User transparent software malfunction detection and reporting

[0001] This application is a continuation of application Ser. No. 09/265,839, filed on Mar. 10, 1999 now abandoned.

[0002] 1. Field of the Invention

[0003] The present invention relates to software malfunction detection and reporting.

[0004] 2. State of the Art

[0005] Software malfunction detection and reporting tools are known. One such tool is Microsoft's “Dr. Watson.” Dr. Watson is a debugging tool designed to provide software programmers with detailed information on the internal state of the Windows operating system when a Unhandled Application Exception (UAE) occurs. Dr. Watson must be running at the time a UAE occurs to extract the internal information from the system.

[0006] Dr. Watson uses comparatively little memory and does not affect the performance of Windows. A software programmer therefore has no reason not to install Dr. Watson, especially if a UAE has occurred before. After Dr. Watson is installed, information is collected when a UAE occurs and written to a special file (DRWATSON.LOG) located in the Windows directory. In addition, a Dr. Watson dialog box will appear, prompting the user to enter a description of what the user was doing prior to receiving the error. A developer may choose to start Dr. Watson automatically each time Windows is started, thus allowing critical information to be collected each time a UAE occurs. After several UAEs have been logged, the log may, if desired, be sent to a remote location for diagnosis.

[0007] Dr. Watson is a diagnostic tool, not a cure for a problem. Having Dr. Watson will not prevent an error from occurring, but the information in DRWATSON.LOG often helps developers determine the cause of the error.

[0008] Another tool, the Microsoft Diagnostics (MSD) program, is designed to assist Microsoft customers and Product Support Services (PSS) technicians in solving problems with Microsoft products. MSD identifies system configuration information such as the BIOS, video card type and manufacturer, installed processor(s), I/O port status, operating system version, environment settings, hardware devices attached, and additional software running concurrently with MSD.

[0009] MSD is often used in conjunction with Dr. Watson to provide information on hardware configurations and UAEs. Error reports may be generated that include information from both the MSD program and Dr. Watson.

[0010] As may be appreciated from the foregoing description, existing software malfunction detection and reporting tools are targeted to software developers. Significant computer expertise and manual interaction is required in order to use these tools effectively. In particular, such tools are not well suited for beta users. Although beta users are supposed to report crashes and malfunctions, a beta user may experience a malfunction but, in the day-to-day rush of business, fail to report it. A need exists for a software failure and detection tool well suited for beta users such that a software developer can obtain malfunction reports and diagnostic information easily and reliably.

[0011] The present invention, generally speaking, “instruments” an arbitrary software program, without changing the software program, to automate malfunction detection and reporting. Although users can be invited to enter a description of what the user was doing prior to receiving the error, report generation and transmission to a remote server can be fully automatic and transparent to the user. In the case of beta testing, therefore, a software developer is guaranteed to receive all pertinent information about malfunctions of an application without having to rely on “fallible humans” for this information. The effectiveness of beta testing, in terms of ultimately contributing to an improved product, is therefore greatly increased. Various kinds of malfunctions may be detected and reported, including an application “crashing,” becoming “hung,” etc.

[0017] Referring now to FIG. 1, a block diagram is shown of an automated malfunction detection and reporting system in accordance with an exemplary embodiment of the present invention. Communication of a malfunction report may be accomplished in any of various different ways, for example via a LAN connection, with data collection occurring inside a corporation, via the Internet, of simply by means of the generation of an ASCII report which is then inserted into an email outbox. In accordance with one advantageous embodiment, to be presently described, a user machine is assumed to have installed software realizing a “polite agent” that communicates messages of abbreviated length with a remote server using a “virtual connection,” e.g., an “opportunistic” Internet connection. Such a polite agent is described in U.S. application Ser. No. 09/138,403 filed Mar. 10, 1999, incorporated herein by reference. The polite agent technology described in published International Patent Application WO 97/07656, also incorporated herein by reference, may also be used with the present invention.

[0018] The agent is provided with a list of active processes, or “target applications, with which the agent is to interact, e.g., to perform automated malfunction detection and reporting. One such target application is shown in FIG. 1. As illustrated, a target application will typically include a main body of executable code (.EXE), various Dynamic Link Libraries (.DLL), various Custom Control and COM objects (.OCX), etc. The terms EXE, DLL, OCX, etc., are specific to the Window operating system. In the case of other operating systems, other terminology is applied to similar concepts. The invention may be applied in connection with any operating system.

[0019] When a process is started on the user machine, the agent “hooks” (installs code realizing an application hook for) that process. The application hook then identifies the process either as a process of interest, i.e., one of the agent's active processes, or as a process not of interest. In the latter case, the application hook removes itself. The function of the application hook is to intercept and filter messages from the target application and communicate messages of interest to the agent, which maintains a timestamped event log. The application hook and the agent communicate through a hook support thread using operating system messaging facilities, e.g., in the case of the Windows operating system, SendMessage( . . . ) and PostMessage( . . . ). For purposes of malfunction detection and reporting, messages of interest include menu selections, loading of a DLL, etc. These events are timestamped by the agent and entered into the event log.

[0020] Note that multiple target application may be active at the same time, in which case each target application would have its own application hook, hook support thread and timestamped event log.

[0021] The application hook allows relatively high-level activity information concerning the target application to be captured. Such activity information is readily intelligible to the average software programmer, in contrast to the much more arcane information gathered by known malfunction detection and reporting tools, such as stack frames, pointers, etc.

[0022] In addition to the application hook, two addition threads are created on a per-application basis, a Micro Debugger thread and a Check Alive thread.

[0023] The Micro Debugger thread is registered with the operating system as the debugger for the target application using the appropriate operating system call, e.g., DebugActiveProcess( . . . ). This call waits until an exception occurs in the target application and then returns control to the Micro Debugger. The target application therefore runs under the control of the Micro Debugger. The Micro Debugger is notified by the operating system when a “crash” (UAE) occurs. The Micro Debugger communicates this information to the agent, performs cleanup for the target application, and terminates the target application. The agent then solicits additional information from the user concerning the malfunction.

[0024] Check Alive is an optional feature that may be specified when the instruction to monitor a particular application is given. The purpose of the Check Alive thread is to detect when the target application has hung, i.e., become non-responsive to user input. The Check Alive performs this function by repeatedly “pinging” the target application, i.e., send a message to the target application to see if the target application responds. A timeout period is set for the message such that if the target application does not respond within the timeout period, control is returned to the Check Alive thread. If Check Alive thread detects that the application is hung, it notifies the agent, which then solicits confirmation and additional information from the user.

[0025] After the target application has either crashed or becomes hung, the agent assembles and (optionally) compresses the “crash log” as part of a bug report form and schedules the bug report form for automatic upload to the server using “polite connection” methods or some other connection. An example of a crash log is shown in FIG. 2.

[0026] Referring now to FIG. 3, operation of the Micro Debugger thread will be described in greater detail. Once an application hook has found a target application, a Micro Debugger thread is created for that target application and registered with the operation system as the debugger for the target application process. The Micro Debugger then causes execution of the process to be then interrupted while the operating system enumerates modules that the process has loaded. In the case of the Windows operating system, the operating system identifies each module loaded by the target application thread along with the memory location of that module. When all of the modules have been enumerated, this information is sent to the agent to be logged, and the target application process is allowed to continue running. Any modules loaded or unloaded thereafter will be detected by the Micro Debugger and logged as described previously.

[0027] When an exception occurs, the operating system calls the Micro Debugger, which determines what kind of exception has occurred and where in memory the exception occurred. The Micro Debugger also captures additional context of the exception, e.g., what happened (divide by zero, access violation, etc.) and in which module. The Micro Debugger then sends this information to the agent to be stored in the event log. Finally, the Micro Debugger terminates the target application. Optionally, prior to terminating the target application, the Micro Debugger may perform cleanup for the target application. In order to do so, the Micro Debugger will typically require some detailed knowledge of the target application. Given such knowledge, the Micro Debugger could, for example, save a user's work prior to terminating the target application such that the user's work is not lost.

[0028] Referring to FIG. 4, operation of the Check Alive thread will be described in greater detail. Once an application hook has found a target application, a Check Alive thread is created for that target application. The Check Alive thread then enters a loop in which the following actions are performed. The Check Alive thread sends a message with timeout to the target application. The timeout period may be, for example, 5 seconds. The Check Alive thread then waits for a response from the target application. If a response is received within the timeout period, the Check Alive thread waits for a period of time (e.g., about 15 seconds) before sending another message with timeout. This manner of operations continues for so long as the target application is running and responsive.

[0029] If a response is not received from the target application within the timeout period, either the Check Alive thread or the agent attaches a dialog box to the application window asking the user to confirm that the target application has hung. If the user concurs, then the process is terminated. The agent then creates a crash log and schedules it for upload to the server as described previously.

[0030] Automatically collecting malfunction information on the server without the need for user intervention greatly facilitates the job of a software developer to find and fix program bugs. A beta test coordinator, for example, may remotely log onto the server and view malfunction information. The beta test coordinator may find, for example, that of 200 installed copies of a program, five copies have crashed within a week's time. Based on this information—the reliability of which is assured—priorities may then be set to resolve the problem.

[0031] It will be appreciated by those of ordinary skill in the art that the invention can be embodied in other specific forms without departing from the spirit or essential character thereof. The presently disclosed embodiments are therefore considered in all respects to be illustrative and not restrictive. The scope of the invention is indicated by the appended claims rather than the foregoing description, and all changes which come within the meaning and range of equivalents thereof are intended to be embraced therein.