fix: collect nested deps for action Goal/Result/Feedback hashing

[YuanYuYuan]

Fixes the nested-dependency-collection defect #334 reported.

A community pull request (#334, draft) from @sathak93 reported a real defect and diagnosed and fixed it correctly, but shipped no test coverage. This preserves their original commit (cherry-picked, authorship intact) and adds the regression test the diagnosis needed on top.

calculate_send_goal_hash, calculate_get_result_hash, and calculate_feedback_message_hash in crates/hiroz-codegen/src/resolver.rs each insert the action's Goal/Result/Feedback TypeDescription into the deps map used to compute its RIHS01 type hash, but never call collect_nested_deps to also pull in that type's own nested/referenced message dependencies — unlike plain service hashing, which already does this (resolver.rs:181-182).

Before and after

	today	with this change
action Goal/Result/Feedback with no nested custom message	correct hash (empty nested-deps set is also the correct set)	unchanged
action Goal/Result/Feedback referencing a nested custom message	deps omits the nested type; RIHS01 hash is wrong	collect_nested_deps pulls in the nested type, matching the plain-service code path

What the test does

crates/hiroz-codegen/tests/action_nested_deps.rs (new) uses tf2_msgs/LookupTransform's Result (geometry_msgs/TransformStamped transform, tf2_msgs/TF2Error error), a real fixture already packaged under assets/jazzy/ — no new asset files needed. TransformStamped nests std_msgs/Header and geometry_msgs/Transform, and Transform itself nests Vector3/Quaternion, so this exercises multi-level recursive nesting, not just a single level.

	today	with this change
test_lookup_transform_get_result_hash_includes_nested_deps	did not exist; the equivalent unfixed code computes RIHS01_d21f340b... (whippet job 4383)	new; pins the corrected hash RIHS01_3cd17157... (whippet job 4384), computed over the complete dependency set

Existing action-hash coverage (action_hash_check.rs's Fibonacci test) could not have caught this — Fibonacci's Result (int32[] sequence) has no nested custom message type.

Verified standalone: cargo test -p hiroz-codegen --lib --tests — all green (whippet 4385). cargo clippy -p hiroz-codegen --all-targets -- -D warnings — clean.

Breaking changes

None to the public API. The RIHS01 hash for any action whose Goal/Result/Feedback references a nested custom message type changes — this is the fix, not a regression: the old hash was computed over an incomplete dependency set and would not interoperate with a real rmw_zenoh_cpp peer using the same action definition.

[Copilot]

🟡 Changes recommended

Formatting must be corrected, untested Goal and Feedback branches covered, and the unrelated #158 reference disambiguated.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Fixes RIHS01 action hashing by including transitive message dependencies.

Changes:

• Collects nested dependencies for Goal, Result, and Feedback hashes.
• Adds a real-world GetResult hash regression test.

File summaries

File	Description
crates/hiroz-codegen/src/resolver.rs	Recursively collects action component dependencies.
crates/hiroz-codegen/tests/action_nested_deps.rs	Pins the corrected LookupTransform result hash.

Review details

Suppressed comments (1)

crates/hiroz-codegen/src/resolver.rs:688

• This Feedback path is not exercised with nested dependencies: the new test uses LookupTransform, whose feedback section is empty, and asserts only the Result service hash. Please add a nested-feedback case (the bundled test_msgs/NestedMessage.action fixture covers this shape) and pin feedback_message_hash so this independent branch cannot regress.

        self.collect_nested_deps(&feedback_desc, &mut deps);

• Files reviewed: 2/2 changed files
• Comments generated: 3
• Review effort level: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[CLAassistant]

All committers have signed the CLA.

[YuanYuYuan]

Hi @sathak93! This PR preserves your original commit (cherry-picked with your authorship intact) on top of a regression test, so the CLA check now needs a signature from both of us. Could you sign the CLA via the link below when you get a chance?

https://cla-assistant.io/ZettaScaleLabs/hiroz?pullRequest=339