# Schema8 iterative type checker — local execution gate

Codex-956atf `/root`, Windows, 2026-09-11. The isolated working schema8 core at
`.local/analysis-core-worktree-20260911`, baseb96ff393, now passes its corrected
logical-depth gate, all28 tests selected by `analysis::`, and existing numeric26.
Strict core all-target Clippy with `-D warnings` also passes. This is static core
admission/type checking, not implemented source8/evaluator/REST/GUI parity.

## Failure and correction

The first warning-free test compile completed in3m34s. Sixteen selected tests
passed, then the required depth128 test aborted with Windows STATUS_STACK_OVERFLOW
`0xc00000fd`. Its executable SHA256 was
`548850D3606C9B4A58859677E86A1C1813EB6E71785C529F771E04125E7A33AF`.
The original failure log is retained. No measured native stack-frame size or
precise machine-code cause is claimed from the source-level diagnosis alone.

The corrected checker replaces recursive expression calls with at most128 heap
continuation frames and one returned type. It visits ordered children one at a
time, retaining expected-type inference, parameter declaration order and existing
meter precharges. The required depth128 was retained; no test stack increase,
smaller boundary or hidden profile workaround was used.

The exact boundary regression now also covers mixed Binary/If/Call/Power/Unary
depth128/129, contextual empty lists through If/Call at128, and a32/33-definition
call graph. Admission precharge regressions separately cover wide expression
children, call-key/argument lookup and bounded relation-subject traversal.

## Completed commands

All commands used Rust/Cargo1.98.1 Windows GNU/LLVM, one Cargo job, the isolated
manifest and root's `target` directory; no worker compiled concurrently.

| Scope | Actual result |
| --- | --- |
| `cargo test --locked --manifest-path .local/analysis-core-worktree-20260911/Cargo.toml --target-dir target -p oa-core --lib analysis::tests::logical_depth_and_call_depth_have_exact_boundaries -- --exact --test-threads=1 --nocapture` | Exit0; warning-free compile2m06s;1 passed,92 filtered;0.05s. |
| Same prefix, `--lib analysis:: -- --test-threads=1` | Exit0;28 passed,65 filtered;7.96s. Includes25 new analysis tests and3 existing topology-analysis tests. |
| Same prefix, `--lib numeric:: -- --test-threads=1` | Exit0;26 passed,67 filtered;0.82s. |
| `cargo clippy --locked --manifest-path .local/analysis-core-worktree-20260911/Cargo.toml --target-dir target -p oa-core --all-targets -- -D warnings` | Exit0;3m37s. |

Selected counts overlap and must not be added to infer a complete93-test pass.
No full core integration or whole-workspace test result is implied.

## Provenance and remaining acceptance

The corrected file SHA256 values are:

| Path below isolated core src | SHA256 |
| --- | --- |
| `analysis/typecheck.rs` | `831F14346BBAE83FBDEEB744F9F29460FAB13A8315645FAA6CE94D0B60B1516F` |
| `analysis/tests.rs` | `9FB30B28E305CB00E7C39204912F8311E96AA92D0334CE4A73FB08FADDFE83C5` |
| `analysis/admission.rs` | `5F85EAA868AC43C4819C98556281D8226F9C0E2A4B11E5C97CF24D855D2127AE` |

Immutable `.local/analysis-core-iterative-gate-20260911/` contains the corrected
33,854,976-byte executable, verified libunwind.dll, initial failure log, exact
depth/full-filter/numeric logs and274-input working-source manifest. Corrected
executable SHA256:
`2B4EE8BA63B20415B352E084314CC7504E778C093598B9564C6EA4F56BC921C9`.
Manifest SHA256:
`28D6CEAE5C52765F5DB5CB411D7E42EDAB91359C9E351663315BC4B7F1FF2B7A`.
The completed Clippy log is preserved beside this immutable evidence separately;
SHA256 `6FE529BBA6A761FEDACB1A3933BB86E4FB644D8CC192F57FB0F45EFA5B9ABFFF`.

[Independent correction review](release-analysis-core-iterative-review-20260911.md)
found no material new correctness/resource defect and verified all three frozen
source hashes. Review SHA256:
`329F9673D39F735174B075984CDD42E3935CAE0983A15B5DBB0A54C4169AD87A`.
Default Serde recursion protection
remains deliberately retained; full depth128 transport needs the previously
specified bounded original-byte decoder, lifecycle and output closure. Source8
grammar, evaluation, actual property/check workflows and maintained model library
semantics remain required product work. Do not merge this milestone as completed
analysis functionality or infer release acceptance from these static tests.