Kpa-clawbot/CoreScope v2.8.0

v2.8.0 — The Testening

on GitHub

🧪 The Testening

This release transforms MeshCore Analyzer from a "push and pray" codebase into a properly tested, CI-gated project. 99 commits, 950+ tests, and every push must pass before reaching production.

Test Infrastructure

• 950+ automated tests across 12 test files — unit, integration, E2E, spec-driven, and golden fixtures
• CI/CD pipeline gates every deploy: backend tests → frontend E2E → coverage report → deploy (only if all pass)
• Playwright E2E tests run against a local server before deploy — broken frontend blocks production
• Code coverage tracking — backend (88%) and frontend (42%) measured and reported on every push
• Dynamic badges in README auto-update: test counts, backend coverage, frontend coverage, deploy status
• Istanbul-instrumented frontend — Playwright exercises code, window.__coverage__ extracted, merged with backend coverage
• Golden fixture tests — 20 real production packets frozen as regression tests. If decoder output changes, tests break immediately.
• Spec-driven decoder tests — verified against MeshCore firmware C++ source (Mesh.cpp, Packet.h, AdvertDataHelpers.h)

AGENTS.md — AI Agent Guide

New project-level AGENTS.md derived from analyzing all 685 commits:

• 4.3x fix ratio identified and documented — rules to prevent it
• Must read firmware source before implementing protocol features
• All tests must pass before push, every feature adds tests
• Cache busters in same commit, verify API shape before building UI
• Plan → human sign-off → implement (no more yolo)

Features

• Packet Filter Language — Wireshark-style filter engine (type == GRP_TXT && snr > 5)
• Node Aging System — 🟢 Active / ⚪ Stale with visual fading on map markers, configurable thresholds
• Hash Issues Detection — flags nodes with inconsistent hash sizes, links to firmware bug + fix
• Sortable Node Columns — click any header to sort, persisted to localStorage
• Deep-linkable Analytics — #/analytics?tab=collisions&section=inconsistentHashSection
• 8 Preset Themes with WCAG AA contrast verification

Bug Fixes

• Decoder: encrypted payload field sizes matched to firmware source (dest 1B, src 1B, MAC 2B — was 6+6+4)
• Decoder: advert flags are a 4-bit enum type, not individual bit flags (room servers were misclassified)
• 4 missing payload types added: GRP_DATA, MULTIPART, CONTROL, RAW_CUSTOM
• Room server detection now works (52 room servers correctly identified)
• Node list shows actual last-heard time from all traffic, not just adverts
• 200 packet cap removed from WebSocket handler
• Resolve-hops API calls eliminated from packets page (client-side only)
• Map markers always use role color (not gray when hash_size missing)

Performance

• All nodes fetched once, filtered client-side (was 200 limit with server filtering)
• Hop resolution fully client-side — no per-packet API calls
• WebSocket live updates no longer cap at 200 packets

Coverage Breakdown