Implementation Review Summary

Intent

Fix the Linux glibc build failure introduced when libamdgpu_top 0.11.4 renamed FdInfoStat::get_all_proc_usage to update_proc_usage in a patch release, and harden the version constraint so a future patch surprise cannot recur.

Findings Addressed

• No findings. The PR is a minimal, surgical fix that matches the issue scope exactly.

Remaining Items

• None. The known pre-existing clippy warnings in src/doctor/checks/{privileges,amd,furiosa,gaudi}.rs were verified to exist on main before this PR and are out of scope (separate cleanup issue territory).

Verification

• All stated requirements implemented
  • Cargo.toml line 76: libamdgpu_top = "=0.11.4" (exact pin) under cfg(all(target_os = "linux", not(target_env = "musl"))).
  • src/device/readers/amd.rs:572: fdinfo.update_proc_usage(&proc_index); — call site renamed.
  • Cargo.lock: libamdgpu_top 0.11.3 → 0.11.4 with the expected transitive libdrm_amdgpu_sys 0.8.12 → 0.8.13 bump and nothing else.
• No placeholder/mock code remaining — pure rename + version bump; no stubs.
• Integrated into project code flow
  • src/device/readers/mod.rs:48 (cfg-gated pub mod amd)
  • src/device/reader_factory.rs:48 (cfg-gated use ... amd)
  • src/device/reader_factory.rs:100 (readers.push(Box::new(amd::AmdGpuReader::new())) inside the linux arm guarded by has_amd())
  • The renamed call lives inside impl GpuReader for AmdGpuReader (get_processes), which is the live runtime path, not feature-gated dead code.
• No other call sites remain — grep -rn 'get_all_proc_usage' src/ returns zero matches; the surrounding usage of FdInfoStat::default(), fdinfo.proc_usage, proc_usage.usage.vram_usage, and proc_usage.usage.gtt_usage is unchanged and remains valid against 0.11.4.
• Project conventions followed
  • Inline comment style on the dependency matches existing patterns (cf. russh lines 57–60, zstd/flate2 lines 49–53), references the issue number, explains rationale concisely, and notes "Re-evaluate when bumping."
  • Branch name fix/issue-205-libamdgpu-top-rename follows branch-conventions.
  • Commit message fix: pin libamdgpu_top to =0.11.4 and rename update_proc_usage call follows commit-conventions (proper fix: prefix, imperative mood, references Fix build failure with libamdgpu_top 0.11.4 — method renamed (get_all_proc_usage → update_proc_usage) #205 in body, English).
• Existing modules reused where applicable — N/A; mechanical rename to upstream's new name.
• No unintended structural changes — diff touches exactly three files (Cargo.toml, Cargo.lock, src/device/readers/amd.rs), 8 additions / 6 deletions total. No moves, renames, or refactors outside scope.
• Tests pass — cargo test --release: 21 passed, 14 ignored. cargo fmt --check passes. cargo build --release succeeds on macOS host.

Notes on AMD-target verification

The AMD reader is gated by cfg(all(target_os = "linux", not(target_env = "musl"))), so the macOS build host cannot exercise the renamed call directly. CI on a Linux glibc target is the source of truth for compile verification of that path; this is a known limitation, not a defect of this PR.

PR Finalization Complete

Summary

• Tests: Skipped per scope — the renamed method is upstream library code; no AMD hardware available in CI to exercise the integration path. No fake tests added.
• Documentation: Zero matches for get_all_proc_usage in any .md file. No CHANGELOG file exists in the project, so no entry was needed. No docs changes required.
• Lint/Format: cargo fmt --check passes with no diff. cargo clippy --all-targets shows 7 pre-existing unused-import warnings in src/doctor/checks/{amd,furiosa,gaudi,privileges}.rs — these exist identically on main and none are in the three files touched by this PR (Cargo.toml, Cargo.lock, src/device/readers/amd.rs). Out of scope, not touched.

Final gate results

No changes were needed. The PR is ready for merge.