Parser compatibility

braid parses output from btrfs-progs, cryptsetup, util-linux, smartmontools, NUT, and ethtool. The fragile parser/safety set (btrfs-progs, cryptsetup, NUT, smartmontools, ethtool) is pinned by default; util-linux is host-provided because lsblk --json is a stable structured contract. All parsed tools remain fixture-covered because parser contracts can still drift when tool versions change. Two validation lanes exist:

Stable lane (pinned and stable contracts)

• just test-parsers — focused parser canary. Exercises selected CLI-reachable parsers against live tool output in VMs (including braid-status-ups, the NUT canary); it is not an exhaustive inventory of every CLI parser path.
• just test-rust — validates golden fixtures for the full parser set, including parse_upsc. Fixture-backed coverage stays current only after running just capture-all-fixtures when parser-critical tool versions change (e.g. nixpkgs bump).
• Fixture refresh is a separate obligation: just test-parsers passing does not guarantee the TUI-only parse_lsblk_json, the unused parse_btrfs_scrub_status_per_device, or parse_cryptsetup_luks_dump are compatible with the current toolchain. parse_cryptsetup_luks_dump is CLI-reachable through braid replace’s existing-LUKS target-capacity preflight. The registered replace-new-already-luks and replace-enroll-existing-luks VM checks exercise that fail-closed path against live cryptsetup output in the full VM suite, but neither check belongs to the focused test-parsers recipe.
• parse_smartctl (the SMART health parser) is reachable from both the TUI and the braid status CLI command, so it is no longer TUI-only. It is still not covered by the live VM canary, though: virtio disks emit no usable SMART, so just test-parsers cannot exercise it. Its drift canary is the stable-only smartctl golden fixture (see the smartctl-fixtures note below).
• Fixtures in cli/tests/fixtures/nixos-26.05/ are committed and authoritative. NUT fixtures live in cli/tests/fixtures/nixos-26.05/upsc/ (and the unstable mirror); they are produced by just capture-ups-fixtures, which boots a dedicated NUT VM with per-state dummy-ups drivers (see tests/capture-ups-fixtures.nix).
• smartctl fixtures are stable-only by design. VM virtio disks do not emit useful SMART data, so just capture-all-fixtures does not regenerate smartctl-sata-with-temperature.json, smartctl-nvme-healthy.json, or smartctl-selftest-*.json. smartctl-sata-with-temperature.json is a one-time physical-drive capture; the NVMe and self-test fixtures are hand-authored (see cli/tests/fixtures/nixos-26.05/README.md). The tool-versions VM test checks that smartctl resolves to a /nix/store/ path on the VM’s PATH and that its self-reported version matches pkgs.smartmontools.version, but it does not detect nixpkgs version bumps because both sides advance together. On any nixpkgs bump that touches smartmontools, manually review and refresh smartctl-selftest-*.json against the new ata_smart_self_test_log.standard JSON shape, smartctl-nvme-healthy.json against the new nvme_smart_health_information_log shape, and smartctl-sata-with-temperature.json against the new health/temperature JSON shape (smart_status, temperature, ata_smart_attributes).
• ethtool WoL fixtures are hand-authored / no-live-capture. VM virtio NICs do not emit useful Wake-on-LAN data, so just capture-all-fixtures does not regenerate ethtool output. The doctor wake_on_lan parser is covered by hand-authored Rust unit fixtures, and wrapper provenance is covered by the override-based VM tests in tool-versions and braid-auto-suspend.

Parser-critical tool versions are the pinned nixpkgs versions of btrfs-progs, cryptsetup, nut, smartmontools, and ethtool, plus host-provided util-linux’s lsblk --json stable contract. Treat any change to the nixpkgs node in flake.lock, any flake.nix change that alters the nixpkgs input, or any change to braid.packages.{btrfsProgs,cryptsetup,utilLinux,nut,smartmontools,ethtool} as a required fixture-refresh event.

When parser-critical tool versions change, run:

1. just capture-all-fixtures
2. just test-rust
3. just test-parsers
4. On any bump that moves btrfs-progs or the kernel (any nixpkgs node change moves both): just test-repro repro-btrfs-scrub-limit-bounds-rate – the live-tool behavior lock for the bounded-rate scrub throttle the scrub-window repro tests rest on. This step is manual and required: repro-* checks are excluded from .#checks, so no CI workflow runs it.

Unstable lane (tracked forecast)

Early-warning lane for upstream parser/output drift. Unstable failures signal upcoming changes, not a contract violation. Fixtures in cli/tests/fixtures/nixos-unstable/ are committed so upstream output changes are visible in git history, but they are non-authoritative.

• just test-all-unstable – VM tests against nixos-unstable. Covers CLI-reachable parsers against live tool output but does not cover the full parser surface (TUI-only parsers, unused parsers, smartctl).
• just capture-all-fixtures-unstable + just test-rust-unstable – covers btrfs/cryptsetup/util-linux/NUT against unstable tool output via golden fixtures. Missing fixtures fail (not skip).
• smartctl and ethtool have no unstable fixtures. Unstable capture/test coverage intentionally covers btrfs/cryptsetup/util-linux/NUT only; see the Stable lane for why smartctl fixtures are stable-only and how to refresh them on smartmontools bumps, and why ethtool WoL output is hand-authored instead of live-captured.

Full unstable canary workflow:

1. just test-all-unstable
2. just capture-all-fixtures-unstable
3. just test-rust-unstable