messages are ServerToClientMessage, with next()/all() keyed by type so each returns that message's shape.

No double casts in src/ either. window.YT and onYouTubeIframeAPIReady are declared on Window; the test's fake player is a class forwarding to the spies the tests inspect. WSContext, FollowContext and PlayerContext export their defaults (defaultWSValue, defaultFollowValue, defaultPlayerValue), which the tests import instead of reading React's private _currentValue. handleMentionKeyDown takes a MentionKey — the key fields it reads — so the test builds a plain object; useUserDumpFeed declares sentinelRef as the callback ref it is. In src/test/setup.ts the IntersectionObserver fake implements IntersectionObserver, and report(isIntersecting, top?) builds complete entries in place of partial ones cast in five files; the Animation fake implements Animation; the canvas getContext stub goes in through defineProperty. The XHR fake's last request is optional behind a getter that fails if none was made, the keyframes are checked with matchers, and the roleless legacy session is built by leaving the role out. One test changes meaning slightly: "leaving the view…" relied on entries with no position at all, and now leaves and re-enters at the same height, which is what it describes. The one cast left is in api/test/ws.test.ts: deno.json's "dom" lib hides Deno's WebSocket overload that takes handshake headers, and dropping that lib would break the editor's view of src/; it's commented and names the real signature.

Also: the notifications test answered a Response-typed promise with a plain object, which failed `tsc -b`; it uses the fetch stub's deferred() now. ESLint didn't read the `deno-lint-ignore no-explicit-any` comments, so `eslint .` failed on three files; those anys are gone. `deno fmt` over api/ and src/ (not the compiled src/locales/*.js); the reflow left all 332 message ids intact, and the .po files only have their source line references updated.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-27 16:05:29 +00:00
khannurien
c8bf0d583c v3: the tests learn to expect trouble, and seven bugs they now catch are fixed — a review of the two test-suite commits found them written mostly for the happy path; each fix below comes with the test that fails without it. Infinite scroll: a later page that failed turned loadingMore back off, which made a new IntersectionObserver, which reports a sentinel still in view at once — so an offline feed asked again as fast as the request could fail, on Search, Index, the followed feeds, user dumps and user playlists alike. useInfiniteScroll now waits 1s, 2s, 4s… up to 30s while the sentinel stays put, counted from the last load so re-renders don't restart it, and asks at once again when the sentinel leaves the view or is pushed down by new items; the fake observer in src/test can report an element in view on observe(), as a browser's does. Playlist covers: the row was updated and playlist_updated sent before the file was written, so a client fetching image?v=<new> in between cached the old cover (or a 404) under the new URL for an hour, and a failed write left image_mime set with no file behind it. The file now goes in first, written beside and renamed over, and only then does the row change. Playlist order: the server wanted every member named, but GET hides private dumps the requester can't see, so a playlist holding someone else's dump that went private — or one a moderator reorders around the owner's private dumps — could never be reordered. The requester now orders what they can see and hidden members keep their slots; the visibility rule lives in api/lib/playlist-members.ts, shared by GET, reorder and the socket. playlist_dumps_updated also sent every member's id, private ones included, to a public playlist's whole audience; each viewer now gets the ids they can see, plus a dumpCount of all members so playlist cards keep the server's count. File dumps: replacing the file and changing visibility took two requests, so a failed second one left the new file live under the old visibility; PUT /api/dumps/:id/file takes isPrivate and applies both in one UPDATE, and it now broadcasts like any edit (it sent nothing, so other viewers never saw a new title). Notifications: one that arrived while the list was loading was dropped, then marked read by the read-all after the load, so it never showed; it's kept and merged in, once. Ranges: bytes=0-0,5-6 got a 416; several ranges, another unit or an unparseable header are ignored now and the whole file sent with a 200, as RFC 9110 allows — the old test asserted the 416.
All checks were successful
Build and Publish Docker Image / build-and-push (push) Successful in 49s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-27 15:13:57 +00:00
khannurien
54e86c4366 v3: the frontend gets a test suite, and the bugs it turned up are fixed — deno task test:web runs 612 tests over src/ in about 11s: Vitest on jsdom with Testing Library, built on the app's own Vite config so the SWC/Lingui transform and the .po loader are what ship. src/test/ holds the harness: a routing fetch stub (queued once-responses, fallback defaults), a FakeWebSocket the tests open, feed and drop, jsdom polyfills, fixtures, and two render helpers — one over hand-set context values, one over the real provider tree (auth, socket, player, follows, categories, chat). It covers every util and hook, every provider (the socket protocol end to end: votes and likes with ack/timeout/error rollback, chat sends/echoes/retries/read markers/reconcile, reconnect backfill), the components (header and search history, cards, Markdown, thumbnails, media and global player, the text editor's mentions/emoji/uploads, comments, the new-dump modal, the chat), and every page.
All checks were successful
Build and Publish Docker Image / build-and-push (push) Successful in 48s
Fixed along the way. Dumps: clearing a dump's note sent `comment: undefined`, which JSON drops, so the server kept the old one; it's sent as null now. Replacing a file dropped a visibility change made in the same save (the file endpoint doesn't read it); it follows as a separate update. Renaming a dump onto a new slug kept showing the old title until reload. The thumbnail picker and delete on the edit page failed without a word, the latter as an unhandled rejection. Playlists: an edit that kept the slug waited on the socket's echo to show, and a new cover never appeared (the upload's answer was ignored, a refused one said nothing); covers were also served stale for an hour, since `max-age=3600` met an unversioned URL — setPlaylistImage now bumps updated_at and every cover URL carries `?v=` (playlistImageUrl). Removing a dump, reordering, and deleting a playlist ignored non-OK answers, leaving page and server apart; they restore, reload or report. Profile: clearing "Who am I?" sent undefined and came back on reload; your own upvoted list (and the upvoted page) was empty until the socket's welcome, and forever without one; "View all" under Following linked to a route that doesn't exist and is a Load more now. Chat: Escape on "Delete this message?" closed the whole chat — ConfirmModal takes Escape first, in the capture phase. Player: a pause or play from a media key, headset or the OS wasn't reflected (MediaPlayer follows the element's own play/pause events), and the seek bar tracked dragging with mouse events only, so on touch the thumb fought timeupdate. Feeds: a guest on /~/followed got a blank page; a server error in the followed feed read as "Could not connect". Search: a failed later page threw away the results shown. Notifications: a failed mark-read replaced a loaded list with an error. Silent failures now say so: deleting a comment, the voters/likers "Load more" (stuck on "Loading…"), and the password-reset request, which said "sent" whatever happened.

Notifications also stop hiding what was new: the bell still clears on opening the page, but the dots and the "N new" count stay for the visit instead of vanishing the moment the mark-read lands, and one arriving while the page is open is marked read too, so the bell doesn't count it.

Accessibility: Modal and ConfirmModal are dialogs (role, aria-modal, labelled/described); playlist membership rows are checkboxes that work from the keyboard; a dump's title, and the profile's email and description, can be opened for editing from the keyboard (the pencil shows on focus); the category editor's ✓ and the global player's video button have names. Player, rich-content and download labels were hard-coded English and are translated, with the new French strings. Lint: fast-refresh and render-purity rules are off for src/test/ only, where probes hand their context out on purpose.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-27 14:35:41 +00:00