Stack trace for access violation

I am trying to write some general exception-handling logic for my C++ code, so that for *any* exception I log a stack trace to a file, and backup some data before the app crashes.

I am using an access violation as a test case, as in the following code:

Code:

---

try{
*((char*) 0) = 1;
  } catch (exception &e) {
          recorderror(e);
  } catch (...) {
          logerror("exception caught");
  }

---

First, I had to set the /EHa flag in Visual Studio to enable this exception to get caught at all. But now it is getting caught at the 'catch (...)' where I seemingly have no information about the exception.

Inside the 'catch (...)', how can I determine what type of exception was caught and how can I generate the stack trace? I did some searching around and can't seem to find a solution.

I found a way to get the stack trace at the point inside the catch, but that doesn't help because this is code that is going to enclose my entire program. I need the stack trace from the point of the original exception.


Thanks!

A quick search of the Web brings up Mixing SEH and C++ Exceptions. Perhaps it would be useful in your context.

Quote:

---

Originally Posted by locksleyu

I am trying to write some general exception-handling logic for my C++ code, so that for *any* exception I log a stack trace to a file, and backup some data before the app crashes.

---

Logging a stack trace is a good idea. I think backing up data is a bad one. Once your program has received an exception (and I mean SEH, not C++ exceptions which are synchronous) there is absolutely no way of telling just how badly things are screwed up.

You might think you are backing up some critical data, but might instead be overwriting the backup (which was intact) with complete garbage. The worst thing a crashing program can do is take down other programs and data along with it.

If you've already decided you're going to crash, then just crash immediately. If the data is that critical, the app should be regularly backing it up during NORMAL execution, not once things go wrong.