Thomas Orozco

This commit introduces a **new capability** to limit the total console output from Buck2 actions and tests, preventing excessive log generation that can overwhelm CI systems. It adds the `BUCK2_CONSOLE_OUTPUT_LIMIT` environment variable, which, when set, caps the cumulative stdout/stderr emitted by both `SimpleConsole` and `SuperConsole` in the **Buck2 client**. When the limit is exceeded, further output is suppressed, improving **performance and resource management** in verbose build environments. This change affects the **console display logic** and includes **comprehensive new tests** to verify the output truncation.

This commit introduces a **new capability** to log the **wall-clock completion time** for **top-level build targets** in Buck2, addressing limitations where existing critical path data might be incomplete for timed-out targets. It achieves this by adding a `wall_clock_completion_ms` field to `TopLevelTargetMetrics` in `buck2_data.proto` and populating this metric within the **`buck2_server_commands`** module. This enables more accurate **attribution of overall build runtime** to individual targets, significantly improving **telemetry and performance analysis**. The change involves **refactoring** in **`buck2_build_api`** to introduce a `Timed` wrapper for build results, and updates to **`buck2_bxl`** and **`buck2_test`** to correctly process these new timed outputs.

This commit **refactors** the **`buck2_action_impl`** module to improve type safety and correctness within the `actions.run` method. It updates the `weight` and `timeout_seconds` parameters in `app/buck2_action_impl/src/context/run.rs` from `i32` to `u32`, ensuring they are always treated as **positive integers**. This **maintenance** change also replaces an unchecked cast with a safer `.try_into().unwrap()` conversion, enhancing the robustness of parameter handling for `analysis_actions_methods_run`. The change prevents potential issues with negative values being passed to these parameters, leading to more reliable action execution.

This commit provides a **documentation update** for the `timeout_seconds` parameter within the `app/buck2_action_impl/src/context/run.rs` file. It clarifies the purpose, behavior, and best practices for using `timeout_seconds`, outlining both recommended approaches and potential pitfalls. This **maintenance** effort aims to improve developer understanding of this critical parameter, ensuring its correct and effective application within the **buck2_action_impl** module. The update helps prevent misconfigurations and promotes more robust action execution by detailing its do's and don't's.

This commit introduces a **new capability** for **Buck2 rule authors**, allowing them to specify an execution timeout for `run` actions via the new `timeout_seconds` parameter in `ctx.actions.run()`. Previously, only test actions supported such time limits. The specified timeout is seamlessly integrated into the **action execution pipeline**, flowing from `UnregisteredRunAction` to `CommandExecutionRequest.with_timeout()`, which then leverages existing timeout enforcement mechanisms for **remote, local, and worker execution**. This enhancement significantly improves build reliability by preventing `run` actions from hanging indefinitely.

This commit introduces a **new capability** to disable cgroup assignment for the Buck2 daemon specifically within end-to-end tests, addressing issues with inaccurate resource utilization metrics. It modifies the **Buck2 resource control** logic in `buck2_resource_control/src/spawn_daemon.rs` to respect a new `BUCK2_TEST_DISABLE_DAEMON_CGROUP` environment variable. The **e2e testing framework** is enhanced by adding a `disable_daemon_cgroup` parameter to `buck_test` and `buck_fixture` in `tests/e2e_util/buck_workspace.py`, allowing tests to explicitly control this behavior. This **test infrastructure enhancement** significantly improves the accuracy of **resource utilization metrics** reported by Buck2 e2e tests, particularly for large BXL tests, by preventing the daemon's resource usage from being incorrectly attributed. Existing resource control tests are updated to explicitly maintain their current cgroup assignment behavior.

This commit introduces a **new feature** to the **`stamp_build_info`** mechanism, enabling it to **respect the C++ toolchain's `cxx.cache_links` settings**. The `stamp_build_info` function, located in `prelude/linking/stamp_build_info.bzl`, will now consider the C++ toolchain's cache upload configurations when embedding or generating build information. This **enhancement** ensures that the stamping of build metadata aligns with the C++ toolchain's caching strategy, leading to more consistent behavior and potentially improved efficiency for cached builds.

This commit **fixes a bug** in the **`python_binary`** rule where the **`cxx.cache_links`** setting, specifically the `allow_cache_upload` attribute, from the CXX toolchain was not being respected. Previously, if `allow_cache_upload` was not explicitly set on the `python_binary` target, it would incorrectly default to `False` instead of consulting the toolchain's configured default. The fix modifies the `python_binary_impl` function in `prelude/python/python_binary.bzl` to correctly retrieve the CXX toolchain's `allow_cache_upload` setting. This ensures that the intended cache upload behavior is consistently applied for Python binaries that link C++ code, resolving an issue where caching might have been inadvertently disabled.

This commit introduces a **new capability** to **truncate error message content** within **build reports**. It adds an `unstable_truncate_error_content` boolean field to the `CommonBuildOptions` protobuf message, enabling users to control the verbosity of error messages. The **CLI parsing logic** in `buck2_client_ctx` is updated to define and process this new `truncate-error-content` option, which is also initialized for `bxl` commands. This feature provides more control over **build observability** and report size, with corresponding updates to documentation and help output for various `buck2` commands.

This commit **fixes** an issue in the **`buck2_test` orchestrator** where worker failures were erroneously reported as "cancelled" tests, leading to inaccurate test results. Previously, a mechanism to prevent race conditions during Buck2 shutdown inadvertently caused worker failures to be miscategorized. The change in `app/buck2_test/src/orchestrator.rs` removes this incorrect classification and introduces a **liveliness check** within the `report_execution_outcome` function, ensuring that test outcomes are only reported if the test run is not genuinely cancelled. This **bug fix** significantly improves the accuracy and reliability of **test reporting**, providing clearer insights into the true reasons for test failures.

This commit implements a crucial **memory optimization** within the `buck2_test` **orchestrator** by **deduplicating `ActionKey`s** during critical path tracking. Previously, the collection of these keys without deduplication in the `critical_path_action_keys` function led to severe Out-Of-Memory (OOM) errors, consuming hundreds of gigabytes of RAM for large test jobs. By introducing an intermediary `HashSet` to deduplicate these keys, this **performance fix** drastically reduces memory consumption from over 280 GB to approximately 28 GB. This change prevents OOMs, allowing critical path analysis to complete successfully and significantly improving the stability and scalability of the test orchestrator.

This commit performs a significant **refactoring** to ensure **`QueueStats`** are consistently captured for **all Remote Execution (RE) actions**, addressing an issue where statistics were not correctly recorded for non-cancelled actions. It introduces `ExecuteResponseWithQueueStats` and moves the `QueueStats` definition into its own dedicated module, `app/buck2_execute/src/re/queue_stats.rs`. This **maintenance** and **bug fix** impacts the **`buck2_execute`** client and **`buck2_execute_impl`** executor logic, specifically `RemoteExecutionClient::execute_with_progress` and `ReExecutor::execute_action`. The change provides more **accurate and comprehensive telemetry** for remote build actions, improving the reliability of performance monitoring and analysis across the entire RE pipeline.

This commit introduces a **new capability** to track and report the **queue time for cancelled Remote Execution (RE) actions** within the build system. It modifies the `re/client.rs` to collect `QueueStats` specifically when a remote execution command is cancelled, passing this duration as `CommandExecutionMetadata` through the `execute/manager.rs` and `executors/re.rs` components. This enhancement improves the **observability** of the RE system by providing more detailed **performance metrics** for actions that do not complete. Consequently, it aids in understanding the efficiency and bottlenecks of the remote execution infrastructure.

This commit **fixes a reporting inconsistency** by ensuring that `CommandExecutionKind` is correctly captured and reported for **cancelled command executions**. It **enhances the `buck2_execute` module** by modifying the `CommandExecutionStatus::Cancelled` variant in `result.rs` to include `execution_kind` and updating `cancel` and `cancel_claim` functions in `manager.rs` to populate this field. The change also updates the `local.rs` and `re.rs` executors within `buck2_execute_impl` to correctly pass this information during cancellation, while `buck2_test/src/orchestrator.rs` is adjusted to ignore the new field. This provides more accurate and complete **command execution status reporting**, offering better insight into *what* was being executed even when an action is terminated prematurely.

This commit performs a **refactoring** within the **CommandExecutionManager** to simplify its execution cancellation API. It removes previously optional parameters from the `cancel` and `cancel_claim` functions in `buck2_execute/src/execute/manager.rs`, making them mandatory. This change was enabled by updates to both the **Local Executor** and **Remote Executor** in `buck2_execute_impl`, which now consistently provide these parameters. The update streamlines the **execution cancellation logic** by ensuring all necessary data is always passed directly, improving code clarity and maintainability across the build system.

This commit **fixes a broken test** within the **`buck2-help-env` command's testing suite**. It addresses a regression by **updating the golden files** `buck2-help-env.golden.txt` and `buck2-help-env-testing.golden.txt`. The update specifically incorporates the `SANDCASTLE_INSTANCE_ID` variable into the expected output, ensuring the tests accurately reflect current environment variable handling. This **maintenance** task restores the reliability of these **core help command tests**, preventing false negatives due to outdated expected outputs.

This commit introduces a **fix** to the **buck2_execute** system, specifically enhancing its diagnostic capabilities. It ensures that **command execution metadata**, including timing information, is properly propagated and logged even when an action is cancelled, such as during a test run timeout. The `cancel_claim` function in `buck2_execute/src/execute/manager.rs` was updated to accept this optional metadata, and the `buck2_execute_impl/src/executors/local.rs` executor now passes the `timing` data during cancellation. This **maintenance** change addresses a critical observability gap, allowing developers to identify and diagnose persistently slow tests that previously went unrecorded due to premature cancellation.

This commit **enhances telemetry** for **test discovery events** by introducing and populating a `target_label` field. Previously, this crucial identifier was logged for test execution but omitted during the discovery phase, leading to an inconsistency in data reporting. The change involves **modifying the `buck2_data` protocol buffer definitions** to include `target_label` in `TestDiscoveryStart` and `TestDiscoveryEnd` messages, and then **implementing the population of this field** within the `run_test_stage` method of the `buck2_test` orchestrator. This provides **more complete and consistent logging**, improving the ability to track and analyze test discovery processes.

This commit introduces a **new capability** to directly propagate the `sandcastle_instance_id` within **command execution reports**, enhancing **observability and data reliability**. It modifies the **`buck2_data`** protobuf definition to include `InlineCommandExecutionEnvironmentMetadata` and updates the **`buck2_execute`** module to capture the `SANDCASTLE_INSTANCE_ID` environment variable and embed it into `CommandExecutionReport` protos. This ensures the Sandcastle instance ID is reliably available for individual actions, even when `buck2_builds` records are incomplete due to timeouts or OOMs, which previously prevented linking `buck2_test_actions` to their Sandcastle instance.

This commit introduces a **new capability** to **optimize the deserialization of large output directories** by enabling **eager fingerprinting** of Remote Execution (RE) output trees and **limiting concurrent downloads**. It achieves this by adding `OutputTreesDownloadConfig` and `OutputTreesDownloadSemaphore` to the `buck2_execute` module, which control these behaviors. The **action execution pipeline** in `buck2_build_api` and `buck2_execute_impl` is updated to utilize this configuration, with `buck2_server` parsing new buckconfig options like `fingerprint_re_output_trees_eagerly` and `output_trees_download_semaphore_size`. This **performance enhancement** allows for more economical handling of large outputs, reducing resource consumption and improving efficiency during build processes.