Mark Fields

This commit **deprecates** the `IIdCompressorCore` interface within the **`id-compressor` package**, preparing for its internalization in a future release. This **API evolution** addresses its unintended exposure by introducing new public utility functions, `serializeIdCompressor` and `toIdCompressorWithCore`, to replace its functionality. The change includes adding detailed deprecation warnings and migration guidance in documentation and API reports, while suppressing warnings in internal usage across `container-runtime` and test utilities. This **refactoring** ensures a cleaner public API surface and provides a clear migration path for consumers ahead of a breaking change in version 2.100.

This commit introduces a new **`resetUnfinalizedCreationRange` API** within the **`id-compressor`** module, representing an **improvement** for robust ID management. This method allows the `IdCompressor` to effectively discard ID ranges that were allocated but not finalized due to a dropped connection, by resetting its internal state. The **`container-runtime`** now integrates this API into its `replayPendingStates` mechanism, ensuring that previously unfinalized IDs are properly re-processed and finalized upon reconnection. This change significantly enhances the **consistency and reliability of ID allocation** across network disconnects, streamlining the state replay process by eliminating the need for an extra op.

This commit **refactors telemetry logging** within the **`container-runtime`** module to **reduce logging overhead and noise** by consolidating boot-time Garbage Collection (GC) related events. It **removes the `GCFeatureMatrix` and `GarbageCollectorLoaded` telemetry events**, which previously fired on boot. Instead, relevant GC configuration details are now integrated into the existing `ContainerLoadStats` event, leveraging a new `serializedConfigs` getter on the `GarbageCollector` class and updating the `IGarbageCollector` interface. This **telemetry optimization** centralizes critical boot-time GC data, making event consumption more efficient and streamlined.

This commit **corrects an inaccurate comment** within the **`container-runtime` package**, specifically in `containerRuntime.ts`. It updates the documentation for the `computeCurrentDirtyState` function to accurately reflect its logic, which was previously described in an inverted manner regarding the container's dirty state. This **documentation fix** improves code clarity and ensures developers correctly understand how the function determines if a container is dirty, preventing potential misinterpretations of the runtime's state management.

This commit **improves test stability** by **skipping** the `opCriticalPath` **benchmark test** within the **end-to-end test suite** when run against the **Routerlicious (R11s) driver**. A `beforeEach` hook in `packages/test/test-end-to-end-tests/src/test/benchmark/opCriticalPath.time.spec.ts` now conditionally disables this test for `'r11s'` or `'routerlicious'` configurations. This **maintenance change** addresses observed **flakiness** specific to these environments, ensuring more reliable CI/CD pipeline results without impacting the test's intended client-side performance measurement.

Bump deps on benchmark to 0.52.0 (#26146)

Benchmark: Bump version in main to 0.53.0 ahead of the 0.52.0 release (#26132)

Benchmark Tool: Add changelog ahead of 0.52.0 release (#26131)

This commit introduces a **fix** to the **container-runtime**'s error handling, specifically within the `ChannelCollection`'s `processMessages` method. It now **distinguishes between different causes of "Duplicate DataStore Id" errors**, classifying them as `DataProcessingError` when the duplicate ID refers to an unbound local DataStore, versus `DataCorruptionError` for already processed DataStores. This **prevents transient, single-client failures from being incorrectly reported as permanent data corruption**, improving the accuracy of telemetry and avoiding unnecessary container invalidation. The change includes new test cases in `channelCollection.spec.ts` to verify the correct error types and a refactor to `DataProcessingError.create` in `telemetry-utils` for more flexible error creation. This ensures that only truly unrecoverable issues are flagged as permanent corruption, providing a more accurate representation of container health.

This commit delivers a crucial **bug fix** to address intermittent benchmark failures where the "benchmark end" event was not emitted if an error occurred during test execution. It introduces new utility functions, `emitResultsMocha` and `captureResults`, to robustly handle function execution, measure time, and capture results or exceptions within the **benchmark runner** framework. The **`mocha/customOutputRunner`** and **`mocha/runner`** modules are refactored to leverage `emitResultsMocha`, ensuring that the critical 'benchmark end' event is consistently emitted and errors are properly captured. This significantly improves the reliability and reporting accuracy of **benchmark tests** by preventing silent failures.

This commit performs a **revert** of PR #25915, specifically targeting changes within the **test versioning utilities** (`packages/test/test-version-utils`). The **maintenance revert** removes the previously introduced `SharedTree DDS` compatibility logic, related package loading simplifications, and adjustments to DDS factory handling in `compatUtils.ts` and `testApi.ts`. This action is taken to **address "Full Compat" failures in the test pipeline**, effectively **unblocking internal test pipelines** by restoring the prior stable state of the test infrastructure.

This commit **fixes an incorrect assertion** `0xb7b` that would prematurely trigger within the **`container-runtime`'s `outbox` module**. Previously, the `flush` method could be called reentrantly during incoming op processing, such as during connection state transitions, leading to the assertion firing even when there were no actual operations to flush. The **bug fix** introduces a check in `packages/runtime/container-runtime/src/opLifecycle/outbox.ts` to ensure the assertion only fires if there are pending operations, preventing unnecessary runtime errors. This improves the **robustness of the runtime's operation processing** by correctly handling reentrant `flush` calls. New test cases have been added to verify the `flush` method's behavior under these reentrant conditions.

This commit **enhances test coverage** for the **`dataStoreRuntime`** module by adding a new test case to `packages/runtime/datastore/src/test/dataStoreRuntime.spec.ts`. The test specifically asserts the content of the error message thrown when `FluidDataStoreRuntime.getChannel` cannot find a requested channel. This is a **temporary stop-gap** to ensure the stability of this specific error string, as external partners currently rely on inspecting this message to detect "channel not found" scenarios. This is particularly crucial for handling **`MultiFormatDataObject`** probing, where distinguishing this error from others is essential for proper functionality.

This commit **fixes a bug** in the **`id-compressor` runtime package** where `logger` and `sessionId` parameters were incorrectly dropped in overloaded `deserialize` methods and the `IdCompressor` constructor. This issue led to the `logger` being `undefined` for components like the `Summarizer`. The fix ensures proper parameter passing and initialization, improving the reliability of logging and session management within the `id-compressor` subsystem. New test cases were added to `idCompressor.spec.ts` to verify the correct wiring of the `logger` and parameter handling, alongside updates to existing performance and utility tests.

This commit introduces a **fix** to the **test logging mechanism** within the **Mocha test setup**, specifically addressing an issue in `packages/test/mocha-test-setup/src/mochaHooks.ts`. Previously, environment-specific properties like `displayName` were incorrectly overwriting the `details` field in test logs, obscuring actual code-generated details. The change renames the logging property dictionary and moves the serialization of these environment properties to a new, dedicated `testEnvProps` field. This ensures the `details` property is no longer squashed, significantly improving the **accuracy and clarity of test log analysis** by correctly separating environment context from test-specific details.

This commit performs a **cleanup and refactoring** by **removing the deprecated `bind` method** from the `IFluidHandleInternal` interface and its concrete implementation, `RemoteFluidObjectHandle`. This change primarily affects the **Fluid runtime utilities** and associated handle management, enforcing the deprecation introduced in version 2.40. To accommodate this removal, test infrastructure in `packages/dds/tree` and `packages/test/test-end-to-end-tests` was updated, adjusting mock and test handle classes like `MockSharedObjectHandle` and `TestFluidHandle` to no longer reference the removed method. The overall scope is to streamline the handle API and remove obsolete functionality from the codebase.

This commit **refactors** the **handle binding** mechanism in preparation for the removal of `IFluidHandleInternal.bind`, a deprecated interface method. It moves the core `bind` implementation logic from `FluidObjectHandle` to `SharedObjectHandle` within `packages/dds/shared-object-base`, where it is more appropriately located. Additionally, various **test suites** in `packages/dds` and `packages/test` are updated to remove direct `bind` calls or introduce dummy `bind` functions for **serialization** scenarios. This **maintenance** work ensures continued functionality during the transition, streamlining the handle system and paving the way for a cleaner API by eliminating deprecated functionality.

This commit significantly **enhances test coverage** for the **Staging Mode** feature by introducing comprehensive unit tests for key components like **`ContainerRuntime`** and **`PendingStateManager`**. The new tests validate `ContainerRuntime`'s staging mode entry/exit, `commitChanges`, and `discardChanges` behaviors, alongside `PendingStateManager`'s `replayPendingStates` and `popStagedBatches` logic under various scenarios. Crucially, this work also includes a **bug fix** for an **infinite loop** in `PendingStateManager.popStagedBatches` that could occur when discarding changes with empty staged batches, improving the **stability** of the staging mechanism. Beyond testing, minor **refactoring** and **maintenance** adjustments were made to related files, collectively ensuring the **reliability and correctness** of the **Staging Mode** implementation.

This commit introduces a **bug fix** to the **Container Runtime**'s **staging mode functionality**. It enhances the `exitStagingMode` API by adding crucial error handling, ensuring that the **container is explicitly closed** if an exception occurs during the staging mode exit process. This prevents the system from being left in an invalid or inconsistent state, significantly improving the **robustness and stability** of the runtime during staging mode transitions.

This **bug fix** addresses an issue within the **local-server-stress-tests** package, specifically in the `maybeSaveFluidOpsTofile` function located in `localServerStressHarness.ts`. Previously, this function would throw an error if a Fluid container never attached, leading to test failures. The commit now adds checks to gracefully handle **detached containers**, ensuring that operations are either saved correctly or an empty array is written. This improves the robustness and stability of the **local-server-stress-tests** by preventing crashes when containers are not attached.