// Copyright 2013 The Chromium Authors. All rights reserved. // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. #ifndef CONTENT_PUBLIC_BROWSER_TRACING_CONTROLLER_H_ #define CONTENT_PUBLIC_BROWSER_TRACING_CONTROLLER_H_ #include "base/debug/trace_event.h" #include "content/common/content_export.h" namespace base { class FilePath; }; namespace content { class TracingController; // TracingController is used on the browser processes to enable/disable // trace status and collect trace data. Only the browser UI thread is allowed // to interact with the TracingController object. All callbacks are called on // the UI thread. class TracingController { public: enum Options { ENABLE_SYSTRACE = 1 << 0, ENABLE_SAMPLING = 1 << 1, }; CONTENT_EXPORT static TracingController* GetInstance(); // Get a set of category groups. The category groups can change as // new code paths are reached. // // Once all child processes have acked to the GetCategories request, // GetCategoriesDoneCallback is called back with a set of category // groups. typedef base::Callback&)> GetCategoriesDoneCallback; virtual void GetCategories( const GetCategoriesDoneCallback& callback) = 0; // Start recording on all processes. // // Recording begins immediately locally, and asynchronously on child processes // as soon as they receive the EnableRecording request. // // Once all child processes have acked to the EnableRecording request, // EnableRecordingDoneCallback will be called back. // // |filter| is a filter to control what category groups should be traced. // A filter can have an optional '-' prefix to exclude category groups // that contain a matching category. Having both included and excluded // category patterns in the same list would not be supported. // // Examples: "test_MyTest*", // "test_MyTest*,test_OtherStuff", // "-excluded_category1,-excluded_category2" // // |options| controls what kind of tracing is enabled. typedef base::Callback EnableRecordingDoneCallback; virtual bool EnableRecording( const base::debug::CategoryFilter& filter, TracingController::Options options, const EnableRecordingDoneCallback& callback) = 0; // Stop recording on all processes. // // Child processes typically are caching trace data and only rarely flush // and send trace data back to the browser process. That is because it may be // an expensive operation to send the trace data over IPC, and we would like // to avoid much runtime overhead of tracing. So, to end tracing, we must // asynchronously ask all child processes to flush any pending trace data. // // Once all child processes have acked to the DisableRecording request, // TracingFileResultCallback will be called back with a file that contains // the traced data. typedef base::Callback)> TracingFileResultCallback; virtual bool DisableRecording(const TracingFileResultCallback& callback) = 0; // Start monitoring on all processes. // // Monitoring begins immediately locally, and asynchronously on child // processes as soon as they receive the EnableMonitoring request. // // Once all child processes have acked to the EnableMonitoring request, // EnableMonitoringDoneCallback will be called back. // // |filter| is a filter to control what category groups should be traced. // // |options| controls what kind of tracing is enabled. typedef base::Callback EnableMonitoringDoneCallback; virtual bool EnableMonitoring(const base::debug::CategoryFilter& filter, TracingController::Options options, const EnableMonitoringDoneCallback& callback) = 0; // Stop monitoring on all processes. // // Once all child processes have acked to the DisableMonitoring request, // DisableMonitoringDoneCallback is called back. typedef base::Callback DisableMonitoringDoneCallback; virtual bool DisableMonitoring( const DisableMonitoringDoneCallback& callback) = 0; // Get the current monitoring configuration. virtual void GetMonitoringStatus(bool* out_enabled, base::debug::CategoryFilter* out_filter, TracingController::Options* out_options) = 0; // Get the current monitoring traced data. // // Child processes typically are caching trace data and only rarely flush // and send trace data back to the browser process. That is because it may be // an expensive operation to send the trace data over IPC, and we would like // to avoid much runtime overhead of tracing. So, to end tracing, we must // asynchronously ask all child processes to flush any pending trace data. // // Once all child processes have acked to the CaptureMonitoringSnapshot // request, TracingFileResultCallback will be called back with a file that // contains the traced data. virtual void CaptureMonitoringSnapshot( const TracingFileResultCallback& callback) = 0; protected: virtual ~TracingController() {} }; } // namespace content #endif // CONTENT_PUBLIC_BROWSER_TRACING_CONTROLLER_H_