Files
ARCADIA-Studio/harness/experiments/scenario-native-motion-oracle-b10b28c-20260913/README.md
T

54 lines
7.5 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Native saved-motion comparator: b10 build and admission checks
The separately reviewed scoped-motion comparator built successfully against **b10b28cefd1d0794073ba241a0c8de9a83e85f68** and passed its original exact-mode and new scoped-motion self-checks. This archive establishes comparator admission behavior only. It contains no native action or saved native motion comparison and makes no independently exact native endpoint claim.
## Actual results
| Check | Result |
| --- | --- |
| Builder `-ValidateOnly` | 5 positives, 5 fingerprint refusals, 3 source-membership refusals; both scripts parse |
| Checker `-ValidateOnly` | 2 positives and 17 receipt/runtime refusals |
| Standalone `rustc --edition=2024 -D warnings` | Exit 0; empty compile log; 963 unique before/after inputs unchanged |
| Original `self-check` | Exit 0; all 7 named checks passed; input stability and receipt admission passed |
| New `motion-self-check` | Exit 0; 4 positive fractional cases and 26 refusal cases; input stability and receipt admission passed |
Build session was `82593`; both self-check commands ran sequentially in session `29149`. The [validation summary](original/validation-summary.json) preserves exact commands, source/binary/runtime identities and actual outcomes. Build compilation ran September 13, 2026, 15:22:22–15:22:25 UTC; exact and motion self-checks ran at 15:23:43–15:23:49 UTC. Complete closed receipts and raw logs are preserved under [build-b10-01](original/build-b10-01/build-receipt.json), [exact-self-check-01](original/exact-self-check-01/receipt.json), and [motion-self-check-01](original/motion-self-check-01/receipt.json).
Build proof hashes and every self-check output hash were verified after execution. Archive preparation verified all 30 copied originals (593,653 bytes) against their source bytes. No compiler or oracle was rerun during archival. Historical preparation wording inside the [original README](original/README.md) and preparation manifest is preserved; these later closed receipts establish the actual completed results.
## Contract and limitation
`scoped-motion BASE ACTUAL OUT VIEW_UUID TARGET_UUID FIELD DIRECTION` takes independently chosen view and target IDs, `lane.x` or `message.send_y`, and `increase` or `decrease`. It requires exactly one existing target in both admitted incident Projects, finite bounded endpoints and a finite strictly correctly signed delta. It restores only Actual's selected coordinate to Baseline's value and compares complete canonical Project serialization. All other scalars, signed zero, array order, semantics, occurrence addresses and saved layouts must remain exact.
Success explicitly states coordinate scope and direction accepted, with `independently_exact_final_coordinate=false`. Reported endpoint, delta and float bits are observations, never independent expected values. This does not establish motion distance or pointer tracking; combine it with native visual observations at two zooms and exact saved-Project `same` checks for Undo, Redo and reopen. Native camera, runtime, selection and history-stack behavior need separate evidence. Controller/core tests supply predetermined fractional arithmetic evidence; mounted/App tests supply integer displacement and transaction evidence.
Original exact `same`, `repair-c`, `lane`, `message` and `self-check` Rust paths were retained unchanged except early dispatch into the two new modes. The earlier spacing comparator's files remain untouched. Distinct build/run schemas prevent reuse of the older comparator's receipts for this executable.
## Prerequisites and omitted bytes
The [published-prerequisite mapping](published-prerequisites.json) links 70 exact prior artifacts matching the current build input capture. Their checked local published bytes match their sizes/hashes. This avoids duplicating unchanged GUI source captures, gates, verifier, environment and fingerprint records:
- [GUI build, native verifier, environment and original comparator ancestry at 04129a75](http://217.177.34.108:3000/sorokin.oleg.edu/ARCADIA-Studio/src/commit/04129a75f3c27c9488b86b797525c36887293c93/harness/experiments/scenario-comparators-gui-b10b28c-20260913/README.md).
- [b10 focused validation and LIST at b09d9d5](http://217.177.34.108:3000/sorokin.oleg.edu/ARCADIA-Studio/src/commit/b09d9d506347e9a00353ed86bedb98702a426142/harness/experiments/scenario-validation-focused-b10b28c-20260913/README.md).
- [b10 full desktop and strict results at f69897b](http://217.177.34.108:3000/sorokin.oleg.edu/ARCADIA-Studio/src/commit/f69897bbb85506c0e00ced4332e328606f4e98b1/harness/experiments/scenario-validation-full-strict-b10b28c-20260913/README.md).
The new build's complete before/after manifests are preserved despite equal content because their original hashes close this build receipt and contain the new comparator bindings. Transitive dependency/toolchain cache bytes are excluded; original metadata remains in those manifests. No transitive cache rehash is claimed for archival.
| Omitted artifact | Bytes | SHA-256 |
| --- | ---: | --- |
| New comparator EXE | 22,814,208 | `f5e78fe9ca3c865ab69e57303bebb4db28499dfebf54baf1dc1eb602d381a0db` |
| Adjacent `libunwind.dll` | 190,464 | `aa282fa47bb2c4e210cea5e470cdbdca58568f0be0bc73e404d86ab9cd95ef6a` |
| Existing desktop GUI | 116,721,152 | `7ece85c6012ae0ae75fc56ed3c3aadb8bc116448ad1782be422ef34a629dc3ce` |
| Existing core rlib | 100,198,044 | `aff76c3ac9fa142acb7dc962f0c04184a8871d16452f82a33e5bd44bf6c8f340` |
| Existing serde_json rlib | 2,702,534 | `8ce86ee2a06cc40c051f3ebe8aea24e7efd49bf46953b59ef0df16c18869c7d7` |
The [archive manifest](archive-manifest.json) identifies every copied original and the two excluded original-directory binaries. Local `.gitattributes` prevents any line-ending conversion of evidence, including source and recipes; force-add ignored logs when publishing.
## Reproduction on another machine
The [original recipe](original/README.md) records the exact host-specific command proposal actually used. Preserve the `oracle/main.rs` / parent fixture relationship for `include_str!`. Read the frozen [build](original/oracle/build.ps1) and [check](original/oracle/check.ps1) scripts before preparing separate local copies. Do not edit archived originals or run into archived output directories.
On another machine obtain production b10, its dependencies and compatible compiler/linker/runtime, then complete fresh local five-stage native gates using a reviewed locally adapted verifier. Supply fresh local `SourcePin`, `Gate`, verifier path/hash, runtime DLL, actual matching desktop-to-core-to-serde fingerprint-linked rlib paths and Windows import-library directory. Absolute paths, environment scripts and dependency filenames require explicit local adaptation; historical receipt hashes do not authorize relocated artifacts. Keep every binding and refusal guard. The runtime bytes remain explicitly pinned; a different runtime requires a separately reviewed recipe change, not weakened admission.
Run both `-ValidateOnly` modes, build into a fresh directory, review the resulting executable and receipt hashes, then pass those new literal hashes to the checker for both `self-check` and `motion-self-check` against the original fixture. Retain new receipts, proof manifests and raw logs. A reproducible source build can produce a different executable hash on a different host; never substitute it for this historical artifact. Actual scoped-motion checks additionally require independently planned direction/scope and captured native before/after files; self-check success does not supply those results.