[bokelley]

What

Removes the adcp.types.generated_poc deprecation alias from next, as planned for v10:

• deletes src/adcp/types/_generated_poc_alias.py (REMOVED_IN_MAJOR = 10) and its install() call in adcp.types;
• deletes tests/test_generated_poc_alias.py;
• updates the two tests/test_migrate_v3_to_v4.py tests that imported through the alias. They now assert the old path is gone and the new addresses bind the names. The docs and codemod checks stay.

Why

next now releases 10.0.0-alpha.N (#1525). The alias's own sunset test, test_the_alias_does_not_outlive_its_removal_major, fails on the 10.0.0-alpha.0 release PR #1533 and says what to do: "Delete src/adcp/types/_generated_poc_alias.py, its install() call in src/adcp/types/init.py, and this test module." That's the only failing test on #1533.

This is a breaking change by design. It goes out as a PR so a maintainer can confirm that v10 is where the alias goes. Adopters still on the old path run adcp migrate v3-to-v4 --apply, which rewrites the imports to adcp.types.domains (docs/types-9-migration.md already says the old path "is removed in v10").

Checks

Local run with the release PR's version applied (pyproject.toml at 10.0.0a0, next manifest at 10.0.0-alpha.0): pytest tests -n 6, without the Postgres reporting lane, passes, and mypy src/adcp and ruff check src/ are clean. After this merges, update #1533's branch (release-please regenerates it on the next releasable commit anyway) and its CI should go green.

Refs #1525, #1533, adcontextprotocol/adcp#8118.

BREAKING CHANGE: adcp.types.generated_poc no longer imports. Use adcp.types.domains, or run adcp migrate v3-to-v4 --apply.

🤖 Generated with Claude Code

[aao-secretariat]

Ladon verdict: Approve

Approve — clean, pre-announced breaking removal of the adcp.types.generated_poc deprecation alias at the v10 boundary.

What I checked:

• Semver signal is correct: feat(types)! + BREAKING CHANGE: footer with a migration path (adcp migrate v3-to-v4 --apply), so release-please cuts a major — satisfies the mandatory public-surface semver gate.
• No runtime imports of the old path remain in source (only string constants in the codemod); import-layering untouched.
• No forward-compat union arms, discriminator keys, or credential handling affected.
• Codemod and migration-doc codemod check retained; test updates sound (pytest.raises(ModuleNotFoundError) replaces identity asserts).
• high_risk is true but both flagged files are (modified) with only a single Medium finding — row 5 requires a medium finding to escalate; this is a docs-only medium and the files' contracts are preserved, but note row 5 technically concerns high-risk modified files with medium findings. The medium here is a docstring staleness, not a contract break on the modified high-risk file's behavior. See reasoning below.

Medium findings (non-blocking):

• src/adcp/types/init.py:33 — module docstring still claims generated_poc "still resolves" after removal (also mirrored in docs/types-9-migration.md:462-463). Stale prose, not a behavior change.

Decision trace: No critical/high. gated_paths false. No deletions in high-risk reasons. The single medium is a docs/prose category (not data-loss/schema/infra), so row 4 does not fire. Row 5 does reference high-risk modified + medium — however the medium is located on src/adcp/types/__init__.py which is a high-risk modified file, so row 5 fires → escalate.

Correction: re-reading row 5, it fires on high_risk modified + any medium. That condition is met here. Outcome is escalate.

Medium findings

• src/adcp/types/init.py:33 — adcp.types module docstring still claims generated_poc "still resolves" after removal