eval: corpus recall 3/9 → 4/9 — self-contained loaded-subscription fixture (P-012)

[PhysShell]

What

Ratchets the corpus benchmark recall 3/9 → 4/9 by fixing a fixture that was understating our own strongest class.

screentogif-loaded-subscription is a subscription leak — the class the extractor is best at — yet it scored a miss. The cause wasn't the extractor: the reduction's before.cs subscribed to _viewModel.ShowErrorRequested but never declared VideoSourceViewModel, so the type-aware extractor couldn't bind the += to an event and honestly emitted OWN050 (unresolved) — a note, not a verdict. The fixture, not the analyzer, was the problem.

Making the reduction self-contained (a minimal VideoSourceViewModel with the three events + a StatusBand stub, mirrored in after.cs) lets the extractor resolve it and flag the leak — warning-tier, an injected DataContext source it can't prove outlives the window — exactly as it does on the full ScreenToGif repo. The matching -= in after.cs's Window_Closing keeps the fix silent.

The proof (benchmark scorecard, CI)

screentogif-loaded-subscription  before[caught: OWN001]  after[clean]
...
benchmark: 4/9 bugs caught in real C# · 9/9 fixes clean · 0 false positive(s) on fixes (recall floor 4)

loaded-subscription flipped MISSED → caught: OWN001; specificity stays perfect (9/9 clean, 0 FP). The corpus-benchmark floor is raised to --min-recall 4 so it can't regress.

Lesson recorded

A benchmark fixture that references an undeclared type silently degrades to OWN050; self-contained fixtures (like the samples) measure honestly. Documented in corpus-benchmark.md + P-012.

The remaining 5 misses are genuine frontend extraction gaps — pool double-return (OWN003) / use-after-return (OWN002), the interprocedural ownership-handoff, a cross-method use-after-dispose, and a region-escape shape — the tracked recall backlog the floor ratchets up to as each lands.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Rg8kSk1YT14x7A1vo5zgED

---

Generated by Claude Code

Summary by CodeRabbit

• Chores

  • Raised the benchmark precision/recall threshold for real-C# corpus testing.

• Tests

  • Updated benchmark test fixtures to ensure self-contained validation and improved result accuracy from 3/9 to 4/9 recall.

• Documentation

  • Updated benchmark documentation to reflect improved test results and clarify remaining extraction gaps.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 8e75f4bb-c6c9-420a-b852-e95bce6be639

📥 Commits

Reviewing files that changed from the base of the PR and between d7069ec and 7224bd3.

📒 Files selected for processing (5)

• .github/workflows/ci.yml
• corpus/real-world/screentogif-loaded-subscription/after.cs
• corpus/real-world/screentogif-loaded-subscription/before.cs
• docs/notes/corpus-benchmark.md
• docs/proposals/P-012-bug-corpus-mining.md

---

📝 Walkthrough

Walkthrough

The screentogif-loaded-subscription corpus fixture gains minimal in-file stubs for VideoSourceViewModel and StatusBand in both before.cs and after.cs, making it self-contained so the extractor can bind the event subscription and produce a verdict. The CI --min-recall gate is then ratcheted from 3 to 4, and corpus-benchmark.md and P-012-bug-corpus-mining.md are updated to reflect the new 4/9 recall result.

Changes

Corpus fixture fix and recall gate ratchet

Layer / File(s)	Summary
Corpus fixture: add self-contained type stubs corpus/real-world/screentogif-loaded-subscription/before.cs, corpus/real-world/screentogif-loaded-subscription/after.cs	Both before.cs and after.cs receive minimal in-file stand-ins for VideoSourceViewModel (event declarations: ShowErrorRequested, HideErrorRequested, CloseRequested) and StatusBand (stub Error/Hide methods), eliminating the unresolved external type dependency that previously caused OWN050 instead of a real verdict.
CI gate ratchet and documentation update .github/workflows/ci.yml, docs/notes/corpus-benchmark.md, docs/proposals/P-012-bug-corpus-mining.md	--min-recall is raised from 3 to 4 in the corpus-benchmark job. Both documentation files are updated to reflect the new 4/9 recall, explain the fixture-driven ratchet, and enumerate the remaining 5 genuine frontend extraction gaps.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• PhysShell/Own.NET#50: Modifies the same corpus-benchmark CI job's scripts/benchmark.py invocation by setting the --min-recall gate to 3, which this PR ratchets up to 4.

Poem

🐇 A stub here, a stub there,
The ViewModel's events now declared with care.
OWN050 no more — the leak is found!
From three-of-nine to four, recall is sound.
The gate is raised, the benchmark proud,
This bunny hops and cheers out loud! 🎉

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 14.29% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main change: increasing corpus recall from 3/9 to 4/9 by fixing the screentogif-loaded-subscription fixture to be self-contained, with explicit reference to the proposal (P-012).
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/recall-loaded-subscription

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}