# Semantic integration verification

Status: verified Windows integration. Merge `332e46b`, formatting correction
`c6db9d4`. This checkpoint does not complete native acceptance, structured semantic
review or the release. The lead retains overall state and native ownership.

## Inputs and preserved work

Merged frozen semantic branch `6ad3cb92df21a5e67c318e31c751fc6a61555ef6` into
DATA `1a07475` (combined direct process/scenario production `60d162c`). Every Rust
change merged automatically. Only two working-memory documents conflicted: both
architectural decisions were retained; DATA's `STATE.md` was preserved unchanged
for the lead's next overall checkpoint.

The integration contains source/schema 5 and explicit historical migration, the
closed typed semantic review core, first equal-constructor acknowledgment, three
REST query/preview/apply operations, current OpenAPI/reST, and scalar desktop
review milestones 1–2. It preserves process continuation, direct scenario editing,
stable scenario viewport, compact controls and recursive value browsing.
Constructor/keyed/ordered/reset/multi-source desktop authoring remains outside the
scalar milestone. No native input or external deployment was performed here.

## Windows verification

Windows, Rust/Cargo 1.98.1. Exclusive target:
`D:/CodexBuilds/Capella2-data-target`. Logs and generated check artifacts:
`D:/CodexBuilds/Capella2-data-target/semantic-integration`.

Run from `D:/CodexBuilds/Capella2-data-integration`:

```powershell
$env:CARGO_TARGET_DIR='D:/CodexBuilds/Capella2-data-target'
cargo test --locked --offline --workspace
cargo clippy --locked --offline --workspace --all-targets -- -D warnings
cargo build --locked --offline --workspace --release
cargo fmt --all -- --check
```

All gates passed. Workspace tests: **718 passed, 0 failed, 1 intentionally ignored**
inheritance scaling probe across 39 test summaries: 353 core, 317 desktop, 48 server.
This includes the integrated scenario/process/value-browser regressions, all 11
scalar semantic desktop cases, source5 migration/writeback and real HTTP operations.
Strict all-target workspace Clippy had no warnings. Optimized workspace build passed
in 2m11s. Logs: `workspace-tests.log`, `workspace-clippy.log`, `workspace-release.log`.

The initial formatting check found only the order of two HTTP test module imports.
Commit `c6db9d4` applies rustfmt to that file; the repeated whole-workspace check
passed (`workspace-fmt-after.log`). No production-code correction was necessary.
The ordinary unprivileged Python launcher was denied by the local environment;
the authorized scoped verification process succeeded. No blocker remains here.

## Normative and actual service contracts

With existing Docutils 0.23 and JSON Schema libraries:

```powershell
$env:PYTHONPATH='D:/CodexBuilds/Capella2-docutils;D:/CodexBuilds/Capella2-contract-tools'
python crates/oa-server/tests/semantic_schema.py --check
python crates/oa-server/tests/semantic_http_contract.py D:/CodexBuilds/Capella2-data-target/release/arcadia-server.exe
```

The closed generated catalog check passed: 13 families, 84 scalar leaves and 184
schemas match checked-in OpenAPI. The HTTP script launched the optimized service
on loopback with an isolated temporary store and terminated it after verification.
**184 schemas valid; 19 actual/typed values checked against served and checked-in
contracts; 9 obsolete/malformed shapes rejected.** It verifies import/migration,
query, preview, apply, typed baselines, one-sided selectors and stale rejection.
Log: `actual-semantic-http.log`. This is actual executable transport evidence.

Strict `docs/language.rst` and `docs/api.rst` generation passed with
`docutils.core.publish_file`, `writer_name='html5'`, and settings
`{'halt_level': 2, 'report_level': 2}`. Outputs went into the evidence directory.
Both generated documents exactly match tracked HTML after line-ending
normalization. The full core suite also parsed the complete normative examples
and exercised typed historical source/JSON, included-file writeback and reopen.

## Frozen binaries and handoff

Both optimized executables are frozen in
`D:/CodexBuilds/Capella2-data-target/semantic-integration/freeze-c6db9d4`:

- `arcadia-studio.exe` SHA256
  `38AF477646BCC7096E5C71D95D827BF24F9E7ECA1E310DCD7D4AD258ED9E55A0`
- `arcadia-server.exe` SHA256
  `C420DFBC30FDA6935E306DF12C310739066389066B878F44D4470E5DD5B1529E`

The directory also contains `SHA256.txt`. No desktop process was launched. The
build target is idle and returned to the lead. Linux integrated checks, native
semantic acceptance, independent scalar-review follow-ups and the structured
semantic desktop milestone remain separate required work.

## Scalar review correction integrated on Windows and Linux

The lead's complete Linux baseline run at `d114a5d` is confirmed from the logs in
`D:/CodexBuilds/Capella2-data-linux-target`: **721 passed, 0 failed, 1 intentionally
ignored** inheritance probe, across 39 summaries. Strict workspace all-target
Clippy and the optimized workspace build also completed successfully. The source
logs are `semantic-workspace-test.log`, `semantic-workspace-clippy.log` and
`semantic-release-build.log`. These were root-run checks, followed by this task's
log inspection; they are not newly rerun workspace checks after the scalar fix.

Cherry-picked the reviewed scalar correction `5a51a7572e8edee6b0bb85fe6aded3a8380e0edd`
as `e7fa061`. The affected forms and semantic files merged automatically, preserving
DATA's recursive browser and direct behavior changes. The Linux checkout fetched
DATA locally and fast-forwarded to the exact same commit. No core or server code
changed. The correction preserves unfinished scoped inputs during structural edits,
uses the combined review candidate for dependent controls, and disambiguates keyed
fact headings with qualified paths and identity. See `release-semantic-gui.md`.

Post-fix verification passed on both platforms: **322 desktop tests passed, zero
failed or ignored**; strict all-target desktop Clippy; whole-workspace formatting;
and optimized desktop builds. Windows build took 1m38s; Linux took 1m33s. The five
additional test functions and expanded cases exercise scoped draft retention,
dependent type/storage/mechanism controls, exact source/history and qualified
fact headings alongside all existing desktop tests.

Commands used on Windows from DATA, with the existing assigned target:

```powershell
$env:CARGO_TARGET_DIR='D:/CodexBuilds/Capella2-data-target'
cargo test --locked --offline -p oa-desktop
cargo clippy --locked --offline -p oa-desktop --all-targets -- -D warnings
cargo fmt --all -- --check
cargo build --locked --offline -p oa-desktop --release
```

Matching commands inside Ubuntu 24.04 WSL:

```sh
cd /mnt/d/CodexBuilds/Capella2-data-linux-review
export PATH=/home/sorokinoleg/.cache/arcadia/rust/bin:$PATH
export CARGO_TARGET_DIR=/mnt/d/CodexBuilds/Capella2-data-linux-target
cargo test --locked --offline --config target/linux-cargo.toml -p oa-desktop
cargo clippy --locked --offline --config target/linux-cargo.toml -p oa-desktop --all-targets -- -D warnings
cargo fmt --all -- --check
cargo build --locked --offline --config target/linux-cargo.toml -p oa-desktop --release
```

Logs share the names `scalar-fix-desktop-test.log`,
`scalar-fix-desktop-clippy.log`, `scalar-fix-fmt.log` and
`scalar-fix-desktop-build.log`. Windows logs are in the target's
`semantic-integration` directory; Linux logs are directly in its assigned target.
Both command sessions completed with exit status 0.

Separate native trial directories under
`D:/CodexBuilds/Capella2-release1-01a06e2e` now contain frozen `e7fa061` executables:

- `ui-semantic-windows-e7fa061/arcadia-studio.exe` SHA256
  `3E46F001711948FB60E3F65C49810DF5C177228D5DFA609ED9DF3F8C96081F34`
- `ui-semantic-linux-e7fa061/arcadia-studio` SHA256
  `94C39037AF47C95FA6EE6515C3EB3B4FFBAC59730286E6EDB5F14B16E7874826`

Each has a `verification.json` manifest. Both frozen hashes equal their build
outputs, and Linux executable access was verified without launching it. Both
checkouts and build targets are returned to the lead. No native input was used,
and no STATE, TASK_LEDGER, README or other native ledger was staged by this task.
Scalar native acceptance and the remaining structured review milestone still
belong to the lead's release plan.