# Change log — review pass (rev 1 → rev 2)

Original submission preserved byte-for-byte at `output/original_submission/`
(before any review edits; captured by `snapshot_v1.py`).

## Site code — unchanged (no corrections needed)
`review_hash_compare.py` confirms `output/site/{app.js,index.html,style.css,server.py}`
are SHA-256 identical to rev 1. Supporting checks re-run during review:
- `test_server.py`: all 14 handler cases pass again (3×201, 10×400, old endpoint 404; stored records = valid payloads only).
- `verify_copy.py`: server copy identical to input `site/server.py`; Decimal recomputation of all four brief examples passes (268.80 / 537.60 / 725.76 / 1935.36).
- New browser checks (trusted server): whitespace-only name rejected ("Please enter your name.", 390 px); 4 seats = CAD 967.68 (1280 px); seats "08" accepted as integer 8, success shown and record stored as `seats:8` (1280 px); post-cleanup smoke load 390 px shows CAD 725.76 with overflow false.

## Corrected: test_server.py (data-safety defect found during review)
Rev 1 deleted any pre-existing `output/site/submissions.jsonl` at start and end.
During review this destroyed the newly written "Edge Case" browser record before
it could be preserved; it was recovered from the rev-1 snapshot, which had
captured it moments earlier. Rev 2 backs up pre-existing bytes and restores them
in a `finally` block, removing only its own test records. Restore path explicitly
tested: staged a copy of the evidence file in `output/site/`, re-ran the script,
got "pre-existing submissions.jsonl restored unchanged" and byte-identical content.

## Corrected: preserve_evidence.py (robustness)
Rev 1 crashed when `output/site/submissions.jsonl` was absent and overwrote the
evidence file. Rev 2 is idempotent (no-op when absent) and appends with
line-level dedupe. Re-run verified.

## Evidence updated
`output/evidence/browser_submissions.jsonl` now contains 3 valid records
(added Edge Case, seats 8, merged from the snapshot copy via
`review_merge_evidence.py`). Delivered `output/site/` re-cleaned to exactly
4 files. `output/handoff.md` updated: review-pass browser table, corrected
persistence-evidence description, leading-zero ("08") behavior noted, review
outcome section added.

## Added helper scripts (documentation of review actions)
`snapshot_v1.py`, `review_merge_evidence.py`, `review_hash_compare.py`.
