.forgejo/workflows/check.yml runs on every push to main/v1 (+ manual
dispatch): nix flake check --no-build (full module-system eval incl.
the downstream template through mkFlake), py_compile of
nomarchy-theme-sync, and bash -n over tracked .sh files. The always-on
net under direct-to-main pushes — first slice of the ROADMAP lock-bump
CI item.
Scoped to the eval tier deliberately: the instance's runner is an
act_runner docker container (no systemd, no /dev/kvm — established
from the legacy repo's .gitea/workflows/check.yml, which ran 57 times
on it), so the checks.* VM suite and real builds can't run there. A
commented vm-checks job documents the KVM-runner upgrade path; the
legacy workflow's container gotchas (nixbld setup for the single-user
installer, sandbox=false for Stylix IFD, Nix pinned 2.31.5 vs
lazy-trees, no JS actions past node20) are carried over verbatim in
the header. docs/TESTING.md §1b documents what a green run does and
does not mean.
Verified: V0 locally (the same check commands, minus the container
Nix install) + YAML parse. A real green run depends on the runner
still being registered — not API-visible unauthenticated, so that is
queued as [human] BACKLOG item 20.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Gitea only scans .gitea/workflows and .github/workflows; it does not scan
.forgejo/workflows (that path is Forgejo-only). The CI was therefore
dormant on a Gitea instance regardless of whether Actions was enabled.
git mv to .gitea/workflows/ (scanned by both Gitea and Forgejo). Updated
the live references in STRUCTURE.md (Gitea + act_runner), AGENT.md, and
today's ROADMAP entry; left the dated historical Shipped entries as the
changelog they are.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
nix flake check --no-build only evaluates the four default configs, so a
bug that only fires when a nomarchy.* toggle is flipped sails through —
exactly how the vscode option rename and the impermanence systemd-stage-1
assertion both reached main undetected.
Add bin/utils/nomarchy-eval-matrix: layers each opt-in scenario (16
toggles, both system and home halves) onto nixosConfigurations.default
via extendModules and forces system.build.toplevel.drvPath (which forces
assertion checks); also forces all 21 palettes through the system-side
Plymouth base00->RGB math that the home-only allThemeVariants never
reaches, plus both standalone homeConfigurations. Per-scenario results
come from builtins.tryEval so one failure doesn't mask the rest; verified
it goes red on a re-injected assertion failure and a bad option ref, green
on the current tree.
Wired into .forgejo/workflows/check.yml after flake check (jq via nix
shell so it doesn't depend on the runner image). AGENT.md §5 now tells
future agents to add a scenario when they add/rename an opt-in option.
nomarchy-docs-scripts: filter systemd-service declarations (.services.
nomarchy-) and denylist the eval-matrix repo tool so neither shows as a
dangling "missing" reference; also retroactively clears the latent
nomarchy-flathub-init drift from the gaming-flathub commit.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Audited every entry in `installer/hardware-db.sh` against
`inputs.nixos-hardware.nixosModules` and found **21 of 43 entries (49%)
referenced modules that don't exist** in the upstream attribute set —
those installs would fail at eval time with "attribute not found"
errors on real hardware. Specifically:
- Framework 13 per-gen: nixos-hardware uses `framework-11th-gen-intel`,
not `framework-13-11th-gen-intel`. Fixed all four generations.
- Framework 13 AMD AI 300: `framework-amd-ai-300-series` (no "13-").
- Framework Intel Core Ultra: added `framework-intel-core-ultra-series1`.
- Framework 16 AMD AI 300: added `framework-16-amd-ai-300-series`.
- Framework generic fallback now uses the `framework` umbrella module.
- ThinkPad X1 Carbon: modules are `lenovo-thinkpad-x1-Nth-gen`,
not `-x1-carbon-genN`. Fixed gens 6/7/9/10/11; added X1 Nano.
- ThinkPad P14s: requires arch+gen suffix; switched to the AMD gen3/4/5
modules (the prior `lenovo-thinkpad-p14s` had no attribute).
- Surface Pro 6/7/8/10: all share `microsoft-surface-pro-intel`. Pro 9
keeps its dedicated module. Pro 3 fixed to `-pro-3`. Surface Book
2/3 and Intel-based Surface Laptop 3/4/5: no nixos-hardware module
— rows dropped; generic chassis+cpu+gpu detection still emits
sensible `common-pc-laptop`.
- ASUS ROG Strix G513 → `asus-rog-strix-g513im` (correct attr name).
- ASUS ROG Zephyrus GA403 didn't exist — dropped. Added `ga402x`,
`gu603h`, `g533zw`.
- ASUS Zenbook generic `asus-zenbook-ux` was non-existent — dropped
(too vague; available modules are per-model like `asus-zenbook-ux481`).
- Dell Latitude 5400 / 7480: no modules — replaced with the existing
`dell-latitude-7420`, `7430`, `7490`.
Added:
- ROG Ally / Ally X support (`asus-ally-rc71l` for `RC71L`,
`RC72LA`, and the "ROG Ally" product string). nixos-hardware
currently ships one module for both revisions.
Documented (in a footer comment) the devices nixos-hardware doesn't
cover so they're known-unsupported rather than accidentally missing:
- Valve Steam Deck → Jovian-NixOS as a separate flake input.
- Snapdragon X laptops → aarch64 only; Nomarchy installer is x86_64.
- Raspberry Pi → same as above.
Bug discovered along the way: the DB's pipe-separated row format
collides with bash regex alternation. A row like
`Microsoft|Surface Pro (10|8|7|6)|_|module` parses as 7 fields, with
"7" extracted as the module name. Surface Pro variants are now one
row per version.
CI gate added (`.forgejo/workflows/check.yml`): a new step extracts
every 4th-pipe-field from `HARDWARE_DB` and `comm -23`s it against
`inputs.nixos-hardware.nixosModules`. Any future entry pointing at a
non-existent module fails CI with a clear error. Closes the regression
class entirely.
Verified locally: bash -n + shellcheck --severity=error pass on
hardware-db.sh; the CI step's exact commands pass against the new DB.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Pillar 7 first step. `.forgejo/workflows/check.yml` runs on every push
to main and every PR. Three sequential checks in one job:
1. `nix flake check --no-build`
Catches eval regressions: broken option references, missing imports,
stale module argument shapes. The same command AGENT.md tells humans
to run by hand before declaring a change done.
2. `bash -n` + `shellcheck --severity=error` over every `nomarchy-*`
bash script.
Mirrors what `.githooks/pre-commit` does locally, but across the
whole tree on every push — so a branch that bypasses the hook (via
`--no-verify` or a fresh clone without `core.hooksPath` set) still
gets gated. Severity is capped at error to match the hook; the long
tail of style/info warnings can be cleaned up incrementally.
3. `docs/SCRIPTS.md` drift check.
Regenerates the audit doc to a temp file and `diff`s against the
committed version. Fails loudly with the fix command if a script
add/remove/rename didn't include the regeneration step.
Dry-run results on the current tree:
- `nix flake check --no-build`: pass (only pre-existing warnings).
- shellcheck across 159 scripts at severity=error: pass.
- SCRIPTS.md drift: clean.
Activation:
Forgejo Actions isn't enabled on the repo yet, so the workflow lands
dormant. To activate: enable Actions on the repo in Forgejo's settings
and register a `forgejo-runner` on any Docker-capable Linux host. The
workflow uses `ubuntu-latest` and installs Nix itself via
`DeterminateSystems/nix-installer-action`, so no special runner image
is needed.
Deferred to a follow-up batch (needs binary cache infra):
- Building ISOs in CI (`nomarchy-installer`, `nomarchy-live`, default).
- Release pipeline (`vYY.MM.x` tags publishing ISOs as artifacts).
- `nixosTest` per palette with golden-image screenshot diffs.
`docs/STRUCTURE.md` now documents `.forgejo/` and `.githooks/` so future
agents and contributors can find both.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
