214 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
khannurien	4ea192e0b1	v4: an uploaded audio file plays from its cover everywhere, and a card's play badge sits in the middle of its thumbnail — a link played from its thumbnail in every view, but an uploaded file only did on a list card: a journal tile opened the dump, an audio file's page showed a round button by its waveform and never its uploaded cover, and an uploaded video's page was a black box until played. The list card's badge sat in the thumbnail's bottom-left corner, where it looked misplaced, and clicking a playing file restarted it. The new usePlayFile hook builds the player's item and the file's artwork (fileArtworkUrl) for the card, the journal tile and the dump's page, and plays, or pauses once the file is the current one. A card's badge, ▶ or ⏸, is now centred, as is the ↗ on a link out; over a placeholder it takes the place of the glyph. An audio file without a cover shows a placeholder tinted from its file name, not a bare 🎵. In the journal, audio counts as having art (hasThumbnail), so it is an image tile, its glyph raised above the title, and a file tile plays and pauses the file. On the dump's page, the cover (96px, the card's badge on it) plays the file beside the shared Waveform, which replaces FilePreview's copy of it; a video takes the server's still as its poster. 674 frontend tests pass, and tsc, eslint, deno fmt, doc:lint and i18n:compile --strict are clean; checked in screenshots in Smooth light at 1280 and 390px, Brutalist dark on the journal at 390px and Kanagawa dark at 390px, playing. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 22:33:33 +00:00
khannurien	2c031b0102	v4: the header drops the username link — on wide screens it stood beside the avatar, which leads to the same profile, so only Playlists stays there. The avatar's menu still opens with @username first. 670 frontend tests pass, and tsc, eslint, deno fmt and i18n:compile --strict are clean; checked in a screenshot in Smooth dark at 1400px. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 22:23:46 +00:00
khannurien	886216749e	v4: the avatar menu shows at every width and holds Log out, so a profile keeps only Invite; the profile's dump cards lay out again — your own profile had four buttons in a row, Invite someone, Settings, Manage and Log out, Invite taller than the rest, and on wide screens it was the only way to reach /settings or /manage. The avatar menu now leads the header's icon row at every width, after the name and Playlists text links on wide screens, and ends with Log out under a separator, muted until hovered. The profile keeps Invite someone alone. The menu's button outweighs .app-header-nav button and the themes' button rules (Brutalist and GeoCities list it among their flat buttons), so it reads as one of the menu's links. Since b9ff9c4, a card under 472px is laid out by the narrow-card container query, but the profile columns' dense-card rules still placed the thumb, body and vote themselves, more specifically: the thumb lost its row span, so the title fell under it and every card was a tall, mostly empty box. Those rules now only make the card denser (a smaller thumb, a 2-line title, smaller type) and leave placement to the container query. 670 frontend tests pass, and tsc, eslint, deno fmt, doc:lint and i18n:compile --strict are clean; checked in screenshots in Smooth dark at 1400 and 390px, Brutalist light at 390px, GeoCities light and Kanagawa dark at 1400px, in French. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 22:22:15 +00:00
khannurien	10c9cf4269	v4: a playlist's cover is a banner over its header, and the edit form picks it in the same shape — the cover was a 72px square beside the title, which left the title and description a narrow column under a mostly empty band: on a phone, in Kanagawa's wide mono, a long title broke into six lines. The cover now spans the header's width above the title, cropped to 3:1 and no taller than 14rem, inside the header's padding so outlined themes keep their ring around it; the title, buttons, description and meta take the whole width under it. The edit form's Cover picker takes the same shape (.playlist-cover-banner, shared with .playlist-detail-img), and with no cover is a dashed 3:1 placeholder. ImagePicker's size is now optional, so a class can size it, and its radius takes any CSS length, here --radius-lg; a picker sized by aspect-ratio sets min-height: 0, or its image would hold it at its own height. 669 frontend tests pass, and tsc, eslint, deno fmt and doc:lint are clean; checked in screenshots in Kanagawa dark (French, 385px), Brutalist at 390px with and without a cover, and Smooth at 1100px, the page and the form. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 22:10:06 +00:00
khannurien	5acdf79c8d	v4: the playlist and dump edit forms end in one bar of actions, a panel of its own — the playlist editor drew rules between its fields and its actions, and on a phone its Delete, Cancel and Save wrapped into a staircase, then stacked (a0c8b1d), neither of which held together. Both forms are now built as DumpEdit was: surface panels 3px apart (.form-panel, from .dump-edit-panel), the last a bar (.form-panel--bar, from .dump-edit-footer) holding Delete on the left and Cancel and Save on the right, compact (btn-sm), on one line. The new DeleteButton says "Delete" and names what goes to assistive tech ("Delete playlist", "Delete dump"); in a bar too narrow for the three labels in every face and language (a container query, under 18rem), it shows 🗑 instead. While edited, the playlist's header drops its own card for the form's panels. The phone stacking rule is gone. An auto-sized editor also re-fits when its width changes with its text unchanged: opening the form makes the page tall enough for a scrollbar, which narrowed the description after it was measured, so it overflowed; and it hides its overflow below its max-height, where a scrollbar could only come of rounding, and drops the resize grip, which the next keystroke would undo. 669 frontend tests pass, and tsc, eslint, deno fmt, doc:lint and i18n:compile --strict are clean; checked in screenshots in Kanagawa dark (French, 385 and 360px), Smooth light at 1000px and DumpEdit in Smooth at 390px. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 22:05:43 +00:00
khannurien	b9281da9c9	v4: the avatar menu opens from the avatar's left edge — since 7531be3 the avatar leads the header's icon row, but its menu still hung from its right edge, so on a phone it ran off the left of the screen. It now hangs from the left edge and opens over the icons. Checked in screenshots in Smooth at 390px and Brutalist at 320px, in French, and Kanagawa at 1100px. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 21:58:15 +00:00
khannurien	7531be300e	v4: the header's controls are line icons, and the avatar leads them — 🔍, ✕, 💬 and 🔔 were four emoji in four palettes beside the avatar and the "+", none of them the theme's. They are now stroked icons (NavIcons.tsx, which the search bar's submit and ChatFab share) in the text colour, their bodies washed with the accent; the wash deepens on hover, and fills while something is unread, with a small dot in place of the red count (the count stays in the button's label and the tab's title, and ChatFab keeps its own). Themes set the line through --nav-icon-stroke, -linecap, -linejoin and -tint-strength: Brutalist heavy and square-cut, Times hairline in ink (its grayscale filter no longer needed on them), GeoCities fat and soaked. The search toggle was a ghost button with its own hover; it now shares the chat button's and the bell's rule, background and borders in every theme. The avatar sat among the icons, its hover box larger than theirs and 1.5px higher (a block .user-menu left a line-box gap under the inline-flex trigger): it now leads the row, where the name and Playlists links stand on wide screens, so the icons run unbroken to the "+"; it is 32px, and hover rings it with the accent instead of a box. 669 frontend tests pass, and tsc, eslint and deno fmt are clean; checked in screenshots in all six styles, light and dark, at 390px, at rest and unread, and Smooth and Brutalist at 1280px. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 21:56:19 +00:00
khannurien	a0c8b1d509	v4: auto-sized text editors no longer scroll, and on a phone an action row with a destructive button stacks — an autoResize editor set its height to scrollHeight, which leaves out the borders that a border-box height counts, so the text always overflowed by a couple of pixels: a playlist's description, a .form textarea since cf89c18, showed a scrollbar that ran under the emoji button. The height now adds the borders. On a phone, Delete, Cancel and Save didn't fit one line in French or in a wide face, and wrapped into a staircase: Delete alone on the left, Cancel and Save under it on the right. A .form-actions row with a .btn-danger now stacks under 512px, with Cancel and Save across the top (Save taking the room left) and the destructive button full width under them; the playlist editor and DumpEdit's footer both take it. 669 frontend tests pass, and tsc and deno fmt are clean; checked in screenshots in Kanagawa (French, 385px, with a long description) and DumpEdit's footer in Smooth at 390px. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 21:51:38 +00:00
khannurien	cf89c18eb9	v4: a playlist's edit form is a .form like the dump's, and its dumps keep the header's edges — the editor laid the cover beside its fields with display: contents, which on a phone left them a squeezed column, and drew its own inputs: the title and the slug both in bold large type, the description in another size, only the slug labelled, and a segmented control styled apart. It now stacks the shared fields, each under its label (Cover, Title with its counter, Link, Description, the visibility toggle), over the action row, as DumpEdit does, and takes the header's whole card. The dumps' drag handle and remove button took width from the cards, which sat inset from the header on both sides; they now sit in the page's side padding (--page-gutter, which .page-content now names), so the cards line up with the header at any width. Undo shows as a small btn-primary over the card it would bring back, instead of a third control beside it. CountedInput, VisibilitySegmented and the .playlist-edit-* rules had no other users and are gone. 669 frontend tests pass, and tsc, eslint, deno fmt, doc:lint and i18n:compile --strict are clean; checked in screenshots in all six styles at 390px, Kanagawa and Smooth at desktop widths too, Undo in Smooth and GeoCities. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 21:48:12 +00:00
khannurien	45f39a959b	v4: /settings and /manage take after /notifications — the three pages the user menu and the bell lead to share one frame: a 680px column (PageShell's new narrow), a title with its icon (PageHeader: ⚙️, 🛠️, 🔔), then groups under a small capitalized label ruled to the column's end (PageGroup, the notifications' day label), their content on surface cards (.page-card, the notification items' surface). Settings drew big section headings over controls laid on the bare page, the full width wide, and Manage the same around the category manager; each section is now a card whose setting rows are ruled apart and keep their labels lined up (a subgrid of the card's columns). The notifications' column, centered by auto margins in a flex column, shrank to its content (380px with short ones); it fills its width now. Its own header, group and bell rules gave way to the shared ones, and Times, Brutalist and GeoCities dress .page-card as they dress the notification items. In the narrower column, Kanagawa's monospace no longer fit the six styles on one row, and French labels leave even less room: a segmented control of more than four options now makes two even rows at any width, as it did on phones. 670 frontend tests pass, and tsc, eslint, deno lint, deno fmt and doc:lint are clean; checked in screenshots in all six styles at 1280 and 390px, Smooth in light too, with the account's style and the notifications mocked so the shared test account was left as it was. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 21:40:10 +00:00
khannurien	1fda26a399	v4: /settings, /manage and /notifications title themselves alike, with a bare heading above their sections — Settings and Manage borrowed the edit form's .form-page-header, a panel rounded only on top to cap the panels under it, so on a page without them it ended in a square edge, and its title was the size of the section headings below. Notifications had a heading of its own. All three now take .page-header and .page-title: a bare h1 a size above the sections' h2, with what goes with it (the unread count) at the row's end; the edit forms keep their panel header. Times and Melange set their display type on .page-title, and the .profile-subpage-* rules, unused since the profile's pages became its tabs, are gone. (The CSS went in with d683ac0, which a concurrent session committed from the shared tree.) 21 Settings, Manage and Notifications tests pass, and deno fmt is clean; checked in screenshots in Smooth at 1280 and 390px. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 21:28:28 +00:00
khannurien	45e004b465	v4: the slug and rename fields lose their hints about old links — "Links to the old address keep working." under the dump and playlist link field, and "Their old name keeps leading to them…" under the moderator's rename field, described a server guarantee nobody filling the form needs to read. Only the fields' errors show under them now; UsernameField loses its hint prop. 670 frontend tests pass, tsc and i18n:compile --strict are clean. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 21:28:01 +00:00
khannurien	437eea0df3	v4: every inline editor ends in a btn-primary Save and a btn-secondary Cancel — comments drew their own .comment-submit-btn and .comment-cancel-btn, the profile's email its own .profile-email-btn pair, the bio a plain-text form-cancel, and the chat's edit box shrank the shared buttons with a rule of its own; they were four sizes, two orders of chrome, and the comment button's white label sat on Smooth's yellow accent. They all take the shared classes now, and the new .btn-sm is the one compact size, for the comment forms, the chat edit box and the email editor; the bio and the dump title stay full size. .btn-secondary gets a disabled state, as btn-primary has. The bespoke rules are gone, with GeoCities' bevel for the comment buttons, which btn-primary and btn-secondary already have. 670 frontend tests pass, and tsc, eslint and deno fmt are clean; the comment, reply, chat edit, email and bio editors were checked in screenshots in Smooth, Kanagawa, Brutalist and GeoCities at 1100 and 360px. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 21:27:26 +00:00
khannurien	d683ac0548	v4: the dump title editor's Cancel is a button, beside Save — it was the grey text link of the full forms' action rows, next to a filled Save, where every other inline editor pairs Save with a bordered Cancel; it now takes btn-secondary, the same height as btn-primary. Save and Cancel sit in .dump-title-actions and wrap together, so on a phone they go under the input as a pair instead of Cancel dropping to a line of its own. 30 Dump page tests pass, and tsc, eslint and deno fmt are clean; checked in screenshots in Smooth, Kanagawa, Brutalist and GeoCities at 1200 and 360px. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 21:23:28 +00:00
khannurien	b9ff9c417e	v4: phone layouts hold down to 300px in every style, Kanagawa's wide monospace included, and the lint is clean — the floating chat and "+" buttons hid the end of every page (Save on the dump edit page, the last setting), since the page kept room for one button and they stack two; it keeps room for both now, read from one --fab-size. A 1fr page grid grew to its widest unbreakable content, and the edit forms' action bars didn't wrap, so at 320px the dump edit page scrolled sideways and the playlist edit's Save stuck out of its card; the grids take minmax(0, 1fr) and .form-actions wraps. The profile header's info column was a hidden horizontal scroller under a fade, which cut the username, the inviter, the email and the Log out and Invite buttons; it wraps instead (long names and emails break anywhere), and the avatar shrinks to 70% on phones. TabBar pins its leading content (the feed's category switcher) at the row's start while the tabs scroll under it, where scrolling to the active tab used to push it out of view; its edge fades show only while tabs are hidden past that edge, so the last tab is no longer dimmed once scrolled to, and a left fade marks a scrolled row. Tabs are narrower on phones, which lets Kanagawa's search tabs fit at 375px. The global player puts its track controls on a row of their own on phones, where they left the title three letters, and an audio bar narrower than 420px drops its volume slider for the waveform. The profile bio lost the margins it stacked on the page's gap. A dump card is now a container and lays out by its own width rather than the viewport's, since a playlist's is narrowed by its drag handle and remove button and titles broke every few letters: under 472px it is a grid where the vote ends the date line, as on the journal's cards, instead of hanging under the thumb, with a spacer row keeping it off the title; under 300px the thumb takes the top row and the text the full width. The lint errors that predated this are fixed: a useless fragment and two async callbacks without await in tests (they now await advanceTimersByTimeAsync), EmailKit's renderer gets the kit as an argument instead of aliasing this, its figure regex drops the literal no-break spaces \s already matches, digest-service's highlightOf takes NotificationData narrowed per case instead of any, and eslint ignores coverage/. 254 backend and 670 frontend tests pass, and tsc, deno check, deno lint, deno fmt, eslint and doc:lint are clean; every page was checked in screenshots at 300 to 480px in Kanagawa and Smooth. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 21:19:43 +00:00
khannurien	58c95dc4f3	v4: the notifications page restores its list and scroll on Back, and shows what it missed — it now saves its loaded pages and scroll position as it is left and restores them on Back, through useFeedCache, useScrollSave and useScrollRestore like the feeds and the profile's tabs; the saved list keeps the ones the server confirmed read as read, so coming back doesn't highlight them again. The page never refetched after a reconnect, so notifications sent while the socket was down never showed, and the next read-all marked them read unseen; a reconnect, and every restored list (after a reload the socket hasn't said yet whether anything is new), now fetch the first page again, and what is new goes on top with the socket's arrivals, those alone sorted newest first so the server's order stays. Marking read now follows the unread notifications shown, each sent once, instead of each socket message, and a list with none unread sends nothing. The arrivals shift the server's offset pages, so a next page could repeat the last one's tail: each notification shows once. The next page loads as the end scrolls into view, with FeedFooter, as every other page list does, and the loading and error states go through ListState. The "N new" pill was a number before a separately translated "new", which French made "3 nouveau"; it is one plural message, "3 nouvelles". Dates past a week and the tooltip use the app's locale rather than the browser's. The mention, comment and like icons had no colored circle; they have one now, and the decorative emoji are hidden from screen readers. 670 frontend tests pass (new: loading on scroll, the save on leaving, the restore with what came meanwhile, the reconnect backfill), and tsc, deno fmt, deno lint, doc:lint and i18n:compile --strict are clean; the restore was checked in the running app at phone width. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 19:54:31 +00:00
khannurien	0a16d3c96c	v4: users can't rename themselves, /settings lays out again, and its selectors fit a phone — PATCH /api/me refuses a username with a 400 and changes nothing else in that request; a name changes only through PATCH /api/users/:key/username, which needs user:manage, from /@name/manage. The old name still stays an alias for good. renameUser lost its 30-day limit and its exempt flag, and RENAME_LIMITS holds the slug limit alone. /settings shows the username as plain text, and the rename hint left the catalogs. The settings grid sized its control column to its widest content, the rename hint, which stretched every control under it; that column now takes the remaining width and each control keeps its own size. A segmented control wraps instead of overflowing; in a settings row on a phone it takes the full width, and one of more than 4 options (the styles) lays out in two even rows. .btn-border turned red on hover, so Settings, Manage, Retry and Back looked like Log out; it is now neutral (Smooth's accent is red too), and only Log out takes the new btn-border--danger, muted until hovered red. A <button> and a <Link> .btn-border are now the same height. API.md and the module docs say who renames; the backend tests rename through the moderator route and check a self-rename is refused, and the malformed-name check moved to the manage tab's test. 254 backend and 667 frontend tests pass, and tsc, deno check and doc:lint are clean; Smooth, Brutalist, GeoCities and Kanagawa were checked in screenshots at desktop and phone widths. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-07 06:50:43 +00:00
khannurien	4468438ba7	v4: every page and API URL follows one of two rules, written down in API.md — pages are a hierarchy where @name is a user and any other root word a category or an app page, tabs and views are plain segments and a default view is the bare path; API resources are /api/<plural>/<key>, a user's key being an id or @name, the caller's things under /api/me, media a sub-resource, lookups by another key a query filter. ~ is gone: the feed is /, /hot, /new, /journal, /follow and /music/hot, a profile /@bob with its overview and the tabs dumps, upvoted, playlists (her own only), follow (users, and /follow/playlists), invitees and manage (role and name, for user:manage), which replace the separate UserDumps, UserUpvoted and UserPlaylists pages; /settings holds the account (now with the username) and /manage the category manager, both linked from the user menu and one's profile. Dumps and playlists live under their owner, /@bob/dumps/<slug> and /@bob/playlists/<slug>, named by slug alone: migration 0021 re-slugs them to the bare slugify(title), unique per owner and kind (the oldest keeps it, clashes get -2), renames the categories whose slug v4 reserves (hot, follow, settings…), and rewrites this site's v3 links in dump notes, comments, chat, and playlist and profile descriptions. A slug changes only when edited, through the edit forms' new link field, which proposes slugify(title) as the title changes; a username changes from /settings or, for user:manage, from /@name/manage. Every slug or name left behind stays an alias of its item or user for good (dump_slug_aliases, playlist_slug_aliases, user_aliases), so old links redirect without chaining and nobody else can take one; slugs change 3 times a day per item, names once in 30 days, moderators exempt. An alias, or a name in another case, answers the item with its current path: the pages put it in their URL (useCanonicalPath, without fetching again), and the OG middleware answers a 301. shared/paths.ts builds and parses every page path, for the client's links, the emails' (digest-service lost its own builders and its id fallbacks, and links /follow and /settings) and the server's matchers (OG tags, self previews, backlinks, self-url); shared/slug.ts holds slugify and the format categories share. Dumps and playlists carry ownerUsername and a required slug, and notifications join their users' current names and their items' paths each time they are sent or read, so renames reach them and a deleted item's has no link. The API was renamed per the plan's table: POST /api/users and /api/sessions, /api/password-resets, /api/me (with settings, password, avatar, media-token, playlists?contains=, follow, feed?from=, notifications), /api/users/:key/{dumps,upvoted,playlists,follow/users,follow/playlists,invitees,avatar} and by path /api/users/:key/{dumps,playlists}/:slug, /api/dumps?url= and ?category=, /api/dumps/:id/{file,thumbnail,metadata}, /api/playlists/:id/cover, /api/remote/{preview,image,bandcamp,youtube} (one router), and /api/unsubscribe, the v3 POST /api/users/unsubscribe staying for the digests already sent. A stored default feed tab "followed" reads as "follow" once. dumps.slug and playlists.slug stay nullable, since SQLite can't add NOT NULL without rebuilding tables others reference; every writer sets them and unique (user_id, slug) indexes hold them. 254 backend tests (new: migration 0021 on a v3 fixture, slug aliases, swaps and rate limits, username renames, aliases and limits, reads by path and their privacy, the OG 301s, both unsubscribe paths, v4-shaped digest links with no ids, notification joins, shared paths and slugs) and 668 frontend tests (new: the profile's tabs and follow views, Settings, Manage, the link field, canonical URLs on the dump, playlist and profile pages, bare default feed paths) pass, and tsc, eslint, deno check, deno lint, deno fmt and doc:lint are clean; the migration was also run on a copy of the development database. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-06 16:08:59 +00:00
khannurien	fd0d81fe66	v4: the emails' words live in the catalogs alone and their look is read from the site's stylesheets — every email message was written as i18n._({ id, message }), so its English sat in the template and again in en.po, and api/mail/styles.ts held a copy of each theme's palette, fonts and radii that a test chased. The templates now name each message by its id only (i18n._("email.reset.title", values)) and its English is written in locales/en.po, its French in fr.po; the four labels the settings tab and the unsubscribe page share with the emails read the same ids through @lingui/react's Trans instead of restating the English. i18n:extract runs without --overwrite, which would write the id where the English is, and i18n:compile runs --strict, so a language lacking a message fails the build; it caught email.digest.hl.mentionIn, never translated into French. The new api/mail/theme-tokens.ts reads src/tokens.css and src/themes/<style>.css and resolves a token for one scheme (var(), light-dark(), color-mix(), translucent colors laid over the card), and styles.ts names tokens for every color, GeoCities' rainbow, Melange's spice and shadow, Kanagawa's syntax hues, the font stack and the corner radii, keeping only what is the email's own (type scale, title, frame, divider, avatar rounding). The copies had drifted: GeoCities' stack had lost Marker Felt and its light emails drew the dark rainbow, which now also recolors its dividers on the "auto" scheme. The pills take the site's badge color, now the --color-badge-bg token that the private badge reads too. The Dockerfile copies src/tokens.css and src/themes/ into the runtime image, which reads them at load. The catalogs were re-extracted, having been stale against the code. 240 backend tests (new: the token reader, the dark rainbow; the palette-copy test is gone) and 661 frontend tests pass, and tsc, deno check and doc:lint are clean; sendDueDigests' empty-digest test fails now and then, as it did before. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-05 18:21:45 +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	89014639cb	v4: settings are stored on the server, starting with the style, the color scheme and the digest switches — every setting lived in the browser's localStorage, so it didn't follow an account to another browser and the server couldn't read it, which the digests need (whether to send each, and in which style). user_settings (migration 0019) holds one row per user who changed anything, the defaults otherwise, and GET/PATCH /api/users/me/settings read and change it (settings-service, model/settings, validators/settings, shared/settings). AuthProvider reads the settings into the session alongside the user, on mount and on focus, so a change made elsewhere arrives; useSettings changes them at once and puts them back if the server refuses. ThemeProvider now sits inside AuthProvider: a signed-in user's choice comes from their settings and is mirrored to localStorage for index.html's first-paint script, a visitor's stays in the browser, and an account with no choice yet takes the one this browser had, once. Until anyone picks a style, the instance's GERBEUR_DEFAULT_STYLE applies, written into index.html's default-style tag along with email-enabled (src/config/instance.ts), and a browser no longer stores the default as if it had been chosen, so changing the instance's style reaches everyone who never picked one. The settings tab gets an Email section with the weekly digest and monthly recap switches, disabled with a note when the server has no SMTP. The style and scheme names moved to shared/theme.ts in the previous commit; the default feed tab and sounds stay in the browser for now. 223 backend tests (new: settings defaults, patches, isolation, refusals and cascade; index.html's new tags) and 658 frontend tests (new: useSettings, the account's theme winning and a browser's carried over, settings read into the session, the digest switches) 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 21:20:57 +00:00
khannurien	f20ca83f0c	v4: every email is rendered by its template in api/mail/ on one layout that takes the instance's style, and sent one way — the welcome email was built and sent inside the register route and the reset link in its service, each writing Markdown, running marked, and calling a nodemailer wrapper whose cc, bcc, replyTo and send result nothing used. Now api/mail/ holds one template per email (welcome.ts, password-reset.ts) on layout.ts's EmailKit, which draws each block with inline styles in a style from styles.ts: the six site styles' palettes, copied from src/themes/, and what makes each recognisable in a mail client (Brutalist's hard border and shadow, GeoCities' ridge and rainbow rules, Times' underlined ink links, Kanagawa's colored quotes, dates and wordmark, Melange's spice bar). On the "auto" scheme the light palette is inlined and the dark one swapped in by a prefers-color-scheme block; a chosen scheme sends only its own. User-written text is escaped where it lands. strings.ts holds every email's words in English and French. GERBEUR_EMAIL_LOCALE picks the language and GERBEUR_DEFAULT_STYLE the instance's style, which the welcome and reset emails are drawn in; GERBEUR_WELCOME_EMAIL_BODY still replaces the welcome's Markdown, whose blocks now take the style too. email-service sends a rendered email with its recipient and headers, refuses to load when GERBEUR_SMTPS_URL is set without GERBEUR_FROM_EMAIL (sends used to fail one by one), and takes a test transport, through which the tests now read what was sent. The welcome send moves to user-service's sendWelcomeEmail. The style and scheme names move to shared/theme.ts, which the server now needs. README, mod.ts and .env.example describe mail/ and the new variables, and drop the "notifications, invites" email that never existed. 221 backend tests pass (new: the layout's schemes, styles and escaping, both templates, the welcome on sign-up, a reset link from the email that works), and deno check, deno lint, doc:lint, deno fmt, tsc and eslint are clean. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-04 21:11:54 +00:00
khannurien	5f0cf86fe6	v3: the dump page's "In collections" list only shows the owner's private badge — every card there said "public", which says nothing on a list of playlists holding the dump. PlaylistCard takes privateBadgeOnly, passed through PlaylistCardList, which hides the visibility badge except on a private playlist of the signed-in user's; the other lists of playlists keep both badges. 5 PlaylistCard tests pass (new: privateBadgeOnly), and tsc, eslint, doc:lint and deno fmt are clean. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-30 21:19:06 +00:00
khannurien	f3f07ed1e3	v3: every doc opens with the role, then the mechanism, then the details — the docs were written bottom-up: what a thing does first ("Declares the three contexts WSProvider fills…"), details that only make sense once its purpose is known next, and the reason it exists last ("…so that a feed doesn't re-render"), so a reader had nothing to hang the first sentences on. The README's rule caused it, since a module doc "opened with what the module does"; it now opens with the module's role in the app, and the value-side phrasing of a type ("Returned to Y by X") comes once the role is said, with WSContext as the before and after. WSContext.ts is rewritten that way throughout, and so are the docs elsewhere that buried their purpose in a trailing "so that" (the migrations, og, image-proxy, public-address, useFading, useFeedCache, the drafts), opened on a generic verb that hid it ("Holds", "Declares", "Loads", "Defines", "Types": the contexts and providers, the shared types, the models, the hooks), or were label fragments ("The page of a dump…", "A row of tabs."). A function whose verb is its action ("Stores session"), a props or result type, and a protocol message ("Sent by X when Y") already say their role and are left as they are. Only comments changed, and the catalogs only move their references; deno fmt, deno lint and doc:lint are clean. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-29 21:42:08 +00:00
khannurien	2af64721c7	v3: the invite tree spaces every row the same — a child sat 0.2rem under its parent but siblings 0.4rem apart, since each row's spacing was split into top and bottom padding and a last child dropped its bottom half. Every row now has one gap above it (--invite-tree-gap), the connectors run through it from the same variable, and the user link is a block, so a theme's line height can't push the avatar off the connector's center. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-29 21:28:43 +00:00
khannurien	07ac4bc629	v3: every paged list loads its pages one way, PlaylistDetail is split like the profile, and the comments keep the README's rules — Index, FollowedFeed, useUserDumpFeed, UserPlaylists, UserListPopover, Notifications, Search and the profile's followed users each hand-wrote "fetch the next page, append it, clear loadingMore on failure". A list the API pages is now a PagedList whose next page useLoadMore appends, and a list that loads on its own goes through usePagedList, which is useFetchState (now taking an initial value, as a restored cache, that starts loaded) plus useLoadMore. Pages are read by pageUrl with readPage (a refusal throws) or readFirstPage (a refusal is empty, as fetchFirstPage was), signed-in reads go through authFetch, and useFeedCache's saveState takes the list. Notifications keeps the socket's arrivals as their own list merged at render, as Index does with recentDumps, and marks them read in one effect. PlaylistDetail (963 lines) only puts the pieces together (146): usePlaylist holds the playlist, its live sync and its edits, PlaylistHeader, PlaylistEditForm and PlaylistDumpList are one file each, and LiveDumpCard is the vote wiring DumpCardList and the playlist both use. What it re-invented is shared: the load is useFetchState, a dump going private and back is usePositionAwareSync, deletes and edits are useDumpListSync (which also keeps comment counts, which the page didn't). A removal is sent when its fade ends, or as the page is left, so an undo sends nothing: the removal/re-add/reorder dance, the pending-undo order and the 80 ms debounce of its echoes are gone. A refused reorder goes back to the server's last order and says so, instead of reloading the page. The comments written since the documentation pass broke its rules in places: module docs as a label and a colon ("The "dumps" tab of a profile: …"), noun-phrase fragments on types, fields and constants, stale lists of callers, over-long lines; they are sentences now, and the ones the refactor made wrong (PlaylistDetail as useFading's and FollowPlaylistButton's caller, "the profiles" as usePositionAwareSync's only users) say what is true. 650 frontend and 214 backend tests pass (new: usePagedList, useFetchState's initial, pageUrl/readPage, the deferred removal, a dump returning at its place in a playlist), and tsc, eslint, doc:lint, deno check, deno lint and deno fmt are clean. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-29 21:19:33 +00:00
khannurien	69d9bf899b	v3: the profile page is split by one rule, and what it re-invented is shared across the app — UserPublicProfile was 1,949 lines because it held its data, six tabs, seven subcomponents and hand-rolled copies of things the app already had or repeated elsewhere. The page now only picks the tab and puts the pieces together (253 lines); useProfile holds the user and the three lists with their live sync, cache and scroll; each tab and header piece is one file under components/profile/; the followed and invitees tabs load on first open through the new useFetchState, keyed by username, which keeps them across tab switches. Every piece taken out replaces its copies: DumpCardList, which reads the votes from WSContext, is every list of dump cards (the index feeds, followed, search, related dumps, UserDumps, UserUpvoted, the profile), so Index no longer threads vote props into the feeds; PlaylistCardList every list of playlist cards; FeedFooter the eight sentinel/loading-more/end blocks; ListState the loading/error/empty chains; Segmented the settings toggles, RoleSwitcher and the playlist edit form, with SegmentedField built on it; LoadMoreButton the three load-more buttons; NewDumpButton UserDumps' and the profile's; useScrollRestore four copies of the scroll restore, and the profile uses useScrollSave instead of saving its whole cache on every scroll. useFading takes an onFaded callback, which replaces PlaylistDetail's copy of its timers; useUpvotedList is the vote tracking UserUpvoted and the profile each had, and a withdrawn upvote now leaves the list and the cache once faded (it was only hidden, and came back after a tab switch). usePlaylistMemberships is the membership load and toggle AddToPlaylistModal and DumpCreateModal each had. pagedList.ts (PagedList, firstPage, restoredList, fetchFirstPage, withFirst), fetchPublicUser and dataOr are shared by the profile, UserPlaylists and useUserDumpFeed; a refused side list still shows empty, as documented, and only the user fails the page. FollowedFeed's two sections are one local hook, and its readPage is expectOk. A card is its owner's when the item is the signed-in user's, everywhere: the profile's upvoted column marked every dump as the owner's on one's own profile. No message changed; the catalogs only move their references. 643 frontend tests pass (11 new for useFetchState, useFading's onFaded and pagedList), and tsc, eslint, doc:lint, deno check and deno fmt are clean. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-29 20:50:49 +00:00
khannurien	80f3292c8f	v3: fix the frontend's findings, and the backend pieces they needed — each by a rule the code now keeps everywhere. Votes and likes go through one useOptimisticToggles in WSProvider: clicks on one item before any ack share one baseline, which a failure or the timeout restores (a second click used to cancel the first one's rollback), only the last ack's count is applied, a votes_update or comment_likes_update waits while this tab has toggles in flight, and a welcome reconciles them against the server's sets (a vote cast offline left its count one too high). lastLikeEvent, which nothing read, is gone. Follows are refetched when the new follows_changed message, sent to the follower's sockets on every follow write, says another tab changed them; a failed load keeps the Follow buttons disabled and is retried 10 s later instead of leaving them disabled for good (success: false) or enabled over empty sets (network error); follow and unfollow resolve to whether they worked, and the button says when they didn't. CategoriesProvider follows the same rule, so a network error no longer turns every category URL into a 404. Every list is kept by one rule: useDumpListSync and usePlaylistListSync never add an item, and usePositionAwareSync remembers the place of any item that leaves a list, whatever took it out, and puts it back when its next event shows it again; with dump_created now setting lastDumpEvent, a dump going private and back returns to its place on another user's profile, which never happened before, and an updated playlist missing from a list is no longer prepended. useFeedCache keys every cache by viewer (the profile pages restored the owner's private dumps to whoever went back) and reads it again when its key changes (another profile in the same page showed the previous one's list). Every response of a list's first page or next pages goes through expectOk (Index, UserPlaylists, useUserDumpFeed and Notifications read body.data unchecked). ... Smaller fixes: the redirects from / and /<slug> keep the navigation state, so the index hides the dump DumpEdit just deleted; Index's mainFetchDone ref, which blocked nothing, is gone; DumpEdit refuses someone else's dump up front, as the API does; the Dump page takes a handed-over dump once and clears it from the history entry, so a reload fetches; a newly upvoted dump is fetched with the viewer's token; the notifications page sends read-all once per notification; the profile's followed and invitees tabs finish loading when left mid-fetch; UserListPopover no longer closes when its own list scrolls; the vote and like counts open their lists for visitors too (a disabled button swallowed the click; they are aria-disabled now); the removal of a playlist's dump forgets it (an edit brought it back), a failed removal puts it back where it was, an undo waits for the removal in flight, and a refused undo says so; PlaylistCreateForm tells whether the dump made it into the new playlist; the cover preview's object URL is revoked; the muted icon has its cross; the bell no longer stays rung after two notifications within 700 ms; a stale invite check can't overwrite a newer one; a stored theme or color scheme is checked; the default feed tab is one store every reader follows, across tabs; the file fields enforce the size they state, derived from DUMP_MAX_FILE_SIZE_BYTES; formatBytes, friendlyFetchError and ImagePicker are translated (Intl names the units), and ImagePicker ignores keys while uploading; visited dumps and playlists keep a week of visits and are parsed once; a chime is claimed under a Web Lock, so two tabs can't both play it; a date in the future reads as "now". Dead code goes: PageShell's centerSlot, overLimit, the exports of isSafeHex and hueFromString, the comment tree's token prop and ChatContext's isChatOpen and closeChat. The French catalogue has the 11 new messages. 214 backend and 632 frontend tests pass; tsc, deno lint, ESLint and doc:lint are clean. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-29 20:08:00 +00:00
khannurien	a36a4cac6b	v3: fix the backend's correctness findings — each by a rule shared across the code rather than one more special case. Link previews: readTextCapped decoded every page as UTF-8, so a Latin-1 page came out garbled unless it went through the curl fallback, which re-encoded it on its own; pages are now decoded in the charset their Content-Type or <meta> declares, and the curl fallback hands the bytes over as they are. The image proxy served and cached for a day an error answer that claimed an image type, and read a body without Content-Length whole before checking its size; it now refuses an error status and reads through the same capped reader. The icons task's Twemoji fetch gives up after 15 s instead of holding the start up, and extractLinkHref, which only its tests used, is gone. Backlinks matched a linked URL against the raw url column, so https://x.com/ missed a dump posted as https://x.com; they now compare canonical URLs, as reposts do. The four LIKE searches put the query in raw, so % and _ were wildcards and the mention search for "a_b" found "axb"; lib/like.ts escapes them. Every list query now treats a malformed row the same way, a 500 through model/rows.ts's parseRows, where ten of them silently dropped it; the playlist dumps query uses the model's column list rather than a hand-written copy, and the aliased list is derived from the plain one. The vote and like services commit through transaction() instead of their own BEGIN and ROLLBACK, which would have rolled back a caller's transaction. ... Validation: the dump validators refuse the nulls their shared types don't allow (a null title or note on creation, a null URL on edit); a playlist title of spaces is refused and titles are stored trimmed; usernames and emails are checked by the same USERNAME_RE and EMAIL_RE in shared/validation.ts on sign-up, on PATCH /api/users/me (which checked other limits) and in the sign-up form. Mentions are read by shared/mention.ts's mentionedUsernames, which skips code and URLs as Markdown shows them: https://mastodon.social/@bob or a `@bob` in code notified bob; the editor's suggestions derive their trigger from the same pattern instead of a copy. Following an unknown user is a 404 rather than a foreign-key 500, and following a private playlist a 404 rather than a 403 that confirmed it exists. `ws` and `emojibase` are reserved category slugs, since the server answers them before the app. GET /api/users/me no longer turns every error into a 500. Dead code goes: adminOnlyMiddleware, verifyEmailTransport, listUsers, deleteUser with disconnectUser and the force_logout message only it sent, PATCH /api/notifications/:id/read (with markOneRead), isSlugAvailable, the unreachable 500 in fetchComment, and the `if (!ctx.state.user)` checks behind authMiddleware. 214 backend and 627 frontend tests pass. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-29 19:35:53 +00:00
khannurien	14e3ab125b	v3: fix the privacy and security findings of the comment pass — private dumps' broadcasts, mentions and attachments reached people who can't see them, and a few smaller holes. A dump's every write now goes through ws-service's broadcastDumpChange(before, after), which decides the audience: a private dump's creation, edits and deletion reach its owner's sockets alone (its deletion id went to everyone before), going private is dump_deleted for everyone else and dump_updated for the owner (whose other tabs dropped it), going public dump_created for everyone else. Mentions in a private dump, a comment on one, or a private playlist's description notify no one until it goes public (mentions are deduplicated per context, so going public notifies them once), and a moderator's edit of a comment, chat message or playlist mentions in the author's name. A link preview of a private playlist is refused, as one of a private dump already was: the preview is stored in the linking dump, which others may see. Attachments record the type of the resource that links them (migration 0018 backfills it); one linked by a private dump or playlist, or a comment on a private dump, is served only with its owner's media token, which Markdown now adds to attachment URLs, and one whose resource is gone answers 404 and is purged at startup. The Bandcamp provider claimed bandcamp.com.evil.example (the pattern wasn't anchored) and now reuses bandcamp-tralbum's anchored one; the YouTube provider took any site's ?list= for a YouTube playlist and missed m., music., youtube-nocookie and /live/ links, and now shares canonical-url's rules. Sign-in does the same scrypt work for an unknown username as for a wrong password, so timing no longer tells which exist. Usernames and emails are unique regardless of case (migration 0017, which refuses to run over existing clashes and names them): "Bob" and "bob" could both sign up and a mention picked one, emails had no constraint, and a reset request in another case silently sent nothing. The Open Graph tags escape the site name and build their URLs on PUBLIC_URL and FRONTEND_URL instead of the request's origin, which a proxy may rewrite. ... Error statuses follow one rule: APIException takes a status and derives the code from it (BAD_REQUEST 400, UNAUTHORIZED 401, FORBIDDEN 403, NOT_FOUND 404, CONFLICT 409, RATE_LIMITED 429, SERVER_ERROR 500, BAD_GATEWAY 502), so a code can't contradict its status. That fixed real bugs: a wrong current password on change-password answered 401, on which authFetch signs the user out; editing or deleting someone else's comment or chat message answered 401 rather than 403; one invalid dump edit answered 422. A taken username is 409, like a taken email. Along the way: the database is created from schema.sql by connection.ts when the file doesn't exist, so every task works on a fresh checkout (dev and serve never ran db/init.ts, which is gone); the startup purge compared ISO dates with SQLite's datetime() as text and kept anything from the current UTC day, and now compares ISO cutoffs; the seeded admin's hash comes from lib/password.ts (split out of jwt.ts, which needs the JWT secret) rather than a copy of its format; avatar, thumbnail and playlist-cover uploads go through replaceUploadedFile, which writes beside the old file and renames over it before the record announces it (a failed avatar write lost the old avatar; a thumbnail was announced before it existed); upload.ts derives its size message from MAX_IMAGE_SIZE_BYTES and stops re-exporting the upload folders; ALLOWED_IMAGE_MIMES, which duplicated detectImageMime, is gone; the SPA fallback no longer calls next() after answering; a reply must answer a comment on the same dump (a foreign or unknown parent was a 500 or a cross-dump thread). 211 backend and 628 frontend tests pass. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-29 15:19:07 +00:00
khannurien	0d4e344d77	v3: comments say what the code can't — the documentation pass wanted JSDoc on every export and every field of an exported interface, which filled the code with restatements ("The dump.", "Whether it is loading.", "The props of X.", "Renders the X.") and module docs that told a reader nothing ("The hot and new feeds, and the props every main feed of the index takes."); some were wrong as well as vague. Every comment in api/, shared/ and src/ is rewritten under one rule, now stated in the README's "Comments" section: every module and export still has a JSDoc, and it gives context (who uses it and where), usage (what a caller must know: a precondition, what it leaves to the caller, a side effect) or its place in the chain (where its input comes from, where its output goes); a field, parameter, private function or local gets one only when it adds to its parent's, and context about a whole type goes on the type; a type's doc is phrased from the value's side ("Returned to Y by X when Z", "Passed to Y by X"); full sentences, as short as they can be, no label-colon-list; // goes wherever the code isn't obvious and says why, and the section banners and narration go. Every claim was checked against the code, which turned up some forty comments that were wrong: DumpCard's claimed it was only used by the hot and new feeds, the WebSocket welcome claimed to be a socket's first message (presence comes first), UpdateUserRequest said null removes a description (it keeps it), deletedDumpIds leaves out that it also holds dumps made private, THEME_COLOR claimed to mirror index.html's, and the like. The routers' JSDoc now says where they are mounted and what becomes of a thrown APIException. doc-lint.ts lets a field without a doc through; a module or export without one still fails. No code changes; 204 backend and 628 frontend tests pass, and doc:lint, deno check, deno lint and tsc are clean. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-29 14:31:14 +00:00
khannurien	42e0356d21	v3: the docs follow deno doc's grain — the index page was a hand-kept list of some 250 links into deno doc's output layout (./api/main.ts/index.html), policed by mod.test.ts; deno doc has no module index of its own beyond all_symbols.html, which lists every module by path with its summary. mod.ts is now a prose overview of the three parts and their layers, linking that list once, and mod.test.ts is gone. Re-exporting everything from mod.ts, as deno doc's tutorial does for a library, was tried and doesn't fit: export * drops the 18 routers (default exports) and makes 17 names ambiguous, and deno 2.9 renders export * as namespaces with the root's module doc and nested ones empty, losing every module doc. A router's page showed one line: its endpoint list sat on the default symbol, whose summary was cut at the colon introducing it; the list is now the module doc of all 18 routers, and the router's JSDoc says so. 22 summaries ended in a colon (the routers, BandcampPlayer, migration 0016, useDumpListSync, usePlaylistListSync) and read cut short wherever deno doc shows a first paragraph alone; they end in a full stop, doc-lint.ts now fails on a colon summary, and the README states the rule. No behaviour changes; 204 backend tests pass. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-28 19:28:24 +00:00
khannurien	563daa720c	v3: src/ is documented like api/ and shared/, and the reference covers the whole repository — the frontend had no module docs, JSDoc on some components and hooks and not on others, props typed inline in half the components and in private interfaces in the rest, // comments on declarations, and no return types on 147 exported functions, so deno doc --lint over src/ found 637 errors. The rule api/ follows is applied to every non-test module of src/: a /** … @module */ block, JSDoc on every export and every field of an exported interface, /** */ for declarations and // for notes, verbs for functions ("Renders…" for a component), parameters named in the prose, {@link} only for symbols in scope, and a page's JSDoc names its route and who may see it (App.tsx's module doc lists the route table). For deno doc to show them, every exported function states its return type: components React.JSX.Element (or | null, or a portal), memo and forwardRef components their exotic types, contexts React.Context<…>, and a hook returning several things a documented Use<Name>Result interface (useAuth, useRequiredAuth, useFading, useEmojiTrigger, useMentionAutocomplete, useDumpCardLink, usePlayRichContent; FeedCacheResult, YouTubePlaylistControls and UseApiFormReturn are renamed to match). Every component that takes props takes an exported, documented <Name>Props interface — the 64 inline prop types are lifted into one, DumpFeed gets DumpFeedProps extending MainFeedProps. React types go through the global React namespace: the four default imports of react are gone, since deno doc can't follow them. Two duplicates go too: useMentionAutocomplete's UserResult was shared/'s UserSuggestion, and JournalCard re-exported JournalShape for no one; i18n.ts no longer re-exports Lingui's i18n, which its three users now import from @lingui/core. deno task doc and doc:lint now take src/ (but its tests and the generated locales), the index page gains a Frontend section linking every module by layer and domain, mod.test.ts walks src/ as well, and the docs workflow runs on changes to src/. deno doc doesn't document npm packages, so it takes a type imported from one — react-hook-form's FieldValues in the form fields, frimousse's SkinTone — for a private one; doc:lint now goes through doc-lint.ts, which fails on every diagnostic but those, 25 of them. The README states the frontend's part of the rule. No behaviour changes; 205 backend and 628 frontend tests pass, and tsc, deno check and deno lint are clean. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-28 07:31:22 +00:00
khannurien	b85a71d86e	v3: shared/ is documented like api/ — deno doc --lint over api/ and shared/ goes from 385 errors to none — every module of shared/ opens with a /** … @module */ block; every export and every field of an exported interface has JSDoc, each request body naming the endpoint that takes it and each response its endpoint; the // comments that described declarations (the mention pattern, the permission union and table, four socket messages, the error message's optimistic-action context) are /** */ now, and the socket protocol's heading comment, which sat on no declaration, is its module doc. The lint wanted explicit types on MENTION_RE and DUMP_MAX_FILE_SIZE_BYTES. Along the way: the chat edit is allowed to moderators too, which the new ChatEditMessage doc says; UpdateUserRequest says that PATCH /api/users/me takes only its description and email; playlist-service's reorderPlaylist @throws is rewrapped. mod.ts and the README no longer call shared/ undocumented. No behaviour changes; 205 backend tests pass. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 21:13:07 +00:00
khannurien	d62e4a9883	v3: the backend is documented for deno doc, one way in every file, and the docs build and publish themselves — comments were in three shapes: file headers as // lines, as an untagged /** */ (which deno doc drops, only an @module block counts) or missing; 43 exports described by a // deno doc can't see, two JSDoc blocks sitting on the wrong declaration and one duplicated; endpoints commented in 5 of 18 route files, three ways, on router.get() calls no doc tool reads. The rule, applied to every non-test module of api/, migrations included: each opens with a /** … @module */ block; every export has JSDoc, and so does every field of an exported interface; a comment describing a declaration is /** */, // only notes code; functions open with a verb, everything else with a noun phrase; parameters are named in the prose, {@link} only for symbols in scope; whatever throws to its caller says so with @throws (for an APIException, the statuses and when); a router's JSDoc lists its endpoints, method and path, who may call it, what it does. deno doc --lint goes from 528 errors to none in api/ (shared/ keeps its 385; it is next). The lint also wanted explicit types on config.ts's settings and a few return types, JsonLdResult exported, and the routers typed Router. deno task doc builds the HTML reference into docs/ from every module of api/ and shared/ but the tests, deno task doc:lint lints it; the index page is the module doc of a root mod.ts (deno.json's exports), which introduces the backend, shared/ and the frontend to come and links every module, and mod.test.ts fails when one is missing. The docs workflow builds it on every push to main that touches them and publishes it to the gh-pages branch, like publish-assistant's. Two defects found along the way are fixed: a private playlist's cover was served to anyone who had its id or slug, and now needs its owner's media token, as private dump files do (playlistImageUrl adds it for private playlists only, so public cover URLs stay the same for everyone); og.ts and static.ts each filled index.html's placeholders, and now share lib/index-html.ts, so the OG middleware picks up a rebuilt index.html instead of keeping the first it read. The README states the documentation rule. 205 backend and 628 frontend tests pass. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 21:07:40 +00:00
khannurien	1702370447	v3: coverage tasks for both suites — deno task test:coverage runs the Deno tests of api/ and shared/ with coverage and writes the report to coverage/api, deno task test:web:coverage runs the Vitest suite with @vitest/coverage-v8 (at vitest's own 5.0.2) into coverage/web. Each prints a table and leaves an HTML report and an lcov.info. They write to separate folders because deno test --clean empties its whole output directory, which would take the frontend report with it. The frontend report counts src/ without src/test/ and the generated src/locales/; the backend report still lists api/test/helpers.ts and setup.ts, since deno.json has no coverage exclude and a separate deno coverage --exclude pass merged the raw profiles into different figures. No behaviour changes. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 20:30:05 +00:00
khannurien	d392be7cb8	v3: api/ gets one layout rule and keeps to it, and whatever crosses the wire is defined once, in shared/ — api/model/interfaces.ts and src/model.ts declared the same 40 types by hand and had drifted: the error envelope, email required on one side and optional on the other, Invitee/InviteTreeEntry and LinkPreview/UrlPreviewResponse for one shape each. The validation limits, the role permissions, the chat length cap, the @mention pattern and the Bandcamp and YouTube player responses were copied too, behind "keep in sync" comments. shared/ now holds them once, one file per domain (api, category, chat, comment, dump, follow, mention, notification, permissions, player, playlist, search, user, validation, ws), and imports nothing from either half. The server sends Dates; the client reads Wire<T>, the same type with every Date an ISO string, and turns it back with src/utils/shared/deserialize.ts. APIErrorCode becomes a const object, since the frontend's erasableSyntaxOnly rejects an enum. The shared types showed two real gaps: the avatar upload and GET /api/users/me return the full self user and the frontend read them as public users, and a stored session may predate media tokens, so the client keeps it as its own Session type. The backend rule: each folder holds one kind of module, each file is named after its domain. routes/ parse and answer; services/ hold the logic and every SQL query (lib/auth.ts's token check moves to user-service); remote/ is the only place that fetches from other sites — the guarded fetch and the HTML scraping out of rich-content-service, public-address, the Bandcamp and YouTube resolvers, the link providers, the image proxy (fetch, checks and cache, out of routes/preview.ts) and the Twemoji glyphs (out of lib/site-icons.ts). model/ holds per table a row type, its guard and its mappers; validators/ the request guards; db/connection.ts the handle and transaction(), db/startup.ts the purges and default admin that db.ts ran on import, called at the same moment. Tests mirror the source one file per module, test/<layer>/<module>.test.ts, with shared/'s under shared/test/ (deno task test runs both); the test that covered the bandcamp and soundcloud providers together is two now, and the backend can() has its own, so 202 tests. deno lint src/ is clean too: window.* becomes globalThis.*, and test bodies and act() callbacks that never awaited lose their async. The README states the rule. No behaviour changes; the catalogues only move their source references. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 20:19:48 +00:00
khannurien	e9f6a1f23e	v3: deno check src/ passes again — the tests import userEvent by name — 71 errors, all the same one: @testing-library/user-event 14.6.7 ships ESM-syntax declarations without "type": "module", so Deno reads them as CommonJS and the default import is the module namespace, on which setup() and click() don't exist. Vite resolves it the other way, which is why the suite still ran. The package exports userEvent by name as well; the 40 test files that imported it use that now. No behaviour changes; the 627 tests pass and the lint is clean. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 19:49:49 +00:00
khannurien	2b6e68391b	v3: src/ gets one layout rule and keeps to it — each layer holds one kind of module, and every layer is split into the same domains — the previous pass sorted components into folders but left hooks, helpers and a barrel inside them, kept hooks/, utils/ and contexts/ flat, and still grouped most tests by theme. The rule: components/ holds only components (.tsx), contexts/ holds FooContext.ts + FooProvider.tsx pairs, hooks/ every hook, utils/ every non-React module, pages/ one component per route and nothing else. Each of components/, contexts/, hooks/ and utils/ is split into the same domains — auth, category, chat, comment, dump, editor, feed, form, layout, player, playlist, profile, social, theme, ws — plus shared/ for what belongs to no one domain, so the chat is components/chat, contexts/chat, hooks/chat and utils/chat. components/ui becomes components/shared to match. Out of components/: useApiForm goes to hooks/form, chat's message.ts and time.ts to utils/chat (with chatDraft and chatText, now draft.ts and text.ts), the emoji skin tone to utils/editor, the remembered volume to utils/player; player/icons.tsx is PlayerIcons.tsx. components/form/index.ts is gone: everything is imported from its own file, FormProvider straight from react-hook-form, and expectOk sits beside friendlyFetchError in utils/shared/api.ts (apiError.ts), since pages that have no form used it too. Out of pages/: the index's feed views, tab bar and presence row move to components/feed; HotFeed and NewFeed, identical but for the sort, are one DumpFeed with an order prop, which also takes over MainFeedProps from pages/index/types.ts. App is a named export like everything else. Tests mirror the tree one file per module — test/<layer>/<domain>/<Module>.test.tsx — so the themed files (hooks/misc, data, editor, sync; utils/misc, text, storage, remote, journal; contexts/providers; pages/auth and UserSubpages; layout/Fabs) are split by subject, the setup each piece needs kept and the rest dropped; the floating-button test becomes one test per button, 627 tests now. The README states the rule. No behaviour changes; the catalogues only move their source references. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 19:44:14 +00:00
khannurien	07d0344be7	v3: src/components sorted into feature folders, the four largest files split, and the component tests mirror the new layout — 48 components sat in one flat folder, ChatModal ran to 1301 lines, DumpCreateModal to 983, and the tests were grouped by theme, so there was no telling which component had none. Components now live in category/, chat/, comment/, dump/ (with dump/create/ for the new-dump wizard), editor/, form/, layout/ (header, page shell, floating buttons), player/, playlist/, profile/, social/ (vote, like, follow, the who-voted popover) and ui/; FeedTabBar and PresenceRow, used by the index page alone, move to pages/index/. ChatModal gives up ChatMessageItem, PendingItem, ChatEditBox and ReplyRef, the message limits and text helpers (chat/message.ts), the day and clock labels (chat/time.ts) and the draft storage (utils/chatDraft.ts, next to dumpDraft.ts). DumpCreateModal gives up StepIndicator, LocalFilePreview, TitleField and DuplicateNotice; CommentThread gives up CommentForm and CommentNode; TextEditor gives up the emoji picker popup and its skin-tone storage; MediaPlayer gives up its icons (which FilePreview imported from it), Waveform, SeekBar and the remembered volume. NewPlaylistForm, a button that opens a modal holding the form, is NewPlaylistButton now, and Thumbnail, ThumbnailPlaceholder, FilePreview and RichContentCard use named exports like everything else. The tests are split one file per component under src/test/components/<folder>/ (and src/test/pages/index/), the same 625 of them. No behaviour changes; the catalogues only move their source references. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 19:30:28 +00:00
khannurien	de5abeceb2	v3: a performance pass over both halves — indexed queries, streamed files, cached previews, and feeds that stop re-rendering on every tick — a review turned up whole-file reads, unindexed scans and render storms; each fix comes with a test. Database: migration 0016 adds the indexes the hot queries lacked — dumps by created_at, (user_id, created_at) and slug, votes by (user_id, created_at), playlists by slug, users by lower(username) for @mentions, and the child side of every cascading foreign key (comment replies, a user's comments, follows both ways, chat messages, invites, reset tokens); the feeds, dump and playlist pages, upvotes and follower fan-out all plan as index searches now, with no temp-table sort. SQLite runs in WAL with synchronous=NORMAL and a busy timeout, and a transaction() helper wraps the multi-row writes (a dump and its categories, category sets, backlinks, attachment links, playlist reorders, follower notifications) so they commit once instead of once a row. Lists no longer run a COUNT(*) per page: they fetch limit + 1 rows and hasMore comes from the extra one; total stays only on search, the one place it's shown. Files: /api/files read the whole upload (up to 50 MB) for every range request a player makes while seeking; api/lib/serve-file.ts streams the requested bytes from disk, with an ETag and 304s, and serves avatars, attachments, covers and thumbnails too. File and thumbnail URLs carry the file's version (size and last change), so the current one is cached for a year and anything else revalidates — a same-size replacement used to stay stale, and a video's still was cached immutable forever under a URL that never changed. Feeds showed image dumps at full size: /api/thumbnails now makes a 960px preview with ffmpeg (GIFs and unreadable images go out as they are), one run per output however many cards ask at once, and ffmpeg is probed once per process instead of spawned per request. Uploads stream to disk; replacing a file writes beside it and renames over, a hard link keeping the old one for rollback instead of holding both in memory; generated stills and previews go when the file is replaced or the dump deleted, and a custom thumbnail with its dump. Outbound: fetchWithTimeout's deadline now covers the body — a server that answered at once and then dripped could hold a create open indefinitely — and pages are read to 4 MB at most; metadata is cached five minutes so the create form's preview and the post that follows share one fetch (refresh-metadata still goes to the page); proxied images sit in a 32 MB LRU. Broadcasts serialize once rather than per socket, and index.html is kept in memory until the file changes. Frontend: currentTime lived in PlayerContext, so every card re-rendered — and re-parsed its Markdown — four times a second while anything played; the playhead has its own PlayerTimeContext now. The socket context is split three ways (feed, chat via useWSChat, presence via useWSPresence), so typing, presence and chat messages leave the feeds alone. DumpCard, JournalCard and Markdown are memoized; card images use the server preview and load lazily; the feed cache is written when the page is left rather than after every pause in scrolling; the theme context value is memoized. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 16:35:51 +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. ... 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. ... 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. ... 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
khannurien	25601edaf8	v3: the backend gets a test suite, and the bugs it turned up are fixed — deno task test runs 186 tests over api/ in about 18s, against a throwaway database and uploads directory (GERBEUR_DB_PATH and GERBEUR_UPLOADS_DIR now override the defaults; api/test/setup.ts sets them before any app module loads), with requests sent through the whole middleware stack in-process via app.handle, outbound fetches and DNS stubbed, and the socket tests on a real Oak listener since an upgraded request has no response for handle() to return. It covers auth (token purposes, revocation after a password change, live roles, login rate limiting, invites claimed once even under parallel registrations, password resets), dumps end to end (create, privacy, by-url matching on the canonical key, backlinks, categories, edits, file replacement, thumbnails, refresh, delete), file serving headers and ranges, attachments, avatars, comments and likes, playlists, follows and feeds, notifications and mentions, categories, search, chat cursors and read markers, link previews and the SSRF guard on every redirect hop, the Bandcamp and YouTube resolvers, migrations (a fresh schema is left untouched; the data rewrites in 0004/0005 and 0010–0013 do what they say), CORS, OG tags, static caching, and the WebSocket protocol. Fixed along the way: an unknown category id on a new dump returned 400 but kept the dump row, and for a file dump deleted the file while keeping the row; an edit or file replacement with one was half-applied before the 400 — categories are now checked before anything is written. A suffix range (bytes=-3) was read as an end offset and got a 416 instead of the last three bytes. Reordering a playlist accepted a list naming a member twice. verifyPassword threw on a stored hash without a salt instead of answering false. And a request with no Accept-Encoding, which RFC 9110 lets a server treat as accepting anything, was sent the brotli sibling; curl and bots now get the plain file. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 12:31:05 +00:00
khannurien	16b510764c	v3: the frontend ships compressed, caches for good, and stops reaching for other people's servers — a review of the bundle's size found the server was the biggest cost. Oak's send() compressed nothing and sent every hashed chunk with max-age=0, so a first visit to the home page moved about 710 KB of JS and CSS where 200 KB would do, and every later one revalidated some thirty files. The build now writes .br and .gz next to every JS, CSS, JSON and SVG file (a precompress plugin, brotli 11), which send() already serves when the browser accepts them; /assets and /emojibase are cached for a year as immutable, index.html is no-cache, and static responses carry Vary: Accept-Encoding. A missing file under either prefix is a 404 rather than index.html with a 200, which an open tab after a deploy used to try to run as a script; the app now reloads on vite:preloadError, at most once a minute. manualChunks sent everything in node_modules to one vendor chunk, so react-hook-form and frimousse loaded with every page though only the lazy form and editor chunks use them, and its markdown rule missed hast, vfile, unified and property-information while every feed loads Markdown through DumpCard anyway; only react, react-dom, scheduler and react-router get a chunk of their own now, the rest is left to the bundler. Up-front JS goes from 133 to 96 KB gzip, the home page from 178 to 160. Fonts: index.html blocked rendering on a Google Fonts stylesheet, opening two more hosts and handing each visitor's IP to Google; they come from @fontsource now, with src/fonts.css declaring only the latin and latin-ext faces (the packages' own CSS declares every subset, which added 6 KB gzip to the stylesheet), so Vite hashes the woff2 files into /assets and a browser still only downloads its theme's font. The families are the variable builds ("Saira Variable" and so on); Fraunces keeps its opsz, SOFT and WONK axes. Emoji: frimousse fetched emojibase-data@latest from jsDelivr, unpinned and third-party; emojibase-data 17.0.0 is a dependency now, its data and messages for each lingui locale are emitted under /emojibase/17.0.0 (served from node_modules in dev), and the picker is pointed there. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 11:44:08 +00:00
khannurien	bf28e49c34	v3: left and right step through emoji results too — with a :query picker open, only up and down went to the grid; left and right moved the caret in the textarea, so the first press of either did nothing visible. Every arrow now steps into the grid, landing one along from the automatic first match; Escape gives the caret back. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 11:29:44 +00:00
khannurien	85d0703077	v3: the editors stop fighting the person typing — a review of every edit box turned up an Enter that sent a chat message mid-IME, an emoji picker that grabbed focus on ":D" and never went away, markdown that folded every line break, and limits that were counted but not kept. TextEditor: an Enter confirming an IME candidate no longer also sends, saves or picks. The : trigger wants two letters (Unicode ones too), not one, so emoticons, "note:" and "12:30" type through, and it keeps focus in the textarea — the picker's search follows what's typed, Enter or Tab takes the best match (a name equal to the query, then one starting with it, rather than frimousse's first in emoji order, which made ":rocket" an astronaut), the arrows step into the grid and typing there carries on in the textarea, and Escape closes just the picker, not the chat around it. It closes on a press outside, which it never did, and neither autocomplete fires inside code or a ``` fence. A ☺ button in each editor's corner opens the picker for mouse users, with a skin-tone selector that's remembered, and its search takes focus only once the picker is placed (a hidden field can't). Emoji, mentions and pasted images go in through execCommand, so Ctrl+Z takes back the insertion instead of wrecking the undo stack. Uploading an image made the textarea disabled, which dropped focus for good; it's read-only now, says "Uploading image…" under the box rather than only in an empty one's placeholder, reports a failed upload instead of swallowing it, and holds Enter until it's done. Any dropped file is caught — a non-image used to be opened by the browser in place of the page, draft and all — and a paste carrying text (a spreadsheet's cells come with a picture of themselves) pastes the text. The mention list follows the caret (a stale lookup can't close a newer one, lookups abort), wraps, scrolls its selection into view, closes on blur, and is a real listbox with aria-activedescendant. The counter shows from 80% of the limit by default rather than a standing "0 / 5000". Limits: chat Send and edit Save refused nothing over 2000 and let the server bounce it; they're disabled past it now. RichTextField's bare maxLength rule failed without a message, so an over-long "Why?" made Save do nothing and say nothing; it says so now. Inline editors answer Escape through a new onCancel (comment reply and edit, profile description, playlist description and title), the playlist editor takes Ctrl+Enter and can't be saved without a title, and saving a comment unchanged no longer stamps it edited. Markdown: a single newline is a line break, so a message written over three lines stops reading as one; links into the app route in place instead of opening a tab; headings are sized for prose instead of 2em page titles, and dropped from card teasers; GFM tables scroll in their own box and task lists lose the bullet; images load lazily. A chat message of only an emoji or three is shown large. Themes: brutalist and geocities give the new buttons no chrome, and geocities' ghost-button list, split in two by the placeholder-thumbnail rules in 1cb904d, is whole again, so its modal ✕, emoji cells and the picker's ✕ are flat as intended. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 10:24:30 +00:00
khannurien	5fd1d364da	v3: the header stops cropping itself and the search stops losing its place — a review of the header, search bar and presence row turned up tabs cut off at both ends, avatars that could only be all there or gone, and a search that forgot what it was showing. Between 860px and about 1000px, and again just past 1150px where the username and Playlists links arrive, presence plus tabs ran wider than the header's centre column; the slot couldn't shrink, so the column clipped it on both sides, and logged in FOLLOWED went missing with one person online. The container queries meant to shed presence under 460px and tabs under 280px never fired, overridden by later rules of the same weight, and no fixed width would have been right anyway: the tabs run from 314px (nyt, English) to 485px (melange, French). The slot can now shrink, and the presence row sizes itself through a new useRoomBeside, which measures the room the tabs leave it in the header and in the row under it: as many avatars as fit, at most six seats, the rest folded into a "+N" counter (square in brutalist), and never fewer than an avatar and the counter — below that the tabs scroll instead. So presence shows at every width, phones included, where a 512px rule used to hide it. When the header's tabs do scroll, they get the app's right-edge fade in the header's own colour, only while they overflow, and the active tab is scrolled clear of it. The avatars fan out on hover only where there's a pointer, so a tap doesn't leave them spread, and the label-less "+" button squares up with the icons beside it on small screens. Search: the ✕ on results opened directly ran navigate(-1) out of the app and now goes home; after several searches it stepped back one query at a time and now returns to where the search began in one step, carried in the entry's state, which useTabParam now keeps across result tabs; those tabs (/search/~/users) weren't recognised as the results page at all. On results the bar stays open holding the current query, including after a new search or Back, where it used to collapse and empty. The input no longer grabs focus on a results page, so a phone's keyboard doesn't cover the results, and submitting puts it away. Escape closes an opened search and returns focus to its button, or on results reverts an edited query, overriding the search input's own clear-on-Escape. The two search bars share one draft, so crossing the breakpoint keeps it. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-27 10:00:15 +00:00
khannurien	bfce3186f8	v3: the frontend stops trusting what it last saw — a review of src/ turned up counts that fell to zero, stale copies winning over fresh ones, sessions that tabs trampled on each other, and a player queue that outlived its listener, and this closes them. Voting or liking something with no live count yet started from 0, so a dump with five votes read 1 on the click and, with the socket down, 0 for good once the undo fired; the buttons now pass the count they show, which the optimistic update and its undo start from. After a reconnect the chat re-read its history as the truth, but the merge let the live copy of a message win, so an edit made while the socket was down never showed; the resync now replaces live copies with what it fetched. Every dump_updated added the dump to the live list if it wasn't there, so retitling a months-old dump put it at the top of New, Hot and Journal for everyone, while one that had arrived live kept its old title; an edit now refreshes a live entry in place and adds nothing, and the server announces a dump going public as a dump_created, which is the one case that is an arrival. The new-dump modal let a slow preview for the previous link land on the new one, title suggestion included; changing the link now aborts both lookups. Markdown found mentions by rewriting the source text with a rule that, unlike the server's, didn't skip an @ after a word character, so an email address linked its domain, a URL like mastodon.social/@bob was split, and @ in code came out as literal link syntax; mentions are now found by a remark step over text nodes only, by the server's rule. Changing the password in one tab stored fresh tokens that any other open tab wiped the moment its old token drew a 401; the session now lives behind utils/authStorage.ts, tabs follow each other through storage events (so a logout also signs out every tab), a 401 only forgets the stored session if it still holds that token, and in-place session updates go through a new updateSession() rather than re-reading localStorage. The persisted player queue carries media-token URLs yet survived logout for the next person to sign in; it is now stamped with its owner, not restored for anyone else, cleared on logout or account switch (a guest's queue survives signing in), and older unstamped queues holding a file are dropped. Playlist checkboxes in the new-dump and add-to-playlist modals flipped whether or not the server agreed and now wait for it. CommentThread and the Dump page's loads use authFetch like the rest, so a 401 there signs out consistently; isTokenExpired has one home; and the useScrollLock imports carry their extension so deno check passes. ... Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-24 20:24:38 +00:00