Fix: bidirectional Quiet<->LowPower fallback in platform_profile setters

[f4mrfaux]

Summary

On hardware where the kernel ACPI platform_profile sysfs exposes only one of Quiet / LowPower for the same semantic profile (different ASUS DMI quirks expose different names — e.g. GA503QR exposes quiet balanced performance, never low-power), setting the absent variant fails with:

org.freedesktop.DBus.Error.NotSupported:
RogPlatform: platform_profile: (low-power) not supported

even when the equivalent IS in choices. The existing fallback in set_platform_profile_on_battery / set_platform_profile_on_ac only handled Quiet -> LowPower, and the bare set_platform_profile D-Bus property had no fallback at all.

This PR makes the alias bidirectional everywhere it matters:

• set_platform_profile (asusd/src/ctrl_platform.rs:469) — substitutes Quiet <-> LowPower when the requested variant isn't in platform_profile_choices but its semantic equivalent is.
• set_platform_profile_on_battery (~line 523) and set_platform_profile_on_ac (~line 562) — extended from one-direction to bidirectional, so the canonical name (the one the kernel exposes) is what gets persisted to /etc/asusd/asusd.ron.
• select_power_profile_for_source (~line 260) — extended to normalize both directions at apply-time, fixing stale-config rehydration after reboot or AC/BAT transitions.

Background — relation to #648

The CLI-string-parsing portion of GitLab issue #648 was fixed by GitLab MR !226 (commit 9366b0e) which made low-power / lowpower / LowPower / low_power all parse to PlatformProfile::LowPower. That fix is necessary but not sufficient: the parsed enum then reaches set_platform_profile, which checks choices.contains(&policy) and rejects on hardware exposing only the kernel-renamed sibling name. This PR closes that follow-up gap.

EPP / behavior preservation

EPP outcomes are unchanged. Both Quiet and LowPower already map to the same source in:

• get_config_epp_for_throttle (asusd/src/ctrl_platform.rs:254-255) — both → profile_quiet_epp
• impl From<PlatformProfile> for CPUEPP (rog-platform/src/cpu.rs:209-210) — both → CPUEPP::Power

So substitution does not affect what EPP gets written; only what name is stored in config and which sysfs string the kernel sees.

Test plan — executed locally on patched daemon

Tested live on GA503QR (linux-g14 7.0.5, /sys/firmware/acpi/platform_profile_choices = quiet balanced performance) via a swap-in-place patched daemon (systemctl stop asusd; systemd-run /target/release/asusd):

#	Command	Pre-patch	Post-patch	Result
1	asusctl profile set LowPower	NotSupported (low-power)	kernel=quiet, active=Quiet	✅ FIX
2	asusctl profile set Quiet	works	works	✅ no regression
3	asusctl profile set Balanced	works	works	✅ no regression
4	asusctl profile set Performance	works	works	✅ no regression
5	asusctl profile set NotARealProfile	CLI parser rejects	CLI parser rejects	✅ no regression
6	asusctl profile set LowPower -b	broken / wrong config	config persists Quiet (canonical), applied	✅ FIX
7	asusctl profile set LowPower -a	broken / wrong config	config persists Quiet (canonical), applied	✅ FIX

Config persistence verified — /etc/asusd/asusd.ron after set LowPower -b -a:

platform_profile_on_battery: Quiet,
platform_profile_on_ac: Quiet,

(both stored as Quiet, the available canonical name, not LowPower).

cargo check -p asusd and cargo build --release -p asusd both clean on ogc/main tip.

What we did NOT validate locally

This hardware exposes quiet balanced performance only — never low-power. So the reverse direction of the bidirectional fallback (a user passing Quiet on a hypothetical low-power-only machine) was reasoned-through (matches the existing one-direction precedent) but not hardware-tested. The patch is a strict no-op when the requested profile IS in choices, so on dual-choice or LowPower-only hardware the behavior should be unchanged from before this PR.

Reviewer notes

• Pre-existing behavior left untouched: set_platform_profile writes config (line 480 in main) before validating against choices. The alias substitution doesn't widen that pre-existing window — the substitution only triggers when the validation would otherwise reject. Worth a separate cleanup PR if desired.
• No tests exist under #[cfg(test)] for set_platform_profile or the alias paths (verified via rg). Happy to add unit tests in a follow-up if maintainers prefer.

🤖 Generated with Claude Code