Debugging startup crash of a universal app


You'll need a few tools:

 

Start the NP profiler and select your app. Select "Run under Debugger (without the profiler)" and select WinDbg from the Profiler Options.

Click on Start Profiling. WinDbg will start up and immediately kick off execution of your application.

 

Immediately, load SOSEX by executing:

.load [path\to\sosex\sosex]

 

Set WinDbg to break on CLR exceptions by executing:

sxe clr

 

Continue execution of your app under WinDbg until you hit the first chance exception. Once there, you can list the stack trace by executing:

kP

 

Or list the managed stack trace (with access to IL) by executing:

!mk

 

This should tell you where your application is crashing. The first chance exception will typically give you an address in memory of the instruction that failed, and if it's in managed code, it'll tell you the IL+number where it happened. That number is the number of IL instructions into the method where it failed. You can list the IL for the method by clicking on the IP of the function, or typing !muf [memory address]

So for example, if the exception tells you that it's thrown from IL+001e you can go to the method that threw and see what is in that space:

In this example, the exception was thrown when we called Window.get_Content.


Comments (1)

  1. Sebastian (@lowleveldesign) says:

    Thanks for the post. Seems that SOSEX is dead and won't support new versions of .NET 🙁 A nice replacement to consider is netext (netext.codeplex.com).

Skip to main content