| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| ... | |
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The per-feature design notes were merged into docs/MESHBAY_DESIGN.md and
deleted from the tree on 2026-09-11, but ~230 comments across the three
packages still named them — usually written `docs/musicbay.md §3.2`, as
though the file were still in docs/. A reader had to know §16 existed to
resolve any of them. They now name the section directly.
Every mapping comes from §16, the concordance, which already records where
each old section landed: musicbay -> §9.8, mediacenter -> §9.7 for the
Videos app and §6.5 where the subject is derived data, photos -> §9.9,
auth-confirm -> §3.6, refactoring-search -> §9.11, invite-pairing-v1 ->
§3.4, per-node-identity-v1 -> §3.2, captcha -> §7.7, chat-sender-keys ->
§4.5, apps/refactor-groups -> §9.1–§9.4, desktop-client-v1 per section.
Bare citations of the same documents (`draft-v6 §2.11`, `§4.8`, `§3.4`)
are retargeted too: those collide with real section numbers in the design
document, so leaving them would have been worse than the named form.
Four cases the concordance does not cover, each decided rather than guessed:
Sub-item references into documents that no longer exist — mediacenter's
`§3.3 row 4`, `§3.4b/c`, `apps.md §3b` — name rows and sub-items §9.7 and
§9.2 do not reproduce. The module-level citation stays; the sub-item
pointer is dropped.
The V-findings keep their labels but lose the dead `§10.1/` prefix.
§13.8 lists V1–V13 as per-application open items, which is not what the
labels mean in these comments, so pointing them at §13.8 would have been
a false citation.
`apps.md §5`'s virtualization requirement has no counterpart anywhere in
the design document. The requirement is stated in the comment itself, so
the citation is dropped rather than aimed at a section that does not say
it.
Comments that attributed a *sentence* to an old note — musicbay's "several
thousand files" example, its "what got measured" note, its measured
~11%/~26% cover-art figures, the "original no root, whole shared tree"
call — state the fact without attribution now. §9.8 does not contain those
sentences and citing it for them would have been wrong.
CLAUDE.md's "a reference to a document that no longer exists" row now says
the concordance is for git history and out-of-tree material; the code cites
sections directly.
Verified: 2851 passed, 4 skipped. The 12 errors in the run are the Firefox
leg of test_sticky_header.py's browser harness, which is broken at the
browser level on this machine — headless Firefox (snap) dies with
`[GFX1-]: RenderCompositorSWGL failed mapping default framebuffer`, renders
nothing, and the probe exits `{"error": "no measurement"}` after its full
90s wait. Chrome runs the same 12 assertions in 3.2s and passes. Nothing
here can affect it: every changed line in style.css is inside a comment.
Also checked: ast.parse on every changed .py, `node --check` on every
changed .js, the /* */ balance in style.css, and that no changed line
exceeds the width its file already used.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Seventeen comments across the Windows packaging targets pointed at
C:\Users\admin\devel\light-client.md and msix-installer.md -- absolute
paths on one developer's machine, unreadable to anyone else who clones
this repository and unverifiable by any test here.
The surrounding prose already carried the substance in every case, so
these are removals, not rewrites, with two exceptions where the pointer
was doing real work:
electron-builder.msix.yml's header told the reader to go read §4 and §8
first. It now states the fact directly (an AppX/MSIX install never
elevates, by design, so the package carries none of installer.nsh's
elevation logic) and says the open items are called out at each
declaration below -- which they already were, at `capabilities` and
`customExtensionsPath`.
The two "msix-installer.md §8" citations become "an open item" / "the
other open item", beside the description of the item that was already
there.
Paragraphs the removals left ragged are re-wrapped.
Verified: node suite 1403 pass / 4 skip (71 of them test_packaging_win.py),
both electron-builder configs still parse as YAML, node --check on main.js.
The .ps1 edits are inside <# #> comment headers plus one deleted Write-Host
in a block that keeps two others; no pwsh on this machine to parse them.
Not touched, and much larger: ~230 comments elsewhere in the codebase cite
per-feature design notes (musicbay.md, mediacenter.md, auth-confirm.md and
twenty more) that were merged into docs/MESHBAY_DESIGN.md.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A node is a machine in somebody's house, so with a handful of groups one is
always off. Search treated that as the exception and charged the reader for it.
Measured on twelve groups against a virtual clock, a live group answering in
200 ms and a dead one taking the full deadline: one node down in the first batch
put a blank page and a progress bar in front of the reader for **ten seconds**,
while two groups of that same batch had answered in two hundred milliseconds and
nine others had not been dialled at all. Four down, spread out — the shape a real
reader has — was ten seconds to the first result and **forty** to the last.
Two causes, neither of them the connection deadline. Results were drawn once a
batch was complete rather than as each arrived, so an index already in hand waited
on a node that was not answering. And the batches were sequential, so a dead group
did not merely cost its own deadline, it postponed every group behind it.
So the three is a ceiling on concurrency and never a batch — `inFlight` starts the
next group the moment one ends — and each index is drawn when it lands.
That alone is not enough, and the tests say where it stops: a ceiling still lets
silent nodes hold every place at once, and with four of twelve down the last three
live groups still waited out a deadline. So the browser remembers which groups
were silent and dials them last, which puts all eight on screen in 600 ms. The
list is advisory and rewritten from what each sweep saw: a private window, storage
that refuses, a first visit or a node that has come back all fall through to the
hub's own order, cost one sweep, and correct themselves.
The "n groups unreachable" line also waited for the sweep to finish, which is the
one moment it is no longer needed. It now appears as they are found.
First result, before and after, twelve groups: one down 10 s → 200 ms; four down
10 s → 200 ms with every reachable group on screen by 600 ms; three down and
listed first 10 s → 200 ms from the second visit on.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Opening a group from the sidebar has no deadline of its own and gets the
transport's: 30 s for the DataChannel, 30 s for each request after it. Search
wrapped the same `connect()` in a flat 10 s, and that 10 s had to cover the hub
round trip, ICE gathering (capped at 4 s in transport.js), DTLS, the channel
opening and the handshake's own round trips. On a phone on 4G the budget was met
by luck rather than margin, and the same group then failed in Search while it
opened from the sidebar, against the same node. The budgets were inverted: the
phase full of round trips had a third of what one request on an open channel got.
Raising the number would have been the wrong repair. `fetchAllIndexes` fans out
three at a time and waits for the slowest of each batch, so a page of unreachable
groups costs batches x the deadline in spinner: a bigger number taxes every dead
group for the sake of the live ones.
So the deadline measures stalling. A node that is not there reports nothing and
still fails in `SEARCH_STALL_MS`, unchanged at 10 s, which is what keeps the
fan-out where it was. A node that answers ICE, then opens a channel, buys another
window at each step, up to `SEARCH_MAX_MS` — a deadline that only ever resets has
none, and a node that answers and then goes quiet would otherwise never be given
up on.
The transport reports those steps through `onConnectProgress`, set by the one
caller that imposes a deadline of its own. `connected`/`completed` is the signal
and not `checking`, because the first means a candidate pair answered and the
second means this side is still trying addresses that may all be dead. A caller's
callback cannot break the connection it is reporting on.
The tests run the shipped `connectToGroup`, lifted out as text, against a fake
clock — a real one would make each scenario a minute and blur the only thing
worth asserting, which is when the deadline fires. Dead node: 10 s. Slow but
moving: connects at 20 s where it used to fail at 10. Answers then stops: 18 s.
Progress that never finishes: the 30 s ceiling. Two dead nodes: two windows, both
transports closed. Checked against the flat deadline, which fails three of them.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Both described panels the desktop client would grow of its own. §8.3 settled
that differently: one UI source means anything written for the client is
something the browser then lacks, so both were built in `static/` instead and
the client ships them.
D5 is `node-page.js` over the operator ops — status, groups, roots with eject,
plug and add, roster, peers, pairing, audit export. D6 is `create-group-page.js`,
which detects the node, provisions it, takes the group's name and roots,
initialises the key and pairs. The service unit is the one genuinely native
part, and the desktop main process already runs it: `systemctl --user` on Linux,
Task Scheduler on Windows.
Left standing in §15.2 they read as work outstanding, which is how a plan
outlives the decision that replaced it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The last two handlers on the loop. Both were synchronous, so a member creating a
folder on a root that had spun down held the node for the spin-up, exactly as a
chunk read did.
Where a check and an act belong together they are now one call rather than two
awaits, and the single disk thread is what makes that atomic: `_mkdir_if_absent`
so two members creating the same name cannot both find nothing there and have
the second `mkdir` raise where a refusal was meant, and `_rmdir_if_empty` for
the reason the caller already re-tested emptiness — the first test happened
before a round trip to the operator's browser, and a file can land in between.
Two awaits would reopen that window one size smaller.
The guard is now the whole class rather than the calls that were fixed. It walks
the module's syntax tree and fails on any filesystem call outside the handful of
functions written to be run through `off_disk` — a new handler that stats a root
inline would pass every measured test, because those exercise the handlers that
exist today. Checked by putting a call back: it names the function and the line.
It leaves ffmpeg's own scratch files out, listed rather than silently allowed:
they are under `tempfile.mkstemp` on the system disk, not on a group root, so
they are not what spins down — but they do read a whole transcode into memory
from the loop, and the day that matters it is a different measurement from this
one.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The rest of AV9's disk half. Serving a file left the loop in the commit before
this one; two paths were still on it.
**The availability poll.** `RootSet.refresh_availability` stats every root, and
eleven call sites reached it from `async def` — the reconcile loop among them, on
a timer. On a sleeping disk that is a stall once per tick, and the stat is also
what keeps the disk awake, so a node paid spin-up for a library nobody was
reading. All eleven now go through `off_disk`, `Root.is_live` included.
**The upload write.** `open`/`write`, and the resolve, the stat, the free-name
search, the rename and the unlink around it. This one could not simply be
awaited: the handler was synchronous, so nothing could come between the
`chunk_index != state.next_index` check and the `advance` that answers it, and
that is the whole of the chunk-ordering rule. Awaiting the write opens the gap —
chunk 1 arriving while chunk 0 is in the disk thread reads a position that has
not moved and is refused as out of order, so an upload would fail on a slow disk
and nowhere else. Verified, not assumed: without the lock the new ordering test
refuses three chunks of four.
So the check, the write and the advance are one critical section again, under a
lock held **per group**. Not per session: `partial_uploads` lives in the group
context so a reconnecting client finds its upload where it left it, which means
two sessions of one member share the position of one `.part` file. Arrival order
is preserved by construction — the dispatcher creates one task per message as it
arrives, tasks start in creation order, and the lock is the first thing each one
waits on, so its waiters queue in arrival order too.
`_do_file_upload` is a coroutine now, which is why forty-two test call sites gain
an `await`. Their outcomes are unchanged, file by file, against the run before
the change.
`test_ops.py` asked which public coroutines `ops` exposes and got `off_disk`,
imported rather than defined there. It now asks for the ones written in the
module, which is what its own docstring means; all forty-three operations are
still checked.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A root that has spun down, or that lives on a network mount, answers its first
syscall in seconds rather than microseconds. Made from the event loop, that
stalls the whole node: no other group is served, no stream is fed, no chat
message is delivered and the hub socket is not read, for as long as the platter
takes to come back. It was found from the other end — a client's connection
attempt timing out on a node with one member, while the disk woke up — and it is
AV9's lesson with the disk in the place of the mail server.
Every filesystem call on a group's content now goes through `off_disk`, onto a
thread that belongs to that group's root set. The stat goes with the read: a stat
is what *wakes* a sleeping disk, so offloading only the read would move the stall
rather than remove it, and the read would then find the disk already awake.
`_locate` is the one place allowed to call `entry_abs_path`, which is
`Path.resolve()` and therefore syscalls too.
Five handlers touched: a chunk request, an audio transcode, a subtitle request, a
video stream and a delete. The delete became a coroutine, which its one caller
already was.
One worker per root set, not a pool and not one for the node. One worker for the
same reason the indexer's executor has one — two interleaved reads of a spinning
drive seek-thrash rather than go faster — and it keeps two threads from being
inside the same file at once, which is what makes the `seek`/`read` pair safe
without a lock. Per root set, because a node serves several groups and their
roots are not all on the same volume: a single worker would put one group's
sleeping USB drive in front of another's SSD, which is this symptom one level
down. The thread is created on the first read, so a group nobody downloads from
never starts one, and the daemon stops them all on the way out.
The tests measure rather than read: a ticker counts its own wake-ups beside a
request made slow on purpose, and a handler that blocks the loop takes every one
of them with it. Put either call back inline and both tests report zero wake-ups,
which was checked before they were trusted.
QUIC still reads on its loop. Its handler is synchronous by construction, no
client speaks it, and its distance from parity is already recorded in the design
document (§15.3, L3); moving it is part of bringing it to parity, not of this.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The two documents disagreed with the code they describe. §15.2 still listed
playlists and Chromecast subtitles as not built, both shipped and tested; the
protocol document's header said wire version 3.2 beside its own §13 saying 3.3,
which is §12's "a number restated away from its definition is a number that will
drift" happening to the document that states the rule.
The packaged install is no longer an open question: the artefacts have been taken
to a running hub and node on Ubuntu 26.04 (.deb), Fedora 44 (.rpm) and Windows 11
(.exe). What is left of D11 is signing the binary, which is a different question
from whether it installs.
The migration row overstated its own risk. Nothing replays the revisions behind
the first packaged release — they were development no installation ever ran — so
the exposure is the next migration, not the chain.
The disk-I/O row gains the consequence that was measured rather than reasoned: a
root waking from standby stalls the loop long enough for a client's WebRTC
connection to time out, on one node with one member. `refresh_availability` is
named beside the two chunk paths, because a stat on a sleeping disk is what wakes
it while reading nothing.
A blank line had also split §15.3's table in two.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A 1080p film is decoded by whoever watches it and re-encoded by the node
when no browser can decode the source. Both were on the CPU, and on an
Atom or Celeron mini-PC neither reaches real time — which is what
`transcode_incompatible_video` exists to refuse. This adds the mechanism
that makes refusing it unnecessary.
Node — `hwaccel.py`: VA-API on Linux, Quick Sync or NVENC on Windows,
established by encoding 1080p and reading the file back with ffprobe.
Nothing is accepted that does not produce the exact profile and level
`stream_init` announces, since the client checks that string before it
trusts a byte: an encoder that wrote another level would make the node's
own codec string a lie, and ffmpeg takes `-level 4.1` and `-level 41`
from h264_qsv without saying which it understood. Three modes per
stream — hardware decode and encode, hardware encode alone, libx264 —
demoted per source codec, because a GPU that decodes HEVC may have no
decoder for MPEG-4 Part 2 and only asking it finds out. A mode that
fails is detected on an empty stdout before `stream_init` goes out, so
the viewer sees one working stream and never an error.
Client — Chromium ships VA-API off on Linux. It is enabled where a
render node and a driver are present, then verified through
`navigator.mediaCapabilities`: a no moves to the next GL backend on the
next launch and an exhausted list drops the switches, so a renamed
feature cannot pass for a feature that is on and `--ignore-gpu-blocklist`
cannot survive on a machine it did not help. Feature lists now merge
rather than overwrite — `appendSwitch` replaces the value, and a second
caller would have silently cancelled the mDNS switch aiortc depends on.
Packaging — the drivers are weak dependencies on all four formats, so a
machine without a GPU installs exactly as before. `dpkg -i` and
`rpm -ivh` ignore weak deps; `packaging/README.md` now says so. Windows
needs no driver: the bundled ffmpeg already carries h264_qsv and
h264_nvenc, and a re-pin that dropped them would cost every low-power
Windows node its hardware encoding silently.
AMD on Windows (AMF) and macOS (VideoToolbox) are named gaps, not
oversights: neither could be tried anywhere in this project, and both
re-encode in software as before.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Most of what the relay says repeats without bound — a line every fiftieth
fragment for the length of a film, one per dropped fragment whenever a client
falls behind — and it lands in the terminal the app was started from. Those go
through `debuglog` now, restored with:
NODE_DEBUG=cast-relay npm start
The loss of fMP4 framing stays on console.warn. It recovers by rescanning, so
nothing throws and no other part of the system hears about it; that line is the
only trace that the picture on the television is missing a piece. Every other
failure here throws, and the renderer already reports the rejection.
The logging guard now covers both cast modules, each with the one line it is
meant to keep audible.
Also corrects the header comment, which still claimed CORS was granted on the
subtitle path only. It has been on both since the stream needed it too.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The receiver reports its state on a timer, so `player status: PLAYING` repeats
for as long as a film runs and buries everything else in the terminal the app
was started from. The progress lines now go through Node's own `debuglog`,
which costs nothing when disabled — the message is never formatted — and is
restored with:
NODE_DEBUG=cast-chromecast npm start
The socket error stays on console.error. It fires when the connection to the
receiver dies and is the only trace of why a cast stopped; routing it through
the same switch would make the one failure worth seeing the one that
disappears.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The relay forwards the node's fragments untouched, and those begin at zero
at the seek point. The player never notices because its SourceBuffer is given
`timestampOffset = start`; a receiver has no equivalent, so the cues are
shifted by `-start` before they leave, recomputed at every restart of the
relay. Sent as they are, a subtitle would be out by the whole seek.
The document is served from the relay's own port at /subs.vtt, behind the same
token as the stream and with CORS: a receiver fetches a side-loaded track with
XHR from its own origin, and without the headers it fails as a network error
with nothing on screen to say so. The URL carries a version because a track is
cached by address — changing the cues behind a fixed URL leaves the previous
language showing.
Cues that end before the stream begins are dropped rather than clamped, so a
line from before the seek cannot appear over the first frames after it.
The relay is plain Node, so the tests start it and fetch from it rather than
reading its source.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Verifying the feature across a library turned up two faults, one of which
broke it outright.
**The extraction budget was a constant and the cost is not.** Extracting
demuxes the whole container, so the file sets the price: measured at 9.8 s per
GB on a library held on an external disk — 36 s for a 3.9 GB title, 71 s for a
7.3 GB one. Against a flat 60 s that worked on most of a library and failed on
the big films, which to whoever is watching one is indistinguishable from a
broken feature. `Avatar.The.Way.of.Water` timed out every time. The budget is
now three times the measured rate per gigabyte, floored at the old 60 s and
capped so no container can pin a transcode slot indefinitely. The client's
own timeout follows it: the node always answers, so that one is a backstop
against a silent peer rather than a deadline for the work.
**Two clicks ran two extractions.** The cache is consulted on the way in, so a
second request arriving while the first was still running missed it and ran
its own — seen in the log as two identical extractions of one 4.3 GB file
overlapping, each holding a transcode slot and reading the file end to end.
Latecomers now wait on the answer the first is producing.
The in-flight entry is registered *before* the first await, not after. The
first version registered it after the probe, two concurrent requests both got
past the check while neither had registered, and the test for it failed —
which is the only reason this note can be written from the right side.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The previous fix read ffprobe's key frames and took the last one at or before
the request. It was wrong twice, and a viewer felt the difference: subtitles
went from 5 s early to 2–3 s late.
Matroska's Cues index only some keyframes, so an index seek backs off to an
indexed one that the frame list does not single out. And the landing point
moves with **which streams are mapped**, because the container is positioned
where every mapped stream has data — on the reported title, a seek to 4913.7 s
landed at 4909.863 with video alone and at 4907.236 with the second audio
track mapped beside it. The frame scan gave the first number; the stream
delivered the second; the gap was 2.65 s, and the measured audio displacement
in the served stream was 2.65 s.
So the node asks ffmpeg instead: the same seek, the same mapping, one copied
frame under `-copyts`, and the answer read back off the result. 0.06–0.07 s,
cheaper than the scan it replaces. The `-ss` argument stays at the request, so
the bytes served are exactly the ones served before — only the number naming
them changes. The probe runs after the audio track is resolved, because it
cannot be right before that is known.
An answer after the request, or further before it than any real keyframe gap,
is discarded in favour of the old label: a number wrong by seconds beats a
fabricated one.
Found by decoding the served stream and locating its first frame in the
source, which put it at 4907.213 s against an announced 4909.863 s. The test
does the same thing rather than comparing the announced number against a
second reading of the same probe.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A cached track replied without leaving a line, so the journal showed nothing
for a viewer who had asked for subtitles and been served instantly. That made
"no request in the log" look like evidence nobody had asked — which it was not
entitled to be, and which cost a wrong conclusion about whether a fix had been
exercised at all.
The request is logged on arrival now, before the cache is consulted, and a
cache hit says so with the size it served. Second time in this feature that a
silent success was read as an absence; the first was an attach-time state that
was correct while the screen stayed empty.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Copied video can only start on a keyframe, so `-ss t` with `-c:v copy`
delivers the keyframe at or before `t`. The node reported `t` anyway and the
client builds `SourceBuffer.timestampOffset` out of that number, so everything
downstream believed the picture stood a few seconds further along than it did.
That was a wrong label while only the scrubber read it — it is recorded as
such, low priority, in the design document. Subtitles made it a wrong answer:
their cues carry the source's own absolute timestamps, so the mismatch put
every line on screen before it was spoken. Reported from real use on an H264
title, where seeking to 600 s, 2650 s and 5000 s lands on keyframes 0.82 s,
1.56 s and 4.64 s earlier.
The copy path now resolves the request to that keyframe, seeks to it, and
reports it. The bytes delivered do not change — ffmpeg lands on the same frame
either way — only the number that names them. The look-up reads the thirty
seconds before the request and cost 0.12–0.51 s on that title, which is the
price already costed in §15.3 and never paid. Re-encoded video is untouched:
it can begin exactly where it is asked to, and does.
Two details worth their lines. The keyframe is passed back to ffmpeg at six
decimals, because rounding a keyframe's own timestamp down puts it before the
frame it names and selects the previous one — the same fault again, smaller.
And the ffprobe window ends past the request, since an interval whose end is
the request never emits a keyframe sitting exactly on it: the resume position
is the one place a viewer asks for the same instant twice, and it would have
been answered a whole GOP early.
The test decodes the first frame served and matches it against the source
frame at the position announced, rather than comparing `start` to an expected
number — both sides of that comparison would be reading the same ffprobe and
would agree by construction.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Reported as "I click a subtitle and nothing appears", on three films. Nothing
was broken. The track selected was the container's forced track, which carries
signage and foreign dialogue only: measured on the film in question, 30 cues
and 77 seconds of text across 2h32 — 0.8% of the running time, against 1559
cues and 41.8% for the full track sitting beside it under the same language
tag. At all three positions tested there was genuinely no cue to show; the
full track would have shown one at two of them.
So the defect is that the menu could not say which was which. The label used
the container's title tag, which said "Forced" on that film and says nothing
at all on most, and no other field was carried. The disposition is the half
that is always there: `probe_video` now reads `forced` and `hearing_impaired`,
`stream_init` carries them, and the label states them in the reader's own
language rather than repeating an English word a muxer happened to type.
The node fixture grows a forced track with no title, because a title would
let the old code pass. The label harness's `t` stub took a parameters object
unconditionally and threw on a key that has none — a fixture narrower than
production, fixed here rather than worked around.
Also removes the activeCues probe that found this. It answered its question:
mode showing, cues 30, active 0, none due at that instant.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Attach-time state — one track, showing, cues parsed — was all correct while
nothing appeared on screen, so it was measuring the wrong thing. A re-render
can replace the <track> and reset a mode nothing sets again, and a cue list
that does not cover the playhead looks identical to one that does.
The probe samples activeCues for ten seconds alongside the playhead, the mode,
and the cue that ought to be on screen at that instant, so "no cue is active"
and "a cue is active and is not painted" stop being the same observation.
Verified beforehand that the mechanism itself is sound: a real Chrome driven
over CDP, fed through MediaSource with timestampOffset 2690 and given a track
appended after playback started, reports activeCues 1 on the cue bracketing
the playhead.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A subtitle button that spins for ever and never shows a track could not be
told apart from a node that answered, a blob that never arrived, or a track
attached with no cues in it: the whole path was silent. Every step of it
happens on someone else's machine, over a link, against a file that may be
gigabytes, so the only question worth asking when it does not finish is which
step did not — and nothing recorded that.
It now logs the request, the node's answer with hash, size and elapsed time,
the fetched blob and its chunk count, the attachment, and the live TextTrack's
mode and cue count. A failure says how long it took before failing.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Two faults, both found while a hang could not be diagnosed from the node's
own journal.
`except Exception` does not catch `CancelledError`, which derives from
BaseException. A cancelled extraction therefore sent no reply and logged
nothing at all, leaving the client waiting on something nothing would ever
answer — the one shape that cannot be reported by whoever hits it. It now
answers, logs, and re-raises so the cancellation still propagates.
And the handler logged nothing on any path, so the journal could not even say
whether a request had arrived. It now names the file, the track, the file size
and the free slot count on entry, the moment the transcode slot is taken, and
the duration and byte count on the way out — the size and duration because the
extraction is a whole-file demux and its cost is set by the file, not by the
subtitle.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The documented setup was `pip install pytest pytest-asyncio aiosqlite`, three
lines under a rule saying not to install packages that are not declared in a
pyproject.toml. All three packages declare a `dev` extra, and the hand-written
list missed what was in it.
It cost three invisible things. Without `piexif`, `test_enrich_photo.py` fails
at import and the whole node suite stops at collection — not one test, the
run. Without `pytest-timeout`, `timeout = 60` in pyproject.toml is an unknown
config option, so the protection its own comment describes ("a hung test must
not wedge the run") was absent, announced only in a warning among the
deprecation notices. Without `ruff`, the project's linter is not installed.
`aiosqlite` never needed naming: it is a runtime dependency of the hub and the
node and arrives with the `-e` installs.
Verified with the corrected line: 2804 tests collected, 2800 passed, 4
skipped, and no unknown-option warning — nothing exceeds the timeout that is
now actually enforced.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
The comment claimed a <track> appended after playback started is not shown by
its `default` attribute. Chrome shows it: in a headless run the TextTrack read
back "showing" before the effect had touched it, with its cues already parsed.
The effect stays, for the two reasons that are true — `default` says nothing
about turning subtitles off again, and an explicitly assigned mode means the
same thing in every engine — but the reason written beside it was not one of
them.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
MSE decodes no in-band text track, so a subtitle cannot ride inside the
fragmented MP4 the player is fed. The node extracts one track whole, converts
it to WebVTT and caches it under its own hash; the client pulls that blob
through the ordinary file_req/chunk path and hangs a <track> on the video
element — the same indirection as a TMDB poster or an audio transcode, which
is what makes a film's subtitles extracted once in the life of the file rather
than once per viewing. Whole-file also makes the cues absolute, so a seek and
an audio-language change both leave the track untouched.
**The ordinal counts every subtitle stream, including the ones never listed.**
Only text codecs are offered: a bitmap track (PGS, VOBSUB — about a fifth of a
real library) has no path to WebVTT without OCR, and one extracted anyway
yields a header with no cues, which is a menu entry that shows nothing and
reports no error. Numbering the survivors of that filter would give a
PGS/SRT/SRT file the ordinals 0 and 1 for its text tracks and `-map 0:s:0`
would then extract the PGS — the same trap `AudioTrack.ordinal` exists for,
one level deeper. A fixture whose first subtitle stream cannot be decoded pins
it, and the handler checks membership of the probed list, never a range.
Additive and MINOR: the selector is drawn from `subtitle_tracks` in the node's
own `stream_init` and from no version number, so `subtitle_req` is never sent
to a peer that would not answer it. The floor stays at 3.0.
Also here: a failed extraction never touches playback, a superseded reply
cannot install its blob over a newer choice, and `_languageName` is shared
with the audio labels — lifted by both label harnesses, since a lift that
names one function stops covering the rule the moment logic moves out of it.
Tests: 9 node (tracks told apart by the words in the extracted cues, not by
tags), 10 client. Full suite green: 1545 node/common, 1252 hub.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The audio-track selector was drawn with `volume` — the same glyph the music
player uses for its level control, and the same one the browser draws a few
dozen pixels below it, since the video player mounts `<video controls>` and
gets the native volume speaker with it. One drawing, two meanings, in one
window.
`speech-pair` is two offset bubbles: the same line, said twice, which is what
the button chooses. Nothing crosses anything — the set is stroked with no
fill, and two overlapping outlines read as a tangle at 20 px — and it leaves
the conventional subtitle glyph free for the button planned beside it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UGY17EPph5LsLzePPXhUVc
|
| |
|
|
| |
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
names its parameters
`pw_version` 4: Argon2id 64 MiB, t=3, lanes=4 — RFC 9106's second recommended
setting. A v3 hash (256 MB) still verifies at its own parameters and is
rewritten at the new ones on the next sign-in, through the rehash path that
already existed.
Why not more. The verifier matters against an offline attacker holding the
database; online guessing is bounded by the sign-in lockout. That attacker pays
the client's 600 000 PBKDF2-SHA512 iterations and the hub's Argon2id per guess,
since `auth_key` is 256 bits and cannot be searched directly. Memory above
64 MiB multiplies that cost by a constant — at most 16 at 256 MB, less with
PBKDF2 counted — while the hub pays the same memory at every sign-in, one
derivation at a time. Measured on meshbay.org: 450 ms at 256 MB, 105 ms at
64 MiB, so a burst of sign-ins clears about four times faster.
Changing the current version exposed a latent lockout. `hash_password` always
used the current version's parameters, while the raw-password scheme recorded
`pw_version = 2` — harmless while versions 2 and 3 shared their parameters,
and with version 4 every legacy registration and v1→v2 rehash would have
stored a 64 MiB hash labelled 256 MB, which nothing could then verify. Seventeen
tests caught it. `hash_password` now takes the version it is hashing for.
The OpenSSL deadlock between two concurrent `lanes=4` derivations is the same at
64 MiB, so Argon2 stays on its single worker. The loop-stall test measures
against a v3 hash, because half of a 45 ms inline derivation is too close to
scheduling noise to be a reliable bound.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LcF3QKWii7uQ2kSyXErzCt
|
| |
|
|
|
|
|
|
| |
A seek now begins picture and sound together at the keyframe, but stream_init
still reports the position that was asked for. §15.3 says what the fix costs
and why it has not been made.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
-ss before -i cannot trim copied video, which must begin on a keyframe, but
accurate_seek did trim the re-encoded audio to the exact request. Every seek
on a copied stream therefore opened with a GOP-wide hole in the audio and ran
a GOP out of sync afterwards — 9.979s on a real film with a 10s keyframe
interval. Accurate seeking is now off wherever video is copied, and stays on
where it is re-encoded, which is the only path that could already begin where
it was asked to.
Every timestamp was correct throughout, which is why nothing caught it; the
tests assert on decoded audio and on frames compared against the source.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
`{ ..._auth }` after the await spreads a null _auth to {}, so the renewal
stored a token with no username and no userId. The app renders the
signed-in interface from that and throws on user.username[0] — a blank
page on every load, in localStorage, until the site's data is cleared.
The sign-out wins the race now, and loadAuth treats an identity-less
object as signed out so a browser already holding one heals itself.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The streaming path mapped 0:a:0 unconditionally, so a dubbed film played in
whichever language was muxed first and the others were unreachable. The node
now enumerates the tracks in stream_init and honours audio_track in
stream_req; switching is the seek path, since one ffmpeg carries one track.
MNP 3.2, additive: the player draws its selector from the node's own list and
never from a version number, so an older node is never asked for a track it
would ignore. MNP_MIN_SUPPORTED does not move.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
boot-guard.js is a classic script loaded before the module graph, so it
survives the graph failing to link. If #app is still empty after ten
seconds it names what failed and offers a reset of this origin — cache,
storage, databases and the service worker, which clearing the cache does
not touch.
Two real defects found building it: openDB never settled when an upgrade
was blocked by another tab, and a connection it gave up on stayed open
and squatted the database.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
| |
The ceiling is 1200 tracks on a library that repeats and 660 on one that
never does; the message said 1000, which would send a reader back into the
wall at 660. It says 500, measured at both ends.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
| |
A DataChannel send() throws above the max-message-size the node advertises
(aiortc: 65536), so the 1 MB body cap was unreachable and each body push
swallowed the difference in a bare catch. Cap at 62 KB after sealing, name
the playlist in the sync result, keep syncing the rest.
Measured: 1000 tracks seal to 53 276 bytes, so the ceiling is ~1200.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
| |
The six MNP 3.1 ops were the only new ones never crossing a channel in a
test. Two cases on the existing aiortc harness: a round trip read back on
a second connection, and a 256 KB body. Found that the 1 MB body cap is
unreachable from a browser — docs/playlists.md §15.3.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
| |
Sync moved to the shell; the Search-page gap it lists is the bug that
was fixed. And two devices converging has now been seen working.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
| |
`max-height: calc(100vh - 16px)` says how tall the menu may be and nothing
about where its bottom lands, so one opened partway down the window ran
past it and its last rows scrolled out of reach. Measure from where the
panel was placed, and re-measure when a submenu opens.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
| |
The dismiss-on-scroll listener is on the capture phase, because `scroll`
does not bubble — so it also heard the menu scrolling itself, and a long
tracklist closed the moment it was wheeled. Filter on the event's origin.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Reported from a phone: signing in with the same account showed no
playlists. syncWith was called from exactly one place in the interface,
so creating a playlist, deleting one, removing a track and saving the
queue all wrote to IndexedDB and stopped there. The store pushes itself
now, coalesced, so a new mutation cannot forget to.
Silence was the real defect. The node audited only successes, so a
refusal left no trace and user_blob_list none at all; the background
push swallowed its reason; the interface said nothing. All three report
now, and "Sync now" says what happened either way.
An unreadable blob on a node was treated as a fetch failure and returned
before the push — permanent, once the node held anything. It is an
absence: the client is the authority, and it gets overwritten.
A sign-in reconciles whatever this browser already holds, a pending push
is flushed when the page goes away, and a push that did not land is
retried once.
no_key is spelled out: a client that signs in with its remembered device
key only ever has a bundle key persisted before the playlist subkey
existed, and an AES handle is non-extractable.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
An artist with one album got a heading and one cover on a row that fits
five, and a library is mostly single-album artists. Consecutive singles
share one grid, in place, so the page stays in artist order.
Each pooled cover keeps its artist's name above it in the same type as
a section heading. Dropping it was the first version and it was wrong:
scrolling then alternates between artists written large and small.
Measured on the probe's fixture: 4208px to 1895px, and a walk of the
page reaches all 21 covers instead of 9.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fourteen places the proposal was wrong, listed rather than edited away:
the MNP bump was 2.1 and is 3.1; the kind pattern refused the one
playlist every account has; confirm() is the house style and prompt()
is what throws; submenus expand in place; deflate is 4.5x and the
first fixture's 23x was a property of the fixture.
Adds the hazard the proposal missed entirely: two wrappers that take
two arguments and forward two, so every "add to queue" arrived as a
plain play.
The acceptance section now says what is checked and what is not —
no real daemon round trip, no two-browser convergence, no Firefox.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
One button in Music's sticky toolbar — load, create, delete, remove a
track, sync now — and "add to playlist" on every cover and row. Both
surfaces share one list, read from the manifest, so they open instantly
with every node offline and no body is fetched until one is wanted.
Submenus expand in place rather than flying out: the account menu's
language list already does this, and a flyout has nowhere to go at 400px.
The tracklist under "remove a track" loads when it is expanded.
A name is typed into a field. Electron has no prompt — it throws.
Also splits the two playback failures: a decode failure belongs to that
file and keeps the bounded counter, a connection failure belongs to the
group and skips all of its queued tracks at once. Six dead tracks are
one more than the bound, which is where a playlist would otherwise stop.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
playlists.js is IndexedDB, WebCrypto and a transport, and node has no
IndexedDB — so it is driven in Chrome against a node stubbed to record
what it was handed, which is also how what leaves the browser is
checked to be sealed.
Sync asks the node what it holds (user_blob_list) rather than comparing
against the merged watermark, which says nothing about that node: the
first version pushed every body on every sync. A tombstoned playlist's
body is deleted as each node is reached, or the quota fills with graves.
The database version and its stores stay in hub-client.js — two modules
opening one database at versions of their own is a VersionError thrown
at whichever runs second.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
playlist-merge.js and playlist-crypto.js have no imports and are run by
their tests, which is the only real evidence this feature can have.
Two revision counters per playlist, not one: a rename on one device and
a track added on another both write n+1, and a single counter makes two
edits that do not overlap collide.
One Argon2 run at sign-in, two handles. The AES handle is imported
non-extractably, so nothing can be derived from it — hence a second
import of the same bytes as HKDF rather than a derivation.
Measured: ~270 bytes a track, deflate worth 4.5x on realistic data, so
the 1 MB body cap holds about 17000 tracks.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
One row per playlist plus a manifest, so starring a track rewrites that
playlist rather than the whole collection. blob_enc is a BLOB, not
base64 TEXT: these run to hundreds of kilobytes.
user_id comes from the session and never from the message; kind is
validated against a pattern; every cap refuses with a stated reason
rather than truncating.
Additive, so MNP_MIN_SUPPORTED does not move — a 3.0 node answers
"unknown message type" and the client writes to the next one it reaches.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The player's queue could only be replaced: every onPlayQueue reset
tracks/order/pos together. It becomes one reducer (queue-ops.js) with
an `op`, because two appends batched into one tick cannot both read the
track count out of separate useStates.
A shared pop-up menu (menu.js) carries the three verbs, on right-click
and on a dots button. A track row is now a div holding two buttons: a
button cannot contain a button.
Found by the browser probe: both music wrappers took two arguments and
forwarded two, so every "add to queue" arrived as a plain play.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The 256 KB cap contradicted its own worked example; entries were also
missing `size`/`name`, without which a stored track cannot be fetched
at all. Storage becomes a manifest plus one blob per playlist,
compressed, so starring a track no longer rewrites the collection.
Adds the UI pass that was missing: the queue verbs, the menus, and the
one real code change behind them — the player's queue can today only
be replaced, not appended to.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
| |
Folders were always ordered by name, whatever the column and direction, so
reversing the name sort or sorting by size moved only the files. A folder's
size is what it holds and its date its newest file; ties go to the name.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
|
| |
Into the folder on screen, under the Upload button's rule. A name already
there, or one the node would refuse, cancels the whole drop with a message.
Folders are recreated level by level; files go out a few at a time. The
in-flight upload guard is keyed by folder and name.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
| |
|
|
|
|
|
|
| |
Previous/next arrows in the pinned toolbar, on group pages and in Search.
Page size is an account preference (Settings → Defaults), 50 by default,
10 to 200 in steps of 10.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|