Air horn: the recording's first blast wobbles on a ~0.15 s cycle, so the
old 0.30 s loop jumped every lap (the repeating beat). Held, it now loops
exactly one cycle (found by self-similarity), levelled and crossfaded;
a release crossfades into the recording's own ending, so a click is a
short blast that trails off; each press cuts a still-sounding blast in
25 ms, so taps stay separate. Exact at any block size.
Loops start on the nearest beat (djay's Quantize), and resizing a running
loop keeps its start and keeps looping. Grabbing a waveform zooms out
only a little, gently, and drags at the zoom it was grabbed at. The sync
button just says Sync.
- Records in the top corners, each deck's controls beside them facing the
middle, the zoomed A-over-B waveforms in the centre, and a whole-song
overview per deck below (click to jump, drag to move) with the
crossfader between them.
- Loops per deck: Loop plus a 1/32 to 8 beat stepper; a loop starts on
the next beat and shows on both waveforms.
- Sync sets this deck to the other's tempo once, and stays lit only while
that still holds.
- The air horn plays like a real one: every press a fresh blast, held it
carries the note, let go it stops.
- Opening and closing DJ mode slides a picture over the final layout
instead of re-laying-out the window every frame (the flicker).
- Tests: garbage is collected on the main thread after each test, which
stops a rare segfault in the full run.
- The shelf is revealed in place: the decks don't stretch or squeeze as
the window parts.
- The DJ button is the top row's last box (where bpm was), so it stays
put; bpm moved under it.
- Tempo, vocals and the crossfader snap to their middle and show it. The
vocals slider has icons, lined up with its ends and middle.
- Sync sticks: a synced deck follows the other's tempo, glows and says
Synced until its own tempo fader moves. The ½×/1×/2× switch re-syncs,
and the BPM line says when it's counted half or double.
- A back-to-the-start button per deck.
- The air horn is a real recording (CC0): tap for the phrase, hold to
stretch the first blast.
- The zoomed waveforms ease out while a deck is paused or held.
- DJ audio follows a change of output device instead of going silent.
- A new playlist folder is named in place, scrolled into view.
- Visualizer bars: choose how many, 1 to 1024 (one per pixel at most).
- The volume slider is as tall as the scrubber.
- Preferences › DJ: vocals model (Best / Balanced / Fast) and gentle CPU
use (a quarter of the cores, lowest priority — measured only ~20%
slower). Splits are cached per model; a deck uses the best one.
- Prepare for DJ on a playlist or songs: BPM and vocals ahead of time.
- Each deck: a volume fader on the outer edge, labelled beside Tempo; a
½×/1×/2× switch for half/double-time counting.
- Drag a waveform to scrub (backwards too).
- Crossfader: a handle twice as tall, a middle line, a snap to centre;
narrower. "calculated, saved" is gone from the BPM line.
- Keep pitch shows a checkmark. An air horn to hold.
- The visualizer draws the mix; the sidebar art hides in DJ mode.
- The Library button has the transport buttons' 8 px corners.
Return or a double-click on a song started the regular player behind the
decks, with its controls greyed out and no way to stop it. DJ mode now
suspends the player. Sync to other track was only re-enabled when its own
deck changed, so loading the other deck afterwards left it grey; it's now
enabled whenever its deck has a song and explains what it can't do.
View > DJ Mode (or the turntable button under BPM) parts the window under
the transport boxes and opens a dark shelf: a record per deck (drag a song
on to load it; drag the record to scratch), play/pause and Sync to other
track beside it, a ±8% speed fader on each edge, Keep pitch, the two
waveforms stacked around one playhead, a vocals/instrumental slider and
the crossfader. Deck A is blue, Deck B orange; loaded songs are tagged in
the lists.
Its own engine (lintunes/dj): whole-song decode, tempo refined to 0.01
BPM plus the first beat (exact on synthetic kicks), WSOLA time-stretch or
varispeed, a QAudioSink on its own thread. Sync matches tempo and, while
both play, nudges the beats into line. A missing BPM is calculated and
saved. Vocals come from Demucs, run once per song in the background and
cached; Demucs lives in its own venv.
Now Playing: tap the art for it full screen, the title or artist line
for its album; shuffle, repeat and a new speakers button are icon pills.
The main menu is a list (Now Playing on top, Settings at the bottom, six
rows to a screen). The speakers button lists this Rabbit plus every
AirPlay speaker and Chromecast on the Wi-Fi: AirPlayRemote is the
desktop's RAOP sender in Java (verified from the R1 against
shairport-sync), CastRemote speaks the Cast protocol directly to the
Default Media Receiver and serves the file over HTTP (not yet tried on a
real Chromecast).
Held is now PRESSED_DARKER 125 (between Material 3 and GNOME), sunk, on
every button. Transport boxes no longer carry the face themselves (a
QFrame has no pressed state): each button does, so shuffle, repeat and
bpm sink while held and the play box's tiles act as separate buttons with
no grey hover patch. Scrollbar thumbs darken further while dragged. The
seek bar's groove and knob are twice as tall. The album art meets a 1 px
line in the lists' grey, without the pale blend that doubled it.
AirPlay: when LinTunes' own sender can't connect (or airplay_backend is
"cli"), Music Assistant's cliairplay is downloaded once, refused unless
its SHA-256 matches the pinned one, and streamed through.
The playlist list gets the panels' 8 px corners. The empty album-art panel
pops up as a plain button under the cursor and sinks darker while held.
The big art window keeps the art's shape after a resize and paints the
art to cover, so it never shows a background.
pyatv's handshakes play nothing on a WiiM; Music Assistant's cliairplay
plays on it through libraop's RAOP client. airplay/raop.py reproduces that
client's behaviour on the wire: one RTSP session with auth-setup, an
AppleLossless ANNOUNCE and RECORD carrying RTP-Info, OPTIONS keepalives,
a 2 s buffer, NTP-locked timestamps, a start that sends the first two
seconds at once, the timing responder and resends. Non-Apple speakers
stream through it; pyatv only finds them. Verified against shairport-sync
(decoded tone, flush on pause/seek, volume, metadata). Round 88's pyatv
compat hooks are gone.
- Typing in the library search now also narrows the playlist list to names
containing every word; folders stay when they or something inside match,
open while filtering, and clearing puts back the folders you had open.
- AirPlay: an unreadable /info reply no longer fails the stream (pyatv died
on shairport-sync's).
- Transport: a repeat button (off / all / one) between shuffle and the
visualizer. Repeat-all wraps the walk (a shuffled one is reshuffled per
lap); repeat-one replays a song that finishes, Next/Previous still move.
BPM button and the now-playing minimum shrink to make room.
- Clicking the streaming tile to stop casting/AirPlay no longer also opens
the big visualizer (the release after the stop counted as a click).
- AirPlay: pyatv's AirPlay 2 session is silent on a WiiM, and its AirPlay 1
pairs then talks plaintext, which the WiiM hangs up on. Non-Apple
speakers now get cliraop's handshake: auth-setup, no pairing, ALAC
(uncompressed frames). airplay_protocol "2" keeps pyatv's AirPlay 2.
tiat.place connects but stays silent; pyatv streams to it with its AirPlay 2
path, which the fake receiver never exercised. Log every stream's start,
first audio and outcome; config.json airplay_protocol (auto/1/2) forces a
version and airplay_debug turns on pyatv's own debug log.
Every song is decoded by the ffmpeg CLI and streamed over RAOP, so ALAC
and AIFF play too (only FairPlay is refused). pyatv has no pause or seek
for RAOP, so both restart the stream at the heard position. The picker
lists every AirPlay device on the network, greyed with the reason when
LinTunes can't stream to it (video-only TVs, PIN pairing). Casting and
AirPlay are mutually exclusive; the visualizer tile shows the AirPlay glyph
and a click there stops it. Needs: pip install -e . (adds pyatv).
The cones move only on the plugin's own 25 fps frames again: driving them
from fresh audio every refresh made bass-heavy songs twitch (a jump on 84%
of refreshes against 41%). The continuous turn, texture crossfade and the
GPU big window stay.
The tile: a click opens it big (and it stays blank while big); hovering
shows a colour wheel (colour / gray / off) and up/down arrows (previous /
next visualizer). No more double-click or click-to-cycle. The big album
art closes on a click or Esc, like the big visualizer.
The plugin still steps at its own 25 fps, frame for frame; between frames
the turn continues, textures crossfade, and the cone follows a preview of
the newest audio, so beats land a refresh sooner. The big window replays
the original's GL calls through OpenGL (QPainter took ~40 ms a frame at
1080p); without GL it falls back. Gray mode is only the gray-mode colour on
the panel background. The sync slider goes to -50 ms (sooner).
Reverse-engineered from its PowerPC binary and checked against the original
running under emulation: textures byte for byte and every triangle, over
3,600 frames. 25 fps like the plugin; its colour maps and wave images are
extracted from its own code; tests/kaleidostrobe_golden.npz was recorded
from the original.
Measured once per file with ffmpeg's ebur128 (cached on this machine),
applied as a gain on the output volume; the next songs are measured
ahead and an unmeasured one glides to its level. Toggle in Controls and
Preferences → General.
Export Web Mix gets an 'Upload to web server' box: the mix is built in a
cache staging folder and mirrored with rsync --delete into <path>/<slug>/
on the host set in Preferences → Integrations → Web Server (config.json,
this machine only). An ssh check asks before replacing an existing mix;
a status window follows check, export and upload. The web player gains a
shuffle button (and the S key).
Get Info has a Volume slider (−100%…+100%): iTunes' own per-song volume
adjustment, which the library already imported but never used. It scales
the volume slider while that song plays; a boost can't pass full volume.
Works on one song or several (Mixed when they differ), and undoes.
Get Info's Previous/Next buttons are gone: since the combined multi-song
editor they were always disabled (and blank in some themes).
The laser settings draw the rig — speaker and mirror, laser, beam, and the
screen at its distance and angles — as a small textbook line drawing that
labels distance, angles and dot size in place of the sliders' numbers.
Preferences section headings are bold and 2 px larger than the text.
Each visualizer style's settings page has a Preset drop-down and a
"Save as Preset…" button. The drop-down lists the built-in Default, your
presets, and "Edit Presets…" (rename in place, delete); it reads "Custom"
whenever the settings match no preset. Presets sync with the preferences;
the per-machine sync delay is never part of one.
Laser: a two-handle "Speaker range" sets what the speaker under the mirror
plays (36 dB/octave edges; the noise floor is measured per range so treble
alone still shows). Bass-only gives clean loops; treble-only is a shimmer
along one line, as a membrane far above its resonances really draws.
The afterimage is now a phosphor buffer, dimmed each frame and rounded
down so it always fades to nothing — constant cost however long the fade
(drawing every past sweep ran the big window past a frame).
Line width is a number box stepping by 0.5; radial dots have a size slider.
Sync: decoding runs ahead of the speakers — by up to ~0.8 s after a GUI
stall, when buffers arrive in a burst — and the visualizer drew the newest
audio. It now keeps timestamped history and draws what the player says is
playing, plus a per-machine Sync delay (config.json) for latency Qt can't
see, like Bluetooth.
Every style has one Fade slider (0 = off); bars now leave fading steps.
Laser: afterimage time, line width down to 1 px, glow on/off, tip dot
on/off and size (a resting beam always shows its dot), and an oscilloscope
dot grid under the beam. Visualizer sliders run the full window width.
Highlight: click a swatch for a full color picker; the color picked
replaces that slot (right-click restores it).
Laser: the screen's distance, pitch, yaw and roll are sliders (the beam ray
is intersected with the turned screen; the drawing shrinks only when a
steep screen would push the figure out of view), and the afterimage can be
turned off. The membrane is now a slack 50 Hz balloon whose split rocking
pairs land in the bass, so it draws loops ~60% of the time instead of ~25%.
Radial: fading trails can be turned off; the number of frequency bands and
volume steps are sliders; ghosts fade from the dot's own color into the
background; a 4096-point FFT and interpolation end the flat bass arcs; and
"Even out bass and treble" takes the song's overall tilt out so the dots
sit on a circle.
Preferences → Appearance → Visualizer picks Bars (as before), Radial dots
(the EQ in a circle: angle is frequency, radius is volume, moving dots
leave light-gray ghosts that fade over a set time), or Laser (a mirror on
a speaker membrane bouncing a beam onto a screen: simulated membrane
modes, D·tan(2θ) reflection, dwell-weighted brightness). Double-clicking
the panel opens it big; a click closes it.
No identify dialog pops up mid-download any more: songs are looked up
quietly as they land and reviewed together once the last one is in, as
rows in the app's columns. Songs that are one album get the album-wide
fields once, unified, plus a cover from the album-art search. Ticked
rows apply as one undo step; unticked ones get the ordinary dialog
after. Popups wait for LinTunes to be the app in front.
A Spotify track, album or playlist is read off Spotify's public embed
page, each song found on YouTube by the result closest to Spotify's
length, downloaded with the song flags, tagged with Spotify's
artist/title/album/track number, then imported and identified.
Pasting a song from a YouTube Mix (watch?v=…&list=RD…) imported song
after song: yt-dlp takes the whole list by default, and a Mix runs to
thousands of rows and loops back on itself, so the linked song came in
again and again with nothing on screen saying why.
Now anything but a plain song link is listed first (--flat-playlist,
streamed into the dialog). One song goes straight through. Several become
a checklist: a song-in-a-list link ticks only its song and offers Just
This Song without waiting, and a playlist link ticks everything. Rows are
de-duplicated by id. The chosen songs download by their own pages, never
the list link, and more than one gets a per-song progress window (Hide
keeps downloading; the status-bar line brings it back).
The play journals keep running totals per track, which is enough for "most
played ever" and nothing a Wrapped asks: top songs of this year, minutes
listened, which month, what time of day, what got skipped. Those all come
from one timestamped line per listen, so LinTunes now keeps that, while
Preferences -> General -> "Keep a play history for my yearly Wrapped" is on.
history/<machine>/<YYYY-MM>.jsonl, one writer per folder like the journals:
local start time with its offset, ms actually heard (Player adds position
steps under 1.5 s, so seeks and pauses don't count), how it ended, where it
played and from what, and a name/artist/album snapshot that outlives a
delete. Under 5 s is noise and never written.
andTunes 0.3.0 writes the same lines on the Rabbit when sync says the
desktop wants them, and sync copies each month home when the device's copy
is longer. "Share this year's play history with friends" puts a trimmed
copy (no ids, playlists or devices) in each friend's cassette folder, daily
in November and December and monthly otherwise.
Spectral-flux onsets + autocorrelation in numpy, decoded by the ffmpeg CLI;
no new dependency. Runs on a worker with the status-bar progress bar, for
one track or many, and the batch is one Ctrl+Z.
On Wayland Qt paints the title bar through its adwaita decoration plugin,
which hardcodes pure white in light mode and ignores the palette. Every
window now goes frameless and gets an app-drawn bar in the palette's Window
color, so it follows the Background setting. Buttons follow GNOME's
button-layout; moving, snapping and edge resizing still go through the
compositor. Windows grow by the bar's height so their contents keep the
size they had.
Opening Connections called find_device() on the GUI thread, which lists the
gvfs-MTP mount; on a wedged mount that blocks forever in an uninterruptible
FUSE wait and the app had to be force-quit. DeviceWatcher now probes on a
daemon thread (one at a time, STUCK after 3 s, late answers still land), the
menu reads its cached answer, and a stuck mount shows "Rabbit isn't
responding…" with a Reset Connection button (kill gvfsd-mtp, remount via
gio). Tests can no longer reach the real mount (conftest patches
find_device), which also takes the suite from ~200 s to ~22 s.
- Rabbit Sync Settings shows the sidebar's folder tree. A folder has its own
tristate checkbox; a fully ticked folder is stored as device_sync.folder_ids
and follows the folder, so playlists added to it later sync too.
- The album-art pass compares covers against the sizes the planner's device
scan already read (plan.art_sizes) instead of an MTP stat per album, caches
"no art" as a .none marker, and only unlinks covers that are really there.
The summary reports covers actually updated rather than every cover.
- A quiet tink when a Rabbit sync, export or URL import finishes, with an
on/off switch (sound_effects, on by default).
- Preferences is now General / Appearance / Integrations, a list on the left
and the page on the right, opened big enough for the tallest page.
The sidebar gets 8px of bare window on its left, and the Library button
spans exactly the playlist list below it, scrollbar included. The Library
button, search box, playlist name, times and playhead all stand on one
floor 5px above the lists (text by its baseline). The top row floats:
equal space before, between and after the controls, with the now-playing
panel a fixed share of the width (700px at 1920). The header row grows if
the app font makes the Library button taller than the strip.
The Library button, the view's header (search box or playlist name) and
the seek row now share one band under the transport boxes, instead of the
seek row hanging alone with a header strip below it. The now-playing panel
caps at 560px and the gaps between controls share the rest evenly; more
room under the menu bar. Long playlist names shrink up to 2pt, then elide.
- A followed playlist (the UFO in friend mode) appears in my sidebar with
a purple UFO: mixed in with my playlists alphabetically, or inside a
folder named after the friend (per-friend, in Sync Settings). It's never
a Playlist of mine, so it stays out of the library and its merges.
- It plays from the cache. Player's queue accepts friend-track keys,
resolved through Cassette: they scrobble to Last.fm but don't count in
my play journal. Songs still on their way are grey and skipped.
- The cassette column there is hidden by default (right-click the
header); pressing one takes the song into my library — straight from
the cache if it's there, otherwise by asking for it.
- Sync Settings per friend: keep followed songs separate (default) or add
them to my library. Unfollowing (right-click) frees their cache.
- The bottom bar says when the friend's folder last synced.
- Sending: a friend's requests ∩ what I still offer is copied into my
outbox (via a .part name Syncthing never sees); whatever they no longer
request is deleted, so the outbox cleans itself up.
- Receiving: a finished file I cassetted is imported exactly like Add to
Library — organized into Artist/Album, tags read, date added now, none
of their listening history — and a song wanted only for a followed
playlist lands in the cache. Then requests.json is rewritten without
what arrived and without what they stopped offering.
- Safe to repeat: Syncthing temp files ignored, Syncthing asked whether a
file is whole, and an exact match already in my library is never
imported twice.
- Folders are watched (re-armed after each rename), with a slow fallback.
- file_importer split into stage_file (off the GUI thread) and
track_from_file (date added = now).
- pytest --syncthing now runs a real invite → share → request → deliver →
cleanup between two Syncthing instances.
- A ▾ cut from the right end of the Library button (shown once there's a
friend) opens a popup of friends. Choosing one enters friend mode: the
button becomes "<Name>'s Library" in light purple, the sidebar lists
their shared playlists with a UFO follow toggle on each, and the content
area is their library — browse only, nothing plays or edits.
- A cassette column on the far left and cassette buttons on every artist
and album: pressed grey = send me this, medium grey on hover to let go.
An artist cassettes all its albums and tracks (a snapshot).
- Bottom bar: Hide tracks I have (exact matches: artist, album, title,
length within 2 s); close matches show light grey. Size of what's
cassetted against free space: orange when Syncthing would stall at its
free-space floor, red when it won't fit.
- Save and Close writes requests.json; Cancel discards; the ▾ or quitting
with changes asks Save / Don't Save / Cancel.
- Connections ▸ Share Library with a Friend… makes a one-time invite code
(device id, display name, token) and the inviter's send-only folder.
- Connections ▸ Add Friend Library… takes the pasted code: adds the
inviter, shares our folder, and accepts theirs receive-only when it's
offered back.
- Syncthing hides an unknown device's folders (verified on 1.30), so the
inviter probes a new knock only while an invite is open: added with
nothing shared, completed on the right token, otherwise removed and
never probed again. Used and cancelled codes go nowhere.
- Sync Settings lists open invites (Cancel Invite) and gives each friend a
page: avatar, plain-language status with a next step, last seen, last
synced, Remove Friend. Removals are staged until OK.
- scripts/cassette_pair.py runs throwaway Syncthings on loopback;
pytest --syncthing drives a real three-machine handshake.
- Multi-select Get Info keeps the art square; a pasted cover goes into
every selected track (shared cover shown, else "mixed artwork").
- Paste Artwork button + Ctrl+V anywhere outside a text field; a caption
says what happened. Clipboard reads that don't decode are retried and
never staged; every attempt is logged at INFO.
- Album art search queries Deezer alongside iTunes and ranks by album
match (iTunes has no copy of Digable Planets' Reachin' at all).
- embed_artwork shared by Get Info and Download Album Art; single-track
Get Info now invalidates the MPRIS art cache.
Import from URL now gets past YouTube's "Sign in to confirm you're not a
bot": a run that downloads nothing for that reason retries once with
yt-dlp's --cookies-from-browser (Firefox first), and a browser that worked
is remembered in this machine's config.json so later imports send the
cookies from the start.
A single Right-arrow press skipped 1,676 tracks in 81 s and hard-froze the
desktop. On Wayland key repeat is generated by the client until the
compositor delivers the release, and the key filter treated every repeat as
a new press. Each skip sent gnome-shell an MPRIS Metadata whose xesam:artist
was a plain Python list, which marshals as "av" instead of "as". gnome-shell
logged an error for every one (50k/min) and reloaded the cover, and that load
is what kept it from ever delivering the release.
Space/Left/Right now act once per press and swallow the repeats.
xesam:artist goes out through mpris.string_array, which build_properties_changed
now shares for invalidated_properties. The new test round-trips the signal
into GDBus, which is what gnome-shell validates with, and fails 'av' == 'as'
against the old code.
trav couldn't play the first track of his playlist. Shuffle off gave him song
two; shuffle on gave him a random one. The player was right both times:
library.json named a song whose file was not on the device, and skipping an
unplayable file advances one slot. Tapping any song that *is* there plays
exactly that song, shuffle or not.
The file was missing because a gvfs-MTP mount can hold a phantom directory --
one it lists happily while the device has no such folder. Every write into it
fails EIO, and mkdir(exist_ok=True) sees the phantom and does nothing, so it
never heals; only remounting clears it. That folder was new because Round 52's
retag moved the file under a new artist.
So a copy that raises OSError now costs that song, not the sync, and the song
is left out of the m3u and library.json. Aborting cost 2,400 songs for one
folder; counting it present anyway put a song in the manifest that isn't on
the device, which is the bug you could hear. The names come back in the
summary and a dialog rather than vanishing.
Planning is a worker now. plan_andtunes_sync is pure and writes nothing, but
it walks every file on the device, and over MTP that is thousands of round
trips -- on the GUI thread it froze the window and GNOME offered to kill
LinTunes, which is how a sync got force-quit halfway through.
And the sync stopped going quiet at the end. "Album art 501/501" is emitted
before the last album, and then _write_index, _ship_buttons and
_prune_empty_dirs ran silently: 501 exists() stats over MTP, ~840 KB of
writes and a full tree walk, with the progress line frozen. They report now,
and _write_index reads Art/ in one listing instead of a stat per album.
andTunes 0.2.2: the wheel scrolls the other way in lists, and smoothly --
a detent adds to a pixel debt that a Choreographer callback pays off a
fraction per frame, so one detent eases to a stop and a fast spin blends into
one movement instead of teleporting a row at a time. Volume is unchanged.
Claude-Session: https://claude.ai/code/session_01SKXUgsBBwe3qaHEjeV8ubP
The sync was fine. Everything it put on the Rabbit was there and correct --
2,432 media files, 208 covers, a manifest whose every field type-checks
against the parser -- and andTunes still answered "No library / Couldn't read
library.json". Fifteen tracks took all 2,439 songs down with them.
Library.group() keyed albumsByKey case-folded and artistsByName raw-case, and
only ever built the Artist inside `if (album == null)`. So the second track of
an album whose artist is spelled differently found the album already there,
skipped the block that would have made the artist, and dereferenced the null
that came back. Six artists in trav's library are spelled two ways: RJD2/Rjd2,
Toro Y Moi/Toro y Moi, FatBoy Slim/Fatboy Slim, LOVING/Loving, Land Of The
Loops/Land of the Loops, Salami Rose Joe Louis/salami rose joe louis. It had
worked until the 12th because that sync was the first to carry both spellings
of one of those albums.
Both maps fold case now, and the artist is fetched-or-made before the album
block and held, so no lookup left in group() can come back null. Artist gains
a key the way Album always had one, and ListActivity navigates by it --
ALBUM_ROW already passed album.key, so artists just stopped being the
exception.
The device filesystem is case-insensitive, so those two spellings are one
folder there. plan_andtunes_sync compared exact strings and saw every such
file as stale *and* missing, deleting and re-copying it over MTP on every
sync forever; the diff and the collision rule both fold now. plan.stale still
carries the device's own spelling, since that is what _delete_stale unlinks by.
Verified on the Rabbit: 2,439 songs listed, RJD2 one row of 12 songs.
Claude-Session: https://claude.ai/code/session_01SKXUgsBBwe3qaHEjeV8ubP
trav's "Dionne Farris - I Know" kept identifying as Jay-Z, against ID3 tags
that plainly said otherwise. AcoustID was right; the ranking threw the answer
away.
The fingerprint matched one AcoustID result at 0.97, and six recordings hang
off it: Dionne Farris twice, plus Jay-Z, Marisela, New Atlantic and David
Essex, all of whom recorded a song called "I Know". A result's score belongs
to the *audio*, so every linked recording carries it however wrong the link
is. With the scores tied, ranking fell through to the duration bucket, where
Jay-Z's 222.7 s beat Dionne's 227.3 s against a 224 s file. The tags never got
a vote: the hint sat below duration in the sort key.
So the lookup now asks who submitted each link. `sources` joins LOOKUP_META —
475 people linked that audio to Dionne Farris, 6 to Jay-Z, 1 each to the rest
— and _link_tier sinks anything under a tenth of the strongest link in the
same result. The share is relative, never an absolute count, and a missing
count ranks as real: an obscure song's true link may have two submissions
against a stray's one, and rounds 45-46's payloads rank unchanged.
And it asks what the file already says. artist_hint_for gathers the artist
tag, the album artist and the artist in the filename; _artist_agreement counts
the words shared with a candidate's credit, placeholders dropped. Like every
hint since round 46 it only chooses among what AcoustID returned.
New key order: stray tier, artist agreement, duration bucket, hint overlap,
release rank — who, which take, which release. Artist above duration is the
whole fix; duration still separates two takes by one artist.
Verified live against the reported file: the proposal is now I Know — Dionne
Farris — Wild Seed - Wild Flower (1994), track 1, with all five mis-tagged
artists off the dropdown. The real response is pinned in tests/test_round52.py.
Claude-Session: https://claude.ai/code/session_01SKXUgsBBwe3qaHEjeV8ubP
andTunes phase 3. The R1 is now one more machine in the Round 38 play-count
model, and nothing lands on it that Android can't play.
Play counts: andTunes 0.2.0 counts a play on a natural finish (LinTunes'
rule) and keeps Music/andTunes/plays/andtunes-<install id>.json in exactly
the per-machine-totals journal shape, written beside the old one and renamed
over it. Each sync first folds it into <data_dir>/plays/ with a new
play_journal.merge_totals, the per-track max: the file now has two homes and
both desktops may bring it back, so the max converges and an older copy can
never pull a count down. Nothing is written when nothing moved, and
PlayJournal.load needed no change.
Format gate: the planner reuses export/web_support.conversion_for outright,
since Android's MediaPlayer decodes the browser's set. FairPlay is refused
and reported; ALAC, AIFF and oddities land as FLAC, converted once into a
per-track cache and size-diffed after that. With no ffmpeg, the export's
"sync without them?" question. A file ffmpeg can't read costs that song,
not the sync.
"Sync Playlist to Rabbit (Auxio)" is retired from the menu; device_sync's
helpers stay because export and andTunes import them.
The dev fixture grew a real ALAC track, a Protected AAC .m4p and an "Odd
Formats" playlist. Verified on the R1 with it: AIFF and ALAC arrived as
FLAC and played, the FairPlay track was refused, four plays came back on the
next sync and each track's effective count rose by exactly one, and a third
sync copied nothing.
Found along the way: USB re-enumeration can wedge gvfsd-mtp, after which
anything touching the mount (find_device, the GUI tests) hangs in an
uninterruptible wait. CLAUDE.md now has the recovery.
Claude-Session: https://claude.ai/code/session_018ZCVBTJRFJ2XfMshu2gtUv
The Android half of andTunes had been paused on a 1 GB toolchain download
(Gradle wrapper, Android Gradle Plugin, Kotlin). None of that was needed:
platform 33 and build-tools 34 were already in ~/Android/Sdk, and an app that
links no libraries needs five SDK steps, not a build system. andtunes/build.py
runs aapt2 -> javac -> R8 -> zipalign -> apksigner and produces a 61 KiB APK.
andTunes 0.1.0 is framework-only Java: a six-tile menu that loads nothing, a
streaming library.json parse grouped into artists and albums in one pass, one
ListActivity for every list (songs and artist songs open with Shuffle all,
albums carry art thumbs, search as you type), a Now playing bar under every
list, and a white Now Playing screen with square art, a slim scrubber,
prev/play/next, shuffle and repeat, and no back button. A foreground
PlaybackService owns the queue: audio focus, becoming-noisy, MediaSession and
a MediaStyle notification, end of queue pauses, missing files skip, and the
last queue and position come back on launch. Black on white, and every size
doubled for the R1's override density. Cold start to the menu: 313-338 ms.
The scroll wheel turned out to be DPAD_UP/DOWN (stock Generic.kl), not
volume. The app takes both in dispatchKeyEvent: lists scroll a row per
detent, and the menu and Now Playing get volume.
LinTunes side: Connections > Install andTunes on Rabbit... installs the
committed APK over adb and grants all-files access, or copies it to Download/
over MTP when there's no adb. Sync fills Buttons/ with default artwork only
where a file is missing, so trav's own art is never overwritten.
Verified against the dev library only, synced to the R1 and driven with
adb input + screencap.
Claude-Session: https://claude.ai/code/session_01KSsYtRn4WZx6xj4GEdVPt9