The lease was set once to now + LOCK_TTL_SECONDS (120 s) with no
renewal. Tasks that legitimately take longer — slow CDN downloads,
ugoira encodes, rate-limited batch forwards (a 100-message channel copy
waits ~4 min on the per-chat token bucket) — had their lease expire
mid-run; the 30 s expiry sweep flipped the row back to pending and
another worker processed it again, double-sending.
run_with_lease now drives the handler through tokio::select! and
refreshes locked_until every 30 s while it runs. The heartbeat lives in
the same future as the handler, so a panicking worker still lets the
sweep recover the row (no leaked task keeping the lease fresh forever).
A task only reaches the queue after a failed send, so the fresh attempt
never ran post_send_actions (edit-before-forward prompt / channel
forward) — it failed before that point. The old guard skipped
post_send_actions for resumed tasks (batch_index > 0 or sent ids
present), which meant any send that needed a retry after partial
progress silently lost its forward and edit prompt.
post_send_actions is now run unconditionally on a successful queue send;
it executes exactly once, after the whole sequence completed.
AGENTS.md: document the twitter API entity decode and the /test report
HTML-decoded display; add the missing db.rs / media_sender.rs /
rate_limit.rs module rows; fix the statics location (handlers/statics.rs);
refresh test counts (~115, twitter live 5, pixiv api.rs 1, photo heavy
test); versioning convention now includes README.en.md.
README.md / README.en.md: add TELOXIDE_PROXY to the env variable list.
Twitter's syndication and GraphQL APIs return tweet text and display
names pre-escaped for HTML (> < & '); the caption builder
escaped the text again, so sent messages showed literal entities (e.g.
>^ω^< came back as >^ω^<). from_syndication_json now decodes the
API text before storing it — both the syndication path and the
TWITTER_AUTH_TOKEN GraphQL fallback route through it — so the caption
escapes exactly once and renders correctly.
The /test report is a plain-text message but printed the pre-escaped
caption and render fields; it now HTML-decodes them for display so the
report shows the rendered text.
Deleted tweets answer the syndication endpoint with HTTP 200 and a
TweetTombstone (no `errors`, no `id_str`). The body classifier only
knew the `errors` shape, so tombstones fell through to the
`no id_str -> Sensitive` branch and degraded to an empty result,
making the bot reply "No media found" for a deleted tweet.
- extract parse_syndication_body(); tombstone shape -> NotFound
- propagate NotFound from the TWITTER_AUTH_TOKEN GraphQL fallback
instead of swallowing it into empty_fetched
- unit tests for all body classes + live regression test on a real
tombstoned tweet
P0 — level rework + redaction:
- info now carries only lifecycle, per-post business results (sent /
forwarded / copied / template applied), admin actions and anomalies
(upload fallback, retry enqueue; dead-letter stays error).
- Per-request detail moved to debug: message/command logging, URL
extraction, fetching/fetched, link-cache hits, media-group batch sends,
queue processing (enqueue/processing/completed/rescheduled), photo
processing (downscale/transcode), inline queries, sensitive-tweet note.
- Full user-submitted URLs and message text now appear only at debug; at
info and above links are printed via the normalized cache key.
P1 — request correlation:
- handlers::log_key() maps a URL to its normalized post key
(twitter:<id> / pixiv:<id> / bsky:<handle>/<rkey>). The whole lifecycle
of one link (fetch -> send -> cache -> fallback) now logs [key=...], so
multi-worker logs can be correlated by grepping the key.
Convention documented in AGENTS.md.
site::fetch retried every PixivError, so a bad/expired token (403) or a
deleted artwork (404) burned all 3 attempts with backoff against pixiv's
API for nothing. Add PixivError::Status(u16) — the app-API calls now
surface the HTTP status — and retry only the transient classes: network
errors, 429 and 5xx. 4xx / Api (token errors) / Json / NoAuth are
returned immediately. The classification is a pure helper
(fetch_error_is_retryable) with unit tests.
The old stop only set an atomic flag checked between jobs: a worker
blocked in recv() never woke (the channel was never closed), and queued
jobs were neither drained nor abandoned in a defined way despite the
"drains up to 256 jobs" comment. Now stop_url_workers sets the flag,
drops the sender so blocked recv() calls wake with None, and awaits the
worker JoinHandles (each finishes its in-flight job first). main awaits
it inside the existing 30s shutdown timeout.
Smaller/faster production binary (verified: cargo build --release -p
xmedia-bot builds clean with the new profile). panic=abort is
intentionally not set — queue workers and db closures rely on JoinHandle
catching panics, which abort would defeat.
Full English translation of README.md — features, quick start, webhook
deployment (domain + IP-only with acme.sh), env table, command table.
The Chinese README stays the primary one.
A typo in EDIT_MESSAGE_TTL_SECONDS / WEBHOOK_PORT etc. used to silently
fall back to a default, so the bot ran with different behavior than the
operator intended (or failed much later on a bare .expect). Unparseable
values now log a loud warning naming the variable; invalid webhook
settings still surface as a hard .expect in webhook mode.
The stop sequence awaited the queue workers, which can be mid-download
(30s client timeout) or mid-ugoira encode (minutes). A stuck worker would
hold shutdown forever; now the process logs and exits after 30s.
download_to_temp buffered the full bytes, wrote them to a temp file, and
prepare_photo then read the whole file back from disk. The bytes are
already in memory — pass them through (download_to_temp now returns
(file, bytes)) so photo processing never touches the disk for input.
Adds the bytes dependency to xmedia-bot (already in the lock via x-media).
download_media_limited buffered the whole frame zip (cap 512 MB) in
memory before extraction, spiking RAM for large ugoira. New
site::download_media_to_file streams chunks straight to a temp file with
the same Content-Length / stream cap checks, and ugoira_video now opens
the zip from disk inside spawn_blocking. Adds FetchError::Io for local
write failures (hand-rolled error pattern preserved).
The download-and-reupload fallback downloaded each batch item serially,
so a 9-item batch took 9× the slowest download. Items are now prepared
concurrently (bounded to 3 in-flight downloads + photo processing) via a
JoinSet, then the group is uploaded in its original order; the per-item
logic moved into prepare_upload_item. Temp files stay alive until the
group request completes. A failing item still aborts the batch (the
JoinSet drop cancels the remaining prep tasks, as before).
A long tweet text or a pixiv artwork with many tags can exceed Telegram's
1024-char caption cap for HTML parse mode, turning an otherwise fine send
into a permanent 400. truncate_caption() cuts at a char boundary (never
splitting a multi-byte rune or an HTML entity like &) and appends an
ellipsis. Applied in caption_with / caption_from_fields, the link-cache
re-send path and the inline-query captions.
Telegram's sendMediaGroup requires the first item to be a photo when a
group mixes photos and videos; the previous code kept the source-site
order, so a mixed post with a video first (twitter media order is not
guaranteed) would 400 permanently. photos_first() stable-sorts photos
ahead of videos/animations before chunking; within-kind order is kept.
Every DB operation (queue lease/enqueue, chat_state get/set, link_cache
read/write) used to open a fresh connection — including the busy timeout
and WAL pragma — then close it, on every message, URL job and callback.
Replace with DbPool: a tiny pool (4 connections max, semaphore-bounded
concurrency for backpressure) whose with_conn() method runs the closure on
a pooled connection inside spawn_blocking. Steady-state cost of an
operation is a list pop + semaphore acquire instead of a connection open.
The docker workflow only builds/pushes; tests were a local responsibility.
Add .github/workflows/ci.yml with two layers:
- test: cargo fmt --check + cargo clippy --workspace --all-targets -D
warnings + cargo test --workspace (fully offline, no secrets) on every
push/PR, including forks.
- live: the #[ignore]d live-network tests plus the pixiv token-gated
tests, run on schedule / manual dispatch / tag pushes only (fork PRs
cannot read repository secrets), with PIXIV_REFRESH_TOKEN /
TWITTER_AUTH_TOKEN injected and continue-on-error for flaky sites.
Test gating (documented in AGENTS.md):
- live-network tests now carry #[ignore = "live network: ..."] (twitter 3,
bsky 2, pixiv bogus-token 1) and run via -- --ignored live.
- pixiv token tests early-return when PIXIV_REFRESH_TOKEN is absent or
empty (an unset GitHub secret arrives as ""); test_fetch previously
failed without a token.
Also fixes the three clippy assertions_on_constants warnings in send.rs
(required for -D warnings).
A task whose media is a locally produced file (pixiv ugoira MP4, bsky HLS
remux MP4) references a path inside a tempfile TempDir owned by Fetched.
The retry ran after that TempDir was dropped, so the file was already gone
and the retry always failed permanently ("local media file missing") — and
the upload fallback even tried to GET the local path as a URL.
Keep the temp dirs in a process-wide registry (Fetched::take_keep_alive ->
send::KEEP_ALIVE) that is only released when the task settles (sent or
permanently failed); the upload fallback now uploads local files directly
instead of attempting to download them.
Restart-mid-queue still loses the files (documented behavior in
input_file_for) — only the in-process retry path is fixed here.
Telegram fires an inline query on every keystroke and every prefix of a
pasted URL (status/12, status/123, ...) matches the site patterns, so
typing used to trigger a full 3-attempt fetch per keystroke. Answer only
after the query has been stable for 800ms, dedupe repeats through
Telegram's inline cache (explicit cache_time 300), and let a repeat of a
query that produced no answer retry the fetch.
YoursFunny