Both browser tiers are blind to `(hover: hover)` gating, in different
ways and without failing: CDP media emulation does not reach ui-test's
iframe, and e2e's phone specs reach phone width with setViewportSize,
which changes no media feature but width. Written down with what does
work — a device-descriptor context — because the next person to gate an
affordance this way will otherwise re-derive it, and the tempting
conclusion from a green suite is that the gate is covered.
The bottom bar's title, artist and "Playing from X" all navigate. In a
bar sized for a bar they are a few characters of text, which is not a
touch target — and explore-link holds its navigation for one
double-click interval and drops it if a second click arrives, a gesture
that exists so double-clicking a row can play it and that means nothing
on touch.
Below the shell's phone breakpoint the three render as plain text. The
words are unchanged: the source line still says where the queue came
from, because dropping the link is the change and dropping the
information would be a different and worse one. The cover art already
carries the phone-only button that opens the full-screen Now Playing
view, which is where the links live.
This is in JS rather than in the stylesheet because what changes is the
content, not its appearance — no CSS rule takes a click handler off an
element. matchMedia is read in connectedCallback for the reason the
reduce-motion query beside it already is, so a test can answer it first.
Two smaller things. PHONE_QUERY moves out of track-list.ts into
utils/breakpoints.ts: it was a private const when one component needed
it, and a second reader is where a copy starts drifting from index.css.
And `phone` joins geometryKey(), because crossing the breakpoint swaps a
link for a bare string and the marquee travels a distance read from
measuring it — the words being identical either side is not the same as
the box measuring the same.
Closes#61
The play button on a home shelf's cover cards is revealed by :hover, and
a touch long-press synthesises a hover state in the WebView — so on a
phone it flashed into view during the 500ms hold that
utils/long-press.ts is measuring for a context menu. A control appearing
because the user was reaching for a different one.
It is gated on `(hover: hover) and (pointer: fine)` rather than on width,
so it is absent on any touch device and present on a desktop with a small
window. A phone user taps the album and plays from the detail view, so
nothing replaces it.
The default outside the query is display:none, not opacity:0. An
opacity-0 button still takes taps and is still in the accessibility tree,
so leaving the reveal as the only guarded part would keep the hit area
for a control the phone can never show.
The test asserts the parsed stylesheet rather than rendering as a phone,
and says so: CDP's Emulation.setEmulatedMedia does not reach this tier's
iframe, so matchMedia still answers `hover: hover` after it is set. The
regression worth catching is someone hoisting the rule back out of the
query as a tidy-up — a change no desktop assertion can see.
Closes#68
`in_library = 1 AND local_*_id IS NULL` was a fixed point.
upsertBatch's conflict clause is `MAX(in_library, excluded.in_library)`,
so it can only ever raise the flag, and pruneStaleLocalCrossReferences —
which its own comment calls the only place a removal from the library is
reflected back into the index — was gated on the id being present. So
nothing in the app could clear such a row, ever: a permanent claim of
ownership with no local row to check it against.
The gate is now the flag *or* the id, for all three entity types. A NULL
id fails the existence test on its own, so this needs no second clause to
say what "not owned" means.
Nothing in the tree writes that shape today — collectLibraryEntities sets
both together — which is why this is worth closing rather than leaving:
the exposure is a database written by a version whose local-id columns
were populated differently, and the next writer that sets the flag
without an id, which nothing structurally prevents and which this shape
made permanent rather than merely wrong until the next scan.
The test seeds the row with raw SQL on purpose. upsertBatch writes a zero
LocalArtistID as literal 0, and 0 satisfies `IS NOT NULL`, so the old
gate already caught that shape — a fixture built through the upsert
cannot reproduce this at all. NULL is what the artifact importer and any
older writer leave behind, the columns being nullable with no default.
Reverted against the old gate, it fails on all three types.
Closes#118
dev-headless.sh checked the PID in *this* worktree's .dev/app.pid and
nothing else, so an app orphaned by a deleted worktree went on listening
with nothing left to stop it — `make dev-stop` only kills the pid it
wrote. The new app then started, failed to bind, exited, and every
subsequent curl and playwright-cli call went to the other process: the
harness reported facts about an app nobody asked for.
That fails a long way from its cause. It presented as "no such table:
libraries" against a *freshly created* YJ_HOME, which reads exactly like
applySchema or staleshape.go having gone wrong, with a zero-byte app.log
beside it saying nothing.
The startup wait cannot catch this, because its health check is satisfied
by any app on the port — which is precisely the failure — so the check is
before the launch and refuses rather than warns. It names the holder's
pid, cmdline and /proc/<pid>/cwd, which is what identifies the checkout
and says "(deleted)" for the case this exists for. It does not suggest
`make dev-stop`: the PID-file check has already passed, so by
construction dev-stop does not know about this process and would report
success while changing nothing. --port already covers the legitimate
second-app case.
The second, cheaper guard the report asks for goes in after the wait:
"the port answered" is not "the app we started answered", so a dead
APP_PID at that point is now an error with the log tail rather than a
success message about somebody else's process.
Closes#119
The codegen-check hook was `go generate` followed by a bare
`git diff --name-only`, which is the whole unstaged worktree rather than
the generators' output. So a commit whose staged changes were fine failed
whenever anything unrelated sat unstaged — notes, a plan document, the
next commit's files — reporting "Generated code is out of date" and then
a diffstat of files no generator has ever written. `make generate` fixed
nothing, because nothing was stale, so the message sent you looking for a
codegen problem that did not exist. Splitting one piece of work into
several commits is exactly the shape that triggers it.
The tree is snapshotted either side of `go generate` and only what moved
across it is reported. That is deliberately a snapshot rather than the
list of generated paths the issue offers as the other option: a fourth
generator is one //go:generate line away, and a path list is a second
place to remember it.
Two things it has to get right. The comparison is a *symmetric*
difference, because generation can push a file into the unstaged set or
pull it out of one — a hand-edited generated file that the generator puts
back is stale generated code just as much as a source change that
outdates it, and comparing one direction reports it as current. And the
snapshot is content, not names, or a generated file that was already
dirty and is then rewritten further keeps its name on both sides and
slips through.
Closes#131
`claim` is the one step the workflow requires before the first edit, and
it failed outright on a token scoped to the work it does: `me()` calls
`GET /user` purely to name the assignee, and that endpoint needs
read:user. So the documented process was blocked by its own tooling, and
the fallback was to do the assignment, the label and the comment by hand
— which is the half-made claim `claim` exists to prevent.
GITEA_USER short-circuits the lookup, so least privilege is enough. The
lookup stays as the fallback because it is right when the scope is there
and needs no setup. Failure is now actionable and says both remedies,
and it still happens before any of the three halves are mutated.
Closes#130
The commit-analyzer reads the type and ignores the scope, so `fix` is a
patch whatever sits in the brackets. Two commits touching nothing but
.gitea/workflows/unclaim.yml were written `fix(ci):` and cut v0.2.1 and
v0.2.2 -- real releases, published to Arch, Homebrew and the APK
registry, containing no user-facing change.
CLAUDE.md already warned that a mistyped feat ships a minor version.
That was not enough, because this was not a mistyped type: `fix` was
chosen deliberately, in the belief that the (ci) scope qualified it.
The version bump is the small half, which is why this gets a paragraph
rather than a clause. A merge to main starts two workflows; if
release.yml then pushes a tag, that tag push starts four more --
arch-package, homebrew-formula, android-apk and desktop-assets -- on a
runner with capacity 1, where the APK build alone is tens of minutes
and publishes a signed artifact to a public registry. So a mistyped
type is six workflow runs, not an odd-looking changelog.
`make release-dry` answers this before the merge instead of after, and
is cheaper than any one of those runs.
The two releases are staying: they are already published, and a version
that vanishes is worse for whoever pulled it than one that turns out to
be empty.
Closes#111
`go test -race ./...` saturates every core for ~47s, and the UI tier
it was sharing them with is a real Chromium with wall-clock timeouts.
So the browser lost, at random: setup took 106s inside the hook
against 63s standalone, and a different suite failed on each run --
three failing to fetch setup.ts from Vitest's own dev server once, a
15s "did not mount itself" the next time -- against a suite that
passes 898/898 five times running on its own.
That reads as "your branch broke the frontend" when nothing is wrong,
which is the most expensive kind of false negative: the next person
bisects a change that was never at fault. It cost two pushes here
before the summary line gave it away.
Sequential costs about 15s.
Closes#128
Five faults found while auditing the play/pause and position path for
a desktop report of the pause icon showing over a seek bar that was
not moving. They are one commit because they are one file's worth of
tangled state, and two of them do not compile apart.
The finished callback did not know which chain it came from. It is
dispatched as a goroutine from the beep callback and then queues for
p.mu, so a user pressing Next in the last second of a track had it
wake up holding the lock for a player that had loaded something else
-- and rewind it, stop it, and hand a stale finish to the queue's
auto-advance. updateStreamers now stamps a chainID and the callback
carries the one it was registered with. (#123)
It also emitted PlaybackFinished and PlaybackStateChanged(stopped)
*after* releasing p.mu, alone in this file, so a Play() taking the
lock in that gap emitted `playing` first and the stale `stopped`
landed last -- the button showing play over a track that was audibly
running. Both emits are back under the lock. (#123)
A source that failed mid-track was reported to the queue as a natural
end, so a broken file auto-advanced in silence and was counted as
played. The handler takes the reason now: the player cannot name the
track, because the metadata is the queue's, so the queue emits
PlaybackFailed and skips recording the play. (#123)
p.format was assigned once, in the constructor, to the *speaker's*
rate, and never again -- so it claimed 44.1 kHz for every file. The
replay-after-finish path resamples from it, meaning a finished track
played a second time was resampled from a rate the decoder never
produced: audibly wrong speed and pitch, and the length and position
fallbacks wrong with it. The fixtures are 22050 Hz, which is what lets
a test see this at all. (#124)
p.trackLengthMs was written only when the database had a row and
cleared only by UnloadTrack, so a file with no row inherited the
previous track's duration -- and every position report is scaled by
it, so the bar reported one track's progress on another's scale.
(#125)
Queue.OnPlaybackFinished indexed q.tracks[currentIndex] having checked
only that the queue was non-empty. currentIndex is -1 whenever the
queue has been exhausted, and onQueueExhausted deliberately leaves the
finished track loaded -- so playing it from there and letting it end
panicked, on a goroutine with no caller to recover it. (#126)
The position readers guarded the decoder with the speaker lock, which
the read-ahead goroutine has no reason to hold and never takes -- so
Position() raced readAhead's Stream() on every position emit, once a
second for the whole of playback. srcMu is the lock that excludes that
goroutine, and taking it naively deadlocks, because seekLocked already
holds it and then emits the landing position from inside that region.
seekSourceLocked is that region extracted, so the lock is released
before anything is emitted. Found by the race detector, via the test
added here for the chain guard: the existing suite never loads a file
outside the integration guard, so make test was green over it. (#127)
OnPlaybackFinished picks up //wails:ignore along with its error
parameter: v3's generator segfaults on a bound method taking an error,
and this was never IPC. That removes a binding the frontend could have
called to force an auto-advance.
Closes#123Closes#124Closes#125Closes#126Closes#127
BufferedStreamer.Stream treated an empty ring buffer as a momentary
underrun and answered with silence and ok. That is right while the
read-ahead is still going to deliver something, and two of its three
exit paths left it never going to: a Close, and a source returning
(0, true) in a loop. Neither set done, so the ring drained and every
call after it was silence claiming to be audio, for the life of the
process.
Nothing above this type could tell that from healthy playback. The
beep.Seq chain never ended, so the player stayed in Playing with the
button showing pause; the decoder's position never moved, so the 1 Hz
report pinned the seek bar at a constant -- and since every report
resets the bar's interpolation, the report actively suppressed the one
thing that would still have moved it. A frozen bar over a track that
was not playing, with no watchdog anywhere to notice.
Every exit now marks the stream done, and the silence fill is bounded
by a duration *and* a run of calls. It needs both. Wall clock is the
real measure, because the speaker paces itself and a stall is a
question about time -- but a caller draining in a tight loop makes
hundreds of calls in microseconds and would outrun a duration alone.
A call count alone is the opposite failure, and not a hypothetical
one: the first attempt used one and spent the whole budget before the
read-ahead goroutine had been scheduled once, ending a perfectly good
stream at sample zero and breaking TestBufferedStreamer_BasicStream.
Err is plumbed out at the same time, because a drained source and a
failed one both arrive as (0, false) and are not the same event.
Reading it is a separate change; without it there is nothing to read.
Closes#122
Second defect in the same workflow. The shell fix took -- the step ran
under `bash --noprofile --norc -e -o pipefail` -- and got one layer
further before failing:
curl: (77) error setting certificate file: /etc/ssl/certs/ca-certificates.crt
ubuntu:24.04 ships no CA bundle, and --no-install-recommends skips the
ca-certificates that curl recommends, so curl came up unable to verify
TLS against our own Gitea.
This was avoidable by reading the repo rather than reasoning about it:
ci.yml (twice), desktop-assets.yml, android-apk.yml and release.yml all
spell out `ca-certificates curl ... jq` for exactly this reason. The
convention was written down five times already.
Validated in the real image this time rather than by extracting the
script and running it on the host, which is what missed this: the step
now succeeds inside `docker run ubuntu:24.04` against a scratch issue --
label present, 204, label gone -- and the previous version reproduces
`curl: (77)` in the same image. Both checked, then the scratch issue was
deleted.
The DELETE also keeps its response body now and prints it on a non-204.
Whether the automatic token carries issue-write scope is still unproven,
because both failures happened before the API call, and "403" without
Gitea's own sentence would cost another merge to interpret.
Closes#102
The workflow shipped in #103 and failed on every close, on its second
line, before reaching the API:
shell: sh -e {0}
/var/run/act/workflow/0.sh: 2: set: Illegal option -o pipefail
Inside `container:` the act runner selects sh, not bash, and
`set -o pipefail` is a bashism. homebrew-formula.yml carries the same
line without trouble because it runs with no container, on the host
image where bash is the default -- so "another workflow does it" was
not the evidence it looked like, and the comment now says so where the
next person will read it.
pipefail is kept rather than dropped for POSIX's sake: the lookup is
`curl -sSf ... | jq`, so without it an API error yields empty output,
an empty label id, and a cheerful "nothing to do" on every close. A
silent no-op is the one outcome worse than a failing job here.
Validated end to end against scratch issues rather than by reading it:
with the label present the step returns 204 and the label is gone, and
against an issue that never carried it the step also returns 204 and
exits 0 -- which is what makes it safe to run on every close rather
than only claimed ones.
Still untested: whether secrets.GITEA_TOKEN carries issue-write scope.
The old run never got far enough to find out. If it 403s, the fix is
one line -- secrets.PACKAGE_TOKEN, which is a user PAT.
Closes#102
A `Closes #N` footer closes the issue on merge and leaves
`Status/In Progress` on it, because Gitea's auto-close touches state
and nothing else. #100 was closed and simultaneously marked as being
actively worked on. `scripts/issue.sh close` does drop the label, and
is exactly the call the footer exists to avoid making.
This hooks the close rather than the merge. Stripping the label in the
PR would work and would be a per-PR habit, which is what the footer
removed in the first place; `issues: [closed]` covers the footer,
issue.sh close and a click in the web UI alike, and asks nothing of
anyone at any of them.
Reopening deliberately does not restore the label: reopening says the
work was not finished, not that somebody is at a keyboard now.
Two costs, both stated in the file rather than discovered later. The
runner has capacity 1 and is shared with an index build that can hold
it for three hours, so this is not instant -- stale for an afternoon
beats stale forever, which is what it was. And it is an eighth
workflow, so CLAUDE.md's count moves with it.
The audit stays, because a workflow that silently stops firing is the
failure mode this area has already produced once:
./scripts/issue.sh list --state closed --label "Status/In Progress"
Closes#102
CLAUDE.md said the Closes list was unreliable and to close by hand. It
is unreliable for a specific reason, and the rule can say what works.
Gitea parses commit messages that reach main. It does not parse the PR
body, which closes something only if the merge happens to copy it into
the merge commit message. Both halves were measured here: #83's merge
commit carried "Closes #9, #13, #14, ..." and closed five of the ten,
because a comma list is only partially matched; #93's merge commit body
was a lone Reviewed-on: trailer, so #92 stayed open behind a perfectly
correct Closes line in the PR description.
So the keyword goes in the commit body as a footer, one issue per line.
That costs nothing elsewhere -- Conventional Commits allows a footer,
commit-check only regexes the subject, and semantic-release reads the
type from the subject, so no release decision changes. The existing
rule that the issue number stays out of the subject is untouched and
was never about the body.
The verification step stays, because a squash or a hand-edited merge
message still drops the footer.
This commit is the experiment: if #98 and #100 close when this branch
merges without anyone touching them, the mechanism is confirmed.
Closes#98Closes#100
They do not merely lag it, they contradict it, which is worse than
absent: the only way to find out one is wrong is to trust it.
docs/dev/roadmap.md, 1648 lines. Its "Current State" describes an app
with basic playback, a track list and an album grid. Phases 1, 2, 3 and
5 have shipped entire -- playlists, shuffle and repeat, shortcuts,
virtualised lists, search, custom columns, smart playlists, the
MusicBrainz client, autotag. Its Decision Log records "Testing:
Deferred" against 836 Vitest tests, a Playwright suite on two engines
and race-detector passes in three build configurations.
.pi/journal.md, "what happened and what's next", frozen at plan 005 and
still saying everything from phase 1 onward is uncommitted. That is the
tracker's job now, and a work log that is wrong about what is committed
is a hazard rather than a stale file.
docs/dev/overview.md, a shallower and partly incorrect CLAUDE.md
architecture section. README pointed at it and points at CLAUDE.md now.
docs/dev/config-suggestions.md, eight suggestions of which one survived
-- the rest were overtaken by rewrites: applyDefaults exists, scan
concurrency is configurable with SSD/HDD detection, and httphandler.go
and the WriteHeader-after-render bug it described are gone entirely.
Nothing is lost: the genuinely unbuilt work was filed first, as #94
device sync, #95 layout customisation, #96 AcoustID and #97 the config
setters that take no lock.
Refs #98
Work has been starting from a chat message and a plan file, so two
people could pick up the same thing and neither could see the other.
The tracker is where that is visible.
Search before starting, claim before the first edit -- not before the
commit, since the point is that the other person can see the work is
taken while it is being done. If no issue covers it, open one first:
that is what makes the tracker a description of the project rather than
a description of the past.
The conventions were already right and are written down rather than
reinvented -- the Kind/Area/Priority/Platform/Reviewed/Status taxonomy,
its exclusive scopes, #73 as the roadmap, real Gitea dependencies for
hard blockers, and PR #83's body shape.
What #83 also demonstrated is that a Closes list closes nothing
reliably: it listed ten and five of them sat open in main for a
fortnight. So closing is a step you take and verify, not a keyword you
trust.
.planning/ stops being a queue and keeps design documents and measured
history -- NOTES.md, the audits, the completed plans and the arguments
in them. plans/pending/ is gone, because a plan nobody is executing is
an issue; everything unimplemented in it is now #85-#91, and each
completed plan says which issue carries its remainder. autotag.md is
kept as a historical record, marked stale where the scoring overhaul
overtook it.
The commit grammar is unchanged and is load-bearing for a different
reason, so the issue number lives in the branch name and the PR body
rather than the commit subject.
Refs #92
Issues become this project's source of truth for what is wanted and what
is already being worked on, which puts "search the tracker" at the top of
every task rather than occasionally. Fifty-odd open issues make that a
real lookup, and a lookup nobody can remember the shape of is a lookup
that gets skipped -- the same way the CI log endpoint cost two sessions
to a tool that 404s.
Text reaches the API as JSON and never as shell, which is why the
formatting half is its own Python file: an issue body is arbitrary prose
carrying backticks, quotes and $, and every attempt to build that JSON
inside the shell ends in nested quoting nobody can verify. Same reasoning
that keeps release notes out of gitea-release.sh's argument list.
Claiming is an assignment, a label and a comment together, because any
one alone is a claim somebody has to go looking for. It resolves the
comment before it mutates anything -- reading it afterwards is how a
claim ends up half-made, with the issue saying it is taken without saying
by what work -- and refuses outright if somebody else holds it.
Three API shapes are pinned here because each fails quietly:
- Labels are resolved to ids rather than posted as names. Gitea accepts
a list of unknown names with 200 and applies none of them, so a typo
reports success and does nothing.
- The dependency endpoint takes a whole IssueMeta, not an index. A body
of {"index": 88} answers 404, which reads exactly like a Gitea build
without the feature.
- close drops Status/In Progress, or a claim outlives the work.
Refs #92
The conditions were joined with " AND " and nothing else, so a smart
playlist could only ever narrow: "jazz released after 1960" was
expressible and "jazz or blues" was not, which is most of what anyone
reaches for a second rule to say.
`RuleSet.Match` is "all" or "any", and an empty match is "all" — which
is what every playlist saved before the field existed carries, so an
upgrade cannot silently widen one. ParseRuleSet rejects anything else
rather than falling through to AND, since a playlist quietly returning
the wrong tracks is worse than one that refuses to be saved.
Under OR each condition is parenthesised and under AND it is not: AND
is the tighter operator, so an OR-join has to protect a condition
carrying a top-level AND of its own — `days_since_played less_than` is
two predicates belonging to one rule.
The editor shows the choice as a sentence with the control in the
middle, and hides it while there is one rule: with nothing to combine,
all and any are the same query.
Closes#35
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Every other key in the MPRIS metadata map can be omitted safely,
because a client reading it renders a track with no title as a track
with no title. Art is different: KDE's applet treats an absent
mpris:artUrl as no news about the art and keeps drawing whatever the
last track had, so playing something without a cover left the previous
album's sleeve on screen — which reads as the wrong track playing
rather than as missing artwork.
The map's construction moves out of UpdateMetadata into a pure
metadataMap so it can be asserted on at all: everything else in this
file needs a live session bus, which is the same reason the Android
contract lives in an untagged file.
Closes#41
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`q.source` was written by SetQueue and cleared in exactly one place,
Clear, so no append path touched it: adding a track to a queue built
from an album left the page still offering "Playing from <that album>",
and since the source is persisted alongside the queue state the wrong
label outlived the session that earned it.
Every add and insert path drops it now. Removing and reordering
deliberately do not — a queue with a track taken out of it is still
that album, and the link still goes somewhere true. Only the arrival of
a track from elsewhere makes the claim false.
The delta event carries the source for the same reason it carries the
current index: an append emits nothing else, so the frontend would keep
the label it was last given until something forced a full state.
Closes#14
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The album page's tracklist badges read `libraryStatusFor(false,
track.mbid)` at render time, which is a dependency on `downloadStore`
that Lit cannot see. The page did subscribe to that store, but its
callback only assigned `canDownload` and `isRequested` — neither of
which a *track* request changes — so no reactive field moved and the
component never re-rendered. The request was filed, the plus stayed a
plus, and clicking again cancelled it.
The other three hosts rendering these badges have always asked for the
repaint in the same place, which is what made this one look correct on
inspection.
Closes#33
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Every parser in `backend/metadata` is header-only -- a few hundred
bytes and return -- so on a spinning disk a scan is not waiting on CPU
or on bytes, it is waiting on the head to arrive. Two things follow,
and the drive says which.
**How many reads should be in flight.** This was a flat 2 for anything
rotational, which is a pre-NCQ assumption: a modern SATA disk reports a
queue depth of 32 and reorders outstanding reads into the order its
head passes over them, and was being handed a quarter of what it can
use. It gets 4 now. A drive that reports 1 -- a USB bridge, a pre-2004
disk -- services one command at a time in the order given, where every
extra worker is one more seek competing for one head and the scan gets
*slower* the harder it is pushed; that keeps 2.
**And that the next seek should already be queued.** A prefetch stage
between the walk and the workers issues `POSIX_FADV_WILLNEED` over the
first 512 KB of each file -- enough for an ID3v2 tag carrying cover
art, or FLAC's STREAMINFO and PICTURE blocks. The buffered channel *is*
the lookahead: the goroutine runs 16 files ahead of the workers,
hinting as it goes, so the read a worker needs has been in flight for
sixteen files' worth of parsing by the time it asks. Rotational only;
an SSD gets the channel back unwrapped and pays nothing, since it has
no seek to hide and already has one worker per core.
`workersForProfile` is the policy on its own so it can be tested
against drives this machine does not have, and the scan logs the
device, its rotational flag and its queue depth, so the decision is
inspectable rather than inferred.
Also: `ScanConcurrency` has been a validated three-value config field
with exactly one caller, passing the constant `auto` -- so choosing
`ssd` or `hdd` by hand did nothing at all. It reads the config now.
The two modes overrule detection about the *disk* and not about its
queue, since a user who picks `hdd` on a queueing drive still wants
that drive's queue used.
What is not here is inode-ordered dispatch. It needs the streaming walk
restructured to buffer per directory, and with queueing the drive is
already reordering what the hints put in front of it; that wants a
measurement on real hardware before the complexity.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MeQt5hgXg5YGoNZQ9ozG7L
`deviceForPath` scanned `/sys/block` comparing device numbers and, when
no entry matched exactly, took the first one whose *major* agreed. Every
SATA disk is major 8. A filesystem's `st_dev` is its **partition**, so
the exact match never hits for anything on one, and the fallback then
resolved `/dev/sdb3` to whatever `/sys/block` listed first -- which is
alphabetical, which is `sda`.
On the machine this was found on that is a Samsung SSD sitting next to
the 6 TB spinning disk the library is actually on, so
`IsRotationalDisk` answered false and the scanner ran one worker per
core across a drive with one head. Matching on major alone cannot be
right on any machine with two disks, which is the case this exists for.
It goes through `/sys/dev/block/<major>:<minor>` instead -- a symlink
the kernel maintains to the device's own sysfs directory -- and climbs
to the parent when that turns out to be a partition. One readlink, no
scan, no ambiguity. The dev_t decode goes with it: Linux packs 12 bits
of major and 20 of minor split across the word, and masking the low
byte of each is right only for the first 256 of either.
`ProfileForPath` returns what the scanner needs to ask next, and the
new half is `queue_depth`: how many commands the drive will accept and
reorder at once. A SATA disk with NCQ enabled reports 31 or 32 and one
without reports 1, which is the difference between concurrency helping
and hurting. An absent file is read as "queues", because everything
that does not publish it -- NVMe, virtio, device-mapper -- is a device
where concurrency is fine.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MeQt5hgXg5YGoNZQ9ozG7L
`newServiceFixture` stops auto-pick from starting a grab, because none
of its tests is about the download and a detached `go m.grab(...)`
racing `t.TempDir()`'s cleanup is how they fail. It did that with
`MaxSizeMB: 1` -- and the size gates read `Candidate.TotalSize`, which
real providers fill and the fake leaves at zero. Zero is under every
ceiling, so the guard never fired and the race it was written to
prevent kept happening, roughly one run in fifteen:
TempDir RemoveAll cleanup: unlinkat ... : directory not empty
The guard is a format the fake never produces. Thirty consecutive
whole-package runs, none.
`TestManualDownloadSatisfiesRequestOnSuccess` was relying on the guard
being broken -- it is the one test here that wants the download -- so
it now clears the preferences itself rather than depending on a bug.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MeQt5hgXg5YGoNZQ9ozG7L
Three faults, one subsystem, and the middle one is why a request that
looked obviously satisfiable came back refused.
**The guardrails were in megabytes, which cannot mean anything.** 300 MB
is a generous FLAC single and a suspiciously small boxset, and whoever
fills the field in has no idea which release the pipeline will apply it
to. `MinKbps`/`MaxKbps`/`PreferredKbps` are the same statement divided
by how long the music is, so one number holds across a nine-minute EP
and a three-hour opera. The runtime comes from `Download.Expected`,
which every anchored request already carries, so this costs no lookup;
the rate is audio bytes over that, falling back to the mean stated
per-file bitrate when the runtime is unknown. Artwork is excluded from
the numerator, or a folder with 30 MB of scans reads as a better rip.
An unknown runtime *passes* the window rather than failing it: the
window is a statement about quality, and refusing everything the moment
MusicBrainz is missing a track length would be a silent embargo.
`MaxFileSizeMB` survives as a separate ceiling, still in megabytes on
purpose -- it is a question about disk space, and it has to apply to a
candidate whose bitrate cannot be worked out at all.
**Auto-pick required daylight over the runner-up**, 0.08 on the
combined score, and so fired hardest in the case it was never written
for: a popular album turns up five *correct* copies, all matching the
tracklist at 95%+ and differing only in format and seeders, their
scores land within a point of each other, and it refused forever on the
grounds that the choice was the user's. It was not. There was no
question about what to fetch, only about which copy -- and abundance is
the condition under which that matters least. A candidate no longer has
to beat the field, only clear the bars on its own terms; where several
do, ranking puts the one closest to the preferred bitrate first.
That tie-break needed the preference to carry weight or it would have
been decorative in a new unit: `BitrateFit` was 0.05 against format's
0.42, so asking for 320 and being handed a FLAC every time was the
designed behaviour. When a preference is set the weights shift to fit
0.40 / format 0.20 / bitrate 0.10, taking it off the two heuristics
that exist as stand-ins for the preference the user has now given.
Health and priority are untouched. And the fit spans 0.5 to 1.0 rather
than 0 to 1, so a preference can promote the copy that matches it and
can never push the others under `minQuality` -- turning "I like 320"
into "never take anything else" silently is what `MinKbps`/`MaxKbps`
are for, out loud.
**And a refusal quoted numbers that passed.** The request list built its
message from `ranked[0]` -- the best candidate *before* the guardrails
and before the lead check -- so a request killed by the size window, or
by having too many good copies, reported "best of 12 found is not a
confident enough match (match 96%, quality 88%)". `AutoPickVeto` names
the gate that actually refused, and `AutoPickable` is that returning
empty.
Existing configs: the old `MinFileSizeMB`/`PreferredFileSizeMB` are not
migrated. A number meaning "300 MB" cannot be reinterpreted as a rate
without knowing the album it was aimed at, so carrying it over would be
inventing an intent nobody expressed. Those two fall back to no window,
which is the permissive default and what a fresh install gets;
`MaxFileSizeMB` carries over unchanged, because a ceiling on bytes
still means exactly what it did.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MeQt5hgXg5YGoNZQ9ozG7L
The feature was renamed to requests and the copy was not. The badge on
every Explore card and track row still offered "Want track X", the
album page's button read "Want this" / "Wanted", the artist page's
release menu said "Want This", and the Downloads empty state told the
user to look for a control by a name nothing rendered.
The `queued` badge is a bookmark rather than an hourglass. An hourglass
says "wait, this is under way", which overstates what a request is:
nothing may be downloading, nothing may ever be found, and the list is
somewhere a user can leave one indefinitely. A bookmark says the honest
thing -- it is on your list -- and reads as the opposite of the plus
that put it there, which is what a toggle's two states have to do.
The backend's `'wanted'` request state is deliberately untouched: it is
a stored enum, not copy.
Also removes a dead duplicate branch in the badge's `render()`. The
first `if (this.actionable)` returned before the ring was built, so a
partly-held album that could still be requested drew a plus instead of
its progress arc.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MeQt5hgXg5YGoNZQ9ozG7L
`favCtrl.iconName` returned the solid glyph in both states, so "not a
favourite" was a filled heart in a duller colour and the only thing
separating the two states was hue. That fails outright for anyone who
cannot tell the two colours apart (WCAG 1.4.1), and reads as
"everything is a favourite" to everyone else.
`iconFor(favorited)` returns the outline or the fill, and the nine
`<wa-icon>` call sites split into the two cases they always were. The
three that show a *state* -- the mini player, the phone's now-playing
view, and the sidebar's marker for the favourites playlist itself --
pass it. The rest are context-menu items, which are actions rather than
states and take the outline `iconName` still returns.
`track-list` and `album-dropdown` already had this right, from inline
SVG paths of their own; this is the same rule for the call sites that
go through the icon library. `regular/star` is vendored to go with
`regular/heart`, which was already there.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MeQt5hgXg5YGoNZQ9ozG7L
The three card grids -- albums, artists, genres -- laid out with
`justify: 'center'` and a fixed 8px gap and padding, which gives the
row a fixed width and pushes everything left over to the two margins.
Measured on a 1440px window: cards 16px apart inside 78px of nothing
down each side. The outside was five times the inside.
`utils/grid-spacing.ts` computes one number instead, from what the row
could not spend on another card: the same value between two cards,
between two rows, and down each edge. That window now reads 30px
outside against 34px between, and it holds at any width.
The virtualizer has a word for this -- `justify: 'space-evenly'` with
`gap: 'auto'` -- and it cannot be used. It fits `floor(width /
cardWidth)` columns without reserving the gap it is about to need, so a
width one card short of exact leaves seven cards a pixel apart. On the
window above it would fit 7 columns with 1px between them. Deciding the
column count here is what puts a floor under the spacing.
Two consequences. The layout is rebuilt when the container width
changes the spacing rather than only when the cover size changes, so
each grid observes its own scroller -- keyed on the spacing, or every
pixel of a drag rebuilds a layout that comes out the same. And
`cover-grid`'s ScrollManager took `GRID_GAP`/`GRID_PADDING` as
constants, which stopped describing anything the moment the spacing
became elastic: it asks the host for the geometry now, since a scroll
position rebuilt from a stale 8px lands in the wrong row.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MeQt5hgXg5YGoNZQ9ozG7L
Explore's album art was almost entirely missing: 5 of 24 cards on the
shelves had a cover, and those five were the ones already on disk.
The Cover Art Archive answers `front-250` with a 307 to an Internet
Archive storage node, and those nodes are slow. Measured against the
twelve albums on Explore's own shelves, a successful fetch took 14-16 s
and a failing one 13-17 s, against a client timeout of 10. So every
live fetch died, and a timeout writes nothing and says nothing -- which
is why this reads as "Explore has no album art" rather than as a slow
upstream. The timeout is 30 s, chosen to clear the measured range: the
fetch is off the critical path, so waiting costs nothing and giving up
early costs the whole page.
Two things beside it, both found on the way.
`writeCache(mbid, nil)` has recorded "the archive has no art for this"
as an empty file since it was written, and nothing has ever read it
back: `readCache` returns "" for an empty file, which is
indistinguishable from a miss. So every art-less release group was
re-fetched from CAA on every render that asked about it. A third of the
shelves are art-less, so that was a third of the page spending a live
request to be told again what the last one said. `knownMissing` reads
it, on both the release-group and the release path.
And the frontend marked a failed fetch as permanently answered for the
session, so a timed-out cover never retried within it. It drops the
marker instead; a genuine 404 is now answered from disk, so re-asking
one costs nothing.
Measured after: 23 of 24.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MeQt5hgXg5YGoNZQ9ozG7L
A branch push and its pull request are the same commit. With
`branches: ['**']` alongside `pull_request:`, opening a PR booked four
runs -- check and e2e for the branch, then both again for
refs/pull/N/head -- and this host has capacity 1, shared with an index
build that can hold it for three hours. PR #1's own checks queued two
runs deep behind exactly that.
`pull_request` covers feature branches. `main` stays because a
post-merge run is the record of the trunk's health, and now that main
refuses direct pushes it happens exactly once per merge.
The trade is that a branch pushed with no PR open gets no CI. That
matches the workflow this repo just committed to, and the signal returns
on the same commit the moment a PR exists.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
Two guardrails for the 2026-08-17 incident, and one is not about CI.
index-artifact.yml's `push` trigger was commented out that day with a
note to restore it once the rebuild completed. Restoring it is the bug.
A refresh is individually cheap, which is what made the trigger look
free; what it actually did was put an unattended job that mutates the
only copy of a ~205 GB catalog on the same trigger as an ordinary code
change, on a runner with capacity 1. The rule the file now states is the
general one -- a job that mutates state which cannot be rebuilt in ten
minutes is triggered deliberately -- so the next such job has somewhere
to look. The cron and workflow_dispatch lose nothing: indexbuild resumes
from its checkpoint either way.
Note what no branching or PR gate would have caught here. That change
was green on its branch, green on the merge and green on main; the fault
existed only against the persistent /cache database, which no fixture
reproduces. Code is gated by CI, irreplaceable state by refusing to
touch it and by docs/index-cache.md's restore.
The other half is the mismatch that started this: two harnesses reading
two files. AGENTS.md is a symlink to CLAUDE.md and skill-check asserts
the symlink rather than comparing contents, because a copy would satisfy
every other check in this repo while silently drifting -- which is the
failure being prevented. The same check now scans CLAUDE.md for make
targets, which it never did: 27 targets named in the file agents trust
most, none of them verified. Coverage goes 19 -> 46.
Scanning prose meant the line-start rule needed a fence. "Two green
branches do not / make a green merge" wrapped onto a line beginning
`make a` and duly failed on a target called `a`. Inside a fence it is
code; outside one it is a sentence that broke there, and a check that
fails on reflow gets disabled rather than fixed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
Two mechanisms shipped today whose invariants are not visible from the
code, and one of them has already cost a rebuild.
Credits: why join phrases are assembly instructions rather than
disassembly ones, why credited_name is stored per row instead of joined
from artists, why the lookup is keyed on the recording MBID (and so
needed no local table), why an absent credit is cached as an answer,
and why the decomposition comes from a third dump at all — the
canonical dump has no join phrases and the JSON dumps overlap a real
library by zero rows. The measurements that justify the feature are
here too, including the correction that the "3 of 2,823" figure behind
plan 013 measured our own writer rather than any library.
The stale-shape repair gains the paragraph it should have shipped with:
retiring a Cache table is a build-tag decision, because the app
downloads its catalog and cmd/indexbuild derives it. Written as what
happened rather than as advice, since it dropped the real CI catalog on
its first run and the shape mismatch it found was there by design.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
The stale-shape repair dropped the CI catalog on its first run:
retiring a table ... table=explore_index
reason="column entity_type is TEXT, schema declares INTEGER"
index maintenance mode=build reason="no completed import yet"
The mismatch was real and the drop was correct by the app's rule: a
client's catalog is *downloaded*, so a wrong shape costs a minute of
re-fetching the artifact, while keeping it costs every Explore read.
It is the wrong rule for one database. cmd/indexbuild's catalog is not
downloaded, it is what the artifact is cut from — the only way back is
the ~205 GB dump stream the /cache volume exists to avoid. And that
database is deliberately kept in the older encoding, which
`fix(indexexport): read an index older than the binary` exists to
tolerate, so the shape does not match by design and would have been
dropped on every run.
retireLibraryTables, right beside it, never touches the catalog for
exactly this reason. The repair reached past that protection because it
runs inside database.NewDB, which cmd/indexbuild also calls.
So the policy is a build tag, which is how this project already tells
the index tools apart (runtime_indexbuild.go, servicestartup.go,
dumpbuild_stub.go): Cache tables are rebuilt in the app and never in
cmd/indexbuild. Owned and Derived are still repaired in both — that is
the half this database can safely discard, and retireLibraryTables
already discards it.
The residual trade is deliberate: a future explore_index column will
now fail the index job loudly on applySchema rather than silently
costing it a 205 GB rebuild. A human should decide that one.
TestTheCatalogSurvivesAStaleShape is the accident, symptom first, with
the shape the real database is in — every current column, ids and
entity type still text. It fails with "the catalog was retired" when
the policy is flipped back.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
TestManagerEndToEndAutoPick waits for StateComplete and then asserts
that staging was released and the library was rescanned. Those happen
*after* the state is recorded: manager.go sets StateComplete, then
satisfies the request, then releases staging, then scans. So waiting on
the state is not waiting on either assertion, and on a loaded machine
the worker is descheduled in between and the test reads the world one
step too early:
manager_test.go:209: staging not released: 1 dirs remain
manager_test.go:218: library scans = 0, want 1
It passed alone every time and failed three times under a full-suite
run, which is the signature of a test race rather than a broken
manager — nothing here is wrong except what the test chose to wait on.
It blocks pushes, since the pre-push hook is exactly the loaded run.
It polls for the side effects now, through the waitFor this package
already has and already uses for the same reason one file over
(service_test.go waits for a request to become satisfied after the same
StateComplete).
Not reproduced on demand: eight spinners and -count=5 did not provoke
it with or without the fix, so this rests on the ordering being plain
in the code rather than on a red-to-green demonstration.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
The phone shell's now-playing view landed on main while the credit
rendering was being written, so it arrived with the one call site that
still showed a multi-artist credit as a single link with the other
artists as punctuation inside it.
It is the same fix as the other ten: render from the parts, fall back
to the single link when there are fewer than two. The subscription is
what makes it show up at all — credits arrive after the track does, so
the name already on screen has to be re-rendered when they land.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
`applySchema` is CREATE ... IF NOT EXISTS and there is no migration
chain, so a *changed* table never migrates: the statement silently
no-ops against the old shape. Two plans had already landed on that, and
neither showed up in a test because a fresh install is perfectly
healthy.
- 014 added `total_tracks` to explore_index and to `indexRowFields`,
the projection every explore read uses, so every search, browse,
artist page and album page failed with "no such column: total_tracks"
on any database that already had a catalog.
- 013 reshaped audio_files, so applySchema could not run at all and the
app did not open.
staleshape.go runs before applySchema and drops what disagrees, so the
create is a create. It parses sql/schemas/ for the expectation rather
than writing the column list down a second time, and it notices a
changed *type* as well as a missing column — 013 moved mbid TEXT to
BLOB, which no ALTER could express and which SQLite will not coerce, so
a query against 16 raw bytes returns no rows rather than an error.
Only Authored tables are exempt. Cache is rebuildable by definition,
Owned is what a rescan rebuilds (plan 013's stated "delete and
rescan"), and a table the schema no longer describes at all goes too --
013 left seven behind plus schema_migrations.
Three things in it are load-bearing, and each was a bug first:
- The parser read `UNIQUE(mbid)` as a column, which made a healthy
catalog look stale. That would have retired it on every launch and
cost every user an artifact download per start.
- The drops are one transaction with defer_foreign_keys. Those legacy
tables reference each other, so any order fails on whichever goes
first; turning foreign keys off instead would suppress
playlist_tracks.audio_file_id's ON DELETE SET NULL and leave entries
pointing at ids a rescan reissues to *different songs*. Nulled
entries are empty; stale ones are wrong, and wrong quietly.
- The order is sorted, so a failure reproduces. Map order is random,
and the foreign-key bug passed its own regression test on two runs in
three until the order was fixed.
Verified against a real pre-013 install: it opens, its 22 playlists
survive, 1,887 linked playlist entries become 0 rather than dangling,
and the legacy tables are swept.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
Every artist name in the app went through `artistLink(name, mbid)`, so
a track credited to several artists rendered one link and the rest as
punctuation — "2Pac feat. Snoop Dogg" linked 2Pac and left Snoop Dogg
as text inside it.
`creditLink(parts, fallbackName, fallbackMbid)` renders the credit from
its parts: one link per credited artist, join phrases as plain text
between them. The link boundaries are known by construction, which is
the point — locating a name inside the stored credit string would
reintroduce the mismatch the catalog exists to avoid, since that string
may come from the file's tags while the parts come from MusicBrainz and
the two disagree for ~1 in 3 multi-artist credits.
Fewer than two parts falls through to the previous behaviour exactly,
so a single-artist credit, a file with no recording MBID and a catalog
that has not answered yet all render as they did before. Nothing tries
to split the fallback string: "Simon & Garfunkel" is one artist, which
is why primaryArtist() does not split on "&" either.
The lookup is keyed on the recording MBID, which both sides already
carry — a catalog row has one and so does a local file — so one binding
serves Explore and the library's own lists, and no local table is
needed for this.
credit-store.ts, and three things in it are load-bearing:
- A miss is cached as an empty array. The backend returns nothing for a
single-artist credit, which is ~87% of tracks, and caching only the
hits would re-request the rest on every render forever.
- request() is per-row and coalesces into one call per frame. A
virtualized list cannot hand over "the whole list": 50,000 rows would
be 100 queries for the ~30 on screen.
- It is an LRU with a counted retainedChars probe, because a cache that
grows with use is a leak with a schedule.
The virtualized lists push requestUpdate() into the virtualizer rather
than only the host, since its rows come from its own properties — a
host update alone would leave them exactly as they were. now-playing
marks its geometry dirty instead, because the marquee measures the text
it is about to scroll.
track-list keeps the single link while a search term is active: the
highlight spans are computed against the flat credit string, and
mapping them onto decomposed parts is a different problem.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
A track credited to more than one artist has exactly one navigable
artist in this app and the rest are punctuation. `primaryArtist()`
string-parses the credit, strips a " feat. " clause and discards the
guest; it deliberately does not split on "&", "with" or "," because
those live inside real artist names.
Measured on a real 26,069-file library plus an 80+80 MusicBrainz
sample: 13% of recordings are multi-artist upstream, while only 0.86%
of files carry any structured multi-artist tag — mp3 carries zero
files with multiple MUSICBRAINZ_ARTISTID across 19,840. Of 1,286 files
saying "feat.", 90% have nothing structured behind it, and a sample of
80 such files was multi-artist in MB 80 times out of 80.
CLAUDE.md justified plan 013's removal of the credit tables with "3
credits of 2,823 listed more than one artist". That measured our own
*writer* — cachedLinkArtist was called once per credit, so a
collaboration could never have been recorded. Dropping the join table
was still right on cost; the evidence for "multi-artist is rare" was
not.
A credit is ordered parts and the credit string is derived from them,
so join phrases are assembly instructions, not disassembly ones.
Nothing here reconstructs a credit by searching a name inside a credit
string: the stored text may come from tags while the parts come from
the catalog, and those disagree for ~1 in 3 multi-artist credits.
Where it comes from, after two dead ends: the canonical dump CI
already streams has no join phrases and no as-credited names, and the
JSON dumps cover 153,691 recordings of ~35M with *zero* overlap
against a real library. So mbdump.tar.bz2 — 7.1 GB, ~13.7 min in
pure-Go bzip2, whose members are alphabetical, which is what lets one
pass resolve an entity's credit without buffering 35M recordings.
- artist_credit_part / artist_credit_ref, multi-artist credits only:
a single-artist credit is already explore_index's own artist_name.
- Column layouts verified against the real 20260815 export;
ErrDumpShape makes a wrong guess a failed build, not a wrong catalog.
- The pass runs on every mode, not just a build. The job picks its mode
from the index's own state, and a complete import means "refresh",
which never enters the importer — so credits could otherwise only
arrive via a rebuild that re-downloads ~205 GB. It reports whether it
populated anything, which is what flips `changed` and republishes.
- The importer asks whether an artifact carries the tables, on the
writer where `core` is attached, so the artifact already published
still imports.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
21 conflicts, all from the same cause: three features were developed on
both lines and this branch's copies are the ones adapted to v3's
bindings and to the file-shaped schema. Resolutions:
- `frontend/wailsjs/` stays deleted — v2's generated bindings, replaced
by `frontend/bindings/`.
- remove-from-library, `library-status.ts`, the requested-badge spec and
its component test: took this branch's copies, which differ from
main's only in calling `pruneEmptyEntities`/`CountAudioFiles`,
importing `@go/download/models.js`, and staging a real UUID for the
catalog's `CHECK(length(mbid) = 16)`.
- `GetFilePathsByRecordingMBIDsByLibrary` dropped: it joined
`recordings`, which no longer exists, and `library_id = 0` answers
both scoped and unscoped now. `GetAudioFilesByPaths` was already here.
- The album page, the artist page and the library badge kept this
branch's versions, which supersede main's: ownership asked once from
the files, the partial-completeness ring, and the request action.
- Docs: no migration chain (013) over main's two-file column rule and
its pre-1.0 squashing note, both of which 013 retired. Kept main's
`CreateSmartPlaylist` read-pool example, which is a real second
instance of that bug.
Verified on the merge result, not on either parent: lint clean in all
three build configurations, `make test` green in all three, 776 Vitest
tests, `tsc --noEmit`, bindings-check and skill-check clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
Plans 013 and 014, the album page that prompted them, and the smaller
fixes they turned up. Changelog, largest first.
## The local library is shaped like files, not like MusicBrainz
`audio_files` carries its own tags and points at `albums` and
`artists`; `file_genres` is the one real many-to-many. `recordings`,
`release_group_recordings`, `artist_credit`, `artist_credit_artist`,
`recording_genres`, `release_groups` and `release_to_rg` are gone from
the local side, and with them a six-way join in every read, a
`MIN(release_group_id)` subquery in eleven queries and a
first-credited-artist subquery in nine. Measured on a real 25,966-file
library, every many-to-many that model expressed was 1:1 in the data.
- Ownership is a file. `GetFilePathsByRecordingMBIDs`,
`LibraryMBIDIndex.CheckMBIDs`, `collectLibraryEntities` and
`pruneStaleLocalCrossReferences` all join `audio_files`, so the 812
orphaned recordings, 216 release groups and 260 artists that library
carried are now structurally impossible.
- One projection: every track query selects from the `track_metadata`
view, one row type, one mapper. Nine hand-rolled copies had drifted
far enough to report different years on different screens.
- `library_id = 0` means every library, so each list query exists once
instead of scoped and unscoped with a branch at every call site.
- No migration chain. `sql/schemas/` is the one description of the
shape; `sql/migrations/`, `applyMigrations` and `schema_migrations`
are squashed away, along with the drift between them that had sqlc
generating against a stale schema.
- `database.InsertTestTrack` is the one test seeder; twenty test files
had been assembling the old FK chain each in its own order.
## The catalog stores its ids as bytes
`explore_index`'s three 36-char MBID columns and its entity-type text
are 16 raw bytes and a small integer. The table and its six indexes go
780 MB to 405 MB on a real 2,052,200-row catalog, which is why a fresh
install is ~0.6 GB rather than ~1.0 GB.
- `backend/explore/mbid.go` is the only place the encoding is known;
everything above it speaks dashed strings.
- `CHECK(length(mbid) = 16)` makes a stringly write fail at the insert
rather than silently returning no rows, since SQLite does not coerce
between TEXT and BLOB.
- The importer asks the artifact what encoding it carries and converts
on the way in, so the artifact already published keeps working and no
format bump is needed.
- `indexRowColumns`/`scanIndexRow` replace four copies of a 22-column
list, and `TestStoredEncodingRoundTrips` sweeps every read path.
## An album page that says how much of the album is yours
- One question, asked once: is there a file. `filePaths` is filled by a
single batched lookup when the tracklist settles, and the badge, the
Play count, the dimmed rows and every menu item read it — replacing
four claims of decreasing confidence that could show a green tick on
an album whose every action did nothing.
- Play, Play 7 of 12, or no play button at all.
- `total_tracks` on `explore_index` (~2 bytes over 400,677 release
groups) and on `audio_files` from tags that have always carried it:
a complete MBID-matched album now makes no catalog call at all, where
it used to spend the most expensive request the app makes.
- A merged cluster shows the running order the most releases agree on,
and the version list marks the release you own rather than standing a
synthetic entry in for it.
- `AlbumReleasesFailed`: a slow fetch is no longer reported as a failed
one by a 12-second timer.
- Rows not in the library are dimmed in place (with `aria-disabled`)
instead of the owned ones wearing a green tick and a legend.
## Caches and cover art get ceilings
- Only the three tiers of a cover are stored; the full-resolution copy
nothing rendered was 1,134 MB of a 1.4 GB covers directory.
- One artist portrait is downloaded and the rest are remembered as
URLs — 4.1 GB of a 5.3 GB cache was candidates no code path reads.
- `browsedArtBudget` and `httpCacheBudget` bound what an age cannot:
the same install held art for 5,770 artists in a 1,301-artist
library.
- `OrphanedArtistImagesJob` joined a bare MBID onto a sharded
directory, so it deleted the rows that were the only record of the
files it left behind. `explore.ArtistImageDir` is that layout's one
definition now.
## The autotag queue asks whether there is work
`tagging_items` was a row per album folder, not a queue, and no query
read the `tag_status` column that held the answer. The four queue
queries ask the files, which matters most where it is least visible:
`startPrefetch` was scoring every album in a tagged library against
MusicBrainz.
## Phantom playlist tracks resolve in place
An M3U8 imported before its files leaves phantom rows; they now match
by path and fall back to position, keep their place in the playlist
when resolved, and pair best-first so two phantoms cannot claim the
same file.
## Playing a track plays the list it is in
Double-click, and Play on a single row's menu, queue the list as
displayed with `startIndex` on that row — the album page and the track
list used to queue one track and discard the album around it. A
multi-row selection still plays exactly itself.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
CLAUDE.md gains a Packaging section for the four Taskfile facts the
recipes just needed — wails3 on PATH by bare name, no -ldflags on
`wails3 build`, bin/ not build/bin/, and bundling as its own step —
plus how build/'s platform metadata generates from build/config.yml and
what that refresh overwrites.
Its lifecycle, bindings, harness, events and CI sections were still
describing v2. The events one matters most: the rule to emit through
events.Emit survives, but its justification is now the weaker one, and
saying so is the point of the migration. v2's runtime.EventsEmit
log.Fatalf'd on any context not carrying the runtime; v3's emit takes
no context at all, so what is left to pin is that one emit path is what
lets emitStatus drop an unchanged payload for every caller at once.
README told a contributor to `go install wails/v2/cmd/wails` and
apt-get libgtk-3-dev/libwebkit2gtk-4.1-dev; the CLI is vendored and the
stack is GTK4 + WebKitGTK 6.0. Two comments claiming Xvfb and one
claiming frontend/wailsjs go with them.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
Neither packaging/arch/PKGBUILD nor the Homebrew formula had been run
since Phase 1, and both were still calling v2's CLI: `wails3 build`
takes -tags, -obfuscated and -garbleargs and nothing else, so
`-clean -trimpath -ldflags` fails at the flag parser. Both also
installed from build/bin/, which is v2's output path — v3 writes to
bin/, and build/ is tracked build assets now.
Three more things the tree needs that neither recipe had. The tasks
invoke `wails3` by bare name, so scripts/toolbin has to be on PATH or
the build dies at its first sub-task. `wails3 build` has no -ldflags at
all, and build:native computes BUILD_FLAGS in its own vars: so a CLI
variable cannot override it — LDFLAGS_EXTRA is appended inside the
production -ldflags string instead, on linux and darwin alike, empty by
default so make build-dev/build-prod are unchanged. And bundling is a
separate step from building: `task build` produces a bare binary on
both platforms, so the formula's macOS path runs `task package`.
The build assets were the scaffold's, not this app's. Info.plist named
CFBundleExecutable `yjref` and com.example.yjref, nfpm packaged
./bin/yjref, the .desktop template said "A yjref application" — an .app
built from that plist would not have launched. They generate from
build/config.yml, whose info block had never been filled from
wails.json either; `wails3 task common:update:build-assets` is the fix.
nfpm's homepage and license are not derived from it and are set by
hand, which is noted in place, and the refresh regenerates build/ios
and build/android, which this repo does not carry.
arch-package.yml's pacman list moves to webkitgtk-6.0/gtk4 to match the
PKGBUILD's depends(): makepkg installs nothing itself, so a mismatch
fails at link time rather than at check time.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
`make sandbox`, `make dev`, `make build-dev` and `make build-prod` all
died with "/bin/sh: wails3: command not found". `wails3 dev` and
`wails3 task` are supervisors: they run the scaffold's Taskfile tree,
which invokes `wails3` by bare name in 54 places across four files. The
CLI is a vendored Go tool by design (plan 009, D3 — a global install
would be this build's first undeclared dependency), so that name did
not exist.
scripts/toolbin/wails3 execs `go tool wails3`, and the Makefile
prepends that directory only for the targets that start a supervisor.
Rewriting 54 scaffold call sites would be churn to redo on every
scaffold refresh; nothing global is installed either way.
The shim does not cd. The first version did, to be sure `go tool` found
the module — it does not need to — and that silently discarded the
`dir:` a task had set, so generate:icons failed with "open
appicon.png: no such file or directory" against a file that was there.
Three things the build path needed once it got that far:
- `frontend/package.json` gains `build:dev`, which build:frontend runs
under DEV=true and which did not exist.
- Vite binds 127.0.0.1. It defaulted to `localhost`, which resolves to
`[::1]` only here, while wails3 dev's asset proxy dials IPv4 — so the
first request for the dev server was refused and the first paint
raced a retry. Zero proxy errors after.
- The icons and the .desktop file are generated on every build.
icons.icns/icon.ico are deterministic from our appicon.png (verified
by regenerating), so the regenerated pair is committed and the churn
ends; .task/ and the .desktop file are ignored.
Also corrects a claim: build-prod strips and trims but does **not**
UPX-compress — that was v2's `-upx` flag. Phase 1 recorded UPX as
still working, but neither build target had been run.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
make e2e is green on chromium: 92 passed. The harness is rebuilt on
what v3 actually offers, and three of the four things it replaced turn
out to be better than what they replaced.
The headless launch is v3's own server mode. scripts/dev-headless.sh
ran a `-tags dev` binary whose app_dev.go parsed -devserver/-assetdir
out of os.Args; that file went with v2, so the harness had no server at
all. `-tags dev,server` is a first-class mode and needs no display, so
Xvfb is gone from the script and from CI.
The bridge hooks two places, neither of them EventsOn. Inbound is
window._wails.dispatchWailsEvent, wrapped by pre-creating the object
the runtime keeps and putting an accessor on the one property.
Outbound is fetch: v3 routes every runtime call through one POST, so
the bridge sees binding calls and event emits from any module, needs no
walk of an object graph, and cannot miss a call made before it looked.
__yjEvents.call posts to that endpoint by method name, so it depends on
nothing in the app's bundle and works on a page with no init script.
That is what lets seed-sandbox.sh drop playwright-cli entirely — it
drove AddLibrary through a browser only because window.go was v2's one
way in — and with it a global npm install and a second Chromium in CI.
measure.mjs and one spec lose their window.go walks and read the
bridge's log instead; e2e/support/method-ids.mjs derives id -> name
from frontend/bindings/ (phase 6b option 1, so it cannot go stale
silently). Plain .mjs because measure.mjs runs under bare node and one
derivation beats two that can disagree.
Four bugs surfaced, and the migration is how.
The cross-service wiring never ran headless. It hung off
Common.ApplicationStarted, which server mode never emits —
setupCommonEvents is an explicit no-op there — so the queue had no
TrackLoader and playing a track changed the queue and then silently did
nothing. It is a service registered last now (backend/startup.go):
services start in registration order, which is the ordering the wiring
needs, in every mode.
Six specs called SetQueue with 3 of its 4 arguments. v2 accepted that
and filled the gap; v3 answers "expects 4 arguments, got 3".
requested-badge's cleanup read window.go and returned early on
`if (!svc)` — the silent cleanup its own comment was written to
prevent, one migration later. It posts to the runtime endpoint now,
which any page can do.
SearchIndex.Search trusted a startup latch, so rows a spec staged
afterwards were unsearchable and three specs passed only when an
earlier one happened to flip it. shelves.go fixed exactly this and left
hasCatalogRows behind; the search path now uses it as the fallback,
with the latch still the fast path.
Two spec edits are deletions of assertions about v2. harness.spec
checked Object.keys(window.go) and that a bad call *hung*; it now
checks the real runtime is loaded and that the backend rejects with a
TypeError naming the argument. album-actions asserted a tracklist
legend that dcc40b1 deleted on main — that spec has been failing since,
and what replaced it is covered in frontend/test/components.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
v2 installed two globals and the fake replaced both. v3 has neither —
the runtime is an npm module and the generated bindings call into it.
What it has instead is better: setTransport() is a public seam for
replacing the IPC transport, and *every* runtime call goes through it,
so the fake is smaller than v2's and covers strictly more.
The event dispatcher is no longer mirrored at all. v2's fake
reimplemented desktop/events.js — the listener list, maxCallbacks
expiry, the reverse iteration — because there was no way to reach the
real one; emit() now goes through window._wails.dispatchWailsEvent,
which is the entry point the backend's own push uses. What is mirrored
instead is one line of Go: how EventManager.Emit packs variadic data
into an event's single data field. Registration and unregistration are
the public Events API. The one non-public thing left is the listener
registry, aliased in vitest.config.mts and used only by
listenerNames() — a test asks whether importing a store subscribed it,
which nothing public can answer.
A binding carries a method ID, not a name, so the fake derives the
ID -> path map from the generated tree: FNV-1a over the FQN, with the
Go type's casing recovered from each package's index.ts, which is the
only place it survives (library/library.ts cannot tell you it is
FrontendUtil). The map has to be complete rather than lazy because 21
assertions read calls() with no argument and compare the whole list.
Two things had to move that are not the fake.
fixture() drains microtasks between two renders: a v3 binding settles
several hops later than v2's, and tests were already written as though
fixture() meant "mounted and loaded". Microtasks and not a timer,
which would hang under the suites that install fake ones.
tracklist-store keeps its defaults on an empty answer instead of
emptying the column list. GetTrackListColumns substitutes
DefaultColumns only when the whole config section is missing; a section
that exists with no columns returns nothing. Until now this was
accidental — the binding was typed Column[], an absent answer arrived
as undefined, and .map threw into the catch.
757 tests pass across all 63 files. They are run in batches: a single
browser session dies partway through the 58 it queues, which reproduces
unchanged at the pre-migration commit and is a resource limit on this
machine rather than anything here.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
frontend/wailsjs/ is deleted and frontend/bindings/ takes its place —
a real TypeScript module tree nested by Go import path, generated by
wails3's static analyser rather than by building the app and running
it. The @go alias absorbs the constant prefix, so a call site imports
'@go/library/library.js' and the codemod over all 93 sites was a
specifier rewrite plus splitting @go/models' namespaces into one
import per package.
The 12 SetContext bindings and the fake `context` model are gone, as
Phase 2's ServiceStartup port promised: 272 methods across 12
services, none of them plumbing.
@runtime/runtime is now a local shim (src/wails/runtime.ts) over
@wailsio/runtime, so the 22 EventsOn imports are untouched. It
unwraps v3's WailsEvent into v2's callback shape, which is exact here:
nothing in backend/events passes more than one data argument, and v3
only packs arguments into a slice when there is more than one.
v3 tells the truth about two things v2 lied about, and that is most of
the diff. A Go nil slice really does arrive as JSON null, and a Go
named string type really is an enum; v2 typed them as T[] and string.
utils/binding.ts states the app's actual contract — an absent list is
an empty list — once, at the boundary where it is true, and also drops
the CancellablePromise the app never cancels. Four test fixtures
widen an enum field back to its value union.
Not done, and Phase 5's to fix: frontend/test/support/wails-fake.ts
still fakes window.go, which v3 does not have, so `make ui-test` is
broken and harness.test.ts fails to compile on EventsEmit. That test
also asserts v2 ordering that no longer holds — v3's Events.Emit calls
the backend and does not notify in-page listeners at all.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
The commit before this removed the tag from the Makefile, lefthook,
both packaging recipes and CI, but left it in CLAUDE.md's "Running
tests" section and the yellowjacket-dev skill — which are the copies a
coding agent actually runs, so a stale tag there is worse than one in
prose. skill-check does not catch this: it verifies that documented
make targets exist, not that documented go commands do.
The historical mentions in .planning/ and .pi/journal.md are left
alone; they are records of what was true then.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
Phases 2 and 3 of plan 009, plus the parts of phase 1 that could not
land before them. Nothing in the tree imports wails/v2 any more; all
three lint and test configurations are green and `go build .` produces
a running binary.
The point of the migration is one file. backend/events/emit.go probed
ctx.Value("events") — a v2-*private* context key — to decide whether
emitting was safe, because runtime.EventsEmit called log.Fatalf on a
context without the runtime and took the process down with it. v3's
emit takes no context, so that is now application.Get() == nil. D1
held: events.Emit keeps its ctx as the WithSink test seam, and all 45
call sites and 7 test files are untouched.
The bootstrap splits into application.New + Window.NewWithOptions +
Run. Ten bound services implement ServiceStartup instead of being
handed a context by hand from OnStartup, which also stops ten
SetContext methods being exported as bindings. jobs.Registry and
explore.SearchIndex keep theirs — neither is bound, so converting them
would be churn for no binding removed.
Four things differed from the plan and are written up in it: GPU policy
moved to the per-window LinuxWindow options rather than surviving on
LinuxOptions; there is no OnStartup/OnDomReady option, so app-level
wiring hangs off ApplicationStarted; application.NewService is generic,
so FEBindings []any could not survive (the binding generator is a
static analyser and would have seen nothing); and the quit veto had to
be restructured, because v3's dialog answers on a callback rather than
returning the button, so ShouldQuit vetoes, asks, and quits again from
the callback.
Window state saving moves to a WindowClosing hook — the size has to be
read while the window still exists, and v3's OnShutdown has neither
context nor window. backend/logging is deleted rather than ported:
v3 takes a *slog.Logger directly, so the v2 logger.Logger adapter had
no caller left.
Phase 1's tail rides along, now that it can: the Makefile's wails
invocations, all 50 webkit2_41 sites, lefthook, both packaging recipes
and ci.yml's apt lists. v3 builds against GTK4 + WebKitGTK 6.0, which
Arch and ubuntu:24.04 both ship, so the tag is a deletion rather than
a translation.
Phase 4 is next and the branch is not usable until it lands: the app
builds, but frontend/wailsjs/ is v2's tree and nothing regenerates it,
so the frontend cannot reach the backend yet.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
The plan assumed build/ was free and that GTK4 was a preference. It
was not free — this repo used it as ignored build output — and GTK4 is
not available on the dev machine, which breaks `go tool wails3`
outright rather than merely changing which webkit is linked.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
Phase 1 of plan 009. The app still builds and runs on v2 — nothing in
main.go or backend/ has moved yet — but the v3 CLI, its pinned runtime
and its build-asset tree are now present, which is what Phase 2 needs.
- wails/v3 v3.0.0-beta.8 pinned per D5, and wails3 added to the `tool`
block beside the v2 CLI per D3. Both are vendored; neither is a
global install.
- build/ now holds the v3 build assets, copied wholesale from a
`wails3 init` scaffold rather than hand-written. That collides with
this repo's existing use of build/ as ignored build *output*, so
.gitignore narrows to build/bin/ and bin/ and the assets are tracked.
The mobile platforms are not carried: this is a desktop player
(MPRIS over D-Bus, beep, XDG paths) and cannot target them.
- build/config.yml's info block is filled from wails.json, which stays
for now because the v2 CLI still reads it.
- Taskfile.yml defaults PACKAGE_MANAGER to pnpm, since the scaffold
assumes npm and frontend/package.json.md5 is part of the dep-caching
scheme.
One deviation from the plan worth recording. webkitgtk-6.0 is not
installed on this machine, so the default GTK4 path is unavailable and
the gtk3 fallback is in use. That also means `go tool wails3` does not
work — the CLI itself fails to compile without webkitgtk-6.0 — while
`go run -tags gtk3 .../cmd/wails3` does. The Makefile rewrite in the
next commit has to account for that, and it goes away once the GTK4
dependency is installed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
Three things on the Explore surfaces, all about not asking twice.
A portrait already on disk costs no network call. explore-view seeded
only from the library store — owned artists, which on a catalog search
is nearly none of the results — and sent everything else to
GetArtistImageURL, the resolving entry point, one await at a time.
GetArtistImagesCachedPaths asks the disk about every unresolved artist
in one call, and only what it does not answer reaches the resolver,
in parallel.
The artist page's two sections both wanted PrefetchReleases and each
called it, so the most expensive call the app makes was issued twice
for an overlapping set on a 1 req/s limiter. They are collected and
sent once on a microtask, and prefetchRequested stops the cold-artist
refetch re-asking for what it already asked for.
The release cards — most of the artist page — had no context menu at
all. They have one now on both release shapes, normalised to a
ReleaseMenuTarget when the menu opens so the union does not reach the
action handlers. It is a discriminated union rather than one nullable
field per kind because the panel is shared with the track menu: that is
what keeps aria-label moving with the target, which is the fault
cover-grid shipped. Which items appear is three different questions —
playback is gated on a local album id, not on "owned", and the request
needs a catalog MBID, so it is absent for a library-only release.
Note on the docs: the CLAUDE.md and NOTES.md prose here was
reconstructed after a mishandled `git stash --keep-index` destroyed the
uncommitted originals. One NOTES.md section is marked as incomplete
where its text could not be recovered.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
The post-scan backfills share MusicBrainz's rate limiters with every
page the user can open, and both were FIFO — so a thousand-artist
enrichment put an album page behind an hour of queued work.
WithBackgroundLane/WithBackgroundPriority add a slower second lane: a
marked wait takes no token while any interactive wait is outstanding.
It is a context marker rather than a parameter because a backfill calls
the same client methods a detail page does. A long backfill also has to
be visible and stoppable, so jobs.KindCatalogEnrich registers both with
progress and cancel — after the work is counted, since these passes are
a no-op on every launch once the library is covered.
What it does not fetch is the point. It ran for hours against a
900-artist library and marked nothing, because three of the four things
it did per artist were work nobody asked for: similar artists, which
the artist page already resolves on view, and a full GetArtistImage
(fanart.tv, TheAudioDB, Wikidata, Wikipedia, ten portraits) reached
only to warm the MB artist lookup EnsureArtistRels does alone. It was
also serial across artists while every limiter is per-host and idle.
The marks are a table rather than more explore_index columns, because
artifactimport merges by column list and a flag added there is a second
place to remember. BrowseReleaseGroupsAll pages to exhaustion, where
the old call silently cut a prolific artist at 100 release groups.
One portrait is downloaded now; the rest are remembered as URLs.
resolveAllSources downloaded every candidate, up to ten, full size,
while nothing reads anything but primary.jpg — 5.3 GB measured on a
real cache, 4.1 GB of it unreachable. OrphanedArtistImagesJob is why
that survived: it joined the bare MBID onto the images directory, but
artist directories are sharded under a two-character prefix, so it
named a path that never existed and deleted the rows that were the only
record of the files it left behind.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
Every write goes through one connection — MaxOpenConns(1), because
SQLite has one writer — and a background pass can hold it for a long
time. The player and the queue wrote inline from paths that hold their
own mutexes, so a contended writer did not merely slow persistence
down: SetQueue blocked in LoadFile's saveState and then in
persistState, while holding q.mu and p.mu.
That is the exact shape of the report: the track changed and the
transport sat at paused, nothing appeared in the queue, and the play
button did nothing because Queue.Play waited on the same held q.mu.
Diagnosed by profiling the running app — 91% of its CPU was
BackfillLibraryDiscographies → upsertBatch, with four of its six
workers parked in sql.(*DB).conn.
Jobs now run in submission order on one goroutine per component, each
carrying its own snapshot. A job must not touch the component's fields
— it holds no lock and the state has moved on — which is why
persistTracks clones. SaveState still flushes and waits, because that
is the one caller for which the row has to exist on return.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
RemoveFromLibrary deletes the audio_files rows the way the scan's own
orphan cleanup does and records each path in excluded_paths. The
exclusion is not an enhancement: without it the next scan finds the
file, sees no row and imports it again, so the button undoes itself.
The soft scan compares files on disk against rows in the database, so
surveyAudioFiles and countAudioFiles both take the exclusion set —
otherwise an excluded path makes the two disagree forever and queues a
full scan on every launch. Deleting a row cascades to queue_tracks, so
the removal calls the same CompactQueue hook RemoveLibrary does.
Also lands the requested badge: library-status-indicator is a button
again where it can act, utils/library-status.ts states once what owning
and wanting mean, and the long-declared queued state finally has a
producer.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UDCbcCZQepnpSQYJ6SxxZm
The album page asked MusicBrainz how many tracks an album has, because
the only total it had was the length of the tracklist it was already
showing — a tautology for a library copy. The denominator was on disk
all along: metadata has read the "5/12" totals off every file since
forever and discarded them. They persist to
release_group_recordings.total_tracks now, and a complete, MBID-matched
album makes no catalog call at all.
Around that:
- AlbumReleasesFailed, so a slow browse is no longer reported as a
failed one. The page inferred failure from a 12s deadline, against a
browse queued behind up to eight prefetches on a 1 req/s limiter.
- Tracks not in the library are dimmed in place rather than the owned
ones carrying a green tick, which is also what let the "loading
catalog" banner go.
- A partly-owned album draws the release, not the part, so the missing
tracks are visible and Play can say "9 of 12" truthfully.
- The version dropdown appears only when tracklists actually differ,
and the version you own is marked by name instead of being replaced
by a synthetic "Your Library" entry.
- A merged cluster shows the running order the most releases agree on,
not whichever pressing the browse returned first — which is what made
a correctly matched album claim it was unlinked from MusicBrainz.
Also carries in-progress work from earlier sessions that shared these
files: the queue source link, autotag mixed-bag grouping, the mix
feature and its schema, and the config general page.
Committed with --no-verify: every pre-commit check was run by hand and
passed, but bindings-check refuses to run while frontend/wailsjs is
dirty and counts *staged* as dirty, so it cannot pass on any commit
that updates the bindings. Verified separately by regenerating and
diffing against the staged content.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NSmYeXS3k9xw3MnMPoCjvP
The durable "I asked for this" record was called Want, and the one-shot
search-and-grab attempt was called Request — names that didn't match
what either actually did. Want is now Request, and the old Request/Item
is now Download/DownloadItem, with a table-rename migration
(download_wants -> download_requests, old download_requests ->
download_downloads) safe against both fresh installs and existing data.
Every anchored manual download now upserts/reuses a durable Request
before running, so a "download now" that finds nothing is picked up by
the background reconciler automatically instead of just failing with
no trace — the gap that caused this session's repeated "no candidates
found" failures on the same album.
Also adds auto-download guardrails (file-size min/max with a preferred
target, allowed file types) that gate what the pipeline may grab
unattended, live-editable from a new settings section. The frontend's
wanted-view becomes downloads-view, with a new Downloads tab showing
attempt/transfer history that previously had no UI at all.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Y2Agd9af5hE7qzti2ackiS
Autotag: detect "junk drawer" folders with no artist/album consensus
and split them into synthetic per-cluster groups instead of forcing
one match on an unrelated pile of tracks; repair tagging_items rows
left behind by a prior scan orphan-cleanup gap.
Explore: fix an exact artist-name search being drowned out by its own
catalog entries in intent-prior scoring, and prune stale in_library
bookkeeping left behind when a referenced library row is deleted.
Download: fix a multi-library regression where every import failed
with "no library root configured" — the importer resolved the
library root from a legacy single-library config field that nothing
populates in the current multi-library model. It now resolves the
destination library per-request from the request's own library_id.
Also widen the Soulseek search window (12s -> 20s), measured against
real request history to be missing available peers on live queries.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Y2Agd9af5hE7qzti2ackiS
Ships the fresh-start schema cleanup: rebuilt explore catalog index
pipeline (dump import, artifact fetch/build, incremental listen-count
refresh), a new download subsystem (Lidarr/Prowlarr/qBittorrent/SABnzbd/
slskd/yt-dlp providers, staging, reconciliation, wanted list), and the
supporting schema/query/store changes across backend and frontend.
Also includes two smaller follow-ups: bump the central index's
rebuild-after cadence from 90 to 180 days, and remove the Explore
"library only" online/offline toggle entirely (frontend-only, no
backend counterpart) rather than carry unused UI/state.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Y2Agd9af5hE7qzti2ackiS
actions/checkout is a JS action and needs node inside the job container,
which the golang image does not carry — the step failed with
"exec: node: executable file not found in $PATH". Clone with git and the
package token instead, matching arch-package.yml.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Add a central job registry that library scans and search index builds
report into, so background work is visible instead of buried in the
settings page.
- backend/jobs: registry with per-job ring-buffer logs, capability-driven
controls, and one coalesced JobsChanged snapshot at 4Hz
- pause survives restart via a job_state table; a paused scan is adopted
back on launch and skipped by the soft scan
- top-bar indicator, popover, details drawer and a Jobs page replacing
the config page's scan UI; per-library start/stop retained
- scan timing breakdown moves into the job log, Full rescan to the Jobs
page; delete the orphaned library-manager component
Also add cmd/indexbuild and cmd/indexexport so the explore index can be
built once centrally rather than by every install, which today streams
~205GB from the ListenBrainz spark dump on first run. indexbuild picks
build/refresh/rebuild from index state; the Gitea workflow runs it on
push, weekly, or manually and publishes only when content changed.
fresh-install no longer defaults YJ_HOME under /tmp: it is tmpfs on most
distros, and the import needs ~6GB of real disk.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
wails names the bundle after outputfilename (yellowjacket.app), not the
hardcoded YellowJacket.app the formula assumed. Glob for build/bin/*.app
and its inner executable so casing/rename can't break `brew install`.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds the Homebrew tap formula and Gitea release-sync workflow, plus
smartplaylist materialize-on-creation, volume-control scroll/drag, and
smartplaylist cover-art batch loading.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Smart playlists now evaluate and snapshot their rules at creation time
instead of only lazily on first open, so the playlist list can show a
real track count in place of the "Smart" label. A one-time idempotent
startup sweep backfills snapshots for smart playlists created before
creation-time materialization existed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Build-from-source formula for macOS/Linuxbrew, mirroring the Arch
PKGBUILD. A Gitea workflow recomputes the tarball checksum on each
version tag and syncs the formula into the homebrew-yellowjacket tap
repo, so releases need no manual formula edits.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Wheel over the volume icon steps volume by 5. The slider now updates
live on drag (@input) instead of only on release, debounced 60ms to
avoid spamming SetVolume. A local pendingVolume tracks intent so rapid
events accumulate and UI stays responsive ahead of the backend echo.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The autotag overhaul added cover-art and MusicBrainz-ID columns to
leanTrackQuery to support the new track-row styling, reintroducing the
per-row correlated subquery anti-pattern (artist_mbid) plus a cover_art
join inside the whole-library derived table. Both ran for every track
before WHERE/LIMIT, so smart-playlist evaluation cost scaled with
library size rather than result size — several seconds for a 500-track
playlist that was previously sub-second.
Move these presentation-only fields into a batched fetchArtwork pass
keyed by the matched recording_ids, mirroring the existing fetchGenres
batch. Cost is now proportional to results. Add TestEvaluate_ArtworkEnrichment
(no prior coverage of these fields) and an artwork_ms debug metric.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add packaging/arch/README.md covering the one-time GPG key import
(public key C061B6267CF9D820), pacman.conf repo block, and install, plus
notes on the Never fallback, debug-package skip, versioning, and repo
priority.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
TestBufferedStreamer_BasicStream flaked under -race (got 1256/1512 vs
1000 expected). The streamer injects silence frames by design when its
ring buffer momentarily underruns; under the race detector the consumer
outran read-ahead and received mid-stream 256-sample silence frames. The
collection loop only skipped leading silence, so those frames were
counted as data.
Skip all zero frames, matching the test's own drain loop — real samples
always start at 1.0, so any zero is injected silence, never source data.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Remove the protect-main pre-push guard so main can be pushed directly,
and update the CLAUDE.md git-workflow note to match.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Enrich owned artists whose discography hasn't been fetched yet in a
bounded, resumable background pass so their wider catalogue is searchable
offline right after a scan, instead of only on first artist-page view.
Keyed off the persistent discog_fetched flag via LEFT JOIN, so already-
enriched artists never reappear and the run is a cheap no-op once every
owned artist is covered. Capped at discogBackfillMaxPerRun per run and
routed through discogSF to avoid double-fetching an artist a concurrent
interactive EnsureArtistDiscography is handling. Invoked on both scan
completion (OnStartup) and OnDomReady to resume a capped/interrupted run.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
yonlu
