transform: avoid field-name collisions for bound methods

[medvednikov]

Fixes #28784.

Summary

• Stop selector field-type inference from falling back to an unrelated struct's same-named field after the receiver has been resolved to a known struct.
• Preserve direct and promoted-field lookup, and retain the existing fallback when receiver struct information is unavailable.
• Add a focused transformer regression for value/pointer receivers and the unknown-receiver fallback.
• Add language regressions covering the reported optional fn () callback assignment, a callback that reads receiver state, reassignment to a different receiver, invocation after assignment, and resetting the callback to none.

Diagnosis

In the report, Model has an on_done method but no on_done field. resolve_selector_type_uncached nevertheless falls through to lookup_unique_field_type('on_done'), which can supply the ?fn () field type from Process. A method value must not acquire the optionality of an unrelated field merely because their names match.

After receiver-specific field and promoted-field lookup fails for a known struct, leave field inference unresolved so the existing checker-backed expression-type fallback can supply the method's function type. The compiler change is four added lines, including its explanatory comment.

Validation status

Draft: the diagnosis and patch have been reviewed statically, but reproduction and runtime validation are not yet complete.

• Verified the committed diff against 9bfa1dc3f3455d5733d5c226662b6aaba8544f2c: only the four-line compiler change and the two new regression files are present.
• No V build, formatter, or tests were run. The execution sandbox has no V compiler, and repository/bootstrap downloads are blocked (Could not resolve host: github.com / raw.githubusercontent.com). Repository reads and branch commits were made through the connected GitHub API.
• No CI, build configuration, or unrelated source files were changed.

Commands still requiring an environment with V available:

./v self
./v fmt -verify vlib/v/transform/type_propagation.v vlib/v/transform/selector_method_field_collision_test.v vlib/v/tests/options/option_fn_bound_method_field_name_collision_test.v
./v -silent test vlib/v/transform/selector_method_field_collision_test.v
./v -silent test vlib/v/tests/options/option_fn_bound_method_field_name_collision_test.v
./v -silent vlib/v/compiler_errors_test.v
./v -silent test vlib/v/

The focused regressions should also be checked against the parent revision to establish a failing-before/passing-after result before marking this ready for review.

[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 2026-09-19T14:15:57.911689Z	ff12b02	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: cfd8f7f0fc

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".

[chatgpt-codex-connector]

Preserve type resolution through nested embeddings

When a field is promoted through two or more embedded structs, this return incorrectly classifies it as a possible method and produces no type. embedded_field_for_promoted_field in vlib/v/transform/struct.v checks only whether the immediate embedded struct directly declares the field, so a valid selector such as Baz.x in vlib/v/tests/structs/nested_struct_embed_selector_test.v (Baz -> Bar -> Foo.x) reaches this branch; before this change, the unique-field fallback returned int. Direct callers of resolve_selector_type, including specialized array/loop inference paths, can consequently lose the field type when a checker annotation is unavailable, so the known-struct early return should occur only after a recursive promoted-field lookup.

Useful? React with 👍 / 👎.

[medvednikov]

Addressed in ff12b02.

Replaced the one-level promoted-field lookup with the existing recursive struct_field_path_for_field helper, then resolve the field type from the final owning struct. Pointer embeddings are unwrapped for the owner lookup, and the known-struct guard still prevents borrowing an unrelated same-named field's type.

Added a checker-free regression covering eight combinations of value/pointer receivers, value/pointer embeddings, and scalar/array fields, plus nearer-field and receiver-field shadowing. The original method/field collision regression remains in place.

Verified that the committed diff matches the prepared two-file patch. V compilation, formatting, and runtime tests remain unrun locally: the sandbox has no V compiler, and bootstrap downloads fail DNS resolution. Leaving the thread open for re-review.