Files
ARCADIA-Studio/harness/experiments/scenario-desktop-focused-20260913/README.md
T

4.1 KiB

Scenario desktop focused gates — 2026-09-13

Completed Windows evidence for aed07176 and its one-line test correction 694efd6e. This archive establishes the results below, not a full desktop-suite or native acceptance result.

Source / stage Actual result Original records
aed07176 — desktop all-targets check Failed, Cargo exit101, E0609 in navigation_union_tests.rs:443 Log, receipt
694efd6e — desktop all-targets check Passed, Cargo exit0; 584 inputs unchanged Log, receipt
694efd6e — desktop all-targets test listing 965 distinct actual test names; listing is not execution of those965 tests Log, receipt and full membership
694efd6e — focused desktop tests 12 passed,0 failed,0 ignored,953 filtered out Log, receipt, exact selection
694efd6e — full suite, strict Clippy, application build NOT RUN after replanning No receipts or result claims for these stages

The failed check attempted expected.project on Arc<CompiledSourceProject>. The correction reads expected.decoded.project; it changes only the navigation/history regression, with no production change between these pins. The corrected regression was included among the12 executed tests. The selection also covers saved Scenario dispatch/destination refusal and SVG export boundaries. This does not establish saved Scenario renderer support: the later combined renderer source68a81c3 adds separate geometry/typed-activation work, and its combined gate supersedes the unexecuted694 full-suite/Clippy/build stages. This archive contains no result for that later source.

Independent verification binds37 original artifacts byte-for-byte, including scripts, preparation notes, package-clearance provenance, logs, receipts and input captures. All receipt-bound script/input hashes and prepared-script hashes were checked. Three aed captures and seven694 captures each retain exactly584 identical records within their attempt. Across attempts only the corrected test file differs. A fresh enumeration and hash read of the current694 worktree also matched all584 records. Raw listing and focused result lines independently match their receipts and exact12-test selection. Receipts record completed processes and terminal exit codes; no gate was launched while archiving.

The original runner records locked/offline arguments, sequential stage dependencies, explicit test names, input pre/postflight and timestamps. The capture script defines the exact584-input scope. Original preparation READMEs and JSON still say “prepared” and describe planned stages; those historical bytes are retained, while the table above states actual completion. Machine-local paths inside original records identify the execution environment; portable evidence links are relative here, and code attribution uses pinned repository links. Running the archived scripts unchanged is not a portable rerun procedure.

Raw records use local -text attributes to preserve their original bytes in Git. Existing core-gate evidence, native navigation/rename evidence and future combined renderer gates remain separate acceptance scopes.

The test executable binding records bytes, SHA256 and mtime after execution, before parent-owned package clearance. Both raw list/focused logs name this exact test executable. This is not a pre-execution digest and is not an application/GUI build claim.