Scenario validation gate at b10b28c
Prepared for source b10b28cefd1d0794073ba241a0c8de9a83e85f68 in C:/Users/User/Documents/Git/ARCADIA-Studio/.local/scenario-spacing-ytr83r-20260913, using parent-owned C:/Users/User/Documents/Git/ARCADIA-Studio/target. This directory contains a new gate recipe, not passing test evidence. Parent dispatches each stage after its predecessor closes successfully. Preparation invoked no Cargo, rustc, test executable, cleanup, or GUI.
Source and inventory
Independent Git and rg enumeration agrees on 612 files in crates, docs, examples, build, harness/fixtures, harness/experiments/textual-authoring-20260911, and root Cargo.toml/Cargo.lock/rust-toolchain.toml. expected-inputs.json pins their actual bytes; external-inputs.json separately binds toolchain/env.ps1 and presence/absence of relevant root, worktree, ancestor and Cargo-home .cargo configuration. These are the declared captured scope, not a claim to hash every possible tool or registry input.
The previous actual8bf listing has1089 unique names. The new committed app::scenario_details_tests module adds six exact names; expected-tests.json therefore requires a fresh1095-name listing, 301 focused passes, and the same four explicitly named ignored tests in full execution. Expected counts are source-derived until the stages run. No old executable is accepted as this source's executable. source-provenance.json preserves the old listing identity, failed strict provenance, and the one-line b10 cfg(test) correction.
Ordered commands
Run in separate PowerShell processes, sequentially:
pwsh -NoProfile -File C:/Users/User/Documents/Git/ARCADIA-Studio/.local/scenario-validation-gates-b10b28c-20260913/strict-preflight.ps1
pwsh -NoProfile -File C:/Users/User/Documents/Git/ARCADIA-Studio/.local/scenario-validation-gates-b10b28c-20260913/run.ps1 -Stage list
pwsh -NoProfile -File C:/Users/User/Documents/Git/ARCADIA-Studio/.local/scenario-validation-gates-b10b28c-20260913/run.ps1 -Stage focused
pwsh -NoProfile -File C:/Users/User/Documents/Git/ARCADIA-Studio/.local/scenario-validation-gates-b10b28c-20260913/run.ps1 -Stage all
pwsh -NoProfile -File C:/Users/User/Documents/Git/ARCADIA-Studio/.local/scenario-validation-gates-b10b28c-20260913/run.ps1 -Stage clippy
pwsh -NoProfile -File C:/Users/User/Documents/Git/ARCADIA-Studio/.local/scenario-validation-gates-b10b28c-20260913/run.ps1 -Stage build
Strict preflight runs desktop+server all-targets Clippy with locked/offline dependencies and -D warnings. It takes the baseline exactly once, captures612 inputs before/after, records toolchain versions once, and requires no test executable. It is a cheap early check, not a substitute for the later strict stage. A failed preflight prevents LIST admission. Existing evidence is never overwritten.
Regular LIST runs cargo test -p oa-desktop --all-targets --locked --offline -- --list. Its actual Cargo Running line establishes the single test executable path/hash/bytes. FOCUSED uses three bounded exact-name invocations of cargo test -p oa-desktop --bin arcadia-studio --locked --offline -- --exact ... --test-threads=1. Their selections are107,101,93 names (conservative command lengths11883,11906,10844, all below12000). Actual per-process filtered expectations are988,994,1002. Aggregate focused.log is byte-concatenated raw batch logs; there is no fabricated single-process794-filtered summary. Each batch must execute its exact unique names once, pass all, and preserve the listed executable and source/recipe inputs. The union must be precisely301 unique selected tests.
ALL runs desktop --all-targets serially; it requires exactly the fresh1095 listed outcomes, the four named ignores, no failures, and zero filtered. CLIPPY covers desktop+server all targets with -D warnings. BUILD creates the desktop GUI binary and binds its actual bytes, while retaining the exact test executable chain. Every regular stage verifies its exact prior receipt, raw output, source captures, transitive recipe hashes, toolchain record, and executable identity. Each before/after capture rechecks frozen Git identity, clean worktree, membership and file hashes.
Retained evidence and limits
No server test rerun is requested: actual57 passing server results at37be410 are retained in source-provenance.json. Core/server production and manifests remain byte-identical to37; server/tests/http.rs only reorders an unchanged module declaration. Both strict stages still check server all-targets. This does not claim new server execution.
Old8bf full/failed-strict and5d4 failed-preflight records remain in their original directories and are referenced by hash; no prior logs or receipts were copied as new results. The5d4 preflight failed Cargo101 on the now-test-only Scene::sequence wrapper with612 unchanged inputs. No GUI/native acceptance follows from gate preparation or from an earlier pin.
verify-preparation.ps1 performs AST checks and17 synthetic checks against the actual list/aggregate parser and batch helpers, including old-inventory, missing/duplicate/ignored outcomes, incorrect filtered counts and duplicate batch union. preparation-verification.json records that preparation only; it does not run a compiler or product test. Do not rerun it over existing evidence. All static recipe files are frozen before regular LIST.