case151 — Provider-agreement matrix (corroboration grows with evidence)¶

Field	Value
Verdict	🟡 COMPATIBLE_WITH_RISK
Category	Risk
Platforms	Linux
Flags	Bad practice
Detected `ChangeKind`s	—
Source files	`examples/case151_xcheck_provider_matrix/`

Verdict: 🟡 COMPATIBLE_WITH_RISK · Cross-check: private_header_leak · Mode: single-release audit · Evidence tier: L2

What it demonstrates¶

"Better results from the combination" as a measurable output property. The same PRIVATE_HEADER_LEAK finding is recorded with a different provider list depending on how much evidence is available — the §6.8 provider-agreement matrix.

Fixture	Evidence present	Providers recorded for the finding
`thin.abi.json`	public-header AST only	`public_header_ast` (1)
`snapshot.abi.json`	header AST + L5 source graph	`public_header_ast`, `source_index` (2)

Both fixtures flag the same leak — the finding does not change — but the rich fixture's source graph corroborates it with a second, independent provider. That provider list is the available corroboration signal ScanResult records.

Scope. This case asserts the provider list differs. Deriving a per-finding confidence tag from the provider count (so 1-provider corroboration renders a weaker tag than 2) is a separate reporting enhancement, not part of this corpus.

Reproduce¶

abicheck scan --binary libdemo.so -H include/ --audit                 # thin: 1 provider
abicheck scan --binary libdemo.so -H include/ --audit --sources .     # rich: + source_index

Fix¶

Same as any private-header leak (see case144): opaque-handle the internal type or install its header.

Source files¶

snapshot.abi.json
thin.abi.json