USB-C power event bursts could repeatedly restart and SIGTERM the settling oneshot until systemd reached its start limit. Let successful runs return inactive and use coalescing start jobs for immediate and delayed reapplication.
Verified: V0 nix flake check --no-build, module parse, and git diff --check; V2 focused battery-charge-limit KVM test with a 12-event active-run burst. V3 hardware item #101 remains queued.
Restore the old flake's external-only workspace handoff, protect closed-lid cable removal with a verified low-level logind inhibitor, and use monitor-added as the settled audio reprobe boundary.
Verified: V2 — nix flake check --no-build; docking-ux, option-docs, and template-sot checks; KVM tools/monitor-fallback.nix lifecycle test.
V3 pending: closed-lid BenQ round 4 in agent/HARDWARE-QUEUE.md.
Real dock QA exposed pointer-bound menus, an absent default rescue watcher, the unsafe zero-output panel-off path, missing keyboard/audio hotplug coverage, and unpainted new outputs.
Route Rofi to the focused output; replace panel disable with safe workspace migration; always run and reconnect the display watcher and repaint hotplugged outputs; expose a default per-device keyboard picker; handle audio card availability transitions.
Verified: V1 — nix flake check --no-build; checks.docking-ux real Home Manager artifact build; option-docs/template-sot; Nix parse, py_compile, and git diff --check. V2 unavailable: no readable KVM. V3 pending in agent/HARDWARE-QUEUE.md.
check.yml has been permanently red since the runner container was
capped at 2 GB (uncapped, it OOM'd the whole 4 GB VPS). Root cause,
measured with GNU time (eval-cache off): `nix flake check --no-build`
walks every output in ONE evaluator process and peaks at ~6.0 GB RSS
on this repo — the checks.* suite alone is ~15 runNixOSTests, each a
full NixOS eval, all accumulating in one heap.
New tools/ci-eval.sh: identical coverage (all checks.* drvPaths, both
nixosConfigurations toplevels, the HM activationPackage — enumerated
dynamically, no drift), one fresh nix process per output so the heap is
freed between outputs. Cold per-output peaks: worst check 0.99 GB
(hardware-toggles), nomarchy toplevel 0.78 GB, HM 0.60 GB; the one
outlier is nomarchy-live at 2.69 GB (live-ISO eval), which fits the
2 GB cap only via the container's swap allowance (docker's default
--memory-swap is 2x memory). check.yml and bump.yml both call the
script now; bump.yml also gains max-jobs=1/cores=2 (it had no limits,
so its V1 build gate could spawn parallel builders in the capped
container). ROADMAP's stale "no CI today" corrected: bump.yml landed
8fded63 on schedule 2026-07-06.
Verified: V1+ — ci-eval.sh green locally end-to-end (3m22s, tree peak
2.75 GB = nomarchy-live's process); per-output peaks measured
individually. The decisive proof is the next Actions run on the VPS
itself — watch the nomarchy-live step; if it OOMs there, the runner
needs swap allowed: --memory=2g --memory-swap=6g.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Bernardo promoted the PROPOSED item live: with fingerprint PAM on,
sudo/login should accept whichever factor comes first instead of
pam_fprintd's wait-for-the-reader-then-password. Stock PAM cannot
express parallel factors (linux-pam#301), so this packages
pam-fprint-grosshack v0.3.0 (pkgs/, pinned from GitLab — the
field-standard fprintd fork), source-reviewed before packaging: every
failure path (no reader, no prints, fprintd absent/hung, timeout,
password typed) returns PAM_AUTHINFO_UNAVAIL and falls through; a
typed password is only ferried via PAM_AUTHTOK to the stock
`auth sufficient pam_unix.so … try_first_pass` rule — the module never
validates passwords itself, so it cannot lock out password login.
New option nomarchy.hardware.fingerprint.parallel, default TRUE (the
better UX is what opting into fingerprint PAM buys; false = stock
sequential). Wiring swaps the modulePath of stock fprintd's rule slot
(mkForce) so the sufficient-before-pam_unix ordering is inherited, not
recomputed. README + downstream template rows added.
Verified: V2 — checks.hardware-toggles extended to three nodes, green:
parallel node asserts the grosshack auth line precedes pam_unix in
/etc/pam.d/sudo and that with NO reader a correct password still
passes sudo while a wrong one fails (the lockout-safety invariant);
seqpam node gets stock pam_fprintd and no grosshack; nopam gets
neither. flake check + option-docs + template-sot green.
V3 pending (HARDWARE-QUEUE, AMD dev box): the real type-or-touch race,
fprintd-stopped fallback, hyprlock/greeter after a fingerprint win.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Bernardo's question: with "Laptop screen off" active, does the panel
come back when the dock is yanked? VM answer (new maintainer harness
tools/monitor-fallback.nix, softGL desktop): no — Hyprland 0.55.4
leaves the session with zero active outputs and never re-enables a
soft-disabled monitor. rescue_blackout in the display hotplug watcher
now re-enables every disabled output (immediate, retried, idempotent
keyword + toast) whenever a monitorremoved leaves nothing active;
independent of the profile auto-switch and its settings gate.
Sharp edge documented rather than hidden: the zero-output state itself
crashes Hyprland in the VM ~4/5 runs (aquamarine CBackend::dispatchIdle
ABRT within ~0.5s — faster than any userland rescue; a control run
removing the external with the panel ACTIVE survives 5/5, so the state
is the trigger, not headless removal; crash reproduced with the rescue
provably inert, so not caused by it). Rescue ships as defense-in-depth:
it recovers every surviving case, and the worst case on real DRM is
session-to-greeter on a re-lit panel instead of a black brick. Rider
fix: `grep -c .` exits 1 at count zero — the ||-guard on that pipeline
silently disabled the rescue exactly when it mattered.
Verified: hazard V2 (VM reproduces the dead state); control
discriminator 5/5; flake check green. V3 pending: real-dock
undock-while-off in HARDWARE-QUEUE (run with nothing important open;
crash on real DRM = rework the row to mirror or wait for a Hyprland
bump).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Bernardo's dock test (the #87 V3): audio stayed on EasyEffects→speakers.
Root cause, established live: WirePlumber's stored default
(default-nodes state, written by any past explicit pick) outranks the
priority.session rules, so #87 alone can never move the default on a
real machine; EasyEffects 8 follows the default device (kcfg default
true, verified by watching its output links). Fix: nomarchy-dock-audio
(modules/home/dock-audio.nix, nomarchy.dockAudio.enable, default on) —
a pactl-subscribe watcher that set-default-sinks new dock-class sinks
(regex data imported from modules/nixos/dock-audio-rules.nix, single
source) with a toast, plus a startup sweep for login-while-docked that
never overrides a manual in-dock pick. Unplug falls back by priority.
Menus: Display gains docked rows — "Laptop screen off · everything →
<ext>" (live-only disable; a persisted one could black-screen an
undocked boot), "Screen on · <name>", "Move workspace · → next
monitor", "Swap workspaces" (exactly-2 gate); Audio gains a self-gated
"Send output → <dock sink>" quick row. All self-gate on real outputs.
Verified: V2-equivalent on live hardware (no dock available) — watcher
followed a fake dock-named null sink and fell back on unload, control
run without the watcher stayed pinned (proving the stored-default
diagnosis), EE output links moved mid-playback; generated menu bash -n
green, dock-row jq exercised on real pactl JSON, dispatchers probed on
live Hyprland; flake check + option-docs + template-sot green.
V3 pending: real-dock retest ×2 in HARDWARE-QUEUE (watcher keys on
new-sink events; pre-existing nodes that only flip route availability
would need the queued diagnosis data).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
With settings.autoCommit on, only menu/theme mutations were committed
(theme-sync pathspec-limits to theme-state.json by design), so hand
edits to system.nix/home.nix and lock bumps stayed forever-dirty.
New internal nomarchy-autocommit (nomarchy-lifecycle): live-reads the
same flag, commits everything dirty as "nomarchy: auto-commit before
<pull|rebuild|home switch>" with the swept file list in the body,
fallback identity, never fatal (|| true at call sites). Called at the
top of all three lifecycle commands; before the ff-only pull it also
un-dirties the tree. Exposed via passthru.autocommit + package export;
guarded by checks.lifecycle-autocommit (sandbox repos: sweep, no-op on
clean/off/non-repo). Sync sweep: README, RECOVERY, ROADMAP, rofi
toggle, control-center prompt, theme-sync docstring, JOURNAL.
Verified: V1+ — flake check --no-build green; the new check builds;
end-to-end run of the built nomarchy-rebuild (stubbed sudo/rebuild)
against a copy of a real dirty ~/.nomarchy: 4 files → one commit,
clean tree. No VM boot — no session-visible surface.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Bernardo ran the real power-cycle on the T14s (LUKS + btrfs @swap
swapfile): hibernate → power off → single LUKS unlock → session
restored. BACKLOG #76 deleted; HARDWARE-QUEUE resume entry pruned
(outcome recorded in the remaining no-swap-notify entry, JOURNAL, and
the ROADMAP ✓-entry); Decisions hibernation note marked shipped. The
no-swap Hibernate notify check remains the one open leftover.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Raise priority.session for HDMI/DisplayPort (1100) and USB (1050) ALSA
outputs so the default sink follows dock plug/unplug; built-in analog
wins again when externals disappear. Rules live in
modules/nixos/dock-audio-rules.nix; checks.dock-audio covers the contract.
Verified: V2 — flake check --no-build; checks.dock-audio green.
V3 pending: HARDWARE-QUEUE #87 real HDMI/USB-dock hotplug.
Explicit HandleLidSwitchDocked=ignore under services.logind so closing
the lid with an external display connected does not suspend. Undocked
lid path left to upstream defaults. checks.clamshell-logind asserts the
baked logind.conf and active logind; real multi-monitor lid V3 queued.
Verified: V2 — flake check --no-build; checks.clamshell-logind green.
V3 pending: HARDWARE-QUEUE #86 lid+external on a laptop.
HARDWARE.md §9 lists T14s Gen 4 + Latitude 5410/5310 with DMI/profile
notes and a "send your DMI line" PR invite. Split dock life #84 into
#86 clamshell + #87 dock audio (stable new IDs).
Verified: V0 (docs/backlog only).
If the installer left /var/log/nomarchy-hm-preactivate.log and no Home
Manager generation exists, doctor reports ✖ with the recovery one-liner
(home-manager switch --flake ~/.nomarchy -b bak). Clears once a
generation is present. checks.doctor covers both arms.
Verified: V2 — checks.doctor green; local smoke.
nomarchy-what-changed wraps nvd into "N added, M removed, K updated"
summaries. nomarchy-rebuild / nomarchy-home print full nvd and toast
the one-liner; System › What changed? offers toast + floating report.
checks.what-changed fixtures nvd via NOMARCHY_NVD.
Verified: V2 — flake check --no-build; checks.what-changed green;
local smoke on real HM generations.
One-shot user unit fires a dismissible "You're set" notification on the
first graphical session (SUPER+M / SUPER+T / SUPER+? + Network + Doctor),
then writes settings.firstBootShown into the flake checkout. Skips the
live ISO hostname (hosts/live.nix keeps its own toast). Toggle:
nomarchy.firstBootWelcome.enable. checks.first-boot covers the gate.
Verified: V2 — flake check --no-build; checks.first-boot green; local
shim smoke; downstream-template-home ships the unit.
Report-only cycle_count + retained capacity (charge_* or energy_*) as
% of design on system batteries; soft warn below 70% of design; skip
when attrs missing. NOMARCHY_POWER_SUPPLY_ROOT for the checks.doctor
fixture (test_power has no cycle/design attrs).
Verified: V2 — nix flake check --no-build; checks.doctor green (fake
BAT0 cycles/% + Device-scope ignore); local BAT0 smoke (energy_*).
Set BOTH nomarchy.nightlight.latitude/.longitude and the night light
switches backing unit: wlsunset computes sunrise/sunset from the
location daily (the fixed .sunrise/.sunset are ignored;
.temperature feeds the night temp). Everything user-facing is
unchanged — same nomarchy-nightlight toggle script, Waybar moon, and
live-state ExecCondition gate; the unit name is a single let-binding
so all three follow together. Declarative-only on purpose: coordinates
are machine config like the keyboard layout, so home.nix (commented
template example), not a menu writer.
checks.nightlight-geo guards the wiring at pure eval: rendered
wlsunset ExecStart carries the coords + night temperature, the
ExecCondition gate landed on the swapped unit, hyprsunset stays off in
geo mode.
Verification: V1 — nix flake check --no-build exit 0 (evals the new
asserts + option-docs sync); built downstream-template-home (default
hyprsunset path unregressed). V3 pending: visible warm shift at the
location's night + instant toggle/off-persistence on the swapped unit
(HARDWARE-QUEUE § Any machine).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Push #76's swapfile-resume residual into a VM as far as headlessly possible.
A full hibernate→resume from a RUNTIME swapfile turned out not to be
headless-reachable: systemd reports CanHibernate=no for a swap that isn't
declared in the boot config, and a runtime-computed resume_offset can't be
fed into the boot cmdline (the genuine chicken-and-egg). So the check now
verifies, bounded and deterministically, what the partition test
(checks.hibernate) can't and what a VM CAN prove: the installer's btrfs
@swap swapfile is built correctly and its offset is valid — NOCOW
mkswapfile, file-type swapon, a good `map-swapfile -r` offset the kernel
accepts (/sys/power/resume_offset), and zram (prio 100) above it.
The literal encrypted-swapfile power-cycle + LUKS-initrd unlock stays a V3
laptop check (the dev box runs exactly this layout, /proc/cmdline verified).
(First cut had an unbounded wait_for_shutdown that hung on the CanHibernate
refusal; rewritten with a thread-bounded wait + a CanHibernate pre-check,
which gave the clean verdict. Device is /dev/vdb without useNixStoreImage.)
Verification: V2 PASS (nix build .#checks.x86_64-linux.hibernate-swapfile,
fresh, exit 0); flake check --no-build green. #76 residual narrowed in
BACKLOG/ROADMAP.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bernardo rightly pushed back on deferring #79 to a real session: a
simulated sunset IS VM-testable. Add checks.auto-theme, a runNixOSTest
that installs nomarchy-theme-sync, points NOMARCHY_DEFAULT_THEMES at the
presets, stubs the rebuild via NOMARCHY_REBUILD (a marker), configures
the day/night pair + schedule, then WALKS THE VM CLOCK across the
boundary: 19:55 -> 20:05 -> 20:06 -> next-day 07:30, asserting the active
theme goes summer-day -> night (1 rebuild) -> night (idempotent, still 1)
-> day (2 rebuilds).
This isolates the auto decision + apply + rebuild-trigger; the actual
home-manager switch is stubbed (that generic apply path is exercised by
every manual theme change). The only on-hardware residual is the user
timer firing on its OnCalendar schedule, which systemd-analyze validates.
#79 drops its [blocked:hw] tag — it's V2-verified. ROADMAP updated;
lesson recorded in MEMORY (don't default timer/session features to V3:
stub the generic step, simulate the clock).
Verification: V2 PASS (nix build .#checks.x86_64-linux.auto-theme, fresh,
exit 0); V0 flake check green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds an "Auto theme (on/off)" row to Look & Feel and an `autotheme`
submenu: toggle the feature, pick the day/night themes (from
`nomarchy-theme-sync list`), and set the sunrise/sunset times (HH:MM,
validated). This is the menu surface over settings.autoTheme, completing
the auto time-of-day theme feature (slices: CLI / timer / menu).
Enabling writes the flag and defaults the summer-day/summer-night pair
--no-switch, then runs `auto --force`, so a single rebuild both installs
the timer and applies the correct theme for now. Disabling is instant
(--no-switch): the flag flips and `auto` self-gates, so the lingering
timer no-ops until the next rebuild. Day/night/time writes are live
--no-switch (read by `auto` at tick time).
Verification: V1 — HM generation builds (bash -n on nomarchy-menu at
build) + bash -n on the built menu. Logic test with stubbed
rofi/theme-sync/notify: selecting the toggle from off runs exactly `set
day/night/enable ... --no-switch` then `auto --force` — the intended
one-rebuild path. V0 flake check green. All three #79 slices shipped;
the live timer switch across sunset is a real-session (V3) check on a
Nomarchy machine.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
nomarchy-doctor gains a read-only hibernate/sleep section (the highest-ROI
Day-2 gap, a natural #76 follow-on): resume device on the kernel cmdline
(+ resume_offset for a swapfile), disk swap >= RAM, zram active, and any
suspend/hibernate error in the previous boot's kernel log. Self-gates
(skip on swap=0) and emits only ok/warn/skip, so it never fails the sheet.
Adds `hibernate` to the checks.doctor probe list so the row is asserted.
Fixes a set -euo pipefail hazard the first VM run caught (exit 1, no
verdict): a no-match grep inside $(...) and a `cond && action` both abort
under set -e — now guarded with || true / || echo 0 / if. Recorded in
MEMORY as a reusable gotcha for writeShellScriptBin scripts.
Verification: V2 PASS — checks.doctor VM test green (fresh build, exit 0;
asserts the hibernate row renders, healthy run stays 0). V1 — ran on real
LUKS+@swap hardware: all three ok rows render (resume set, swap 34G >= RAM,
zram 16G) and the sheet reaches its verdict. V0 bash -n + flake check.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
theme.nix now resolves nomarchy.theme.iconThemePackage from a small
icon-pack registry (Papirus/Tela/Qogir/Reversal/Numix, matched by name
prefix). Default themes name Papirus-*, so the resolved package is the
plain papirus-icon-theme derivation — byte-identical closure, zero added
MB. Only when a theme's `icons` names a set from another pack is that
pack (and only it) symlinkJoined alongside papirus. stylix.nix consumes
it via mkDefault. Adding a pack is one iconPacks row.
Deliberately NOT shipping more packs by default: papirus alone is already
~1 GiB and each extra pack adds ~hundreds of MB, so variety is opt-in per
install (set `icons` in theme-state.json), not a default closure cost.
Documented in templates/downstream/home.nix + ROADMAP § Icon themes.
Verification: V1 (HM activationPackage builds). Eval proofs — default
resolves to plain papirus (zero delta); icons="Tela-dark" ->
nomarchy-icon-themes symlinkJoin (papirus+tela) with the name flowing,
without building tela; all 5 pack attrs exist in the pinned nixpkgs;
template-sot + option-docs + flake check green. No default visual change
(same papirus), so no screenshot; the opt-in render is downstream.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add checks.hibernate: a runNixOSTest modeled on nixpkgs' hibernate.nix,
wired to the actual risk #76 introduces — coexistence with the zram swap
that oom.nix now enables by default. It imports oom.nix, puts the disk
swap at a lower priority than zram with an explicit boot.resumeDevice,
and uses systemd-initrd. The test drives a full
hibernate -> power-off -> resume -> crash cycle: a volatile ramfs marker
survives resume but NOT a crash-boot (proving genuine resume), and it
asserts zram sits at priority 100 while the disk swap is lower, so the
hibernate image lands on disk rather than volatile zram.
What QEMU genuinely can't do — feed a boot-time resume_offset for a btrfs
swapfile (offset only exists after first-boot allocation) or drive a
LUKS-initrd unlock — stays a V3 laptop check (HARDWARE-QUEUE, dev box
already runs that exact layout), plus the no-swap notify in a real
session. With this, all V0/V1/V2 work on #76 is done; #76 collapses to a
[blocked:hw] stub and the design record moves to ROADMAP.
Verification: V2 PASS — nix build .#checks.x86_64-linux.hibernate built
the driver fresh and exited 0. nix flake check --no-build green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
#69: audit-theme-design tags hue/CVD/ANSI-family noise on white,
vantablack, lumon, hackerman, matte-black, miasma as [identity] with a
legend — do not retune identity palettes into traffic lights.
#70: import-palettes keeps surface≠overlay when color0==color8, derives
light UI chips from base (ANSI-black trap), gentler light mantle; roles
are first-class. No bulk re-import of shipped theme JSON.
Verified: V0 (py_compile + synthetic import fixtures + audit report).
Add docs/VISION.md (product themes, v1.0 bar, agent slice hints) and
README maps under docs/ and agent/. Point ROADMAP, BACKLOG, LOOP,
CLAUDE, HARDWARE, and the root README at the split: VISION = intent,
BACKLOG = executable queue, ROADMAP = design history, .claude = harness
only. Mark v1.0 track in BACKLOG PROPOSED for human triage.
Slice (c), the last of the docs review: the install/first-run story
read end to end against the tree. Facts held (mkFlake args match the
table, disko's @snapshots + hibernation swapfile, .#default/.#me
naming); the one seam — README §3 never pointed the manual path at the
template README's first-run steps — got a connective paragraph. Item 6
deleted from BACKLOG; ROADMAP § Full docs review flipped to ✓.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Slice (a) of the docs review: 9 undocumented nomarchy.* options got
README rows or inline .leaf mentions (autoTimezone, camera IR-hide,
intel.guc, amd.pstate/vaapi, package, system.stateFile, restic.paths).
The diff is now a flake check: option names are eval-walked from the
four option-declaring modules (config halves stay lazy) and compared
to the tables by tools/check-option-docs.py — undocumented options and
stale table rows both fail. Negative-tested against the pre-fix README
(reports exactly the 9 gaps). Item 6 split into slices a–d in BACKLOG.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Follow-on to the swaync fix (a643391): the generated Waybar CSS colored
the module row and #window @subtext and dim states @muted, and the
generated rofi theme put @fg text on @surface chips — invisible where
palettes use those roles as "on-surface" (flexoki-light: subtext==base,
surface==text; gruvbox: muted/base = 1.27). flexoki-light has no
whole-swap, so it ran this CSS. Generated surfaces only — the summer
whole-swaps are hand-authored and untouched.
- waybar: secondary text → alpha(@text, 0.85); dim states (inactive
workspaces, DND bell, muted volume) → alpha(@text, 0.5).
- rofi: @dim → text+"80", @surface → text+"1a" (#RRGGBBAA fg tints).
- check-theme-contrast.py pairings/docs now cover all three generated
surfaces; status glyphs (good/warn/bad on base, worst 2.05) are
deliberately not held to a text ratio — raising them is palette
design, a human decision.
Verified: V0; V1 (checks.theme-contrast green over 21 themes;
downstream-template-home builds; rendered waybar style has zero
@subtext/@muted/@surface; rendered custom.rasi shows fg-derived
tints). Honesty: NOT visually verified — V3 queued (flexoki-light
eyeball + dark-theme regression spot-check).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Palette roles aren't uniform: summer-day and flexoki-light use subtext
as text-ON-surface (== base) and surface as a dark chip (== text), so
swaync's @subtext-on-@base body text and @surface/@text buttons were
self-colored (caught on the Latitude).
- body/summary: @text on @base — the one text pairing that passes in
every palette (survey: worst WCAG ratio 5.18 across all 21 themes;
subtext/base bottomed at 1.00 twice, muted/base at 1.27).
- hover rows + widget-title buttons: alpha(@text, 0.1) tints —
contrast by construction; subtext/surface defines dropped.
- NEW checks.theme-contrast (tools/check-theme-contrast.py): asserts
the hex-on-hex text pairings across themes/*.json, cheap (no VM).
The audit also found the generated waybar/rofi CSS shares the bug
class, live on flexoki-light (no whole-swap) → BACKLOG item 27.
Verified: V0; V1 (checks.theme-contrast builds green, 21 themes;
rendered services.swaync.style has zero @subtext/@surface). Honesty:
NOT visually verified — V3 queued (notify-send on summer-day).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Recovers the crashed iteration #7 (items 21+23, Latitude QA findings):
- nomarchy-waybar supervisor (waybar.nix, exec-once'd from
hyprland.nix): any waybar exit respawns the bar; crash-loop guard
(5 fast exits -> critical notify + stop); TERM trapped for a real
stop. Fixes the bar-less session after a theme-switch crash.
- nomarchy-theme-sync prefers a clean `pkill -x waybar` (supervisor
restart with fresh config+style) over the crash-prone in-place
SIGUSR2 reload when the supervisor is running; SIGUSR2 stays as the
unsupervised fallback.
- sys-rebuild: snapshot-first system rebuild against the CURRENT lock
(no `nix flake update`) — the no-update twin of sys-update;
README §3/§5 + motd list it.
Verified: V0 re-run green after crash recovery (flake check --no-build
+ py_compile); V1/V2 per the crashed session's journal entry (HM
renders exec-once=nomarchy-waybar; headless software-GL VM:
SIGKILL -> new pid, clean kill -> respawn, SIGUSR2 -> alive).
V3 pending on the Latitude (queued in HARDWARE-QUEUE.md).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
SUPER+SHIFT+K → hyprctl switchxkblayout current next, rendered only
when the session layout has a comma — the same gate as the Waybar
language indicator. Bind data is a separate multiLayoutBinds list in
keybinds.nix (still one source); hyprland.nix and the rofi cheatsheet
gate on the identical condition, so bind and cheatsheet cannot drift.
`current` targets the focused keyboard, so per-device-overridden
boards (single layout) are a no-op.
Parity: hyprland/language (with the keyboard glyph) added to both
summer waybar.jsonc whole-swaps + #language in their CSS. Static JSON
cannot eval-gate, so on summer themes the module shows even with one
layout — parity-over-minimalism, deliberate. Found en route:
summer-night has an idle_inhibitor the generated bar lacks (reverse
gap) — folded into backlog item 13.
Also — agent/BACKLOG.md gains items 21-24 from Bernardo's Latitude
5410 hardware QA (waybar crash on theme switch + no respawn; nameless
network-menu rows; a no-update sys-rebuild; Back-everywhere audit),
queued ahead of everything; HARDWARE-QUEUE's snapshot re-test now
carries the "update to >= a47aa3a + relogin" precondition, since
findings 5/6 match the pre-polkit-agent behavior exactly.
Verified: V0 (flake check, both summer JSONs parse) + V1 both
directions (default single-layout build renders no bind; us,de via
extendModules renders bind AND cheatsheet row). V3 cycle check queued.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Re-diagnosis of the "btrfs-assistant 2.2 segfaults on launch" bug, and
the fix is not the planned nixpkgs override:
1. gdb: the crash is btrfs_util_subvolume_iterator_next() —
libbtrfsutil's UNPRIVILEGED subvolume-iteration path in btrfs-progs
6.17.1 (upstream-fixed after 6.17.1, kdave/btrfs-progs 886571653).
Symbol versions were checked and match: not ABI/link drift.
2. VM A/B on real btrfs: as root it works (exit 0); as a user it
segfaults (139). The pkexec launcher runs it as root — so the GUI
path was fine all along, except…
3. …the distro shipped NO polkit authentication agent: every pkexec
prompt in the session failed silently. That is the actual root
cause of "the GUI doesn't open".
Shipped: hyprpolkitagent (Hyprland's Qt agent — the prompt is
Stylix-themed) via exec-once in hyprland.nix; `nomarchy-menu snapshot`
prefers btrfs-assistant-launcher again, with the nomarchy-snapshots
fzf flow kept as fallback (and for SSH); checks.snapshot-gui guards
the root path + offscreen GUI event loop on a real btrfs volume so a
lock bump can't silently regress it. No btrfs-progs patch: all our
flows are root-side; the unprivileged fix arrives with a lock bump.
Verified: V0 (flake check) + V2 (snapshot-gui check, executed green) +
V1 (HM generation builds; hyprland.conf carries the agent exec-once;
generated menu passes bash -n with the launcher wiring). Remaining V3
queued: first on-hardware themed polkit prompt.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
modules/nixos/oom.nix — running out of memory now kills the offending
process (with a desktop notification via systembus-notify) instead of
freezing the desktop for minutes until the kernel OOM killer fires.
earlyoom over systemd-oomd, deliberately: oomd kills whole cgroups,
and a Hyprland session runs as ONE scope (nothing here spawns per-app
systemd scopes, unlike GNOME) — under pressure oomd would take out the
entire desktop to save it. earlyoom kills the single largest process
before the thrash point. nixpkgs default-enables oomd in an inert
state (no slices monitored); it is disabled outright so there is one
owner. Session plumbing is --avoid-listed (unanchored — NixOS wrappers
rename comm to .foo-wrapped); no --prefer tuning, largest-RSS
selection already finds the hog. All mkDefault; opt out with
services.earlyoom.enable = false (README note added).
Verified: V0 (flake check) + V2 — new checks.oom-protection
runNixOSTest, executed: a chunked allocator (686 MB peak, 1 GB VM) is
SIGTERM'd by earlyoom in 0.1 s, a bystander unit survives (the
process-level granularity the module exists for), and systemd-oomd is
asserted inactive. 25 s runtime.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
settings.autoCommit — toggled from System › Auto-commit (self-gated on
the flake being a git repo) — makes every nomarchy-theme-sync apply/set
also git-commit theme-state.json in the downstream flake, so settings
history is `git log`. Off by default.
Design: the flag is live-read by the tool on each write (nothing in Nix
consumes it → instant toggle, no rebuild, no new option); the commit is
pathspec-limited to theme-state.json so unrelated dirty files are never
swept up; it fires when the flag is on before OR after the write, so
the disable-toggle itself lands in history; same-value writes no-op
(diff against HEAD); missing git identity falls back to
Nomarchy <nomarchy@localhost>; `bg next` is deliberately excluded.
Rider fix: `get` now prints booleans JSON-style (true, not Python's
True) — un-sticking the System menu's "Auto timezone (on/off)" label,
whose `= true` comparison could never match. All shell consumers
already normalise true|True, so no other behavior change.
Verified: V0 (py_compile, nix flake check) + V1 (HM generation builds,
generated nomarchy-menu passes bash -n, wiring present in the rendered
script) + a 7-assertion sandbox-repo round-trip covering enable/set/
same-value/disable/post-disable/apply paths and the identity fallback.
Remaining: on-machine check queued in agent/HARDWARE-QUEUE.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A git-tracked agent/ directory so AI agents can iterate on the distro
unattended (runner-agnostic: /loop, headless claude -p, or a fresh
manual session — all state lives in the checkout, per the distro's own
philosophy):
- LOOP.md — the iteration protocol: orient → pick one BACKLOG task →
verify up the V0–V3 ladder → commit+push main → record. Safety rails
(v1 untouchable, no force-push, no surprise lock bumps) and
stop-and-escalate conditions.
- BACKLOG.md — the forward half of docs/ROADMAP.md reworked into a
prioritized queue (5 NOW / 6 NEXT / LATER / PROPOSED / Decisions);
ROADMAP.md stays the design/decision record + shipped log.
- GOALS.md — the four pillars (stable > reproducible/zero-hidden-state >
effortless config > beautiful), quality bars, non-goals.
- CONVENTIONS.md — coding/design rules (in-flake state, menu placement,
Waybar whole-swap parity, toggle-vs-package, no formatter).
- MEMORY.md — curated hard-won lessons (VM recipes, btrfs-assistant
segfault watch, rofi/WirePlumber/hyprlock gotchas).
- HARDWARE-QUEUE.md — every pending V3 on-hardware check collected from
the ROADMAP, with exact steps, split by machine.
- JOURNAL.md — append-only iteration log, seeded with this bootstrap.
Plus a root CLAUDE.md entry point and README/ROADMAP pointers.
Verified: V0 — docs-only; nix flake check --no-build green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Phase 3 of the in-flake-state philosophy (mirroring night-light + the
keyboard-layout watcher): a System ▸ Display menu item picks an output's
resolution and persists it into the flake instead of a runtime-only tweak.
- rofi.nix: the `display` case reads `hyprctl monitors -j` (auto-skips the
chooser with one output), lists the output's advertised modes (rounded
refresh, deduped, highest-first) plus the preferred/highres/highrr
tokens, applies the pick INSTANTLY via `hyprctl keyword monitor` —
keeping the output's current position + scale, so only the resolution
changes — then persists it with `nomarchy-theme-sync set
settings.monitors.<name> <mode> --no-switch` (no rebuild).
- theme.nix: seed `settings.monitors = {}` (output-name -> "WxH@R") so a
sparse/older state file still evaluates; exposed via nomarchy.settings.
- hyprland.nix: `resolvedMonitors` overlays settings.monitors onto
nomarchy.monitors by name — a declared output keeps its other fields
with only the resolution swapped, an undeclared output (wildcard-only,
e.g. the laptop panel) becomes a fresh rule. So a later rebuild bakes
the menu pick into the generated monitor rule — the monitor twin of the
keyboard graduation. The menu pick wins over a hand-set resolution (it's
the explicit live action).
Monitor connector names are dot-free, so a direct dotted set-path is safe
(no whole-map read-modify-write like the keyboard watcher needs).
VM-verified via a headless-Hyprland nixosTest (run + removed; too heavy
for a CI gate): the graduation baked the rule into the generated
hyprland.conf, the set/get round-trip persisted, and `hyprctl keyword
monitor` actually changed the reported resolution. Only the interactive
rofi picker isn't headlessly driveable (same caveat as the keyboard watcher).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Completes the remaining "Distro branding, round 2" work.
GRUB theme: hosts/live.nix builds nomarchyGrubTheme whose background is
the same composed splash as the isolinux/BIOS splash (accent logo on
base), with a palette-coloured boot menu and accent timeout bar. Derived
from nixos-grub2-theme only to reuse its bundled DejaVu .pf2; the stock
NixOS logo is dropped (ours is in the background). Wired as
isoImage.grubTheme so UEFI boot matches BIOS.
distroId: set system.nixos.distroId = "nomarchy" so os-release is honest
(ID=nomarchy, ID_LIKE=nixos). Safe — switch-to-configuration builds its
"is this NixOS?" guard from the configured distroId (and /etc/NIXOS is
the fallback), so rebuilds keep working. isNixos→false blanks the
nixos.org URLs, so extraOSReleaseArgs restores them to the project.
Adds checks.distro-id (runNixOSTest): boots distroId=nomarchy and runs
switch-to-configuration dry-activate green; os-release eval-verified from
the real downstream config; grub theme built + structure-verified.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add checks.battery-charge-limit (runNixOSTest): the test_power module
fakes a Mains adapter, toggling ac_online emits a real power_supply
uevent, and the test asserts the udev rule restarts the oneshot (changed
InvocationID). Verifies the re-apply *trigger* at runtime, beyond the
earlier eval check; the sysfs write still needs real hardware. Also
records the SSH_AUTH_SOCK rendered-artifact verification in the roadmap.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The agent's SSH socket was only exported via the zsh integration, so a
GUI client launched outside a shell (rofi launcher, autostarted apps)
couldn't use gpg-agent for SSH. Add a home.sessionVariables export
resolving the socket with gpgconf --list-dirs agent-ssh-socket — the
same lookup the shell integration uses, so the two can't drift — which
reaches GUI apps via the login shell that starts Hyprland (the path
NIXOS_OZONE_WL already takes).
Eval-verified: the var resolves to the gpgconf substitution; the home
config still builds.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The threshold was written boot-only, so a firmware that clears
charge_control_end_threshold on unplug would silently lose the cap
until the next reboot. Add a services.udev rule on the mains adapter
(ATTR{type}=="Mains", vendor-neutral) that restarts the oneshot via
systemctl --no-block on every AC state change. Uses restart (not
try-restart) so it re-applies even if the boot run was inactive.
Eval-verified: rule present when the charge limit is on, absent when
off; downstream-template-system still evaluates.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
