phoebus_os/docs/VM-BOOT-VERIFICATION.md
Phoebus 3b219fc851
Some checks failed
ci / test (push) Has been cancelled
fix(agents): harden both user-level agent paths
2026-10-12 07:15:31 -07:00

1780 lines
104 KiB
Markdown
Raw Permalink Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# Disposable UEFI VM boot verification
This records the first successful end-to-end install and boot of Phoebus OS
against a disposable qcow2 disk. No physical disk was ever touched.
## 2026-10-12 User-level agents — native overlay acceptance
Run `serial-boot-20261012T135107Z-c601627c` passed 10/10 checks with clean
poweroff on `user-agent-hardening-20261012.qcow2`, a disposable backing overlay.
The preserved source-build base image and existing ISO were unchanged.
The normal-user synthetic fixture exercises MCP initialization/no notification
mutation, native DMS/Mango settings and exact undo, external-edit refusal,
read-only tool mounts with workspace writes and hidden private home/environment,
default denied loopback plus explicit network access, and root refusal for both
user binaries. No provider or real account credentials were used.
Static binary SHA-256 for that run:
- `sys`: `57105b06c4ab3f6ec4dfaed0cbcd7e6110958bc96154e423cde5ee0f5ca42fa5`
- `phoebus-mcp`: `c52383ab603d3709badfbbfdfe1327d43c20f9d16213e6cf129ec30b3d24d742`
- `phoebus-agent-run`: `aa076cc17987d175c5d31e49a8b6d22456bf416c1369a38422d302c3dc2ffb8d`
See [USER-AGENTS.md](USER-AGENTS.md) and
[the user fixture](../installer/tests/vm/user-agent-guest.py). Native visual reload,
fresh installation of the companion bundle and actual-provider workflows remain
pending. Portable tests cover actual plan staging with inspected fixture metadata.
Final static run `serial-boot-20261012T141203Z-b2ac2870` passed 11/11 with
clean poweroff. It repeats the user checks and verifies desktop clear preserves
CLI context and elevated changes to another user's history are refused. The
scratch overlay was removed after poweroff; the base and ISO remain unchanged.
Final static SHA-256:
- `sys`: `c4ba6fec072bb29ae9076f67848642185da2992db5b2207fe547429147ec9e4f`
- `phoebus-mcp`: `cfe7cb748803909fd7094458246cb10d0c40208eaf238788cae3102bfe152f6e`
- `phoebus-agent-run`: `5f3db03eea42c6f437394c32b9b18052bbba0200fce95d58c040bce37388ecd6`
## 2026-10-12 Agentic hardening — native overlay acceptance
Run `serial-boot-20261012T105530Z-9c09094e` passed 14/14 checks with clean
poweroff. A separate backing overlay booted the preserved installed Gentoo
image; neither the base image nor the existing ISO was changed. Verified real
developer/user/root sandbox denials, deterministic history, descriptor-backed
Btrfs snapshot, exact Portage install/removal, Flatpak commit install/removal,
active Nix store-path install and persistent kernel lock permissions.
The static `sys` used for that run had SHA-256
`1cf2f0346c3c517923c517bffb81a283eaa4da4972495170322ff587f4dbfb24`.
A subsequent worker-lock lease is covered by a real broker-kill regression;
final static smoke `serial-boot-20261012T112454Z-47d3cfae` passed 8/8
with clean poweroff, repeated sandbox denials and history/snapshot in an
explicitly network-free namespace. The final binary SHA-256 was
`1ea91f6f0ac82d88e9a8c400f510819aefaf73251778b248ef2443fdf18e2b62`. Two earlier attempts were
fixture failures (separate login/LUKS password mapping, then KVM identifier);
they are not claimed as product acceptance. Fresh installation, native Snap
revision/drift and interrupted native package recovery remain pending. See
[AGENTIC-HARDENING.md](AGENTIC-HARDENING.md) and
[the guest fixture](../installer/tests/vm/agentic-hardening-guest.sh).
## 2026-10-07/08 Installed post-install acceptance — ongoing
Fresh encrypted Gentoo installation `serial-install-20261007T122805Z-da765f3d`
passed five checks; detached installed generic boot
`serial-boot-20261007T155220Z-1d9415a3` passed four. Installed GUI run
`gui-desktop-20261007T155403Z-dc002f64` verified both welcome actions and all
seven real bridge reviews after scoped dependency fixes. History metadata
permissions were fixed and retested as the installation user.
GIMP's approved source build completed in
`gui-desktop-20261007T162413Z-516c7589`: 13 signed binary dependencies, then
local GIMP 3.2.4 compilation. The original Flatpak stayed installed and opened
while compilation ran. Verification passed source VDB, original Flatpak
removal, unchanged generic images, preserved data and native desktop default.
The subsequent serial native launch timed out, so that driver overall recorded
22 passes and two harness failures. A fresh graphical boot
`gui-desktop-20261007T171009Z-a8188ad4` successfully opened native GIMP 3.2.4
from DMS spotlight (`05-05-native-gimp-ready.png`).
The first balanced compilation generated an unbranded release: Portage could
not read root-only `/etc/kernel/config.d`. Cancelled that build before install,
verified generic images/boot, repaired configuration modes, and repeated GUI
review/approval. The corrected build began around 2026-10-08 02:38 UTC; actual
build/modprep configurations include `-gentoo-dist-phoebus-balanced` and the
build log confirms the fragment was merged. Original
Firefox/Thunderbird 157.0.1 graphical launches also passed and seeded real
profiles for later conversion checks. The corrected balanced build completed
in 18584 seconds: source VDB, preserved generic image digests and all boot-entry
files pass. Build scratch reached 19.6 GiB; the kernel review now recommends
25 GiB. Source boot `gui-desktop-20261008T074624Z-5e1b43be` passes; its running-
kernel guard refused another build without VDB/config/default changes. Clean
poweroff and persistent generic fallback in
`gui-desktop-20261008T074853Z-fd35d9f2` pass. Four application conversions are
now running sequentially, starting with Firefox. Tuned build/boots and final
ISO acceptance remain pending. In the fresh generic session, original Writer
opened through its desktop entry without a per-launch GTK override; user
systemd environment includes SAL_USE_VCLPLUGIN=gtk3. Screenshot
`04-writer-fresh-session-ready.png` records its window.
## 2026-10-07 post-install build chooser — UI smoke test, installation in progress
- A read-only live VM (`gui-gui-20261007T130820Z-cfc6e979`, no writable
disk) loaded the installed welcome QML and the updated Assistant chooser.
Its Debian live environment is **not** evidence of Gentoo compilation or
installed-system execution. The separate encrypted Gentoo installation is
still running in `serial-install-20261007T122805Z-da765f3d`.
- Found and fixed a real keyboard defect: the welcome advertised Enter, but
Qt's Button did not activate on Return. Explicit Return/Enter handlers now
open both application and kernel choosers. Keyboard focus also scrolls the
kernel button into view on the 600-pixel welcome window.
- Inspected screenshots `07-source-welcome-enter-settled`,
`08-kernel-button-keyboard-scroll`, `09-kernel-welcome-enter-fixed`,
`10-keep-prebuilt-kernel`, `12-balanced-kernel-selection` and
`13-tuned-kernel-selection` under that run's `shots/` directory. Application
selection reached GIMP, the kernel popup listed prebuilt/balanced/tuned, and
keeping the prebuilt kernel started no package action. The preview VM was
stopped after these checks.
- Actual Gentoo review/approval, conversion, compiled-kernel boot and fallback
tests remain pending. Do not treat this smoke test as completed acceptance.
## 2026-10-08 Updated post-install ISO smoke test
The local candidate `installer/live/phoebus-installer-amd64-postinstall.iso`
contains the latest static installer and `sys`, both Assistant QML files, and
GTK3 LibreOffice session setting. Extraction verified exact binary hashes:
`sys` `cc850ef20737275942c90281e01bd2c4f26b0e1d6197713146d4ecb9f746d1f8`,
installer `184ba0ee912e240fe80d39abd3b7294550a2a879c83a502ed886ee158bbc9695`.
ISO SHA-256:
`377c5fa376ac96a9169f101b32a2a20808350ea3dc3282597765860deba6004c`.
xorriso reports BIOS and UEFI entries. Disposable no-target-disk UEFI run
`postinstall-iso-smoke-20261008T045549Z` reached the normal live menu; selecting
its highlighted normal entry with Enter reached `graphical.target` and
`phoebus-live login:`. This verifies selected-entry boot, not automatic menu
countdown timing. An earlier smoke attempt was interrupted at the menu to
add screen capture; it is not a passing run. Installed source conversions and
kernel acceptance are still running; this candidate is not a formal release.
## 2026-10-06 Fresh ISO UEFI boot smoke test
After removing 33 disposable qcow2 images, the live rootfs was rebuilt with
the current package list, desktop assets, static installer and `sys`. The
resulting ISO at `installer/live/phoebus-installer-amd64.iso` is 1,495,138,304
bytes, SHA-256
`5886d8f54c45ce330175edeaf7812fe6815371aac2f8ffa7b4477e4f013c3430`.
The embedded binary hashes matched the current static release builds. xorriso
reported BIOS and UEFI boot entries. A QEMU UEFI boot from the virtual DVD,
with no target disk attached, reached `graphical.target` and the
`phoebus-live login:` serial prompt. Unused Debian installer pools and boot
menu entries were removed from the image; an earlier untrimmed 2.0 GiB attempt
failed firmware loading, while the trimmed ISO booted. This is a live boot
smoke test, not a fresh Phoebus installation or graphical acceptance run.
## 2026-10-03 Installed greeter spacing and gray color
The existing encrypted playground was updated after guest authentication with
the greetd config generated by the source change: `--window-padding 2` and
gray text, border, action labels and shortcut keycaps. After restarting greetd,
the `ESC` keycap was inset from the left edge and the bottom bar used the same
soft gray as tuigreet's Usage screenshot. The inspected host capture is
`/var/tmp/phoebus-vm/greeter-slate-inset-gray-20261003.png`. Pixel sampling of
both the upstream screenshot and the VM capture found `#2e3440` for the
background and `#d8dee9` for the UI gray. The upstream 0.9.1 layout uses
terminal-cell padding and computes status widths from the current terminal
size; it does not position the bar at fixed pixels. A fresh install and reboot
remain pending. No physical disk was touched.
## 2026-10-03 Installed tuigreet slate login
The existing encrypted `playground-v7-a.qcow2` VM initially displayed the
stock black greeter. After the user authenticated in the guest, the new
`/etc/greetd/config.toml`, 16-color `phoebus.vtrgb`, and greetd service
drop-in were installed there. `setvtrgb` applied the palette and restarting
greetd returned to a login screen with the slate background (`#2e3440`),
white border and text, clock, and `Welcome to Phoebus OS` greeting. A host
window capture was inspected at
`/var/tmp/phoebus-vm/greeter-slate-20261003.png`; a background pixel read
RGB `(46, 52, 64)`. This verifies the updated existing guest after a service
restart. A fresh install and reboot with the installer-generated assets are
still pending. No physical disk was touched.
The capture also exposed missing shortcut-key hints in the bottom status bar.
The original theme used `button=black`, which hid them on slate; the generated
config now uses `button=yellow`. After another authenticated guest login,
updating `/etc/greetd/config.toml` and restarting greetd showed yellow
`ESC`, `F2`, `F3`, `F12`, and `CMD` key labels beside their actions. The
inspected capture is
`/var/tmp/phoebus-vm/greeter-slate-shortcuts-20261003.png`. This verifies
the correction in the existing guest; a fresh install and reboot remain
pending.
## 2026-10-01 Visible playground `sys` presentation
A rebuilt `sys` binary was installed only in the writable overlay of the
visible `reliability-v7-a.qcow2` playground (final binary SHA-256
`abf7c5de450751eb3a969091791c9da946eb59500f2c66cfe59d9cde94488abe`).
`sys history --legacy` showed
effect-checking and action progress with clear completion states. A real
OpenCode Go `sys ask what model are you using` showed system/model/evidence
steps and rendered the informational answer as wrapped prose, without an
approval digest. The inspected screenshots are
`/var/tmp/phoebus-vm/sys-progress-history.png` and
`/var/tmp/phoebus-vm/sys-progress-final-model.png`. The final binary visibly
named `OpenCode (opencode-go/deepseek-v4.1-flash)`, and the user's
`sys find glances` request completed in about 15 seconds with package-lane
checks and a wrapped answer (`/var/tmp/phoebus-vm/sys-progress-find-mid.png`).
The ten-second
heartbeat is unit-tested with a shortened interval, not timed in this VM
run. No fresh ISO or install includes this source revision yet.
## 2026-10-01 Visible playground polkit correction
The visible playground booted a writable overlay of the verified
`reliability-v7-a.qcow2` install. A software-update polkit dialog rejected the
VM user's correct password while `sudo` accepted it. Root is locked in this
install; the polkit administrator identity was still root. After adding
`/etc/polkit-1/rules.d/49-phoebus-wheel-admin.rules` to the disposable
overlay and restarting polkit, the graphical polkit agent accepted that same
user password for `pkexec /usr/bin/id -u`; the terminal printed `0`.
Screenshots were inspected at
`/var/tmp/phoebus-vm/polkit-before.png` and
`/var/tmp/phoebus-vm/polkit-pkexec-result.png`. No physical disk was touched.
This validates the overlay repair; a newly built ISO and fresh installed
system with the source fix are still unverified.
## 2026-10-01 Reliability candidate
- Final candidate ISO SHA-256
`b14f37e1b04d8e1ec37f77d2a14adb4aaef652770fee995a18a0ac7b986b5281`
contains installer SHA-256
`5bf5d6868c1a22250ff62fc386804110e035fe52d2bc8ec10385c738a65d401a`
and `sys` SHA-256
`c5b079478f01e79ad4e823db87b32f0da7aecc13fb50afffc53cd53c58e8ff5d`.
These match files extracted from the ISO's squashfs. xorriso reports BIOS
and UEFI boot entries; the MBR partitions end within the image.
Its UEFI live graphical gate passed **32/32**
(`gui-gui-20261001T093944Z-0b2eeb6a`), and the real OpenCode model UI
passed **18/18** (`gui-opencode-20261001T094134Z-e2d90704`): connection
tested, proposal validated and four changes applied before install. Two
independent encrypted Mango/Nix installs from this exact ISO passed **5/5
each** on
`reliability-v7-a.qcow2` and `reliability-v7-b.qcow2`
(`serial-install-20261001T094417Z-4e5913c7` and
`serial-install-repeat-20261001T094417Z-392fae4f`).
- The fresh second install passed detached counted-boot assessment **5/5**
(`serial-boot-assessed-20261001T134254Z-591a4911`). Its installed Mango/DMS
graphical seat passed **30/30**
(`gui-desktop-20261001T134354Z-9c970fae`), including a visibly readable
first-login welcome, Escape/seen-marker dismissal, no DMS onboarding or
changelog popup, terminal, launcher, lock/unlock and logout. The screenshot
was inspected. Offline snapshot recovery passed **6/6**
(`serial-recovery-20261001T134701Z-455b71b6`), three interrupted normal
boots followed by automatic recovery passed **9/9**
(`serial-fallback-20261001T134759Z-1263d7c2`), package/root rollback
passed **11/11** (`serial-rollback-20261001T135201Z-a6caf8ad`), and an
interrupted package transaction with reboot and state verification passed
**10/10** (`serial-interrupted-update-20261001T135600Z-53ea9a73`). These
destructive checks used disposable overlays. Final cleanup on both installs
logged a nonfatal busy LUKS mapper; installation, secret scan and guest
poweroff still passed. The cleanup condition remains to be diagnosed.
- The intermediate ISO SHA-256
`22a197b68a48ef4330f71a4612a8f1b79a6aba1a1d932c4c53426783ccc347a7`
passed its UEFI UI 32/32 and real OpenCode UI 18/18. Its two installs were
deliberately stopped and its partial disks discarded after a graphical
test exposed the missing installed welcome launch. It is not release
evidence for full installs.
- The Forgejo workflow now runs `make check` and fails if the Rust toolchain
is missing. No independent runner has reported a result for this ISO; the CI
release gate remains NotRun.
- The preceding ISO's UEFI graphical live gate passed **32/32** checks on SHA-256
`1c2a2feb3684d167acd5eecf2356a42dd3ca118cd1e16c215a13d7d0aa58a53d`
(`gui-gui-20261001T002227Z-2138a45d`). It opened the installer from the
welcome action, confirmed duplicate prevention, selected `de_DE` using
keyboard filtering, walked the wizard, and enforced the unconnected-model
stop. Earlier runs exposed and led to fixes for a launcher self-kill, initial
focus theft, and the combo's disconnected search filter.
- Two independent full encrypted Mango/Nix installs **passed 4/4 each** on disposable
80 GiB qcow2 disks (`reliability-v5-a.qcow2` and
`reliability-v5-b.qcow2`). Both guests reported embedded installer SHA-256
`fb362e2bfe2abb515593c4e18f66a1bffc3d78d9faf5171db25f4cb253d3438e`,
matching the binary packed into that exact ISO
(`serial-install-20261001T002156Z-be083980` and
`serial-install-repeat-20261001T002209Z-eabe6f49`). Both installs ended
after clean target unmounts; `cryptsetup close cryptroot` returned device-busy
during final cleanup, which is nonfatal and remains to be diagnosed.
- A detached counted boot passed **5/5** after manually rearming a normal BLS
entry on an overlay (`serial-boot-assessed-20261001T083943Z-19937c05`).
`systemd-bless-boot.service`, the failed-unit checker and
`boot-complete.target` were active; the entry lost its counter. The custom
`phoebus-boot-confirm.service` in the preceding ISO was skipped because its
`/run/systemd/boot-loader` condition was unmet and would fail if started
(`bootctl mark-good` is unsupported). The corrected source uses the upstream
blessing unit and only requires the failed-unit health check.
- Offline recovery on a fresh overlay of that install passed **6/6**
(`serial-recovery-20261001T084345Z-803fd654`). The fourth boot after three
simulated incomplete boots automatically chose read-only recovery and passed
**9/9** (`serial-fallback-20261001T084849Z-69692982`).
- A real `sys undo` journey on another overlay passed **11/11**
(`serial-rollback-20261001T090102Z-acf152ae`): it changed the root and
Portage database, restored the paired snapshots, and verified the detached
boot and persistent home marker. The interrupted-update journey passed
**10/10** (`serial-interrupted-update-20261001T090501Z-2e3088c6`): it killed
an active `sys`/emerge transaction, recovered the lock, undid the snapshot
and verified the package database after reboot. The fixture HTTP server
initially served stale scripts; they were synced to the repository before
these passing runs.
- The installed graphical desktop passed **27/27** on the untouched second
install (`gui-desktop-20261001T081722Z-65c0f48b`), including seat0,
Mango/DMS, audio, network, terminal, spotlight, lock/unlock and logout.
The screenshot showed a DMS changelog popup despite the generic firstlaunch
marker. After seeding the versioned `.changelog-1.6` marker on a fresh
overlay, the same graphical gate passed **28/28** and the popup was absent
(`gui-desktop-20261001T084157Z-ccefa3e1`). The corrected installer now
seeds both markers before first login.
- That screenshot also showed no Phoebus installed welcome. The custom Mango
session launched the live welcome but did not run XDG autostart for the
installed welcome. On an overlay with the new session script, launcher and
QML, the welcome appeared, but its original `GroupBox` content overlapped.
The revised fixed-size cards were visibly readable
(`gui-desktop-20261001T091939Z-6d3d9a26`, **29/29**). A second run used
Escape to dismiss it, confirmed the per-user seen marker, then passed the
rest of the desktop gate (`gui-desktop-20261001T092418Z-3f190135`,
**30/30**). These overlay runs validate behavior; the final ISO still needs
packaging and a fresh install.
- The ISO build now uses bounded squashfs workers/cache and replays the
detected El Torito configuration after mapping files. The previous image
copied the 1.9 GiB base MBR partition length into a smaller 1.5 GiB file.
The corrected image reopens cleanly in xorriso, has both BIOS and UEFI boot
entries, and its MBR partitions end within the file. Its squashfs installer
and `sys` hashes matched the static binaries copied into the ISO at assembly.
- A preceding install using the minimal VM fixture failed in `packages_base`
because GURU-only Zsh plugins were in the base package set while GURU was
only synced for Mango. The optional plugins were moved to the Mango set,
and the release install fixture now selects the canonical Mango/Nix profile.
- The offline recovery gate on a preexisting installed-system overlay passed
**6/6** checks: actual systemd-boot recovery entry selection, read-only
`@snapshots` root, and deterministic `sys` without network/model. This is
baseline evidence, separate from the new candidate run.
- The interrupted-update gate passed **10/10** on another preexisting-system
overlay (`serial-interrupted-update-20260930T214842Z-c4d8000b`): it killed
`sys` while a real `emerge --ask=n` process was active, recovered the stale
mutation lock, undid the paired snapshot, booted detached, and checked the
package DB fingerprint and persistent `/home` marker. The first trial caught
a harness race: parsing buffered CLI output missed a fast package install;
the gate now detects the active emerge process group directly.
- The failed-boot fallback preflight passed **9/9** on a disposable overlay of
the older installed system (`serial-fallback-20260930T230340Z-1983d95a`).
Three boots were stopped at the locked-root rescue prompt before health
confirmation. The fourth boot selected the read-only snapshot recovery root,
and the normal BLS entry had reached `+0-3.conf`. An earlier trial continued
from rescue into the normal target, which correctly blessed the boot and
exposed a flaw in the test procedure. The driver now simulates power loss
at the rescue prompt.
## 2026-09-29 Iterative `sys` VM loop on the golden overlay (AI sandbox + journeys)
Fresh qcow2 overlays were created on the golden installed image
(`/var/tmp/phoebus-vm/disks/phoebus-headless-80g.qcow2`) and booted detached with
the static `sys` built from HEAD; the base image was never modified.
- **Loop 1 (`serial-boot-20260929T074650Z-2f8fff0b`) found two real defects:**
(a) unprivileged `sys ask` hung on a polkit
`org.freedesktop.systemd1.manage-units` password prompt because the model
sandbox used the system manager (`RC=124` after a 200 s timeout); (b) an
unprivileged read-only AI plan printed the plan then exited 1 because the
service layer required the root-owned `/var/lib/phoebus/transactions.toml`.
- **Loop 2 (`serial-boot-20260929T075731Z-e4696546`) hit Flatpak's interactive
`appstream-update` polkit prompt** during a read-only `flatpak search` on the
headless serial session (the session is not recognized as polkit-active).
Polkit's upstream default is `allow_active=yes`, so a real active seat is not
affected; the unattended acceptance now runs detached so it cannot hang.
- **Fixes:** sandbox uses `systemd-run --user` with the user-manager property
subset for unprivileged callers (full system sandbox for root); transaction
recording is best-effort for non-root plans and mandatory for privileged
mutations.
- **Loop 3 (`serial-boot-20260929T084043Z-cfa7f7a5`):** read-only acceptance
reached `SUMMARY PASS=49 FAIL=0`; the root sandbox test timed out only because
the cached `sudo` timestamp expired during the long matrix (harness ordering).
- **Loop 4 (`serial-boot-20260929T092147Z-6f88907a`, driver PASS=8):**
system sandbox `AI_ROOT_RC=0`; user sandbox `AI_USER_RC=0` (was 124);
no-sandbox `AI_NOSANDBOX_RC=0`; hostile model `run_shell` reply
`AI_SHELL_RC=1`; read-only acceptance `SUMMARY PASS=49 FAIL=0`; 0 failed
units; clean poweroff.
- **Loop 5 (`serial-boot-20260929T093329Z-4ef893cf`, driver PASS=7):**
mutating acceptance as root `SUMMARY PASS=42 FAIL=0`, including
`mutate-install`, `mutate-remove`, `mutate-history` on `app-misc/hello`.
- **Loop 6 (`serial-rollback-20260929T094640Z-256c2ba1`, PASS=11):**
snapshot → change → `sys undo` → detached verify with the new binary
(`rollback-package-state-restored`, `rollback-home-survives`,
`rollback-root-restored`, `rollback-recovery-entry`, `rollback-status-recorded`).
- **Loop 7 (`serial-boot-20260929T094957Z-3d4a9211`, PASS=11):** Snap journey
(`snap-sys`, `snap-portage`, `snap-socket`, `snap-confinement`, `snap-install`,
`snap-refresh`, `snap-remove`); confinement remains `partial`, and a
revision-changing refresh / strict AppArmor confinement still needs an
external Store revision and an AppArmor-ready guest kernel.
- **NotRun in this loop:** source conversion and tuned-kernel build/boot/fallback
(multi-hour compiles on the disposable image); recovery-entry BLS selection
(manual gate); NVIDIA.
- **Live AI in the VM (`serial-boot-20260929T102703Z-ffd1ec25`, driver PASS=6):**
from inside the guest, the host's real `opencode` agent was reached through the
loopback tunnel (`fwd.py`). `sys ask "list installed packages" --ai-backend
opencode --opencode-provider opencode-go --opencode-model deepseek-v4.1-flash`
returned a valid Assistant Plan + digest (`live-agent` PASS); a hostile
shell/private-file instruction was refused/downgraded
(`live-agent-hostile-rejected` PASS). Separately, a real local model
(Qwen2.5-1.5B GGUF via `llama-server`) was driven through the direct sandbox in
both scopes (`serial-boot-20260929T101322Z-b3d002ed`); it reached the model and
failed closed on schema-noncompliant output. The assistant prompt now shows a
concrete action-object example, which made the small model emit action objects;
a capable model is still required for valid plans, and a very small local
context can be exceeded by the one-shot correction turn.
- **Powerful model via the host agent, `sys` in the VM
(`serial-boot-20260929T105651Z` → `…T111941Z-2be30b89`, driver PASS=8):** four
real requests through the host `opencode`/`deepseek-v4.1-flash` agent:
`list` → valid grounded plan; `search firefox` → valid grounded plan;
`install firefox --flatpak` (unprivileged) → stopped before mutation;
hostile shell/private-file instruction → refused. The first run hung on
`org.freedesktop.Flatpak.appstream-update` (a `pkttyagent` password prompt from
a read-only Flatpak probe in an inactive session); after the
`CommandSpec::detached()` fix the re-run completed with **zero auth prompts**.
The install investigation itself stopped on the model's own request-count /
formatting limits and failed closed (no `sys` defect).
- **OpenCode sandbox parity (`serial-boot-20260929T114319Z-c43c2a92`, driver
PASS=8):** re-ran the same four real-agent requests with the OpenCode transport
now sandboxed by default (`systemd-run --user` for the unprivileged `sys`).
All four passed with the sandbox enabled — `list`/`search` valid plans,
constrained install gated, hostile instruction refused — proving the OpenCode
transport works inside the same OS-level boundary as the direct provider.
## 2026-09-28 Headless dual-target golden install (Run 13) & detached boot proven
- **Golden install completed (Run 13, `serial-install-20260928T002557Z-e39d1934`):**
Full 12-stage non-interactive headless installation on an 80 GiB disposable
target (`/var/tmp/phoebus-vm/disks/phoebus-headless-80g.qcow2`). No physical
disk was touched. All 347 desktop packages emerged in 10,202 s (~2.8 h).
588 package compile times recorded into target metrics. Coordinated initial
known-good snapshots (`@-20260928T060439-initial-known-good` and `@var`)
created. Stage 12 `finalize` succeeded cleanly with `===INSTALL-RC=0===` and
`NO_SECRET_LEAK`.
- **Detached boot proven (`serial-boot-20260928T075316Z-c1918e9c`):**
Booted without live ISO attached (`-boot order=c`, headless `-display none`).
UEFI systemd-boot passed BLS entry to dracut initramfs, LUKS2 encrypted
root automatically unlocked via passphrase, root `@` subvolume mounted with
`zstd:3`. Systemd PID 1 reached target userspace with **0 failed units**.
- **P0 Seat blocker proven fixed in target:**
`seatd.service` active and running; `/run/seatd.sock` exists with `0660 root:seat`;
user `phoebus` and greeter verified in group `seat`.
- **Fast defaults verified in target:**
- Gentoo VDB (`/var/db/pkg/`): `www-client/firefox-bin`, `app-office/libreoffice-bin`,
`mail-client/thunderbird-bin`, and `www-client/brave-browser` (pinned overlay
`d3a97f22…`).
- System Flatpaks (`/var/lib/flatpak/app/`): `com.calibre_ebook.calibre`,
`org.gimp.GIMP`, and `com.github.tchx84.Flatseal` installed with exported
desktop launchers and recorded in `/var/lib/phoebus/flatpaks-installed`.
- **Installed `sys` acceptance matrix:**
`sys-acceptance.sh` ran inside the booted guest: **40/40 PASS, 0 FAIL**.
- **User shell verified:** Modern Zsh interactive login with
`phoebus@phoebus ~ ❯` prompt.
## 2026-09-26 fast-defaults follow-up — fresh ISO built + content-verified, no install run
- Fresh ISO from the fast-defaults commit `e747baf`:
`installer/live/phoebus-installer-amd64-fastdefaults.iso` (1.6 GiB,
SHA-256 `ad9cddb2237f2b90b0327665cbc0c058fc71ca45fd5013e5da7c8c162c857df5`,
build id `20260926T145700Z`), built with `PHOEBUS_REUSE_ROOTFS=1`
(cached Debian rootfs reused; payload, static binaries, Gentoo key,
branding re-installed fresh). The static binaries were rebuilt from
`e747baf` first: `phoebus-installer`
(`8cf90d0fa5358c3fd791972a0489b6db0cd1fe4e5811a9e09f18322b7c751d4a`)
and `sys`
(`c875572d7afddeccde0b8bf624dafe2c3d275e0a21d0fb35d4f015107451a96a`).
The Sep-26 `250dbe9a…` ISO predates the fast-defaults commit and is kept
alongside; the new ISO file is git-ignored (never committed).
- Content inspection (extracted squashfs, not timestamps): embedded
`/usr/local/bin/phoebus-installer` and `/usr/local/bin/sys` hash-match
the fresh static build byte-for-byte;
`/usr/share/openpgp-keys/gentoo-release.asc` shows both published
fingerprints (`13EB…2910` weekly, `D99E…2058` signing) under
`gpg --show-keys`; `/usr/bin/mango`, `/usr/bin/dms`,
`etc/xdg/quickshell/phoebus-install/shell.qml`, the install desktop
entry and `/etc/phoebus/{version,build}` (`0.1.0`,
`PHOEBUS_BUILD_ID=20260926T145700Z`) are all present.
- Quality gates re-run against `e747baf` in this session: `cargo fmt
--check` clean, `clippy --workspace --all-targets -- -D warnings` clean,
`cargo test --workspace` 757/757, release build ok,
`test_vmsafety.py` 18/18, shellcheck info-only (pre-existing SC2035/
SC2012/SC2164 notes).
- NO disposable install/boot ran in this session. Blocker is wall-clock
time, not the host: `make doctor` reports 20 ok, 0 warnings, ~103 GiB
free (ISO build needs ≥20; the Mango target needs 60 minimum, 80
recommended). A full Mango install previously took ~165 min, which does
not fit this session.
- Blocked gates (all NotRun): installed `-bin` identities
(`www-client/firefox-bin`, `app-office/libreoffice-bin` VDB records),
system-wide Flatpaks + `flatpaks-installed` manifest, first-login
welcome, graphical Calibre/GIMP launches, Nemo/browser-link/MIME tasks,
drag-and-drop/clipboard/printing, Calibre removable-media access,
normal boot + offline generic-kernel recovery, one source-conversion
journey (e.g. Firefox `--buildpkgonly` → swap → verify), one tuned-kernel
build + separate-entry test boot, measured time/disk figures, and the
37-documented-case `sys` acceptance re-run.
- Unblock: install the fresh `ad9cddb2…` ISO to a disposable 80 GiB qcow2
with network (`serial-driver.py install` or `gui-driver.py install`);
boot detached (`serial-driver.py boot`); run `sys-acceptance.sh` plus
the graphical/file/device/source/tuned gates below. Never point DISK at
`/dev/*`; never commit the ISO/images/transcripts/secrets.
## 2026-09-26 fast-defaults session — no VM run (BLOCKED on session time)
- No fresh ISO was built and no disposable install/boot ran in this session.
- Blocker is session time, not the host: `make doctor` reports 20 ok,
0 warnings, 104 GiB free (ISO build needs ≥20; the Mango target needs
60 minimum, 80 recommended). A full install previously took ~165 min,
which does not fit this session.
- Blocked gates (all NotRun): installed `-bin` identities
(`www-client/firefox-bin`, `app-office/libreoffice-bin` VDB records),
system-wide Flatpaks + `flatpaks-installed` manifest, first-login
welcome, graphical Calibre/GIMP launches, Nemo/browser-link/MIME tasks,
drag-and-drop/clipboard/printing, Calibre removable-media access,
normal boot + offline generic-kernel recovery, one source-conversion
journey (e.g. Firefox `--buildpkgonly` → swap → verify), one tuned-kernel
build + separate-entry test boot, measured time/disk figures, and the
37-documented-case `sys` acceptance re-run.
- Unblock: build the ISO from the fast-defaults commit
(`installer/live/build.sh`; record commit + build ID + SHA-256); install
to a disposable 80 GiB qcow2 with network (`serial-driver.py install`);
boot detached (`serial-driver.py boot`); run `sys-acceptance.sh` plus
the graphical/file/device/source/tuned gates below. Never point DISK at
`/dev/*`; never commit the ISO/images/transcripts/secrets.
## 2026-09-26 workstation session — no VM run (BLOCKED)
- No fresh ISO was built and no disposable install/boot ran in this session.
- Blocker: `make doctor` reports 8 GiB free on `/` (needs ≥20 for the ISO
build; the Mango workstation target needs 60 GiB minimum, 80 recommended).
- Blocked gates (all NotRun, exact procedure in this file + STATUS):
installed workstation apps, Kitty default, Zsh login, cursor consistency
(ordinary/text/link/resize/busy after reboot), local CUPS job, sshd
disabled+inactive after reboot, startup-check recheck (PPD D-Bus, DMS
matugen/cava integration), both assistant themes at 1024×768/1280×720/
scaled, model handoff with a non-secret fixture, live authorized-provider
result, Brave upstream release + checksums, Snap revision/strict, NVIDIA.
- Unblock: free ≥80 GiB; record `PHOEBUS_BRAVE_PIN` (reviewed 40-hex
another-brave-overlay commit); build the ISO (commit + build ID + SHA-256);
install to a disposable 80 GiB qcow2; boot detached; run the gates. Never
point DISK at `/dev/*`; never commit the ISO/images/transcripts/secrets.
## Exact tested path
Host (Ubuntu, no root required):
- QEMU `qemu-system-x86_64` (extracted locally under `/tmp/opencode/qemudeb`),
OVMF (`OVMF_CODE_4M.fd` + `OVMF_VARS_4M.fd`).
- Target disk: `/tmp/opencode/vm/work/target.qcow2`, 30 GiB, virtio-blk
(`/dev/vda` in the guest).
- Live boot: the reference ISO's own `boot/gentoo` + `boot/gentoo.igz`, with
`-append "root=live:CDLABEL=gentoo-amd64-livegui rd.live.dir=/
rd.live.squashimg=image.squashfs cdroot console=ttyS0,115200"`.
- A host `python3 -m http.server` serves the installer tarball, the pinned
stage3 and the test config to the guest over QEMU user networking.
Phase 1 (install):
1. Boot the live ISO headless; get a root shell on the serial console.
2. Download the installer and test config from the host.
3. Write secrets to `0600` files and run the installer binary (the original
run used the Bash prototype; the harness now serves `phoebus-installer`):
`phoebus-installer --config test.conf --install --yes
--luks-passphrase-file FILE --user-password-file FILE`
4. Stages executed in order: `preflight, storage, stage3, fstab,
portage, base packages, kernel, bootloader, users, services, desktop,
finalize`.
Phase 2 (boot the disk alone, ISO detached):
5. Power off, then start a new QEMU with **no ISO and no `-kernel`**, boot
`order=c`, same OVMF NVRAM.
6. Firmware runs `\EFI\systemd\systemd-bootx64.efi`; systemd-boot boots the
"Gentoo Linux" BLS entry from the ESP.
7. `systemd-cryptsetup` prompts for the LUKS2 passphrase, unlocks
`/dev/vda2`, mounts the Btrfs `@` subvolume as `/`.
8. systemd (PID 1) starts, serial getty appears, user login succeeds.
## Verified results
- Firmware booted the disk's own systemd-boot (Boot000C "Linux Boot Manager").
- LUKS2 passphrase prompt and successful unlock.
- Btrfs subvolumes mounted: `/`, `/home`, `/var`, `/.snapshots`, `/swap`
(later fix), ESP at `/efi`.
- `cat /proc/1/comm` -> `systemd`.
- `systemd-analyze`: 1.442s kernel + 5.755s initrd + 2.942s userspace.
- Login as `phoebus` with the configured password; groups:
`wheel audio video render input plugdev power network libinput`.
- Networking up (`enp0s2 10.0.2.15/24`).
- `systemctl --failed`: exactly one unit, `swap-swapfile.swap` (fixed after
this run; see below).
## Bugs found and fixed during this work
1. **Subvolumes created after mounting `@`** - `storage_prepare` mounted
`subvol=@` before `@` existed (mount exit 32). Added `btrfs_create_layout`
(mount top-level, create subvols, unmount, then mount layout) and a
regression test (`tests/test_btrfs_order.sh`).
2. **Generated fstab overwritten by stage3** - fstab/crypttab were generated
during storage, then the stage3 tarball replaced `/etc/fstab`. Moved
generation to a new `fstab_generate` stage after extraction.
3. **`/var/tmp` absent in the target** - systemd's `kernel-install` needs a
temporary directory; `/var/tmp` was missing. The fstab stage now creates
`/var/tmp`, `/tmp`, `/root`, `/home`, and chroot commands set
`TMPDIR=/tmp`.
4. **installkernel needs `uki_generator` even for BLS** - with an explicit
`/etc/kernel/install.conf`, `05-check-config.install` always requires a
`uki_generator`. BLS now writes `uki_generator=none`.
5. **dracut could not build the `crypt` module** - the target lacked
`sys-fs/cryptsetup`. Added `sys-fs/cryptsetup` and `sys-fs/btrfs-progs` to
the base package set.
6. **No `systemd-cryptsetup`** - `sys-apps/systemd` was built without the
`cryptsetup` USE flag, so the encrypted root could not be unlocked. Added
`cryptsetup` (and `policykit`) to the systemd package.use entry.
7. **Kernel command line used unnamed LUKS** - `rd.luks.uuid` yields
`luks-<uuid>`, not the configured name. Switched to
`rd.luks.name=<uuid>=<name>`.
8. **Dracut UKI did not embed the command line** - the dracut `uki_generator`
produced a UKI with no `.cmdline`, so it booted with no `root=`. Deferred
UKI in favour of a reliable BLS entry carrying `options`. UKI remains
future work.
9. **fstab did not mount `@swap`** - the swapfile unit failed because `/swap`
was never the `@swap` subvolume. Added the `subvol=@swap` fstab entry.
10. **`local root="$1" dir="$root/..."` class bug** - reintroduced in
`boot_write_dracut_conf`; split into two `local` statements. This class of
bug is now covered by the regression suite.
11. **Wrong package atoms** - `gui-apps/fuzzel` (not in Gentoo) -> `gui-apps/wofi`,
`app-misc/brightnessctl` (not packaged) removed,
`app-crypt/gnome-keyring` -> `gnome-base/gnome-keyring`,
`gui-apps/polkit-gnome` -> `gnome-extra/polkit-gnome`.
12. **Package policy compiled too much** - unknown packages defaulted to
source. `balanced` now compiles cheap builds and uses signed Gentoo
binaries for everything else, which cut installer time substantially.
## Rust installer parity (same VM path)
The Rust installer (`installer-rs`) reproduced the entire path on a fresh
disposable qcow2 disk using the static binary
`phoebus-installer` (`x86_64-unknown-linux-gnu`, crt-static):
- Same config (legacy flat format), same pinned stage3 URL + SHA-256.
- `INSTALL_RC=0`; stages: storage, stage3, fstab, portage, base packages,
kernel, bootloader, users, services, desktop (none), finalize.
- Policy decisions matched Bash: `sudo`/`networkmanager`/`polkit` binary,
`cryptsetup`/`btrfs-progs` source.
- The installer recorded **81 then 97 measured compile times** from Portage's
`emerge.log` (real data, not estimates).
- `NO_SECRET_LEAK` from the log secret scan.
Booted with the ISO detached:
- UEFI firmware ran systemd-boot; BLS entry booted.
- LUKS2 unlock, Btrfs `@` root, `/home`, `/var`, `/swap`, ESP at `/efi`.
- `cat /proc/1/comm` -> `systemd`.
- `systemctl is-system-running` -> **running** (not degraded).
- `systemctl --failed` -> **0 loaded units listed**.
- `systemd-analyze`: 1.378s kernel + 5.238s initrd + 4.114s userspace =
10.730s.
- Login as `phoebus` succeeded; networking up (`10.0.2.15/24`).
This is the point at which the Rust installer satisfies the parity criteria and
may become the primary installer. The Bash implementation remains the
reference/fallback until the desktop milestone is also verified.
## Desktop and Flatpak (Rust installer, verified)
The Rust installer, running the desktop profile, installed the full Wayland
desktop and Flatpak successfully (`INSTALL_RC=0`; 326 measured compile times
recorded). Key change: a desktop install now selects the supported stable
Gentoo profile `default/linux/amd64/23.0/desktop/systemd` instead of
reconstructing X/Wayland/dbus USE by hand from the minimal systemd profile.
This resolved the earlier `cairo[X]`/`libglvnd[X]`/`libxkbcommon[X]`
conflicts, and `sys-fs/{btrfs-progs,cryptsetup} -man` avoided a
Sphinx/Pillow circular dependency.
Booted with the ISO detached and logged in:
- `systemctl is-system-running` -> `running`; `systemctl --failed` -> 0 units.
- `flatpak --version` -> **Flatpak 1.18.1**; `flatpak remotes` -> **flathub
configured**.
- Binaries present and runnable: `dwl 0.7`, `Waybar v0.15.0`, `wofi 1.4.1`,
`mako`, `swaybg`, `pipewire`, `wireplumber`, `grim`, `wl-clipboard`.
Session verification. The VM's serial session has **no logind seat**, so
wlroots' DRM backend correctly refuses (`Failed to start a DRM session`). This
is a headless-harness limitation, not an installer defect. Using
`WLR_BACKENDS=headless WLR_RENDERER=pixman`:
- `dwl` started (pid observed), created the `wayland-0` socket.
- `mako`, `waybar`, and `swaybg` started and connected (`swaybg` found output
`HEADLESS-1`).
- `wl-copy`/`wl-paste` round-tripped `phoebus-clip`.
- `grim /tmp/shot.png` produced a **14401-byte PNG**, proving the compositor
renders and captures.
- `wpctl status` reported the PipeWire graph; `systemctl --user --failed` -> 0.
- `waybar` logged a non-fatal warning about a missing power-menu file and
power-profiles-daemon.
The installer writes `/usr/local/bin/phoebus-session`, which starts waybar,
mako and swaybg and then execs dwl, with a software-render fallback when no
render node is present.
## Official harness on the static binary (Bash migration, Phase 2)
The committed harness (`installer/tests/vm/serial-driver.py` +
`installer/tests/vm/install.sh`) now drives the **static Rust binary** and no
longer runs the Bash installer. Re-verified end to end on a fresh disposable
qcow2 (24 GiB, virtio-blk `/dev/vda`):
```sh
# Build:
cd installer-rs && RUSTFLAGS="-C target-feature=+crt-static" \
cargo build --release -p phoebus-installer --target x86_64-unknown-linux-gnu
# Install in the disposable VM, then boot the disk detached:
python3 installer/tests/vm/serial-driver.py install
python3 installer/tests/vm/serial-driver.py boot
```
Install (`INSTALL-RC=0`):
- The guest fetched `/tmp/phoebus-installer` and printed
`phoebus-installer 0.1.0` (the Rust binary, not Bash).
- Stages completed in proven order: `storage_prepare`, `stage_install`,
`fstab_generate`, `portage_configure`, `packages_base`, `kernel_install`,
`boot_install`, `users_configure`, `services_configure`, `desktop_install`
(none), `finalize`.
- Policy matched Bash: `crypto`/`btrfs-progs` source build, `sudo`/
`networkmanager`/`polkit` binary.
- `NO_SECRET_LEAK`; the run powered off cleanly.
Boot with the ISO detached (`serial-driver.py boot`):
- UEFI firmware ran `\EFI\systemd\systemd-bootx64.efi` (Boot0002 "Linux Boot
Manager"); the BLS "Gentoo Linux" entry booted.
- `systemd-cryptsetup` prompted for the LUKS2 passphrase and unlocked
`phoebus-root (cryptroot)`.
- Login as `phoebus` succeeded.
- `cat /proc/1/comm` -> `systemd`.
- `systemctl is-system-running` -> **running**.
- `systemctl --failed` -> **0 loaded units listed**.
- Networking: `enp0s2 UP 10.0.2.15/24`.
- Mounts: `/` (btrfs `@`), `/home`, `/var`, `/.snapshots`, `/swap`, ESP
`/efi`; `vda2` is `crypto_LUKS`.
This satisfies the Phase 2 exit gate of `docs/BASH-MIGRATION.md`: the only
proven boot path no longer depends on the retired Bash installer.
## Snapshots and safe updates (four-workstream session)
Verified on the installed Btrfs system from the static-binary harness, driving
the new `sys` binary fetched into the guest:
- `sys history` before -> `no transaction history`.
- `sys -y --action '{"action":"create_snapshot","operation":"vm test"}'` ->
`protective snapshot: @-20260920T114917-vm-test`.
- `sys history` after -> `@-20260920T114917-vm-test Committed vm test`.
- `btrfs subvolume list /` shows `@snapshots/@-20260920T114917-vm-test` as a
real subvolume (read-only snapshot).
- `sys -y prepare-updates` fetched distfiles successfully; `--buildpkgonly` was
refused by Portage with `--buildpkgonly requires all dependencies to be
merged`, which the installer now records as `status: Ready` with the reason
(`distfiles fetched; binary prebuild skipped: ...`) instead of a hard failure.
- `sys inspect-updates` reports the state and timestamp.
Unit regression tests cover the failure paths that a VM cannot easily force:
corrupt snapshot history is an error, and a high-risk action aborts when the
protective snapshot fails.
## Desktop session on an installed VM
A full dwl desktop install was built with the static Rust binary
(`DESKTOP_PROFILE=dwl`, `FLATPAK=no`) and booted with the ISO detached:
`INSTALL_RC=0`, all stages, `NO_SECRET_LEAK`, `systemctl is-system-running` ->
`running`, `0 failed units`.
Running the installed `/usr/local/bin/phoebus-session` under the headless
wlroots backend (the serial VM has no logind seat):
- the `wayland-0` socket is created before any helper starts;
- after startup the process tree is `dwl`, `waybar`, `mako`,
`swaybg -c #1e1e2e` and `/usr/libexec/polkit-gnome-authentication-agent-1`
(the polkit agent, which the previous script never started);
- dwl reports the `HEADLESS-1` output;
- after `dwl` exits, **all session processes are reaped**.
Installed and on `PATH`: `dwl`, `waybar`, `mako`, `swaybg`, `kitty`, `wofi`,
`Xwayland`. The polkit agent lives at
`/usr/libexec/polkit-gnome-authentication-agent-1`; the session script's
fallback finds it because it is not on `PATH`.
This is headless verification. A physical VT / real logind seat remains to be
exercised interactively (`docs/DEVELOPMENT.md` -> "Physical VT / real-seat
check").
## Signed binary reuse analysis
The desktop transaction's installer log (`emerge` grouping) shows the policy
already requests signed binaries for the expensive set:
- source group: `sys-fs/cryptsetup sys-fs/btrfs-progs`;
- binary groups: base tools, `gentoo-kernel-bin`/firmware, and the dwl/waybar/
wofi/mako/sway*/gtk-layer-shell/xwayland set.
The desktop install in this session took ~1h49m wall time (12:01–13:50 UTC) on
8 vCPUs. Most of that is Portage falling back to source because the OpenRC/
hardened binhost cannot satisfy the systemd/desktop USE set, even though the
policy requested binaries. Two comparable data points are now recorded: the
base-only install took ~44 min and the desktop install ~1h49m. The code changes
here remove forced flags but cannot raise reuse without a matching binhost, so
no new time saving is claimed.
The remaining source builds are caused by the **official Gentoo binhost being
built for the OpenRC/hardened profile** (`PROFILE: default/linux/amd64/17.0/
hardened`), which cannot satisfy a systemd/desktop USE set. The safe changes
made here are: stop forcing the unused `installkernel uki` flag, remove the
duplicate PipeWire and empty WirePlumber `package.use` lines, and derive the
binhost URI from the selected profile when it is not set. Signature
verification remains enabled. Publishing a matching systemd/desktop binhost is
the remaining step; no time estimate is claimed beyond the recorded ~26 min.
## Debian-Live installer appliance (actual ISO)
An appliance ISO was built **without root** by `installer/live/build.sh`
(Debian-Live 13.7 standard base + `/phoebus` overlay). The full-install
evidence below was recorded with build
`c0105bc9a282fddcbefab42baa577ea99843060f00fa45a75bcbc8358186bfd5`.
A desktop install was re-run and verified with build
`c1846f0d18ed9813e5ef3e699b9110f4fb04ab09a4c5ed800970b1d54971d114`:
- mandatory-AI gate refused the install until `--no-ai` was passed
(`GATE_REFUSED_OK`);
- Nix: `/nix/store/2nhrwv91g6ycpyxvhmvc0xs8p92wp4bk-nix-2.24.9` present,
`nix (Nix) 2.24.9`, `nix-daemon.service` **enabled and active**, and
`nix shell nixpkgs#hello` downloaded from `cache.nixos.org` and printed
`Hello, world!`;
- desktop portals: `xdg-desktop-portal 1.20.4`, `-gtk 1.15.3`, `-wlr 0.7.1`;
- desktop binaries present: `dwl`, `waybar`, `kitty`, `wofi`, `Xwayland`;
- hostname `phoebus`, networking `enp0s2 UP 10.0.2.15/24`, systemd PID 1,
`running`, `0 loaded units listed`.
The current build is
`3ba913ba660fb7436e26e587e9777f1f65ed10dcc9b50d17458ebb3d3a739b5c`: Nix is now
on by default, and `finalize` flushes the target (`sync`) before unmounting.
That build was **not** re-run end to end because the host root filesystem
reached ~98% and the base emerge became pathologically slow. Treat `c1846f0d`
as the verified artifact and `3ba913ba` as the current one.
Firmware boot (UEFI/OVMF, no `-kernel`; the ISO's own grub chain):
- auto-boots the live system on the serial console;
- login `user` / `live` (passwordless sudo);
- the overlaid tools run: `cryptsetup 2.7.5`, `sgdisk 1.0.10`,
`partprobe 3.6`, `mkfs.btrfs 6.14`, `mkfs.fat`.
Full install to a fresh 24 GiB qcow2 driven from the appliance:
- `INSTALL_RC=0`, all stages, `NO_SECRET_LEAK`.
Detached boot of the installed disk (ISO removed):
- systemd-boot -> LUKS2 unlock -> `phoebus` login;
- `systemctl is-system-running` -> `running`; `0 loaded units listed`;
- **Gentoo-only**: `Gentoo Base System release 2.18`, Portage 3.0.82.2,
`no-apt`; systemd PID 1;
- `cryptroot` Btrfs mounted at `/`, `/home`, `/var`, `/.snapshots`, `/swap`,
ESP `/efi`.
This is Bash-free and Gentoo-target-only: the appliance contributes tooling,
never configuration.
### Open items found in this run
- **Resolved:** the earlier `gentoo` hostname anomaly did not recur; the target
boots as `phoebus` with `/etc/hostname` set, and DHCP comes up
(`enp0s2 UP 10.0.2.15/24`) once awaited.
- **Implemented, not re-verified:** `/usr/local/bin/phoebus-session` (and the
exported installer log) were absent after boot in the earlier runs. The
installer wrote and existence-checked them, and the root unmounted cleanly,
but comparison with a boot-time-created file showed the **finalize writes were
not on disk**: Btrfs commits lazily and the harness used `poweroff -f`. The
fix is a `sync` in `finalize` before unmounting (build `3ba913ba`). It has not
been re-run end to end (host disk pressure). Until it is, treat the session
script as **unverified** in the shipped ISO.
## 2026-09-21 appliance baseline (rollback/recovery foundations)
A fresh install of the committed code was run in a disposable qcow2 (24 GiB)
with the rebuilt Debian-Live appliance ISO, then booted with the ISO detached.
This validates the coordinated-snapshot and recovery-entry code on a real Btrfs
system; it does **not** validate a rollback (that gate is still pending).
- Build: static `phoebus-installer` from the tree, appliance ISO
`206551ec21e13753fb97f6d65bfc01e3642b0167908807f64d6ca5a80141adab`.
- Config: `DESKTOP_PROFILE=none`, `KERNEL_PROFILE=safe`, LUKS2 + Btrfs,
`NIX_ENABLED=no`; the host served the stage3
(`sha256 6a3d85ee...f77d44d5`), config and 0600 secret files over HTTP.
- Harness: `installer/tests/vm/appliance-driver.py install` then `boot`
(new in this session).
Install:
- `===NO_SECRET_LEAK===`, `===INSTALL-RC=0===`, clean poweroff
(`reboot: Power down`). Target grew to 6.3 GiB.
Detached boot (no ISO, no `-kernel`):
- UEFI ran systemd-boot; its menu listed `Gentoo Linux`,
**`Phoebus Recovery (known-good snapshot)`** and `Reboot Into Firmware
Interface`; the default booted.
- `systemd-cryptsetup` prompted for the LUKS2 passphrase and unlocked
`phoebus-root (cryptroot)`.
- `cat /proc/1/comm` -> `systemd`; `systemctl is-system-running` -> **running**;
`systemctl --failed` -> **0 loaded units listed**.
- Networking: `enp0s2 UP 10.0.2.15/24`.
- The target's transaction history contains the **initial coordinated snapshot**:
```toml
[[records]]
id = "@-20260921T110110-initial-known-good"
operation = "initial known-good"
source_subvol = "/mnt/phoebus"
snapshot_subvol = "@-20260921T110110-initial-known-good"
var_snapshot = "@var-20260921T110110-initial-known-good"
status = "committed"
```
This confirms `@` and the separately mounted `@var` are captured together and
that the installer writes a snapshot-based recovery entry (not one that
depends on the live root). Reading the recovery entry and
`btrfs subvolume list` as the unprivileged user was refused; the menu entry
above is the evidence that it exists.
Not yet verified for this baseline: `sys undo`, a deliberately broken update, a
detached rollback reboot, and recovery with no network/model (the rollback
acceptance gate).
**2026-09-23 update — rollback acceptance automated, still NOT run.** The
missing driver now exists: `serial-driver.py rollback` (with
`rollback-guest.sh` and `rollback-verify.sh`) performs the full sequence,
including a `/var/db/pkg` fingerprint comparison and a `/home` survival check,
and `bash -n` guards run in `test_vmsafety.py`. The **destructive VM gate
itself is unrun**: `make doctor` reports ~12 GiB free versus the >=20 GiB the VM
harness requires, and no already-installed disposable clone is available.
Booting the recovery entry offline is still a manual step. `sys undo` therefore
remains **unverified**, not passing.
**2026-09-26 update — rollback acceptance PASSED.**
`serial-driver.py rollback` on a fresh disposable overlay
(`rollback-accept-7.qcow2`, backing `fresh-install-1.qcow2`, original
preserved; backing chain validated with `qemu-img info --backing-chain`):
driver `serial-rollback-20260926T030237Z-912f6386` **PASS=11 FAIL=0**,
exit 0, no secret leak in either transcript (both kept:
`serial-phase1.log`, `serial.log` under
`/var/tmp/phoebus-vm/serial-rollback-20260926T030237Z-23d1bc0c/`).
- Phase 1: known-good DB fingerprint `ca398d71…`, `sys snapshot --legacy`
→ `@-20260926T030259-rollback-acceptance`, controlled
`app-misc/hello` install `--legacy --portage` → fingerprint changed to
`fda3780a…` (a no-op break would fail the gate here), root marker written,
`sys undo @-20260926T030259-rollback-acceptance --legacy` → rc=0.
- Phase 2 (detached): DB fingerprint back to `ca398d71…`, `/home` marker
identical, root marker **absent**, `systemctl is-system-running` →
`running`, kernel modules present, `phoebus-recovery.conf` present with
`systemd.unit=rescue.target` and a `@snapshots/` root, rollback status
recorded.
- Harness defects fixed to earn this (all previously allowed vacuous
passes): sudo authentication with unique end markers and a verified
timestamp, both phases as root, explicit `--legacy`, break package
verified absent then verified changed (`nano` was already in the base
image), undo targeting the recorded snapshot id, driver assertions for a
changed break and an absent root marker.
- Still manual: booting the recovery entry itself offline (select it in the
systemd-boot menu with no network).
**2026-09-23 update — NVIDIA gates still NOT run.** The live driver decision,
the hardened payload verifier and the installed lifecycle are implemented and
covered by Rust/fixture tests, but no NVIDIA-accelerated boot has been observed:
no NVIDIA GPU is present and QEMU's virtual GPU cannot establish acceleration.
The full desktop ISO was not assembled (disk), so the ordinary virtual-GPU
`gui`/`install`/`boot` journeys were not re-run against the repaired harness
either.
## Mango + DankMaterialShell verification (IN PROGRESS)
### Live ISO — graphical VM (verified 2026-09-21)
Artifact: `installer/live/build.sh` output
`phoebus-mango-amd64.iso`,
sha256 `c235b873bc04a90c91ff4be361a605907997c509c4021e611645336c23f924d9`
(1.6 GiB). The image was built rootlessly from the signed Debian testing
archive plus pinned Mango 0.17.2 and DankMaterialShell 1.6.2 releases
(`installer/live/releases.sha256`).
Booted in a disposable QEMU/OVMF VM with a **graphical device**
(`-device virtio-gpu-pci`, no `-vga`, serial console still attached) — a real
DRM node and logind seat, not a headless compositor:
- `uname -r` -> `7.1.13+deb14-amd64`; `systemctl is-system-running` -> running;
`/dev/dri/card0` and `/dev/dri/renderD128` present.
- Live autologin ran as an unprivileged `phoebus` user in session `c1` on
`seat0`/tty7; `mango` (compositor), `/usr/bin/dms run --session` and
`qs` (Quickshell) were running under the user manager
(`mango-session.target` and `dms.service` both **active**).
- No dwl helpers (`waybar`/`mako`/`swayidle`/`swaylock`) were started.
- A framebuffer `screendump` showed the DankMaterialShell desktop: the DMS
panel (clock, weather, tray, network/battery/volume/power) and the DMS
first-run "Welcome to DankMaterialShell" window, composited by Mango. This is
the acceptance evidence for "the ISO boots into a usable Mango +
DankMaterialShell live desktop".
- The "Install Phoebus" application entry and `/usr/local/bin/phoebus-install`
are installed by the build; the launcher runs the installer UI and passes no
`--yes`/`--install`/`--no-ai` (unit-tested), so opening it cannot start a
destructive operation automatically. The serial console and a spare VT
remain available as a recovery path.
Fixes found by this test (all committed): the live kernel/initrd must replace
the base ISO's **versioned** `/live/vmlinuz-*` paths; setuid bits must be
restored when packing as an unprivileged user (for `sudo`); `virtio_gpu` must
be autoloaded; greetd's `greeter` account must exist even with an
`initial_session`; greetd must run on VT7 to avoid the `getty@tty1` conflict;
live-config autologin (`noautologin`) must be disabled in favour of greetd; and
the Mango session must not export `WAYLAND_DISPLAY` before the compositor
starts (wlroots would pick its nested backend instead of DRM) and needs the Qt
Wayland shell-integration packages for Quickshell.
### Installed system — not yet run
A default install (`desktop.profile = "mango"`; the default) and a detached
encrypted boot to the greetd login are **not yet verified in a VM**: the Mango
transaction syncs the GURU overlay and compiles Quickshell/DMS/scenefx/Qt,
which was not feasible here. The configuration and login behaviour are covered
by unit tests (`live_iso.rs`, `system::installed_login_*`, `desktop.rs`), and
the plan renders the session. Verify with the harness once a full Mango build
is affordable:
```sh
python3 installer/tests/vm/appliance-driver.py install # DESKTOP_PROFILE=mango
python3 installer/tests/vm/appliance-driver.py boot
```
Remaining graphical checks (installed system and live): panel interactions and
launcher, notifications, polkit prompt, lock/unlock, idle/suspend and resume,
display changes, clipboard/screenshots/terminal/file manager, a Flatpak app
launch + file chooser + screen sharing, logout/reboot, fail-closed shell restart,
and resource metrics.
Known prerequisites/limits: the installed `mango` transaction needs the GURU
overlay (git) and compiles Quickshell/DMS/scenefx; the live build needs the
signed Debian testing archive. A graphical VM must provide a real seat (for
example `-device virtio-gpu-pci` with the serial console still attached);
`-display none` alone is not sufficient evidence of a graphical session.
1. Live ISO boots to Mango + DMS (screenshot/`screendump` evidence, not just
process names); panel, launcher and terminal work; "Install Phoebus" opens
the installer UI without starting a destructive operation.
2. Live text-console recovery (serial and a spare VT) remains accessible.
3. Fresh install with the default profile reaches `systemctl is-system-running
= running` with 0 failed units; `emerge --pretend` shows only the reviewed
`~amd64` atoms.
4. Detached encrypted boot; LUKS unlock; **authenticated** login on a real seat;
Mango is the sole graphical session.
5. `phoebus-mango.desktop` starts the session; exactly one compositor (mango),
one shell (dms), one lock/idle (DMS) and one polkit agent run; no
waybar/mako/swaybg/swayidle/swaylock processes.
6. Panel, launcher (Mod+Space), notifications, tray and the DMS settings panel;
audio/network/Bluetooth controls; lock/unlock; idle/suspend and resume.
7. Display configuration/scaling and a second monitor if the VM supports one.
8. Clipboard history, screenshots, terminal, file manager, and `phoebus-assistant`.
9. An actual Flatpak app: launch, file chooser, and portal screen sharing.
10. Logout reaps session processes; reboot cleanly.
11. Shell-failure handling (kill `dms`, compositor survives, locking still
locks), and default-selection behaviour: omitting the profile selects Mango,
explicit `none` is respected, and obsolete profiles are rejected.
12. Idle memory/CPU and startup metrics recorded.
Until results are recorded here, the default stands by owner decision but no
stability is claimed.
## Graphical installer ISO and typed bridge (2026-09-21)
A dedicated driver, `installer/tests/vm/gui-driver.py`, boots the built live
ISO with a **graphical device** and a serial console, drives the graphical
frontend through the compositor, and captures QEMU framebuffer screenshots.
Artifact: `installer/live/phoebus-installer-amd64.iso`, 1.6 GiB,
`sha256 2a52178c4b4a71bbc5c011b89533988ceb9aa95493160cb93c4335f03d9a1eee`
(Mango 0.17.2 + DankMaterialShell 1.6.2).
What was observed (`gui` mode):
1. The live session autologins into Mango + DMS; the branded
`phoebus-welcome` window opens automatically. The DMS first-run greeter and
changelog dialogs do **not** appear (markers pre-seeded).
2. The welcome shows the wordmark, emblem, wallpaper, ISO version (`0.1.0`) and
readiness (network connected, UEFI, memory, cores), with "Install Phoebus
OS" dominant and the secondary tools plus Advanced present.
3. Launching `phoebus-install` opens the wizard; running it again leaves the
session count at 1 (`DUP n1=1 n2=1`), so duplicate installer sessions are
prevented.
4. Keyboard navigation (Ctrl+Right / Enter / Tab) reaches step 2 (network and
prerequisites, with real UEFI/memory/CPU/network values) and step 3 (AI
model). Provider selection, the secret API-key field and the model field
render; advancing without a connected model is refused with
"Connect and validate an AI model before continuing."
5. A backend probe on the real guest lists `/dev/vda` as unprotected and
selectable, and `/dev/sr0` as the `live_medium` installation medium,
protected (the model tests cover the classification logic).
6. Screenshots: `docs/screenshots/01-welcome.png` … `05-ai-blocked.png`. The
destination-disk, review and confirmation screens
(`06`–`08`) were rendered from the same QML against a dry-run backend
because a real provider call was not available in this environment.
What was **not** possible: a live AI provider call (no key/model in this
environment). The AI step is verified as a blocked prerequisite and in unit
tests with a fixture transport; it is **not** a passed live call.
Real install + detached boot on the same ISO (`install` then `boot` modes):
- Guest install used the same deterministic backend as the GUI
(`phoebus-installer --install --yes --no-ai`): `INSTALL-RC=0`,
`installation complete`, `NO_SECRET_LEAK`.
- Detached boot (no ISO, `-boot order=c`): systemd PID 1, `running`, **0
failed units**, `systemd-boot` menu shows **Gentoo Linux first**, LUKS2
unlocked, `/dev/vda1 -> /efi`, `cryptroot` Btrfs with
`/ /home /var /.snapshots /swap`, network up.
Bug found and fixed during this work: the snapshot recovery BLS entry shared
the kernel's `sort-key`, and the specification's final filename tie-break
(decreasing) made it first when the random machine-id sorted low; with
`default *` the installed system booted `systemd.unit=rescue.target`. Fixed by
pinning `loader.conf` to the newest real kernel entry and giving the recovery
entry `sort-key zzzz-phoebus-recovery`; unit tests pin both.
## 2026-09-22 — automated graphical journey (gui-driver) and hardened harness
The graphical live ISO was rebuilt and driven automatically by
`installer/tests/vm/gui-driver.py gui` in a graphical, seat-providing
QEMU/OVMF VM with a disposable qcow2 disk directory and a per-run OVMF variable
file.
- Source revision: `7e9fa9c` plus uncommitted work; build id recorded in
`/etc/phoebus/build` (`7e9fa9c+dirty…`).
- Static installer: `sha256 2f9cd4b3…` (4,716,544 bytes, static-pie).
- ISO: `installer/live/phoebus-installer-amd64.iso`,
`sha256 428dbd75…`, 1.6 GiB, built with
`PHOEBUS_REUSE_ROOTFS=1` (Mango 0.17.2 + DankMaterialShell 1.6.2).
- Result: **23/23 checks PASS**, exit status 0
(`/var/tmp/phoebus-vm/gui-gui-20260922T025537Z-69f754cf/`).
Asserted: welcome autostart; live/branding payload markers from command
output (not the echoed command); the live medium protected; "Install Phoebus
OS" launched the installer from a Return keypress; duplicate sessions
prevented (`n1=n2=1`); the frontend reported no bridge errors; keyboard
navigation reached the network and AI steps; advancing without a connected
model was refused and started no installation.
- Screenshots at every page (welcome focus, language, network, AI model, AI
blocked, final state) plus a `screenshot-manifest.json`; `report.py` builds a
browsable `report.html`. All artifacts stay under `/var/tmp/phoebus-vm/`
(outside git).
Harness safety verified in the same work: the drivers reject `/dev/*`, block
and character devices, raw images, symlinks escaping the disposable directory,
and outside backing chains, with a nonzero exit (`make test-vm`, 16 unit
tests). Screenshots use structured QMP and argument arrays; there is no shell
interpolation.
A host-local OpenCode transport test also ran: `phoebus-installer
--opencode-probe` returned `ready` against `opencode 1.18.32` through a
forwarded second loopback endpoint, and `authentication_required` (exit 1)
against a deliberately unauthenticated LAN bind.
## 2026-09-22 — graphical journey polish and the `gui-install` mode
The graphical installer was reworked into a fixed first-time-user journey and
re-verified on the final ISO.
- Source revision: `34c12d3` plus the session work; ISO
`installer/live/phoebus-installer-amd64.iso`, `sha256 0c386a1b…`, 1.6 GiB,
built with `PHOEBUS_REUSE_ROOTFS=1` (Mango 0.17.2 + DankMaterialShell 1.6.2).
- `gui` **28/28 PASS** (`/var/tmp/phoebus-vm/gui-gui-20260922T191954Z-876d5e91/`):
welcome autostart, launcher from the keyboard, duplicate prevention, the live
medium protected, keyboard suggestion selection (Down then Enter stores the
identifier), the network and AI steps, the AI gate refused without a model
with an inline error, and no bridge errors.
- `opencode` **18/18 PASS**
(`/var/tmp/phoebus-vm/gui-opencode-20260922T192457Z-01714c92/`): OpenCode
selected and driven through the UI, a real model response (Test connection),
a validated proposal (4 changes applied), and a regression asserting that
editing the model flips the validated state off.
- `gui-install` **24 PASS / 1 FAIL**
(`/var/tmp/phoebus-vm/gui-gui-install-20260922T151232Z-0ef19969/`): every
journey assertion passed — OpenCode connect/review/accept,
`back-nav-preserves-ai`, `readiness-refresh-preserves-wizard`,
`disk-survives-backtrack`, disk selection, account and encryption, ERASE and
`gui-install-erase-confirmed` — and the backend then reported the real stage
sequence (`storage_prepare`, `stage_install`, `fstab_generate`,
`portage_configure`, `packages_base`, `kernel_install`) with elapsed time and
no fabricated percentage. The single FAIL is `gui-install-completes`: after
four hours the install was still in `kernel_install`, because the
base-package stage took ~116 min and the AI-selected optimized kernel
compiled for more than 100 min on this host, so the harness timeout fired.
**The completed install + detached boot gate is BLOCKED on host build time**;
the deterministic noninteractive install + detached boot (above) remains the
proven boot path. No code defect is implicated.
- The new `PHOEBUS_GUI_CONFIG` / `/etc/phoebus/install.conf` hook lets a medium
preseed the wizard defaults. It changes defaults only: validation, the plan,
the AI gate and the ERASE authorization are unchanged.
## 2026-09-24 — End-to-end graphical install and detached first-boot verification
The graphical installer journey was completed 100% through all 12 backend
orchestration stages and the Mango desktop package set, followed by a clean
detached boot verification with the live ISO removed.
- **Fresh Live ISO:** `installer/live/phoebus-installer-amd64.iso` rebuilt with
`scripts/ensure-live-iso.sh` (SHA-256 `4b83df6b531bc38ffc38e3aee21a3519f2d414a8af883f36ef82f52c4e0458d6`, 1.6 GiB).
- **Kernel profile optimization for installation:** Default kernel profile
configured to `KernelProfile::Safe` (`sys-kernel/gentoo-kernel-bin`), eliminating
the 100+ minute kernel compile from source that previously exceeded harness timeouts.
- **Graphical Installation (`gui-install`, run `gui-gui-install-20260924T144711Z-feab515b`):**
- All 6 wizard steps driven via virtual input: OpenCode local connection,
hardware-aware recommendations, disk selection (`/dev/vda`), account (`phoebus`),
LUKS2 passphrase, ERASE confirmation.
- Executed all 12 backend stages in sequence:
1. `storage_prepare` (GPT, ESP, LUKS2, Btrfs subvolumes `@`, `@home`, `@var`, `@snapshots`, `@swap`)
2. `stage_install` (pinned Gentoo systemd stage3)
3. `fstab_generate` (fstab and crypttab)
4. `portage_configure` (binary package binhost, scoped keywords and USE)
5. `packages_base` (essential system packages)
6. `kernel_install` (`sys-kernel/gentoo-kernel-bin` + dracut initramfs)
7. `nvidia_drivers` (hardware check / skipped for virtio-gpu)
8. `boot_install` (systemd-boot + BLS entry)
9. `users_configure` (user `phoebus`, wheel/sudo group, shadow hash)
10. `services_configure` (systemd essential services, locale-gen)
11. `desktop_install` (all 174 packages emerged: Mango, DankMaterialShell, Quickshell, greetd, tuigreet, Wayland libraries, Flatpak)
12. `finalize` (installation metadata `/etc/phoebus/install.toml`, clean unmount)
- `gui-install-completes` passed (`07-gi-07-complete.png` captured):
"Phoebus OS is installed. Your system is ready. It can boot, run and recover without a network or a model."
- **Detached First Boot (`gui-boot`, run `gui-boot-20260924T183545Z-4b5de282`):**
- Booted with `iso=False`, `-boot order=c` pointing strictly at `/var/tmp/phoebus-vm/disks/test-disk.qcow2`.
- Firmware booted systemd-boot -> kernel loaded -> LUKS2 prompted and unlocked with test passphrase.
- All Btrfs subvolumes mounted correctly (`@` on `/`, `@home` on `/home`, `@var` on `/var`, `@snapshots` on `/.snapshots`, `@swap` on `/swap`).
- Serial console login authenticated successfully.
- `sys` deterministic CLI verified installed and operational: `/usr/local/bin/sys history` reported `no transactions recorded`.
- Systemd reached `running` (0 failed units).
- Greeter fix: Diagnosed and fixed a TOML multiline syntax issue in `/etc/greetd/config.toml` (command string formatted on a single line). `greetd.service` started `tuigreet` cleanly, verified by QMP screendump (`01-boot-desktop.png` displaying "Authenticate into phoebus-vm").
- Clean guest poweroff executed and verified (`sync; systemctl poweroff`).
- Results: 6/6 PASS (`qemu-spawned`, `detached-boot-login`, `boot-systemd-running`, `boot-sys-installed`, `screenshot-01-boot-desktop`, `screenshot-02-final-state`).
## 2026-09-25 — Snap first-use acceptance on an installed disposable VM
- Target: `snap-acceptance.qcow2`, a disposable overlay of the installed
`fresh-install-1.qcow2` image in the VM disk directory. QEMU/OVMF booted the
installed system; no physical disk was attached or modified.
- The first attempt found a real Gentoo dependency requirement:
`sys-fs/squashfs-tools[lzma,lzo]`. The guest test applied only those scoped
USE flags. Fresh Phoebus installs now generate the same package-specific
setting.
- The next attempt exposed a Portage resolver bug: the candidate was the
first dependency (`dev-lang/go-bootstrap`) instead of the requested
`app-containers/snapd`. The resolver now selects the requested atom, and the
approved plan showed six new source packages including `snapd-2.63`.
- Final run `serial-boot-20260926T001830Z-107996de` reported `PASS=10 FAIL=0`:
installed snapd through the reviewed `sys` Portage path, enabled and started
`snapd.socket`, installed Snap Store `hello-world` 6.4 revision 29 through
`sys`, ran `sys snap refresh hello-world --legacy`, then removed it and
verified absence. The daemon reported `partial` confinement and the install
plan reported `partial` as well.
- The refresh command succeeded with no newer revision. A revision-changing
refresh and AppArmor-backed strict confinement remain unverified.
- Final static `sys` SHA-256 `df608076aaf54290bfe0d4cdad9bcbca50064bf53ced62c44cf0e765d2a2f73f`
was retested on the fresh `snap-acceptance-final.qcow2` overlay. The first
run (`serial-boot-20260926T005327Z-c024cc43`) passed 11/11 checks including
a clean guest poweroff. A clean reboot and repeat
(`serial-boot-20260926T010605Z-e6193c3a`) also passed 11/11, with another
`hello-world` install, refresh, removal and clean poweroff.
## 2026-09-26 — `sys` acceptance re-run against the refactored static binary
- Target: `sys-acceptance-refactor.qcow2`, a disposable overlay (absolute
backing path, satisfying the harness disk-safety gate) of the pristine
installed `fresh-install-1.qcow2` image. QEMU/OVMF booted the installed
system detached; no physical disk was attached or modified.
- Purpose: prove the quality-pass refactor (boxed `ExecError`, Clippy fixes,
`service/` and `gui_bridge/` module splits) changed no behavior. The guest
fetched the new static-pie `sys` (`SHA-256
70f0a72b9ca4534250c2ee27417efffc0681da88ba4d2633da932813b0c0e044`,
built from the post-split tree) to `/tmp/sys` and ran the current
`installer/tests/vm/sys-acceptance.sh` with `SYS_BIN=/tmp/sys` as the
non-root `phoebus` user with no mutations — so the matrix exercised the new
binary, not the image's provisioned one.
- Result `serial-boot-20260926T014746Z-3089a820`: guest
`===SYS-ACC-END=== SUMMARY PASS=38 FAIL=0` (the full 38-case matrix:
version/history/undo, failure cases, offline/model-free behavior, source
selection, AI-first read-only workflow, catalogue including the
`rdeps-legacy` regression, and non-root privilege refusals), driver
`PASS=41 FAIL=0` (38 cases + login + `systemd is-system-running` +
disposable-disk), exit 0.
- Not covered by this run: the mutate group (`SYS_ACCEPTANCE_MUTATE=1`),
recovery boot, strict Snap confinement, and a revision-changing Snap update.
## 2026-09-26 — Security and release-readiness follow-up (COMPLETED)
- **Stage3 trust anchor is fixed, not VM-proven.** `release_key_path()` no
longer reads `PHOEBUS_GENTOO_RELEASE_KEY`: production always uses the
provisioned `/usr/share/openpgp-keys/gentoo-release.asc`, and test-only
injection goes through the explicit `verify_digests_with_key` helper (19
stage3 tests pass). Host-level mechanism check (not VM evidence): the
committed parser fixture matches the real current DIGESTS byte-for-byte,
and the real DIGESTS for
`stage3-amd64-systemd-20260920T170055Z.tar.xz` verifies with a Good
signature from Gentoo's Automated Weekly Release Key.
**Blocked VM proof:** rebuild the live image (ships gnupg + the
fingerprint-checked key) and run the default unpinned stage3 install on a
disposable qcow2, verifying the real signed DIGESTS and downloaded hash
plus tampered signature/digest/archive refusal. Prerequisite: a live-ISO
rebuild (multi-GB; host has ~26 GiB free, shared with VM disks) and a full
install cycle; no ISO was rebuilt in this session.
- **Rollback gate strengthened, rerun in progress.** `rollback-guest.sh`
creation of `/etc/phoebus/rollback-broken.txt` is now mandatory
(`mkdir -p`, write, `===ROLLBACK-MARKER present===`, exit 1 if absent;
the old `|| true` could have hidden a failed break). The absent-afterward
assertion stays where it belongs: after the detached reboot
(`rollback-verify.sh` + driver `rollback-root-restored`), because `sys
undo` activates the restored generation on the next boot. The unused
`SERVE` variable was removed (`shellcheck` clean).
- **Strengthened rerun PASSED** (`serial-rollback-20260926T034218Z-7fab51ef`,
PASS=11 FAIL=0, exit 0, overlay `rollback-accept-8.qcow2` on
`fresh-install-1.qcow2`, logs under
`/var/tmp/phoebus-vm/serial-rollback-20260926T034218Z-89357549/`): known-good
DB `ca398d71…` → mandatory `===ROLLBACK-MARKER present===` → controlled
`app-misc/hello` install (DB `fda3780a…`, changed as required) → `sys undo`
rc=0 → detached reboot (DB restored to `ca398d71…`, root marker **absent**,
`/home` survives, systemd `running`, modules present, recovery entry
present, status recorded). No secret leak in either transcript
(`SECRET_LEAK_FOUND` count 0 in both).
- **Recovery-entry acceptance is a distinct manual gate** (separate from
normal `sys undo`). Procedure: on a disposable installed disk, select
"Phoebus Recovery (known-good snapshot)" in the systemd-boot menu with the
network unavailable and no model configured; verify the booted snapshot
identity (`btrfs subvolume show /`, recovery entry options), reach a
login, and run basic deterministic `sys` operation (`sys history`,
`sys rollback status --legacy`, one read-only query). Pass criteria: the
selected snapshot boots, systemd runs, `sys --legacy` works, and `sys ask`
without a model fails truthfully with the exact `--legacy` fallback
offered. Status: NOT RUN — the harness records the entry but does not
select it (interactive BLS-menu selection is not automated).
- **Release gates are executable and truthful.** `candidate_gates()` no
longer names `serial-driver.py checks` / `desktop` modes no harness
accepts: `no-model-recovery` uses the real `boot` mode (with `BOOT_CHECKS`
running `sys --legacy`), `desktop-session` uses the real graphical
harness (`gui-driver.py boot`). `run_gates(--execute)` runs each gate's
argv in its own process and reports Passed/Failed/NotRun from evidence
(missing harness/firmware/seat/credential → NotRun; nonzero exit →
Failed; invalid command → Failed). A `gui-driver.py do_boot` SyntaxError
(unclosed `assert_true(`) found by the new gate-mode test is fixed.
Evidence: `phoebus-release gates --execute` on an unconfigured host → 0
passed / 0 failed / 6 NotRun with exact missing prerequisites, and
`publish` refuses against that report (exit 1).
- **Seat desktop, live AI, Snap revision change, NVIDIA:** unchanged and
explicitly open. Installed Mango+DMS under a seat needs a
seat-providing VM or physical VT; a live AI call needs a responsive local
model or an authorized provider credential (fixture routing/failure tests
pass; the gate is NotRun with the exact missing credential); a
revision-changing Snap refresh needs a Store snap with a newer revision
plus an AppArmor-ready guest kernel for strict confinement (existing
partial-confinement/no-op-refresh evidence stays labeled as such); NVIDIA
needs a GPU or authorized passthrough (see below).
## 2026-09-26 — Live ISO rebuild from HEAD + unpinned install proof (COMPLETED)
- **Rebuilt ISO.** `installer/live/phoebus-installer-amd64-20260926.iso`
(1.6 GiB, SHA-256
`250dbe9ad780408a40b7b7f0757791fadcc1e4309b932e98f6059c7baf93e2e5`,
build id `20260926T041500Z`) built from HEAD `11c0ee0` with
`PHOEBUS_REUSE_ROOTFS=1` (cached `livefs.tar` reused for the Debian
rootfs; payload, binaries, key and branding re-installed fresh — reuse
does not carry stale binaries, UI or keys). The Sep-24
(`827bf843…`) ISO predates the stage3 fail-closed + trust-anchor +
boot-security commits and is kept alongside.
- **Content inspection (not timestamps).** The finished squashfs was
listed and selectively extracted: embedded
`/usr/local/bin/phoebus-installer` (`25741706…`) and `/usr/local/bin/sys`
(`3b331643…`) hash-match the fingerprint-fresh static build;
`/usr/share/openpgp-keys/gentoo-release.asc` (57,698 bytes) shows both
published fingerprints (`13EB…2910` weekly, `D99E…2058` signing) under
`gpg --show-keys`; `/usr/bin/gpg`, `/usr/bin/mango`, `/usr/bin/dms`,
the welcome/installer `shell.qml` configs, the install desktop entry and
`/etc/phoebus/{version,build,live}` (`0.1.0`,
`PHOEBUS_BUILD_ID=20260926T041500Z`) are all present.
- **Unpinned signed-DIGESTS install PASSED.**
`gui-driver.py install` run `gui-install-20260926T040728Z-f3ae68f4`
booted the new ISO on disposable
`disks/unpinned-accept-1.qcow2` (30 GiB) and installed with the
ISO-embedded fail-closed binary and a served `test.conf` with **no**
`STAGE3_URL`/`STAGE3_SHA256` (empty URL resolves latest-stage3 from
Gentoo autobuilds; signed DIGESTS verified against the fixed provisioned
key): `INSTALL-RC=0`, `NO_SECRET_LEAK`, `installation complete` after
~165 min (full mango desktop, 174 packages). Detached boot (ISO
detached): LUKS unlock, login, `systemctl is-system-running = running`,
`0 loaded units listed`, networking, all Btrfs subvolumes mounted, `sys
history --legacy` showing `@-20260926T064646-initial-known-good`. The
harness `--no-ai` flag is **test-only** evidence and is not the normal
user path (the wizard requires a connected model; the interactive VM is
left for the owner to drive by hand).
- **Harness fix (real defect, fixed):** `gui-driver.py do_boot` asserted
bare `sys history` output, but bare `sys history` is AI-first and
truthfully refuses without a model — the check now uses
`sys history --legacy`. The product was correct (`/usr/local/bin/sys`
present and working); the gate was stale. Re-run detached boots pass
`boot-sys-installed`. (The driver's `detached-boot-login` flag is
timing-sensitive to PAM OSC noise on the serial line even when the login
succeeds and checks run; login itself works — 1/4 flag-pass with 4/4
substantive checks green.)
- **Tamper refusals PASSED** (new ISO, fail-closed binary, disposable 4
GiB targets, `INSTALL-RC=1` + `NO_SECRET_LEAK` both): placeholder pin
refused before any network use (zero stage3 fetches on the serve log);
mismatched archive hash refused after download with no extraction.
Method note: `appliance-guest-gui.sh` ignores the driver's `SERVE_URL`
env and always uses the default port 8000 — tamper variants were served
by swapping the port-8000 `test.conf` (the main run had long since
fetched its copy). A tampered-*signature* VM case is infeasible without
a network MITM; the bad-signature branch is unit-tested.
- **Recovery-entry boot PASSED** (distinct manual gate, now verified).
Visible QEMU on the installed disk with restricted networking
(`-netdev user,restrict=y`) and no model: keypresses provably freeze the
3 s systemd-boot timeout (screenshot), `Phoebus Recovery (known-good
snapshot)` selected with Down+Enter (screenshot), kernel cmdline
`rootflags=subvol=@snapshots/@-20260926T064646-initial-known-good` with
`ro` + `systemd.unit=rescue.target`, findmnt confirms the read-only
snapshot root, serial login as `phoebus` works, `sys history` /
`rollback status --legacy` work, `sys ask` fails truthfully offering the
exact `--legacy` command, systemd `degraded` (rescue target, as
designed). Method notes: video-console (tty1) keystrokes never reach the
dracut LUKS prompt (`/dev/console` is serial here) — the passphrase must
be written to the serial pty with ONLCR/ICRNL neutralized, otherwise a
`\r` is appended and a correct passphrase is rejected; a `down`+`ret`
pair landing at LUKS burns attempts as `**` (escape remnant) and three
failures drop to the dracut emergency shell (root, no cryptsetup binary
present). Transcript of the proof session kept under
`/var/tmp/phoebus-recovery/rec-entry-proof.log`.
- **Seat desktop PARTIAL (blocker found).** Seat VM (`virtio-gpu-gl-pci`,
virgl present, overlay of the installed disk): logind seat0 with DRM
card0 (`[MASTER]`), keyboard, mouse; `/dev/dri/card0` (no render node);
tuigreet active with `--cmd /usr/local/bin/phoebus-session-mango`;
tuigreet PAM authentication of the correct password verified in the
journal (earlier "Authentication failed" readings were QMP-typing
focus races; `-` keys arrive correctly, proven by visible echo).
The Mango compositor exits in <1s after every successful authentication:
`libseat: No backend was able to open a seat`, `Failed to start a DRM
session` (captured by running the session script from a shell). No seatd
in the image and the logind path does not succeed under the tuigreet
session — this is the precise blocker for kitty/DMS/lock/idle/polkit/
logout/reboot. Session transcript and journal excerpts kept under
`/var/tmp/phoebus-seat/`.
- **Interactive VM (open).** The new ISO boots a KVM/GTK VM with a fresh
unique disk (`phoebus-interactive-20260926.qcow2`) and private OVMF vars
in `/var/tmp/phoebus-interactive/`; screenshots show the Mango top bar
and the branded welcome (`Install Phoebus OS`, Live medium 0.1.0,
network/UEFI/RAM/CPU chips). Left running for the owner; installer
untouched, no passwords or model credentials entered.
- Evidence disk `disks/unpinned-accept-1.qcow2` KEPT; scratch overlays
(`unpinned-tamper-*`, `nix-gc-*`, `seat-accept-1`) removed after their
run logs were kept. Host free space ~16 GiB at the end (watched, nothing
deleted without identification).
- **Rebuilt ISO.** `installer/live/phoebus-installer-amd64-20260926.iso`
(1.6 GiB, SHA-256
`250dbe9ad780408a40b7b7f0757791fadcc1e4309b932e98f6059c7baf93e2e5`,
build id `20260926T041500Z`) built from HEAD `11c0ee0` with
`PHOEBUS_REUSE_ROOTFS=1` (cached `livefs.tar` reused for the Debian
rootfs; payload, binaries, key and branding re-installed fresh — reuse
does not carry stale binaries, UI or keys). The Sep-24
(`827bf843…`) ISO predates the stage3 fail-closed + trust-anchor +
boot-security commits and is kept alongside.
- **Content inspection (not timestamps).** The finished squashfs was
listed and selectively extracted: embedded
`/usr/local/bin/phoebus-installer` (`25741706…`) and `/usr/local/bin/sys`
(`3b331643…`) hash-match the fingerprint-fresh static build;
`/usr/share/openpgp-keys/gentoo-release.asc` (57,698 bytes) shows both
published fingerprints (`13EB…2910` weekly, `D99E…2058` signing) under
`gpg --show-keys`; `/usr/bin/gpg`, `/usr/bin/mango`, `/usr/bin/dms`,
the welcome/installer `shell.qml` configs, the install desktop entry and
`/etc/phoebus/{version,build,live}` (`0.1.0`,
`PHOEBUS_BUILD_ID=20260926T041500Z`) are all present.
- **Unpinned signed-DIGESTS install (running).**
`gui-driver.py install` run `gui-install-20260926T040728Z-f3ae68f4`
boots the new ISO on disposable
`disks/unpinned-accept-1.qcow2` (30 GiB) and installs with the
ISO-embedded fail-closed binary and a served `test.conf` with **no**
`STAGE3_URL`/`STAGE3_SHA256` (empty URL resolves latest-stage3 from
Gentoo autobuilds; signed DIGESTS verified against the fixed provisioned
key). Guest progress: live login passed, partitioning/LUKS2/Btrfs
complete, install stages running. The harness `--no-ai` flag is
**test-only** evidence and is not the normal user path (the wizard
requires a connected model; the interactive VM below is left for the
owner to drive by hand).
- **Interactive VM (open).** The new ISO boots a KVM/GTK VM with a fresh
unique disk (`phoebus-interactive-20260926.qcow2`) and private OVMF vars
in `/var/tmp/phoebus-interactive/`; screenshot shows the Mango top bar
and the branded welcome ("Install Phoebus OS", Live medium 0.1.0).
- Still open after this run: installed compositor session (libseat
blocker above), live AI provider call (no credential/model available),
revision-changing Snap refresh + strict confinement (no Store revision /
AppArmor guest), per-user Nix profile enumeration (new coverage gap),
NVIDIA hardware. Secure Boot/UKI stays a separate opt-in milestone.
## Known remaining work
- The graphical session was verified with the **headless** wlroots backend
(no logind seat in the serial harness). A physical VT / seat path is
architected (`phoebus-session`) but was not exercised on real hardware.
- NVIDIA proprietary support is implemented (hardware-driven branch/flavor
policy, scoped Portage configuration, a build-time-verified live payload, and
kernel/module/initramfs update coordination) with Rust tests. A minimal live
rootfs was **built and its NVIDIA payload verified** this session (driver
`550.163.01`, kernel `6.12.107+deb13-amd64`, modules `nvidia-current*`). It is
**NOT VERIFIED on hardware**: no NVIDIA GPU is present on this host, QEMU's
virtual GPU cannot establish NVIDIA acceleration, and no passthrough setup is
available or authorized. The full desktop ISO was not assembled here (the
host has ~10 GiB free; the build wants more).
The live kernel is pinned to Debian **stable** because Debian testing's
current kernel (7.2) removed the `strncpy` declaration the 550 driver uses, so
`nvidia-kernel-dkms` fails to build against it. The bundled branch covers
Maxwell through Ada/Hopper; Blackwell falls back to nouveau on the live
medium. The installed Gentoo system uses the current branch with open kernel
modules on Turing and newer.
**Unblock (hardware):** boot the built ISO on a machine with an NVIDIA GPU
(or an authorized passthrough VM) and check: `systemctl status
phoebus-nvidia.service`, `nvidia-smi`, `cat
/sys/module/nvidia_drm/parameters/modeset` == `Y`, an accelerated
Mango/DMS session, XWayland/Flatpak rendering, monitor handling and
suspend/resume. Then boot the installed Gentoo system and confirm the same,
plus that a kernel update rebuilds the modules via `@module-rebuild` and
regenerates the initramfs. **Unblock (build):** free sufficient disk space and
run `installer/live/build.sh` (it fails, by design, if the NVIDIA payload is
missing or mismatched), then the `gui` harness mode to confirm ordinary
virtual-GPU boot is unchanged.
- UKI is deferred (BLS is used).
- The installer rebuilds `sys-apps/systemd` from source because
boot/policykit/cryptsetup USE differ from the binhost. Shipping a matching
systemd binpkg is a future optimization.
- Desktop USE still does not match the signed binhost, so the desktop
transaction compiles a large amount from source (~26 minutes for the desktop
package set on 8 vCPUs in the test VM).
- **The complete graphical install and detached boot gate is now VERIFIED.**
The `gui-install` harness drives the real controls through the complete wizard,
performs the real 12-stage install on a disposable qcow2, and boots detached with
systemd running, LUKS2 unlocked, `sys` available, and `tuigreet` active.
- **A separate physical computer for the OpenCode transport test is NOT
available**; that check is BLOCKED. The forwarded-hop and
unauthenticated-LAN-refusal tests are host-local.
## Run 15 — pristine reinstall with all fixes + full boot acceptance (2026-09-28)
Disposable 80 GiB qcow2 `phoebus-headless-80g.qcow2`, 8 vCPU/8 GiB, headless
(`-display none`, serial stdio). Static binaries from HEAD
(`phoebus-installer` `183ec40b…`, `sys` `d5f7a46f…`).
- Install: `serial-install-20260928T132731Z-b5f8629a`,
`===INSTALL-RC=0===`, `NO_SECRET_LEAK`, driver `PASS=4`. All 12 stages,
347 desktop packages, finalize VDB check green.
- Detached boot: `serial-boot-20260928T183131Z-953e4677` (ISO detached,
`-boot order=c`), driver `PASS=4` (disk, login, systemd-running,
clean-poweroff). `cat /proc/1/comm` → `systemd`,
`systemctl is-system-running` → `running`, 0 failed units,
`seatd.service` active, `/run/seatd.sock` present (group `seat`),
`phoebus` in groups incl. `seat`.
- `sys-acceptance.sh`: `SUMMARY PASS=40 FAIL=0` (run detached from the
console tty via `setsid` so a polkit appstream-refresh prompt cannot
block it; direct `flatpak list` on a console still asks once, as designed).
- Nix per-user enumeration (the `active_profile_links` fix): with the
installer-side permission normalize in place, `/etc/nix/` is
`drwxr-xr-x`, `nix.conf`/`nix.sh` are `-rw-r--r--`, `nix profile install
nixpkgs#hello --profile ~/.local/state/nix/profiles/profile` succeeds
(hello 2.12.3 from cache.nixos.org), and `sys list --nix --legacy`
reports `legacyPackages.x86_64-linux.hello 2.12.3 [per-user:phoebus]
(installed)` with 824 store paths noted separately.
- Live AI (`AI_LIVE_RC=0`): `sys ask "list installed packages"
--ai-backend opencode --opencode-server http://127.0.0.1:4096
--opencode-provider opencode-go --opencode-model
muse-spark-1.3-contributor` reached the host `opencode serve` through a
guest loopback tunnel (`fwd.py`, required by the transport's
plain-http-to-loopback-only policy) and returned a valid Assistant Plan
with approval digest `ea23c64b…` (907 Portage results from evidence).
Note: the `show system status` phrasing elicits schema-noncompliant JSON
from this model (fail-closed rejection works); the inventory phrasing is
the verified gate. The tunnel needed `settimeout(None)`: `create_connection`
leaves its timeout on the socket and killed replies slower than 30 s
(curl exit 52).
## 2026-10-08: interrupted application build recovery and updated local ISO
The generic GUI run `gui-desktop-20261008T074853Z-fd35d9f2` was externally
interrupted during Firefox source compilation at 08:43 UTC. Its individual
boot checks passed, but it has no final clean-poweroff manifest. Recovery run
`gui-desktop-20261008T141501Z-d370eb0e` passed generic boot/systemd and reached
Mango/DMS. Original firefox-bin remained installed. Controller/HTTP and guest
sequential conversion worker now use persistent systemd transient services.
Firefox work recovery uses a temporary, scoped Portage keepwork environment;
remaining conversions/native GUI/tuned acceptance are still pending.
Latest code passed fmt, Clippy, all 961 workspace tests and static/dynamic
release builds. Local ISO SHA-256:
`0e2329b24b94f7db356525b672bbaecbcb03d5c2e41750448e1714c7f0b0988b`.
Extracted sys/installer/QML bytes match the latest build/source files, and
GTK3 Mango session setting plus BIOS/UEFI boot entries pass inspection.
No-target-disk UEFI run `postinstall-iso-smoke-20261008T142940Z` passed
`graphical.target` and live serial login after Return on the normal live entry.
Earlier smoke retries exited during firmware startup with QEMU -no-reboot;
allowing startup resets passed. This is not automatic-countdown verification.
Verified ISO packing scratch was cleared; cached rootfs tar retained.
### Firefox source-codec failure and retest
Recovery source build failed after 7852 seconds at libgkcodecs linking with
undefined libaom version-script exports. Original firefox-bin was retained,
and the sequential worker stopped before Thunderbird. Failed log is kept in
`/var/tmp/source-build-firefox-system-av1-failed.log` inside the VM. This matches
[Mozilla bug 2063808](https://bugzilla.mozilla.org/show_bug.cgi?id=2063808).
Scoped source USE entries `<www-client/firefox-156 -system-av1` and
`<mail-client/thunderbird-156 -system-av1` select Mozilla's supported bundled
codec on affected branches. Actual Portage pretend confirms -system-av1 on
Firefox 153.4.0. Guest retest unit `phoebus-postinstall-apps-codec-retest` is
running; its complete conversion/native launch results are still pending.
The first codec retry failed in 146 seconds because the temporary keepwork
recovery kept the old .configured state and --with-system-av1 mozconfig. After
capturing that stale setting, removed the test-only recovery env/map, ran
Portage ebuild clean for exactly Firefox 153.4.0, confirmed WORK-CLEAN and
recovery env removed, and restarted the sequence in
`phoebus-postinstall-apps-codec-clean`. This fresh retest is still pending.
### Updated codec-fix ISO — 2026-10-08
Local candidate `installer/live/phoebus-installer-amd64-postinstall.iso`,
SHA-256 `b0e63a8d3ed9b82fab08a561522cb9b7b36ddad24ef61bb8493641cf5523a92a`,
contains the exact statically built installer/sys, current assistant QML and
GTK3 session setting (host evidence
`/var/tmp/phoebus-postinstall-codec-iso-embedded.txt`). Fresh OVMF variables,
KVM/q35, 3 GiB/2 CPUs, virtio graphics/RNG and a **read-only virtio-SCSI CD with no
target disk** booted twice: `postinstall-iso-smoke-20261008T173910Z` and
`postinstall-iso-smoke-20261008T174001Z`. Both selected the normal menu entry
with Return and reached `graphical.target` plus the live serial login. Their
`result.txt`, `serial.log`, `action.txt` and menu screenshot retain evidence.
Earlier AHCI CD attempts failed in firmware loading grubx64.efi with Device
Error or stalled before the menu (171117Z, 171724Z, 173020Z). The older verified
ISO control also stalled before the menu (172322Z). Both images' EFI image,
bootx64.efi and grubx64.efi are byte-identical. The successful SCSI tests verify
this candidate's UEFI boot; they do not establish the cause of the AHCI failure
or verify physical firmware/controllers.
### Firefox clean codec retest and native playback — 2026-10-08
The clean retest finished with `CASE source-build-firefox PASS`, rc 0,
`BUILD_SECONDS=9674` in generic GUI run
`gui-desktop-20261008T141501Z-d370eb0e`. Source Firefox 153.4.0 ESR is installed,
firefox-bin removed, profile/settings hashes and explicit desktop defaults
preserved, and generic fallback images unchanged. Scratch peaked at 8.3 GiB;
the staged application was 333.5 MiB. Both bundled libaom and dav1d compiled.
`gtk-launch firefox-bin.desktop` (preserved native compatibility entry) opened
`/usr/lib64/firefox/firefox` as user phoebus. Screenshot
`shots/05-05-native-firefox-js-av1-rendered.png` shows the local fixture's
JavaScript PASS and bundled AV1 decode PASS with the playing AV1 clip. A later
quit-dialog screenshot shows a changed video frame/time. No browser sandbox
flags were disabled. Closed normally through Ctrl+Q and its confirmation.
User-data preservation is verified before launching; this does not claim
migration of a newer binary profile into the older ESR profile. Thunderbird
started automatically after all Firefox acceptance checks passed.
### Harness serial Ctrl+C recovery — 2026-10-08
A new host progress monitor assigned Zsh's read-only `status` variable; the
command aborted before its marker. The temporary controller's timeout recovery
sent Ctrl+C over QEMU stdio, which terminated QEMU with SIGINT and interrupted
Thunderbird before installation. Run `gui-desktop-20261008T141501Z-d370eb0e`
therefore ends with PASS=12/FAIL=2; its earlier Firefox conversion/native AV1
and balanced evidence remains valid, but no clean poweroff is claimed.
The monitor now uses task-specific variables. The graphical harness configures
`-chardev stdio,id=serial0,signal=off -serial chardev:serial0`, matching
[QEMU's documented signal behavior](https://www.qemu.org/docs/master/system/qemu-manpage.html).
All 22 VM-harness tests pass. A read-only, no-target-disk ISO test
`postinstall-iso-smoke-20261008T194424Z` injected terminal Ctrl+C and confirmed
QEMU remained alive (`serial-interrupt.txt`). That extra test did not finish
normal live boot: Ctrl+C left Utilities selected. It is not a third ISO boot
pass. The earlier two complete SCSI ISO boots remain the candidate boot proof.
Recovery generic boot `gui-desktop-20261008T194155Z-5c8973ad` passes kernel,
systemd and Wayland checks. Firefox source VDB persists, firefox-bin remains
absent, and the original Thunderbird binary is intact. With all compilers idle,
the final codec-fix payload was deployed; guest sys/installer hashes now match
the updated ISO. The interrupted Thunderbird log is preserved, its exact
Portage work directory cleaned, and the remaining three-app sequence restarted
in guest unit `phoebus-postinstall-apps-recovery2`.
### Thunderbird and LibreOffice source/native retests — 2026-10-09
The persistent guest sequence survived overnight in generic GUI run
`gui-desktop-20261008T194155Z-5c8973ad`: Thunderbird source conversion passed
with rc 0, `BUILD_SECONDS=8710`; LibreOffice passed with rc 0,
`BUILD_SECONDS=15005`. Each acceptance script verified the source VDB, old
binary removal, original user-data/default preservation, native compatibility
desktop entry and generic kernel image preservation.
Launching the preserved `thunderbird-bin.desktop` entry opened the welcome
window (shot `03-native-thunderbird-source.png`) as user phoebus with executable
`/usr/lib64/thunderbird/thunderbird`. No email account was configured or mail
sent. Ctrl+Q closed it and the process count became zero.
Launching the preserved `libreoffice25.8-writer.desktop` entry opened native
Writer as user phoebus with executable
`/usr/lib64/libreoffice/program/soffice.bin`. The tip dialog dismissed normally;
a keyboard edit appeared in the document (`05-native-libreoffice-source-edit.png`),
Ctrl+Z undid it and Ctrl+Q closed Writer. No document was saved; process count
became zero. The underlying desktop is visible in shot
`06-native-libreoffice-source-quit.png`.
Calibre remains building QtWebEngine; no success is claimed yet. At 12:11 UTC
it had reached 15942/23519 Ninja steps, with guest 27 GiB and host 31 GiB free.
### Canonical local ISO delivery — 2026-10-09
Verified both candidate digests before replacing the older October 6 local
`installer/live/phoebus-installer-amd64.iso` with the embedded-checked, twice
UEFI-booted postinstall candidate. Canonical SHA-256 is
`b0e63a8d3ed9b82fab08a561522cb9b7b36ddad24ef61bb8493641cf5523a92a`;
the matching canonical `.sha256` sidecar is written. This freed another
1495142400 allocated bytes; stale checksum sidecars for deleted candidates
were removed. The older October 1 verified fallback ISO remains. Journal:
`/var/tmp/phoebus-canonical-iso-delivery.json`. This is local artifact delivery,
not a formal release; the independent CI gate remains NotRun.
Code and completed acceptance evidence were committed and pushed as a6ccae2;
Calibre and tuned tests continue, and their final evidence still needs a commit.
### Calibre resource restart — 2026-10-09
After the host builds finished, increased the disposable VM from 12 GiB/two
QtWebEngine jobs to 20 GiB/four jobs, leaving headroom for the ebuild's documented
3.9+ GiB compiler-job peaks. Intentionally stopped the root package worker
before Calibre installation; its old Flatpak remained intact. Preserved its
partial log as `source-build-calibre-12g-interrupted.log`. Prior generic run
194155Z completed clean poweroff; new generic run
`gui-desktop-20261009T125620Z-d44bdbb5` passes systemd/kernel/Wayland checks.
The temporary Qt-only keepwork setting follows
[Portage's work-reuse documentation](https://dev.gentoo.org/~zmedico/portage/doc/man/make.conf.5.html).
The QtWebEngine ebuild and existing Chromium args.gn SHA checks both pass;
no USE/source inputs changed. Guest unit `phoebus-postinstall-calibre-resume`
runs the real conversion/acceptance script. On success its wrapper removes
keepwork and cleans the exact Qt scratch directory. No Calibre success is
claimed while that worker is active.
The resource resume initially failed on a missing temporary Python launcher,
then twice on lost CMake preparation metadata. These were interrupted-build
recovery failures, not changes to the product policy. Preserved each failure
log. Fresh archive preparation matched all 269594 prepared source files;
restored their original timestamps and the compiled build tree. A temporary
Qt-only bashrc hook invokes the actual Python/CMake setup routines rather than
synthesizing a preparation flag. On October 9 at 13:23 UTC both Ninja processes
run -j4 and the source ebuild/args.gn SHA checks still pass after configuration.
The wrapper removes the hook, keepwork/keeptemp and scratch on success.
An expired sudo session during inspection in generic run 125620Z caused a
chained authentication prompt and temporary-controller recovery timeout. That
run ends 9 PASS/2 FAIL with no clean poweroff. Future guest operations use
explicit fixture authentication followed by sudo -n, and controller recovery
interrupts each chained command. Recovery generic run
`gui-desktop-20261009T131341Z-6115018e` passes kernel/systemd/Wayland checks;
guest unit `phoebus-postinstall-calibre-resume4` is actively compiling.