diff options
| -rwxr-xr-x | archsetup | 44 | ||||
| -rw-r--r-- | tests/installer-steps/test_configure_hibernate_delay.py | 131 | ||||
| -rw-r--r-- | tests/installer-steps/test_orchestrators.py | 2 | ||||
| -rw-r--r-- | todo.org | 121 | ||||
| -rw-r--r-- | working/velox-reinstall/velox-reinstall-runbook.org | 154 |
5 files changed, 450 insertions, 2 deletions
@@ -1689,6 +1689,7 @@ essential_services() { configure_randomness configure_networking configure_power + configure_hibernate_delay configure_backlight_access configure_ssh_server configure_fail2ban @@ -1851,6 +1852,49 @@ UDEVEOF error_warn "triggering leds udev rules" "$?" } +configure_hibernate_delay() { + # How long a suspended machine waits before hibernating. Read only by + # systemctl suspend-then-hibernate, so it does nothing until something + # asks for that — a desktop panel's idle stage, a lid rule, or a hand + # invocation. + # + # Unset, the delay depends on the hardware: a machine with a battery + # hibernates on a low-battery alarm at an hour nobody chose, and one + # without a battery falls back to systemd's 2h default. 90 minutes + # makes it a stated ceiling instead — on a laptop systemd still + # hibernates early if the battery gets there first, since it runs both + # timers and takes whichever fires soonest. The ceiling is long enough + # that stepping away costs a screen unlock rather than a passphrase + # and a full resume, and short enough that a laptop in a bag stops + # holding its encryption keys in RAM (a suspended machine does; a + # hibernated one does not). + # + # Harmless on a machine that cannot hibernate: nothing reads it there. + # Numbered 60- because systemd reserves 10-40 for vendor drop-ins under + # /usr and 60-90 for local ones under /etc; a 10- file here would sort + # below a vendor drop-in and quietly lose to it. + # $1 is the drop-in directory, defaulting to the system's so tests can + # run against a temp dir. + local confdir="${1:-/etc/systemd/sleep.conf.d}" + local dropin="$confdir/60-hibernate-delay.conf" + + action="setting the suspend-to-hibernate delay" && display "task" "$action" + + mkdir -p "$confdir" 2>> "$logfile" || { error_warn "$action" "$?"; return 1; } + cat > "$dropin" << 'SLEEPEOF' 2>> "$logfile" || { error_warn "$action" "$?"; return 1; } +# Wait 90 minutes after suspending before hibernating. +# +# Unset, the delay depends on the hardware: a machine with a battery +# hibernates on a low-battery alarm, one without falls back to systemd's +# 2h default. 90 minutes keeps a short absence cheap (a screen unlock, not +# a passphrase and a resume) while making sure a laptop left in a bag +# drops its encryption keys out of RAM rather than holding them for hours. +[Sleep] +HibernateDelaySec=90min +SLEEPEOF + chmod 644 "$dropin" 2>> "$logfile" || error_warn "$action" "$?" +} + configure_power() { # Power diff --git a/tests/installer-steps/test_configure_hibernate_delay.py b/tests/installer-steps/test_configure_hibernate_delay.py new file mode 100644 index 0000000..95f2dcb --- /dev/null +++ b/tests/installer-steps/test_configure_hibernate_delay.py @@ -0,0 +1,131 @@ +"""Test configure_hibernate_delay — the suspend-to-hibernate handoff. + +The desktop panel can upgrade its idle suspend stage to +suspend-then-hibernate, but the delay between the two is systemd's +HibernateDelaySec, which lives in a root-owned drop-in and so cannot be a +panel setting. Left unset the delay depends on the hardware: a machine +with a battery hibernates on a low-battery alarm at an hour nobody chose, +and one without falls back to systemd's 2h default. + +90 minutes is the seeded value: long enough that stepping out of a meeting +costs a screen unlock rather than a passphrase and a resume, short enough +that a laptop in a bag is not holding its encryption keys in RAM all +afternoon (a suspended machine is; a hibernated one is not). + +Method: sed-extract configure_hibernate_delay from the real `archsetup`, +point it at a temp config dir, and assert on the file it writes. + + python3 -m unittest tests.installer-steps.test_configure_hibernate_delay +""" + +import os +import re +import stat +import subprocess +import tempfile +import textwrap +import unittest + +REPO_ROOT = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..")) +ARCHSETUP = os.path.join(REPO_ROOT, "archsetup") + + +def run(confdir): + script = textwrap.dedent(f"""\ + logfile=/dev/null + action="" + display() {{ :; }} + error_warn() {{ echo "WARN: $1"; return 1; }} + source <(sed -n '/^configure_hibernate_delay() {{/,/^}}/p' "{ARCHSETUP}") + configure_hibernate_delay "{confdir}" + echo "RC=$?" + exit 0 + """) + return subprocess.run( + ["bash", "-c", script], capture_output=True, text=True, timeout=10, + ) + + +def rc_of(r): + m = re.search(r"^RC=(\d+)$", r.stdout, re.M) + assert m, "no RC line in output: %r / %r" % (r.stdout, r.stderr) + return int(m.group(1)) + + +def body(confdir): + with open(os.path.join(confdir, "60-hibernate-delay.conf")) as f: + return f.read() + + +class ConfigureHibernateDelay(unittest.TestCase): + # ------------------------------------------------------------ normal ---- + def test_writes_a_sleep_drop_in_with_the_delay(self): + with tempfile.TemporaryDirectory() as d: + r = run(d) + self.assertEqual(rc_of(r), 0) + text = body(d) + self.assertIn("[Sleep]", text) + self.assertIn("HibernateDelaySec=90min", text) + + def test_drop_in_is_world_readable_not_writable(self): + with tempfile.TemporaryDirectory() as d: + run(d) + mode = stat.S_IMODE( + os.stat(os.path.join(d, "60-hibernate-delay.conf")).st_mode) + self.assertEqual(mode, 0o644) + + def test_explains_why_the_value_is_not_left_to_systemd(self): + # The number is a judgement call, so the file has to carry its + # reasoning: whoever changes it next needs the tradeoff, not just + # a number to overwrite. Assert on the tradeoff's own terms — + # "contains a hash" would pass on any comment at all. + with tempfile.TemporaryDirectory() as d: + run(d) + text = body(d).lower() + self.assertIn("2h default", text) # what unset would give + self.assertIn("encryption keys", text) # why not longer + self.assertIn("passphrase", text) # why not shorter + + def test_drop_in_sorts_above_vendor_drop_ins(self): + # systemd reserves 10-40 for /usr and 60-90 for /etc. A 10- file + # in /etc sorts below a vendor 20- file and silently loses to it. + with tempfile.TemporaryDirectory() as d: + run(d) + name = os.listdir(d)[0] + prefix = int(name.split("-")[0]) + self.assertGreaterEqual(prefix, 60) + self.assertLessEqual(prefix, 90) + + # ---------------------------------------------------------- boundary ---- + def test_running_twice_leaves_one_correct_drop_in(self): + with tempfile.TemporaryDirectory() as d: + run(d) + first = body(d) + r = run(d) + self.assertEqual(rc_of(r), 0) + self.assertEqual(first, body(d)) + + def test_absent_directory_is_created(self): + with tempfile.TemporaryDirectory() as d: + nested = os.path.join(d, "etc", "systemd", "sleep.conf.d") + r = run(nested) + self.assertEqual(rc_of(r), 0) + self.assertIn("HibernateDelaySec=90min", body(nested)) + + # ------------------------------------------------------------- error ---- + @unittest.skipUnless(os.geteuid() != 0, "root ignores directory write bits") + def test_unwritable_directory_warns_and_does_not_crash(self): + with tempfile.TemporaryDirectory() as d: + confdir = os.path.join(d, "ro") + os.mkdir(confdir) + os.chmod(confdir, 0o500) + try: + r = run(confdir) + self.assertIn("WARN:", r.stdout) + self.assertNotEqual(rc_of(r), 0) + finally: + os.chmod(confdir, 0o700) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/installer-steps/test_orchestrators.py b/tests/installer-steps/test_orchestrators.py index 34c46d1..c2eb710 100644 --- a/tests/installer-steps/test_orchestrators.py +++ b/tests/installer-steps/test_orchestrators.py @@ -28,7 +28,7 @@ ARCHSETUP = os.path.join(REPO_ROOT, "archsetup") ORCHESTRATORS = { "essential_services": [ "configure_randomness", "configure_networking", "configure_power", - "configure_backlight_access", + "configure_hibernate_delay", "configure_backlight_access", "configure_ssh_server", "configure_fail2ban", "configure_firewall", "configure_service_discovery", "configure_job_scheduling", "configure_package_cache", "configure_snapshots", @@ -45,7 +45,126 @@ below): input-side-spec.org (DRAFT, four decisions open). * Archsetup Open Work -** TODO [#A] Move secrets out of public dotfiles → private repo + combined personal ISO :feature:security:dotfiles: +** TODO [#A] Reseat velox input-cover ribbon — phantom power button :bug:velox:hardware: +DEADLINE: <2026-08-14 Fri> +:PROPERTIES: +:CREATED: [2026-08-13 Thu] +:LAST_REVIEWED: 2026-08-13 +:END: +Machine off, lift the input cover (Framework QR-guided procedure, 5 +fasteners), reseat its ribbon connector to the mainboard — disturbed in the +2026-08-13 board swap. Root cause of every "mystery reboot" that day: +chassis flex (flash-drive touch, ethernet bump, lid partially lowered) +fired phantom power-button presses — journalctl -b -1 showed "Power key +pressed short." → orderly logind poweroff, then the glitching button +powered it back on. While in there, reseat the USB expansion cards too — +the flaky slot (two hard resets, one no-enumeration) is likely the same +flex problem. +THIRD SYMPTOM (2026-08-13 evening): touchpad delivers ZERO input events — +15s synchronized libinput debug-events capture while swiping caught +nothing, though i2c enumeration and a driver rebind handshake are clean. +Signature of a dead interrupt line on the same ribbon. Keyboard + power +LED lines work; BT mouse is the interim pointer. +ESCALATED 2026-08-13 21:00: a fourth event killed the machine THROUGH the +shield. Previous boot's journal ends mid-line (tailscaled chatter) with no +shutdown sequence at all — a hard power cut, not logind acting. So the +glitch now reaches the EC/hardware power path, which no software setting +can intercept. The reseat is the only fix, and this is a +lose-work-without-warning failure mode, not an inconvenience. +Interim shield (already live): /etc/systemd/logind.conf.d/powerkey.conf +sets HandlePowerKey=ignore — phantom presses log but do nothing; EC-level +10s hold still force-cuts. Consider keeping it even after the repair. +Verify after reseat: flex the chassis edges + partially lower the lid, then +grep the journal for new "Power key pressed" lines — zero means fixed. +Must be done before the Sunday flight — a phantom press mid-travel with the +shield on is survivable, but the connector should not be trusted at 30,000 +feet on the loose setting. +** DOING [#A] Velox reinstall — DR test of archangel + archsetup :velox:chore: +DEADLINE: <2026-08-15 Sat> +:PROPERTIES: +:CREATED: [2026-08-13 Thu] +:LAST_REVIEWED: 2026-08-13 +:END: +Mainboard swapped Intel→AMD (Ryzen AI 9 HX 370); new NVRAM has no boot entry. +Decision: full reinstall via archangel+archsetup, run deliberately as a +disaster-recovery drill before the Sunday flight. Runbook (live checklist): +[[file:working/velox-reinstall/velox-reinstall-runbook.org][working/velox-reinstall/velox-reinstall-runbook.org]] +Done 2026-08-13: ISO rebuilt (archangel-2026-08-13, archsetup baked with AMD +microcode detection, velox profiles at /root/, .ai/inbox excluded — build.sh +edits pending commit in archangel), contents verified, dotfiles swept clean of +Intel assumptions. +Finding folded in: velox's truenas backups silently stopped ~Jul 6 (newest is +DAILY.0 Jul 6; wolf.conf.gpg from Jul 29 is in NO backup). Salvage pass in the +runbook is therefore REQUIRED before partitioning, and the fresh install must +fix + verify the backup timer (runbook Phase 5). +** TODO [#B] Truenas session-host VM for long-running agent sessions :feature:tooling: +:PROPERTIES: +:CREATED: [2026-08-13 Thu] +:LAST_REVIEWED: 2026-08-13 +:END: +A small VM on truenas (TrueNAS SCALE KVM) as the home for long-running / +away-mode agent sessions. The case, per Craig 2026-08-13: truenas is the +only machine on ethernet, so network recovery after an outage is automatic +(wifi hosts may never reassociate unattended); it's UPS-backed through +blip-to-hours outages; it has the Comet KVM for out-of-band recovery; and +appliance uptime discipline means it doesn't reboot for workstation +reasons. Tonight's live demonstration of anchor-staleness risk (39 min +unlogged during a bare-metal recovery) is the motivating incident — the +session's durability equals the anchor's lag at interruption, so a host +that doesn't get interrupted is worth real money. +Costs to engineer around: a third environment to keep synced (repos, +rulesets, tailnet identity); credential provisioning — GATED on the +secrets-repo work (the [#A] secrets task above): the VM should be the +secrets bundle's second consumer after the personal ISO, not another +hand-copied key sprawl; a firm RAM carve-out so builds don't fight the +ZFS ARC; headless only — desktop-coupled sessions stay on ratio/velox. +Build deliberately AFTER the vacation, not before Sunday. +Companion idea (cheaper, complementary): put ratio on the UPS. +** TODO [#C] screen-lock test suite red on ratio :bug:test:dotfiles: +:PROPERTIES: +:CREATED: [2026-08-13 Thu] +:LAST_REVIEWED: 2026-08-13 +:END: +tests.screen-lock.test_screen_lock fails 21 of 23 (+1 error) on ratio, +verified pre-existing with unrelated changes stashed (2026-08-13). Breaks +the make test green bar for every dotfiles commit until triaged. Suspect +environmental (the suite exercises hyprlock/relaunch behavior that may +need session state this shell lacks) or a regression from a recent +screen-lock commit — diagnose, then fix or mark/skip with a reason. +Grading: Major severity (blinds the pre-commit gate for the whole repo) × +every-commit frequency on this machine, but test-infra only, no user +impact = [#C] judgment call rather than matrix-dictated. +** TODO [#C] Keyboard backlight binding + boot default :feature:dotfiles:velox: +:PROPERTIES: +:CREATED: [2026-08-13 Thu] +:LAST_REVIEWED: 2026-08-13 +:END: +Velox's kbd backlight (chromeos::kbd_backlight since the AMD board) boots +at 0 and the dotfiles carry no keyboard-brightness keybinding at all (swept +2026-08-13 — never existed). Add: Hyprland binds (XF86KbdBrightness* or a +chord) driving brightnessctl -d "chromeos::kbd_backlight", a sane boot +default, and a udev rule granting the video/input group write access so it +works without sudo (a bare ssh session got EPERM). Check whether Fn+Space +(EC-handled on Frameworks) already cycles it — if so the bind is a +complement, not the only path. Ratio: n/a (desktop). +** TODO [#B] Hibernate in the settings dial power actions :feature:dotfiles: +:PROPERTIES: +:CREATED: [2026-08-13 Thu] +:LAST_REVIEWED: 2026-08-13 +:END: +Add hibernate alongside suspend/lock in the settings module's dial power +actions. Sequencing (Craig confirmed the dial placement 2026-08-13): +1. Prove hibernate on velox first — systemctl hibernate through a real + resume; the chain (LUKS swap p3, keyfile-in-initramfs, encrypt+resume + hooks, resume= on the ZBM cmdline) went live with the 2026-08-13 + reinstall but is untested on this AMD board. +2. Then consider suspend-then-hibernate as the default lid behavior + (systemd sleep.conf HibernateDelaySec) — hibernate's savings with no + button at all; possibly a "deep sleep" toggle in the module. +3. Then the dial action itself. +Note: ratio has no swap partition, so hibernate stays velox-only until +ratio gets one; the dial entry should degrade gracefully where there's no +resume target. :PROPERTIES: :CREATED: [2026-08-11 Tue] :LAST_REVIEWED: 2026-08-11 diff --git a/working/velox-reinstall/velox-reinstall-runbook.org b/working/velox-reinstall/velox-reinstall-runbook.org new file mode 100644 index 0000000..02671e2 --- /dev/null +++ b/working/velox-reinstall/velox-reinstall-runbook.org @@ -0,0 +1,154 @@ +#+TITLE: Velox Reinstall Runbook — DR Test of archangel + archsetup +#+AUTHOR: Craig Jennings +#+DATE: 2026-08-13 + +Context: velox's mainboard swapped Intel → AMD (Ryzen AI 9 HX 370, Radeon +890M, 96GB RAM). Old SSD intact but the new board's NVRAM has no boot entry, +and velox is ZFS root + ZFSBootMenu, so a stock Arch USB can't even read the +pool. Decision: full reinstall via archangel + archsetup, run deliberately as +a disaster-recovery test of the ISO and scripts before the Sunday flight. +Recent backup in hand; ratio available as the working machine. + +Fallback ordering if the test finds a real gap: +- Before partitioning starts: the old system is intact — the ZBM repair + route (efibootmgr entry pointing at the ZBM loader on the ESP, then + amd-ucode swap in a chroot) is still available. +- After partitioning: the floor is a manual Arch install; the backup makes + that survivable. + +* Phase 0 — Preflight on ratio (agent-driven, done before you leave the desk) + +- [ ] Rebuild the ISO with archsetup baked in: the 2026-08-02 ISO predates + the microcode vendor-detection fix (archsetup, 2026-08-08) and was built + without ARCHSETUP_DIR at all. + #+begin_src sh + cd ~/code/archangel && sudo ARCHSETUP_DIR=~/code/archsetup ./build.sh + #+end_src +- [ ] build.sh fixes before the final rebuild (archangel repo): + - rsync exclude for =.ai= (keeps =archsetup/.ai/private-design/= — the + credential audit — off the portable USB stick). + - copy =installer/velox-*.conf= to =airootfs/root/= so the machine profile + is on the ISO at =/root/velox-zfs.conf=. +- [ ] Verify the ISO carries: =/code/archsetup= (with =install_cpu_microcode=), + =/root/velox-zfs.conf=, no =.ai/private-design=. Loop-mount or unsquashfs + spot-check. +- [X] USB ready (done 2026-08-13 15:25): the new ISO was copied to the Ventoy + drive, sha256-verified against the source, and the 2026-04-09 + 2026-06-16 + archangel ISOs removed. Boot the stick and pick + =archangel-2026-08-13-vmlinuz-6.18.43-lts-x86_64.iso= from the Ventoy menu. + +* Phase 1 — UEFI setup on velox (BIOS screen, before any boot) + +- [ ] Disable Secure Boot. Mandatory — the ZFS kernel modules are unsigned; + the new board ships with it enforced by factory default. +- [ ] Set the system clock. The board swap reset the RTC to 2025-01-01; + a wrong clock breaks TLS and pacman signature checks in the live env. + Rough accuracy is fine — NTP tightens it once networked. +- [ ] While you're in setup: check boot-order UI shows the USB. + +* Phase 2 — Salvage pass (live ISO, BEFORE running the installer) — REQUIRED + +NOT optional insurance. Verified 2026-08-13: velox's newest truenas backup is +DAILY.0 = 2026-07-06 — five weeks stale. The backup timer on velox broke +around Jul 6 (truenas itself only went dark Jul 24, and it's back now; ratio +and mybitch backed up today). Everything since Jul 6 exists only on the old +SSD — including =wolf.conf.gpg= (created Jul 29), which is therefore in NO +backup at all. This pass also keeps the repair fallback alive until +partitioning starts. + +- [ ] Network up (=nmtui= or ethernet), then confirm clock: =timedatectl=. +- [ ] Import the old pool read-only and unlock: + #+begin_src sh + zpool import -N -o readonly=on -R /mnt zroot + zfs load-key zroot # passphrase prompt + zfs mount zroot/ROOT/default + zfs mount -a 2>/dev/null # home datasets etc.; ignore failures + #+end_src +- [ ] Push a full fresh backup to truenas over the LAN — mirror the layout + the backup job uses (etc + home), into a clearly-named one-off dir: + #+begin_src sh + rsync -aHAX --info=progress2 /mnt/etc /mnt/home \ + truenas:/mnt/vault/backups/velox/pre-reinstall-2026-08-13/ + #+end_src + (=/usr= is in the regular backups but is all reinstallable — skip unless + paranoid. The 96GB-RAM board will not be the bottleneck; the LAN is.) +- [ ] Spot-check the copy landed: =wolf.conf.gpg=, =.ssh=, =.gnupg=, newest + files in =~/documents= and =~/downloads=. +- [ ] Check for uncommitted repo work and either push or note it: + =~/.emacs.d= (known: the auto-dim-other-buffers.el unresolved merge), + =~/.dotfiles=, anything under =~/code=. +- [ ] Export cleanly: =cd /; zfs unmount -a; zpool export zroot=. + +* Phase 3 — Install (the actual DR test) + +- [ ] Review the profile, then run the installer: + #+begin_src sh + less /root/velox-zfs.conf # FILESYSTEM=zfs, HOSTNAME=velox, single nvme + archangel --config-file /root/velox-zfs.conf + #+end_src + Note: the profile's ZFS_PASSPHRASE / ROOT_PASSWORD are the =welcome= + placeholders — fine for install; both change post-install (=zfs change-key + zroot= for the pool, =passwd= for root). +- [ ] Record every rough edge as a DR-test finding — that's the point of + running it this way. Anything that needs a manual nudge gets a todo entry + in archangel or archsetup afterward. +- [ ] Reboot into ZBM → boot the new environment. + +* Phase 4 — archsetup (first boot of the installed system) + +- [ ] Log in as root, network up, then verify the clock synced. +- [ ] Get archsetup — two paths, test the offline one since this is a DR + drill (the online curl path is the everyday alternative): + #+begin_src sh + # offline: mount the install USB and copy the baked tree + mount /dev/disk/by-label/ARCHANGEL* /mnt 2>/dev/null || mount /dev/sdX1 /mnt + cp -r /mnt/code/archsetup /root/archsetup && cd /root/archsetup + ./archsetup + #+end_src +- [ ] Expected on the new board: =install_cpu_microcode= detects + AuthenticAMD and installs amd-ucode (verified 2026-08-13, 7/7 tests). + Podman socket, camera udev rule, tlp radio state, ZFS /tmp mask are all + in the installer now — none need manual application afterward. +- [ ] archsetup clones + stows dotfiles. The velox host tier has no Intel + assumptions (swept 2026-08-13); maint's capability probe runtime-detects + amd-pstate. + +* Phase 5 — Post-install restore + verification + +- [ ] Restore from backup (credentials, ssh keys, gpg, user data). The + secrets-bundle-in-ISO design is not built yet — manual restore is the + known gap, not a test failure. +- [ ] WireGuard: decrypt + re-place =wolf.conf.gpg= at =~/.config/wireguard/=; + re-import the NM profile (autoconnect off, as before). +- [ ] Change the placeholder passwords: =passwd=, =zfs change-key zroot=. +- [ ] PSR workaround — REQUIRED on this board. The Ryzen AI 300 has a known + idle instability (Panel Self Refresh hangs/reboots the machine; hit during + the live session 2026-08-13). Add =amdgpu.dcdebugmask=0x610= to the + installed system's kernel command line — velox boots via ZBM, so set it on + the pool: =zfs set org.zfsbootmenu:commandline="... amdgpu.dcdebugmask=0x610" zroot/ROOT/default= + (keep the existing args; append). Revisit after a BIOS update ≥3.05 or a + kernel that fixes PSR on Strix Point — track via the Framework issue + tracker (SoftwareFirmwareIssueTracker #110). +- [ ] New-hardware spot-checks: + - =journalctl -k | grep -i microcode= — amd-ucode applied. + - =cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver= — expect + amd-pstate(-epp). + - wifi + bluetooth up (new board radios), touchpad behavior, camera. + - =glxinfo -B= / =vulkaninfo --summary= — Radeon 890M on RADV. +- [ ] Fresh clones automatically carry the post-purge rewritten git history — + closes the clone-reconcile rider from 2026-08-11 without action. +- [ ] Fix and verify the backup timer on the fresh install — it was silently + broken since ~Jul 6. After the first manual run succeeds, confirm a new + DAILY.0 appears under =truenas:/mnt/vault/backups/velox/=. Diagnose why it + broke (timer unit dead? mount failure? credential?) if the old journal + survives in the salvage copy. +- [ ] Update the machine-identity memory: velox is now AMD (amd-pstate), + both daily drivers AMD. Fix the stale =intel_pstate= comment in + =airplane-mode= line 6 while at it (cosmetic). +- [ ] File every DR-test finding in the owning project's todo. + +* Timing + +Today is Thursday; the flight is Sunday. Target: Phases 0–4 tonight or +Friday, leaving Saturday as pure buffer. If the install stalls past Friday +evening, cut losses to the manual-install floor. |