Files
Panama/tests/quickshell/brightness-helper-contract.sh
T
Gabriel Brown 816ea68f9a Read DDC/CI from the AUX bus, not the EDID bus
With I2C access granted, the helper still found nothing: every connected
monitor was probed on the wrong bus.

A connector has two. The `ddc` symlink points at the classic I2C line
that carries EDID on HDMI and DVI. DisplayPort carries DDC/CI over the
AUX channel instead, which appears as a child directory of the
connector. Both exist on a DP connector and both resolve, so the wrong
choice looks entirely reasonable and simply finds no monitor: this
machine's DP-2 has ddc -> i2c-5, where ddcutil reports "No monitor
detected", while its AUX child i2c-9 answers VCP 0x10 immediately.

This is the guess the previous commit said was unverified, and it was
wrong in the way that mattered. Enumerating from sysfs is still right --
it gives the connector name Hyprland uses and skips empty connectors --
but it has to prefer the AUX child and fall back to the symlink.

The fixture now mirrors sysfs properly: /sys/class/drm/<connector> is a
SYMLINK to the real device directory, and `find` does not follow the
path it is given. A fixture built from plain directories passes whether
or not the code resolves the symlink first, which is a test that agrees
with itself rather than with the kernel.

Verified on hardware: the Kuycon P20 reports 100%, accepts 70 and
returns to 100, and the media keys move it through codex's OSD path.
Both bus-selection mistakes are now caught by the contract.

Claude-Session: https://claude.ai/code/session_01BRvzt4H8XXLPVH5MyYdk9L
2026-08-18 09:51:55 -04:00

164 lines
7.0 KiB
Bash
Executable File

#!/usr/bin/env bash
# panama-brightness enumerates monitors from sysfs and speaks DDC/CI to them.
#
# The parts worth pinning down are the ones that decide whether a slider appears
# at all, and whether it appears attached to the right screen:
#
# * only connectors with something plugged in are probed, because probing an
# empty bus costs a timeout each and there are fourteen of them here;
# * a panel that cannot report brightness is omitted rather than shown as a
# control that does nothing;
# * the connector name matches what Hyprland calls the output, since the UI
# joins on it to get the monitor's description;
# * no I2C access produces the udev command that fixes it, not "no displays".
#
# Runs entirely against fixtures. Real monitors are never touched: both the
# sysfs root and the device root are redirected, and ddcutil is replaced on PATH.
set -uo pipefail
repo_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
helper="$repo_dir/config/dot/quickshell/scripts/panama-brightness"
fail() {
printf 'brightness helper contract: %s\n' "$1" >&2
exit 1
}
fixture="$(mktemp -d /tmp/panama-brightness.XXXXXX)"
trap 'rm -rf "$fixture"' EXIT
mkdir -p "$fixture/drm" "$fixture/dev" "$fixture/bin" "$fixture/i2c"
# Two connectors with a monitor, two without. DP-2 answers DDC; DP-3 is
# connected but does not implement brightness. HDMI-A-1 and DP-1 are empty and
# must never be probed at all.
# A connector has an EDID bus (the `ddc` symlink) and, on DisplayPort, an AUX
# bus that appears as a child directory. DDC/CI rides the AUX channel where one
# exists, and the `ddc` line answers nothing on a DP connector even though it
# still resolves -- so the aux argument here is what a real DisplayPort monitor
# looks like, and omitting it is what HDMI and DVI look like.
make_connector() {
local name="$1" ddc_bus="$2" status="$3" aux_bus="${4:-}"
local device="$fixture/devices/$name"
# /sys/class/drm/<connector> is a SYMLINK to the real device directory, and
# this fixture mirrors that rather than using a plain directory. It matters:
# `find` does not follow the path it is given, so code that searches the
# unresolved path finds nothing while appearing to work anywhere the entry
# happens to be a real directory.
mkdir -p "$device"
printf '%s\n' "$status" >"$device/status"
mkdir -p "$fixture/i2c/i2c-$ddc_bus"
ln -sfn "$fixture/i2c/i2c-$ddc_bus" "$device/ddc"
[[ -n "$aux_bus" ]] && mkdir -p "$device/i2c-$aux_bus"
mkdir -p "$fixture/drm"
ln -sfn "$device" "$fixture/drm/$name"
return 0
}
# DP-2 is the DisplayPort case: its EDID line is bus 5, which answers nothing,
# and its AUX child is bus 9, which does. Choosing bus 5 here finds no monitor
# at all, which is exactly the bug this pins down.
make_connector card1-DP-1 4 disconnected
make_connector card1-DP-2 5 connected 9
make_connector card1-DP-3 6 connected
make_connector card1-HDMI-A-1 7 disconnected
# has_accessible_bus only needs one readable/writable node to exist.
touch "$fixture/dev/i2c-5"
# Stub ddcutil. Records every bus it is asked about so the test can prove the
# disconnected ones were skipped. Bus 6 refuses, standing in for a panel without
# VCP 0x10.
#
# Bus 5 reports its brightness out of 200 rather than 100. Most panels do use
# 100, which is exactly the problem: with a maximum of 100 the scaling
# arithmetic is the identity, so a helper that ignored the reported maximum
# entirely would pass every assertion. 200 makes reads and writes that skip the
# conversion visibly wrong.
cat >"$fixture/bin/ddcutil" <<'STUB'
#!/usr/bin/env bash
bus=""
args=("$@")
for ((i = 0; i < ${#args[@]}; i++)); do
[[ "${args[$i]}" == "--bus" ]] && bus="${args[$((i + 1))]}"
done
printf '%s\n' "$bus" >>"$DDCUTIL_PROBE_LOG"
for arg in "$@"; do
if [[ "$arg" == "setvcp" ]]; then
printf 'set %s %s\n' "$bus" "${args[-1]}" >>"$DDCUTIL_SET_LOG"
exit 0
fi
done
case "$bus" in
9) printf 'VCP 10 C 120 200\n'; exit 0 ;;
*) exit 1 ;;
esac
STUB
chmod +x "$fixture/bin/ddcutil"
export DDCUTIL_PROBE_LOG="$fixture/probes.log"
export DDCUTIL_SET_LOG="$fixture/sets.log"
: >"$DDCUTIL_PROBE_LOG"
: >"$DDCUTIL_SET_LOG"
run_helper() {
PATH="$fixture/bin:$PATH" \
PANAMA_BRIGHTNESS_DRM_ROOT="$fixture/drm" \
PANAMA_BRIGHTNESS_DEV_ROOT="$fixture/dev" \
"$helper" "$@"
}
# ── Enumeration ──────────────────────────────────────────────────────────────
listing="$(run_helper list)"
jq -e . >/dev/null 2>&1 <<<"$listing" || fail "list did not emit JSON: $listing"
[[ "$(jq -r '.displays | length' <<<"$listing")" == "1" ]] \
|| fail "expected exactly one controllable display, got: $listing"
[[ "$(jq -r '.displays[0].connector' <<<"$listing")" == "DP-2" ]] \
|| fail "the connector name must match Hyprland's output name: $listing"
# The AUX bus, not the EDID bus its `ddc` symlink points at.
[[ "$(jq -r '.displays[0].bus' <<<"$listing")" == "9" ]] \
|| fail "the display was mapped to its EDID bus instead of its DisplayPort AUX bus, where nothing answers: $listing"
# 120 of a maximum of 200 is 60%.
[[ "$(jq -r '.displays[0].value' <<<"$listing")" == "60" ]] \
|| fail "brightness was not read as a percent of the reported maximum: $listing"
[[ "$(jq -r '.error' <<<"$listing")" == "" ]] \
|| fail "a successful listing must not carry an error: $listing"
# A connected panel that cannot report brightness is dropped, not listed.
jq -e '.displays | map(.connector) | index("DP-3") == null' >/dev/null <<<"$listing" \
|| fail 'a display without VCP 0x10 was listed as controllable'
# ── Disconnected connectors are never probed ─────────────────────────────────
if grep -qxE '4|7' "$DDCUTIL_PROBE_LOG"; then
fail "a disconnected connector was probed -- each empty bus costs a timeout: $(tr '\n' ' ' <"$DDCUTIL_PROBE_LOG")"
fi
# ── Writes scale to the reported maximum ─────────────────────────────────────
run_helper set 9 40
[[ "$(tail -1 "$DDCUTIL_SET_LOG")" == "set 9 80" ]] \
|| fail "set did not scale to the display's maximum: $(cat "$DDCUTIL_SET_LOG")"
run_helper set 9 500
[[ "$(tail -1 "$DDCUTIL_SET_LOG")" == "set 9 200" ]] \
|| fail "an out-of-range percent was not clamped: $(cat "$DDCUTIL_SET_LOG")"
# ── No I2C access explains itself ────────────────────────────────────────────
rm -f "$fixture/dev"/i2c-*
denied="$(run_helper list)"
[[ "$(jq -r '.displays | length' <<<"$denied")" == "0" ]] \
|| fail "displays were reported without I2C access: $denied"
grep -q 'udevadm' <<<"$(jq -r '.error' <<<"$denied")" \
|| fail "the no-access error must name the command that fixes it, got: $(jq -r '.error' <<<"$denied")"
printf 'brightness helper contract: PASS\n'