[bokelley]

Summary

Backport of #1797 (SDK 10) to the 9.x line, where #1784 reproduces identically: from adcp.types.domains.formats import Slot raised ImportError when it was the first adcp import in a process, because the domain roots imported every schema module eagerly and formats/canonical/_base.py re-entered formats/canonical through core/__init__ while half built.

• consolidate_exports.generate_domain_exports writes each domain root as a PEP 562 module: __all__ and a TYPE_CHECKING import block keep the static surface, _DECLARED_IN names each name's schema module, and __getattr__ imports the declaring module on first read and caches the binding. An unknown name raises AttributeError, so submodule imports still fall through. Nothing is renamed; every name still resolves; import * still binds everything.
• The 24 domain roots are regenerated. make validate-generated ✓, make lint ✓, make typecheck-all ✓. Surface, layering, lazy-types, public-API and codegen suites pass (365 tests).
• New tests/test_domain_roots_import_in_any_order.py: each domain root and the reported entry points import first in a fresh interpreter.

Non-breaking: import-order fix only.

🤖 Generated with Claude Code

[aao-secretariat]

Ladon verdict: Approve

Approve — backport of #1797 to the 9.x line making 24 domain package roots lazy (PEP 562) to fix import-order re-entrancy (#1784).

The only non-generated source change is consolidate_exports.generate_domain_exports (scripts/consolidate_exports.py:972-1011), which emits a TYPE_CHECKING static-import block, _DECLARED_IN, and a guarded __getattr__/__dir__ matching the established types/__init__.py pattern. The AttributeError fall-through preserves submodule imports and import *. Regenerated roots preserve the public surface exactly (core: base __all__=1440, head __all__=1440, _DECLARED_IN=1440 in lockstep), so there is no public-API break and fix(types): is the correct semver signal. The new tests/test_domain_roots_import_in_any_order.py exercises the reported import-order paths in fresh interpreters.

No blocking or medium findings. high_risk is false, gated_paths is false, no author team gate. Although review_decision is REVIEW_REQUIRED, the gated-paths hard gate (row 2) only applies when gated_paths is true — it is not here — so this falls through to row 9.

[aao-secretariat]

Ladon verdict: Approve

Approve — clean backport of #1797 making the 22 domain package roots lazy (PEP 562) to resolve the #1784 import-order cycle.

Checked:

• Generator (scripts/consolidate_exports.py): __all__ and the new _DECLARED_IN map both derive from the same rows dict, so every exported name resolves via __getattr__; AttributeError fallthrough preserves submodule imports.
• The runtime __getattr__/__dir__ sit correctly under if not TYPE_CHECKING:, with the static surface in the TYPE_CHECKING block — preserving type-checker coverage per the repo's PEP 562 facade invariant.
• Public surface preserved: no renames, __all__ unchanged, non-breaking — no semver signal required.
• New fresh-interpreter import-order test suite is thorough.

No critical/high/medium findings. Consistent with the prior approve on head 12297a9.

Decision path: high_risk=false, gated_paths=false (no gate applies), no no-auto-approve team match, zero findings → decision table rows 1–8 do not fire → row 9 approve.