this fix the nextcloud login + drive selector (chroot)
fix also invitation / magic link
also correct the UX: once user has logged in nextcloud, show an explicita page
- fix issue with selected drive and user logout/login via another user
(was raising a "404 not found")
- normalize all localStorage to "oxi-" prefix
- add a specific frontend/AGENTS.md for frontend part (stop increasing the global AGENTS.md)
add `include_in_photo_index` and `include_in_music_index` policies
both true for default personal drive
photo is implemented
music is not yet implemented
The e2e CI job ran the legacy `scenarios/*` specs against the vanilla `static/`
frontend that upstream has since removed, so it could never pass. Point CI at
this repo's SvelteKit SPA suite (tests/e2e/spa) and wire up what it needs:
- CI: build the release binary with `--features plugins` (the admin Plugins-tab
specs exercise the WASM runtime) and run `npm run test:coverage`, building the
instrumented SPA with COVERAGE=1 VITE_E2E=1 so the server serves the
data-testid-instrumented build the specs drive.
- Coverage harness: target 127.0.0.1 instead of `localhost` (which resolves to
::1 first on CI runners while the server binds IPv4, so readiness never
connected) and poll `/ready` for webServer readiness; tee start-server-spa.sh
output to a log surfaced by an always-run CI step for diagnostics.
- Files page: restore a persistent breadcrumb home link (buildCrumbs returns
only the path folders, so there was no "go home" affordance), and fix the
`?file=` deep-link race where the viewer→URL effect stripped the param before
the listing loaded — a bookmarked preview link now opens the viewer.
All 101 spa specs pass locally.
Add an end-to-end and unit test suite for the SvelteKit frontend:
- Playwright e2e specs (tests/e2e/spa) with a throwaway container stack,
codegen scenarios, and an Istanbul-based coverage report pipeline.
- Vitest unit tests across API endpoints, components, stores and composables.
- `data-testid` hooks on interactive elements (AppShell, FileViewer,
ShareDialog, search, photos, files breadcrumbs, login/Nextcloud flows,
public share pages) so the e2e suite can target them deterministically.
- Serve the SPA app-shell CSP from a <meta> policy (svelte.config.js) plus a
middleware that skips the CSP header on HTML; move the Nextcloud Login Flow
v2 grant page to the SvelteKit /nextcloud/login route.
- `just front-codegen` recipe and start-server-spa.sh harness.
Make the test environment robust and consistent:
- Install a deterministic in-memory localStorage/sessionStorage in the Vitest
setup so storage behaves identically across Node versions (Node 26 ships a
native Web Storage global that otherwise shadows jsdom's).
- Pin devenv to Node 26 + PostgreSQL 18 and pin every CI job to Node 26.3.0
so the dev shell and CI run the same toolchain versions.
Repair the API/WebDAV (hurl) suite, which had drifted from the backend:
- Migrate the removed `/api/folders/{id}/listing` endpoint to `/resources`
(cursor-paginated `{items:[{resource_type,resource}]}` shape) across the
batch-copy, grants, nested-group, and WebDAV NC tests + the dav_helpers
wipe routine.
- Stop photos_etag from uploading the dedup-tracked fixture so the dedup
blob-lifecycle test can own its content-addressed blob exclusively.
- dedup_create now asserts the idempotent same-content re-upload (201 +
existing file id) instead of the stale 409 expectation.
Generated coverage reports, nyc output and the e2e server runtime data dir
are gitignored rather than committed.
Videos now get a thumbnail generated eagerly server-side on upload, through
the same WebP/blob-hash pipeline as photos — instead of the old browser path
that only ran when the Photos grid first rendered a video tile, re-downloaded
the whole video to seek a frame, and PUT 3 JPEGs back (and produced nothing at
all for HEVC/.mov, which a browser <video> cannot decode).
- New VideoFramePort (application) + FfmpegVideoFrameService / NoopVideoFrameService
(infrastructure): shell out to the system ffmpeg (no compile-time libav dep),
extract one representative frame as PNG, bounded by its own semaphore + a
per-process timeout + kill_on_drop. Noop when ffmpeg is absent/disabled, so
videos degrade gracefully to no thumbnail.
- ThumbnailRefreshHook.on_file_created routes video/* to
generate_video_thumbnails_background: stream the (decrypted, reassembled) blob
to a size- and time-bounded temp file on the data volume, extract a frame, and
reuse the shared render_and_persist_all_webp helper — so video thumbnails are
WebP, blob-hash keyed (dedup'd) and content-negotiated, exactly like photos.
- GET thumbnail serves the video's WebP to every client (byte-sniffed
Content-Type); a genuine miss returns 204.
- Config: OXICLOUD_ENABLE_VIDEO_THUMBNAILS (default true, needs ffmpeg detected
at startup) + OXICLOUD_FFMPEG_PATH / _CONCURRENCY / _TIMEOUT_SECS / _MAX_MB.
- Dockerfile installs ffmpeg in the runtime image.
- Frontend: drop the client-side generateVideoThumb/frameFromVideo re-download
path; the server is now the source of truth.
Benchmark (examples/bench_video_thumbnails.rs, needs ffmpeg): 4/4 codecs incl.
HEVC/.mov produce a thumbnail server-side (was 0% for HEVC); ~50-70 ms/frame in
the background; ~3.9 KB preview WebP; up to ~23x less per-first-view transfer on
the test corpus (far more on real multi-MB clips). Methodology in
benches/VIDEO-THUMB.md.
Hardening from an adversarial review: video render holds the decode_semaphore
like the image path; the ffmpeg scale filter bounds both dimensions; the blob
stream has a timeout; the temp file lives on the data volume; the size cap uses
saturating_mul.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ShareDialog (~15 KB JS) and MoveDialog (~5 KB JS) were statically imported by the
files, favorites, recent and shared routes, so they downloaded on every visit
even if the user never opened a share/move dialog. Convert them to the existing
lazyComponent pattern (as already used for FileViewer/WopiEditor): the chunk is
fetched the first time the dialog is opened.
The Vite manifest confirms both flip from static to isDynamicEntry. This defers
~26 KB raw / ~9.5 KB gzipped (JS + CSS) off the initial load of those four routes.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Checkpoint of the in-progress frontend toolchain work (Vite pinned to ^6 after
the 7/8 rolldown build break, eslint-plugin-svelte v3 navigation/reactivity
fixes, CI/Dockerfile/manifest updates) together with three UI performance
optimizations (verified on the Vite 6 build):
- Critical CSS: move auth.css/music.css off the global path into their route
chunks (login/device/nextcloud-login, music) -- -25% gzipped critical CSS
(~5.4 KB) on every non-auth/non-music page load.
- relativeTimeAgo: cache the Intl.RelativeTimeFormat (was rebuilt per call, once
per row per render) -- 22.7x faster date formatting in large lists.
- Virtualize search results and grouped trash (list view) via VirtualList -- DOM
rows mounted stay ~constant (~27) instead of O(N) (94.6% fewer for 500 hits).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
OIDC SSO login intermittently ended on a 403 "Invalid or expired OIDC state
— possible CSRF attack" even though the login had already succeeded
server-side.
Root cause: the (now-removed) legacy vanilla-JS frontend registered a
`/sw.js` service worker that, with navigation preload enabled, double-fetched
the top-level navigation to `/api/auth/oidc/callback`. The OIDC `state` is
single-use, so the first callback consumed it and logged the user in while
the duplicate (~0.4s later) found the state gone and returned the 403 the
browser rendered.
Backend — idempotent callback: after a successful web login, remember
`state -> exchange_code` in a short-lived (120s) cache. A duplicate callback
whose state was already consumed now replays that same redirect instead of
403-ing, returning the cached result directly without re-running the IdP code
exchange (the authorization `code` is single-use too). Keyed by the
unguessable 32-byte state, so it adds no new attack surface and fixes the 403
for everyone — including browsers still running a stale legacy service worker.
Frontend — evict the stale worker: the current SvelteKit app registers no
service worker, so fresh clients can't double-fire. But a browser that
previously loaded the legacy frontend still has `/sw.js` registered and
controlling pages (and `/sw.js` now 404s, so vendor self-cleanup is
inconsistent). killLegacyServiceWorker() runs first in the root layout's
onMount: it surgically unregisters only `/sw.js` workers, drops only the
legacy `oxicloud-cache-*` caches, and reloads once (guarded).
Fixes#510.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two robustness fixes behind the recurring "folder upload stuck at ~93%" reports.
1. Skip non-regular files up front. A copied s6/runit service tree contains
FIFOs (e.g. supervise/control named pipes) that report a size but BLOCK
FOREVER when the browser reads them — the deterministic ~8-files-short that
no retry/watchdog tweak could fix. uploadBatch/uploadTree now probe each
file's first chunk against a 3 s timeout (partitionReadable), upload only the
readable ones, and report the rest: "N uploaded · M skipped (not regular
files)". Progress runs over the uploadable count, so it reaches 100% instead
of parking at 93% while a lane hangs on a pipe.
2. Auto-reload on a new deploy. svelte.config.js polls _app/version.json
(60 s); the root layout reloads itself when the deployed build changes —
unless an upload is in flight — so an open tab can't keep running stale code
after a rebuild (the recurring "my fix isn't applied" trap).
npm run check: 0 errors, 58 tests pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The legacy-frontend removal dropped the deprecated /api/folders/{id}/listing
route, but folders.ts still called it, so every folder view 404'd
("listing failed: 404"). Complete the migration: fetchFolderListing now pages
through the cursor-paginated /api/folders/{id}/resources feed and rebuilds the
combined {folders, files} listing the views expect.
- Pages through next_cursor (limit 200) and splits mixed resource items by
resource_type. 403 still throws; the 304/ETag fast-path is gone (that feed has
no whole-listing ETag) so the in-memory folderCache is the only revalidation.
- Favorite/share badge sets aren't carried by /resources, so they come back
empty for now (no star / share badge until wired from /favorites + /shares).
- folders.test.ts updated to the paginated shape.
npm run check: 0 errors. 58 tests pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Follow-up to the connection-exhaustion fix. Rather than routing large files to
plain uploads (which kept STORAGE dedup but gave up delta's re-upload bandwidth
savings), keep delta for every file >= 8 MB and instead cap each worker's
concurrent connections so a few large files uploading at once can't blow past
the browser's ~6-per-host budget and starve the small-file plain uploads.
- deltaWorker.js: serialize negotiate (at most one in flight per worker) and
drop chunk-PUT concurrency 2 -> 1, so each worker holds ~2 connections max.
- deltaUpload.ts: revert the 64 MB threshold back to 8 MB — every large file
gets sub-file dedup again. (Storage dedup was never affected: BLAKE3 + CDC +
ref-counting run server-side for plain and delta uploads alike.)
With main upload concurrency at 2, total in-flight upload connections stay <= ~4.
npm run check: clean, 58 tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
With WASM finally enabled, large files (e.g. 32 MB logs) started running the
delta worker, which opens SEVERAL concurrent requests each (overlapping
negotiate batches + chunk PUTs). A few of those running at once blew past the
browser's ~6 connections-per-host limit, so plain uploads of the small files
queued with zero bytes sent until the 30 s stall watchdog cancelled them — the
upload "stuck at 4% / 94%" with N (pending) XHRs in the Network panel. The
session-refresh request got starved too (the spurious 401s).
- Raise the delta-worker threshold to 64 MB (new DELTA_WORKER_MIN_SIZE) so
typical large files take a single-connection plain upload. Delta's payoff is
sub-file dedup on RE-upload; on a first upload it is pure connection overhead.
Client-side instant-hashing still only reads files < 8 MB into memory.
- Lower upload concurrency 3 -> 2, leaving headroom under the 6-connection
budget for session refresh/poll and the occasional delta worker.
npm run check: clean, 58 tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The auth screens rendered the OxiCloud cloud mark with viewBox "120 120 280 280",
whose left edge (x=120) cropped the cloud's left side (the path starts at x≈107).
Align them to the AppShell (logged-in) logo's viewBox "95 67 320 320" — same cloud
path — so the mark is fully visible and centred with proper padding inside the
badge, matching the in-app logo everywhere.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause of folder uploads "freezing at ~95%": the global Content-Security-
Policy `script-src` was `'self'` + inline-script hashes with NO
`'wasm-unsafe-eval'`. Chromium therefore blocked `WebAssembly.instantiate`
("Wasm code generation disallowed by embedder"), so the vendored BLAKE3/FastCDC
WASM threw on instantiation — both on the main thread (instant by-hash uploads
and the batch dedup check) and inside the delta-upload worker. Every file then
fell back to a plain byte upload, and the backend logs showed 0 check-batch /
0 negotiate calls. Large files (32 MB service logs) compounded it and the
session token expired mid-upload, so the last handful failed.
- web/mod.rs: add `'wasm-unsafe-eval'` to `script-src`. WASM-only, safe variant
— does NOT enable `eval()`/`new Function()`. Restores instant uploads, delta
(sub-file dedup), and the client hashing the idempotent re-upload relies on.
- deltaUpload.ts: liveness watchdog on the delta worker. A healthy worker posts
progress sub-second; if it goes silent for 20 s it is wedged (WASM init or
chunking hung without throwing) — disable delta for this file AND every later
one so they fall straight through to a plain upload instead of each burning
the full 120 s+ delta timeout. Defense-in-depth so a broken WASM path can
never again freeze an upload for minutes.
cargo test: pass. npm run check: clean, 58 tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Re-uploading a partially-uploaded folder used to surface hundreds of spurious
"already exists" failures, and a file the watchdog aborted (or one the server
committed just before the client gave up) was lost.
Backend — save_file_with_blob_impl (the shared write path for both plain and
by-hash uploads): on a name conflict (23505), if the existing non-trashed file
holds byte-identical content (same folder, same name, same blob hash), return
that file as success instead of erroring. A different-content clash still
conflicts. Re-upload / re-sync becomes a clean no-op for everything already
stored — only the genuinely missing files transfer.
Frontend — uploadWithRetry: each file gets one automatic retry on a transient
failure (quota is never retried). With backend idempotency, retrying an
already-stored file is an instant no-op and a stalled/aborted file gets a real
second chance, so a folder upload self-completes instead of leaving gaps.
cargo test: 448 passed. npm run check: clean, 58 tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A folder upload with several large files could appear frozen for ~2 min: a few
concurrent uploads stalled and the old 120s per-file timeout neither aborted the
request (leaving zombie XHRs that exhaust the browser's per-host connection pool)
nor recovered quickly.
- uploadFileWithProgress now self-aborts on a stalled connection: the deadline
resets on every upload-progress tick (a slow but *moving* transfer is fine),
and once the body is sent the server gets a fixed window to respond; on a stall
xhr.abort() frees the connection immediately — no zombie, no cascade.
- Lower upload concurrency 4 -> 3 to reduce server contention from large
concurrent uploads.
- The outer per-file timeout is now just a generous backstop for a wedged delta
worker / by-hash request.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Backend:
- POST /api/dedup/check-batch — returns the subset of submitted whole-file
BLAKE3 hashes the caller already owns, in one query (user-scoped,
anti-enumeration via idx_files_blob_hash). Lets a client learn which of N
files it can skip with a single round trip.
(dedup_service::user_owned_blob_references, dedup_handler, routes) + tests.
Frontend — upload pipeline:
- Instant ("by-hash") upload for content the caller already owns: hash every
in-band file, ONE /api/dedup/check-batch, create the owned ones with zero
content bytes, upload only the rest. Covers all sizes below the 8 MB delta
threshold (delta handles larger files). vendor/hashWasm computes the
whole-file BLAKE3 on the main thread.
- Resilient parallel uploads: bounded concurrency (4) + a per-file deadline,
so one stuck/slow/failing file no longer freezes the whole batch — it blocks
only its own lane and times out / is skipped while the rest proceed. Quota
exhaustion stops the run early; partial results are reported ("N uploaded,
M failed").
- Folder uploads (uploadTree) show live bell progress + a final result and go
through the same dedup + parallel pipeline.
- Storage bar ("Almacenamiento") refreshes after uploads/deletes
(session.refresh) instead of showing the stale login value.
Frontend — i18n / UI fixes:
- Fix literal {{count}} and {{percentage}}/{{used}}/{{total}} (param-name
mismatches) in the selection toolbar and storage line; add es strings.
- Remove the underline on user-menu link rows.
Benchmark (uploadStrategies.bench.test.ts) compares baseline / per-file / batch:
the batch collapses N per-file probes into one check (e.g. a WAN 1000-file run
drops from 1700 to 1001 round trips) while matching per-file's byte savings.
Also includes in-progress group virtual-description i18n work present in the
working tree (groups.ts, ResourceList, locale `groups` keys).
Verified: cargo clippy -D warnings (clean), backend 448 tests; frontend
npm run check (clean), 58 unit tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Bundles the backend+frontend delta-upload (content-dedup) feature with a
batch of frontend fixes from this session.
Upload / dedup:
- Client-hashed delta & instant upload (deltaUpload, hashWasm vendor shim)
- Backend dedup batch endpoint (dedup_service, dedup_handler, routes)
- session store owned-hash helpers; unit tests + upload-strategy bench
Frontend UI/UX:
- Colour file-type icons in grid/list (per-type tinted tiles + glyph hue)
- Robust thumbnail fallback; PDFs now show their type icon (backend
generates no PDF thumbnails) instead of a blank tile
- Fix PDF preview: load via a same-origin blob: iframe — the API URL is
blocked by the global X-Frame-Options: DENY in the browser's framed
PDF viewer, matching the existing CSP `frame-src blob:` design
- Groups: localized virtual-group description (no DB schema-note leak),
add nav.groups to the 15 missing locales, fix primary-button contrast
- Repoint --color-text-light → --color-on-accent (was faint grey on accent)
- Nudge the admin role badge off the user-menu header divider
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Backend — tail latency & throughput:
- FileContentCache, image transcode, and search now use moka single-flight
(try_get_with / get_or_load) so N concurrent misses for the same key
collapse to one disk read / transcode / query instead of a thundering herd.
Microbenchmark (128 concurrent on one hot key): 128 loads / p99 ~1023ms
before vs 1 load / p99 ~32ms after.
- DB: configurable per-statement timeout on the primary pool
(OXICLOUD_DB_STATEMENT_TIMEOUT_SECS, default 30; maintenance pool exempt) so
a runaway query can't pin a connection and starve the pool.
- DB: background pool-saturation monitor
(OXICLOUD_DB_POOL_MONITOR_INTERVAL_SECS) that WARNs as the primary pool nears
exhaustion — the early signal before tail latency cliffs.
- mimalloc: set MIMALLOC_PURGE_DELAY=0 (Dockerfile + compose) so freed pages
return to the OS and RSS tracks the live working set; benchmarked on
musl/aarch64 at ~400MB reclaimed vs 0MB with the default.
Frontend — UI / i18n fixes:
- i18n: fix literal "{{count}}" and "{{percentage}}/{{used}}/{{total}}" in the
selection toolbar and storage line — the call sites passed param names that
didn't match the locale placeholders; unify on `count` and pass the storage
template its params. Add es files.selected_count.
- sidebar: hide the drive picker when there's only one drive (the redundant
"Personal" row); remove the coloured left accent on the active nav item.
- logo: stop clipping the cloud's left bulge — viewBox recentred on the cloud's
true bbox with proportional SVG size so it keeps the same rendered scale.
- user menu: drop the default <a> underline on the link rows.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- photos: generate + upload the three video-thumbnail sizes in parallel via
Promise.allSettled instead of a sequential await loop; previewData is still
captured before its upload so the local preview survives a failed upload.
- useOwnerCache: resolve the batch, then apply a single reactive assignment
to `#names` instead of spread-copying the record once per id (fewer copies
and fewer derive re-runs on large resolves).
- admin: migration status polling slowed from 2s to 5s.
- files store: soft cap (10k) on the per-item selection toggle. Bulk
"select all" in the views is intentionally left uncapped.
The three tiers had collapsed to one identical value (#5e6a78 / #9fadbe),
so hints, metadata, timestamps and placeholders were all indistinguishable.
Re-separate them into three perceptibly distinct steps, each verified WCAG
AA >=4.5:1 on every page/surface/hover/muted background in both light and
dark mode. Emphasis maps to contrast: muted is strongest (general secondary
text), faint the weakest (timestamps/placeholders) sitting on the AA floor;
secondary stays clearly stronger than all three. The dark floor is binding,
so the dark steps separate upward (lighter = more emphasis).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013J5koaCrHDMvS7uwpawLBN
Edouard Vanbelle