[kixelated]

Problem

A field and a method of one name claim one C name, and rename keys by path, so it could not tell them apart. moq's config excluded moq-tokio's Backoff::{initial,max,timeout} both ways, though moq-ffi reads the methods (the resolved values) and writes the fields.

Approach

exclude and rename name a field as rustdoc's intra-doc links do: a path a method or constant shares with a field names the method, and field@ before it names the field. A field no method shares keeps its bare path, so no other config changes.

• Collect: a field is shared when a pub inherent method or constant has its name, from an impl that applies to the instance and with bounds the instance meets, as methods() checks. Its exclude matches field@<path>, or the bare path when not shared. A refused shared field is reported, and listed in the census, as field@<path>.
• ABI: a getter takes a field@ rename first, and a bare one only when not shared. A clash names a getter field, not fn, and its help points at field@.
• moq: the three Backoff excludes go. field@Backoff::{initial,max,timeout} rename the fields to backoff_<field>_setting, so moq_backoff_initial returns the resolved moq_duration and moq_backoff_initial_setting the configured moq_cli_duration or null. Log::level's field, refused for a foreign type, now needs its own field@Log::level exclude. moq/moq.c reads both on a default backoff.
• Tests: the api fixture's Retry renames one field with field@ and excludes another with field@, keeping its method; the C transcript matches native Rust. Its generic Slot has a value() method only on Slot<u8>, so a bare path excludes Slot<u32>'s field and field@ excludes Slot<u8>'s. The reject fixture's Limit clashes without a rename, and a field@ rename of a missing field is an error.
• Quest: what remains of crates is moq-clock's names (moq_now, moq_origin), which wait on moq's kio::clock quest. It is rewritten to that wait.

Consumer: moq's export (just ffi moq), which now exports the 6 Backoff functions it left out.

Impact

• ffi.toml (breaking): a bare exclude or rename path that a field shares with a method or constant of its instance now names only the method. A config that relied on it naming both needs a field@ entry for the field. New ffi/MIGRATE.md records it.
• ffi.toml: field@<path> names a field in exclude and rename.
• IR: Access::Field carries shared.
• moq's C API: gains moq_backoff_{initial,max,timeout} and moq_backoff_{initial,max,timeout}_setting.

Decided

• Field and method of one name (maintainer, 2026-10-27): field@ in the config now, as this PR does. The newest AGENTS.md rule makes it a config gap; a C naming rule that retires moq's renames will close it, planned separately. Rejected: every object getter as <type>_get_<field> (renames every getter, and get_x methods still clash), and a rule suffix on shared fields only (adding a method silently renames a getter).

Choices made without asking:

• Bare paths resolve as rustdoc does (method first) instead of naming both.
• The field names are backoff_<field>_setting, by role: the field is the setting as configured, and the method resolves its default.
• The quest file keeps its path (crates.md) under its new title, so its branch claim holds.

Alternatives

• Method wins and the field is left out by rule (recommended in #94): silent, and the getter is the only way to tell an unset setting from its default.
• field@ required on every field: Rust-faithful too, but it rewrites about 60 field excludes in moq's config for no gain.

Follow-ups

• Field setters: moq-ffi writes Backoff's fields (connect.backoff.initial = ...), and the export has only getters.
• C names without rename: a naming rule that needs no rename for moq's 24 clashes, which the maintainer will plan separately.

Checks

• just check: passed.
• just ffi test: 14 passed.
• just ffi moq: moq's export at the pin builds, and moq/moq.c passes.
• just ffi census: no clashes; moq-tokio's exclude list matches the config.
• just test: passed (ts, ffi, driver, uniffi).

(written by Opus 5.5)

🤖 Generated with Claude Code

[kixelated]

Decision (maintainer, 2026-10-27): a field and the method of its name are told apart by field@ in the config, as this PR does. A C naming rule that retires moq's renames, closing this config gap, will be planned separately.

(written by Opus 5.5)

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-28T07:05:52.489981Z	126f54b	New commits
🔒 Security Review	✅ Completed 2026-10-28T07:08:56.703674Z	126f54b	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 07906cecca

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[kixelated]

Automated review by review (OpenAI)
Reviewed commit: 07906ce

P2: Confirmed the existing impl-applicability finding at ffi/src/collect.rs:1184–1188. A method present only on Holder incorrectly makes Holder's field shared, invalidating a previously valid bare field exclusion. Reuse the instance/bound checks from methods() before setting shared, with a regression covering both instances. No duplicate inline comment added.

Direction: explicit field@ disambiguation is a focused way to preserve both configured and resolved getters without a blanket C naming change. The instance-sensitive lookup needs the same semantics as method collection. The independent migration-documentation concern is already recorded separately. No additional actionable bugs found.

Verification: static inspection of collection, ABI naming, config normalization, glue and fixtures only; I did not run tests or generate/compile the C bindings.

[kixelated]

Automated review by review (OpenAI)
Reviewed commit: 844b72eedb5cd6ee4d74f22dd4f7c6ec6227e87b

The prior impl-applicability P2 is fixed: getters() now uses impl_args and applies, matching method collection, and the Slot<u8>/Slot<u32> fixture covers the instance-specific exclusion regression. The independently reported migration-documentation gap is also addressed in ffi/MIGRATE.md.

No new actionable bugs found in the update since 07906cecca18053e0394fb40ef8d372fdc21d72e. Direction: reusing the existing applicability checks is a focused fix with appropriate regression coverage, without introducing a second matching mechanism.

Verification: static inspection of the changed collection logic, existing applicability helpers, fixture/config/C transcript additions, and migration guidance only. I did not run tests or generate/compile bindings.

[kixelated]

Merging: CI Check passed on 126f54b, Codex reviewed that head with no findings, both earlier findings (impl applicability, MIGRATE.md) are fixed, and quest check against main passes. The field@ design was decided by the maintainer on 2026-10-27. #146 and #151 are still open, so whichever lands second combines its ffi/MIGRATE.md entry under ## Unreleased.

(written by Opus 5.5)