Zeina Migeed

This commit **adds a new failing test case** to the `pyrefly` project, specifically targeting the interaction of `yield-from` expressions with **union return annotations**. The test, located in `pyrefly/lib/test/yields.rs`, is designed to reproduce a **reported bug** (issue #2976) where `pyrefly` incorrectly handles such generator type signatures. This **maintenance** work improves the **test coverage** for complex generator scenarios, laying the groundwork for a subsequent **bug fix** in the type checker's analysis of `yield-from` statements.

This commit **fixes a bug** preventing **annotated legacy type variable definitions** (e.g., `T: TypeVar = TypeVar("T")`) from being correctly recognized as type parameters within the `pyrefly` type checker. Previously, these were misidentified as `NameAssign` bindings, but now the **`AnnAssign` handler** in `pyrefly/lib/binding/stmt.rs` correctly routes them to create proper `Binding::TypeVar` bindings. Furthermore, the **type solving logic** in `pyrefly/lib/alt/solve.rs` has been updated to use `check_type`, ensuring the special type representation of these variables is always preserved. This change eliminates **downstream `not-a-type` errors** when these type variables are used in annotations, ensuring **accurate type inference and error reporting** for complex generic type constructs.

This commit **reclassifies the type of unresolved imports** within the `pyrefly` type system, changing them from `AnyStyle::Explicit` to `AnyStyle::Implicit`. This **semantic correction** in the **binding builder logic** (specifically in `pyrefly/lib/binding/stmt.rs`) accurately reflects that unknown types from missing modules are not intentionally `Any`. While this has **no immediate impact on error output**, it establishes a **correct semantic distinction** for future type-checking logic that differentiates between explicit and implicit `Any` types. A related test in `test/errors.md` is updated to reflect the new `Unknown` classification.

This commit **adds a new failing test case** to the `pyrefly` type inference system, specifically within `pyrefly/lib/test/assign.rs`. The test demonstrates a potential **bug** where an explicitly **annotated variable's type** (e.g., `int`) is incorrectly "erased" or overwritten by `Any` when a value of type `Any` is assigned to it. This **test addition** highlights an issue in `pyrefly`'s type preservation logic, indicating that the declared type is not being maintained as expected for downstream analysis.

This commit **fixes a false positive "bad-return" error** in the **pattern matching exhaustiveness checker** within the `pyrefly` static analyzer. It **corrects the logic for narrowing operations** by ensuring that variable bindings for each narrowed name are properly looked up, rather than incorrectly reusing the original match subject expression. This **bug fix** introduces a new helper function `build_narrow_entries` in `pyrefly/lib/binding/bindings.rs` and refactors its integration into `pyrefly/lib/binding/pattern.rs` and `pyrefly/lib/binding/stmt.rs`. The change **improves the accuracy of exhaustiveness checking** for `match` statements, particularly when dealing with attribute subjects, and includes **new test cases** in `pyrefly/lib/test/pattern_match.rs` to validate the corrected behavior.

This commit **fixes noisy timeline events** generated by the **V1 ranking workflow**'s label management in `v1_analysis.py`. Previously, the workflow used a "clean slate" approach, removing and re-applying all managed labels on every run, which created redundant add/remove events in issue timelines and unnecessary GitHub API calls. This **bug fix** replaces that with a **diff-based label update** mechanism, where the script now fetches the current label state and only applies changes (additions or removals) when the desired state differs. This significantly **reduces unnecessary API traffic** and ensures issue timelines remain clean and informative by preventing duplicate label operations.

This commit **adds a new test case** to the `pyrefly` library's **pattern matching** module, specifically within `pyrefly/lib/test/pattern_match.rs`. The test verifies that **alias capture** in `match` statements does not incorrectly leak or affect other branches, ensuring proper scope and isolation of captured variables. This is a **quality assurance** measure to prevent a potential **bug regression** related to variable binding behavior, thereby enhancing the reliability and correctness of the **pattern matching** functionality.

This commit **adds a new test case** to `pyrefly/lib/test/pattern_match.rs` to document a **false positive `bad-return` error** within the **`pyrefly` static analyzer**. The test demonstrates that `pyrefly` incorrectly reports a missing return when **OR patterns (`|`)** are used in `match` statements on enum attributes, even if all enum variants are exhaustively covered. This **test addition** highlights a regression in the **`pyrefly` pattern matching analysis** and will help ensure future fixes correctly handle such exhaustive enum matches.

This commit provides a **bug fix** for a **PyreFly panic** by preventing duplicate entries in `exported_names` during the processing of `invalid_all_exports` within the **binding logic**. It introduces a check in `pyrefly/lib/binding/bindings.rs` to ensure the integrity of exported symbols, specifically addressing issues related to duplicate entries in `__all__` when augmented assignment is used. A new test case in `pyrefly/lib/test/scope.rs` verifies this correction. This change enhances the **robustness and stability** of PyreFly's module analysis by preventing crashes caused by specific module export patterns.

This commit provides a **bug fix** for `pyrefly`'s **type inference and overload resolution** when accessing descriptors through generic types with bounded type variables. Previously, `DescriptorBase::ClassDef` would incorrectly discard `TypeVar` information, leading to "No matching overload found" errors for overloaded `__get__` methods because `objtype` was inferred as a concrete type instead of a generic one. The **fix** modifies the `DescriptorBase` enum and its construction in `pyrefly/lib/alt/class/class_field.rs` to preserve the `ClassBase` wrapper, ensuring correct `TypeVar` propagation. This **resolves a critical issue affecting real-world code**, specifically enabling `pyrefly` to correctly analyze **Pydantic's `model_fields`** when used with generic models, as validated by new test cases.

This commit **adds a new failing test case** to the `pyrefly` type checker, specifically targeting a **false positive bug** in its **override detection logic**. The test, located in `pyrefly/lib/test/class_overrides.rs`, simulates a scenario involving subclassing the built-in `str` type with `LiteralString` overloads, which currently triggers an incorrect override detection. This **maintenance work** is a critical step towards identifying and resolving a subtle type checking error, thereby improving the accuracy and reliability of `pyrefly`'s static analysis for complex inheritance patterns.

This commit **fixes false positive `bad-override` errors** in the **type checker's override validation logic** by introducing a mechanism to **filter out inapplicable parent overloads**. Previously, the system would incorrectly require subclasses to satisfy *all* parent overloads, even when the `self` parameter type of an overload was incompatible with the subclass type, leading to erroneous diagnostics when, for example, subclassing `str` and overriding methods like `upper()`. The new `filter_overloads_for_override` function in `pyrefly/lib/alt/class/class_field.rs` ensures that only relevant overloads are considered during method binding. This **bug fix** significantly **improves the accuracy of type checking** and **reduces noise for developers** extending classes with complex overload signatures.

This commit **adds a new test case** to the **type system** within the `pyrefly` library, specifically targeting the behavior of `Literal` types. The test demonstrates that when iterating over a tuple of `Literal` types using `enumerate()`, the `Literal` types are unexpectedly promoted to their base types (e.g., `Literal["a"]` becomes `str`), whereas direct iteration preserves them. This **maintenance** addition, located in `pyrefly/lib/test/literal.rs`, improves **test coverage** and documents a specific nuance in `enumerate[T]`'s TypeVar solving mechanism, impacting **type inference accuracy** for code utilizing `Literal` types with `enumerate`.

This commit introduces significant **enhancements to the primer classifier** to address inconsistencies and improve classification reliability. It implements a **self-critique pass** where the LLM verifies its own reasoning for factual accuracy, and a **majority voting mechanism** for verdict assignment, making 5 independent calls to reduce non-determinism. Furthermore, **cross-project consistency enforcement** ensures similar error patterns receive uniform classifications across different projects. These **feature improvements**, along with an **upgrade to the `claude-opus-4-6` model** in `scripts/llm_transport.py`, aim to provide more robust and consistent evaluations of code changes within the **primer classification system**.

This commit **adds a new failing test case** to `pyrefly/lib/test/descriptors.rs` to reproduce a known issue. Specifically, it targets a bug related to **overloaded descriptors with bounded type variables**, as documented in issue #1083. This **maintenance** task ensures that the identified problem, which also affects Pydantic-based code, is captured by the test suite, preventing regressions and guiding future fixes. The test's current failure highlights an existing defect in the **descriptor handling logic** within `pyrefly` that requires resolution.

This commit **fixes a display bug** in the **`primer_classifier`** verdict output, specifically addressing the rendering of raw JSON reasons. It introduces a new `_format_reason()` function in `scripts/primer_classifier/formatter.py` to transform JSON reason dictionaries into **readable, labeled sections** for improved user experience. Concurrently, the `classify_with_llm` function in `scripts/primer_classifier/llm_client.py` is updated to **ensure the `reason` field is consistently a string**, serializing dictionary values to prevent downstream parsing issues. This **bug fix** ensures that classifier verdicts are always presented clearly and consistently to the user, enhancing the readability of LLM-generated explanations.

This commit **adds a new failing test case** to the `pyrefly` library's **generic restrictions test suite**. This **test addition** specifically targets an issue with `AnyStr | None` passthrough when used with class methods, aiming to reproduce a reported bug. By demonstrating the incorrect behavior, this test is crucial for validating the problem and will serve as a regression test once the underlying bug is fixed. It helps ensure robust type handling within the `pyrefly` framework.

This commit **fixes a race condition** within the **`mypy_primer` testing infrastructure** that previously led to **false primer deltas**. The `scripts/pyrefly_primer_wrapper.sh` script, which handles `pyrefly` initialization and dependency setup, was not safe for concurrent execution when `mypy_primer` ran old and new `pyrefly` versions simultaneously. A **directory-based lock** has been implemented in the wrapper script to serialize these critical setup steps, ensuring that both instances correctly establish their environments. This **bug fix** prevents erroneous dependency mismatches and significantly improves the reliability and accuracy of `mypy_primer`'s delta reporting.

This commit **adds comprehensive documentation** for the **issue ranker pipeline** to `scripts/issue_ranker/AGENTS.md`. The new markdown document details the architecture, usage, and key findings of the ranker, providing essential context for its operation. This **documentation update** is a **maintenance task** aimed at improving the usability and consistency of the **issue ranker agent**. By centralizing this information, it can be automatically loaded into Claude sessions, thereby streamlining interactions and enhancing the efficiency of the ranking process.

This commit performs **maintenance** by **removing extraneous `echo` statements** from the `scripts/pyrefly_primer_wrapper.sh` script. Previously, these statements were causing wrapper messages to be printed, which **polluted the output of the primer diff** messages. This change **cleans up the script's output**, ensuring that the `primer diff` messages are concise and free from irrelevant logging, thereby improving the clarity of the diffs.