Commit Graph
8 Commits
Author SHA1 Message Date
peter c5f121139b fix: every first watch state survives a write beside it, and a folded Watch later keeps its pin
A season marked watched, a pull from Plex and a YouTube resume point made their first row by a read
then an add, and failed on the unique key when a first checkpoint or the watched mark made it in
between. They go through the same insert-then-read as the checkpoint (the resume point is one
upsert); a pull that meets a local change still waiting for its push leaves it.

Folding duplicate Watch later lists moved their videos but not a Home pin, which the cascade then
deleted; the first pin of a dropped copy moves onto the kept list.

Marking one notice read trimmed before its own flag was flushed, so the trim left one past the cap.
The route test now crosses the cap and reads the result from a session of its own.
2026-10-20 01:56:30 +02:00
peter 8206384c91 test(fence): owner filters in any form, statement writes under any name, core commits by list
The owner-filter check missed an aliased() model, an import alias, .in_() and
filter_by; four live inline filters on aliased VideoState go through owned_by now.
Cross-unit writes are seen under any import name, through sa.update and via
__table__. Core modules may commit only where CORE_COMMITS names the reason;
trim_read and audit.gc flush now and their callers commit. The pin target, the
debug-session header and a link's target fetch through owned_row. The docs
claim what the fence holds, and say what is beyond it.
2026-10-20 00:45:21 +02:00
peter ad6ccbac05 refactor(core): the core's owner filters go through userscope, each held by a second account
Tokens, notifications, quota, debug sessions and the reset-link burn filtered
their owner inline. Ten of the fifteen filters had no test that noticed one
going missing; each now has a test where another account's rows must not
count, show or change.
2026-10-19 23:36:01 +02:00
peter b1832afd18 style: apply ruff format across backend + scripts (line-length 100)
Mechanical, whole-tree `ruff format` pass — no behaviour change. Recorded in
.git-blame-ignore-revs (next commit) so git blame skips it.
2026-08-23 07:43:06 +02:00
peter 80ea1918e3 fix(downloads): derive title clamp from the column + real-Postgres regression
Address review findings on the title-clamp: TITLE_MAX_LEN now reads the length straight
from Notification.title so the clamp can't drift from the column a migration would change
(single source of truth). Add a db-lane test that actually commits a >255-char title into
Postgres (end-to-end proof the truncation is gone, not just the clamp logic) and one that
asserts the constant tracks the column width.
2026-08-16 19:40:17 +02:00
peter 83326ecf53 fix(downloads): clamp notification title to column limit so download_gc can't wedge
A downloaded FB/IG reel's "title" is the whole caption (up to ~1900 chars). It flowed
untrimmed into Notification.title (String(255)); at the batched download_gc commit that
raised psycopg StringDataRightTruncation in pass 1, aborting the whole run so expiry
deletion, LRU eviction and orphan reap never executed (13/110 prod assets already exceed
the limit). Clamp the title to 255 (with an ellipsis) centrally in create_notification —
defending all six callers — since the full title already reaches the client via `data`
and the column is only an English fallback. Add a regression test.
2026-08-16 19:24:51 +02:00
peter 3dfa35f396 perf(notifications): bulk-delete trimmed read notifications (C-4.1)
trim_read looped a db.get()+db.delete() per id (N+1). The ids are already in hand, so
one DELETE ... WHERE id IN (...) replaces the loop.
2026-07-21 21:47:26 +02:00
peter b9a3a9012d feat(notifications): durable per-user inbox (P1) + maintenance schema
Add a server-backed notification center that coexists with the client-side
transient bell: a per-user `notifications` table (type/title/body/data JSON/
read/dismissed), a `/api/me/notifications` CRUD API (list, unread_count, read,
read_all, dismiss, clear), and a left-nav inbox module with a live unread badge
polled via useLiveQuery. Known types render trilingual text from type+data
(English stored text is the fallback); read rows are trimmed past a soft cap.

Also adds the schema the maintenance job builds on: videos.list?part=status
columns (embeddable/privacy_status/upload_status) and the validation lifecycle
columns (last_checked_at, unavailable_since, unavailable_reason).

Page-id validation is centralized in one PAGES source of truth (isPage) so the
new page survives reload without a second allowlist to keep in sync.
2026-06-18 03:20:17 +02:00