docs(evidence): task-12 namespace-wide _PRW wake-device enumeration

.omo/evidence/task-12-ryzen-7000-x670e-compat.txt

@@ -0,0 +1,120 @@
+Task: todo 12 of ryzen-7000-x670e-compat
+Title: Namespace-wide _PRW wake-device enumeration in acpid
+Date: 2026-08-05
+
+Objective
+---------
+Replace the fixed laptop-oriented wake-device candidate list in
+acpid (drivers/acpid/src/wake.rs:176-280) with a namespace-wide walk
+that discovers every device exposing a _PRW method, so AMD desktop
+GPP* paths are discovered alongside legacy Intel laptop paths
+(PC00/CNVW/lid). All _PRW evaluation continues to flow through the
+serialized AML executor.
+
+Implementation
+--------------
+- drivers/acpi-rs/src/aml/namespace.rs
+  Added collect_prw_device_paths() which walks the AML namespace
+  recursively and returns every device/object path that has a _PRW
+  method. The walk skips duplicate paths and preserves device
+  ordering within each branch.
+
+- drivers/acpid/src/acpi.rs
+  Added AcpiContext::enumerate_prw_device_paths() that exposes the
+  namespace walker to acpid. It returns the full path list and
+  leaves actual _PRW evaluation to the existing serialized AML
+  executor (evaluate_acpi_method / aml_eval), so the serialized
+  executor requirement (todo 10) is honored.
+
+- drivers/acpid/src/wake.rs
+  Replaced the fixed candidate list in WakeRegistry::enumerate()
+  with a call to enumerate_prw_device_paths() plus a HashSet guard
+  to collapse any duplicates. Arm/disarm semantics and the LG Gram
+  wake-device paths are preserved because the namespace walk
+  includes them when _PRW is present.
+
+Tests
+-----
+Host-runnable tests added in drivers/acpi-rs/src/aml/namespace.rs:
+1. collect_prw_discovers_intel_and_amd_devices_with_no_duplicates
+   - Builds a synthetic namespace with:
+       \PC00.CNVW._PRW  (Intel laptop Wi-Fi / Bluetooth wake)
+       \GPP0.GPP7._PRW  (AMD desktop PCIe root port / device wake)
+       \_SB.PCI0.LPCB.EC0.LID0._PRW (legacy lid wake)
+   - Verifies all three are discovered and no path appears twice.
+
+2. collect_prw_ignores_devices_without_prw
+   - Builds a namespace where some devices have _PRW and sibling
+     devices do not.
+   - Verifies only _PRW-bearing paths are returned.
+
+RED/GREEN
+---------
+RED (before implementation): the two new tests failed to compile
+because collect_prw_device_paths() and the AcpiContext wrapper did
+not exist; the fixed candidate list in wake.rs also could not
+produce the AMD GPP0/GPP7 paths.
+
+GREEN (after implementation):
+
+$ cargo test -p acpi
+...
+test aml::namespace::tests::collect_prw_discovers_intel_and_amd_devices_with_no_duplicates ... ok
+test aml::namespace::tests::collect_prw_ignores_devices_without_prw ... ok
+...
+test result: ok. 22 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
+
+$ cargo check -p acpid
+Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.10s
+(54 pre-existing warnings, none related to the new code)
+
+Build verification
+------------------
+Ran the canonical check-sweep:
+
+$ ./local/scripts/build-redbear.sh --allow-dirty --check-sweep redbear-mini
+
+- Check-sweep covered all 52 packages in the target; individual logs
+  are under /tmp/redbear-build-state.HkdE2r/logs/check-sweep-*.log.
+- The base check-sweep log (check-sweep-base.log) shows only
+  pre-existing warnings; no errors were introduced by the _PRW
+  changes.
+- The subsequent full build phase proceeded past recipe compilation
+  and failed later with:
+
+    installer: failed to install: No space left on device (os error 28)
+
+  This is an environment/disk-space failure unrelated to the acpid
+  change and unrelated to the previously observed malformed-patch
+  issue in rust-native build artifacts.
+
+Commits
+-------
+Base submodule (branch submodule/base):
+  2db58838 feat(acpid): namespace-wide _PRW wake-device enumeration
+  (3 files changed, 156 insertions(+), 38 deletions(-))
+
+Parent repo (branch 0.3.2):
+  The local/sources/base gitlink already points to commit
+  61d196433ed9e8414eb1bf7a204642d8392639ae, which includes the
+  namespace-wide _PRW commit 2db58838 as its parent. No additional
+  gitlink change was required because the parent HEAD already
+  recorded that submodule hash.
+
+Constraints honored
+-------------------
+- Did not change arm/disarm semantics in wake.rs.
+- Did not hardcode board-specific paths; all wake candidates are
+  discovered from the AML namespace.
+- Did not introduce unwrap()/expect() in library code.
+- Did not edit config/redbear-full.toml or local/cache/pkgar/.
+- Left the unrelated pcid working-tree changes from another session
+  untouched (they remain uncommitted in the base submodule).
+
+DoneClaim
+---------
+Namespace-wide _PRW wake-device enumeration is implemented,
+tested, and committed in the base fork. The acpid crate
+compiles cleanly and the host-runnable acpi tests pass.
+The change is ready for the next full image build once disk-space
+is available.