Question

I’m having an odd problem with my vs debugger. When running my program under the vs debugger, the debugger does not break on an unhandled exception. Instead control is returned to VS as if the program exited normally. If I look in the output tab, There is a first-chance exeption listed just before the thread termination.

I understand how to use the “Exceptions” box from the Debug menu. I have the break on unhandled exceptions checked. If I check first-chance exceptions for the specific exeption that is occuring, the debugger will stop.

However, it is my understanding that the debugger should also stop on any ‘Unhandled-Exceptions’. It is not doing this for me.

Here are the last few lines of my Output tab:

A first chance exception of type 'System.ArgumentOutOfRangeException' occurred in mscorlib.dll
The thread 0x60c has exited with code 0 (0x0).
The program '[3588] ALMSSecurityManager.vshost.exe: Managed' has exited with code -532459699 (0xe0434f4d).

I don’t understand why the exception is flagges as a “first chance” exception when it is unhandled.

I believe that the 0xe0434f4d exit code is a generic COM error.

Any Ideas?

Metro.

Answer 1

If you’re on a 64-bit OS, there’s a pretty good chance you’re being bitten by an OS-level behavior that causes exceptions to disappear. The most reliable way to reproduce it is to make a new WinForm application that simply throws an exception in OnLoad; it will appear to not get thrown. Take a look at these:

1. Visual Studio doesn’t break on unhandled exception with windows 64-bit
   • http: // social.msdn.microsoft.com/Forums/en/vsdebug/thread/69a0b831-7782-4bd9-b910-25c85f18bceb
2. The case of the disappearing OnLoad exception
3. Silent exceptions on x64 development machines (Microsoft Connect)
   • https: // connect.microsoft.com/VisualStudio/feedback/details/357311/silent-exceptions-on-x64-development-machines

The first is what I found from Google (after this thread didn’t help), and that thread led me to the following two. The second has the best explanation, and the third is the Microsoft bug/ticket (that re-affirms that this is “by design” behavior).

So, basically, if your application throws an Exception that hits a kernel-mode boundary on its way back up the stack, it gets blocked at that boundary. And the Windows team decided the best way to deal with it was to pretend the exception was handled; execution continues as if everything completed normally.

Oh, and this happens everywhere. Debug versus Release is irrelevant. .Net vs C++ is irrelevant. This is OS-level behavior.

Imagine you have to write some critical data to disk, but it fails on the wrong side of a kernal-mode boundary. Other code tries to use it later and, if you’re lucky, you detect something’s wrong with the data …but why? I bet you never consider that your application failed to write the data—because you expected an exception would be thrown.

Jerks.