# veilor-os — Test Method Changelog

Append-only log of changes to `test/TESTING.md`. Each entry: date, the
veilor-os version it first applied to, what changed in the procedure,
and *why*. The why is the load-bearing part — without it this file
becomes a list of opinions.

Entries are newest-first.

---

## 2026-05-05 · v0.5.27 · TESTING.md created

**Change:** First version of the canonical procedure document.

**Why:** Through v0.2 → v0.5.26 we'd been reproducing the test
procedure ad-hoc each time, which meant test runs were uncomparable
and regressions were caught only by accident. The v0.5.26 → v0.5.27
debugging session surfaced the LUKS-cmdline bug, the GRUB rebrand
gap, the gum-cursor render glitch, and the fbcon KMS issue all in a
single VM run — but only because the test happened to walk every step
in order. Codifying the steps means the next regression is caught the
same way reliably.

The procedure documents the **hybrid VM method** explicitly: Claude
drives every step it can via QEMU monitor `sendkey`, the human types
LUKS + admin passwords directly into the QEMU window because plymouth
ignores synthesised keystrokes. Past trail (14+ failed sendkey
variants) is the source of truth for that limitation; do not re-fight
that battle without first rereading the trail.

The procedure also separates **VM** (cheap iteration, catches install
logic) from **real hardware** (mandatory for tag, catches firmware /
KMS / GPU). Future releases must produce a `test/test-runs/` report
for each before tagging.

**Files added:**
- `test/TESTING.md` — canonical procedure
- `test/METHOD-CHANGELOG.md` — this file
- `test/test-runs/` — per-run reports go here (template lands with
  first real run, currently empty)