Files
ARCADIA-Studio/harness/release-navigation-440-windows-20260913.md
T

88 lines
5.5 KiB
Markdown

# Navigation and selection: closed Windows gates at 440
The integrated source **44095f78c28b79d6d085863dc928adba8e24d955** passed the
following actual Windows checks. These results advance automated verification;
native and package acceptance remain pending.
| Gate | Actual closed result |
| --- | --- |
| Desktop listing | 1098 tests, one identified test executable |
| Desktop full | 1095 passed, 0 failed, 3 explicit manual cases ignored |
| Desktop strict all-target Clippy | Exit 0, no diagnostics |
| Cross-feature navigation | 17 passed, 0 failed/ignored, 1081 filtered |
| Mounted preview-scroll group | 4 passed, 0 failed/ignored, 1094 filtered |
| Source-symbol group | 33 passed, 0 failed, 1 manual case ignored, 1064 filtered |
| Focused captured rename Escape | 1 passed, 0 failed/ignored, 1097 filtered |
| Core listing/full | 472 listed in 44 binaries; 471 passed, 0 failed, 1 manual probe ignored |
| Core strict all-target Clippy | Exit 0, no diagnostics |
| Server listing/full | 56 listed in 4 binaries; all 56 passed, 0 failed/ignored |
| Server strict all-target Clippy | Exit 0, no diagnostics |
| Matching GUI and server build | Exit 0; two Cargo-identified unoptimized development binaries |
The [archive](experiments/navigation-440-windows-20260913/README.md) preserves
original logs, start/end and exit receipts, commands, input maps, environment,
admissions and runner source. Its generated readback verifies that the 1098
desktop members are exactly the previously published 1082 plus 16 new selection
cases, all passing. Full and focused runtime membership matches the actual
listing and uses the same recorded executable. Core/server membership matches
per executable, including empty binaries; their full runs have no filtered cases.
The three desktop manual exceptions remain unexecuted by these gates:
- `app::source_rename::timing_tests::accepted_release1_prepare_commit_sync_and_confirm_timing`
- `app::source_rename::timing_tests::layout_timing_tests::accepted_release1_constructor_vs_reuse_through_next_highlight`
- `app::source_symbols::tests::source_symbols_100_modules_1000_definitions_10000_uses`
Their source annotations describe isolated Windows timing/process-memory trials.
The core exception, `inheritance_chain_validation_scaling_probe`, also explicitly
requires manual invocation. Linux-only fault injection is outside this Windows
listing. No manual performance result is inferred from these ordinary test runs.
The desktop test executable is 150,299,136 bytes, SHA256
`55fdef506646e65fbb965ceba3cc1b8e5911a8dd018c6ca71cdfb50feb017be6`.
The GUI build is 117,171,712 bytes, SHA256
`95f0c3e9ea1521dbc555e5330f26e105bd9065b9fb4207980f6d7ae62d2ef99a`;
the server build is 39,728,128 bytes, SHA256
`da3f42cfa70f7daf3ea714bc196c2365024e91584ef13cdf07c89cbcc227f9d3`.
Executables are identified here but not committed or delivered by this evidence
archive. This is not an optimized-release or packaged-binary claim.
Every test/lint gate captured 581 unchanged committed inputs and the same source
HEAD before/after. The build captured 615 unchanged inputs, adding scripts to
that scope. All captured bytes match both tested 440 and publication base
355f6d4d244e6b57fd89a8afd80a2dacdea69ba1. That equivalence is limited to the
recorded paths; AGENTS.md and evidence/documentation outside those paths may
differ. It does not transfer results to a later production change. The final
root worktree contained the authorized AGENTS.md update while retaining HEAD440.
The recorded environment is Windows x64 gnullvm, Rust/Cargo 1.98.1, LLVM-Mingw
Clang 23.1.1, jobs 1, no CARGO_TARGET_DIR override. Build artifacts record opt-level
0, no debuginfo, debug assertions and overflow checks enabled. Exact invocations
and tool identities remain in the original records. Reproduction at the tested
source uses the configured Windows toolchain:
```text
cargo test --locked -p oa-desktop --bin arcadia-studio -- --list
cargo test --locked -p oa-desktop --bin arcadia-studio -- --test-threads=1
cargo test --locked -p oa-desktop --bin arcadia-studio navigation_union -- --test-threads=1
cargo test --locked -p oa-desktop --bin arcadia-studio preview_scroll_tests -- --test-threads=1
cargo test --locked -p oa-desktop --bin arcadia-studio app::source_symbols -- --test-threads=1
cargo test --locked -p oa-desktop --bin arcadia-studio actual_captured_escape_closes_only_its_priority_owner_and_retains_worker -- --test-threads=1
cargo clippy --locked -p oa-desktop --all-targets -- -D warnings
cargo test --locked --offline -p oa-core --all-targets -- --list
cargo test --locked --offline -p oa-core --all-targets -- --test-threads=1
cargo clippy --locked --offline -p oa-core --all-targets -- -D warnings
cargo test --locked --offline -p oa-server --all-targets -- --list
cargo test --locked --offline -p oa-server --all-targets -- --test-threads=1
cargo clippy --locked --offline -p oa-server --all-targets -- -D warnings
cargo build --locked --offline -p oa-desktop -p oa-server --bins --message-format=json-render-diagnostics
```
The native Details-to-Preview offset loss documented in
[issue 27, comment 914](http://217.177.34.108:3000/sorokin.oleg.edu/ARCADIA-Studio/issues/27#issuecomment-914)
remains an acceptance finding at this source. Existing passing scroll tests do
not assert preservation of the pre-Details native offset before scrolling again.
No GUI images, native acceptance, later correction, package, source8 or Scenario
candidate acceptance is included here. Those contributions retain their own
source and evidence boundaries. STATE and release status are unchanged.