If the application you are profiling stops responding, fails to load, or closes unexpectedly, the profiling process may have caused the profiled application to crash. If this happens, you may not receive profiling results for your session, or you may receive incomplete results captured before the crash.

If your profiling session returns no results, or incomplete results, but does not hang or crash, see Troubleshooting missing results.

Troubleshooting all crashes

If the application crashed but did not return the above error message, try the following approaches. They are listed in order of how likely they are to resolve common causes of crashes.

Force your application to use .NET 4

Many crashes can be avoided by rebuilding the application as a .NET 4 executable, and profiling it using the Attach to a .NET 4 process option in Application Settings.

For instructions on reconfiguring your application for use with this profiling method, see Forcing your application to use .NET 4.

The Attach to a .NET 4 process approach profiles your application when it is already running, rather than launching a new instance of the application. Because attaching to a running process does not change the way the application's code executes, it is less likely to cause a crash.

Run ANTS Performance Profiler from the Visual Studio menu

If the crash occurs in a profiling session launched directly from ANTS Performance Profiler, try profiling using the Visual Studio ANTS Performance Profiler add-in instead.

For instructions see Using the Visual Studio add-in.

Delete corrupt third-party PDBs

The following third-party assemblies sometimes contain corrupt .pdb files, which can cause crashes during profiling:

• Microsoft.Practices library
• AjaxControlToolkit

For instructions on finding and deleting corrupt .pdb files, see Troubleshooting PDB problems.

Troubleshooting web application crashes

If the application you are profiling is a web application, try the following two steps.

Profile using another web server application

Some applications that crash under profiling can be profiled successfully if run on another type of server application:

• If you are profiling in IIS, switch to the ASP.NET built-in web development server.
• If you are profiling in the ASP.NET web development server, switch to IIS.

Note that changing environments will work only if the server application you select supports all your application's features:

• Applications using integrated pipeline mode, impersonation, or HTTPS will not run on the built-in web-development server.
• Applications that cannot operate in a restricted security context may not run in IIS.

Profiling results obtained in from applications running on the web development server may also differ from performance in a production environment.

IIS: ensure you are profiling on the correct port

• If you are using IIS 6 or IIS 7, select Unused port and choose a port that is not used by IIS.
  Note that this will not work if your application's code binds to a specific port.
• If you are using IIS 5, or if you are using IIS 6 or 7 and your application binds to a specific port, select Original port.
  IIS will restart so that the profiler can attach to the port.

Troubleshooting error messages

StackOverflow Exception

During line-level profiling, if the application you are profiling is very large or the profiling session lasts a long time, a stack overflow can occur. If this happens, the application will crash and may record a "StackOverflow Exception" error in the log. (For instructions on finding the ANTS Performance Profiler logs, see Log files.)

To reduce the risk of an overflow, open the Tools menu, click Options, and select Simplify very complex stack traces to save memory.

If this does not prevent the crash, try profiling in sampling mode: on the Application Settings dialog box, set theProfiling mode to Sample method-level timings (fastest, least detail, no hit counts).

Getting more help

If the steps above do not prevent the profiled application from crashing, please contact Redgate support, including, if possible, a log of the application crash.