Files
ARCADIA-Studio/harness/experiments/analysis-mixed-cli-20260925/RUNTIME-PLAN.md
T

5.4 KiB

Mixed CLI runtime reconciliation plan

Parent owns Cargo, executable runs and the eventual checker invocation. inspect_runtime.py is data-only: raw JSON/log/ZIP/file reads plus read-only Git commands. It must not execute archived helpers, compiler or product binaries and refuses overwriting its final readback. Do not invoke before final gates are complete.

Baseline: public-read source bab6f96a5b, immutable archive SHA256 1c20b7bffdaee9b1139b0e922d4366f6ac739f8567040e8f4723dc19e3974656. It proves299 core outcomes,10 public-read external tests,65 selected ordinary integration results and three doctest passes (with the preserved narrow historical recorder spelling exception). Existing source_cli's12 names come from baseline source only: this checkpoint newly executes that suite, so no prior passing-runtime claim is made for it.

Require exact full302 names/outcomes: all299 baseline plus the three authored mixed_discovery names. Require focused mixed_discovery exactly3. Eleven integration list/run pairs: new mixed_source_files5, mixed_source8_read10, seven ordinary selections65, source_cli12 and mixed_cli6. Compare names, raw outcomes and artifacts, not totals alone. Require three docs and the exact two existing compile-fail name mappings; current recorder must succeed. Require core+CLI strict Clippy, inherited fmt bytes, actual CLI build, ordinary check/rename-preview2 and mixed validJSON/unresolvedJSON/validhuman3 against that same built binary.

The final checker verifies each declared input against immutable Git and current raw bytes, launcher template/source pin, environment, command/filter/package, receipt/raw log/hash, artifact origin/current bytes and source cleanliness. Preserve historical receipts separately without treating their overwritten binaries as current. The reviewed executable helpers are a8ae7e9c4cb15a87720e364772c3239eef1fe6f4753b1f74b1f9e5482c7bcdad (ordinary) and78b612d631175032c7a7819105e03262e21630a1139be62ae78bbe9b5d387a7b (mixed). The mixed helper's UTF-8 string literals may be read using AST literal evaluation; neither helper is imported or executed by the checker. Reconstruct expected whole reports from authored inputs and preserve exact files/directories/streams.

Initial f232482 source evidence must remain: six integration receipts, mixed_source_files5/5 and mixed_cli6/6, plus mixed_cli Command glob-shadow warning and source_cli10/12 with only complete_alarm_example_preview_then_write_and_fixed_point and rename_shows_missing_identity_and_separate_source_migration failing. The corrected tests retain full-text/model/fixed-point oracles and align only historical header expectations to current schema8. Record final exact repair/source hash; do not normalize the initial failures into passes.

The initial recorder recipe omitted source_cli's actual include_str dependency harness/experiments/textual-authoring-20260911/alarm-response.arcadia. Parent added exactly this input and pinned final f10490b13b. Final helper SHA25615d3bae6c4235f20a4b5ae523e9cde38842118397bf1accbf5837fb18de356b4, normalized template0349a19e62b9c57231e6ef3abad30643f39d7ed2fa4ac1fcab4b314159a39d52, declares567 inputs. Preserve original512dcc82-template receipts with566 declarations; any separate historical Git-blob confirmation is not a retroactive raw-input capture.

Exact old stack controls

The baseline299 inventory contains eleven named core stack controls. Require each actual final outcome to pass:

  • language::analysis_source::index::tests::deep_parse_index_plan_and_failed_build_cleanup_on_128k_stack
  • language::analysis_source::index::tests::maximum_index_paths_compare_lookup_debug_and_drop_on_128k_stack
  • language::analysis_source::mixed::project::ordinary_index::rename::emission::acceptance::tests::scoped_small_stack_actual_deep_wide_verification_and_drop
  • language::analysis_source::mixed::project::ordinary_index::rename::emission::tests::scoped_small_stack_changed_unchanged_and_failure_drop
  • language::analysis_source::mixed::project::ordinary_index::rename::tests::deep_and_wide_equality_and_late_refusal_drop_on_small_stack
  • language::analysis_source::mixed::project::ordinary_index::tests::small_stack_success_and_late_header_failure_leave_no_partial_owner
  • language::analysis_source::mixed::project::tests::index_tests::success_and_late_failure_drop_on_small_stack
  • language::analysis_source::mixed::project::tests::inherited_storage_failure_and_success_drop_deep_syntax_on_128k_stack
  • language::analysis_source::mixed::tests::mixed_deep_typed_success_and_rejection_cleanup_on_128k_stack
  • language::analysis_source::tests::every_parent_shape_rejects_depth129_and_cleans_children_on_128k_stack
  • language::nesting_tests::deep_declaration_parsing_and_positions_use_a_bounded_rust_stack

Also require deep_structural_writeback_runs_on_a_small_thread_stack in actual source_nested_updates integration results. These establish their finite exercised scopes; no universal stack headroom, setup-stack qualification or Linux/depth128 claim follows. Any parent-captured static frame evidence is supplemental and bound to its actual binary, not substituted for worker execution.

No Linux, GUI/edit/save/history adoption, atomic snapshot or universal stack/RSS claim follows. Missing records remain incomplete gates; preserve evaluator problems before correcting them. Parent decides acceptance after independent readback.