Jeremy Braun

fix oss build

This commit introduces a **new capability** to **propagate tracing spans** across **I/O executor threads** within the `buck2_execute` module. It ensures that the `current_span` context is correctly proxied, allowing `ThreadPoolIoRequest` objects to track their originating span when they create new spans. This **feature enhancement** improves **observability and debugging** by maintaining a consistent tracing context for operations that offload work to I/O thread pools, preventing broken traces. The changes are primarily implemented in `app/buck2_execute/src/execute/blocking.rs` by adding parent ID tracking and a span proxy mechanism.

This commit introduces a **new capability** to **proxy the current tracing span** across the **deferred materializer's** asynchronous command processing thread. It modifies methods within `app/buck2_execute_impl/src/materializers/deferred.rs` to pass the `CURRENT_SPAN` when sending various materialization and invalidation commands. Concurrently, `app/buck2_execute_impl/src/materializers/deferred/command_processor.rs` is updated to include an optional `SpanId` in the `MaterializerCommand` enum and to correctly propagate this span for `Declare`, `InvalidateFilePaths`, and `Ensure` operations. This enhancement ensures that tracing context is maintained across thread boundaries, significantly improving the **observability and debugging** of deferred materialization workflows within `buck2_execute_impl`.

This commit introduces a **new capability** to **propagate synchronous tracing spans across thread boundaries** within the **`buck2_events` tracing system**. It adds the `maybe_proxy_current_span` function and a `RestoreSpan` helper struct to `app/buck2_events/src/dispatch.rs`, enabling the `CURRENT_SPAN` context to be correctly "proxied" to work delegated to other threads. This **enhancement** ensures that all operations, even those performed asynchronously, maintain their proper parent-child relationships in traces, significantly improving **observability and debugging** by eliminating "orphan" spans with `parent_id = 0`. The change focuses on span context propagation rather than timing, which is handled separately.

This commit introduces a **new capability** to the **`buck2_events` tracing system**, enabling the `current_span` to be correctly propagated across asynchronous **Tokio tasks**. It adds the `SpanProxyAsync` struct and its methods, including `new`, `new_with_span`, and `poll`, to `app/buck2_events/src/dispatch.rs`. This allows spans created within a spawned Tokio task to automatically set their `parent_id` to the originating `current_span()`. Consequently, this significantly improves the accuracy of **distributed tracing** and **event correlation** for asynchronous operations, ensuring a complete and coherent view of execution flow.

This commit **adds a new test case** to the **materializer** component, specifically documenting the current behavior where **materialization spans** are generated **without a `parent_id`**. This **preparatory test** formalizes an existing characteristic, setting the stage for a future improvement (D93334098) that will address this limitation. Additionally, the `tests/e2e_util/helper/utils.py` module sees a **refactoring** of `filter_events` and `json_get` to enhance test utility by allowing them to return the root JSON object when a match is found. This work is primarily **maintenance** and **test infrastructure improvement**, ensuring future changes to span generation are properly validated.

This commit **refactors** the management of **Python `preload_deps`** by moving them from individual `python_binary`/`python_test` targets into the **Python toolchains**. A new `preload_deps` attribute is added to `PythonToolchainInfo`, and a helper function `python_attr_preload_deps()` is introduced to consolidate these dependencies from both the target and the toolchain. This **maintenance** change **simplifies the build graph** by reducing redundant dependency declarations and centralizes the configuration of common preloaded dependencies. It impacts various **Python build rules** and **linking/packaging logic** across `prelude/python` modules, including `native.bzl`, `omnibus.bzl`, `make_py_package.bzl`, and `python_binary.bzl`. The primary goal is to eliminate unnecessary graph edges and improve consistency in how preloaded dependencies are applied.

This commit introduces a **new `default_deps` attribute** to the **`cxx_toolchain` rule** and extends the **`CxxToolchainInfo` provider** within the `prelude/cxx` module. This **infrastructure enhancement** allows CXX toolchains to declare a list of default dependencies, which can then be automatically added to targets utilizing that toolchain. It is a **preparatory change** to facilitate a future shift towards managing common CXX dependencies directly within the toolchain definitions. This update affects the core definition and metadata propagation of CXX toolchains, laying the groundwork for more streamlined dependency management across the build system.

This commit **fixes a bug** in a **test** within the `buck2_execute_impl` module by correcting the syntax for a sleep command. The test, located in `app/buck2_execute_impl/src/executors/local.rs`, was using an incorrect sleep command for **PowerShell on Windows**, which could cause test failures. The change updates the command to `Start-Sleep -Seconds`, ensuring the test executes reliably on Windows environments. This **test fix** improves the overall stability and correctness of the build system's test suite.

This commit **adds new tests** to the **`uquery`** system, specifically within `tests/core/query/uquery/test_uquery.py`, to illustrate the current behavior of the string `'None'` (and `'none'`) when used as target patterns and function names. This **test addition** also updates the `TARGETS.fixture` with corresponding test data, ensuring comprehensive coverage for these specific cases. The work serves as **maintenance** by documenting existing behavior, particularly in anticipation of a future change (D97162342) that will alter how these specific strings are processed.

This commit introduces the **`None` pseudo-keyword** into the **Buck2 query DSL**, allowing query arguments to explicitly accept `None` values where applicable. It involves significant changes across the **`buck2_query` and `buck2_query_parser` components**, including parsing `None` as a distinct `Expr::None` variant, updating the query evaluator to handle `QueryValue::None`, and modifying query functions to accept `None` for arguments that opt-in via `accept_none()`. This **new capability** enhances the flexibility of query arguments and improves error reporting for missing arguments by providing more specific context. The `buck2_query_derive` crate was also updated to support this change, ensuring proper code generation for query functions.

This commit introduces a new **`QueryValueDepth`** enum to semantically represent query traversal depth, replacing previous `Option<i32>` parameters in **`deps()` and `rdeps()` functions**. This **refactoring** centralizes the "large values are unbounded" logic, previously duplicated, into the type system itself, significantly improving clarity and maintainability within the **`buck2_query`** module. It also enables the `None` keyword to naturally signify "unbounded depth" for query arguments, simplifying usage across **aquery, cquery, and uquery** contexts in BXL. This change makes depth handling more robust and explicit throughout the query system.

This commit introduces a **new capability** by adding a `hashed()` method to the **`SmallSet`** data structure within the `starlark_map` module. This method consumes the `SmallSet`, pre-computes a hash from all its elements in iteration order, and returns a `Hashed<SmallSet<T>>` wrapper. The primary impact is to **enable `SmallSet` instances to be used in contexts that require the `Hash` trait**, which was previously not directly supported. This **feature addition** significantly expands the utility and flexibility of `SmallSet` for downstream consumers in hash-based collections and operations.

This commit **resolves a bug** in the **dependency view checking mechanism** where `check_within_view` could incorrectly fail for attributes with default values. It **enhances the `buck2_node::attrs::attr` module** to collect and store default allowed dependencies during attribute creation, and **modifies `buck2_interpreter_for_build::nodes::check_within_view`** to properly consider these defaults. This ensures that default `TargetLabels` are correctly evaluated against view rules, **preventing false-positive view check failures** and improving the accuracy of dependency validation within Buck2.

This commit **refactors** the **attribute coercion and traversal logic** within the `buck2_node` and `buck2_interpreter_for_build` modules to make the `PackageLabel` parameter **optional**. Specifically, methods like `traverse` now accept an `Option<PackageLabel>`, preparing the system for future scenarios where a package context might not be immediately available during attribute coercion. It introduces a new error type and improved error handling for `Source` attributes that are coerced without a package, ensuring clearer diagnostics. This change is a **preparatory step** for an upcoming modification (D96159747) that will introduce paths where the package is `None`, enhancing the flexibility and robustness of attribute processing.

This commit performs a **refactoring** within the **`buck2_client_ctx`** module to enhance **type safety** and better align with **Rust's ownership model**. Specifically, several `finalize` methods across `events_ctx` and various `subscribers` (e.g., `event_log`, `re_log`, `recorder`) have been updated to consume `self` or `Box<Self>` instead of taking a mutable reference. This change semantically indicates that the object is no longer usable after `finalize` is called, ensuring its proper destruction. The **`EventSubscriber` trait**'s `finalize` method signature was also updated, requiring all implementors to adapt to the new consumption pattern. This improves the clarity of object lifecycle management within the event handling system.

This commit **adds a comprehensive doc-comment** for the `deps` query function within the **Buck2 query engine**. Specifically targeting `app/buck2_query/src/query/syntax/simple/functions.rs`, this **documentation improvement** details the function's parameters and usage, adapted from existing `rdeps` documentation. It serves as a **maintenance** task to enhance the clarity and usability of the **Buck2 query language**. This change directly benefits users by providing clear instructions for a core query function, making the query system easier to understand and correctly employ.

This commit performs a **refactoring** within the **Python build system**, renaming the `python_toolchain` attribute to `_python_toolchain`. This change ensures **attribute name consistency** between the rule definition and its anonymous rule usage, aligning with common Starlark conventions for internal attributes. Specifically, the `process_native_linking_rule` definition and its implementation in `prelude/python/linking/link_helper.bzl` were updated, along with its usage within `_convert_python_library_to_executable` in `prelude/python/python_binary.bzl`. This internal adjustment improves the **maintainability and clarity** of the **Python linking rules** without altering external behavior or public APIs.

This commit performs a **refactoring** to correct a **typographical error** and improve clarity within the **Buck2 BXL query system**. It **renames the `deps` parameter to `depth`** in the `deps` method signatures across the `BxlCqueryFunctions`, `BxlUqueryFunctions`, and `BxlAqueryFunctions` traits defined in `app/buck2_build_api/src/query/bxl.rs`. Correspondingly, the parameter name is updated in the `deps` method implementations for these query types within `buck2_query_impls`, ensuring consistency. This **typographical fix** clarifies the parameter's true meaning, enhancing the **internal API readability** and correctness for developers interacting with the query functions.

This commit introduces a **new capability** for the **`buck2 audit config` command**, enabling a new `--output-format=inline_section` option. This format prints configuration values with their section name inline, making it significantly easier for users to search for specific `section.key` patterns. The change involves defining the new output format on the client side in `app/buck2_cmd_audit_client/src/config.rs` and implementing the detailed rendering logic, including updates to functions like `print_value` and `render_value`, within the server-side `app/buck2_cmd_audit_server/src/config.rs`. This enhancement improves the usability and searchability of configuration output, with corresponding updates to golden test data.