`Path.resolve()` is a filesystem call — it lstats every component of the path.
`_resolve_dlc_path` and `safe_join` both re-resolved their ROOT on every single
call, and those run once per song, per art fetch, per scanned row.
Found while profiling a 2-fps report: on a real 50,944-song library the server
was issuing ~23,500 stat/lstat calls per second, re-walking the same three
parent directories over and over, and burning ~50% of a core doing it. It is
worst exactly where big libraries live — the library was on an NTFS-3G (FUSE)
mount, where every stat is a userspace round trip through mount.ntfs-3g (itself
visible in top). The cost was the constant re-resolution, not the work.
A root is fixed for the life of the process, so resolve it once
(safepath.resolved_root, lru_cache). Measured, 5,000 lookups against a real
library path:
before: 15,264 stat syscalls (54.2 ms)
after: 277 stat syscalls ( 0.8 ms) 55x fewer
Containment is unchanged, which is the part that matters:
- safe_join still resolves the CANDIDATE on every call — following its
symlinks IS the zip-slip / traversal defence, so it is never cached. Only
the server-owned root is.
- _resolve_dlc_path keeps its lexical containment check (deliberately does not
follow symlinks, so junction-mounted libraries keep working).
Tradeoff, documented on resolved_root: if a root's symlink is re-pointed at a
NEW target while the server runs, the old target stays in effect until restart.
Fine for a library path fixed at startup; the cache is keyed on the Path, so
switching library dir is a different key.
Tests: root resolved once across 500 lookups (the regression), a different root
is a different entry, and the containment contract re-pinned — traversal,
Windows drive-absolute, backslash, NUL, empty, and a symlink escaping the root
must still be refused. Full suite green.
* fix: remote transcription posts to /transcribe, not /align (stem-splitter#17)
transcribe_vocals_remote() POSTed the vocal stem to /align. That endpoint is FORCED ALIGNMENT —
"here are the lyrics, tell me when each word is sung" — and its `text` field is required. We have
no lyrics; transcribing them is the entire point. So the server rejected every request with a 422
from FastAPI's validation layer, before its handler ever ran, and remote transcription has never
worked for anyone.
It now posts to /transcribe (added in feedBack-demucs-server#14), which takes only the audio.
`language` moves from the query string to the FORM BODY, where the server actually reads it
(Form("")). As a query param it was silently ignored, so an explicit hint did nothing and
Whisper's auto-detection quietly decided instead — loading the wrong wav2vec2 aligner. It
"worked", it was just wrong, which is the failure mode that hides for months.
Error bodies are no longer cut at 300 chars. The body IS the diagnosis: a 422's JSON names the
field it rejected, a 500's traceback answers on its LAST line. Both got decapitated — which is
part of why this stayed invisible for so long. The message explaining the bug was inside the part
that got cut.
Nothing caught any of this because every test of this module tested the MAPPER, fed a hand-written
dict. The mapper was always fine. The request was never exercised, and the request was the bug.
tests/test_lyrics_transcribe_remote.py now pins it: the endpoint, the form field, the multipart
upload, the bearer token, an instrumental returning no lyrics rather than an error, and a 404
saying the server is too old. Verified they FAIL against /align + params.
Signed-off-by: topkoa <topkoa@gmail.com>
* fix: make the error-body cap an actual bound; correct the docstring's endpoint
- _err_body() appended the truncation marker AFTER slicing to _MAX_ERR_BODY, so the result could
exceed the cap it exists to enforce (4014 chars for a 4000 bound). A cap that is only a
suggestion surprises exactly the callers who trust it — a log line, a job record persisted to
disk and re-read on every load. The marker now fits inside the bound.
It also stripped after measuring, so a short JSON body followed by kilobytes of trailing
whitespace got truncated: real content cut to make room for blanks. Strip first, then measure.
- The public docstring still advertised /align — the exact contract this PR exists to change, in
the one place a reader would look for it. It now says what the function does and why, and that
an older server answers 404.
Found by Copilot and CodeRabbit on #959.
Signed-off-by: topkoa <topkoa@gmail.com>
* fix: keep the exception line when truncating — the tail is the answer
_err_body() kept only the HEAD of an over-long body. On a traceback the last line is the
diagnosis, and the docstring said exactly that while the code threw it away: a 4000-char window
holding "Traceback (most recent call last)" and none of the exception is a window onto nothing.
Same mistake as the 300-char cap it replaced, one level up — cutting off precisely the part the
function exists to preserve.
Head AND tail now, both inside the bound: two thirds head (what was being attempted), one third
tail (what actually went wrong), with the marker between them. Verified the test FAILS against
head-only truncation.
Found by Copilot on #959.
Signed-off-by: topkoa <topkoa@gmail.com>
* fix: every failure out of transcribe_vocals_remote() is a RuntimeError; 404 says why
The docstring promised one failure mode — RuntimeError — and the caller (_maybe_transcribe_lyrics)
catches exactly that so one song's failed lyrics don't take down the batch around it. But a DNS
failure, a timeout, a reset connection or an unreadable stem escaped as requests.RequestException
or OSError, walked straight past that handler, and turned "this song's lyrics failed" into "the
whole batch died".
A 404 now explains itself. Bare "404" sends someone hunting for a typo in their server URL; the
real answer is that their server predates /transcribe, and we are the only ones in a position to
know that.
Found by Copilot on #959.
Signed-off-by: topkoa <topkoa@gmail.com>
---------
Signed-off-by: topkoa <topkoa@gmail.com>
* fix(sloppak): the full mix is a stem — drop the invented `original_audio` key (#933)
Core read, served, and depended on `original_audio:` — a top-level manifest key
this repo invented in #583 that the feedpak spec never defined. The format
already had a home for the pre-separation mixdown: it is a stem. feedpak 1.15.0
(feedpak-spec#53) RESERVES the id `full` for it, so read it from there.
The key existed to work around a bug in our own reader. The packer's comment
said so plainly: "we must NOT list the full mix as a playable stem — the player
sums every entry in `stems` and does not gate playback on `default`, so a listed
full mix plays on top of the stems". Faced with a reader that would double the
song, the packer put the mixdown outside `stems` and invented a key to point at
it. The fix belongs in the reader, and that is what this is.
load_song() now partitions the stem list: `full` comes out as
LoadedSloppak.full_mix, the instruments stay in .stems. Nothing that sums stems
or draws one fader per stem can see the mixdown, so retaining it is safe — which
is what lets the packer put it where the format says it goes.
- ws_highway: `song_info` gains full_mix_url / has_full_mix. The old
original_audio_url / has_original_audio remain as deprecated aliases for one
release so an older stems plugin keeps working (#945).
- `stems` on the wire, and stem_ids / stem_count in the library index, are now
INSTRUMENT stems only — a separated pack that retains its mixdown no longer
advertises a bogus "full" chip or an inflated stem count.
- enrichment: fingerprint against the mixdown wherever it lives. This widens
coverage — _song_audio_file() previously returned None for any pack without
the invented key, so fingerprinting silently did nothing for nearly every pack.
- sloppak: `original_audio:` is still READ as a deprecated fallback, because
every pack in the wild carries it and would otherwise lose its pristine mix.
tools/migrate_full_mix_stem.py rewrites those packs into the spec shape
(original/full.ogg -> stems/full.ogg, add the `full` stem at default:off, drop
the key); the fallback and the aliases die with #945.
The spec gate keeps the debt honest: the grandfather entry now tracks #945, and
the gate fails if it goes stale.
Verified: spec gate OK (4/4, incl. ingesting the spec's new example pack that
retains `full`); 2493 python tests, 995 js tests; migrator round-tripped over
real packs from the library and the results pass the spec's reference validator.
* fix(migrate): discover directory-form packs instead of silently skipping them
iter_packs() searched only files, so a directory-form pack (`song.sloppak/`, the
authoring shape) was walked INTO and never yielded — silently missed by a run
that's meant to be exhaustive. Discover suffix-named directories too (yielded
whole, not descended into), and route packs through migrate_pack/verify_pack.
Directory packs are REPORTED as `dir-form-unsupported`, not rewritten in place:
a single-file pack is replaced atomically (a fully-built temp archive swapped in
with one os.replace), but a populated directory can't be swapped that way, so an
interrupted in-place rewrite could leave an authoring pack half-migrated. The
status is a problem status, so it counts against the run's exit code and shows
in the summary — the operator re-packs or migrates it as a `.feedpak` instead of
it vanishing from the report. Addresses a CodeRabbit review finding.
Signed-off-by: Kris Anderson <topkoa@gmail.com>
* fix(migrate): verify requires an explicit `off` on a retained full mix
verify_zip accepted any non-truthy `default` on a multi-stem `full` (missing,
empty, boolean, `false`/`no`/`0`, malformed) as "ok". But core defaults an ABSENT
`default` to True — ON (lib/sloppak.py: `s.get("default", True)`) — and treats an
empty/unrecognized string as ON too, so a migrated-shape pack whose `full` stem
has a missing or blank default beside instrument stems would actually play the
mixdown on open and double the song. verify was certifying that as safe.
Require an explicit normalized `off` beside instrument stems: `on`-ish values are
reported `full-stem-default-on` (actively plays), everything that is not a
normalized `off` is reported `full-stem-default-not-off`. The migrator already
writes the literal `off`, so its own output is unaffected; this also certifies
the pack is in the tool's canonical, most-portable shape. The len>1 gate is kept,
so a sole `full` stem (which IS the audio) is not policed.
Adds parametrized coverage for missing / empty / boolean / off-ish / malformed
defaults, and a sole-full-stem case. Addresses a CodeRabbit review finding.
Signed-off-by: Kris Anderson <topkoa@gmail.com>
---------
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Co-authored-by: Kris Anderson <topkoa@gmail.com>
* fix(sloppak): bound the unpack cache, and add a way to read a song without unpacking it
A tester's sloppak_cache reached 60 GB from an 1800-song library — his entire
library, unpacked, none of it played. Stems are already-compressed audio, so an
unpacked pack is ~1.1x its zip: the cache is a second, DECOMPRESSED copy of every
song it touches. It had no size cap, no LRU, and no cleanup of any kind — not even
when the song itself was deleted.
Two halves:
1. resolve_source_dir() now evicts least-recently-used songs to stay under a cap
(FEEDBACK_SLOPPAK_CACHE_MAX_MB, default 4 GB ≈ 130 songs of recency; 0 disables).
The sweep runs on unpack — the only moment the cache grows — so it can't drift.
An evicted song is dropped from _source_cache too: get_cached_source_dir() is
the only thing media.py consults before falling back, so a stale path there
would 404 every stem for the rest of the process instead of re-unpacking.
get_cached_source_dir() now also verifies the dir still exists, which makes
"just delete sloppak_cache/ to reclaim disk" safe advice.
2. read_member_bytes() reads ONE file out of a pack without unpacking it — the
same trick read_cover_bytes() uses so the library grid doesn't explode every
pack to show a cover. Unpacking a whole song to read a few KB of JSON is ~45x
write amplification; doing it in a loop over the library is what produced the
60 GB. rig_builder's library-wide tone batch is the caller that did exactly
that (fixed separately); this gives it, and everyone else, the right primitive.
Eviction is concurrency-safe: unpacks run 2-at-a-time, so a dir being written is
marked in-flight and the sweep skips it — checked and rmtree'd under one hold of
the guard, and the marker is released even if the unpack raises (a leaked marker
would make that dir permanently un-evictable).
read_member_bytes normalizes both the requested path AND the archive's stored
member names through safe_join, taking the last match — so './arrangements/x.json',
backslash members from Windows tooling, and duplicate members that normalize to
the same path all read back exactly as unpack-then-read did. Zip-slip is rejected
before anything is opened.
Tests: tests/test_sloppak_unpack_cache.py. All bite-tested (reverted each fix,
watched it fail) — including one that was passing vacuously: a freshly-unpacked
dir is the most-recently-used, so the LRU never reaches it and the in-flight race
test proved nothing until the packs were sized to force the sweep that far.
* test: split semicolon-joined statements (E702)
CodeRabbit on #950. Style only; no behaviour change.
* feat(library): effective genre falls back to MusicBrainz enrichment
Converted packs rarely carry a genres manifest key — on Byron's real
library 1188/1190 songs had no genre, starving the genre facet and the
career passport rack (2 usable genres) while song_enrichment already
held MB genres for 636 matched songs.
The effective-genre expression now resolves: per-song override → pack
genre → json_extract(enrichment.genres, '$[0]') for MATCHED rows only
(review/failed candidates could carry the wrong recording's genres).
Same fast-path gating as before: the plain indexed column is used
unless overrides or enrichment genres actually exist; stand-in DBs
without the table degrade via the OperationalError guard. Career
passports pick this up automatically through _effective_genre_expr().
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* docs: changelog names manual rows in the enrichment fallback
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
CodeRabbit follow-up on #942:
- add_play_seconds() now stamps last_played_at (like touch_position):
an unscored play that ran to the natural end WAS played — recent /
Continue ordering must see it. Resume position stays untouched.
- stats-recorder post() treats non-2xx as failure: a 4xx/5xx JSON error
body parsed as an object read as success, silently dropping the
accrued seconds instead of re-queuing them.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Career v2, WS2. Nothing measured play time before (the achievements
plugin's final-position shortcut double-counts loops and mis-reads
seeks). Now:
- stats-recorder.js accrues WALL-CLOCK seconds across song:play/resume ↔
pause/stop/ended spans (single spans clamp at 2h against suspend
inflation) and piggybacks them as `seconds` on the POSTs it already
sends; failed POSTs restore the accumulator; a session reset flushes
first so time can't re-attribute to the next song/arrangement.
- POST /api/stats accepts optional `seconds` (finite, 0 < s ≤ 6h) on the
scored and position branches, plus a new seconds-only branch for
unscored plays that ran to the natural end — banks time WITHOUT
touching the resume position (song:ended must not overwrite Continue)
and still counts as playing today for the streak.
- song_stats gains additive idempotent `seconds_total`; record_session/
touch_position accrue, new add_play_seconds() for the seconds-only
path; the legacy-encoding stats merge sums seconds across duplicates.
- Passports surface it: "14.2 h in Blues" under the badge stamp and on
the shelf cover sub-line — a true fact that only grows, never a
target or a meter (Stage 5 post-cap, per the career design).
Tests: seconds accrual/validation/seconds-only branch (stats API),
per-instrument-and-genre summing (career), fmtHours formatting (vm).
Full suites: pytest 2480, JS 1165.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Follow-ups to #829 (CodeRabbit's review nit + the consumer adoption the PR
body promised):
- freqs_to_midis: reject non-finite frequencies (NaN/Infinity) — a provider
handing one through would otherwise raise inside int(round(...)) and 500
GET /api/tunings. Tests cover nan/inf/-inf alongside the existing garbage
cases.
- v3 instrument badge: TUNING_NOTE now prefers the exact integer midis the
server serves (tuningMidis) over reconstructing the note from the lowest
string's frequency via log2 against a hardcoded 440 — which can land a
semitone off at non-440 reference pitches. Frequency path kept as the
fallback for older cached responses.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MS2YFb6UUSwJVV6CmEa25i
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
The runtime stylesheet moves to CONFIG_DIR. static/tailwind.min.css is never written again.
━━━ TWO DIFFERENT THINGS WERE SHARING ONE PATH ━━━
static/tailwind.min.css a BUILD ARTEFACT. Committed, image-baked, generated by scanning
the in-tree plugins only. CI's tailwind-fresh check verifies it.
the RUNTIME sheet PER-INSTALL STATE. Additionally scans whatever the user installed
into FEEDBACK_PLUGINS_DIR, so it differs machine to machine.
Writing the second over the first meant that MERELY RUNNING THE DEV SERVER from a git checkout
silently modified a tracked file. `git add -A` then swept a 100KB reshuffle of minified CSS
into the commit and ci/tailwind-fresh went red with a diff that explains nothing — on a PR
whose real change touched no Tailwind classes at all. It also wrote app state into the app
directory, which is read-only in some deploys.
A new route serves the runtime sheet when there is one and falls back to the committed one
otherwise. It is registered BEFORE the /static mount, which would otherwise swallow the path.
━━━ A PERSISTED SHEET MUST NOT OUTLIVE ITS REASON (Codex [P2] x2) ━━━
1. THE USER REMOVES THEIR PLUGINS. Startup only rebuilds when user plugins exist, so nothing
would ever overwrite the stale sheet — and it still carries classes for plugins that are
gone. With no user plugins the COMMITTED sheet is complete by definition. Guarded.
2. THE APP IS UPGRADED, and my first guard for this was WRONG. I compared mtimes. Codex: that
is not a freshness signal across install methods — archives and container images routinely
PRESERVE SOURCE MTIMES, so a just-shipped stylesheet can carry an OLDER timestamp than a
runtime sheet a user built days ago. The mtime check then calls the stale one FRESH and it
masks the new core CSS indefinitely — permanently, if no Tailwind toolchain is present to
trigger a rebuild.
Freshness is decided by CONTENT now. Each runtime build stamps a sidecar with the sha256 of
the committed sheet it was made from. Core ships new CSS -> that file changes -> the hash
changes -> the runtime sheet is correctly judged stale. Timestamps only gesture at the
question that hashing answers.
Falling back to the committed sheet is always safe: at worst it lacks a just-installed plugin's
classes for the seconds until the async rebuild lands.
VERIFIED END TO END. Ran the real dev server with 3 plugins installed: it rebuilt Tailwind over
them (123,291 bytes), wrote the sheet + sidecar to CONFIG_DIR, still served /static/
tailwind.min.css at 200 — and `git diff` on the tracked file came back CLEAN.
8 tests. Bite-tested: reverting to the shared path fails 3, dropping the staleness guards fails
2 more.
pytest 2425, pyflakes 0, Codex 0.
Closes#911
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The guard in startup_events() read:
if getenv_compat("FEEDBACK_DEMO_MODE") or getenv_compat("FEEDBACK_DEMO_MODE") == "1" \
and not _DEMO_JANITOR_STARTED:
`and` binds tighter than `or`, so that is `A or (B and C)`. The not-already-started half
never ran when the env var was truthy — the only case that reaches it at all. A second
startup started a SECOND janitor thread, overwrote the handle, and shutdown then joined
only the last: the first leaked and kept firing registered hooks hourly, forever.
The guard now lives INSIDE start_janitor(). A caller cannot get operator precedence wrong
if there is nothing left for it to get wrong.
━━━ THREE WAYS TO WRITE THIS GUARD WRONG. I HIT ALL THREE. ━━━
1. NO GUARD — the original bug. Double-start, orphaned thread.
2. GUARD ON THE FLAG (`if _DEMO_JANITOR_STARTED: return`). Codex [P2]. stop_janitor()
DELIBERATELY leaves that flag True when a hook outruns its join timeout, so that a later
startup cannot spawn a janitor beside a live one. But the hook usually finishes a moment
later: the thread exits and the flag is stale. A flag-keyed guard then refuses to start a
replacement for the rest of the process — demo cleanup silently dead. (The original bug
accidentally MASKED this by always starting.)
3. GUARD ON LIVENESS ALONE (`if thread.is_alive(): return`). Codex [P2], second pass. A
timed-out stop leaves the old thread ALIVE BUT DOOMED — its stop event is set and it
exits as soon as its current hook returns. Treating that as a running janitor skips the
replacement, and we are back at (2) a second later.
So: a janitor counts as running only if its thread is alive AND it has not been told to stop.
━━━ AND EACH JANITOR NOW OWNS ITS STOP EVENT ━━━
start_janitor() used to `_DEMO_JANITOR_STOP.clear()` a single SHARED Event. Start a
replacement while a doomed thread is still finishing a hook and that clear RESURRECTS it: it
loops back to wait(), sees the flag cleared, and carries on. Two janitors — the exact bug we
started from. A fresh Event per janitor makes it impossible; the old thread waits on its own
event, which stays set, so it can only exit.
Env semantics UNCHANGED, verified across every value ("", "1", "0", "true", "false", "off"):
the old expression and demo_mode_enabled() agree on all of them. The only behavioural change
is the idempotency fix.
FOUR tests, and each of the three wrong guards fails a different subset:
no guard -> 2 fail (double start; orphaned thread)
guard on the flag -> 2 fail (never restarts after a timed-out stop)
liveness alone -> 1 fail (no replacement for a doomed janitor)
liveness + not-stopping -> all pass
pytest 2416, pyflakes 0, Codex 0.
Closes#902
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
lib/demo_mode.py (342). server.py 1,870 -> 1,649.
The read-only request guard (its 96-entry blocked-route table + the middleware) and the
hourly session janitor (registry, hook runner, thread). Bodies VERBATIM.
THE MIDDLEWARE NEEDS `app`, SO THE MODULE TAKES IT. _demo_mode_guard is an
@app.middleware("http") and cannot exist without an app object. Rather than have a module
under lib/ reach for a global, it exposes install(app) and server.py — which owns the app —
hands it over. The janitor is symmetrical: start_janitor() / stop_janitor(), called from
server.py's startup and shutdown hooks, where the process lifecycle actually lives.
register_demo_janitor_hook IS PART OF THE PLUGIN CONTRACT. It is a key in plugin_context,
so plugins hold it as a LIVE REFERENCE from setup(). server.py imports this exact object
and puts it in the dict unchanged — identity preserved, and
tests/test_plugin_context_contract.py (#898, merged) fails if that ever stops being true.
This is the first carve that guard has actually protected.
━━━ stop_janitor()'s ORDER IS LOAD-BEARING ━━━
The obvious way to write it — clear the "started" flag, then join — is WRONG, and I wrote
it that way first. server.py's original deliberately returns EARLY, leaving
_DEMO_JANITOR_STARTED True and the thread handle intact, when the thread outlives the join:
# Leave _DEMO_JANITOR_STARTED True so a new janitor is not
# spawned by a subsequent startup while the old one is alive.
Clearing the flag first quietly reintroduces exactly the double-janitor leak the flag
exists to prevent. Preserved byte-for-byte, and the reason is now written down at the
function rather than only at its single call site.
━━━ A BUG MOVED VERBATIM, ON PURPOSE ━━━
if getenv_compat("FEEDBACK_DEMO_MODE") or getenv_compat("FEEDBACK_DEMO_MODE") == "1" \
and not _DEMO_JANITOR_STARTED:
`and` binds tighter than `or`, so this is `A or (B and C)` — the not-already-started
re-entry guard is DEAD whenever the env var is truthy, which is the only case that runs.
A second startup leaks a janitor thread (the handle is overwritten, so shutdown joins only
the last). Verified. Preserved exactly and filed as issue #902: a carve whose whole value
is being provably behaviour-neutral is not the place to change behaviour.
pyflakes caught three more missing imports on the way in (uuid, warnings x2). Five carves,
ten missing imports, every one a NameError on a live path.
pytest 2399, pyflakes 0, Codex 0.
Refs #48
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
lib/scan.py (326). server.py 2,098 -> 1,870.
The background scan, its spawn ProcessPoolExecutor, and the kick/runner plumbing that
serialises passes. Bodies VERBATIM except the seam reads.
Everything shared is read LATE off appstate — the same contract every module in
lib/routers/ uses, and it is not cosmetic: tests monkeypatch CONFIG_DIR and swap meta_db,
so a value captured at import time pins the wrong one for the life of the process.
CONFIG_DIR -> appstate.config_dir
meta_db -> appstate.meta_db
_default_settings -> appstate.default_settings()
_stat_for_cache -> appstate.stat_for_cache()
━━━ THE SCAN STATUS IS REBOUND, NOT MUTATED ━━━
_background_scan does `global _scan_status; _scan_status = {**INIT, ...}` at every stage
transition. It REPLACES the dict; it never updates it in place. So nothing may hold that
dict by value — a reference captured once goes permanently stale at the first stage change
and would report "listing" forever while the scan ran to completion.
Hence `scan.status()`, a getter, and hence appstate publishes scan_status as a CALLABLE.
appstate.py already said so in a comment; this is the code that makes it true. (Same for
the plugin_context entry, which was already `lambda: dict(_scan_status)` — late-bound, so
it survives the move unchanged. The contract test from #898 covers it.)
━━━ appstate.server_root: A TRAP CLOSED PERMANENTLY ━━━
_background_scan seeds the builtin content, which needs the directory holding server.py.
`Path(__file__).resolve().parent` is correct in server.py and silently WRONG anywhere under
lib/ — it yields lib/, which holds no docs/ or data/ — and it fails by finding NOTHING
rather than by raising, so the seeds would just quietly never run.
lib/builtin_content.py (#900) closed that by taking the root as a parameter. This adds the
other half: server.py publishes it ONCE as appstate.server_root, so no module under lib/
ever has a reason to derive it. Documented at the slot.
pyflakes caught two more missing imports on the way in (loosefolder_mod, enrichment) —
each a NameError on a live scan path, and the suite would have handed them over one failure
at a time. It stays part of every server.py slice.
TESTS. The two scan fixtures (test_settings_api::scan_module,
test_feedpak_extension::scan_server) patched server._make_scan_executor to swap the spawn
pool for an in-process ThreadPool; they now patch it on lib/scan.py. Worth noting WHY that
still works: the fixtures re-import `server` per test, but `scan` stays cached in
sys.modules — and it picks up the fresh CONFIG_DIR anyway, because the appstate reads are
late-bound. The seam is doing exactly the job it was built for.
━━━ TEST ISOLATION: A REGRESSION THE CARVE ITSELF CREATED (Codex [P2]) ━━━
background_scan() deliberately NEVER sets running=False — ownership of that flag lives in
_scan_runner, so a kick_scan() racing the terminal write cannot observe a stale False and
start a second runner. Correct in production.
But the scan fixtures call background_scan() DIRECTLY, skipping the runner. That was
harmless while the state lived on `server`, which the fixtures RE-IMPORT per test. It is
NOT harmless now: `scan` stays cached in sys.modules across sys.modules.pop("server"), so
the status dict OUTLIVES the test. One direct call leaves the shared scanner marked
"running" forever, and every later scan or rescan returns "already in progress" and quietly
does nothing.
Verified: after a direct call, kick_scan() returns False and starts no scan at all.
The suite passed anyway, on ordering luck — which is exactly how this class of bug ships.
tests/conftest.py::reset_scan_state now snapshots and restores lib/scan.py's module state
around the two fixtures that drive it directly.
pytest 2398, pyflakes 0, Codex 0.
Refs #48
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
lib/builtin_content.py (321 lines moved). server.py 2,418 -> 2,098.
The calibration/diagnostic sloppaks and the starter library: _copy_builtin_packs,
_write_builtin_pack, the two seed helpers, their source tables, and the seed marker.
━━━ THE ONE SIGNATURE CHANGE, AND WHY THE CARVE IS UNSAFE WITHOUT IT ━━━
server.py has:
def _feedBack_server_root() -> Path:
return Path(__file__).resolve().parent
That is correct IN server.py: the repo root in dev, resources/feedBack when bundled — the
tree that actually holds docs/ and data/.
Move that body into lib/ unchanged and it keeps working, silently, and returns lib/. There
is no docs/diagnostics under lib/, so every seed would find nothing, log "source missing"
at debug, and return. Nothing raises. Nothing fails. The starter library simply never
appears, and the calibration sloppak is never seeded — on a fresh install, in the field.
A verbatim move whose MEANING changed because __file__ did.
So this module cannot compute a root: `server_root` is a PARAMETER, and server.py — the
only place that legitimately knows where it lives — passes it in. The trap is now
structurally impossible rather than merely avoided. (_copy_builtin_packs already took the
root that way; the two seed helpers now do too.)
Everything else is byte-identical. CONFIG_DIR is read late as appstate.config_dir and the
DLC root through dlc_paths._get_dlc_dir — the same seam every router in lib/routers/ uses,
late-bound because tests monkeypatch it.
━━━ PYFLAKES FOUND THREE MISSING IMPORTS THE TESTS WOULD HAVE FOUND ONE AT A TIME ━━━
The moved code uses `secrets`, `stat` and `tempfile`; none was in my import block. Each is
a NameError on a live path. `python3 -m pyflakes` names all three in one shot — this is the
Python twin of the no-undef gate that guarded every frontend carve, and it should run on
every server.py slice from here.
It also flagged a PRE-EXISTING one I deliberately did not touch: server.py's
TuningProviderRegistry.get_merged() calls `logger.exception(...)` in an except handler and
there is no `logger` in the module (it is `log`). So a raising tuning provider takes down
the merged-tunings call for everyone, with a NameError naming the wrong problem. Filed as
issue #899 rather than smuggled into a carve whose whole value is being behaviour-neutral.
The constants lost their underscore prefix: they cross a module boundary now (the seed
tests read them), so `_BUILTIN_STARTER_SOURCES` was a lie.
pytest 2397, pyflakes 0, Codex 0. Guarded by the plugin_context contract test (#898).
Refs #48
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The artist page + external-links payload (/api/artist/{name}/page, /links,
/links/refresh) plus their exclusive helpers (_artist_links_payload,
_artist_links_from_mb, the URL-slot table) move to lib/routers/artist.py. Bodies
verbatim except @app->@router and the seam reads (meta_db->appstate.meta_db,
CONFIG_DIR->appstate.config_dir, _default_settings->appstate.default_settings).
MB link enrichment is reached as enrichment.X; the URL-safety validator is
imported from lib/library_registry.py. No new seams.
server.py: 2,507 -> 2,413 (-94).
Verified: pyflakes clean; route set unchanged (143); full pytest 2395 passed.
eslint 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Song audio (/audio/{f}), the local-audio-path resolver (/api/audio-local-path),
and raw sloppak-member serving (/api/sloppak/{f}/file/{rel}) — plus the shared
_resolve_sloppak_local_file helper — move to lib/routers/media.py. Bodies verbatim
except @app->@router and the cache/static path seams (AUDIO_CACHE_DIR->
appstate.audio_cache_dir, STATIC_DIR->appstate.static_dir, SLOPPAK_CACHE_DIR->
appstate.sloppak_cache_dir — all already in the seam). No new slots.
The two test fixtures that redirect STATIC_DIR to a temp dir now also patch
appstate.static_dir (the moved routes read the seam, not server's global).
server.py: 2,638 -> 2,507 (-131).
Verified: pyflakes clean; route set identical (143), all unique-path (no
catch-all, no shadowing); full pytest 2395 passed (the audio-local-path +
sloppak-file-traversal cases). eslint 0. Boot smoke: /audio, /api/sloppak/{}/file
serve 404 for unknown, 0 tracebacks.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The 14 /api/enrichment/* routes — status, kick/cancel, per-song state, the
Match-Review queue (accept/reject/pick/search/rematch/refresh/states), and the
AcoustID fingerprint identify endpoints — plus the route-exclusive candidate
sanitizer move to lib/routers/enrichment.py. Bodies verbatim except @app->@router
and the seam reads (meta_db->appstate.meta_db, CONFIG_DIR->appstate.config_dir).
The enrichment engine (transport, matcher, worker, upload caps) already lives in
lib/enrichment.py from the earlier subsystem move and is reached as enrichment.X.
No new seams.
server.py: 2,925 -> 2,638 (-287).
Verified: pyflakes clean; route set identical (143); full pytest 2396 passed (the
enrichment route + Match-Review + identify cases, which fake the network on the
enrichment module). eslint 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The library query surface (songs/albums/artists/stats/genres/tuning-names/
practice-suggestions), the provider list/art/sync endpoints, and collection CRUD
move to lib/routers/library.py. The registry itself — LibraryProviderRegistry,
LocalLibraryProvider, SmartCollectionProvider, the collection-provider lifecycle
(_sync/_unregister_collection_provider), and the shared query/collection helpers
(_library_filter_args, _split_csv, _sanitize_collection_rules, _safe_art_redirect_url,
the filter-key sets) — moves to lib/library_registry.py.
The PLUGIN CONTRACT is untouched: server.py still constructs the singleton
(LocalLibraryProvider needs meta_db), still exposes register_library_provider /
unregister_library_provider to plugins via plugin_context (with the per-plugin
ownership scoping in plugins/__init__.py), and injects the registry + local
provider into appstate. The router reads appstate.library_providers /
appstate.local_library_provider at call time; the provider classes are duck-typed
so no plugin imports a base class. Acyclic: library_registry imports
routers.art (for LocalLibraryProvider.get_art) + appstate, never server.
server.py: 3,692 -> 2,925 (-767).
Verified: pyflakes clean; ORDERED route table identical set (library block mounts
at one site — all exact/specific-path, no catch-all, no shadowing); full pytest
2397 passed (incl the plugin register/unregister + collection-as-provider tests).
eslint 0. Boot smoke: /api/library + providers list "local"; a created collection
surfaces as a `collection:N` provider through the seam; collection CRUD clean.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Upload/delete, the catalog-metadata write-back, user-meta, overrides, gap-fill,
and the per-song info payload (11 routes) move to lib/routers/song.py with their
exclusive helpers (the atomic upload commit + the song-IO lock, the upload caps,
the gap-fill proposal builders). Bodies verbatim except @app->@router and the
seam reads: meta_db->appstate.meta_db, art_override_paths->appstate.art_override_paths,
and the scan/ingest helpers that stay in server.py (the scan lifecycle owns them)
-> new appstate seam callables: kick_scan, invalidate_song_caches, stat_for_cache,
and scan_status() (a getter — the underlying dict is reassigned). The gap-fill
MBID/ISRC regexes are reached as enrichment.X; _MULTIPART_OVERHEAD_SLACK (shared
with the staying AcoustID-identify route) moves to lib/enrichment.py beside
_ACOUSTID_MAX_UPLOAD_BYTES.
ROUTE ORDER: song_router mounts AFTER art_router — get_song_info's catch-all
`/api/song/{filename:path}` would otherwise shadow `/api/song/{path}/art*`
(Starlette matches first-registered; the :path converter is greedy).
server.py: 4,478 -> 3,692 (-786).
Verified: pyflakes clean; ORDERED route table preserves the specific-before-catch-all
invariant; full pytest 2397 passed (incl the art/cover 304 + CAA-fetch tests that
caught the shadowing before it was fixed). eslint 0.
BEHAVIORAL — needs an on-device pass (upload a sloppak, edit metadata write-back,
delete a song) before merge.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GET/POST /api/settings, /api/settings/reset, and the two-phase atomic
export/import bundle (/api/settings/export|import) move to lib/routers/settings.py
with their exclusive helpers (the relpath allowlist validator, the atomic writer,
the library-DB snapshot + sqlite integrity gate, the config-type validator, the
bundle schema). Bodies verbatim except @app->@router and the seam reads:
meta_db->appstate.meta_db, CONFIG_DIR->appstate.config_dir,
_running_version->appstate.running_version(), and _default_settings->
appstate.default_settings (the canonical defaults builder stays in server.py —
the scan + artist-links code share it — and is injected as a new seam callable).
server.py: 5,539 -> 4,478 (-1,061).
Verified: pyflakes clean (bar the pre-existing File/safe_join/tuning_name/ET);
route table IDENTICAL (143); full pytest 2397 passed (154 settings cases incl the
export→import round-trip + library-DB snapshot/restore + relpath-allowlist SSRF/
traversal guards, retargeted onto the settings module). eslint 0.
BEHAVIORAL — needs an on-device settings export→import round-trip sign-off before
merge (do not merge on green CI alone).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The six song-art routes — GET /api/song/{f}/art, .../art/cover-search,
.../art/candidates, POST .../art/upload, .../art/url, DELETE /api/art/{f}/override
— plus their exclusive helpers (the ETag/304 response machinery, _save_art_override,
_url_host_is_internal, _fetch_art_url + the art size/redirect caps) move to
lib/routers/art.py. Bodies verbatim except @app->@router and the seam reads:
meta_db->appstate.meta_db, ART_CACHE_DIR->appstate.art_cache_dir, and the three
shared art helpers that stay in server.py (used by the song/delete routes too)
-> appstate.<callable> (_song_pack_art_exists, _art_override_paths — already
seam-injected for the enrichment worker — plus a new art_safe_name slot). The
CAA / release-search transport lives in lib/enrichment.py and is reached as
enrichment.X. LocalLibraryProvider.get_art now calls art_router.get_song_art.
server.py: 5,988 -> 5,540 (-448).
Verified: pyflakes clean (bar the pre-existing File/safe_join/tuning_name/ET);
route table IDENTICAL (143); full pytest 2399 passed (33 art serve/candidates/
override/url cases incl the SSRF-guard _url_host_is_internal + _fetch_art_url
size-cap tests, retargeted onto the art module); test_packaging 43; eslint 0.
Boot smoke: /art 404, /art/candidates 404, DELETE /override 200 from the router.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(server): move the metadata-enrichment subsystem into lib/enrichment.py (R3)
MusicBrainz / Cover-Art-Archive / AcoustID transport, the match-scorer glue, and
the background enrichment worker (~930 lines, 61 defs) leave server.py as one
cohesive unit. Bodies are verbatim; the only changes are seam reads:
meta_db / config_dir / sloppak_cache_dir / art_cache_dir -> appstate.<slot>
_song_pack_art_exists / _art_override_paths (stay in server.py for the art +
delete routes) -> appstate.<callable> (new seam slots, injected by reference)
_env_flag -> env_compat.env_flag_compat (the existing identical helper)
_artist_title_from_filename -> imported from metadata_db (its home)
the User-Agent VERSION lookup: Path(__file__).parent ->
Path(__file__).resolve().parents[1] (lib/enrichment.py -> app root)
server.py drives the worker through the module (import enrichment; the routes +
scan lifecycle call enrichment.X). Tests that faked the network on `server`
(_mb_http_get, _enrich_network_enabled, _caa_http_get, ...) now patch the same
names on `enrichment` — the module attribute is resolved at call time, so one
setattr reaches both the routes and the worker's internal callers. Acyclic:
enrichment imports appstate/appconfig/dlc_paths/metadata_db/mb_match/
acoustid_match/sloppak/loosefolder, never server.
server.py: 6,917 -> 5,988 (-929).
Verified: pyflakes clean (bar the pre-existing File/safe_join/tuning_name/ET);
route table IDENTICAL (143); full pytest 2400 passed (140 enrichment/art cases
incl the offline-safety + transport-error-pauses-pass contracts that fake the
network); test_packaging 44 passed (enrichment.py resolves under lib/); eslint 0.
Boot smoke: /api/enrichment/status + POST /kick serve from the new module.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test: reset enrichment worker state between tests (CodeRabbit)
lib/enrichment.py now owns the worker, and it stays imported for the whole
session while the `server` fixtures pop-and-reimport `server` — so the cancel
Event / status dict / caches would leak across tests, and a stale `_enrich_cancel`
could short-circuit a later direct `_background_enrich()`. An autouse conftest
fixture clears that process-global state before each test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test: tighten the enrichment-reset fixture (CodeRabbit)
Narrow the import guard to ImportError (not blind Exception, BLE001), and stop
clearing _caa_index_locks — it's guarded by _caa_index_locks_guard, so an
unlocked clear() would race a still-alive worker, and its per-release mutexes
carry no test state.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The merged-tuning-catalog route moves to lib/routers/tunings.py, verbatim except
@app->@router, CONFIG_DIR->appstate.config_dir, and the two seam substrates it
needed:
- lib/appconfig.py — the pure config.json reader `_load_config` (used by ~11
server sites + future config-reading routers). server.py re-imports it, so
those call sites and any `server._load_config` test reference are unchanged.
- appstate.tuning_providers — the TuningProviderRegistry instance injected by
reference (a stable object mutated in place via register()/unregister()), so
the router reads the same registry plugins populate through plugin_context.
The instance stays defined in server.py, so `server.tuning_providers` still
resolves — zero test retargets.
The tuning constants (DEFAULT_REFERENCE_PITCH, TUNING_PRESET_MIDIS,
freqs_to_midis) already live in lib/tunings.py and are imported directly.
server.py: 6,960 -> 6,917.
Verified: pyflakes clean (bar the pre-existing unused `tuning_name` import);
route table IDENTICAL (143); full pytest 2400 passed (110 tuning/config cases);
eslint 0. Boot smoke: /api/tunings serves referencePitch + tunings + tuningMidis.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The three /api/diagnostics/* routes (export, preview, hardware) plus their
exclusive payload-cap helpers and the `_diag_*` normalisers. Bodies verbatim
except @app->@router, CONFIG_DIR->appstate.config_dir, _running_version()->
appstate.running_version() (a new seam slot; the impl stays in server.py where
the settings region also calls it), and the builtin-plugins lookup in
_diag_plugins_roots: Path(__file__).parent -> Path(__file__).resolve().parents[2]
(routers -> lib -> app root; plugins/ ships at the app root in every packaging
path).
The pure caps/normalisers (_diag_cap_console/_dict/_contributions,
_diag_coerce_bool, _diag_normalize_include, _DIAG_MAX_*) are re-exported from
server.py so the existing `server._diag_*` / `server._DIAG_*` tests keep
resolving — none of them monkeypatch these, so no test retargets.
server.py: 7,216 -> 6,960.
Verified: pyflakes clean (bar the intentional re-export lines); route table
IDENTICAL (143); full pytest 2399 passed (77 diag/packaging + 122 diagnostic-
matched cases incl the cap/coerce/normalize suites); eslint 0; Codex pending.
Boot smoke: /hardware, /preview, and POST /export (200 application/zip) all
serve from the new router location.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GET /api/version + its exclusive _safe_http_url URL validator. Bodies verbatim
except @app -> @router and the VERSION-file lookup: Path(__file__).parent (the
app root when this lived at the top level) -> Path(__file__).resolve().parents[2]
(routers -> lib -> app root). VERSION ships at the app root in every packaging
path (Dockerfile COPY VERSION /app/, desktop bundle).
server.py: 7,275 -> 7,214.
Verified: pyflakes clean; route table IDENTICAL (143); pytest 2401 passed (33 in
test_version_endpoint, incl the URL-validation + env-override cases); eslint 0.
Boot smoke: /api/version returns the real version + validated source/license URLs.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The last of the progression cluster. XP award (1 route) + per-song practice stats
(record/recent/best/top/per-song, 5 routes) — meta_db-only apart from the two
seam accessors record_stats uses (get_progression_content, builtin_diagnostic_
filename, both landed by shop/progression). Bodies verbatim; @app -> @router,
meta_db -> appstate.meta_db, _as_int from metadata_db, _clean_str from reqfields.
Three scattered source blocks (xp, the stats block, and the separated
/api/stats/{filename:path} which the /api/library/practice-suggestions route
splits off) are assembled into one module and mounted once. Registration order
is preserved WHERE IT MATTERS: the /api/stats/{filename:path} catch-all is
assembled LAST inside the router, so it still can't shadow the fixed /recent
/best /top paths — verified against the live route table (recent/best/top all
precede the catch-all) and the route SET is identical to origin/main (143).
server.py: 7,478 -> 7,275.
Verified: pyflakes clean; route set identical + catch-all-last; pytest 2401
passed (113 across song_stats/profile/progression); eslint 0. Boot smoke:
/stats/recent /best /top all 200 (not shadowed), xp/award 200.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
4 routes (overview/add-paths/onboarding/events, spec 010) + their EXCLUSIVE
helpers (_goal_ui_progress, _progression_overview 112L) + the
_PROGRESSION_EVENT_TYPES whitelist. Bodies verbatim; @app -> @router,
meta_db -> appstate.meta_db, _clean_str from reqfields.
The two SHARED server accessors read through the seam: get_progression_content
(added for shop #851) and builtin_diagnostic_filename (new slot — a trivial
const-returning fn shared with the stats router's api_record_stats). Both are
injected via the second appstate.configure() after their defs (the import-top
configure runs before them). The cache + fns stay in server.py, so
test_progression_api's server._progression_content patch is untouched — 0 retarget.
server.py: 7,798 -> 7,594.
Verified: pyflakes clean; route table IDENTICAL (143); both seam accessors wired;
pytest 2401 passed (63 in test_progression_api); eslint 0. Boot smoke: GET
/api/progression 200 (drives _progression_overview + both accessors), events
400 on bad body.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The progression-content substrate: `_get_progression_content` (a lazy,
double-checked-locking content cache) is now published into the appstate seam as
a CALLABLE. The cache global + lock + the function stay in server.py (startup
uses it, and test_progression_api patches `server._progression_content`
directly), so ZERO test retargeting — routers just call
`appstate.get_progression_content()`.
Because the accessor is defined at server.py:1152 but the import-top configure()
runs at :346, a second `appstate.configure(get_progression_content=...)` publishes
it right after the def (configure is idempotent/additive).
First consumer: routers/shop.py (3 routes: buy/equip/list). Bodies verbatim;
@app -> @router, meta_db -> appstate.meta_db, _clean_str from reqfields,
_get_progression_content() -> appstate.get_progression_content(). This unblocks
stats/progression/profile next (all share the accessor).
server.py: 7,880 -> 7,845.
Verified: pyflakes clean; route table IDENTICAL (143); pytest 2401 passed
(test_progression_api's server._progression_content patch still works via the
kept cache); packaging guard; eslint 0. Boot smoke: GET /api/shop 200 (drives
appstate.get_progression_content), buy 400 on bad body.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
work keeper-chart prefs (3) + favorites toggle + tags list + saved toggle +
session/continue — 7 routes across 5 domains, all meta_db-only (0 setattr, 0
helpers). Bodies verbatim; @app -> @router, meta_db -> appstate.meta_db,
_clean_str from reqfields.
These were scattered singletons with no natural neighbor, so they're grouped as
"small library/user-state endpoints" and mounted once. All paths are distinct
and non-overlapping, so registering them together doesn't change routing:
verified the route SET is identical to origin/main (143) AND that no moved path
shadows or is shadowed by another (order-independence check).
server.py: 7,880 -> 7,833.
Verified: pyflakes clean; route set identical + order-independent; pytest 2401
passed; packaging guard 45; eslint 0. Boot smoke: tags/session GET 200,
favorites/saved toggle (400 on missing filename), work/charts 200.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Unblocked by the DLC-path substrate (#843): chart's only server-module deps are
now app + meta_db (both seam); _get_dlc_dir/_resolve_dlc_path come from dlc_paths,
sloppak/loose detection from the shared lib modules. 4 routes (split/unsplit/
work/fileinfo), meta_db-only otherwise. 0 setattr targets, 0 helpers to relocate.
Bodies verbatim; @app -> @router, meta_db -> appstate.meta_db. include_router at
the original site; 143-route table identical to origin/main. No test retargeting.
server.py: 8,003 -> 7,909.
Verified: pyflakes clean on the router; no new undefined/dead in server.py; route
table identical; pytest 2401 passed (74 across work_charts/context_menu/
group_filter/packaging); eslint 0. Boot smoke: chart/work 200, chart/fileinfo
resolves the real pack path through _resolve_dlc_path.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(ws_highway): swallow a mid-stream WebSocketDisconnect quietly
Pre-existing (byte-identical to origin/main), flagged by CodeRabbit on #844: the
outer try's only guard was `except Exception as e: log.exception("highway_ws
unhandled error")`. The inner `except WebSocketDisconnect` covers ONLY the
post-`ready` keep-alive loop, and the drum_tab/notation sends have their own
localized guards — but the ~13 other streamed sends (beats, sections, notes,
chords, anchors, …) fall through to the blanket handler. Since
WebSocketDisconnect is an Exception subclass, a routine tab-close mid-load was
logged as an error at whichever send was in flight.
Fix: a dedicated `except WebSocketDisconnect: return` before the blanket handler,
matching the two localized guards and the lib/ coding guideline.
tests/test_ws_highway_disconnect.py drives highway_ws with a fake websocket that
raises WebSocketDisconnect on the first streamed send (`loading`), over a real
minimal sloppak. Negative-checked: removing the guard makes it fail (the
disconnect is logged as `highway_ws unhandled error`); the fix passes. Uses a
raw handler on `feedBack.server` rather than caplog, since configure_logging()
reroutes that logger through structlog where caplog doesn't observe it.
Full suite green. Closes the review thread on #844.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test(ws_highway): assert streaming stops after disconnect (exactly one send) — CodeRabbit
The stub now raises only on the first send and the test asserts ws.sends == 1,
so a handler that caught the disconnect and kept streaming would fail. Still
negative-checked: removing the guard fails the test.
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The single largest handler in server.py — the 902-line /ws/highway/{filename}
chart streamer — plus its 3 exclusive helpers (_pick_smart_arrangement,
_sanitize_authors, _sanitized_song_offset). server.py: 9,008 -> 8,003 (-1,005,
the biggest single R3 cut).
Clean move despite the size: the handler's only server-module deps are the 4
path constants + 3 exclusive helpers + app + log. Everything else it uses is
either NESTED inside the handler (_evict_audio_cache, _fill_scale_degree,
_manifest_entries, _tone_names, _xml_rank, _send_keepalives) or imported from
the shared lib modules (song/audio/sloppak/drums/notation/dlc_paths/metadata_db).
- Path constants read through the appstate seam: added static_dir /
sloppak_cache_dir / audio_cache_dir slots (config_dir already there);
server.py configures them. Bodies otherwise verbatim (@app.websocket ->
@router.websocket, PATHS -> appstate.*, log -> module logger).
- sloppak_cache_dir IS setattr-patched, so the 3 test_highway_ws_* suites now
also `setattr(appstate, "sloppak_cache_dir", ...)` next to their existing
server patch. _sanitize_authors unit tests import it from routers.ws_highway
(it moved). No other test churn.
- Removed 23 now-dead imports from server.py (song/audio/drums/notation/
bisect/contextvars/structlog/WebSocket*/_arr_smart_sort_key) — diffed against
the origin/main unused-import baseline so only NEWLY-dead ones went.
owns_tmp (assigned, never read) moved verbatim — it's pre-existing dead on
origin/main too; left as-is to keep the move faithful.
Verified: route table identical to origin/main (143, paths/methods/order);
handler body verbatim spot-checked; pyflakes clean (server has no new
undefined/dead); pytest 2400 passed; packaging guard 53; eslint 0. Boot smoke:
the highway WS streams the full chart (song_info/beats/sections/notes/chords/
notation/anchors/drum_tab -> ready) BYTE-for-byte the same message sequence as
origin/main across arrangements 0/1/2, zero tracebacks.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The keystone for the path-dependent routers (ws_highway, song, audio-local-path,
sloppak): the "where do the song files live + safe containment" layer leaves
server.py so a router can reach it.
- `_resolve_dlc_path` (pure — args only) moves verbatim.
- `_get_dlc_dir` reads the env-derived paths through the appstate seam
(appstate.dlc_dir/dlc_dir_env/config_dir) instead of module globals, so
lib/dlc_paths.py does no import-time IO.
- appstate gains `dlc_dir`/`dlc_dir_env` slots (config_dir already there);
server.py configures them. All three are env-derived, so a setenv+reimport
fixture reconfigures them for free — ZERO setattr retargeting.
- server.py RE-EXPORTS both (`from dlc_paths import _get_dlc_dir,
_resolve_dlc_path`), so its 24+16 call sites AND the tests that reach
`server._get_dlc_dir()` / `server._resolve_dlc_path()` directly
(test_dlc_junction, test_highway_ws_*) resolve unchanged — no test edits.
server.py: 9,085 -> 9,008.
Verified: _resolve_dlc_path byte-identical to origin/main; _get_dlc_dir's only
change is the three path identifiers -> appstate.*; pyflakes clean; route table
identical (143); pytest 2407 passed (test_dlc_junction + both highway_ws suites
green via re-export); packaging guard 52; eslint 0. Boot smoke: library scans
(8 songs, _get_dlc_dir), a real song resolves + serves art (_resolve_dlc_path),
and the highway WS reaches `ready` with notes.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two pre-existing cover-upload issues CodeRabbit flagged on #841 (verbatim move,
so correctly not fixed there):
1. One `except Exception as e` wrapped BOTH the PIL decode and the img.save/
tmp.replace, returned 400 for both, and echoed `e` — so a disk/permission
failure was mislabeled a client error and could leak a filesystem path. Now:
decode/validation -> 400 "Invalid image" (generic); save/replace failure ->
logged 500 "could not save cover" (no detail).
2. A shared `{pid}.png.tmp` let two concurrent uploads clobber each other's temp
file. Now a unique `tempfile.mkstemp` in the cover dir, atomic replace to
publish. Plus an existence re-check just before publishing so an upload that
raced a playlist delete can't leave an orphan cover.
mkstemp itself is INSIDE the try (Codex catch): an unwritable dir / full disk
raises there and is the same persistence failure as save/replace, so it hits the
logged generic-500 path instead of escaping as an unhandled 500. Cleanup guards
`tmp is not None` for the mkstemp-failed case.
Did NOT add a full per-playlist critical section (CodeRabbit's "heavy lift"):
FeedBack is single-user (Principle I), so a cover upload racing a delete on the
same id can't happen — documented in the code rather than building a lock
framework for a precluded race.
tests/test_playlist_cover_errors.py pins all of it; negative-checked three ways:
the old single-except-400 shape fails the 500 + no-leak-400 tests, and moving
mkstemp back outside the try fails the temp-creation-500 test. Fix passes 5/5.
Full suite 2405 passed; boot smoke: valid cover 200, bad image -> generic
"Invalid image" 400, no .tmp litter.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The biggest router yet — 12 playlist routes + custom covers — and the first that
needs the config-path seam. server.py: 9,302 -> 9,085 (-217).
Three causally-linked pieces, all required for playlists:
- `config_dir` joins the appstate seam (the plan always put the path constants
there; deferred in S3, needed now). It's env-derived, so the ~49
pop-and-reimport fixtures reconfigure it for free — ZERO setattr retargeting.
STATIC_DIR/SLOPPAK_CACHE_DIR (patched via setattr) stay in server.py until a
router that reads them is extracted, and get retargeted then.
- `_clean_str` (pure request-field sanitizer, 14 callers) -> lib/reqfields.py;
server.py imports it back. Unblocks wanted/saved/collections/profile/... later.
- routers/playlists.py: bodies verbatim, `@app`->`@router`, `meta_db`->
`appstate.meta_db`, `CONFIG_DIR`->`appstate.config_dir`, `_clean_str` from
reqfields, `_ART_CACHE_HEADERS` as a local const (art keeps server.py's).
The two exclusive cover helpers (_playlist_cover_path/_url) move with it.
include_router at the original site; full 143-route table identical to
origin/main. One test retarget: test_playlists_api called
`server._playlist_cover_path` directly -> now imports it from routers.playlists
(reads appstate.config_dir, which the `server` fixture configures).
Verified: pyflakes clean; route table identical; pytest 2401 passed (28 in
playlists+collections+appstate); packaging guard 51 (auto-picked up reqfields);
eslint 0; boot smoke drives create/rename/add-song/cover-upload/serve/delete —
the cover writes 1.png under CONFIG_DIR THROUGH appstate.config_dir and serves
200 with an mtime cache-bust token; a wrong-typed name field still 400s via
_clean_str; demo untouched.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two pre-existing races in the loops routes, flagged by CodeRabbit on #839 (the
verbatim extraction moved them unchanged from server.py, so they correctly
weren't fixed there):
1. save_loop computed `COUNT(*)` OUTSIDE meta_db._lock, then inserted inside it.
Two simultaneous unnamed POSTs read the same count and both mint "Loop N".
Fix: one lock scope around COUNT + INSERT.
2. list_loops read the shared single connection (check_same_thread=False) with
no lock, so it could overlap a POST/DELETE commit. Fix: read under the lock,
like every writer.
Low severity in context — FeedBack is single-user (Principle I), so concurrent
unnamed-loop POSTs essentially can't happen — but each fix is one lock scope.
tests/test_loops_concurrency.py pins both with a threading.Barrier that releases
16 workers into save_loop at once. Negative-checked: reverting the COUNT back
outside the lock fails the uniqueness assertion 5/5 runs; the fix passes 3/3.
pytest 2400 passed; on-device two unnamed POSTs -> ['Loop 1','Loop 2'].
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Third router. Practice loops (saved A/B regions per song): GET/POST/DELETE
/api/loops, meta_db-only (0 setattr targets, 0 helpers to relocate per
router_scan.py). Bodies verbatim; @app -> @router, meta_db -> appstate.meta_db.
include_router at the original site; 143-route table identical to origin/main.
server.py: 9,337 -> 9,301.
No test retargeting (test_demo_mode only names the paths in middleware regexes).
Verified: pyflakes clean; route table identical; pytest 2398 passed; packaging
guard green; eslint 0; boot smoke drives POST (auto-names "Loop N") / GET / DELETE
/ missing-fields error, and demo mode 403s both writes while allowing the read.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(server): extract the artist-alias routes into routers/artist_aliases.py (R3)
Second router, picked by router_scan.py: the artist-aliases / Tidy-up (P4) group
ranks at 0 monkeypatch.setattr targets and 0 helpers to relocate. 5 routes
(list/set/merge/delete aliases + raw-artist picker), all meta_db-only.
Bodies verbatim; only @app.<m> -> @router.<m> and meta_db -> appstate.meta_db.
include_router mounts at the original site; full 143-route table identical to
origin/main (paths, methods, order). No test retargets: test_artist_alias drives
via TestClient(server.app) + server.meta_db, neither of which moved.
Verified: pyflakes clean on the router; no new undefined in server.py; JSONResponse
still used in server.py (not dead); pytest 2398 passed (18 in test_artist_alias);
packaging guard green; eslint 0; boot smoke drives all 5 routes end-to-end
(set ACDC->AC/DC, read back, 400 on missing fields, delete).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs: credit both router extractions in the size-exemptions rationale (CodeRabbit)
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The packaged desktop app died at startup:
File ".../Resources/slopsmith/server.py", line 71, in <module>
import appstate
ModuleNotFoundError: No module named 'appstate'
feedback-desktop's scripts/bundle-slopsmith.sh copies a HARDCODED list of core
files into the bundle -- server.py, VERSION, lib/, data/, static/,
plugins/__init__.py. The root-level appstate.py (#833) and routers/ (#834)
shipped fine in Docker, passed every test, and were silently dropped from the
packaged app.
Both now live under lib/, the one core directory every packaging path already
copies wholesale -- Dockerfile `COPY lib/`, docker-compose.yml, and the desktop
bundler's `cp -r lib` -- and that all three put on sys.path (on Windows via the
embeddable-Python ._pth, where PYTHONPATH is ignored: build-windows.sh writes
`../slopsmith` and `../slopsmith/lib`). No feedback-desktop change and no new
release are needed for this to take effect.
lib/ is also the CORRECT home under Principle V, and always was once the design
settled: with the injection seam, appstate.py constructs nothing and does no
import-time IO, and a route module only builds an APIRouter. The premise that
forced root placement -- "appstate opens sqlite at import" -- stopped being true
when configure() replaced ownership. The Dockerfile / .dockerignore /
docker-compose.yml entries added for the root layout are reverted; nothing else
in core changes (git mv, so --follow survives).
tests/test_packaging.py is the guard: it walks server.py's module-level imports,
keeps the ones resolving inside this repo, and fails if any lives outside a
directory the packagers copy -- with the ModuleNotFoundError spelled out. So the
next root-level core module can't ship broken. Negative-checked: restoring
appstate.py to the root fails it; the message names the file and the four
packaging files a root module would have to teach.
(It also has to skip `origin in {"built-in","frozen"}` -- on 3.14 the frozen
stdlib reports origin="frozen", and Path("frozen").resolve() lands inside the
repo, which flagged `os` and `stat` as first-party.)
Verified: the bundler's copy replicated exactly into a temp dir and booted with
PYTHONPATH=<bundle>:<bundle>/lib -- `import server`, `import appstate`,
`import routers.audio_effects` all resolve, appstate.meta_db is server.meta_db,
143 routes. The same simulation against origin/main reproduces the production
ModuleNotFoundError. pytest 2398 passed (2348 + 50 new); route table still
identical to origin/main (paths, methods, order); docker build context reaches
lib/appstate.py and lib/routers/ with no __pycache__; native uvicorn boot smoke
serves /api/version, /api/library, the moved audio-effects router, and all three
migrated plugins' src/ graphs; eslint 0 errors.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Move-only, same shape as the MetadataDB extraction. The core-owned song/tone ->
audio-effect-provider routing index leaves server.py for a flat lib/ module.
The class body is byte-identical; server.py reconstructs exactly from
origin/main minus the cut range plus the import-back and the call site.
server.py: 9,705 -> 9,433 lines.
The only non-verbatim change is the constructor seam: `__init__` takes
`config_dir` instead of reading the module-level CONFIG_DIR, so the module does
no IO at import (Principle V). The `audio_effect_mappings` singleton stays in
server.py -- no route, no test, and none of the `monkeypatch.setattr(server, ...)`
targets move. No import went dead.
Verified: pyflakes clean on the new module; no new undefined name in server.py;
pytest 2341 passed; eslint 0 errors; boot smoke drives the extracted DB
end-to-end (POST a mapping -> GET reads it back -> audio_effects.db lands in
CONFIG_DIR, proving the config_dir seam) and all three migrated plugins still
serve their src/ module graphs.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Move-only. The library metadata cache -- the `MetadataDB` class (4,018 lines)
plus the query helpers it owns (keyset paging cursors, the tuning grouping key,
smart-arrangement naming, tag normalisation, the startup DB-restore swap) --
moves out of server.py into a flat `lib/` module. Every moved block is
byte-identical to its server.py original; server.py is exactly origin/main
minus the six cut ranges, minus the now-dead `import contextlib`, plus the
import-back block and the constructor call site.
server.py: 14,037 -> 9,705 lines.
The one non-verbatim change is the seam that lets the class leave server.py:
`MetadataDB.__init__` now takes `config_dir` explicitly instead of reading the
module-level CONFIG_DIR, so `lib/metadata_db.py` does no IO at import
(Principle V). The `meta_db` singleton stays in server.py, so `server.meta_db`
(282 refs) and `server.app` (67 refs) resolve unchanged and no route moves.
None of the 114 `monkeypatch.setattr(server, ...)` targets moved.
Logging still goes through the `feedBack.server` logger, so log filters and
caplog assertions resolve to the same logger object.
`tests/test_settings_export_library_db.py` imports `_apply_pending_db_restore`
from metadata_db (the test moves with its subject); no other test changed.
Verified: pyflakes clean on the new module (zero undefined names, zero unused
imports) and no new undefined name in server.py; pytest 2341 passed;
node --test 1030 passed; eslint 0 errors; uvicorn boot smoke serves
/api/version, /api/library, and all three migrated plugins' src/ module graphs
(stems, studio, editor -> 200).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(gp2rs): write beat times at 6-decimal precision
The editor/timeline derives per-bar BPM from beat spans
(bpm = beats*60/span), which amplifies rounding: at millisecond
(3-decimal) precision a constant-tempo GP import (e.g. 140 BPM) shows a
spurious per-bar "tempo drift" of ~0.05-0.7 BPM because most bar lengths
don't land on a ms boundary (worse for fast/odd meters). gp2rs computes
these beat times exactly from the GP tempo map, so the only precision
loss is the ebeat/startBeat format string. Writing them at 6 decimals
(microseconds) makes the derived tempo match GP's authored value.
Verified on GP5 imports (Highway to Hell 116, Equivalence 140, Living
After Midnight 138): the derived per-bar BPM collapses from two drifting
values to the single authored constant.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013JgxKh99UAeQqmhzSc73tv
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* test(gp2rs): compare ebeat times by value, not string
The 6-decimal beat-time write makes _assert_ebeats' exact-string compare fail
("0.500" vs "0.500000"). These tests only assert spacing, so parse both sides
to float — precision-agnostic, no need to rewrite every parametrized list.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Byron Gamatos <xasiklas@gmail.com>
The tunings catalog is served as frequencies scaled to the reference pitch,
so every consumer that needs note identities (the v3 instrument badge's
TUNING_NOTE, plugins converging on the host profile model) reconstructs MIDI
numbers client-side via log2 — a rounding footgun at non-440 references, and
N copies of code the host can run once.
Add `tuningMidis` to the response: the same catalog keyed instrument-count →
name → absolute open-string MIDI notes (low → high). Built-ins come straight
from TUNING_PRESET_MIDIS (no float round-trip at all); provider-contributed
entries are inverted from their frequencies at the served reference via the
new freqs_to_midis() (the inverse of open_midis_to_freqs, garbage-guarded).
Purely additive — referencePitch/tunings are unchanged.
Tests: every built-in round-trips at 440; round-trip holds at 430/432/444/450
(the exact case client-side reconstruction drifts on); garbage rejected.
Claude-Session: https://claude.ai/code/session_01MS2YFb6UUSwJVV6CmEa25i
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(drums): capture velocities alongside times in unmapped-percussion reporting
Both drum converters opt-in out_unmapped capture (convert_drum_track_from_midi,
convert_drum_track_to_drumtab) gain an index-aligned `velocities` list next to
`times`, carrying each dropped note real dynamics — MIDI velocity verbatim; GP
velocity with the same 1-127 gate as mapped hits, falling back to the 100
import default. A hand-mapping UI (the editor unmapped-notes dialog) can then
restore mapped notes at their source dynamics instead of flattening to v:100
(editor-side consumer: feedBack-plugin-editor#111).
The GP path chronological sort now reorders times and velocities in LOCKSTEP
so multi-voice measures cannot silently reassign dynamics. Additive: callers
that ignore the new key are unaffected.
Tests: extended tests/test_midi_import_drums.py + tests/test_gp2rs_drums.py
(alignment, lockstep sort, out-of-range fallback) — 26 passing.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JEoFeTPSnz4NpwwCG52hnu
* docs(gp2rs): clarify velocity-default comment, mark dead-path fallback
- The mapped-GP velocity comment conflated GP's authoring default (95,
Velocities.default) with the drumtab render default (100,
DEFAULT_VELOCITY in lib/drums.py) used when `v` is omitted. Clarify
both defaults and that only the latter applies to omitted hits.
- Mark the `else: times.sort()` fallback in the unmapped-percussion
time/velocity sort as belt-and-suspenders — times and velocities are
always appended together under the same len<100 guard, so lengths
can't actually diverge.
No behavior change; comment-only maintainability nits from PR review.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: ChrisBeWithYou <chris@rifflarr.local>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
convert_midi_track_to_keys_wire builds its own inline tempo map and
divides by the raw midi.ticks_per_beat at two sites (the tempo-table
precompute and the tick_to_seconds closure). A malformed header with
division == 0 raised ZeroDivisionError, and an SMPTE division (which
mido returns as a NEGATIVE signed short) produced negative/garbage
note times.
Guard the divisor with `ticks_per_beat if ticks_per_beat > 0 else 480`
so both the zero and negative cases fall back to the SMF default. The
`> 0` form (not `or 480`) is required because a negative value is
truthy and would slip past `or`. Positive-division behavior is
unchanged.
Follow-up to #796, which fixed the same class of bug in the newer
convert_midi_tempo_map / _build_tick_to_seconds path.
Adds two focused tests: division == 0 no longer crashes and emits a
non-negative time, and a negative/SMPTE division yields sane
non-negative times.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(midi): convert_midi_tempo_map — extract tempos, time signatures, beat grid
The keys/drums MIDI converters always built a tempo-aware tick->seconds
map internally (to bake note times) and then discarded it — and never
read time_signature meta at all — so every MIDI import landed with no
bars, no measures, and an implied 4/4 no matter what the file said.
New lib/midi_import.py helper convert_midi_tempo_map(midi_path,
track_index) extracts the grid a .mid actually carries:
- tempos: {time, bpm} per tempo event (deduped per tick, 120 default)
- time_signatures: {time, ts: [num, den]} — the song-timeline shape
- beats: one row per beat on the editor grid shape — numbered downbeats
with a den hint, measure:-1 interior beats; the beat unit follows the
active signature (6/8 = six eighth-note rows per bar)
Event scope mirrors _build_tick_to_seconds: SMF type 0/1 merge meta
from all tracks, type 2 reads ONLY the chosen track (independent
timelines — callers must never share one grid across type-2 tracks).
Mid-bar signature events (ill-formed but seen in the wild) apply at the
next bar boundary. All times compute from absolute ticks through the
cumulative tempo table and round once at emit — rounding error never
accumulates with song length. A bar-count safety valve guards malformed
files. Consumer: the editor's multitrack MIDI import (tempo-seed
dialog, feedBack-plugin-editor roadmap Phase 3).
Tests: tests/test_midi_tempo_map.py — 10 cases driving the REAL
function against real .mid files built with mido (no stubs): default
grid, tempo bends, 500-bar rounding-drift check, 4/4->3/4 and 6/8
signatures, mid-bar signature deferral, duplicate-tick last-wins,
type-2 meta isolation from a bogus sibling track, empty files, grid
coverage bounds. Full MIDI-adjacent suite green: 55 passed
(test_midi_tempo_map + test_midi_import + test_midi_import_drums +
test_gp2midi) under the project venv.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JEoFeTPSnz4NpwwCG52hnu
* fix(midi): make convert_midi_tempo_map robust — real division guard, symmetric tempo default, single-pass meta
- Push the ticks-per-beat fallback into _build_tick_to_seconds (the single
place ticks route through), guarding on `> 0` so a division==0 (malformed)
or negative SMPTE-division header no longer raises ZeroDivisionError or
walks the beat grid into negative times. Mirror the guard at the
convert_midi_tempo_map beat_ticks site. The local `or 480` was cosmetic
before — the closure still divided by the raw division.
- Seed tempos_out with a 120 BPM row at time 0 when the first set_tempo
lands after tick 0, symmetric with the (0, 4, 4) time-signature default,
so the sidecar matches the grid the head of the song actually used.
- Collapse the duplicated meta_source/note_source lists into one
source_tracks walked in a single pass (meta collection + end_tick).
- Fix a weak assert in test_mid_bar_signature_applies_at_the_next_boundary
(operator-precedence `(A and B) or C`) to assert den == 4 outright.
- Add tests: non-positive division (0 + negative SMPTE), first tempo after
start seeds 120@0, explicit SMF type-0 file, and the _TEMPO_MAP_MAX_BARS
safety valve.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: ChrisBeWithYou <chris@rifflarr.local>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
The whisperx->transcribed alias runs before the membership check, so the literal whisperx never reaches _ALLOWED_LYRICS_SOURCES (same reason sng is omitted). Remove the dead entry. Per CodeRabbit review on #799.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
The feedpak spec (§7.1) defines the lyrics_source vocabulary as {authored, transcribed, user}, but the reader only accepted the legacy {xml, notechart, whisperx, user} set and silently downgraded anything else to "xml". A spec-compliant writer (e.g. the stem_splitter plugin, which emits transcribed for WhisperX-produced lyrics) therefore lost its provenance badge.
Widen the allowed set to the union of the spec vocabulary and the legacy values so both validate, and alias the legacy whisperx engine name to the spec transcribed so existing packs normalise to the spec badge.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
librosa.sequence.dtw's default step sizes permit pure horizontal/vertical
moves; on songs whose chroma is self-similar for long stretches the flat
cost surface let the path collapse (minutes of score onto one audio frame),
so auto-sync produced monotonic-but-garbage sync points and the per-bar
warp imported charts badly out of sync while reporting success.
Use the standard music-sync step pattern [[1,1],[1,2],[2,1]] (local tempo
ratio bounded to 0.5x-2x), falling back to unconstrained steps if the
global length ratio makes it infeasible.
Validated on the reported song (138 BPM tab, YouTube audio): coarse points
now track 1:1, refine holds slopes 0.77-1.04, warped downbeats hit onset
peaks at 3.3x background energy.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
