diff options
Diffstat (limited to 'scripts/post-rebuild-check')
| -rwxr-xr-x | scripts/post-rebuild-check | 124 |
1 files changed, 113 insertions, 11 deletions
diff --git a/scripts/post-rebuild-check b/scripts/post-rebuild-check index 24e99a4..2013b26 100755 --- a/scripts/post-rebuild-check +++ b/scripts/post-rebuild-check @@ -1,6 +1,6 @@ #!/bin/sh # SPDX-License-Identifier: GPL-3.0-or-later -# post-rebuild-check - the five checks a rebuilt machine actually needs. +# post-rebuild-check - the eight checks a rebuilt machine actually needs. # # A rebuilt machine looks finished and isn't. Five gaps surfaced on velox # within two days of the 2026-08-13 reinstall, and three of them LOOKED @@ -25,6 +25,14 @@ # DoT/DNSSEC validation this machine's DNS runs on, so nothing # resolves -- including the NTP pool that would fix the clock; velox # deadlocked exactly this way 2026-08-19 and needed a second device) +# 7. hypridle installed but not running (nothing then triggers idle lock +# or suspend, so a laptop runs until its battery is gone -- which is +# how velox reset the RTC that caused check 6's deadlock in the first +# place; a caffeine remembered from an earlier boot is the known cause) +# 8. a working repo cloned from the read-only https endpoint (correct +# for a stranger with no key on the server, wrong for this machine, +# which finds out at the first push with a 403 -- velox's dotfiles +# remote sat that way for four days after its rebuild) # # The .gitignore rule in check 4 is what scopes it: a tooling path is only # expected where the project's own .gitignore names it, so a project that @@ -58,6 +66,10 @@ # the special value MISSING = no NTP daemon active # PRC_CHRONY_CONF path to chrony.conf (a fixture, under test) -- the # confdir it names is what decides which drop-ins count +# PRC_IDLE_DAEMON pgrep output for hypridle; "" = installed but not +# running, the special value MISSING = not installed +# PRC_REPO_REMOTES newline list of "path origin-url"; an empty URL +# means origin could not be read # PRC_SYSTEMCTL path to the systemctl binary (a fake, under test) # PRC_SYSTEMCTL_TIMEOUT seconds to allow each systemctl call (default 5) # @@ -69,10 +81,11 @@ usage() { cat <<'EOF' post-rebuild-check - verify a rebuilt machine is actually finished -Runs the six checks that caught velox's 2026-08 reinstall gaps: failed +Runs the eight checks that caught velox's 2026-08 reinstall gaps: failed units, present-but-inert user units, orphaned *.example configs, missing -per-project tooling state, the signal-cli registration, and whether time -sync can recover from a wrong clock without DNS. +per-project tooling state, the signal-cli registration, whether time sync +can recover from a wrong clock without DNS, whether anything still +triggers idle lock and suspend, and whether the working repos can push. Usage: post-rebuild-check [--help] @@ -94,6 +107,7 @@ CHECK_FINDINGS=0 FINDING_LINES="" signal_missing="" ntp_missing="" +idle_absent="" # Every systemctl call is bounded. A wedged user manager spins and answers # nothing -- seen live on velox 2026-08-17, where `is-enabled`, `cat`, and @@ -190,7 +204,7 @@ while IFS= read -r line; do unit=${line#*:} finding "$scope unit failed: $unit" done < "$STAGE" -report "check 1/6: failed units" +report "check 1/8: failed units" # --- 2. user unit files present but not enabled --------------------------- @@ -280,7 +294,7 @@ while read -r name state; do esac finding "unit file present but not enabled: $name ($state)" done < "$STAGE" -report "check 2/6: unit files" +report "check 2/8: unit files" # --- 3. *.example files whose real sibling is missing --------------------- @@ -325,7 +339,7 @@ while IFS= read -r root; do [ -e "${ex%.example}" ] || finding "example without its real file: $ex" done < "$WORK/examples" done < "$WORK/roots" -report "check 3/6: local files" +report "check 3/8: local files" # --- 4. gitignore-mode projects missing their tooling --------------------- @@ -379,7 +393,7 @@ todo.org todo\.org inbox inbox EOF done < "$WORK/projects" -report "check 4/6: project tooling" +report "check 4/8: project tooling" # --- 5. signal-cli registration ------------------------------------------- @@ -406,7 +420,7 @@ if [ "$signal_missing" = 1 ]; then elif [ -z "$signal_missing" ] && [ -z "$accounts" ]; then finding "no signal account registered — agent-text relays into this machine, so paging breaks for the whole fleet" fi -report "check 5/6: signal registration" +report "check 5/8: signal registration" # --- 6. NTP can recover a wrong clock without DNS ------------------------- # @@ -483,7 +497,95 @@ else finding "every NTP source is named by hostname — a wrong clock breaks DNS, so nothing can resolve them and the clock stays wrong" fi fi -report "check 6/6: NTP bootstrap" +report "check 6/8: NTP bootstrap" + +# --- 7. the idle daemon survives session start ---------------------------- +# +# A laptop that never sleeps has no symptom until the battery is gone, so +# nothing surfaces this without being asked. On velox 2026-08-19 hypridle +# started cleanly at 15:29:48 and `settings restore` killed it six seconds +# later, replaying a caffeine stored in an earlier boot. The machine ran +# 11h40m fully awake on battery, died when it flattened, and reset its RTC -- +# which took DNS down with it, the same deadlock check 6 exists for. The +# desktop looked correct throughout. +# +# Behavioural on purpose: this asks whether the daemon is alive, not why it +# might not be, so a stale caffeine, a crash, and a broken config all surface +# the same way. Gated on hypridle being installed, because that is what marks +# a machine as using it -- archsetup installs it only for Hyprland, so a +# headless or dwm box would otherwise report a finding on every run. + +if [ -n "${PRC_IDLE_DAEMON+set}" ]; then + idle_pids=$PRC_IDLE_DAEMON + if [ "$idle_pids" = "MISSING" ]; then + idle_pids="" + idle_absent=1 + fi +elif command -v hypridle >/dev/null 2>&1; then + # pgrep exits non-zero with no match, which is the not-running case rather + # than a probe failure, so the || keeps `set -e`-style callers out of it. + idle_pids=$(pgrep -x hypridle 2>/dev/null) || idle_pids="" +else + idle_pids="" + idle_absent=1 +fi + +if [ "$idle_absent" = 1 ]; then + : # hypridle is not part of this machine -- nothing to check +elif [ -z "$idle_pids" ]; then + finding "hypridle is installed but not running — nothing triggers idle lock or suspend, so this machine stays awake until its battery is gone; a caffeine remembered from an earlier boot is the known cause" +fi +report "check 7/8: idle daemon" + +# --- 8. working repos cloned from the read-only endpoint ------------------ +# +# archsetup clones the user's own archsetup and dotfiles from +# https://git.cjennings.net/..., which serves anonymous clones and refuses +# pushes. That default is correct for a stranger installing archsetup -- they +# have no key on the server -- and wrong for this machine, which has to push. +# ARCHSETUP_REPO / DOTFILES_REPO override it, but only where they are +# configured: a curl|bash install, or a rebuild from a stock ISO, takes the +# default straight back. +# +# Nothing about the tree shows it. The clone is complete and ordinary, and the +# machine finds out at the first push, with a 403 -- which is how velox's +# dotfiles remote was found on 2026-08-17, four days after its rebuild, by +# which time the same rebuild's shallow clone had already answered a +# credential-history question wrongly. +# +# Only the read-only endpoint is flagged. An https remote elsewhere may be +# perfectly pushable through a credential helper, and guessing about hosts +# this machine does not own would stand noise in front of the real findings. + +if [ -n "${PRC_REPO_REMOTES+set}" ]; then + repo_remotes=$PRC_REPO_REMOTES +else + repo_remotes="" + for repo in "$HOME/code/archsetup" "$HOME/.dotfiles"; do + # -e not -d: a worktree or submodule .git is a file naming the gitdir. + [ -e "$repo/.git" ] || continue + # A repo with no origin still gets a line, with an empty URL, so the + # loop below reports it rather than skipping it into a silent pass. + repo_url=$(git -C "$repo" remote get-url origin 2>/dev/null) + repo_remotes="${repo_remotes}${repo} ${repo_url} +" + done +fi + +stage "$repo_remotes" +while IFS= read -r repo_line; do + [ -n "$repo_line" ] || continue + repo_path=${repo_line%% *} + repo_url=${repo_line#"$repo_path"} + repo_url=${repo_url# } + case "$repo_url" in + "") + finding "$repo_path: origin could not be read — the remote was not checked" ;; + https://git.cjennings.net/*|https://cjennings.net/*) + finding "$repo_path: origin is the read-only endpoint ($repo_url) — git push returns 403; set the ssh form, or ARCHSETUP_REPO/DOTFILES_REPO before installing" ;; + esac +done < "$STAGE" +report "check 8/8: repo remotes" # --- summary -------------------------------------------------------------- @@ -491,5 +593,5 @@ if [ "$TOTAL_FINDINGS" -eq 0 ]; then echo "all checks clean" exit 0 fi -echo "$TOTAL_FINDINGS finding(s) across 6 checks" +echo "$TOTAL_FINDINGS finding(s) across 8 checks" exit 1 |