2.5 KiB
Prepared Scenario desktop gates
Pinned source: aed07176ab in .local/scenario-layout-ytr83r-20260913. Target: repository root target, with debug output. All Cargo stages use locked/offline mode and execute from the pinned Scenario worktree. These scripts have not executed a desktop gate.
inputs-complete-before.json already contains an independent fresh hash capture of all 584 files; every record matches the completed core attempt3 after-manifest. Do not rerun capture -Stage before into this directory. The capture script re-enumerates the complete same scope before and after each stage and refuses count, path, byte, hash or commit changes.
Parent owns package clearance. run.ps1 requires package-clearance.json with base, cargoExitCode=0, resolved absolute target, dependenciesRetained=true. The provided parent clearance is accepted through its target field. No files-removed minimum is required. No script deletes artifacts or invokes cleaning.
Dispatch one stage at a time in this order:
- run.ps1 -Stage check
- run.ps1 -Stage list
- run.ps1 -Stage focused
- run.ps1 -Stage all
- run.ps1 -Stage clippy
- run.ps1 -Stage build
Each later stage requires the preceding stage's successful receipt and unchanged inputs. Check uses --all-targets --verbose; list and all use the identical desktop --all-targets scope. Focused uses the 11 full names from expected-tests.json with --exact and one test thread. No --nocapture is used, preserving deterministic harness result-line parsing; failure details remain in the raw test output. All also runs one test thread and compares every executed/ignored test identity against the actual listing. No total desktop test count is assumed.
Expected new membership is 6 dispatch +5 SVG export: 7 tests in app::saved_scenario_dispatch_tests (5 dispatch and2 caller), 1 in workflow::saved_scenario_destination_tests, and3 exact diagram::tests SVG tests. Both list and focused require all11 exact names. Focused also requires exactly11 executed/pass results and zero failures/ignored. Existing SVG/view-options tests are covered by all.
Raw per-stage logs, argument arrays, timings, exit codes, script/selection/clearance/baseline hashes, before/after captures, listing and execution memberships are retained. Build additionally records executable bytes/hash/mtime. Files are never replaced; failures must retain their directory and use a newly prepared attempt for rerun. Script creation is preparation, not compilation or test acceptance. No GUI is launched.