Files
Panama/tests/quickshell/displays-contract.sh
T
Gabriel Brown 1a91d00f2c Stop two contracts from latching a broken desktop state
Both tests run against the live session and capture "what it was before"
so they can put it back. Neither checked that what they found was sane,
so one interrupted run poisoned every run after it -- and because each
subsequent run faithfully restored the bad value, the desktop stayed
broken while the failure looked like an ordinary flake.

displays-contract left the monitor at scale 1.25 after a failed revert.
The next run recorded 1.25 as the original and restored the desktop to
it. It now reads the shipped scale out of monitors.lua and refuses to
run when the live display disagrees. A failure to parse that value is
fatal rather than skipped, because silently skipping the check is how
the laundering happened in the first place.

focus-session-expiry kills and restarts the shell mid-session, so an
interrupted run leaves caffeine on with nothing left to turn it off. The
next `focus start` recorded "previously on", handed it back on expiry,
and failed the assertion that caffeine ends off -- identically, forever,
with the desktop unable to idle or lock the entire time. It now refuses
to start unless caffeine is already off, which is the only state in
which the test can tell "restored correctly" from "never released".

Both guards name the exact command to recover with. Verified each fires
on a dirty state and passes on a clean one; caffeine was found latched
on this machine and has been released.

Claude-Session: https://claude.ai/code/session_01BRvzt4H8XXLPVH5MyYdk9L
2026-08-18 07:48:02 -04:00

320 lines
13 KiB
Bash
Executable File

#!/usr/bin/env bash
# Display configuration.
#
# This is the only setting in Panama that can leave the user unable to SEE the
# screen well enough to undo it: a mode the panel cannot show, or a scale that
# makes everything unreadable, is not recoverable through the UI that caused it.
#
# So the property under test is not "can it change the resolution" but "does an
# unconfirmed change always come back". A regression here is not a broken
# feature, it is a user staring at a blank monitor.
#
# * an unconfirmed change reverts on its own, and stores nothing
# * a confirmed change is what writes to the settings store
# * a mode, scale, rotation, or output the compositor did not offer is refused
# before anything is applied
#
# The compositor is the live one -- there is no way to test this otherwise --
# but preferences are isolated, and every path restores the display it started
# from.
set -euo pipefail
repo_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
harness="$repo_dir/config/dot/quickshell/displays-harness.qml"
service="$repo_dir/config/dot/quickshell/services/Displays.qml"
page="$repo_dir/config/dot/quickshell/modules/settings/DisplaysPage.qml"
settings_page="$repo_dir/config/dot/quickshell/modules/settings/SettingsPage.qml"
monitors_lua="$repo_dir/config/dot/hypr/monitors.lua"
fail() {
printf 'displays contract: %s\n' "$1" >&2
exit 1
}
# Keep is unavailable until compositor readback exactly matches the request.
for contract in \
'property var pendingRequested:' \
'property var revertExpected:' \
'property bool revertVerificationActive:' \
'property int revertGeneration:' \
'readonly property bool canConfirm:' \
'function matchesRequest(' \
'function scalesForMode(' \
'function isScaleClean('; do
rg -Fq "$contract" "$service" || fail "display service contract is missing: $contract"
done
rg -Fq 'enabled: Displays.canConfirm' "$page" \
|| fail 'Keep is enabled before the display change is verified'
rg -Fq 'options: Displays.scalesForMode(' "$page" \
|| fail 'scale choices are not filtered for the active resolution'
rg -Fq 'property string selectedOutput:' "$page" \
|| fail 'connected outputs cannot be selected'
rg -Fq 'options: Displays.monitors.map(' "$page" \
|| fail 'the output selector is not populated from connected displays'
rg -Fq 'id: revertVerifyTimer' "$service" \
|| fail 'automatic restoration has no bounded readback verification'
rg -Fq 'if (root.busy)' "$service" \
|| fail 'the display service accepts a new apply while another operation is busy'
# Stored JSON is untyped at field level, so the Lua startup consumer is the
# final validation boundary and must support every named output it accepts.
for contract in 'valid_mode' 'valid_scale' 'valid_transform' 'pairs(displays)'; do
rg -Fq "$contract" "$monitors_lua" || fail "monitor startup validation is missing: $contract"
done
# SettingsPage headers are genuinely pinned outside its scrolling surface.
python3 - "$settings_page" <<'PY' || fail 'SettingsPage header is not pinned outside the Flickable'
import sys
text = open(sys.argv[1], encoding="utf-8").read()
loader = text.find("id: pinnedHeader")
flickable = text.find("id: pageScroll")
if loader < 0 or flickable < 0 or loader > flickable:
raise SystemExit(1)
PY
MONITORS_LUA="$monitors_lua" lua - <<'LUA' || fail 'monitor startup accepted invalid persisted geometry or ignored a named output'
package.preload["prefs"] = function()
return {
get = function()
return {
["DP-2"] = { mode = "not-a-mode", scale = -1, transform = 99 },
["HDMI-A-1"] = { mode = "1920x1080@60", scale = 1.5, transform = 1 },
["BAD OUTPUT"] = { mode = "1920x1080@60", scale = 1, transform = 0 },
}
end,
}
end
local calls = {}
hl = { monitor = function(value) table.insert(calls, value) end }
assert(loadfile(os.getenv("MONITORS_LUA")))()
local by_output = {}
for _, value in ipairs(calls) do by_output[value.output] = value end
assert(by_output["DP-2"].mode == "4500x3000@60")
assert(by_output["DP-2"].scale == 1.5)
assert(by_output["DP-2"].transform == 0)
assert(by_output["HDMI-A-1"].mode == "1920x1080@60")
assert(by_output["HDMI-A-1"].scale == 1.5)
assert(by_output["HDMI-A-1"].transform == 1)
assert(by_output["BAD OUTPUT"] == nil)
assert(by_output[""] ~= nil)
LUA
if [[ "${PANAMA_DISPLAYS_STATIC_ONLY:-0}" == "1" ]]; then
printf 'displays contract: PASS (static)\n'
exit 0
fi
config_home="$(mktemp -d /tmp/panama-displays-config.XXXXXX)"
run() { XDG_CONFIG_HOME="$config_home" qs -p "$harness" "$@"; }
status() { run ipc call displays-test status; }
# The shipped geometry, read from the Hyprland config rather than from the
# running compositor.
#
# Everything below captures "original" from what it observes at start, which is
# correct only if the display is already in a good state. A previous run that
# failed mid-revert leaves the display changed, and the next run then captures
# THAT as the original and faithfully restores the desktop to a broken value.
# One flake becomes permanent. So refuse to run from a state that does not match
# what the config says, rather than laundering it.
shipped_scale="$(sed -n 's/^local shipped_scale *= *\([0-9.]*\).*/\1/p' \
"$repo_dir/config/dot/hypr/monitors.lua" | head -1)"
[[ -n "$shipped_scale" ]] || fail 'could not read the shipped scale from monitors.lua -- the guard below depends on it, and skipping it silently is how a dirty baseline gets laundered'
if [[ -n "$shipped_scale" ]]; then
live_scale="$(hyprctl -j monitors | jq -r '.[0].scale')"
if ! awk -v a="$live_scale" -v b="$shipped_scale" 'BEGIN { exit !(a == b) }'; then
fail "the display is at scale $live_scale but the config ships $shipped_scale -- refusing to capture a dirty state as the baseline. Restore it first: hyprctl eval 'hl.monitor({ output = \"DP-2\", mode = \"4500x3000@60\", scale = $shipped_scale, transform = 0 })'"
fi
fi
original_mode=""
original_scale=""
original_transform=""
original_width=""
original_height=""
original_refresh=""
monitor_name=""
monitor_state() {
hyprctl -j monitors | jq -c --arg output "$monitor_name" '.[] | select(.name == $output)'
}
display_is_restored() {
local current
current="$(monitor_state)"
[[ -n "$current" ]] || return 1
jq -e \
--argjson width "$original_width" \
--argjson height "$original_height" \
--argjson refresh "$original_refresh" \
--argjson scale "$original_scale" \
--argjson transform "$original_transform" \
'.width == $width and .height == $height
and ((.refreshRate - $refresh) | fabs) < 0.01
and ((.scale - $scale) | fabs) < 0.001
and .transform == $transform' <<<"$current" >/dev/null
}
restore_display() {
[[ -n "$original_mode" ]] || return 0
hyprctl eval "hl.monitor({ output = \"$monitor_name\", mode = \"$original_mode\", scale = $original_scale, transform = $original_transform })" >/dev/null \
|| return 1
for _ in $(seq 1 50); do
display_is_restored && return 0
sleep 0.2
done
return 1
}
stop_harness() {
# Kill by PID, never `pkill -f displays-harness`: that pattern also matches
# any shell whose command line contains this script's text, which includes
# the invoking shell itself.
[[ -n "${harness_pid:-}" ]] && kill "$harness_pid" >/dev/null 2>&1 || true
rm -rf "$config_home"
}
cleanup() {
local status=$?
trap - EXIT
if ! restore_display; then
printf 'displays contract: FAILED to restore %s to %s scale %s transform %s\n' \
"$monitor_name" "$original_mode" "$original_scale" "$original_transform" >&2
status=1
fi
stop_harness
exit "$status"
}
trap cleanup EXIT
XDG_CONFIG_HOME="$config_home" qs -p "$harness" --daemonize >/dev/null
harness_pid=""
for _ in $(seq 1 40); do
run ipc show 2>/dev/null | rg -q '^target displays-test$' && break
sleep 0.1
done
run ipc show 2>/dev/null | rg -q '^target displays-test$' || fail 'test IPC target did not start'
harness_pid="$(run list | awk '/Process ID:/ { print $3; exit }')"
refresh_fixture="$(run ipc call displays-test refreshIdentityFixture)"
jq -e '
.count == 2
and .modes == ["[email protected]", "[email protected]"]
and .selected == ["[email protected]"]
' <<<"$refresh_fixture" >/dev/null \
|| fail "59.94 Hz and 60.00 Hz lost their distinct selection identity: $refresh_fixture"
for _ in $(seq 1 50); do
[[ "$(status | jq -r .count)" != "0" ]] && break
sleep 0.1
done
state="$(status)"
monitor_name="$(jq -r .name <<<"$state")"
[[ -n "$monitor_name" ]] || fail "no display was detected: $state"
original_mode="$(jq -r .mode <<<"$state")"
original_width="$(jq -r .width <<<"$state")"
original_height="$(jq -r .height <<<"$state")"
original_refresh="$(jq -r .refresh <<<"$state")"
original_scale="$(jq -r .scale <<<"$state")"
original_transform="$(jq -r .transform <<<"$state")"
[[ "$(jq -r .modes <<<"$state")" -gt 0 ]] || fail 'the display reported no usable modes'
# ── Anything the compositor did not offer is refused before applying ─────────
while IFS= read -r kind; do
[[ "$(run ipc call displays-test applyBad "$kind")" == "false" ]] \
|| fail "an invalid $kind was accepted"
[[ "$(status | jq -r .awaiting)" == "false" ]] \
|| fail "an invalid $kind left a change pending"
done <<'KINDS'
mode
scale
transform
output
dirtyScale
KINDS
# The display must not have moved for any of those.
now="$(status)"
[[ "$(jq -r .scale <<<"$now")" == "$original_scale" ]] || fail 'a refused change still altered the scale'
# An immediate Revert may race both the apply process and its first readback.
# It must queue until both are clear, then verify the original generation.
target_scale=$(awk -v s="$original_scale" 'BEGIN { print (s == 1.25) ? 1.5 : 1.25 }')
[[ "$(run ipc call displays-test applyScale "$target_scale")" == "true" ]] \
|| fail 'the immediate-revert fixture could not apply'
run ipc call displays-test revertChange >/dev/null
immediate_reverted=false
for _ in $(seq 1 60); do
if display_is_restored && [[ "$(status | jq -r .awaiting)" == "false" ]]; then
immediate_reverted=true
break
fi
sleep 0.2
done
[[ "$immediate_reverted" == true ]] \
|| fail 'an immediate Revert raced the apply/readback and did not restore the display'
# ── An unconfirmed change reverts on its own and stores nothing ──────────────
[[ "$(run ipc call displays-test applyScale "$target_scale")" == "true" ]] \
|| fail 'a valid scale change was refused'
applied=false
for _ in $(seq 1 30); do
[[ "$(monitor_state | jq -r '.scale')" == "$target_scale" ]] && { applied=true; break; }
sleep 0.2
done
[[ "$applied" == true ]] || fail 'the scale change never reached the compositor'
[[ "$(status | jq -r .awaiting)" == "true" ]] || fail 'an applied change is not awaiting confirmation'
[[ "$(status | jq -r .canConfirm)" == "true" ]] || fail 'an applied change was never verified by compositor readback'
# Wait out the countdown. This is the whole point of the contract.
reverted=false
for _ in $(seq 1 120); do
if [[ "$(monitor_state | jq -r '.scale')" == "$original_scale" ]]; then
reverted=true
break
fi
sleep 0.5
done
[[ "$reverted" == true ]] || fail 'an unconfirmed change did NOT revert -- this would strand a user on an unreadable display'
[[ "$(status | jq -r .awaiting)" == "false" ]] || fail 'the pending state survived the revert'
[[ "$(status | jq -r .overridden)" == "false" ]] || fail 'an unconfirmed change was written to the settings store'
# ── A confirmed change is what writes ────────────────────────────────────────
run ipc call displays-test applyScale "$target_scale" >/dev/null
[[ "$(run ipc call displays-test confirmChange)" == "false" ]] \
|| fail 'Keep accepted a display change before compositor readback'
verified=false
for _ in $(seq 1 30); do
[[ "$(status | jq -r .canConfirm)" == "true" ]] && { verified=true; break; }
sleep 0.2
done
[[ "$verified" == true ]] || fail 'the confirmed change never became safe to keep'
[[ "$(run ipc call displays-test confirmChange)" == "true" ]] \
|| fail 'Keep refused a verified display change'
sleep 0.6
[[ "$(status | jq -r .awaiting)" == "false" ]] || fail 'confirming did not clear the pending state'
[[ "$(status | jq -r .overridden)" == "true" ]] || fail 'confirming did not store the change'
store="$config_home/panama/settings.json"
jq -e --arg m "$monitor_name" '.displays[$m].scale != null' "$store" >/dev/null \
|| fail 'the confirmed change is not in the settings store'
# ── Forgetting clears it ─────────────────────────────────────────────────────
run ipc call displays-test forget >/dev/null
sleep 0.6
[[ "$(status | jq -r .overridden)" == "false" ]] || fail 'forget did not clear the stored display setting'
restore_display || fail 'the final cleanup could not restore and verify the original display'
original_mode=""
stop_harness
trap - EXIT
printf 'displays contract: PASS\n'