Implementation Review Summary

Intent

Fix seven visual polish issues in the local-mode Activity panel and device list rendering on Apple Silicon: border off-by-one, P/E gauge alignment, thermal badge removal, hostname suppression, ANE always-on, NPU scaffolding, and Processes header restyle.

Verification

• All stated requirements implemented
• No placeholder/mock code remaining
• Integrated into project code flow
• Project conventions followed
• Existing modules reused where applicable
• No unintended structural changes
• Tests pass (325 unit + 385 integration, 0 failures)

Detailed findings

All seven issue requirements are correctly and completely implemented:

Fix 1 - Border off-by-one (activity_panel.rs:214): Changed 0..(inner_width + 1) to 0..inner_width, making the bottom border emit panel_width columns matching top border and content rows. Correct.

Fix 2 - P/E gauge shared bar_width (activity_panel.rs:376-378): Pre-computes shared_bar_width using p_block_width.max(e_block_width), then passes it to both draw_cluster_line calls. The function signature was updated to accept bar_width: usize directly. Correct.

Fix 3 - Thermal badge removal (gpu_sparkline_panel.rs): Removed thermal_pressure_badge helper, thermal_badge computation in build_rows, and thermal_pressure from test fixtures. All badge fields set to None. GPU Temp row now has badge: None. Correct.

Fix 4 - GPU Temp sparkline width (resolved by Fix 3): With badge_len = 0 for all rows, sparkline_width is now identical across GPU Util, GPU Mem, and GPU Temp. Correct.

Fix 5 - Hostname suppression (4 renderers + frame_renderer.rs + ui_loop.rs): Added show_hostname: bool to print_gpu_info, print_cpu_info, print_memory_info, print_storage_info. All call sites pass true in remote mode and false in local mode. format_hostname_with_scroll is properly gated inside the conditional. Hostname scroll animation in ui_loop.rs is skipped in local mode for both GPU and CPU hostname updates. Chassis renderer is untouched per spec. Correct.

Fix 6 - ANE always-on + NPU scaffolding (gpu_sparkline_panel.rs): has_ane_data renamed to show_ane_row, now returns detect_apple_silicon(state) (no power gate). show_npu_row added, returns false with doc comment pointing to src/api/metrics/npu/common.rs. Both are wired into gpu_content_rows and build_rows. Tests cover ANE-idle case and NPU scaffolding. Correct.

Fix 7 - Processes header restyle (process_renderer.rs:62-74): Replaced bare "Processes:\r\n" with styled title rule using Cyan for prefix/title and DarkGrey for trailing dashes. Still emits exactly 1 row, so RESERVED_HEADER_ROWS = 4 is unchanged. Correct.

Minor observations (LOW, not blocking)

• #[allow(clippy::unnecessary_wraps)] on show_npu_row is a no-op since the function returns bool, not Option/Result. Harmless but unnecessary.
• Comment at process_renderer.rs:194 still says "Processes:" title -- could say "styled title rule" to match the new implementation. Purely cosmetic.

Remaining Items

None -- all seven fixes are correctly implemented with no issues found.

Security & Performance Review -- PR #166

Scope

Reviewed all 9 changed files (171 additions, 111 deletions) for security vulnerabilities, performance regressions, and code quality issues.

Findings

FIXED -- LOW: Dead clippy allow attribute (commit 3759566)

#[allow(clippy::unnecessary_wraps)] on show_npu_row in src/ui/gpu_sparkline_panel.rs was a no-op. The unnecessary_wraps lint only fires on functions returning Option or Result, not bool. The attribute was misleading about the function's semantics. Removed.

INFO -- LOW: Minor block-width formula inconsistency

In src/ui/activity_panel.rs, the p_block_width formula at line 376 (cores.len() + cores.len() / 4) does not subtract 1 for the trailing-group-of-4 case, while the blocks_printed formula at line 426 does. This means shared_bar_width is 1 column shorter than necessary when the core count is a multiple of 4. The lost column becomes right-side padding, so the visual output is correct and panel width is maintained. No fix needed -- the inconsistency is cosmetic (1 pixel of bar) and the conservative direction (shorter bar, not overflow).

INFO -- Pre-existing: Byte-index slice on Unicode header

src/ui/process_renderer.rs:174 uses header_format[horizontal_scroll_offset..] which is a byte-index slice. The header contains multi-byte Unicode sort arrows (3 bytes each). If horizontal_scroll_offset lands mid-character, this would panic. This is pre-existing code not introduced by this PR, but worth tracking for a future hardening pass.

Verified Clean

• cargo test -- all 325 tests pass (0 failures)
• cargo clippy -- -D warnings -- no warnings
• cargo fmt --check -- clean
• All show_hostname call sites in frame_renderer.rs are consistent: true for remote-mode renderers, false (or !view_state.is_local_mode) for local-mode renderers
• saturating_sub is used correctly throughout for width calculations -- no integer underflow risk
• No new unwrap() calls on fallible data paths
• No memory leaks: hostname scroll animation correctly skipped in local mode (no growing offset map entries)
• No security concerns: pure UI rendering code with no user input parsing, no file I/O, no network access

Overall Assessment

This PR is well-written. The changes are straightforward UI polish with proper use of saturating_sub for safe arithmetic, correct gating of hostname animations to avoid wasted work, and clean separation of local/remote rendering paths. One dead attribute was removed.