| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Reported from use: a note created with only a title then saved left a db.json
entry with no file beside it, and reopening it said the file was missing.
Cause: FileView.setText("") on a path that view has not loaded creates no
file at all and reports no error. The index write (always non-empty JSON)
was fine, so only the body was lost, which is why it looked like a missing
file rather than a failed save. Bodies are now written by a Process running
cat > <file> with the text on stdin: verified against empty and non-empty
payloads, consecutive writes on one Process, and a store with no db.json at
all. The exit code is the real outcome, so a refused write reports it.
Second report in the same flow: the editor stays editable after the first
click, so typing more and then pressing Commit & push dropped the edit, the
push carrying only what the first click wrote. The second click now writes
the current title and body before it commits and pushes.
|
|
The confirm was on the second click, after the first had already removed the
file and the entry, so the 'Confirm?' label promised a check that had
already happened and one stray click lost the note. The first click now only
arms the button; the second removes and pushes the removal.
|
|
A review of the branch found four ways the page could report success for
work that never happened, all reachable in ordinary use.
Writes: setText is asynchronous unless blockWrites, and its
saved()/saveFailed() do not fire for every write (a second setText to the
same path loses the signal, which stranded the callback), so a save was
reported before it landed and a failed db.json or body write was never
seen. Both views now block and carry the outcome, so writeDb and writeBody
know the result when they return; a failed create therefore leaves no index
entry. The create path also loads the body before the index, so a failed
body write cannot leave an entry with no text.
Commits: the sync chain swallowed every commit failure with '|| true', so a
failing pre-commit hook or a signing failure pushed an unchanged HEAD and
reported success. It now stages, commits only when something is staged, and
lets 'set -e' fail the chain, with stderr folded in for the whole chain
rather than just the push.
Duplicate notes: the 'awaiting commit' state lived only in a 3 second
timer, so a slower second click re-entered create() and appended a second
note. It is now a page property that outlives the timer, and the button
follows it.
Preview: the shared Switch writes checked itself, dropping the binding, so
the switch stopped following page.preview; resynced on change, the way
StatusRow.qml does.
Also validates a note's file name as digits before read, write or rm, and
clears the index error once the index reads again.
|
|
The editor, and the write and git path behind it. Save is two clicks and
each does one thing: the first writes the note and only then arms the
button as 'Commit & push?', so a quick second click cannot commit a note
the first has not written yet.
The write path refuses when db.json cannot be read as the index. That is
not hypothetical: git pull --rebase --autostash resolves a remote db.json
change against a local uncommitted one, the stash pop conflicts, git still
exits 0, and the file is left full of merge markers. Reading that as 'no
notes' and writing it back deleted every title in a probe; Notes.valid now
distinguishes it from an empty store and the write reports instead. parse
stays tolerant for display. Stamping lastpull also only follows a pull
that worked, and only when git is on at all.
Db reads that a write is based on go through reload() plus waitForJob():
blockLoading does not make reload() blocking, contrary to the note in
AGENTS.md, and reading text() straight after reload() returns the old
contents. A body is written with a plain setText, which needs no prior
load; only reading text() before a load would.
|
|
A drawer tile and page for bash-notes: browse, write, edit and delete
notes in the page, sharing notes.sh's store so both tools work on the
same notes.
Git follows notes.sh's shape, pull before write, commit and push after,
with two departures the page needs: an unconfirmed Save leaves db.json
uncommitted (write and commit are two clicks here, one invocation there),
so the pulls carry --autostash, and lastpull is stamped after the pull
rather than before it, which notes.sh does and then ignores the failed
pull that follows.
|
|
The canvas: a To .md ended any other way than its own Stop (the footer
Stop, a question, a drop, a daemon restart, an older daemon refusing the
command) left the card at Working forever. The footer Stop goes through
stopOp(), asking and dropping end the operation first, a disconnect ends
it, and an error while nothing else runs is taken as its outcome. Each
fileops run gets its own Process, so a stopped run's late answer is never
read as the next one's. To .md takes one PDF, as one file is written.
Compress... no longer flips back when file(1) refines a kind, long errors
are elided, and a desktop with no documents folder gets
Documents/Assistant rather than ~/Assistant.
fileops: an animated GIF to PNG or JPEG left one file per frame under
names the rename never found; the first frame is taken. JPEG flattens
transparency onto white. Zipping a folder holding the output folder read
its own growing archive without end; only regular files outside it are
added, which also skips FIFOs and dangling links. qpdf warnings no longer
fail a merge. A folder's free name keeps its dots, a lone .gz or .xz is
decompressed rather than refused, and a dotfile's archive is not hidden.
Tools die with fileops even when it is killed outright, a killed run's
leftovers are swept on the next run, a stop between fork and bookkeeping
cannot miss the child, and a late stop cannot add a second answer.
Error lines lose bsdtar's doubled prefix and its delayed-exit noise.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
One CLI for converting, extracting, merging and compressing staged files,
answering with one JSON line. Results are written under a temporary name
and renamed to the first free one, so a failure or a stop leaves nothing
half written, and a stop kills the tool's whole process group. This
ffmpeg has no libx264 and no AAC encoder, so MP4 is H.264 through VAAPI
with MP3 audio, and audio extraction copies AAC, MP3 and Opus tracks as
they are. The test builds its inputs and checks every result's type.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Local conversions, extraction, merge and compression through a small
fileops CLI, results in Documents/Assistant, and To .md through the
daemon's new write command.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Canned questions per staged kind in the file state, stored in Prefs with
built-in defaults so a future settings page can edit them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Checked by hand: resume rule, island drop and its refusal message, the file
state, Chat about it, the picker, and the Escape order.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
A system file dialog would open under the overlay that holds the keyboard,
so the picker is drawn in the card, over the transcript, from
Qt.labs.folderlistmodel.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Three card states: an entry line, a file state offering what to do with
dropped files, and the chat. A drop always starts a new task, and asking
about a file is one of the choices rather than the only one.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
All nine visual checks passed by hand.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
A Luma-style overlay on SUPER+Home, inside the desktop process, as a pure
view over desktop-assistant's socket. It replaces the voice overlay in
assistant/ and stands in for llamachat while it is tested.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
A drawer page, reached from the user page, listing every global git key and
editing, adding and removing values through git config itself, so the file
keeps its comments and quoting. --fixed-value lets each write target one
value of a multi-valued key.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
The git sync was meant to include user.signingkey and the spec had
dropped it, claiming nothing reads a GPG key. git does. The page offers
the secret keys gpg holds, plus None, and writes the fingerprint.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
passwd in a terminal already does it, AccountsService's SetPassword needs
admin rights and a hash, and a change desyncs the login keyring.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
A drawer row greets the login user and opens a page that edits name,
email, avatar, language and location, written where other programs
already read them: AccountsService (and through it the passwd GECOS
LibreOffice reads), KDE's emaildefaults and the global git config, with
the surname shortened to an initial for git.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
A git hash starting with a letter (3.1-a59215d) ordered behind its base,
because any trailing alphabetic chunk counted as a pre-release; only
alpha, beta, rc, pre and dev do now. Trailing zeros (1.14 against
1.14.0) compare equal instead of reading as an upstream bump. A bad line
in the slackpkg blacklist is skipped rather than crashing the script and
leaving the drawer with nothing. An AppImage keeps its key over a
flatpak of the same name, as the spec orders them, and flatpak list has
a timeout. A Slackware upgrade compares version and build, so an arch
change alone is not a row. The spec's prose now matches the code: the
upstream comparison includes the installed version.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Drafting the code found two holes in the spec. Awaiting build was
limited to the user's own recipes, so an SBo package whose recipe had
caught up but was not yet built (nessus) showed in no section; it now
covers any recipe. And "the newest ChangeLog line decides" would never
call a package new once it had been upgraded after being added; only
Added and Removed lines count now. The prefix rule for kernel suffixes
is dropped as redundant: version order already ranks a suffixed version
newer, which is never behind.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
The conky updates card copies nvtable's table, where installed-behind and
repo-behind share a row and are told apart by colour, and it measured
"shipped" as the .info version: gamescope and kitty read as ready to
install while the binary repo still held the old build. The drawer module
sorts each tracked package into one action, from PACKAGES.TXT of the
binary repo rather than the recipe: upgrade here, awaiting build, upstream
newer, app updates. Slackware is compared mirror against installed, not
against slackpkg's local ChangeLog, which `slackpkg update` replaces
without installing anything; new and removed packages come from the
mirror's ChangeLog. NFS reads go through a timed-out child because the
mount is hard.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
The drawer keeps its look and takes over: the island's controls panel goes,
since each of its tiles opened a drawer page that already existed, and SUPER
alone opens the drawer instead. The tile grid scrolls inside a three-row box
so new tiles do not eat the notification space. New: a performance page
reusing the island's Perf service, a cliphist clipboard page replacing the
rofi list, and a quiet mode in the status registry that mutes Claude and
opencode sounds through one notify-sound script while their popups stay.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
On screen the spiral alone read as loose: separation keeps the spiral's
gaps, and the previews filled 24 to 30% of the screen. Thirty rounds now
pull every window 15% closer to the centre and push them apart again,
which fills 35 to 55% and makes each preview up to 1.5x larger. Twenty
windows lay out in about 4ms.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
layout() seeds windows on a golden-angle spiral, most recent at the centre,
pushes overlapping pairs apart along the axis that costs less screen, and
fits the result with one scale so sizes stay relative. The spec's first
seed, the windows' real positions, was prototyped and dropped: every window
on this desk fills its monitor, so they started in one pile, and pushing a
pile apart made a line that still overlapped after 300 passes and filled
8 to 15% of the screen. The spiral converged in 1 to 55 passes. Comparing
overlaps in raw pixels stacked wide windows into a column, which is why the
axis choice divides by the screen's width and height.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
The window-switcher grid becomes a Mission Control scatter: previews keep
their real shape and relative size, seeded from their real positions and
pushed apart, then fitted with one scale. A top-right hot corner on DP-1
opens it after a 250ms glow that warns before it fires, and stays inert
while gaming. shared/Status.qml is not linked into window-switcher because
a second instance would run its own game poll and react to presentation
twice; the corner reads status.gaming directly.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
IconImage cannot be recoloured without an effect layer, so the severity
shows as a dot on the icon's corner. A stale public IP is a dash with no
age, as the Errors section already said.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Sub-project 4. An icon in the status pill opens a card grid of CPU,
RAM, temperatures, GPU, disks and network, read from the same /proc,
/sys and hwmon sources conky uses. The icon warns on temperatures,
RAM, disks and a public IP that drifts from a pinned address, since
the VPN depends on the ISP's fixed IP. CPU load is left out of the
icon because spikes would flicker it. The four conky widgets are
removed once the panel is verified beside them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
One JSON envelope on stdout for every calctl command, the event as an
argument rather than stdin, a get command because the editor edits the
series rather than the occurrence clicked, and a pure Cal.js for the
date logic so it is tested under node like the island's other helpers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Sub-project 3. The clock pill opens a month grid beside the selected
day's agenda, and an editor that writes the vdirsyncer calendars
through a small Python tool, calctl, rather than khal: khal 0.11 has no
JSON output and keeps a cache that is a second view of the same files.
Recurrence stops at a plain FREQ and UNTIL, and the form owns one alarm;
anything richer in an existing event reads as custom and survives an
edit untouched.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Edit mode let a right-click and the wheel through to the tile, so a
right-click opened the drawer and scroll changed volume, against the
spec. The drag area now takes both buttons and the wheel.
A scroll up on a level already above the cap, such as a mic set to 120%
elsewhere, snapped it down to the cap. The cap now only stops a step up.
Prefs logs a failed write, which printErrors off had silenced: the new
order would otherwise vanish at the next restart with no trace. Its
parsed object is renamed from data to doc, so it cannot shadow a base
property of that name.
The spec now says what was built: JSON.parse rather than a JsonAdapter,
and the helper list as it is.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Planning found two places where the spec diverged from the island it
extends. Icons go through the icon theme like every other island icon,
and drag is one MouseArea per tile with a pure drop-slot helper, which
is less code than DragHandler and DropArea and can be tested under node.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Sub-project 2. A grid of seven toggles opened from the status pill and
by a bare SUPER release, with an edit mode to reorder, hide and show
tiles. The visible list lives in a JSON file under the state home, so a
tile added to the code later starts hidden rather than appearing
unasked. The drawer keeps the detail: right-click on a tile opens its
page.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
The sliver was the accent at rest, and udt had snapped the accent to the
same colour as the mic indicator (#36d190), so an idle sliver read as a
live microphone. It is Theme.overlay at rest now; captures keep their
colours whatever the wallpaper does.
The flag's WheelHandler never fired. MouseArea.onWheel, which SoundPage
already uses, does; scroll now switches layout, one step per notch.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Twelve tasks from pure logic to waybar removal. The spec is corrected on
two points found while reading the code: Quickshell's PwNode has no running
state, so mic and screen detection poll pw-dump like the waybar script did,
and the VM dot rule now agrees with its own table on saved VMs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
The same keyboard appears as a different device over USB and over its
dongle, so the main device is resolved per event, never cached.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Sub-project 1 of the move toward a Modesty-style integrated desktop: a
top-centre island on DP-1 hidden behind a sliver, a clock pill and a status
pill that slide out of it, a sound panel, and the removal of waybar.
Records the decisions taken while brainstorming: the island floats over
windows with no exclusive zone, privacy capture recolours the sliver rather
than revealing the pills, the clock pill flicks the desktop icon, app icon and
app name on workspace change and app open but not on focus change, the volume
icon has no gestures beyond opening its panel, and the flag images move into
this repo. Modesty is GPLv3, so only its ideas are used.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
fanfictioner runs sd-cli one-shot, not through sd-server, so the sd row
read down while a CLI run held VRAM and generated. A running sd-cli now
reads up and busy, its model marked (sd-cli), and outranks the web UI.
The model is the first --diffusion-model/-m: sd-cli's prompt follows and
is free text that may contain "-m".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Six TSV lines per sweep: llama and sd engines, four apps. llama's slot
probe passes autoload=false, since a plain /slots?model= loads the model
into VRAM. sd-server has no progress endpoint, so busy is its log
having been written in the last 5s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Engines (llama.cpp, sd.cpp) read-only, apps get start/stop through their
existing controls. Records that /slots?model= autoloads an unloaded model
unless autoload=false is passed, found when a design probe reloaded one.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
The list comes from hyprctl binds -j on every opening, so it is always
what Hyprland has loaded, loop-generated binds included. Labels come
from each bind's description option, since hyprctl reports Lua actions
as an opaque __lua handle.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
The daemon already publishes its phase, state and countdown as files under
XDG_RUNTIME_DIR, and the status module already shells out to it for
presentation mode. Reading those files makes the traffic two-way for the
cost of three FileViews, with no new interface to invent.
Into status/ rather than a module of its own: breaktimer state is one more
thing the desktop is doing, which is what that page already is, and
Status.qml is already where breaktimer.sh is called from.
The config file is the other half, and stands alone. Every tunable is a
shell variable in a tracked script, so the installed copy and the committed
one have already drifted over three sound paths pointing into a home
directory. One sourced file fixes that with no parser and no format.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
|
Design for three new appearance tabs: a hyprsunset-qt-equivalent profile
editor, a hypridle timeout/enable editor that preserves the file's comments,
and an icon/cursor selector with live previews. Records the preview
mechanisms probed on this box (GTK icon lookup, hyprcursor .hlc unzip,
xcur2png) and scopes the hardcoded-theme propagation out as follow-ups.
|
|
The daemon drops every image hint today, so a screenshot and an application image never arrive. libnotify 0.8.8 sends the content image via -i/--icon (the image-path hint) while -n/--app-icon is app_icon. The spec adds the image contract field, daemon hint handling and theme-name resolution, the balloon preview, and a local-only inline image policy.
|
|
The freedesktop spec says -1 means the server decides and 0 means never; the
spec and plan had them reversed, which the shipped code inherited and the
live handover test caught. Corrected to match the spec, libnotify's -1
default and dunst.
|
|
The freedesktop expire_timeout is the notification's lifetime, and --wait
and -b clients block until the daemon closes it. So the daemon emits
NotificationClosed at expiry and frees the client, but keeps the entry in
the live queue: the drawer lists it until it is dismissed or evicted, inert
because its client is gone. History holds only what was dismissed or
evicted.
The reserved space is therefore scrollable, since the live queue can hold
20 at once, and a row past its expiry is read-only.
|
|
The second of the two specs the status registry named. A Go daemon owns
org.freedesktop.Notifications and holds the state; quickshell renders it.
The daemon and notifyctl live in a separate repo, the balloon shell and the
drawer's notification centre live here.
Rendering is split because quickshell has no generic D-Bus module, so
something outside QML has to own the bus name. Files are the interface in
the direction the renderers read, the same convention the registry set, and
notifyctl is the one surface back.
DND suppression lives in the balloon shell, not the daemon: the drawer's
reserved space lists every live notification including the suppressed ones,
because a list the user opened is not an interruption. This supersedes the
registry spec's assumption that the daemon would read dnd.
|
|
F1: after a click the shared Switch writes checked directly, dropping
StatusRow's declarative binding, so an external mode change from waybar
or statusctl no longer moved the switch. Resync via onValueChanged.
F2: the spec claimed hyprctl clients counts idle inhibitors, which is
false for a layer-surface inhibitor; replace with the behavioral
hypridle marker check.
F3: README said the check covers both watch states; it exercises the
activated report and the absent-file down report.
F4: give breakProc an onExited check so a failed breaktimer.sh verb is
visible instead of silent.
F5: drop StatusTile's unused required st property and its injection.
F6: comment the startup-order limit on dndBeforePresentation.
|
|
Desktop modes as state: do not disturb and presentation mode, with one
place that owns them. First of two specs; the notification daemon that
replaces dunst is the second and consumes the registry's dnd mode rather
than owning it. The registry comes first because DND belongs to the
desktop rather than to the notification daemon, and because the daemon is
the larger and riskier component.
The design that needed deciding was not the toggle but the coupling.
Presentation mode sets DND, so DND has two writers, and turning
presentation mode off has to restore what the user had rather than
clearing it. That rule is why this is a registry and not two booleans.
State is one file per mode in XDG_RUNTIME_DIR, matching the convention
breaktimer already writes on this machine. That directory is /run/user/1000,
a tmpfs, so a reboot clears every mode with no cleanup code. Logout without
reboot is less certain and the design does not lean on it: elogind runs
here and its pam module documents removing the runtime directory at last
logout, but the same manual says the module does nothing when the system
was not booted with elogind as init, which on Slackware it is not. If
removal does not happen, a mode survives a logout, which is the same
behaviour as surviving a shell restart.
FileView covers both directions: atomicWrites for the write, watchChanges
for the watch, no shell-out. External watchers need close_write,moved_to
because an atomic write arrives as a rename, the same trap the mail watcher
hit with Xapian.
Confirmed while designing: the compositor advertises
zwp_idle_inhibit_manager_v1 and waybar's built-in idle_inhibitor already
drives it here, so the inhibitor is asserted over Wayland rather than D-Bus.
That module is replaced rather than demoted, because it owns its inhibitor
object and cannot display state owned by anything else; left alongside the
registry it would assert a second inhibitor and idle would resume only when
both released.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A7ThHHh5iTYbVfp3rNAkw2
|
|
Absorbs the kdeconnect-indicator into the drawer: status, battery,
pairing, ping, clipboard send, file share, refresh and mount. State is
polled from a qdbus6 helper script while the drawer is open, since
quickshell 0.3.1 has no generic D-Bus module.
The spec notes the phone-to-PC clipboard is the daemon's plugin, not the
indicator's, so removing the indicator does not touch it.
|
