26 lines
1.5 KiB
Markdown
26 lines
1.5 KiB
Markdown
# Verifier control evidence
|
|
|
|
The independently reviewed verifier accepted the unchanged real archive and
|
|
refused all 15 corrupted copies. [Review](REVIEW.md) describes each actual refusal,
|
|
the preserved initial timeout and the limits of this consistency evidence.
|
|
|
|
`summary.json` indexes all 16 closed cases. Each `cases/NAME/` retains the exact
|
|
control result, payload manifest and actual stdout/stderr. The separate
|
|
`interrupted-known-failure-cause/` preserves the timed-out attempt's manifest and
|
|
streams; it has no closed result and is not counted as a successful control.
|
|
The initial and successor drivers and successor aggregate are preserved exactly.
|
|
Historical paths in those receipts identify the original local execution.
|
|
|
|
Corrupted ZIP copies remain local scratch and are omitted here to avoid repeating
|
|
the complete evidence 17 times. The drivers define the mutations; their payload
|
|
manifests retain the exact changed member hashes. The recorded ZIP hashes identify
|
|
the original generated copies, not files supplied in this directory. Raw source
|
|
gate evidence is in the adjacent original evidence.zip. None of these controls
|
|
executed archived helpers or the subject test binary.
|
|
|
|
The successor changes only the timeout from 180 to 900 seconds and selects the
|
|
remaining thirteen cases. Its result is combined with the first three closed
|
|
controls; the interrupted attempt remains distinct. All 49 watched original
|
|
files were unchanged at control closure. Later additions of these publication
|
|
notes do not change the original evidence.zip, manifest.json or verifier.
|