3.0 KiB
Scenario reorder destination experiment evidence
Question, result and bounded decision. All new code remains under cfg(test); this checkpoint enables no production group reorder command and claims no native UI acceptance. Ordinary group spacing was delivered separately.
Tested source: cae7a22297d09810c412dbdf4ce5d8a28164b247. The three Windows GNU/LLVM captures use Rust/Cargo 1.98.1, stable-x86_64-pc-windows-gnullvm, one build job and CARGO_INCREMENTAL=0. Each captures 814 unchanged input files; exact inventory and Git-blob comparisons pass. Full environment, command, process, output and source-proof records are inside the archive.
cargo test -p oa-desktop reorder_ --message-format=json -- --nocapture
cargo fmt --all -- --check
cargo clippy -p oa-desktop --all-targets --message-format=json -- -D warnings
Tests: 22 passed / 0 failed / 0 ignored, 1,573 filtered, including all five session and four slot experiment functions. The other thirteen are existing name-filter matches; this is not full Scenario/desktop testing. Format and strict all-target desktop Clippy each exit 0. No Linux or native process was run for this test-only increment.
checks/evidence.zip: 3,244,466 bytes, SHA-256 b713a5c8c03f4cb4f4a6884bde7eb081fca00d1f5ebf85427ce14bc03acae166. Three closed runs, zero failures, 27 logical files, 23 stored members, one deduplicated source capsule and three exact source proofs. These are all attempts for this increment. Previous spacing failures remain in that contribution's separate evidence.
After checkout, with Python 3.11+:
python -B harness/scenario-reorder-interaction-evidence-20260924/checks/verify.py
The verifier checks every archive/member pin, captured source capsule, command/result relationship and source-proof receipt. It does not rerun Cargo or the original Git-blob comparison. Reused helper README examples and the manifest's scenario-selection-cargo-gates kind are historical naming; actual run names here are reorder-interaction-test-1, reorder-interaction-format-1 and reorder-interaction-clippy-1.
The independent review records the original specification ambiguities, bounded cache correction and static/runtime evidence audit. The slot experiment notes retain original expectations followed by actual executed findings. Preparation or “pending execution” statements in their earlier sections describe that earlier phase; later closed receipts and appended results supersede them.
The result rejects unchanged nearest-Y picking as the only method for choosing exact semantic destinations: admitted equal-Y slots leave one anchor unreachable. It does not accept a chooser's usability, reject every possible graphical handle, prove all nested/append destinations, establish Windows pointer quantization, or claim that offscreen targets cannot be reached after a camera change. See the report for remaining App and native requirements.