﻿# UtiliVera Duplicates 0.9.0 engine-test evidence

- Generated UTC: `2026-08-15T12:55:35.6594050+00:00`
- Frozen EXE: 353,792 bytes
- File/Product version: `0.9.0.0`
- Frozen EXE SHA-256: `5F59977C69A25CCCB9569D1ED3F4F2D8043E2336CCFCD4C14933A623C754D727`
- Command: `powershell -NoProfile -ExecutionPolicy Bypass -File run-tests.ps1`
- Result: **73/73 PASS**, exit code 0

## Evidence boundary

The suite recompiles the current audited engine and test sources listed below. It does not claim to introspect compiled source from the EXE. Frozen EXE command-line behavior and installed-payload identity are separately exercised by `installer-test-evidence-0.9.0.json`; the independent round-eight report records the clean-copy 73/73 run and frozen CLI checks for this exact EXE hash.

## Tested source identity

| Relative path | Bytes | SHA-256 |
|---|---:|---|
| `src/DuplicateEngine.cs` | 83642 | `456A2CA03718B2D539335B302D28EE8ADC2C27163DFE5DA555CA6B3DB57F2E86` |
| `src/SafeCleanup.cs` | 19282 | `BC85574E719847F6939EC4900D23479C95C0CF92047C4D08E47F08AFE3683B1C` |
| `src/CommandLine.cs` | 20307 | `AB8FB3C84B9C01CF51ED2BFA050960D55262E508E3C2EFC4EC77C1F852806332` |
| `src/SessionStore.cs` | 3231 | `90226B19BAE1111A286C3C6D7F0CA3CB2168BE98C2B1E96A28E9E227C3959311` |
| `src/ProfileStore.cs` | 1345 | `6D4426AD38CBE7C2B9F4BA7B489CAE2D0EFFD98893A9CD33363603E522870C24` |
| `src/AppSettings.cs` | 3937 | `A82C8ADB5A2979CF1871CC741E8D1EADD5780B267C0FDF8C7384007426F4482F` |
| `tests/EngineTests.cs` | 41469 | `C5487E490543CF5307ABC559F5675F34C8166452EC9C0736797C4EB274435885` |
| `run-tests.ps1` | 1220 | `F5872A3FD7917724AA12EEEAD98F9FC37D5E774C865649B9217A5E1D4B45CF6D` |

## Complete named result output

```text
PASS quick hash same content
PASS SHA-256 same content
PASS byte verification same content
PASS byte verification rejects collision candidate
PASS enumeration honors excluded folder
PASS content and zero-byte duplicate groups
PASS different names group by content
PASS reference folder protection flag
PASS hard links do not inflate reclaimable bytes
PASS scan progress reporting
PASS planner keeps reference copy
PASS cleanup candidates reverify byte equality
PASS planner never removes every copy
PASS selection assistant explains and keeps newest copy
PASS selection assistant keeps oldest copy
PASS selection assistant keeps shortest preferred path
PASS selection assistant keeps a name-pattern match with reason
PASS cleanup plan de-duplicates physical identities
PASS verified quarantine transaction
PASS cleanup retains protected original
PASS ledger restoration and integrity
PASS extension filter
PASS clean corpus has no scan errors
PASS bulk cleanup spans all exact groups and remains fully restorable
PASS verified hard-link replacement reclaims space while retaining both paths
PASS hard-link replacement is fully reversible from its recovery ledger
PASS hard-link crash fixture is created
PASS pending hard-link crash rolls back without data loss
PASS post-link crash fixture is created
PASS post-link pre-ledger-flush crash restores an independent copy
PASS perceptual hash survives resize and JPEG encoding
PASS perceptual hash separates different composition
PASS similar-image groups are review-only and never claim automatic savings
PASS similar-image mode finds true 90-degree rotations
PASS similar-image clustering rejects transitive bridge false positives
PASS approximate matches cannot enter automatic cleanup
PASS same-name review mode groups different content without enabling cleanup
PASS empty-file audit mode isolates exact empty duplicates
PASS unique-file mode verifies content before reporting unique files
PASS duplicate-folder mode verifies recursive trees and file hashes
PASS archive mode finds identical content across ZIP-based formats
PASS archive mode reads duplicate content inside native 7Z archives
PASS archive mode reads and hashes content across complete multi-volume RAR sets
PASS audio fingerprint ignores differing ID3 metadata
PASS audio fingerprint ignores differing ID3v1 tail metadata
PASS FLAC fingerprint ignores metadata blocks
PASS Ogg fingerprint ignores Vorbis comment packets
PASS audio metadata mode matches artist and title across different payloads
PASS decoded acoustic fingerprint matches the same audio across sample rates and channel layouts without grouping a different tone
PASS video review mode rejects duration-only false positives with different payloads
PASS video fingerprint ignores filename and container-header metadata
PASS video perceptual-frame Hamming distance is exact at threshold boundaries
PASS media metadata reports WAV codec sample rate and channels
PASS media metadata reports a stable video codec fallback
PASS advanced-name mode ignores copy suffixes, punctuation and case
PASS 500-file parallel scale corpus remains exact
PASS persistent hash cache accelerates unchanged candidate files without changing results
PASS cache invalidates changed size and timestamp safely
PASS hash cache prunes stale paths after a successful scan
PASS hash cache can be cleared explicitly
PASS scan pause and resume preserves the session and result set
PASS 1024-plus-character paths are discovered and verified
PASS 1024-plus-character paths complete quarantine transaction
PASS 1024-plus-character quarantine ledger restores original path
PASS 1024-plus-character paths complete verified hard-link replacement
PASS 1024-plus-character hard-link ledger restores independent copies
PASS paths beyond the Windows 32K limit produce an explicit scan error
PASS application data and quarantine tree are self-excluded
PASS settings round-trip preserves Unicode-safe modes, storage policy and locations
PASS settings fall back to the durable backup after corruption
PASS named profile migrates saved modes, storage policy and reference roots
PASS JSON result session reloads and revalidates unchanged files
PASS JSON session detects drift and disables automatic cleanup
TOTAL PASS 73
```
