通过

How to get Exception.StackTrace line numbers on Windows Phone

  • 项目

(update: I've written an updated blog post about this. Also I made it available on NuGet as package AsyncStackTraceEx. And published the source code under MIT license on Github.)  

Scenario: I've written my app and released it to users. Some of them have reported crashes but I don't know where. So I released an update which captures unhandled exceptions and invites the users to email these to me, in the hope that I can figure out what's wrong.

Problem: On Windows Phone it's not possible to deploy the PDB alongside the EXE/DLL. Without it, Exception.StackTrace is unable to provide line-numbers, and so all my phone error-reports come back without line-numbers and I don't know how to debug.

Solution: I'll write a small helper which adds those line numbers to exception stack-traces, without needing PDBs.

 

How it works

Here's a side-by-side comparison of the output.

PDB-based Exception.StackTrace(the highlighted parts are absent on phone since they depend on PDBs) Using my helper-library(the highlighted parts are present even on phone)
   at VB$StateMachine_3_BarAsync.MoveNext() in Class1.vb:line 24--- End of stack trace from previous location where exception was thrown ---   at TaskAwaiter.ThrowForNonSuccess(Task task)   at TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)   at TaskAwaiter.GetResult()   at VB$StateMachine_2_FooAsync.MoveNext() in Class1.vb:line 19--- End of stack trace from previous location where exception was thrown ---   at TaskAwaiter.ThrowForNonSuccess(Task task)   at TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)   at TaskAwaiter.GetResult()   at VB$StateMachine_1_TestAsync.MoveNext() in Class1.vb:line 14--- End of stack trace from previous location where exception was thrown ---   at TaskAwaiter.ThrowForNonSuccess(Task task)   at TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)   at TaskAwaiter`1.GetResult()   at VB$StateMachine_0_Button1_Click.MoveNext() in Class1.vb:line 5    at Test.BarAsync   at Test.FooAsync()#BarAsync in Class1.vb:19   at Test.TestAsync()#FooAsync(True) in Class1.vb:14   at Test.Button1_Click() in Class1.vb:5

 

I noticed that all the exceptions I ever cared about either arose in async operations in the framework, or in my own async methods. That let me use a fluent syntax, like this: (I've given the examples in C#, to balance out the fact that implementation is in VB...)

private async void Button1_Click(object sender, RoutedEventArgs e)
{
    try
    {
        await TestAsync().Log(); // so exception's stacktrace will show "Button1Click() in MainPage.xaml.cs:53"
    }
    catch (Exception ex)
    {
        MessageBox.Show(ex.Log()); // this retrieves the version of the stacktrace that includes line numbers
    }
}

Whenever I await something, I stick on ".Log()" at the end. This ensures that line numbers are preserved for any exception that comes out of the awaited task. And rather than retrieving the Exception.StackTrace property, I retrieve the stack-trace through my Exception.Log() extension method. This cleans up the async callstack into something more readable, and inserts back all those line numbers. I also added two other optional overloads of await-point logging, to get richer information in the exception's async callstack; I use that richer information to (optionally) indicate which API I'm calling, and what arguments I'm passing to it. After all, it's often the arguments to a method that are crucial to understanding why it threw the exception.

await BarAsync(b).Log("BarAsync"); // so the exception stacktrace shows "...#BarAsync in MainPage.xaml.cs:61"
await FooAsync(true).Log("FooAsync", true); // so the exception stacktrace shows "...#FooAsync(true) in MainPage.xaml.cs:72"

Comments

  • Anonymous
    August 15, 2013
    Nice trick! However it seems to me that the implementation is more complex than it should... Here's my attempt at making it simpler: gist.github.com/.../6250080 (I also change the name of the method to AsyncTrace)

  • Anonymous
    August 19, 2013
    That's nice. What you observed was that it's better to add into the Exception.Data dictionary, rather than mess around with weak reference list as I was using. I didn't know about Exception.Data :) ... You also decided not to bother correlating it with the actual exception stack trace, nor cleaning up the exception stack trace, nor providing extra data. Fair enough design decisions!

  • Anonymous
    October 18, 2013
    Thanks Lucian and Thomas for sharing your ideas. The first time I saw something like this it was on "Tango Master"  (a windows phone app). The application crashed, and when I restarted it then app asked for sending a crash report. Once I saw it, I implemented my own approach for providing the same functionality, and now that I see your ideas I have some code changes that It will provide more info. Kind regards, Herber