acdream/docs/reviews/coverage-ledger.md

7.9 KiB

Coverage ledger

Audit baseline: 15539a22a67f8d915d88f8b1d8126cd55eedda6e.

The baseline commit and path-set fingerprints make the reviewed file universe reproducible without copying thousands of paths into this document:

| Scope | Fingerprint (git ls-files ... | git hash-object --stdin) | Files/lines at baseline | |---|---|---:| | src/**, tools/** | b79a0390fad9a3e86de765e49f869720045bbc94 | 1,325 C# files / 416,628 lines | | tests/** | 34b425e434a7a3980e556a1ce445f2a40ae196fd | 1,260 C# files / 402,131 lines | | docs/**, root authorities, memory/** | 46b6c109b8b986a9425c46e66cbe6967e5ae3d46 | 1,127 tracked paths |

The repository has 4,018 tracked paths: 1,365 under src, 1,352 under tests, 1,109 under docs, 152 under tools, 15 under memory, one WorldBuilder gitlink, and 24 root/editor/CI paths.

Owned production and tool code

“Inventory complete” means the file set, project references, package references, size distribution, and solution membership were collected. “Reviewed” means every file in the row participated in the all-file structural, comment/reference, configuration, failure-handling, and dependency scans, with manual review of entry points, public boundaries, exceptions, and size/ complexity hotspots.

Area C# files Lines Inventory Deep review
src/AcDream.App 628 205,466 complete complete
src/AcDream.Bake 11 1,884 complete complete
src/AcDream.Cli 6 1,736 complete complete
src/AcDream.Content 31 7,533 complete complete
src/AcDream.Core 275 78,848 complete complete
src/AcDream.Core.Net 98 21,781 complete complete
src/AcDream.Headless 35 8,859 complete complete
src/AcDream.Launcher 15 3,342 complete complete
src/AcDream.Launcher.Core 50 13,661 complete complete
src/AcDream.Platform 2 206 complete complete
src/AcDream.Plugin.Abstractions 8 147 complete complete
src/AcDream.Plugins.Smoke 1 34 complete complete
src/AcDream.Runtime 112 62,894 complete complete
src/AcDream.UI.Abstractions 37 6,294 complete complete
tools/* 16 3,903 complete complete

The non-C# owned tooling surface contains 84 tracked executable scripts: 38 WinDbg/CDB command files, 30 PowerShell scripts, and 16 Python scripts across the repository root and tools/. These are included in the tool review even though they are not represented by the C# path-set count above. Initial execution/readability triage is complete; support classification and exact prerequisites are classified in F-004. All 13 tracked tool projects were also built individually: eight succeed and five fail (three absent relative DatReaderWriter references and two source-API drift errors). Two of the eight passes depend on a developer-specific absolute project path that happens to exist on the audit machine. Script triage found absolute Windows paths in 55 of 84 files and explicit developer-home paths in 43.

Size is already an audit dimension: 83 owned C# files are at least 1,000 lines, 24 are at least 2,000, and nine are at least 3,000. Size alone is not a finding; cohesion, ownership, and change coupling will determine which files are flagged.

Test source

Test area C# files Lines Inventory File-by-file review
AcDream.App.Tests 515 174,271 complete complete
AcDream.Bake.Tests 5 665 complete complete
AcDream.Cli.Tests 1 178 complete complete
AcDream.Content.Tests 29 6,120 complete complete
AcDream.Core.Net.Tests 115 26,976 complete complete
AcDream.Core.Tests 398 99,802 complete complete
AcDream.Headless.Tests 18 9,573 complete complete
AcDream.Launcher.Core.Tests 27 8,615 complete complete
AcDream.Launcher.Tests 7 2,123 complete complete
AcDream.Platform.Tests 2 224 complete complete
AcDream.Runtime.Tests 91 64,102 complete complete
AcDream.UI.Abstractions.Tests 46 8,583 complete complete
Fixture/helper projects and tests/Fixtures 6 899 complete complete

The exact baseline set is every path returned by git ls-files 'tests/**/*.cs'; its fingerprint is recorded above. There are 68 test C# files at least 1,000 lines, 16 at least 2,000, and nine at least 3,000. Parameterized rows and generated/live-DAT gates will be audited through their data sources as well as their declaring methods.

Documentation and repository infrastructure

Area Inventory Review
Root README.md, NOTICE.md, AGENTS.md, CLAUDE.md complete complete
docs/README.md, milestones, roadmap, issues complete complete
Architecture and divergence authorities complete complete
Active M4 plans/specifications complete complete
Historical plans, research, and audit records complete complete classification/link scan
memory/ and documented claude-memory/ links complete complete
Solution/project graph and CI complete complete
External/generated/reference material complete complete

Initial inventory facts

  • AcDream.slnx contains all 14 owned src projects, 16 test/fixture projects, and only one of the 13 tracked tool projects (RetailTimeProbe).
  • references/WorldBuilder is the sole tracked reference gitlink and is not initialized in this worktree. No other documented reference checkout is present.
  • Five tracked tool projects reference a nonexistent references/DatReaderWriter project; two encode a developer-specific absolute path.
  • The repository-instruction audit skill named at AGENTS.md:725 is absent, as is the claude-memory/ tree linked by the documentation map.
  • The solution Release build was executed at the baseline: it succeeds with 26 warnings, all emitted by test projects.
  • Every test assembly was executed. Eleven complete normally in the combined run; Launcher.Core passes 338/338 alone but deadlocks in the combined run. Current reproducible totals are 14,747 passing and 77 reported skipped, plus one duplicate theory row dropped during discovery.
  • All 11,458 declared fact/theory-like methods were syntax-inventoried for assertion signals, diagnostic naming, literal tautologies, conditional execution, skips, and timing primitives. The conditional-execution pass found 314 method-level empty returns in 294 methods across 107 files; 271 of those methods silently pass when an asset or opt-in environment gate is not available. This cross-file triage supplied the exception set for the completed semantic follow-up below.

Completion evidence

The semantic follow-up is complete. It manually reviewed every exceptional class produced by the all-file matrix: no-oracle methods and their helpers, empty returns, explicit/custom skips, environment and installed-DAT gates, source-text assertions, sleeps/real-time deadlines, random/concurrent tests, literal tautologies, duplicate data, diagnostic naming, dead-stack tests, and the open flake ledger. The exact actionable catalog is in test-quality-audit.md; valid no-throw and deterministic-repeatability tests were deliberately excluded from the “useless” label.

Documentation coverage used a full 1,014-file Markdown relative-link scan and manual reading of the authority chain, current campaign/status documents, repository instructions, memory entry points, issue/divergence ledgers, CI, release/update paths, and representative historical records. Historical research was classified and mechanically checked rather than line-edited as if it were current architecture.

Build/test execution, package-vulnerability lookup, project/tool builds, repository-size/provenance census, secret-pattern scan, and managed-stack capture for the launcher hang provide executable corroboration. No source, test, configuration, existing documentation, commit, or external system was modified by this audit.