test(hibernate): #76 checks.hibernate-swapfile — verify swapfile+offset in a VM

Push #76's swapfile-resume residual into a VM as far as headlessly possible.
A full hibernate→resume from a RUNTIME swapfile turned out not to be
headless-reachable: systemd reports CanHibernate=no for a swap that isn't
declared in the boot config, and a runtime-computed resume_offset can't be
fed into the boot cmdline (the genuine chicken-and-egg). So the check now
verifies, bounded and deterministically, what the partition test
(checks.hibernate) can't and what a VM CAN prove: the installer's btrfs
@swap swapfile is built correctly and its offset is valid — NOCOW
mkswapfile, file-type swapon, a good `map-swapfile -r` offset the kernel
accepts (/sys/power/resume_offset), and zram (prio 100) above it.

The literal encrypted-swapfile power-cycle + LUKS-initrd unlock stays a V3
laptop check (the dev box runs exactly this layout, /proc/cmdline verified).

(First cut had an unbounded wait_for_shutdown that hung on the CanHibernate
refusal; rewritten with a thread-bounded wait + a CanHibernate pre-check,
which gave the clean verdict. Device is /dev/vdb without useNixStoreImage.)

Verification: V2 PASS (nix build .#checks.x86_64-linux.hibernate-swapfile,
fresh, exit 0); flake check --no-build green. #76 residual narrowed in
BACKLOG/ROADMAP.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

agent/BACKLOG.md

@@ -39,12 +39,21 @@ not Nomarchy, so it can't close it). Steps in HARDWARE-QUEUE. Design record in
   `@swap` swapfile + `resume_offset` wiring; `swapSize=0` = no-swap opt-out.
 - `docs/MIGRATION.md` enable-hibernation runbook (existing machines).
 - `rofi.nix` Power → Hibernate: kept unconditional, **notifies on failure**.
-- `checks.hibernate` VM test (V2 PASS) — hibernate→resume→crash round-trip;
-  zram (prio 100) doesn't swallow the image, disk swap carries it.
+- `checks.hibernate` VM test (V2 PASS) — hibernate→resume→crash round-trip
+  (swap partition); zram (prio 100) doesn't swallow the image, disk carries it.
+- `checks.hibernate-swapfile` VM test (V2 PASS) — the installer's btrfs `@swap`
+  **swapfile** setup: NOCOW `mkswapfile`, file-type swapon, a valid
+  `map-swapfile -r` offset the kernel accepts (`/sys/power/resume_offset`),
+  zram above it. (Bounded proof of the swapfile+offset the partition test
+  can't reach; a full runtime-swapfile hibernate isn't headless-reachable —
+  systemd `CanHibernate=no` for undeclared swap.)
 
-**V3 remaining:** real hibernate→resume on the LUKS+@swap swapfile+offset
-layout (QEMU can't feed a boot-time offset / drive LUKS-initrd) + the no-swap
-notify in a real session. Steps queued in HARDWARE-QUEUE. Close #76 on PASS.
+**V3 remaining (narrowed):** the literal encrypted-swapfile power-cycle —
+`Hibernate → power off → unlock LUKS → resume session` on the LUKS+@swap
+layout — plus the no-swap notify in a real session. QEMU can't feed a
+runtime-computed `resume_offset` into the boot cmdline, and the dev box runs
+exactly this layout (`/proc/cmdline` verified) but isn't Nomarchy. Steps in
+HARDWARE-QUEUE. Close #76 on PASS.
 
 ## LATER
 

agent/JOURNAL.md

@@ -17,6 +17,28 @@ Template:
 
 ---
 
+## 2026-07-11 — #76 checks.hibernate-swapfile (VM-verify the swapfile+offset)
+- **Task:** Bernardo — push #76's swapfile-resume residual into a VM (like
+  the auto-theme sunset trick).
+- **Did:** Tried a full runtime-swapfile hibernate→resume test; it hit a
+  bounded, definitive **`CanHibernate=no`** — systemd won't hibernate to a
+  swap that isn't declared in the boot config, and a runtime-computed
+  `resume_offset` can't reach the boot cmdline (the genuine chicken-and-egg).
+  First cut also had an **unbounded `wait_for_shutdown`** that hung → killed
+  it, bounded the wait in a thread (180s) + a `CanHibernate` pre-check, which
+  is what surfaced the clean verdict. Pivoted `checks.hibernate-swapfile` to
+  what a VM *can* prove deterministically and the partition test can't: the
+  installer swapfile **build + offset** — NOCOW `mkswapfile`, file-type
+  swapon, valid `map-swapfile -r` offset accepted by `/sys/power/resume_offset`,
+  zram above it. (Device fix: `/dev/vdb` once `useNixStoreImage` was dropped.)
+- **Verified:** **V2 PASS** (fresh build, exit 0). BACKLOG/ROADMAP updated;
+  #76 residual narrowed to the literal encrypted power-cycle (dev box runs it).
+- **Lesson:** the sunset-trick spirit applies but not always the full cycle —
+  when systemd/kernel gates the behaviour (CanHibernate), verify the
+  *construction* deterministically and leave only the physical cycle to HW.
+  Always bound `wait_for_shutdown` (no timeout param → thread + join).
+- **Next suggestion:** #76/#79 both `[blocked:hw]`; QA / PROPOSED otherwise.
+
 ## 2026-07-11 — Code-review fixes (parallel review of the session batch)
 - **Task:** Bernardo asked for parallel work; spawned a fresh-eyes opus
   review of `2055842..HEAD` (static). It found the batch sound — 2 real LOW